SMITHSONIAN CONTRIBUTIONS to ASTROPHYSICS

SMITHSONIAN

CONTRIBUTIONS

to

ASTROPHYSICSVOLUME I

SMITHSONIAN

INSTITUTION

WASHINGTON

D. C.

Contents of Volume 1

No. 1. New Horizons in Astronomy. Pages i-x, 1-181. Pub-lished December 19, 1956.

No. 2. Papers on Reduction Methods jor Photographic Meteors.Pages 183-243. Published May 13, 1957.

Smithsonian

Contributions to Astrophysics

VOLUME 1, NUMBER 1

NEW HORIZONS IN ASTRONOMYedited by FRED L. WHIPPLE

DIRECTOR, ASTROPHYSICAL OBSERVATORY

SMITHSONIAN INSTITUTION

flfe

SMITHSONIAN INSTITUTION

Washington, D. C.

1956

UNITED STATES

GOVERNMENT PRINTING OFFICE

WASHINGTON : 1956

For sale by the Superintendent of Documents, U. S. Government Printing OfficeWashington 25, D. C. - Price $1.50

Smithsonian Contributions to Astrophysics

The Smithsonian Institution's Astrophysical Observatory is passingthrough a period of reorganization. The scientific headquarters havebeen moved to Cambridge, Mass., where a close working association withthe Harvard College Observatory is being developed. These changes, aswell as the many advances in science and technology, force us to reevaluatethe basic scientific policies and goals of the Astrophysical Observatory.

With the present rapid progress in basic science and its application totechnology, research in the fundamental astrophysical processes of thesun, earth, planets, and interplanetary medium is going forward at arapid pace. Concurrently, our mushrooming technology has become moreand more sensitive to these phenomena of the solar system, heretoforethought to be of only academic interest.

In keeping with these developments, the Astrophysical Observatorymust broaden the scope of its activities while continuing its long-estab-lished tradition of research—a tradition that began with the pioneeringstudies on aerodynamics by the observatory's founder, Samuel P. Langley,and continued during years of active investigations under Dr. C. G.Abbot in solar and terrestrial phenomena and their interrelationships.Thus, our future program will include not only radiation, its variations,and its effects on our planet, but also other phenomena that affect theearth and its atmosphere. We find, for example, that corpuscular radia-tion from the sun is associated with solar activity as well as with suchgeophysical phenomena as the aurorae, night air glow, terrestrial magne-tism, the ionosphere, radio transmission, and other observed phenomena.Meteoric bodies, dissipated from comets or chipped from asteroidalsources, not only produce meteors in our atmosphere and provide samplesof cosmic material for our museums but also ionize the high atmosphereand contribute to its chemical composition, scatter sunlight in the zodiacallight, and provide us with a cosmic ballistics laboratory of ultravelocityprojectiles.

The Astrophysical Observatory, therefore, faces widening horizons.It will broaden the scope of its studies of solar phenomena and atmos-pheric effects to include related phenomena within the solar system suchas meteorites, hypervelocity ballistics, and the zodiacal light. It willcooperate in exploiting new methods of research and technology such asnuclear processes and artificial satellites. To strengthen its effort theAstrophysical Observatory will not carry out its research in isolation;it will coordinate its work closely with that of other astrophysicists andgeophysicists who are studying similar problems.

Although research in astrophysics continues to expand, no correspond-ing growth has occurred in the avenues of publication. Observationalresults are often so compressed in print that they cannot be analyzed in

detail by independent investigators; and advances in technical proceduresor instrumentations are often presented with such brevity that their valueto the potential user is small.

In an effort to affect these trends and to provide a proper communi-cation for the results of research conducted at the AstrophysicalObservatory, we are inaugurating this new publication, Smithsonian Con-tributions to Astrophysics. Its purpose is the "increase and diffusion ofknowledge" in the field of astrophysics, with particular emphasis onproblems of the sun, the earth, and the solar system. Its pages will beopen to a limited number of papers b\r other investigators with whom wehave common interests.

We earnestly hope that Smithsonian Contributions to Astrophysics willserve to promote a greater understanding, appreciation, and enjoyment ofthat part of the universe in which we are privileged to live.

FRED L. WHIPPLE, DirectorAstrophysical ObservatorySmithsonian Institution

Cambridge, MassachusettsNovember 14, 1956

A grant from the National Science Foundationtoward the publication of New Horizons inAstronomy is gratefully acknowledged.

New Horizons in Astronomy

Preface

The project of collecting a series of papers oriented toward newhorizons in astronomy began in the 1954-55 Panel of Astronomy to theNational Science Foundation. The initial impetus came largely from theenthusiasm of Dr. Otto Struve, who has presented some of his thoughtsin a paper entitled "The General Needs of Astronomy" (Publ. Astron.Soc. Pacific, vol. 67, p. 214, 1955). The resulting deliberations led to theestablishment of an ad hoc committee on the "Needs of Astronomy," ofwhich I was made chairman. I felt that some of the goals of this com-mittee could be attained by publishing a collection of papers by leadersin the various fields of astronomy who would present their concepts ofresearch projects most likely to advance the science of astronomy duringthe next decade.

The purpose is at least fourfold:To provide material that might assist the National Science Foundation,

in its allocation of funds, to support the astronomical research most likelyto be productive.

To help younger astronomers choose the most fertile fields for activeresearch.

To encourage established astronomers to pause for a moment andreflect on the most advantageous planning in their own fields of research,and to supply them with a broader insight into the activities in otherfields of astronomy, possibly related to their own.

To encourage research in new or neglected fields of astronomy, par-ticularly those that border on other areas of physical science such asgeophysics, electronics, chemistry, fluid dynamics, and the like.

The publication of New Horizons in Astronomy has been supportedjointly by the Smithsonian Institution and the National Science Founda-tion. The Astrophysical Observatory shares with the National ScienceFoundation the hope that these papers may promote intensified effortsand increasingly significant results in astronomical research.

The hearty cooperation of the many busy scientists who took time fromtheir other duties to write these papers should establish a debt of gratitudein the minds of all those who will profit by these contributions.

I am deeply indebted to Mr. Paul H. Oehser, Mr. Ernest Biebighauser,and Mr. John S. Lea, of the Editorial and Publications Division of theSmithsonian Institution, and to Mrs. Lyle Boyd, of the Harvard CollegeObservatory, for their great interest and effort in this project; otherwisethese papers would not now rest in your hands.

FRED H. WHIPPLE

ContentsPage

Techniques and Instrumentation

Optics: I. S. Bovven 1Solar Instrumentation: J. W. Evans and R. B. Dunn 5Astronomical Seeing and Scintillation: Geoffrey Keller 9Spectroscopy: Charlotte E. Moore 13Radial Velocities: R. M. Petrie 15Astrometry: K. Aa. Strand 21Photoelectric Astronomy: A. E. Whitford 25Interferometers in Radio Astronomy: B. F. Burke and M. A. Tuve . 31

Related Sciences

Rocketry: R. Tousey 39Computing Machines in Astronomy: Paul Herget 45Turbulence: S. Chandrasekhar 47Astroballistics: Richard N. Thomas 49Electrodynamics of Fluids and Plasmas: Max Krook 53Hydromagnetic and Plasma Problems: Eugene N. Parker 59

Solar System

Earth's Magnetism: Walter M. Elsasser 67Time: William Markowitz 73Geographical Positions: John A. O'Keefe 75Celestial Mechanics: G. M. Clemence 79Meteorites: John S. Rinehart 81Meteors: Fred L. Whipple 83Minor Planets: Paul Herget 87Planets, Satellites, and Comets: Gerard P. Kuiper 89Airglow and Aurora: A. B. Meinel 95Solar-Terrestrial Relationships: Weather and Communications:

Walter O r Roberts 99Solar Physics: Leo Goldberg and Donald H. Menzel 103

Stars and the Galaxy

Spectral Classification: W. W. Morgan . 115Stellar Atmospheres: Marshal H. Wrubel 119Aerodynamic Problems in Stellar Atmospheres: Richard N. Thomas. 123Emission Nebula*: Lawrence H. Aller 127

381014—56 2

CONTENTS

PageStars and the Galaxy—Continued

Double and Multiple Stars: Otto Struve 131Stellar Dynamics: U. Van Wijk 135Variable Stars: C. Payne-Gaposchkin 137Interstellar Matter: Lyman Spitzer, Jr 13921-Centimeter Research: Bart J. Bok 141Some Observational Problems in Galactic Structure: Harold Weaver. 149Galactic Clusters and Associations: A. Blaauw 159Globular Clusters Observed through a Crystal Ball: W. A. Baum. . 165Stellar Structure and Evolution: M. Schwarzschild 177Radio Sources: John D. Kraus 179

Techniques and Instrumentation

OpticsBy I. S. Bowenl

Geometrical optics is one of the oldest ofsciences. Most of the basic laws have beenknown for centuries, yet in the past two orthree decades advances in optical designs haverevolutionized astronomical techniques. Anoutstanding example of this is the Schmidtcamera. Until the discovery of the principleof this instrument in 1931, camera opticsproviding critical definition over a field greaterthan two or three degrees in diameter werelimited to focal ratios of d/f =1/5 or less.With the Schmidt camera critical definitioncan be achieved over wide fields at focal ratiosas fast as dff=l. This 10- to 100-fold gain inspeed has opened up many new fields inastronomy.

Furthermore it has been possible to design,for special purposes, many modifications of theSchmidt cameras, some of which achieve evengreater gains in speed and field. These includethe Baker Super-Schmidt for meteor photog-raphy and the solid-block and thick-mirrorSchmidts, the Schmidt-aplanatic sphere com-bination, and the twice-through corrector platesfor spectroscopic applications. There are un-doubtedly many additional modifications thatshould be explored. For example, the focalsurfaces of most of these Schmidt systems arecurved with a radius approximately equal tothe focal length. Thin glass plates can bebent to radii greater than 18 inches and simplefield flatteners yield satisfactory results incameras with focal ratios up to about d/f=l.However, to exploit fully the still higherspeeds in the shorter focal lengths it will benecessary to develop either field flatteners thatare effective at the higher focal ratios or to finda flexible emulsion base that is stable againstchanges in temperature and humidity.

Another urgently needed development is anoptical system that will provide a wide field of

1 Director of Mount Wilson and Paloraar Observatories.

critical definition for telescopes of very largeaperture. Schmidt telescopes have been suc-cessfully made with apertures up to 48 incheswith a focal length of 120 inches. However, inorder to avoid chromatic aberrations the focallength must increase as the 3/2 power of theaperture. Thus the minimum focus of a 200-inch Schmidt is 1,000 inches. The length ofthe tube would have to be twice this, or 2,000inches; i. e., three times that of the presentHale telescope. This would make the cost ofthe telescope tube and the dome to house itprohibitive.

Unfortunately the field of good definition ofa high-speed paraboloid is very small; less than}{ inch in diameter for the Hale telescope inwhich d/f= 1/3.3. Corrector lenses designedby Dr. F. E. Ross and placed in front of theplate have successfully increased this area ofgood definition to a diameter of 3 or 4 inches.This still falls far short of the 14 X 14-inch area ofcritical definition attained with the 48-inchSchmidt, which operates at the even higherspeed of d/f= 1/2.5. Because of this smallfield, the use of large telescopes such as the200-inch is impractical for most survey pur-poses.

Another important need at the present timeis an objective prism or similar device that maybe used for the wholesale classification of spectraof stars of magnitude fainter than can now bereached. Thus, because of the deterioration ofthe definition of the spectra by "seeing," it isnot feasible to use an objective prism on tele-scopes of appreciably more than 10-foot focallength. Furthermore, the limit on the exposuretime set by the sky background is essentiallythe same with an objective prism as for directphotography. If the usual dispersion of 200-300 A/mm is used and the spectra are widenedto 0.2 or 0.3 mm, the starlight is spread over anarea 1,500 to 4,000 times that of a star image,

SMITHSONIAN CONTRIBUTIONS TO ASTROPHYSICS

with the result that the limiting magnitude forthe spectra is 8 or 9 magnitudes below the limitfor direct photography. For a 10-foot focallength the limiting magnitude for direct pho-tography isMph=21 and therefore for spectros-copy Mph=12 or 13. This falls far short of themagnitudes that may be reached with slitspectra. Thus with the 200-inch telescope a1-night exposure enables one to reach the 15thmagnitude at 38 A/mm and the 18th magnitudeat 170 A/mm. At the present time there is noequipment available for picking up unusualfaint objects for detailed study at these higherdispersions with a slit spectrograph. The antici-pated development of the image tube, discussedbelow, will render this problem even more seriousby pushing the limit of the slit spectrographfrom 2 to 5 magnitudes still fainter.

One suggested solution of this problem is theuse, as a multislit, of a high-contrast positiveprint of a given star field taken with the sametelescope. With a wide-angle telescope and aspectrograph having wide-angle collimator andcamera optics, a large number of spectra couldbe obtained simultaneously without interferencefrom the sky background. Unfortunately, ifSchmidt optics are used throughout, the fieldflatteners necessary to bring the curved fields ofthe telescope and collimator into coincidencedestroy the off-axis collimation. Some otherwide-angle system must be devised for thepurpose.

The efficiency of slit spectrographs would besubstantially increased if larger gratings couldbe made or if gratings could be ruled which pro-vide higher angular dispersion and still retain ahigh concentration of light in one order. Inorder to avoid interference between collimatorand camera optics, this grating should prefer-ably be of the transmission type. The so-called"venetian-blind type" described in "Astronom-ical Spectrographs: Past, Present, and Future,"in ' 'Vistas in Astronomy,'' is one possible solutionif the necessary ruling techniques can bedeveloped.

Another problem that needs investigation isthat of seeing. Presumably very little can bedone to reduce turbulence in the upper atmos-phere. However, in solar instruments much ofthe disturbance is caused by the heating of theoptical parts and their surroundings by the sun-

light which is incident on them. Undoubtedlythe local air turbulence due to this could be sub-stantially reduced by proper light shields andby artificial cooling of parts that cannot beshielded. Useful studies could also be made todetermine the most effective method for thermo-stating an optical instrument such as a spectro-graph without setting up internal air currentsand turbulence that often cause serious deteri-oration of the image.

Every large telescope is essentially a collectorof light which is focused on some type ofreceiver. The properties of this receiver playa very large role in the proper design of thetelescope. Thus up until the present centurypractically all astronomical observations weremade visually; i. e., the human eye was thereceiver. The range of wavelengths to whichthe eye has appreciable sensitivity is very small(1,000-1,500 A) and the field of view of criticaldefinition is also very limited. Furthermore,the lens of the eye is a part of the opticalsystem and sets an upper limit to the size ofthe cone of light that can be concentrated onthe sensitive surface. For this type of receiverthe simple two-lens refractor operating at focalratios of d/j'=1/15 to 1/20 was ideal. As aresult nearly all large telescopes constructedin the last half of the 19th century were of thistype.

By 1900 photographic procedures were rapidlyreplacing visual observations. The silver-halidegrains, however, are sensitive in the ultravioletto beyond the limit of transmission of theatmosphere while the more recent developmentof dyes has produced plates that are sensitivewell out into the infrared range. Also, in orderthat an image be recorded on a photographicplate it is necessary that a certain minimumnumber of quanta fall on a unit area of theplate. This limit is such that in order tophotograph faint galaxies and nebulosities ina feasible exposure time it is necessary to use alight collector with a focal ratio of at leastdjf=l/5. This fact resulted in a sudden shiftearly in the present century to the use of thereflector, which provides the complete achro-matism and greater speed necessary for theefficient use of the new type of receiver.

Not only were the properties of the photo-graphic plate major factors in causing the shift

NEW HORIZONS IN ASTRONOMY

from refractors to reflectors, but they haveplayed an important role in fixing many of thedesign criteria for both the reflector itself andthe auxiliary equipment, such as spectrographs,to be used with it. One of these properties isthe linear resolving power of the plate. Thisresolving power sets the minimum focal lengthof the main telescope or of the spectrographcamera if a given angular resolving power orwavelength resolution, AX, is to be achieved.Furthermore, in all of these instruments thisplate resolving power is the primary factorconsidered in setting up optical design andconstruction tolerances. Another property ofthe plate, namely, that it can record a verywide field at one time, led to the development ofvarious wide-field optics culminating in theSchmidt camera. Similarly the necessity ofconcentrating a certain minimum of light perunit area in order to obtain an image has beenthe chief cause of the design and constructionof optical systems of extreme speed («?//= 1 orgreater) for spectrograph cameras.

It is now becoming apparent that anotherchange in the receiver for astronomical ob-servations is in the offing. Thus photoelectricsurfaces have been developed, and are in usefor television and other purposes, that have aquantum efficiency from 10 to 100 times ashigh as that of the best photographic plates.The electrical circuits used in some of thesereceivers permit the reduction or eliminationof a uniform background of the type caused byscattered starlight and the permanent aurora,which sets the present limit in direct photog-

raphy of the sky. Indeed it now appears thatin the foreseeable future the limit of detectabil-ity of a faint star image will approach the theo-retical limit. This limit is set by the necessityof collecting a larger number of quanta fromthe star than the probable random fluctuationin the number of quanta received from the skybackground in the area of the sky covered bythe object.

These developments are still in such an earlystage that it is too soon to predict their possibleeffect on telescope design. Thus the linearresolving power will depend on the final typeof receiver adopted. I t is also quite probablethat the new receiver will not require, like thephotographic plate, a minimum concentrationof quanta per unit area for a satisfactory record.One criterion for this will depend on the pos-sibility of eliminating spurious thermal electronsby refrigeration or other means. In any casethe much greater efficiency of the photo-electric process will permit the record of agiven object to be obtained in very much lesstime than that required by the photographicprocess. The shift in these two properties willprobably eliminate the necessity for optics ofextreme focal ratio such as the types recentlydeveloped for both direct and spectroscopicobservations. Regardless of the propertiesof the photoelectric receiver as finally developed,substantial changes in telescope and spectro-graph design may be expected, if for no otherreason than the elimination of many of thepresent restrictions set by the properties of thephotographic plate.

Solar InstrumentationBy J. W. Evansx and R. B. Dunn

The technical problems of solar observationare decidedly different from those in any otherfield of astronomy because of the abundanceof light and the large apparent diameter of thesun. The latitude for the exercise of ingenuityin securing more and more refined informationis practically unlimited, and affords a kind ofsatisfaction rarely found in any other field.The next 25 years promise to be the mostexciting and the most productive since the for-mulation of the laws of spectrum analysis byKirchhoff. Since 1945, the conception andconstruction of new instruments and techniqueshave accelerated markedly, as have the theoret-ical advances which designate the criticalobservations we need to determine the physicalcharacter of the sun. We have barely begunto get acquainted with the new tools and toexploit their possibilities, but the results arealready impressive.

Most of the new developments are projectionsof older devices and methods, with refinementsthat greatly increase their power. Two ofthem, however, introduce entirely new methodsthat are sure to be exceedingly fruitful: theobservation of the far ultraviolet spectrum ofthe sun from rockets, and the recording of solarradio emission.

The rocket observations are a fine exampleof cooperative effort in several fields to solve avery difficult technical problem. The record-ing instrument is either a spectrograph designedto work in the 900- to 3,000-angstrom range oran X-ray ionization chamber for the 1- to5-angstrom region. The recorder must bemounted in the rocket with a device which keepsit pointed at the sun with an accuracy of a fewminutes of arc, in spite of lively gyrations inits supporting platform. Finally, the rocketitself must be capable of carrying its load toan altitude of 100 km or more in order to rise

1 Sacramento Peak Observatory, Sunspot, X. Mex.

above the regions of the earth's atmospherethat absorb the solar radiations in these wave-length regions and prevent their observationfrom the ground.

This brief description can give no concept ofthe difficulties encountered in the work, norof the years of patience and imagination thatwere required to overcome them; but the re-sults so far have more than justified the effort.The expected spectroscopic features duly ap-peared, but some very unexpected emissionlines were also recorded. The expected linesstrengthen our confidence in the broad featuresof existing solar theory, and the new linespresent new conditions which any valid theorymust satisfy. The rocket observations haveunveiled a hitherto inaccessible part of thesolar spectrum, but the quick glance we havebeen allowed sharpens the appetite for moredetails, more quantitative measurements, anda further extension to still shorter waves.Rockets can still do a great deal, but therecan now be little doubt that a man-made satel-lite eventually will be our platform for solarresearch above the absorption of the terrestrialatmosphere. New difficulties will appear andbe overcome, and new data on the solarspectrum will further delineate the structureof the sun. Such advances, inevitably, willraise new questions in place of the old onessolved, but they will concern questions of de-tails which we cannot ask at present becausewe do not know enough.

It was discovered during World War IIthat the sun was the source of occasionalradio noise strong enough to interfere withradar operations. This discovery has led toa new branch of solar astronomy which we havejust begun to explore in its general outlines.Instead of the 2-to-l range in wavelength ofthe visible spectrum, the observable radioemission from the sun covers a range of at

SMITHSONIAN CONTRIBUTIONS TO ASTROPHYSICS VOL.1

least 1,000 to^l. This provides a tool foranalysis in depth. Shortwaves in the centi-meter range originate near the visible surfaceof the sun. As we go to longer waves, thepoint of origin rises through the corona toheights of some hundreds of thousands ofkilometers for the longest waves of 10 metersor so. Like the light spectrum, the natureof the radio spectrum depends on physicalconditions at its point of origin and in thehigher layers traversed by the radiation inleaving the sun. The interpretation of muchthat has been observed is not yet clear, butthere can be no doubt that solar radio astronomywill occupy an extremely important place infuture research.

The design and construction of new solarinstruments at observatories all over the worldcontinues at an ever-increasing pace. Someof them, like the Babcocks' magnetograph,are the response to a particular need. Othersare new versions of tried-and-true devices,like spectrographs, for which there are sure tobe many uses not necessarily specified in ad-vance. In a field where we have still so muchto learn, we should plan our new equipmentwith care in order to avoid unprofitable duplica-tion.

In considering solar instrumentation, bothpresent and future, we should always keep iamind our real object: to find out somethingabout the sun. Specifically, we want the statis-tics, photometry, and physics of the solaratmosphere, and a good theoretical modelthat explains all the observations. We mustapproach the problem from every possibledirection. Routine daily survey programs,detailed studies of active regions, criticalobservations designated by theoretical models,new types of observations—all should be usedto solve the problem.

Each new instrument acquired should fill aspecific need. A careful investigation is oftenrequired to determine whether the plannedobservation is already being carried out else-where. Especially in planning a survey pro-gram, the observer should ask himself: ShouldI make an effort to take a few prominencemovies, when other observatories have collectedsuch gross quantities of film that they defyany sort of reduction ? Will my daily survey of

sunspots contribute significantly to the dailycoverage obtained elsewhere? Can the solaractivity be adequately described from presentsolar surveys? In other words, the observershould have a specific problem in mind whenbuilding the instrument and when taking theobservations.

In instrumental development there is roomfor every type of genius. To get an ingeniousidea and show that it works is not enough.The inventor, or someone else, must stilldevelop the instrument to the point where anycapable observer can use it routinely to obtainreliable data. A typical example of instru-ment development is the coronagraph. Lyotshowed that the coronagraph could be built,and he proceeded to build one. He used itwith the utmost skill to solve several importantcoronal problems, and, incidentally, demon-strated its capabilities. Since then a greatdeal of work has resulted in developing theinstrument, as a coronagraph, and in makingit more convenient to operate. It is hopedthat due to additional scale, resolution, photom-etry, and systematic use, Lyot's ideas willtell us even more about the sun than did hisfirst observations. In some cases the con-veniences themselves make the difference be-tween taking or not taking the critical observa-tion.

The development of the completed instru-ment is very often the result of a team effort.We need new gadgets and new ideas, but inthe solar field we also need people who takeexisting ideas and instruments and add suchvital factors as photometric standards, directphotoelectric photometry, differential photom-etry, seeing cancellation devices, and conveni-ences that speed the operation.

As elsewhere, solar instrumentation is gettingso complex that we need specialists in the opti-cal, electronic, and mechanical aspects and indata reduction, who will cooperate with the solarphysicist to produce the final result. It is nolonger feasible for a person to be skilled in allfields necessary to the solution of a problem.He must work with a group, and the groupmembers must have mutual confidence in eachother. A great deal of intercommunication isessential. Finally, the group must be compe-tently directed and coordinated.


The value of some very simple solar observa-tional device in an educational institutionshould never be underestimated. An exampleis a prominence birefringent filter. The directcontribution to research on the sun may be smallor none, but the stimuli provided by such fa-cilities may be decisive in attracting capable stu-dents to solar research, which greatly needsskilled workers.

Until a few years ago solar research was main-ly concerned with broad average characteristics.One studied the spectrum without regard forsmall variations from point to point on the sun.The chromosphere was treated as a homogene-ous atmosphere of uniform temperature. Morerecently the importance of the local differenceshas been recognized and the tendency has beentoward a finer grained analysis, in which manysmall areas of the solar surface must be studiedseparately. The result is an enormous increasein the volume of information required, since wenow have to consider many small areas insteadof a single large one. Gathering data, reducingthem to useful numerical quantities, and analyz-ing the greatly increased volume of such nu-merical quantities are already acute problems,and will become more serious as time goes on.

In gathering data, the unaccustomed pinchof the basic limitation of information containedin the available light from the sun is alreadybeing felt. A telescope receives light from agiven area on the sun in the form of discretequanta at an average rate of n per second, in apurely random fashion. This means that thenumber of quanta collected by the telescope ina given second differs from n in a random fash-ion, with a mean percentage error of -j=-

Hence a 1-second observation can at best de-termine the brightness of the area with thispercentage accuracy, and the larger n is, themore accurate the observation. When we tryto observe smaller and smaller areas of the sunwe are, in effect, reducing n in proportion, andthe fundamental errors of observation increase.Unfortunately we cannot make up for this bysimply observing for a longer time. Seeing isnever good enough to yield an undistorted andunsmeared image of the sun for more than afraction of a second, and the longer our observa-tion the more our small area becomes confused

with light from neighboring areas. The idea ofa scarcity of sunlight is perhaps a novel one, butmay not seem so surprising when we considerthat we would like right now to take the lightfrom an area of 0.01 square millimeter in a300-mm solar image, spread it out into a visiblespectrum nearly a hundred yards long, andthen determine the intensity in an area of 0.01square mm of this spectrum. The one thing wecan do is to increase efficiency in the use of thequanta we receive. The photographic plateutilizes only one in a hundred of the incidentquanta. Modern photoelectric surfaces areabout 10 times as efficient, and we have alreadybegun to take advantage of this improvement.The problem of the future is to find still moreefficient photoelectric surfaces, and to developaccurate and convenient methods for recordingcontinuously the photoelectrons emitted fromeach point of the surface.

Once the observational data are secured, theproblem of reducing them to numerical quan-tities must be faced. This has always been alaborious process, and the vastly greater volumeof data now required demands a radically newmethod of automatic reduction. The mag-nificent spectra coming from the new vacuumspectrograph at the McMath-Hulbert Observa-tory emphasize the problem decisively, and Itake them as a clear-cut example. TheMcMath-Hulbert observers find that both theFyaunhofer lines and the solar continuumshow marked variations in the directionperpendicular to the dispersion. These varia-tions are due to spectroscopic differences inthe small regions imaged on the slit of thespectrograph. They consist in wiggles andchanges of width in the lines, and changes inthe brightness of the continuum. Thus thespectrograms have variations which must bemeasured in two directions instead of one, alongthe spectrum and across the spectrum. Nowsuppose we want to exhaust the informationcontained on such a plate over a wavelengthrange of only 5 angstroms. Let us say that theslit was 1 cm long, that the resolving power ofthe telescope was 0.2 mm along the slit, andthe resolving power of the spectrograph was0.01 A. Complete information, therefore, callsfor the determination of the light intensity at500 points along the spectrum for each of 50

s SMITHSONIAN CONTRIBUTIONS TO ASTROPHYSICS

points along the slit. In other words, we haveto make 25,000 intensity determinations for acomplete analysis, instead of the 500 that wouldsuffice if there were no appreciable variationsalong the slit. In this instance the newdimension has multiplied the labor of analysisby a factor of 50. The information is there inthe plate, but the labor of extracting it hasbecome prohibitive.

The case of the wiggly lines is only one ex-ample of the trend toward the study of finerdetails on the solar disk, in the chromosphere,and in the corona. The result is an unprece-dented mass of information in the coded formof photographic density, magnetic variationsin a tape, or some other analog record. To beof any use, these records must be decoded andpresented in numerical form. Once this hasbeen accomplished, we are confronted withendless pages of numbers, all ready for interpre-tation by the unhappy theoretician. Classicalmethods of data handling will be quite in-adequate to utilize more than a tiny fraction ofthe information contained in the observations,and future progress depends very heavily onfinding the means for breaking this bottleneck.

These problems of data reduction and analysison a large scale are by no means limited tosolar astronomy. Industrial and military needs

have stimulated the development of informationtheory, the science of extracting the maximumpossible information from a minimum of rawdata in the least time. Devices for automati-cally performing various standard operations indata reduction, and computers which handlethe numerical results with unimaginable speed,have been devised. The theoretical and phys-ical machinery for the solution of the astronom-ical problem are now at hand, and the comput-ers at least are already vigorously at work forastronomers in several fields. But as yet theautomatic reduction of astronomical data tothe point where they are ready to be fed intoa computer has hardly been touched. Theproblem for the immediate future, therefore, isto develop the devices for bridging the gapbetween the output of a telescope and the inputto a train of automatic reducing equipmentalready available. The output of such a trainis numerical information, ready either for thecomputer or for direct analysis by the astro-physicist if the quantity is manageable. Theimportance of this problem should be em-phasized. Unless it can be solved the effec-tiveness of many of the instrumental develop-ments for solar research will be limited to afraction of their real potential.

Astronomical Seeing and Scintillation1

By Geoffrey Keller

Anyone who has made observations with largetelescopes is only too painfully aware that themost serious limitations to their performance areusually connected in one way or another withproblems of astronomical seeing. Images ofstars are much larger than one would expectthem to be were their size determined only bythe aberrations of the telescope itself. Detailsof extended objects, such as the moon and thesun, are badly smeared when the seeing is poor.As a case in point, Baum (1955) gives the aver-age image diameter seen with the 200-inch tele-scope at about 2.5 seconds of arc, whereas in theabsence of bad seeing conditions the perfectlyadjusted telescope should give images of di-ameter of the order of 0.04 seconds of arc. Thetotal amount of light received from a bright starby a telescope of moderate size also fluctuatesin an irregular fashion, and in a manner whichcan easily be shown to be due not to intrinsicfluctuations in the star but to the fact that thestarlight must pass through regions of theearth's atmosphere wherein the density variesirregularly. Protheroe (1955a) has shown thatwith a telescope of 12% inches aperture thefluctuations in brightness of a bright star willaverage around 10 percent of the mean, and willhave frequencies ranging up to a thousand cyclesper second. In pioneering investigations in thisfield Mikesell, Hoag, and Hall (1951) obtainedsimilar results.

Similar effects are encountered in radio as-tronomy, and here also the responsible agencyis a region of nonuniform index of refraction inthe earth's atmosphere. Whereas in the case ofoptical scintillation the principal effects appar-ently are caused in the troposphere and lowerstratosphere, the majority of the effects on radio

1 "Scintillation" is used here to mean only the changes with time ofthe brightness of the telescopic image of a star or other source. "Seeing"is construed to mean the degree of change in the position, shape, or sizeof a stellar linage or of a detail of an extended object.

1 Perkins Observatory, Delaware, Ohio.

waves seem to be caused by nonuniform electrondensities in the ionosphere. Ryle and Hewish(1950) have observed fluctuations in the inten-sity of radio sources at meter wave lengths ofthe order of 10 percent and fluctuations in theapparent positions of the radio objects of theorder of a minute or so of arc. Such variationsgive rise to serious observational problems, and,though some observational uncertainties due toscintillation and bad seeing may be overcome bymaking repeated measurements, the loss in effi-ciency is very great.

The most extensive efforts that have beenmade to date to minimize the effects of theearth's atmosphere on astronomical observa-tions have consisted of more or less empiricalcomparisons of various possible sites for estab-lishing new observatories. When a very largeinvestment of this sort is to be made, it becomeseconomically feasible to spend a considerablesum on extended tests. Such tests were madeprior to the construction of the Hale telescopeat Mount Palomar, and currently a new seriesof tests is being made in connection with asearch for a site for a proposed American Ob-servatory. Such surveys generally use tele-scopes and record the fluctuations in size andbrightness of a star such as Polaris (chosenfrequently because of its relatively fixed positionin the sky).

Another potential means of reducing seeingeffects is through the use of instruments mount-ed on high-altitude balloons, rockets, and even-tually extraterrestrial satellites. These methodsmight eliminate our seeing problems, but beforeadopting them we should know, for example,how high the instrument platform must bein order to be safely above the troublesomelayers of the atmosphere. In the case of aradio telescope, the answer is likely to be ofthe order of hundreds of miles. There areobvious difficulties with such an approach

9

10 SMITHSONIAN CONTRIBUTIONS TO ASTROPHYSICS VOL. 1

when we attempt to apply it to large telescopes,so that it will probably be many years or evencenturies before the bulk of astronomicalobservations can be made at extra-atmosphericstations.

Having picked a site where the disturbancescaused by intrinsic seeing and scintillation areat a minimum, the astronomer is next interestedin knowing how best to design his telescopeand observatory so that thermal disturbancesin the air caused by their structure may beminimized. A number of theoretical and ob-servational studies of the behavior of scintil-lation have been made at the Perkins Observa-tory that seem to point more and more to theconclusion that the local thermal irregularitiesof the air may be the predominant cause ofbad image distortion (as distinguished frombad scintillation). The work of the astronomersat Paris and Haute-Provence seems to haveled them to similar conclusions, so that thenew 193-cm telescope for the observatory atHaute-Provence is being very carefully de-signed to minimize the temperature irregular-ities in the air in the tube and just outsidethe dome (see Couder, 1953). Should theseefforts be successful, a major improvement inthe performance of large telescopes may result.

The importance of minimizing irregularrefraction in the gas inside of auxiliary astro-nomical equipment has been well demonstratedby the experience of observers with the new52-foot vacuum spectrograph at the McMath-Hulbert Observatory (McMath, 1955). Thesteadiness and sharpness of the spectral lineswere greatly increased as the spectrograph wasevacuated. Doubtless there are many otherpieces of auxiliary equipment whose perform-ances could be greatly improved by a similarprocedure.

Although the writer knows of no successfulinstruments, so far devised, to compensate for(rather than eliminate) the effects of seeing,some have been envisioned and might well beinvestigated (Babcock, 1953). Any device whichattempts to compensate for seeing must de-pend, first, on knowledge of what sort of opticalirregularities exist in the air; this requires studyof the distortion in the wave fronts in starlightwhich has just passed through a layer of irregu-larities. The device must then make, presum-

ably by electro-optical methods, compensatingadjustments in tie optical system in the tele-scope. Thus, if the air in a portion of the lightpath is unusually warm, the corresponding indexof refraction will be abnormally low and thevelocity of light abnormally high. As a result,a wave front tends to advance farther than itshould as it passes through this region. Thecompensating device attempts, in effect, tempo-rarily to move the surface of the primary mirrorfarther back in the area where this portion of thewave front is reflected, so that after the reflec-tion of the wave front the irregularities havebeen ironed out. Methods of this type havecertain drawbacks. For example, we must ob-serve objects that are sufficiently bright to giveenough photons per second to activate thedevice which controls the compensator; yet,enough photons must be left over to makepossible the primary astronomical observation.

In the preceding paragraphs we have lookedat the problems of astronomical seeing from thepoint of view of the typical astronomer, towhom seeing is an unmitigated evil that mustbe avoided or removed by whatever means cometo hand. To the atmospheric physicist andmeteorologist, the situation may be quite thereverse. The fluctuations in the light of abright star, known as scintillation, are known tobe caused by the motion across the telescopeobjective of a complex pattern of lights anddarks, known as the shadow pattern or shadowbands. This shadow pattern is caused by thepassage of starlight through atmospheric irregu-larities which must occur at a considerableheight. These irregularities diffract the lightand cause interference and reinforcement of thewave front at various points along the ground.A fairly complete theory of the relation betweenthe atmospheric irregularities and the patternof lights and darks in the shadow pattern hasbeen developed by van Isacker (1953). With areliable means of studying the structure of theshadow pattern, much may be said about thepattern of the density variations in the air itself.A suitable method for observing the shadowpattern has been developed by the author(1955) and by Protheroe (1955b). There seemsto be considerable promise that by the use andextension of this method we might be able toobtain considerable information about the

NEW HORIZONS IN ASTRONOMT 11

nature of clear air turbulence, its motions, itsheight, and how these quantities vary withmeteorological conditions. At the same time,more empirical studies of the relationshipbetween scintillation and wind velocity arebeing made (Gifford and Mikesell, 1953;Gifford, 1955; Protheroe, 1955a; and Mikesell,1955).

Studies of the scintillation of radio objectsare also being made, and much informationabout the structure of the ionosphere is beingobtained (Pawsey, 1955).

To summarize, the study of astronomicalseeing and scintillation is of vital interest tothe astronomer in the same sense that studiesof optics and radio technology are of importanceto him. There is much to be learned about thesubject, and there is considerable promise thatwith more understanding will come ways andmeans of greatly improving the performance oftelescopes. Although seeing may be a nuisanceto the astronomer, it also is a tool for makingboth fundamental and routine studies of theearth's atmosphere.

A few observatories in this country andabroad are making studies of both kinds.Much exciting work remains to be done and itis very much to be hoped that additionalastronomers can be persuaded to take aninterest in the subject.

References

BABCOCK, H. W.1953. Publ. Astron. Soc. Pacific, vol. 65, p. 229.

BAUM, WILLIAM A.1955. Sky and Telescope, vol. 14, pp. 264, 330.

COUDER, ANDRE.1953. Publ. TObe. Haute-Provence, vol. 2, No. 53.

GIFFORD, FRANK, JR.

1955. Bull. Amer. Meteorol. Soc., vol. 36, p. 35.GIFFORD, F., AND MIKESELL, A. H.

1953. Weather, vol. 8, p. 195.KELLER, G.

1955. Journ. Opt. Soc. America, vol. 45, p. 845.Also, Contr. Perkins Obs., ser. 2, No. 5.

MCMATH, R.

1955. Sky and Telescope, vol. 14, p. 372.MIKESELL, A. H.

1955. Publ. U. S. Naval Obs., ser. 2, vol. 17, pt. 4.MIKESELL, A. H., HOAG, A. A., AND HALL, J. S.

1951. Journ. Opt. Soc. America, vol. 41, p. 689.PAWSEY, J. L.

1955. Radio star scintillation due to ionospherefocusing. Physics of the ionosphere.The Physical Society, London.

PROTHEROE, W. M.

1955a. Contr. Perkins Obs., ser. 2, No. 4.1955b. Journ. Opt. Soc. America, vol. 45, p. 851.

Also, Contr. Perkins Obs., ser. 2, No. 6.RTLE, M., AND HBWISH, A.

1950. Monthly Notices Roy. Astron. Soc. Lon-don, vol. 110, p. 381.

VAN ISACKER, J.

1953. Publ. Inst. Roy. Me'teorol. Belgique, ser. B,No. 8.

SpectroscopyBy Charlotte E. Moore

The outlook for future research in spec-troscopy is highly promising. The rocketspectroscopy of the solar spectrum, for example,extends our horizon to the shortwave region,and thus provides a far more complete pictureof solar spectroscopy than ever before. TheO vi solar emission lines and other observationsof equally far-reaching significance demonstratethe astrophysical importance of further study ofhigh-ionization spectra in the laboratory (John-son, Malitson, et al., 1955). Ultraviolet stand-ards of wavelength must be established in theshortwave region, and more term values mustbe worked out, particularly in spectra of themore abundant elements. Fe iv is an outstand-ing example of a neglected spectrum urgentlyneeded by the astrophysicist. The coronalspectrum, the nebular spectra, and now thespectra of distant galaxies, whose lines areshifted further and further to longer waves, allshow that further study of ultraviolet spectra isextremely important.

In the longwave region, beyond the photo-graphic range, the development of suitable heatdetectors has extended the astrophysical vista,as well as that of the laboratory, to the farinfrared, and microwave spectroscopy hasextended it still further. With the astrophysi-cal range from Lyman alpha (1215 A) to the21-cm wavelength of hydrogen, our studies canno longer be limited to selected regions of the"visible" spectrum.

Among high-dispersion stellar spectra thesunspot spectrum is one of the most challenging.Hundreds of spot lines have not yet beenmeasured; many of them are doubtless ofmolecular origin. Furthermore, no homo-geneous set of sunspot spectrograms exists.During the coming sunspot maximum, high-dispersion sunspot spectrograms should betaken with the same spot, over the whole

1 National Bureau of Standards, Washington, D. C.

spectral range. This is a taxing but vitalprogram.

Our knowledge of stellar spectra is far fromcomplete. Any study of abundances requirescorrect identifications of the lines, together witha knowledge of the excitation and ionizationenergies of the separate atoms and ions.P. W. Merrill has recently reported that in theSe star, R And, there are some 275 unidentifiedabsorption lines between 4000 A and 6900 A.J. Greenstein has observed several hundredunidentified lines in v Sgr. (see Merrill, 1956).In the solar spectrum about 30 percent of thelines are still of unexplained origin. These andother similar examples demonstrate how muchwork is yet to be done in spectroscopy. Thisneed is reflected in the persistent demand,among astronomers, for a new, large, revisedmultiplet table that extends from the farultraviolet to the extreme infrared.

These demands impose upon the laboratoryspectroscopist the task of observing andanalyzing more atomic spectra, and also ofselecting spectra to be studied that anticipatefuture astrophjsical needs. The use of modernsources, such as electrodeless discharges andinfrared detectors, is one of the most promisingapproaches to the subject. Earlier descriptionsof atomic spectra are not only lacking in faintlines but are incomplete, also, because of mask-ing by oxide bands. The electrodeless dischargehas the advantage in atomic work that it canbe controlled in such a way as to suppressmolecular lines. More than 1,200 Fe i lineshave been identified in the solar spectrum byprediction from the known atomic energy levels.A suitable laboratory source of excitation woulddoubtless excite these and many more faintlines. Kiess has already observed a numberof the predicted Fe i lines in selected regions.EdleVs (1953) observations of CA n in theinfrared have resulted in the identification of

13

14 SMITHSONIAN CONTRIBUTIONS TO ASTROPHYSICS

four new multiplets among solar lines. Formany familiar spectra there exists no completeor homogeneous set of observations. Modernequipment and techniques could be used togood advantage in improving this situation.

The spectra of the rare earths are the mostconspicuous group that needs study. Withinthe next 3 years it is planned to enlist theservices of all available experts in attackingthis difficult problem. Volume 4 of "AtomicEnergy Levels" will be devoted entirely torare-earth spectra. This program should pro-vide the astrophysicist with a wealth of newdata. In order to disentangle the intricaciesof these spectra, the assistance of theoreticalinvestigators to predict the positions of theterms of various configurations is needed.This problem leads to use of digital computersin solving large matrices. Here is a useful fieldof activity in theoretical problems that parallelsthe laboratory and astrophysical research.

The comments just given apply chiefly toatomic spectra, but a wide vista appears, also,in molecular spectroscopy. For example, theuse of controlled flames as a source makes itpossible to bring out certain band spectra andsuppress others. Precise laboratory measure-ments and analyses, particularly for diatomicmolecules of the more abundant elements, aresadly needed. The writer is continually receiv-ing requests for a critical compendium ofmolecular data paralleling the volumes on

"Atomic Energy Levels." In solar and spotspectra alone this subject is far from completedand promises interesting results in futureastrophysical research.

Spectroscopy embodies, also, the study ofline intensities, equivalent widths, and curvesof growth—all rewarding subjects for futureresearch, particularly with high-dispersion stel-lar spectra. Just as in the case of spectrumanalysis, measured laboratory intensities arerequired to increase our knowledge of abund-ances of elements. This phase of the workmust be carried on for years to come.

The door is open and the opportunity is richfor the student interested in spectra. Themore unusual examples of astrophysical identi-fications—nebular lines, coronal lines, solaremission lines of high ionization, and technetiumlines in the S-stars—all challenge the astro-physicist and in themselves more than justifyactive financial support of spec troscopic researchin the laboratory, in theoretical fields, and inthe observatory.

ReferencesEDL£N, B.

1953. Accad. Naz. Lincei Conv. Sci. Fis. Mat.Nat., vol. 11, p. 56.

JOHNSON, F. S.; MALITSON, H. H.; PUBCELL, J. D.;AND TOTJSET, R.

1955. Astron. Journ., vol. 60, p. 165.MERRILL, P. W.

1956. Carnegie Inst. Washington, Publ. 610.

Radial VelocitiesBy R. M. Petrie1

The measurement of the line-of-sight compo-nent of stellar motion ranks among the earlyachievements of stellar spectroscopy. Theprinciple was announced by Doppler in themiddle of the 19th century and was extendedand elaborated by a number of astronomers andphysicists soon after. The first applicationswere made by visual methods and are now ofhistorical interest only. The subject did notattain full stature until the introduction ofphotographic methods during the closing yearsof the century, but it then became at once ahighly successful and valuable technique ofastronomy. Although radial-velocity work isthus about 60 years old, it is still of prime im-portance and promises to remain an indispens-able tool in many branches of astronomy.The pursuit of this work is not successful with-out a considerable expenditure of patienteffort, but the student may look to it to bringsubstantial rewards for his labors. He may,with profit, reflect upon the prophetic words ofHuggins: "It would scarcely be possible, with-out the appearance of great exaggeration, toattempt to sketch out even in broad outline themany glorious achievements which doubtless liebefore this method of research . . . . "

In this article attention will be directed topresent activities and some probable futureneeds. The development of our subject fromthe beginning, outlined above, is described inthe first part of Campbell's "Stellar Motions,"published in 1913.

Spectrographs

After the introduction of photography the mainquestion to be answered was how accurately thespectrograph could record the positions ofstellar absorption lines with reference to someterrestrial comparison source. The Doppler

1 Director, Dominion Astrophysical Observatory, Victoria, BritishColumbia, Canada.

shift is always small and great care is necessaryto avoid spurious instrumental shifts such as,for example, those introduced by mechanicalflexure, temperature variations, and faultyoptical components. The development of re-liable instruments such as those at Potsdam,Mount Hamilton, or Victoria may be studiedin an article by Eberhard (1933). The mainproblems of design have now been solved, andthere exists a considerable number of instru-ments capable of yielding accurate radialvelocities.

Some problems of instrumentation remain asa challenge to the student. The following areexamples of these.

Gratings.—The development of the "blazed"grating of high efficiency opens a field for ap-plication to radial-velocity work in the visualregion. Spectral types of F 5 and later may bestudied effectively in this way, which offers twoimportant advantages. First, the spectra aresimpler in the visual region, so that one canminimize the troublesome problem of using"blended" lines; and, second, many stars arebrighter in the visual than in the usual photo-graphic region. A successful grating spectro-graph has been constructed and used by Merrill(1931), and of course the powerful coud6spectrographs of the largest telescopes employgratings as the dispersive units. There is stillroom, however, for experiments in the mosteffective use of gratings in radial-velocity workwith moderate dispersion.

Long-focus instruments.—The great power ofmodern coude spectrographs lends itself well toradial-velocity investigations requiring the high-est precision. It is not yet known with cer-tainty whether the long-focus collimator andwide-slit spectrograph fed by a combination ofthree to five mirrors realizes the potential ac-curacy of the very high dispersion. We wantto know, on the one hand, whether the wide

15


slit (relative to the stellar image) has limitedthe accuracy because of the introduction ofguiding errors; on the other hand, we want toknow whether the multimirror system has pre-served full "identity of source" between stellarand comparison spectra. Experimental workto answer these questions needs to be under-taken as a preliminary to radial-velocity workof the highest precision. As an example of theinherent possibilities, one may refer to Adams'(1941) spectroscopic determination of the solarparallax.

Comparison spectra.—Present practice gen-erally is to use an arc, or spark, in air, with themost convenient metals being iron or titanium.These sources are fairly satisfactory but are notentirely constant unless a number of precautionsare taken. It is possible that a better sourcecan be found in some modification of a dischargetube such as the hollow cathode type recentlydescribed (Edlen, 1955).

The greatest obstacle to the study of stellarmotions through radial velocities is the slowrate at which spectrograms and measures ac-cumulate. The slit spectrograph, while accu-rate, is very wasteful of starlight and ordinarilyutilizes only a few percent of the incident light.As a result we are, for example, almost totallyignorant of the radial velocities of stars ofapparent magnitude 10 and fainter. Fiftyyears of work have yielded velocities for some15,000 stars, but this sample is much too smallfor galactic studies, and current progress is slow.A major effort and a new attack are obviouslyrequired if the study of stellar dynamics is notto be hampered by lack of observations.Fortunately there is now some hope of im-proving matters through two different currentdevelopments.

Slitless spectrographs.—The idea of removingthe spectrograph slit and collimator and thusphotographing with one exposure the spectra ofof a field of stars is an old one. An account ofthe early experiments with objective prisms isgiven by Millman (1930).

Until recently the objective-prism work fellshort of its goal because of the presence of fielddistortions and instrumental errors exceedingmany times the Doppler shift. However aseries of important experiments by Fehrenbach(1947, 1948) have resulted in the development

of an objective prism free from distortion, andstudies by Treanor (1948) and Schalen (1954)have verified and extended this result. Awholesale accumulation of radial velocities offaint stars is now in prospect; continued experi-mentation and application are sure to bringrich rewards. Already the "mean errors" havebeen reduced to about ± 6 km/sec, and this issufficient for application to many problems ingalactic motions and structure.

Image convertors.—Another promising de-velopment is the image convertor, which iscapable of high efficiency and may ultimatelyreplace the use of photographic plates in re-cording stellar spectra (Argyle, 1955). Con-siderable experimentation is now going on toimprove this technique and to apply it to stellarspectroscopy. The subject is well worth theattention of the student who wishes to makea contribution to radial velocity work.

Measurement

Doppler shifts are, generally speaking, verysmall and their successful measurement callsfor painstaking care and experience. Thelabor of measuring the stellar and comparisonlines on several hundred spectrograms is con-siderable and often hampers the progress of aprogram.

The measurements were made originallywith the aid of a direct-vision microscopemounted above a traveling micrometer, or, inthe Hartmann comparator, spectra werematched against suitable "standards," againwith the aid of a microscope. The fatigueaccompanying this process was lessened, someyears ago, by the introduction of projectionmethods (Petrie, 1937; Redman, 1939). Afurther improvement was made at Victoriawhen a new type of projection instrument(Petrie and Girling, 1948) allowed one tomeasure the Doppler shift directly upon theprojected image and, by virtually eliminatingthe "reduction," gave a substantial increase inefficiency. Indeed, so useful is this projectioncomparator that the measurement and reduc-tion of spectrograms is no longer a seriousobstacle to the completion of a heavy radial-velocity program.

The obvious advance to be made now is thedevelopment of an automatic machine that

NO. 1 NEW HORIZONS IN ASTRONOMY 17

will remove the subjective errors and allow animpersonal definition of the precise position ofa stellar line relative to the comparison spectrum.A machine of this sort has been built by Johnson(1949), and another of different type by Hos-sack (1953). In spite of these ingeniousmachines there is still room for experimenta-tion and development. The technical aidsalready exist and it is fairly certain that thepresent generation is the last that will be re-quired to measure Doppler shifts with handand eye. A real triumph awaits some studentequipped with a sound knowledge of modernelectronic aids.

A word of caution may be permitted here.An automatic machine does only half the jobif it merely presents one with the measuredpositions of stellar and comparison lines. Thecalculation of a radial velocity from the enginereadings requires nearly as much time as theoriginal measurement. The ideal machine,therefore, will measure the shift of the spectralline from its zero-velocity position and so allowthe rapid deduction of the radial velocity.

Wavelengths

The radial velocities deduced from the measuredpositions of spectral lines are, of course, verysensitive to the accuracy of our wavelengths.We recall that in the photographic region anerror of 0.1 A is equivalent to 7 km/sec. Afirm knowledge of wavelengths is a prerequisiteto a significant knowledge of motions.

The practical difficulties are considerable.Radial velocity programs usually force one toemploy a spectroscope of low dispersion andpurity; with such an instrument all spectralfeatures in types later than A 0 are blends oftwo or more lines. The effective zero-velocitywavelength of a feature depends upon therelative intensities of the blending lines, thespectrographic resolution, and the amount ofstellar rotation and other line-widening in-fluences. Even in early-type spectra difficultiessometimes arise because of blends, stellarrotation, and Stark effect.

At Victoria a good deal of attention has beendevoted to the matter of effective wavelengthsand it is believed that satisfactory methods andvalues have been established (Petrie, 1946,1947, 1948, 1953: McDonald, 1948; Wright,

1952). The problems must be faced anewwith each new spectrograph and vigilance isnecessary to ensure that the deduced radialvelocities are, in fact, motions and not spuriousdisplacements arising from uncertain, or in-correct, wavelengths.

In spite of the work already done on thissubject, more is required. We do not, forexample, have a satisfactory set of standardwavelengths for O-type spectra and cannot,therefore, be entirely certain of their motionsin star "associations" nor of their apparent"red shift" as members of galactic clusters.And, at the other end of the temperature se-quence, we lack effective wavelengths for low-dispersion spectra of N stars. This is of someimportance since we need to know the spacemotions of a larger number of these (apparently)faint objects.

Applications

The application of radial velocities leads us intothe subjects of stellar dynamics, galactic struc-ture, binary stars, etc. These topics arebeing discussed elsewhere in this compilationand the writer will therefore merely point outsome of the possibilities awaiting the astronomerwho devotes himself to radial velocities. Theworker who undertakes to supply the funda-mental observations of stellar motions can besure that his contribution is of permanent valueto astronomy. He must therefore be patientand content to see his work gradually increaseas his programs and plans are completed.

The general problems of stellar motion re-main with us: solar motion, star streaming,galactic rotation, and mean velocities in differ-ent parts of the galaxy. The need is alwaysfor more observations and observations of moredistant stars. For a general description ofpresent-day need one should refer to the reportof the International Astronomical Union Sym-posium on "Co-ordination of Galactic Re-search" (Blaauw, 1955).

The measurement of galactic rotation is ofprime importance. We require a more accurateknowledge of Oort's constant A in the neigh-borhood of the sun. Current programs of Bstars at the Radcliffe and Dominion Astro-physical Observatories will supply about 1,000new velocities, mostly within a distance of

18 SMITHSONIAN CONTRIBUTIONS TO ASTROPHYSICS VOL.1

1,500 parsecs of the sun. These, together withexisting observations, should give a gooddetermination of the coefficient of differentialrotation in our part of the galaxy.

The more distant regions are quite im-perfectly studied. With the recognition ofspiral arms about 2,000 parsecs away, itbecomes an exciting task to discover whatvelocity effects exist at the edges of the armsand between them. This means we mustobserve large numbers of spiral-arm objects—early B stars, Cepheids, and supergiants oftypes B 8 to A 3. At the same time we mustmeasure the speeds of the interarm populationwhich will probably consist of stars of Popula-tion II together with a number of stars thathave wandered out of the spiral arms. Ob-servations of the early-type stars will alsoreveal the motions of the interstellar matter.

A problem of considerable interest is thestellar velocity distribution perpendicular to thegalactic plane. This may be found from radialvelocities of stars near the galactic poles. Theobservation of the brighter stars of types A 0-F 5within 10° of the north Galactic Pole is now near-ing completion at Victoria. A companion programof fainter stars is proceeding at Mount Hamilton.In this branch of stellar motion work much re-mains to be done. Extensive observations arerequired among the later-type stars of the diskpopulation and, on the other hand, of the faintblue stars of (probably) Population II. Fromthe velocities of the disk population we candetermine the density of matter in the galacticplane region, in the neighborhood of the sun.

The above proposals involve observations ofmany stars and are rather long-term projects,but much may be done in other directionswhere shorter programs will bring valuableresults. Some of these are suggested verybriefly in the following paragraphs.

Star clusters and associations.—These fascinat-ing groups are of first importance in studies ofstellar luminosity and evolution. Radial-veloc-ity observations are needed to verify clustermembership and to calculate space motions.Careful and extensive observations of the"aggregates" and expanding associations mustbe made before we understand the kinematicalproperties of these groups. The study of

clusters and associations is sure to yield im-portant and exciting information.

Stellar atmospheres.—Recent work on super-giant stars and special eclipsing stars, such as31 Cygni and f Aurigae, shows how radial-velocity measures reveal the existence of atmos-pheric motions. This is a little-explored fieldand requires study with highest possible disper-sion. Comparisons between giant and main-sequence stars of the later spectral types willbe valuable in promoting knowledge of thestability of stellar atmospheres.

Variable stars.—Experience has shown thatvariable stars should be studied with simul-taneous spectroscopic and photometric meas-ures. The observations of Sanford (1952) andAbt and Sanford (1954) on the Cepheid variableW Virginis and of Struve (1954) on BWVulpeculae are good examples. The work tobe done here is limitless, considering the greatrange of stellar variation. To be most fruitful,the highest possible dispersion must be used.

Binary stars.—The problems of binary starsare legion, but perhaps no other field offers morechallenges calculated to advance our knowledgeof stars.

Visual binaries should be observed continu-ously so that, ultimately, the combination ofspectroscopic and micrometric data will tell usthe dimensions and luminosities of the compo-nents. Similarly, eclipsing binaries should beobserved assiduously, especially those withdeterminate light elements and those showingthe spectra of both components. Our knowl-edge of stellar masses, radii, and luminosities,as well as further extensions and verification ofthe mass-luminosity relation, must come fromthis work. Radial velocity observers will findit a rewarding task.

The bewildering array of spectroscopic anom-alies exhibited by the close binaries has beenbrought to our attention by Struve (1950).The observation and interpretation of radial-velocity curves are an endless challenge. Onesuspects that the most potent observationalaids will be required to advance our under-standing of these spectacular phenomena.

Stellar satellites.—The discovery of planetarysystems is probably just within the powers ofthe most precise radial-velocity measures thatcan be made with existing equipment. This

NEW HORIZONS IN ASTRONOMY 19

has been shown by Struve (1952), who proposedthat a search be made. It may be that a majorastronomical discovery will result from themeticulous observation of a number of ournearby and bright stars.

ReferencesABT, H. A.

1954. Astrophys. Journ., Supplement, No. 3.ADAMS, W. S.

1941. Astrophys. Journ., vol. 93, p. 11.ARGYLE, P. E.

1955. Journ. Roy. Astron. Soc. Canada, vol. 49,p. 19.

BLAATJW, A.1955. Int. Astron. Union Symposium No. 1.

Cambridge University Press.EBEBHARD, G.

1933. Handbuch der Astrophysik, vol. 1, pt. 1,p. 299.

EDLEN, M. B.1955. In Draft reports of the International Astro-

nomical Union Dublin meeting, August-September 1955, p. 104. CambridgeUniversity Press.

FEHRENBACH, C.

1947. Ann. d'Astrophys., Paris, vol. 10, p. 306.1948. Ann. d'Astrophys., Paris, vol. 11, p. 35.

HOSSACK, W. R.1953. Journ. Roy. Astron. Soc. Canada, vol. 47,

p. 195.JOHNSON, H. L.

1949. Astron. Journ., vol. 54, p. 190.MCDONALD, J. K.


MERRILL, P. W.1931. Astrophys. Journ., vol. 74, p. 188.

MlLLMAN, P.1930. Circ. Harvard Astron. Obs., No. 357.

PETRIE, R. M.1937. Journ. Roy. Astron. Soc. Canada, vol. 31,

p. 289.1946. Journ. Roy. Astron. Soc. Canada, vol. 40,



p. 213.1953. Publ. Dominion Astrophys. Obs., vol. 9,

p. 297.PETRIE, R. M., AND GIRLING, S. S.


REDMAN, R. O.1939. Monthly Notices Roy. Astron. Soc. Lon

don, vol. 99, p. 686.SANFORD, R. F.

1952. Astrophys. Journ., vol. 116, p. 331.SCHALEN, C.

1954. Arkiv fur Astronomi, vol. 1, p. 545.STRUVE, O.

1950. Stellar evolution, chap. 3. Princeton Uni\Press.

1952. Observatory, vol. 72, p. 199.1954. Publ. Astron. Soc. Pacific, vol. 66, p. 329.

TREANOR, P. J.

1948. Monthly Notices Roy. Astron. Soc. Lon-don, vol. 108, p. 189.

WRIGHT, K. O.

1952. Publ. Dominion Astrophys. Obs., vol. 9,p. 167.

AstrometryBy K. Aa. Strand*

Astrometry deals with the space-time be-havior of the stars, and thus belongs to theclassical field of astronomical studies. Whilethe early investigations were directed mainlytowards establishing a suitable frame of refer-ence for determining the complex motions of theplanets, studies of the positions and motions ofindividual stars, as well as of the various stellarsystems, gradually developed.

If one examines the observational data accu-mulated in astrometry, the record is indeed im-pressive. During the past 150 years more than370 catalogs of meridian observations of starshave been published. We have monumentalastrographic catalogs like the Carte du Ciel, theAstronomische Gesellschaft catalog (AGK2),and the Yale zone catalogs, to mention only thelargest. The Yale General Catalogue of StellarParallaxes lists the distances for nearly 6,000stars measured since the beginning of this cen-tury, and the Burnham and Aitken double starcatalogs collect more than half a million microm-eter measures, the result of a century's work.

Considering this impressive observationalrecord, it is not surprising that a person whohas not worked in astrometry usually assumes,mistakenly, that most of the work has beencompleted. Our knowledge of such propertiesof the stars as their masses, luminosities, andmotions depends directly or indirectly upon theavailable astrometric data, and any improve-ments to be made in these parameters will de-pend upon further work in the field.

What future work is needed in astrometry tobroaden our general knowledge of astronomy?A whole series of problems confronts us. Theyrange in size from those that can be investigatedonly by close cooperation among many observ-atories to those that can be solved by a fewastronomers in a^single observatory; and in

'̂ Director, Dearborn Observatory, Northwestern University, Evans-ton, HI.

381014—56 3

complexity the problems range from setting upan accurate reference system of stars coveringthe entire sky to determining the parallax of anindividual star.

A series of conferences held during the pasttwo years at Groningen, Evanston, Pulkovo,and Brussels has resulted in general agreementthat, to solve many present-day problems, as-tronomers must undertake an international pro-gram of meridian circle observations. The pro-gram, starting in 1956, will consist of observa-tions of nearly 21,000 stars brighter than 9.1magnitude north of declination —5°. The totalnumber of observations, including those re-quired of the fundamental stars, will amount to260,000, since each star is to be observed 10times. Twelve observatories will participate,and by using timesaving devices in both theobservational and reduction phases of the pro-gram, they expect to complete it in 5 years.American astronomers should note with somepride that the U. S. Naval Observatory willcarry out a substantial part of the observationsand will also supply, on request, the reductionsto the day and to the equinox 1950 for all thestars observed. The U. S. Naval Observatorywill also compile the final catalog of positions forthese stars.

What information will this large undertakinggive us ? It will provide a reference system ofstars with greater density and precision thanany previous system; and it will make possiblethe solution of problems that could not besolved without it, some of which may be sum-marized:

1. The program will provide a reference sys-tem for the AGK3, the second photographic ob-servation of the stars in the Astronomische Ge-sellschaft Katalogs, which when compared withthe AGK2 will lead to accurate proper motionsfor approximately 180,000 stars brighter than11.5 photographic magnitude. Dr. Vyssotsky

21


of the Leander McCormick Observatory and hisassociates will supply spectra for approximately60,000 stars not previously classified.

2. About 11,500 of the stars in the referencesystem will be used in the Russian program onstellar proper motions relative to extragalacticnebulae.

3. The proper motions determined from theAGK2 and AGK3 will provide the systematiccorrections that are needed to make use ofmany of the older star catalogs.

4. The results from the AGK3 will providethe necessary data to reduce to absolute theproper motions obtained from a reobservation ofthe Carte du Ciel plates.

5. The stars from the AGK3 could be used asreference stars for the Lick Observatory pro-gram, which expects to determine proper mo-tions of stars with respect to extragalacticobjects.

It is gratifying to note that it is still possibleto undertake large-scale programs requiringcareful international cooperation, for which as-tronomy is conspicuous among the sciences.However, it should be observed that the burdenof American participation in the AGK3 restssolely upon the U. S. Naval Observatory.

Because the Lick program of deriving propermotions of stars with respect to extragalacticobjects will yield important scientific results, itshould receive the support necessary for its com-pletion. The Lick program will provide abso-lute proper motions of stars down to the 17.5photographic magnitude, resulting in data thatcan be expected to yield corrections to theadopted values of precession and lead to newdeterminations of solar motion and galactic ro-tation. The most distant stars in the Licksurvey will have distances of more than 2 kilo-parsecs and thus reach beyond the region of themaximum circular velocity, as indicated by theOort model. This study should settle the pres-ent discrepancy between the data on galacticrotation, as obtained from proper motions andfrom radial velocities.

If the Lick survey is to attain the neededminimum accuracy, a time interval of approxi-mately 20 years must elapse between first- andsecond-epoch plates. This means that the com-plete repetition of the Lick plates will not begin

until 1967. How soon after that date the pro-gram will be completed depends upon the meth-ods of measurement and of reduction. It isquite clear, however, that the program is of suchmagnitude that it requires the utilization of thelatest developments in automatic or semiauto-matic equipment to complete it without pro-longed delay.

Since the value of the Lick program wouldbe much enhanced if it could be extended tothe southern sky which is not observable atthe Lick Observatory, serious considerationshould be given to extending the program tothe Southern Hemisphere. In view of thesmall number of observatories south of theequator and the magnitude of the problemsthat already confront them, this program prob-ably should be undertaken by an observatoryin the Northern Hemisphere. It remains tobe investigated whether a telescope identicalwith the Lick 20-inch Carnegie astrographshould be constructed or whether, as suggestedby Dr. Luyten at the Meeting of the Inter-national Astronomical Union in Dublin, alimited number of plates taken with the Brucetelescope would serve the purpose.

Photographic astrometry with long-focusrefractors is traditionally a field in whichAmerican astronomers have held the leader-ship; here, also, we face a series of problemswhich must be solved for further progress inastronomy.

It has long been known that the trigonometricparallaxes are affected by systematic errors;whether further statistical investigations of theexisting observations will provide additionalinformation on this problem remains doubtful.The systematic errors have, of course, the mostserious effect upon the small parallaxes; butthe absolute magnitudes of giants and sub-dwarfs, which are scarce in the solar neighbor-hood, are therefore derived from a relativelysmall number of small parallaxes and thus arealso affected by large systematic errors. Toimprove this situation it is essential that manyof the small parallaxes be repeated, and thatthe observational series for the individualstars be increased from the 20 plates or lessused in the very early work to 50 plates oreven more. These determinations would add


to our knowledge of the absolute magnitude ofthe stars just mentioned, and would also serveas important statistical material for furtherinvestigation of the systematic errors in trig-onometric parallaxes.

The stars of low luminosity in the solarneighborhood, discovered in the Luyten propermotion survey, are another group for whichwe need determinations of trigonometric paral-laxes. The great majority of these stars,late-type M-dwarfs, subdwarfs, and whitedwarfs, are considerably fainter than any wecan observe with the telescopes now being usedfor parallax work. These telescopes cannotreach stars much fainter than the 12 th mag-nitude, but most of the known white dwarfsare fainter than the 14th. For this reason,the commission on parallaxes and propermotions of the International AstronomicalUnion recommended at the recent meeting inDublin that observatories possessing largereflectors seriously consider a program ofdetermination of trigonometric parallaxes offaint stars.

Other aspects of the use of long-focus re-fractors are the positional study of the starswithin 10 parsecs by an extended series ofplates to discover perturbations for "singlestars," and the subsequent orbital analysisof these perturbations. Quite recently such astudy led to the visual discovery of a binary(Ross 614) in which the companion has only8 percent of the sun's mass.

When we examine the relation of astrometryto the double stars, we find a series of problemswhich deserve the attention of modern astron-omy. For the double stars within 20 parsecs,it is possible to derive stellar masses with anaccuracy of 25 percent or better; to attain thisaccuracy, however, visual and, whenever pos-sible, photographic observations of the relativemotions are needed as well as extended series ofplates from which the mass ratios and parallaxescan be derived with great precision. Much in-formation in regard to stellar masses has beenobtained in this way over the past 15 years, butmuch more is needed. This group of binariesconstitutes only about l/1000th of the approxi-mately 40,000 known binary systems. To ob-serve them all would clearly be an impossible

task; nevertheless, many of them must be ob-served if we wish reliable data on such questionsas statistical distribution of the orbital elementsin binaries and combination of spectra in phys-ical pairs. These are only two of the questionsthat have a bearing on stellar evolution andwhich we cannot answer by studying the smallgroup of known binaries within 20 parsecs ofthe sun.

Problems of membership and internal motionsin clusters will become even more important asthe time interval between early- and late-epochplates increases the accuracy. The early Cartedu Ciel plates and those taken with the largerefractors shortly after the beginning of thiscentury are especially important in thatrespect.

One of the most recent additions to astrom-etry, the study of the age of the O-B associ-ations from their expansion, deserves high pri-ority, and it is to be hoped that the large merid-ian programs in connection with the AGK3 willstill permit the stars selected for this study tobe observed with utmost precision. A similarprogram on Cepheid variables, to trace the po-sitions of the spiral arms in our galactic system,also deserves special attention.

Lack of manpower to cany out the many im-portant problems just outlined remains a prob-lem. We may summarize briefly what we re-quire in instrumental developments to solve thisquestion. The staff of the U. S. Naval Observ-atory has, over a period of years, introducedtimesaving devices that have increased thespeed as well as the accuracy with which thestars are observed, and because of the use ofIBM machines the reductions of the observa-tions are no longer lagging behind as in the past.

In photographic astrometry the use of auto-matic devices in guiding, plate transport, andtiming has resulted in increased accuracy incertain problems, but the real problem here is todevelop equipment that will effectively shortenthe time required for measurement and reduc-tion of the photographic plates. One measuringmachine, allowing semiautomatic measurementsand operated in connection with IBM machines,is already in use. Since the operating costs ofthis system are prohibitively high for most

2 4 SMITHSONIAN CONTRIBUTIONS TO ASTROPHYSICS VOL.!

American observatories, it seems desirable to tention. However, those outlined show that inlook into other recent technical developments spite of, or rather because of, the huge amountwhich might aid in reducing the time spent at of work done in the past, there are still manythe measuring machine without requiring a interesting and important problems on whichhighly expensive computing system. progress in astronomy depends; therefore, they

In this brief report it is not possible to men- deserve the attention of the active researchtion all phases of astrometry that deserve at- workers in astronomy.

Photoelectric AstronomyBy A. E. Whitford l

The past two decades have seen increasinguse of photoelectric methods in all branches ofastronomy where quantitative measurement ofradiation is important. This trend seemsdestined to continue. In the years ahead thephotoelectric cathode may become the mostwidely used radiation detector in the astronom-ical tool kit; indeed, it may supplant the photo-graphic plate in the same way that the platesupplanted the human eye 50 to 60 years ago.

A brief review of the links in the informationchain along which knowledge about the externaluniverse must pass will show the reasons forbelieving that photoelectric methods mayassume a dominant place. The light beam froma distant luminous source, which carries all theinformation the astronomer will ever haveabout that object, arrives at the surface of theearth after some losses in the interstellarmedium and in the earth's atmosphere. Theatmosphere adds an unwanted backgroundglow, and, at times of bad seeing, diffuses thebeam somewhat. The potential informationin the light beam suffers a loss and degradationbefore it enters the astronomer's instrumentthat are for the most part beyond his control.Some day the astronomer may get above theatmosphere, but he will not get beyond theinterstellar medium. Meanwhile, he mustmake sure that his earthbound instruments dothe best possible job of extracting all of theinformation remaining in the light beam.

The astronomical telescope has long sincereached a state of perfection where the tran-quility of the atmosphere, rather than thequality of the optics, determines the imagequality. The astronomer may, in theory,gather more information by increasing theaperture, but the engineering difficulties and thecost both rise steeply for any considerable in-crease in size over that of existing telescopes.

Wasbborn Observatory, University of Wisconsin, Madison, WJs

The final link in the information chain, theradiation detector at the focus of the telescope,is therefore the element which must be scruti-nized for possible lack of efficiency. If theefficiency is low, improvements can be countedas equivalent to increasing the aperture ofexisting telescopes.

The various practical detectors in actual usecan be evaluated by comparing them with ahypothetical ideal detector that recorded thearrival of each quantum, and did it with 100-percent efficiency for all accessible wavelengthsof the astrophysical spectrum. The photo-graphic plate cannot be given a very high markin such a comparison. Since it is a saturable,nonlinear device, no unique quantum efficiencycan be assigned. But in the range of densitiesfavorable to contrast-judgments by the humaneye, only about 1 quantum in 1,000 is recordedby producing a developable plate grain. Itsability to integrate a long exposure is the reasonfor its superiority over the eye in detectingfaint stars and recording low surface brightness;the eye makes its decision in about % second.Within that % second the eye, as shown by Rose(1948), is a better detector than the photo-graphic plate. The eye cannot, however,deliver the plate's objective record of thousandsof details or the extremely precise positionalinformation needed for astrometry.

The cesium-antimony cathode is the one mostwidely used in astronomical photometry. Ithas a useful range of 3100A to 6000A. Theaverage quantum efficiency at the peak near4400A is 12 percent, and exceptional tubes goas high as 25 percent. There is thus a superi-ority of a factor of the order of 100 over thephotographic plate in the initial informationmade available in a given time from a givenangular area on the sky. As presently used,only one chosen element or area may be meas-ured at a time; if photometric information on

25


many objects in one field is needed the photo-graphic plate is a great timesaver. But thesuperior quantum efficiency of the photo-electric cathode is the basic reason why itseems destined to play an increasing role inmeasurements of astronomical radiation.

In order for the favorable quantum yield tobe useful, there must be no instrumental back-ground or amplifier noise large enough todegrade appreciably the information containedin the released photoelectrons. Existing tech-niques succeed quite well in approaching thisideal and further efforts can be expected toproduce only small improvement. There is notmuch to choose between any of the threecurrently used methods of registering the outputof a phototube: current measurement with aload resistor, charge integration with a storagecondenser, and charge integration by pulsecounting. At the lowest intensities integrationmethods will probably find increasing use.

The future uses of the photoelectric methodin astronomical photometry are of course notentirely predictable. The only really safeprediction is that new and ingenious applica-tions of the method, not now foreseen, willsurely be devised. A projection of the pres-ently successful techniques suggests activityin three main areas. In general, these exploitthe basic advantages of the photoelectricprocess: high quantum yield, linear response,and wide spectral range.

Intensity comparisons: Magnitude systems

The range of intensity between the sun andthe faintest stars yet detected is of the orderof 1020. Important astronomical conclusions,such as the cosmic distance scale, depend onthe precision with which this range is covered.The superior accuracy of the photoelectricmethod rests in part on its higher quantumyield, in part on its linearity. Elimination ofthe step calibration needed by the nonlineardetectors such as the photographic plate notonly saves a great deal of time but also removesa fertile source of accumulated error. Anintensity range of 10* to 106 can be coveredwith one phototube and a single optical system.

The task of calibrating magnitude systemsis already largely a photoelectric province.

Photographic interpolation between photo-electrically established standards in the fieldof a star group such as a globular cluster(e. g., as by Arp, Baum, and Sandage, 1953)will continue to be used where photoelectricmeasurements of one faint star at a time wouldbe too time-consuming. But for transfer ofstandards from Selected Areas, photoelectricmethods appear to be needed for good accuracy.The time is rapidly approaching when everyobservatory which does any photometry atall will think of the plateholder and the photo-electric photometer as routine interchangeableattachments to the telescope, each to be usedas occasion demands.

Since photoelectric methods require the useof the telescope for only one object at a time,and the reductions are also quite time-consum-ing, magnitude measurements have thus farbeen restricted to specific classes of stars, suchas B-stars and Cepheid variables, and to theestablishment of standards in Selected Areasor star groups under special study. A deter-mined effort might provide instrumental andcomputing aids which would make mass pro-duction of magnitudes more economical. Thesemight include better coordinate mechanismsfor rapid identification and centering of stars,automatic subtraction of sky readings, auto-matic computation of extinction and correctionof the star signal for it, a logarithmic amplifierto give direct reading magnitudes, and directprocessing of telescope output onto data storagerecords such as punched cards. Then it mightnot be too enormous a task to do what Irwin(1955) suggests, and undertake to provideaccurate photometric data for all the stars,say, in the Henry Draper Catalogue. A massiveattack by one observatory could accomplishit; it would be a public-spirited endeavor, tobe viewed as a repayment of the scientific debtowed to previous generations for the patientdetermination and compilation of star positions,spectral types, radial velocities, and propermotions. I t would at least correct the presentincongruous situation where the simplest andmost fundamental datum about a star, itsbrightness, is often the least reliable entry inthe catalog.


Color systems and spectrophotometry

The advantage of the photoelectric cathodein color studies stems partly from its widespectral range, partly from its linearity. Thesimple ratio of the intensity through two colorfilters gives a good color index without the needof establishing two magnitude scales as in thephotographic method. Because the photo-electric color index comes from a differentialmeasurement, the internal accuracy is higherthan for the magnitudes themselves. A two-color index provides the minimum amount ofinformation about a star. The wide spectralrange of a photoelectric cathode permits threeor even five narrow-range colors with a cesium-antimony cathode. The range may be extendedto 10,000A or more with a cesium-oxide cathodeand the number of color filters extended to sixor seven. The next step is to scan the wholerange with an exit slit on a spectrograph. Theobserving time per star goes up because of thelimitation that only one element of the spectrumcan be observed at one time. High-resolutionphotoelectric scans of short portions of the solarspectrum have been made, and similar scans willundoubtedly be attempted for the brighteststars. Over the whole range, however, a re-solving power of about 350 gives a reasonableobserving time per star and adequate delineationof the main features of the continuous spectrum.

The color of a star is already recognized as anecessary part of a photometric determinationof its distance in order to evaluate the inter-stellar reddening. For early-type stars, at least,three colors on the U-B-V system of Johnsonand Morgan can give a measure of reddeningwhen the spectrum is not known. Extension ofthis powerful method to later-type stars, andto the longer wavelength portions of the spec-trum where the interstellar absorption is less,would be a great help in galactic and extra-galactic distance determinations.

Determination of the absolute energy distri-bution in normal Type I stars will be worthy ofconsiderably more work over the whole rangethat can be covered with existing photoelectriccathodes. From such studies can come bettermodel atmospheres, and from these, in turn,higher accuracy in quantitative studies of nu-clear abundances from line strengths. Com-

parison of these "normal" stars with Type IIstars and peculiar stars by multicolor or scan-ning methods should be most informative. Ex-amples would be subdwarfs, metallic line stars,and pulsating stars of all kinds. Differences ofcomposition or structure may be expected toleave clues in the continuous spectrum; accu-rate photometry will be needed.

Composite sources are another fruitful fieldfor multicolor studies. Globular clusters andgalaxies are examples. The resolving power indistinguishing between various syntheses of theover-all energy distribution from hypotheticalmixtures of known stellar types may never bevery high. But a combination of spectra, ofcounts of brightest stars of identifiable types,and of over-all energy distribution may lead tofairly definite conclusions. The diffuse natureof these objects precludes anything but lowresolution spectrophotometry by the scanningmethod, and for faint objects multicolor filterphotometry may be the best that can be done.Their low contrast with the sky can be handledcleanly by the differential subtraction methodsso easily applied in photoelectric photometry;and photoelectric accuracy will be needed.Photometry of distant galaxies for diameter andmagnitude (as suggested by Baum, 1953) andfor energy distribution (Whitford, 1953) willprovide evidence of the state of affairs at amuch earlier era in the evolution of the universe,and will thus be of interest to those concernedwith cosmogony.

Differential measurements

Light curves of eclipsing binaries providedone of the early demonstrations of the precisionattainable by photoelectric methods. Theaccuracy was due partly to the linear response,and partly to the differential comparisonof variable and comparison star, thus minimiz-ing variable factors in sky and apparatus.The fundamental information about starsderivable from precision light curves of eclipsingvariables will continue to offer attractive pos-sibilities to the photoelectric observer with asmall- or medium-sized telescope. For recon-naissance purposes other methods will continueto be of value, but the precision needed forderiving properties like limb darkening requiresphotoelectric accuracy.

2S SMITHSONIAN CONTRIBUTIONS TO ASTROPHYSICS VOL.1

Linear response is a great aid in all mannerof differential measurements because simplesubtraction is possible. Alternation of fieldsonce a minute or oftener is one way of doingthis. The flicker or a. c. method is another,and it may give a better elimination of unwantedextraneous factors. One example is the meas-urement of very faint stars or galaxies whichadd only a few percent to the sky radiationthat comes through the smallest usable focal-plane diaphragm. Nebulosity of low-surfacebrightness would be another example. Ex-tension of flicker methods to this type ofmeasurement would be a logical development.

Wherever it is possible to "put a tag" onthe wanted signal, it can usually be lifted outof an overwhelming background by differentialmethods, often involving a. c. amplification.One example would be the polarization ofstarlight (Hiltner, 1949, and Hall, 1950);another would be measurement of solar magneticfields (Babcock, 1953) where the modulatingelement again involves polarization. Theseflicker methods have been highly developed inradio astronomy where the signal may be 100to 1,000 times smaller than the unavoidablenoise. Further use of the technique in photo-electric astronomy seems likely.

Possible advances in technique

If photoelectric astronomy continues underthe restriction of one picture element at a time,what developments might extend its range ofusefulness ? At the head of the list would comea more sensitive cathode for the red and infra-red where the quantum efficiency is now 20times poorer than for the blue. This regionof the spectrum may assume increasing im-portance for two reasons: Interstellar absorp-tion is greatly reduced, and discovery levelor penetrating power correspondingly increased;and at very large red shifts, extragalactic systemscarry not only their energy maximum but alsoall the parts of the spectrum ordinarily studiedinto the red region. A high-efficiency photo-electric detector would greatly facilitate in-vestigations in these areas. A new tri-alkalicathode (Sommer, 1955) promises considerableimprovement in the red as far as 7500A.Germanium phototransistors may be developedsufficiently to fill the need as far as 16,O0OA.

Some reduction in observing time can berealized by using two or more phototubes, fedby beam-splitters, or by several different out-puts of a spectrograph. The most excitingdevelopment in photoelectric astronomy, how-ever, would be an image tube that wouldpreserve the high quantum yield of a cathode,while responding simultaneously yet separatelyto all the bits of information in the variousarea elements of the image plane. Such adevelopment seems to be a definite possibility.The techniques being explored, as explainedby Baum, have been reported by Struve(1955). Work is actively in progress inFrance, England, and the United States.

In its simplest form, that first introduced byLallemand, the image tube consists of an elec-tron lens system for focusing the acceleratedphotoelectrons in an imago on a photographicplate at the end of the tube opposite to thecathode. Every electron sensitizes a plategrain, thus preserving the high quantum effi-ciency of the cathode. The same number ofdevelopable grains is achieved in a much shortertime than would be needed in direct exposure ofthe plate to light. A fine-grain plate may beused without loss of speed, thus increasing thenumber of bits of information in a given imagearea before saturation sets in. As yet, exposuretimes adequate for showing an intensity levelas low as the night sky have not been possible,and the anticipated gain in threshold detectionlevel has therefore not been realized.

A proposed alternative scheme involvesstorage of the released charge on a mosaic, asin the image orthicon. This is later wiped offby a scanning electron beam, and the effecton the beam is the signal from that element.

While a great deal of hard work remains,there appear to be no insuperable difficultiesin the way of final success. If such a tubecomes, what will be the first applications?Reduction of prohibitively long exposure timesin photographic spectroscopy would be one ofthe first. All-night exposures would be re-duced to a few minutes. Spectra of faintgalaxies, so important in considering cos-mological questions, would become fairly rou-tine. On direct pictures of the sky a detectionlimit 2.5 magnitudes fainter should result,because 100 times as many countable events


go into the decision as to whether one pictureelement (say the size of the seeing tremor disk)has received a statistically significant excessof photons from a star during the exposure

•time.While high quantum efficiency and accurate

imaging can be realized in an image tube, itmay be doubted whether the linearity and otherphotometric advantages of a single phototubecan be preserved through the whole process.Nevertheless a successful image tube wouldbring many observations, now utterly unat-tainable, into the realm of the possible. Inthe rich harvest that would follow its intro-duction, there would almost certainly be someunexpected discoveries.

Conclusion

As one looks ahead, it seems clear that thenext decade will see increasing use of the photo-electric process in astronomical photometry.Wherever two quantities of radiation have tobe compared, the photoelectric method will bethe preferred method if time permits. Shouldthe image tube come into common use, a revolu-tion in astronomical observing methods mayensue. It is not impossible that the wholefield of the detection and measurement ofradiation may pass from the present state of

low efficiency to something approaching theideal state where every quantum is counted.It would then resemble the present state ofoptics as an astronomical tool: no revolutionaryimprovements could be expected. But eventhen, as now, astronomical advances willdepend quite as much on the skill and ingenuityof the astronomer as on the tools he has athis disposal.

ReferencesABP, H. C , BATTM, W. A., AND SAN DAG E, A. R.

1953. Astron. Journ., vol. 58, p. 4.BABCOCK, H. W.

1953. Astrophys. Journ., vol. 118, p. 387.BAUM, W. A.

1953. Astron. Journ., vol. 58, p. 211 (abstract).HALL, J. S.

1950. Publ. U. S. Naval Obs., No. 17, pt. 1.HlLTNER, W. A.

1949. Astrophys. Journ., vol. 109, p. 471.IBWIN, J. B.

1955. Sky and Telescope, vol. 14, p. 132.ROSE, A.

1948. In Marton, ed., Advances in electronics,ch. 3. Academic Press, New York.

SOMMER, A. H .1955. Rev. Sci. Instr., vol. 26, p. 725.

STBUVE, O.

1955. Sky and Telescope, vol. 14, p. 224.WHITFOBD, A. E.

1953. Astron. Journ., vol. 58, p. 49 (abstract).

881014—56 4

Interferometers in Radio AstronomyBy B. F. Burke * and M. A. Tuve x

Only a faint glimmer of an idea yet exists asto how astronomical objects can radiate, appar-ently with a fair degree of conversion efficiency,electromagnetic waves some tens of centimetersor even meters in length. The discovery phaseis probably over; only a few such objects areknown, but these are enough to face us un-equivocally with the problem. "Plasma oscil-lation" is an easy phrase, and where the electrondensities are high, as in the solar atmosphere,these words seem to offer a plausible explana-tion, but when one is confronted with theproblems of boundary conditions and the mech-anism for escape of the radiation to free spacethe plausibility disappears. Additional difficul-ties arise for the low apparent electron densitiesin gas clouds. Strong and perhaps turbulentmagnetic fields combined with violent stream-ing motions, perhaps even with shock waves,may resuscitate the hypothesis, but theory hasyet to rescue the observer in this field.

Progress toward solution of this problem maystill lie in the hands of the observer, however,as there exist only a dozen or so sources whichhave been specifically located and for which"color" information (in the radio spectrum), forexample, has begun to be available. When atleast several hundred sources, thermal and non-thermal in their radiation characteristics, havebeen cataloged and their radiation propertiesdescribed in some detail, the puzzles may di-minish. Large arrays of antennas, mainlydipoles or helices, connected and used as inter-ferometric devices, are the principal technicalbasis for obtaining observations at the longerwavelengths. Parabolic reflectors of any rea-sonable size cannot be expected to give thenecessary precision of> location in the sky; at20-cm wavelength even a 500-foot parabola

1 Department of Terrestrial Magnetism, Carnegie Institution ofWashington, Washington, D. C.

would still cover an area 5 minutes of arc indiameter. Arrays of great size and interfer-ometer spacings of as much as a mile are simpleby comparison to such a dish.

It should not be surprising if a new branch ofscience raises difficulties in its early days, andradio astronomy has not been disappointing inthis respect. In the past few years, this youngbranch of astronomy has provided a largevariety of both physical and astronomicalproblems. A striking feature of the earliestinvestigations has been that most of the in-trinsically strong sources of radio noise arerelatively insignificant optical objects, a circum-stance that adds to the difficulty of examiningin detail their physical behavior. The handfulof optically identified discrete sources all exhibitone feature in common, however, which pro-vides some indication of the nature (and diffi-culty) of the problem, for in all cases we seemto be dealing with highly turbulent gas whoseconstituents are in high states of ionization.Other general or special requirements mayexist, and they will have an important influenceon any theory of how such a tenuous butviolently agitated gas becomes a relatively effi-cient emitter of radio noise. For example, thepolarization of the light from the Crab Nebula(pi. 1) provides evidence for the existence of amagnetic field and large quantities of highlyenergetic electrons, which also could be efficientradiators in the radio spectrum, but whetherthis is a general characteristic of all radiosources or a peculiarity of this one object onlythe future will tell. I t appears certain, how-ever, that the necessary highly turbulent gascan be produced in a variety of ways. In ourown galaxy, cataclysms such as supernovaexplosions are known means of providing aturbulent medium that is capable of strongradio emission; but although all discrete galactic

31


sources may possibly be supernova remnants,many of them may also represent an entirelynew class of objects.

The extragalactic sources, which are intrin-sically more powerful emitters of radio noise,require a correspondingly more violent origin.The collision of two galaxies seems to be asufficiently drastic occasion, but there are pe-culiar galaxies such as M87 which evidently re-quire some other means of excitation.

The existence of two different populations ofradio sources raises the important question ofhow to distinguish between them, and how toestablish, if possible, a distance scale. Theonly reliable method at present seems to be iden-tification with a visual object, but as time goeson this method must prove increasingly difficult.Even now it is clear that if a pair of collidinggalaxies as violent as the Cygnus source exists,but 10 times farther away, we have little hopeof observing it visually even though it wouldstill be an observable radio source. Similarly,many galactic objects, conspicuous in the radioregion, are undoubtedly so obscured by dustthat optical observation will be difficult. Inboth cases, however, a distance scale would cer-tainly bring to light interesting new problems.For example, the distribution in space of extra-galactic radio sources may provide a clue to im-portant cosmologies! problems, since it is prob-able that many observable radio sources aresituated at great distances, approaching thelimits of the visible universe. So far, only sta-tistical methods have been employed, and atpresent there appears to be a serious discrepancybetween observations. The results of the Cam-bridge interferometer radio survey indicate anunusually large number of faint sources, imply-ing an actual increase in apparent spatial densityof radio sources in the universe at great dis-tances. On the other hand, preliminary obser-vations by Australian radio astronomers showa number-magnitude relation that implies anearly uniform space density of radio sources.The resolution of this direct observational con-flict is certainly one of the most pressing prob-lems in radio astronomy today. The solutionmay lie in an understanding of the techniquesinvolved, for both surveys utilized interfero-metric techniques but in radically differentforms.

All the sources we have been considering arecharacterized by their nonthermal nature, sincenonequilibrium processes appear to be involved.An equivalent black body in order to exhibit anequal surface "brightness" in the radio regionnot only would have to be at an extraordinarilyhigh temperature but also would have to havean equivalent temperature that rises with in-creasing wavelength. In the meter-wave re-gion, where many sources are sufficiently in-tense to be seen with relatively simple equip-ment, they appear superimposed on a smoothbackground of radio noise, most of it evidentlyof galactic origin.

Interferometric techniques are particularlyvaluable in removing the effects of the back-ground, "labeling," as it were, the signals origi-nating in discrete sources. Figure 1 illustratesthe principles of the phase-switching interferom-eter first described by Ryle (1952). In theupper drawing, two antennas are arrangedsymmetrically to form the radio analog ofthe Michelson interferometer. If a radio sourceis directly overhead, or is off-axis by an amountsufficient to result in a path difference ofan integral number of wavelengths, the twosignals constructively interfere. Therefore, byswitching the phase of one antenna at a regularrate the signal from a discrete radio source ismodulated, while the slowly varying back-ground is not, and we1 can distinguish the noisefrom such sources by looking for a modulatedcomponent in the total signal received. Thesystem has other advantages as well. If asource is at a half-power point of the inter-ference pattern, the signal received will be thesame whether a half wavelength has been in-serted or not, and consequently at these po-sitions the signal from a discrete source is un-modulated. As a result, position measurementscan be made by a null method, with muchgreater precision than is possible with a pencilbeam. Furthermore, the relative amplitude ofthe interferometer signal as a function of an-tenna spacing gives a measure of the angularsize of the source, effectively constructing theFourier transform, just as in the optical Michel-son interferometer. It is possible, therefore, toobtain information with relatively small an-tennas that otherwise could be obtained onlywith a single antenna of prohibitively large size.

BURKE AND TUVE PLATE 1

The Crab Nebula. Photograph from the Mount Wilson and Palomar Observatories(200-inch photograph).


I I I I •>v

ANTENNAA

CABLEIRECEIVERI CABLE ANTENNA

B

DIRECTIONAL RESPONSE WITH EQUAL CABLELENGTHS; SIGNALS FROM OVERHEAD SOURCE ADD

1 1 1 1L _

1 1 < •

at oI RECEIVER! -g- X

HALF WAVE DIFFERENCE IN CABLE LENGTHS:SIGNALS FROM OVERHEAD SOURCE CANCEL

FIGURE 1.—Phase-switching interferometer.

There is danger in the method, however,which must be guarded against; if more thanone source is being received, the interferometerwill measure the combined effect of the two,and, unless one reconstructs the Fourier trans-form by going to sufficiently large spacing, asingle source will appear to be observed wherein reality there are two. The effective gain ofthe interferometer is the geometric mean ofthe gains of the two antennas, and at first sightit would appear that if we simply make theantennas larger, and thus see fewer sources perbeam width, this difficulty would be circum-vented. It is true that the confusion of strongersources may be resolved by this technique, butthe increased antenna gain then makes it possibleto detect a still larger number of sources perbeam width, and since one naturally wishesto examine these, also, multiple spacings arestill required. At the longer wavelengths,particularly, where a pencil beam of the orderof a minute of arc or less requires an array ofprohibitively large size, interferometric tech-niques are indispensable, but they must beapplied with caution and full knowledge of thedifficulties involved.

The study of the sun is an interesting specialcase, for the radio emission at longer wave-lengths provides a useful tool for probing thecorona; here interferometers have alreadyproved their value, providing us with a qual-itative picture of the sun at a number of differ-ent wavelengths. The detailed appearance ofthe sun, particularly at the longer wavelengths(greater than 4 m) needs much more study,and there is little doubt that interferometrictechniques must prove necessary.

It was remarked earlier that the effectivesensitivity pattern of a pair of antennas usedas an interferometer is equal to the geometricmean of the gains of the individual antennas.This fact is utilized in the special case of the MillsCross (Mills and Little, 1953), which, in reality,is a zero-spacing, phase-switched interferometerutilizing two linear arrays at right angles. Fig-ure 2 gives an illustration of the principle.Each array has a fan-shaped sensitivity pattern,but when connected as a phase-switched inter-ferometer the effective sensitivity is the geo-metric mean of the two antenna patterns, whichis a pencil-beam whose resolving power is equiv-alent to a square array containing many more

34 SMITHSONIAN CONTRIBUTIONS TO ASTROPHYSICS VOL.!

MAIN LOBEOF X ^

PRINCIPAL LOBE WHENPATTERNS ARE MULTIPLIED

MAIN LOBEOF "B"

ARRAY 7C ARRAY

OUTPUTS TO RECEIVERFIGURE 2.—Antenna diagram of the Mills Cross interferometer.

elements. At the longer wavelengths this pro-vides us with a technique for producing a pencil-beam with simplified arrays, and the future willundoubtedly see arrays based on this principlebut of far larger dimensions than those in usetoday, since a pencil-beam is unquestionablythe best way to investigate extended sources.As in most experimental and observationalfields, the advantages of this method are ac-companied by disadvantages as well, for thissystem is troubled by spurious responses aris-ing from its unusual diffraction pattern. I tis possible for a strong source, passing welloutside of the main pencil beam, to appear asa weak source, and great care must be exercisedto avoid, if possible, or at least to identify suchspurious responses. (See fig. 3.)

The longer wavelengths, of the order of a

meter or more, have been emphasized sincethe observation of a large number of sourcesat these wavelengths can probably be accom-plished more conveniently by interferometrictechniques than by an equivalent array, whilethe measurement of angular size of smallsources is probably possible by no other means.The phase-switching technique has the furtheradvantage that the noise originating in discretesources is "labeled" by modulation, and if, inthe future, it is possible to investigate theintrinsic properties of source noise this featuremay be invaluable in distinguishing the desiredsignal from the galactic noise background.All present interferometric systems have weak-nesses, however, which must be appreciatedif the important work on weak radio sourcesis to have reliable significance.


PATHOF

STAR

FIGURE 3.—Detailed antenna pattern of a Mills Cross interferometer.


In our concern with equipment it is importantthat we not lose sight of the larger issues. Thecosmological problem has been outlined, but otherimportant questions also exist. What is thecause of radiation in the case of those galaxiesthat are not, apparently, colliding with any-thing? What is the significance of the dis-tribution of radio sources in our own galaxy?And, what are the physical conditions necessaryfor the emission of radio noise from large,turbulent gas clouds? Are magnetic fields

essential? The answers to these questions,and others as well, depend on an increase inthe information concerning discrete radiosources. Larger interferometers will certainlyplay a role in the gathering of this information.

ReferencesMILLS, B. Y., AND LITTLE, A. G.

1953. Australian Journ. Phya., vol. 6, p. 272.RTLB, M.

1952. Proc. Roy. Soc. London, ser. A, vol. 211,p. 351.

Related Sciences

RocketryBy R. Tousey *

For many years astronomers the world overhave dreamed that sometime it might becomepossible to discover the nature of the extremeultraviolet and X-ray spectrum of the sun andthe stars. Research in this field becamepossible at the conclusion of World War IIas a result of the development in Germany ofthe V-2 rocket, which was capable of carryingscientific equipment to altitudes far above themain absorbing regions of the earth's atmos-phere. The V-2 rocket and the AmericanViking and Aerobee rockets which followedhave opened the field of rocket astronomy.The cost of this research is large, but the stakesare very high, for it seems probable that thekey to many unsolved astrophysical problemsmay be found in the nature of the spectrum ofthe radiation of very short wavelengths emittedby the sun and the stars.

The difficulties of conducting astronomicalresearch from rockets are many. The spaceavailable for instrumentation in all but thelargest rockets is rarely more than 10 cubicfeet, and the payload is of the order of only afew hundred pounds. The time available forobservation is 2 to 5 minutes. I t is difficultto keep instruments pointed at the celestialobject under study since rockets roll and turnin an unpredictable fashion. Flight prepara-tions are time consuming and complicated, sothat the firing time is often determined bycircumstances other than those that are opti-mum for the experiment.

In spite of the vicissitudes of research fromrockets, during the past several years greatadvances have been made toward the solutionof the problems involved in rocketry; yet, thereis much still to be desired. Astronomers aremost anxious that the work be carried ahead asvigorously as possible.

1 U. S. Naval Research Laboratory, Washington, D. C.

Improvements needed in rocketryAlthough the V-2 rocket was capable ofcarrying a tremendous payload to an altitudeof 100 miles, which is adequate for manyexperiments, it was extremely complicated,difficult to launch, expensive, and more oftenthan not it misfired. Two American rocketswere designed to replace the V-2 for upperatmosphere research: the Aerobee, a small,comparatively inexpensive rocket with a peakaltitude of about 120 km; and the Viking,similar to the V-2 hi complexity and size,capable of 200-km altitude, and stabilizedduring flight. A number of Vikings wereflown with successful experiments, but thetrend toward economy has forced scientiststo use and improve the Aerobee rocket forresearch.

The Aerobee uses an acid-aniline fuel, islaunched with a booster from a tower, andcosts at present about $30,000. This rocket isquite satisfactory for many experiments. Itcan be launched at sea, though not as easilyas is desirable. The early type of Aerobee didnot reach altitudes sufficient for many experi-ments. The Aerobee-Hi, now in production,will ascend to 200 to 300 km with payloads be-tween 150 and250 lbs.; thus,it has performancecharacteristics that are satisfactory for almostall types of research. Although it is a muchsimpler and cheaper rocket than the V-2 orViking, the Aerobee is still too expensive to beused as frequently as would be desired inastronomical research.

The preparation of a rocket and the launch-ing are complicated. Furthermore, at landbases such as the White Sands Proving Grounds,the scientist is limited to firing at a particulartune, when the range is ready and clear and thewinds are favorable, even though he may wishto fire, say, during the onset of a solar flare.

39


The improvement most needed in rocketry,therefore, is a really inexpensive rocket thatcan be made ready and flown without elaboratepreparations and precautions. Recent ad-vances encourage hope that such a rocket,employing a solid fuel propellant, can be devel-oped in the next few years. A rocket that willcarry 100 pounds to 300 km and can be launchedby a few persons, perhaps by the scientiststhemselves, at a cost of perhaps $5,000 no longerseems unattainable. Such a rocket might bestbe launched at sea, or at least flown out to seafrom shore. This would avoid complicatedand restricting safety requirements. Sea re-covery methods would have to be worked out,since most astronomical experiments are photo-graphic. This involves the sealing and flota-tion of the section containing the instrumenta-tion and the development of a system for locat-ing the floating section.

For flights over land, recovery of photographicfilm may be achieved simply through the useof a strong casette, but ordinarily the equip-ment is completely ruined. Recent advancesin parachute techniques, however, make itpossible now to salvage the equipment forfurther study or reuse. The cost of the para-chute assembly must be greatly reduced,perhaps to the neighborhood of $500 to $1,000,before its general use will be warranted econom-ically.

One of the most difficult and importantproblems in rocket astronomy is that of keepingthe equipment accurately pointed at the celes-tial object being studied. Although it may bepossible to stabilize a rocket in flight, the preci-sion of equipment stabilization required formany experiments can be far more easily andcheaply provided by stabilizing only theinstrument. So far, attempts at pointing havebeen limited to the sun. The most successfulof several "sun-followers" is the "BiaxialPointing Control" developed by the Universityof Colorado under Air Force sponsorship anddescribed by Stacey, Stith, Nidey, and Pietenpol(1954). This device kept a spectrograph flownin a U. S. Navy Aerobee on February 22, 1955,pointed at the sun within an average error of± 1 minute of arc during a total exposure timeof 2 minutes. The cost of this equipment is

high, therefore safe recovery is mandatory ifmany experiments are to be conducted.

Further development of pointing controls inthe direction of lower cost and greater load-carrying ability and simplicity is very importantfor rocket astronomy. With such controls thestudy of the bright stars from rockets maybecome possible. This will probably require apointing control that operates from the moon,or the sun, and the earth's magnetic field, orpossibly from simply the moon and the sun;the star itself may provide the final signal forprecise pointing.

The fields of rocket astronomy

The experiments already conducted from rock-ets cover many fields of research in astron-omy and the physics of the upper atmosphere.The work already accomplished has been mostrecently summarized by a collection of paperspresented at the Gassiot Committee meetings(Boyd and Seaton, 1954). Many fascinatingnew experiments have been proposed. In thissection an attempt is made to cover some of themost interesting.

The solar spectrum.—The sun is certainly themost important and the least difficult celestialobject to study. From the rocket work con-ducted by the Naval Research Laboratory(Tousey, 1953a, 1953b; Wilson et al., 1954;Johnson et al., 1955), by the University ofColorado (Rense, 1953), and by the Air ForceCambridge Research Center (Jursa et al., 1955),its ultraviolet spectrum is now known in con-siderable detail to 977 A. Friedman (1955) andcoworkers (Havens et al., 1955; Friedman andChubb, 1955; Byram et al., 1955), using photoncounters, have measured the sun's X-rayemission in several wavelength bands between100 A and 5 A, and also have obtained muchinformation about the Lyman alpha line ofhydrogen and the dissociation of oxygen in theupper atmosphere.

Down to 2085 A the ultraviolet solar spec-trum resembles, generally, the visible spectrumin having many deep Fraunhofer lines. Theintensity continues to fall below that of a 6,000°K black body, and at 2100 A has reached abrightness temperature of approximately 5000°K. The photon counter measurements indicate


4,050° K at 1450 A and 3,900° K at 1250 A.At these wavelengths the continuous spectrummust come from the very coolest layers of thesun's atmosphere. At 2085 A strong continu-ous absorption of unknown origin sets in;although nearly obliterated, the Fraunhoferlines can still be seen, the lowest so far identifiedbeing at 1700 A. The continuum itself hasbeen photographed to 1550 A.

Below 1650 A the solar spectrum consists, inthe mam, of emission lines. So far, 32 emissionlines have been identified. These include Ly-man alpha and beta of hydrogen, the tworesonance lines of O VI at 1032 A and 1038 A,one resonance line of N V at 1239 A, the lineof He II at 1640 A corresponding to Ha, andvarious (mostly) resonance lines of silicon andcarbon up to three times ionized. The intensityof Lyman alpha has been measured on severaloccasions, both photographically and withphoton counters, with results varying from 0.1to 1 or more ergs/cm2/sec, suggesting that theemission in this line may be quite variable,depending on solar conditions. The soft X-rayintensities measured with photon countersappear to be roughly those expected from acorona at 750,000° K. They have been shownby Havens, Friedman, and Hulburt (1955) tobe quite sufficient to produce the E-layer of theionosphere.

There remains a great deal of fascinating re-search on the sun to be done from rockets.In this field only the surface has been scratched.While the data already obtained bear closelyon the structure of the sun, much more needs tobe learned to make possible the formulation ofa complete picture of the solar atmosphere.The solar spectrum from 977 A to 100 A is stillcompletely unknown, and from 100 A to 5 A ithas been measured only in broad chunks. Inthis spectral range lie the resonance lines of mostof the highly stripped atoms present in thecorona. It is now possible to study this re-gion. The requirements are a good pointingcontrol and a rocket that will spend severalminutes above 200 km. A scanning spectro-graph with a photon counter detector andtelemetering may be the simplest approach, atleast to obtain information without greatresolution. In the end, however, spectrumphotographs should be obtained.

Special studies of the sun.—The solar spectrumis most easily studied by using the light comingfrom a considerable area of the solar disk.I t is important, however, to learn exactlywhere on the disk the various emissions orig-inate, and how their intensities depend on theheight in the solar atmosphere. Several ap-proaches to this problem have been proposed.

Some information can be obtained with astigmatic spectrograph on whose slit the sun'simage is held stationary to within 1 minuteof arc with a pointing control. Spectraobtained in this fashion by Johnson et al.(1955) show, for example, that the O VI lineat 1032 A has a greater ratio of limb to centerof the disk intensity than does Lyman beta at1026 A, as would be expected. More measure-ments of this sort would make it possible toimprove greatly the model of the chromos-phere and corona.

Ideally one would like to fly a spectro-heliograph and to measure the intensity dis-tribution over the disk and corona of all lines inthe vacuum ultraviolet. For the Lyman alphaline, which has very high intensity, such aninstrument is feasible. A possible design waspublished by Behring et al. (1954). SeveralLyman alpha spectroheliographs or cameraswill be flown in the next year or two. Followingthis, weaker lines of higher excitation energyshould be investigated, including the ultimatelines of He I and II at 584 A and 304 A,respectively.

Another method of studying the distributionof Lyman alpha over the disk, suggested byFriedman, employs a photon counter of anarrow field of view, sensitive to this line only.By using a television scan of about 20 linesresolution, a crude picture could be transmittedto ground.

Once these experiments have been successfulit becomes of immediate interest to study thespectrum and the distribution of the spectrumover the disk under disturbed solar conditionsand during an intense flare. The solid propel-lant rocket and sea recovery are almost aprime requisite for this purpose because theymake it possible to fire at almost the instant aflare takes place. Though this may also bepossible with Aerobee-Hi, it will be difficultindeed, owing to the safety requirements


associated with launching from ground andbecause of the great complications surroundingthe use of an acid-aniline fuel.

A first approach to the question of the changeof ultraviolet and X-ray emission during aflare is planned by Friedman and his coworkers.They will instrument a number of "Rockoons."These small rockets are carried aloft at sea withballoons, and launched from an altitude of80,000 feet; they reach approximately 80 km,which is sufficient for studies of Lyman alphaand hard X-rays. When they are launched atsea they can be watched during the entire dayand fired at the moment a flare appears. Eachrocket will carry a photon counter sensitive toLyman alpha only, another to detect thepossible presence of 1-2 A X-rays, and a thirddevice to search for a burst of 7-ray Brem-strahlung that may be produced in a shockwave in the solar atmosphere associated withthe flare. It is also possible to have severalRockoons ready to launch in sequence, a fewminutes apart, to study the time history ofthe radiation from the flare. Such experi-ments can be done far better from a simplesolid propellant rocket, since this would carrythe instruments much higher than the Rockoon,to altitudes where soft X-rays and regions ofthe ultraviolet other than Lyman alpha canalso be studied. Furthermore, if no flareshould appear during the day the rockets wouldnot be launched, and so would not be lost, asare Rockoons.

Still another interesting experiment would beto study the extreme ultraviolet and X-ray solarspectrum during a solar eclipse. It is not outof the question to do this spectrographicallywith solid propellant rockets launched at sea. Afurther development would be required to enablea pointing control to track accurately in spite ofthe enormous change in solar light. I t wouldalso be quite possible to photograph the flashspectrum, and the results would be most im-portant for deducing the correct model of thechromosphere.

Friedman has proposed studying the sunwith photon counters and Rockoons during aneclipse. Here one would hope to determine thedistribution of intensity in the corona of radia-tion in several bands of the X-ray region, andthe distribution of Lyman alpha with height in

the chromosphere and corona. This experi-ment would be much better performed with asolid propellant rocket, since the balloon maydrift out of the eclipse path prior to totality.

The stars.—It would, of course, be of tre-mendous interest to use rockets to study theshort-wavelength spectrum of the stars. Thoughtheir faintness in the visible makes the problemappear extremely difficult, certain of the O andB stars are so very hot that their extreme ultra-violet and X-ray spectrum may well be suffi-ciently intense. Exploratory investigations canbe accomplished with photon counters, and hereagain Friedman has planned several experi-ments, both by day and by night. For thefirst experiments the sky will be scanned, as therocket rolls and yaws, by a Lyman alpha photoncounter of high sensitivity and also by certainother ultraviolet and X-ray counters. In laterexperiments the counter will be directed with apointing control. By day the situation is com-plicated by the expected presence of scatteredLyman alpha radiation from the sun, which initself is worth studying. At night, however,Lyman alpha and X-rays may be expected fromstars, nebulae, and other galaxies, and there iseven the possibility of detecting these radiationsin the zodiacal light. The relation between theF-corona and the zodiacal light could be studiedby visible or ultraviolet light photometry of thesky from a rocket at altitudes where the Ray-leigh sky background is very dim.

Many experiments can be imagined if anaccurate stellar pointing control were developed.For example, A. Boggess and J. E. Milligan havesuggested photographing the stars in the ultra-violet with a Schmidt camera and objectiveprism. The energy distribution of ultravioletspectra, worked out from these spectra, wouldbe of great interest in connection with stellaratmosphere studies. I t might also be possibleto extend knowledge of the wavelength depend-ence of interstellar reddening into the ultra-violet. Ultraviolet spectroscopy of the planetsmight well give information concerning possibleabsorbing atmospheric constituents. Perhapsmost exciting of all, however, is the possibilitythat the spectra of many of the stars could bephotographed in the vacuum ultraviolet.

Particles.—There are other fields of astronom-ical research which can be pursued by means of


rockets. One of these is the study of micro-meteorites. These tiny particles are thought tobombard the outer atmosphere in such abun-dance that it may be possible to detect acous-tically their impacts on a rocket, or even tocollect a sample and parachute it to earth forstudy. Experiments of this sort are beingconducted by both the Navy and the Air Force.

Another field is the analysis of the primarycosmic radiation. This is difficult at low alti-tudes because of the modifications and showersthat take place due to the stopping effect of theair. At rocket peak altitudes, however, itshould be possible to study the real primaryradiation.

Upper atmosphere physics.—Finally, there isthe large field of research on the physics of theearth's upper atmosphere, a field claimed alikeby astronomy, geophysics, and meteorology.Rocket work has already solved the problem ofthe origin of the E-layer, and when higherflying rockets are available the radiation respon-sible for the F-layers will certainly be dis-covered. Direct measurements of the composi-tion of the upper atmosphere have been made,and the density of ions of various sorts has beenmeasured with mass spectrometers. The elec-tron density through the E and Fi layer hasalready been measured from rockets by a radiopropagation method by Seddon (1954) andcoworkers (Seddon et al., 1954). This workshould be carried on at many points over theearth and under varying ionospheric condi-tions. The pressure and density of the atmos-phere have been measured from rockets andthe work should be continued at many placesover the earth and at different seasons. Itmay be noted that the density measurementsof Friedman (1955), made by studying theabsorption of X-rays by the atmosphere, do notagree with measurements made close to therocket with pressure gages. This interestingconflict of results must be resolved.

Finally, it is hoped that the origin of auroraecan be explained by means of rockets. It isexpected that during the International Geo-physical Year rockets can be launched withinthe auroral zone and flown into actual aurorae.The nature of the incoming particles will bestudied. Emission of Lyman alpha will also be

looked for, since it may be expected from re-combining protons and electrons.

Outside the auroral zone spectacular displaysare infrequent, but there is still present the nightairglow, an emission of light from the upper at-mosphere. The origin of the various emissionsin the airglow and their possible connection withthe aurora are not well understood; and opinionsdiffer concerning the altitudes at which thevarious spectral lines are emitted. Experi-ments from rockets designed to study directlythese emissions are underway; they should becontinued until the airglow is well understood.

For all these experiments the development ofthe solid fuel propellant rocket is most impor-tant. Since many experiments will be required,at special times and locations, a cheap andsimple rocket is essential.

Satellite

A satellite vehicle is simply a special type ofrocket that is projected at a velocity proper toplace it in a stationary orbit. I t is especiallysuited to experiments that require a long ob-serving time. In the foreseeable future, how-ever, the space and weight limitations on instru-mentation are very severe, and there can be nothought of recovery.

There have been proposed several importantexperiments that cannot be done with ordinaryrockets and may be feasible from a satellite.Perhaps the simplest experiment is the studyof solar ultraviolet and X-ray radiation over along period of time in order to determine thetype of intensity variations that may occur.Such variations are considered to be a cause ofsudden changes in the state of the ionosphereand radio transmission, and may also have animportant effect on weather. Photon counterswould be used.

Friedman and coworkers have proposed asearch of the sky for radiation hot spots bymeans of detectors in a satellite. The Lymanalpha line would be the radiation chosen forthe first experiments because of its probablehigh intensity and possible correlation withknown regions of high concentration ofhydrogen.

The earth's magnetic field could very wellbe studied from a satellite. Since the satellite's

44 SMITHSONIAN CONTRIBUTIONS TO ASTROPHYSICS VOL.]

orbit would cover a large fraction of the earthduring a day, it would be possible to study thevariation of the earth's magnetic field withlatitude and longitude, and, if the orbit wereelliptical, also with altitude. Experiments ofthis sort are important in connection with ttfeChapman-St0rmer ring current. An actualexperiment from a satellite to determine themagnitude of the ring current has been proposedby J. P. Heppner.

The satellite would, of course, be ideallysuited to the study of micrometeorites so smallthat they do not penetrate far into the earth'satmosphere but which probably ionize theupper atmosphere in process of being stopped.Even more important, however, may be thestudy of primary cosmic rays of energies solow that they do not penetrate far into theatmosphere. Particles of this sort are believedto exist and, in fact, many of them may comefrom the sun. The satellite also offers thepossibility of studying the distribution inspace from which cosmic radiation arrives, andperhaps of providing clews to the origin of thismysterious radiation.

ReferencesBEHRING, W. E.; JACKSON, J. M.; MILLER, S. C , JB.;

AND RENSE, W. A.1954. Journ. Opt. Soc. America, vol. 44, p. 229.

BOYD, R. L., AND SEATON, M. J., EDS.1954. Rocket exploration of the upper atmos-

phere. Interscience, New York.BYRAM, E. T., CHUBB, T. A., AND FRIEDMAN, H.

1955. Phys. Rev., vol. 98, p. 1594.FRIEDMAN, H.

1955. Ann. de Geophys., vol. 11, p. 174.FRIEDMAN, H., AND CHUBB, T. A.

1955. In Report of the 1954 Cambridge Con-ference, pp. 58-62. The Physical Soci-ety, London.

HAVENS, R. J., FRIEDMAN, H., AND HULBURT, E. 0.1955. In Report of the 1954 Cambridge Con-

ference, pp. 58-62. The Physical Soci-ety, London.

JOHNSON, F. S.; MALITSON, H. H.; PURCELL, J. D.; ANDTOUSEY, R.

1955. Astron. Journ., vol. 60, p. 165.JURSA, A. S., LEBLANC, F. J., AND TANAKA, Y.

1955. Journ. Opt. Soc. America, vol. 45, p. 1085.RENSE, W. A.

1953. Phys. Rev., vol. 91, p. 299.SEDDON, J. C.

1954. Journ. Geophys. Res., vol. 59, p. 463.SEDDON, J. C, PICKAR, A. D., AND JACKSON, J. E.

1954. Journ. Geophys. Res., vol. 59, p. 513.STACEY, D. S.; STITH, G. A.; NIDEY, R. A.; AND PIETEN-

POL, W. B.1954. Electronics, vol. 27, p. 149.

TOUSEY, R.1953a. In Kuiper, ed., The sun, p. 658. Univer-

sity of Chicago Press.1953b. Journ. Opt. Soc. America, vol. 43, p. 245.

WILSON, N. L.; TOUSEY, R.; PURCELL, J. D.; JOHNSON,F. S.; AND MOORE, C. E.

1954. Astrophys. Journ., vol. 119, p. 590.

Computing Machines in AstronomyBy Paul Hergetl

The impact of computing machines uponscientific research in recent years is not fullyappreciated outside of a small group of actualmachine users. When an electronic computa-tion center is installed in a scientific or technicalinstitution, almost invariably a workloadquickly develops that requires more than theavailable running time on the computer. Thisis a "snowballing" situation in which the dem-onstration of what can be done on one problemsoon leads to the exploration of similar pos-sibilities in related problems.

The next effect is that nearly all nondefenseand pure research projects are crowded out, forlack of priority. The National Science Founda-tion has recognized this situation and calledan ad hoc committee in February 1955 to con-sider the need for computing facilities inuniversity research centers in general. Astron-omy was represented by Schwarzschild andHerget. More recently other groups haverecognized the need to stimulate the develop-ment of computation centers in universities.

The need for electronic computing machinesin classical astronomy is mentioned in anotherpaper in this series, "The Minor Planets."As additional examples, we may note thatBrouwer, Clemence, and Eckert, using thesemachines, have completely recomputed themotions of the outer planets and Herget hasprepared a uniform system of coordinates forEarth and Venus. Clemence has recomputeda new general theory for Mars which is nearingcompletion. In studying the secular variationsof the elements of the major planets, it is nowfeasible to compute second order effects simplyby recomputing the coefficients under theappropriately varying conditions at intervalsof, say, 1,000 years. An extensive computa-tional program is also a significant part of the

1 Department of Astronomy, University of Cincinnati, Cincinnati,Ohio.

meteor studies currently being conducted byWhipple.

In astrophysics, one large group of problemsconcerns stellar models. The equations ofstate and equilibrium conditions for stellarinteriors form a system of ordinary differentialequations in one independent variable, theradius. Similar mathematical but differentphysical conditions exist in stellar atmospheres.Recently Henyey has elaborated on the prob-lem of stellar interiors by taking into accountthe evolutionary changes within the star.Again, if one deals with pulsating stars, notonly the stellar radius but also the time mustbe used as an independent variable. Thenthere are such specialized problems as the searchfor a mechanism that can explain the occurrenceof spicules on the sun.

Another new field of computational problemsis now arising from considerations of magneto-hydrodynamics. These are even more difficultbecause they involve partial differential equa-tions. The whole field of "wave equations"is of this same type. Much work has been doneby L. C. Green, L. H. Thomas, and others, butit is always necessary to restrict the equationssomewhat because even the largest computersin existence are still inadequate to deal withthis problem in its entirety. This problem isof interest not only for the light astronomicalatoms but more or less for the whole periodictable up to the fissionable and heaviest elements.

Examples of problems in stellar atmospheresmay be found in the Proceedings of the NationalScience Foundation Conference held at IndianaUniversity in October 1954. Other typicalproblems are given in the report of the JointMeeting of the American Astronomical Societyand the Association for Computing Machineryheld at Ann Arbor, Mich., June 1954. Oneinteresting problem was the computation of thehistory of a globular cluster (in two dimensions)

48


in which the individual stellar trajectories werecarried forward simultaneously until, eventu-ally, two out of a hundred stars evaporatedfrom the cluster.

Optical design, lens computations, and ray-tracing have been revolutionized by the devel-opment of high-speed computers. Thus thecost of experimenting with a novel design isgreatly reduced, since all the effects can nowbe so readily computed. Highly efficient meth-ods have been developed, especially by Baker,but there is still room for further study as thecapabilities of newer machines are improved.

The design of radio antennas for specialpurposes is another new computational problemin astronomy.

It is significant that, in nearly every one ofthese astronomical problems, the researcherhas been able to develop his own facility inthe use of an electronic computer as a researchtool even though he has not been primarily acomputer. The use of computing machines inastronomy will undoubtedly grow exponentiallyas each successful solution breeds more prob-lems, and each new experience inspires moreideas.

TurbulenceBy S. Chandrasekhar1

An aspect of theoretical investigations in as-tronomy and geophysics that is not often recog-nized is that matter and motions, in the large,may exhibit patterns of behavior that one mightnever suspect if one restricted oneself to matterand motions in the small. While several ex-amples can be cited to illustrate this, the moststriking is that which requires all motions insystems with large linear dimensions to be tur-bulent; for turbulence will be the natural stateof motions if the Reynolds number is sufficient-ly large. And since the Reynolds number con-tains the linear dimension of the system as afactor, it is clear that hydrodynamical motionswhich occur in astrophysics and geophysics mustnecessarily be turbulent. Indeed, "turbulence"is the refuge under which astronomers mostcommonly seek protection whenever they do notunderstand a particular phenomenon. At pres-ent, however, our understanding of the nature

« Yerkes Observatory, Williams Bay, Wls.

of turbulence is no greater than was our under-standing of the dynamical theory of gases, be-fore Maxwell and Boltzmann.

Since a deductive fundamental theory ofturbulence is a prerequisite for a rational com-prehension of many astrophysical and geo-physical phenomena, an attempt to follow cur-rent developments in the theory of turbulenceshould be an integral part of astronomical edu-cation. The need is the greater in view of theincreasing importance of magnetism in astro-nomical thinking; we have only to recall in thisconnection the problems of interstellar polari-zation and magnetic fields, the synchrotronradiation from the Crab Nebula, t ie galacticcorona, and the still unsolved problem of theearth's magnetic field.

Hydrodynamics and hydromagnetics arebound to play central roles in future astronom-ical developments and the part of turbulence isonly one aspect of this.

47

AstroballisticsBy Richard N. Thomas 1

Over the postwar years a small group ofastronomers have found themselves acting asmidwives at the birth of a new subject, astro-ballistics, conceived on the common bordersof astronomy, aerodynamics, ballistics, andchemical reaction-kinetics. Its multiple originprovides an excellent example of the preoccu-pation with crossfield research in contemporaryastronomy. Its evolution exhibits the trend,beginning with spectroscopy, which is changingastronomy from a wholly observational andtheoretical science to one demanding extensivelaboratory investigations. Its future dependson our success in establishing this crossfieldwork; that is, on our continuing to produce agroup of investigators who feel at home in atleast the overlapping parts of the four areas—astronomy, aerodynamics, ballistics, and chem-ical reaction-kinetics.

Origin and scope

The motion of planets and planetoids was foryears a conventional part of study in astronomy.In other papers of this group appear descriptionsof some of the exciting new data and theorieswhich have deflected our interest in theseobjects from their kinematic to their physicalproperties. We recognize, however, that we canhardly claim a satisfactory knowledge of ourplanetary system until we can also specifythe properties of those objects so small thatthey cannot be observed individually. Ingeneral, our information on these objects iswholly statistical, coming from studies of theircollective scattering effect on sunlight, actionas a resisting medium, etc. One notable excep-tion is the study of meteors; for these, the earthserves as a sampling box, sweeping up theinterplanetary material, with the earth's at-mosphere as a fluorescing medium, respondingto the entrance of each meteor. If we com-

1 Harvard College Observatory, Cambridge, Mass.

pletely understood all the processes that occurwhen a solid object moves through a gaseousatmosphere, we would have a fine method forinferring the corresponding properties of eachmeteor observed. We do not yet have suchan understanding, however, and the science ofastroballistics has arisen from our attempts toremedy this lack. Our present knowledge ofmeteors, as a part of the interplanetary medium,is discussed elsewhere in these papers. Herewe consider the meteor only as it represents asolid object moving through a gaseous atmos-phere, under one variety of conditions.

When a solid moves through a gas, severalthings may happen. The solid loses momen-tum to the gas and is decelerated. The amountof deceleration depends critically upon themass, shape, and velocity of the body, and uponthe composition and thermodynamic state ofthe gas. The solid loses energy as it deceler-ates, and the energy contributes heat to bothbody and gas. The partition of heat betweenthe two depends critically upon the velocity,shape, composition, surface structure, andthermodynamic state of the body, and upon thecomposition and thermodynamic state of thegas. The heat transfer to the solid may resultin thermal radiation and ablation (either bymelting or by vaporization) from the surface, aswell as in a simple temperature rise throughoutthe interior. The heat transfer to the gas mayresult in excitation and ionization, and thus inradiation, if the temperature rises sufficiently.Moreover, the material ablated from the solidmay interact chemically with the gaseous at-mosphere, serving either as a heat source orsink, depending upon the reaction.

The astronomer can make at least three typesof measures on a meteor moving through theatmosphere: (1) He can measure directly timeand position, and so obtain velocity anddeceleration; (2) he can measure the total radi-

48


ation, and possibly its spectral distribution; and(3) by radar, he can infer the ionization pro-duced by the meteor. Ideally, he would like tohave available a set of tables which he couldenter with these data, and so read off the phys-ical characteristics of a single body which alonecould have produced such a disturbance. Atworst, he might expect to obtain a range of suchbodies, and this range would represent his un-certainty of the properties of the particular me-teor observed. These properties he could com-pare with those of rare meteors that were largeenough and slow enough for the bodies to sur-vive passage through the atmosphere, and berecovered as meteorites.

Unfortunately, when we attempt to find oreven to prepare such a set of tables, we find adisheartening situation. Meteors impact on theearth's outer atmosphere at speeds ranging be-tween 10 and 75 km sec"1. The ballistician andaerodynamicist, to whom we might expect toturn for information, have just since the warbecome really conscious of the velocity rangethat is of interest to the meteor astronomer.They, in turn, have looked to the astronomer,hoping for insight into just those problems wepose to them. Only during the last war did theaerodynamicist develop a keen interest in speedsas high as that of sound, ~% km sec"1. Whilethe ballistician customarily dealt with speedsseveral times that of sound, his knowledge wasprimarily empirical, and his empirical investi-gations were limited by bis accelerating devices.The experimental investigations reported in theunclassified literature do not deal with objectsof known geometry whose speed exceeds ~ 2 kmsec"1. At these speeds, interest lies in air-resistance; aerodynamic heating of the body isnegligible. By analogue devices such as thewind tunnel, deceleration properties of bodies ofknown shape at equivalent airspeeds of ~ 3 kmsec"1 have been studied. Theoretical investiga-tions have made rapid progress in interpretingand correlating these empirical studies. On thewhole, we may say that the deceleration prob-lem is in a most satisfactory state, not only inthe range covered by the experiments but, forpurposes of extrapolation, considerably beyondthat range. But deceleration measures alone donot suffice to permit an inference of the physicalproperties of the meteor. For that, one must

turn to luminosity and ionization studies; andhere, the cupboard is essentially bare.

Ballistic and aerodynamic studies of objectsmoving at such high speeds that ablation occursare virtually nonexistent, at least so far as wecan infer from the unclassified literature. Afew simple free-flight experiments have beenmade, with velocities reaching ~ 6 km sec"1 atmaximum. We find an equally small numberof wind-tunnel studies, involving always sub-stances requiring but low heat transfer forablation. From a theoretical standpoint, thesituation is equally bad. The study of theaerodynamic flow pattern about a body that islosing mass at its leading surface has hardlybeen discussed. Indeed, the most satisfactorytheoretical work on any aspect of the fieldprobably lies in a few studies, made in collab-oration by astronomers and chemists, relatingto ablation in the meteoric regime—highvelocity and very low gas density.

When we consider the data available onexcitation and ionization caused by the ablatedmaterial ejected into the gaseous atmosphere,we find the bleakest situation of all. In the late1920's and early 1930's theoretical atomicphysics was striking out into the areas openedby the new quantum mechanics, and experi-mental physics was entering those areas openedby improved instrumentation. Problems ofradiative transition probabilities, while notcompletely solved, were at least understood.On the other hand, problems of collisionalexcitation were vague, expecially those in-volving exchange reactions, and particularlyatom-atom inelastic collisions. A number ofpreliminary experimental investigations werereported in the energy range of interest to theastronomer, < 1 kev. Unfortunately, from ourpresent standpoint, the early 1930's were alsothe era of the crucial experiments in the theninfant science of nuclear physics. The focus ofinterest rapidly shifted, and concern with thisarea of low-energy atom-atom collisions di-minished. An extensive base of data uponwhich to draw never developed, and it is onlyin recent years, among relatively small groupsof physicists and chemists, that interest in theseproblems for their own sake has reawakened.The astronomer's interest arises largely fromhis meteor studies, and occasionally from


phenomena in stellar atmospheres where, how-ever, his chief concern lies in electron-atominelastic collisions. As the instrumental rangeincreases, particularly in studies of very strongShockwaves, the aerodynamicist also is in-creasingly preoccupied with the details of theseatomic collision processes, excitation, ioniza-tion, and thus radiation.

To categorize and collate the variety of prob-lems just described, the science of astrobal-listics developed. Its province includes thestudy of those phenomena arising out of themotion of a solid through a gas at such speedsthat ablation occurs. Interest in and data onvarious of these phenomena arises in astronomy,aerodynamics, ballistics, and chemical physics.Investigations that combine the points of viewof the several sciences involved will have abetter perspective and are more likely to solvethe various problems than a study based onany individual science alone.

Evolution of the science

The field of astroballistics, as such, developedwhen astronomers recognized that it wouldprobably be necessary to set up a systematicprogram that linked astronomical observationsof meteors with laboratory studies in ballisticsand aerodynamics. The immediate postwarperiod found ballisticians and aerodynamicistsstill preoccupied mainly with measures of thesystem of aerodynamic forces on a missile oraircraft and, to a lesser extent, with problemsof strong Shockwaves. A number of advancesin experimental techniques, however, made itpossible to begin, at least, to bridge the gapsbetween meteor observations and the dataand theory of ballistics and aerodynamics.

A certain amount of development of newlaunching devices, and considerable improve-ment of the old, plus the use of low melting-point materials, together have made it possibleto simulate ablation from meteors. Newlydeveloped free-flight firing ranges permit studiesunder varying conditions of the atmospherethrough which the solid moves. Advancesin high-speed photography allow us to makeinferences of the spectral distribution of lumi-nosity along the axis and trail of an artificialmeteor, which we have been able to accelerate

up to speeds of 6 km sec"1. Microwave equip-ment facilitates studies of the ionizationcaused by the same type of artificial meteor.Theoretical studies already begun, based onthe insight developed in these first investiga-tions, will help guide subsequent research.Thus, we are laying a solid foundation fromwhich to study the problems described earlieras largely unsolved. We should, however,stress one aspect of all these investigations:the results of these first attempts have beeninvariably crude and preliminary, and progressin extending them has been dishearteninglyslow. We will return to this point. Atpresent, the one satisfying feature the resultshave in common is the simple fact that theyexist.

These first experimental investigations hadone very stimulating outcome. The firstspectral observations, though crude, demon-strated that it is possible to observe detailsof the chemical reactions involving ablatedmaterial and atmospheric atoms. Meteor ob-servations have never provided such data,primarily because of the unfavorable conditionsof (1) high velocity, which inhibits the ex-othermic chemical reactions expected when ametallic spray hits the atmosphere; and (2)low atmospheric density, which permits con-siderable diffusion before conditions are suchthat the reaction may occur. In the laboratory,such inhibition and diffusion effects may nowbe studied. Thus, a very fertile method forstudying chemical reaction kinetics appears.Even more, all these occur in a high-speedstream of gas, so that we have an approach toan aerodynamic flow problem of excitingpossibilities. For the astronomer, who is justnow acquiring a deep interest in gas-dynamicalproblems in stellar atmospheres, such observa-tions are invaluable in providing a physicalinsight otherwise inaccessible. In the lab-oratory, it is hard to provide gas streamshaving sufficient internal energy that atomicand molecular excitation change appreciablyand observably. Consequently, such data asthe above become highly desirable. The re-mark made above, however, should be re-peated; these studies are chiefly preliminary.The field has just been opened.


Needs and future of astroballisticsThe gross problems of astroballistics may beclearly delineated. A very great deal ofexperimental work must take place, both toprovide the empirical constants required in anytheoretical or interpretive work and to estab-lish the basic physical feeling for the problemswithout which the theory cannot proceed. Atpresent, facilities for such work are largelynonexistent or inadequate in comparison withour needs. Consequently, the field can developonly slowly.

For a long time the chief characteristic ofAmerican astronomy has been the size, quality,and quantity of its telescopes. Astrophysics inAmerica has grown largely through the outputof observational data. Several spectroscopylaboratories are contributing invaluable workon wavelengths and transition probabilities.Aside from these, however, we have no cor-responding astronomical laboratories devotedto the study of astroballistics, of conditionsunderlying particle growth in the interstellarmedium, and of magnetohydrodynamic proper-ties of plasmas and the like. It is of coursetrue that some phase of each of these subjects

may be under investigation at a dozen placesthroughout the country. But in how manyof these places is there an astronomer workingas a member of the research group, both con-tributing and gaining information ? How muchof the information gained is accessible in theopen literature? We should recognize twoproblems here. First, we have need for anational astrophysical laboratory that can oper-ate with considerable latitude to investigateproblems such as those in the field of astrobal-listics. Its staff should comprise astronomersas well as scientists from "related" fields, and itshould include both permanent members andthose on a temporary basis. Second, we musturge our graduate schools to help produce thestaff for such an institution by emphasizing intheir program the experimental aspects ofastronomy as much as the observational.Astroballistics is hardly unique in requiringthese supports; it is, after all, but a small fieldin astronomy. The needs of astroballistics,however, are a prototype of those existing incontemporary astronomy in general, and thefuture progress of the field may serve as asignificant example.

Electrodynamics of Fluids and PlasmasByMaxKrook1

Sufficient evidence has accumulated to es-tablish the widespread occurrence of large-scale electrodynamic phenomena in astro-physical contexts. To understand such phe-nomena and to interpret them quantitativelywe must investigate the origin, nature, andbehavior of cosmic electromagnetic fields andtheir interaction with cosmic matter. Thesetasks naturally involve considerations of fluidelectrodynamics and, from a more general andfundamental point of view, considerations ofthe kinetic theory of plasmas.2 Since theselatter branches of physics are still in a veryrudimentary stage of development, there isconsiderable scope for research not only inastrophysical applications but also in thefundamental physical disciplines themselves.

An appreciable fraction of cosmic materialconsists of ionized, and hence electrically con-ducting, gas. The dynamical behavior of suchconducting fluids is determined in part by forcesexerted by the electromagnetic field on electricalcharge and current distributions in the fluid.But equally, the dynamical behavior of theelectromagnetic field is determined in part bythe fluid motions. In general therefore, con-ducting fluid and electromagnetic field have tobe treated simultaneously and on an equalfooting as interacting components of a system.The study of systems of this kind is termedfluid electrodynamics (or magnetohydrodynamicsor hydromagnetics) when the fluid is treated as acontinuum, and is termed kinetic theory ofplasmas when the fluid is treated microscopi-cally.

Our aim here is to emphasize unsolved prob-lems and to discuss possible future trends. Noattempt, therefore, is made to review the

1 Harvard College Observatory, Cambridge, Mass.1 In this article the term "plasma" will be used in a general sense to

mean any partially or wholly ionized gas.

381014—56 5

literature in any comprehensive way. In adiscussion of this kind, it is not possible to dojustice to the individual contributions of themany authors in these fields, but some publishedreviews which include extensive bibliographiesare listed at the end of this article.

Fundamental equations

In the kinetic theory of plasmas, the state of asystem is specified by the following:

(1) State-variables for the electromagnetic field.These are the usual parameters of the Maxwell-Lorentz

theory; i. e., electric intensity E, magnetic intensity H,

charge density p«, current density J.(2) State-variables for the matter. These are

parameters which give a statistical description of themicroscopic state of the plasma.

With any dynamical process in a plasma isassociated some characteristic length L, andsome characteristic time T. If L and r areboth "sufficiently large," the plasma may betreated as a continuum whose dynamic andelectrodynamic properties are specified by aset of material constants; e. g., dielectric con-stant K, magnetic susceptibility /x, kinematicviscosity v, electrical conductivity <r (whichmay have directional properties), etc. In thismacroscopic theory, applicable also to con-ducting liquids, the state of the fluid is specifiedby the usual variables of fluid dynamics, i. e.,pressure p, density p, temperature T and flow

velocity©.The fundamental equations of the continuum

theory are made up of the Maxwell electro-magnetic equations together with the equationsof fluid dynamics, but with the followingmodifications (which represent coupling betweenelectromagnetic field and fluid motion):

(a) The momentum and energy equations of thefluid include terms which involve electromagnetic

53


variables; e. g., a form for the momentum equation is:

where F=external force density, and c=velocity oflight.

(b) The constitutive equation for the currentdensity is altered, e. g., to the form:

On account of the complexity of this systemof equations, it is generally extremely difficultto solve dynamical initial- and boundary-valueproblems of fluid electrodynamics, and especiallyso in cases of astrophysical interest. For thisreason, effort has been concentrated on investi-gating simple and idealized types of problemsand on exploring the consequences of variouskinds of assumption. It is hoped that insightinto the electrodynamic behavior of fluids gainedin this way may shed some light on the realdynamical problems and eventually lead totheir solution.

Some general principles of the subject maybe obtained by analyzing the mathematicalstructure of the fundamental equations. Di-mensional considerations, for example, showthat significant mutual interaction of fluidmotion and magnetic fields can occur onlywhen the "scale" of the system is large enough.For systems of ordinary laboratory size wewould generally require high magnetic intensity,or high electrical conductivity, or low density,or some combination of these factors. Thisexplains why magnetohydrodynamic behaviorappears more naturally in astrophysics (wherelinear dimensions of systems are large) than interrestrial physics.

The development of fluid electrodynamicswould be advanced immeasurably by theelaboration of techniques for performing con-trolled experiments on a laboratory scale andover a wide range of behavior. The experi-ments so far carried out with liquid mercuryand liquid sodium barely touch the fringe ofthe problem; and similar remarks apply alsoto plasma kinetics. Although certain aspectsof plasma physics have been studied extensivelyin laboratory experiments, e. g., with electrical

discharge tubes, many dynamical aspects ofparticular interest in astrophysics remainpractically unexplored.

Ionospheric studies are also capable ofyielding valuable information about plasmadynamics. And conversely, plasma dynamicscan play an important part in interpretingionospheric behavior. The structure of theearth's upper atmosphere can be studiedextensively and accurately by varioustechniques. Quantitative predictions of thetheory can therefore be subjected to morestringent checks in the case of the ionospherethan in most astrophysical cases.

Cosmic magnetic fields

In 1908, Hale discovered the strong magneticfields of sunspots (ii7~2,000 gauss) by utilizingthe Zeeman splitting of spectral lines. Greatimprovement in this technique, particularlyby H. D. and H. W. Babcock, has resulted inthe detection of fluctuating fields in the solaratmosphere ( H ~ l to 20 gauss), and of strongvariable fields in many stars (H~6,000 gauss).The polarization in the light of distant starsand its correlation with reddening have beenattributed to orienting effects of interstellarmagnetic fields on elongated dust particles(H~10~5 gauss). A circumstance which isprobably significant for the theoretical interpre-tation is that astrophysical magnetic fieldsgenerally vary with time.

The coupling of material motions and electro-magnetic field implies that, in a conductingfluid, mechanical energy and electromagneticenergy are convertible one into the other.Since relative motions are commonplace inastrophysical plasmas, we can understand ina rough qualitative way why magnetic fieldsmight exist there. (In a conducting fluid, mag-netic field energy is generally much greater thanelectrical field energy, except possibly undertransient conditions.) Quantitatively, andeven from a detailed qualitative point of view,however, we are still far from having a satis-factory theory of astrophysical magnetic fields.

Order of magnitude estimates are often basedon a so-called equipartition theorem, whichasserts that the average density of kineticenergy \po2 in a system equals the averagedensity of magnetic energy fiH2/8ir. For the


interstellar gas, this would imply fields of theorder 5X10"6 gauss. The theorem has so farbeen verified only in a small number of com-paratively trivial cases, and its real range ofvalidity is not known. Even for systems inwhich the theorem could be true, the equipar-tition state may not be possible because ofvery long relaxation times; this question can bedecided only by examining the mechanismswhich operate to convert mechanical energyinto magnetic energy and vice versa.

Our knowledge of such mechanisms is stillseverely limited. On general grounds it ap-pears that an efficient process for amplifying aninitially small magnetic field might be providedby the interaction of turbulent fluid motionswith the field. Very little progress has yetbeen made in the study of magnetohydrody-namic turbulence, especially in astrophysicallyinteresting cases of compressible fluids. Muchof the difficulty of this subject is due to the factthat the sort of questions in which we areinterested cannot reasonably be answered bycombining a statistical theory of turbulencewith a dynamical theory of the electromagneticfield. When a dynamical theory of turbulenceis eventually developed, great strides will bepossible in the theory of magnetohydrodynamicturbulence.

On account of many uncertainties in ourknowledge of the magnetic properties of inter-stellar dust, our knowledge of galactic magneticfields is correspondingly very inexact. I t istherefore important to look for further observ-able consequences of galactic fields which mightlead to more exact and detailed information ontheir structure and behavior.

One promising line of approach appears to bethrough problems of generation and propaga-tion of primary cosmic ray particles in thegalaxy, although it is still too early to anticipatewhere, or how far, this approach will lead.These problems present a wide and attractivefield for future research.

Preliminary calculations with the Fermi mech-anism have yielded some results which are con-sistent with the observational data and someresults which are not. In this model, which iscapable of some elaboration, particles of suf-ficiently large initial energy increase that energyby successive encounters with magnetic fields

"frozen" into interstellar gas clouds. The low-energy cosmic ray particles are supposed to beinjected into the galactic accelerator by stars,and this is not inconsistent with recent evidencewhich indicates that the sun itself may beejecting low-energy cosmic rays.

It has recently been found that optical, con-tinuous radiation from the Crab Nebula isstrongly polarized. This has been interpretedas synchrotron radiation from electrons inmagnetic fields of order 10~* gauss. An inten-sive study of this and similar nebulae in bothoptical and radio frequencies may well proveinvaluable in our understanding of mechanismsfor generating cosmic rays.

Magnetic fields and other electrodynamicphenomena can play a significant role inproblems of structure, behavior, and evolutionof astrophysical systems. There is considerablescope for research in problems of this kind.The structure of a star may, for example, bemodified appreciably in regions where magneticpressure is comparable with gas pressure; also,magnetic fields may have profound effects onthe stability of configurations. Further, ageneral galactic origin of cosmic rays, as inFermi's model, implies a heavy drain on themechanical energy of gas clouds; this wouldhave a substantial effect on galactic evolution.

Kinematic models

Since solutions of proper dynamical problemsare difficult to obtain, kinematic models canplay an important part in the development ofthe theory. In the kinematic approach weassume that the motion or configuration is ofa particular type and then examine the conse-quences with the fundamental equations. Thedynamo mechanisms proposed by Elsasser andBullard to account for the permanent field ofthe earth are of this kind, as are models oftorsional oscillations of the sun in its ownmagnetic field. The most fruitful line of attackon questions of the origin and behavior of sun-spots appears, at present, to be through in-vestigations of a variety of kinematic models.

In general, kinematic models cannot properlybe accepted as solutions of real dynamicalproblems until their consistency has beenestablished by stability investigations. Suchstability problems are usually rather difficult,

56 SMITHSONIAN CONTRIBUTIONS TO ASTROPHYSICS VOL.!

so the validity of most of the existing kinematicsolutions has remained untested.

Linearized equationsThe nonlinearity of the basic equations is amajor source of difficulty in the theory. Alinearized form of the equations may be usedfor treating motions of small amplitude; e. g.,propagation of magnetohydrodynamic waves.Limited but valuable information can beobtained from the discussion of such problems,especially for systems with boundaries and withgradients of the physical parameters; e. g.,density.

One of the more important applications oflinearized theory is in the investigation ofstability of systems against small disturbances.The stability of a number of simple systems inexternal magnetic fields has been examinedin this way, but very little is known aboutstability of systems for which the magneticfield is wholly internal (cf. kinematic models).

Kinetic theory of plasmas

The material constants <r, v, etc., that appearin the continuum theory are not determinedby that theory but have to be obtained byexperiment or from a different theory. Akinetic theory of plasmas, on the other hand,implicitly includes the phenomena of electricalconduction, viscosity, etc., in its basic equations.Such a theory is in fact invoked to calculate<r, v, etc., for use in the macroscopic theory.In addition, however, there are many cases ofinterest, characterized inter alia by "sufficientlysmall" values of L or T, for which the con-tinuum theory is not valid and for which wethen have to use kinetic theory.

Several problems of a fundamental naturealready occur in the very task of formulatingbasic equations for a kinetic theory of plasmas.The Liouville equation, although useful forcertain formal investigations, is far too complexto serve as a basis for attacking practicalproblems. At the opposite extreme is the verycrude model of a plasma as consisting of inter-mingled positive, negative, and neutral fluidswith frictional coupling. Between these ex-tremes there are many possible formalisms withvarying degrees of generality.

At present, the most fruitful compromise

between generality and mathematical tracta-bility is the theory based on the use of single-particle distribution functions fa(vl, 7, t).(^=velocity of particle of type "a".) Particleinteractions have then to be represented inidealized form—"close" interactions as binarycollisions, and "distant" interactions as averageinternal electromagnetic forces. Static anddynamic effects of particle correlations areneglected in the first instance but allowed forroughly by the introduction, post hoc, of theDebye screening radius. The basic equationsof this theory consist of the Maxwell equationstogether with a set of Boltzmann equations orof Fokker-Planck equations; e. g., for particlesof type "a":

mc

where ea=charge, ma= mass of particle, ( ~ \

is a sum of Boltzmann collision integrals, and

E, H include contributions of internal fields.Investigations of a fundamental character todetermine the range of validity of the theorywould be highly desirable.

In addition to the length L and frequency

u=— which characterize a particular dynamical

process, the plasma is characterized by a set ofintrinsic fundamental lengths and fundamentalfrequencies. Typical frequencies associatedwith particles of type "a" are the plasmafrequency <>>™ = {kxnae^lma]K, the gyro-fre-quency «i*)=——> and the collision frequencies

with particles of type "j8", o#},; (na—particledensity). The modes of dynamical behavior ofa plasma depend intimately on the relativemagnitudes of the various characteristic lengthsand frequencies. These modes have not yetbeen sorted out in a completely satisfactoryway for the general case.

The macroscopic theory can be derived fromthe kinetic theory by the Chapman-Enskogprocedure in the limiting case of "high" density(e. g., the <*£), » all other fundamental fre-quencies), and "small" gradients (e. g., L »all mean free paths). To any type of problemoriginally formulated in terms of the continuum


theory, there corresponds a more general for-mulation in terms of kinetic theory. Thegeneral microscopic problem is of course morecomplex than the special continuum problem.Some simplification is introduced in anotherlimiting case, that of "low" density (neglect ofcollisions), and "small" gradients (linearizedtheory). For a gas of neutral molecules thislimit would correspond to linearized free-mole-cule flow; for a plasma, however, cooperativebehavior persists through the agency of internalelectromagnetic fields.

Linearized kinetic equations

The nonlinearity of the equations is again asource of complication in the kinetic theory ofplasmas. The development of the theory asreported in the literature has been confinedalmost wholly to the linearized theory and, inparticular, to problems of small-amplitudeplasma oscillations. Such oscillations can bestudied comparatively easily in the two oppositelimiting cases of high density (collisionfrequency dominant), and of low density(collisions neglected); the intermediate rangebetween high and low density presents diffi-culties even though the equations have beenlinearized.

For an infinite, uniform plasma in thermo-dynamic equilibrium, dispersion relations (i. e.,relations between frequency « and wave numberk) have been obtained for the above two lim-iting cases. Qualitative and semiquantitativeresults have been found for the intermediaterange by using approximate kinetic equationswhich incorporate the physical properties ofthe Boltzmann equation but are mathematicallymore tractable. In the absence of externalfields, relatively undamped sound waves athigh density go over continuously into relativelyundamped, longitudinal (electron) oscillationsat low density, with heavy damping in theintermediate region. The character of theoscillations in the presence of external fieldshas not yet been studied in complete detail.In this case transverse and longitudinal wavesare in general coupled; magnetohydrodynamicwaves are the high-density analogue of low-density, free electromagnetic waves, etc.

Some attention has also been devoted to theinvestigation of systems consisting of beams

of charged particles interacting with one anotherand with a plasma. The examination of thestability of such systems is involved in thequestion of excitation of plasma oscillationsby such beams.

There is considerable scope for developmentof plasma electrodynamics even within theframework of the linearized theory. For theastrophysical applications, we are particularlyconcerned with boundary-value problems, withthe effects of gradients in the physical condi-tions (density, etc.), and with modes of excitingplasma oscillations; e. g., by supplying mechan-ical energy. An important question, stillunresolved, is that of conditions under whichoscillations in a bounded plasma can providesignificant radiation in radio frequencies. Eventhough the radio emission of objects like theCrab Nebula may eventually be shown to con-sist predominantly of synchrotron radiation,plasma oscillations probably will still be im-portant in accounting for radio noise from othersources; e. g., solar bursts, and for variousother electrodynamic phenomena.

In order to interpret observations of radiofrequency emission by the sun and othersources, we require to know how electromagneticwaves are propagated in plasmas. Such prop-agation problems are usually discussed interms of the magneto-ionic theory. They canbe treated more consistently and generally interms of the kinetic theory dealt with above,although the two methods give concordantresults over a wide range in the small-amplitudecase.

Nonlinear phenomena

Since the fundamental equations are nonlinear,plasmas are capable of types of dynamicalbehavior not subsumed in the linearizedtheory; e. g., the generation of discontinuitiesfrom initially continuous motions. Thesemodes of behavior have not yet been investi-gated in any systematic way.

Many of the astrophysical phenomena towhich we would wish to apply the theory(either microscopic or macroscopic) are soviolent as to make it extremely unlikely thatan adequate treatment of the phenomena wouldbe possible in terms of linearized theory.Nonlinear plasma electrodynamics presents


a whole new range of possibilities in our attackon various astrophysical problems; e. g.,generation of cosmic rays, generation of radionoise, etc. On account of the intrinsic com-plexity of the subject, it will probably benecessary to examine first a large number ofsimple and idealized cases in order to gaininsight into nonlinear modes of behavior.

A rather limited range of nonlinear behaviorcan be described in terms of oscillations of thelinearized theory with cross-modulation that isdetermined by treatment of the nonlinear termsas a perturbation. There are several interestingproblems of this kind concerned with the propa-gation of electromagnetic waves of "moderate"amplitude through plasmas. Some of thesehave a direct bearing on ionospheric propa-gation.

A typical nonlinear phenomenon, not reason-ably representable in terms of cross-modulation,consists in the generation and propagation ofshock waves. Some aspects of magnetohydro-dynamic shocks have been discussed in the liter-ature. To study the physical processes whichmay occur in the presence of shock waves in aplasma, we must examine the detailed micro-scopic structure of shock fronts and their sta-bility. Such investigations must be based, ofcourse, on the full nonlinear equations and aregenerally rather complicated.

Certain variations of the magnetic field of theearth have been attributed by Chapman andFerraro to clouds or streams of ionized gas

ejected by the sun and impinging on the earth'smagnetic field. Such clouds are also supposedto be responsible for various other observedphenomena; e. g., aurorae. Very little is knownabout the dynamical behavior of such clouds andthen* interaction with external fields. Still lessis known about the particular kinds of in-stability in the sun's atmosphere which result inthe ejecting of clouds and about the dynamicsof the ejection process.

Our discussion of problems in the electro-dynamics of fluids and plasmas is by no meansexhaustive. Many aspects have been treatedonly in cursory fashion; other aspects have noteven been mentioned. The development of thesubject has barely begun and the field is a wideopen one.

General referencesALFV£N, H.

1950. Cosmical electrodynamics. Oxford Uni-versity Press.

BIKRMANN. L.1953. In Beckerley, ed., Annual review of

nuclear science, vol. 2, p. 335. AnnualReviews, Inc., Stanford, Calif.

CHAPMAN, S., AND COWLING, T. G.

1952. Mathematical theory of non-uniform gases.Cambridge University Press.

COWLING, T. G.1953. In Kuiper, ed., The sun, p. 532. University

of Chicago Press.ELSASSER, W. M.

1955. Amer. Journ. Phys,, vol. 23, p. 590.ROMPB, R., AND STEENBECK, M.

1939. Erg. Exakt. Natur., vol. 18, p. 257.

Hydromagnetic and Plasma Problems1

By Eugene N. Parker

An ionized gas or plasma possesses twoobvious general classes of dynamical modes asa consequence of its being a mixture of twoelectrically coupled gases. In the electron gasthe velocity of sound is much higher than inthe other constituent, the ion gas, so that ifthe plasma is deformed sufficiently rapidly, theion gas will lag behind the electron gas; thiswill result in separation of charge and in largeelectrostatic forces that tend to restore theinitial relative configuration. The resultingoscillations of this "electric" mode are calledplasma oscillations. In the absence of magneticfields, the oscillations have a fundamentalangular frequency

&>,=(4wieTi/m)>*, (1)

where e is the charge of an electron in electro-static units, m is the electron mass in grams,and n is the number of electrons per cm8. Thefrequency defined in equation (1) is called theplasma frequency, and we might add that inthe presence of magnetic fields there are threepossible frequencies that are algebraic combina-tions of wp with the electron gyrofrequencyBe/me (Westfold, 1949).

If the disturbances initiated in a plasma areslow compared to the plasma frequency, thenobviously the electron and ion gases will movetogether. In this case the plasma is wellapproximated by a conducting classical fluid(Elsasser, 1954, 1955a). Since the motionsof an electrically conducting fluid interactwith magnetic fields, and in astronomicalphenomena are so generally associated withmagnetic fields, this plasma mode may betermed the "magnetic" mode. The study ofthe dynamics of conducting fluids in the

1 Assisted In part by the Office of Scientific Research and the Oeo-Phystcs Research Directorate, Air Force Cambridge Research Center,Air Research and Development Command, V. 8. Air Force.

'Enrico Fermi Institute for Nuclear Studies, The University ofChicago, Chicago, m.

presence of magnetic fields is variously referredto as hydromagnetics, magnetohydrodynamics,and fluid electrodynamics.

Hydromagnetics

The importance of hydromagnetics in astro-physical phenomena may be seen from thefact that much of the region within the galaxy,including stellar atmospheres and interiors, isoccupied by gas in various degrees of ionizationand readily transports electric charge. Bothdirect observation and theoretical inferencenow generally support the belief that much ofinterstellar space, together with a large fractionof the stars, is permeated by magnetic fieldswith an energy density comparable to thekinetic energy density of many of the gasmotions. Thus, the magnetic field stressesare comparable to the Reynolds stresses. Thehigh conductivity of the matter indicates thatit is strongly coupled to the pervading field,so that in treating most fluid motions of astro-physical interest, instead of the hydrodynamicequation,

(2)

one must use the hydromagnetic equations,

dvdf

(VXB)XB. (4)

Here c represents the conductivity of the fluidand M the permeability in mks units.

The complexity of the nonlinear hydro-magnetic equations indicates that the generalproblem must be divided into the several basicphysical processes; consequently, the study ofhydromagnetics involves the investigation ofindividual effects, one at a time. Instead of a


formal general solution of equations (3) and(4), we will ultimately achieve a general under-standing based on quantitative understandingof the many basic concepts (Elsasser, 1955b).

To mention but a few of the physical prin-ciples, let us consider first a fluid of very highconductivity. Then equation (3) reduces to

f=VX(vXB). (5)

From equation (5) it is readily shown that thelines of force move exactly with the fluid.Formally, if C represents a contour moving withthe fluid, and S the area enclosed by the contour,then

(6)

The lines of force retain their identity, since anytwo elements of fluid threaded by the same lineof force will remain so connected. Equation(5) is the same as the Helmholtz vorticityequation for an inviscid fluid. Cauchy's in-tegration gives B as an algebraic function of thederivatives of the Lagrangian coordinates of thefluid particles (Brand, 1947).

If the conductivity is not essentially infinite,then of course the lines of force slip relative tothe fluid. Besides the slipping, however, theylose their identity by continually changingtheir pattern of connection. Thus, for instance,a loop in a long tube of flux may detach itselffrom the tube, giving a magnetic ring and theoriginal, more or less straight tube. Thisdetaching process forms one of the basic stepsin the regenerative hydromagnetic dynamodiscussed below.

The forces exerted on the fluid by the mag-netic field are derivable from the magneticstress tensor, BtB}/n, (i, j=l, 2, 3), representingtension along the lines of force, and the isotropicmagnetic pressure, B2/2fi (Cowling, 1953), inmks units. Thus the total pressure isp-\-B2/2pi.Within a region of magnetic field the gas pres-sure p will tend to be less than it is outside bythe amount B2/2n. The gas density tends tobe less within the field, resulting in the phe-nomenon of magnetic buoyancy (Parker, 1955a).

The tension in the lines of force and the massof the fluid clinging to them indicate that the

lines are analogous to stretched strings, allowingthe propagation of transverse waves with avelocity whose square is equal to the stressdensity B2\\i divided by the material density(Alfv6n, 1950). These hydromagnetic wavesare transverse, and for this reason are generatedmore copiously by the transverse fluid motionsof subsonic hydrodynamics than are thelongitudinal acoustical waves. As the mag-netic energy density becomes large comparedto the kinetic energy density of the fluid, allhydromagnetic motions reduce to hydromag-netic waves (Parker, 1955b). I t is interestingto note, however, that the presence of themagnetic field may enhance the generation ofacoustical waves (Kulsrud, 1955).

The fundamental problem of the convectivestability of a rotating conducting fluid in thepresence of a field has been solved by Chan-drasekhar (1953b; see also 1952, 1953a, 1954a)with remarkable results. Simple Benard cellsdo not occur. Instead, competition developsbetween the rotational and magnetic effectsand results in two unstable regimes. Onemight say, very roughly, that the magneticfield tends to stiffen the fluid, producingstability, and the Coriolis forces tend to pro-duce instability. The results of Chandrasek-har's analytical formulation of this very dif-ficult problem have been given experimentalconfirmation (Nakagawa, 1955).

Chandrasekhar (1955) has also treated theproblem of hydromagnetic turbulence and findsthat there are two statistical modes, with eitherthe fluid motions or the magnetic field dominat-ing at large wave numbers. The analysisshows that there is not a close approach toequipartition of energy between the fluid mo-tion and the magnetic field, but, on the other

hand, it does indicate that K pv2 and B2/2n will

not differ by an order of magnitude.The origin of the stationary dipole field of

earth and that of the migrating solar fieldsresulting in sunspot formation are beginningto be understood on the basis of hydromagneticdynamo effects whereby two magnetic fields ina region are each generated from the other byfluid motions. The so-called toroidal field isproduced by simple shearing from the poloidalfield; the latter is then regenerated by C37clonic

NEW HORIZONS TN ASTRONOMY 61

motions in the fluid, forming loops from thetoroidal field (Parker, 1955c; Elsasser, 1955b).Presumably a sunspot is formed from the solartoroidal field through a combination of mag-netic buoyancy, hydrostatic equilibrium (Par-ker, 1955a), and the observed cooling ofthe interior of the spot, though most of thedetails remain to be treated quantitatively.

Probably nowhere does one find more phe-nomena whose origin so obviously cannot beexplained on hydrodynamic grounds than inthe violent phenomena of solar activity, suchas spicules, active prominences, flares andsurge prominences, the ejection of ion clouds,and cosmic ray bursts. The origins of thesephenomena are today not at all understood,but one presumes that at least some of themadmit of a hydromagnetic origin. The ter-restrial consequences make solar activity aparticularly pressing challenge to hydromag-netic research.

Understanding the terrestrial effects is im-portant, but it has often been suggested thatsolar activity may prove also to be the "littlebrother" to some of the grander phenomenathat result in the Wolf-Rayet stars, flare stars,and Babcock's magnetic variables where prom-inence activity, flares, and sunspots have devel-oped to pathological proportions.

Solar activity apparently ejects field-bearingclouds of gas into interplanetary space. At theearth the clouds are somehow responsible forthe aurorae, magnetic storms, certain iono-spheric disturbances, the configuration of theterrestrial weather pattern, and the observedinterplanetary gas density of as much as 500hydrogen atoms per cm3 (Storey, 1953; Sieden-topf, Behr, and Elsasser, 1953; Behr andSiedentopf, 1953). Further, cosmic ray in-tensity variations not directly due to theimmediate emission of new particles by thesun are now generally regarded (Neher, 1955)as being due to the antics of interplanetaryclouds. It becomes important, therefore, toinvestigate the dynamics of such clouds. If atthe same time the dynamics of ejection fromthe sun can be expounded, we will be able tocheck our ideas by both solar and terrestrialobservations.

Cosmic rays are presumed to be acceleratedby interaction with the magnetic fields carried

by moving matter. The acceleration is pre-sumed to take place in the galaxy and solarflares. One consequence of the high conduc-tivity of most matter is that the electric field inthe frame of reference moving with the fluid isvery small. Thus, it is readily shown that

E = - Y X B , (7)

where v is the fluid velocity, and there can beno large electrostatic acceleration effects. Thegeneral solution of the motion of a chargedparticle in slowing varying hydromagneticfields, as well as calculations of several cases ofabruptly varying fields, indicate that there areno hydromagnetic accelerating mechanismsother than Fermi's mechanism (Fermi, 1949)and the betatron effect (Swann, 1953; Riddifordand Butler, 1952). Since both mechanismsrequire that the particles initially have energiesof the order of 0.5 Bev per nucleon to avoid theenormous ionization losses at low energies, onepresumes that there is some more vigorousmechanism capable of accelerating particlesfrom thermal energies in spite of the tremendousenergy losses. Finding such a mechanism isobvious^ of fundamental importance.

The assumption that cosmic rays are accel-erated throughout the galaxy (Morrison, Olbert,and Rossi, 1954; Unsdld, 1955; Parker, 1955b;Elsasser, 1955a) presents energetic difficulties.Davis (1955b) has suggested that the gas ejectedfrom the sun forms a cavity in the magneticfield of the galaxy capable of trapping particlesin the solar system, though the mechanismapparently cannot account for the cosmic raycutoff with the attendant correlation with thesunspot cycle. A galactic origin of cosmic raysand the sweeping away of galactic cosmic raysby outgoing magnetic interplanetary cloudsseems to give a suitable cutoff, but so little isknown about the nature of interplanetary cloudsand Davis' cavity that the details of the mech-anism are entirely speculative.

Besides the evidence from the polarization ofstarlight (Davis and Greenstein, 1951), it hasbeen shown (Chandrasekhar and Fermi, 1953)that the observed properties of the spiral armsof the galaxy require a magnetic flux density ofabout 10"5 gauss along each spiral arm. Amagnetic field of this strength permeating the

381014—56 «


interstellar medium will be important in deter-mining the characteristics of stars formed bygravitational collapse of the interstellar medium,for, although it does not directly affect Jeans'instability criterion (Chandrasekhar, 1954b),the field will strongly influence the nature of thefluid motions against which work the gravita-tional forces of collapse. Hydromagnetics evi-dently plays an important role in the formation,structure, and evolution of galaxies, thoughlittle has been done on the integrated problem.

The pulsations of some variable stars suggest(Chandrasekhar and Fermi, 1953; Chandrase-khar and Limber, 1954; Ferraro, 1954) thatthe general magnetic fields of the individualstars play an important role in determining theform and period of the oscillations of the stellarbody. The tremendous general stellar fieldthat is required implies a hydromagnetic dynamoof immense power.

Plasma problem

The extraordinary number of modes of oscil-lation of the electron gas in a plasma is exhibitedin Bailey's (1950, 1951) general electromagneto-ionic theory. Starting with a plasma in thepresence of a magnetic field, and using Max-well's equations, the equation of continuity,etc., neglecting damping by collisions, and con-sidering small sinusoidal deviations from uni-formity, Bailey finds a number of steady statemodes with frequencies that are simple alge-braic combinations of the plasma frequency uP

and the gyrofrequency <ae, which is the angularvelocity of an electron in the given magneticfield. Besides these usual modes, however, thereare modes which grow or decay exponentially byenergy transfer to the over-all drift velocity ofthe ions. The presence of the magnetic fieldenhances the exponential growth effects. Sincethe theory neglects the second order terms, onedoes not know to what extent the exponentialgrowth is limited.

The response of a plasma to electromagneticwaves results in an index of refraction that canbe real, imaginary, or complex, and the propa-gation of electromagnetic waves through a non-uniform plasma is complicated and of funda-mental importance (see Pawsey and Smerd,1953). Consider the solar atmosphere in the

absence of magnetic fields. The plasma fre-quency decreases outward so that escapingradio waves of low frequency can only haveoriginated in the outer corona. Higher fre-quencies may originate farther into the atmos-phere. The observed cutoff frequency at thelow end of the frequency spectrum affords anindication of the level of the source in the solaratmosphere; the time rate of change of the cut-off frequency is an indication of the velocity ofthe source. But, at the same time, the re-fraction of the waves in the solar atmospherecomplicates the problem of determining theheliographic position of the source.

The presence of a magnetic field furthercomplicates the propagation, producing, amongother things, a tendency of the transverseelectromagnetic waves to propagate parallel tothe magnetic field, as a consequence of theelectrons' circling around the field. This effectappears in the terrestrial phenomena of atmos-pheric whistlers investigated by Storey (1953).The propagation depends upon the interplane-tary gas density near the earth and so givesinformation on the interplanetary medium.

The anomalous sources of radio noise in thesun and elsewhere are often associated with highvelocity gas or particle streams. The rate ofdecrease with time of the observed frequency ofa given solar burst suggests that the excitationmight be due to clouds that are moving outthrough the solar atmosphere with velocities ofthe order of 600 km/sec (Pawsey and Smerd,1953); intense galactic sources have been identi-fied with galaxies in the process of passingthrough each other (Baade and Minkowski,1954). The radio emission presumably arisesfrom the collisions of the interstellar gas cloudscontained in the interpenetrating galaxies(Greenstein and Minkowski, 1954). The pres-ence of high velocity streaming and the generaldifficulty with thermal mechanisms (requiringtemperatures up to 1O150 K) in the more intensesources suggest the investigation of vigorousmacroscopic phenomena such as the exponenti-ally growing plasma oscillation (Bailey, 1950,1951; Haeff, 1948, 1949; Pierce, 1948; Bohmand Gross, 1949, 1950; Buneman, 1950; Malm-fors, 1950; Blum, Denisse, and Steinberg, 1951;Twiss, 1951). To establish whether plasma


oscillations are responsible for the intensesources, one will have to obtain a detailedknowledge of the streaming which supposedlyexcites the oscillations. Then one must in-vestigate the effects of the nonlinear terms andthe actual boundary conditions found in naturein order to determine to what extent the oscil-lations will grow and whether the electromag-netic radiation generated by the radiations canescape. The question of escape has already beeninvestigated but with conflicting results.

Perhaps the acceleration of cosmic rays isclosely associated with the origin of radionoise; both might seem to require, or areobserved to have, association with violentfluid motions. Hydromagnetic processes seemunable to furnish the injection energies neededfor Fermi's mechanism and for the betatroneffect. Thus, obviously, investigations of vio-lent plasma phenomena should be carried outwith cosmic rays as well as radio noise in mind.

But in contrast to this point of view is therecent discovery that the visible and radiospectrum of the Crab Nebula can only be ex-plained as synchrotron radiation from electronswith energies of the order of 1 Bev, which bydefinition are cosmic ray particles (Shklovsky,1953). The radio emission from the galactichalo seems to require a similar mechanism.Thus, because we are ignorant of the "basic"energy source, cosmic rays become the apparentprime mover.

ReferencesALTVBN, H.

1950. Coemical electrodynamics. Oxford Uni-versity Press.

BAADE, W., AND MINKOWBKI, R.1964. Astrophys. Journ., vol. 119, pp. 206, 215.

BAILET, V. A.1950. Phys. Rev., vol. 78, p. 428.1951. Phys. Rev., vol. 83, p. 439.

BEHB, A., AND SIEDENTOPF, H.1953. Zeitschr. Astrophys., vol. 32, p. 19.

BLUM, E. J., DENISSE, J. F., AND STEINBERG, J. L.1951. Comptee Rendus Acad. Sci., Paris, vol. 232,

p. 483.BOHU, D., AND GROSS, E. P.

1949. Phys. Rev., vol. 75, p. 1851.1950. Phys. Rev., vol. 79, p. 992.

BBAND, L.1947. Vector and tensor analysis. John Wiley

and Sons, New York.

BUNEMAN, O.

1950. Nature, vol. 165, p. 474.CHANDRASEKHAR, S.

1952. Philos. Mag., vol. 43, pp. 501, 1317.1953a. Proc. Roy. Soc. London, vol. 216, p. 293;

vol. 217, p. 306.1953b. Monthly Notices Roy. Astron. Soc. Lon-

don, vol. 113, p. 667.1954a. Philos. Mag., vol. 45, p. 1177.1954b. Astrophys. Journ., vol. 119, p. 7.1955. Proc. Roy. Soc. London, vol. 233, p. 330.

CHANDRASEKHAR, S., AND FERMI, E.1953. Astrophys. Journ., vol. 118, pp. 113, 116.

CHANDRASEKHAR, S., AND LIMBER, D. N.1954. Astrophys. Journ., vol. 119, p. 10.

COWLING, T. G.1953. In Kuiper, ed., The sun, p. 532. University

of Chicago Press.DAVIS, L., JR.

1955a. Summer meeting of the American PhysicalSociety, Mexico City, Mexico, August.

1955b. Phys. Rev., vol. 100, p. 1440.DAVIS, L., JR., AND GREENSTEIN, J. L.

1951. Astrophys. Journ., vol. 114, p. 206.ELSASSER, W. M.

1954. Phys. Rev., vol. 95, p. 1.1955a. Amer. Journ. Phys., vol. 23, p. 590; vol.

24, p. 85.1955b. Summer meeting of the American Physical

Society, Mexico City, Mexico, August.FERMI, E.

1949. Phys. Rev., vol. 75, p. 1169.FERRARO, V. C. A.

1954. Astrophys. Journ., vol. 119, p. 407.GREENSTEIN, J. L., AND MINKOWBKI, R.

1954. Astrophys. Journ., vol. 119, p. 238.HABTT, A. V.

1948. Phys. Rev., vol. 74, p. 1532.1949. Phys. Rev., vol. 75, p. 1546.

KuLBRUD, R.1955. Astrophys. Journ., vol. 121, p. 461.

MALMFOBS, K. G.1950. Ark. Fys., vol. 1, p. 569.

MORRISON, P., OLBERT, 8., AND ROSSI, B.1954. Phys. Rev., vol. 94, p. 440.

NAKAOAWA, Y.

1955. Nature, vol. 175, p. 417.NEHER, H. V.

1955. Spring meeting of the American PhysicalSociety, Washington, D. C , Apr. 28-30.

PARKER, E. N.1955a. Astrophys. Journ., vol. 121, p. 491.1955b. Phys. Rev., vol. 99, p. 241.1955c. Astrophys. Journ., vol. 122, p. 293.

PAWSET, J. L., AND SHERD, 8. F.

1953. In Kuiper, ed., The sun, p. 466. Uni-versity of Chicago Press.

PIERCE, J. R.1948. Bell Syst. Tech. Journ., vol. 28, p. 33.


RlDDIFOBD, L., AND BtJTLEB, S. T . SWANN, W . F . G.1952. Phuos. Mag., vol. 43, p. 447. 1933. Phys. Rev., vol. 43, p. 217.

SlBDENTOPF, H., BBHB, A., AND E L S A S S E B , H . Twi88 R Q1953. Nature, vol. 171, p. 1066. ^ p h R L g 4 4 4 g

SHKLOVSKY, I. S. A

1953. Dokl. Akad. Nauk S.S.S.R., vol. 90, p. 983. UNSOLD, A.9TOBET. L. R. O. 1 9 5 5 - Zeitschr. Phys., vol. 141, p. 70.

1953. Philos. Trans. Roy. Soc. London, ser. A, WBSTFOLD, K. C.vol. 246, p. 113. 1949. Australian Journ. Sci. Res., A., vol. 2, p. 169.

Solar System

Jiartti s MagnetismBy Walter M. Elsasser 1

An explanation of the phenomena of geo-magnetism is of interest to theoretical astro-physics not just because the earth happens tobe a celestial body, but mainly because toreproduce such phenomena in a conventionallaboratory is extremely difficult if not nearlyimpossible. Physical processes in the universeclearly fall into two classes: those that have areadily accessible laboratory equivalent (e. g.,spectroscopy, nuclear reactions), and those thatdo not (e. g., radiative viscosity, large-scaleturbulence). The phenomena we choose todesignate as hydromagnetic, of which geomag-netism is the prime example, belong to thesecond class. The nearest hydromagnetic phe-nomena beyond the confines of the earth arethose on the sun, particularly in sunspots, buttheir study has shown them to be exceedinglycomplex.

Babcock's now classical investigations ofstellar magnetic fields started as a search formagnetic fields in early-type stars. These werechosen because many of them are known torotate rapidly and a connection between rota-tion and magnetic fields was suspected. Theconditions under which stellar magnetic fieldscan be ascertained are extremely stringent: thestars must be bright enough (>7th mag.), andthe average magnetic field over the visible faceof the star must be at least several hundredgauss. Of the stars investigated, 65 had to berejected as spectroscopically inadequate for theanalysis, 35 yielded measurable magnetic fields,and in another 20 magnetic fields are suspectedbut not yet assured (as of 1953, the date ofBabcock's review). As Babcock points out,this is an extraordinarily large yield even froma purely statistical viewpoint, and the presump-tion is justified that very many stars possessmagnetic fields of appreciable magnitude.There is also reason for the belief, based on the

> Department of Physics, University of Utah, Salt Lake City, Utah.

investigations of A. Deutsch, that phenomenaequivalent to sunspots, but sometimes of muchlarger extent, occur in many stars. Perhaps itis useful to recall that the magnetic field is thedynamically essential feature of a sunspot, asevidenced by the fact that the mechanicalforces exerted by the magnetic field exceed allother forces acting in a sunspot.

The main interest in cosmic magnetic fieldscenters on two problems: first, the way inwhich they influence observable astrophysicalphenomena, stellar atmospheres, interstellargas motions, radio noise, the acceleration ofcosmic rays, and others; and second, the mannerof their generation. The first class of problemslies outside the domain of this paper, but isdealt with elsewhere in this volume. Theearth does furnish us with a prime example ofthe mechanism of generation of large-scalemagnetic fields; indeed, there is reason to be-lieve that geomagnetism lends itself somewhatmore readily to the construction of a consistenttheory than do the phenomena in the solarconvection zone which give rise to sunspot andsolar magnetism, not to speak of the magneticfields of stars whose details are barely, if at all,accessible to observation.

The ubiquitous presence of magnetic fieldsin the universe leads one to accept a model ofthe generation of the earth's magnetic fieldwhich demonstrates that magnetic fields can begenerated, and maintained, by inductive am-plification in an electrically conducting fluid ofsufficiently large dimensions. This is thedynamo theory for which a great deal of ob-servational evidence now exists. In keepingwith the nature of this volume we shall notdwell on the evidence in detail, but shallemphasize the basic dynamical features andtheir implications.

The geomagnetic dynamo is located in theearth's liquid core (diameter 0.55 of that of the

67


earth). There seems little doubt that this core,which seismological evidence has shown to befluid, consists of molten iron. It may be shownby an analysis of the observed field that thesources of the geomagnetic secular variation arelocated near the boundary of the core. Thegeomagnetic secular variation is a very remark-able phenomenon; it is the only effect originat-ing in the interior of the earth whose spectrumhas periods of order of some tens up to somehundreds of years, whereas all geological proc-esses are of course measured in millions, at thebest in hundreds of thousands of years.

The earth's magnetic field (if we ignore thevery small component that originates in theionosphere) consists of a dipole inclined by11K° relative to the earth's axis, and of anirregular part made up of all the higher termsof a spherical-harmonic analysis. If the timedependence of the irregular field is considered,it is found that in all probability this irregularfield does not have a component constant intime. The dipole axis also changes in longitudeand latitude, but much more slowly than thehigher spherical harmonics. There is at presenta good deal of evidence from paleomagnetism(the study of the remnant magnetization ofrocks) to the effect that the present anglebetween the geographical axis and the magneticdipole axis is larger than it was in the (r. m. s.)average over past geological periods, and thatthe statistical distribution of these angles iscentered about the geographical axis. Thisresult indicates the reality of a close connectionbetween the earth's rotation and its magneticfield which the dynamo theory postulates.

The secular variation (which is essentiallythe time derivative of the nondipole componentof the geomagnetic field) originates in a rela-tively thin layer at the top of the core. Thereason for this is that a layer of metal acts as ascreen for time-dependent fields, and the periodsof the secular variation are such that they wouldbe effectively screened off by a layer, say, 100km thick. Thus the secular variation cannotgive us information about the interior of thecore where the essential part of the dynamomechanism is located, but it does teach ussomething about the outer layers of the coreand, incidentally, about the relation of the core

and the solid part of the earth, the mantle,surrounding it.

In order to explain the observations it isnecessary to assume that the mantle has a smallelectrical conductivity. This is plausible con-sidering the temperatures of perhaps 2,000-4,000° prevailing in the lower mantle. Some ofthe magnetic field generated in the core onpenetrating into the mantle sets up a mechan-ical force there. As a result of the action ofthis force, there is a small difference in angularvelocity between mantle and core, the mantlerotating slightly faster than the top layers ofthe core. In the stationary state a balance isachieved between the accelerating force justmentioned and an opposing eddy friction, thelatter probably also of a magnetic nature. Thedifferential rotation has been observed andanalyzed; it is known as the westerly driftbecause for an observer stationed at the earth'ssurface, the individual features of the irregularmagnetic field seem to undergo a slow averagemotion toward the west. It has also beenshown that this motion is suffering fluctuationsof moderate size over periods of 10-20 years.These fluctuations appear correlated with thefluctuations in the earth's (more precisely, themantle's) rate of rotation which are directlyobserved in positional astronomy of the moonand planets. Theoretical analysis has led tothe result that these fluctuations in the earth'sangular velocity can be satisfactorily explainedin terms of a slightly variable magnetic couplingbetween core and mantle. The required fluc-tuations of the magnetic fields near the bound-ary of the core are quite in keeping with whatwe might expect on the basis of our generalknowledge of the earth's magnetic field. Thefluctuations may be either in the over-all fieldproducing a torque on the mantle or may bedue to changes in the (magnetic) eddy frictionwhich counterbalances the accelerating torque.In any event, this explanation of an importantastronomical phenomenon seems now to rest ona sound geophysical basis.

As we go into the interior of the core thereappears a type of magnetic field which has nocounterpart on the outside, the so-called toroidalfield. In a pure toroidal field the field lineswould be circles about the earth's axis (the


actual field in the earth is a superposition ofmore than one type and hence more compli-cated). The toroidal field, since it cannot beobserved directly, had to be inferred from so-lutions of the electromagnetic field equations,but its existence does not underlie any seriousdoubt since its existence is mathematically nec-essary. The toroidal field apparently is appreci-ably larger than the dipole field in the interiorof the core, being perhaps of order 20-40 gauss.Its counterpart also exists in the solar convec-tion zone, and there is reason to believe that itprovides the field from which the sunspot fieldsare ultimately formed.

Some 20 years ago C. T. Cowling showed thata fluid in which the motion is confined to merid-ional planes cannot form a dynamo. This re-sult may be generalized to the statement thatno fluid motion of rotational symmetry aboutan axis can maintain a dynamo. There are goodtheoretical reasons, furthermore, to believe thatno two-dimensional motion, that is, no motionin which the fluid particles are confined to sur-faces, is capable of providing dynamo action.Hence dynamos have an essentially three-di-mensional geometry; this makes the structureand dynamics of even the simplest models socomplex that we prefer not to expound themhere in any detail. Instead, we shall confineourselves to a discussion of the essential in-gredients of dynamos, the general theorems onwhich their operation is based and which maketheir existence plausible. So far as we can judge,these principles are most likely to be the samefor all hydromagnetic dynamos, the terrestrialone as well as the solar and stellar ones.

The simplest formal conditions appear in thelimit of infinite electrical conductivity. One canthen show that

ItJ B n d<r= (1)

a theorem which states that the magnetic fluxnormal to a given arbitrary surface remains con-stant in time as this surface is carried with thefluid. This result, first given by Cowling, hassometimes been expressed by saying that themagnetic lines of force are "frozen" in the fluidand are bodily carried along by it. It mayreadily be seen, then, that the magnetic field can

be sheared or twisted locally by the fluid motionso that amplification results. It remains to bedemonstrated that amplificatory processes canbe organized in such a way that they can main-tain the magnetic fields observed on the earth,sun, and stars. This turns out to be an exceed-ingly difficult mathematical problem. Thosewho seek a formal solution after the manner ofexistence proofs for the solutions of differentialequations are likely to remain disappointed.The problem is essentially nonlinear (clearly, nolinear system can exhibit self-sustained ampli-fication) and our understanding of the behaviorof nonlinear dynamical systems is still in an ex-tremely rudimentary state. Those making ananalysis must, for the time being, be contentwith inadequate rigor and must rely to someextent on intuition, keeping in mind that thereare definite phenomena waiting for explanation.The present writer finds this situation ratherchallenging to the theoretician, but he realizesthat it does not always appeal to the moremathematically inclined analyst.

Equation (1) applies only to a medium ofinfinite conductivity. In such a fluid we couldnot have a steadily operating dynamo becausethe continued stretching and twisting of linesof force would (as Bondi and Gold haveremarked) result only in a progressive confusionof magnetic fields, not in a stable pattern.In a fluid of finite electrical conductivity, theright-hand side of (1) is no longer zero. Theratio of the right-hand to the left-hand sideis then measured by a nondimensional quantityknown as the "magnetic Reynolds number,"

T» T TT (n\

where L and V are length and velocity, used inthe manner known from the conventionalhydrodynamic Reynolds number, and 0- isthe conductivity of the fluid (in e. m. u.).Rm is large, that is, the conservation theorem(1) is approximately fulfilled when we aredealing with large linear dimensions. In thecore of the earth and still more in the stare,Rm is as a rule quite large numerically. Thisfact can be expressed in a rather concretephysical language by saying that for largeRm, the time required for spontaneous decayof the electromagnetic fields in the fluid is


long compared to the periods of the mechanicalfluid motion; Rm as defined by (2) can be shownto be indeed the ratio of these two time con-stants. In the laboratory Rm is as a rule numer-ically small; this cannot readily be remediedsince all three factors on the right of (2) areclearly limited in laboratory experiments.Since the functioning of hydromagnetic dyna-mos depends essentially on the field's beingcarried along by the fluid, that is on largeRm, it follows that hydromagnetic dynamosare essentially a matter of cosmic physics;namely, of very large linear dimensions. Thissubstantiates our previous contention thatthe earth is the nearest thing to a laboratoryof bydromagnetism, at least so far as dynamosare concerned.

If the magnetic lines of force are continuouslystretched and twisted, as is the case in theturbulent regime of a conducting fluid, theaverage result may be shown to be amplifica-tion of the existing field. This increase islikely to stop when a state of equipartition isachieved where there is an amount of averagemagnetic energy equal to the average kineticenergy of the fluid motion. Hydromagneticturbulence fields are of great astrophysicalinterest (they have been investigated byChandrasekhar) but it is clear that they do notconstitute dynamos. The latter are called onto produce and maintain observed magneticfields which are of a high degree of regularity;e. g., the earth's dipole field.

We have mentioned before that a fluid canact as a hydromagnetic dynamo only if itsmotions are essentially three-dimensional. Thusa stably stratified fluid where motions can atbest take place in the approximate direction ofequipotential surfaces cannot act as a dynamo.Under the actual conditions of geophysics andastrophysics the fluid must be thermallyunstable and hence in convective motion.We do not know the exact causes of convectionin the earth's core. We are better off withregard to the sun and stars; there we can berather certain that the hydrogen convectionzone is the seat of amplifying processes thatlead to hydromagnetic dynamos.

A relation between rotation and dynamoaction is clearly indicated by the behavior ofthe earth's magnetic field and is corroborated

by much astrophysical evidence. Theoreticalarguments lead to the conclusion that rotationof the conducting fluid is essential for dynamoaction. If a convective layer rotates, the con-vective transport of angular momentum leadsto an average state where the layer rotatesnonuniformly, the outer parts of the layerrotating less rapidly than the inner. If weassume a dipole field in the interior of theearth's core, such a field will be drawn out bythe nonuniform rotation so as to give rise to atoroidal field. The toroidal field in turn shedsmagnetic eddies which, under the influence ofthe Coriolis force, may be shown to behave insuch a way that reinforcement of the "poloidal"(dipole) field results. Through the theoreticalinvestigations of recent years it has becomevery probable that dynamo action results fromthe interplay of mutual amplification of twotypes of field components, the toroidal andpoloidal fields. They correspond roughly tothe two coils, say the stator and rotor of aconventional electrical generator. In the anal-ysis of the geomagnetic dynamo this mechanismappears fundamental; the stellar dynamos areperhaps somewhat more complicated, butultimately should reduce to a similar type ofinteraction.

A conspicuous feature of dynamo mech-anisms of this type is their low geometricalsymmetry. Convection excludes any motionsthat do not have a radial component, and theCoriolis force which is not invariant withrespect to the interchange of east and westleads to patterns of fluid motion that do nothave rotational symmetry about the earth'saxis. The fluid motions required for the ter-restrial dynamo do not seem to have anygeometrical symmetry, and this is probablycharacteristic of hydromagnetic dynamos ingeneral. Even in technical generators thesystem of wires, commutators, etc., forms alow-symmetry structure. In this requirementof low symmetry, hydromagnetic dynamosdiffer from most of the traditional problemsof classical dynamics where it has been thecustom to make problems accessible to solutionby using symmetry operations for their mathe-matical simplification. This situation mightaccount for the slow pace with which thetheoretical explanation of the phenomena of

NO.I NEW H0RIZ0N8 IN ASTRONOMY 7 1

terrestrial and solar magnetism has developed Referencesin the past, in spite of the fact that many of theobservations are far from recent. It is hardly We have omitted a list of detailed referencesnecessary to say that under these conditions since they may be found in the following fairlythe subject offers many challenging opportu- extensive reviews:nities for analytical investigations of funda- E L S A S 8 E B ( W A L T B B M .

mental importance to astrophysics. The going 1950 R e v M o d Fhy8> v o l ^ p Lwill admittedly be difficult and new analytical ^55 Amer. Journ. Phys., vol. 23, p. 590.techniques seem to be called for before one 1956a. Rev. Mod. Phys., vol. 28, p. 135.can hope for major successes. 1956b. Amer. Journ. Phys., vol. 24, p. 85.

TimeBy William Markowitz1

The problems of chief interest concerningtime are the precise nature of the variations inthe speed of rotation of the earth and the re-lation between the gravitational and atomictime scales.

The first problem deals with the basic proper-ties of the earth. The second deals with a fund-amental physical property of the universe. Al-though we do know that variations in the speedof rotation of the earth occur, we are not certainof the details. In fact, within the last few yearsmany of the ideas on the subject have been dras-tically revised. As to the second problem, wehave no idea at all, at present, whether or notthe gravitational and atomic time scales arethe same; this must be decided by experiment.

It has been known for some time that thespeed of rotation of the earth is subject tochanges, which may be classified as periodic, ir-regular, and secular. These variations madethe mean solar second unsatisfactory as thefundamental unit of time for scientific and tech-nical purposes. Accordingly, in September 1955at Dublin, the International Astronomical Un-ion redefined the unit of time as a specifiedfraction of the tropical year for 1900.0. Thenew unit is identical with the second of Ephem-eris Time, and is obtained in practice from ob-servations of the moon.

I shall briefly describe recent developmentsconcerning studies of the rotation of the earthand the instruments and clocks used in deter-mining time.

The yearly and semiyearly periodic termswere found by Stoyko at Paris in 1937 and wereconfirmed at Greenwich. Later, large changesin these terms from year to year were indicated.However, a homogeneous solution based on ob-servations made with the photographic zenithtubes (PZT's) of the U. S. Naval Observatory(Markowitz, 1955) showed little change from

1 Director Time Service, T7. S. Naval Observatory, Washington, D. C.

year to year for the interval 1951-1954. Short-period terms due to tidal action of the moonwere also found.

It was formerly believed that sudden changesin speed of rotation as large as several milli-seconds per day occurred at irregular intervals.Brouwer (1952), however, has been led to thehypothesis that the irregular changes are causedby small, cumulative, random changes in speedof rotation and that the large sudden changesdo not occur.

Tidal friction tends to slowly diminish thespeed of rotation. Holmberg (1952) has recent-ly pointed out that atmospheric tides may in-crease the speed of rotation, an idea due toLord Kelvin. Hence, we do not know whetherthe earth is slowing down or speeding up.

The quartz-crystal clock, which has replacedthe pendulum clock in precise timekeeping, hadbeen developed by 1950 to the point where itcould reliably be used to determine the periodicterms in the speed of rotation of the earth. In1955 the announcement was made that anatomic standard of high precision had been con-structed which can be compared with astro-nomical observation (Essen and Parry, 1955).

A program is now under way to obtain thefrequency of the atomic standard in terms of theEphemeris second as determined at the U. S.Naval Observatory with the dual-rate moon po-sition camera. If the frequency changes in thecourse of time, the gravitational and atomic timescales are not alike. An answer may be ob-tained within 10 years.

The atomic standard will also be used tostudy variations in the speed of rotation of theearth by comparing Universal Time with atomictime.

Thus, we shall shortly be comparing atomictime with astronomical time—Ephemeris andUniversal—to provide information on the prob-lems cited. The atomic standard apparently

73


reproduces the behavior of the cesium atom toone part in 109, and higher accuracies willprobably be attained. At present the fre-quencies that can be obtained from astronomicalobservation have precisions of several parts in109. The probable error of the differencesbetween atomic time and astronomical timeswill be due ahnost entirely to uncertainties inthe latter. Hence, it is the astronomical deter-mination of time which must be improved. Anartificial satellite which does not move withinthe earth's atmosphere could provide anotherdetermination of Ephemeris Time. We mustwait and see, however, what particular satellitesare created and how well the position can bedetermined. In general, we must look forwardto continuing and improving the present tech-niques in time determination.

We shall now consider what needs to be done.The Committee on Needs in Astronomy askswhere financial aid might prove most beneficialin aiding progress. Dr. Struve (1955), in hisarticle "The General Needs in Astronomy,"asks a more basic question; namely, shouldmore attention be given to the type of positionalastronomy carried out at the U. S. NavalObservatory, and should more research be donein this country with respect to the earth as abody?

I should like to answer these questions forthe fields of meridian astronomy and for timedetermination. Both types of observation pro-vide basic astronomical data which can not beobtained in any other way. Hence, researchin both fields should be encouraged in orderthat results of the highest possible precisionmay be obtained.

The determinations of time and of funda-mental positions have become highly specializedfields, and are carried on in the United Statesonly at the Naval Observatory. The observingtechniques have been improved over the years,and other astronomers in the United Stateshave been relieved of the necessity of doingsimilar work. On the other hand, these as-tronomers have become isolated from develop-ments in these fields. In the United States

instruction is not given at any school or observa-tory in time determination or meridianastronomy as currently practiced. The endresult is that, because of the general shortageof scientists and of astronomers as a whole,there is a particularly acute shortage of as-tronomers entering these fields.

For the past few years the Naval Observatoryhas not been able to fill vacancies in the TimeService and Transit Circle Divisions with as-tronomers, and the outlook for the future is notencouraging. What is desired is that enoughastronomers be trained so that work in thesefields can be actively pursued at the NavalObservatory, and so that, in addition, research—at least in theoretical aspects of the problems—can be pursued at some other observatories orschools.

From time to time foreign astronomers andstudents have come to the Naval Observatoryfor study and research. Such visits could alsobe made, with considerable benefit, by Americanastronomers and students. The minimum use-ful period would be about 2 weeks, while forresearch and thesis purposes at least a semesterwould be required.

Hence, if money is to be spent to promoteadvances in these fields, it can best take theform of grants for study at the Naval Observa-tory. Grants could be made to teachers andwriters of textbooks in astronomy and even toa few undergraduates. At the present timemoney is available for research grants, but fewrequests for research in these special fields willbe received until there are more astronomersable to use the grants.

ReferencesBBOUWBH, D.

1952. Astron. Journ., vol. 57, p. 125.ESSEN, L., AND PARRT, J. V. L.

1955. Nature, vol. 176, p. 280.HOLMBEKG, E. R. R.

1952. Monthly Notices, Geophys. Suppl., vol. 6,No. 6.

MABKOWITZ, W.1955. Astron. Journ., vol. 60, p. 171.

STRUVE, O.1955. Publ. Astron. Soc. Pacific, vol. 67, p. 214.

Geographical PositionsBy John A. O'Keefe *

The determination of position on the surfaceof the earth is the subject of two branches ofastronomy; namely, field astronomy and ge-odesy. The latter is not always considered asa branch of astronomy, but Simon Newcombin the Encyclopedia Britannica has given us aprecedent for doing so; and it seems veryreasonable on historical and logical grounds.

It is the function of field astronomy to pro-vide the initial orientation and position for thesurveys which determine geographical position.Geodesy then extrapolates these positions bymeans of measurements on the ground to covera whole country, or even a whole continent,without further reference to the stars, exceptfor orientation. The role of extrapolation ingeodesy is unique, so far as I know, in science.If it were physically possible, the geodesistswould certainly prefer to have only a singleastronomical determination for the whole worldand to rely on geodesy for all other determina-tions. The reason for this is, of course, thatbecause of the irregularity of the earth's crust,the direction of the vertical at any point is notexactly what would have been expected on thebasis of a perfectly ellipsoidal earth. So longas we are thinking only of navigation at seathese errors, which very rarely amount to asmuch as one nautical mile, are insignificant.But when we deal with precise land surveys,we can extrapolate relative positions halfwayround the world from the measurements oftriangulation more precisely than they can befound from on-the-spot measurements by fieldastronomy.

As a consequence, vast nets of triangulationhave developed which must, for the best results,be adjusted simultaneously. The feat of adjust-ing systems of equations that involve severalthousand unknowns has been accomplishedrepeatedly in the last few years.

1 U. 8. Army Map Service, Washington, D. O.

Because he cannot measure vertical angleswith anything like the precision with whichhorizontal angles are measured, the geodesistlives in a mathematical world which is exactlylike the Flatland of Abbot's famous book, solong as it is a question of geographical positions.He knows the lengths of his lines and theangles between them, but he does not know thespace configuration of the surface upon whichhe works—the sea-level surface and its pro-longation un4er the land. The maneuvers bywhich he succeeds in avoiding the unanswerablequestions about the space situation bear aremarkable resemblance to the squirming ofthe relativists when asked how their curvedspace would look from the outside. The re-semblance is a family one, since C. F. Gauss isthe father of both systems of thought.

Of course there are techniques for findingheights, but these are only with relation tothis sea-level surface of unknown form.

The principal application of the techniquesof field astronomy in geodesy is therefore indetermining the form of the sea-level surface ofthe earth, for each measurement of an astron-omical position determines the slope of the sea-level surface at a point. The difference be-tween the astronomically measured position andthe geographical position determined geodeti-cally is, in fact, the slope of the sea-level surfaceat that point (if we assume that the slope atthe initial point was zero). From a large num-ber of such determinations of slope, the formof the sea-level surface can be found byintegration.

Finally, we may also seek to find the formof the sea-level surface of the earth from themeasurements of gravity. There is a famoustheorem, called Stokes' theorem, which con-nects the form of the sea-level surface of theearth with the measured values of the intensityof gravity.

76


Thus the classical approach was to deal interms of known angles and known distances ona largely unknown surface, the determinationof whose form was left to a small group ofgeophysicists, since the mathematicians hadfound ways to bypass the requirement for it.By extrapolation, vaster and vaster systemsof triangulation have been constructed, untilnow we have most of the Western Hemisphereon one system, and most of the Eastern Hemis-phere on another.

The new techniques

The geodimeter is rapidly replacing the invartape for the measurement of baselines. Thegeodimeter is essentially the result of reversingthe Fizeau method of getting the velocity oflight so that, from the known value of thevelocity, the distance is found. The toothedwheel is replaced by a Kerr cell, modulatedby ultra high-frequency voltages. The newtechnique has proved remarkably successfulin trials over the last few years, since it measuresdirectly the triangulation sides. The old invartape method could be employed only on rela-tively short lines, which then had to be relatedto the long lines of the regular chains by anelaborate base-expansion net.

The employment of radar for the measure-ment of great distances is rapidly expanding.The most useful tool at the present time is themodification of the wartime Shoran, which isknown as Hiran. The principal new featureof Hiran is the attempt to reduce the signalintensity correction by the use of gain control.The Mediterranean has been bridged by thismethod from Crete to Libya; the North Seafrom Scotland to Norway; and the Straits ofFlorida from Key West to Cuba.

For yet longer lines, radar techniques donot seem to work. The methods that go overthe horizon depend on the ground wave; andthe velocity of the ground wave depends, itseems, on the nature of the ground. Repeat-ability is rather easily secured; but the inter-pretation of the results offers considerabledifficulty at the present.

The moon is now being employed for meas-ures over the horizon. The parallax of themoon, unlike that of any other celestial object,is large enough to furnish a method for the

determination of place on the earth. Thereare several ways of using this fact; each ofthem deals in its own way with the twindifficulties of the remoteness of the moon andthe irregularity of its surface.

The most perspicuous method is that ofMarkowitz, who simply photographs the moonin a star-field and measures the distances frompoints of the moon's limb to the stars. Simpleas the theory is, its practice was not mastereduntil recently. The light of the moon must bereduced by a filter so that it can be observed onthe same photograph as an eighth magnitudestar, 100 million times fainter. The motion ofthe moon during the time required to obtain animage of an eighth magnitude star amounts toseveral miles; this must be allowed for by aspecial optical device which, in effect, arreststhe motion of the moon relative to the stars fora few seconds. The measurements are madewith great care with a precise measuring engine,and the effects of limb irregularities are takenout by averaging about 30-odd points. Theresult of a pair of plates has a probable error of0.15 second of arc in each coordinate, whichamounts to about 900 feet on the ground be-cause of the moon's remoteness. To obtainmore precise results on relative positions, it isnow planned to obtain a long series of obser-vations.

A second method, which aims to obtaingreater precision in the individual observations,makes use of eclipses. Near the moments ofsecond and third contact, the relative positionsof the sun and moon may be determined fromphotometric measurements of the flash spec-trum or of the Baily beads. Since the limbs ofthe sun and moon are alined in space, there isno disturbance from atmospheric refraction ofthis relation. In addition, the resolutiontheoretically obtainable far exceeds the resolu-tion of the best telescopes.

The weakness of this method is that thephotometric observations are disturbed byscintillations of the beads. This is a seriousmatter because the diminution of brightness atthe sun's limb is not instantaneous; hence it isnecessary to fix an arbitrary level of brightnessas corresponding to the limb. The precisedetermination of the moment when this bright-ness is reached is troubled by the scintillation.


The precision of a tenth of a second in arc whichis required for the success of these methods isthus difficult to reach even for an individualbead. For the elimination of the irregularitiesof the lunar limb, reliance is placed on theaveraging of large numbers of points.

A very similar technique starts from obser-vations of the occultation of stars by the moon.For precise results it is not sufficient to rely onvisual observations because of the relativelylong and uncertain reaction time of the humannervous system. Photographic observationsare likewise excluded because of the relativelylow efficiency of the photographic plate indetecting faint stars in a short time. The onlypractical method has turned out to be photo-electric.

As compared with eclipses, occultations arerelatively sharp and well-defined phenomena,consequently there is relatively little disturb-ance from scintillation. The precision withwhich the angles can be measured is of theorder of 0.005 second of arc, which partly com-pensates for the moon's remoteness. On theother hand, there is no chance to allow for thelunar irregularities by averaging, since a singleoccultation refers to a single point of the moon'slimb. To remove the effects of the irregu-larities, it has been found necessary to placethe observers in such a way that all observersof a given occultation see the star disappearbehind the same point of the moon's limb.This means that the telescopes must be port-able and that special computations must bemade to determine the points at which theyare to be placed. Not more than two occul-tations can be observed per month by a singleastronomical party because of the problems ofmovement and survey.

The lunar methods of determining positiondo not fit into the conventional picture of theold triangulation. They do not fix either thedirection or the distance between the pointsinvolved; rather, they yield condition equa-tions involving both the horizontal and thevertical positions. They represent the ap-proach of a new point of view in geodesy.

The geodesy of the futureA development that will most certainly comewith the future is the three-dimensional attack

on geodetic problems. In aerial photographyas a method of mapmaking, we have alreadyseen the entry of the three-dimensional ap-proach into the actual preparation of thetopographic map. With the development ofaerial triangulation, we are witnessing in ourtimes the gradual but steady superseding of thelower order triangulation. At present, in ad-vanced mapping projects, only the first-ordertriangulation is felt to be required, unlessthere is a question of cadastral mapping.Even for cadastral projects, there are reportsthat the necessary accuracy has been attainedby aerial methods.

The most important change in the geodesyof the future will come, perhaps, from theapplication of the satellite. Its use will pre-sent a genuinely three-dimensional problem, sothat the determination of height with respectto the center of the earth will be on a par withthe determination of ground position.

The satellite can be employed like the air-craft of the Hiran and flare observations.That is to say, its position can be fixed fromone set of ground stations while another set ofpositions in a new area is fixed from the satel-lite. This procedure parallels the procedurein flare triangulation, where the aircraft posi-tion is fixed by one group of stations by inter-section while the positions of the unknownstations are fixed by resection (three-point fix)from the aircraft. However, there is oneimportant difference. In speaking of positionsin the flare problem, we mean only the hori-zontal coordinates; in fact, the vertical coordi-nate is not normally measured. In the satel-lite case, the vertical angle will be measuredalong with the horizontal angle, since the alti-tude above the horizon will be sufficient topermit this; and hence we shall have what thephotogrammetrists call the problems of spaceintersection and space resection.

There is another difference, however, whicharises from the fact that the satellite positioncan be successfully extrapolated forward, orinterpolated, so that the observations of inter-section and resection do not have to be simulta-neous, as they must be in the case of aircraft.This greatly increases the flexibility and rangeof the method.

7 8 SMITHSONIAN CONTRIBUTIONS TO ASTROPHYSICS VOL. a

In addition, the satellite can be used to deter- expressed in latitudes and longitudes. Al-mine more precisely the parameters that fix the though this question could be avoided by thesize and shape of the earth. This subject, which use of rectangular coordinates, as a practicalbelongs under another heading in this book, is matter it is not; and hence we can expect thatyet germane to the discussion of geographical the establishment of definitive values of thepositions because the assumed numerical values earth's semimajor axis and of flattening willof the size and shape of the earth are formally materially assist in the international coordina-involved in all surveys or maps which are tion of survey and map material.

Celestial MechanicsBy G. M. Clemence1

The unsolved problems in celestial mechanicsare very numerous; in fact, relatively few prob-lems have been solved with all the rigor, gen-erality, and elegance that may be desired.

The problems are of widely varying degreesof generality. In the most general case we maystudy the character of the motions of any sys-tem of bodies, real or imagined, under any initialconditions whatever. In the least general case,we must be able to calculate the actual orbit ofa particular body for .a limited time and tocompare it with observed positions of the body;our purpose in this case may be either to findthe body on a future occasion or to study thediscrepancies between theory and observation.

It probably is not possible to say that onesuch problem is more important than another,in an absolute sense. I have selected for dis-cussion here two that seem especially interestingbecause of the many practical applications thatcould immediately be made, once the solutionsbecome known, and also because fast calculatingmachines will be of no assistance in solvingthem. Thus I may disprove the common beliefthat such machines can supply all the answers toproblems in celestial mechanics.

The first may be called the problem of thegeneralized planetary theory. The solar portionof the lunar theory has been brought to a highdegree of formal perfection; the coordinates ofthe moon relative to the earth, as affected bythe disturbing action of the sun, are representedas the sum of trigonometric series, each term ofwhich is strictly periodic with a constant co-efficient. Thus the series may be evaluated forany value of the time whatever. In the develop-ment of the series, advantage is taken of thesmallness of the ratio of the distances of themoon from the earth and sun, with expansionsbeing made in powers of this ratio.

1 Director, Nautical Almanac, TJ. 8. Naval Observatory, Washington,D.C.

The heliocentric motion of a planet under thedisturbing action of another planet is a verydifferent case, since the corresponding ratio isso large in general that expansions in its powersare not practicable. In the methods so far used,the difficulty has been overcome by allowing thetime and its positive integral powers, reckonedfrom an arbitrarily chosen initial epoch, to ap-pear as factors of the trigonometric terms, out-side of the arguments. Thus the terms lose theirstrictly periodic character. As the time in-creases without limit so do the coefficients, andthe series necessarily loses its physical signifi-cance at times very remote from the initialepoch. In the applications of the method so farmade, the practical value is limited to a fewthousand years at most.

The ability to construct a planetary theoryhaving the same generality as the solar' portionof the lunar theory has long been recognized asvery desirable, and several unsuccessful at-tempts have been made to devise a practicablemethod. The failures should not deter anyonefrom taking up the problem again. The quali-fications needed are a sound basic knowledge ofthe principles of celestial mechanics and of theart of practical computation, acquaintance withthe power and limitations of electronic calcu-lating machines, and a readiness to discard thetraditional methods for new and unconventionalones.

The second problem to be mentioned is thatof adapting the conventional planetary theoryto automatic electronic calculating machines.By conventional planetary theory is meant thekind already mentioned, where the time and itspowers appear as factors of the periodic terms.In spite of its limitations the theory is of greatvalue and deserves many more applicationsthan it has had. The principal difficulty inadapting it to automatic calculation is thatthe most discriminating judgment is required

79

80 SMITHSONIAN CONTRIBUTIONS TO ASTROPHYSICS VOUl

at all stages of the work if the amount ofarithmetic is to be kept within reasonablebounds. For example, the actual calculationsinvolved in Hill's theory of Jupiter and Saturn,which occupied him for 8 years, could be per-formed in a very few hours with a modernmachine if it were practicable to incorporateall of Hill's judgments in the automatic pro-gram. But to devise such an elaborate pro-gram might well require 8 years of detailedplanning, with even more liability to error thanwas the case with Hill's own work. Theobvious alternative, to simplify the programat the cost of letting the machine do unneces-sary calculations, is not attractive; it appearsthat with the most simple program conceivablethe machine would do about a million times asmuch arithmetic as is necessary, and that with

the sort of program that could be laid out in ayear the machine would still work for somethousands of hours. Another possibility is tostart and stop the machine a thousand timesin the course of the work in order to give theastronomer an opportunity of exercising hisjudgment. But the most attractive methodwould be to recast the form of the developmentscompletely so as to avoid the numerous multi-plications of pairs of trigonometric series,which have to be integrated term by term, andto substitute for them a more easily controllednumerical representation of the disturbingforces. I t may reasonably be expected, al-though it is not certain, that the new methodwould involve more numerical work than theold; we would willingly accept any increase upto, say, a hundredfold.

MeteoritesBy John S. Rinehart1

Meteorites are bits and pieces of extrater-restrial matter which have survived the rigorsof their passage through the earth's atmosphere.Small ones and some quite large ones arerecovered. They are our only ponderableextraterrestrial material. As such, they are astorehouse of information to the astronomer.

The science of meteoritics embraces a widediversity of technologies: geochemistry, geology,physics, chemistry, aerodynamics, astronomy,astrophysics, mineralogy, and metallurgy.

The meteoriticist carries out a variety ofactivities. He may simply collect, catalog,and display meteorites. He may study theirmetallurgical and crystallographic structuresand cogitate on the cosmological significanceof his findings; he may measure the relativeabundances of chemical elements in meteoritesand try to relate them to the universe as awhole; he may investigate the craters thathave been formed by meteorites striking theearth, or the debris that a meteorite spewedabout as it hurtled through the air.

In contrast to many other branches ofastronomy, meteoritics, in a broad sense, hasbeen seriously neglected by scientific investiga-tors. Although much has been written aboutmeteoritics in recent years, we still lack a wellintegrated concept of the science. The goalof future studies must be to effect this integra-tion. Before this can be accomplished, amultitude of specific problems will have to bedefined and solved.

Where do they come from, these metallicand stony calling cards from outer space?What are they like? How many are there?The answers to these questions will tell us agreat deal about the origin of the solar system,the structure of the major and minor planets, andthe relative cosmic abundances of the elements.

' Astrophysics! Observatory. Smithsonian Institution, Cambridge,M M .

The classification and description of mete-orites is in good order. Meteorites are namedaccording to the locality in which they arefound. They may be classed roughly as irons,stony-irons, and stones. The irons containthree main constituents in varying proportions:kamacite, which corresponds to the a-phase inthe pure iron-nickel alloys (up to 6.2 percentnickel by weight); taenite, corresponding to thepure 7-phase; and plessite, which may consistof a microscopic octahedral arrangement ofkamacite and taenite or may be of a granularnature. If the meteorite contains more than6 percent nickel, kamacite and taenite bandsappear interlaced in an octahedral pattern,known as Widmanstatten figures. Occasionallyan iron will have inclusions of iron sulfide, ironphosphide, diamonds, graphite, and otherminerals. The stony irons are of two kinds,those in which the iron is distributed as nodulesand those in which veins of iron are intertwinedwith siliceous material. The stones consistlargely of siliceous material, through whichare interspersed only flakes or fine veins of iron.

Meteorites have been described in a varietyof ways. X-ray studies have established thecrystal structures of kamacite and taeniteand the crystallography of the Widmanstattenfigures. Neumann bands have been identifiedas mechanically twinned regions. Gross com-positions and densities of a large number ofmeteorites are known. The relative abun-dances of elements, particularly the rarer ones,have been found by the use of neutron activa-tion and isotope dilution techniques.

The classical method of fixing the age of ameteorite is to determine the relative amountsof helium, uranium, and thorium it contains.Originally, investigators assumed that all ofthe helium was derived from radioactivedisintegration of uranium and thorium. Re-cently, the existence of helium generated by

81


cosmic rays has been demonstrated and allowedfor in age calculations. An argon-potassiummethod, which is limited in application tostony meteorites, an isotopic lead compositionmethod, and a strontium-rubidium method areall in use. The apparent ages of the stonyphase of meteorites determined by each methodare in substantial agreement, approximately4X109 yr. Iron meteorites are much younger,3X10* yr, according to the uranium-heliummethod; but the tritium measures by Firemanon recently fallen irons indicate that somebillions of years would have been required forcosmic rays to produce the observed He8.These age discrepancies need resolution. Is itpossible, as Urey has suggested, that radio-active He escapes while cosmic-ray He remains—a consequence of radioactive substancescrystallizing only on grain surface?

Physical metallurgy has contributed sig-nificantly, though not extensively, to ourknowledge of the origin and age of meteorites.The approach here is indirect. Considerationis given to the combined pressure and tempera-ture conditions, aging, and constraints thatcould have produced the observed metal-lurgical structures. Especially important arethe Widmanstatten figures, the local concentra-tion of nickel within elements of the figures,and phase transitions indicative of pressurechanges. Research in this area has been ex-ploratory and fragmentary. The need for morework both on terrestrial alloys and meteoritesis great. The influence of pressure on the phase

diagram of the iron-nickel system is not knownnor have the diffusion coefficients of nickel iniron been established. The segregation ofelements on a microscale is a completely un-touched though most fertile field.

Neumann bands, slickensides, distorted lay-ers, crushed nodules, cleaved surfaces—eachhas its story to tell if we can but read it.Mechanical metallurgy is almost completelyunexplored.

At least 10 authenticated large meteoritecraters exist on the earth. The moon containssome 2,000 craters believed to be of meteoriticorigin. These craters are important geologicaland selenographical features. They are, more-over, positive evidence of encounters betweensubstantial chunks of fast-moving planetaryand interplanetary material. Study of thecraters gives us much insight into the cosmo-logical role of such physical encounters. I t isonly recently that such craters have beenrecognized for what they are, and it is stilldifficult in most instances to state unambiguouslythe velocity and mass of the meteorite thatmade the crater. Indeed, it is frequentlydifficult to know for certain that the crateris of meteoritic origin. If we are ever toknow the rate of accretion of matter by theearth, we must first solve the riddles thatreside in and about the craters. We mustexamine the geologic havoc that has beenwrought, we must collect meteoritic debrisfrom within and without the crater, and we mustconduct high-speed impact tests of our own.

MeteorsBy Fred L. Whipple *

In these days of expanded research poten-tialities, the study of meteors has matured as aresearch area and many of the major problemsof yesterday are now solved; for example, thecontroversial problem of hyperbolic meteors.Both the precise photographic measures ofbrighter meteors and the radar measures of thefainter meteors have failed to prove the exist-ence of hyperbolic meteors. If such meteorsexist, they represent a minority below the 1-percent level. Furthermore, the photographicand visual meteors are clearly of cometaryorigin except for a small fraction not exceeding10 percent that may be of asteroidal origin.Most photographic and radio meteors, how-ever, move in direct orbits of relatively lowinclination to the ecliptic; their aphelia tend tolie beyond the asteroid belt and usually beyondJupiter's orbit, but many of the fainter radiometeors exhibit quite small aphelion distances.The cometary meteoric mass is proved to con-sist of extremely fragile, crumbly material thatis probably of very low density. The spectra byP. M. Millman indicate a composition some-thing like that of stony meteorites, but it is un-likely that any sizable pieces have reached theground. Micrometeorites, on the other hand,may well represent the cometary solids.

Although photographic and radio techniqueshave been exploited to an amazing degree in thepast several years, serious gaps remain to bebridged before we can reach a satisfactoryunderstanding of meteoric phenomena and ofthe origin of these tiny bodies from interplane-tary space. I should like to mention brieflycertain of these difficult observational problemsand devote most of my attention to the almostuntouched areas in laboratory and theoreticalwork. As long as these areas are unexplored,our knowledge of meteors, their nature, and

1 Harvard College Observatory, Cambridge, Mass., and Director,Astrophyslca) Observatory, Smithsonian Institution.

their origin will remain conspicuously in-complete.

In the area of two-station direct photographyof meteors for velocities, trajectories, decelera-tion and light curves, the J. B. Baker Super-Schmidt cameras in the Harvard Meteor Pro-gram have largely completed the basic observa-tional research in terms of their present-daycapabilities. Analysis is being carried alongrapidly and effectively by L. G. Jacchia, R. E.McCrosky, and A. F. Cook. I must point out,however, that no serious effort has yet beendevoted to the use of the television image tubein optical meteor methods. There is everyreason to believe that the sensitivity can be in-creased more than an order of magnitude, sincethe transient character of the meteoric radi-ation is highly suited for the present-day photo-electric techniques. Utilization of the imagetube, on the other hand, requires a major effortsince the telescope and the receiver must beintegrated into a new and specialized form.Such a technique shows promise of reaching tothe ninth magnitude visually for very slowmeteors, and to the sixth apparent magnitudefor the fastest meteors. Applied as in the Har-vard Program, the image tube may accomplisha corresponding improvement (some four mag-nitudes) in the registration of persistent meteortrains shortly after the meteors have passed.Thus, measurement of upper winds as a functionof altitude in the range from 80 to 112 kilome-ters could be systematic and highly precise.

The train type of observation is importantnot only for the understanding of upper atmos-pheric turbulence but also for solving the basicproblem of meteoric masses. The "coastingmotion" in meteor trains presents to date theonly direct method of measuring meteoricmasses, via their momentum transfer to theambient air. Either photographic or image-tube methods should be applied with larger base

83


lines than those of the present stations, about50 rather than 20 miles, to clarify the questionas to the average masses associated with mete-ors of a given brightness and, from a morefundamental point of view, the question of thedensities of cometary solids. A single exampleleads to a density of only 0.05 gm/cm3!

I t has not yet been possible to utilize thefull power of the Baker Super-Schmidt meteorcameras for the photography of faint meteorspectra. Relatively coarse transmission rep-lica-gratings with a high concentration of lightin one first order have not yet become availablefor use on fast cameras of large aperture.Coupled with image tubes, such coarse gratingsmight lead to remarkable progress in the under-standing of the spectra of fainter meteors andof train and wake phenomena for brightermeteors and might make it possible to de-termine the composition of cometary solids.

D. W. R. McKinley at Ottawa, Canada, andA. C. B. Lovell with his colleagues at JodrellBank, England, have made remarkable stridesin measuring the velocities of faint meteors byradio techniques. J. G. Davis has now devel-oped a three-station method for measuringradiants, and, therefore, orbits as well. Only amajor effort utilizing this technique can pro-vide the many data critical to our understand-ing of the orbits and origins of meteors, as wellas their interaction with the atmosphere. Inparticular, it is essential that such a systembe used with a number of receiving stations inorder that observations may be made along thefull length of meteor trails to determine ioniza-tion as a function of distance and height, dif-fusion effects at great altitudes, deceleration inthe atmosphere, precise correction to outeratmospheric velocities and orbits, heights, winds,and other important meteoric observables.Such a program is being undertaken at Harvardunder the technical direction of G. S. Hawkins.This technique should be carried to its prac-tical limit.

Fortunately, many radio groups are studyingthe properties of forward-scatter by meteorsbecause of its importance in communication.In the near future, such information needs tobe collated, summarized, and interpreted witha view to increasing our understanding of the

nature, decay, and diffusion of the electronclouds produced by meteoroids.

There is every reason to hope that theplanned Earth Satellite Program of the UnitedStates and the International Geophysical Yearof the National Academy of Sciences willprovide direct above-the-atmosphere measuresof small meteoroids in space, so far undetectableby any of the meteoritic methods. I t isextremely important that such observationsbe made in order to clarify the relationshipsamong comets, meteors, and the zodiacal lightand the effect of corpuscular radiation on smallparticles in interplanetary space.

Although sound observational techniques forthe study of meteors are being utilized vigor-ously, the opposite can be said of the theoreticaland laboratory research which should now beunderway to consolidate these observationalgains. The basic meteoric phenomenon stillremains unexplained in terms of even anapproximate physical theory. This deplorablesituation exists not only because the problemis difficult, with many facets, but because noserious effort has been directed towards thelaboratory determinations of the fundamentaldata required in such a theory. Desperatelyneeded are measures of atomic and molecularcross sections for dissociation, excitation, ioniza-tion, attachment, and recombination for me-teoritic and atmospheric elements in the energyrange up to 103ev. These quantities are need-ed for such elements as silicon, iron, nickel,magnesium, sodium, calcium, manganese, andothers in air, or, better, in pure atmospheresof nitrogen or oxygen. Such laboratory meas-ures coupled with the remarkable progress ofrecent years in shock-tube research and ultrahigh-velocity phenomena should make it pos-sible to utilize the observations of meteors toestablish the beginning, at least, of a satisfactorybasic theory for meteors. The needs in thearea of astroballistics are elaborated in a paperof this volume by R. N. Thomas.

Artificial meteors, either in evacuated rangesor at moderate altitudes in the atmosphere,could provide extremely valuable informationregarding the meteoric phenomena under con-ditions in which the nature and mass of themeteoric body would be known. Such re-search, started under the direction of J- S.


Rinehart at the Naval Ordnance Test Station,Inyokern, Calif., should be pursued vigorouslyto provide the confirmation of theories basedupon the laboratory measures.

In quite a different area of meteoritic studies,laboratory measures would also be of extremevalue. Corpuscular radiation from the sunalmost certainty plays an important role in thedisintegration of meteoroids in space. Werequire laboratory measurements of the sputter-ing phenomena produced by protons in theenergy range 10*-10*ev on common meteoriticsurfaces—stony minerals as well as nickel-iron.A knowledge of the rate of disintegration underthese circumstances would greatly increase ourunderstanding of the cometary meteoroid.Once such a body has been ejected from acomet, it is subjected to corpuscular radiationwhich causes it to spiral inward towards thesun and also to disintegrate. There seemslittle doubt that cometary particles are themajor source of scattering and diffraction ofsunlight in the solar Fraunhofer corona, in thezodiacal light, and in the Gegenschein. Wecannot integrate our information concerningcomets, meteors, the corona, the zodiacal light,and the Gegenschein until we know the effectof corpuscular radiation on small meteoriticparticles.

Similar laboratory measures of atmosphericcomponents impinging on meteoritic materialwith energies in the range 10-103ev wouldprovide a sound foundation for an understand-ing of the interaction of extremely smallmeteoroids with the upper atmosphere. If thesputtering effects produced under these condi-tions are sufficiently small, then "micromete-orites" can indeed pass through the upperatmosphere without destruction; by black-bodyradiation at temperatures below the meltingpoint they can radiate away the heat derivedfrom atmospheric encounter. Such sputteringinformation would be of extreme value ininterpreting the observations of meteoritic dustin the atmosphere and in deep sea oozes. Itwould also help clarify the question of whethermicrometeorites provide condensation nuclei toinitiate heavy rainfall at extremely greataltitudes.

In the laboratory-ballistic area of measure-ment, particularly needed are the values of such

3S1O14 — 30 7

physical quantities as the heat transfer coef-ficient and the luminous efficiency. Thesequantities are, respectively, the percentage ofthe available translational energy of the me-dium that is transferred to the surface of themeteoric body and the percentage energy ofthe ablated material that is transferred intoobservable radiation. A corresponding quan-tity is the ionization efficiency, measuring thenumber of electrons freed in the meteoricprocess by each atom of the meteoroid. Thesequantities are, naturally, functions of thevelocity and very likely are functions of the airdensity and composition of the meteoroid.

The physics of the meteoric process involvesthe integration of physical data, physicalphenomena, and meteoric phenomena into acoherent whole. The development of such atheory has obviously been the goal of muchresearch. In one area, there is hope that such atheory can be developed without so much sub-sidiary information, viz, the theory of the per-sistent meteor train- The evidence pointsstrongly to the conclusion that meteoric energyis transferred rapidly to some retentive agencyin the atmosphere, such as active nitrogen, andis transformed slowly into radiation after themeteoric phenomenon is ended. Such a theorywill be closely integrated with the problems ofthe physics of the upper atmosphere but doesnot seem to be hopeless in terms of our knowl-edge at the moment.

A number of theoretical problems fall inthe realm of celestial mechanics. The problemof the "Jupiter barrier" requires much moreattention. The Jupiter barrier arises from theperturbational disturbances by a large planetwhich tend to prevent a particle spiraling intowards the sun from reducing its apheliondistance below the perihelion distance of theplanet. Perturbations by close approachesto the planet necessarily throw the particleinto an orbit that will return near the point ofencounter, and thus an orbit with aphelionbeyond the planet's orbit. Many of theperturbational problems of interactions betweenthe planets and small particles of the solarsystem take on a stochastic character and, infact, are soluble by modern stochastic theory.

An interesting problem combining stochastictheory and celestial mechanics concerns E.

SMITHSONIAN CONTRIBUTIONS TO ASTROPHYSICS VOL.1

Opik's suggestion that perturbations of aster-oidal material by Mars are responsible for theexistence today of asteroidal material crossingthe earth's orbit. Here we require a morecomprehensive theory of the perturbativeeffects of a smaller planet upon microparticlesin the solar system.

This enumeration of problems in the fieldof meteors is far from complete; in particularit omits a number of important questions inthe category of correlation-interpretation-theory. These questions involve the relation-ship between the orbits of comets and meteors,the distribution of particle orbits to produce theobserved zodiacal light, the nature of the Ge-genschein, the development of meteor streams,identification criteria for meteor streams ofgreat diffusion or low population, and a number

of problems relating to the orbits of comets,meteors, meteorites, and asteroids.

The author hopes that this report, incompletethough it is, will at least tend to eliminatecomplacency with regard to the present com-pleteness of meteoric research.

ReferencesKAISBH, T. R., BD.

1955. Meteors. Pergamon Press, Ltd., London.LOTBLL, A. C. B.

1954. Meteor astronomy. Oxford at the Clar-endon Press, New York.

POBTEB, J. G.1952. Comets and meteor streams. John Wiley

& Sons, New York.RlNEHABT, J. 8.

1954. Behavior of metals under impulsive loads.American Society for Metals, Cleveland,Ohio.

Minor PlanetsBy PaulHerget1

The techniques and methods of observing theminor planets have not improved substantiallyfor more than half a century. Within the lastdecade, however, the development of electroniccomputing machines has increased the possi-bilities for computational work by a factor ofnearly a thousand. Advances in technologyhave improved photographic emulsions so thatwe can observe planets much fainter than thoseobserved formerly, but there is no immediateprospect of improving methods of blinking,measuring, and reductions based on stellar com-parison positions from star catalogs. Minorplanets will continue to be discovered more orless at random, and each one must be followedalmost individually for several months in orderto secure its orbit and to predict the followingopposition.

One of the most difficult problems for theMinor Planet Center occurs when the observerassigns an erroneous identification to the objectwhose position is determined. As fainter andfainter objects are observed, the likelihood ofmaking such misidentifications increases greatlyand can be avoided only by securing a confirm-ing observation from two to four weeks later. Invery doubtful cases even a third observation isneeded so that an independent orbit can becomputed, if necessary. For this same reason,it is now increasingly important to providemuch more accurate ephemeris and perturbationcomputations than in the past.

In 1950 Kuiper launched a large-scale ob-serving project in which the minor planet zoneof the sky, about 23° on each side of the ecliptic,was observed in the neighborhood of oppositioncontinuously for a period of 20 months. Themain objective of this program was to gatherfairly homogeneous data on the magnitudes of

1 Department of Astronomy, University of Cincinnati, Cincinnati,Ohio.

the minor planets and to provide a basis for in-vestigating theories of their origin. A veryimportant concomitant result was a nearly com-plete survey of all minor planet positions, whichproved to be a very valuable standard of refer-ence whenever doubtful cases occurred in theimprovement of orbits. If the computationswere based on actual observations it should bepossible to locate the doubtful planet on thesurvey plates from its computed position and,thereby, provide still more observational data,unless the planet was below the magnitude limitof the plate. On the other hand, if the supposedobservations were spurious it would be impos-sible to locate an object at the computed po-sition on the survey plates. The value of sucha survey is so great that it should be repeated atleast once every decade. For the ordinary mi-nor planet orbits, with sufficient observations inthe past and with good perturbation computa-tions, this should be sufficient to assure accu-rate ephemerides; and it may have the effect ofdisplacing at least half of the ordinary minorplanet observing that now goes on.

There is room for a special type of observingprogram which could be undertaken by someonewho has a sufficiently large telescope, who likesto sleep through the night, and who is willing towork low on the horizon. The purpose wouldbe to increase the length of the observed arc ofnewly discovered minor planets, both at the endof the first opposition in which they are dis-covered and at the beginning of the next follow-ing opposition. A moderately large telescope,something like a 24-inch reflector, would beneeded because the magnitudes will be unusual-ly faint because of the atmospheric extinctionat low altitudes and the disadvantageous phaserelationship nearer to superior conjunction.Also, many of the newly discovered faint planetswill be in the region of perihelion during the first

87


opposition and, due to their eccentricity, will beat a greater heliocentric distance in the secondopposition.

Recent work at the Naval Observatory andby Kuiper has disclosed new problems relatedto the physical constitution and history of theminor planets. Their solution requires con-current photoelectric observations of individualminor planets, including changes in color,magnitude, and polarization, and extendingover a wide range of phase angles. Someevidence already exists that two classes ofminor planets can be recognized from polariza-tion observations. Laboratory work of asimilar nature with known materials and con-ditions should offer some clues to these prob-lems. Theoretical studies are also desirable.

The Minor Planet Center has experimentedwith all available types of computing machinesin recent years, and it has now concluded thatthe perturbations of all ordinary minor planetsmay be most readily handled either by themethod of variation of vectorial constants orby Hansen's method. The computations re-quired are no longer a formidable problem.Similarly, the problems of preliminary orbitsand differential corrections have been reducedto a matter of minutes. The same resultscould easily be attained for all cometcomputations.

The routine aspects of minor planet computa-

tions are so well in hand that it is now possibleto turn to new and more interesting problems,the solution of many of which will be aided bythe experience gained in the work describedabove. An accurately computed dynamicalorbit of Pallas from the time of its discoveryup to the present will permit an independentand highly reliable determination of the valueof the constant of precession. G. W. Hillprepared a list of 13 planets that are especiallysuitable for determining the mass of Jupiter,provided the observations and accurate per-turbations extend over a period of about acentury; work on this problem has begun.Another program, already well underway, willuse the first four minor planets for determiningthe position of the equinox instead of observa-tions of the sun and the major planets. Thisproject also requires accurate computationsof their orbits. More penetrating researchesmay now be made upon the families of minorplanets originally recognized by Hirayama.

Electronic machines have not yet been usedas effectively for the calculation of generalperturbations as for that of special perturbations,but the problem has been attacked and willundoubtedly result in progress in the nearfuture. The availability of electronic com-puting machines now makes the vistas in allthese fields more attractive than at anyprevious time in history.

Planets, Satellites, and CometsBy Gerard P. Kuiper *

The planets and the satellites must bestudied from distances (—1 astronomical unit)that are intermediate between that of the stars(>105 a. u.) and that of the bulk of the earth(< 10~6 a. u.). Likewise, some of the problemsthey present are intermediate—related both toastrophysics and geophysics—but this situationis rather recent.

The following paragraphs, part of a prefaceto a book now in preparation, may serve as abackdrop on which to view some of the 20thcentury problems of planetary astronomy.

Until the introduction of photography with largetelescopes and the birth of astrophysics, both towardthe close of the 19th century, planetary astronomy—with some exceptions—was astronomy. The system ofthe "fixed" stars was of profound interest philosoph-ically; but to the practicing astronomer it was largely aconvenient frame to which the complex motions of theplanets could be referred. The great discoveries inastronomy were made in efforts to interpret the plane-tary motions. Copernicus, Kepler, Newton, Euler,Lagrange, Laplace, Gauss, and Poincar6, to name butthe greatest, created the concept of modern naturalscience while studying the planets.

The phenomenal growth of astrophysics, and theexciting explorations of the galaxy and the observableuniverse, led almost to an abandonment of planetarystudies. The number of astronomers has always beensmall and for better or for worse has not followed theenormous increases of physicists and chemists. Celes-tial mechanics began to look like a nearly finisheddiscipline in which further progress was hopelesslydifficult. Physical observations of planetary surfaces,particularly of Mars, led to controversies and specula-tions. More and more this branch of planetary work,including the study of the moon, became the topicpar excellence of amateurs—who did remarkably wellwith it. The Memoirs of the British AstronomicalAssociation became the chief record for the develop-ment of planetary surface markings. Astronomers withlarge telescopes were so occupied with the engagingproblems of stars, nebulae, binaries, clusters, thegalaxy and the universe, that astronomy almost en-tirely became the science of the stars.

1 Yerkes Observatory, Williams Bay, Wis.

In some respects the science of planetary as-tronomy has greatly changed from its distin-guished predecessor. Eight- or ten-decimal ac-curacy and a formidable theoretical apparatushave often given way to one-decimal accuracyand order-of-magnitude arguments. Observ-ing programs, which occupied the major observ-atories for decades, have been replaced byisolated observations by single astronomers whobuild new apparatus and use it once or twiceunder optimum conditions. "Astronomical ac-curacy" is a compliment born in an age differentfrom ours.

Nevertheless, there has been no breakdownof standards. What has happened is that anew branch of science was born, an offshootof young and vigorous astrophysics, which wasadded to classical planetary astronomy but wasnot immediately integrated with it. And soit happened that 10-decimal accuracy and1-decimal accuracy existed side by side, in afield now occupied by scientists of very differentinterests and training. The study of planetarymotions and the great problem of the stabilityof the solar system rose slowly and laboriouslyon foundations laid by the great masters.New computational techniques proved increas-ingly valuable in the solution of special prob-lems. Dr. Clemence reviews this basic part ofplanetary astronomy elsewhere in this series.

The program of this 20th century additionconsists of a great variety of things that atfirst sight may seem to lack coherence. Ageophysicist, considering the task of a plane-tary astronomer, might well be skeptical.He could reason that, since the objects of studyare some 10s times more distant than the earth,the obtainable information might be 1010

times smaller, and furthermore, that the ab-sence of material contact or near-contact withthe planets deprives the astronomer of severalsources of information: seismic studies, heat

80


flux measurements from the interior, gravitysurveys, magnetic surveys, study of magneticvariations, volcanism, chemical and crystal-lographic analyses of the crust, oceanographicstudies, detailed atmospheric studies, etc.One might wonder what there is left for theastronomer to do.

Classical or gravitational astronomy, how-ever, has put the planets on a coequal basiswith the earth in some aspects. The mass ofthe earth in absolute units is known no betterthan the masses of Venus, Mars, Jupiter,Saturn, Uranus, and the Moon, with Mercuryand Neptune closely following. Nor do weknow the planetary orbits with an accuracygreatly different from that of the earth'sorbit. Hence, the radiation received from thesun is known equally well for all. The albedosof the planets can be determined with con-siderable precision; hence the equilibriumtemperatures can be computed quite well,even if with somewhat higher precision for someof the planets than for the earth, whose albedois difficult to determine accurately. Theperiods of rotation of the planets can be found(Venus is still an exception), often with con-siderable precision, as well as the planetaryobliquities. With the radiation intensity andthe obliquity known, theoretical climatologymay be applied to the planets provided some-thing can be said about planetary composition.This, too, is within the range of possibilities.I t is immaterial whether the source is 109

miles away or just in front of the slit of thespectrograph, provided enough light is avail-able to make a spectrogram of the desireddispersion during, say, one night's exposure.

There are, however, serious handicaps, notresulting from distance but imposed by Natureitself. In the first place, some of the mostabundant atmospheric constituents are homo-nuclear molecules (H2, N2, O2), and because oftheir symmetry the vibration-rotation spectraare forbidden. Since the electronic absorptionspectra H2 and N2 are in the ultraviolet part ofthe spectrum that is cut off by our atmosphere(X<0.3M), we cannot discover these gases onother planets unless such enormous quantitiesare present that pressure effects destroy thesymmetry of the molecules and make theirinfrared spectrum faintly visible (as on Uranus

and Neptune for H2). Nitrogen in our atmos-phere is a good example of invisibility; even atsunset, when the sun's rays pass through 320 kmatm., four-fifths of which is N2, it does not addtelluric lines to the solar spectrum, even at highdispersion. A second handicap affects the inertor noble gases, He, Ne, A, Kr, Xe. They areof great interest for the study of the origin ofthe planets because they will not have combinedwith other elements during geologic time andmust therefore always have been part of theplanetary atmospheres. (This statement doesnot apply, of course, to radiogenic He4, He3,and A40.) Because these gases range in atomicweight from 4 to 130, we can use them—in thecase of the earth, where they can be observed—to determine empirically how the fraction ofatmospheric constituents lost by evaporationduring the process of planet formation dependson molecular weight. Since the original com-position of the protoplanets, from which theplanets condensed, was very probably the cos-mic composition (essentially the composition ofthe solar atmosphere), and since the abundanceof heavy elements, like Fe, Mg, Si, etc., may beestimated, we may compare the terrestrialabundance table with the cosmic abundancetable to obtain the desired ratio, assuming thatno appreciable fraction of the heavy elementswas lost. Now it would be exciting if onecould apply this approach to other planets.But the very property which makes the noblegases so interesting in tracing the early historyof the earth (i. e., their stability in the atomicstate due to the completeness of their outer-most electronic shell) makes them at presentundiscoverable on other planets. Such highenergies are needed to excite these atoms thattheir absorption spectra are far in the inacces-sible ultraviolet.

I t is clear, then, that two important advancesmay have to await the development of anextraterrestrial observing station. If planetaryspectra could be taken from 200- or 300-kmelevation, entirely new and vital informationcould be gotten on the composition of plane-tary atmospheres. One would then get anundistorted picture of atmospheric composition.The handicaps imposed by Nature would havebeen overcome.

Some of this much-desired information can


probably be obtained with a large telescope ata ground observatory. By introducing com-pensation methods, Hiltner has succeeded inmeasuring polarizations down to a few units inthe fourth place. It should be possible, inprinciple, to do the same in spectrophotometry.This would allow one to observe the Ramanspectra of H3 in the Jovian planets and to makea rough determination of the H2/CH4 ratio.Conceivably even the He abundance mightthus be found. Similarly for the terrestrialplanets, N2 and possibly other gases mightbecome observable, at least if they are moreabundant than the strongly scattering CO2. Iteven seems possible, in principle, that the solarlines Mg II at XX2795.5 and 2802.7 A, includingtheir variable emission components, could beobserved regularly from the ground from theRaman lines in the spectrum of Jupiter, atXX3163 and 3172 A. The problem is probablylargely a question of instrumental development.

Related to the problem of atmospheric com-position are at least four others capable ofempirical pursuit. They are the problems ofatmospheric clouds, the planetary albedo andits wavelength dependence, planetary polariza-tion as a function of the phase angle and as afunction of wavelength, and the spectrum ofthe infrared emission of the planet to space,which at least for the terrestrial planets mustbalance the absorbed solar radiation. For theJovian planets a finite fraction of the radiationemitted probably still derives from gravitationalcontraction and from cooling off of the planetaryinterior.

The atmospheric clouds on Jupiter and Saturnare almost certainly atmospheric condensationproducts, as is true for most of the clouds andhaze on Mars. Occasionally Mars has yellowdust clouds also, while the nature of the cloudcover on Venus is still in doubt. Uranus andNeptune probably have no visible clouds, al-though markings have been reported on Nep-tune. The reality of these markings is not con-finned by observation with large telescopes, orby the model required for the atmospheres ofUranus and Neptune to account for the enor-mous quantities of CH4 and H3 observed. Sat-urn has a general haze cover which forms thebottom of the visible atmosphere. It usually isdivided into a number of distinct belts, parallel

to the equator, of slightly different reflectivityand, sometimes, color. In addition, separateclouds are occasionally seen, perhaps one perdecade on the average; they must, of course, behuge to be visible, with diameters about as largeas the entire earth. They seem to be caused bymajor eruptions from the interior and, becauseof the very pronounced differential rotationwithin the planetary atmosphere, they are soonstretched out into belts parallel to the equator,merging with the belts already present. It isquite interesting how the belt system of Saturnhas changed with time. Jupiter, likewise, showsmajor upheavals in its cloud cover, every decadeor so, not at all related to its position in its ec-centric orbit, and therefore not caused by solarheating. These two planets apparently do nothave a sun-induced meteorology, as do terres-trial planets, but one largely governed by re-leased internal heat. It has been pointed outby meteorologists that internal heating is muchmore efficient in stirring an atmosphere thansolar heating from the outside.

Fascinating and unsolved problems are posedby these observations. Do these occasionalbursts come from the deep interior? Then howdo they get through the mantle of solid hydro-gen usually assumed to surround the densercore ? Or is contraction of the mantle itself thecause ? What is the explanation of the brightlycolored clouds—brick red, chocolate brown,blue, amber, white, etc.—observed especiallyduring these upheavals? Is the Red Spot acloud cap over a floating island, and if so, whatfloats in what? Or is it merely a long-livedvortex in the Jupiter envelope? (It is not at-tached to the core because its rotational speedis variable.) These questions, in turn, are re-lated to the unsolved problem of the depth ofthe gaseous envelope and the hydrodynamics ofthe differential rotation on both Jupiter andSaturn. Of the greatest interest and promise isthe recent discovery of a whitish cloud on Ju-piter emitting radio bursts at irregular inter-vals; the power of the bursts measured at 22 Meis 10* times that of a terrestrial lightningdischarge, but the spectrum is quite different.

The cloud problem for Mars is simple bycomparison, and has been largely clarified by acombination of the brightness at different wave-lengths, polarization measures, and studies of


general behavior. By contrast, the identity ofthe Venus cover is not at all obvious. Differentsuggestions have been made: water droplets,dust, and salt particles, but none of these fitall the observational data on polarization, invisual and infrared light, on the brightnessvariations with phase, on the planet's color,and on the appearance of the cloud belts. Inconversations with Professors W. E. Grath ofBonn and P. Harteck of Rensselaer PolytechnicInstitute, the possibility was discussed that theparticles are polymerized C3O2, formed as aresult of the impact of solar ultraviolet radia-tions on an atmosphere largely composed ofCO2. This possibility will be checked by labo-ratory experiments. Here is an exciting prob-lem that, once it is solved, will also deepen ourunderstanding of the history of the earth, sosimilar to Venus in dimensions and mass.

A curious problem is that of the planetaryradiation to space. For a planet or a satellitewithout an atmosphere, there is no difficulty.Black-body radiation will be a good approxima-tion for such an object, covered as it probablyis with fine powder or debris after its longexposure to a vacuum, to solar radiation varyingwith the rotation, to cosmic rays, and to inter-planetary debris. But how does a planet likeUranus or Neptune establish its temperaturebalance with the absorbed solar radiation?What atmospheric constituent radiates effec-tively between 20M and 100/*, where the maxi-mum of the emission should occur? WithH2 a homonuclear molecule and CH4 a sym-metrical molecule, and NH3 and H2O frozenout, a situation arises which is not unlike thatin the upper layers of the earth's atmosphere.There, too, appreciable quantities of (in thiscase ultraviolet) solar radiation are absorbed,but the atoms cannot dispose of this energyby radiating in extensive infrared bands. Theresult is that the atmosphere has a hot upperfringe at nearly 10 times the mean equilibrium-radiation temperature of the earth ( ^ 250° K).

The thermal radiations in the radio-frequencyrange will be of special interest since they willpresumably arise from the solid surfaces, andtherefore give a measure of the depths of theatmospheres. Thermal radio emission of Marswill determine the mean daily temperature and,

by implication, the surface temperature atnight.

Enough has probably been said to illustratethe wide variety of planetary problems, manyof them unsolved. One could mention others:precise diameter determinations and resultingplanetary densities; their interpretation interms of models based on the physics of matterunder very high compression; determinationof the moments of inertia and the resultingrefinements of the planetary models; the relatedproblem of planetary composition; and relationof the interior to the envelope for the Jovianplanets. In a separate class is the problem ofthe dark Martian features and their variations,and the question as to whether they are organicor inorganic.

Satellites present a host of problems of theirown. Needless to say, the moon is foremostamong them, but they are all different andinteresting. Titan, with its orange color andits atmosphere of methane; Triton, with itsslightly yellowish color and its probable at-mosphere, of unidentified composition; Io,with its bright orange color, but lack of at-mosphere; the snow satellites of Saturn; thegreatly different mean densities and, hence,compositions of the satellites; the great varietyof albedos, from near unity for the innersatellites of Saturn to 0.07 for the moon; thedifferences in albedo sometimes found fordifferent hemispheres of the same body, ex-ceptionally large for Iapetus; the spots on theJupiter satellites and on the moon; etc. TheRings of Saturn are unique and of unrivaledbeauty in a large telescope; their snowy com-position seems reasonably well established.One can understand why Jupiter does not havesuch a ring. The high density of Io indicatesthat the surroundings of Jupiter were quitehot at the time Io formed; and, even if somesnow later condensed within the Roche limitof Jupiter, it would almost certainly haveevaporated subsequently by solar radiation.Saturn's ring appears to be just safe fromsuch destruction. But why don't Uranus andNeptune have rings? That is very puzzlingand further work must be done to lower thethresholds set by present observations.

Comets are much stranger objects thanplanets or satellites. Their ephemeral nature


shows that they "do not belong" near theearth; that they were not formed there. Theexposure to strong solar heat and corpuscularradiation causes the comets to disintegraterapidly, and show heads and long tails that areamong the weirdest structures seen in Nature.No wonder pamphlets were written in past ageson whether the appearance of a comet mightmean the death of a king. Comets are unad-justed to their surroundings, as we observethem near the earth; for this very reason theydemonstrate, by their violent reactions, certainproperties of these surroundings, as the oldand adjusted planets and satellites no longer do.Comets have thus become of the greatestinterest in the study of the properties of solarradiations and interplanetary magnetic fields,as well as for their own sake. The unpredicta-bility of the occurrence of a "new" comet andthe irregularity of the stimulating solar beamsadd to the excitement when a bright comet

appears, which has not been very often inrecent years. A good bright comet, like thatof 1882, would be a major scientific event.

Problems of evolution and origin are as com-pelling in solar-system studies as they are instudies of stars and galaxies. It seems as ifsome of these are being solved. On the otherhand, a good look at the problems in geophysicswill temper any excess optimism. Much ofwhat we would like to know about planets,satellites, and comets is hidden from our view;and even if we could see it, explanations wouldnot follow automatically, as geophysics shows.Clearly, one must be satisfied with limitedobjectives and with explanations and theoriesof a general kind. But because of the manyrepercussions these problems have on geophysicsand related sciences, any advance will also havea general interest. The limits imposed bytelescopic equipment and other practical con-siderations have by no means been reached.

381014—56-

Airglow and AuroraBy A. B. Meinelx

The fact that the background of the moonlessnight sky is not entirely dark has been knownfor centuries, while references to the auroraeextend back into the dim recesses of Nordicantiquity. Yet only since the turn of the 20thcentury have these phenomena been the subjectof systematic study. The presence of anemission spectrum for the aurora was recognizedas early as 1867 when Angstrom observed andmeasured the wave length of the green auroralline. The presence of an emission spectrumfrom the light of the night sky was noted asearly as 1895 by W. W. Campbell, who used anew visual spectrograph at Lick Observatorymany years before the spectrum was redis-covered photographically by V. M. Slipher atLowell Observatory. The light of the night skyas a detached phenomenon of physical interestwas recognized by the pioneering investigationsof Fabry in 1910, V. M. Slipher in 1919, LordKayleigh in 1920, and Dufay in 1923. For ahistorical account of early night sky and auroralresearch, the reader is referred to the articles byDufay (1928), Fabry, Dufay, and Cojan (1934),Dejardin (1936), Elvey (1942), Dufay (1938),and Swings (1949).

Many young scientists may wonder whyairglow and aurorae are included in a report onnew horizons in astronomy. Historically, bothphenomena are noteworthy because of theirastronomical consequences, although both arisesolely within the gaseous envelope of the earth.To most astronomers the light of the night skyis a nuisance to be tolerated by necessity, sincelittle can be done about it. However, in spiteof the unpleasant problems they make for theobserver, the light of the night sky and theaurorae are phenomena of wide astronomicaland astrophysical interest.

The term "light of the night sky" is used to

'National Astronomical Observatory, and Yerkes Observatory,Williams Bay, Wls.

describe a complex flux of radiation composedof both terrestrial and extraterrestrial sources.The extraterrestrial component arises from:(a) interplanetary scattering of sunlight (zodi-acal light), (b) interstellar scattering and emis-sion excited by stars, and (c) the integratedradiation of individual unresolved stars. Theatmospheric component of the light of the nightsky arises from: (a) scattering of starlight andextraterrestrial radiation in our atmosphere,(b) atomic emissions in the form of mono-chromatic emission lines, (c) molecular emis-sions in the form of emission bands and possiblycontinua, and (d) Cerenkov radiation andN2

+ emission from cosmic rays. Assigning toeach source a percentage of the total light of thenight sky can only be done with uncertainty sinceit is difficult to separate clearly some effects.In the visible region the ratio of extraterrestrialto terrestrial must be approximately unity.

Upon certain occasions aurorae are super-imposed on the night sky. The auroral radia-tions are caused by two basic mechanisms:(a) heavy-particle excited atomic and molecularemissions (in the homogeneous arc phase), and(b) electrical-discharge excited atomic andmolecular emissions (in the rayed-structurephase). However, most of the light may beproduced by secondary electrons in both cases.

To describe the appearance of the light ofthe night sky on the spectrum of a faint astro-nomical object, such as a distant nebula, it isbest to portray each spectral region separately.In the region sensitive to the 1030^-0 emulsion,the XX4800-3100 spectrum is characterized bymany bandlike features superimposed upon acontinuum which is strong in the blue andrapidly disappears shorter than X3900. As aconsequence of this spectrum it is difficult todetect weak astronomical features such as thebroad, but shallow, H and K lines in the spec-trum of faint galaxies.

05


The region <X3500 is composed of groupingsof bands of increasing intensity and contrastdown to the atmospheric cutoff. These bandsappear distinct and shaded toward the red, incontrast to the diffuse Xi and X2 bands in theX4500 region. No emission lines of atomicorigin occur in this region except during auroralactivity.

The visible region of the airglow spectrum,XX6500-4200, is dominated by four sharp emis-sion lines with relatively weak bands and acontinuum near the blue end. The lines, inorder of intensity, are the famous auroral lineat X5577, two red nebular lines at X6300-6364 (all three due to forbidden 01 emission),and the sodium doublet at X5893. The latteris generally so weak that it is only recorded onlow-dispersion spectrograms and as a singleline. The other features are weak OH bandsand an unidentified continuum that must beheavily contaminated by s tar l ight , e tc .Whether this "continuum" is real and ofatmospheric origin is still unknown. On spec-trograms sensitive to the deep red (i. e., 103a-Femulsions) the OH bands near X6560 and X6800become prominent features.

Spectrograms of astronomical sources in thevisible often show additional lines that couldpuzzle the new observer. These lines are dueto scattering by the atmosphere of neon andmercury radiations from adjacent cities. Therelative intensities generally appear anomalous,especially for neon in the red because of the"sunset" reddening over long air paths due torapid changes in the absorption coefficient.

In the infrared, the airglow becomes intenseas a result of a strong band system due to theOH radical. The most intense bands of thissystem are 4.5/*, but the bands at X10,000 andin the I -N emulsion photographic region(XX8900-6500) are strong enough to cause dif-ficulty in some astronomical observations.

The auroral spectrum shows the same atomicline features as the airglow, with the additionof a rich assortment of strong molecular bandsin the red and infrared and violet regions.The absolute intensity level can reach a factorof 10* times that of the airglow. Weak auroraeoften make themselves evident on astronomicalspectrograms only by enhancement of the X6300[01] lines and the appearance of the X3914 N2

+

band, which appears as a line owing to theoccurrence of the band head in the red branch.

Both the airglow and aurorae are phenomenathat are potentially important to the under-standing of the physical and chemical propertiesof the upper atmosphere. Our atmosphereprovides us with a nearby astrophysical labora-tory where we can observe a gaseous atmosphereover a wide pressure range under the influenceof electromagnetic and particle radiation froma star. Unfortunately, the interpretation ofthese phenomena has proved so complex thatwe are still in doubt with regard to many of thequestions that investigators have struggledwith. To begin with, both phenomena, air-glow and aurora, are difficult to study withresolution sufficient for detailed analysis. Theairglow is very weak, necessitating the use ofthe fastest possible spectrographs. Even so, asingle exposure sufficient to resolve molecularbands may require all the dark hours for severalmonths. Although the aurora is much brighter(up to 104), it is transient in both time andspace, sometimes lasting for the entire nightbut more frequently only a few minutes. Quiteoften the lifetimes of interesting features aremeasured in tenths of a second. Newly de-veloped photoelectric techniques may offer hopein attacking such problems.

For some years, the major airglow problemhas been the identification of the emissions.This problem is approaching what is thought tobe a solution, although X, and X2 bands ofLord Rayleigh in the blue are still completelyunidentified since no one has yet resolved theirrotation structure, if they possess any. Amongother interesting problems are the heights of theemissions of the night glow and the excitationmechanisms of both.

In the case of heights, very ingenious instru-ments have been devised to measure the patchi-ness and motions of the airglow patches. Esti-mates of the heights based upon the dependenceof brightness on zenith distance and made byaveraging many nights of observations withcareful accounting for absorption and scatteringhave been disappointingly at variance. Thereason may lie in the fundamental assumptionsof a statistically uniform, thin, emitting layer.

In the case of excitation mechanisms, someemissions have been thoroughly explained, while


others present unanswered questions. Theaurora is in particular interesting since the spec-trograph has demonstrated the influx of ener-getic protons, presumed to be of solar origin.The precise role of the protons in direct excita-tion of the atmospheric gases taxes the presentcapacity of quantum mechanics, since accuratewave-function descriptions of the molecular en-ergy states are still meager. Laboratory experi-ments stimulate the imagination, and whensuch experiments have been carefully devisedthe results have not been disappointing.

In the immediate future, the program for theInternational Geophysical Year will attempt togather data on a world-wide scale with the bestmeans uncovered in individual research studies.The reason is that one-point observations, evenin great detail, leave phases of the problem pre-sented by the aurora unanswerable. The syn-optic observations that will be obtained shouldopen a new door to the interpretation of solar-terrestrial relations; however, as in the past,the unlocking of these secrets will be a challeng-

ing but rewarding task for names not yet on thepages of the record.

General referencesCHAMBERLAIN, J. W., AND MBINEL, A. B.

1954. In Kuiper, ed., The earth as a planet. Uni-versity of Chicago Press.

DEJARDIN, G.1936. Rev. Mod. Phys., vol. 8, p. 1.

DUFAT, J.1928. Bull. Obs. Lyon, vol. 10, No. 9, p. 81.1938. Bull. Obs. Lyon, vol. 3, No. 16.

ELVEY, C. T.1942. Rev. Mod. Phys., vol. 14, p. 140.

FABBT, C , DUFAT, J., AND COJAN, J.1934. Etude de la lumidre du ciel nocturne.

Revue d'Optique, Paris.MBINEL, A. B.

1951. Phys. Soc. London Rep. Prog. Phys., vol.14, p. 121.

SWINGS, P.1949. In Kuiper, ed., The atmospheres of the

earth and planets, ed. 1. University ofChicago Press.

SWINGS, P., AND MEINEL, A. B.1951. In Kuiper, ed., The atmospheres of the

earth and planets, ed. 2. University ofChicago Press.

Solar-Terrestrial Relationships:Weather and Communications

By Walter Orr Roberts*

Relatively few of the many branches ofastronomy can be said to have applications ofdirect practical importance. The field of solar-terrestrial relationships is an exception. Twodecades of practice have shown the radiooperator that solar variations can make thedifference between success and failure in gettinga message through. Today, we may look forimprovements in practical weather forecastingbecause of new discoveries made on the commonborder between meteorology and astronomy.Just how far away these advances are willdepend, at least in part, on the magnitude andthe quality of the effort made in this field.

The effects of the sun on radio communica-tions are well known, even if poorly understood.The influence of solar activity on weather isless apparent, and the reality and nature of theeffects are still somewhat controversial. Enoughclues exist, however, to convince a substantialgroup of research workers that changes in thesun's shortwave emanations and in its corpus-cular radiation are reflected in worldwideweather patterns, and that practical gains inboth long- and short-term rainfall forecastingand temperature forecasting for large areas ofthe earth can be expected when these connec-tions are better understood. The fact that thesun exhibits rather stable long-term trendsimproves the prospects for season-in-advanceweather forecasts based on variations in solaractivity.

The sun's effects on weather obviously mustbe on a large scale. Several factors in the so-called "general circulation" of the earth'satmosphere give evidence of such solar in-fluences. The results of these large-scalechanges, in terms of surface rainfall and

• High Altitude Observatory, Boulder, Cote.

temperature, depend very much on local ele-ments such as topography and latitude; aworldwide circulation change that producesrain at one spot may produce drought inanother. The ties of sun to local weather aretherefore most complex.

The mechanism responsible for responses ofweather to solar activity clearly does not in-volve substantial changes of the sun's totalenergy; rather, it involves changes in selectedspectral ranges where only a relatively smallportion of the sun's total energy output isconcentrated. It is also clear that the mostpronounced weather responses are likely to befound at the upper levels of the earth's atmos-phere, where the sun's variable energy isprincipally absorbed. However, meteorolog-ical observations at such levels are still rathersporadic. Improvements in all types of weatherforecasting are likely to come from research onthe nature and causes of persistence and suddenchanges in the dynamics of the general circula-tion of the atmosphere. The motions of theupper levels and the possible nature of themechanisms that start disruptions in the pat-terns of momentum and mass transport in theearth's atmosphere require particular attention.Dynamic meteorology is still in its infancy, butthe solar astrophysicist has much to contribute.Ten years ago there seemed no real prospectthat major advances in long-term weatherforecasting would result from solar observa-tions; today they promise advances of in-calculably vast practical significance.

Without much exaggeration it can be statedthat modern solar-terrestrial research beganwhen R. C. Carrington of the Royal Observa-tory in England made his classic observation ofa great solar flare visible in integrated light at


11 a. m. on September 1, 1859. At the instantof the flare, a fluctuation occurred in thestrength of the earth's magnetic field. Onthe 2 days following, a violent magnetic stormwas recorded by magnetometers at the KcwObservatory. Balfour Stewart, the directorat Kew, concluded that the terrestrial magneticdisturbances were related to the flare; however,the idea was so unbelievable that most researchworkers dismissed the sequence of events as acoincidence.

The connection between flares and magneticstorms is now amply demonstrated, althoughthe problem remains of explaining just how asolar flare produces these effects and the evenmore spectacular phenomena in the auroraand ionosphere. In the years since Carring-ton's great flare, a whole new science hasdeveloped to solve the problems common tosolar physics and to terrestrial atmosphericphysics.

Both sides of this joint attack have greatlyenriched our observational knowledge. Thespectroheliograph, the coronagraph, and thesolar radio telescope have been the astronomicalmilestones. Geophysically, the night sky spec-trophotometers and ionospheric sounding de-vices have competed in importance with earth-based, airborne, and rocket-borne cosmic-raycounters, with photometers, and with magne-tometers operated with ever-increasing regular-ity from an ever-increasing diversity of stations.Great advances, particularly in meteorology,have come from the simple fact that we nowcan prepare daily or twice daily hemisphere-wide charts of airflow and temperature dis-tribution at a wide range of altitudes fromwhich large-scale dynamical studies can bemade.

In every facet of terrestrial atmosphericphysics, tantalizing connections between sunand earth have been found. But many piecesof the jigsaw puzzle remain to be put in place.Advances of great importance in ionosphericradio communications and in meteorology willundoubtedly appear when we have a satisfac-tory theoretical explanation that fits togetherthe many known solar-terrestrial relationships.

A specific example may help to illustratethe expected course of progress. We now knowthat world weather patterns sometimes respond

in a rather clear-cut way to changes in thecharacter and level of solar activity as measured,for instance, by the average intensity of theemission of the sun's green coronal line. Al-though our understanding of the solar physicsof the situation is very imperfect, we infer thatthe observed large changes in the coronalemission probably signify large changes in thetotal ultraviolet or X-ray emission of the sun.Nonetheless, the total changes of energy ofthe sun's emission must be very small, probablymeasured in thousandths or millionths of thesolar constant. Not only do the world weatherchanges require vastly greater energy than thesun's variations can supply but terrestrialphysics suggests that the level at which theX-ray or ultraviolet energy is absorbed mustbe so high that it is difficult to imagine that itcould exert a significant influence all the waydown into the weather sphere, far below. Yetthe problem stands, a challenge and a mystery.No one can doubt that real progress in fore-casting will come when we have a satisfactoryphysical explanation of the link, whether itbe a trigger mechanism for amplifying infinites-imal solar changes or some yet unknown cause.

The multitude of phenomena of the iono-sphere and the many variations of the effectswith latitude and longitude are equally enig-matic problems. Few astronomers or geophysi-cists today doubt that the earth is subject tohighly irregular and unpredictable "showers" ofsolar corpuscles, or that these corpuscles controlimportant effects in aurorae, the ionosphere, andearth magnetism. Yet, in spite of the brilliantvictories of the Chapman-Ferraro theory of sucheffects, we know that the theory is only a crudebeginning, and that it is grossly inadequate toaccount for major phenomena that are well ob-served. We particularly need new research toimprove our predictions for communications inpolar latitudes.

Many simple, but major, questions of solar-terrestrial relationships are yet unanswered.For example, do cosmic rays come from the sun ?A few short years ago it was hard to find a work-ing cosmic-ray physicist who would admit thatthe sun as much as gently modulated the cosmicrays that were believed to come in all directionsfrom outer spaqe. Astronomers generally as-sumed, without any serious doubts, that per-


haps well over half of the energy of the universewas bound up in these mysterious high-speed par-ticle emissions. Now the view prevails that amajor part of the low-energy cosmic rays comefrom, or are at least strongly controlled by, thesun; and how many of the more energetic cosmicrays also come from nearby, rather than fromthe farthest reaches of space, is an open question.

The observational-theoretical picture of solaractivity needs integration. The vast complexof sunspots, solar magnetic fields, flares, plages,prominences, and the corona needs to be weldedinto a consistent and understandable unit.From such a unified picture of solar activity,and of the quiet sun as well, will come better

understanding of the sun's radiational and cor-puscular emissions, particularly those emissionsstopped high in the earth's atmosphere. Atthe moment, it is hard to say whether the solarphysicist or the rocket and satellite launcherswill be first to specify the nature of the sun'svariable emissions as seen from an altitude, letus say, of 250 kms. In any event, the ad-vances in the decades immediately ahead willbe spectacular, and the practical stakes in long-and short-range weather forecasting may wellsurpass the very important gains that canreliably be expected in the field of communica-tions forecasting.

Solar PhysicsBy Leo Goldberg1 and Donald H. Menzel3

At the present time, astronomy is in a re-markable period of rapid growth wherein therate of accumulation of knowledge is higherthan at any time in the past. In such adynamic situation, existing facilities rapidlybecome obsolete and the new instrumentsrequired for continued development of thescience are generally larger and more expensivethan their predecessors. Unfortunately, thetraditional sources of financial support of largeastronomical facilities have been seriouslydiminished. Few private donors are in aposition to contribute gifts of the magnitudenecessary for modern research installations,and the larger foundations appear to be con-centrating their primary efforts in other areas.Thus, apart from occasional gifts or contractsfrom industrial concerns, the burden of supportmust fall upon Government agencies, whichcan function wisely only if they are guided bythe collective, considered judgment of re-sponsible scientists. This means that as-tronomers as a group must assume the re-sponsibility for deciding which large-scalefacilities are urgently needed for the continueddevelopment of their science, and they mustalso solve the organizational problems con-cerned with the cooperative use of thesefacilities.

We do not imply that all future astronomicalresearch should be directed by planning com-mittees. On the contrary, we strongly believethat the best new ideas and the initiation ofnew fields of investigation will come fromindividuals or small groups of investigatorswho must continue to receive the strongestpossible financial support from the NationalScience Foundation, the Office of Naval Re-search, and other Government agencies. Theregular program of grants and contracts by

' The Observatory, University of Michigan, Ann Arbor, Mich.• Harvard College Observatory, Cambridge. Mali.

such agencies supports a wide variety of studiesand is frequently responsible for so-called"seed" research, in which the expenditureof relatively few thousands of dollars may leadto the germination of a whole new field ofstudy. The next step in the nurturing of theseedling may well require the construction ofequipment costing hundreds of thousands oreven millions of dollars. It is at this pointthat planning on the national level must enterto evaluate the potentiality of the study withreference to astronomy as a whole. It wouldbe unfortunate indeed if the question of costshould be a major factor in any decision againstembarking in some new, exciting, and otherwiseworthwhile project. This country is capableof extending far greater financial support thanit has in the past to provide an enormouslyexpanded effort in basic research. The reallyserious bottleneck in astronomy is the currentshortage of topflight astronomers. We mustthere fore guard against wasting preciousscientific manpower in large but relativelyunproductive projects, especially those fi-nanced by public funds. The shortage ofastronomers is a national problem which shouldbe a major concern of the National ScienceFoundation.

No branch of astronomy has progressed morerapidly since 1940 than has solar physics.We do not except radio astronomy, which, asa special technique of tremendous powerrather than a separate branch of astronomy,has supplemented optical methods in contribut-ing importantly to virtually every field ofsolar research.

Following a period of phenomenal growth atthe beginning of the century, sparked by theinvention of the spectroheliograph and the solartower and by improvements in spectroscopicgratings, solar research progressed rather slowlythrough the 1920's and early 1930's. Largely

103


through the efforts of George Ellery Hale,observational solar physics had advanced muchmore rapidly than the theoretical interpretationof the data. This gap, however, began to closein the late twenties, with the development oftheories of ionization equilibrium and of atomicstructure.

Although by 1930 a certain number of suc-cesses had been achieved from the applicationof the new physical theories to the observa-tions, they were far outnumbered by manyconspicuous failures. In 1930, numerous road-blocks filled almost every avenue of solarphysics. It is useful to survey these barriersfrom the vantage point of 1956, both to notethose that have already been removed, or are inthe process of being eliminated, and to assesshow far we may now expect to move down theroad before new obstacles appear.

The photosphere

One of the most fundamental problems insolar physics concerns the structure andcirculation of the gaseous layers comprisingthe photosphere, the region in which thecontinuous spectrum is formed and in whichmost of the Fraunhofer absorption takes place.The solution of this problem calls for the con-struction of a model photosphere in whichthe chemical composition and the distributionof temperatures, pressures, gas velocities, andmagnetic fields are completely specified every-where in the photosphere. In theory, we canderive all of this information from observationsof the continuous spectrum and the Fraunhoferlines. Practically, however, the problem isfrightfully complicated. For example, mea-surements of solar limb darkening and spectralenergy distribution will yield the distributionof temperature with depth if we know whatprocesses produce the continuous spectrum.We can then calculate the pressure distributionif the atmosphere is in hydrostatic equilibrium.This assumption is certainly wrong if the at-mospheric gases are in violent, turbulent motioncaused by convective instability or electro-magnetic forces. To construct an accuratemodel we must evaluate the relative contribu-tions of radiation and convection in thetransport of energy through the atmosphere.

Stability depends on the temperature gradient

and the chemical nature of the atmosphere.The temperature gradient, in turn, must besufficient to drive the solar radiation throughthe semiopaque gas. In an idealized atmos-phere, where the atmospheric gases are arti-ficially stirred and where no heat is transferredfrom or into the moving volume, the tempera-ture gradient must follow the so-called adiabaticlaw. Such an atmosphere is in neutral equi-librium. Exchanging equal masses of gas atdifferent atmospheric levels produces no effecton the temperature distribution. As the massfrom the lower level rises and expands, itassumes the same volume, pressure, andtemperature as that of the mass it replaces.

As long as the temperature decreases upwardat a rate slower than that specified by theadiabatic condition, the atmosphere will bestable. Radiative transfer will predominateand the flux of energy will be essentially con-stant over the entire solar disk. However, ifthe natural gradient is steeper than theadiabatic, a mass of gas displaced upwardbecomes warmer than its surroundings. It willcontinue to rise, like a hot-air balloon. Analo-gously, a mass displaced downward becomescooler than its surroundings and thus descends.An atmosphere, so constituted, becomes un-stable and convection sets in.

When convection is mild, the energy con-veyed upwards will essentially equal that con-veyed downwards. Transfer by radiation maystill predominate. However, as convectionincreases, nonlinear effects enter so that we nolonger have a precise balance. We shall thenhave a net outward transport of energy bymoving airmasses. The convective and radia-tion fields may be said to "couple" together.The radiative flux at the outer boundary willvary from point to point over the solar disk insuch a way as to maintain a constant average.The convective field will carry momentum aswell as energy and thus will produce either dimi-nution or increase of the effective gravity, withconsequent distention or compression of theatmospheric gases.

Unsold has shown that convective instabilitytends to exist in two separate levels of the solaratmosphere, one below the photosphere and theother near the top of the reversing layer. Thelower one arises from ionization effects of hvdro-


gen, the dominant chemical constituent of thesolar atmosphere. If a volume consistinglargely of ionized gas is displaced upward, itexpands and cools. The cooling effects lead toelectron capture. The released energy raisesthe temperature above the value prescribed bythe adiabatic relation, and thus we find a hydro-gen convective zone. The second zone isassociated with the region where the atmosphereis transparent except in the absorption lines.The "blanketing effect" of the absorbing atomsagain changes the temperature gradient insuch a way as to induce convective instability.

When we turn to the Fraunhofer spectrum forinformation on chemical composition and gasmotions, we encounter even more difficultproblems. The shape and intensity of a solarabsorption line depend in a very complicatedmanner on a variety of factors, of which theabundance of the responsible element is onlyone. The "f-value," "line strength," or ab-sorptive power of an atom for a given radiationis also significant. The population of atoms inthe two levels responsible for the line formationis also extremely important. The Boltzmannequation fixes these populations if the mediumis in thermodynamic equilibrium. The temper-ature is the determining parameter. Likewise,the Saha equation, under similar conditions,fixes the amount of dissociation. However, thevery fact that energy is flowing through themedium requires departures from this equilib-rium condition. An exact determination of thesignificant populations theoretically requiresadvance knowledge of all the atomic constants,such as areas for absorption of radiation orcollisional excitation, and the appropriatetransition probabilities. The equations of sta-tistical equilibrium are most cumbersome anddifficult to solve. Further progress wouldseem to depend on the development of newtechniques for the obtaining of approximatesolutions near thermodynamic equilibrium.

We need to know the mechanism that theatoms employ for removing energy from the line.The process may involve pure absorption,coherent scattering, or noncoherent scattering,and it is usually not obvious which mechanismor combination of mechanisms operates for agiven line at a given height in the atmosphere.The character of the line depends critically on

the assumed atmospheric model. Third, wemust know or determine what processes causethe broadening of the spectral lines. The mostimportant broadening mechanisms include ran-dom Doppler effects caused by thermal agita-tion or by macro- or micro-turbulence, colli-sions, Stark effect, and radiation damping.Widening from hyperfine structure and Zeemaneffect may not always be negligible. Theparameters necessary for calculating the Starkand collisional broadening are very difficult tocome by, theoretically or experimentally.

Twenty-five years ago only a rough beginninghad been made towards the quantitative inter-pretation of the solar spectrum. Virtually allanalyses of the Fraunhofer spectrum werebased on the idealized assumption that thecontinuous spectrum originated in a sharplybounded radiating surface, the photosphere, andthat the lines were produced in a cooler super-imposed layer, the reversing layer, which wassupposed to be at constant temperature andpressure. Only very crude guesses were avail-able for most f-values and line-broadeningparameters, and accurate intensity measure-ments had been made for but a mere handful ofsolar lines. Nevertheless, H. N. Russell (1929)accomplished a remarkable feat in achievingwhat he has called the "zero-th approximation"to the chemical composition of the sun, espe-cially in view of the fact that only recently haveastrophysicists displayed any general agreementthat the "first approximation" has finally beenreached.

No one doubted the existence of large tem-perature and pressure gradients in the solarphotosphere, but attempts to construct a modelsolar atmosphere were frustrated by a total lackof knowledge concerning the origin of the con-tinuous spectrum. Calculations by Menzel andPekeris (1935) gave too low a figure for theabsorption by transitions involving the negativehydrogen ion. In 1938 Wildt (1939) noted theexistence of a bound state of H —, and suggestedthat photoelectric ionizations and free-freetransitions of this negative ion were the majorcauses of the sun's continuous opacity. Wildt'ssuggestion was confirmed by the work ofChandrasekhar and associates (Chandrasekharand Breen, 1946) and Chalonge and Kourganoff(1946).


Numerous calculations of model solar photo-spheres followed. Even now, we must admitthat the photospheric model is known only inthe first approximation. Although mostworkers agree on the temperature distributionin the intermediate layers, the structure of thedeepest and uppermost layers remains uncer-tain. Since most of the radiation from theseregions is absorbed by the overlying gases, wemust use theory rather than direct observationfor the analysis. In consequence, the locationand extent of the convective zone are notprecisely known. Indeed, the magnitude ofwhat appear to be convective processes in theouter atmosphere suggests that even the inter-mediate layer may not be free from effects ofthe hydrogen zone. Radiation from the upperlayers can best be observed at the extreme limbof the sun, where terrestrial atmospheric seeinghampers observation. The blanketing effect ofthe absorption lines and possible departuresfrom thermodynamic equilibrium further com-plicate the problem.

Additional difficulties of even more basiccharacter stand in the way of a definitive modelof the photosphere. Just a few years ago itseemed possible that a straightforward analysisof measurements of solar limb darkening andabsolute spectral energy distribution would de-termine an accurate solar model. I t was hopedthat the model could be tested and further re-fined by calculation of absorption-line profilesin terms of a specific chemical composition ofthe atmosphere. For several reasons, this hopenow seems to have been optimistic. First, noexisting model has succeeded in reproducing indetail the profiles of strong absorption lines. Inparticular, the observed central intensity hasalways been greater than the calculated, but thediscrepancy has usually been attributed to in-perfections in the theory of line formation.Various devices such as fluorescence effects andnoncoherent scattering have been employed tofill in the line center, but none has proved quitesatisfactory. Second, the empirical dependenceof the continuous opacity upon wavelength, de-rived from Pierce's (in press) latest measures oflimb darkening, differs significantly from thevalues calculated in terms of currently acceptedelectron pressure distributions and the theory ofcontinuous absorption by neutral hydrogen at-

oms and negative hydrogen ions. Errors evi-dently exist either in the model or in the theoryof the continuous absorption, or perhaps in both.Third, no spherically symmetrical model of thephotosphere appears capable of representing thecenter-to-limb variation in the intensities ofboth low and high excitation lines of Fe I(Bohm, 1954). Nor can the center-limb vari-ation in the CO line intensities be accounted foron such a model (Newkirk, 1953). Similar re-sults have been obtained by Elste (1955) atGottingen and by others.

The evidence is accumulating that the line-intensity observations can be reconciled withtheory only when account is taken of temper-ature and velocity fluctuations in the photo-sphere and chromosphere. The recognition oftemperature fluctuations is not new, since theyare reflected in the brightness fluctuations ofthe granules, but their influence on the Fraun-hofer spectrum represents a second-order effectwhich could not be detected before the adventof modern techniques of observation and analy-sis. I t is now generally agreed that the phe-nomenon of the granules is a consequence ofturbulent convection in the photosphere andthat, on the average, the bright and presumablyhot granules are ascending and the cooler, inter-granular regions are descending in the photo-sphere. If the convection is sufficiently violent,however, coupling of the mechanical and radi-ative fields may occur, with nonlinear effects.The optical thickness of the atmosphere mightvary significantly between the bright and darkareas, with an appreciable change of line in-tensity.

Another consequence of the turbulence is theexcessive broadening of the centers of theFraunhofer lines as compared with that to beexpected from thermal Doppler effect at atemperature of 6,000° K. The root-mean-square turbulent velocity derived from indi-vidual line profiles and from the curve of growthis about 2 km/sec. Richardson and Schwarzs-child (1950) supposed that the radial velocitiesof individual granules could be obtained fromthe measurement of minute Doppler shifts frompoint to point along the length of Fraunhoferlines, under conditions of high spectroscopicresolution and exceptional seeing. Carefulmeasurement of selected Mount Wilson plates

NEW HORIZONS IN A8TRONOMY 107

did indeed reveal irregularities in the lines,which, upon measurement, yielded a root-mean-square velocity of about 0.4 km/sec. This effectwas interpreted as arising from the macroturbu-lence of relatively large elements, the highervelocity of 2 km/sec being attributed to themicroturbulence of elements smaller than thethickness of the "reversing layer."

It had been supposed that the observationof the random Doppler shifts was limited bythe resolution of the solar image; i. e., by theexternal seeing. Recent work by McMath(1956) and McMath, Mohler, and Pierce(1955) at the McMath-Hulbert Observatoryhas revealed, however, that limitations inspectroscopic resolution are at least of equalimportance and that air turbulence within thespectrograph may "wash out" the smallDoppler effects. On McMath-Hulbert photo-graphs made with the vacuum spectrograph,all photospheric lines show a characteristic"zigzag" appearance even when the seeingquality is less than average. Further, whenthe seeing is good, the lines fluctuate in in-tensity from point to point along the solardisk. Thus the temperature and velocityfluctuations in the solar atmosphere may nowbe directly observed in the Fraunhofer spectrum,at least for elements about 2 seconds of arc indiameter or larger.

Several inferences may be drawn from theseresults, the most general of which is that theinterpretation of photospheric and chromo-spheric observations on the basis of sphericalsymmetry should now be abandoned, and theobservational emphasis directed towards de-termining the nature of the velocity andtemperature fluctuations in the solar at-mosphere. The impact of this new develop-ment upon the National Science Foundation islikely to be very great. First, important fundswill be required both to modernize existingsolar spectrographs and to construct new ones.Second, an enormous quantity of new andvaluable data will become available for analysis,which will necessitate funds for assistance inreduction and analysis. Third, new impetuswill be provided for the development andrefinement of image-tube techniques. At pres-ent photographic exposure times are not lessthan 10-20 seconds, whereas the image tube

may be expected to reduce exposures to a smallfraction of a second. Such a reduction wouldgreatly reduce the effects of atmosphericscintillation and improve our determinationsof the lifetimes of small-scale solar phenomena.Finally, it is already apparent that spectroscopicresolution has far exceeded the image resolutionand that limitations in external seeing qualitywill seriously hinder the discovery of the detailsof solar atmospheric structure.

Poor external seeing is caused both by dis-turbances inherent in the atmosphere and bylocal disturbances of the air produced by solarheating of the telescope and mirrors. The Nation-al Science Foundation panel for a national as-tronomical observatory, under the chairmanshipof R. R. McMath, has recently considered thisproblem and has concluded that two coursesof action should be taken. First, the presentsite survey for the national observatory shouldbe extended to include observations of daytimeas well as nighttime seeing. Second, an in-vestigation of instrumental seeing should beundertaken at the earliest possible momentwith the goal of designing a solar telescope inwhich heating effects would be minimized.It is probable that the solution of the problemof telescopic seeing is now the most importantinstrumental challenge in the field of solarphysics.

The vacuum spectrograph or other spectro-graphs that may be designed to eliminateinstrumental turbulence will initiate a newphase of research in solar spectroscopy that willtake many years to run its course. The endresult should be a much finer elucidation ofthe structure of the solar atmosphere, includingits composition and circulation, than is possibleat present. As a valuable byproduct we mayalso expect a better understanding of themechanisms of absorption-line formation. Aspreviously mentioned, the fact that the observedcentral intensity of a very strong line is alwaysgreater than that calculated from theory hasbeen a major difficulty. However, it is alreadyobvious from the vacuum spectrograms thatsome if not all of the discrepancy betweentheory and observation will vanish when weallow for fluctuations in observed centralintensity, which have generally been blurredby poor seeing inside the spectrograph. In


addition, theorists will have to include effectsof mechanical as well as radiations! transferof energy.

Finally, it may not be out of place to pointout that if temperature fluctuations are presentin the solar photosphere they undoubtedly occuralso in stellar atmospheres, especially in thehighly turbulent atmospheres of the supergiantstars.

Chromosphere and corona

By definition, the beginning of the chromo-sphere is the upper boundary of the photosphere.It is also the transition region, perhaps 20,000km thick, in which the electron temperatureincreases from its photospheric boundary valueof 4,000°-4,500° K to 1,000,000° K or higher, inthe inner corona. I t is therefore of strategicimportance, since within its confines probablyoccur the processes that maintain the corona atits high temperature. More than 20 years agoMenzel (1931) called attention to the anomalyposed by the coexistence in the flash spectrumof emission lines of both neutral metals andionized helium. Menzel pointed out that theobserved intensity of He+X4686 could not beexplained by a temperature of 6,000° K even ifthe chromosphere were composed of pure he-lium, and suggested that the required ionizingradiation might come from a hypotheticalultraviolet excess in the solar spectrum.

The problem posed by Menzel is still para-mount for theories of chromospheric structure,even though the solution in terms of an ultra-violet excess has been ruled out. Since the1930's, at least three important developmentshave paved the way for a renewed attack on thechromospheric problem. First, the discoveryof the high-temperature corona inevitably re-quires the existence of large temperature gradi-ents somewhere in the chromosphere. Second,the observation from rockets of the far ultra-violet solar spectrum has shown that the hypo-thetical ultraviolet excess does not exist.Third, the observations clearly show that thechromosphere is not a static phenomenon andthat hydrodynamic or aerodynamic processesplay an important role in fixing its physicalstate.

Evidence bearing on the structure of thechromosphere comes from observation of the

optical spectrum during eclipses, from motionpictures in Ha obtained with large corona-graphs, from observation of the radio emissionat high frequencies, from observation of thecentral intensities of certain strong Fraunhoferlines and of the XI0830 absorption line of He I.from observation of emission lines in the rocketultraviolet, and from observations of the earth'sionosphere. Interpretation of the observationsfrom the conventional point of view of spheri-cally symmetrical chromospheric models leadsto serious discrepancies. In particular, one setof observations has seemed to require a tempera-ture for the low chromosphere in the neighbor-hood of 5,000°-6,000° K, whereas another setseems to demand much higher temperatures ofabout 20,000° K or greater.

A possible path out of the dilemma waspointed out in 1949 by Giovanelli (1949), whosuggested that the chromosphere is nonuniformand that it consists of alternate "hot" and"cold" columns. Hagen's (1954) measurementsof the center-limb variation of the radio emis-sion also point to a model of this general type.Strong support for the nonuniform model hasrecently been provided by the detailed analysisof flash spectra by the Harvard-High AltitudeObservatory group (Athay, Evans, and Roberts,1953; Athay, Billings, Evans, and Roberts,1954; Athay and Roberts, 1955; Pecker andAthay, 1955; Matsushima, 1955; Athay, Men-zel, Pecker, and Thomas, 1955; Athay andThomas, 1955, 195Ga, 1956b; Athay and Men-zel, 1956; Billings, Cooper, Evans, and Lee,1954) and by studies of the structures ofFraunhofer lines at the McMath-Hulbcrt Ob-servatory (McMath, Mohler, Pierce, and Gold-berg, 1956; Mohler and Goldberg, 1956).Although the exact geometry and physicalproperties of the hot and cold regions havenot yet been clearly established, interestingworking hypotheses have been suggested whicliclearly indicate the lines along which futureresearches should be planned. Observations ofthe flash spectrum at future eclipses will con-tinue to play a leading role in the investigationsof the chromospheric structure, but they maynow be supplemented by additional new tech-niques which have recently been developed.Of great importance are observations of micro-wave radio emission, which should be carried


out with large antennae of the highest possibleresolving power. The accurate measurementof emission profiles of solar lines in the rocketultraviolet will also be crucial for the testing ofchromospheric models. Finally, the Fraun-hofer spectrum itself constitutes a rich butlargely neglected source of chromosphericinformation. The interpretation of the limbspectrum is complicated by the two-dimensionalintegration, both radial and tangential, of thechromospheric emission. On the disk, the lineof sight integration is only one dimensional.Indeed, the complex structures of the X10830helium line and of the cores of the hydrogen andionized calcium lines, as revealed by the vacuumspectrograph, are undoubtedly a consequenceof the temperature fluctuations in the chromo-sphere and should be analyzed from this pointof view.

The dynamic nature of the chromosphere isgraphically portrayed by the Ha motion pic-tures recently obtained by R. B. Dunn ofHarvard and the Upper Air Research Observa-tory with the 15-inch Sacramento Peak corona-graph. These remarkable photographs recordthe rapid changes that occur in the irregularstructure of the outer chromosphere. Sincethe chromosphere is basically a dynamic phe-nomenon, with kinetic transport of bothenergy and momentum, a true model of thisregion of the solar atmosphere should depictand account for the internal motions. Onemay question whether any static solution ofthe fundamental equations exists. Small per-turbations about such a hypothetical staticsolution, for example in terms of a linearcombination of normal modes, seems out ofkeeping with the actual observations. Instead,we suspect that the mean energy transport,determined by the solar energy generation,can never be fulfilled in a purely radiativeatmosphere.

The solution will depend on the magneto-hjdrodynamic equations of motion in theirnonlinearized form. The Bhatnagar-Gross-Krook (1954) equation appears to be a usefulsubstitute for the much more complex Boltz-mann transport equation. The simplificationrepresents a new approach to problems of gasdynamics in which the distribution of velocities

no longer follows Maxwell's law. Instead,the equation represents the velocities as lyingbetween the two extremes of isotropic scatteringwithout change of velocity, or of reemissionunder thermodynamic equilibrium at the tem-perature described by the root-mean-squarevelocity. The equation bears a close resem-blance to the Milne-Eddington equation ofradiative transfer, with solution between theextremes of monochromatic radiative equil-ibrium and local thermodynamic equilibrium.

The solution of the transfer problem, espec-ially when electric currents and magneticfields are present in the medium, can rarelybe expressed in terms of known functions.Instead we shall have to determine families ofsolutions, with ranges of boundary values andinitial conditions. In general, these calcula-tions will require the use of high-speed comput-ing machines. The National Science Founda-tion may well be called on to furnish funds forthe development of models of the dynamicchromosphere.

Closely associated with structural problemsof the solar atmosphere is the question of originof the solar radio emission. It is well knownthat an ionized gas, with specified electrondensity, possesses the so-called "plasma" oscilla-tion frequency equal to the "critical frequency"at which the phase-velocity becomes equal tozero. Many persons have supposed that nat-ural oscillations of such a medium will lead tothe emission of radio waves on the plasmafrequency. A detailed analysis indicates, how-ever, that the plasma oscillations, which re-semble sound waves in their longitudinalcharacter, cannot be converted into transversewaves by means of boundary conditions alone.In short, plasma oscillations cannot producetransverse radio waves. When the mediumcontains a magnetic field, however, the rigidityof the medium associated with magnetismresolves any longitudinal wave into vibrationshaving one transverse component. A dis-turbance propagated from regions of high intoregions of low density tends to have its vibra-tions amplified. In this way, as Krook andMenzel have shown, a magnetohydrodynamicwave at low density will eventually escape fromthe sun's outer boundary as a true electro-

110 SMITHSONIAN CONTRIBUTIONS TO ASTROPHYSICS V«L.

magnetic wave, with polarization determinedby the orientation of the internal magneticfield.

Low-frequency radiation tends to be trappedor attenuated in an atmosphere containingions and electrons. Thus, only higher fre-quencies are able to penetrate from the chromo-sphere through the overlying coronal regions.Similarly, a given disturbance rising through thesolar corona permits lower and lower radiofrequencies to escape. Study of the changingspectrum of solar radio noise, especially incombination with optical data, should givevaluable information about the solar atmos-phere. J. P. Wild's (1953) dynamic spectrumanalyzer presents the observational facts ingraphic form. The Cornell measures of solarradio noise, as analyzed by H. W. Dodson(1953), have already yielded significant cor-relations with flare and other solar activity.

The corona itself poses many difficult prob-lems. There is little doubt that the streamers,domes, and arches owe their basic structure togeneral and local magnetic fields. Magneticfocusing must be important in certain of thesephenomena. The forces associated with electriccurrents undoubtedly play a significant role indetermining the motions of prominences andthe general condensations that occur from thesolar corona. The detailed theory of fluid gasdynamics necessary to account for these phe-nomena will undoubtedly be very complicated.

Solar activity

The term "solar activity" includes all types ofphenomena of a transient nature except thesmall-scale fluctuations in velocity and temper-ature associated with granulation. Typical ofthese phenomena are sunspots, faculae, plageregions, flares, filaments, prominences, andcoronal disturbances. Solar activity occursthroughout the solar atmosphere; e. g., the sun-spots and faculae in the photosphere, the flaresin the chromosphere and corona, and the prom-inences largely in the corona. No attempt willbe made in this brief survey to summarize theproblems related to solar activity in any sys-tematic way. The entire subject is still in ahighly chaotic state, hardly less so at presentthan in the 1930's. One reason for this stateof affairs is the aforementioned inherent di-

versity and complexity of solar phenomena. Asecond reason is that only in rare cases have themost modern spectroscopic techniques been ap-plied to the study of solar details. The changeson the solar surface are so gross and easy toobserve that solar observers have been temptedto make large numbers of observations with in-struments of relatively low resolving power,either visually or on uncalibrated plates.

The rapid acceleration of observing programsduring the International Geophysical Year givesto those participating in the program the re-sponsibility of maintaining high standards fortheir records. It is to be hoped that the needfor large masses of data will not distract theattention of solar physicists from the really basicproblems of solar activity, which pertain to theorigin and physical nature of solar phenomena.In the end, a complete understanding of theeffects of solar radiation on the earth is morelikely to come from the solution of these basicproblems than from correlations developed be-tween one or two observed solar quantities andvarious geophysical data.

Knowledge of the physical state of the variouskinds of solar phenomena must precede correcttheories of their origin. In the light of recenttechnical developments—which include the per-fection of gratings and the magnetograph bythe Babcocks, the construction of the vacuumspectograph by McMath, and the developmentof a large coronagraph by the Harvard-AirForce group on Sacramento Peak—we can lookforward to a fruitful period of quantitativeanalysis of the spectra of solar details during theforthcoming sunspot maximum.

An understanding of the detailed structure ofsunspots, including the three-dimensional dis-tribution of temperatures, pressures, gas ve-locities, and magnetic fields, should be a pri-mary target during this period. There alsoseems to be no reason why high-dispersionstudies of flare spectra should not lead to know1-edge of the vertical temperature and pressuredistributions in the flare-emitting layers, whichcan provide insight into the mechanisms oforigin both of the flares themselves and of theultraviolet radiation responsible for ionosphericdisturbances. In this connection we considerthat the physical investigation of faculae andthe associated "plages faculaire" has been


grossly neglected. It is well established thatflares break out only in regions occupied by theplages. The latter, which are also regions ofabnormal magnetic activity, are generally lo-cated above the photospheric faculae. Theremust exist, then, certain physical propertiesassociated with faculae and plages that favorthe outbreak of flares; yet the properties ofthese regions have hardly been studied.

Our understanding of the physics of prom-inences is equally weak. The development ofthe motion-picture technique by McMath andby Lyot in the early 1930's has been followedby large numbers of reasonably accurate meas-urements of prominence motions, chiefly in Haradiation. Some of the observed motions havebeen explained plausibly, but nevertheless onlyqualitatively, on the basis of the combinedlaws of hydrodynamics and electrodynamics byAlfv6n and by Menzel, but a detailed quantita-tive description of prominence behavior is stillsomewhere in the future. Here again, thedevelopment of a successful theory is retardedby uncertain knowledge of the temperatures,pressures, turbulent motions, and magnetic fieldswithin the prominences. The first three ofthese quantities are readily within the reach ofmodern instruments and theory. But one mustavoid the temptation to tackle these importantproblems with equipment yielding marginaldata that can be interpreted only on the basisof doubtful assumptions or even wishfulthinking.

Finally, although we have devoted majorattention to observational trends in solarphysics, we have not wished to minimize theimportance of support by the National ScienceFoundation and other Government agencies fortheoretical research of varying shades of purity.The so-called theoretical projects that are mostlikely to receive support today are those thatinvolve detailed computations— usually by high-speed calculators—based on theories already inexistence. In short, practical results are guar-anteed, but such projects are no more theoreticalthan those that involve the measurement ofstellar magnitudes or of spectral line intensities.We do not question the value of accurate com-putations to astronomy and solar physics.However, the National Science Foundationshould also consider how its support might

stimulate the birth of new theories as well asfoster applications of existing ones.

For example, the solution of many problemsof solar structure or solar activity waits uponfurther developments in the field of generalizedgas dynamics concerning the motions of high-temperature gases in the presence of varioustypes of force fields. Indeed, the United Statesappears to be lagging in this important field,which has applications to physics, geophysics,and industry as well as to astrophysics.

Several symposia on this general subject havealready been held and have been useful both interms of summarizing what little knowledgeexists in this area and in stimulating newthinking. If a really top-notch group of in-vestigators could be assembled to work togetherfor a relatively long period—for months, or evenyears, instead of a few days—to study theproblem in its broadest aspects, we couldexpect significant new advances. For a prob-lem of this magnitude, support would benecessary for a number of years, during whichheadquarters of the project would be main-tained at a particular university with somerotation of personnel. A center of this sortdevoted to theoretical problems of fluid gasdynamics could be as important to astronomyand to other branches of science as a cooperativeobserving facility.

ReferencesATHAY, R. G., BILLING*, D. E., EVANS, J. W., AND

ROBERTS, W. O.

1954. Astrophy.s. Journ., vol. 120, p. 94.ATHAT, R. G., EVANS, J. W., AND ROBERTS, W. O.

1953. Observatory, vol. 73, p. 244.ATHAT, R. G., AND MENZEL, D. H.

1956. Astrophys. Journ., vol. 123, p. 2S5.ATHAT, R. G., MENZEL, D. H., PECKER. J.-C, AND

THOMAS, R. N.1955. Astrophys. Journ. Suppl., vol. 1, p. 505.

ATHAT, R. G., AND ROBERTS, W. O.1955. Astrophys. Journ., vol. 121, p. 231.

ATHAT, R. G., AND THOMAS, R. N.1955. Astrophys. Journ. Suppl., vol. 1, p. 491.1956a. Astrophys. Journ., vol. 123, p. 299.1956b. Astrophys. Journ., vol. 123, p. 309.

BHATNAQAB, P., GROSS, J., AND KROOK, M.1954. Phys. Rev., vol. 94, p. 511.

BILLINGS. D. E., COOPER, R. H., EVANS, J. W., ANDLEE, R. H.

1954. Electronics, vol. 27, p. 174.B6HM, K.-H.

1954. Zeitschr. Astrophys., vol. 35, p. 179.


CHALONGE, D., AND KOURGANOFF, V.1946. Ann. d'Astrophys., vol. 9, p. 69.

CHANDRASEKHAR, S., AND BREEN, F.1946. Astrophys. Journ., vol. 104, p. 430.

DODSON, H. W.1953. In Kuiper, ed., The sun, p. 692.

ELSTE, G.1955. Zeitschr. Astrophys., vol. 37, p. 184.

GIOVANELLI, R. G.1949. Monthly Notices Roy. Astron. Soc., vol.

109, p. 293.HAGEN, J. P.

1954. Jouru. Geophys. Res., vol. 59, p. 158.MATSU8HIMA, S.

1955. Astrophys. Journ. Suppl., vol. 1, p. 479.MCMATH, R. R.

1956. Astrophys. Journ., vol. 123, p. 1.MCMATH, R. R., MOHLEB, O., AND PIERCE, A. K.

1955. Astrophys. Journ., vol. 122, p. 565.MCMATH, R. R., MOHLER, O., PIERCE, A. K., AND

GOLDBERG, L.


MENZEL, D. H.1931. Publ. Lick Obs., vol. 17.

MENZEL, D. H., AND PEKERIS, C. L.1935. Monthly Notices Roy. Astron. Soc, vol.

96, p. 77.MOHLER, O., AND GOLDBERG, L.

1956. Astrophys. Journ., vol. 124, p. 13.NEWKIRK, G.

1953. Thesis, Univ. Michigan.PECKER, J.-C, AND ATHAT, R. G.

1955. Astrophys. Journ., vol. 121, p. 391.PIERCE, A. K.

. In press.RICHARDSON, R. S., AND SCHWARZSCHILD, M.

1950. Astrophys. Journ., vol. I l l , p. 351.RUSSELL, H. N.

1929. Astrophys. Journ., vol. 70, p. 11.WILD, J. P.

1953. In Kuiper, ed., The sun, p. 676.WILDT, R.


Stars and the Galaxy

Spectral ClassificationBy W. W. Morgan *

In a classical review titled "Some Problemsof Sidereal Astronomy" prepared by H. N.Russell in 1917, the main object of astronomyis stated as being. . . the development, on the basis of collected facts,of satisfactory theories regarding the nature, mutualrelations, and probable history and evolution of theobjects of study.

Russell suggests that before the stage offormation of a general theory is reached, twocourses may be followed:

(1) to collect masses of information, as accurateand extensive as possible, by well tested routinemethods, and leave it to the insight of some fortunateand future investigator to derive from the accumulatedfacts the information which they contain regardingthe general problems of the science; (2) to keep thesegreater problems continually in mind, and to plan theprogram of observations in such a way as to secure assoon as practicable data which bear directly upondefinite phases of these problems.

The present survey gives an approximatepicture of the research situation in spectralclassification in 1956, a comparison withRussell's survey, and some suggestions of theprobable directions for future work.

The Henry Draper Catalogue, which waspublished between 1918-1924, remains the mostimportant fundamental body of data in Rus-sell's first category. Russell's conclusions are,to a large extent, based on the Harvard work.

The interval 1917-1955

Two fundamental conclusions were drawn byRussell from the study of the lines of stellarspectra: (a) that almost all of the observedlines are identifiable with known elements;and (b) that the vast majority of stellar spectrafall into a single, continuous linear sequence.Later work has confirmed fully Russell's firstconclusion; the second requires some qualifica-tion.

1 University of Chicago, Chicago, III.

As early as 1897, Miss Antonia Maury hadreached the conclusion that—. . . a single series was inadequate to represent thepeculiarities which presented themselves in certaincases, and that it would be more satisfactory to assumethe existence of collateral series.

She devised a system of 22 "groups" whichformed a sequence; most of the "groups" werethen divided into "divisions," which, in turn,introduced a second dimension. I t was fromthis pioneering work of Miss Maury that thefield of spectroscopic parallaxes derives, throughthe later work of Hertzsprung, Russell, Adams,and Kohlschutter.

The situation in 1920 was, therefore, approx-imately as follows: spectral classification wasconsidered to be primarily a one-dimensionalproblem; in addition, differences in luminositywere observable by means of certain sensitiveline-ratios. The early work of Miss Mauryhad not been adopted as standard, nor was itcontinued on any appreciable scale.

If a single development could be consideredof primary importance in spectral classificationsince 1920, it would be the realization of thefundamental two-dimensional nature of theproblem. The critical work here was that ofBertil Lindblad, who developed criteria forthe separation of stars of differing luminosityby use of plates of low dispersion; this lastproperty permitted the extension of the spectro-scopic parallax method to great numbers ofstars too faint to be observed with higherdispersion.

Another fundamental development was theformalizing of the earlier work (in particular,that of Lindblad) in the two-dimensionalYerkes system of Morgan and Eeenan. As faras the actual criteria of classification are con-cerned, there was little new in the latters'system; the principal development lay in therecognition that spectral classification per se

115


is a two-dimensional process and that thedesignation of the "luminosity class" shouldbe by means of an empirical notation similarin nature to the ordinary "temperature type."

The process of determining spectroscopicabsolute magnitudes was thus divided into twostages: (1) the determination of a spectraltype and luminosity class; and (2) the calibra-tion of the empirical luminosity classes in termsof absolute magnitude. The first of theseoperations is purely a description of the spectra.This description is a fundamental property ofthe latter, and there are no compelling reasonswhy it would have to be revised with the passingof time. In this respect the second operationcan be improved with the passing of time, dueto more and more accurate calibration databecoming available.

Another development in the interveningyears has been the devising of methods forsegregating certain kinds of stars by means ofspectra of exceedingly low dispersion. Themost important applications of these methodshave been to the use of objective-prismspectrograms.

In 1950 attention was called to the possibilityof classifying members of certain "naturalgroups" on objective prism spectra. Forexample, a single criterion is sufficient for thesegregation of a well-defined group of early Amain-sequence stars; this is carried out bymeans of the Balmer series of hydrogen. If agroup of stars is selected having the character-istic of maximum intensity for the Balmerlines, it is found that the stars in this grouppossess the characteristics of closely similarabsolute magnitude and spectral type. Amongthe most interesting of the groups identifiableon objective-prism plates is that of the bluegiants. Nassau and collaborators at Clevelandhave prepared such a catalog of northern bluegiants, and this work has been extended tosouthern declinations and fainter northernstars by Haro and associates at the NationalAstrophysical Observatory in Mexico.

Of comparable importance has been thedevelopment of a method of classification oflate M giants in the infrared spectral region byNassau and collaborators. A catalog of thebrightest of these objects has been publishedrecently, but a beginning has scarcely been

made on this most promising field for futuregalactic studies.

A most important development has been inthe field of classification of the faint inhabitantsof dust clouds—the T Tauri stars—by Joy(Mount Wilson), Herbig (Lick), and Haro(Tonantzintla). The work on these strangesystems of stars is also in its infancy.

Present and future problemsThis topic will be considered under threegeneral heads: (1) astrophysical interpretation;(2) spectral classification as applied to problemsof galactic structure and stellar evolution; and(3) instrumental needs.

(1) Astrophysical interpretation.—Very prom-ising beginnings have been made in the past fewyears in devising semiquantitative methods ofclassification for stars in certain restricted areasof the HR diagram. Some of these have madeuse of measures of line depths and equivalentwidths to furnish more objective data for classi-fication. Of especial importance is the workof Greaves, Baker, and Wilson at Edinburgh, ofChalonge and associates in Paris, and of Hos-sack at Toronto. Accidental accuracy of a highorder has been achieved. The principal difficul-ty here is probably in devising a frame of refer-ence sufficiently precise to make full use of theextremely accurate individual results obtainedby these investigators.

It is from investigations such as these that wewill be able to decide how uniquely the stars canbe classified in a two-dimensional array, andwhat the relationship is between the "normal"and the "peculiar" stars. The practical limitattainable in the accuracy of spectroscopic par-allaxes will also be determined from high-dis-persion investigations.

(2) Spectral classification as applied to prob-lems of galactic structure and stellar evolution.—The application of the spectroscopic parallaxmethod to large numbers of fainter stars is nowquite feasible. The onty thing needed for agreat increase in our knowledge of the arrange-ment of the stars in space in the neighborhoodof the sun is more investigators; the method andequipment are now quite adequate for the work.New programs will probably be of two kinds:(1) work on the O-associations, galactic clusters,and other objects associated with spiral struc-

NEW HORIZONS ENT ASTRONOMY 117

ture; and (2) investigations of the amorphousdisk and halo population.

An interesting group for further investigationin the second category is that of the late Mgiants, which are most important for the de-termination of the kinematical properties of thedisk population. However, for their observa-tion at great distances it will be necessary todevise infrared techniques for measurement ofradial velocity and absolute magnitude.

Some of the most dramatic developments ofthe 30-odd years since Russell's survey havebeen in the field of stellar evolution. Thetheoretical work in this field, up to 1951, hasbeen well summarized by Stromgren (1952).This work, together with later developments,now makes possible the application of Russell'ssecond criterion quoted at the beginning of thissummary. It is now possible to formulateprecise observational experiments concernedwith various aspects of the great generalproblem of stellar evolution.

One of the most inviting practical problemsof the future is the invention of a system ofspectral classification according to stellar age,or relative evolutionary state.

(3) Instrumental needs.—The principal in-strumental need for present and projectedobservational problems in this field is probablynot for new or larger telescopes but for moreefficient spectrographic equipment for use withexisting instruments.

The requirements for such "maximum effi-c iency" spectrographs have been outlinedclearly by Bo wen (1947, 1952), who has alsodesigned and supervised the construction ofinstruments of this type which are now in usewith the Mount Wilson-Palomar reflectors.Special attention should be called to the new60-inch Cassegrain spectrograph designed byBowen which has recently been put into opera-tion. The faint limiting magnitude obtainablewith this instrument illustrates how great a gainin efficiency is possible with reflectors in thesecond category of size. Special mention shouldalso be made of the exceedingly efficient nebularspectrograph designed and constructed underthe direction of Mayall for the 36-inch Crossleyreflector of the Lick Observatory.

There are now several large and most usefulobjective prism Schmidt cameras in operation.Especially important new instruments of thistype have been put into operation recently atHamburg-Bergedorf, the University of Mich-igan, and Vanderbilt University. Exceedinglyefficient slitless spectrographs have been de-signed and constructed recently by Herbig(Lick) and Meinel (University of Chicago).

ReferencesBOWEN, I. S.

1947. Publ. Astron. Soc. Pacific, vol. 59, p. 253.1952. Astrophya. Journ., vol. 116, p. 1.

STROMGREN, B.1952. Astron. Journ., vol. 57, p. 65.

381014—56

Stellar AtmospheresBy Marshal H. Wrubel'

The outer region of a star in which the opticalspectrum is formed is referred to as the atmos-phere. This definition delimits a shell lyingbetween the deeper regions, which cannot bedirectly observed, and the higher regions, whichcontribute very little energy to the opticalspectrum (but which may be all-important inthe radio spectrum). Although it is convenientto treat this shell separately, it is certainly notindependent of the layers lying below andabove it.

Any discussion of stellar atmospheres logicallybegins with the sun. Here is a star on ourdoorstep; and the variety and accuracy of thesolar data we can accumulate far surpasses thatfor any other star. We may examine theradiation received from every point on thesun's disk and for every wavelength thatpenetrates the earth's atmosphere.

In spite of centuries of solar observation,new techniques continue to introduce newphenomena. To mention only two of the mostinteresting: the Babcocks' magnetograph,which gives a tracing of the atmosphere'smagnetic field point by point; and the McMath-Hulbert vacuum spectrograph, which has re-vealed a remarkable complexity of line struc-ture. Both of these innovations are triumphsof observational technique and will createchallenges to the theoretician.

The theoretical models that have been ex-plored until recently may be characterized asbeing in "hydrostatic and radiative equilib-rium"; that is, the atmosphere is supported byhydrostatic pressure and radiation is the solemeans of energy transport. The logical chainis this: the temperature distribution is foundeither theoretically or by limb darkeningmeasurements, and it depends on the broadcontinuous absorption characteristics of the gas.A complete physical description of the atmos-

1 Indiana University, Bloomlngton, Ind.

phere follows from the temperature distributionby using the hydrostatic equation, the absorp-tion coefficient, the ionization equation, etc.The resulting "model atmosphere" combinedwith the theory of the formation of absorptionlines is used to predict the observed spectrum.The only adjustable parameters of such a solarmodel are the abundances of the elements whichare determined by a comparison with observa-tions.

Although the reasoning is straightforward,many problems arise even in the static model.To list only a few:

(1) The temperature distribution of theupper layers is uncertain because of thedifficulty of observation close to the limb of thesun.

(2) The abundance of the second mostabundant element, helium, cannot be deter-mined directly since it is effectively "deadweight" in a low-temperature atmosphere.

(3) The fundamental data of transitionprobabilities for many lines are still missing.The great emphasis on nuclear and molecularphysics in recent years has attracted physicistsaway from research in atomic spectra. Apossible solution is for some young astronomersto learn the laboratory techniques and to do thework themselves.

(4) The theories of the shape of the line-absorption coefficient, i. e., broadening, absorp-tion versus scattering and noncoherent proc-esses, are not in satisfactory form. Of partic-ular interest in this connection is the recent useof a shock tube at the University of Michiganto obtain laboratory data on broadening.

(5) The problem of predicting the profileof an absorption line formed when radiationpasses through many layers of different physicalcharacteristics has been solved only for veryspecial cases.

(6) The cumulative effect of many absorp-119


tion lines (blanketing effect) is difficult toevaluate.

However, overshadowing all these difficultiestoday is the problem of describing a dynamicatmosphere. The theoretician no longer canavoid including large-scale motions in his solarmodels but the kinematical description is farfrom clear. One important source of informa-tion is the "granulation"; but the details areobscured by "seeing" in the earth's atmos-phere. It is to be hoped that Schwarzschild'sunorthodox plan to photograph the sun fromabove the "seeing" will be successfully carriedout in the near future.

Although simple models (such as those usedby de Jager) involving temperature inhomo-geneities in adjacent columns of the atmospheremay prove serviceable for a while, eventuallywe must have a model obeying the hydro-dynamical laws (better still the hydromagneticlaws) and including the transport of energy bya combination of convection and radiation.

Increasing use will undoubtedly be made ofhigh-speed electronic computers to obtainnumerical solutions to these difficult problems.Computers can certainly not create ideas, butthey can explore the consequences of variousassumptions to an extent previously impossible.It is a challenge to find efficient machine pro-cedures for calculating model atmospheres andline profiles under a variety of physical con-ditions. Computer techniques will probablybe second nature to astrophysicists a fewacademic generations hence; at the moment,however, theoreticians have been very con-servative in adopting these new methods.

As if the existing observations do not providesufficient problems, astronomers have watchedwith great interest as rocket spectrographs andradio telescopes have broadened the limitsof the observable spectrum. Lyman a has atlast been photographed and even soft X-rayshave been detected. Observations such asthese will have a profound effect on futuremodels of the sun.

There is very little generally accepted theoryconcerning the spectacular transient solarphenomena. Many solar observatories are in-volved in the accumulation and classificationof data from which theories will eventuallyarise, and a highly coordinated effort in this

direction is being planned for the forthcomingInternational Geophysical Year. Many prob-lems in this particular field are treated in-dividually in other articles of this volume.

The techniques developed to describe thesolar atmosphere have been applied withsome success to "normal" atmospheres of otherspectral types. In some cases it is possibleto obtain information that the sun will notprovide directly: for example, the abundanceof hydrogen relative to helium can be foundin principle from the spectra of hot stars.Unfortunately, no conclusive value can as yetbe given.

Problems immediately arise because of ob-servational limitations. There is just notenough light to work with. Photoelectrictechniques for the absolute spectrophotometryof stars are currently being developed by severalobservers and should yield important astro-physical data. Nevertheless we have onlyindirect measurements of limb darkening, andthese for only a few eclipsing binaries.

As the available light goes down, so does thepossible dispersion, and observations becomeless reliable. Some objects are out of reach ofall except the largest reflectors.

Surveys can be carried through by usingrough curve-of-growth methods based on ideal-ized models, but eventually we need detailedmodel atmospheres for a wide variety of spectraltypes. These are constructed by using theoret-ical temperature distributions and are testedby requiring the flux to be constant at everydepth. However, it is difficult to apply thiscriterion in early-type stars where the fluxis jagged due to the absorption edges of hydro-gen and helium.

As we go toward the cool stars, the im-portance of the blanketing effect grows and apremise upon which other models are foundedbreaks down—it is not possible to separatethe lines from the continuum. Entirely newtheoretical techniques will have to be combinedwith laboratory data on the absorption of mole-cules before we will have as detailed an inter-pretation of the atmospheres of stars of latespectral type as we have for the sun.

Nature has provided a unique tool for study-ing the outer regions of late-type giants.Binary systems like 31 Cygni, in which a small


hot component is gradually eclipsed by theextended atmosphere of its giant companion,provide a wealth of detailed information. Theobservations indicate a "network of prom-inences" far more extensive than that of thesun; and if an understanding of a dynamicatmosphere is important in the sun, it is vitalhere.

As the spectroscopic peculiarities becomemore unusual, the extrapolation of normal tech-niques becomes less successful. Thus, ourpresent understanding of the atmospheres ofpulsating variables, magnetic stars, Wolf-Rayet stars, etc., is rudimentary at best.Nevertheless, the tendency is to make quantita-tive measurements whenever possible and notto be satisfied with visual estimates.

It is obviously impossible to cover all aspectsof such a varied subject in a short survey.Fortunately several detailed discussions of solarand stellar atmospheres have recently appeared,and references to individual papers will befound there.

References

AIXER, L. H.1953. Astrophysics. Ronald Press, New York.

BEER, A., BD.1955. Vistas in astronomy. Pergamon Press,

London and New York.ClIANDRASEKHAR, S.

1950. Radiative transfer. Oxford UniversityPress, London.

HYNEK, J. A., ED.1951. Astrophysics. McGraw-Hill, New York.

KoURGANOFF, V., AND BUSBRIDGE, I. W.

1952. Basic methods in transfer problems. Ox-ford University Press, London.

KUIPER, G. P., ED.

1953. The sun. University of Chicago Press.UNSOLD, A.

1955. Physik der Sternatmospharen. Reviseded. Springer, Berlin.

WOOLLEY, R. V. D. R., AND STIBBS, D . W. N .1953. The outer layers of a star. Clarendon

Press, Oxford.WRUBEL, M. H., ED.

1954. Proceedings of the National Science Foun-dation conference on stellar atmospheres.Indiana University. Bloomington, Ind.

Aerodynamic Problems in Stellar AtmospheresBy Richard N. Thomas*

The term "cross-field research" occurs withincreasing frequency in astronomical studies.This new emphasis does not imply that as-tronomy, in the past, failed to utilize advancesin other sciences; rather, it indicates our grow-ing awareness that the physical problems ofmatter studied by the observational astron-omer are not far removed from those studied bythe laboratory scientist. Classically, astro-physics has dealt with matter under condi-tions that were strange and abnormal, com-pared to those in the terrestrial laboratory.Astronomy has been an observational science.The objects of its inquiry have been obscuredby the earth's atmosphere and situated at greatdistances; and the phenomena studied wereincapable of reproduction in the laboratory.Of all the "cross-field" liaisons that exemplifyour changing views, none is more interestingthan that between modern aerodynamics andastrophysics.

Modern astrophysics not only concerns itselfwith the common problems in the two sciencesbut also recognizes that the data required tosupport a theory or an anticipated futuredevelopment in one field may already exist inthe other. Also, the analytic methods thatare gradually being developed for a specificproblem in one field may already be commonknowledge in the other. For example, theadvent of high-altitude rockets, satellites, andthe possibility of extraterrestrial manned ob-jects will break the barriers imposed by theterrestrial atmosphere and the astronomicaldistance scale. Yet a satisfactory developmentof ultraspeed aerodynamics requires carefulattention to the problems of ablation from asolid moving through a gas, of radiative energyloss from a gas compressed rapidly to a highdegree, of the attendant ionization and chemicalreactions in the gaseous atmosphere, and of the

1 Harvard College Observatory, Cambridge, Mass.

methods of description of a gas under conditionsdeparting drastically from local thermodynamicequilibrium.

These problems are all of vital interest toastrophysics because our observations showthat such configurations occur in astronomicalobjects. Several of these problems, those con-nected with the ablation from a rapidly movingsolid object, form the fundamental basis ofmeteor astronomy and have led to the creationof astroballistics, a new field comprising bothastronomy and aerodynamics, which is de-scribed elsewhere in this series of articles.Others of these problems seem to occur withincreasing frequency as we acquire more andbetter data on stellar atmospheres and arethus faced with the need to construct a theorysufficiently precise to explain the data. Inthis review, we want to elaborate on certainaspects of stellar atmospheres and to emphasizethe continued interplay of astrophysical theoryand observation with aerodynamic theory andexperimentation, in a kind of checkerboardpattern. The exciting aspect concerns thecommon physical conditions occurring in thestellar atmosphere and the aerodynamic experi-ment. The frustrating aspect is the lack offacilities for experimental astrophysics, as itbecomes more and more evident that we canactually duplicate in the laboratory physicalsituations of direct astronomical interest.

Ten years ago, theorists concerned withstellar atmospheres were completely preoc-cupiod with the notion of a gaseous atmospherewhose properties were fixed by the threeassumptions: radiative transfer of energy, localthermodynamic equilibrium, and the absenceof any velocity fields large enough to affectthe structure of the atmosphere. The discoveryby Schwarzschild, 50 years ago, of the over-whelming efficiency of radiative energy trans-port, relative to convective, completely dom-

123


inated all thinking in subsequent years. Thelater recognition of atmospheric convectivezones in stars where hydrogen was not mainlyionized was of minor consequence; even inthese zones radiation dominates the energytransfer. There were, however, three "types"of atmosphere where "contrary" evidence wasbecoming difficult to rationalize away: theatmospheres of the variable stars, where thenotion of a velocity field had long existed butwhere the exploration of the physical con-sequence of its existence had been rudimentary;the atmospheres of certain supergiants, wherethe evidence for large velocity fields had existedfor 10 years but nothing had been done in theway of incorporating them into an atmosphericmodel; and the solar chromosphere, whereconsiderable evidence of velocity fields and ofnonequilibrium phenomena caused the break-down of virtually all our classical notions of astellar atmosphere.

There are two ways in which a velocity fieldcan perturb an atmosphere: through a transferof momentum or a transfer of energy. Whatlittle thinking on the influence of velocity fieldsthat did take place in these earlier days wasconcerned largely with the momentum effect.Velocity fields were simply imposed as an addedpressure term, and the thermal and radiativeproperties of the atmosphere were supposed toremain invariant. Unfortunately, the velocityfields discussed are often of such size that, if thetemperature of those atmospheric regions wherethe velocity fields exist are not higher than wewould infer from our customary theories, themotions must be supersonic. Such a situationmade it difficult to introduce the experience ofthe aerodynamicist in describing velocity fieldsin a compressible gas because such supersonicmotion, particularly of the quasirandom sort re-quired by the astronomical data, is in aerody-namics always accompanied by Shockwaves andby the rapid conversion of mass motion intothermal. Thus we had the unhappy situation ofan empirical fact (large velocities) forced into anartificial framework (no dissipation of me-chanical energy) because astronomers were re-luctant to abandon the atmospheric modelswhich had hitherto worked so well. Moreover,there were absent many of the features onewould expect to occur in a rapid production of

thermal energy—emission lines, excesses ofultraviolet radiation, etc. The situation, there-fore, was far from clear cut.

(If the reader expects the above history to beclimaxed by a happy solution in this review, hewill be disappointed. If these problems hadbeen solved there would be little point in record-ing their apparently illogical history. We dwellon the above pattern of thinking because, un-happily, much of it is today's pattern of think-ing also. At present we can report only onefforts to resolve these difficulties, not on theirsuccess.)

On the other hand, when the thermal dissi-pation of the mechanical energy of the velocityfield was included, the effect on the atmosphericmodel was, and is, likewise far from clear. Thisenergy transfer represents the most serious andthe most fascinating problem of all. First, anysuch energy transfer establishes a cyclic process,a mechanical energy source and a radiative sink.Such a situation is ideal for setting up a de-parture from the conditions of local thennody-namic equilibrium. Second, such energy transferintroduces the possibility of a perturbation onthe radiation field produced by the star. Con-sequently, the obvious subject for investigationis the problem of the energy transfer associatedwith velocity fields in stellar atmospheres.

The solar chromosphere offers the most nat-ural starting point for an investigation of thissort. First, from a general point of view, thereis one strong boundary condition on this investi-gation of aerodynamic effects in stellar atmos-pheres: the "classical" stellar atmosphere mod-els have been remarkably successful in repro-ducing the features of the continuous spectrumin the visual region. How can we introducesuch things as Shockwaves without affecting thecontinuous spectrum? In the sun, the largevelocity fields seem confined to the outer at-mosphere, where the total mass is too small tomake appreciable contribution to the continu-um. We may, however, determine how fardown toward the photosphere the effect of thechromospheric velocity fields extends by look-ing at weaker and weaker lines and the wings ofthe strong lines. With this information, we maybe able to answer the question of what observ-able consequence would result from an increaseby some factor in the energy of the velocity field.


Or, alternatively, what would we have to do toturn the solar atmosphere into that of a Wolf-Rayet star, where the whole atmosphere seemsa composite of aerodynamic motions?

Second, with a view to finding the origin ofthese velocity fields, the closeness of the sunand our ability to resolve its disk are of enor-mous value. Many astronomers believe thatthe chromospheric velocity fields simply reflectthe motion of the solar spicules, which aresupposed to arise from the granulation, whoseorigin lies in the hydrogen convection zone.Some of these astronomers regard magneto-hydrodynamic effects as essential in the produc-tion of spicules, from granules; others believethe transition can be effected from the aero-dynamic behavior of a compressible gas in agravitational field alone; and still others believethat the connection between spicules andgranules has yet to be established. Yet in allthese points of view several experimental andtheoretical aerodynamic problems stand out—problems which we can investigate jointly inastronomy and in aerodynamics: (1) What isthe behavior of a supersonic jet when radiationprocesses form a major dissipative mechanism ?Recent solar work on the stability of anionized gas against radiative cooling suggeststhat these stability considerations may fix, toa major degree, the permitted kinetic tempera-ture of the gas in the jet. (2) How does asupersonic jet composed of an ionized gas be-have in a magnetic field, with the field (a)parallel to the jet axis, and (b) perpendicular?(3) Suppose we have a velocity field, with time-dependent terms, in a gaseous atmosphere in agravitational field. How rapidly do the acous-tic waves, produced by this field and propa-gated upward, die out? What is the inter-ference pattern ? (4) If we superpose a fluctu-ating magnetic field on this velocity field withtime-dependent terms in an ionized medium,can induction effects accelerate a nontrivialamount of gas? (5) Last but hardly least inour incomplete list of problems, How do wespecify the spectroscopic state of the gas underthe environment of these four problems, andothers in the same physical sequence ?

Let us return for a moment to the suggestionthat spicules are the source of the mechanicalenergy which causes the chromosphere and that

381014—66 10

spicules arise from granulation. In this case,all stars with atmospheric convection zonesshould have chromospheres, that is, all main-sequence stars of spectral class later than~^40.But what of the earlier spectral classes?Should we expect no extended atmosphereshere, no velocity fields? Yet the existence ofdifferential velocities between emission andabsorption lines in stars of classes O and Bis well known. In a more extreme case, werefer again to the Wolf-Rayet stars, where thewhole atmosphere seems composed of a systemof jet-ejections, or streams of outgoing and in-going material. Whence arises this motion?Aside from the possibilities of the hydrogenconvection zone (and possibly a helium con-vection zone in the hotter stars ?) and inductioneffects in time-dependent magnetic fields, wehave two other sources of mechanical energy:stellar rotation and stellar pulsation.

I t is interesting to note that the first sugges-tion that aerodynamic turbulence might arisein stellar atmospheres arose from the observa-tion of the high Reynolds numbers associatedwith stellar rotations. Furthermore, we notethat "shell" stars are most often those having arapidly rotating nucleus. What is the aero-dynamic state of such a rapidly rotating gaseousatmosphere ? Does the star set up a system ofaerodynamic turbulence, anisotropic for thelargest scale motion, with the greatest velocitiesalong the direction of rotation? This wouldconform to laboratory experience. Or, is therotation velocity so high (greater than ourlaboratory experience) that rings of matter areejected? This problem has hardly beentouched, astronomically or otherwise.

The concept of stellar pulsation is relativelyold; the implications of stellar pulsation,relatively new. Given a gaseous atmosphereof arbitrary temperature and density distribu-tion, and an impressed oscillation of some ther-modynamic variable at some level in theatmosphere, what happens? Remember, weare not dealing with the terrestrial atmosphere,with temperature ~ zero so far as activationof internal degrees of freedom are con-cerned. We need a terrestrial disturbanceof ~Mach 10 to cause any but the most trivialluminosity. In the star, the ambient temper-ature ranges upward from~4,000°, and even


for Mach 2 one expects to find an observablechange in ionization and excitation. A literalinterpretation of velocity-curves would suggestAf>2 in a wide class of variable stars. In-deed, in some stars emission lines appear oversome portions of the cycle, and we mightinterpret their occurrence as implying theexistence of localized regions of high kinetictemperature. On the other hand, some stars,whose velocity variations imply compressionwaves of sufficient amplitude to be shock-waves, do not show these emission lines.Should we interpret these differing results assuggesting that the compression wave justdoes not have a sufficiently long path to developinto a shock before it is out of the visibleatmosphere?

Actually, two physical parameters seem toplay a major role in aerodynamic phenomenain the atmospheres of variable stars: the extent

of the h3Tdrogen convection zone and the extentof the atmosphere. The convection zoneappears to determine the phase relation betweenthe luminous flux and pressure at the "origin"of the disturbance; the extent of the atmospheredetermines the character of the disturbance,and, in a sense, has a self-amplifying aspect.The more the dissipation of energy in the at-mosphere, the more extended the atmosphere,hence the greater the chance for a shock todevelop and dissipate energy.

All these problems are tantalizing. Fewhave been investigated, none solved. Ourgreatest progress has been in our finally recog-nizing what the problems actually are. Ourgreatest challenge is the need to develop somephysical insight to the solution of these prob-lems which lie on the most advanced frontiersof both astrophysics and aerodynamics.

Emission NebulaeBy Lawrence H. Aller 1

Galactic nebulae that show an emission linespectrum are of two types. Those associatedwith the type I population are the so-calleddiffuse nebulae; those associated with thetype II population are the planetaries. Thediffuse nebulae are normally irregular in form,often of low density and surface brightness,and sometimes of considerable size (e. g.,Orion, 30 Doradus in the Large MagellanicCloud, and NGC 604 in the TriangulumSpiral). The planetaries are often symmetricalin form and smaller than the diffuse nebulae.Many have a higher surface brightness andhigher density than the better known diffusenebulae.

Types of observations

Because of their different characteristics, thekind of equipment and technique best suitedfor the one type of emission nebulae is notnecessarily best for the others. In studies ofdiffuse nebulae, wide-angle cameras of highspeed and detectors sensitive to low surfacebrightnesses are desirable. For studies of theplanetaries a large scale (and therefore a largetelescope) is essential, since the angular di-ameters of all but the largest of these objectsamount to only a few seconds of arc. Theprincipal types of observation that can bemade are the following:

(a) Direct photographs with the aid of nar-row band-pass filters to isolate monochromaticnebular radiations. In particular, interferencefilters have proved very useful in picking upfaint Ha emission regions throughout theMilky Way. When most of the nebular radi-ation is concentrated in a few strong, well-separated lines, the filter technique is extremelyvaluable for studying problems of stratificationand structure.

1 University of Michigan Observatory, Ann Arbor, Micb.

(b) Photographs with the aid of Fabry-Perot etalons and suitable plate-filter com-binations are extremely valuable for studyingthe internal motions in extended diffuse neb-ulae. This technique, introduced by Fabryand Buisson more than 40 years ago andfurther developed by Baade, Goos, Koch, andMinkowski, has been applied with great successby Georges Courtes at Haute-Provence. Stud-ies may be made with relatively small wide-angle telescopes that will yield considerableinformation on the mass motions and internalturbulence. It is important that such in-vestigations be carried out not only in Ha butalso in the forbidden radiations of [O II] and[O III] whenever possible.

(c) Spectrograpbic studies fall essentiallyinto three categories. First are slit spectro-graphic studies of the line and continuousspectra to obtain line identifications and in-tensities from which the physical state in thenebula or nebular filament may be inferred.For this type of problem a so-called nebularspectrograph with high speed and low dis-persion is necessary for all but the brightestobjects. Planetary nebulae have been studiedrather intensively, particularly by Bowen andhis coworkers, but the spectra of diffuse nebulae(except Orion and a few special objects) havenot been adequately investigated. Spectro-graphs of high dispersion may be used forstudies of the internal motions in small, brightplanetaries. Olin Wilson's multislit has provedparticularly valuable in such studies. Finally,slitless spectrograms are essential for the studyof monochromatic images of small nebulaewhen radiations that fall close to one anotherare to be separated; e. g., the [N II] X6548,6584 and Ha 6563 Lines.

(d) Polarization measurements are importantin these objects where scattering by small solid

127


particles occurs and in radio sources. Electronscattering is rarely a factor in the continuousspectra of diffuse nebulae.

(e) Radio-frequency observations of the free-free emission from diffuse nebulae have beenmade by Haddock and his coworkers. Even-tually, with the development of larger antennae,it should be possible to measure the emissionfrom the brighter planetaries as well. Theseobjects have high surface brightnesses but smallareas so that their total flux is small. The21-cm data supply information on those regionsof space where hydrogen is neutral.

Observable parametersThe following data can be obtained for typicalemission nebulae:

(a) The apparent size, form, and structureCan be measured in each of the stronger mono-chromatic radiations.

(b) The surface brightness and isophotic con-tours can be measured for each of the strongermonochromatic emissions.

(c) The wavelengths, identifications, and in-tensities of the lines and the energy distributionin the continuous spectra can be measured fortypical nebulae.

(d) The distances of diffuse nebulae may befound from the apparent magnitudes, colorexcesses, and luminosity classes of their illumi-nating stars. The distances of the planetarieseannot be established so easily, and for themoment we shall have to be content withstatistical distances based on the association ofthese objects with the central bulge of ourgalaxy and external galaxies.

(e) The relationship between the structureand internal motions can be established from acomparison of isophotes of slitless spectrogramsor monochromatic photographs and radial veloc-ity measurements. I t is possible to establishwhether the motions are orderly expansions orturbulent streaming, and to correlate the motionwith the appearance of the object. O. C.Wilson has given a rather complete picture forthe planetaries. Similar studies for the diffusenebulae should prove worthwhile.

Interpretation of the observational data interms of physical parameters

The densities and temperatures of the radiatinggases may be established from the surfacebrightness and from spectroscopic data. If thenebula is smooth and unobscured and anestimate of the electron temperature is avail-able, a measurement of the surface brightnessin Ha or preferably in the Balmer continuumwill provide an estimate of the electron density.Both diffuse nebulae and planetaries are fre-quently highly obscured. If a measurement ofthe Balmer decrement is available, an estimateof the space absorption can be found readily.Albert Boggess III has shown how one maycombine optical measurements of surface bright-ness in Ha with radio-frequency measurementsto get the space absorption and densities forcertain diffuse nebulae for which Balmer decre-ments have not been measured.

The forbidden lines provide estimates of theelectron density N« and give us our best guessesconcerning the electron temperatures T(. If allthe radiations originate in the same strata onemay obtain N ( and T« from a comparison of theforbidden line intensities of [O III] and [N II],for example. Often, however, the [N II] and[O III] lines are predominantly produced indifferent regions of the nebula. Hence, someother method is needed. A number of yearsago, Ufford, Van Vleck, and I called attentionto the fact that the intensity ratio of the 3726and 3729 [O II] lines varied from one nebula toanother, apparently depending on the density.Seaton calculated improved target areas for thecollisional excitation of the [O II] lines so thatquantitative estimates of the density could bemade from the observed line intensities. Oster-brock has recently derived densities in the fila-ments of the Network Nebula, in the OrionNebula, and in other objects from the X3726/X3729 data.

Some outstanding problemsAt the moment, the physical nature of theexcitation of the bright line spectra of the gase-ous nebulae appears to be well understood. The


continuous spectra of normal, purely gaseousnebulae are less well understood, but seem to beinterpretable in terms of recombination of hy-drogen and helium ions with electrons and thedouble-photon emission. Further quantitativemeasurements of the nebular continua areneeded to test various theoretical suggestions.

Further calculations of transition probabilitiesfor forbidden lines, especially those of iron invarious stages of ionization, are needed. Con-tinuous absorption coefficients should be calcu-lated for abundant ions, and further targetareas for the collisional excitation of metastablelevels should be computed. The accuracy ofthe available theoretical target areas ought tobe improved.

Numerous theoretical problems connectedwith the structure and especially the kinematicsof gaseous nebulae remain to be formulated andsolved. For example, what holds together thethin filaments in such an object as the NetworkNebula? It is generally assumed that magneticfields are responsible, but the character and par-ticularly the origins of these fields remain ob-scure. Any real understanding of the forms ofdiffuse nebulae must await some fundamentalprogress in the aerodynamics of compressibleionized gases moving at supersonic velocities inthe presence of magnetic fields. Classical hydro-dynamics and turbulence theories will requiremuch amplification and extension before theycan be applied to the diffuse nebulae andplanetaries.

The relation between the diffuse gaseousnebulae and the solid particles of the inter-stellar medium requires further theoretical andobservational study.

Perhaps the most exciting problems are thoseposed by the gaseous nebulae that are strongradio sources. Oort and Walraven recently in-terpreted the continuous spectrum of the CrabNebula as "synchrotron" radiation from fastelectrons moving in a magnetic field. Possiblysuch radiation is always or nearly always associ-ated with strong radio-frequency emission,although powerful radio sources do not neces-sarily emit a strong optical continuum. Theorigin of the magnetic fields and the source ofthe energy of the particles is not explained. I tis not sufficient to blame it on the supernovaprocess; the detailed mechanism must be workedout.

New equipment and observational techniqueshave expanded and are continuing to expandour empirical knowledge of emission nebulae.Theoretical interpretations of the forms andkinematics are lagging behind. The problemsposed are among the most difficult in theoreticalastrophysics, a fact which makes the subject ofbright-line nebulae all the more challenging.

Some suggested referencesThe following books and review articles con-tain further references to the literature:ALLER, L. H.

1956. Gaseous nebulae. Chapman and Hall,London.

VAN DB HULST, H . C , BD.1955. Second symposium on cosmical aerodynam-

ics held at Cambridge, 1953. Publishedfor Int. Astron. Union by North HollandPublishing Co. and by Interscience Press,New York.

WURM, KARL1951. Die planetarische Nebelu. Akademische

Verlag, Berlin.

Double and Multiple StarsBy Otto Struvex

The emphasis at the present time is uponproblems connected with the origin and evolu-tion of double stars. G. P. Kuiper's discussionin three articles in the Publications of the Astro-nomical Society of the Pacific2 in 1935 and 1955forms a sound basis for future work. He attrib-utes the origin of double stars to the formationof separate condensations in a primordial inter-stellar cloud which possessed too much angularmomentum, or too much mass, for the forma-tion of a single star. In this respect the originof a multiple system does not differ significantlyfrom that of a star cluster. But the later evo-lution of a binary differs from the evolution of aloose cluster. A visual double star is a verystable formation and is not likely to changevery greatly during many hundreds of millionsof years, while a galactic star cluster may bedispersed in the Milky Way in an interval of tenor a hundred million years (though at least onecluster, NGC 67, is believed by H. Johnson andA. Sandage to have an age of 5X10* years).

Since we have good reasons for believing thatsome galactic clusters are very young (the asso-ciations of Orion and f Persei, the double clusterin Perseus, NGC 6231, etc.) while others aremuch older (Pleiades, Hyades, Praesepe, NGC725 and NGC 67), it would be of great interestto know whether any binaries in these clustersshow differences in their properties that mightbe related to the differences in their ages. Asyet, our information concerning all types ofbinaries in galactic clusters is too fragmentaryfor us to draw any conclusions.

Of special interest is the question of the evo-lution of close binary systems. From Kuiper'swork on /3 Lyrae, we know that two types ofinstability may occur, with matter escapingfrom one or both components either through the

1 Department of Astronomy, University of California at Berkeley,Calif.

» Vol. 47, pp. 15, 121, 1935; vol. 67, p. 887. 1955

bottleneck near the inner Lagrangian point, Llt

of the critical zero-velocity surface (instabilityof type A), or through an opening at the firstouter Lagrangian point, Zj (instability of typeB). Kuiper thought that the instability of fiLyrae is a combination of types A and B, butmore recent work seems to indicate that bothcomponents are unstable and lose matter in theform of two streams moving in opposite direc-tions through the bottleneck near L\. Therecan be little doubt that this phenomenon pro-duces the secular change in the period of 0Lyrae. The loss of matter in /3 Lyrae, if thestreams are ultimately expelled into space, mustamount to about 5X10" gr/sec in order to ex-plain the increase of the period. Since the massof the star is of the order of 103* grams, the lifeexpectancy of /3 Lyrae should be of the order of100,000 years.

No other star is known that closely resembles/3 Lyrae (in itself a good indicator of the shortduration of its particular stage of evolution).However, many other systems have gaseousrings or streams and many of them, accordingto F. B. Wood, have changing periods, especiallywhen at least one component is a large subgiantwhose volume fills its loop of the critical zero-velocity surface.

One of the most pressing needs in this field isa systematic, prolonged study of the periods ofall suitable eclipsing and spectroscopic binaries.The photometric observations would be rela-tively simple, but should be made with photo-electric means. Results of great value wouldbecome available in a few years; but even moreimportant would be a list of accurate periodsduring the next 10 years which could later becompared with similar lists at intervals of about25 years. We cannot start too soon with sucha program!

The spectroscopic observations could, formany systems, be carried out with telescopes

131


of moderate size. We need periodic deter-minations of the velocity curves of all thebrighter spectroscopic binaries. Here againresults may be expected immediately, not onlywith regard to the periods but also with regardto changes in the orbital elements. Algol,Spica, 5 Orionis, and many more should bereobserved at least once every 10 years.

One of the most puzzling questions in thefield of double-lined spectroscopic binaries isthe interpretation of the tendency of manysystems to show a strengthening of the violet-displaced absorption lines relative to the red-displaced components, irrespective of whichof the two stars happens to be approaching.This phenomenon was discovered by MissMaury at Harvard in the spectra of n Scorpiiand V Puppis. But later observations atMcDonald Observatory failed to show thisvariation. Yet, independently, we found thatit was conspicuous in Spica and in several othershort-period spectroscopic binaries of earlyclass B or O. A recent study of v Scorpii byS. J. Inglis shows that it is present in thehelium lines but that it may be reversed insign in the hydrogen lines. I had previouslyassumed that there is more gas in front ofthe advancing hemispheres of each componentthan in front of the receding hemispheres.However Inglis' result suggests that it maybe caused by a difference in ionization andexcitation. As far as I know, there are nophotoelectric measurements that would helpus in explaining this phenomenon. Since itoccurs only in close systems (it is also con-spicuous in many W Ursae Majoris systems)I suspect that it is in some way related to theexistence of gaseous streams.

S. S. Huang has recently asked whetherthere might not be accumulations of gaseousmaterial at those two Lagrangian points whichcorrespond to the equilateral triangle solutionsof the restricted three-body problem. Unlessthe mass ratio of the binary is very differentfrom one, these solutions are not stable, butthere could still be regions of increased densityin the gaseous ring at L% and L4. I t is possiblethat a more careful study of the eclipses of theemission features in the hydrogen lines of RWTauri, RW Persei, SX Cassiopeiae, U Sagittae,

and many more would enable us to locate theregions in which these emission lines areproduced. Joy's original interpretation of RWTauri in terms of a ring around the brightercomponent of the binary, or my later interpre-tation in terms of a ring of gas surrounding theentire system, may have to be modified in thelight of Huang's idea.

During the interval 1940-1950 severalMcDonald astronomers made a superficial sur-vey of the velocity curves and spectroscopicfeatures of about 100 eclipsing variables. Mostof the systems presented no peculiarities andthe results were sufficient for statistical studies,but a considerable number yielded results ofexceptional interest. A few of these starshave since been investigated more exhaustively(U Cephei, U Sagittae, SX Cassiopeiae, RXCassiopeiae, RZ Scuti), but quite a numberhave not yet been scrutinized. I can mentionhere only a few of them. WX Cephei (J.Sahade and C. U. Cesco) shows amazingchanges in the intensities of the two Ca IIabsorption components. They are unlike thechanges discussed above. DN Orionis has anexceptionally small range of velocity and,therefore, a very small mass function. It isone of the most peculiar binaries in this respect,resembling XZ Sagittarii (J. Sahade). Bothmust possess subgiant components of very smallmass (about 0.1 or 0.2©) but of large di-ameter. These components violate the con-ventional mass-luminosity relation. AW Pegasiand V 377 Centauri (Sahade) are even morepeculiar, and we have been unable to interprettheir spectra. Details about these and othersystems may be found in the Astroph3rsicalJournal or in the Contributions from theW. J. McDonald Observatory.

The famous visual double star 7 Virginishas been suspected by Lick astronomers tovary in light. Since both stars can be observedspectroscopically with high dispersion, a testof each component (or of the integrated lightof both) for variability would be of interest.Any visual binary whose orbit is known shouldbe investigated at least in integrated light forvariability. The discovery of an RR Lyraeor another pulsating component in a double star,whose masses are known, would be a major


triumph. It might even be possible from thecolors (or spectra) and magnitudes to pick outpromising pairs, but the chances for successcan hardly be estimated. I recommend thisimportant search only to those who haveother problems that are certain to give sig-nificant results.

A few eclipsing binaries must be observed asnearly continuously as is possible, both photo-e lec t r i ca l ly and spectroscopically. Amongthese I should list e Aurigae, W Cephei,/S Lyrae, UX Monocerotis (R. Lynds suspectsthat the small, bright, and hot component ofthis system is a short-period pulsating variable,while the other component is a large and coolsubgiant). No spectrographic equipment isnow available for such continuing observations.I therefore endorse Th. Dunham's project forthe construction of a new spectrographictelescope.

I hope that someone will have observed thelight curve of « Aurigae over a wide range ofwavelengths. It is important to know whetherthe depth of the eclipse (a totality began Dec.15, 1955, ending 330 days later) is really thesame in all wavelengths.

What is the dispersion of the mass-luminosityrelation, expecially at the upper end? Ifnuclear evolution proceeds without appreciablechange in mass, we should expect some scatteramong the O- and B-type stars. Most of ourpresent information comes from the velocitycurves of double-lined binaries, and this prob-lem is looked after by the astronomers at Vic-toria. Special attention should, however, bedevoted to such stars as J. S. Plaskett's massivebinary, in which the more massive componenthas the weaker absorption lines; the profilesof the lines of this component show irregularchanges.

It is not yet clear whether the yellow and redgiants in Population I obey the conventionalmass-luminosity relation. K. O. Wright's workon Capella seems to indicate that it does, andmy own earlier measurements suggested onlya relatively small departure. D. M. Popper'swork on f Aurigae and other similar systemsmay soon solve this problem. It may helpus in constructing the evolutionary tracks ofmassive stars, but here again we must be

prepared to find that the giants lose mass byexpansion. A. J. Deutsch's work on the ex-panding shell of the system of a Herculis sup-ports this hypothesis.

What is the nature of the so-called Trumplerstars? Some of them occur in binary systemsbut we have as yet no firm knowledge of theirmasses. As Trumpler has shown, their lineshave large red shifts which, if interpreted bythe theory of general relativity, would givevery large masses (or very small radii). Thereare some reasons for believing that their massesare not more than about 75o> But how dowe then explain the red shifts?

Are all post-novae and all SS Cygni starsclose binaries? And why are there so manyWolf-Rayet stars which are components ofclose binary systems? Are all of these objectsthe products of the evolution of binaries (asHuang suggests in a recent article in the Astro-nomical Journal8) and are the RR Lyrae starsthe products of nuclear evolution of singlestars (or of wide visual binaries) uncomplicatedby the loss of mass resulting from the instabilityof a close binary? It is not yet clear why lossof mass into space (as distinguished from theformation of streams within a binary, or theexchange of mass between the components ofa binary) should be significantly different.We must assume that once a stream of gas isgenerated in a binary it is somehow expelledfrom the entire system, and that the absenceof such streams in single, slowly rotating starsimpedes their loss of mass by "corpuscularradiation."

It would be unreasonable to conclude thisreport without stressing the importance ofcontinuing routine observations of all classesof double and multiple stars. In preparinghis program, an investigator should not beconcerned solely with the solution of immediateproblems but should devote a part of his timeto the accumulation of accurate observationswhich will bear fruit long after his death.The most important astronomical problems—those of evolution of stars, of perturbations indouble star orbits, of slow periodic or irregularvariations, and many more—require tens orhundreds of years of persistent effort. It is

•VoLftl, p. 40, 1956.


especially difficult to sustain an enthusiastic every astronomer has profited from the observa-attitude on the part of a scientist if he does tions of his predecessors; he should thereforenot expect to gather the harvest resulting be willing to pass on to his successors a storefrom his labor. Nevertheless, we must all of good observations to make up for what hefollow the dictum of Ejnar Hertzsprung: has inherited from former generations.

Stellar DynamicsByU. Van Wijk1

Although stellar dynamics is largely a theo-retical field, anyone working in it has a stronginterest in observational results concerning themotions of the stars—their radial velocities andproper motions.

One of the more active areas of stellar dy-namics has been the study of the interactionbetween stars and interstellar clouds insofar asit affects the motions of the stars. In papersby Spitzer and Schwarzschild (1953) and byOsterbrock (1952), the effect of randomly dis-tributed interstellar clouds has been considered.One may hope that the 21-centimeter bandobservations will ultimately lead to a morespecific idea about the distribution of inter-stellar clouds and that the effect of nonrandomdistributions, such as a spiral arm, may thenbe considered.

A field which is basically quite fruitful butwhich has been lying fallow recently is that ofspherical clusters and, particularly, the escapeof cluster stars. The papers of Spitzer (1940)and Chandrasekhar (1943) are still the basicreferences, and it may well be that furtherprogress will strongly depend on the readyavailability of electronic calculators.

An interesting recent development is the1 Princeton University Observatory, Princeton, N. J

attempt to interpret the present kinetic prop-erties of the stars in terms of the conditionsunder which they were formed. Examples ofthis approach are the papers by Belzer, Gamow,and Keller (1951) and van Wijk (1956). Herewe approach the border line between stellardynamics and the study of galactic structure.But two subjects which are just beyond thisborder line may still be mentioned here becauseof their intimate connection with our presentsubject: the work on pseudointegrals of motionby Lindblad and the study by Blaauw (1952)of the dissolution of very loose galactic clusters.

ReferencesBELZER, J., GAMOW, G., AND KELLER, G.

1951. Astrophys. Journ., vol. 113, p. 166.BLAAUW, A.

1952. Bull. Astron. Inst. Netherlands, vol. 11,p. 414.

CHANDRASEKHAR, S.1943. Astrophys. Journ., vol. 98, p. 54.

OSTERBROCK, D.1952. Astrophys. Journ., vol. 116, p. 164.

SPITZER, L., JR.1940. Monthly Notices Roy. Astron. Soc. London,

vol. 100, p. 396.SPITZER, L., JR., AND SCHWARZSCHILD, M.

1953. Astrophys. Journ., vol. 118, p. 106.VAN WIJK, U.

1956. Astron. Journ., vol. 61, p. 277.135

Variable StarsBy C. Payne-Gaposchkin*

Variable stars play a dual role in astronomy.First, they furnish standards of luminosity anddistance. Second, their own physical proper-ties throw light on energy sources, the processesof pulsation and explosion, and the circum-stances of stellar formation.

Standards of luminosityThe RR Lyrae stars, Cepheids, novae, andsupernovae have been the most importantindicators of distance. Their average lumi-nosities (mean for the pulsating stars, maximumfor the explosive ones) are known with an un-certainty of a few lOths of a magnitude. Weare now concerned with determining the disper-sion of luminosity for stars of these groups.

The absolute magnitudes of RR Lyrae starswill emerge from critical comparisons of thecolor-magnitude arrays of globular clusters.Probably they will be found to be related to themetal richness of the cluster stars. Thestriking differences between the noncluster RRLyrae stars of our neighborhood and those nearthe galactic center, when interpreted in thelight of the data from globular clusters, willfurnish a new approach to the dimensions ofour galaxy. These studies, which requirespectroscopic analysis and photoelectric ac-curacy, can be made only with the largesttelescopes. The extension of this work to thenearby, bright globular clusters of the SouthernHemisphere would be most important.

The dispersion in the absolute magnitudesof Cepheid variables and its relation to colorand spectrum is likewise an investigation tobe carried out in the Southern Hemisphere.The Cepheids in the Magellanic Clouds furnishperhaps the most urgent and profitable problemin applied variable star astronomy; the studyshould be made concurrently with photo-electric and spectroscopic equipment. I t should

> Harvard College Observatory, Cambridge, Mass.

be borne in mind, however, that the MagellanicCepheids may differ in detail from galacticCepheids of like period. The study of colorand spectral variation of galactic Cepheidsshould therefore be pressed, with especialattention to the effects of selective absorption.Here, again, material from the SouthernHemisphere is of particular urgency; therich Carina region seems comparatively freeof absorption, and radial velocities of Cepheidsin this area will fill out our information on therelation of the Cepheids to galactic rotation,and thus on the dimensions of the galacticsystem.

The absolute magnitudes of novae haverecently been placed on a firm basis by thelight curves of the novae in Messier 31. Prob-ably the novae in other nearby galaxies willbe the most valuable means of extending thisinformation.

The absolute magnitudes of supernovae,the distinction between the two types, and theirrelationship to ordinary novae must againrest on studies made with the largest telescopes.

Physical problems

Detailed study of the brighter variablestars is continually revealing new types ofvariation. Future progress in variable starastronomy calls for understanding the relation-ships of these types to one another and of thephysical processes involved in them. Whilethe discovery of variable stars should certainlycontinue, it seems probable that systematicstudy of chosen typical specimens is moreimportant than indiscriminate search for faintvariables. However, it should not be assumedthat all undiscovered variable stars are faint;even among the naked-eye stars there may stillbe a number of important examples hithertoundetected. Means for tracing the variationsof brightness of stars that are spectroscopically

137


interesting, such as shell stars and Of stars,should not fall into abeyance; in this regardthe maintenance of adequate sky patrols(not necessarily with powerful instruments,as the profitable stars are likely to be brighterthan the tenth magnitude) is a project worthyof continued support in both hemispheres.

It is important that the variable stars chosenfor systematic investigation be studied withthe greatest available precision. Brightnessand color should be measured photoelectrically.Spectra should be photometrically analyzed,and all spectra should be suitably standardizedfor this purpose. Estimates of the intensitiesof absorption and emission lines are not goodenough for modern purposes. I t should beemphasized that the red and near infraredare now as accessible as the photographicregion, and often provide data that are equallyor more significant.

Lastly, a star that is varying should bestudied concurrently by photometric and spec-troscopic methods; this is of especial importancewhen the variation is not strictly periodic.

Eclipsing stars.—Of unique importance be-cause of their contribution to our knowledge ofstellar masses and dimensions, eclipsing starshave entered a new era with the coming ofphotoelectric photometry. Any eclipsing starthat gives a two-spectrum velocity curve and agood light curve adds to our fundamentalknowledge of stars.

Several groups of eclipsing stars give unusualpromise. The combinations of a blue star anda red giant with an extensive atmosphere areimportant in the study of stellar envelopes; inaddition to the well-known specimens now beingactively studied, others are undoubtedly to bebe found in lists of "composite spectra."

The remarkable analogies between the short-period UX Ursae Majoris, the system of NovaHerculis, and the pair that constitutes AEAquarii invite a search for such stars, whichwill throw light on the processes of stellarexplosions.

Periodic variables.—The classical Cepheids andthe parallel sequence of Population II variablesare rich in possibilities. Concurrent studies ofbrightness, color, spectrum, and radial velocitywill be needed before the interrelationships ofthese groups are clarified. Careful spectro-

photometry will probably reveal differences ofcomposition. Equally important is the devel-opment of the physical theory that bears on thepropagation of waves through the envelopes ofthese groups of stars.

The long-period and other red variables in-volve similar theoretical problems. Theirspectra also invite quantitative observationalwork: the identification and intensities ofbright lines are of obvious importance, andquantitative study of the absorption spectra(admittedly beset with grave observationaldifficulties) is also a problem of the firstimportance.

Explosive variables.—The most difficult prob-lem, which can be attacked only with thelargest telescopes, involves the physical proc-esses of the supernovae, and especially theunsolved problem of their spectroscopic changes.

The study of the novae proper has alwaysbeen hampered by the unexpectedness of theoutbursts. Too often, alas, a nova has beenstudied from a sense of duty, and each contribu-tion thus tends to be isolated from relatedstudies. Every suitably equipped observatoryshould have a nova program prepared in ad-vance, and the supplies necessary to put it intoaction. The program should from the firstinclude the red and near infrared, and especialattention should be paid to photometricstandardization.

The U Geminorum stars are so faint that thenecessary spectroscopic studies require power-ful equipment. The likelihood that all aredouble stars and the possibility that many arespectroscopic binaries of short period makethem of exceptional interest. Photometricstudies might reveal that some of them areeclipsing stars; in addition, it would be impor-tant to learn whether the recently observedrapid changes of SS Cygni are a common featurein the rest of the class.

Irregular variables.—The faint irregular vari-able stars in nebulosity are destined to play animportant role in ideas of the origin and develop-ment of stars. The discovery of these "Tassociations" and the concurrent spectroscopicand photometric study of their behavior areextremely important provinces of variable starastronomy.

Interstellar MatterBy Lyman Spitzer, Jr.1

The study of interstellar matter may bedivided into two main parts. Firstly, there isthe measurement of the interaction betweeninterstellar material and electromagnetic radia-tion. The objective of this research is todetermine, from the observed radiation reachingthe earth, the properties of the materialbetween the stars. While this study is pri-marily observational, the interpretation of theobservations in terms of the number of absorb-ing atoms or of the properties of the absorbingsolid particles may involve much theoreticalanalysis.

Secondly, there is the physical analysis ofthose processes that occur in interstellar space.In this study the density, composition, andspatial distribution of interstellar material areobtained from the observational research de-scribed above, and the various physical proces-ses that occur are deduced from the laws ofnature, as determined in terrestrial laboratories.Thus, for example, the internal temperatureof the small solid particles or grains is com-puted theoretically, the interactions betweenatoms and grains are analyzed, and the methodsby which interstellar material can condenseinto a new star are deduced.

At present our information on material inspace is seriously limited both by observationalinaccuracies and by a lack of information onthe relevant laws of nature. During the nextfew decades significant advances in both theserespects may be anticipated.

New experimental techniques promise agreat increase in observational accuracy. Ofparticular importance is the imminent develop-ment of photoelectric spectrophotometry. Thecombination of a photoelectric cell and aspectrograph will have many important ad-vantages in interstellar research. One pos-sibility, which is already being explored by

1 Princeton University Observatory, Princeton, N. J.

Stromgren at the Yerkes Observatory, is theaccurate color measurement for stars of allspectral types. Accurate measures of absorp-tion lines yield very precise determinations ofspectral type, and hence of the intrinsic stellarcolor. The color excess, or deviation betweenthe actual color and the intrinsic color, maythen be found for any star. With this develop-ment we should be able to determine thedensity of small solid particles at points in thesky very close together, and thus obtain a moredetailed picture than is now possible of thedistribution of interstellar material.

Another suitable program for photoelectricspectrophotometry is the precise measure ofinterstellar line profiles. With this techniqueit is possible to determine both the number ofabsorbing sodium and calcium atoms in theline of sight to each star and the precise velocitydistribution of the atoms. Still another pos-sibility is the measurement of much fainterinterstellar lines than are customarily observed.Thus, detailed information can be obtained onthe spatial distribution of such interstellaratoms as neutral Ca and ionized Fe and Ti.At present the interstellar lines produced bythese atoms can only barely be detected in avery few stars.

Radio astronomy is another very importantobservational development in this field. As thesize of radio telescopes increases more and more,the information about the distribution of neutralhydrogen will become more and more detailed.Information about other types of atoms mayalso be obtained, perhaps, with radio tech-niques. Evidently our exploration of inter-stellar clouds will become much more preciseand much more extensive in the years to come.

This increasing detail of information, andthis very great increase in the accuracy of inter-stellar maps that can be drawn, will naturallylead to an expanded effort on the analytical and

139


theoretical front. For example, data are grad-ually accumulating on what might be calledinterstellar meteorology—the great problem ofthe motions and transformations of the inter-stellar clouds. As in terrestrial meteorology,this subject must include not only ordinarydynamics but also the effects of radiation, bothemitted and absorbed. Changes of state, in-volving the combination and dissociation ofmolecules and atoms, must also be considered.These phenomena are probably influenced pro-foundly by the presence of both large-scalemagnetic fields and cosmic rays, which have nocounterpart in terrestrial meteorology. Evi-dently the comprehensive analysis of the manysimultaneous processes will be a vast and chal-lenging adventure.

Another theoretical problem concerns thephysical chemistry of grains. The boundaryconditions in this problem are simple. Weknow approximately the conditions under whichatoms in interstellar space gather together toform solid particles. However, the nature ofthese particles is completely unknown. In par-ticular, the detailed structure of the grains, theforces holding them together, and, in fact, allexcept the grossest properties of these smallsolid particles are completely uncertain. A

precise, elaborate, and ingenious laboratoryapproach, together with a considerable amountof theory, is probably required to shed light onthe situation. The interactions of grains withatoms, including the rate at which the grainsgrow and are destroyed and the various selec-tive ways in which they interact with differenttypes of elements, are of particular interestastronomically. Another important problemconcerns the magnetic properties of such grains,the solution of which also probably demands acombination of observation and theory.

A central theoretical problem is, of course,the one of evolution. How do clouds evolveover a period of a billion years or more, and, inparticular, how are new stars formed from theinterstellar medium? The detailed answer tothese questions is perhaps the most difficultpart of the entire field, and from a logical stand-point should perhaps be postponed until therest of the subject is fully developed. In viewof the very great interest in this particularproblem, however, it appears certain that dur-ing the next few decades much new work willbe done. While no final results may be ex-pected in this area for some time, it seems verylikely that ingenuity and hard work will producemany glittering and intriguing hypotheses.

21-Centimeter ResearchBy Bart J. Bok '

The discovery of the 21-centimeter line ofneutral hydrogen by Ewen and Purcell (1951)and the subsequent confirmation of this dis-covery by Dutch and Australian groups (Mullerand Oort, 1951; Pawsey, 1951) marked thebirth of a new era in the study of our MilkyWay system and of the universe of galaxies.That radio radiation with a wavelength of 21cm emanating from neutral atomic interstellarhydrogen might be sufficiently strong to berecorded was first suggested by van de Hulst(1945) and was predicted independently byShklovsky (1949). But not until the actualdiscovery of the line did astronomers begin torealize how powerful a tool for future researchhad become available.

Research in the 21-cm field may at presentbe divided roughly into three broad areas.The first concerns the spiral structure of ourgalaxy, and has yielded spectacular first results.Really detailed analysis, however, has onlybegun, and will undoubtedly be fruitful infuture studies of spiral structure. The secondcategory is that of related optical and radiostudies of the fine structure of the interstellargaseous medium in our galaxy. We shall listbelow some of the topics of research mostimportant here, but it is already clear that the"21-cm astrophysicist" has waiting for him amultitude of specific research problems. Thethird category is that of extragalactic 21-cmresearch, one in which the work has barelystarted but which obviously will attain increas-ing importance as the reach of our equipmentincreases.

One should bear in mind that up to thepresent all but one of the five principal radiotelescopes that are active in the 21-cm field areparaboloids with apertures of the order of 25to 30 feet, the one exception being the 50-foot

1 Harvard College Observatory, Cambridge, Mass., and Common-wealth Observatory, Canberra, Australia.

paraboloid at the Naval Research Laboratoryin Washington. Also, most of the work untilnow has been done with electronic equipmentwith bandwidths of 40 and 15 kc/sec (equivalentto resolutions in radial velocity of 8 and 3km/sec) and very few observations have beenmade with bandwidths between 5 and 10kc/sec (equivalent to radial velocity resolu-tions of 1 to 2 km/sec). Within a year or sothere should be in operation at least a halfdozen radio telescopes with apertures of 60 to100 feet, all of them steerable and with surfacetolerances sufficiently strict to permit full useat the 21-cm wavelength. Research at 21 cen-timeters has just been started with two largeradio telescopes—the 83-foot instrument atDwingelo in Holland and the 60-foot George R.Agassiz radio telescope of Harvard Observatory(see the recent article by Heeschen, in press).Several of these instruments apparently willhave electronic equipment capable of record-ing 21-cm profiles with a bandwidth of 5 kc/sec(equivalent to a resolution in radial velocity of1 km/sec). The group at the California Insti-tute of Technology is planning to use two 90-foot antennas as an interferometer with variableseparation of the two component antennas.Moreover, three very much larger paraboloidsintended in part for use in 21-cm research arenow either under construction or on the draw-ing board: the 250-foot paraboloid for JodrellBank, the 140-foot paraboloid for the Inter-University Radio Observatory now under con-sideration by the National Science Foundation,and the large instrument planned for Australia.

Spiral structure of our galaxy

The Leiden group has contributed mosteffectively to the study of the spiral structureof our galaxy. The early work by the Leidengroup culminated in the famous joint paper byvan de Hulst, Muller, and Oort (1954) which

141


contains the picture of the spiral structure ofour galaxy in its broad features as observablefrom a Northern Hemisphere radio observatory.For the Southern Hemisphere a preliminarysurvey has been completed by Christiansenand Hindman (1952), and the most recentresults were presented by Carpenter (1955)at the Troy meeting of the American Astro-nomical Society. Since the original Dutchsurvey was completed, there has been con-siderable activity in the study of the spiralstructure for the part of our galaxy betweengalactic longitudes 40° and 200°; that is, fromCygnus to well beyond Sirius. The mostextensive research has been done by Westerhout(in press), who has derived and analyzed 21-cmprofiles for the entire section between galacticlatitudes —10° and +10°. To supplement thework of Westerhout, we have the publishedand unpublished results of Matthews (1955)and of Davis at Harvard and of Heifer andTatel (1955) at the Carnegie Institution. Allthese researches indicate that the simplifieddescriptions of three primary spiral arms in thissection, the Orion Arm, the Perseus Arm, andthe Distant Arm, are only very rough first ap-proximations. The available evidence sug-gests major branching of the principal spiralfeatures; it is evident, furthermore, that all ofthe spiral features are definitely not preciselyin the same plane. Comparable detailed studiesfor the Southern Hemisphere are still lacking,but the first results by the Australian groupssuggest that here, also, complexity rather thansimplicity are to be found.

The study of the spiral structure in the partof our galaxy interior to our sun offers a varietyof complex problems. The principal source ofdifficulty is that, even on the simple assumptionof purely circular galactic rotation, a featurewith a given radial velocity may be producedby hydrogen clouds at either one or anotherdistance; or, more generally, that a particularfeature may well be the blurred combinedimpression produced by features at two differentdistances from the sun. The early resultsobtained by the Leiden group have beensummarized in a paper by Kwee, Muller, andWesterhout (1954); and, more recently, Schmidt(in press) has obtained useful new evidence. AtHarvard, Heeschen and Howard (1956) have

begun a study rather similar in scope to that ofSchmidt. It seems established that one InnerArm is present which presumably is identicalwith the optically discovered Sagittarius Arm,but further evidence exists for the presence ofsome structure in the internal neutral-hydrogenconglomerate, close to the center of our galaxy.

Two major problems arise in any analysis ofspiral features from 21-cm profiles. First, alarge group of problems relates to the probableshape of the curve that gives circular velocityof rotation as a function of distance from thegalactic center. In 21-cm research we have nodirect way of obtaining distances, and we canonly assign to a given observed feature a dis-tance based on its observed radial velocity ofapproach or recession. If our assumed formfor the curve relating distance from the galacticcenter to circular velocity of rotation is inerror, then all of our derived distances obvi-ously will be in error also. A second and relatedcause for difficulties lies in the fact that, on thebasis of 21-cm observations alone, we have noway of disentangling effects of peculiar motionsfrom those caused by purely circular galacticrotation.

We should be able to outline on a fairlyhomogeneous basis the general features of thespiral structure of our galaxy once the southernMilky Way observations are completed andanalyzed as have been the profiles for the north-ern Milky Way by the Leiden observers.Great interest attaches to the future clarifica-tion of the details of spiral structure, but hereprogress will be slow at best, and, initially atleast, somewhat uncertain. We have alreadyindicated that the dynamical study of the lawof circular galactic rotation is imperative ifwe wish to derive meaningful distances for thefeatures found through our 21-cm observations.To deal conclusively with problems of thebranching of spiral arms and with fine structurephenomena, our only hopeful approach appearsto be through coordinated optical and radiostudies of sections of the Milky Way, one byone. For proper analysis of 21-cm features,we shall need extensive and precise opticaldata—for example, for OB stars and associationsin the area of the sky covered by a particular21-cm study. For proper interpretation of the21-cm features, we should at all times attempt

• a t NEW HORIZONS IN ASTRONOMY 143

to obtain radio profiles for a fairly wide rangeof galactic latitudes and for points in the areaof investigation spaced as close together asappears to be reasonable for a particular radiotelescope. Preliminary work in this field byMatthews (1955) has already demonstratedthat the results of closely spaced surveys willbe well worth the effort, and that correlatedradio and optical studies will provide insightinto the fine details of spiral structure. Fur-thermore, a great deal of new information shouldbe obtainable regarding the fine structure ofthe interstellar medium and the associationbetween interstellar hydrogen and certainclasses of stars. A related problem of greatinterest is the relative distribution of neutralhydrogen and cosmic dust inside a spiral arm,or branch of an arm.

Astrophysical studies

The 21-cm astrophysicist is essentially aspectroscopist who has for study only onesingle spectrum line. Fortunately, the micro-wave spectrographs presently in use are reallyhigh-dispersion instruments and thus muchcan be learned from a single 21-cm profile.The 21-cm astrophysicist realizes that he isthe fortunate fellow who has first achievedaccess to the all-pervading medium of neutralhydrogen of our galactic system, which meansprobably as much as 95 percent of all theinterstellar hydrogen. In this section, we shalldescribe first some astrophysical problems thathave proved practicable for paraboloid anten-nas with apertures of 20 to 30 feet, then turnto problems whose solution is aided by antennaswith apertures of 50 to 80 feet, and concludewith some astrophysical problems that we mayattempt to study with antennas of 100-footand larger apertures.

The work of Lilley (1955) and of Heeschen(1955) at Agassiz Station illustrates the firstgroup—astrophysical problems that may bestudied with antennas of relatively smallaperture. For the Taurus Complex of darknebulosity, Lilley found generally far strongersignals for fields well inside the dark complexthan for relatively unobscured fields borderingupon the complex. It thus became evidentthat the large dust complexes have excessneutral hydrogen gas associated with them.

Lilley found the ratio between the averagedensities of gas and of dust to be close to100:1 inside the Taurus Complex. Lawrence,Menon, and I (Bok et al., 1955) used the 24-foot paraboloid to study some very denselyobscured spots inside the Ophiuchus andTaurus Complexes. We found no markeddifferences in 21-cm signal strength when wecompared intensities for the highly obscuredfields with those found for neighboring fieldsstill inside the complexes but definitely lessobscured. We deduced from our observationsthat the large dust complexes are regions ofmarkedly high concentration of interstellaratomic neutral hydrogen, but that the dis-tribution of the atomic hydrogen inside thecomplex does not mirror precisely that of thecosmic dust. Existing 21-cm observationssuggest that the neutral hydrogen is moreevenly distributed than the cosmic dust, butno really significant conclusions can be drawnuntil the really large paraboloids become avail-able for studies of this nature. A carefulsearch for neutral hydrogen concentrations by21-cm interferometer techniques should proveof great interest in connection with this prob-lem; a negative result would be fully as im-portant as a positive one. When the necessaryobservations have been made, their interpre-tation will still present a good many difficulties.Our lack of information regarding the presenceof molecular hydrogen complicates analysisconsiderably, and as yet the possible effectsof variations in temperature inside the largedust complexes are by no means understood.The assumption that the interstellar mediumof neutral atomic hydrogen possesses one singlecharacteristic temperature is at best a verycrude one, and in future research temperaturevariations undoubtedly will have to be con-sidered along with density variations.

A second area in which 21-cm studies havealready made important contributions is thatof the study—by combined optical and radiotechniques—of regions of special interest, suchas the Orion region. The recent work byMenon (in press) illustrates how much we canlearn through detailed 21-cm studies. Op-tically, the Orion region is known by itsseparate features: the Great Nebula in Orion,with a total estimated mass not in excess of


1,000 solar masses; the Trapezium Cluster andvarious OB associations and aggregates;Barnard's Luminous Arc of very great dimen-sions; and, finally, the large dust complex thatis evidently obscuring much of the region. The21-cm observations by Menon show that all ofthese features are imbedded in a large "bowl ofneutral hydrogen jelly," with a total massprobably of the order of 60 to 100 thousandsolar masses. Menon's observations suggestthat this entire mass is in slow rotation andthat it is, very likely, expanding at the rate of8 to 10 km/sec. Here the 21-cm approach hasalready added a dimension to our earlier un-related optical studies. The astrophysical andcosmogonical importance of the new approachis obvious.

The third group of problems that are provingamenable to study by 21-cm techniques dealswith the fine structure of the interstellarhydrogen medium. The studies by Lawrence(in press) and by Heeschen and Drake (1956)serve to illustrate two different approaches.Lawrence, working in close collaboration withMunch, has studied at the Agassiz Station21-cm profiles for fields centered upon stars forwhich Munch, at Mount Wilson and Palomar,had found multiple optical interstellar lines.Two established coincidences between radialvelocities of 21-cm features and strong inter-stellar absorption lines indicate that the sameinterstellar gas clouds may be responsible.Further study of these phenomena should proveexceedingly rewarding, but we can hope toarrive at a clear understanding through 21-cmstudies only with radio telescopes of largeapertures with associated electronic equipmentof very narrow bandwidth. The research ofHeeschen and Drake refers to the generaldirection of the Pleiades Cluster. They find adouble-peaked 21-cm profile—the radial veloc-ity of one peak coinciding almost precisely withthat of the Pleiades stars, and that of theother peak being very nearly equal to the meanradial velocity for the strongest components ofthe interstellar lines. In all these investigationsof the fine structure of the interstellar medium,one must insist on high angular resolution ac-companied by narrow bandwidth; withoutnarrow bandwidth, the whole study of thedistribution of individual gas-cloud velocities

becomes impracticable. Once we have per-fected our equipment, success in breaking downthe interstellar medium into its basic compo-nents should be within our grasp.

The greatest contribution so far to 21-cmresearch by the 50-foot paraboloid of the NavalResearch Laboratory has been in the study ofabsorption features observed in distant discreteradio sources. Hagen, Haddock, McClain andLilley (Hagen, et al., 1955; Lilley, 1955) havestudied 21-cm profiles and related phenomenain the continuum at wavelengths of 21 cm andless for a number of discrete sources, some ofthermal origin, others not. The first relevant21-cm profiles have exhibited certain amazingabsorption features which tend to confirm thecloud or turbulent nature of the interstellargaseous medium. For the future study ofthese profiles it will prove profitable to insist onthe highest attainable angular resolution ac-companied by very narrow bandwidth of therecording equipment. Almost every one ofthese sources studied has presented us with adifferent type of problem. The neatest ab-sorption features are probably shown by theCassiopeia source, and it seems difficult toescape the conclusion from the 21-cm observa-tions by the Naval Research group and byDavies (in press) and his associates at JodrellBank that the source is in the Perseus Arm ratherthan within 1,000 parsecs of the sun. Thestudy of the source at or near the galacticcenter—or at least in the direction between oursun and the galactic center—offers a host ofpossibilities for future research and here verymuch greater angular resolution than is attain-able to date is urgently required. We shallreturn in the next section to the Cygnus source,which exhibits absorption features that origi-nate far beyond the bounds of our own galaxy.But the brief references that we have givenabove indicate that 21-cm techniques offergreat opportunities for future research. Thecontemplated interferometer for 21-cm re-search, with 2-component 90-foot steerableparaboloids, should make possible an entirelynew approach to the study of discrete sourcesand their absorption features. Large parabo-loid antennas with apertures in excess of 100feet also promise very significant contributions.In cases where the 24-foot antenna is useless, the


50-foot antenna begins to suggest; but the 100-foot antenna should be able to show the way andthe larger antennas should really be able to solvesome of the critical galactic astrophysicalproblems of 21-cm research that are ahead of us.

Extragalactic research

To the Australian group at Sydney (Kerr,Hindman, and Robinson, 1954) and theirassociates (Kerr and de Vaucouleurs, 1956)goes the credit for having detected and studiedthe first 21-cm radiation from beyond our ownMilky Way system. Their studies of theMagellanic Clouds, which have been blendedskillfully with the optical researches carriedon at Canberra (Buscombe, Gascoigne, and deVaucouleurs, 1954), have demonstrated con-clusively that the 21-cm technique offersentirely new possibilities for the study ofgalaxies beyond our own Milky Way system.These studies have already made importantcontributions both to the dynamics of theMagellanic Clouds and to our knowledge ofpopulation distributions. Until it was foundthat the 21-cm signals from the Small and fromthe Large Magellanic Clouds are comparablein strength, many had doubted that the SmallCloud had anywhere nearly as much neutralhydrogen as the Large Cloud. Through the21-cm study we are learning a great deal aboutthe dynamics of the two Clouds. Their rota-tion seems well established, and the massesderived from rotational velocities are in goodagreement with those estimated in other ways.Studies of the distribution of random velocitiespromise to add materially to our knowledge ofthe dynamics of the gaseous interstellar mediumin these two galaxies.

The second big contribution to date to thestudy of radiation from outside our galaxywas made by Lilley and McClain (1955) intheir work on the principal absorption featurein the discrete source in Cygnus. They haveestablished that an absorption feature in thespectrum of the Cygnus source is present atprecisely the frequency predicted from theoptically known velocity of recession, producedby the red-shift. The absorption featureobviously originates in the galaxy inside whichthe Cygnus source itself is imbedded. Notonly is the work of Lilley and McClain im-

portant in establishing the presence of a 21-cmneutral hydrogen medium surrounding theCygnus discrete source, but the agreementbetween the optically predicted red-shift veloc-ity and the radial velocity determined by radiotechniques has important consequences forcosmogonical theory.

The 21-cm radiation from the AndromedaNebula and other nearby galaxies remains asyet undiscovered. Special equipment to aidin its detection and measurement is now underconstruction in Holland and elsewhere. Theimportance of 21-cm studies for individualgalaxies of the Local Group can hardly be over-rated. If we can attain angular resolutionsufficient to separate areas inside spiral armsfrom those between the arms, very significantcontributions can be made to the solution ofthe whole group of problems related to thedynamics of galactic rotation and of spiralstructure. We shall then have also whollynew evidence on the distribution of neutralatomic hydrogen in spiral and other galaxiesand new light will be thrown on the problemsof stellar birth and evolution. For the future,21-cm research of nearby galaxies, notablythose with spiral structure, has a high priority.

Quite recently Heeschen (in press) has suc-ceeded in detecting 21-cm radiation associatedwith a cluster of galaxies. At a frequency near1387 mc/sec, corresponding to a recessional radialvelocity of close to 7000 km/sec, he found evi-dence for a weak but consistently present signal(&T=2°K±0.°2) from the direction of theComa cluster of galaxies. The specificallydetermined radial velocity of recession for thesame cluster is 6680 km/sec, which appears toindicate satisfactory agreement. The disper-sion in radial velocity derived from the 21-cmobservations appears to be of the order of 500km/sec, somewhat smaller than the optical valueof about 800 km/sec. The total estimated massof neutral atomic hydrogen is a little less than10" solar masses. Heeschen's result was ob-tained with the 24-foot radio telescope atAgassiz Station. Information regarding thedistribution of neutral hydrogen inside thecluster can be obtained only with instrumentsof greater aperture, which afford higher angularresolution.

A potentially very profitable approach to the


problems of intergalactic hydrogen is throughthe study of absorption features. Field andHeeschen have recently attempted to observein the spectrum of the Cygnus source absorp-tion features produced by the general mediumof intergalactic hydrogen between the Cygnussource and our sun. The results with presentequipment are inconclusive, but the analysisof the available data shows that very rarefiedabsorption features might be detectable, evenfor the low densities that presumably prevail inthe intergalactic medium.

We should at this point refer again to theinterferometer array now under constructionby Bolton and his associates at the CaliforniaInstitute of Technology. A variable-spacinginterferometer, with two basic componentsof 90-foot aperture each, should make acces-sible for study a number of discrete sourcesrunning literally into the hundreds. Fromthe absorption features, much informationshould be obtainable—information relating notonly to the clouds between our sun and thesediscrete sources but also concerning the dis-tances of individual discrete sources.

Concluding remarks

I t is now just about 5 years since Purcell andEwen first succeeded in detecting the 21-cmline of neutral hydrogen. In the interveningyears much progress has been made, and Idoubt if many would have dared to predict in1951 that by 1956 equipment equaling in excel-lence present-day radio telescopes and electronicrecording equipment would have become avail-able at several institutions. But although wemay look with some satisfaction upon ouraccomplishments to date, in reality 21-cmresearch has only just begun.

I have purposely confined my remarks to21-cm research in a most limited sense. I havestudiously avoided referring to the importanceof studies in the continuum at wavelengths of50 cm and less, which must become an integralpart of all 21-cm studies. I have also omittedall reference to other spectrum lines in themicrowave range, notably to the line of deu-terium, the discovery of which would open upan entirely new field for research of the inter-stellar medium by radio techniques. I t wouldbe very helpful if, instead of having to confine

ourselves to the single 21-cm line, we could haveavailable additional lines, preferably those thatmight give clues to the presence of atoms otherthan hydrogen, and components of bands thatwould permit the study of molecules, such asOH and H2. The field of galactic and extra-galactic microwave spectroscopy is indeed afruitful one for future exploration.

ReferencesBOK, B. J., LAWRENCE, R. S., AND MENON, T. K.

1955. Publ. Astron. Soc. Pacific, vol. 67, p. 108.BUSCOMBB, W., GASCOIGNE, S. C. B., AND DE VAU-

COULEURS, G.1954. Australian Journ. Sci., vol. 17, p. 3.

CARPENTER, M. S.1955. Paper presented at Troy meeting of the

Amer. Astron. Soc.CHRISTIANSEN, W. N., AND HINDMAN, J. V.

1952. Australian Journ. Sci. Res., ser. A, vol. 5,p. 437.

DA VIES, R. D.. Paper presented at Jodrell Bank Sympo-

sium 1955, in press.EWEN, H. I., AND PTTRCELL, E. M.

1951. Nature, vol. 168, p. 356.HAGEN, J. P., LILLET, A. E., AND MCCLAIN, E. F.

1955. Astrophys. Journ., vol. 122, p. 361.HEESCHEN, D. S.

1955. Astrophys. Journ., vol. 121, p. 569.. Astrophys. Journ., in press.

HEESCHEN, D. S., AND DRAKE, F. D.1956. Astron. Journ., vol. 61, p. 5.

HEESCHEN, D. S., AND HOWARD, W. E., III.1956. Astron. Journ., vol. 61, p. 6.

HELFER, H. L., AND TATEL, H. E.1955. Astrophys. Journ., vol. 121, p. 585.

KERR, F. J., AND DE VAUCOULEURS, G.1956. Australian Journ. Phys., vol. 9, p. 90.

KERR, F. J., HINDMAN, J. V., AND ROBINSON, B. J.1954. Australian Journ. Phys., vol. 7, p. 297.

KWEE, K. K., MULLER, C. A., AND W E S T E R H O U T , G.1954. Bull. Astron. Iiist. Netherlands, vol. 12,

No. 458, p. 211.LAWRENCE, R. S.

. Paper presented at Jodrell Bank Sympo-sium 1955, in press.

LILLET, A. E.1955. Astrophys. Journ., vol. 121, p. 559.

LILLET, A. E., AND MCCLAIN, E. F.1955. U. S. Naval Res. Lab. Rep. No. 4689.

MATTHEWS, T. A.1955. Publ. Astron. Soc. Pacific, vol. 67, p. 112.

MCCLAIN, E. F.1955. Astrophys. Journ., vol. 122, p. 376.

MENON, T. K.. Paper presented at Jodrell Bank Sympo-

sium 1955, in press.

NO.I NEW HORIZONS IN ASTRONOMY 147

MlJLLER, C. A., AND OoRT, J. H . VAN DB HULST, H. C.1951. Nature, vol. 168, p. 357. 1945. Nederlands Tijdschr. Nat., vol. 11, p. 210.

PAWSBY, J. L. V A N D E HULST, H. C, MUUJSR, C. A., AND OOHT, J. H.1951. Nature, vol. 168, p. 358. 1Q54 B u l l A s t r o n I n 8 t Netherlands, vol. 12,

SCHMIDT, M^ N O y

. Paper presented at Jodrell Bank Sympo-sium 1955, in press. WESTEBHOUT, G.

SHKLOVSKT, I. S. • Paper presented at Jodrell Bank Sympo-1949. Astron. Zhurn., vol. 26, p. 10. sium 1955, in press.

Some Observational Problems inGalactic Structure

By Harold Weaver1

The past decade has brought many new in-sights into the problems of galactic structure.New observational techniques—optical andradio—have produced results that were thoughtfar beyond reach only a few years ago. Wenow realize the importance of the variousstructural units of the galaxy (clusters, associ-ations, arms, etc.) as groups of evolving starsof different ages. Data on the structure andthe kinematic properties of the galaxy providebasic information for studies of stellar evolu-tion; clusters and associations provide tests ofstellar-model evolutionary calculations. I t isclear that we must relate our galactic studiesto those of other nearby resolved galaxies.Knowledge will advance most rapidly if weunderstand how the structural details of ourgalaxy relate to those of the sequence of galaxies.

New approaches

Several recent observations, to a considerabledegree, have reoriented the whole trend ofthinking about galactic structure and haveforcefully indicated the need for new approachesto the problems. Among these observationsare the following:

1. Baade's (1944) decisive demonstrationthat the space and physical-specification vari-ables in the space-density function D(M, S;R, 9, z) are strongly correlated.

The stellar space-density function D(M, S;R, 0, z)dMdSdV will play an important partin the discussion that follows; it indicatessymbolically the space density of stars in ab-solute magnitude range M to M-\-dM, spectral-type interval S to S-\-dS, and volume elementdV, centered at galactocentric position R, 0, z.Other or additional variables can be used for

1 Leuschner Observatory, University of California, Berkeley, Calif.

physical specification; color is frequently sub-stituted for spectral type. If we fix the spacevariables and represent the distribution graph-ically we obtain the Hess (1924) diagram.

The space distribution of nonstellar materialmay be represented by Dt(R, 0, z), which refersto particles of kind i.

For R small or z large in D(M, S; R, 6, z) onefinds pure or essentially pure Population II.If to M, S are assigned ranges encompassingintrinsically bright O and early B stars, Baadefound that in extragalactic systems the locus ofmaximum of D(M, S; R, 0, z) traces out spiralarms where Population I stars and much inter-stellar dust is to be found. Population I,occurring in regions where z is small and R notsmall, appears as an addition to, and inter-mingled with, the general substratum of Popu-lation II stars forming the disc of the system.

2. Morgan, Sharpless, and Osterbrock's(1952a, 1952b) use of Baade's spiral tracertechnique to produce the first picture of spiralstructure in the galaxy.

3. Oort, van de Hulst, and Muller's (1952)observation that the 21-cm radiation of hy-drogen indicates hydrogen concentration inspiral arms whose form and position confirmand extend the work of Morgan and hisassociates.

4. Ambarzumian's (see Kourganoff, 1951)recognition of stellar associations and of theirimportance in galactic and stellar evolutionproblems.

In the following discussion we shall con-centrate attention on the problems that fallprincipally under 1 and 2; those that fallprincipally under 3 and 4 will form the subjectmatter of other papers. Effort will be madeto avoid duplicating material included in

149381014—56 -11


Blaauw's (1955) excellent summary of theInternational Astronomical Union symposium,"Coordination in Galactic Research," whichshould be read by all investigators in this field.

In this discussion we shall critically examinethe results of some recent researches in galacticstructure; and, rather than stress the agree-ments, we shall stress the discrepancies betweenobservation and theory and between theobservations themselves. While this approachwill accomplish our purpose—to expose areaswhere problems need to be worked on—itwill show the field of galactic research in anuntrue light. The reader must guard againstgaining the false impression that little hasbeen accomplished in the field. The fact is,rather, that much remains to be done.

Form of the armsThe form of the spiral arms first indicated byMorgan and his associates, using emissionnebulosities, and by Oort, using hydrogenconcentrations, becomes increasingly complexand less easily understood as additional observ-ational data accumulate. Further work, bothobservational and theoretical, in detailed de-lineation of the spiral arms by stars and bygas is required. The Morgan-Meinel-Johnson(1954) technique of very short spectra pro-vides a powerful method of discovering distantspiral tracers. Southern Hemisphere observa-tions of the 21-cm radiation are especiallyneeded; 21-cm observations of high resolvingpower in angle and in frequency are needed inboth hemispheres. Such observations will,among other things, provide data for determin-ing the forms of the gas condensations, and thedegree of association between OB stars and gas.

The distance scale within the galaxyThe distance scale presently employed in tracingspiral structure by means of OB stars is correctonly to the extent that (1) absolute magnitudesand colors of the MK classification system arecorrect, and (2) the ratio of total interstellarextinction to color excess is known.

The distances of the hydrogen condensationsare correct only to the extent that (1) anaccurate differential-rotation function Aw(R) isavailable, and (2) the interstellar gas is freefrom large-scale peculiar motions.

Stellar distance scale.—The present scale ofabsolute magnitudes (Keenan and Morgan,1951) is not entirely independent of galacticrotation constants (first order theory only).These, in turn, are dependent upon an olderabsolute magnitude scale and on correctionsfor average interstellar extinction. Work of afundamental nature is here necessary to elim-inate the possibility of a closed circuit ofsystematic errors that do not easily revealthemselves. Fundamental absolute-magnitudescales are required—scales derived from stellarassociations, clusters, moving clusters, andstars of large parallax (with the latter twotypes of objects serving for establishment ofzero point). The degree of agreement betweenmain sequence magnitudes (and also luminosityclass IV, III, . . ., magnitudes) derived fromdifferent associations, clusters, and so forth,will be of extreme interest; comparisons withfield stars in the solar neighborhood will like-wise be of importance. Several investigatorsare engaged in observations of this type andare producing very interesting results.

Observations of the ratio of total interstellarextinction to color excess are contradictory.On the basis of both photographic and photo-electric studies (Baade and Minkowski, 1937;Stebbins and Whitford, 1945; Sharpless, 1952),the Orion Nebula has been pointed out aspossessing an extinction law appreciably differ-ent from that of other interstellar material.A photographic spectrophotometric study ofO and B stars by Divan (1954) indicatesprecisely the same extinction law in the Orionregion as in all other regions investigated.On the other hand, very accurate photoelectricobservations by Johnson and Morgan (1955)indicate an "abnormal" ratio of extinction tocolor excess for the Cygnus region in the sensethat relatively more ultraviolet extinction ispresent in that region than in the Cepheus-Monoceros region. Space variations of theextinction law complicate photometric dis-tance calculations and pose difficult problemsfor the theoretician attempting to explain thedistribution of particle size in the interstellarmaterial. The observational determinationand subsequent theoretical discussion of thewavelength dependence of interstellar extinc-tion requires further attention.


Hydrogen distance scale.—The differentialrotation function Au(R) for R^>R0 must bederived from stellar motions or from the massdistribution in a model galaxy; circular motionis assumed in either case. If derived fromstars, Aco(i?) is dependent upon the accuracyof the stellar distance scale. For R<C.R0 thehydrogen observations themselves permit, inprinciple, the establishment of Aw(R).

A 21-cm observation in the direction of theinner part of the galaxy yields a »max which, ina galaxy in pure rotation and with a monotonicw(R) function, arises from the point that, onthe given line of sight, is nearest the galacticcenter. Precisely,

(1)lia Sin A

In practice, there may be observationaldifficulties in the measurement of the »„,„described in equation (1) because of localpeculiar motions of the gas, random motionsor cosmic dispersion within the gas, andvariations of gas density along the line ofsight.

For R—Ro small, and with the definitionA=— )£ Rou'o, equation (1) may be approxi-mated by the expression,

Ro=-2A(1 —sin X) sin X (2)

On the basis of linear theory one can findRo, provided vmn has been measured and theOort A parameter is known.

Van de Hulst, Muller, and Oort (1954) usedequation (2) to compute Ro with an adoptedvalue of .4=19.3 km/sec per kiloparsec, whichis a mean of various A values derived on lineartheory. It is likely that A is overestimated.(Weaver, 1955a, 1955b, 1955c.) From four»max values they found i?0=8.26 kpc; they em-ployed the value i?0=8.2 in further calculationsof Aw(i?) from equation (1). The scale of

is fixed by Ro; systematically, theor <a(R) curve found from the »„„«

observations is, then, no more precise than theRo or, if we look back one step in the reasoning,than the A value and the linear theory onwhich it was based.

Van de Hulst, Muller, and Oort checked Ro

against Baade's (1954) value for the same881014—58 12

parameter, 8.16 kpc, derived from the distribu-tion of RK Lyrae variables in the region of thegalactic nucleus. The excellent agreement ofthe two values does not necessarily confirmtheir correctness. Baade's estimate could beappreciably changed by a small revision of theinterstellar extinction correction employed. Achange of 0.25 mag in the extinction would alterRo by a kiloparsec.

The crucial need of an accurate Ro is ap-parent; further investigation is imperative.

The uncertainty in the galactic distance scaleis also indicated by inconsistency in the cur-rently quoted values of Ro, «o, and Fo. Van deHulst, Muller, and Oort (1954) found «o=26.4±2.5 (m. e.) km/sec per kiloparsec, a valuederived by differencing A and B. Weaver andMorgan (1956) employed proper motions of 79Cepheids in mathematically unbiased equationsto find «o=23.2 ±5.0 (m. e.) km/sec perkiloparsec. These values combined with Ro=8.2 kpc yield, respectively, Vo=216 km/secand 190 km/sec. These are not fundamentaldeterminations since they represent the productof two independently derived quantities.

Mayall (1946) found Fo=200 km/sec fromradial velocities of 50 globular clusters. Thisis a fundamental determination, though itprobably does not represent the true value ofVo, being smaller than that value by thedifference between the circular velocity of thegalaxy at Ro and the corresponding value for thesystem of globular clusters. A direct observa-tional determination of this difference does notexist. Mayall (1946) attempted to evaluateit by determining the solar motion with respectto the extragalactic nebulae. From severalsolutions he concluded Vo=300 km/sec. Huma-son and Wahlquist (1955) have derived anessentially identical value. Any peculiar mo-tion of the galaxy is included in this result,which therefore does not permit a strong con-clusion that Vo is as great as 300 km/sec, butstrongly indicates that F0>200 km/sec. Fricke's(1949a, 1949b) value of Vo from high velocitystars, also a fundamental determination, like-wise indicates F0>200 km/sec. Drawing ananalogy between our galaxy and the AndromedaNebula, we reach a similar conclusion fromMayall's (1951) velocity measurements inAndromeda. An inconsistency thus appears.


The product J?o«o is smaller than the Vo valuederived by independent fundamental methods.Examination of Ro and w0 is required. Of thetwo quantities, Ro is, percentagewise, the moreuncertain.

The u(R) curve derived by Kwee, Muller, andWesterhout (1954) from 21-cm observationsshows waves which, they suggest, representdepartures from circular motion which arecorrelated with position with respect to spiralstructure. The general question of systematicdeviations from circular motion over sizable

GALACTICCENTER

gen. For the stars, accurate photometric dis-tances and radial velocities can be derived forobjects nearer than r0, an observationally setlimit. Over the region of space thus defined(shown in fig. l,a) the motion pattern can, atleast in principle, be derived and regional de-partures from circular motion deduced. Hy-drogen observations provide no distance scaleindependent of velocity; they permit study ofdepartures from circularity only over the vol-ume of the galaxy in which tWx occurs (shownin fig. 1,6) in the range Ri<R<Ro. The

\ CENTER /

FIGURE 1.—The plane of the paper represents the plane of the galaxy, a, Crosshatched area indicates the region over which starspermit the study of departures from circular motion, b, Crosshatched area indicates the region over which 21-cm radiationof hydrogen permits study of departures from circular motion.

volumes of the galaxy merits investigation on abroad basis. Observations of both hydrogenand stars are desirable. Of particular interestwill be tests for similarity of kinetic propertiesof stars and gas in the same volume of space andfor a correlation between gas and stellar motionsand spiral structure as suggested by Kwee,Muller, and Westerhout.

Study of departures from circular motion islikely to prove easier with stars than with hydro-

galactocentric distance Rx is fixed at the pointat which strong turbulence in the interstellarhydrogen sets in. In the case of hydrogen, alarge peculiar motion occurring in the line ofsight, but not within the indicated volume,could obscure a smaller peculiar velocity withinthe indicated volume and thus confuse the ob-servation. To a lesser degree, variations in thedistribution of random velocities in the hydro-gen also obscure the situation. Such difficulties


do not exist for the stellar observations. Hydro-gen observations, on the other hand, extend toa much greater distance than do stellar ones.Comparison of flmax values derived at corre-sponding angles on either side of the galacticcenter will be of very great interest.

Association of spiral tracers and hydrogen

No adequate test has yet been made of the com-monly accepted hypothesis that arms or struc-tures delineated by very early type stars arecoincident in space with hydrogen gas densitymaxima. There are indications that these maynot always be coincident (Weaver, 1953; van deHulst, Muller, and Oort, 1954). OB stars mayappear where no hydrogen density maximumoccurs.

Several tests of this hypothesis can be pro-posed. The simplest involves investigation ofthe simultaneous occurrence of hydrogen max-ima and concentrations of early type stars overthe surface of the sky. This test was inde-pendently proposed by W. W. Morgan duringa colloquium given at Berkeley, Calif., in Novem-ber 1955. An alternative test involves theagreement of velocity maxima of the radialvelocity distributions of the OB stars (treatedas particles in a star gas) and of the hydrogengas, properly weighted to match the distribu-tion of the stars in longitude. A third testinvolves the determination of the closeness ofassociation in space of maxima in the spacedensity functions of OB stars and of interstellarhydrogen as found from the 21-cm observa-tions. Distances of the stars are to be derivedphotometrically; distances of the hydrogenmaxima are to be determined from a rotationfunction based on the same photometric dis-tance scale used for the stars. Differentassumptions are involved in the three cases;all three tests should eventually be made.

Composition of the arms

Understanding of the physics of the spiralarms will require, among other data, knowledgeof their composition. Nearby extragalactic sys-tems will undoubtedly play an important rolein composition studies. Photographs of M 51published by Zwicky (1955) are instructive;analogous photoelectric measurements would beinvaluable. Isophotes and isocolor contours,

even of a rudimentary kind, for details in a fewproperly chosen extragalactic systems would beof great interest. A start on this problemcould be made by measuring magnitude andcolor differences between arm and interarmregions of some of the most resolved extra-galactic systems.

Spiral tracers are highly concentrated in thearms; data on the space distribution of later-type Population I stars, B 3, B 5, . . . , in andnear the arms is lacking. We have no knowl-edge of how prominent the arms would appearif the galaxy were viewed through a series ofhypothetical filters, each of which would permitus to see stars in only a small spectral-typerange, B 3, B 5, . . . . The ordinary method ofanalysis of star counts does not offer muchpromise of producing such a picture of thegalaxy. Rather, we should obtain the picturewith the aid of accurate individual photometricdistances derived for each member of a largesample population of stars. An empirical modelof the space distribution of any one particulartype of star could then be constructed. Datasuch as density gradient perpendicular to thegalactic plane, in the galactic plane, and per-pendicular to the arm, and so forth, could beinferred from the model for each stellar type.

An approach through such an empiricalmodel avoids the concepts of a continuousspace-density law and of a space-invariantluminosity function. I t does not involve thesolution of an integral equation for the space-density law or the correction of the apparentspace-density law with the aid of an averageinterstellar extinction relation in a specifieddirection. Each star is treated individually;maximum resolution is achieved. A limit ofcompleteness in absolute magnitude M(r) canbe specified at each distance; the space densityof intrinsically faint stars is not inferred froma few intrinsically bright ones as when theintegral equation is solved.

Extensive observations are required for sucha program; a cooperative venture involvingseveral observers and telescopes might proveadvantageous. Stromgren's (1951, 1954) ac-curate photoelectric method of spectral classifi-cation should play an important part in anysuch plans. Stromgren's results indicate thatthe photometric probable error of one observa-


tion of a predicted absolute visual magnitudeis less than 0.2 mag; in distance estimate aprobable error of approximately 15 percentmight be expected . With the 82-inchMcDonald telescope, observations could be ex-tended to apparent magnitude 14. Furtherexploratory work in the use of this importantobserving technique is called for. The resultsobtained should provide data for the establish-ment of an empirical space distribution as atest.

The need of an accurate and fundamentalcalibration of the absolute magnitudes andcolors of the MK spectral types and luminosityclasses as well as an accurate value of the ratioof total extinction to color excess is reemphasizedby a program of this type.

Variations in the arms

Within the spiral arms are found variations inthe space density of stars and in the stellarmixture. Observational data leading to abetter understanding of such variations wouldbe of considerable value.

A pronounced example of the first kind ofvariation is found in the longitude range 110°to 145° adjoining the rich Perseus star cloud.In this range of longitude there exists a deepminimum in the number of OB stars (Weaver,1953); there is essentially a discontinuity inthe Perseus arm as delineated by early-typestars.

Variations in the content of the arms fre-quently appear. The Perseus cloud contains Ostars and late-type supergiants (Bidelman, 1943,1947); classical Cepheids are found within it.The Orion association contains principally mainsequence stars of type O 6 and later; it containsno supergiants of type later than B (Sharpless,1952, 1954). The Scutum association, whichhas M 11 as its nucleus, contains main sequencestars of type B 9 and later, and ordinary giants(Russell, 1953). Gurzadian (1949) has pointedout many cases of this sort.

A preliminary study of variations in thearms might be made on the basis of galacticclusters and Trumpler's cluster types.

Luminosity functions and related quantities

An observing program of the type outlined inthe preceding sections would provide detailed

information on the space variation of D(M,S) R, 6, z) over the M-range and r-rangeaccessible to observation. However, such aprogram now represents only a possible futuredevelopment; information on stellar popula-tions which will provide data for stellar evolu-tion studies must be provided more rapidly.

Among observable quantities that providesuch information with less completeness thanD(M, S; R, 6, z) but that are much moreeasily obtained are the following:

1. The luminosity function, L(M; R, 6, z),which is the integral of D with respect to S.

2. Integrated magnitudes, colors, or spectra,which represent, essentially, integrals of theluminosity function over specific wavelengthranges and which refer to a defined volumeelement of the galaxy or of some extragalacticsystem.

3. The ratio of number of stars brighter thansome fixed M and within a defined volumeelement to the total luminosity of all starswithin the same volume element (Baum andSchwarzschild, 1955). The wavelength rangesin which M and the total luminosity are meas-ured must be specified.

4. The ratio of mass to luminosity (measuredin a specified wavelength range) of the materialwithin a defined volume element of the galaxyor of an extragalactic system.

5. The ratio of luminosities of stars con-tained within some defined volume clement ofthe galaxy or an extragalactic system measuredin several different specified wavelength rangesto the luminosity of Uio same stars in somestandard wavelength range.

Basically, the integrated quantities representcertain average properties of the stellar popula-tion in the defined volume element. The ratioscompare one portion of the population withanother portion of the same population. Theparticular ratio taken fixes the portions of thepopulations compared. All of these observablequantities need further investigation and appli-cation within our own galaxy and extragalacticsystems.

The luminosity function of stars characteriz-ing the solar neighborhood has been determinedby van Rhijn (1936), by Luyten (1939, 1941),and others. Van Rhijn's function shows aminimum at absolute magnitude -f 8; Luyten's


does not. The reality of this feature requireschecking. The test might be made with theaid of late-type dwarfs discovered spectro-scopically (Vyssotsky, 1943, and coworkers,1946, 1952). With the present functioningof Schmidt telescopes, intrinsically faint starsdiscovered spectroscopically should play anincreasingly important role in the determina-tion of L(M; Ro, 0, 0).

The van Rhijn luminosity function pre-sumably characterizes Population I; the ex-actness of the characterization may bearfurther analysis. The method of derivingthe luminosity function entails the use ofvolume elements of different sizes for differentluminosity ranges. The very bright stars areobserved over a volume element of great size,a cylinder of possibly a kiloparsec radius; thevery faint stars are observed within a sphericalvolume element of a few parsecs radius centeredon the sun. The highly luminous stars areknown to be nonrandomly distributed in thegalactic plane; indications are (McCuskey,1951) that the fainter stars also vary in mixtureover moderate distances from the sun. Thevan Rhijn luminosity function was derived whenthere was still belief in the universality of thefunction. There is no certainty that it refersto a homogeneous population of stars. Thefunction is, for the faint stars, essentiallyL(M; R, 0, 0); for the bright stars it is someaverage L(M), the average being taken overa large and somewhat uncertainly specifiedvolume element.

An important step towards a clearer under-standing of the Population I luminosity func-tion could be taken by determining the lumi-nosity functions of a number of representativePopulation I objects; namely, galactic clusters,star clouds, associations. Such luminosityfunctions could be determined with adequateaccuracy by statistical means, provided anaccurate magnitude sequence was available foreach case. Of particular interest would be astudy of variations in the form of the luminosityfunctions for stellar groups of different Trump-ler type or different age, different size, and soforth. The possibility of forming from sucha sample a mean function representative ofPopulation I in the solar vicinity should beinvestigated. Variations in different groups

or changes through evolution may, of course,make the establishment of such a mean functionimpossible on the basis of present knowledge.

Luminosity functions of typical PopulationII would also be of great interest. Importantresults for luminosity functions of globularclusters have recently been given by Sandage(1954) and by Tayler (1955); for nebulae byBaum and Schwarzschild (1955), by Baum(1955), and by Roberts (1955, and in press).For these latter cases indirect methods usingintegrated magnitudes and multicolor observa-tions were employed.

Variations have been found in luminosityfunctions of typical Population II systems justas they have in Population I systems. Anelliptical nebula appears to contain a muchlarger percentage of dwarfs than a globularcluster. The luminosity function of the dwarfbranch of a globular cluster appears remarkablysimilar to that of the stars in the solar neigh-borhood; a typical galactic cluster contains amuch smaller percentage of dwarfs than doesthe solar neighborhood. Comparisons of thissort based upon many more cases than are nowavailable will be of very great interest.They may well provide the key to an understand-ing of the mixture of stars we find in the neigh-borhood of the sun, particularly if informationof the space density of stars of different spectraltypes and luminosity classes in and betweenthe arms can be derived.

The local systemInterpretation of space density results in theregion of the sun in terms of a simple pictureof spiral arms may prove difficult because ofthe interference of the "local system" (see Bok,1937, for earlier references). The observationsby Nassau and Morgan (1951) demonstratethat the distribution of bright OB stars followsthe outline of the local system. Heeschen andLilley (1954) have found the local system inthe 21-cm radiation of hydrogen. The struc-ture likewise appears among both the bright anddark nebulae. A thorough modern study of thelocal system would be of interest; it would alsobe of value as an aid in disentangling thelocal system and what one ordinarily thinks ofas spiral structure.


Other aspects of the galaxyFor an understanding of the galaxy a knowl-edge of the interstellar material is no lessimportant than that of the stars. Problemsof some aspects of interstellar matter have beendescribed where necessary: the degree of as-sociation of interstellar hydrogen and spiraltracers, motions of the hydrogen, and the valueof the ratio of interstellar extinction to red-dening. However, general problems of theinterstellar medium and of the 21-cm radiationof hydrogen fall outside the scope of thisreview. They will be found in the reviews ofthose subjects. Likewise, problems of motionswithin the galaxy have been introduced onlywhere necessary since they form the subjectmatter of a separate paper. In actual practice,the study of the distribution of stellar orinterstellar material and the study of kinematicscannot be separated if a complete view of theproblem is to be gained. The distribution ofmaterial and the motions of the material arecomponent parts of one problem that must beconsidered as a whole.

ReferencesBAADE, W.

1944. Astrophys. Journ., vol. 100, p. 137.1954. Quoted in Bull. Astron. Inst. Netherlands,

vol. 12, p. 117; an earlier value will befound in Pub. Obs. Univ. Michigan, vol.10, p. 7, 1951.

BAADE, W., AND MINKOWBKI, R.1937. Astron. Journ., vol. 86, pp. 119, 123.

BAUM, W. A.1955. Astron. Journ., vol. 60, p. 153.

BAUM, W. A., AND SCHWABZSCHILD, M.1955. Astrophys. Journ., vol. 60, p. 247.

BlDELMAN, W. P.1943. Astrophys. Journ., vol. 98, p. 61.1947. Astrophys. Journ., vol. 105, p. 492.

BLAAUW, A.

1955. Int. Astron. Union Symposium, No. 1.BOK, B. J.

1937. The distribution of the stars in space. As-trophysical Monographs, p. 97. Uni-versity of Chicago Press.

DIVAN, L.

1954. Thesis, Univ. Paris, 1954, and Li6ge, Colloq.Inter. d'Astrophysique, 1954, p. 538.

FHICKE, W.

1949a. Astron. Nachr., vol. 277, p. 241.1949b. Astron. Nachr., vol. 278, p. 49.

GURZADIAN, G. A.1949. Astron. Journ. Acad. Sci. U.S.S.R., vol. 26,

p. 329 (for abstract see Astron. NewsLetter No. 54, p. 1, 1951).

HEESCHEN, D. S., AND LILLET, A. E.1954. Proc. Nat. Acad. Sci., vol. 40, p. 1095.

HEES, R.1924. Seeliger Festschrift, p. 265. Springer,

Berlin.HUMASON, M. L., AND WAHLQUIST, H. D.

1955. Astron. Journ., vol. 60, p. 254.JOHNSON, H. L., AND MORGAN, W. W.

1955. Astrophys. Journ., vol. 122, p. 142.KEENAN, P. C , AND MORGAN, W. W.

1951. In Hynek, ed., Astrophysics, p. 12. Mc-Graw-Hill, New York.

KOTJRGANOFF, V .1951. Quelques documents sur la structure de la

galaxy. (Inst. d'Astrophys., Paris, June,1951. Trans, by C. Payne-Gaposchkinin Astron. News Lett., No. 64, 1952.)

KWEE, K. K., MULLER, C. A., AND W E S T E B H O U T , G.1954. Bull. Astron. Inst. Netherlands, vol. 12,

p. 211.LUTTBN, W. J.

1939. Publ. Astron. Obs. Univ. Minnesota, vol. 2,No. 7.

1941. Ann. New York Acad. Sci., vol. 42, p. 201.MATALL, N. U.

1946. Astrophys. Journ., vol. 104, p. 290.1951. Publ. Obs. Univ. Michigan, vol. 10, p. 19.

MCCUSKBT, S. W.1951. Publ. Obs. Univ. Michigan, vol. 10, p. 67.

MORGAN, W. W., MEINEL, A. B., AND JOHNSON, H. M.1954. Astrophys. Journ., vol. 120, p. 506.

MORGAN, W. W., SHARPLESS, S. L., AND OSTERBROCK,D. E.

1952a. Astron. Journ., vol. 57, p. 3.1952b. Sky and Telescope, vol. 11, p. 134.

NASSAU, J. J., AND MORGAN, W. W.1951. Pub. Obs. Univ. Michigan, vol. 10, p. 43.

OORT, J. H., VAN DE HULST, H. C , AND MuLLEB, C. A.1952. Trans. Int. Astron. Union, vol. 8, p. 512.

ROBERTS, M. S.1955. Publ. Astron. Soc. Pacific, vol. 67, p. 327.

. Astron. Journ., in press.RUSSELL, J. A.

1953. Astron. Journ., vol. 58, p. 89.SANDAGE, A. R.

1954. Astron. Journ., vol. 59, p. 162.SHARPLESS, S. L.

1952. Astrophys. Journ., vol. 116, p. 251.1954. Astrophys. Journ., vol. 119, p. 200.

STEBBINS, J., AND WHITFORD, A. E.1945. Astrophys. Journ., vol. 102, p. 318.

STROMGREN, B.1951. Astron. Journ., vol. 56, p. 142.1954. Astron. Journ., vol. 59, p. 193.


TATLEB, R. J.1955. Astron. Journ., vol. 59, p. 413.

VAN DB HULST, H . C , MuLLEH, C. A., AND OOBT, J. H .1954. Bull. Astron. Inst. Netherlands, vol. 12,

p. 117.VAN RHIJN, P. J.

1936. Publ. Astron. Lab. Groningen, No. 47.VTSSOTSKT, A. N.

1943. Astrophys. Journ., vol. 97, p. 381.VTSSOTSKT, A. N., JANSSEN, E. M., BT AL.


VTSSOTSKT, A. N., AND MATEER, B. A.1952. Astrophys. Journ., vol. 116, p. 117.

WEAVER, H. F.1953. Astron. Journ., vol. 58, p. 177.1955a. Astron. Journ., vol. 60, p. 202.1955b. Astron. Journ., vol. 60, p. 208.1955c. Astron. Journ., vol. 60, p. 210.

WEAVER, H. F., AND MORGAN, H. R.1956. Astron. Journ., vol. 61, p. 268.

ZWICKT, F.1955. Publ. Astron. Soc. Pacific, vol. 67, p. 232.

Galactic Clusters and AssociationsBy A. Blaauw l

Research on galactic clusters and associationscan be divided into two categories: the proper-ties of the individual objects, and the propertiesof the system of the objects as a whole. Thefirst aim of these investigations is, of course, toarrive at a knowledge of the present conditionof the individual clusters and associations andof their present motions and distribution inspace. But the ultimate purpose is to under-stand these conditions as a result of the originand evolution of the clusters and associations.This problem is immediately connected withthat of the history of the galaxy and with thatof stellar evolution. This double aspect makesthe study of clusters and associations partic-ularly significant.

There is no sharp division between galacticclusters and associations. This is best illus-trated by the fact that certain objects, such asNGC 2244, may be classified under either ofthese categories. Most associations must haveshort lifetimes and will have dispersed amongthe general population of the galaxy within oneor two revolutions of the galaxy (a few hundredmillion years). Galactic clusters are morestable objects but this need not mean that theirorigin is essentially different from that of theassociations. They may simply be remnantsof the rich central parts of some associations withvery small internal motions, that had a negativeenergy from the outset, or have disposed of asmall positive energy by the ejection of partof their stellar content. Stellar associationsoffer an opportunity to study the properties ofthe very youngest stars, and they seem to beone of the best places to look for informationon the process of star formation itself. Galacticclusters offer a variety of star samples, each ofwhich informs us about the properties of starsformed simultaneously and at the same loca-tion in the interstellar medium, under certain

»Yerk«s Observatory, Williams Bay, Wls.

common circumstances. A vast amount ofobservational data has been collected in thepast for the galactic clusters, most of which willserve as a basis of more refined work madepossible by the improved present and futureobservational techniques. Much less has beendone on the associations, whose importancehas only recently become clear. However,what has been done observationally so far,especially in the field of photometry, meets thestandards required for our present demands.

Having in mind that research on galacticclusters and associations can contribute to ourunderstanding of the evolution of the galaxyand of the stars, we shall consider three majortrends in this research, in the three sectionsfollowing.

Individual properties of the stars

Individual properties of the stars in clustersand associations may be: the absolute magni-tude, the color, the spectral type, the angularmomentum, the duplicity or multiplicity, andthe motion of the star with respect to the centerof the cluster or association. Such studies areimportant because the stars in the cluster orassociation can be assumed to be of approxi-mately equal ages, to have formed at the samelocation in the interstellar medium, and,perhaps, to have uniform chemical composition.This last assumption is uncertain since theabundance of interstellar grains may havevaried within the cloud complex in which thestars were formed. In principle, knowledge ofthese properties of the stars in a large numberof clusters and associations would provide anideal basis for studying (1) the properties ofstars of equal age but of different mass andangular momentum; (2) the changes, with age,not only of the stars' position in the HR dia-gram but also of their rotational and duplicityproperties.

159381014—56 -13


Trumpler's extensive investigations of galac-tic clusters have revealed the large variety ofpatterns in the HR diagrams and this, combinedwith modern concepts of stellar evolution,shows that these objects represent a wide rangeof evolutionary stages. The modern equivalentof these diagrams, based on accurate photo-electrically measured colors and magnitudes,has already allowed more detailed discrimina-tion of the properties of different stars in thesame cluster, and more refined comparison ofthe color magnitude arrays of different clusters.I t has, moreover, the great advantage of reach-ing considerably fainter apparent—and, hence,also absolute—magnitudes than the spectro-scopic method.

While color and absolute magnitude are themost obvious quantities to choose in definingthe physical properties of a star, they do notdescribe them completely. Initial chemical com-position and angular momentum cause vari-ations of the array in the color absolute-magni-tude diagram for clusters of equal age. Weshould consider also the duplicity or multiplicityof the stars and their motions in the cluster.These will probably have no relation to thecolor-magnitude arrays, but they must have animportant bearing on the process of the forma-tion of the cluster from the interstellar mediumand will have to be taken into account in subse-quent studies of the evolution of the clusters.There may be other properties of the stars in ad-dition to those listed above, which, though notyet easily accessible, may provide importantclues; the magnetic fields might be among these.

There can be little doubt that the first desider-atum for future work on galactic clusters is thedetermination of accurate color absolute-mag-nitude arrays for a large number of clusters.These should provide the basis for subsequentstudies of the secondary effects mentionedabove, which may be feasible only for a limitednumber of stars and a limited number of clus-ters. These latter could then be selected so asto represent the most useful variety of evolu-tionary stages.

A critical step in the determination of thecolor absolute-magnitude arrays is the conver-sion of apparent magnitudes into absolute mag-nitudes. There are only a few clusters for whichwe have geometrically determined distances:

the Hyades, Ursa Major, and a Persei clusters.The use of differential galactic rotation as anindicator of distance has not produced valuabledistances in the past, but it may do so in thefuture for the young clusters if these follow thespiral structure of interstellar hydrogen closelyand if radial velocities determine uniquely towhich arm the clusters belong. H. L. Johnson'sapproach to the problem of the absolute magni-tude determination, based on the assumption ofidentical "zero-age main sequences," may proveto be very useful provided sufficient evidencecan be produced to show that the chemicalcomposition does not affect the position of themain sequence. One must bear in mind, how-ever, that this approach eliminates the detec-tion of composition differences unless these areof a more pronounced character in the rapidlyevolving part of the color absolute-magnitudediagram than in the slowly evolving part; thedetermination of the absolute magnitude scaleshould be based, preferably, on the latter.

The extension of the color-magnitude arraysto faint stars may require the identification offainter cluster members than are known atpresent. This holds even for the nearestclusters, the Hyades being a good example ofa cluster whose faint members are still unknown.It will in general be the case for the poorerclusters observed against the rich backgroundof the Milky Way. So far, observations ofthe faint stars in clusters have usually beenconfined to a sample considered representativeas far as the magnitudes and colors are con-cerned, with little attention being given to theestimation of the total numbers of faint mem-bers. Yet, this is an important item if wewant to consider the density of the populationalong the main sequence at zero age for differ-ent clusters. That differences in this respectexist follows from the comparison of NGC752 and Praesepe. Such differences are prob-ably related to the mass density and the stateof motion in the interstellar medium from whichthe cluster originated. Their observation mayprovide interesting data on the formationprocess of the clusters.

Existing data on the identification of faintcluster members can in many cases be con-siderably increased by the measurement of newproper motions in the fields of the clusters.


The large collections of photographs of galacticclusters, taken at several observatories 30 ormore years ago, are becoming more and moreuseful for this purpose. In many cases, workof this kind done in the years preceding WorldWar II cannot be considered to be up to datefor the future projects mentioned above.Comparison of the old epoch plates with modernones may give proper motions of several to10 times the weight of those available at present.It is gratifying to know that several observa-tories that possess fine collections of old plateshave already started comparing these withmodern ones, whereas others have expressedtheir interest in undertaking programs of thiskind. International Astronomical Union Com-mission 37 on star clusters and associations atthe Dublin Assembly has adopted a resolutionto the effect that observatories which possesscollections of early epoch cluster plates presentlists of these to the president of Commission37 and indicate to what extent they are in aposition to collaborate on proper motion work.

Observations of the stellar rotation velocitiesand of the frequency of spectroscopic andvisual binaries have, in the past, been givenrelatively little attention. Yet they would seemto offer most interesting prospects for futureresearch. Why are there few or no spectro-scopic binaries among the Pleiades althoughvisual doubles are plentiful, whereas spectro-scopic binaries are frequent in other clusters,notably in those containing many early-typestars? Are we facing here an evolutionaryeffect or is this difference directly related todifferent initial conditions in the interstellarmedium at the time the clusters were born?Whatever the answer may be, it is obviousthat a systematic study of the frequency ofrotational velocities and of the duplicitycharacteristics in clusters of a variety of agesand appearances will ultimately give valuableinformation on the formation problem.

The foregoing remarks referred in the firstplace to the work on galactic clusters. Althoughthe basic problems are very much the same forthe associations, the emphasis in future workwill probably be somewhat different. Theassociations on the whole are much youngerthan the clusters so that the stars have evolved

very little, while the internal velocities aregenerally large and very likely much the sameas at the time of formation of the stars; there-fore, they can indicate the initial propertiesof the stars and may yield clues to our under-standing of their process of formation. Theobservations made so far have been concernedmainly with the photometry (absolute mag-nitudes and colors), the proper motions, andradial velocities, while a beginning has beenmade with the systematic investigation of theduplicity properties. But all these referredonly to the brightest part of the population andthe stars of spectral types about B5 and earlier.Very little is known as yet of the occurrenceof fainter, later-type stars. Stars down tospectral type A0 along the main sequencehave been surveyed in the Orion association, inIII Cep, and in the most concentrated partsof a few other objects. But in two of thenearest associations, II (f) Perseus and ILacerta, we are still ignorant about the mem-bers of types B5 and later. Yet these twoassociations are favorably located at relativelyhigh latitudes where the confusion by thegeneral galactic background is less serious thanin most other cases. Surveys of the late B andlater-type stars in associations will suffer froma few drawbacks: the area over which the starshave to be classified is much larger than forthe clusters—for the nearest associations ofthe order of 100 square degrees—and, sincethe internal velocities are large, members andfield stars cannot be distinguished uniquely.The membership may have to be based onstatistical evidence, like that, for instance,showing that the concentration in the sky fora particular type of star is greater in the regionof the association than outside it. The sizeand direction of the proper motions may helpwhen the expanding character is pronounced,but then it can be used only in the outer regionswhere the density of the population is small.In spite of these difficulties, knowledge of thefainter membership of the associations mustbe considered one of the most urgent goals ofresearch on these objects. The first approachshould, of course, be by Schmidt telescopes orlarge-field astrographs equipped with objectiveprisms.


Dynamical studies

Dynamical studies of clusters have, in the past,been mainly concerned with systems in astate of approximate equilibrium, and havedealt with the relation between internal velocitydistribution and mass distribution, the timesof relaxation, and the rate of escape of starsfrom the cluster. Investigators usually as-sumed that the clusters are approximately theage of the universe, several billion years.Recently acquired knowledge of the ages ofclusters derived from increasing informationon stellar evolution offers the possibility fora fresh approach to dynamical studies ofclusters. Various questions come to our mind.We notice that the distribution of stars in theHyades is quite irregular. We estimate itsage at about 109 years. Can we reconcilethese two observations, estimate how manystars must have escaped from the Hyades, andpredict its future constitution? Suppose weindeed collect data on the ages of a large num-ber of clusters as suggested in the precedingsection, can we explain the numbers of clustersof various ages as a result of the predicted rateof disintegration ?

The dynamics of associations presents prob-lems of a somewhat different kind. Especiallyin the youngest objects, the forces acting on thestars are probably mainly due to the interstellarmatter present in the region of the association.If we make the reasonable assumption thatonly a small fraction of the interstellar matterhas condensed into stars, its mass may verylikely exceed the total mass of the stars manytimes. Probably the most promising aspectof the study of the motions in associations isthat it will give us direct information on thestate of motion of the generating interstellarcloud complex, and it is intimately connectedwith the complicated problems of the dynamicsof cosmic clouds. I t is clear, however, that wemust obtain the most precise data on themotions of individual stars. For the brightestones (m<C6) we will depend on continuedmeridian observations. For the fainter stars(around m=9) we must rely on photographicdeterminations of new positions. These willrequire special programs involving large num-bers of plates covering the limited regions of

the associations in order to attain high accuracyfor the individual stars. In addition to theproper motion work we must consider radialvelocity observations as one of the principaldesiderata for future work. This has beenmentioned also hi connection with the detectionof spectroscopic binaries.

The system in the galaxy

Investigations of the space distribution and thekinematics! properties of clusters have shownthat they occupy a flat subsystem in the galaxywith peculiar motions of the order of 10 km/sec.Only in recent years have the more accuratephotometric distance determinations revealedthe connection between open clusters and thespiral structure.

For future work in this field, also, the ac-cumulation of a large number of accurate color-magnitude diagrams with the inferred knowl-edge of the ages of many clusters forms thebasic requirement. Once this information hasbeen obtained one can study the distributionaland kinematical properties for different agegroups. For the youngest clusters the evidencealready obtained for the distribution along thespiral arms will be strengthened, and it will beinteresting to see where, along the spiral struc-ture of interstellar gas, the formation of clustersused to take place. Data on proper motions,already referred to, and mean radial velocitieswill provide information on the velocities ofthese objects at the time of formation.

The most intriguing of these studies may,however, well be those dealing with the some-what older objects. Shall we be able to recon-struct from their present distribution andmotions their distribution at the time of forma-tion ? This may be possible for the slow movingclusters whose ages are now of the order of onerevolution of the galaxy (200 million years).If we assume their space velocities to be of theorder of 5 km/sec, they have traveled duringtheir lifetime a distance of about 1,000 parsecsfrom their origin. This is probably not too farto allow an approximate reconstruction of thedistribution of the original locations. Thus wemay be able to study the secular changes in thespiral structure of the galaxy, a problem whichwe otherwise could tackle only by studying thespace distribution of the A-type stars. These


are, of course, much more numerous, but theclusters have the advantage of being detectableover a larger area.

These studies require measurements of theproper motions and of the radial velocities ofthe brightest stars in the clusters, in additionto the photometric data. It may not be toooptimistic to expect that present and futureobservational equipment may collect these datafor several hundred clusters.

Studies of the space distribution of the asso-ciations are well under way at present and areof great interest because they outline the spiralstructure. Also radial velocities of the associa-tion are being measured and they are essentialin supplementing the studies of spiral structurebased on the 21-cm hydrogen emission. Workon the proper motions, of special value for thenearer objects, will be greatly facilitated by therecently planned repetition of the AGK2.

Globular Clusters Observed througha Crystal Ball

By W. A. Baum1

At the focus of a large telescope, a globularcluster is one of the most magnificent sights inthe sky. One can see thousands upon thou-sands of stars all crammed together within afew minutes of arc. Embedded like jewels hereand there in the sparkling swarm, the brighterstars run the full gamut of color; the brightestare brick red while others, slightly less bright,are distinctly blue. Toward the center of thecluster, the population becomes fantasticallydense. It is amusing to consider what our skywould look like if the earth were associatedwith a star in the nucleus of a globular cluster.

Perhaps the best way to judge what thefuture has in store will be to examine thepresent state of our knowledge and somepresent lines of speculation to see the gapswhich need to be filled.

Globular clusters constitute a very well de-fined class of objects, all pretty much alike intheir general features. You could hardlymistake one for any other type of system. Outof more than a hundred observable globularclusters in our own galaxy, only two or three arestill debatable cases.

The distribution of globular clusters in thegalaxy is almost spherically symmetric aboutthe galactic nucleus, whereas stars like thosein the solar neighborhood, along with gas anddust, are restricted to a highly flattened disk-shaped region. This difference in distributionis associated, as one might expect, with adifference in angular momentum; globularclusters probably do not as a group share to anylarge degree in the rotation of the galactic disk,but they have large individual radial velocities.Except in the immediate region of the galacticnucleus, the number of globular clusters per unit

1 Mount Wilson and Palomar Observatories, Pasadena, Calif.

volume decreases as the inverse third or fourthpower of the distance from the nucleus, andthere is no well-defined outer boundary to theirdomain. I t would be very valuable to have amore complete and accurate analysis of thedistances and velocities of all known globularclusters in the galaxy, but the task is a big one.

The stars in globular clusters are an entirelydifferent breed from those with which we arefamiliar in the solar neighborhood. They havedifferent chemical abundances and they con-sequently follow different patterns of evolutionas they age. These differences can be deducedfrom spectra and from color-magnitude dia-grams. Stars of the globular-cluster kind aresometimes referred to as Population II. Theyare found not only in the clusters but also asunattached vagabonds wandering through thesame regions of the galaxy (nucleus and halo)frequented by the clusters. They are foundsimilarly in the halo of M 31 and in nearbydwarf galaxies. In general they seem to favorregions which are free of dust. Since it doesnot necessarily follow, however, that starsinhabiting a dust-free region are always thesame kind (i. e., same chemical abundances andassociated patterns of evolution), one cannotsay with certainty whether stars of the globular-cluster kind do or do not constitute the pre-dominant population in elliptical galaxies; ifthey do, then they are probably the mostnumerous kind of stars in the universe. Thisis a fundamentally important question onwhich more work is needed.

Color-magnitude diagrams

One of the best ways to characterize the starsthat compose a stellar system is to plot theircolor indices (or their spectral types) against

165

166 SMITHSONIAN CONTHIBUTIONS TO ASTROPHYSICS

- 6

My

- 4

- 2

0

+2

• 4

+6

+8

_

—

_

—

i i i

\

\\\\\\\ ^

fIIfMI3

1 1 1

1 1

MI3

M3

\\

\\ J

X

i i

i i i i i i i i

—

—

—

^^^^ —

AA

MI3/I

/ 1J 1/ J

M3

—

\ —w \\ \

M3 \

X N POP. i

\ MI3 ^ N

i i 1 I I N I I-.4 +.4 +JB -t-1.2 +1.6

FIGURE 1.—Schematic color-magnitude diagrams for M 3 and M 13. Sequences for the solar neighborhood are sketched withdashed lines for comparison. The diagrams have been adjusted vertically so that the mean of the cluster-type variables(circle) in M 3 falls at M,=0 and so that the evolutionary breaks from the main sequence fall at the same bolometric magnitude(dotted line).


their magnitudes. In contrast to the diversityof color-magnitude diagrams found amongopen clusters which inhabit the galactic disk,the color-magnitude diagrams of globularclusters all have basically similar features.This does not mean that they are identical.There are indeed important differences amongthem, but the differences mainly pertain to thepositions of the sequences rather than to theexistence or nonexistence of various sequences.

Color-magnitude diagrams for two well-known globular clusters, M 3 and M 13, aresketched in figure 1. The color indices spanan extremely broad range; the bluest stars haveindices around —0.5 on the PQ— V system whilethe reddest are around +1.5. The brighteststars in globular clusters have absolute photo-visual magnitudes of about —3, while thefaintest are presumably much below the presentlimit of observation at the bottoms of thediagrams shown here.

In the lower part of each diagram, fromM , ~ + 4 downward, is a segment of the mainsequence. Stars which formerly lay a littleabove M f ~ + 4 on the main sequence haverecently evolved away from it and theypresently form the upper branches of thediagram. Stars which started from somewhatbrighter levels of the main sequence, sayabove M , ~ + 3 , have long since completedtheir evolutionary sequences and passed onto the graveyard. When a star evolves awayfrom the main sequence it evidently movesinto the subgiant region at the right of themain sequence and climbs upward into thered-giant branch. Since the total ascent isabout seven magnitudes, the star is consumingfuel several hundred times faster during itsred-giant phase than during its childhood onthe main sequence. As a result it doesn't livevery long as a red giant. When it starts torun short of hydrogen, the star is believed tomove toward the left along the horizontalbranch and then dribble downward at theextreme left of the diagram to the graveyard.Its death throes are suspected of being occa-sional nova outbursts, after which it collapsesinto a degenerate white dwarf and cools itselfinto oblivion.

The time required for the color-magnitude

diagram to reach its present configuration isuncertain, but it apparently is of the samegeneral order as the age of the universe judgedfrom the recession of galaxies, say, 5 to 10billion years. Each star probably spends morethan half of its life on or near the main sequence,where it uses up only 10 to 15 percent of itshydrogen. As it proceeds up the verticalbranch, its rate of rise is roughly proportionalto the luminosity, and its radius expands asthe square root of the luminosity. By thetime it reaches M,~0, the star is evolving quitefast. I t requires only about 2X108 years togo the rest of the way to the top of the diagramand about an equal length of time to traversethe horizontal branch. The nova stage, ifindeed that is the star's actual route to thegraveyard, must be less than 109 years' dura-tion. Beyond that, its evolutionary scheduleis unknown.

Near the top of the red-giant branch, thestar starts to fluctuate small fractions of amagnitude at a relatively slow rate, its ampli-tude increasing as the tip of the branch isapproached. In the same region of the color-magnitude diagram one occasionally finds abonafide long-period variable, which suggeststhat the fluctuation may build up briefly tolarge amplitudes as the hydrogen-spent starstarts to collapse and slide back down the red-giant branch. In any case, it apparentlyquiets down again as it travels rapidly towardthe left along the horizontal branch. Thensuddenly it breaks into another fit of oscilla-tion, this time as a cluster-type variable. Itsperiod and light curve are relatively steady,but they are believed to undergo gradualsecular changes as the star pursues its evolu-tionary course through this region of thediagram. Present data suggest that the period,starting initially at about one day, becomesprogressively shorter, while the amplituderises to a maximum and then falls again. Thisfall in amplitude is accompanied by a changein the shape of the light curve which may beassociated with a change in the mode of pulsa-tion. At the blue edge of the variable-starregion the star quiets down again until it slidesoff the blue end of the horizontal branch. Ifthe star then passes through a nova stage, one


might count it as the third fit of instabilityalong the evolutionary track, but it evidentlydiffers in kind from the other two.

The foregoing paragraphs outline the generalfeatures of globular-cluster diagrams and theirevolutionary interpretation as we understandthem today. While certain basic facts nowseem to be fairly definite, it should be empha-sized that much of this information is only ten-tative and some of it is even speculative.Moreover, there are a few items which wereomitted altogether, because we haven't thefoggiest notion how to fit them into the over-allevolutionary picture. These include the long-period Cepheids (between M,^ — 1 andM,~*—5) found in some of the clusters, theplanetary nebula in M 15, and the occasionalblue stars found above the horizontal branch ina few clusters.

Altogether, there are so many points onwhich more photometric material is needed thatit is difficult to suggest what should be under-taken first. In some cases, the question islargely one of observing more stars or samplingmore clusters. This is true, for example, if onewishes to understand the nature of red-giantfluctuation, the properties of the red long-periodvariables, or the role of the long-period Cepheids;it is also true of investigating the differencesbetween cluster diagrams, which we shall dis-cuss shortly. In other cases, the problem israther to improve the accuracy of existing data.I t has sometimes been tempting to overrate thefinality of the present material and to neglectthe need for further investigation.

Among the various branches of the clusterdiagrams, the main sequence is observationallythe most difficult. Recent developments inphotoelectric photometry, however, have madeit possible to measure extremely faint stars withcomplete freedom from the scale errors thathave hampered photographic efforts. As aresult of such developments, it has been foundthat the stars comprising the main sequence ofa globular cluster do not fall on the ordinarymain sequence with which we are familiar.They are subdwarfs. This result is of cos-mogonic significance because it leaves littledoubt that globular clusters are made from adifferent recipe from that of the majority ofstars in the solar neighborhood; a disparity in

age alone would hardly seem sufficient toaccount for it.

There are some fascinating differences be-tween the color-magnitude diagrams of variousglobular clusters. Features which differ notice-ably from cluster to cluster include: (1) theposition of the main sequence, (2) the position(colorwise) of the top of the red-giant branch,(3) the shape and tilt of the red-giant branch,(4) the number of cluster-type variables, (5)the populousness of the horizontal branch atthe right of the cluster-type variables, and (6)the shape and tilt of the horizontal branch.The last four of these features seem to bemutually related; if one intercompares all of thecolor-magnitude diagrams which have beenphotoelectrically calibrated on the same colorsystem, he finds that they can be arranged intoa series (call it series A) in which these lastfour features tend to change mutually andprogressively from one diagram to the next.The number of cluster-type variables therebybecomes a handy index for categorizing globularclusters and their diagrams. It is rather puz-zling to note, however, that the second featurein the list above suggests arranging the dia-grams into a different series (call it series B),and that it is this latter arrangement whichspectroscopic observations tend to support.Data bearing on the first feature in the listabove are still too scanty to tell us whichseries, if either, the main sequences will favor,but a prediction can be ventured: if series Atruly represents the effects of chemical abun-dances upon evolution, the arrangement of themain sequences ought to be in accord with it.

Actually, series B has somewhat the appear-ance of being series A folded in the middle, butthe number of clusters investigated is toosmall to provide more than a hint of any suchrelationship. If one intercompares a group ofcluster diagrams arranged in order accordingto series A, the tops of the red-giant branchestend to be progressively bluer until one comesto M 92, after which the red-giant branchesare progressively redder. Both extremes ofseries A might thus be associated with the redextreme of series B, and we could then forgetabout series B having any independent sig-nificance. This happy unification seems torun into grief, however, if we attempt to


interpret abundances from the spectra andsimultaneously to invoke abundances to accountfor series A. If abundance differences arereserved for explaining series A, and if B cannotbe associated with it, then perhaps we shouldseek a phenomenon which does not affect thegeneral evolutionary pattern but which doesaffect the swollen atmospheres associated withthe stars during their red-giant phases. Thefact that all the giants of a cluster are affectedto somewhat the same degree suggests thepossibility of an environmental phenomenon,such as surface accretion during passage of thecluster through the galactic plane or nucleus,but such environmental effects have not beenregarded with much favor. A third possibilityfor explaining the existence of the two series isbased upon the fact that abundance differencesthemselves are not necessarily one-dimensional;in principle there are as many parameters asthere are elements in the periodic table. Wecan imagine, for example, that some of thefeatures of the color-magnitude diagram (maybethose abiding by series A) might be sensitiveto the hydrogen/helium ratio, while others(maybe those abiding by series B) might besensitive to the hydrogen/metals ratio. In anycase, the problem of interpreting these series isa very challenging one.

Photometric observations of high precisionare badly needed in more globular clusters ifthe differences between them are to be fullyunderstood. Although many color-magnitudediagrams of various sorts exist, only a few havebeen calibrated photoelectrically and adjustedaccurately to a standard color system. Thereare plenty of favorably situated clusters whosediagrams would be valuable. Table 1 is a listof clusters which lie 20° or more from thegalactic plane. Twenty-four of them are inthe Southern Hemisphere while only 13 are inthe Northern Hemisphere; however, severalof the southern clusters can be reached fairlywell by northern telescopes. The modulusrepresents a rough estimate of the magnitudeof the horizontal branch; for the upper branchesto be well defined, the color-magnitude diagramshould extend about two magnitudes fainterthan the modulus without serious deteriorationof accuracy. Since integrating techniques nowenable one to obtain photoelectric observations

to about the same limit that can be reachedphotographically, most of these clusters canbe reached with telescopes of modest aperture ifthey are adequately equipped. The approxi-mate number of known variables, the integratedspectral class (or alternatively the integratedcolor index), and the approximate distance fromthe galactic nucleus are all parameters bywhich clusters can be grouped and compared.We have already cited the number of variablesas an index to the form of the color-magnitudediagram according to series A. We shall returnin a moment to a discussion of the relationshipof these integrated properties to the distanceof the cluster from the galactic nucleus.

Among the clusters in table 1, color-magni-tude diagrams suitable for intercomparingthe upper branches have thus far been pub-lished for NGC 4147, M 3, M 5, M 13, M 92,M 2, M 10, and M 15. Work is now in progresson one or two others, but the majority haven'tbeen touched. It should also be mentionedthat there are several clusters which are toointeresting to neglect in spite of the fact thatthey lie at less favorable galactic latitudesthan those listed in table 1. Outstandingexamples with low moduli are w Centauri, M 4,M 22, and NGC 6397. Clusters lying close tothe galactic nucleus include NGC 6304, NGC6316, NGC 6355, NGC 6453, and NGC 6522.There are also two or three massive galacticclusters of great age, similar perhaps to M 67,whose color-magnitude diagrams may help toclarify evolutionary differences between haloand disk populations.

It is much more difficult to reach significantsamples of the main sequences than similarlyto reach the upper branches. It requires alarge telescope and a great deal of photoelectricobserving time under first-rate conditions.Thus far, M 13 is the only cluster for which themain sequence has been firmly located. Sometentative photoelectric results were used forincluding M 3 in figure 1, and similar materialalso now exists for M 15. Earlier appearancesof normal main sequences in the photographic-ally extrapolated diagrams for M 3 and M 92have turned out to be spurious. It may beseveral years before a definitive amount ofdata on the relationships of globular-clustermain sequences can be assembled.


TABLE 1.—Globular clusters for which |6|>20c

Clutter designation a 1950 t 1950

NGC 104 47 Tuc 00 2l!°9 - 7 2 21288 00 50.2 - 2 6 52362 01 00.6 - 7 1 07

1261 03 10.9 - 5 5 251841 04 52. 5 - 8 4 051851 05 12. 4 - 4 0 051904 M 79 05 22.2 - 2 4 342419 07 34.8 +39 004147 12 07. 6 +18 494590 M 68 12 36.8 - 2 6 295024 M 53 13 10.5 +18 265053 13 13. & +17 575272 M 3 13 39.9 +28 385466 14 03. 2 +28 465634 14 27.0 - 0 5 455694 14 36.7 - 2 6 195824 15 00.9 - 3 2 535897 15 14. 5 - 2 0 505904 M 5 15 16.0 +02 166171 M107 16 29.7 - 1 2 576205 M 13 16 39. 9 +36 336218 M 12 16 44 6 - 0 1 526229 16 45.6 +47 376254 M 10 16 54.5 - 0 4 026341 M 92 17 15.6 +43 126684 18 44. 1 - 6 5 146752 19 06.4 - 6 0 046809 M 55 19 36.9 - 3 1 036864 M 75 20 03.2 - 2 2 046934 20 31. 7 +07 146981 M 72 20 50.7 - 1 2 447006 20 59. 1 +16 007078 M 15 21 27.6 +11 577089 M 2 21 30.9 - 0 1 037099 M 30 2137.5 - 2 3 25

NGC7492 23 05.7 - 1 5 54IC4499 14 52.7 - 8 2 02

Estimated modulus(M)

mag

14. 515. 715. 7

16. 319.417.016. 116.716. 415.716. 417. 117.7

16.215.216.214. 815.217. 715.215.2

14.214. 518. 116.917.018.815. 916. 215.517.9

Number of variables

82

14

35

364

284210

18618

70

972415

121

216

12

11513120661739

Integrated spectrum

F5F3F5A5A6A8

F2

F4A9F5

F7G2F2F7F6GOA5

GlF9G2FlFOFOA7

Distance fromgal. nucleus

kpc8

1512

20652413191713171922

87694

284

10

54

2515163912139

32

The foregoing discussion of color-magnitudediagrams is presented in terms of ordinarytwo-color photometry. Much can be added,however, by extending observations to three ormore colors. In particular, the addition ofultraviolet enables one to compare the relativestrengths of the Balmer regions in variousstars. This can be done for instance byplotting a diagram of U-B against B-V.Results in M 13 show that the red giants andsubgiants have an ultraviolet excess of about0.05 magnitude, while the very blue stars onthe horizontal branch have very pronouncedultraviolet deficiencies. Rather oddly, thereis a lonesome B 2 star lying about two magni-tudes above the horizontal branch in M 13which does not exhibit this ultraviolet defi-ciency. In M 3 and NGC 4147, the ultravioletexcess of the red giants and subgiants is severaltimes greater than in M 13. Similar informa-tion for other clusters would be very valuable;it might provide an important clue to the reason

for the differences between their color-magni-tude diagrams.

Multicolor photometry and spectrophotom-etry have not yet been applied to individualstars in globular clusters, and the results ofany multicolor exploration are likely to payhandsome rewards. Although the useful rangeof a photometer becomes limited to brighterobjects when the color bands are made narrower,the bright stars in globular clusters are stillwithin easy reach of conventional six-colorobservations with a telescope of modest aper-ture. On the other hand, narrow-band spec-trophotometry will be quite difficult.

SpectraSpectroscopically, the individual stars in glob-ular clusters are rather difficult objects; hardlyany of them are brighter than 14th magnitudein the blue region. Until recently, the spectrahave consequently been of quite low dispersion.These have been suitable, however, for spectral


classification and for detection of the grosserpeculiarities of the cluster stars. It was alreadyrecognized a decade ago, for example, thatcyanogen absorption is abnormally weak andthat the hydrogen lines are a little more en-hanced than in normal stars of the sameabsolute magnitude.

During the last few 3rears, the fast newcoud6 cameras at Palomar have made it pos-sible to reach some of the brighter cluster starswith a dispersion of 38 A/mm and, in twocases, of 18 A/mm. Because of the very largecollimator-to-camera ratios for these disper-sions, the slit of the Palomar spectrograph canbe made wide enough to receive nearly all of astar image during good "seeing" without causinga loss of photographic resolution in the spec-trum. The first four such spectra at 38 A/mmwere obtained in 1952 with the idea of check-ing spectroscopically on the photometricallyobserved difference between the red-giantbranches of M3 and M92, which at that timewere the only two clusters whose color-mag-nitude diagrams could be intercompared on astandard system. These spectra not only con-firmed the reality of an intrinsic differencebetween these two clusters but also showedquite dramatically that the red giants in M92are very extraordinary objects. They aremarkedly different from any other known stars.Additional spectra of the same dispersion weresoon obtained in M92 and, subsequently, inmost of the globular clusters for which inter-comparable color-magnitude diagrams nowexist as well as in several others not yet photo-metrically investigated.

The most striking peculiarity of the red-giant spectra in some of the clusters is theextreme weakness of the metallic lines. Super-ficially, these spectra appear to be of verymuch earlier types than the color indices of thestars suggest. In M 92, for instance, themetallic lines seem to have roughly the strengthsusually found in F-type stars, whereas the colorindices would normally be associated withearly K stars. This disparity largely vanishes,however, if the spectral classification is basedupon the ratios of certain metallic lines insteadof^upon their absolute strengths. In otherwords, the excitation temperature is much lessdiscordant with the color temperature if it is

inferred, as it should be, from the ratios of linestrengths instead of from the absolute strengthsthemselves.

A few particular ratios deserve mention. Asthe excitation temperature increases, the ratioof X4254 to X4250 should decrease, whereas theratio of X4247 to X4250 should increase. Thefeature at 4254 A is a resonance line of neutralchromium (0.0 volt excitation potential), thefeature at 4247 A is a blend of scandium II witha high excitation line of neutral iron (3.3volts), and the comparison line at4250A arisesprincipally from a low level of neutral iron(1.6 volts). Another useful ratio which de-creases with increasing temperature is Cr I4275 to Fe I 4271, and another ratio whichincreases with increasing temperature is Ti II4394 to Fe I 4405.

M 92 is apparently an extreme case of thepeculiarities cited above, although M 15 runsit a close second. In some of the other clusters,such as M 3 and M 13, these effects are relativelymild. Within any one cluster, the spectra allshow roughly the same degree of peculiarity.In general, this degree of peculiarity appears tobe correlated with the redness of the red-giantbranch (i. e., with series B) in the sense thatthose clusters whose red-giant branches liethe farthest toward the right in the color-magnitude plane are the least peculiar spectro-scopically.

All of the spectroscopic evidence suggeststhat the weakness of the metallic lines in someof the cluster spectra is a bona fide abundancephenomenon. Quantitative abundance esti-mates have been made on the basis of spectraobtained at 18 A/nun in M 92 and M 13, whichrepresent opposite extremes. Equivalentwidths of suitable metal lines in an M 92spectrum are found to be about 60 percent ofthose in an M 13 spectrum, implying that themetals in M 92 are at least a factor 10 lessabundant than in M 13. Line ratios in thesesame spectra assure us that the excitationtemperatures in M 92 and M 13 are aboutthe same in spite of the difference in the colorindices; perhaps the color difference is due tothe fact that metallic lines chop differentamounts out of otherwise similar continua.

As pointed out earlier, however, the use ofabundances to account for spectroscopic differ-


ences (and therefore series B) poses a dilemma,because we also feel obliged to call uponabundances to explain the various evolutionaryfeatures which behave according to series A.Possible ways around this dilemma, such asallowing more than just the hydrogen/metalsratio to vary, have already been mentioned.

Integrated propertiesIn addition to the classification of color-magnitude diagrams, there are several otherinformative parameters which pertain to eachcluster as a system or as a unit. These includeintegrated luminosity or magnitude, integratedcolor index, integrated spectral-energy distri-bution, integrated spectral class, mass, mass-luminosity ratio, distance, reddening, positionin the galaxy, motion, spatial profile, and so on.

Since the integrated luminosity and colorrepresent the summed effects of all of themembers of a cluster, one should in principlebe able to subtract the contributions of theresolved stars from the total observed lumi-nosity, and thereby deduce how much light iscontributed by members which are too faintto be individually resolved by present telescopes,say, fainter than mPf=23. In practice, how-ever, uncertainties in the luminosity function(i. e., in the numbers of stars per magnitudeinterval) are much too large and the quantitysought is much too small for such a procedureto succeed. Direct measurements of thisunresolved background by itself appear to showsome indications of a substantial white-dwarfpopulation, but the uncertainties are presentlytoo large to justify much confidence.

As one might expect from the differencesbetween color-magnitude diagrams, the frac-tions of the total light contributed by thevarious branches differ somewhat from clusterto cluster. Typically, about half of the visuallight comes from the red giants above Af,=0,perhaps 20 percent each from the horizontalbranch and from the subgiants, and theremainder from the main sequence downward.In view of the broad range of color indices ineach cluster diagram, it is not surprising thatthe integrated color indices of globular clustersare not identical. One cannot correctly deter-mine the reddening of an individual cluster

simply by comparing its color index with anunreddened mean.

The integrated spectral-energy distribution ofa stellar system such as a globular cluster is, ofcourse, quite different from that of a single starhaving the same color index. This means that,regardless of data-reduction procedures, theordinary color index of a cluster is rather sensi-tive to the actual color response of the photo-meter used.

The spectral energy distribution of a globularcluster is also very different from that of anelliptical galaxy, indicating that the color-magnitude diagrams must be fundamentallyunlike unless the distribution of stars in the un-evolved region of the cluster diagram is radicallydifferent from that for an elliptical. If the dia-grams are assumed to be of the same form, onecan show that the ratio of the number of K starsto the number of F stars on the main sequenceof an elliptical galaxy would need to be literallyhundreds of times greater than on the main se-quence of a globular cluster in order to accountfor the observed disparity in their integratedspectral-energy distributions. Since the presentdata can be fitted equally well by assuming un-like diagrams, the identification of the pre-dominant population in ellipticals remains anopen question. The importance of further workon this problem has already been stressed.

For much the same reason that the color indexis sensitive to the wavelength bands used for theobservations, the estimation of the integratedspectral class of a stellar system such as a glob-ular cluster is dependent upon the spectral re-gion utilized. The blue stars on the horizontalbranch play a bigger role at short wavelengthsthan at long wavelengths. Integrated spectraof 50 clusters have been obtained by throwingthe clusters sufficiently far out of focus to smeara blend of many stars uniformly over the spec-trograph slit. The spectral types listed in table1 were derived from integrated spectra obtainedin this manner. Some of these spectra are cur-rently being reexamined in an effort to takebetter account of the peculiarities now recog-nized; the present tendency is to assign themlater spectral types than were formerly esti-mated.

An attempt has been made to determine the

NEW HORIZONS EN ASTRONOMY 173

mass of M 92 from the dispersion in radial ve-locities among red giants in its nucleus. Thiswas done by applying the virial theorem tovelocities measured on 23 Palomar spectro-grams of 15 stars. Some of the red giants forwhich two spectrograms were obtained showevidence of intrinsic velocity variations, whichis not surprising in view of the photometricvariations already described. As a result ofthese variations, the uncertainty in the truevelocity dispersion is disconcertingly large.Nevertheless, with some recent refinements intheory the results yield an estimated mass of1.4 X 10s solar units, the upper limit being per-haps twice that value. The correspondingmass-luminosity ratio is of the order of unity,which is several times smaller than the averagefor our own galaxy as a whole and is about ahundred times smaller than that of an ellipticalgalaxy. The interpretation of these compara-tive ratios in terms of differing populations,luminosity functions, and nonluminous content4'are not yet clear.

The distances of globular clusters are con-ventionally estimated by assigning the cluster-type variables a mean absolute magnitude ofzero, or by comparing the brightest stars withthose belonging to other clusters whose variableshave been calibrated. I t should be rememberedthat the assignment of zero absolute magnitudeto the cluster-type variables is only a rough ap-proximation which should not be taken tooseriously when color-magnitude diagrams areintercompared. In figure 1, for example, thecluster-type variables are a half magnitude apartbecause the diagrams were adjusted to place theevolutionary breaks from the main sequences atthe same bolometric magnitude.

Since most globular clusters inhabit regionswell outside the absorbing material in thegalactic disk, the photometric moduli (blue)should, on the average, be reduced by 0.25esc b in order to estimate true distances. Whilea statistical correction of this sort is usefulwhen one is dealing with large-scale problemsinvolving many clusters, it does not provide areliable estimate of absorption for individualclusters due to the probable patchiness of theabsorbing blanket in which we are imbedded.Unfortunately, other criteria of reddening and

absorption such as comparison of U, B, V datawith the solar neighborhood or matching ofcolors of the cluster-type variables are also opento question. Consequently, the existence ofaccurate photometric data would not by itselfguarantee a precise distance for a cluster evenif the absolute magnitude of the cluster-typevariables were exactly known.

The fact that globular clusters differ signifi-cantly from one another in their diagrams andin their over-all characteristics suggests thatthey probably were not all formed under iden-tical circumstances. I t is therefore of interestto consider what clues we have as to their timesand places of origin during early epochs of thegalaxy. When the integrated characteristicsof globular clusters are examined in terms oftheir positions in the galaxy, some fascinatingcorrelations stand out immediately. In par-ticular, the number of cluster-type variables,which is an index for classification according toseries A, appears to have a marked correlationwith the distance from the galactic plane. Ifwe examine all clusters which have beensearched for variables, which are not badlyobscured, and whose distances are fairly wellknown, we find that the variable-poor clusters(say, those with 0 to 3 variables) favor regionsmuch closer to the galactic plane than theothers, hinting that they may have been formedduring a slightly later stage in the dynamicrelaxation of the galaxy. The sample of clus-ters in table 1 is not well suited for illustratingthis effect because of its cutoff at 20° latitude,but the effect can still be seen; the mean dis-tance from the galactic nucleus to the variable-poor clusters listed in the table is 9 kpc, whilethe mean distance for the variable-rich clustersis 19 kpc.

It follows, of course, that the various char-acteristics which are correlated with the numberof variables tend to show this same dependenceon galactic distribution. For example, the in-tegrated spectral type, which is influenced bythe relative shape and populousness of theupper branches of the cluster diagram, revealsan effect of this kind. Among the clusterslisted in table 1, those possessing later spectraltypes (say, F 7 to G 2) have a mean distanceof 11 kpc from the nucleus, while those pos-


sessing earlier types (say, A 5 to F 6) have amean distance of 21 kpc from the nucleus.Again, the effect would be more distinct ifclusters in the 10° to 20° zone were added tothe sample.

If the centroid of the system of globularclusters in our galaxy is assumed to coincidewith the galactic nucleus, one can check thezero point of the cluster-type variables againstthe present estimate of the sun's distance fromthe nucleus (about 8 kpc). The variable-richclusters seem to be quite well centered on thenucleus if the conventional value of M=0is adopted for the variables, but the variable-poor clusters appear to prefer a slightly brightervalue, perhaps half a magnitude. Whilescarcely more than qualitative due to thesmallness of the sample, this result is nicelyconsistent with figure 1.

There are several observable globular clustersinhabiting the immediate vicinity of thegalactic nucleus. One or two of these may betransients just passing through, but others aresuspected of being permanent residents and ofbeing quite peculiar. I t would be especiallyinteresting to see how they fit into the over-alltype-distribution picture described above.

The individual space motions of globularclusters in the galaxy are unknown becausethey are too distant to have reliable propermotions. Nevertheless, some general conclu-sions can be drawn from their radial velocities,which range from —360 km/sec to +290km/sec. As already remarked, the globularclusters probably do not share to any largedegree in the general rotation of the galaxy.The present evidence suggests that the rota-tional velocity of the cluster system is less thana third of the rotational velocity of the disk.I t may be difficult to pin this figure down moreclosely due to the limited number of clustersand to their large individual motions. Rel-ative to the clusters, the sun's orbital velocityis about 200 km/sec in roughly the same direc-tion derived from differential rotation. Thecluster system shows no indication of a generalcontraction or expansion.

Efforts have been made to determine whetherglobular clusters favor orbits of high ellipticitypassing through the galactic nucleus or whether

they travel in approximately circular orbits atconstant radii from the nucleus. Althougharguments have been advanced for both models,the present radial velocity data do not make aclear case for either. Probably the truth issomewhere in between, and perhaps the situa-tion is further complicated by the form of thepotential function of the galaxy. In anycase, the time required for a cluster to makeone loop or one oscillation is of the samegeneral order as the period of galactic rotation;hence, each cluster has passed through thegalactic plane quite a few times since its birth.

The spatial distribution of the stars in aglobular cluster has been the subject of severalobservational and theoretical studies. Obser-vationally, the data consist of star counts or sur-face brightness measures as a function of radiusfrom the cluster center. To a first approxima-tion, these quantities vary as (constant+r2)""2,and the corresponding integrated luminosityrapidly approaches a well-defined asymptoticlimit for large r. This two-dimensional profileis very different from that of an ellipticalgalaxy, whose integrated luminosity continuesto increase as log (constant+r) and whichshows no tendency for a limit except thatimposed by the domain of its neighbors.

Since the profile of a globular cluster is amanifestation of internal dynamic conditions,it is related to the numbers of stars havingvarious masses, to their mass-luminosity ratios,and to the manner in which energy is parti-tioned. We know that nearly all of the lightof a globular cluster comes from stars above itsmain sequence and that their masses should allbe of the same order because they all evolvedfrom the same region of the main sequence;thus a single mass-luminosity ratio applies fairlywell to the surface-brightness profile. This hasbeen checked by noting that the difference inintegrated color between inner and outer regionsof a cluster is generally small and that the radialspread of the red-giant population does notdiffer significantly from the spread of the sub-giants and the blue stars. However, since theexchange of energy in the central region of acluster is believed to be sufficient for equi-partition, one should expect the profile for themain-sequence membership to be broader than


the profile for the brighter stars, and there isobservational evidence in support of this.Differing amounts of internal energy per unitmass from cluster to cluster probably accountfor their differing degrees of over-all compact-ness.

The origin of globular clusters is still shroudedin mystery. It seems fairly evident that most ofthem in our own galaxy were formed during anearly epoch in its evolution. We do not know,however, whether they condensed out of inter-stellar material originally in the domain theynow inhabit or whether they were created indenser central regions and ejected due to thehigh turbulent energy which may have existedat that time. M 87, which is a very massiveelliptical in the Virgo cluster of galaxies, pro-vides a possible clue to this question. It hastwo unusual features: there are an astonishingnumber of globular clusters associated with it,and there is a peculiar blue jet streaming fromits nucleus with several pronounced globulesalong it. Is it possible that M 87, because ofits unusual mass and energy, has a slow rate ofdynamic development and is still in the stageof making globular clusters? Perhaps therange of integrated color indices of the M 87clusters will help to decide whether any of themcan be young, but the problem is observa-tionally difficult. It is also of interest to notethat the radial distribution of the globularclusters in M 87 is approximately the same asthe distribution of the nonclustered population(judged from the surface brightness profile).Since the relaxation time of a galaxy is theoret-ically long compared with the average age ofthese clusters, this result implies that the condi-tions of formation have been such as to producesimilar distributions since the beginning.

In closing, we should not neglect to considerthe value of globular clusters as tools in otherastronomical problems. Their usefulness insurveying the structure of our own galaxy isalready evident from earlier remarks, but theirimportance in extragalactic studies also deservescomment. Globular clusters belonging to othergalaxies are convenient distance indicators.Their luminosities can be used for intercom-paring the distances of galaxies as far as theVirgo cluster of galaxies, which serves as a

cosmic yardstick for everything beyond. Arecent estimate based on the globular clustersassociated with M 87 places the Virgo clusterof galaxies at 10 megaparsecs. If new resultsfor more distant clusters of galaxies are adjustedto this scale value, one obtains a Hubble con-stant of 150 km/sec per megaparsec, whoseinverse is 6.5 billion years. Much more workis needed on this problem, not only in connec-tion with the photometry of globular clustersin remote galaxies but also in connection withestablishing a better luminosity function bywhich the clusters belonging to different galaxiescan be compared. We must also worry aboutwhether the luminosity function for clusters isnecessarily the same in all kinds of galaxies.

A fair amount of observational work has beendone on the globular clusters in the Andromedagalaxy (M 31). In some respects they providea better sample than those in our own galaxyfor investigating colors, luminosities, compact-nesses, motions, and distribution. The brightestare around 15th apparent magnitude, whichcorresponds to M ~ — 9. They are close enoughthat we can even compare their profiles withthose of clusters in our own galaxy to obtain anobscuration-free check on the distance of M 31.

The present paper outlines some of theproblems in a very broad field. So many dif-ferent workers have had a part in building ourpresent state of knowledge that it would havebeen foolish to attempt the incorporation ofanything approaching a complete system ofbibliographical references into the present text.An excellent 80-page bibliography, by Helen B.Sawyer, already exists (Publ. D. Dunlap Obs.,vol. 1, p. 383, 1947), but anyone actually plan-ning to work in the field will also want tobecome acquainted with at least part of thework done since that date. Recent publica-tions and current programs include work bythe following: Arp, Artiukhina, Baade, Baum,W. Becker, Belserene, Bernstein, Cuffey,Deutsch, Eggen, Fatchikhin, Florya, Gaposch-kin, Gascoigne, Greenbaum, Greenstein, Hoag,von Hoerner, Huang, Idlis, Irwin, H. L. John-son, Kholopov, I. King, Kron, Kukarkin,Lohmann, Mayall, Melbourn, Minin, Morgan,Munch, Oort, Oosterhoff, Osterbrock, Parenago,

176 SMITHSONIAN CONTRIBUTIONS TO ASTROPHYSICS TOUI

Reddish, Roberts, Roman, Rosino, Sandage, of these activities are summarized and refer-Savedoff, Sawyer-Hogg, Schurer, Schwarzschild, enced in the 1948, 1952, and 1955 reports ofShapley, Stebbins, Swope, Tayler, Walker, Commission 37 of the International Astronom-Wahlquist, Wallerstein, Whitford, Wilkins, ical Union (see Trans. Int. Astron. Union,O. C. Wilson, van Woerden, and others. Some vols. 7-9).

Stellar Structure and EvolutionBy M. Schwarzschild 1

The theory of the internal constitution of thestars entered a basically new phase in 1938with the discovery of the particular nuclearprocesses that provide the main energy sourcesfor the stars, and, soon thereafter, with thefairly accurate determination of the reactionrates of these processes.

It had long been known that a star of givenmass and chemical composition has a uniqueequilibrium structure. But it was not possibleactually to determine this unique structureas long as one of the fundamental relations,that for the rate of energy production bynuclear processes, was unknown. Now thatthe necessary nuclear data are available wecan not only determine the internal structureof a star in its initial state when it just startsburning its hydrogen fuel, but we can alsofollow, step by step, in a quantitative non-speculative manner, the evolutionary changescaused by the nuclear transmutations.

Many possibilities are now within our grasp:to interpret the Hertzsprung-Russell diagramin terms of stellar evolution, to determine theages of star clusters and even of individualstars, to sort out the oldest stars and by analyz-ing them to gain insight into the compositionof our galaxy at its earliest phases, and toobtain clues to the origin of the elements bya comparative analysis of the oldest and theyoungest stars.

Only the very first steps have as yet beentaken in this large field of new and fundamentalproblems. The future speed and success ofthe developments will depend not only on thenumber of specialists working in the subjectand on their ingenuity but also, much morethan in earlier decades, on the advances inneighboring scientific and technical areas.Clearly, more highly accurate determinationsof reaction rates for a larger variety of nuclear

1 Princeton University Observatory, Princeton, N. J.

processes will be needed from nuclear physics.Similarly, for many star clusters color-magni-tude diagrams of high precision at faint mag-nitudes are needed from observational astron-omy. In both these fields, which are funda-mental to the theory of stellar evolution,progress has been very rapid and encouragingin recent years. There are, however, threeother neighboring fields which should beactively cultivated if the investigation of stellarevolution in its broadest sense is to progress.

Computing facilities are our first concern.The basic problem of stellar structure and evolu-tion can be formulated mathematically interms of a fourth-order boundary-value prob-lem, whose solution depends on the stellarmass and chemical composition and varies withtime. To find the solution, numerical methodshave to be employed. Much useful prelimin-ary computation can be done with desk ma-chines. The numerical work needed, however,to obtain a sufficient range of solutions withsatisfactory detail and accuracy is so extensivethat it cannot be tackled effectively with deskcalculating machines; large electronic machinesare needed. At a small number of institutessuch large machines have generously been madeavailable for pure astronomical research. Astro-physical research absolutely requires the helpof large electronic computers if it is to growfrom its present promising beginning to thematurity the problems demand.

The second area we need to develop is photo-metric spectroscopy with high dispersion to de-termine the chemical composition of many sam-ple stars. This observational work, which isbasic to many branches of astrophysics butwhich is very tedious and time consuming withtraditional techniques, has progressed veryslowly in recent years despite its importance.These investigations may expand as soon as itbecomes practicable to use electronic image

177


tubes for this purpose. The advent of this newinstrumentation, however, will have its full rev-olutionizing effect only if at the same time alarger number of technicians becomes availableto build and maintain the new instruments andto reduce the data.

The third field is that of the turbulent phe-nomena in stellar atmospheres. Recent inves-tigations have shown that these phenomena areof interest not only for the theory of stellar at-mospheres but also play a decisive role in de-terminations of the stellar radius for many typesof stars; hence they are essential for the com-plete theory of stellar structure. Laboratory ex-periments and theoretical investigations in hy-drodynamics have resulted in great progress,during the last decade, in our understanding ofturbulence. However, in this difficult subjectextrapolation from the laboratory to the starswill always involve some risk; therefore, directobservations of turbulence in stellar atmos-

pheres would be highly valuable. This is, inprinciple, possible for the photosphere of thesun. In practice, however, it is extremely diffi-cult because of interference from the disturb-ances of the earth's atmosphere. In conse-quence, very little progress has been made inthis field during the past 50 years. Now, how-ever, observations from high-altitude vehiclesappear to open up new possibilities; if they canbe realized, the final major difficulty in thetheory of the stellar interior may be resolved.

At present, no serious block seems probable inthe necessary development of any of these fields.Thus, research in the theory of stellar structureand evolution presents many unsolved, but solv-able, basic problems and may result in a greatadvance in answering objectively certain cosmo-logical questions. The rate of successful prog-ress in this field will depend essentially on theingenuity and the number of workers who applythemselves to it.

Radio SourcesBy John D. Kraus *

Perhaps you have seen a distant landscapecompared in two photographs, one taken inordinary light and the other in infrared. In theordinary photograph the distant objects arelargely hidden by haze, but in the infraredpicture the distant hills and valleys stand outclearly. Infrared rays are longer than lightrays and more easily penetrate the haze.Radio waves are millions of times longer thanlight and have an even greater power ofpenetration than infrared. Because of thisfact, radio—applied to astronomy—cutsthrough the cosmic haze caused by the cloudsof gas and dust in interstellar space andpresents a view of vast regions that will alwaysremain hidden to ordinary telescopes. Radiohas, in effect, opened a new window on theuniverse.

It was about 1873 that Maxwell postulatedthe possibility of radio waves. About 15years later Hertz produced them and demon-strated their transmission over a distance of afew feet. In 1901 Marconi received radiosignals across the Atlantic and an era of world-wide radio communication was begun. Thirtyyears later Jansky detected radio waves fromthe center of our galaxy and initiated a newmethod for exploring the cosmic universe. Itwas 60 years from Maxwell to Jansky, from aman's idea to its cosmic application. Andtoday, some 20 years after Jansky's discovery,this application of radio to astronomy israpidly developing into a new science calledradio astronomy.

Radio observations of outer space are madewith an antenna and a radio receiver. Theantenna gathers and focuses the radio waves asdoes the lens of an optical telescope, while thereceiver detects and records the signal much asdoes the photographic plate of an optical

1 Ohio State University, Columbus, Ohio.

telescope. By analogy, the antenna and re-ceiver may be called a radio telescope.

As observed with a radio telescope, the skylooks strangely different. None of the familiarstars of the night sky are observable, but intheir places are many radio sources formingtotally unfamiliar constellations. Some of thesesources can be identified with supernova rem-nants and other gaseous nebulosities and withextragalactic nebulae. There are a coupledozen sources thus identified, at least tenta-tively, but there are hundreds more notidentified with any optical object and the bulkof them may never be. It is as though radioand optical telescopes "see" two differentuniverses with only an occasional objectobservable by both. And those objects thatcan be detected by both appear to be of differentsizes and shapes in the two telescopes. Forexample, the sun has a visual diameter of aboutone-half degree, but at certain radio wave-lengths the diameter measures several degreesbecause the radio waves originate in the upperregions of the corona or solar atmosphere.

If our eyes were sensitive to radio wavesinstead of to light, the winter sky would appearsomewhat as suggested in figure 1. Thecontours are lines of equal surface brightnessshowing the radio background. The back-ground is a maximum along a ridge close to theplane of our galaxy or Milky Way system. Onthis background are dots and circles that showthe position of some of the strongest or brightestradio sources. Their brightness is indicated bya radio magnitude scale analogous to the visualmagnitude scale. Some of the sources are of anextended type with diameters of more than 1°.These are shown by the open circles. Othersare more localized or pointlike, and these areindicated by the solid dots. A few sourceshave been identified with optical objects and

179

SMITHSONIAN CONTRIBUTIONS TO ASTROPHYSICS

•TVROSETTE NEBULA

N6C2244// /

- 4 0

FIGURE 1.—Map of the radio sky at a wavelength of 124 cm made with the radio telescope at Ohio State University(KoandKraus, 19SS).

for these the optical designation is given. Notefor example the Crab Nebula (Ml), which is asupernova remnant, or the gaseous nebulaedesignated IC 443 and NGC 2244. However,many of the sources on the map have not beenidentified with an optically observable objectand this is true for most radio sources, partic-ularly the weaker ones not shown on the map.

The identification of radio sources withoptical objects is a challenging problem thatrequires close cooperation between radio andoptical astronomers. Accurate radio positionand size determinations are vital in this work.

One of the outstanding achievements of radioastronomy was the identification of the secondbrightest radio source (other than the sun)with a faint pair of colliding galaxies at a dis-tance of some 200 million light-years. Team-work by radio astronomers at Cambridge,England, and Sydney, Australia, and opticalastronomers at Mount Palomar was requiredfor several years before the identification wasmade (Baade and Minkowski, 1954). If thisobject were at 10 times the distance it wouldstill be a readily observable radio source butwould then be at or beyond the range of the


largest optical telescope. It is this fact whichsuggests that radio telescopes may open to ex-ploration vast regions of space that lie beyondthe range of any optical instruments.

The known radio sources are now numberedin the hundreds, with lists published by Ryle,Mills, Bolton, Brown, Haddock, and OhioState University. However, those that arereliably established are much less numerousand those identified with optical objects areonly a dozen or two (Pawsey, 1955).

The work of radio exploration has barelybegun and great discoveries lie ahead as largerradio telescopes are built and more radio sourcesare found. Many of these sources may emittoo little light for optical detection or may behidden by clouds of cosmic dust, while manyothers may be at distances beyond the range

of any optical telescope. Studies of thesesources, individually and collectively, are boundto supplement and alter our purely opticalpicture of the universe in a most significantway. From Maxwell to Hertz, to Marconi, toJansky, to the outermost reaches of our universeare big steps, but they do not end there andnew advances are being made daily. Perhapsyou may become the radio astronomer whowill make the next big step.

ReferencesBAADE, W., AND MINKOWSKI, R.

1954. Astrophys. Journ., vol. 119, p. 206.Ko, H. C , AND KBAUS, J. D.

1955. Ohio State Univ. Radio Obs. Rep. No. 4,June 22; in Sky and Telescope, vol. 14,p. 370.

PAWSEY, J. L.1955. Astrophys. Journ., vol. 121, p. 1.

o

SMITHSONIAN CONTRIBUTIONS to ASTROPHYSICS

Documents

Transcript of SMITHSONIAN CONTRIBUTIONS to ASTROPHYSICS