New evidence for extraordinarily rapid change of the geomagnetic field during a reversal

ISSN : 0028-0836

Accès :

00006056-199504200-00013

Nature

New evidence for extraordinary rapid change of the geomagnetic fieldduring a reversal

Auteur(s) : Coe, R. S.; Prev ot, M.; Camps, P.

Numéro : Volume 374(6524), 20 April 1995, pp 687-692

Type de publication : [ARTICLE]

Editeur © 1995 Macmillan Magazines Ltd.

Institution(s) :

Earth Sciences Department, Univ ersity of

Calif ornia, Santa Cruz, Calif ornia 95064, USA

(R. S. Coe).

Laboratoire de Geophy sique et Tectonique,

Univ ersite de Montpellier 2, URA CNRS 1760,

34095 Montpellier Cedex 05, France (M.

Prev ot, P. Camps)

AbstractPalaeomagnetic results f rom lav a f lows recording a geomagnetic polarity rev ersal at Steens Mountain, Oregon

suggest the occurrence of brief episodes of astonishingly rapid f ield change of six degrees per day . The ev idence is

large, sy stematic v ariations in the direction of remanent magnetization as a f unction of the temperature of thermal

demagnetization and of v ertical position within a single f low, which are most simply explained by the hy pothesis that

the f ield was changing direction as the f low cooled.

The Steens Mountain rev ersal record, dated radiometrically at 16.2 My r bef ore present [1], has long held the

reputation as the best recording of a polarity transition by lav a f lows [2]. A detailed study [3-5] f ound 56

distinguishable f lows with stable direction of magnetization intermediate between normal and rev ersed. Figure 1

shows this record, which can be div ided into two parts. In the f irst, the f ield mov es f rom stable rev ersed polarity

through transitional directions to temporary normal polarity . In the second, the f ield retreats f rom the short-liv ed

episode of normal directions back to intermediate directions that are similar to those attained bef ore, and then

successf ully completes the transition to normal polarity directions v ia a dif f erent route.

Ovid: New evidence for extraordinary rapid change of the ... http://ovidsp.tx.ovid.com.biblioplanets.gate.inist.fr/spa/o...

1 sur 13 23/06/2009 14:16

Figure 1. The Steens Mountain directional record showing the three large jumps or gaps (encircled numbers). The

projection is equal area, and each point is a directional group that represents one to nine consecutiv e f lows with

indistinguishable directions. Stars denote normal and rev ersed geocentric axial dipole directions. Filled (open)

sy mbols are plotted on the lower (upper) hemisphere. a, First part of record: f ield direction mov es f rom stable

rev ersed to transitional to temporary normal. b, Second part of record: f ield direction mov es f rom temporary normal

to transitional to stable normal.

An interesting aspect of the record is the three large gaps, two of which occur in the second part Figure 1.

Because we did not observ e obv ious signs at the outcrop of a lapse in eruption rate at any of the gaps, and

because f lows within two of the gaps y ielded anomalously scattered directions f or no apparent reason, we suspected

originally that the f ield may hav e changed unusually rapidly during the time of the gaps [3-5]. Re-study of the f low

in the f irst gap y ielded ev idence strengthening our suspicion, suggesting that the f ield may hav e changed

extraordinarily rapidly (3 degrees or 300 nT per day ) [6]. Howev er, we could not completely rule out unremov ed

v iscous magnetization as a possible alternativ e explanation. In this Article we f ocus on the second gap, recorded in

a parallel section 1.5 km to the north, where the f ield jumped about 80 degrees f rom westerly to southeasterly

declinations Figure 1(b). As we shall show later, v iscous contamination is not a tenable explanation f or the surprising

results we obtained, nor do sev eral other alternativ e thermal and rock-magnetic hy potheses stand up to scrutiny .

The new sectionTo carry out this study we mov ed 250 m north along the clif f f rom the original section A, tracing two prominent

f lows A38 and A39 (directional group 21 of Figure 1) that outcrop continuously , to a place where the underly ing f lows

are particularly well exposed and could be sampled v ertically (section D, Figure 2). There, these f lows are laterally

continuous on a scale of 100 m, in contrast to some of those in the same stratigraphic position at the original

location, which pinch out within a f ew metres of the sampled sections. Thus some of these f lows are probably not

precisely the same units as at the original section, although a cov er of rubble between the two places prev ented us

f rom v erif y ing this directly . Nonetheless, just as was the case at the prev ious locality , the two f lows D41 and D40

immediately underly ing the prominent marker f lows A39 and A38 f all within the same directional gap and display

anomalously scattered directions of high-temperature remanence, which smear out between the tightly clustered

directions of the f lows immediately below and abov e Figure 2. Moreov er, in contrast to samples f rom their margins

and f rom other f lows, many samples f rom the interior of f lows D41 and D40 f ail to reach a stable endpoint during


2 sur 13 23/06/2009 14:16

thermal demagnetization; instead, their directions tend to rotate f rom southerly towards westerly declinations as the

temperature progressed bey ond 300 degrees C Figure 3(a).

Figure 2. Natural remanence directions af ter thermal demagnetization to 500 degrees C of samples f rom f lows within

the directional gap (shaded), which display anomalous scattering, and f rom pre-jump and post-jump f lows. At the new

section, where each f low was sampled along a v ertical prof ile, f low D41 and D40 directions smear out all the way

between the directions of the underly ing and ov erly ing f lows. At the original section [3], where horizontal cov erage

was stressed more than v ertical sampling, f low A42 and A41 directions are nonetheless signif icantly smeared, with

Fisher [38] precision parameter k = 29 as compared to k = 196 and k = 156 f or the pre- and post-jump f lows below

and abov e, respectiv ely .


3 sur 13 23/06/2009 14:16

Figure 3. a, Vector component diagrams illustrating how the direction of natural remanent magnetization (NRM) of

samples f rom the interior of f low D41 (right) rotates f rom southerly toward westerly declinations during progressiv e

thermal demagnetization abov e 300 degrees C, whereas that f rom samples near the margins (centre) and f rom lower

f lows (lef t) decay straight towards the origin with constant direction. Points are the projection of the remanence


4 sur 13 23/06/2009 14:16

v ector on the horizontal (open circles) and N-S v ertical (f illed circles) planes, and the numbers beside them are

demagnetization temperatures in degrees C. Arrows denote unblocking temperature interv als used in the conductiv e

cooling computation [16] summarized in the text. b, Angular dev iation of remanence af ter thermal demagnetization to

500 degrees C as a f unction of v ertical position of samples in f lows through the second directional jump (lef t);

equal-area projections showing the f ull direction of samples f rom f lows D41 and D40 af ter demagnetization to 500

degrees C (right). Angular dev iation f or each sample is the angle measured f rom the pre-jump direction (D42) toward

the post-jump direction (A39) af ter projecting its direction onto the plane containing the f low D42 and A39 mean

directions. Dif f erent sy mbols denote dif f erent v ertical prof iles within a horizontal distance of 10 m. Note the large

swing in direction in f low D41 towards and away f rom A39 and the smaller such swing in f low D40.

Ev en more remarkable, howev er, is the sy stematic v ariation of direction of high-temperature remanence of

samples within f lows D41 and D40 as a f unction of their v ertical position Figure 3(b). From bottom to top in the lower

two-thirds of f low D41, the direction mov es f rom near that of underly ing f low D42 (prejump direction) about 60

degrees towards that of the ov erly ing f low A39 (post-jump direction), and then back to the f low D42 direction again.

The swing in direction is corroborated in sections 2 m to the south and north, which ov erlap the bottom and top of the

principal section. A similar but smaller swing occurs in the lower one-third of f low D40, but abov e that all samples

exhibit stable endpoint directions that are close to those of the ov erly ing f low.

Rapid-field-change hypothesisA simple hy pothesis to explain these results is that when f low D41 was cooling through its magnetic blocking

temperatures, the geomagnetic f ield was mov ing f rom the pre-jump direction recorded by f low D42 towards that of

the post-jump direction recorded by f low A39. This accounts v ery naturally f or the change in remanence direction

away f rom that of the ov erly ing f low towards that of the underly ing f low during progressiv e thermal demagnetization

of indiv idual samples Figure 3(a). It also explains the observ ed v ariation in direction of high-temperature remanence

with v ertical position in the f low Figure 3(b), because the more slowly cooled interior of f low D41 records directions

that are f urther f rom that of the underly ing f low.

The similar but less pronounced swing shown in the lower part of f low D40, which picks up approximately where

that in f low D41 leav es of f , suggests that it was erupted shortly af ter D41 was emplaced, probably ev en bef ore

D41 had completely cooled. The samples higher in D40 appear to hav e been remagnetized by the ov erly ing f low

A39, because they all maintain directions close to those of A39 at ev ery demagnetization step abov e 350 degrees

C. They also exhibit titanomagnetite-granulation textures in polished thin section under ref lecting light that are

characteristic of hy drothermal alteration [7]. This texture is absent in the lower third of D40 and all of D41,

suggesting that f low D40 may hav e protected D41 f rom being remagnetized by the hy drothermal action of the much

thicker A39.

Experimental observ ations of cooling basaltic lav a on the island of Hawaii demonstrate that simple conductiv e

calculations can describe the ev olution of temperature with time reasonably well, especially when an ef f ectiv e

thermal dif f usiv ity of k = 6 X 10 sup -7 m2 s sup -1 is used to take account of av erage v esicularity and other

second-order f actors [8]. The characteristic cooling time f or a thin sheet of constant thickness and thermal

dif f usiv ity held at constant temperature at its boundaries is t = a2/k, where a is half the thickness [9]. For f low D41,

which is 1.3 m thick, this giv es only 8 day s, suggesting an astonishingly f ast rate of change of f ield direction of the

order of 10 degrees per day . This rapid change apparently did not occur as a result of v anishingly low geomagnetic

intensity during the directional jump. Palaeointensity experiments conducted prev iously [5] y ielded estimates of the

transitional f ield intensity that decrease f rom 10 to 7 micro Tau f or the three pre-jump f lows, remain at 7 micro T f or

the two anomalous f lows in the jump (samples f rom the quickly chilled f low margins), and increase f rom 4 to 13

micro T f or the three post-jump f lows ov erly ing the jump. Thus the ancient f ield strength during the cooling of

anomalous f low D41 was probably about 7 micro T, which would require a rate of change of total f ield around 1 micro

T per day to account f or the directional rate.


5 sur 13 23/06/2009 14:16

Alternative hypothesesThis rate is so extraordinary , of the order of 1,000 times f aster than that directly observ ed in secular v ariation

of the geomagnetic f ield [10], that we are obliged to consider caref ully any alternativ e hy potheses that might

account f or our results. These are of two general classes: thermal (that is, simple baking or multiple cooling units)

and rock magnetic (that is, anisotropic remanence acquisition, v iscous contamination or chemical remagnetization).

Baking. Hy potheses inv olv ing simple baking by the ov erly ing f low are beset by serious dif f iculties. Thermal

demagnetization experiments show that only the upper two samples of f low D41 were signif icantly remagnetized by

the ov erly ing f lows, exhibiting appreciable angular dev iations of remanence between 300 and 500 degrees C f rom

the pre-jump towards the post-jump direction. Bey ond 500 degrees C thermal demagnetization, the directions of the

six upper samples are v ery similar to the pre-jump direction, which implies that the upper 60 cm of the f low was not

reheated abov e that temperature. Below 60 cm, the directions display the peak in angular dev iation with depth Figure

3(b) that giv es rise to the rapid-f ield-change hy pothesis. For this signature to hav e been caused instead by baking,

the middle part of the f low would hav e to hav e been reheated to higher temperature than the upper 60 cm, which is

of course unrealistic.

Multiple cooling units. More f lexibility is af f orded by the hy pothesis that f low D41 was erupted as more than one

cooling unit. To account f or the directional signature of f low D41, any later-erupted unit would hav e to intrude the

f irst horizontally , thereby delay ing cooling of the interior relativ e to the margins. Examples of such behav iour are

known f rom observ ations of modern eruptions, but unless intrusion occurs bef ore most of the f irst unit has

cry stallized the two units do not weld at their boundaries and the compound nature of the f low is apparent (G. P. L.

Walker, personal communication). No internal boundaries of this kind are present in any section of f low D41 that we

hav e examined, suggesting that if any such intrusion occurred. it was when the interior of the f irst unit of D41 was

well abov e the solidus temperature. Thus most of the two units would still hav e cooled as one through the

remanence blocking range. This conclusion is reinf orced by microscopic examination of polished thin sections of

samples in f low D41 f or indications of relativ e cooling rate. The av erage maximum size of titanomagnetic cry stals

shows a simple trend, with its maximum in the central part Figure 4(a). Moreov er, the proportion of dendritic

titanomagnetite, indicativ e of degree of undercooling, is maximal (near 100 percent) at the top and bottom

boundaries and decrease progressiv ely inwards.

Figure 4. Magnetic properties of samples (solid lines) through the critical f low D41 as a f unction of v ertical distance

abov e the bottom contact. Angular dev iation of remanence (see Figure 3(b)) also shown (dotted line) f or ref erence.

a, Av erage maximum grain size (micrometer) of titanomagnetite cry stals (not size of remanence-carry ing magnetite

particles--see text). b, Curie point Tc and v iscosity coef f icient V. c, Hy steresis parameter ratios indicativ e of

magnetic grain size (Jrs /Js = saturation remanence/ saturation magnetization; Hrc /Hc = remanent


6 sur 13 23/06/2009 14:16

coerciv ity /coerciv ity ).

Magnetic anisotropy . Pref erred orientation of magnetic minerals and their easy axes of magnetization can

def lect the direction of thermoremanence away f rom the ambient f ield [11,12]. To cause the large dev iations of

remanence directions, howev er, this mineral f abric would hav e to be exceptionally strong in the interior and

weakened progressiv ely toward the boundaries. On the contrary , the f abric of f low D41 is extremely weak

throughout and incosistent in orientation: susceptibility anisotropy is small (less than 0.5 percent) and bears no

relation to the angular dev iation of remanence. Furthermore, thermoremanent magnetization imparted in the course

of palaeointensity experiments [5] to f our samples of the anomalous f lows f rom the original section dev iated f rom

the applied f ield by no more than a f ew degrees.

Viscous contamination. Our experience with Steens and other basaltic lav as is that the v iscous ov erprint

impressed by the present normal polarity Brunhes epoch is ef f ectiv ely eliminated af ter demagnetization at

temperatures ranging between 150 and 300 degrees C. The ov erly ing f low direction and that of the normal axial

dipole f ield are f ar apart in the second jump, making it implausible that Brunhes v iscous remanence could play any

role in the observ ed pattern of directions in f low D41. To test the possibility that a more ancient unremov ed v iscous

component (imparted by thermal enhancement of magnetic v iscosity af ter the thick ov erly ing f lows A39 and A38

were emplaced) could explain the v ertical distribution of remanence direction in f low D41, we measured the v iscosity

index, def ined as the ratio of v iscous remanence acquired in the ambient f ield in approximately two weeks to the

stable natural remanent magnetization. It is low throughout the f low (generally less than 10 percent) and not

correlated with the angular dev iation of remanence Figure 4(b), indicating once again that v iscous remanence is not

a v iable explanation.

Chemical remanence. A similar but less easily discounted class of hy potheses is that a moderate rise in

temperature or hy drothermal activ ity accompany ing the emplacement of post-jump f lows A39 and A38 might hav e

caused some kind of partial chemical remagnetization of f low D41. Because chemical alteration can produce

remanence with unblocking temperatures much higher than the temperature at which it occurs [12], a chemical

remagnetization could hav e unblocking temperatures that strongly ov erlap those of the original thermoremanence.

But as there is no a priori reason why such ef f ects should be peaked at the centre of a 0.7-m zone of lav a in the

lower half of the f low rather than increasing steadily upwards or being concentrated along unit boundaries or cracks

at the top and bottom, this interpretation needs to be supported by mineralogical observ ations or rock-magnetic data

to be conv incing.

Despite considerable ef f ort, we hav e f ound no such ev idence to support this hy pothesis. A detailed account of

our observ ations and magnetic experiments will be published elsewhere, but we summarize here the most important

results. In thin section under the petrographical microscope, f low D41 appears to be a common basalt, rich in

plagioclase and oliv ine. Oliv ine is partially altered to iddingsite. Titanomagnetite grains are subdiv ided by lamellae of

the usual trellis ty pe, ty pical of deuteric oxidation (that is, oxidation during original cooling) at high temperature.

Thermomagnetic curv es in low f ield indicate single Curie points close to 555 degrees C and are rev ersible within

10-15 percent. This is consistent with the impression, gained by microscope observ ations, of a simple magnetic

mineralogy dominated by iron-rich titanomagnetite. There is no signif icant v ariation of Curie point with v ertical

position in the f low Figure 4(b). About 5 percent of the natural remanence has blocking temperatures abov e 600

degrees C, signalling the presence of titanohaematite, which is also a ty pical product of deuterix oxidation.

Magnetic hy steresis parameters should also prov ide clues whether magnetochemical changes occurred that

could hav e caused the angular dev iations in f low D41. In particular, the ratios of saturation remanence to saturation

magnetization and remanent coerciv ity to coerciv ity Figure 4(c) are v ery usef ul indicators of ef f ectiv e magnetic

grain size. They demonstrate that the magnetic grains in f low D41 are pseudo-single-domain or interacting single-

domain, not multidomain [13]. They are smallest at the top and bottom and reach a maximum 85 cm abov e the


7 sur 13 23/06/2009 14:16

base, well abov e the anomalous angular dev iations of direction and somewhat abov e the height at which the largest

titanomagnetite grain size was observ ed microscopically . There is no contradiction with this latter observ ation,

because the lamellae produced by the deuteric oxidation/exsolution process subdiv ide the titanomagnetite host

grains. Thus the ef f ectiv e magnetic grain size is controlled by this process, rather than by the ov erall size of the

titanomagnetite grain. It is important as well to emphasize that f or magnetite grains like ours hav ing Jrs /J sub s

(saturation remanence/saturation magnetization) ratios larger than 0.2, the blocking and unblocking temperatures of

any partial thermoremanent magnetization (PTRM) are the same [14]. Ev en af ter allowance f or the dif f erence in rate

of cooling in nature and rate of heating in the laboratory [15], f or such f ine-grained particles the blocking temperature

in nature of a giv en PTRM cannot be much larger than the unblocking temperature in the laboratory .

To sum up, we hav e f ailed to unearth any v ariation in composition, phase or grain size of magnetic minerals as

a f unction of v ertical position that would suggest that magnetochemical changes occurred pref erentially in the lower

third of f low D41 during the emplacement of post-jump f lows A39 and A38. Although this negativ e result does not

prov e that the anomalous dev iations in remanence directions we f ound there could not hav e been caused by some

chemical mechanism, as y et unknown and f or which we can f ind no clue, this is obv iously not a satisf y ing

explanation. Moreov er, as described abov e, we did f ind both microscopic and magnetic ev idence f or

magnetochemical alteration of the upper 60 percent of f low D40 by the ov erly ing f low or f lows. This alteration

decreased in intensity downward, as would be expected, and did not af f ect the lower part of D40 or any of D41.

Finally , we point out that f low B51 in the f irst directional gap, f or which we reported earlier [6] a similar pattern of

angular dev iation to D41 in the lower half (though not in the upper half because baking was much more

pronounced--see Figure 4 in [6]), has quite dif f erent magnetic properties f rom f low D41. Its magnetic mineralogy is

more complex, with most samples exhibiting low-, intermediate- and high-temperature Curie points, and its magnetic

grain size and magnetic v iscosity are considerably larger. Theref ore, it is unlikely that one rock-magnetic explanation

could account f or the v ertical distribution of directions in both f lows, whereas the single hy pothesis that the

geomagnetic f ield direction was changing extremely rapidly as they cooled can do so v ery naturally .

Conductive cooling calculationsFor these reasons we return to the hy pothesis of rapid f ield change during cooling to enquire whether it can

explain quantitativ ely the thermal demagnetization results f rom f low D41. A calculation using the analy tical error

f unction solution [9] f or conductiv e cooling of a thin lav a sheet emplaced instantaneously at 1,140 degrees C on a

thick pile of already -cooled lav a, and with its upper surf ace held at constant ambient temperature, predicts that the

sample in f low D41 taking the longest time to cool to 500 degrees C (about 6 day s) would be the one which display s

the largest angular dev iation in Figure 3(b). This agrees with the rapid-f ield-change hy pothesis. Numerical

computation employ ing a more realistic model, which allows f or temperature-dependent dif f usiv ity , latent heat of

cry stallization released ov er a range of temperatures, and v ariable v esicularity with depth in f low D41, shows f or a

range of cases that the last sample to cool to 500 degrees C would be either that with the largest angular dev iation

or the next lower sample, and that the cooling time would be between 6 and 16 day s [16]. Figure 5 shows a f it of this

model to well def ined remanent directions of PTRM in f low D41 f or unblocking interv als abov e 300 degrees C. One

to f our (on av erage, three) PTRMs f rom all samples in the three v ertical prof iles through the f low are included. The

model assumes that as the f low cooled, each element of PTRM f aithf ully recorded the instantaneous f ield direction,

and, as discussed abov e, that the blocking temperatures f or each element are equal to their unblocking

temperatures during thermal demagnetization. The time elapsed f rom when the f low began to cool until it cooled

through the upper and lower blocking temperature of each PTRM is calculated and plotted against the angular

dev iation of PTRM direction. Thus the total magnetic recording interv al plotted is the span of time ov er which

v arious samples in f low D41 were cooling through magnetic blocking temperatures between 680 and 300 degrees C.


8 sur 13 23/06/2009 14:16

Figure 5. Conductiv e cooling analy sis by Camps et al. [16] of the palaeomagnetic data reported here, assuming the

rapid-f ield-change hy pothesis. The analy sis y ields a rate of change of geomagnetic f ield of 6 +/- 2 degrees per day ,

as suggested by shading (see text). Angular dev iation of direction obtained by thermal demagnetization of samples

f rom f low D41, measured f rom the underly ing (D42) f low direction toward the ov erly ing (A39) f low direction, is

plotted as a f unction of time elapsed since beginning of cooling (see text). Each bar indicates the time interv al

during which each direction of magnetization was blocked. Dashed bars correspond to samples near the f low top that

were reheated by the ov erly ing f low. (Figure adapted f rom [16].)

The general trend in Figure 5 is consistent with a simple model in which the f ield remained stationary during the

f irst half of the magnetic recording interv al and then mov ed f rom the pre-jump towards the post-jump direction at a

rate of 6 +/- 2 degrees per day . The f it to the data is reasonably good. Only sev en PTRMs lies of f the general trend.

Fiv e of those are shown as dashed bars in Figure 5 because they are f rom samples in the upper 22 cm of the f low

that were clearly remagnetized by baking, as discussed earlier, and so can be excluded. This is probably the case

f or the sixth, which is the lowest-temperature PTRM f rom the next sample down, still only 30 cm below the top. We

hav e no explanation f or the sev enth outlier, which is the 525-575 degrees C PTRM f rom the sample deep within the

f low that shows the peak in angular dev iation in Figure 3(b).

There are two main aspects of our data that cannot be f itted by the idealized cooling models we hav e employ ed.

The f irst, just mentioned, is the sharply peaked f orm of the angular dev iation as a f unction of v ertical position


9 sur 13 23/06/2009 14:16

Figure 3(b). We cannot dismiss it as an artef act due to a single outly ing point because it is corroborated by two

other samples f rom partially ov erlapping sections. The second aspect is the quantitativ e relationship of the angular

dev iations in the lower third of f low D40 to those in D41. The magnetic data suggest that D40 cooled quicker, and

baked the top of D41 less, than the simple models predict. Prev ious workers hav e also noted the surprisingly small

extent to which lav a f lows hav e reheated those underly ing them [17]. This indicates that the assumption of cooling

proceeding f rom an initially unif orm temperature throughout the f low just af ter emplacement must be signif icantly

incorrect. Such would be the case if appreciable heat were lost during the process of emplacement, as more

sophisticated modelling indicates [18], or if a later pulse of magma were injected into a prev ious f low unit erupted

shortly bef ore, as mentioned brief ly abov e. Indeed, the slight changes in slope of the maximum-grain-size curv e

Figure 4(a) around 25 and 100 cm abov e the base of f low D41 could conceiv ably be a result of such a two-stage

emplacement process. These complications notwithstanding, the general compatibility of thermal demagnetization

data f or f low D41 with the simple conductiv e cooling model is an argument in support of the rapid-f ield-change

hy pothesis.

Broader implications of rapid field changeIf this hy pothesis is correct, it may imply that rapid jumps of f ield direction occur many times during rev ersal

of polarity . If only three occurred during the rev ersal recorded at Steens Mountain, corresponding to the three gaps

labelled in Figure 1, and if it took sev eral thousand y ears to complete, as believ ed to be the cases f or this rev ersal

[3] and f or rev ersals in general [19-20], the probability of one of the 56 transitional lav a f lows cooling through its

magnetic blocking range during one of these jumps is a f ew tenths of a per cent. The Steens Mountain section

records two such occurrences. Thus it would either hav e to be an exceptionally rare recording, or many other rapid

changes of f ield direction occurred that were not recorded. Perhaps there were extended interv als when the f ield

jumped repeatedly , either randomly or in an oscillatory manner, or perhaps many isolated jumps occurred. Because

of the f ragmentary nature of lav a f low records and the decade or longer time resolution of sedimentary records,

these possibilities are dif f icult to prov e or disprov e.

Whether such rapid jumps of the geomagnetic f ield might be generated externally is an open question. It has

been suggested that during the rev ersal recorded at Steens Mountain, when the internally generated f ield was weak,

enhanced external f ield activ ity due, f or example, to greater penetration of charged particles f rom the Sun [21]

might somehow cause the jumps. Although attractiv e because of their ability to produce rapid f ield change, a

dif f iculty inherent in all such external mechanisms is that the f ast swing in direction occurred between relativ ely

stable beginning and ending directions. In other words, at the second directional gap, the f ield must hav e maintained

the pre- and post-jump directions f or sev eral tens of weeks at a minimum, enough time f or two f lows spanning 28 m

and f iv e f lows spanning 18 m Figure 2 to be erupted and cool. It is hard to understand how such long-liv ed directions

anchored to the Earth could be produced by the external f ield, and ev en harder to imagine how the external and

internal f ields could work in concert to produce the jump in direction.

On the other hand, one must ask whether magnetic screening due to the electrical conductiv ity of the mantle

would permit such a rapid f ield change to be observ ed at the surf ace if it were generated internally . The conductiv ity

of rocks near the surf ace is so low that their screening is negligible, but electromagnetic-induction studies hav e long

suggested that conductiv ity increases signif icantly with depth in the mantle [22,23]. Sev eral recent studies suggest

that this increase may occur at the phase transition at 660 km depth, reaching an av erage v alue of the order of 1 S

m sup -1 in the lower mantle [24-27], which is lower by a f actor of 10-1,000 than believ ed earlier. Laboratory

measurements of conductiv ity of presumed lower-mantle minerals at high pressure and temperature are consistent

with these low estimates [28] or are ev en lower [29,30]. For such a mantle conductiv ity prof ile (O S m sup -1 abov e

660 km depth, and 1 S m sup -1 below), we calculate that the characteristic smoothing time [31,32] f or an impulsiv e

change in magnetic f ield of spherical harmonic degree 1 and 6 at the core-mantle boundary is 20 and 6 day s,

respectiv ely . We consider that a degree of more than one is more realistic because such rapid changes at greatly

diminished f ield intensity suggest a local, non-dipolar source. Thus the rates of change inf erred f rom our data are


10 sur 13 23/06/2009 14:16

not incompatible with present estimates of mantle conductiv ity . Moreov er, it is widely held that mantle conductiv ity

is laterally heterogeneous on a large scale [33-37], which would allow such rapid signals to be observ ed at the Earth's

surf ace abov e regions of low conductiv ity ev en if the av erage were considerably greater than 1 S m sup -1.

Concluding remarksThe large-amplitude directional dev iations of remanent magnetization within single lav a f lows reported here, and

their unusual dependence on both v ertical position and unblocking temperature, are unique in our experience. Their

occurrence in two f lows at two dif f erent locations and stratigraphic positions where large jumps occur in the Steens

Mountain record is probably not coincidental. Putting aside f or the moment that f ield changes 1,000 times more rapid

than observ ed f or the modern f ield are implied, the hy pothesis that the rev ersing f ield changed sev eral tens of

degrees in direction during the cooling of these f lows is by f ar the most natural and satisf y ing explanation of our

observ ations. Ev en though such extraordinary rates of change are inescapable under this hy pothesis, it still seems

pref erable to any rock-magnetic explanation that we can imagine, each of which prov es to be sev erely contriv ed

when examined closely and tested with measurements of magnetic properties.

This is not to suppose that geomagnetic rev ersals take place much more quickly than the sev eral thousand

y ears currently supposed, but rather to suggest that polarity transitions may be punctuated by episodes of

extraordinarily rapid f ield change. An internal origin is not excluded by recent estimates of mantle conductiv ity ,

whereas an external mechanism cannot easily account f or the relativ ely long-standing pre- and post-jump directions.

Thus such rapid signals may well originate in the outermost core.

Receiv ed 8 September 1994; accepted 10 March 1995.

ACKNOWLEDGEMENTS. We thank G. P. L. Walker and G. Baker f or adv ice and help in the f ield, and J. Glen

f or assistance with computer draf ting. This work was supported by the CNRS-INSU Dy namique et Bilan de la Terre,

the US NSF Div ision of Earth Sciences, and the NATO International Scientif ic Exchange program.

REFERENCES1. Baksi, A. K. Geophy s Res. Lett. 20, 1607-1610 (1993). SFX Bibliographic Links Document Delivery

[Context Link]

2. Watkins, N. D. Geophy s. J. R. astr. Soc. 17, 121-149 (1969). [Context Link]

3. Mankinen, E. a., Prev ot, M., Gromme, C. S. & Coe, R. S. J. geophy s. Res. 90, 10393-10416 (1985). [Context

Link]

4. Prev ot, M., Mankinen, E. A., Gromme, C. S. & Coe, R. S. Nature 316, 230-234 (1985). [Context Link]

5. Prev ot, M., Mankinen, E. A., Coe, R. S. & Gromme, C. S. J. geophy s.Res. 90, 10417-10448 (1985). [Context

Link]

6. Coe, R. S. & Prev ot, M. Earth planet. Sci. Lett. 92, 292-298 (1989). SFX Full Text [Context Link]

7. Ade-Hall, J. M., Palmer, H. C. & Hubbard, T. P. Geophy s. J. R. astr. Soc. 24, 137-174 (1971). [Context Link]

8. Shaw, H. R., Hamilton, M. S. & Peck, D. L. Am. J. Sci. 277, 384-414 (1977). SFX Document Delivery

[Context Link]

9. Jaeger, J. C. in Basalts, the Polderv aart Treatise on Rocks of Balsaltic Composition Vol. 2 (eds Hess, H. H. &


11 sur 13 23/06/2009 14:16

Polderv aart, A.) 503-536 (Interscience, New York, 1967). [Context Link]

10. Langel, R. A. in Geomagnetism Vol. 1 (ed. Jacobs, J. A.) 249-512 (Academic, London, 1987). [Context Link]

11. Uy eda, S., Fuller, M. D., Belshe, J. C. & Girdler, R. W. J. geophy s. Res. 68, 279-291 (1963). [Context Link]

12. Stacey , F. D. & Banerjee, S. K. The Phy sical Principles of Rock Magnetism (Elsev ier, Amsterdam, 1974).

[Context Link]

13. Dunlop, D. J. J. geophy s. Res. 91, 9569-9584 (1986). [Context Link]

14. Bolshakov , A. S. & Shcherbakov a, V. V. Izv estiy a, Phy s. Solid Earth 15, 111-116 (1979). [Context Link]

15. Dodson, M. H. & McClelland-Brown, E. J. geophy s. Res. 85, 2628-2637 (1980). [Context Link]

16. Camps, P., Prev ot, M. & Coe, R. S. C.r. Acad. Sci., Paris 320, Ser. Ila (9) (in the press). [Context Link]

17. Audunsson, H. & Lev i, S. J. geophy s. Res. 93, 3480-3496 (1988). [Context Link]

18. Stasiuk, M. V., Jaupart, C. & Sparks, R. S. J. Geology 21, 335-338 (1993). [Context Link]

19. Kristjansson, L. Geophy s. J. R. astr. Soc. 80, 57-71 (1985). [Context Link]

20. Bogue, S. W. & Merill, R. T. A. Rev . Earth planet. Sci. 20, 181-219 (1992). [Context Link]

21. Starchenko, S. V. & Shcherbakov , V. P. Dokl. Akad. Nauk. SSSR 321, 69-74 (1992). SFX Document

Delivery [Context Link]

22. Runcorn, S. K. Trans. Am. geophy s. Un. 36, 191-198 (1955). [Context Link]

23. McDonald, K. L. J. geophy s. Res. 62, 117-141 (1957). [Context Link]

24. Constable, S. J. Geomag. Geoelect. 45, 707-728 (1993). [Context Link]

25. Bahr, K., Olsen, N. & Shankland, T. J. Geophy s. Res. Lett. 20, 2937-2940 (1993). [Context Link]

26. Schultz, A., Kurtz, R. D., Chav e, A. D. & Jones, A. G. Geophy s. Res. Lett. 20, 2941-2944 (1993). [Context Link]

27. McLeod, M. G. J. geophy s. Res. 99, 13577-13590 (1994). [Context Link]

28. Shankland, T. J., Pey ronneau, J. & Poirier, J.-P. Nature 366, 453-455 (1993). SFX Bibliographic Links

Document Delivery [Context Link]

29. Li, X. & Jeanloz, R. J. J. geophy s. Res. 95, 5067-5078 (1990). [Context Link]

30. Li, X. & Jeanloz, R. J. J. geophy s. Res. 96, 6113-6120 (1991). [Context Link]

31. Backus, G. E. Geophy s. J. R. astr. Soc. 74, 713-746 (1983). [Context Link]


12 sur 13 23/06/2009 14:16

32. Gubbins, D. & Roberts, P. H. in Geomagnetism Vol. 2 (ed. Jacobs, J. A.) (Academic, London, 1987). [Context

Link]

33. Roberts, R. G. Geophy s. J. R. astr. Soc. 78, 547-572 (1984). [Context Link]

34. Roberts, R. G. Geophy s. J. R. astr. Soc. 85, 583-600 (1986). [Context Link]

35. Schultz, A. & Larsen, J. C. Geophy s. J. R. astr. Soc. 88, 733-762 (1987). [Context Link]

36. Stacey , F. D. Phy sics of the Earth (Brookf ield, Brisbane, 1992). [Context Link]

37. Jeanloz, R. A. Rev . Earth planet. Sci. 18, 357-386 (1990). [Context Link]

38. Fisher, R. A. Proc. R. Soc. A217, 295-305 (1953). [Context Link]

Copyright (c) 2000-2009 Ovid Technologies, Inc.

Tout accès ou utilisation d’OvidSP implique que vous acceptez de respecter les termes et conditions, ainsi que toutes les lois afférentes. Si vous

n’êtes pas d’accord avec ces termes vous ne devez pas utiliser ce site.

Version: OvidSP_UI02.01.02.102, SourceID 40419


13 sur 13 23/06/2009 14:16

New evidence for extraordinarily rapid change of the geomagnetic field during a reversal

Documents

Transcript of New evidence for extraordinarily rapid change of the geomagnetic field during a reversal