Bent silicon crystal in the Laue geometry to resolve x-ray fluorescence for x-ray absorption...

REVIEW OF SCIENTIFIC INSTRUMENTS VOLUME 74, NUMBER 11 NOVEMBER 2003

Bent silicon crystal in the Laue geometry to resolve x-ray fluorescencefor x-ray absorption spectroscopy

A. J. Kropf,a) R. J. Finch, J. A. Fortner, and S. AaseChemical Engineering Division, Argonne National Laboratory, Argonne, Illinois 60439

C. Karanfil, C. U. Segre, J. Terry, G. Bunker, and L. D. ChapmanIllinois Institute of Technology, Chicago, Illinois 60616

~Received 3 April 2003; accepted 17 August 2003!

A highly strained, curved silicon crystal in the Laue geometry has been used as a large-area x-rayfluorescence analyzer for x-ray absorption spectroscopy. The analyzer is able to resolve theLafluorescence lines for neighboring actinide elements. A large gain in the signal to background ratiohas been demonstrated for small quantities of Np in the presence of U, with the U fluorescence peakapproaching 1000 times the magnitude of the off-peak background. ©2003 American Institute ofPhysics. @DOI: 10.1063/1.1618014#

mpratythvaorrooeaicoa

apLa

re

seithr

eid

r

s

nhelto

trolr

geck-st-ionor

it-

n-

is

r ofd xre-the

t. Ale-ma

INTRODUCTION

Measuring x-ray fluorescence from a sample is a comon method of obtaining x-ray absorption spectrosco~XAS! data from dilute samples and other sample configutions for which the transmission absorption measuremenimpractical. The primary limitation to using XAS for maninteresting environmental and biological problems ispresence of a high level of background radiation from ariety of sources, such as elements with lower energy abstion edges, radioactive elements, or scattered radiation fother beam interactions with the sample. Several methhave been developed to address these limitations, buthas its own strengths and weaknesses. A highly bent silx-ray analyzer in the Laue geometry has some notablevantages compared to other detection methods. This pdiscusses experiments that demonstrate how a bentanalyzer~BLA ! may be used to resolve theL fluorescence xrays of neighboring actinide elements, permitting measuments that would be either difficult or impossible using othtechniques.

ANALYZER DESIGN

The design concept of the BLA has been discuspreviously.1,2 The essential property of such an analyzerthat bending the crystal solves the mismatch betweennarrow acceptance angle of a perfect crystal and the ladivergence of fluorescence from the sample.1 The shapeneeded to preserve the correct angle of incidence betwthe fluorescence x rays and the diffraction planes over a wangle is a logarithmic spiral.3,4 This is represented in polacoordinates by

r ~u!5r0 cos~x2uB!exp@ tan~x2uB!u#, ~1!

wherer0 is the bending radius of the crystal atu50, x is theasymmetry angle between the reflecting planes and the

a!Author to whom correspondence should be addressed; [email protected]

4690034-6748/2003/74(11)/4696/7/$20.00

Downloaded 19 Sep 2008 to 134.253.26.11. Redistribution subject to AI

-y-is

e-p-mdschn

d-erue

e-r

dse

ge

ene

ur-

face normal, anduB is the Bragg angle of the reflectioplanes for the x-ray wavelength of interest. Except for tcase in whichx is equal touB and the logarithmic spirabecomes a circular section, the correct form is difficultproduce accurately with dynamical bending.1 Therefore, wehave used a fixed bender form, a computer-numerical-con~CNC!-milled aluminum block, over which the silicon wafehas been bent.

In addition to matching the Bragg condition over a larangle, bending the crystal increases the bandwidth, or roing curve width, making the analyzer nearly an ideal adjuable bandpass filter, with the exceptions of only absorptand a lowered reflection efficiency for extreme bending. Fa highly strained silicon wafer~in the kinematical limit!, theexpected resolution of the BLA for a reflection can be wrten as

Du~T!5T

r F tan~x6uB!11

2~11n!sin 2x

7tanuB~cos2 x2n sin2 x!G , ~2!

whereT is the thickness of the wafer andn is the Poissonratio.5 This formula clearly breaks down asr approachesinfinity ~i.e., a flat crystal! and the bandwidth,Du, ap-proaches the intrinsic linewidth of a perfect crystal. The eergy resolution is

DE

E5Du cot~uB!. ~3!

Since the BLA is nonfocusing, a large area detectorrequired. The diffracted x-rays are scattered by 2uB relativeto the x rays incident upon the analyzer. For an analyzeany reasonable size, this results in overlap of the diffracterays with the x rays coming directly from the sample. Thefore, Soller slits must be placed between the analyzer anddetector to absorb radiation from the sample source poinkey to the BLA’s ultimate success lies in the correct impmentation of the Soller slits.il:

6 © 2003 American Institute of Physics

P license or copyright; see http://rsi.aip.org/rsi/copyright.jsp

thp

end-uninio

ghen

cetorcnder

einngthned

elimedlen

engie

th

dthtr2

tarrg

fofo

anthr

ndeavean

thebe

rption

inof

en-aue

nceil-b-to

LAup

m-re-

c-les,

thee-ount

se

xi-

ofeakisthec-l.A:

--y-entx-ofeatory.-Rh-

t of-ed

4697Rev. Sci. Instrum., Vol. 74, No. 11, November 2003 Bent silicon Laue crystal for XAFS

ANALYZER SOLUTIONS FOR FLUORESCENCE X-RAY ABSORPTION FINE-STRUCTURESPECTROSCOPY „XAFS…

There are many detection methods that fill niches inx-ray absorption spectroscopist’s toolbox. A simple aproach, which is highly effective in many cases, is to usfluorescence ion chamber together with Z-1 or Z-2 filters aSoller slits to reduce the elastic and Compton scattererays entering the ion chamber.6 This technique has the advantages of large solid angle, essentially unlimited corate, and moderately high efficiency, without imposing strgent limits on the source spot. Due to the use of anchamber~or a scintillation detector in current mode!, theStern–Heald configuration suffers from a lack of enerresolution, making it very difficult to measure spectra in tpresence of lower energy background radiation that caneffectively be filtered.

One solution to the problem of multiple fluorescenlines is a multielement solid-state detector. These detechave the advantages of high efficiency, moderate eneresolution, and the ability to collect the entire fluorescenspectrum simultaneously. However, even at secogeneration synchrotron sources, these detectors are genoperated at near saturation~,1 Mcps/channel!. The currentgeneration of solid-state germanium detectors is ovwhelmed by the photon flux available at third-generationsertion device sources. Count-rate limits notwithstandithese detectors are an extremely useful tool; as suchprovide a benchmark that must be well surpassed by anydetector in order for the new detector to be considerevaluable addition.

Multilayer mirror analyzers have recently been devoped as an alternative way to overcome the count-rate ltation of solid-state detectors.7 These analyzers are best usfor low-energy XAFS~,12 keV! since the decreasing angof incidence at higher energy reduces the angular acceptaThe energy resolution, which is reasonably good at lowergies, is poorer than a solid-state detector at higher enerA multilayer mirror analyzer imposes strict limitations on thsource geometry. For high resolution and efficiency,source must be a line or a point.

An alternative approach is to design a narrow bandwidetector around diffractive optics. The traditional approacto use Bragg diffractive optics in the Johannson geomewhich places a crystal, with a surface ground to radiusRand then bent to radiusR, on a Rowland circle of radiusR. Ifproperly made, resolution approaching the intrinsic crysbandwidth can be achieved at largeR. These detectors suffefrom decreasing solid-angle coverage with increasing eneas well as posing a difficult manufacturing processachieving a correctly ground surface. The mechanismmaintaining the proper alignment of the sample, crystal,detector can be complicated and relatively large. Onother hand, one crystal may be used for quite a wide enerange, except when designed for near back-reflection cotions. While excellent for very high-resolution work, thsolid angle is typically very small, reducing the practicefficiency of a spectroscopic analyzer. A Bragg diffractioptic can impose even tighter limits on the source size


e-adx

t-n

y

ot

rsgye-

ally

r--,eywa

-i-

ce.-

es.

e

hisy,

l

yrrde

gyi-

l

d

shape than multilayer mirrors. On the other hand, sinceBragg analyzer is focusing, much scattered radiation caneliminated with minimal shielding.

At first glance a BLA would be best suited for higheenergy fluorescence because of the unavoidable absorof the silicon wafer, making multilayer mirrors~or Braggoptics! and Laue diffractive techniques complementarytheir energy coverage. However, with the availabilityhighly polished thin wafers~,50 mm!, efficient analyzerscan be made for x-ray fluorescence as low as 5 keV. A geral comparison between the curved Bragg and the bent Lgeometry is difficult given the large parameter space. Sithe highly bent silicon BLA increases bandwidth and tunabity at the expense of reflectivity, when also considering asorption, the BLA would seem to compare less favorablythe Bragg analyzer. Nevertheless, the efficiency of the Bcan be kept above 25% while the larger solid-angle makesthe difference. In order for a Bragg analyzer to attain a coparable solid-angle requires large crystals, higher-orderflections and smallR, reducing the reflectivity significantlybelow one in many cases.5

EXPERIMENT

Measuring theL-edge fluorescence XAFS in mixed atinide samples is a challenging problem. For these sampexcellent energy resolution is required, as theLa1 fluores-cence lines for U, Np, and Pu are separated only by;350eV; theLa2 lines are separated by less than 150 eV fromneighboringLa1 lines. The highest resolution solid-state dtectors are not capable of resolving these lines at useful crates~i.e., short pulse-shaping times!. With very long pulse-shaping times~.4 ms!, one can reduce the electronic noiand potentially achieve a full width at half maximum~FWHM! as narrow as 220 eV, but at the cost of low mamum count rates~less than 50 kcps/channel!. This resolutionis still inadequate for resolving the fluorescence linesneighboring actinides, since the Gaussian tails of the pwill still constitute a large background signal. When oneattempting to discriminate against a large background,detector will spend an overwhelming fraction of the colletion time measuring photons that are ultimately not usefu

We have studied two series of samples using the BL~1! a surrogate Y2O3 /SrO mixture for testing and calibration, and~2! Np in a U3O8 matrix.8 Once we have demonstrated sufficient sensitivity, we plan to study corrosion laers from drip tests on approved testing materials of spuranium oxide nuclear fuel that are primarily uranium oides, but which also potentially contain small quantitiestransuranic elements.9–11 All measurements reported wercarried out on the MR-CAT undulator beamline, 10-ID,the Advanced Photon Source, Argonne National LaboratA cryogenically cooled Si~111! double-crystal monochromator was used to select the incident energy. Either ancoated or an uncoated ultralow-expansion~ULE! glass mir-ror was used to reject the higher order harmonic contenthe beam.12 Kirkpatrick–Baez mirrors, positioned downstream of the harmonic rejection mirror, provided a focusbeam smaller than 10mm310mm for configuration B


or 14.1

4698 Rev. Sci. Instrum., Vol. 74, No. 11, November 2003 Kropf et al.

Downloaded 19 Se

TABLE I. BLA configurations used for the measurements. The resolution was measured and calculated fkeV x rays.

Analyzerdistance~mm!

Si waferthickness

~mm!Detector

~number of units!Beammask

Solid angle~crystal area!

~sr! Soller slits

Measuredresolution

~eV! ~FWHM!

Calculatedresolution

~eV!

A 180 200 ion chamber pinhole 0.040 No 145 114B 330 140 scintillation~1! focused 0.003 No 75 43C 201 140 scintillation~2!/

germanium~9!pinhole 0.091 Yes 105 68

m

, t

.3

xtbco

io

te-a

.0

va

isin

fosea

thpur

glu

nr

d

rgt

t the

yslar

ac-rgyer-

de-thehat

theSr

n

~Table I!, while only x-ray slits masked the unfocused beafor the other configurations.

Several Laue analyzers were compared. In all casesSi wafer had a 100& surface normal and the111& planeswere used as the diffracting planes, resulting in a 35asymmetry angle. The typical area was a 30 mm3120 mmrectangular window cut into the bender form to allow therays to pass through to the detector. Table I describesdifferent configurations. The direct x-rays were rejectedeither Soller slits or physical separation by limiting the aceptance of the analyzer by means of a lead mask. In cfigurations A and C, the limited area of the detectors~mul-tiple scintillation detectors, solid-state Ge detectors, or anchamber with a 7.5 cm diam window! resulted in only afraction of the total analyzer area being used. The calcularesolution is from Eq.~2!, while the actual resolution is measured as in a real experiment, collecting x rays acrossextended two-dimensional~2D! region of the analyzer face

In configuration C, for which Soller slits were used, 7mm long stainless steel slits were spaced at 1° inter~about 3.6 mm at the analyzer center!. Figure 1 shows apicture of the analyzer. The motion-control system consof a 25 mm travel, motor-controlled slide for movementthe focal direction, analyzeru motion with smaller than0.0008° step size, and fine control~,0.01°! of the Soller slitangle.

RESULTS

Figure 2 shows a scan of intensity versus energyconfiguration C~sample series 2! as the analyzer angle iscanned to diffract energies greater than the incident benergy (E.17 700 eV) to below the ULa lines. The samplewas 0.5% Np~by weight! in U3O8. The elastically scatteredpeak is small, a result of positioning the analyzer alongx-ray polarization vector to minimize the elastic and Comton scattering components. Below the elastic energy, thenium Lb lines sit on the Compton background. ThreeLblines are resolved in this plot; a small fourth peak is sugested at 1.05°. Configuration B, with higher energy resotion, resolves the small fourth ULb line ~not shown!. TheNp La1 peak is prominent, even at this low concentratioand the ULa1 andLa2 peaks are resolved. The FWHM fothis scan in the region of the Np and ULa fluorescence linesis about 105 eV. The ULa1 peak has a slightly reduceamplitude due to saturation of the scintillation detector~op-erated in the photon-counting mode!. The optimal analyzer-to-sample distance was satisfied only over a small enerange, and the focal distance was not scanned along with

p 2008 to 134.253.26.11. Redistribution subject to AI

he

°

hey-n-

n

d

n

ls

ts

r

m

e-a-

--

,

yhe

angle. Consequently, the analyzer was positioned so thahighest resolution was achieved for the NpLa lines, notice-ably sacrificing resolution for the elastically scattered x raand U Lb fluorescence lines. Nonetheless, this particubent crystal can clearly be used over at least a 5 keV range,although perhaps not over the entire length. This is of prtical interest, since the only way to cover a broad enerange is to manufacture analyzer/slit combinations with ovlapping energy ranges.

For comparison, two spectra of the Y2O3 /SrO mixture,one taken with a conventional single-element germaniumtector using a 0.5ms pulse-shaping time and the other withe BLA in configuration C, are shown in Fig. 3. Clearly, thpeak width from the Ge detector is much larger. Somewsurprisingly, even though the Sr and YKa lines are wellseparated for both detectors, the background rejection ofBLA is better. The resolution of the Ge detector for theKa fluorescence line is about 380 eV~0.5 ms digital pulseshaping!. The BLA resolution is good enough to show a

FIG. 1. Photograph of the bent Laue analyzer.


erta

e

nor;

edtivi-

herhereore

eV,

redtotherlythea-

redt

l-ent.efi-thed,

r-for

enndbleen-

g inelyem-theesn-theutp-,rtergh2°

t


asymmetry in theKa peak, reflecting theKa1 and Ka2

lines, although they are not resolved.The resolution and reflectivity of the bent silicon waf

were checked for configuration C by placing the bent crysin the direct beam~,50 mm beam size! at u50° @from Eq.~1!# and measuring the extinction of the beam as the incidbeam energy was scanned~Fig. 4!. This is equivalent torocking the analyzer angle, but ensures that beam positiothe crystal does not change. Without considering the abstion of the silicon, the efficiency is 46% for configuration Cthat is, the crystal diffracted 46% of everything not absorbWhen considering the absorption of the silicon, the reflecity is only 26%; the silicon wafer absorbs 43% of the incdent beam.

When looking at the extinction from a small spot on twafer, as in Fig. 4, the resolution is much better than fohighly divergent incident beam. This is partially due to tlateral divergence accepted by the analyzer in the fluocence measurement, but it is also due, in part, to form erreither from anticlastic bending or from form errors in th

FIG. 2. Bent Laue analyzer spectrum for 1 part Np, 160 parts U~configu-ration C!; incident beam energy: 17 700 eV.

FIG. 3. Fluorescence spectrum for Y2O3 /SrO comparing a single elemenfrom a germanium detector to the bent Laue analyzer, configuration C.


l

nt

onp-

.-

a

s-s,

bender itself. The measured resolution was about 70compared to 65 eV, calculated from Eq.~2!. For comparison,a 40mm wafer was bent over the same form and measuwith a 25mm incident beam size; the resolution improved25 eV, compared to an calculated value of 17 eV, andabsorption decreased, improving the reflectivity to nea35%. Not considering the absorption by the silicon wafer,reflectivity was 47%, nearly the same as for the thicker wfer, as expected.

The measured value for the reflectivity can be compato the multilamellar model for Laue diffraction of a bencrystal. Using theXOP ~version 2.0! software package,13,14

the reflectivity for this crystal, ignoring absorption, was caculated to be 45%, in good agreement with the measuremIt is worth noting that since the crystal is diffracting in thkinematical limit, the absorption-free Laue reflection efciency for a less extreme bend could be larger than 50%,typical maximum for Laue reflection from an unstraineperfect crystal.

Proper alignment of the Soller slits is also very impotant. Figure 5 shows the intensity versus Soller slit angleconfiguration C for 1° slits and 2° slits~every other slotfilled!. It can be seen that if the slits are misaligned by ev0.5°, the signal will suffer, reducing the peak-to-backgrouratio and requiring longer collection times for a compararesult. For a particular analyzer to be used over a largeergy range, the slits must be spaced far apart, resultininsufficient background rejection, or they must be closspaced and movable. In this instrument, the entire slit assbly is rotated around a point on the analyzer surface nearcenter of the crystal. Although the virtual source point movwith energy, the slit spacing allows for some nonideal aligment. Extrapolating the intensities to zero shows that for1° slits, the diffracted beam is not entirely cut off until abo2.5° tilt from the maximal intensity. For 2° slits, this accetance is about 4.5°. Since 2uB is in the neighborhood of 16°less effective slits, either from greater spacing or sholength, would run the risk of passing the direct beam throuto the detectors. For this energy range, between 1° and

FIG. 4. Extinction of beam vs incident energy for 40 and 140mm siliconwafers.


wo

acrly

escr

wehigey

ded

me

thepo-

isand

ure-

gle-

tion

tore-thein-reso-beion-

forgeshean-ri-40

so-on

ec-m-ion-thent

on,


separation appears to be a reasonable compromise betbackground rejection and tunability. For lower energy flurescence lines with a largeruB , the slit spacing could beincreased or the slit length shortened.

Figure 6 shows the XAFS spectrum of the Np in U3O8

specimen, the sum of 5 of measurements. The signal to bground is 3.5:1, even though the Np peak is still cleasitting on the tail of the uranium fluorescence peak~as in Fig.2!. This ratio means that the uranium fluorescence has breduced by about 550 compared to the unfiltered beam. Apractical consequence, since reasonably good XAFS specan generally be obtained in the fluorescence mode fosignal-to-background ratio of 1:2, given enough time,could theoretically measure XAFS with no change to tconfiguration for a 1:1100 Np to U ratio. From the edposition and shape, the Np appears to be predominantlthe 14 oxidation state. Figure 7 shows thek2-weighted ex-tended XAFS~EXAFS! with the background removed anthe single-shell fit. Fitting the EXAFS shows six-coordinatNp–O, with a bond distance of 2.2760.02 Å, consistent

FIG. 5. Intensity vs angle for Soller slits~configuration C!. Open circles for1° slits and closed squares for 2° slits.

FIG. 6. Absorption edge for Np in U3O8 ~configuration C!.


een-

k-

ena

traa

s

in

with results for higher Np concentration samples in the saseries, described by Finch and Kropf.8

DISCUSSION

To meet the anticipated requirements of measuringx-ray absorption spectra of mixed actinide samples incorrating a low level of Np in a U matrix, both resolution andbackground rejection are critical parameters. Resolutionrequired to adequately separate the fluorescence lines,background rejection is required to ensure that a measment can be made in a reasonable time, if at all.

We can solve the problem of resolution by optimizinthe choice of Si wafer thickness, diffraction planes, sampto-analyzer distance, and asymmetry angle. In configuraC, we have used a 140mm thick wafer with a bending radiusat u50 of 235 mm. This provides adequate resolutionhave the Np fluorescence line appear on a significantlyduced background of U fluorescence, although not yet toultimate background level. The bending radius could becreased, at the expense of solid-angle coverage, and thelution would be improved. The loss of solid angle wouldpartially compensated for by an increase in the reflectefficiency. It is not yet clear if the CNC-milled form is accurate enough to allow higher resolution or if the tendencyanticlastic bending would eliminate the possible advantaof a larger bending radius. The other way to improve tresolution would be to use a thinner wafer. In this case,ticlastic bending will certainly be worse. It was our expeence, using a bender form with this large a window, that amm wafer was too thin and form errors dominated the relution. These form errors were easily visible in the reflectifrom the bent crystal’s surface.

The second critical parameter, that of background rejtion or maximum peak-to-background ratio, can be accoplished either by shielding or increased analyzer reflectefficiency. Any increase in the reflection efficiency will instantly increase the peak-to-background ratio, increasingsensitivity of the instrument. This is a very good argumefor increasing the bending radius to improve the resoluti

FIG. 7. Isolatedk2 EXAFS data for Np in U3O8 ~BLA tuned to NpLa1) ~fitis betweenk52.5 and 8.5 Å21, single shell!.


bleffisry,n,d

,ls

fore

l oagaeat

orb

unsll

ncmao

tivavonreFS

mnThe

thh

ma

u-ae

athrea

na-ill

ivity.inglong.bebe

er-al-ea-

tosi-evi-s toententess

era-stededm,ys-

om-plen.

slitsbothr ofion. A

nceed xdtor

to-

dtione Uap-

on.ri-

ac-ce.omthe

tedded


but only for dilute samples with the signal level comparato or smaller than the background. As the reflection eciency approaches 50%, not considering absorption, aconfigurations B and C, further increases in the bendingdius will garner diminishing returns in improved efficienclimiting the maximum practical bending radius. In additiothe resulting decrease in the collected solid angle couldcrease the signal-to-background ratio. However, as longall extraneous background sources have been eliminatedbackground due to x rays incident upon the sample will adecrease at anR22 rate.

Several sources of background must be eliminatedthe best possible background rejection. First, the high-eneharmonics in the beam must be eliminated and not allowto hit the sample or to scatter about the hutch. This leveharmonic rejection can typically be achieved only withmirror, which will result in a great deal of x-ray scatterinfrom the mirror surface. Shielding coupled with physicseparation are the best means of eliminating this sourcbackground. A test for the effectiveness of the upstreshielding is to close the beam-defining slits and measurebackground in the detectors. If the signal is not zero, mshielding will be beneficial. The sharpest cutoff can be otained from a low atomic-number surface, such as thecoated glass. Second, air scatter between the beam-defislits and the sample can potentially enter the detectorproperly positioned shielding has not been installed. Finathe most difficult problem is the scatter and fluorescefrom the sample itself, which will scatter or refluoresce frothe Soller slits or scatter from the silicon analyzer. Onlywell-designed set of Soller slits can reject this sourcebackground. The slits described have proven very effecin this regard; peak-to-valley ratios exceeding 1000:1 hbeen achieved. Given this level of background rejectienough time, and a modestly improved resolution compato Fig. 2, it should be possible to measure useful XAspectra of 1 part Np in 2000 parts U.

Only after the background has been reduced to minilevels is it worth considering ways to improve the total courate and, thus, reduce the length of the measurement.most effective method is to focus a larger fraction of tbeam to a spot size smaller than 0.1 mm30.1 mm. In testssuch as the extinction measurement performed for Fig. 4,effect of the beam size on resolution was probed. In tmeasurement, the maximum beam size in the dispersiverection can be approximated by

h5r tan~Du!>rDu, ~4!

wherer is calculated from Eq.~2! andDu is from Eq.~3!. Asthis limit is approached, the square extinction curve becaa triangle, but the maximum extinction was still observedthe tip of the triangle, while only a small penalty in resoltion was observed. For this analyzer, the maximum besize is nearly the thickness of the crystal. Under real msurement conditions, the lateral~nondispersive! extent of thecrystal must be considered as well, since the Bragg diffrtion condition is only exactly satisfied for one wavelengalong a line from one end of the crystal to the other. Thefore, the required beam size will be significantly less th


-in

a-

e-astheo

rgydf

lofmhee-

n-ingif

y,e

fee,d

althe

eisdi-

et

ma-

c-

-n

that calculated from Eq.~4!, depending on the width of theanalyzer and how far away from the design point the alyzer is being used. A beam size larger than the ideal wresult in both decreased resolution and decreased reflectOn the other hand, in the case of a strongly absorbsample, a line source several mm wide could be used, asas the analyzer is oriented normal to the sample surface

After focusing the beam, multiple analyzers can thenconsidered, although the additional analyzers cannotplaced in positions of minimal elastic and Compton scatting. If the additional analyzers do not have a similar signto-background ratio, they may not effectively add to the msurement statistics. Furthermore, it may not be possibleplace additional analyzers in a geometrically optimal potions. The requirements for the source were discussed prously. The geometry in the measurements reported waplace the sample at a 45° angle with respect to the incidbeam, and to orient the analyzer perpendicular to the incidbeam. In this case, the source was a short line of thicknless than 10mm. In order to bring both ends of the analyzinto the diffraction condition for one wavelength simultneously, the sample to analyzer distance had to be adjuonly to a precision of several hundred microns. The relaxrequirement for the depth of field is due to the small beathe thickness of the wafer, and the limited extent of the crtal in the dispersive direction. Continuing from Eq.~4!, con-sidering a line or point source of thickness much less thanh,the depth of field in this orientation is

d5h/sin~a/2!>rDu/sin~a/2!, ~5!

or about 650mm for an analyzer acceptancea525°. If ad-ditional analyzers are not oriented in such a favorable geetry, then the penetration depth of the beam into the sammust be smaller in order to satisfy the depth of field criterio

Both stainless steel and tin-covered stainless steelwere tested. The same stainless steel slits were used ininstances to fill the slots. In the second case, a thin layetin foil was taped to the stainless steel slit over the positwhere the direct beam from the sample would hit the slitsmultielement Ge detector~configuration C! was placed be-hind the slits as the detector. Unexpectedly, no differewas observed in the fluorescence spectrum of the detectrays ~i.e., Fe fluorescence from the slits was not observe!.However, with the increased resolution of the Ge deteccompared to the scintillation detector, the signal-background ratio for the Y in the Y2O3 /SrO sample wasimproved by a factor of 2; i.e., only half of the backgrounobserved was due to Sr fluorescence that the scintilladetectors were unable to discriminate against. Since thand Np fluorescence lines are so closely spaced, thisproach may not produce similar results for that combinati

An alternative approach may be taken for further expementation with background rejection. One could manufture slits with a triangular or sawtooth pattern on the faThis pattern would prevent slit scatter and fluorescence frreaching the detector. Given the comparison betweenstainless steel and tin slits, this approach may be of limiutility. However, for the higher activity samples envisionein the future, a significant amount of lead may be requir


ar

e

rss

deacb

o

reioe-

rte

n

xthe

poanre

rea

-of

noctnicndebndeac

lta-r ofer ape-uf-

ad-

ero-tonrgy,der

-nt

gre,

U.:

, J.

oc.

9:


between the sample and the detector to reduce the bground due to sample radioactivity. Machining an appropate sawtooth pattern into thick lead slits would not possignificant challenge.

We suggest two other applications of the BLA. The fialternative application is for highly crystalline sampleWhen it is not possible to grind a sample to a fine powand when it is not practical to spin the sample, the BLA mbe used to eliminate diffraction peaks from fluorescenmeasurements. Reducing diffraction peak amplitudesthree orders of magnitude will often reduce their impactthe data to insignificance.

Since the BLA has the energy resolution to isolatesingle fluorescence line, it may also be possible to addanother difficult problem: that of a higher energy absorptedge, from elementZ11, limiting the data range. When thBLA was tuned to the ULa1 fluorescence line and the energy was scanned over the Np edge, often an inve~negative-going! absorption edge was observed at the NpL3

edge energy. The magnitude was typically reduced, exteing the usable data range by a few tenths of an Å21, but theproblem was not eliminated. This occurred because the eabsorption of the Np reduced the x-ray flux incident uponU in the sample. This effect will clearly be affected by thsample thickness, with thinner samples suffering less. Atential solution may be to measure a positive-going edgeuse that data to precisely cancel the inverted edge, theextending the usable data range by several Å21.

A bent Laue analyzer with a peak-to-valley ratio of mothan 1000:1 extends the useful spectroscopy limits bymuch as a factor of 100 times compared to current statethe-art multielement germanium detectors, when usedmixed actinides. In general, even when high resolution isrequired, the BLA can offer modestly improved performancompared to a germanium detector, dependent solely onbackground rejection achievable in a particular experimeFurthermore, given a focused beam from an insertion devthe BLA will not be subject to the same count rate limits, aresulting nonlinearities, as the germanium detector, therpotentially offering improved data acquisition times afewer corrections to the data. The tradeoff is increasedperiment complexity and the restriction to single-energy


ck-i-a

t.ryey

n

assn

d

d-

trae

-d

by

sf-

ort

ehet.e,

y

x--

quisition unless both systems can be employed simuneously. It remains to be seen whether a small numbegeneral-purpose analyzers can be manufactured to covwide energy range, or if each experiment will require a scially designed analyzer to provide a performance level sficiently greater than a solid-state detector to justify theditional experiment complexity.

ACKNOWLEDGMENTS

The authors especially thank Thomas Torres~IIT ! andAdvanced Design Consulting, Inc. for their efforts in thproduction of the analyzer/Soller slit assemblies and the ptotype mechanical systems. Use of the Advanced PhoSource was supported by the U.S. Department of EneOffice of Science, Office of Basic Energy Sciences, unContract No. W-31-109-Eng-38. Work performed at MRCAT is supported, in part, by funding from the Departmeof Energy under Grant No. DEFG0200ER45811.

1Z. Zhong, D. Chapman, B. Bunker, G. Bunker, R. Fischetti, and C. SeJ. Synchrotron Radiat.6, 212 ~1999!.

2C. Karanfil, Z. Zhong, L. D. Chapman, R. Fischetti, G. B. Bunker, C.Segre, and B. A. Bunker,CP521, Synchrotron Radiation InstrumentationSRI99: 11th U.S. National Conference, edited by P. Pianettaet al. ~Ameri-can Institute of Physics, New York, 2000!, Vol. 521, p. 178.

3Y. Sakayanagi, Jpn. J. Appl. Phys., Part 221, L225 ~1982!.4D. B. Wittry, W. Z. Chang, and R. Y. Li, J. Appl. Phys.74, 3534~1993!.5E. Erola, V. Etelaniemi, P. Suortti, P. Pattison, and W. ThomlinsonAppl. Crystallogr.23, 35 ~1990!.

6E. A. Stern and S. M. Heald, Rev. Sci. Instrum.50, 1579 ~1979!; E. A.Stern, B. A. Bunker, and S. M. Heald, Phys. Rev. B21, 5521~1980!.

7K. Zhang, G. Rosenbaum, and G. Bunker, J. Synchrotron Radiat.6, 220~1999!.

8R. J. Finch and A. J. Kropf, Mater. Res. Soc. Symp. Proc.757, 455~2003!.9P. A. Finn, Y. Tsai, and J. C. Cunnane, Mater. Res. Soc. Symp. Proc.713,607 ~2003!.

10R. J. Finch, E. C. Buck, P. A. Finn, and J. K. Bates, Mater. Res. SSymp. Proc.556, 431 ~1999!.

11J. A. Fortner, R. J. Finch, A. J. Kropf, and J. C. Cunnane,2003 Interna-tional High-Level Radioactive Waste Management Conference, March30–April 2, 2003, Las Vegas, Nevada~American Nuclear Society, 2003!.

12C. U. Segreet al., CP521, Synchrotron Radiation Instrumentation: SRI911th U.S. National Conference, edited by P. Pianettaet al. ~AmericanInstitute of Physics, New York, 2000!, Vol. 521, p. 419.

13M. Sanchez del Rio and R. J. Dejus, Proc. SPIE3448, 340 ~1998!.14M. Sanchez del Rio, C. Ferrero, and V. Mocella, Proc. SPIE3151, 312

~1997!.


Bent silicon crystal in the Laue geometry to resolve x-ray fluorescence for x-ray absorption...

Documents

Transcript of Bent silicon crystal in the Laue geometry to resolve x-ray fluorescence for x-ray absorption...