Nuclear Instruments

and Methods in Physics Research B 101 (1995) 493-498

__

LiYlOliW

Beam

B

Interactions with Materials&Atoms

!!!I!! ELSEWIER

A new device for XAFS data collection up to 2000 K (or 3700 K under vacuum)

*,

F. Farges a*b, J.-P. ItiC b,c, G. Fiquet d, D. Andrault e aLaboratoire de Physique et Mecanique des Geomateriaux, Universid Marne-la-Vallee (and URA CNRS 734), 2 allee de la butte verte, 93166 Noisy le Grand cedex, France b Laboratoire pour l’utilisation du rayonnement electromagnttique (LURE), Brit. 209 D, Universite Paris&d, 91405 Orsay cedex, France ’ Laboratoire dephysique de la mat&e condensee, Universite Parts 6, 4 place Jussiey 75005 Paris, France 4 Laboratoire de Geologic, Eeole Normale Superieure, 46 allee d’ Bake, 69041 Lyon cedex, France ’ Laboratoire des GPomateriau, Universite Paris 7 (and Institut de Physique du Globe de Parts and (IRA CNRS 734), 4 place Jussieu, 75005 Paris, France Received 27 January

1995; revised form received 14 March 1995

Abstract A new device to collect high- to ultra high temperature X-ray absorption fine structure (XAFS) data is presented. The experimental XAFS data is collected in the energy-dispersive mode using the heating wire technique. The X-ray beam is focused on a 500 pm 0 hole (drilled within a Pt,Rh,, heating wire) that contains the sample (100-300 pm thickness). The wire is heated by the Joule effect. The actual temperature is measured using an optical pyrometer and an electrical power-temperature calibration, based on 10 model compound melting points (T is f 20 K on the average). The maximum possible temperature is that of the wire melting point (- 2000 K for the Pt,Rh,, alloy). However, much higher temperatures can be achieved by the use of Ir, W, Ta (etc.) wires under inert atmosphere (to prevent the wire from oxidation by air). Data collection at all temperatures is rather fast (3 s/spectrum for a 3-13 A-’ k-domain used) and can be performed every 20 K. Therefore, in situ, high temperature kinetics can also be studied as a function of time such as phase transitions, oxidation reactions or melting phenomena. The study of GeO, polymorphs as a function of temperature will be shown to give an example of the possibilities offered by this technique.

1. Introduction High temperature X-ray absorption fine structure (XAPS) spectroscopy is a useful method that helps to understand the structure of molten phases as well as the structural mechanisms that initiate melting [l-5]. This method can also be used, for instance, to trace the evolution of the local structure in crystalline compounds during phase transitions induced by temperature changes [6], but also thermal expansion mechanisms [3,7,8]. Theoretical limitations of the XAFS method due to thermal agitation have now been overcome by the use of several theories [2,5,9] that correct for thermal-induced anharmonicity, like, among others, the cumulant expansion theory [3,4,7,8]. Also, a number of new high temperature devices have recently been used to collect high temperature XAFS

* Corresponding author. [email protected].

fax

+33

1 1 44

2487,

e-mail:

0168-583X/95/$09.50 0 1995 Elsevier Science B.V. All rights reserved SSDI 0168-583X(95)00496-3

spectra [5]. In addition to the review by Filipponi and Di Cicco [5], one can also quote some other furnaces, like the high-temperature fluorescence devices operating at SSRL (Stanford, California) [1,2,4]. In this study, the Mysen-Frantz [lo] heating wire technique has been developed to collect high- to ultra-high temperature (up to 3700 K [ll]) XAPS data in the transmission mode. This micro-furnace, originally designed for Raman diffusion and X-ray diffraction experiments, has been implemented on an energy dispersive XAFS spectrometer. This device allows the collection of a large number of high-quality X-ray absorption spectra (every 20 K and only in few seconds each). The resulting excellent reproducibility of these experiments (XAFS-derived interatomic distances are at the f 0.003 A reproducibility level) allows an accurate description of temperature-induced anharmonicity. Furthermore, time-resolved phase transitions can be followed in situ at high-temperature. To illustrate this, the high temperature behavior of GeO, polymorphs (e.g., rutile-, quartz-, molten and glassy modifications) is presented.

494

F. Farges ei al. /Nucl. Instr. and Meth. in Phys. Res. B 101 (1995) 493498

2. The heating wire cell The heating wire cell (Fig. 1) is that elaborated by Richet and coworkers [lO,ll]. A 70 mm long Pt,,Rhe,, wire has been used to collect data up to 2000 K (the nominal melting point of this alloy is 2070 K). A 0 500 pm hole has been drilled within a flattened section of the wire (1 cm long), in which the sample is positioned. Heating is provided by a stabilized power supply (10 V-80 A). The measurement of the actual temperature within the hole is determined using two techniques: optical pyrometry and Joule effect measurement, as explained now. The optical pyrometer (working at 0.605 pm wavelength) has been optically designed to operate at 45 mm from the heating wire. The diameter of the surface used to measure the relative emissivity of the platinum-rhodium alloy (ca. 0.33) is about 400 p,m. The uncertainty of the temperature measurement was found to be 20 K at 1400 K. The reproducibility and stability of the sample temperature is around 5 K at 1400 K. At 2000 K, gradients within the hole/sample are expected to increase significantly, but they were found to be lower than 10 K. The maximum temperature measurable with that pyrometer is 3000 K. The optical pyrometer used here cannot work properly below 900 K. For temperatures between 293 and 900 K, we calibrate the heating wire in temperature using high purity model compounds for which the congruent melting points are well-known (Fig. 2). The average uncertainty in the temperature measurement is +20 K when using this last method. These power-temperature calibrations have also been performed after the heating experiments to check for possible variations in the wire resistance during XAFS data collection. For instance, for a constant power of 30 W

CaMgSi,O 6 (1664 K).. /*

1 .)

,:’

i BaNO 3 (865 K)

ic acid (460 K) ,

0

10

lowest temperature measurable by IROPtica’ Pyromehy

, / , I I , /

20

30

, , , /

40

50

POWER (W) Fig. 2. Example of power temperature calibration obtained based on several high purity dehydrated model compounds. Melting temperatures

are taken from Ref. [ll].

delivered by the power supply, the actual temperature of an aged wire (used during 10 h) has decreased by ca. 100 K compared to that of a new one.

3. The XAFS spectrometer

used

X-ray absorption spectra are collected in the transmission mode on an energy-dispersive spectrometer (XAS 10 at LURE, Orsay [12]. The storage ring is operating at 1.8 GeV, together with 100-300 injected positron currents. The sample is positioned within the heating wire hole (Fig. 3), exactly where the X-rays are focused by a bent crystal

+

insulating

board

Fig. 1. Schematic drawing (from Ref. [ll]) of the high temperature cell with a silica glass tube for experiments under an inert atmosphere.

495

F. Farges et al. / Nucl. Instr. and Meth. in Phys. Res. B 101 (I 995) 493-498

shunt

0.0010

1100 A

Fig. 3. Principle of the heating-wire experiment on a energy-dispersive spectrometer. The X-rays are focused within the furnace thanks to a bent monochromator crystal (usually Si-(111) and Si-(311)). The intensity of the transmitted beam is measured with a 1024 array of diodes.

Si-111 or 311). With Si-111 crystal, the energy resolution is about 2-3 eV (mainly due to core-hole effects). The transmitted X-ray beam intensity is monitored using an array of 1024 photodiodes located behind the heating cell. To calculate the absorption coefficient, the incident X-ray intensity is measured by the same photodiode array but in the absence of sample. A reference foil (usually a metal) is used as a model compound for which a energy versus absorbance spectrum has been previously collected on a classical XAFS spectrometer (i.e., with a step-by-step monochromator). The comparison of the spectra for the model compound (pixels versus absorbance and energy versus absorbance) allows the derivation of the conversion relation between diodes pixel numbers and energies (usually, a simple first order polynomial). Typical energy range that can be probed at LURE is 5-20 keV (K edges from Ti up to MO; L,,, from Cs to Np). Prior to data collection, the sample powder is packed within the hole. Special care is needed to get a rather homogeneous sample that is not too thick or too thin (absorbance step between 1 and 3). The use of matrices (like boron nitride or equivalent) in which the sample is diluted can be used to adjust easily the transmission. However, in some case, this technique is not recommended because both compounds (matrix and sample) are reacting together, so we preferred to avoid such configuration. Each XAFS spectrum can be obtained within some tenths of a second at a given temperature. In order to increase the signal-to-noise ratio, 32 to 64 scans are averaged together, which makes a total data collection time of several seconds per averaged scan. (usually

We will show the example of data collected at Ge K-edge (11103 eV) in GeO, oxides. High temperature data was reduced using the cumulant expansion method [3,13], the curved wave theory [13] and the standard single-scattering formalism. Despite the multiple-scattering (MS) paths of the photoelectron cannot interfere with single-scattering (SS) paths for the closest pair-correlation on the Fourier Transform @I’), MS effects were estimated using the Feff 5 code [14] and the structure data for the rutile and the quartz modifications [15,16]. The importance of MS effects (in the EXAFS region) were investigated in the 2-5 A range from the central Ge in both compounds. The Feff calculation has been performed assuming first only SS paths and second all MS paths (this last model includes therefore the contribution from SS paths) of the EXAFS (Fig. 4). The difference in the Fourier transform (IT) betwten SS and MS models is negligible, particularly below 5 A from the central Ge (however, a clear MS feature is observable near 5.5 A on the IT) (note that ab initio Feff calculations (SS or MS model) are able to reproduce nicely the experimental data despite a discrepancy in the m-magnitude. This discrepancy is attributed to the presence of some porosity in the sample studied and the uncertainties on the Debye-Waller factors for the ab initio calculation). The influence of MS paths in the Ge K-edge EXAFS data for the rutile are therefore found to be negligible. Then, the use of the simpler single-scattering formalism to model the experimental Ge-0 and Ge-Ge pair correlations is authorized. The room temperature spec-

1

Fig. 4. Effect of multiple-scattering (MS) in the EXAFS region for the rutile polymorph of GeO,, at Ge K-edge, showing the limited MS effects in the experimental spectra. From top to bottom: (a) Fourier Transform (FT) considering only the single-scattering paths of the theoretical EXAFS; (b) ibid. but including also the multiple-scattering paths to the EXAFS; (c) m of the experimental EXAFS data. The similarities between curves (c) and (a) suggests that a simple single-scattering formalism is sufficient to model the experimental signal. Distances are uncorrected for the various backscattering phase-shifts. Calculations were performed using the Feff 5 [14] package and the crystal structure data for the rutile phase 1151. Similar conclusions can be drawn for the quartz modification.

496

F. Farges et al. / Nucl. Instr. and Meth. in Phys. Res. B 101 (1995) 493-498

s

3

159OK~ 13lUlK

2

%

k:

:z:: :: 1WOK 1220 K 1195 K 995K 695 K 295 K _ d

8 d

9;

11150

295 K 550 K 800K 1010 K 1250 K 1315 K 1320 K 1325 K 1400K 1450 K 15OOK

2

4

6

ENERGY (eV)

8

10

12

14

8 d

1

3

5

R (A)

k(A-‘)

Fig. 5. (a), (b), (c): XANES, normalized EXAFS and FT for selected temperatures collected at the Ge K-edge in GeO, rutile. Changes in the spectra are mostly related to increasing disorder effects (range 300-1300 K). Near 1325 + 20 K the Ge coordination decreases from 6 to 4 and the quartz phase appears. At temperatures higher than 1400 K, the liquid phase is obtained, with a short range structure similar to that measured at high temperature in the quartz phase.

is reduced to extract the empirical 0-backscattering amplitude and Ge-0 backscattering phase-shift functions. These functions are used to model the spectra collected at higher temperatures. All spectra were reduced considering the same parameters. tra

4. Samples studied Quartz GeO, (q-GeO,, P4,/mmm, Z = 2) is a reagent grade (Merck GmbH), 99% pure, in which Ge is 4-coordinated by 0 with Ge-0 distances of 1.739 + 0.001 A [16]. This phase is stable only between 1283 K and 1389 K [17], but can be quenched (metastable) easily at room temperature. In contrast, the rutile polymorph (r-GeO,, tetragonal, P3,21, Z = 3) constitutes the stable phase of the GeO, system at room temperature. At ambient pressure, the rutile stability field extends up to 1283 K [15,17,18]. In this phase, Ge is 6-coordinated by 0, with 6 Ge-0 distances between 1.87 and 1.90 i [15]. The rutile modification has been synthesized from the quartz phase, by adding trace amounts (0.05 mol.%) of L&O in the quartzGeO, before heating the mixture at 1200 K [17]. Trace amounts of lithium oxide in the quartz modification of GeO, are known to activate the quartz + rutile phase transition [17] and vice versa (this study). Without L&O, the quartz/rutile phase transitions are particularly sluggish [17-201. However, several lithium germanates were detected (by X-ray diffraction) in the rutile sample after heating, but their relative amounts are far negligible compared to that for r-GeO, (similarly, only trace amounts of the original quartz phase could be detected in this sample).

5. High temperature XAFS data for GeO, polymorphs We have collected 57 spectra between 300 and 1500 K within half an hour. Examples of normalized Ge K-edge X-ray absorption near edge structure (XANES), extended X-ray absorption fine structure (EXAFS) and their FT spectra for the GeO, polymorphs (quartz, rutile, liquid and glass) are presented on Figs. 5a, 5b and 5c, respectively, for selected temperatures comprised between 293 and 1500 K. With increasing temperature, a progressive loss of

f 200

600

1000

TEMPERATURE

1400

(K)

Fig. 6. Evolution of the average anharmonic Ge-0 distance as a function of temperature for the near 1325 K is attributed to a coordination change around Ge (6 + 4), probably related to a rutile + quartz phase transition.

F. Farges et al. /Nucl, Instr. and Meth. in Phys. Res. B 101 (1995) 493-498

2

R=(1.746fO.O03)A

B 9 0 :: 2

+ 1346-9 s + 1271-4 s +11&t-7s + +848-l s

2 :

+ 664-7 576-9 s + 414-7 s + 352-5 s + 317-o s +271-4s + 163-6 s + 55 -8 s + o-3 s

P

T = (1320 f 15) K

ENERGY (eV) Fig. 7. In situ XANEB evolution at 1325 K, as a function of time. Distances are derived from anharmonic curve-fitting of the XAFS oscillations [21]. Distances are consistent with the rutile + quartz phase transition known in this temperature range from other techniques.

next-nearest Ge neighbors contributions is observed. This signal has completely disappeared at temperatures higher than 1400 K. However, the main signal arising from oxygen first neighbors remains detectable up to 1500 K. The XAFS-derived anharmonic distance for the rutile phase is shown on Fig. 6 as a function of temperature. The progressive increase in the average Ge-0 distance (3001300 K range) in the GeO, polyhedra is consistent with an average linear thermal expansion coefficient of 13 f 2 X 10m6 K-’ (in contrast, harmonic models would result in an apparent bond contraction of ca. 0.03 A in this temperature range). Near 1300 f 30 K, the average Ge-0 distance decreases suddenly, suggesting major changes in the average local structure around Ge.

6. In situ observation ture

of a phase transition in tempera-

For temperatures around 1300 f 15 K, major changes in the Ge K-edge XANES for the rutile polymorph are observed as a function of time (Fig. 7). Anharmonic analysis of the EXAFS signal shows a de$rease in the averagt Ge-0 distance, from 1.906 f 0.003 A to 1.760 f 0.003 A at the constant temperature of 1320 f 15 K (Fig. 6). In parallel, we found that the number of O-first neighbors has decreased to 4.0 f 0.3 atoms. Also, the third cumulant shows a severe drop, from 6.4 f 0.3 X 10e3 k to quasi-negligible values, - 8 f 4 X lop7 K- ’ k. These variations in EXAFS-derived parameters are consistent

497

with a t61Ge -+ t41Ge transition, that can be related to the rutile + quartz modification described near 1300 K by X-ray diffraction, Raman and thermodynamic studies [17201. Usually these changes are sluggish [17-201 but the modification has been speed up by introduction of 0.05 mol% of Li,O in the GeO, at room temperature, as explained previously.

7. The liquid and glassy states At 1400 + 30 K, there is a sudden loss in the contribution of the Ge-Ge second neighbors (Fig. SC). However, the average Ge-0 distance for the 0 environment around Ge remain similar to that for quartz-GeO, (t41Ge-0 = 1.755 f 0.003 A). The number of first neighbors around the central atom does not decrease significantly during the melting of GeO, (around 3.8 f 0.3 oxygen atoms around Ge). This result appears in contrast with the observations reported for molten Pb [3], where ca. half of the original Pb-first neighbors could be observed by XAFS in the molten phase. During the melting of GeO,, no such structural inconsistency is found between the measured number of neighbors and the average Ge-0 distance (3.8 atoms with an average Ge-0 distance of 1.755 A). It appears here that the use of a quasi harmonic model for the [ 4]Ge-0 bond is also valid for the molten phase, as for the high temperature quartz modification (unfortunately, there is no other structural data available for molten GeO, to compare our XAFS data with). The use of the cumulant expansion formalism for Ge-0 bonds at high temperature may be authorized by the rather high bond strength (around 1 valence units) of this bond and its little thermal-induced anharmonicity [21]. In contrast, this hypothesis may not be true for much weaker bonds, like metallic Pb-Pb [9]. The quench of the melt has been realized by simply turning off the power supply of the heating wire (room temperature was reached after only a few seconds). The original rutile sample was clearly melted within the hole, forming a menisk with some bubbles. The analysis of the XAFS spectral data for this glass sample shows the presence of 3.7 Ge-0 bonds at 1.740 k 0.003 A, as well as disordered Ge-Ge contributions near 3.15 + 0.05 A. This result is consistent with previous XAFS and neutron diffraction studies on vitreous germania [22,23]. The synthesis of a glassy GeO, from a well-crystallized rutile sample confirms that GeO, was melted at 1400 f 30 K (nominal melting temperature is 1389 K [17]. The structural similarities of the Ge-polyhedra between the glass and the molten phase are therefore striking. This again confirms that the cumulant expansion theory is likely to be applicable for rigid bonds exposed to high temperatures. In another study, we will show that it is possible to generalize this hypothesis to a variety of well known crystalline oxide compounds studied by high temperature XAFS at Ti, Fe, Ni, Ga, Sr and Zr K-edges.

498 8. Perspectives:

F. Farges et al. / Nucl. Instr. and Meth. in Phys. Res. B 101 (1995) 493-498

Dl W.E. Jackson, G.A. Waychunas,

ultra high temperatures?

The use of iridium, tantalum or tungsten wires to collect ultra high temperatures XAFS data is in progress. This setup has been used with success at LURE for ultra high temperature X-ray diffraction studies. The heating wire is placed within a silica tube with a constant argon or nitrogen flow (Fig. 1). Two holes are drilled within the quartz tube in order for the incoming X-rays to penetrate and to exit the tube. The study of samples under extreme conditions allows, for instance, the investigation of Earth sciences oxide materials at ultra-high temperatures (melts, perovskite structures). Also, some systematic XAFS data collection at a variety of edges can be realized in order to better understand the general trends that governs temperature induced anharmonicity in oxides. Finally, this setup can be used for other XAFS spectrometers (with step-bystep monochromators or “quick-exafs” facility) assuming that these devices can focuse the incoming X-rays.

Acknowledgements The authors wish to thank the staff at LURE (and more particularly people at XASlO beamstation: F. Baudelet, E. Dartyge, A. Fontaine and A. Polian) for their help in collecting the experimental data; P. Richet (CNRS, IPGP, Paris), Gordon E. Brown Jr (Stanford University), John R. Rehr (University of Washington, Seattle), A. Filiponi (Universita dell’ Aquila, Coppito), Joel Dion (IPGP, Paris) and an anonymous reviewer for their help in the writing of this paper.

References

G.E. Brown Jr., J. Mestre de Leon, S. Conradson and J.-M. Combes, X-ray absorption fine structure, ed. S.S. Hasnain (Ellis Horwood, Chichester, 1990) p. 298. [31 E.A. Stern, P. Livins and Z. Zhang, Phys. Rev. B 43 (1991) 8850. 141 F. Farges, G.E. Brown Jr., G. Calas, L. Galoisy and G.A. Waychunas, Geophys. Res. Lett. 21 (19941 1931. [51 A. Filipponi and A. Di Cicco, Nucl. Instr. and Meth. B 93 (1994) 302. [61 D. Andrault, J. Peyronneau, F. Farges and J.-P. Itit, Physica B 208/209 (1995) 327. [71 L. Wenzel, D. Arvanytis, H. Rabus, T. Lederer, K. Baberschke and G. Comelli, Phys Rev. Len. 64 (1990) 1765. 181 A.I. Frenkel and J.J. Rehr, Phys. Rev. B 48 (1993) 585. 191 A. Filipponi, J. Phys. (Condens. Matter) 6 (1994) 8415. DOI B.0. Mysen and J.D. Frantz, Chem. Geol. 96 (1992) 321. [ill P. Richet, P. Gillet, A. Pierre, A. Bouhfid, I. Daniel and G. Fiquet, J. Appl. Phys. 74 (1993) 5451. 1121E. Dartyge, C. Depautex, J.-M. Dubuisson, A. Fontaine, A. Jucha, P. Leboucher and G. Tourillon, Nucl. Instr. and Meth. A 246 (1986) 452. [131 J.J. Rehr, R.C. Albers, C.R. Natoli and E.A. Stern, Phys. Rev. B 34 (1986) 4350. [141 J.J. Rehr, Z.I. Zabinsky and R.C. Albers, Phys. Rev. Len. 69 (19921 3397. 1151 W.H. Baur and A.A. Khan, Acta Crystallogr. B 27 (1991) 2133. J. Glinneman, H.E. King Jr, H. Schultz, T. Hahn, S.J. La [161 Placa and F. Dacol, Zeit. Kristallogr. 198 (1992) 179. [I71 P. Richet, Phys. Chem. Minerals 17 (1991) 79. k31 M. Madon, P. Gillet, C. Julien and G.D. Price, Phys. Chem. Mineral. 18 (19901 7. [191 T. Yarnanaka and K. Ogata, J. Appl. Crystallogr. 24 (1991) 111. DOI T. Yamanaka, K. Sugiyama and K. Ogata, J. Appl. Crystallogr. 25 (1992) 11. La F. Farges, G. Fiquet, D. Andrault, J.-P. Itie, Physica B 208/209 (1995) 263. P21 J.A.E. Desa, A.C. Wright and R.N. Sinclair, J. Non Crystal. Sol. 99 (19881 276.

[l] G.A. Waychunas, G.E. Brown Jr., W.E. Jackson Ponader, Nature 332 (1988) 251.

and C.W.

1231 C. Lapeyre, J. Petiau, G. Calas, F. Gautier and J. Gombert, Bull. Mineral. 106 (1983) 77.