High Pressure Research Vol. 26, No. 3, September 2006, 267–276

Study of partial melting at high-pressure using in situ X-ray diffraction D. ANDRAULT*†‡§¶, G. MORARD‡¶
, N. BOLFAN-CASANOVA
, O. OHTAKA#, H. FUKUI∗∗, H. ARIMA‡‡, N. GUIGNOT
, K. FUNAKOSHI††, P. LAZOR‡‡ and M. MEZOUAR
†Laboratoire Magmas et Volcans, Université de Clermont-Ferrand, France ‡Université Denis Diderot, Paris 7, France §Institut de Minéralogie et de Physique des Milieux Condensés, Université Paris 7, France ¶Institut de Physique du Globe de Paris, Paris F-75252, France
European Synchrotron Radiation Facility, Grenoble, France #Earth and Space Science, Osaka University, Japan ∗∗Institute for Study of the Earth’s Interior, Okayama University, Misasa, Japan ††Japan Synchrotron Radiation Institute, SPring-8, Hyogo, Japan ‡‡Department of Earth Sciences, Uppsala University, Sweden (Received 5 April 2006; revised May 2006; in final form 12 May 2006) The high-pressure melting behavior of different iron alloys was investigated using the classical synchrotron-based in situ X-ray diffraction techniques. As they offer specific advantages and disadvantages, both energy-dispersive (EDX) and angle-dispersive (ADX) X-ray diffraction methods were performed at the BL04B1 beamline of SPring8 (Japan) and at the ID27-30 beamline of the ESRF (France), respectively. High-pressure vessels and pressure ranges investigated include the Paris– Edinburgh press from 2 to 17 GPa, the SPEED-1500 multi-anvil press from 10 to 27 GPa, and the laser-heated diamond anvil cell from 15 to 60 GPa. The onset of melting (at the solidus or eutectic temperature) can be easily detected using EDX because the grains start to rotate relative to the X-ray beam, which provokes rapid and drastic changes with time of the peak growth rate. Then, the degree of melting can be determined, using both EDX and ADX, from the intensity of diffuse X-ray scattering characteristic of the liquid phase. This diffuse contribution can be easily differentiated from the Compton diffusion of the pressure medium because they have different shapes in the diffraction patterns. Information about the composition and/or about the structure of the liquid phase can then be extracted from the shape of the diffuse X-ray scattering. Keywords: Melting; Synchrotron radiation; Diamond anvil cell; Multi-anvil press; Paris–Edinburg press

1.

Introduction

Defining melting criteria for high-pressure experiments has always remained complicated because of the limited access to the sample. In the laser-heated diamond anvil cell (LH-DAC), *Corresponding author. Email: [email protected]

High Pressure Research ISSN 0895-7959 print/ISSN 1477-2299 online © 2006 Taylor & Francis http://www.tandf.co.uk/journals DOI: 10.1080/08957950600897013

268

D. Andrault et al.

the prevailing melting criteria are based on optical observations that are facilitated by the diamond transparency to visible light. Upon melting, the sample undergoes changes of its emissivity factor and/or modification of its interaction with the IR-laser, which induces variation of the sample surface thermal emission. Unfortunately, this optical method disables fine analysis of the melting behavior, such as the determination of solidus, liquidus, and/or eutectic lines, and also control of partial melting. These difficulties have been responsible for long and extensive controversies about the properties of the mantle and the core. Even today, the melting diagrams of the two main components of the Earth’s interior, the (Mg, Fe)SiO3 perovskite phase [1, 2] and iron [3, 4], remain ambiguous at high pressures. In a large volume apparatus, like the multi-anvil (MAP) and Paris–Edinburgh (PEP) presses, the main advantage is a better control of the temperature. A very important aspect is that the furnace efficiency is not affected by sample melting, in contrast with what happens using the LH-DAC technique. Also, the sample chemistry can be relatively homogeneous in a large capsule. Determining the sample state requires advanced techniques, however, because of the opacity of the pressure medium and anvils. One way is to check the microstructure of the recovered samples using the scanning electron microscope, for example, but this technique is rather limited, as it does not provide much information about the sample state above the solidus temperature due to liquid re-crystallization upon quenching. Melting can also be determined using in situ electrodes that detect significant changes of the sample electrical conductivity, a quite interesting method to analyze liquid percolation, for example [5]. Nevertheless, in situ diffraction clearly appears to be the best method for the study of melting in large high-pressure vessels. It requires the use of intense X-ray beams, and this type of facility is now available on different synchrotron rings in the world. Studying the liquid structure at high pressure is a very active research field since a few years. Structural changes of the second or first order were, for example, evidenced in liquid phosphorous [6] or liquid GeSe2 [7], using energy-dispersive (EDX) or angle-dispersive X-ray diffraction methods (ADX) in the large volume apparatus. Another very useful technique is the X-ray absorption spectroscopy, which makes it possible to probe the local structure in the liquid phase, as illustrated by a previous study on liquid Ge [8]. These techniques have led to an important literature that is important for our present study, even if our goals are slightly different. Our major interest is indeed more specifically oriented towards the determination of the degree of melting and of the positions of solidus, liquidus, and/or eutectic lines, as a function of pressure and temperature, and to potentially reconstruct the high-pressure melting diagrams from careful in situ experiments.

2.

High-pressure vessels

Three different high-pressure vessels, the LH-DAC, the MAP, and the PEP, were alternatively used for our melting experiments: (1) Pressures from 9 to 22 GPa were provided by the SPEED-1500 double stage MAP installed at the BL04B1 beamline of the SPring8 synchrotron (Japan) [9]. We used tungsten carbide anvils with 3 mm truncations, sitting in the so-called [1 0 0] configuration with cubefaces parallel to the compression axis. Samples consisted of a Fe-alloy with Fe0.9 S0.1 composition intimately mixed with about 50 wt% of SiO2 powder. The sample and MgO pressure marker were loaded in a sintered alumina tube of 1 mm diameter and 0.2 mm wall thickness, itself inserted in a 7 mm Cr-doped MgO octahedral pressure medium. After the loading in the press was completed, the Al2 O3 tube sat parallel to the opening gap between the WC anvils (figure 1). The whole sample assemblage was made available

Partial melting at high-pressure

269

Figure 1. X-ray absorption image of a typical iron-alloy sample loaded in the SPeed-1500 MAP installed on the BL04B1 beamline (SPring8, Japan). The small rectangle indicates the size of the X-ray beam in diffraction mode.

to the X-ray probe by X, Y, Z motors. Temperatures exceeding 1600 ◦ C were provided by resistance heating of two LaCrO3 sheet furnaces located above and below the Al2 O3 tube. Temperature was monitored using a WRe3 /WRe25 thermocouple running parallel to the Al2 O3 sample container. Both sample and junction are located at equal distances from the two sheet furnaces. More experimental details are provided elsewhere [13]. In this configuration, temperature reproducibility is estimated to be better than 30 K. We neglected the effect of pressure on the thermocouple properties. For the MgO pressure marker, we prefer the recent equation of state (EoS) refined from diamond anvil cell experiments [10], as it appears to remain valid from the moderate conditions found in the large volume apparatus [11] to the most extreme pressure found in an LH-DAC [12]. According to the resolution of the X-ray diffraction experiment and to the temperature uncertainty, we estimate that the sample pressure is known with an accuracy better than 0.5 GPa. (2) LH-DAC experiments were performed at the ID27 beamline of the ESRF (Grenoble, France) [14]. We used DAC mounted with diamonds with 300 µm culets, and Re gaskets with a hole of 100 K before some major diffraction lines of the sample disappear from the diffraction pattern for time exposure of a few minutes. This effect is due to the enhanced grain growth of the iron solid phase, especially in the presence of liquid (see [23] for similar grain growth enhancement for silicate phases). Note that the use of ADX makes it possible to follow qualitatively the grain size directly on the image plate; progressive evolution from fine diffraction rings to spotty diffraction pattern evidences grains growth. In the same temperature range, we also observe the appearance of a band of diffuse X-ray scattering, using

Partial melting at high-pressure

271

Figure 2. Schematic diagram for melting in the Fe–FeS system in the 15–20 GPa pressure range, compatible with previous studies [34]. Upon heating, the onset of melting corresponds to either solidus or eutectic temperature. Above the eutectic temperature, the degree of melting increases progressively with increasing temperature, until the sample is completely melted above the liquidus temperature.

both EDX and ADX diffraction methods. Depending on both sample’s initial composition and the shape of the melting diagram, the temperature at which the onset of melting is observed can correspond to the solidus or eutectic temperature (figure 2).

5.

Observation of the liquid phase

Using the PEP, structural investigation of the Fe–FeS eutectic liquid has been carried out (Morard et al., in preparation). For temperatures above the complete melting of the Fe–FeS mixture (above the liquidus), we observe a large band of diffuse X-ray scattering located at ∼6 ◦ (2θ angle), which corresponds to a distance of ∼2.4 Å (figure 3(a)). Because the multichannel collimator is not sufficiently efficient at low angles, some diffraction lines of the pressure medium remain, which can be easily removed to retrieve the main contribution of the liquid. A classical normalization and an FT-treatment [24] yield the radial distribution function of the liquid (figure 3(b)). A typical highly packed metallic structure is fully compatible with the well-defined contributions. The structure is close to that of pure liquid iron, as measured at ambient pressure [25] and high pressure up to 5 GPa [26]. The chemical analysis of the quenched samples highlights a low O concentration in the eutectic liquid. A small reaction rim of 5 µm thick of (Mg,Fe)O appears at the contact between the sample and the MgO capsule. The oxidation of a small amount of Fe is most probably related to O diffusion from the pressure medium to the sample, until the oxygen fugacity (fO2 ) is well buffered all around the sample. The ADX method is clearly preferred for gathering information about the structure of liquids. Still, similar information can also be retrieved from EDX experiments, and liquid structures have been refined successfully using this technique [27]. The process of spectral

272

D. Andrault et al.

Figure 3. PEP experiments using the ADX method: (a) diffraction pattern showing the diffuse X-ray scattering at 17.2 GPa and 1200 K. The dotted line shows the liquid signal after eliminating the diffraction peaks from the MgO capsule; (b) radial distribution function g(r) of the Fe–FeS liquid.

normalization is more complicated, because the X-ray beam intensity profile (intensity versus X-ray energy) should be well known. An important aspect is that it is not always possible to perform ADX diffraction in all high-pressure vessels. In the MAP, the primary anvils do not offer a large angle of aperture, and the use of the EDX method is much simpler.

6.

Quantifying degree of partial melting

Using the MAP and EDX methods at the BL04B1 beamline of SPring8, we collected X-ray diffraction patterns for the Fe0.9 S0.1 as a function of increasing and decreasing temperature with steps of ≤50 K for a pressure of 18.5 GPa. Between ∼1083 and ∼1523 K appears a band of diffuse X-ray scattering, whose intensity varies strongly with temperature (figure 4(a)). Above 1523 K, the band intensity remains constant upon renewed increase of temperature. This intensity behavior is due to a change of the melt fraction as a function of temperature, as the experimental path is crossing a two-phases loop in the Fe–S phase diagram [18]. Note that the coexistence of solid and liquid phases cannot be explained by temperature gradients because of the very small X-rayed volume (∼2 × 10−3 mm3 ). Indeed, the very high thermal conductivity of the sample is likely to enhance temperature homogenization. As the intensity of the diffuse X-ray scattering is directly proportional to the liquid mass and because the sample is stationary relative to the X-ray beam, the intensity of the diffuse scattering signal can be used to retrieve the degree of sample melting as a function of temperature. For this, we normalized intensities relative to the maximum intensity recorded for a fully molten sample (figure 4(b)). For these experiments (and also for the LH-DAC work, given subsequently), the use of a sample made from a mixture of Fe–S and SiO2 is to promote the formation of a well-connected SiO2 matrix that insures a perfect stability of the liquid Fe–S sample relative to the X-ray beam.

Partial melting at high-pressure

273

Figure 4. MAP experiments, using the EDX method: (a) diffuse X-ray scattering during the cooling of a Fe0.9 S0.1 alloy at 18.5 GPa. Maximum intensities at the highest temperatures (top curve) denote a completely molten sample. On cooling (from top to bottom curve), the degree of partial melting decreases with temperature until the background reaches the classical shape of the Compton diffusion at ∼1073 K (bottom curve). Note the intensity change a of second diffusion peak at ∼66 keV (see text). (b) Degree of partial melting extracted from the varying intensity of the diffuse X-ray scattering.

The chemical analysis of the recovered samples show minor amount of O dissolved in the Fe–S sample, and almost no Si, which points out relatively oxidizing fO2 conditions [28]. Note that the presence of SiO2 is not problematic for planetary applications, as SiO2 is one of the main components. Using theADX set up with the PEP at the ID27 high-pressure beamline, equilibrium between solid and liquid could also be produced and studied. The new sintered diamond anvils, because of their transparency to the hard X-rays, offer a large gap between the anvils by translating the press (up to 800 µm at 17 GPa [17]). A scan of the press is performed to test whether the melting affects the whole sample or a part of it remains solid. At a given temperature of 1100 K, for a pressure of 15 GPa, we were able to track the growth of a pure iron crystal by following the increase in intensity of its 311 reflection on the 2D image plate (figure 5). The intensity of the peak can be fitted by a usual Avrami rate equation [29], i.e. a slow increase followed by a sudden rise, used to describe transformation kinetics. It shows that crystallization kinetics

274

D. Andrault et al.

Figure 5. PEP experiments using the ADX method: 2D image plates showing the crystallization with time of a pure iron crystal in a Fe–FeS liquid at 15 GPa and 1100 K. The growing diffraction peak corresponds to the 311 reflection of pure iron. After reaching temperature stability (bottom), same conditions were retained for additional time of 100 s (middle) and 200 s (top).

Figure 6. LH-DAC experiments using the ADX method: diffraction patterns of a Fe–FeS sample recorded with increasing temperature at a nominal pressure of 55 Gpa. Diffraction peaks of the Fe and Fe–S alloy disappear at the advantage of a clear band of diffuse X-ray scattering. PMs point to the diffraction lines of the SiO2 pressure medium.

Partial melting at high-pressure

275

studies are also possible with iron alloys using the PPE, as it has been conducted with silica glass at high pressure using MAP [30] and with dehydration of antigorite at high pressure using PEP [31]. Similar partial melting experiments could be performed in the LH-DAC. We present ADX diffraction patterns in which liquid and solid phases coexist at a pressure of 55 GPa and at several temperatures up to 2700 K. With progressive temperature increase, it has been possible to witness the appearance of the first liquid and retrieve the nature of the coexisting phases (figure 6). Owing to the use of ADX diffraction and to the large surface of reciprocal space covered by the imaging plate, the intensity of the diffraction lines of the solid phases informs quantitatively on the phase contents and degree of melting. Note that the size of the laser spot is most often smaller than the whole sample. Chemical segregation can therefore occur in the sample because of the inevitable presence of thermal gradients [32].

7.

In situ information about changes of the liquid composition

Another interesting feature of the MAP experiments is related to changes of the shape of the diffuse X-ray scattering band with decreasing temperature (figure 4(a)). We already mentioned that the main band, located at ∼55 keV (∼2.4 Å), varies in intensity over a ∼400 K temperature interval. During cooling, we observe a second band of diffuse scattering at ∼66 keV at (∼2.2Å), at which the relative intensity increases with decreasing temperature, and thus with the decreasing amount of liquid in the sample. The main band is well explained by the Fe–Fe atomic pair contribution, whereas the energy position of the second is well explained by the presence of Fe–S pairs in the metallic liquid. At low liquid fraction, the liquid is largely enriched in S, thus displaying a high-intensity contribution at 2.2Å. This second pair could also be explained by the formation of a second, immiscible, S-rich liquid (however, we do not observe sulfur separation in quenched charges). The increase in the number of Fe–S pairs in the liquid with decreasing temperature is expected, given the incompatibility of S with solid iron [33].

8.

Conclusion

In this study, we present different examples showing how in situ X-ray diffraction can be used to monitor and study partially molten materials. The use of different high-pressure vessels, PEP, MAP, and LH-DAC, and of EDX and ADX diffraction techniques provides complementary information. The main method of investigation is based on the intensity of the diffuse X-ray scattering. Moreover, it is possible to retrieve information on the grain size and the dynamic behavior of the sample, which are both closely correlated with the presence of a liquid phase in the sample charge. From an experimental point of view, as often the case, the LH-DAC experiment appears less precise in terms of control of the chemical composition of the sample because of the presence of thermal gradients. Still, this technique allows us to produce extreme pressures to >100 GPa, which no large volume apparatus has yet achieved. The large volume press is ideal for this type of experiments because of its optimal temperature control and reduced thermal gradients. Temperature step of 10 K can be monitored, up to >2000 K, for pressures up to >25 GPa, and therefore a broad range of geophysical interests can be investigated. The PEP is a bit more limited in terms of pressure and temperature, but provides a great advantage of its horizontal opened window (between the two opposite anvils) that is well adapted to the

276

D. Andrault et al.

use of ADX. The normalization process is thus simplified, and the coverage of the reciprocal space is optimized, an important feature when the grain size evolves with time. Acknowledgements We thank W. Crichton, M. Fialin, H. Schultze, T. Yamanaka, C. Sanloup, G. Fiquet, and two anonymous reviewers for help, support, comments, and discussions. This work was supported by JSPS-CNRS exchange programs, Diety-INSU programs, and SPRing8 and ESRF facilities. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34]

D.L. Heinz, E. Knittle, J.S. Sweeney et al., Science 264 279–280 (1994). A. Zerr, Science 262 553–555 (1993). Q. Williams, E. Knittle and R. Jeanloz, J. Geophys. Res. 96 2171–2184 (1991). R. Boehler, Nature 363 534–536 (1993). T. Yoshino, D. Walter and T. Katsura, Nature 422 154–157 (2003). Y. Katayama, T. Mizutani, W. Utsumi et al., Nature 403 170–173 (2000). W. Crichton, M. Mezouar, T. Grande et al., Nature 414 622–625 (2001). S. Pascarelli, T. Neisius, S. de Panfilis et al., J. Synch. Rad. 6 146–148 (1999). W. Utsumi, K.I. Funakoshi, Y. Katayama et al., J. Phys. Condens. Matter 14 10497–10504 (2002). S. Speziale, C.S. Zha, T.S. Duffy et al., J. Geophys. Res. 106 515–528 (2001). Y. Fei, J. Li, K. Hirose et al., Phys. Earth Planet. Inter. 143–144 515–526 (2004). N. Guignot, D. Andrault, G. Morard et al., Earth Planet. Sci. submitted (2006). D. Andrault, N. Bolfan-Casanova, O. Ohtaka et al., Earth Planet. Sci. submitted (2006). M. Mezouar, W.A. Crichton, S. Bauchau et al., J. Synch. Rad. 12 559–664 (2005). D. Andrault, R.J. Angel, J.L. Mosenfelder et al., Am. Mineral. 88 301–307 (2003). E. Schultz, M. Mezouar, W. Crichton et al., High Press. Res. 25 71–83 (2005). G. Morard, M. Mezouar, N. Rey et al., Rev. Sci. Instr. submitted (2006). Y. Fei, M.C. Bertka, L.W. Finger, Science 275 1621–1623 (1997). J.M. Brown, J. Appl. Phys. 86 5801–5808 (1999). M. Mezouar, P. Faure, W.A. Crichton et al., Rev. Sci. Instr. 73 3570–3574 (2002). J. Hammersley, Fit2d User Manual (ESRF, 1996). A.C. Larson, R.B. Von Dreele, GSAS Manual (Los Alamos National Laboratory, 1988). N. Bolfan-Casanova, H. Keppler, D.C. Rubie, Earth Planet. Sci. 182 209–221 (2000). J.H. Eggert, G. Weck, P. Loubeyre et al., Phys. Rev. B 65 174105 (2002). Y. Waseda, K. Suzuki, Phys. Stat. Sol. 39 669–678 (1970). C. Sanloup, F. Guyot, P. Gillet et al., J. Geophys. Res. 107 2272 (2002). K. Tsuji, K. Yaoita, M. Imai et al., Rev. Sci. Instr. 60 2425–2428 (1989). C.K. Gessmann, B.J. Wood, D.C. Rubie et al., Earth Planet. Sci. 184 367–376 (2001). M. Avrami, J. Chem. Phys. 7 1103–1112 (1939). K. Suito, M. Miyoshi, A. Onodera et al., High Temp. High Press. 34 243–250 (2002). J.P. Perrillat, I. Daniel, K.T. Koga et al., Earth Planet. Sci. 236 899–913 (2005). D. Andrault, G. Fiquet, Rev. Sci. Instr. 72 1283–1288 (2001). J. Li, Y. Fei, H.K. Mao et al., Earth Planet. Sci. 193 509–514 (2001). Y. Fei, J. Li, M.C. Bertka et al., Am. Mineral. 85 1830–1833 (2000).