Hydrogenation of LaNi5 studied by in situ synchrotron powder diffraction

Hydrogenation of LaNi5 studied by in situ synchrotron powder diffraction

Acta Materialia 54 (2006) 713–719 www.actamat-journals.com Hydrogenation of LaNi5 studied by in situ synchrotron powder diffraction J.-M. Joubert a a...

375KB Sizes 0 Downloads 9 Views

Acta Materialia 54 (2006) 713–719 www.actamat-journals.com

Hydrogenation of LaNi5 studied by in situ synchrotron powder diffraction J.-M. Joubert a

a,*

ˇ erny´ b, M. Latroche a, A. Percheron-Gue´gan a, Bernd Schmitt , R. C

c

Laboratoire de Chimie Me´tallurgique des Terres Rares, ISCSA, CNRS, 2–8 rue Henri Dunant, F-94320 Thiais Cedex, France b Laboratoire de Cristallographie, Universite´ de Gene`ve, 24 Quai Ernest Ansermet, CH-1211 Gene`ve 4, Switzerland c Swiss Light Source, PSI Villigen, CH-5232, Switzerland Received 24 January 2005; received in revised form 29 September 2005; accepted 29 September 2005 Available online 20 December 2005

Abstract Hydrogen absorption in LaNi5 intermetallic compound has been studied by in situ synchrotron powder diffraction. Fast data acquisition allows the measurement to be performed under dynamical conditions, i.e., while the full reaction proceeds. The major finding of this study is evidence of a non-equilibrium transient c phase never reported at room temperature without prior high-temperature cycling. Other findings are the variations of the extent of the line broadening and changes in the line shape during the absorption and desorption processes. They are explained by a reorganization of the hydrogen-induced dislocations in a and b phases. Ó 2005 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Line broadening; Synchrotron radiation; X-ray diffraction; Hydrides; Dislocation

1. Introduction Metal hydrides reversibly formed by the hydrogenation of intermetallic compounds are of interest from both fundamental and applied points of view. In situ powder diffraction measurements during the hydrogenation reaction offer valuable information on the reaction process. One may follow, for example, the nature and quantity of the phases involved, the lattice parameters, the kinetics and the creation of lattice defects. The most studied intermetallic compound for hydrogen storage has probably been the binary compound LaNi5 (CaCu5-type structure, space group P6/mmm, La in 1a, Ni in 2c and 3g). It has already been the subject of several in situ hydrogenation studies using conventional [1,2] or synchrotron [3] X-ray or neutron [4–8] radiation. These studies have provided evidence of the transformation of a hydrogen-poor solid solution of hydrogen (a phase: LaNi5H0–0.5) into a hydrogen-rich hydride (b phase: LaNi5H6, same metallic sub-structure) *

Corresponding author. Tel.: +33 1 49 78 12 11; fax: +33 1 49 78 12 03. E-mail address: [email protected] (J.-M. Joubert).

with a much larger unit cell by a discontinuous phase transformation. At the first absorption, a very large and anisotropic diffraction line broadening is present in the b phase which has been interpreted by the presence of a high density of dislocations (mainly prismatic dislocations with Burgers vector equal to a, hereafter called E2) [9–12]. Once formed, those dislocations remain in the material and the broadening is always present in both a and b phases during the subsequent cycles. Another major issue has been the detection of a third phase (c phase: LaNi5H3) appearing after high-temperature cycling as probably a non-equilibrium phase [1,2,5]. These in situ measurements have always been made in a step-by-step mode. The sample is partially hydrogenated (dehydrogenated) by adding (removing) a small amount of hydrogen insufficient to complete the reaction. One then waits for a sufficient time for the equilibrium to be reached and the measurement is done at the given step of the reaction. Though very valuable, this type of experiment cannot be called a real in situ measurement, since it is not done while the reaction proceeds but statically at given predetermined equilibrium steps. In particular, it does not allow one to reveal the existence of

1359-6454/$30.00 Ó 2005 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.actamat.2005.09.036

714

J.-M. Joubert et al. / Acta Materialia 54 (2006) 713–719

transitory non-equilibrium phases likely to appear during a fast reaction. In this work, the availability of a microstrip detector able to collect one full high-resolution powder diffraction pattern with sufficient statistics in a very short time (1 s) at the Materials Science Beamline at the Swiss Light Source [13] allows measurements to be performed without interrupting the reaction. This has been done both for absorption and desorption for the first hydrogenation cycles of LaNi5. The reaction can proceed in a time of the order of one minute, so that each accumulated pattern can be considered as a snapshot at a given time. The phases involved, their relative quantity and their lattice parameters, as well as the evolution of the diffraction line profiles and widths are studied. The analysis of the results allows a description of the actual behavior of the sample while the reaction proceeds. 2. Experimental The intermetallic compound LaNi5 was synthesized by induction melting of the pure elements under vacuum and was annealed at 1200 °C over 4 days. The sample was found to be single phase using metallography, conventional X-ray diffraction and microprobe analysis. The latter analysis gives a composition close to the nominal composition LaNi4.99(4). The in situ powder diffraction experiment was performed at the Materials Science Beamline at the Swiss Light Source in Villigen, Switzerland [13]. A reaction-tight cell rated to 25 bar was especially designed for the experiment (Fig. 1). The powdered sample (typical grain size 20 lm) is located in a glass capillary of 20 mm in length and 0.3 mm in diameter (2R) allowing a measurement in the conventional Debye–Scherrer transmission geometry. The sample was mixed with 25 wt.% amorphous carbon

in order to prevent the capillary breaking due to the volume change of the powder during the hydrogenation reaction and in order to decrease the X-ray absorption by the sample. The top of the capillary is open to allow hydrogen flowing inside. The cell is closed by an X-ray-transparent beryllium cap (thickness 0.5 mm) screwed to the cell and sealed with an O-ring. The cell is connected to an external hydrogenation device allowing the pressure of hydrogen (purity 5 N) delivered to the system to be monitored. Due to the connection of the cell to the hydrogen supply, a complete rotation of the cell during the measurement is impossible. The powder averaging during the measurement was therefore ensured using an internal spinner inside the reaction cell. Absorption was carried out by submitting the sample to a hydrogen pressure of 15 bar at room temperature and letting it absorb with its own kinetics in quasi-isobaric conditions. Desorption is made possible by connecting the whole device to a vacuum pump. An optimum X-ray wavelength with respect to the primary intensity and the X-ray absorption in the beryllium ˚ ). The and in the sample was selected (k = 0.71115 A absorption coefficient l of LaNi5 at this wavelength is 380 cm1. The typical penetration depth 1/l is therefore of the order of 30 lm, larger than the particle size (especially after decrepitation), thus allowing a measurement representative of the ÔbulkÕ sample. Owing to the low packing inside the capillary, the effective lR value is close to unity. The beam size at the capillary is 5 mm in length. The measurement is done continuously during the absorption and desorption reactions using a microstrip detector covering an angular range 7–47° with 10,240 channels [13]. The acquisition time for one pattern is set between 5 and 20 s depending on the kinetics of the reaction. The powder diffraction patterns were analyzed by the Rietveld method using the Fullprof program [14]. The

Fig. 1. Views of the hydrogenation cell attached to the goniometric head at the center of the diffractometer: (a) the sample capillary mounted on the axis of the internal spinner, (b) the cell closed with the beryllium cap. Note the copper tube at the rear of the screw allowing the connection to the hydrogen supply.

J.-M. Joubert et al. / Acta Materialia 54 (2006) 713–719

diffraction lines corresponding to the beryllium cap were placed in excluded regions. A pseudo-Voigt function defined by gLðH Þ þ ð1  gÞGðH Þ;

ð1Þ

where L and G represent the Lorentzian and Gaussian components and g the refined mixing parameter, was used to refine the line profiles. The full width at half the maximum H was defined by H 2 ¼ ½U þ DST ðhklÞtg2 h þ Vtgh þ W .

ð2Þ

The anisotropic line broadening was taken into account using the implemented model for a hexagonal structure [15]. Two parameters (Saa and Scc) are defined which describe the strains along the a and c axes. They fully define the function DST(hkl). The values of U, V and W defining the instrumental resolution were obtained from the refinement of a LaB6 standard (SRM 660). 3. Results The results of the refinement in terms of phase content, lattice parameters and induced strains are presented in Figs. 2–4 for the first hydrogenation cycle (absorption 1 and desorption 1) and the absorption of the second cycle (absorption 2). Selected parts of several representative diffraction patterns are shown in Fig. 5. Absorption 1 proceeds relatively slowly. The a phase is progressively transformed into the b phase without a significant incubation time as regards the kinetics of the whole process. A small but significant increase of the a lattice parameter is observed. The absorption in the a phase is accompanied by a weak increase of the strain parameter along the a-direction. The precipitated b phase is strongly broadened (e.g., see Fig. 5(b)), with a Gaussian character (g  0.3) in contrast to the Lorentzian character of the a phase (Fig. 5(a)) (g  1). The b phase lattice is strongly and isotropically expanded compared to that of the a phase (Da/a = 7.7%, Dc/c = 7.7%, DV/V = 24.9%). The lattice

715

parameters slightly decrease as the absorption is progressively completed. Desorption 1 proceeds about 10 times faster than absorption 1. The most striking factor to be noted is the presence of another phase, named the c phase, in the course of the phase transformation from b to a (Fig. 5(c)). This phase has a line broadening of the same order of magnitude as that of the b phase and has a cell volume intermediate between those of b and a phases. The lattice ˚ , c = 4.025 A ˚ ) do not change durparameters (a = 5.280 A ing the desorption. An interpolation between the cell volumes of the intermetallic compound and that of the b phase considering a composition of 6.5 H/f.u. (hydrogen atoms per formula unit) for the b phase leads to an estimated composition of about 3 H/f.u. for the c phase. The maximum phase content (around 10 wt.%) is reached when the a and b phases are present in equal quantities. The c phase then progressively disappears as the b–a transformation is completed. A very significant decrease of the lattice parameters of the b phase, and to a lesser extent of the a phase, is observed while the desorption proceeds. This can be explained by the decrease of the hydrogen content in the two solid solutions (a and b phases). This is particularly evident for the b phase since the lattice parameters are decreasing before any phase transformation occurs between t = 0 s and t = 100 s. It is worth noting also that the strain parameters of the b phase change while the b phase transforms. Saa decreases and Scc increases in such a way that the line broadening becomes isotropic. The a phase appears already strongly anisotropically broadened with a constant broadening during the transformation, comparable with that of the b phase at the end of absorption 1. Unexpected triangular shapes for the diffraction peaks of this latter phase are, however, observed (Fig. 5(d)). Note that the effect is more pronounced for the diffraction lines hk0. The sample was left under vacuum between t = 1000 s and t = 3600 s (results not shown) which explains the discontinuity of the aa parameter between desorption 1 and absorption 2 in Fig. 3.

Fig. 2. Phase content of a, c and b phases as a function of time during absorption 1, desorption 1 and absorption 2 (left to right).

716

J.-M. Joubert et al. / Acta Materialia 54 (2006) 713–719

Fig. 3. Lattice parameters of the a (up) and b (down) phases as a function of time during absorption 1, desorption 1 and absorption 2 (left to right). a parameters (filled symbols) are read on the left-hand scale, c parameters (open symbols) on the right-hand scale.

Fig. 4. Strain parameters Saa and Scc of the a and b phases as a function of time during absorption 1, desorption 1 and absorption 2 (left to right).

Absorption 2 proceeds quite differently from absorption 1. The kinetics is strongly improved, the reaction being 90% completed in 50 s. The c phase is also present with the same characteristics as in desorption 1. The lattice parameters of the a phase increase substantially as expected from the change of hydrogen content in the primary solid solution, i.e., before the phase content is observed to change significantly. In the same way, the lattice parameters of the b phase are observed to change in the solid solution, i.e., after the transformation from a to b. As for the b phase in desorption 1, the Saa parameter of the a phase is observed to decrease substantially also leading to isotropic broadening. A decrease of Scc for the b phase is also observed. The broadening of the b phase is then con-

stant and equal to the broadening observed in the first absorption. The shapes of the b phase always appear with a pronounced Gaussian character. The cycling was pursued in situ up to desorption 3 (results not shown). The same features can be noted: i.e., desorptions 2 and 3 are equivalent to desorption 1, and absorption 3 is equivalent to absorption 2. This concerns the evolution of the lattice and strain parameters and the presence of the c phase. The only noticeable changes concern the improvement of the kinetics and the maximum c phase content. The reaction is 90% completed at t = 150, 25 and 95 s for desorption 2, absorption 3 and desorption 3, respectively. The maximum c phase content reaches 15% during desorption 3.

J.-M. Joubert et al. / Acta Materialia 54 (2006) 713–719

717

Fig. 5. Selected parts of representative diffraction patterns (a) before absorption 1, (b) after absorption 1, (c) at t = 240 s during desorption 1 and (d) after desorption 1. Measured (points) and refined (line) data are presented. Markers indicate the diffraction peaks of the phases involved. Additional lines may be noted in all the patterns at 2h = 12.5°, 17.3° and 17.65° which are due to the sample environment.

4. Discussion The results presented in this work are consistent with previous literature data. The phase transformation from a to b occurring during absorption 1 is in agreement with what has been observed in [6]. This was expected since the reaction does not proceed rapidly. The a phase remains essentially unbroadened while the b phase nucleates with an already high quantity of defects. The same constant behavior of the lattice parameters is observed. Following Pitt et al. [6], this behavior could be attributed to the non-coherent nucleation of the b phase, the particles of this phase moving apart from the particles of the a phase in the process of decrepitation. The extent of the broadening in the b phase after absorption or of the a phase after desorption is also comˇ erny´ et al. patible with previous studies. As an example, C [11] have used the same method for analyzing the strains in a cycled fully desorbed sample, i.e., in conditions close to the a phase at the beginning of absorption 2 in our experiment, and found Saa = 0.76 and Scc = 0.22. The values of the lattice parameters for a and b phases at the beginning and at the end of each absorption and desorption are consistent with previous work [3,11]. Those of the c phase are also consistent with published values [1].

It may be noted after others [1] that this phase has an a lattice parameter intermediate between those of the a and b phases but closer to that of the b phase, and the c parameter quite close to that of the a phase in close correlation with the property of this phase to accommodate the strong difference of the lattice parameters between these two latter phases. The large lattice parameter changes in a and b phases during desorption 1 and absorption 2 can be attributed to the change of hydrogen concentration in the solid solution domains preferentially to a mechanical interaction between the two phases as proposed in [4,6]. The presence of the c phase at room or near room temperature is not observed for the first time. It has already been seen by Matsumoto and Matsuchita [1] at 40 °C after submitting the sample to high temperature, by Akiba et al. [2] at 25 °C after cycling at 95 °C and by Gray et al. [5] after cycling at 100 °C. However, it was not observed at room temperature without high-temperature cycling [2–4,6,8]. This is the major difference observed in the present work. Note that in all the previous studies, the measurements, even if done in so-called in situ conditions, have been done in equilibrium conditions at discrete steps during the absorption or desorption process, thus demonstrating the non-equilibrium nature of this phase. Our experiment is essentially different since the measurement is done in

718

J.-M. Joubert et al. / Acta Materialia 54 (2006) 713–719

non-equilibrium and dynamic conditions during the absorption and desorption. The c phase is not observed during absorption 1, but is always observed afterwards, both in absorption and desorption with similar maximum quantities in both processes up to the third cycle. Its lattice and broadening parameters are neither observed to change significantly during the reactions nor to be significantly different between absorption and desorption. Of course, this result should be understood given the accuracy that can be obtained from the refinement of a phase with a maximum content around 10%. Due to the temperature dependence of the plateau pressure, the sample would not absorb under 15 bar surrounding pressure if its temperature was above 80 °C. At higher temperature, the absorption plateau pressure of LaNi5 is above 15 bar [2], and the reaction would stop until the temperature decreases. The fact that during absorption 3 the reaction is 90% completed after only 25 s proves that the heat of the reaction may be evacuated in the meantime and that the sample temperature does not reach temperatures above 80 °C at any moment. The temperature rise should therefore be less for absorption 2, for which the reaction is 90% completed in 50 s, and one should note that the temperature rise at the first absorption (90% completed in 5500 s) is almost negligible. The c phase formation cannot have been created by the temperature rise during this first absorption and it is already seen at the first desorption, when the reaction is endothermic. Moreover, the fact that the amount of c phase is equal in absorption and desorption when the temperature change is opposite is also an indication that the temperature effect is not the main cause for the occurrence of the c phase. The purity of the hydrogen is not believed either to be a cause for the appearance of the c phase. Therefore, it seems that the only reason for the occurrence of the c phase is the fact that the measurements are done in dynamical conditions under a very short timescale of the order of 10 s, while the timescale for the other experiments was of the order of hours. It is believed that the c phase is a transient, non-equilibrium phase which is always formed but which does not remain in the sample for a sufficiently long time to be seen using conventional measurements. The comparison between Figs. 2 and 4(b) in [1] is striking. The same behavior with a maximum c phase content when a and b phase contents are equal is observed. The fact that the c phase appears in such a static measurement when the sample has been previously cycled at high temperature can be considered. Buckley et al. [16] explain this feature by a c-favorable structure prerequisite to c formation to distinguish the samples cycled or not at high temperature. Since we have observed that the c phase is always present after the first absorption, our interpretation is that previous high-temperature cycling slows down the kinetics of the disappearance of the c phase sufficiently so that it is seen in conventional in situ measurements. The reason why previous cycling at high temperature favors this behavior is still unclear.

The absence of the c phase at the first absorption can be related to the fact that dislocations are not present in both a and b phases (only the b phase contains defects). It could also be explained by the fact that the kinetics is slow since it has been demonstrated that the lifetime of the c phase is much shorter than the time needed for the reaction at this step. Finally, another explanation could be the lack of interfaces between a and b phases if one considers, as in [6], that the grains of the b phase move apart from those of the a phase at the first cycle and that the highest c phase content is observed when a and b phase contents are similar, indicating that the c phase occurs at the interface. Remarkable changes of the strain parameters of a and b phases are observed: a systematic reduction of Saa of either phase is observed in the phase that disappears. A similar reduction of Saa of the b phase was already noticed in [6] but not interpreted. This parameter, representing the broadening on the hk0 lines, has been shown to be directly related to the dislocation density [12]. A reduction of this parameter may therefore be understood as a decrease of the dislocation density in the given phase. This fact may be explained if one considers that the appearing phase nucleates in the compression (for a phase in desorption) or in the dilatation (for b phase in absorption) zones around the dislocations. The dislocations are preferentially kept at the interface between the two coexisting phases, thus playing the role of misfit dislocations. As diffraction is phase sensitive, these interface dislocations are no longer perceptible in effective broadening which explains the reduction of Saa. A schematic representation of this phenomenon is shown in Fig. 6. Also, changes of Scc are observed in the b phase. Both in absorption and desorption, a maximum value is obtained for this parameter when the a and b phases are in equal amounts: therefore when the interface is a maximum. This may be due to the elastic strain field of the two coherent phases with different volumes interacting with each other. One may note that a similar explanation has already been proposed [6]. The effect of this strain field is more visible for the b phase because this phase is much more deformable than the a phase [17,18]. The fact that the effect is more visible for the c-direction is in agreement with the proposed model of hydride precipitate growth along this axis which has been previously shown to be in agreement with the high density of dislocations lying in the prismatic planes [11,19]. The rather unusual triangular shapes of the peaks of the a phase after desorption may be explained by the spatial inhomogeneity of the dislocation density [20,21]. It is worth noting that the triangular shapes are the most discernable on the hk0 lines which is in agreement with the fact that they are due to an inhomogeneous density of the prismatic dislocations (E2) responsible for the broadening of those lines [11]. Formation of the a phase is probably accompanied by a migration of the dislocations with the a–b interface leading to a final distribution which is inhomogeneous. It is possible that dislocation walls are formed at the meeting

J.-M. Joubert et al. / Acta Materialia 54 (2006) 713–719

719

in static and equilibrium conditions, among them the systematic presence of the c phase both in absorption and desorption after the first absorption. We have demonstrated that this phase is a transient non-equilibrium phase, only appearing during a very short time period. Moreover, the profile analysis yields not only information on the dislocation density by means of the study of the line broadening but also on the distribution of the dislocations. Another interesting feature of this type of measurement (though not exploited in this work because of the simultaneous presence of three phases) is the possibility to use the reacted fraction from X-ray diffraction to study the kinetics. In addition, to be as close as possible to isothermal conditions, one obtains the phase amount as a function of time in contrast to volumetric or gravimetric measurements which give hydrogen content. Those two parameters are different due to the presence of solid solutions in metal hydride systems, and it is believed that the joint analysis of both could shed more light on the mechanisms involved in the kinetics of hydride formation. Fig. 6. Schematic representation of a dislocation (a) in a single phase, inducing high apparent strain broadening in this phase, (b) as a misfit dislocation between a and b phases lowering the apparent strain broadening in either phase.

point between a nuclei. This may be related to the proposal in [12] of a significant amount of dislocation dipoles to explain residual changes of lattice parameters. On the other hand, b phase formation in the a phase during absorption results in a final distribution which is more homogeneous. The dislocation density in pure a or b phase is identical as can be guessed from equal values of Saa and Scc. Two hypotheses can be formulated. Either the dislocations grouped in dislocation walls are dispersed during the b phase formation, or they are annihilated during the nucleation of b in the dilatation zone of the dislocations while others are randomly created in the same amount in the rest of the sample during b phase growth. 5. Conclusion The use of transmission geometry, combined with the use of moderately hard X-rays, allows measurements to be performed that are really representative of the sample bulk, in contrast to conventional X-ray experiments which are very sensitive to surface effects. The time resolution of the microstrip detector allows measurements in dynamic conditions. The sample state is monitored using synchrotron powder diffraction while the absorption or desorption reactions proceed, whatever the kinetics. As this type of experiment allows one to probe the sample behavior in conditions close to those applied in the case of a hydrogen gas storage unit, we believe it could be applied to other interesting systems for such applications. Indeed, new features have been evidenced compared to measurements done

Acknowledgment The authors acknowledge the staff of the Materials Science Beamline at the Swiss Light Source, Villigen, Switzerland, particularly to the local contact, Fabia Gozzo. References [1] Matsumoto T, Matsushita A. J Less-Common Met 1986;123:135. [2] Akiba E, Nomura K, Ono S. J Less-Common Met 1987;129: 159. [3] Wu E, Gray EMA, Cookson DJ. J Alloys Compd 2002;330:229. [4] Kisi EH, Gray EMA, Kennedy SJ. J Alloys Compd 1994;216:123. [5] Gray EMA, Kisi EH, Smith RI. J Alloys Compd 1999;293:135. [6] Pitt MP, Gray EMA, Kisi EH, Hunter BA. J Alloys Compd 1999;293:118. [7] Kisi EH, Wu E, Kemali M. J Alloys Compd 2002;330:202. [8] Pitt MP, Gray EMA, Hunter BA. J Alloys Compd 2002;330:241. [9] Wu E, Gray E, Kisi EH. J Appl Crystallogr 1998;31:356. [10] Wu E, Kisi EH, Gray E. J Appl Crystallogr 1998;31:363. ˇ erny´ R, Joubert J-M, Latroche M, Percheron-Gue´gan A, Yvon K. J [11] C Appl Crystallogr 2000;33:997. ˇ erny´ R, Joubert J-M, Latroche M, Percheron-Gue´gan A, Yvon K. J [12] C Appl Crystallogr 2002;35:288. [13] Gozzo F, Schmitt B, Bortolamedi T, Giannini C, Guagliardi A, Lange M, et al. J Alloys Compd 2004;362:206. [14] Rodrı´guez-Carvajal J. In: Proceedings of the XV congress of international union of crystallography, satellite meeting on powder diffraction, Toulouse, France; 1990. p. 127. [15] Latroche M, Rodrı´guez-Carvajal J, Percheron-Gue´gan A, Boure´eVigneron F. J Alloys Compd 1995;218:64. [16] Buckley CE, Gray EM, Kisi EH. J Alloys Compd 1995;231:460. [17] Bereznitsky M, Ode A, Hightower JE, Yeheskel O, Jacob I, Leisure RG. J Appl Phys 2002;91:5010. [18] Hector Jr LG, Herbst JF, Capehart TW. J Alloys Compd 2003;353:74. ˇ erny´ R, Percheron-Gue´gan A, Yvon K. J [19] Joubert J-M, Latroche M, C Alloys Compd 2002;330:208. [20] Groma I, Unga´r T, Wilkens M. J Appl Crystallogr 1988;21:47. [21] Unga´r T, Groma I, Wilkens M. J Appl Crystallogr 1989;22:26.