J O U R N A L O F C H E M I S T R Y Materials The 63 K phase transition of ZrTe3: a neutron diVraction study Ram Seshadri,a Emmanuelle Suard,b Claudia Felser,a E.Wolfgang Finckh,a Antoine Maignanc and Wolfgang Tremel*a aInstitut fu�r Anorganische Chemie und Analytische Chemie, Johannes Gutenberg-Universita�t, Becher Weg 24, D-55099, Mainz, Germany bInstitut Laue Langevin, Avenue des Martyrs, BP156, Grenoble F-38042, France cLaboratoire CRISMAT, ISMRA, 6, Boulevard Mare�chal Juin, Caen F-14050, France Received 13th July 1998, Accepted 9th September 1998 The 63 K phase transition of ZrTe3 has been followed through SQUID magnetisation and neutron diVraction studies.The transition is characterized by a small quenching of the magnetic susceptibility. Contrary to the expectation that the structural transition is associated with a charge density wave, the results of Rietveld refinements of high resolution neutron powder diVraction profiles indicate that below the phase transition the diVerent bonding Te–Te contacts are more equal rather than less.The results are examined with the help of band structure calculations on structures determined at three diVerent temperatures.The picture that emerges supports the view that anion–cation redox competition plays a crucial role in determining not only the structures of these compounds, but also the temperature dependence thereof. Before the realization of high-Tc superconductivity in the long-range coherence, there is considerable faulting along the layered copper oxides, considerable attention was paid to the other directions.physical properties of layered transition metal chalcogenides, Canadell, Mathey and Whangbo9 examined the electronic in particular to the occurrence of phase transitions in systems structure of type B ZrTe3 in considerable detail within the such as NbSe3 associated with the formation of charge density extended Hu�ckel approximation.They were interested particuwaves (CDW).1 Of these chalcogenides, ZrTe3 has had a larly in the diVerence between ZrTe3 and other group IV rather colourful history. First reported by McTaggart and trichalcogenides. They suggested that the unusual properties Wadsley,2 the structure was determined by Furuseth et al.3 of ZrTe3 including the transition at 63 K could be explained who found for MX3 (M=Ti, Zr, Hf and X=S, Se, Te) two only within the B type structure.These conclusions have since diVerent structure types within the P21/m space group. In been undermined in the light of Furuseth and Fjellva°g10 both structure types, the atoms occupy the 2(e) position (x, B, redetermining the structure of ZrTe3 and establishing that it z). The atom positions of the so-called types A and B structures is all or mostly type A and not type B.Recently, photoemission are related to one another through xA=1-xB and zA=zB for and thermopower measurements and detailed electronic strucall the atoms. ZrTe3 was determined as crystallizing in the ture calculations on type A ZrTe3 have been presented.11 type B structure. Through the use of so-called frozen-phonon calculations, the Room-temperature resistivity measurements on ZrTe3 sug- importance of short Te–Te distances (and metrical changes gested that it is metallic or semimetallic.2 This was confirmed thereof ) in determining the shape of the Fermi surface and by Bayliss and Liang who performed optical reflectivity the number of states near the Fermi energy have been estabmeasurements on single crystals4 and found that although the lished.11 Nearly simultaneously a redetermination of the type A material is anisotropic, the strengths of the optical transitions structure and detailed ab initio band structure calculations along the diVerent crystallographic directions are similar.have been presented by Sto�we and Wagner.12 The conclusions Takahashi et al.5 reported transport measurements on single reached by these authors largely coincide with those presented crystals as a function of temperature. They observed an in ref. 11 regarding the ro� le of the short Te–Te contacts in anomaly in the resistivity peaking around 55 K. The magnitude influencing the electronic properties of ZrTe3. of the Hall coeYcient was found to rise sharply at this In the absence of a description of the structure of ZrTe3 temperature.Their initial interpretation was that the anomaly below the transition, and in particular of an idea of what was caused by increased scattering of charge carriers rather happens to the diVerent distances between the atoms, the than a decrease in their density. Below 2 K they found that precise nature of the transition would remain a matter of ZrTe3 becomes superconducting. These authors extended their speculation.To this end, we have examined the evolution of first study, using elastic measurements to complement the the structure of ZrTe3 through the collection and analysis of transport studies.6 While they could confirm the presence of powder neutron diVraction (PND) profiles. We present the the transition at 63 K, they were unable to establish whether rather surprising results here.it arose from a depletion of charge carriers or from increased The occurrence of what seemed to be a CDW transition scattering. The nature of the superconducting state turns out and superconductivity in the same compound has prompted to be rather unusual, being described as not bulk, but filaour interest in ZrTe3.These two diVerent ground states are mentary from an analysis of the excess conductivity.7 The sometimes thought to arise from similar causes but are usually nature of the transition associated with the resistivity anomaly competing. Indeed, in ref. 11 it has been suggested that the was examined through electron diVraction and dark-field structural transition and superconductivity are related to inde- imaging at low temperatures by Wilson and coworkers,8 who pendent features of the band structure.An understanding of found that even at room temperature, some of the ZrTe3 the relations between crystal and electronic structure and crystals displayed structural modulation. Below # 63 K, they physical properties of systems such as ZrTe3 could yield found a phase transition characterized by the appearance of insights crucial to the preparation of new materials with superlattice spots at nearly q=( 1 14, 0, H ).Imaging in the satellite spots suggested that while along the a axis the structure has interesting properties. J. Mater. Chem., 1998, 8, 2869–2874 2869the calculated profiles in suitable reciprocal space directions. Experimental Asymmetry was handled using the Hermite polynomial Sample preparation approach of Be�rar and Baldinozzi.17 During the later stages of the refinement, some peaks were An approximately 15 g powder sample of ZrTe3 was prepared found to be fitted rather poorly.Through trial and error, a by heating well ground, stoichiometric mixtures of the elements small quantity of ZrTe518 was found to be present in the in a sealed, evacuated quartz glass ampoule at 973 K for three sample.This phase was therefore included in all the refinements days. Longer heating times were avoided because an analysis presented here. For the D20 diVractometer, the wavelength of the relevant Ellingham diagrams showed that the thermowas calibrated to be 2.4085 A° using a yttrium iron garnet dynamic product (in reaction with the walls of the quartz glass (YIG) standard, and for the D2B diVractometer the wave- ampoule) would be oxides of zirconium.This rather mild length was calibrated to be 1.5941 A° using an Si standard. heating protocol resulted in samples contaminated by small Cell volumes and their errors were calculated using standard amounts of ZrTe5 (determined from the PND profiles).The formulae.19 Other metrical information was obtained through samples also suVered some texturing. the use of the PLATON97 program.20 Correlations between the diVerent refined parameters were ignored in the calculations SQUID magnetisation of quantities such as interatomic distances and angles. Magnetisation data under a 1 T field were collected on a 0.1174 g sample of ZrTe3 on cooling between 100 and 5 K ire calculations a Quantum Design MPMS 5 magnetometer.The sample was The electronic structure of type A ZrTe3 has been discussed held in a gelatine capsule that was fixed to the end of a in detail in ref. 11 and is discussed only very briefly here. drinking straw. The magnetisation of the gelatine capsule and Tight-binding linear muYn-tin orbital (LMTO) band structure drinking straw were recorded separately and fitted with high methods21 were used within the atomic sphere approximation reliability to the analytic form M=A0+A1/T.This was then (ASA). The calculations were performed on the room- used to correct the data for the sample holder. temperature single crystal X-ray structure of Furuseth and Fjellva°g10 and the structures obtained from the Rietveld Neutron diVraction refinement of the D2B data at 100 and 1.7 K. 2774 irreducible Neutron data were collected on two diVerent diVractometers, k points were used to achieve convergence. D2B and D20 at the Institute Laue-Langevin, Grenoble France. D2B is a high resolution diVractometer and as we Results and discussion shall see the resolution of the powder profiles obtained on it were largely sample limited.Data on D2B were collected at Description of type A ZrTe3 1.7 and 100 K using a wavelength of 1.5941 A° . In order to Despite it having been discussed extensively in the litera- enhance the resolution a 10¾ primary beam collimation was ture,3,9,11,22 we present for completion a depiction of the employed and the monochromator beam divergence was limstructure of type A ZrTe3 in the panels of Fig. 1 focusing on ited by means of a system of slits positioned after the monothose aspects of the structure that we will find most important chromator. The uncertainties in the temperature are with respect to the phase transition. The Zr atoms are eight- throughout <0.2 K.Each data collection took about 12 h. D20 is a diVractometer of lower resolution but working with a fixed sample and detector. As a result, during an experiment (e.g. a temperature ramp) there are eVectively no systematic errors introduced. The special geometry of D20 allows rapid acquisition of diVraction profiles of very low noise and very high precision both in counts and in scattering angle.We also attempted to study the pressure dependence of the structure of ZrTe3 using a hydrostatic Ti–Zr (‘zero matrix’) pressure cell with He as the working fluid. Data were collected at fixed temperature while ramping the pressure up to 1000 bar. Due to certain experimental constraints, the data were of insuYcient quality13 for positional parameters to be extracted.The data were analyzed using the Rietveld method14 as implemented in the XND program.15 We have used a special feature of this implementation which requires some explanation. When several diVraction profiles are collected as a function of some external variable (temperature, pressure, composition etc.), two possibilities exist for their treatment.The prevailing method is to refine each data set independently, extracting structural parameters that are then assembled in order to follow a certain trend. The lesser used alternative is implemented in the XND Rietveld program, allowing the refinable parameters in the structural model to be expanded as a polynomial in the external variable. This results in a considerable reduction in the ratio of refinable parameters to data, providing better refinements from data sets of limited quality.This has been previously employed to examine the composition dependence of structure in some layered manganites.16 The profiles are handled in the reciprocal space as convolutions of Lorentzian and Gaussian functions with their individ- Fig. 1 Structure of type A-ZrTe3 showing (a) the nature of the Zr–Te ual angular dependencies.Preferred orientation (particularly polyhedra, (b) the complete structure projected down [010] with the for the data acquired on the D20 diVractometer under pressure) Te atoms marked and (c) the nature of the Te(2)–Te(3) contacts within the sheets. was treated using first order Legendre polynomials to weight 2870 J. Mater.Chem., 1998, 8, 2869–2874coordinated by Te as shown in Fig. 1(a). These polyhedra are arranged in double sheets stacked along the monoclinic c axis as displayed in Fig. 1(b), where the view is looking down [010]. The basal plane is then defined by rectangular sheets of Te(2) and Te(3) atoms with short and long distances between them as shown in Fig. 1(c) (with a view down [001]).The Te(1) are arranged in buckled sheets around z=0 and z=1. Sheets of Zr are then stacked between the Te sheets in a manner that yields the coordination depicted in Fig. 1(b). The motif of rectangular sheets, with short, alternating Te(2)–Te(3) contacts along the a direction is something that we return to in the discussion of the phase transition. Note that when presented in the manner employed here the structure of ZrTe3 is very much quasi-2D, comprising slabs that are stacked along the monoclinic c axis.We prefer this description over the traditional view that the structure is quasi-1D or chain-derived. Magnetic susceptibility Between 100 and 65 K, the SQUID susceptibility (corrected for Fig. 3 Temperature dependence of the unit cell volume of ZrTe3 as a sample holder contribution and for core diamagnetism23) dis- function of temperature.The filled circles are D20 data and the played in Fig. 2 is eVectively Curie–Weiss. Below this tempera- squares are D2B data. For the D2B data, the error bars are smaller than the symbols. The solid line is a linear fit to data above 63 K and ture, the susceptibility and its inverse display a small plateau, the curve is a quadratic fit to the data between 13 and 61 K.before once again reverting to nearly Curie–Weiss behaviour around 30 K. The phase transition can thus be described as a small quenching of the susceptibility. We postpone the interpretation of this behaviour to after we have discussed the structure. The CDW transition in ZrTe5 takes place above 100 K,24 so the changes observed here cannot be attributed to the small amount of the ZrTe5 impurity that is present.Thermal evolution of the structure We commence with a description of the evolution of the cell volume between 120 and 1.9 K. The D20 data were collected on a continuous temperature ramp from 1.9 to 200 K with data being binned into approximately 4 K intervals. The refinements yielded the cell volume evolution displayed in Fig. 3 as small filled circles with error bars. The squares at 1.7 and 100 K are from refinements of the high resolution D2B data discussed shortly. The error bars for these two data are smaller than the symbols (note that at 1.7 K, the points from D2B and D20 overlap). Between 120 and 65 K, the data can be fitted with high reliability to a straight line.The data point at 61 K deviates significantly from this evolution. Between 61 and 13 K, the volume evolves in a nearly quadratic manner as shown by the fitted curve. The fits allow the transition to be characterized by a very small volume contraction at 63 K of about 0.02%. The evolution of the individual lattice parameters with temperature is displayed in the diVerent panels of Fig. 4. We observe immediately that the clearest changes are along the a Fig. 4 Temperature dependence of the cell parameters of ZrTe3. The points with error bars are from individual refinements of D20 data and the lines are from coupled refinements of D20 data (13 data sets at a time) treated separately above and below the transition. axis. From the preceding discussion of the structure, this would implicate the Te(2)–Te(3) distances in the phase transition.The points with error bars are the values obtained from individual Rietveld refinements of D20 data. The lines are the results of refinement of two structural models against 13 data sets above the transition and (separately) against 13 data sets below the transition, expanding all refineable structural parameters to first order in the temperature.More explicitly, instead of refining some structural parameter p, the parameter is expanded in the temperature according to p=p0+p1(T ). p0 and p1 are then refined against N data sets acquired at the diVerent temperatures T. At the end of the refinements, one Fig. 2 Temperature dependence of the magnetic susceptibility and its obtains p0, p1, and the errors dp0 and dp1.The close correspon- inverse of ZrTe3 under a 1 T magnetic field. The lines are guides to the eye. dence between the points from individual refinements and lines J. Mater. Chem., 1998, 8, 2869–2874 2871obtained from the coupled refinements validates the procedure profile of ZrTe3 (with a small ZrTe5 admixture) collected at 1.7(1) K.The fit between the data and the refined model is not (and especially, the first order expansion). The necessity for this coupling is that the structural parameters thus obtained very good, due to problems of texturing in the sample, and what seems to be high dispersion in the structural coherence lengths are of greater reliability, possessing smaller error bars. Methods for estimating the error from such a coupled refine- in the sample.This is also manifested in the rather large values of the agreement factors; the Bragg and weighted profile R ment have been presented previously.16,25 In the present case they are approximately half as small as the error bars associ- factors. Indeed, refining two ZrTe3 phases with all structural parameters except the lattice parameters constrained, and ated with the parameters resulting from refining the data sets individually. Since the strategy of the coupled refinements allowing for diVerent peak widths corresponding to distinct particle sizes did result in better fits ( lower weighted-profile R yields a temperature dependence of structure (i.e.a state function rather than a state point), we can calculate the factors) but considerably increased the number of refined parameters, leading us to abandon this strategy.Nonetheless, the errors diVerent structural parameters at arbitrary temperatures within the refined state function.Doing so at the temperatures of 50 on the interatomic distances seem to be only slightly larger than those associated with the high-quality single-crystal results pre- and 80 K, representing the structure below and above the phase transition, we obtain the distances presented in Table 1.sented in ref. 10. From the refined Rietveld scale factors, quantitative analysis26 ignoring the Brindley factors27 (which are not The magnitudes of the error bars in this table are clearly too large to discern changes in interatomic distances that might very important for neutron refinements) suggested that the ZrTe5 impurity included in all the refinements presented here was as characterize the phase transition.We note however, that the long and short Te(2)–Te(3) distances are 2.99(5) and high as 15%. Table 2 lists the structural parameters of ZrTe3 obtained from the refinements at the two temperatures, and 2.88(5) A° at 50 K while they are 3.00(9) and 2.88(9) A° at 80 K.The D20 data thus do not support the traditional picture Table 3, the important interatomic distances extracted from the data. For comparison, interatomic distances obtained in the of a CDW transition, following which one would expect a divergence in the distances below the transition. room-temperature single-crystal study of Furuseth and Fjellva° g10 are also displayed.The most significant diVerences between the structures at the three temperatures are in the short Structures at higher resolution and long Te(2) and Te(3) distances, which show a small but Refinement of data fromtheD2B diVractometer at temperatures perceptible divergence between 1.7 and 100 K, through to the around 100 K and 1.7 K yield structures of much higher pre- room temperature.This is plotted in Fig. 6. If the structural cision. Fig. 5 displays the experimental and fitted PND (D2B) changes at 63 K were associated with a CDW-like distortion of the sheets defined by Te(2) and Te(3), the observed metrical Table 1 Interatomic distances obtained from the coupled Rietveld behaviour is precisely the opposite of what would then be refinement of D20 data, calculated at 50 and 80 K expected.Atom 1 Atom 2 D(50 K)/A° D(80 K)/A° Densities of state and the nature of the transition Te(1) Zr 3.09(4) 3.08(7) In Fig. 7 we display a comparison of the Te(2) and Te(3) p- Te(1) Zr 3.13(4) 3.14(7) orbital and Zr d-orbital derived densities of state (DOS) Te(1) Zr(×2) 2.97(3) 3.01(5) Te(2) Zr(×2) 2.98(3) 2.96(6) Table 2 Structures obtained from the Rietveld refinement of D2B data Te(3) Zr(×2) 3.01(3) 3.01(5) at 1.7 and 100 K.Space group P21/m (no. 11) Te(2) Te(3) 2.88(5) 2.88(9) Te(2) Te(3) 2.99(5) 3.00(9) Atom x y z B/A° 2 T=1.7 Ka Zr 0.2878(4) 0.25 0.6654(3) 0.84(5) Te(1) 0.7635(6) 0.25 0.5574(2) 0.73(6) Te(2) 0.4284(5) 0.25 0.1647(3) 0.08(5) Te(3) 0.9068(6) 0.25 0.1600(3) 0.24(5) T=100.0 Kb Zr 0.2882(5) 0.25 0.6648(3) 1.00(5) Te(1) 0.7643(6) 0.25 0.5579(3) 0.84(6) Te(2) 0.4297(5) 0.25 0.1650(3) 0.25(5) Te(3) 0.9068(6) 0.25 0.1601(3) 0.46(5) aa=5.8726(3), b=3.9084(2), c=10.0536(5)A° , b=97.848(3)°.RB= 9.7%, Rwp=8.5%. ba=5.8775(3), b=3.9125(2), c=10.0645(5)A° , b= 97.835(3)°. RB=9.4%, Rwp=8.2%. Table 3 Interatomic distances obtained from the Rietveld refinement of D2B data, at 1.7 and 100 K, compared with those at 298 K Atom 1 Atom 2 D(1.7 K)/A° D(100 K)/A° Da(298 K)/A° Te(1) Zr 3.132(4) 3.132(5) 3.156(2) Te(1) Zr 3.122(4) 3.120(5) 3.140(2) Te(1) Zr(×2) 2.957(3) 2.959(3) 2.956(1) Te(2) Zr(×2) 2.956(3) 2.956(3) 2.939(2) Fig. 5 Rietveld refinement of D2B data acquired at 1.7(1) K. The Te(3) Zr(×2) 2.959(3) 2.966(3) 2.961(2) Te(2) Te(3) 2.816(5) 2.811(5) 2.793(2) data (a), Rietveld fit to the ZrTe3 structure (b), Rietveld fit to the ZrTe5 structure (c), a parasitic peak due to the cryostat (d) and the Te(2) Te(3) 3.057(5) 3.067(5) 3.103(2) diVerence between observed and refined profiles (e) are displayed, as aSingle crystal data of Furuseth and Fjellva°g.10 are markers of the peak positions for the diVerent structures. 2872 J. Mater. Chem., 1998, 8, 2869–2874a rearrangement of some states arising from (or causing) changes in Te(2)–Te(3) distances. More simply, the transition is associated with charge transfer. Rouxel has argued eloquently for the importance of what he calls redox-competition28 in going from oxides to the more covalent transition metal chalcogenides, whereby the observed reduction in the energy diVerence between cation d bands and anion sp bands results in a competition between these for valence electrons.In the previous work from this group11 the importance of such redox competition in the ZrTe3 system was discussed. Access to the low-temperature structures now allows us to show that redox competition not only governs structural principles but perhaps also the temperature dependence thereof.The changes in the DOS indicate that the biggest diVerences occur between the room temperature and 100 K rather than between 100 and 1.7 K as one might expect from the temperature at which the transition is found. However, the band structure calculations are eVectively at 0 K, and the ro� le of phonons is ignored. It is possible that when temperature eVects Fig. 6 Evolution of the short and long Te(2)–Te(3) distances with are taken into account, the situation would correspond more temperature.The 298 K data are from the single crystal study of closely to what is observed. We interpret the quenching of the Furuseth and Fjellva°g.10 magnetic susceptibility as arising from a small decrease in the total density of states (not shown) at the EF on cooling below the transition temperature resulting in a decrease in the Pauli paramagnetic contribution.The sharp drop in the thermopower along the a axis at the transition and below11 [the direction being that of the Te(2)–Te(3) contacts] could be ascribed to increased mobility along this axis. The one question remaining seems to be the symmetry change at the phase transition associated with a structural modulation as observed in the electron diVraction measurements. Electron diVraction is much more sensitive to changes in symmet than is powder neutron diVraction (particularly when the symmetry change is associated with a very large supercell ) while it does not enjoy the metrical precision of neutron diVraction coupled with Rietveld refinement.Indeed, that the symmetry of the low-temperature structure could be diVerent is suggested by the behaviour of the thermal parameters. Those of Zr and Te(1) at 1.7 K are about 80% the values at 100 K, but for Te(2) and Te(3) the changes in B on cooling are larger, the values of B at 1.7 K for Te(2) being about 30% of the value at 100 K and for Te(3) about 50% of the value at 100 K.The use of the P21/m space group with its associated unit cell for the low-temperature refinement is thus only approximate, being the commensurate subcell of some lower symmetry system. We do not expect this to aVect our conclusions regarding the trends in the interatomic distances. The nature of the rearrangement of atoms within the sheets permits the classification of the transition as displacive and second order.29,30 Identifying an order parameter in the structural metric from the present study is however made diYcult by the fact that the Te–Te distances along the a axis are Fig. 7 Partial LMTO densities of state of ZrTe3 calculated using the published 298 K structure of Furuseth and Fjellva°g10 and the unequal in both the high and the low temperature phases.structures determined in this study at 100(1) and 1.7(1) K. The That the low-temperature structure [when we consider only Te(2)-p, Te(3)-p and Zr-d states are shown in a small window around the Te(2)–Te(3) sheet] is less distorted than the high-tempera- the Fermi energy. ture structure is perhaps unusual but by no means novel. Perovskite manganites of the general formula Ln1-xAxMnO3 where Ln is a rare-earth ion and A is an alkaline earth ion obtained from high level LMTO calculations on the structures at three diVerent temperatures.Without going into the details have received much recent attention because of the finding that in these systems the onset of ferromagnetism on cooling of the electronic structure (discussed extensively in ref. 11), we note that the Te(2)–Te(3) p-orbital interaction is s*, and this is accompanied by the phenomena of giant negative magnetoresistance (GMR) being displayed.31 Careful structural interaction results in a partially filled band that crosses the Fermi energy. Changing the distances between the Te(2) and investigations on such GMR manganites have revealed that the delocalization of the eg electron on MnIII below the Te(3) atoms in the sheets should result in some rearrangement in the number of states in the valence band.In fact, the ferromagnetic transition results in the MnO6 octahedra being less distorted below the transition temperature rather than photoemission data (acquired at 298 and 80 K) presented in ref. 11 support the picture of a transfer of weight from the more.32 Similar eVects have been observed as a function of temperature or pressure in some layered manganites.33 In valence band to just below EF, in keeping with what is observed here.The structural transition is thus associated with ZrTe3, the transition in the thermopower and the quenching J. Mater. Chem., 1998, 8, 2869–2874 2873References 1 J.Rouxel, in Crystal Chemistry and Properties of Materials with Quasi-One-Dimensional Structures, ed. J. Rouxel, D. Riedel, Dordrecht, 1986 pp. 1–26. 2 F. K. McTaggart and A. D. Wadsley, Aust. J. Chem., 1958, 11, 845. 3 S. Furuseth, L. Brattas and A. Kjekshus, Acta Chem. Scand. Ser. A, 1975, 29, 623. 4 S. C. Bayliss and W. Y. Liang, J. Phys. C., 1981, 14, L803. 5 S. Takahashi, T.Sambongi and S. Okada, J. Phys. (Paris) Colloq. C, 1983, 3, 1733. 6 S. Takahashi, T. Sambongi, J. W. Brill and W. Roark, Solid State Commun., 1984, 49, 1031. 7 H. Nakajima, K. Nomura and T. Sambongi, Physica B, 1986, 143, 240. 8 D. J. Eaglesham, J. W. Steeds and J. A. Wilson, J. Phys. C., 1984, 17, L697. 9 E. Canadell, Y. Mathey and M.-H. Whangbo, J. Am. Chem. Soc., 1988, 110, 104. 10 S. Furuseth and H. Fjellva°g, Acta Chem. Scand., Ser. A, 1991, 45, 694. 11 C. Felser, E. W. Finckh, H. Kleinke, F. Rocker and W. Tremel, Fig. 8 Isothermal (19 K) pressure dependence of the a, b and c lattice J. Mater. Chem., 1998, 8, 1787. parameters of ZrTe3 obtained from a coupled refinement of D20 data 12 K. Sto�we and F. Wagner, J. Solid State Chem., 1998, 138, 160. during a pressure ramp from ambient pressure to 1 kbar. 13 By quality, we refer to a combination of resolution, statistics (high signal to noise ratios) and dynamic range s=sin h/l that would allow large numbers of parameters to be refined stably and with precision. of the susceptibility, in conjunction with the calculated decrease 14 H. M. Rietveld, J. Appl. Crystallogr., 1969, 2, 5. in the DOS at the EF below the transition suggests that what 15 J.-F.Be� rar, computer code XND version 1.11, ESRF, Grenoble, is observed is similar to the situation in the manganites except France, 1996; J. F. Be�rar, Proceedings of the I.U.Cr. Satellite that the the cause is not an atomic energy level (a Jahn–Teller Meeting on Powder DiVraction, Toulouse, France, 1990; J. F. distortion) but rather the behaviour of the s* band formed Be�rar and F.Garnier, Advanced Powder DiVraction II by Te(2) and Te(3) p orbitals. Conference, N.I.S.T. Gaithersburg, Maryland, 1992; The program is freely available at the URL http://rx-crg1.polycnrsgre. fr/public/xnd/xnd.html. The eVect of pressure on lattice parameters 16 R. Seshadri, C. Martin, M. Hervieu, B. Raveau and C. N. R.Rao, Chem. Mater., 1997, 9, 270. From the discussion so far it is evident that only data of the 17 J.-F. Be� rar and G. Baldinozzi, J. Appl. Crystallogr., 1993, 26, 128. highest quality can reveal details of the phase transition in 18 H. Fjellva°g and A. Kjekshus, Solid State Commun., 1986, 60, 91. ZrTe3. From the isothermal data collected by us on the D20 19 C. Giacovazzo in Fundamentals of Crystallography, ed.diVractometer under pressure, it was not possible to extract C. Giacovazzo, IUCr-Oxford, 1992, pp. 122–124. structural parameters. However, through the use of the data- 20 A. L. Spek, computer code PLATON97, Acta Crystallogr. Sect. coupling strategy using six data sets, it was possible to obtain A, 1990, 46, C34. 21 R. W. Tank, O. Jepsen, A. Burkhardt and O.K. Andersen, The experimental compressibilities along the diVerent directions. Stuttgart TB-LMTO-ASA program, MPI fu� r Festko� rperfor- The assumption is that in the small pressure range studied (up schung, Stuttgart, Germany, 1998. to 1 kbar) the changes are linear. Fig. 8 displays the percentage 22 W. Tremel and R. HoVmann, J. Am. Chem. Soc., 1987, 109, 124. changes in the a, b and c lattice parameters of ZrTe3 at 19 K. 23 K.-H. Hellwege and A. M. Hellwege, Landolt-Bo�rnstein Tables, The points refer to the temperatures at which the data were New Series, Volume 2, Springer-Verlag, Heidelberg, 1966; Note acquired. The error bars on the line are smaller than the points that the correction for core diamagnetism is considerably aVected by how charges are assigned to the diVerent atoms, so the true representing the temperatures. The Te(2)–Te(3) interactions susceptibility might be in some error.along the a axes should be softer than the Zr–Te interactions 24 T. Sambongi, in Crystal Chemistry and Properties of Materials so one might naý�vely expect that this is the axis most sensitive with Quasi-One-Dimensional Structures, ed. J.Rouxel, D. Riedel, to pressure. We find instead that both the a and the b axes Dordrecht, 1986 pp. 281–313. have similar compressibilities but the c axis along which the 25 Briefly, the final error on the refined parameter is independent of sheets are stacked is significantly softer. This is in keeping all terms except those that are of zeroth order in the expansion. For N parameters, we then have dp=ÓN×dp0.R. S. thanks Dr. with the presence of a van der Waals gap between the double J. P. Attfield for a clarification of this point. sheets of the Zr–Te polyhedra and in keeping with the struc- 26ajal, Manual of the FullProf Rietveld Program, tural description that we have chosen. Similar results were Laboratoire Leon Brillouin, CEA Saclay, France, 1997. Available observed from the pressure isotherms at 100 K suggesting that by anonymous ftp from bali.saclay.cea.fr. the transition seems to have no significant eVect on the 27 G. W. Brindley, Philos. Mag., 1945, 36, 347. contractions of the diVerent axes within the resolution of the 28 J. Rouxel, Chem. Eur. J., 1996, 2, 1053. 29 H. D. Megaw, Crystal Structures, A Working Approach, experiment. W. B. Saunders, Philadelphia, 1973. 30 M. T. Dove, Am. Mineral., 1997, 82, 231. Acknowledgements 31 C. N. R. Rao, A. K. Cheetham and R. Mahesh, Chem. Mater., 1996, 8, 2421. It is a pleasure to thank Professor O. K. Anderson and Dr. 32 V. Caignaert, E. Suard, A. Maignan, C. Simon and B. Raveau, O. Jepsen for providing the LMTO codes, and Dr. P. Convert C. R. Acad. Sci. (Paris) Ser. IIb, 1995, 321, 515; P. G. Radaelli, M. Marezio, H. Y. Hwang, S. W. Cheong and B. Batlogg, Phys. of the ILL for help and advice on D20. During the various Rev. B, 1996, 54, 8992. stages of this work, we have received assistance from Dr. 33 J. F. Mitchell, D. N. Argyriou, J. D. Jorgensen, D. G. Hinks, V. Ksenofontov and Mr. F. Rocker (Mainz) and Messrs. C. D. Potter and S. D. Bader, Phys. Rev. B, 1997, 55, 63; J. Torregrossa, L. Melesi and P. Cross (ILL). We thank them. D. N. Argyriou, J. F. Mitchell, J. B. Goodenough, O. Chmaissem, S. Short and J. D. Jorgensen, Phys. Rev. Lett., 1997, 78, 1568. This work has been supported by the Deutsche Forschungsgemeinschaft (DFG) and the Fonds der chem- Paper 8/05427D ischen Industrie. 2874 J. Mater. Chem., 1998, 8, 2869–2874