Using standard prb s

X-ray absorption spectroscopy of single-crystalline „VO…2P2O7:
Electronic structure and possible exchange paths
S. Gerhold, N. Nu¨cker, C. A. Kuntscher,* and S. Schuppler Forschungszentrum Karlsruhe, IFP, P.O. Box 3640, D-76021 Karlsruhe, Germany Naval Research Laboratory, Code 6345, Washington DC 20375 A. V. Prokofiev,† F. Bu¨llesfeld, and W. Assmus Kristall- und Materiallabor, Physikalisches Institut, Johann Wolfgang Goethe-Universita¨t, Robert-Mayer-Straße 2-4, ͑Received 16 August 2000; published 25 January 2001͒ Using polarization-dependent V 2 p and O 1s near-edge x-ray absorption spectroscopy, we studied the un- occupied electronic structure of single-crystalline ͑VO͒2P2O7. It is highly anisotropic, and shows similarities tovanadium oxides like VO2 and V2O5 at the V 2p edge and at the O 1s threshold. The contributions from V-Oand P-O orbitals could be identified. The results rule out the spin ladder model for the magnetic behavior of͑VO͒2P2O7, but are consistent with the alternating chain scenario.
PACS number͑s͒: 78.70.Dm, 71.20.Ps, 75.30.Et were oriented by Laue diffraction to better than 1°, and ͑100͒ ceiving increased attention over the past few years, as the and ͑001͒ faces were prepared; to ensure flat and clean sur- crystal structure1,2 suggests that VOPO may be a realization faces, several slices of about 100 nm thickness were taken of a spin- 1 two-leg ladder3 ͑also see Fig. 1͒. Indeed, first off with the diamond blade of an ultramicrotome prior to the experiments on the magnetic properties of polycrystalline measurements. NEXAFS spectra were recorded at the Na- samples were interpreted as consistent with the ladder tional Synchrotron Light Source ͑NSLS͒ using the NRL/ model.4–6 However, quite early VOPO was also proposed to NSLS beamline U4B. The energy resolution was set to 210 be an alternating antiferromagnetic Heisenberg spin- 1 chain.7 meV at 530 eV. The bulk sensitive fluorescence yield ͑FY͒ This view recently gained support from neutron-scattering detection mode was utilized, giving a probing depth of experiments8,9 that are inconsistent with the ladder model8and corroborate the idea of an alternating spin- 1 chain along the b axis.9,10 The chain model is also supported by electron-spin resonance and susceptibility measurements11 obtainedfrom large single crystals.12 The results from Refs. 9 and 11were further analyzed by extending the alternating chainmodel to include ͑frustrated͒ two-dimensional interchaincoupling.13,14 Obviously, the complex interplay betweenmagnetic properties and the underlying structural and elec-tronic characteristics calls for a detailed investigation of theelectronic structure of VOPO—all the more as up to now justa single such study on VOPO, with a significance limited bythe use of polycrystalline samples, exists in the currentliterature.15 This is quite unlike the situation for vanadiumoxides whose electronic structure has been analyzed in alarge number of in-depth studies.
In this paper we report on near-edge x-ray absorption fine structure ͑NEXAFS͒ experiments at the V 2p and O 1s edgefor VOPO. On single crystals, polarization-dependent mea- FIG. 1. Crystal structure of VOPO corresponding to Ref. 2, surements allow a detailed investigation of the unoccupied viewed along the a axis onto the ͑100͒ plane ͑four unit cells electronic structure and its anisotropy, as well as an assign- shown͒. Vanadium, phosphor, and oxygen atoms are represented by ment of the relevant orbitals near the Fermi energy (EF).
white, gray, and black circles, respectively. The ͑b,c͒ planes are The results, in turn, are used to check if the various exchange stacked along the a axis, so that the pairs of VO6 ‘‘octahedra’’ form paths implied by the different models for the magnetic be- well-separated two-leg ladders which are connected by P2O7 double havior mentioned above are consistent with the electronic tetrahedra. Note that the x and y directions are rotated by 45° with respect to the b and c crystal axes in order to align them approxi- Single crystals of VOPO with a size of several mm3 were mately with the in-plane V-O bonds. The thick line is a sketch of a grown as described in Ref. 12. For NEXAFS, the crystals superexchange path in the alternating chain model.
0163-1829/2001/63͑7͒/073103͑4͒/$15.00 63 073103-1
PHYSICAL REVIEW B 63 073103
probes the local ground-state symmetry of the vanadium ion,
whereas at the O 1s edge it gives the partial unoccupied
density of states with O 2 p character.17,22 The shape of the
V 2 p spectra for Eʈ(b,c) is similar to that observed for
VO2 ; the intensity ratio and the angular dependence of the
first two O 1s peaks below 532 eV is similar to V2O5. This
suggests that the band-structure calculations available for
V
Ref. 23͒ may help to assign states consisting of hy- bridized V 3d – O 2 p orbitals to characteristic features of theVOPO spectra,24 especially as no such calculations havebeen published yet for VOPO.
The polarization dependent NEXAFS spectra fall in two groups with different spectral features: first the spectrum for
Eʈa, i.e., with the polarization perpendicular to the layers,
and second the in-plane spectra with Eʈ(100). Up to 532 eV
the unoccupied electronic structure is isotropic within the
͑100͒ plane since the V 2p and O 1s spectra for Eʈb and Eʈc
are identical in this range ͑see Fig. 2͒ and since the spectrum
for E oriented at 45° between b and c equals the Eʈb,c
average, as mentioned above. Above 532 eV the spectra are
FIG. 2. V 2 p – O 1s FY-NEXAFS spectra for VOPO with the different even within the ͑100͒ plane. Combined with the light polarization E parallel to the crystal axes a, b, and c. Peak
earlier argument that the first two O 1s features are similar to labels P and A at the O 1s edge denote features related to the ͑b,c͒ those of V2O5 spectra this observation strongly suggests that plane and the a axis, respectively. The inset shows P3 and P4 for the features at the O 1s edge below 532 eV are related to Eʈ(100) in more detail.
V 3d – O 2 p orbitals only, and that other contributions likeP 3␴ – O 2p hybrids appear above 532 eV.25 Ϸ1300 Å ͑as calculated from tabulated atomic cross It is useful for a detailed discussion of the O 1s spectra to sections16͒. NEXAFS spectra were recorded simultaneously divide the VOPO oxygen sites into four groups ͑following in the total electron yield mode, with the known limitations the notation of Ref. 15; also see Fig. 1͒: ͑i͒ apical oxygen associated with that mode: ͑i͒ a small information depth of atoms O͑1͒ of the distorted VO6 octahedra with a short bond, Ϸ40 Å ͑Ref. 17͒—as VOPO does not cleave easily, a proper 1.6 Å, and a long bond, dV-O(1b) 2.3 Å; ͑ii͒ pla- in situ surface preparation is not possible and adsorbates, nar oxygen atoms O͑2͒ at the corners of the ‘‘ladders’’ with especially at the O 1s edge, are likely to affect the spectra; one V atom and one P atom as the nearest neighbors and and ͑ii͒ charging effects, since VOPO is an insulator. All dV-O͑2͒Ϸ1.94 Å; ͑iii͒ planar oxygen atoms O͑3͒ forming the experimental FY spectra shown in Fig. 2 are corrected for ‘‘ladder rungs’’ with two V and one P atom as nearest neigh- the incoming photon flux and for self-absorption effects, the bors, and dV-O͑3͒Ϸ2.07 Å; and ͑iv͒ O͑4͒ concatenating two latter by applying an extended correction scheme described PO4 tetrahedra to form the P2O7 group.
elsewhere,18 allowing one to treat a sequence of closely For vanadium oxides like VO2 and V2O5 the first and spaced absorption edges. Simultaneously recorded NiO spec- second peaks at the O 1s edge are ascribed to O 2 p orbitals tra were referred to a NiO standard from electron-energy-loss hybridized with the t2g and eg subgroups of the unoccupied spectroscopy,19 resulting in an absolute energy calibration of V 3d orbitals, respectively; they are separated by an octahe- dral ligand field splitting of about 2.2 eV.21 This picture can The results are shown in Fig. 2 for the orientation of the directly be transferred to the case of VOPO in Fig. 2 ͓label- polarization vector E parallel to the crystal axes a, b, and c.
ing spectral features for E parallel to the ͑100͒ plane by P
Further spectra with E oriented at angles between the axes
and those for Eʈa by A͔: peaks P1 and A1 correspond to t2g
were also recorded to test consistency, and gave the expected states, whereas the features P2 and A2 correspond to eg results: a spectrum for E at 45° between b and c is identical
states. P1 is located at 529.1 eV, P2 is just a shoulder at to the average of the spectra for Eʈb and Eʈc; the Eʈa
530.5 eV, and A1 and A2 are located at 529.2 and 531 eV, spectra extrapolated from the Eʈc spectrum and from spectra
respectively. A peak shift for different orientations like the recorded for E at 45° and 60° between c and a are almost
one between P2 and A2 is not observed for V2O5.20,21 The equal to the directly measured Eʈa spectrum.
ligand field splitting for VOPO determined from the separa- We now compare the present results with NEXAFS spec- tion of the first and second O 1s peaks is about 1.6Ϯ0.2 eV and thus smaller than for the vanadium oxides mentioned 21͒, although VOPO is structurally quite different. However, above. In order to interpret peaks A1 to P2 we can use Ref.
the 3d1 configuration of the vanadium in VO2 is the same, 23 as a guide ͑also see Table I͒. Note that the notation used and the distortion of the VO6 octahedra is very similar in for the orbitals is in the local x, y, and z basis set depicted in V2O5, but there is no continuous network of V-O bonds in Fig. 1, where the in-plane x and y axes are rotated by 45° the ͑100͒ plane for VOPO. NEXAFS at the V 2p edge against the in-plane b and c axes; a is parallel to z. The main PHYSICAL REVIEW B 63 073103
contribution to P1 is due to V 3d TABLE I. Assignment of the observed O 1s near-edge peaks to xy – O( 2,3) 2 p x,y V 3d – O 2 p hybridized orbitals.
from symmetry considerations and in analogy to Ref. 23V 3d xz,y z – O( 1 a ) 2 p x,y cur. P2 is assigned to V 3d x2-y 2 – O( 2,3) 2 p x,y xz,y z – O( 2,3) 2 p z 3z2-r2 – O( 1 a ) 2 p z 3z2-r2 – O( 1 b ) 2 p z lies about 0.5 eV higher in energy than P2, which is, in an 4h environment, possible only for a dominant V-O͑1a͒ hybridization;26 this is consistent with the strong d dependence of the pd␴ hybridization.27 The symmetric oc- currence of x and y labels in all pertinent orbitals reflects the We will now concentrate on the spectral weight above 532 eV. The V 4s p – O 2 p antibonding bands give a nearly extent they are compatible with the magnetic interactions isotropic contribution to the absorption structures, and are implied by the various models for the magnetic behavior of responsible for most of the spectral weight of the main peak VOPO. For a reasonable exchange coupling of two adjacent above 535 eV.15,21 The spectral weight between 532 and 535 vanadium spins there must be sufficient ‘‘hybridization eV and the anisotropic part above 535 eV should be assigned strength’’ as well as unoccupied density of states for all to P 3␴ – O 2p antibonding orbitals ͑cf. Table II͒. O͑1͒ will bonds involved along the exchange path.29 The ladder model not contribute, as it does not bond to phosphor atoms. When requires V-O͑1͒-V paths along the legs of the ladder and trying to unravel possible P-O͑2͒ contributions to polariza- V-O͑3͒-V paths along the rungs. As there are no signs for a tion-dependent NEXAFS along the crystal axes we note that significant V-O͑1b͒ hybridization the V-O͑1͒-V path must be the lines connecting neighboring P and O͑2͒ enclose angles very weak, and thus the spin ladder model can clearly be of about 33°–45° with the b axis, and lie almost perpendicu- excluded by our NEXAFS results. This agrees with the latest lar to a. The most likely hybridization is P 3␴ – O͑2͒ 2p ͓ experiments on magnetic properties.8,9,11,13,14 On the other hand, the model of an alternating chain along b ͑see Refs. 9 y , depending on which specific O͑2͒ site out of the 16 inequivalent ones is being considered͔; accounting and 14͒ requires V-O͑2͒-P-O͑2͒-V and V-O͑3͒-V paths ͑de- for the slightly different angles these orbitals will contribute picted as a heavy solid line in Fig. 1͒. If, as suggested in Ref.
50–70 % to the Eʈb spectra and 30–50 % to Eʈc. We indeed
14, the chains are additionally coupled along a there has to observe two features fulfilling this condition: a well- be a V-O͑2͒-P-O͑4͒-P-O͑2͒-V path as well; its analogs, in- developed peak P3 at 533.7 eV for Eʈb, where for Eʈc only
volving O͑3͒ instead of O͑2͒, can be excluded by noting that a strong shoulder on the low energy side of P4 is present and they would, at the same time, lead to an additional V-O͑3͒-P-O͑2͒-V coupling along c which is experimentally not the prominent feature P5 at 535.9 eV close below the main observed.14 Our NEXAFS data are consistent with these peak for Eʈb. Similarly, the line connecting neighboring P
models, first because all relevant V-O and P-O hybridiza- and O͑3͒ is oriented almost completely along c. The hybrid- tions are shown to be present, and second because the ization can take place with a symmetric combination of R-O͑3͒ hybrids exhibit less pronounced features than the O 2 px and O 2py orbitals and should show up in NEXAFS P-O͑2͒ hybrids as discussed above. A further distinction in mostly in the Eʈc spectrum. Indeed, such a spectral weight
NEXAFS between the individual contributions from the appears as a small peak P4 around 534.2 eV for Eʈc, where
two classes of crystallographically slightly inequivalent the quite broad feature P3 for Eʈb is simply falling off with-
chains30,31 is not possible: although each single absorption out any extra structures, and more prominently around 537.5 process is local the information is averaged over all these eV ( P6). In the same manner, for O͑4͒ only P 3␴ – O͑4͒ 2pz crystallographically inequivalent sites ͑due to the macro- contributions to the Eʈa spectrum are possible: they can be
scopic beam spot size and the large penetration depth͒.
observed as a weak peak A3 at 534.4 eV, and more pro- In summary, polarization-dependent NEXAFS at the nounced at 537.2 eV (A4). The fact that each kind of P-O V 2 p and O 1s edges for single-crystalline VOPO shows that bonds just discussed seems to give rise to spectral weight at the unoccupied electronic structure is highly anisotropic. The two different energies might reflect the ligand field splittingof the V 3d orbitals, mediated by the V-O-P hybridization:the same O 2 p orbitals are involved both in the V-derived t TABLE II. Assignment of the observed peaks at the O 1s edge above 532 eV to P 3␴ – O 2p hybridized orbitals.
g peaks as well as in the P 3 ␴ – O 2 p peaks. This would support the suggestion of coherent molecular orbitals provid- ing the superexchange path via the PO4 groups in VOPO.28Alternatively, the energetically higher peak of each pair could also be due to P 3d – O 2 p hybrids.
With these assignments of the unoccupied electronic structure near threshold one can now test if and to what PHYSICAL REVIEW B 63 073103
orbital character of the features observed close to the O 1s Fruitful discussions with E. Pellegrin, as well as his criti- edge is V 3d derived; states with P 3␴ character start to cal reading of this manuscript, are gratefully acknowledged.
appear at somewhat higher energy. The energetic sequence We thank J.-H. Park and S. L. Hulbert for valuable experi- observed for the crystal-field split e mental help. Research was carried out in part at the National V-O͑1b͒ hybridization is negligible, and thus that VOPO is Synchrotron Light Source, Brookhaven National Laboratory, not a spin ladder. Our results are consistent with the now which is supported by the U.S. Department of Energy, Divi- generally accepted models describing VOPO as an alternat- sion of Material Sciences and Division of Chemical Sci- ences, under Contract No. DE-AC02-98CH10886.
*Present address: 1. Physikalisches Institut, Universita¨t Stuttgart, 17 F. M. F. de Groot, J. Electron Spectrosc. Relat. Phenom. 67, 529
Pfaffenwaldring 57, D-70550 Stuttgart, Germany.
Guest scientist from A. F. Ioffe Physical Technical Institute, St.
S. Gerhold, C. A. Kuntscher, G. Linker, M. Merz, N. Nu¨cker, C.
A. Kuntscher, S. Schuppler, S. Stadler, Y. U. Idzerda, W. H.
1 Y. E. Gorbunova and S. A. Linde, Dokl. Akad. Navk. Nauk.
Tang, J. Gao, A. Attenberger, H. Berger, A. V. Prokofiev, F.
SSSR 245, 584 ͑1978͒ ͓Sov. Phys. Dokl. 24, 138 ͑1979͔͒.
Bu¨llesfeld, and W. Assmus ͑unpublished͒.
2 P. T. Nguyen, R. D. Hofmann, and A. W. Sleight, Mater. Res.
M. Knupfer ͑private communication͒.
Bull. 30, 1055 ͑1995͒.
E. Goering, O. Mu¨ller, M. L. den Boer, and S. Horn, Physica B 3 E. Dagotto and T. M. Rice, Science 271, 618 ͑1996͒.
194–196, 1217 ͑1994͒.
R. Zimmermann, R. Claessen, F. Reinert, P. Steiner, and S.
R. S. Eccleston, T. Barnes, J. Brody, and J. W. Johnson, Phys.
Hu¨fner, J. Phys.: Condens. Matter 10, 5697 ͑1998͒.
Rev. Lett. 73, 2626 ͑1994͒.
F. M. F. de Groot, J. Electron Spectrosc. Relat. Phenom. 62, 111
J. Kikuchi, T. Yamauchi, and Y. Ueda, J. Phys. Soc. Jpn. 66,
23 V. Eyert and K.-H. Ho¨ck, Phys. Rev. B 57, 12727 ͑1998͒.
6 R. C. Eccleston and M. J. Steiner, Physica B 234-236, 544
to VOPO is also limited by the different 3d occupancies of the 7 D. C. Johnston, J. W. Johnson, D. P. Goshorn, and A. J. Jacobson, Phys. Rev. B 35, 219 ͑1987͒.
25 Following the notation of Ref. 15, we denote the P 3sp3 orbitals 8 A. W. Garrett, S. E. Nagler, T. Barnes, and B. C. Sales, Phys.
Rev. B 55, 3631 ͑1997͒.
26 For a contracted VO6 octahedron the 3d3z2-r2 orbital lies higher in 9 A. W. Garrett, S. E. Nagler, D. A. Tennant, B. C. Sales, and T.
energy than the 3dx2-y2 orbital; for an elongated octahedron the Barnes, Phys. Rev. Lett. 79, 745 ͑1997͒.
situation is reversed. For equally strong V-O ͑1a͒ and V-O ͑1b͒ 10 The experiments in Ref. 9 were performed with a matrix of ori- bonds one would expect broadened eg peaks at the same energy position for Eʈa and for Eʈ(100), in contrast to the experimen-
11 A. V. Prokofiev, F. Bu¨llesfeld, W. Assmus, H. Schwenk, D.
Wichert, U. Lo¨w, and B. Lu¨thi, Eur. Phys. J. B 5, 313 ͑1998͒.
27 W. A. Harrison, Electronic Structure and the Properties of Solids 12 A. V. Prokofiev, F. Bu¨llesfeld, and W. Assmus, Cryst. Res. Tech- nol. 33, 157 ͑1998͒.
28 D. Beltra´n-Porter, P. Amoro´s, R. Iba´n˜ez, E. Martı´nez, and A.
13 A. Weisse, G. Bouzerar, and H. Fehske, Eur. Phys. J. B 7, 5
Beltra´n-Porter, Solid State Ionics 32Õ33, 57 ͑1989͒.
29 P. W. Anderson, Phys. Rev. 79, 350 ͑1950͒.
14 G. S. Uhrig and B. Normand, Phys. Rev. B 58, R14 705 ͑1998͒.
30 Z. Hiroi, M. Azuma, Y. Fujishiro, T. Saito, M. Takano, F. Izumi, 15 X. W. Lin, Y. Y. Wang, V. P. Dravid, P. M. Michalakos, and M.
T. Kamiyama, and T. Ikeda, J. Solid State Chem. 146, 369
C. Kung, Phys. Rev. B 47, 3477 ͑1993͒.
16 W. J. Veigele, in Handbook of Spectroscopy, edited by J. W.
31 J. Kikuchi, K. Motoya, T. Yamauchi, and Y. Ueda, Phys. Rev. B Robinson ͑CRC Press, Cleveland, 1974͒, Vol. 1, p. 28.
60, 6731 ͑1999͒.

Source: http://www.physics.montana.edu/magnetism/publications/PhysRev_B_2001_vol63.pdf