PHYSICAL REVIEW B, VOLUME 63, 045115

Electronic structure, bonding, and ground-state properties of AlB2 -type transition-metal diborides
P. Vajeeston,1,2 P. Ravindran,1 C. Ravi1,*, and R. Asokamani1
2

Department of Physics, Anna University, Chennai-25, India Department of Chemistry, University of Oslo, Box 1033, Blindern, N-0315, Oslo, Norway Received 29 July 2000; published 9 January 2001

The electronic structure and ground state properties of AlB2 type transition metal diborides TMB2 TM Sc, Ti, V, Cr, Mn, Fe, Y, Zr, Nb, Mo, Hf, Ta have been calculated using the self consistent tight-binding linear mufn-tin orbital method. The equilibrium volume, bulk moduli (B 0 ), pressure derivative of bulk moduli (B 0 ), cohesive energy (E coh), heat of formation ( H), and electronic specic heat coefcient ( ) are calculated for these systems and compared with the available experimental and other theoretical results. The bonding nature of these diborides is analyzed via the density of states DOS histogram as well as the charge density plots, and the chemical stability is analyzed using the band lling principle. The variation in the calculated cohesive properties of these materials is correlated with the band lling effect. The existence of a pseudogap in the total density of states is found to be a common feature for all these compounds. The reason for the creation of the pseudogap is found to be due to the strong covalent interaction between boron p states. We have made spin polarized calculations for CrB2 , MnB2 , and FeB2 and found that nite magnetic moments exist for MnB2 and CrB2 whereas FeB2 is nonmagnetic. DOI: 10.1103/PhysRevB.63.045115 PACS number s : 71.20. b, 64.30. t, 75.10.Lp

I. INTRODUCTION

The modern scientic and technical revolution is responsible for increasing interest and impetus in the search for materials possessing specic and desired properties. The transition metal borides can be classied as highly refractory and corrosion-resisting compounds,1 and the transition metal compound (Nd2 Fe14B) has been recently applied as a highperformance permanent magnetic material. From a more scientic point of view, the monoborides have INVAR behavior,2 while the diborides of transition metal have a unique combination of properties such as high melting point, hardness, chemical stability, high thermal conductivity, low electrical resistivity, and low work function.3 Since this combination is markedly different from that of parent materials, a great deal of research has been devoted to investigating the interaction between boron and the metal atoms.3,4 Especially the TiB2 compound is a potential despersoid for the development of light-weight high-temperature structural materials. These AlB2 -type transition metal diborides have attracted researchers from different elds and many experimental57 as well as theoretical studies are going on in these materials.811 Although a number of excellent research works have been published on the structure and physical properties of the borides,1214 the bonding nature of these compounds is not very clear yet.12 So, we have tried to explain the bonding nature with density-of-states DOS histogram and charge density plots. Recently, band structure calculations have been performed on AlB2 -type compounds by Wang et al.,12 in which they have explained the chemical stability of these AlB2 -type diborides in terms of band lling principle.12 In the present work, the chemical stability is explained in terms of the band lling principle and also by the density of states at the Fermi level N(E F ). Though the electronic specic heat for these compounds had been reported by coefcient others,12,15 the results of the present theoretical work are more reliable because the calculations are performed through
0163-1829/2001/63 4 /045115 12 /$15.00

k-point optimization procedure. The bulk modulus B 0 of these compounds are also calculated and these B 0 values are compared with the available experimental16 and theoretical17,18 values. As there are no reported theoretical values of heat of formation ( H) for these compounds, an attempt is made to theoretically calculate H values which are compared with the available experimental values.19 We have made spinpolarized calculations for FeB2 , CrB2 , and MnB2 compounds. Mohn et al.14,20,21 have made theoretical studies on magnetic properties of mono-, semi-, and diborides using the augmented spherical wave ASW method. We compare our results with their results and also with other experimental22 results. We present the results of the local-density approximation LDA based investigation, using the self-consistent tight binding-linear mufn-tin orbital TB-LMTO method. The rest of this paper is organized in the following way. In Sec. II, the crystallographic structures of these transition metal diborides are explained. In Sec. III, the details of the calculations are presented. The total and site projected density of states DOS for all the compounds are reported in Sec. IV. In Sec. V the nature of chemical bonding is analyzed with the help of charge density plots. Cohesive properties for all the compounds are given in Sec. VI. In Sec. VII, the chemical stability of these systems is explained. The magnetic properties of the MnB2 and CrB2 are explained in Sec. VIII. A summary of the results is given in the last section.
II. THE STRUCTURAL ASPECTS OF THE TRANSITION METAL DIBORIDES

The crystal structure of AlB2 -type transition metal diborides is designated as C32 with the space group symmetry P6/mmm. It is simply a hexagonal lattice in which closepacked TM layers are present alternative with graphite-like B layers. These diborides cannot be exactly layered compounds because the interlayer interaction is strong even though the TM layers alternate with the B layers in their crystal struc2001 The American Physical Society

63 045115-1

VAJEESTON, RAVINDRAN, RAVI, AND ASOKAMANI TABLE I. The calculated and experimental lattice parameters (a,c in , equilibrium volume in 3 /atom), the TM-B and B-B bond distance in for TMB2 compounds. Compound Present a ScB2 TiB2 VB2 CrB2 MnB2 FeB2 YB2 ZrB2 NbB2 MoB2 HfB2 TaB2 3.154 3.070 2.983 2.949 2.924 2.931 3.314 3.197 3.107 3.081 3.166 3.115 c 3.523 3.262 3.047 3.045 2.950 2.921 3.855 3.561 3.328 3.101 3.499 3.244 Experimental Volume TM-B a 3.148 3.038 2.998 2.969 3.009 3.045 3.290 3.170 3.115 3.041 3.139 3.294 c 3.516 3.22 3.056 3.066 3.039 3.035 3.835 3.533 3.265 3.066 3.473 3.886 10.115 8.876 7.825 7.642 7.281 7.242 12.223 10.502 9.271 8.495 10.124 9.085 2.533 2.409 2.297 2.284 2.242 2.235 2.716 2.564 2.447 2.359 2.530 2.421 1.821 1.773 1.722 1.703 1.688 1.674 1.914 1.846 1.794 1.779 1.828 1.799 B-B

PHYSICAL REVIEW B 63 045115

FIG. 1. Total energy vs number of k points in TiB2 .

ture. The boron atoms lie on the corners of hexagons with three nearest neighbor boron atoms in each plane. The TM atoms lie directly in the centers of each boron hexagon, but midway between adjacent boron layers; each TM atom has 12 nearest neighbor B atoms, six nearest neighbor in plane TM atoms. There is one formula unit per primitive cell and the crystal has simple hexagonal symmetry (D 6h ), whose crystal structure and Brillouin zone are found in earlier publications.2225 By choosing appropriate primitive lattice 1 1 vectors, the atoms are positioned at TM 0,0,0 , B ( 3 , 6 , 1 ), 2 2 and B ( 3 , 1 , 1 ) in the unit cell. The distance between 3 2 TMTM is equal to c. This structure is quite close packed, and can be coped with efciently and accurately by the atomic sphere approximation ASA method.2629 The valence states considered in the present calculation for the boron atoms are 2s, 2p, and 3d and for the transition metals are 3d, 4s, and 4p for the rst series, 4d, 5s, and 5p for the second series and 5d, 6s, and 6 p for the third series. The crystallographic parameters of TMB2 compounds used in the present calculations are listed in Table I.
III. METHOD OF COMPUTATION

WS sphere radii are such that an overlap is below 9%. The average WS radius was scaled so that the total volume of all the spheres is equal to the equilibrium volume of the unit cell. The calculations are semirelativistic, i.e., except spinorbit coupling, all the other relativistic effects are included. The combined correction terms are also included in the calculation. The Brillouin zone k-point integrations are made using the tetrahedron method on a grid of 1456k points in the irreducible part of the hexagonal Brillouin zone IBZ , which corresponds to 27 000 k points throughout the Brillouin zone. The optimized k points versus energy of TiB2 is shown in Fig. 1. We nd that above 1456 k points the energy becomes constant. So, in all our calculations, we use 1456 k points in the IBZ. We have used optimized c/a and equilibrium volume obtained from our calculations for the cohesive properties study. The calculations are done at different cell volumes for each system and the corresponding total energies are evaluated self-consistently by iteration to an accuracy of 10 6 Ry. For the borides which possess Cr, Mn, or Fe as one of the constituents, we have made spin polarized calculation to look for spontaneous magnetization.
IV. DENSITY OF STATES

To calculate the electronic ground-state properties of the transition-metal diborides, we have used the TB-LMTO method of Andersen.27,28 The von BarthHedin30 parametrization is used for the exchange correlation potential within the local density approximation. The LDA theory has been shown by many researchers during the past few decades to give accurate ground-state properties such as equilibrium volume, cohesive energies of elements and solids, bulk modulus, heat of formation, and electronic specic heat coefcient of intermetallic compounds.3134 In the present calculation, we have used ASA. In this approximation, the crystal is divided into space lling spheres, and therefore with slightly overlapping spheres centered on each of the atomic sites. In all our calculations reported here, the WignerSeitz

The site projected and total density of states DOS of TMB2 are shown in Figs. 2 and 3 where the vertical line indicates Fermi level E F . The DOS histogram of TMB2 consists of three parts: a the peak present in the lower energy part of the DOS curve which is mainly due to the localized or tightly bound s electrons of B; b the bonding states of TM-d and B-2p orbitals near the Fermi level; and c the top of the DOS curve due to antibonding states. It is found that the B-s electrons in TMB2 are localized and naturally its effect in bonding is very small. The electrons from TM-d and the B-2p states both contribute to the density of states at the Fermi level. The DOS of TM-d and B-2p are energetically degenerate from the bottom of the valence band to the Fermi level, indicating the possibility of covalent bonding

045115-2

ELECTRONIC STRUCTURE, BONDING, AND GROUND- . . .

PHYSICAL REVIEW B 63 045115

FIG. 2. Total and site projected density of states for 3d transition metal diborides ScB2 , TiB2 , VB2 , CrB2 , MnB2 , and FeB2 .

FIG. 3. Total and site projected density of states for YB2 , ZrB2 , NbB2 , MoB2 , HfB2 , and TaB2 .

between TM and B atoms in all these compounds. However, the spatial separation between TM and B species and the charge transfer effect prevent the hybridization effect. As a result, our charge density studies show that the bonding between TM and B atoms in early transition metals diborides are not dominated by covalent bonding. From the DOS curves given in Figs. 2 and 3, it is clear that these compounds have close similarity. Further, all these compounds possess nite N(E F ) at the Fermi level. Hence, these diborides are said to exhibit metallic behavior in their crystalline state. From our detailed investigation of charge density distribution between various atoms in TMB2 compounds we found that the metallic bonding between TM is the principal cause for the metallicity in these compounds. From the DOS histogram we found that the d-DOS bonding state peak is systematically shifting towards the lower energy side as one moves from Sc to Fe, Y to Mo, and Hf to Ta in the 3d, 4d, and 5d transition metal series respectively due to increase in the number of valence electrons see Figs. 2 and 3 . Although the rigid-band model describes the electronic structures of these diborides fairly well, there are some differences between the band structures as expected. The density of states is largest towards the end of a transition metal series because the antibonding states at the top of the d band are the most localized d states. These antibonding states have the smallest interaction with the neighboring atoms and hence, the smallest energy spread. The height of the DOS peak is lower in the 4d series TMB2 compounds than the 3d

series because the 4d orbitals extend further from the nucleus than the 3d orbitals. Since there is no sizable increase in the atomic volumes of the 4d series TMB2 compounds compared to the corresponding 3d-series TMB2 compounds, the 5d and 4d orbitals on neighboring atoms must overlap more than do the 3d orbitals. This increased overlap leads to a larger interaction between neighbors, and thus to a larger bandwidth and a smaller density of states. The typical feature of the total DOS of these compounds is the presence of what is termed as a pseudogap a sharp valley around the Fermi energy in all these compounds. It is interesting to note that pseudogaps exist not only in crystalline solids35 and amorphous alloys,36 but also in quasicrystals.37 Two mechanisms were proposed for the formation of pseudogap in the binary alloys. One is of ionic origin and the other is owing to hybridization effects. The electronegativity difference between TM and B is low and hence the ionicity does not play a major role on bonding behavior of these compounds the percentage of ionicity is less than 8% in these compounds . Consequently the pseudogap present in TMB2 is believed to be due to covalent hybridization between TM and B atoms. Such a strong hybridization gives not only an important mixing between the states of the conduction bands but also leads to a separation of the bonding states creating a pseudogap. The pseudogap is also observed in all hexagonal-close-packed hcp transition metals38 and this is attributed to d resonance. The formation of pseudogap in TiB2 was believed to be due to the compe-

045115-3

VAJEESTON, RAVINDRAN, RAVI, AND ASOKAMANI

PHYSICAL REVIEW B 63 045115 V. CHARGE DENSITY AND BONDING

FIG. 4. Total density of states for AlB2 , YHg2 , and HfBe2 in the AlB2 structure.

tition between the strong Ti-B hybridization and the Ti 3d resonance.10 Our charge density distribution plots show that there is weaker hybridization between TM and boron atoms in the early TM diborides even though they possess pseudogap feature. This indicates that, apart from TM-B covalent hybridization, some other factor inuences the creation of pseudogap in these compounds. In order to identify the origin of the pseudogap feature in all these compounds, we have made model calculations for compounds with same structures. The total DOS curves for AlB2 , YHg2 , and HfBe2 in the AlB2 structure are shown in Fig. 4. Interestingly, the pseudogap is present in AlB2 also even though it does not have any p-d covalent interaction. Furthermore, the DOS of YHg2 does not show a pseudogap feature even though it has TMTM bond. Another example is HfBe2 , it has Hf(d) Hf(d) interaction as well as Be(s) Be(s) interaction. From Fig. 4, it should be noted that there is no noticeable pseudogap present in HfBe2 . So, we conclude that the presence of pseudogap in AlB2 -type transition metal diborides is mainly originating from B(p) B(p) covalent interaction and the TMTM or TMB covalent interactions are less signicant to the creation of pseudogap. It should be noted that the E F is lying on the pseudogap in TiB2 , ZrB2 , and HfB2 . This is due to the band lling effect since all these three compounds possess same number of electrons per atom.

On the basis of Korringa-Kohn-Rostoker KKR calculations and the x-ray photoelectron spectrum, Ihara, Hirabayashi, and Nakagawa24 proposed that the bonding nature of ZrB2 can be explained by a combination of the graphite bonding model of the boron network and the hcp metal bonding model of zirconium. A tight-binding calculation for TiB2 was performed by Perkins and Sweeney25 and found strong evidence of graphite band structure. From the viewpoint of orbital overlap, Burdett, Lanadell, and Miller13 studied the electronic structure of transition metal borides with the AlB2 structure and found that the interaction of the orbitals of the transition metal with those of the planar graphite-like net of boron atoms and interaction with those of other metals are both important in inuencing the properties of these compounds. Tian and Wang,10 from their electronic structure studies, found that strong interlayer bonding plays an important role in the formation of TiB2 . The anisotropy in the bonding behavior of TiB2 has been studied recently using the orientation-dependent electron-energy-loss spectroscopy along with the band structure calculations.39 The nature of the bonding in transition metal diborides has been described in early works by simplied models which emphasize the role of TM-TM bonding,40,41 B-B bonding,42,43 or TM-B bonding.4447 These approaches are not consistent with each other. Further, two conicting points of view have been advanced regarding the electronic structure and chemical bonding in transition-metal borides. In some studies, it has been assumed that there is electron transfer from boron to the metal41,48,49 while in others it is argued that charge transfer is in the opposite direction.43,5054 Moreover, the conclusions arrived at regarding the nature of chemical bonding in diborides from electronic structure calculations are also contradictory to each other.55,56 So, in order to have deeper understanding about the origin of bonding behaviors in AlB2 transition metal diborides, we have given the three dimensional view of charge density distribution for TiB2 in different planes where Ti-Ti, B-B, and Ti-B bondings are present in Figs. 5 a , 5 b , and 5 c , respectively. From Fig. 5 a , it is clear that the Ti-Ti bonding in TiB2 is not dominantly of covalent nature. Instead we found nite uniformly distributed charge density between Ti atoms as shown in Fig. 5 a indicating that there is a metallic bonding between Ti-Ti in TiB2 . To have better understanding about the nature of bonding between boron atoms, the charge density distribution in the 0002 plane of TiB2 is given in Fig. 5 b . From this gure, it is clear that there is strong covalent interaction between boron atoms in TiB2 . The observation of strong covalent bonding between boron atoms is consistent with experimental studies in the sense that the thermal expansion coefcient measurements on TMB2 compounds show57 that the thermal expansion coefcient in the c direction decreases with increasing radius of the metal atom and that in the a direction changes very little with the size of the metal radius. The charge density distribution between Ti and B atoms in TiB2 is shown in Fig. 5 c . From this gure the Ti atoms are chosen to be in the origin. It should be noted that there is

045115-4

ELECTRONIC STRUCTURE, BONDING, AND GROUND- . . .

PHYSICAL REVIEW B 63 045115

FIG. 5. Valence electron charge density plot for TiB2 in a the 0001 plane where the Ti-Ti bonding is present, b the 0002 plane where the B-B bonding is present, and c the (101 0) plane where the Ti-Ti, B-B, and Ti-B bondings are present where Ti atoms are in the corner. In all plots 50 contours are drawn between 0 to 0.25 electrons/a.u.3 .

nite covalent interaction between Ti and B in TiB2 and this covalent interaction is not as strong as that between boron atoms in TiB2 . The present observation of weaker covalent interaction between TM-B compared with between B-B is inconsistent with the conclusion arrived from the thermal expansion coefcient measurement.57 For all the diborides there is a considerable electron drift towards the boron atoms and this gives rise to a substantial ionic contribution to the bonding. The magnitude of this electron donation decreases from ScB2 to FeB2 in the 3d series. The present observation of nite charge transfer from TM to B in early 3d TMB2 compounds is in agreement with the conclusion recently arrived from the discrete-variational

X method.58 To analyze the variation in the bonding behavior of TMB2 compounds by band lling we have given the charge density distribution for 3d transition metal diborides in Fig. 6. From this gure one can see that the bonding between Sc-B is weaker compared with that between TM-B in other TMB2 compounds. There is a negligible charge density distribution between Sc and B in ScB2 indicating that the Sc-B bond is not dominantly covalent in nature. When considering large electronegativity difference between Sc and B we conclude that there is an ionic bonding between Sc and B. The bonding between boron atoms in all these compounds is of covalent nature and does not change signicantly along this series. This is in agreement with the experimental obser-

FIG. 6. Valence electron charge density plot for ScB2 , TiB2 , VB2 , CrB2 , MnB2 , and FeB2 in the (101 0) plane with 50 contours are drawn between 0 to 0.25 electrons/a.u.3 . In the case of CrB2 and MnB2 , the charge density is calculated from the spin-polarized calculation.

045115-5

VAJEESTON, RAVINDRAN, RAVI, AND ASOKAMANI

PHYSICAL REVIEW B 63 045115 TABLE II. Calculated pre. and experimental expt. bulk modulus (B 0 in Mbar , its pressure derivative (B 0 ), heat of formation ( H in KJ mol 1 ), and cohesive energy (E coh in Ry 1 atom ) for TMB2 compounds. Compound ScB2 TiB2 VB2 CrB2 MnB2 FeB2 YB2 ZrB2 NbB2 MoB2 HfB2 TaB2
a b

vation of very little change in the bond strength within the boron layer by the introduction of larger metal atoms.57 The covalent interaction between transition metals systematically increases with increase of d-band lling. It should also be noted that the TM-TM distance decreases from ScB2 down to FeB2 progressively, which facilitates greater p-d and d-d interaction. Our calculations show that the bonding behavior in TiB2 is the combination of ionic, covalent, and metallic nature. Even though TiB2 has high melting point and large cohesive energy, its bonding behavior is not different from the other compounds. So, the distinguished nature of ground state behavior in TiB2 among 3d TMB2 compounds is originating from the band lling effect, i.e., the falling of E F at the pseudogap. In the case of CrB2 and MnB2 , we found covalent interaction between transition metal and boron. This is the reason why our calculated magnetic moments are much smaller than that of the corresponding transition metals. For FeB2 , the calculation shows strong covalent interaction between Fe and B. As a result, the magnetic moment is completely quenched in this material.
VI. COHESIVE PROPERTIES

B0 pre. 1.91 2.13 1.75 1.56 2.1 2.3 1.41 1.95 1.01 1.6 2.16 1.82

B0 1.85 2.1 1.67 1.68 1.65 1.69 2.05 1.94 1.67 1.71 1.35 1.78

B0 expt. 2.02 2.15 2.12 -

H pre. 231.52 308.38 208.14 181.93 140.04 113.71 101.68 296.81 192.73 131.61 244.20 179.15

H expt. 307 328a 206a 133a 120a 94.14a 105a 322.59b 197a 335.98b 209.20b

E coh pre. 1.17 1.32 1.01 0.98 0.88 0.79 0.88 1.25 0.94 0.83 1.21 0.98

Reference 19. Reference 77.

It is to be noted that as the d electron number increases, the bonding states get lled accompanied by a decrease in the equilibrium volume. The equilibrium volume is the minimum when the band is half lled and beyond this, it increases with the lling of the valence band. This trend can be seen from the equilibrium volumes given in Table I. The cohesive energy of a material is a fundamental property which has long been the subject of theoretical and computational approaches. The chemical bonding is a mixture between covalent, ionic, and metallic bonding and therefore the cohesive energy cannot be determined reliably from simple models. Thus, rst principles calculations based on density functional theory DFT have become a useful tool to determine the cohesive energy of the solids. In this connection, the cohesive energy of TMB2 is calculated by using the expression E coh2
AB AB A B E atom 2E atom
2 E total ,

AB

2 where E total refers to the total energy of the compound at A B equilibrium lattice constants and E atom and E atom are the atomic energies of the pure constituents calculated semirelativistically. To determine the heat of formation, we have rst calculated the total energies of TM elements and B corresponding to their respective equilibrium lattice parameters. At zero temperature, there is no entropy contribution to the free energy, therefore the free energy of formation, or the heat of formation ( H) can be obtained from the following relation: 2 H AB 2 E total

AB

A B E solid 2E solid ,

2 where E total refers to the total energy of TMB2 at equilibrium A B lattice constants and E solid and E solid are total energy of the pure elemental constituents.

AB

The cohesive properties of 3d metal carbides and nitrides59 indicate that the number of valence electrons per atom n e is a useful variable in correlating properties related to the cohesive energy of compounds which have a similar type of chemical bonding. So, for the present study, we use n e (n TM 2n B)/3, where n TM and n B are the number of valence electrons for transition metal and boron, respectively. The calculated values of the cohesive energies and heat of formation of all systems are given in Table II. The systematic errors in total energy due to the use of ASA are canceled signicantly, leading to a reasonably accurate formation energy. The heat of formation energies calculated for these compounds are in good agreement with the experimental values.60,61 In order to understand the role of band lling on the stability of these materials, the n e vs H is shown in Fig. 7 and also a comparison is made with experimental H in this gure. It is interesting to note that the present theoretical values coincide with experimentally reported values for VB2 and early transition metal diborides. On the other hand, the experimental values are higher than the theoretical values for the diborides whose n e is greater than that of VB2 . This is partly due to the noninclusion of magnetic correlation effects in our calculation. Further, the experimental calorimetric measurements were made19 at high temperature and our calculated H value is applicable only for low temperatures. In the case of 4d and 5d series the calculated values are coinciding very well with experimental values. We also note that the 3d compounds usually show the most exothermic enthalpies of formation and that the values tend to decrease systematically as we go from 3d to 4d to 5d metals. Further, the H is higher for TiB2 , ZrB2 , and HfB2 in the 3d, 4d, and 5d series, respectively, which is related to the unique properties such as high hardness, high melting point with maximum corrosion resistance among the compounds in the series. The origin of the unique properties of these com-

045115-6

ELECTRONIC STRUCTURE, BONDING, AND GROUND- . . .

PHYSICAL REVIEW B 63 045115

P 3B 0

1 x x
2

(1 x)

FIG. 7. Number of electrons per atom (n e ) vs heat of formation ( H kJ mol 1 ) in TMB2 compounds.

pounds among the TMB2 compounds is due to band-lling effect which is discussed in detail below. The E F of TiB2 , ZrB2 and HfB2 falls near the minimum in the DOS curve in Figs. 2 and 3. This means that, all the bonding states are lled and the antibonding states are empty for the electron per atom n e 3.33. As a result, the cohesiverelated properties such as the melting temperature and the enthalpy of formation will be higher as shown in Fig. 7. In a rigid-band picture, we would start to populate antibonding states when n e is increased above 3.33e/a and consequently the cohesive-related properties will be reduced. When n e decreases below 3.33e/a one cannot make full use of all the bonding states and that also leads to a reduction in the cohesion. We observe that the lling up of bonding orbitals leads to a positive slope and the lling up of antibonding orbitals leads to a negative slope as shown in Fig. 7. Although a rigid-band approach would be too crude to give direct quantitative estimates of H, for example, it is expected to give the trends as a function of n e . From Fig. 7, it is clear that the rigid band picture works well in these materials.
A. Bulk modulus and its pressure derivative

Experimental bulk modulus values are not available for most of these compounds and this is the rst report on the values of B 0 for those compounds. An universal relationship between the binding energies and distances between atoms has been discovered for bimetallic adhesion,62 chemisorption on metals,63 and metallic cohesion.64 Later Vinet et al.65 proposed a universal model of the equation of state UEOS for all classes of solids in compression, which is claimed to be superior to that of the Birch-Murnaghan EOS.66 If we dene x as (V/V 0 ) 1/3 and H(x) as x 2 P(x)/3(1 x), the ln H(x) vs (1 x) curve should be nearly linear according to their (1 x) and the EOS at a theory; i.e., ln H(x) ln B 0 given temperature can be expressed as

3 1) is the slope of the curve. It is related to where 2 (B 0 the pressure derivative of the bulk modulus B 0 . We have calculated the values of ln H(x) and (1 x) using the pressure-volume P-V data and made the least-squares t in a manner similar to what we have made earlier.31,67,68 In the case of TMB2 , the bonding nature can be regarded as a combination of metallic, covalent, and partly ionic bonding.10 The UEOS is suitable for describing mixed bonding systems.65 To determine the bulk modulus, we have carried out self-consistent calculations for six different volumes within the volume range V/V 0 1.15 to 0.85 which is 15% around the experimental equilibrium volume. The calculated bulk modulus is given in Table II. For materials at the beginning of a transition series, the bonding orbitals begin to ll, leading to increase in cohesion and hence to a decrease in the atomic volume. This decreased atomic volume and increased compression of s orbitals lead to an increase in the bulk modulus. These effects are all maximized near the middle of the transition series,69 when all the bonding orbitals are lled, and then the trend is reversed when the antibonding orbitals begin to be lled. We also identify the same trend in TMB2 compounds currently under consideration. The calculated bulk modulus values are compared with the available experimental values in Table II. From this table, it can be seen that our calculated values are in good agreement with the experimental values. The B 0 value is found to be maximum for 5d HfB2 , followed by 3d TiB2 and 4d series diboride ZrB2 . This trend is in agreement with the reported experimental values. Generally, the compounds with high melting temperature T m are expected to have high B 0 . The melting temperatures of HfB2 , ZrB2 , and TiB2 are 3523, 3313, and 3253 K, respectively. However, the B 0 values are higher for TiB2 than for ZrB2 . The T m is dened by both bulk modulus and the shear modulus.70 So, the observation of no systematic trend between B 0 and T m indicates that the shear contribution varies signicantly among these compounds. To the best of our knowledge, B 0 values are available only for IVB compounds. Neither the theoretical nor the experimental values of B 0 are available for the rest of the compounds under study. It should be noted that the B 0 values of the present calculation are in good agreement with the available experimental values. However, the calculated B 0 value of TiB2 by Tian and Wang10 in the earlier theoretical calculation is an overestimate when compared with the experimental as well as our theoretical work. This variation in the B 0 values between ours and the Wang calculation is due to the k point optimization. So, we believe that our calculated values are more reliable for other compounds also. The pressure derivative of bulk modulus at zero pressure B 0 is a parameter of great physical signicance in high pressure physics. It is related to a few other important thermophysical properties like phase transitions, interphase energy, adsorption energy, etc. .71 Generally, the variation along the period is one of gradual increase, roughly until the middle of

045115-7

VAJEESTON, RAVINDRAN, RAVI, AND ASOKAMANI TABLE III. The calculated density of states at the Fermi level N(E F ) in states Ry 1 f.u. 1 ], electronic specic heat coefcient ( in mJ mol 1 K 2 ), and electron phonon coupling constant ( ) for TMB2 compounds. The results obtained from the spin-polarized calculations are represented by spin . Compound ScB2 TiB2 VB2 CrB2 CrB2 spin MnB2 MnB2 spin FeB2 YB2 ZrB2 NbB2 MoB2 HfB2 TaB2 N(E F ) 11.99 4.27 15.86 34.88 37.89 108.94 20.18 37.28 11.68 3.84 13.95 20.01 3.688 12.92
present exp theo

PHYSICAL REVIEW B 63 045115

2.08 0.74 2.48 6.05 6.57 18.89 3.51 6.46 2.03 0.67 2.42 3.47 0.64 2.24

2.2 1.08 4.84 13.6 4.45 2.33 3.58 1.7

2.48 0.38 3.62 5.84 1.070 8.09 2.38 0.36 2.44 2.76 0.34 2.37

0.058 0.459 0.952 1.247 0.26 0.032 -

the period, followed by a drop for the remaining elements. The increasing trend can be attributed to the increasing level of lling in the bonding d level which reaches a saturation at the middle of the period. The same behavior is not valid for the entire d-block elements. The B 0 is directly related to the electron density in the d-series elements, the element with lowest B 0 will have high electron density. It is valid for only 3d-elements and not for 4d and 5d series in the TMB2 compounds. We compare the B 0 values of the TMB2 compounds with that of constituent transition element. It is found that the B 0 of these compounds is half that of the constituent transition element. This indicates the strong interlayer and intralayer chemical bonding in TMB2 compounds.
B. Electronic specic heat coefcient

tioned above, the calculated values usually underestimate with experimental values, because the electronphonon enhancement factor is not included in the theoretical calculations. But, surprisingly, our calculated values overestimate the values in the case of CrB2 , MnB2 , and FeB2 in the 3d series, NbB2 in the 4d series, and TaB2 in the 5d series earlier works also nd this variation . At least, the discrepancy in the 3d transition metal diborides can be accounted for by the magnetic correlation effect in these compounds. For example, the value is very large compared with the value for MnB2 because of the higher experimental N(E F ). So the spin polarized calculation is made for this compound and it reduces the values considerably. Among these diborides, HfB2 has small value and the VIB comvalue. We have also estimated the pounds have high electron-phonon enhancement factor for these materials using the experimentally reported values by the relation ), where exp and th refer to the experimenexp th(1 tal and theoretical values of electronic specic heat coefcients, respectively. In the case of TaB2 and NbB2 , we have values larger than the experimentally reported obtained values. So, we have not given the values for these compounds in Table III. In order to clarify the theoretically obtained large in these compounds, more accurate electronic specic heat coefcient measurements at low temperatures are needed. In the other compounds, the calculated values are found to be much smaller than those of the superconducting materials. These TMB2 compounds do not obey the Matthias rule,72 which relates the optimal electron concentration for superconductivity 5 and 7 electrons per atom , whereas the electron per atom ratio for these compounds is less than 5. This may be the possible reason for the low values and the nonobservation of superconductivity (T c 1 K) in these compounds.73 The present observation of large values in some of the TMB2 compounds compared with the experimental study is due to the nonreliable experimental value or the failure of LDA.
VII. CHEMICAL STABILITY

The linear term in the specic heat at low temperatures is proportional to the N(E F ). Values of for transition metal mono- and diborides are measured by earlier works.10,15 In Table III, we have compared the electronic specic-heat coefcient obtained from the present study with the available theoretically calculated values and experimental values. The theoretical values are calculated directly from the free elec2 ( 2 /3)N(E F )k B , using the calcutron approximation, lated density of states at the Fermi energy. The calculated values will always be higher than the experimental values since the electronphonon enhancement effect is not taken into account in our calculations. The present theoretical values are found to be in agreement with the other theoretically calculated values as given in Table III. The calculated N(E F ) values are very sensitive to the number of k points used in the calculations. So, the discrepancy between the present results with the earlier reported12 theoretical values is partly due to the k point effect. As men-

It is well known that the chemical stability is associated with melting point. Compounds with high melting points generally have high chemical stability. It is interesting to note that the melting temperature of the elements Sc to Cr in the periodic table increases linearly. On the other hand, the melting temperature of these elements is found to decrease with the addition of boron, where the T m decreases from TiB2 to CrB2 in the 3d series, ZrB2 to MoB2 in the 4d series, and HfB2 to TaB2 in the case of 5d series. This discrepancy can be explained as follows: In the case of transition metals the number of bonding states increases with the increase of d electron number and gets maximum in the middle of the series. Since the lling of bonding states enhances the bond strength, the T m increases while going from Sc to Cr. On the other hand, all the bonding states get lled in IVB-TMB2 compound itself due to the presence of pseudogap at E F . So, the increase in d electrons when going from TiB2 to FeB2 lls the antibonding/nonbonding states. As the lling of antibonding/nonbonding is less favorable for

045115-8

ELECTRONIC STRUCTURE, BONDING, AND GROUND- . . . TABLE IV. The width of the valence band (W v in Ry , width of the bonding states (W p in Ry , W v /W p , and melting point (T m in K for the TMB2 compounds. Compound ScB2 TiB2 VB2 CrB2 MnB2 FeB2 YB2 ZrB2 NbB2 MoB2 HfB2 TaB2 Wv 0.76 0.90 0.96 1.05 1.03 1.01 0.75 0.93 1.18 1.08 0.98 1.10 Wp 0.88 0.89 0.88 0.95 0.95 0.93 0.88 0.88 0.85 0.88 0.93 0.98 W v /W p pre. 0.87 1.01 1.10 1.11 1.09 1.10 0.86 1.05 1.17 1.23 1.05 1.12 W v /W p Ref. 12 0.90 0.99 1.03 1.12 1.13 0.76 0.99 1.02 1.14 1.02 1.14 Tm 2523 32533498 26733020 2473 2261 2373 33133518 31733309 23732648 33733653 33103473

PHYSICAL REVIEW B 63 045115

chemical stability, T m decreases with increase of d electron number. The lower N(E F ) is often related with higher stability and higher melting point.35 This can be understood as follows: If a material has large N(E F ), it means that large density of electrons is present in the highest occupied level, i.e., in the vicinity of the Fermi level. It is not a favorable condition for stability since the one electron eigenvalue sum increases and hence lattice instability will arise easily resulting in lower melting point. On the other hand, if the N(E F ) value is small, it means that the electrons participate in bonding and get localized. As a result, the stability of the material will be larger and will possess higher melting point. The present observation indicates that the electrons at the Fermi level mainly decide the melting behavior of solids. We have analyzed the chemical stability of these compounds using the band lling of the bonding states as follows: As the rigid-band model describes the electronic structure of the diborides rather well, the band lling can naturally be regarded as being responsible for the variations of their electronic structure and related properties. We dene W v , the energy difference from bottom of the valence band to the Fermi level and W p , the energy width from the bottom of the valence band to the pseudogap. Since W v represents the width of the occupied states and W p represents the bonding states, W v /W p can be used to describe the occupied portion of the bonding states. Here, the W p is found to be almost constant, while the W v alone varies for different compounds depending upon the number of valence electrons. The calculated values of W v /W p for TMB2 compounds are listed in Table IV along with those reported in Ref. 12. From this table it should be noted that ScB2 has 9 valence electrons and its bonding states are not fully lled (W v /W p 1). In the case of TiB2 , W v /W p is almost equal to unity. This indicates that all the bonding states are lled and all the antibonding states are empty. From CrB2 to FeB2 the W v /W p factor systematically increases and also becomes greater than unity see Table IV . This indicates that the bonding states are already lled and the antibonding/nonbonding states get oc-

cupied when going from CrB2 to FeB2 . So the chemical stability is in the following order for the 3d transition metal diborides: TiB2 VB2 CrB2 MnB2 FeB2 , in the 4d series ZrB2 NbB2 MoB2 , and in the 5d series HfB2 TaB2 . For comparison we have given the W v /W p obtained by Wang et al. in Table IV along with our results. It should be noted that there is signicant difference between our results and those reported in Ref. 12. Since we have used k-point optimization in our calculations we believe that our results are quantitatively closer to reality. Even though there is quantitative difference between the present result and Ref. 12, from Table IV one can see that qualitatively both results are consistent with each other. The ScB2 and YB2 compounds need more electrons into the band to increase the stability of these compounds. In the 3d-transition metal diboride series TiB2 has enough space to accommodate its 10 valence electrons. So it is expected to be the most stable compound in this series. In the case of isoelectronic compounds along the group where the number of valence electrons remains the same, the chemical stability is in the following sequence: TiB2 ZrB2 HfB2 in the fourth group and VB2 NbB2 TaB2 in the fth group. The remaining compounds also follow this sequence. A similar trend is observed in the other theoretical12 as well as experimental works. The large chemical stability in the 5d series transition metal diborides over the corresponding 3d or 4d transition metal diborides is due to the enhancement in the covalent hybridization between the transition metal and boron, because when we go down the periodic table in a particular group, one can expect the delocalization of valence band due to the screening effect. Hence, compared to Ti, the outermost electrons in Hf are much delocalized. As a result of this, the valence electrons for Hf will have strong covalent interaction with the boron compared with that of Ti. This hybridization effect will separate the bonding states from the antibonding/nonbonding states which leads to low N(E F ). This may be the possible explanation for the high melting point for HfB2 over TiB2 . From the above analysis, the band structure calculations were found to explain the chemical stability of these TMB2 systems successfully.
VIII. THE MAGNETIC PROPERTIES

CrB2 has a complicated helicoidal magnetic structure, as resolved by neutron diffraction measurements.74 But for simplicity, we have assumed the ferromagnetic ordering in our calculation. Early magnetization and susceptibility measurements yielded rather surprising results that MnB2 is a simple ferromagnet with a small saturated moment of 0.25 B /Mn atom49 and 0.19 B /Mn atom75 and an ordering temperature of 157 3 K49 and 143 K.75 FeB2 is metastable in the Fe-B system, and its magnetic properties are not known. However, the possible existence of a magnetic transition in this phase cannot be ruled out in view of the ferromagnetic transitions in the stable iron boride phases Fe2 B and FeB.22,76 The spin polarized calculations are performed for CrB2 , MnB2 , and FeB2 . Filling up of the d band moves the Fermi energy into a region of high DOS where both CrB2 and MnB2 fulll the Stoner criterion. FeB2 also has high DOS at

045115-9

VAJEESTON, RAVINDRAN, RAVI, AND ASOKAMANI

PHYSICAL REVIEW B 63 045115

FIG. 8. The total energy vs unit cell volume curves for MnB2 from the spin-polarized and non-spin-polarized calculation where E E 2412.

FIG. 9. The spin projected DOS for MnB2 in the ferromagnetic phase.

the Fermi level, but we do not observe spontaneous spin polarization in this case. As mentioned above, our charge density analysis shows that there is a strong covalent bonding between Fe and B in FeB2 compared to the other transition metal compounds considered here. So, the observation of nonmagnetic behavior in FeB2 is due to the participation of electrons in the chemical bonding instead of magnetism. The variation of total energy with volume for MnB2 with and without spin polarization are given in Fig. 8. From this curve, it is clear that a large gain in total energy is observed by the inclusion of spin polarization in our calculation. So, our calculation predicts that the ferromagnetic state is energetically more favorable than the nonmagnetic state in MnB2 . It is consistent with the experimental studies in the sense that the magnetization and susceptibility studies suggest49,75 an itinerant band ferromagnetic behavior in MnB2 . However, our calculation yields the magnetic moment of 1.6 B /Mn, and this is much higher than the experimentally observed value. Very recent theoretical studies21 on MnB2 suggest that spin canting structure could be responsible for the small ferromagnetic component found in the experiment. The equilibrium cell volume of the spinpolarized case is 155.632 a.u.3 , 149.345 a.u.3 and that of the nonspin-polarized case is 154.763 a.u.3 , 149.325 a.u.3 for CrB2 and MnB2 , respectively. The enhancement in the equilibrium volume by the inclusion of spin polarization in our calculation is due to magnetovolume effect. The calculated total, site, and spin projected DOS for MnB2 obtained from the spin polarized calculation are shown in Fig. 9. It should be noted that the E F falls on a sharp peak Fig. 2 in the nonspin polarized case of MnB2 see Fig. 2 . On the other hand, the spin polarization splits the Mn-d DOS in such a way that the E F lies in a valley region in the total DOS as shown in Fig. 9. As a consequence of this, we have obtained a small value of compared with that of the nonspin-polarized case which is closer to the experimentally reported value Table III . In the case of CrB2 , we observed a very small magnetic moment of 0.32 B /Cr. Hence, the DOS does not change signicantly when going

from the nonspin-polarized case to the spin-polarized case. As a result, the E F is present in the shoulder of DOS peak which possesses a dominant nonbonding Cr 3d character. This is not a favorable condition for stability. This may be the possible reason for the stabilization of complex helicoidal magnetic structure in CrB2 . Our calculated magnetic moment of CrB2 is found to be close to the experimentally measured74 magnetic moment of 0.5 0.1 B /Cr. The gain in total energy by spin polarization obviously reects the formation energy and hence the recalculated formation 150.43 and energy for MnB2 and CrB2 are 184.24 KJ mol 1 , respectively. The previous calculations on mono-14 and semiboride20 of Fe show a reduction in the magnetic moment on boron addition to Fe with the magnetic moment values of 1.91 B /Fe for Fe2B and 1.12 B /Fe for FeB. From our calculations we found that further addition of boron in FeB is found to suppress the magnetic moment in the FeB2 case. But the opposite trend was observed in the case of MnB2 where Khmelevsky and Mohn21 found that the addition of B in the Mn2 B matrix increases the magnetic moment up to MnB. In the case of MnB2 , we found that the magnetic moment is exactly the same as in Mn2 B, indicating an increase in the magnetic moment on addition of B up to a certain range after which further addition of boron reduces the magnetic moment. In the case of CrB and Cr2 B, no magnetic moment is observed. But in the CrB2 case, we nd a small amount of magnetic moment.
IX. SUMMARY AND CONCLUSION

We have performed rst-principles local density functional electronic structure calculation for the 12 TMB2 compounds using the TB-LMTO-ASA method. The calculated lattice constants are found to be in very good agreement with experimental results. These TMB2 compounds are not exactly layered compounds because of the d-d interaction among the TM and the interaction of TM-d electrons with the p electrons of the B atom present in these phases. The

045115-10

ELECTRONIC STRUCTURE, BONDING, AND GROUND- . . .

PHYSICAL REVIEW B 63 045115

cohesive energy, heat of formation, and bulk modulus for most of these compounds are calculated for the rst time. We list the important conclusions arrived from our calculations. 1 We successfully explained the chemical bonding behavior of AlB2 -type TMB2 compounds from our DOS and charge density analysis. The covalent nature of bonding in these compounds systematically increases with increase of n e along the series. These compounds possess mixed bonding nature such as strong covalent bonding between boron, metallic, and covalent bonding between transition metals, and ionic and covalent bonding between TM and B. 2 The calculated cohesive properties such as heat of formation are found to be in good agreement with the experimental values and the change in trend along the series is explained via the band lling of the bonding state analysis. 3 The calculated electron phonon enhancement factor for TMB2 compounds is very small compared with the superconducting compounds. 4 From the DOS histogram, we found that pseudogap is a common feature for all these compounds. The most popular belief for the origin of pseudogap is that it arises due to bonding between nonmetal and transition metal the p-d nonmetal atoms in most of the intermetallic compounds.78

But, in the present systems, the common origin for the creation of pseudogap in all these materials is due to strong covalent bonding between boron atoms. 5 Calculated bulk modulus and its pressure derivatives are compared with available experimental and other theoretical values. The calculated bulk modulus values are found to be in good agreement with experimental values. 6 The electron band structure total energy studies show that the cohesive-related properties show a maximum in bond strength for TiB2 in the 3d TMB2 compounds and this is due to the maximum lling of bonding states. 7 We found nite magnetic moment in MnB2 and CrB2 from our spin-polarized calculations. Due to strong covalent bonding between Fe and B, our calculation predicts a nonmagnetic behavior in FeB2 .
ACKNOWLEDGMENTS

The rst author P.V. gratefully acknowledges the nancial support from DST, India and R. Vidya and R. Rita for their encouragement at several stages of this work. P.R. wishes to thank P.A. Korzhavyi and L. Offernes for useful discussions.

*Electronic address: ravi@kjemi.uio.no

E. A. Knyshev, V. M. Novgorodtsev, U. S. Plyshevski, V. A. Kobyakov, Z. G. Stepanova, V. V. Svistunov, and A. R. Becketov, J. Less-Common Met. 47, 273 1976 . 2 E. P. Wohlfarth, J. Magn. Magn. Mater. 7, 113 1978 . 3 W. Gordon and S. B. Soffer, J. Phys. Chem. Solids 36, 627 1975 . 4 B. Loundquist, H. Myers, and R. Westin, Philos. Mag. 33, 1187 1960 . 5 V. A. Neronov, M. A. Korchagin, V. V. Aleksandrov, and S. N. Gusenko, J. Less-Common Met. 82, 125 1981 . 6 V. S. Sinelnikova, V. N. Gurin, A. N. Pilyankevich, L. V. Strashinskaya, and M. M. Korsukova, J. Less-Common Met. 47, 265 1976 . 7 S. Otani and Y. Ishizawa, J. Cryst. Growth 140, 451 1994 . 8 T. Lundstrom, B. Lonnberg, and I. Westman, J. Less-Common Met. 96, 229 1984 . 9 B. Champagne, S. Dallaire, and A. Adnot, J. Less-Common Met. 98, L21 1983 . 10 D. C. Tian and X. B. Wang, J. Phys.: Condens. Matter 4, 8765 1992 . 11 R. R. Atri, K. S. Ravichandran, and S. K. Jha, Mater. Sci. Eng., A 271, 150 1999 . 12 X. B. Wang, D. C. Tian, and L. Wang, J. Phys.: Condens. Matter 6, 10 185 1994 . 13 J. K. Burdett, E. Lanadell, and G. J. Miller, J. Am. Chem. Soc. 108, 6561 1986 . 14 P. Mohn and D. G. Pettifor, J. Phys. C 21, 2829 1988 . 15 J. Castaing, R. Caudron, G. Toupance, and P. Costa, Solid State Commun. 7, 1453 1969 . 16 D. E. Wiley, W. R. Manning, and O. Hunter, J. Less-Common Met. 8, 149 1969 . 17 P. S. Spoor, J. D. Maynard, M. J. Pan, D. J. Green, J. R. Hellmann, and T. Tanaka, Appl. Phys. Lett. 70, 9159 1997 .
1

18 19

S. I. Wright, J. Appl. Crystallogr. 27, 794 1994 . L. Topor and O. J. Kleppa, J. Chem. Thermodyn. 17, 1003 1985 . 20 P. Mohn, J. Phys. C 21, 2841 1988 . 21 S. Khmelevskyi and P. Mohn, Solid State Commun. 113, 509 2000 . 22 M. C. Cadeville and A. J. P. Meyer, C. R. Hebd. Seances Acad. Sci. 255, 3391 1962 . 23 S. H. Liu, L. Kopp, W. B. England, and H. W. Myron, Phys. Rev. B 11, 3465 1975 . 24 H. Ihara, M. Hirabayashi, and H. Nakagawa, Phys. Rev. B 16, 726 1977 . 25 P. G. Perkins and A. V. J. Sweeney, J. Less-Common Met. 47, 165 1976 . 26 O. K. Andersen, Phys. Rev. B 12, 3060 1975 . 27 H. L. Skriver, The LMTO Method Springer, Heidelberg, 1984 . 28 O. K. Andersen and O. Jepsen, Phys. Rev. Lett. 53, 2571 1984 . 29 O. K. Andersen, in Methods of Electronic Structure Calculations, edited by O. K. Andersen, V. Kumar, and A. Mookerjee World Scientic, Singapore, 1994 . 30 U. von Barth and L. Hedin, J. Phys. C 5, 1629 1972 . 31 P. Ravindran and R. Asokamani, Phys. Rev. B 50, 668 1994 . 32 S. Otani, M. M. Korsukova, and T. Mitsuhashi, J. Cryst. Growth 186, 582 1998 . 33 M. L. Cohen, Phys. Rep. 110, 293 1983 . 34 A. R. Williams, J. Kubler, and C. D. Gelatt, Phys. Rev. B 19, 6094 1979 . 35 J. H. Xu and A. J. Freeman, Phys. Rev. B 41, 12 553 1990 . 36 A. Pasturel, C. Colinet, and P. Hicter, Physica B & C 132, 177 1985 . 37 J. C. Phillips, Phys. Rev. B 47, 2522 1992 . 38 O. Jepsen, O. K. Andersen, and A. R. Mackintosh, Phys. Rev. B 12, 3084 1975 .

045115-11

VAJEESTON, RAVINDRAN, RAVI, AND ASOKAMANI

39

PHYSICAL REVIEW B 63 045115

60 61

K. Lie, R. Brydson, and H. Davock, Phys. Rev. B 59, 5361 1999 . 40 E. Dempsey, Philos. Mag. 8, 285 1963 . 41 H. J. Juretschke and R. Stinitz, J. Phys. Chem. Solids 4, 118 1958 . 42 G. V. Samsonov, Yu. M. Goryachev, and B. A. Kovenskaya, J. Less-Common Met. 47, 147 1976 . 43 W. N. Lipscomb and D. Britton, J. Chem. Phys. 33, 275 1960 . 44 K. E. Spear, J. Less-Common Met. 47, 195 1976 . 45 B. Post, F. W. Glaser, and D. Moskowitz, Acta Metall. 2, 20 1954 . 46 J. Piper, J. Phys. Chem. Solids 27, 1907 1966 . 47 Y. S. Tyan, L. E. Toth, and Y. A. Chang, J. Phys. Chem. Solids 30, 785 1969 . 48 R. Kiessling, Acta Chem. Scand. 4, 209 1958 . 49 M. C. Cadeville, J. Chem. Solids 27, 667 1966 . 50 L. Pauling, Proc. R. Soc. London, Ser. A 196, 343 1949 . 51 H. C. Longuet-Higgins and M. Roberts, Proc. R. Soc. London, Ser. A 224, 336 1954 . 52 R. W. Johnson and A. H. Daane, J. Chem. Phys. 38, 425 1963 . 53 G. V. Samsonov and B. A. Kovenskaya, in Boron and Refractory Borides, edited by V. I. Matkovich Springer-Verlag, Berlin, 1977 , pp. 5,19. 54 B. Post, in Boron, Metallo-Boron Compounds and Boranes, edited by R. M. Adams Interscience Publishers, New York, 1964 , pp. 301,313. 55 V. M. Anishchik and N. N. Dorozhkin, Phys. Status Solidi B 160, 173 1990 . 56 D. R. Armstrong, Theor. Chim. Acta 64, 137 1983 . 57 B. Lonnberg, J. Less-Common Met. 141, 145 1988 . 58 M. Mizuno, I. Tanaka, and H. Adachi, Phys. Rev. B 59, 15 033 1999 . 59 A. F. Guillermet and G. Grimvall, Phys. Rev. B 40, 1521 1989 .

S. V. Meschel and O. J. Kleppa, Metall. Trans. A 24, 947 1993 . L. A. Reznitskii, Russ. J. Phys. Chem. 41, 612 1967 . 62 J. H. Rose, J. Ferrante, and J. R. Smith, Phys. Rev. Lett. 47, 675 1981 . 63 J. R. Smith, J. Ferrante, and J. H. Rose, Phys. Rev. B 31, 3427 1985 . 64 J. Ferrante, J. R. Smith, and J. H. Rose, Phys. Rev. B 28, 1835 1983 . 65 P. Vinet, J. H. Rose, J. Ferrante, and J. R. Smith, J. Phys.: Condens. Matter 1, 1941 1989 . 66 F. D. Murnaghan, Proc. Natl. Acad. Sci. U.S.A. 30, 244 1944 . 67 P. Ravindran and R. Asokamani, Phys. Rev. B 53, 1129 1996 . 68 C. Ravi, P. Vajeeston, S. Mathijaya, and R. Asokamani, Phys. Rev. B 60, 15 683 1999 . 69 A. K. McMahan, H. L. Skriver, and B. Johansson, in Physics of Solids Under High Pressures, edited by J. Schilling and R. N. Shelton North-Holland, Amsterdam, 1981 , p. 169. 70 F. Mulargia and F. Quareni, Geophys. J. 92, 269 1988 . 71 S. Raju, E. Mohandas, and V. S. Raghunathan, J. Phys. Chem. Solids 58, 1367 1997 . 72 B. T. Matthias, Prog. Low Temp. Phys. 2, 138 1957 . 73 L. Leyarovska and E. Leyarovski, J. Less-Common Met. 67, 249 1979 . 74 S. Funahashi, Y. Hamaguchi, T. Tanaka, and E. Bannai, Solid State Commun. 23, 859 1977 . 75 L. Andersson, B. Dellby, and H. P. Myers, Solid State Commun. 4, 77 1966 . 76 N. Lundquist, H. P. Myers, and P. Westin, Philos. Mag. 7, 1187 1962 . 77 J. Samuel and Jr. Schneider, ASM Engineering Materials Hand Book, Vol. 4 ASM International, Materials Park, OH, 1991 , pp. 790795. 78 M. Weinert and R. E. Watson, Phys. Rev. B 58, 9732 1998 .

045115-12