Scripta Materialia 55 (2006) 811–814 www.actamat-journals.com On similarities and differences of the electronic structure for Cu(II)/O2 and Ag(II)/F1 infinite layer compounds Wojciech Grochala* Department of Chemistry, Warsaw University, Pasteur 1, 02093 Warsaw, Poland Interdisciplinary Center for Mathematical and Computational Modeling, Warsaw University, Pawińskiego 5a, 02106 Warsaw, Poland Received 24 June 2006; accepted 6 July 2006 This work commemorates the 120th anniversary of the discovery of elemental F2 by Moissant We calculate and compare the band structures of a hypothetical difluoride of Ag(II) with flat [AgF2] layers and of known CaCuO2 infinite layer compound. Electronic structure of [AgF2] is strikingly similar to that of Ca[CuO2], but there are subtle differences as well. Calculations suggest that charge-doped [AgF2] materials show good prospects for superconductivity, provided that delocalization of extra charge is avoided. 2006 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Silver; Fluorine; Copper; Oxygen; Superconductivity Fluorides of divalent silver (Ag2+) constitute a fascinating family of materials. Divalent silver is unique to fluorides, and so far it has not been found in other ligand environments. Due to its enormously large electron affinity, it is capable of oxidizing ‘noble’ gas Xe, and vigorously reacts with elemental metals including the most electronegative Au or Pt. It has been postulated – based on a multitude of appealing qualitative and semi-quantitative arguments – that properly doped fluorides of Ag2+ might exhibit high-temperature superconductivity [1]. Recent experimental study of Cs2AgF4 [2] has revitalized interest in layered fluorides of Ag2+. Cs2AgF4 crystallizes orthorhombically [2], and it contains flat [AgF2] sheets separated by two [CsF] spacers; its structure is related to the structure of La2CuO4, parent (undoped) compound of the first oxocuprate superconductor, La2xBaxCuO4 [3]. Prediction of superconductivity in doped fluorides of Ag2+ [1] and recent study of Cs2AgF4 [2] call for careful re-examination of the electronic structure of layered fluorides of Ag2+. In this contribution we calculate and compare the band structures of a hypothetical difluoride of Ag(II) with flat [AgF2] sheets and of known * Tel.: +48 22 8220211x276; [email protected] fax: +48 22 8225996; e-mail: CaCuO2, an infinite layer compound [4]. Difluoride of Ag(II) with flat [AgF2] layers is not purely a theoretical creation; such material might indeed be obtained from genuine binary AgF2 (with its puckered sheet structure) by (i) intercalation of various oxidation-resistant Lewis bases between the sheets, with their concomitant flattening, or (ii) by applying external pressure [3]. We will subsequently denote hypothetical two-dimensional (2D) difluoride of Ag(II) simply as [AgF2]. The CASTEP code was used for all calculations [5]. Our density functional theory calculations utilized the Perdew–Burke–Ernzerhof correlation–exchange functional, typical cut-off of 380 eV, k-point grid of 6 · 6 · 7 or 6 · 6 · 8, and ultrasoft Vanderbildt pseudopotentials. Spin polarization was not enforced. The tetragonal P4/mmm cell with the unit cell vectors of a = b = 3.854 Å and c = 3.200 Å [4] was adopted for CaCuO2, as illustrated in Figure 1. For a hypothetical infinite layer AgF2 we have used an analogous cell with the unit cell vectors of a = b = 4.14 Å and c = 3.50 Å (Fig. 1). Such choice guarantees reasonable values of the intrasheet and the intersheet Ag–F separations, as compared to the known polymorph of binary AgF2. The calculated pressure on the cell is only 0.7 GPa for [AgF2] (compared to 0.6 GPa for Ca[CuO2]), ensuring the validity of this choice. The band structures of a hypothetical infinite-layer [AgF2] material and that of Ca[CuO2], are shown in 1359-6462/$ - see front matter 2006 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.scriptamat.2006.07.028 812 W. Grochala / Scripta Materialia 55 (2006) 811–814 Figure 1. View of the unit cells of a hypothetical polymorph of AgF2 with flat [AgF2] sheets (left) and of the CaCuO2 infinite layer compound (right). Figure 2. The electron density integrated over selected bands of interest is presented in Figure 3. Electronic structure of a [AgF2] is quite similar to that of Ca[CuO2] (Fig. 2), but there are some differences as well. Two characteristic bands with large width can easily be distinguished in the band structure of [AgF2], one from ca. 6.3 eV to 3.8 eV (occupied), and another from 1.9 eV to +1.8 eV (approximately half occupied, due to d9 electronic configuration of Ag2+). Appreciable dispersion and integrated electron density in those bands (Fig. 3 and ESD) clearly point to their origin from the r combination of the d(x2–y2)Ag/p (x, y)F atomic orbitals. The orbitals of nonmetal and transition metal must mix in the bonding manner in the lower band, and in the antibonding way in the associated upper band, as obtained from the usual picture of the d orbitals split in a square planar ligand field [6]. The half-occupied r* band of [AgF2] is a direct analogue of such a band for infinite layer oxocuprates; it Figure 3. The electron density integrated over selected bands in the electronic structure of a hypothetical polymorph of AgF2 with flat [AgF2] sheets (left) and of the CaCuO2 infinite layer compound (right). (A) and (C) – bands crossing the Fermi level, (B) and (D) – bands immediately below those shown in (A) and (C), respectively. Isosurface of electron density was shown for density of 0.02e Å3. is believed that the presence of this band is a prerequisite for superconductivity in doped materials [7]. Significant dispersion of the r and r* bands for [AgF2] indicates appreciable mixing of the transition metal’s d and nonmetal’s p valence functions. This unusual feature makes fluorides of Ag2+ quite similar to their oxocuprate siblings; appreciable ‘hybridization’ (as understood in solid state physics) has been first suggested in Ref. [1] and then confirmed by XPS measurements [8] and – indirectly – by magnetic measurements [2]. Fluoroargentates thus join a fascinating family of ‘strongly correlated materials’, and the world of unusual phenomena, Figure 2. Electronic band structure and electronic density of states (DOS [electrons/eV]) for a hypothetical polymorph of AgF2 with flat [AgF2] sheets (top) and of the CaCuO2 infinite layer compound (bottom). Partial DOS from s, p and d electrons has also been shown, p contribution originating predominantly from nonmetal (F or O), d one from transition metal (Ag or Cu). W. Grochala / Scripta Materialia 55 (2006) 811–814 such as for example superconductivity and giant magnetoresistance. One notices two bands in the 4.7 eV to 2.9 eV energy window; these are p combinations of the in-plane d(xy)Ag orbital and of appropriate p(x, y)F functions (see ESD). The p overlap to F is poor for the majority of transition metal cations, and bonding is polarized towards nonmetal; this allows F to attain net negative charge. The in-plane p-bands have their in-plane p* counterpart, at about 1.9 eV to 0.9 eV. Analogous p- and p* bands are seen for Ca[CuO2], from 5.4 eV to 2.8 eV and from 1.7 eV to 0.6 eV, respectively. The uppermost p* band merits particular attention, as it may become of importance upon hole-doping to the d9 system; note, this band is the highest occupied band of the ‘d8 Ag3+ site’ in the ‘localized’ bonding picture. Subsequent hole transfer from Ag3+ to F, with the predominant hole occurrence on F centers, might be viable [1,9]; this might ultimately lead to the magic electronic state – in analogy to superconducting oxocuprate materials [10]. Four other bands with varying widths fill the energy gap between the r and r* bands; the concomitant crystal orbitals are composed of p or p* combinations of the out-of-plane d(xz, yz)Ag and the p(z)F orbitals (see ESD). The last interesting band propagates from 2.4 eV to 1.3 eV; it is made up mainly from the d(z2) orbitals of Ag. This band should have very small dispersion for a truly 2D case; its noticeable width observed here originates mainly from the intralayer overlap via 2s orbitals of F, and from the interlayer d(z2)/ d(z2) overlap. The respective band is seen at 2.3 eV to 0.9 eV for Ca[CuO2]. As may be seen from this comparison, the general bonding scheme is very similar for two quasi-isoelectronic systems considered, [AgF2] and Ca[CuO2]. But, most importantly of all, what are the differences? And how might they influence the propensity of these systems for charge localization/delocalization upon hole- or electron-doping? We will now attempt to answer these questions. First of all, the presence of Ca2+ between the [CuO2] layers allows for more efficient through-orbital interlayer interactions as compared to [AgF2]; this leads to more complex band structure (particularly for the pand p*-bands, which are most affected by this interaction), and to the appearance of several ‘avoided’ and ‘allowed’ interband crossing regions. Electronic DOS of Ca[CuO2] may be divided into three regions in the energy scale: (i) the first one, from 7.2 eV to 3.7 eV, where the pO contribution predominates [11]; (ii) the second one, from 3.7 eV to 1.8 eV, where the dCu contribution takes over; and (iii) the third one, from 1.8 eV to 0.0 eV, where the pO contribution is nearly equal to the dCu one [12]. Analogous regions may also be detected in the DOS for [AgF2], but their ordering is different: (i) the first region, from 6.8 eV to 4.0 eV, is characterized by nearly equal contribution from the metal and nonmetal states; (ii) the second region, from 4.0 eV to 2.5 eV, is dominated by dAg orbitals; (iii) finally, the third region, from 2.5 eV to 0.0 eV, shows the largest share of the pF states [12]. 813 We have determined the energy of the center of gravity, Eicent , for the metal (i = d) and nonmetal (i = p) states, using the following formula: Eicent ¼ R½ðDOSi ÞEi =RðDOSi Þ; ð1Þ i where E is a particular energy for a given state i, while DOSi is the associated value of partial DOS, and summing goes over the (7.5 eV to +2.5 eV) energy window. eV, The calculated values are: E4d;Ag cent ¼ 2:73 2p;F Ecent ¼ 2:65 eV, for [AgF2]. The corresponding values 2p;O for Ca[CuO2] are: E3d;Cu cent ¼ 2:52 eV, E cent ¼ 2:85 eV [13]. Thus, separation of the metal and nonmetal states is small, but different in both cases; 0.08 eV for [AgF2], +0.33 eV for Ca[CuO2] [14]. The third important difference between [AgF2] and Ca[CuO2] is related to the absolute values of the partial DOS for nonmetal states. The maximum value of DOS2p,O is about 3.2 e/[eV cell], yet that of DOS2p,F is noticeably larger, some 4.3 e/[eV cell]. This difference could be due to either (i) the larger contraction of the valence 2p orbital of F than of O, at a larger F–F than the O–O separation, or/and (ii) the weaker interaction of the valence 2p orbitals of nonmetal via d orbitals of transition metal (smaller overlap, less mixing). Most likely, both effects are operative with comparable weights. And what about ionicity of the Cu2+/O2 and Ag2+/ 1 F bonding? Table 1 shows the results of the Mulliken population analysis and the net atomic charges for [AgF2] and Ca[CuO2]. Seemingly, charge distribution is very different for [AgF2] and Ca[CuO2], as indicated by the Mulliken charges on atoms. However, transition metals show similar population of their valence d orbitals – there is slightly over 0.5 hole in the d-band in each case. Populations of valence p orbitals of nonmetals differ by 0.6e; charge transfer from lone pairs (those perpendicular to the [CuO2] layers) of oxide anions to the diffuse d (less so s) orbitals of Ca (formally unoccupied for Ca2+) may be held responsible for this effect. Correct division of the total electron density between O and Ca is difficult, but it should not affect too much our estimate of the polarity of the Cu–O bonds. If one shifts the discussed 0.6e back from Ca to O atoms, virtually identical populations of transition metal d and nonmetal p orbitals are obtained for [AgF2] and Ca[CuO2]. Thus, the Ag2+–F1 and Cu2+–O2 bonds have pretty similar polarity, an intermediate one between the ‘ionic’ and ‘covalent’ formulations [15]. Table 1. Population analysis of s, p and d electrons, and net atomic charges for [AgF2] and Ca[CuO2] System/atom s p d Total Charge [AgF2] Ag F 0.36 1.96 0.22 5.54 9.43 – 10.01 7.50 +0.99 0.50 Ca[CuO2] Cu O Ca 0.54 1.85 2.05 0.76 4.95 5.99 9.45 – 0.60 10.75 6.80 8.64 +0.25 0.80 +1.36 Important populations are in italics. 814 W. Grochala / Scripta Materialia 55 (2006) 811–814 Last but not least, different strength of magnetic interactions (not accounted for in our calculations) constitutes another important difference between [AgF2] and Ca[CuO2]. The magnetic moment of an unpaired d electron of Cu2+ is about 1.7 1.9 lB; the corresponding moment for Ag2+ is 2.1 2.2 lB. This comparison shows that magnetic interactions between the adjacent transition metal atoms are supposedly larger for [AgF2] than for Ca[CuO2], provided that they occur with equal ease via 2p orbitals of nonmetals [16]. The similar nature of the chemical bonds and of the band structure is a good indication for possible generation of superconductivity in the layered compounds which contain the [AgF2] sheets. But will these similarities suffice? It remains to be seen if genuine 2D [AgF2] materials could be obtained and charge-doped without localization of charge carriers, and if they would exhibit high-temperature superconductivity. Author thanks the ICM and the Department of Chemistry for continuing financial support of this research. All calculations were performed at ICM supercomputers. Dr. Jacek Piechota is gratefully acknowledged for making available his license of Materials Studio. Electronic supplementary data (ESD) accompanies this paper (electronic structure of related hypothetical [AgCl2] and [AuCl2] infinite layer systems, electron density in all bands for [AgF2] and Ca[CuO2] and molecular orbitals of AgF2 4 Þ. Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.scriptamat.2006.07.028. [1] W. Grochala, R. Hoffmann, Angew. Chem. Int. Ed. Eng. 40 (2001) 2743. [2] S.E. McLain, M.R. Dolgos, D.A. Tennant, J.F.C. Turner, T. Barnes, T. Proffen, B.C. Sales, R.I. Bewley, Nature Mater 5 (2006) 561. AOP 10.1038nmat1670, June 2006. [3] W. Grochala, Nature Mater 5 (2006) 513. [4] (a) T. Siegrist, S.M. Zahurak, D.W. Murphy, R.S. Roth, Nature 334 (1988) 231, Superconductivity may be generated by hole- or electron-doping into this compound; (b) M. Takano, M. Azuma, Z. Hiroi, Y. Bando, Y. Takeda, Physica C 176 (1991) 441; [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] (c) M.G. Smith, A. Manthiram, J. Zhou, J.B. Goodenough, J.T. Markert, Nature 351 (1991) 549. M.D. Segall, P.J.D. Lindan, M.J. Probert, C.J. Pickard, P.J. Hasnip, S.J. Clark, M.C. Payne, J. Phys.: Condens. Matt. 14 (2002) 2717. Another occupied band with large width shows up in the electronic structure of [AgF2] in the same energy window as the r band. This band can be approximately assigned to the combination of selected p(x, y) atomic orbitals of F atoms. J.K. Burdett, Chemical Bonding in Solids, Oxford University Press, New York, NY, 1995. W. Grochala, R.G. Egdell, P.P. Edwards, Z. Mazej, B. Z̆emva, Chem. Phys. Chem 4 (2003) 997. H.C. Muller-Rosing, A. Schulz, M. Hargittai, J. Am. Chem. Soc. 127 (2005) 8133. J.K. Burdett, Inorg. Chem. 32 (1993) 3915. Correct population analysis is often very difficult for solids, particularly when the planewave basis is used. However, here we discuss the relative differences in the DOS division for two compounds, rather than the absolute division of DOS for a given compound. In our calculations, various orbital contributions nicely sum up to total DOS (not shown). In fact, states at the Fermi level show small dominance of the valence orbitals of a transition metal. Note, CASTEP automatically sets the energy of the Fermi level to zero. Similar analysis for [AgCl2] (see ESD) yields the position of 3p states of Cl way above that for 4d states of Ag. This result indicates that Ag2+ should spontaneously introduce holes into the Cl(3p) band, thus liberating Cl2. This is in good agreement with chemical intuition and common experience; recollect: binary AgCl2 could not been prepared so far. Population of s and p functions of transition metals is very different for both compounds considered. An overall (s + p) occupation is larger for Cu than for Ag by 0.7e. Cu is forced to use its diffuse valence functions in order to accommodate much negative charge into ½CuO2 2 sheets. This difference between [AgF2] and Ca[CuO2] is not very important; occupation of (s + p) states may easily be affected by intercalation of various electron-withdrawing or electron-donating (formally neutral) species between the [AgF2] layers. Indeed, McLain et al. [2] show that magnetic superexchange constant is pretty large for ferromagnetic Cs2AgF4 (3.8 to 5.0 meV, i.e. 50 K), without precedence among related fluoride systems.
© Copyright 2026 Paperzz