- Full text:
- PDF (eReader) / PDF (Download) (758 kB)

The mechanism of ferromagnetism stability in sandwich clusters V _{n }(C _{6 }H _{6 }) _{n +1 } has been studied by first-principles calculation and model analysis. It is found that each of the three types of bonds between V and benzene (Bz) plays different roles. V 3d _{z 2 } orbital, extending along the molecular axis, is weakly hybridized with Bz's HOMO-1 orbital to form the σ-bond. It is quite localized and singly occupied, which contributes 1 µ _{B } to the magnetic moment but little to the magnetic coupling between neighboring V magnetic moments. The in-plane d _{x 2 - y 2 }, d _{x y } orbitals are hybridized with the LUMO of Bz and constitute the δ-bond. This hybridization is medium and crucial to the magnetic coupling though the δ states have no net contribution to the total magnetic moment. d _{x z }, d _{y z }, and HOMO of Bz form a quite strong π-bond to hold the molecular structure but they are inactive in magnetism because their energy levels are far away from the Fermi level. Based on the results of first-principles calculation, we point out that the ferromagnetism stability is closely related with the mechanism proposed by Kanamori and Terakura [J. Phys. Soc. Jpn. **70 ** (2001) 1433]. However, the presence of edge Bzs in the cluster introduces an important modification and suppresses significantly the ferromagnetism stability. A simple model is constructed to explain the essence of the physical picture.

*ab initio*calculation, tight-binding model

## References

- 1 N. J.Long:
*Metallocenes*(Blackwell Science, Oxford, U.K., 1998). Google Scholar - 2 K.Miyajima, A.Nakajima, S.Yabushita, M. B.Knickelbein, and K.Kaya:J. Am. Chem. Soc.
**126**(2004) 13202. Crossref, Google Scholar - 3 K.Miyajima, S.Yabushita, M. B.Knickelbein, and A.Nakajima:J. Am. Chem. Soc.
**129**(2007) 8473. Crossref, Google Scholar - 4 T.Yasuike and S.Yabushita:J. Phys. Chem. A
**103**(1999) 4533. Crossref, Google Scholar - 5 J.Wang, P. H.Acioli, and J.Jellinek:J. Am. Chem. Soc.
**127**(2005) 2812. Crossref, Google Scholar - 6 A. K.Kandalam, B. K.Rao, P.Jena, and R.Pandey:J. Chem. Phys.
**120**(2004) 10414. Crossref, Google Scholar - 7 Y.Mokrousov, N.Atodiresei, G.Bihlmayer, and S.Blügel: Int. J. Quantum Chem.
**106**(2006) 3208. Crossref, Google Scholar - 8 M. B.Knickelbein:J. Chem. Phys.
**121**(2004) 5281. Crossref, Google Scholar - 9 M. M.Rahman, H.Kasai, and E. S.Dy: Jpn. J. Appl. Phys.
**44**(2005) 7954. Crossref, Google Scholar - 10 V. V.Maslyuk, A.Bagrets, V.Meded, A.Arnold, F.Evers, M.Brandbyge, T.Bredow, and I.Mertig:Phys. Rev. Lett.
**97**(2006) 097201. Crossref, Google Scholar - 11 H.Xiang, J.Yang, J. G.Hou, and Q.Zhu:J. Am. Chem. Soc.
**128**(2006) 2310. Crossref, Google Scholar - 12 J.Kanamori and K.Terakura:J. Phys. Soc. Jpn.
**70**(2001) 1433. Link, Google Scholar - 13 http://www.openmx-square.org/ Google Scholar
- 14 T.Ozaki:Phys. Rev. B
**67**(2003) 155108; T.Ozaki and H.Kino:Phys. Rev. B**69**(2004) 195113; T.Ozaki and H.Kino:J. Chem. Phys.**121**(2004) 10879; T.Ozaki and H.Kino:Phys. Rev. B**72**(2005) 045121. Crossref, Google Scholar - 15 J. P.Perdew, K.Burke, and M.Ernzerhof:Phys. Rev. Lett.
**77**(1996) 3865. Crossref, Google Scholar - 16 J. M.Soler, E.Artacho, J. D.Gale, A.Garcia, J.Junquera, P.Ordejon, and D.Sanchez-Portal:J. Phys.: Condens. Matter
**14**(2002) 2745, and references therein. Crossref, Google Scholar - 17 M. J.Han, T.Ozaki, and J.Yu:Phys. Rev. B
**73**(2006) 045110. Crossref, Google Scholar - 18 A supercell size of 13×13×15 in Å3, larger than the automatically determined supercell size by openMX (with every side being expanded by about one Å), has been used for n=2. We find that the difference in total energy caused by this expansion of the supercell size is only a few tenth of meV. Google Scholar
- 19 If we adopt a standard constrained LDA calculation for estimating U, the obtained value is generally too large, more than 6. 0 eV. An alternative approach was proposed by Solovyev[Phys. Rev. B53(1996) 7158] for the U value in the t2g system of LaMO3 (M=Ti, V, Cr). In this approach, eg states are allowed to screen the Coulomb interaction among t2g electrons. The estimated U value is about 3.0 eV. As the analysis was made with the atomic sphere approximation, this value of U can be directly transferred to OpneMX. Of course, even for the same element, U depends on the system. However, for the present small cluster, screening of d electrons by other degrees of freedom will be less efficient. We therefore expect that U=3.0 eV should not be an overestimationl. Google Scholar
- 20 In order to see the possible variation in the energy difference between FM and AFM states in V2Bz3 depending on the computational details, we have performed some test calculations using OpenMX, VASP (Vienna ab-initio simulation package, http://cms.mpi.univie.ac.at/vasp/) and WIEN2k (http://www.wien2k.at/). VASP is a plane-wave pseudo-potential code while WIEN2k is a plane-wave all electron code. First, we calculated the total energy difference between FM and AFM states (ΔE =EAFM-EFM) for the same ideal structure of V2Bz3 where V–Bz distances are all equal. The results are as follows: ΔE= -0.003 eV (VASP), 0.006 eV (WIEN2k) and 0.007 eV (OpenMX). The result of WIEN2k is close to that of OpenMX both giving positive values, while VASP gives a negative value. As we are discussing a very small energy, it is not easy to identify the origin of the differences in ΔE. One possible reason for the difference, however, is that the core radius used in VASP is 2.3 a.u., being much larger than the value of 1.3 a.u. used in OpenMX. As the present system has a significant charge transfer from V to Bz, smaller core radius is required to guarantee the transferability of pseudopotential. As the all electron calculation by WIEN2k is heavy, the structural optimization was not performed by WIEN2k. With VASP, after structural optimization, ΔE becomes -0.012 eV. However, the energy difference becomes +0.007 eV in the GGA+U with U=5.0 eV. In VASP, U is included in the narrow projector space of PAW (Projector Augmented Wave) method, while U is active in the more extended localizes orbital in OpenMX. Judging from the energy levels, U=5.0 eV in VASP is nearly equivalent to U=3.0 eV in OpenMX. The supercell used in VASP and OpenMX calculations have been checked to ensure the convergency. In Wien2k, a supercell of 12×12×14 Å3 is used, which has been checked by VASP since both Wien2k and VASP fill the vacuum part with plane-waves. In Wien2k, Rmt*Kmax=4 is used to control the cutoff energy with the smallest muffin-tin shere radius (Rmt) of H being 0.93 a.u., which corresponds to a cutoff energy around 250 eV. lmax=10 and Gmax=20 a.u.-1 are used for expansion of charge (also potential) within Muffin-tin sphere and interstitial part, respectively. Google Scholar
- 21 A. D.Becke:Phys. Rev. A
**38**(1988) 3098; C.Lee, W.Yang, and R. G.Parr:Phys. Rev. B**37**(1988) 785. Crossref, Google Scholar