Importance of metal–metal bondings at the interface of MgO and 3d-transition metals

Abstract

First principles cluster calculations of MgO(001)//M(001) (M=Sc to Cu) interfaces have been systematically made using numerical atomic orbitals as basis functions. Detailed analysis of bond overlap populations (BOP) found that the M–Mg covalent bond predominantly determines the interfacial bond-strength. M atoms prefer to be located on top of O atoms, not because the first nearest-neighbor M–O covalent bond is stronger than the M–Mg bond at the same bond-length, but because the M–Mg bond prefers to be at the second nearest-neighbor distance in order to increase the number of effective interfacial bonds from 1 to 4. The equilibrium interfacial bond-length is found to be determined by the balance of M–Mg bond-reinforcement and weakening of the M–M bond. Atomic number dependence of the interfacial bond-strength is also determined by the variation of the M–Mg BOP. It decreases with rising atomic number mainly because of the contraction of M-orbitals. Theoretical electron energy loss near edge structures (ELNES) are given to help interpretation of experimental spectra to be acquired directly from the interface.

Introduction

Synergistic effects of ceramics and metals are often utilized in components for structural and electronic applications. In order to design and control their macroscopic properties, understanding of the interface between ceramics and metals is essential. The interface is of great interest also from fundamental viewpoint, since the manner of atomic bonding changes drastically at the interface. However, our fundamental knowledge for such an interface is still very limited.

High resolution electron microscopy (HREM) works for metal/ceramic interface were started in 1980s1, 2. With the development of the HREM technique, atomic arrangements at the interface have been examined in detail. Most of these works were made for MgO/metal or Al₂O₃/metal, since they were considered to be good examples of metal/ceramic interfaces. Experimental HREM works of the MgO/metal interfaces have been reported by two groups3, 4. Another potentially powerful tool for the investigation of the interface is the electron energy loss spectroscopy (EELS) equipped in a nano-probe transmission electron microscope. It is expected to provide information of local chemical bonding at the interface directly by experiments. However, EELS results from the metal/ceramic interfaces are still immature. Al₂O₃/metal interfaces were examined by a group in Stuttgart[5] and Ikuhara et al.[6] have reported EELS from the MgO/V interface recently.

First principles calculation of MgO/metal interfaces was first reported by Scho ̈nberger et al.[7] using full-potential LMTO (linearized muffin-tin orbital) method for Ti/MgO(00 l) and Ag/MgO(001) interface. Two other results using different computational methods have recently been reported. Li et al.[8] used full-potential APW (augmented plane wave) method for Ag/MgO(001) and Fe/MgO(001) interface. Heifets et al.[9] applied Hartree–Fock type super-cell method using CRYSTAL code for Ag/MgO(001) interface. Conclusions of these three papers are similar to each other. According to them, metal atoms are preferred to be placed on the top of the oxygen atoms of the MgO(001) surface when the orientation relationship is taken to be (001)_M//(001)_MgO and [110]_M//[100]_MgO. The optimum bond length across the interface was found to be ranging from 0.26 to 0.29 nm, which is 20–40% greater than the bond length of O–Mg in crystalline MgO, i.e. 0.21 nm.

Since results of three calculations did not contradict to each other, one may not find clear requirements for another theoretical calculation for a similar kind of interface. We would like to emphasize two new objectives for the present study on the M/MgO(001) interface by a first principles cluster calculation: (1) To gain our physical insight to the interfacial bonding mechanism, overlap populations between atomic orbitals across the interface is quantitatively evaluated. For this purpose, the use of localized basis-functions is advantageous. (2) Not only some specific metals but a series of transition metals (M) as a counterpart of the interface is systematically investigated. The major outcome we would stress in this paper is the finding of the significant contribution of the M–Mg bonding at the interface. It has been widely accepted that the M–O bond is more important than the M–Mg bond, simply because metal atoms prefer to sit on O atoms rather than Mg atoms. However, when metal atoms are present on the O atoms, there are four M–Mg bonds at the distance of approx. 40% longer than the nearest neighbor M–O distance. When the M–Mg bonding is strong, these four bonds are no more negligible. To our surprise, such a simple geometry has not been explicitly considered in previous theoretical calculations except for the one by the present authors for Fe/TiX (X=C,N,O)[10]. Even when the optimum interfacial bond-length or geometries were carefully computed, the origin of the stable atomic arrangement has not been pursued in detail from the viewpoint of chemical bondings.