SURFACE MAGNETISM OF THE CLEAN Ni (111) SURFACE AND OF A Ni MONOLAYER ON Cu (111)

Results of highly precise all-electron local spin density FLAPW calculations on (i) a close-packed 7-layer Ni (111) slab and (ii) a p (1 x 1) Ni monolayer on Cu (111) are presented. For Ni ( I l l ) , we find an enhancement of the magnetic moment for both the surface (to 0.63 pB) and sub-surface Ni layers (from a center layer value of 0.58 pB). In contrast, a slight decrease of the magnetic hyperfine field at the surface is obtained. Although the magnetism is suppressed for monolayer-Ni/Cu (111) due to sp (Cu) -d (Ni)interactions, we find that the Ni overlayer is not magnetically ''dead" but has a sizable but reduced moment of 0.34 pB/atom. One of the remarkable features of transition metal Table I. Theoretical layer-projected magnetic moments surfaces (or transition metals as overla~ers) is the re(in PB) and magnetic h ~ ~ e ~ $ n e fields ( in kGauss) broken tention and, in most cases, the enhancement of magdown into core and conduction electron (CE) contributions for the 7-layer Ni (111). netic moment at the surface compared to bulk due to the reduced symmetry and lower coordination number of the surface atoms [I] . This observation holds for the open (100) and (110) surfaces of the magnetic 3d metals. However, for the experimentally most accessible close-packed surfaces (i.e., (111) orientation), there has been far less theoretical effort devoted to be study of their magnetic structure. This is due to the loss of cubic symmetry near the surface and hence the need for a more general full-potentia1 approach to treat surface related phenomena. Further, for the close-packed surface, there are less well-defined localized surface states compared with those of the more open surfaces. The immediate question thus arises as to whether 2-dimensional magnetism exists for a 3d transition-metal monolayer in contact with a nonmagnetic metal substrate even for the close-packed (111) orientation. One prototypical example is the Ni monolayer grown epitaxially on the Cu (111) substrate due to the close match of their lattice constant. There are several factors which determine the size of the surface magnetic moment, i.e., the localization of the surface d-band, the position of the Fermi level (or charge transfer), and the hybridization of the magnetic Ni-d and nonmagnetic Cu-s,p electrons, which acts to quench the magnetism. In this paper, we present results of self-consistent all-electron local density functional calculations of the surface magnetism of Ni (111) and of a monolayer Ni/Cu (111) using the full-potential linearized augmented plane-wave (FLAPW) method [2] . Both systems are modelled by a single-slab geometry consisting of 7 atomic layers. Relaxation of the interlayer spacing near the surface region is not included. Table I summarizes our theoretical magnetic moments from the surface to center layer of the Ni (111) 7layer slab (the slab possesses inversion symmetry with respect to the center layer). As expected, the surface magnetic moment is enhanced but only by N 10% from the bulk moment. However, due to its close-packed nature, the (111) surface magnetic moment (0.63 p ~ ) is less than that of the (100) surface (0.68 p ~ ) [3] .An unexpected result obtained from our calculation is that the sub-surface Ni (111) layer has essentially the same magnetic moment as that of the surface layer (c.f. Tab. I). This may be a direct consequence of the delocalized nature of the surface states for the (111) surface which interact strongly with the bulk states (i.e., resonant surface states). (Analysis of the electron states shows that the localized surface state near EF exists only in vicinity of in the 2D Billouin zone). Nevertheless, the layer-projected magnetic moment approaches its bulk value (0.56 p ~ ) just two layers beneath the surface-vacuum boundary. Unlike the magnetic moment enhancement at the surface, the magnetic hyperfine field (contact term) remains almost constant for the surface layer compared to its "bulk-like" center layer value (c.f. Tab. I). A decomposition of the total hyperfine field into core and conduction electron (CE) contributions shows that : (1) the core contribution scales precisely with the layer-projected magnetic moment, namely -158 kG/pB ; and (2) the contribution from the conArticle published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphyscol:19888742 C8 1626 JOURNAL DE PHYSIQUE duction electrons decreases in the surface layer, indicating a (somewhat) less covalent type indirect polarization (which is common to ferromagnetic bulk metals) for surface atoms. The increase in the hyperfine field for the sub-surface atoms is mostly due to the enhancement of the moment (and thus the magnitude of core polarization). We now consider the possibility of 2D magnetism for a Ni monolayer on Cu (111). For the (100) surface, both theoretical calculation [4] and experiment [5] have confirmed a magnetic Ni overlayer on Cu (100) with a moment 0.4 pg. However, for the (111) orientation, the results remain controversial due to its close-packed nature. In a early pionneering study, Tersoff and Falicov [6] found that a "magnetic dead" layer for Ni/Cu (111) from their parameterized (from bulk) tight-binding calculation. Our self-consistent all-electron calculation shows that although the magnetism is suppressed for Ni/Cu(lll), the Ni overlayer is not magnetically dead. A sizable but reduced moment of 0.34 pg is found for the Ni overlayer on Cu (111). Since there is no appreciable averlapping between the d-bands on Ni and Cu, the suppression of magnetic moment (compared with that of the surface Ni (111) or its corresponding bulk value) is mostly due to the hybridization between the non-magnetic Cu (s, p) electrons and the Ni (d) electrons, as suggested by Tersoff and Falicov [7] . As in the case of clean Ni surfaces, the reduction of the magnetic moment for Ni/Cu from (100) to (111) orientation is 0.05 pg. The magnetic hyperfine field for the Ni overlayer is supressed to -80 kG due to the decrease of magnetic moment (with the same scale constant 160 kG/pB for core polarization). The -26 kG CE contribution (mainly from states which have 4s components near the nucleus) remains essentially the same as that of the surface layer of Ni (111) with a covalent-type polarization character. Again, this indicates some hybridization between the Cu and Ni although the d-band separation favors localization (and the direct type polarization for valence electrons). Finally, some possible ways for experimentally confirming our theoretical prediction on the magnetism of a Ni morrolayer onCu (111) include (1) electron capture spectroscopy to detect the spatial spin density distribution near the surface region, i.e. a negative spin polarization in the vacuum region but the sign reversed to positive spin character at the surface layer-vacuum boundery ; (2) angular resolved photoemission to measure the exchange splitting (e 0.35 eV) of surface states near EF at the symmetry point M, and (3) Mossbauer or NMR spectroscopy to measure the hyperfine field for which we predict a decrease from its bulk value and to scale roughly with the local moment.