Metal-Induced Gap States at Well Defined Alkali-Halide/Metal InterfacesManabu Kiguchi,1Ryotaro Arita,2Genki Yoshikawa,3Yoshiaki Tanida,4Masao Katayama,3Koichiro Saiki,1,3Atsushi Koma,3and Hideo Aoki21Department of Complexity Science and Engineering, Graduate School of Frontier Sciences, The University of Tokyo,Hongo, Bunkyo-ku, Tokyo 113-0033, Japan2Department of Physics, Graduate School of Science, The University of Tokyo, Hongo, Bunkyo-ku, Tokyo 113-0033, Japan3Department of Chemistry, Graduate School of Science, The University of Tokyo, Hongo, Bunkyo-ku, Tokyo 113-0033, Japan4Fujitsu Laboratories Ltd., Atsugi, Kanagawa 243-0197, Japan(Received 26 December 2002; published 15 May 2003)In order to search for states specific to insulator/metal interfaces, we have studied epitaxially growninterfaces with element-selective near edge x-ray absorption fine structure. An extra peak is observedbelow the bulk edge onset for LiCl films on Cu and Ag substrates. The nature of chemical bonds asprobed by x-ray photoemission spectroscopy and Auger electron spectroscopy remains unchanged, sowe regard this as evidence for metal-induced gap states (MIGS) formed by the proximity to a metal,rather than local bonds at the interface. The dependence on the film thickness shows that the MIGS areas thin as one monolayer. An ab initio electronic structure calculation supports the existence of theMIGS that are strongly localized at the interface.DOI: 10.1103/PhysRevLett.90.196803 PACS numbers: 73.20.–r, 71.15.Mb, 73.40.NsIntroduction.—While there is mounting interest inthe nature of the ‘‘heterointerface’’ (solid-solid interfacesbetween very dissimilar materials), insulator/metal inter-faces are especially intriguing, since they provide fasci-nating possibilities such as metal-insulator transition[1], band gap narrowing [2], and superconductivity [3],as well as technological possibilities such as catalysis,etc. Despite the interest, electronic structures charac-teristic of the insulator/metal interface have not beenstudied satisfactorily, for good reasons. First, well de-fined interfaces are hard to prepare due to the differentnature of chemical bonds. Second, signals from the inter-face are obscured by significant contribution from thesubstrate in conventional experimental methods such asultraviolet photoemission spectroscopy (UPS), inversephotoemission spectroscopy, or electron energy loss spec-troscopy (EELS).Recently, an interesting experimental result was re-ported for insulator/metal systems. Muller et al. studiedf222gMgO=Cu interfaces in fine particles (rather thanin a film) with transmission EELS, and succeeded inobserving metal-induced gap states (MIGS) for the firsttime [4]. The concept of MIGS was first introduced forsemiconductor/metal junctions in discussing the Schottkybarrier at the interface [5], and subsequently applied toinsulator/metal interfaces [6]. MIGS are thought to ac-company metal wave functions whose (exponential) tailspenetrate into the insulating side of the interface.Schintke et al. studied MgO=Ag 001 with scanning tun-neling spectroscopy and found a state in the band gap [7].However, the first point raised above remains: Sincef222gMgO is a polar (hence presumably metallic) surface,whether the prepeak observed for f222gMgO=Cu origi-nates from the polar surface or from the MIGS has notbeen definite, although their study is pioneering. Further-more, f222g MgO=Cu is not a well defined interface dueto a large lattice mismatch. For MgO=Ag 001, the stronghybridization between the O 2p band and the Ag 5spband is expected to dominate the interface [8]. All theseshould obscure the identification of the MIGS in its propersense, i.e., states formed solely by the proximity to ametal rather than by local bonds.In order to clarify this, we propose here to prepare welldefined interfaces by exploiting our studies, in which wehave revealed that some alkali halides grow heteroepi-taxially on metal substrates in a layer-by-layer fashion, sowe end up with atomically well defined insulator/metalinterfaces [9]. Being epitaxially grown, the number ofatomic layers can also be controlled, which helps to probethe nature of the interface state. In a previous study wehave examined the epitaxially grown LiCl=Cu 001, butwere unable to find an indication for the interface state[10], presumably because the methods employed (UPSand EELS) pick up signals from the substrate.Here we adopt near edge x-ray absorption fine structure(NEXAFS) to study electronic structures at the LiCl/metal interfaces, where the second point raised abovecomes in. Namely, we adopt Cl-K edge NEXAFS, whichis based on x-ray absorption by Cl atoms, and henceprovides information on the LiCl film with negligibleinfluences of the substrate. Furthermore, NEXAFS,with its high-energy photons, can probe very deep inter-faces, so suited for obtaining the dependence of the inter-face states on the thickness of the insulating layer.The results obtained here for LiCl thin films on Cu(001)and Ag(001) indeed exhibit clear evidence for the elec-tronic states intrinsic to the insulator/metal interfacewhich are as thin as one monolayer (ML) [1ML2:6A]. We have then compared the experimental resultwith an ab initio density functional calculation, whichPHYSICAL REVIEW LETTERSweek ending16 MAY 20 03VOLUME 90, NUMBER 19196803-1 0031-9007=03=90(19)=196803(4)$20.00 2003 The American Physical Society 196803-1supports the existence of the MIGS that are stronglylocalized at the interface.Experimental.—Epitaxial LiCl films were grown onmetals at 300 K by a Knudsen cell. Cl-K edge NEXAFSwas carried out at BL-11B in the Photon Factory inInstitute of Materials Structure Science [11]. The Cl-KNEXAFS provides information on the unoccupied Cl-pstates[12].Figure 1 shows the Cl-K edge NEXAFS spectra forLiCl=Cu 001 and LiCl=Ag 001 taken at grazing x-rayincidence (15) for various thicknesses of the LiCl layer.All the spectra are normalized by their edge jumps. TwoNEXAFS peaks are observed at 2827 eV (p1)and 2829 eV(p2) for bulk LiCl. Now, a new finding here is that apronounced prepeak (p3) appears just below the bulkedge onset, which is clearly visible for thinner LiCllayers. Hereafter we will focus on the prepeak.There are two points to note. First, the prepeak existseven for a 10 ML LiCl on Cu(001). Namely, although thepeak may seem more prominent with decreasing the filmthickness in Fig. 1, this is an artifact of normalizing
View Full Document