ARTICLES 546nature materials | VOL 2 | AUGUST 2003 | www.nature.com/naturematerials E xploring the ultimate density limits of magnetic information storage, whether on computer hard disks or in MRAMs (magnetic random access memories), requires elaborate tuning of the preferred (easy) magnetization axis, of the magnetic anisotropy energy, and of the magnetic moment in the units used to store a bit. These units are single-domain particles (with diameter d < 20 nm) where the magnetic moments of all atoms are ferromagnetically aligned 1 to form the overall magnetic moment of the particle M, which is also called the macrospin. The preferred orientations of M, and the anisotropy energy barriers K separating them, are given by the delicate balance between several competing energies. These are the magnetocrystalline bulk anisotropy, its surface and step counterparts, and the shape anisotropy, or demagnetizing energy, resulting from the interaction of M with its own dipolar stray field. Unravelling the anisotropy's origin is far from trivial due to the competition between these energies 2 . This is unfortunate because the anisotropy is one of the key quantities: it defines the stability of the magnetization direction against thermal excitation,and therefore the minimum particle size for which non-volatile information storage may be achieved (at 300 K this requires K ≥ 1.2 eV). A further key parameter is the modulus of M, M defining the dipolar stray field used to read and write, but also mediating interactions between adjacent bits. These interactions are minimized for out-of-plane magnetization, and because the ultimate limit of single-particle bits may only be achieved for uniaxial systems, uniaxial out-of-plane systems are best suited to explore the ultimate density limit of magnetic recording 3,4 . Current studies attempting to identify the origin of magnetic anisotropy mainly deal with two model systems. These are colloids or three-dimensional (3D) nanoparticles, and 2D nanostructures created by molecular-beam epitaxy at single-crystal surfaces. For colloid particles, remarkable progress has been achieved in monodispersity 5 , their self-assembly into 2D superlattices 6,7 and in the accomplished anisotropy energies per constituent atom 8 . Despite their promising properties for applications,3D nanoparticles present several difficulties for tracing back the origin of anisotropy. First, although the magnetism of a single particle can be addressed 9 , it is almost impossible to study the morphology of the very same particle in conjunction with its magnetism. Second, the particles frequently have a few atomic layers of oxide at their surface, which is not ferromagnetic

We investigate the effect of H adsorption on the magnetic properties of individual Co atoms on Pt(111) with scanning tunneling microscopy. For pristine Co atoms, we detect no inelastic features in the tunnel spectra. Conversely, CoH and CoH 2 show a number of low-energy vibrational features in their differential conductance identified by isotope substitution. Only the fcc-adsorbed species present conductance steps of magnetic origin, with a field splitting identifying their effective spin as S eff ¼ 2 for CoH and 3=2 for CoH 2 . The exposure to H 2 and desorption through tunnel electrons allow the reversible control of the spin in halfinteger steps. Because of the presence of the surface, the hydrogen-induced spin increase is opposite to the spin sequence of CoH n molecules in the gas phase. DOI: 10.1103/PhysRevLett.114.106807 PACS numbers: 73.22.-f, 32.10.Dk, 75.30.Gw, 75.75.-c Individual surface-adsorbed magnetic atoms exhibit remarkably large orbital moments and anisotropies [1][2][3][4][5]. Like in adsorbed single ion molecules [6][7][8][9][10], their chemical environment can be tailored through exposure to reactive molecules, thus allowing the tuning of their magnetic properties [11]. Among the wealth of available molecules, H 2 is of particular interest. The high reactivity of adsorbed transition metal atoms promotes the dissociation of the H 2 molecule and the formation of metal-H n complexes (n ¼ 1; 2; 3), even at cryogenic temperatures [2,[12][13][14][15][16]. Moreover, the relatively small H desorption barrier allows the reversible control of the number of adsorbed hydrogen atoms, e.g., by desorption through electrons from a scanning tunneling microscopy (STM) tip and adsorption from the gas phase [2,15].The magnetic properties of gas phase transition-metal-H n complexes have been studied by means of ab initio calculations. These calculations reveal a significant change of the magnetic properties through hydrogenation, with a clear tendency of decreasing spin with increasing number of H atoms [17][18][19]. This results from the antiparallel spin alignment between metal and H. In particular, for the case of Co, the spin S ¼ 3=2 of the free atom is reduced to 1 and 1=2 upon the adsorption of one and two hydrogen atoms, respectively [17][18][19].The effect of hydrogen adsorption on the spin of surfaceadsorbed magnetic atoms is largely unknown. Hints that the spin possibly changes upon H adsorption can be inferred from the H-induced appearance [14] or disappearance [13,15] of the Kondo effect. However, for S > 1=2 this can also be caused by a change in magnetic anisotropy [20].Neither the spin nor the anisotropy have been measured in Refs. [13][14][15], while for Co=graphene=Ptð111Þ the adsorption of three H atoms was shown to reduce the anisotropy energy [2].Here we demonstrate that the spin of individual Co adatoms on Pt(111) can be controlled through hydrogenation. This process is reversible as the H atoms can be desorbed one by one with the STM tip. Clean cobalt atoms on Pt(111) have S ≈ 1, an out-...

No abstract

We report on the strengths and limitations of scanning tunnelling microscopy (STM) when used for characterising atomic-scale features of quasi two-dimensional materials, such as graphene and single layers of hexagonal boron nitride, which may present strong corrugations when grown epitaxially on a substrate with a lattice mismatch. As a paradigmatic test case, we choose single-layer and bilayer graphene on Ru(0001), because their STM images show both a long-range moiré modulation and complex atomic-scale distortions of the graphene lattice. Through high-resolution STM measurements, we first determine with high accuracy the moiré epitaxial relations of the single layer and the bilayer with respect to the metal substrate. In particular, we also provide direct evidence for the existence of AA-stacked bilayer graphene domains on Ru(0001). We then demonstrate that the local strain distribution, as inferred from the same STM images, can be affected by large errors, so that apparent giant strains arise in some regions of the moiré as an imaging artefact. With the aid of density functional theory simulations, we track down the origin of these fictitious distortions in the high directionality of the graphene π-orbital density combined with the large corrugation of the sample. The proposed theoretical model correctly accounts for the observed dependence of the apparent strain on the STM tip-sample separation and on the different degree of curvature of the second graphene layer with respect to the single layer.