Magnetic Nanoclusters Put in Order –
Insights with Scanning Tunneling Microscopy

Axel Enders

The impact of nanomagnetism on information technology has been critical in recent years and will remain so in the coming years. Magnetic thin films have been used for data storage in computer hard-disks for decades, and continuous advancements of their properties have led to an amazing increase of the areal storage capacity by many orders of magnitude. The future of magnetic data storage might depend on the development of devices based on patterned layers with ultrahigh densities of discrete magnetic elements and without any moving parts. The ideal magnetic storage device would consist of smallest monodisperse magnetic clusters, densely packed into an ordered monolayer, with stable remanent magnetization at room temperature, accessible switching fields, and negligible interactions. The key to success will be to fabricate hard-magnetic particles of approximately 3 nm diameter, as they still exhibit a stable magnetization direction at room temperature. The magnetization of smaller particles is thermally unstable, whereas big particles lead to a waste of areal density. However, the positioning of nanostructures at surfaces and engineering their properties is a key challenge in nanotechnology.
Our recent advancements in selfassembled growth permit us to fabricate model systems, which are close to the ideal described above. Nanometer clusters are fabricated by buffer layer assisted growth, see Figure 1. The cluster material is deposited on the substrate that is pre-covered with a Xenon layer at 35-40 K. Warming up the substrate causes desorption of the Xe layer. The metal clusters become highly mobile and coalesce, thus growing in size until making contact with the surface. The final cluster size depends on the Xe thickness and on the metal coverage. On flat metal substrates, the clusters are randomly distributed, as can be seen in the scanning tunneling microscopy image in Figure 1.


fig. 1

Figure 1. Fabrication of nanometer size clusters with a noble gas buffer layer. (1) Adsorption of Xe on the metal substrate at T = 30 K. (2) Deposition of metal onto the Xe. (3) Thermal desorption of Xe; clusters make contact with the surface. (4) STM image of Fe clusters on Pt(111) (Image size 100 x 100 nm2).

The properties of such clusters, most importantly here their magnetism, are dependent on electronic hybridization between clusters and substrate and can be tailored by controlling such interactions via the substrate material, surface orientation or with interlayers.
Lateral ordering of the nanostructures can be achieved by directed growth on patterned nanotemplate surfaces. Such surfaces provide well-defined energetic sinks with sub-nanometer accuracy, to guide the nucleation processes and diffusion of nanostructures or clusters. Mechanically stable boron nitride nanomesh layers are used here as templates for the fabrication of ordered cluster layers. This approach makes the fabrication inevitably a two-step process. In a first step, the boron nitride template layer is formed, followed by the deposition of clusters during a second step. The boron nitride nanomesh is formed by thermal decomposition of borazine gas, (HBNH)3, on Rh(111) surfaces in an ultrahigh vacuum chamber. The resulting h-BN layers are atomically thin, electrically insulating, mechanically extremely stable, and show a strain-driven hexagonally ordered corrugation with a periodicity of 3.2 nm (Figure 2a). Deposited clusters preferentially occupy the pits of the BN layer, and approximately 30% of the pits are occupied by clusters after one deposition cycle (Figure 2b). Higher nanomesh filling can be achieved by repeated deposition cycles; after three subsequent cycles already 80% of the pits are occupied (Figure 2c). Comparative studies with STM and Monte Carlo simulations show that the filling is somewhat limited by some mobility of the clusters on the BN layer, which results in assimilation of a small fraction (approx. 15%) of newly deposited clusters by larger clusters that have been deposited during an earlier cycle.


fig. 2

Figure 2. Fabrication of ordered cluster layers with template surfaces, imaged with scanning tunneling microscopy. (a) Boron nitride nanomesh monolayer, imaged with atomic resolution (inset). The distance between the centers of neighboring pockets in the BN layer is 3.2 nanometer. (b) STM image of Cobalt clusters (yellow) on the BN layer (purple) after one deposition cycle. (c) Repeated deposition cycles result in higher cluster coverage; here: 3 deposition cycles. (d) Schematic drawing of an ideal nanocluster layer with potential for application in magnetic data storage.


Besides being a nanotemplate, the BN layer has two additional traits: it decouples deposited nanostructures electronically from the metal substrate, and it is suitable for robust annealing at high temperatures. Tunneling Spectroscopy studies reveal a Coulomb gap of >160 meV in the electronic structure of the Co clusters on the BN layer, proving their electronic decoupling from the rhodium substrate. BN layers thus turn out to be an ideal playground for the study of the electronic interaction of nanostructures with the substrate. The thermal stability could be exploited to fabricate high-anisotropy L10 alloy clusters by post-annealing of initially disordered FePt clusters, as the BN layer is expected to suppress cluster growth via Ostwald ripening.
Evidently, the BN nanomesh has big potential to achieve templating on the nanometer scale and electronic decoupling, and could be of considerable practical importance for FePt based recording media. The high periodicity of the cluster layers on the BN makes them suitable for bit-patterned media where one cluster represents one magnetic bit. Some aperiodicity, as visible in the STM images, can actually be accounted for by multiple-pass reading and writing. Future challenges are in the further improvement of the cluster size distribution and ordering, and in the controlled modification of their magnetic properties.