* Present adress: Department of Chemistry, York University, Toronto, Ontario, Canada

The atomic scale mechanisms of epitaxial nucleation and growth and their influence on the resulting surface morphology have attracted considerable interest. Island shape, island density, and growth morphology depend on the complex interplay of kinetic properties such as adatom flux, adatom mobility, and effective capture rates of adatom clusters, islands, or other surface heterogenities. Consequently, they can be strongly affected by deposition parameters, such as temperature, deposition rate, or, for electrochemical deposition, the overpotential h = E_dep – E_Me/Me^z+ (where E_Me/Me^z+ is the equilibrium potential for metal deposition/dissolution) [1-5]. For instance, it has been shown theoretically [5] and experimentally [6,7] that depending on the deposition parameters step flow growth, layer-by-layer growth via homogeneous nucleation and growth of monolayer islands, or (kinetically) controlled growth of multilayer islands may result. Furthermore, there are strong efforts to exploit these kinetically limited processes for the controlled formation of low-dimensional structures in the nanometer range [4].

Here we demonstrate a novel effect, the potential-induced crossover from two-dimensional step-flow to nucleation and growth of high aspect ratio multilayer islands at substrate steps, which was observed for Ni electrodeposition on Ag(111) and which, in addition to being interesting from a fundamental point of view, provides another way for controlled structuring of metal (electrode) surfaces. Step decoration by 3D islands commences at significantly lower overpotentials than 3D island formation on the Ag terraces and is apparently independent of the terrace width, allowing the preparation of uniform 2D or 3D admetal structures exclusively at step sites. This growth behavior can be explained by a simple model of kinetically hindered 3D growth, where the rate of (heterogeneous) next-layer nucleation is initially low but increases significantly faster with overpotential than the lateral growth rate.

The homebuilt electrochemical STM used in the experiments and the experimental procedures are described in detail in Ref. [8,9]. Ni was deposited from modified Watts electrolyte (10^-2 M H₃BO₃, 10^-4 M HCl, and 10^-3 M NiSO₄), prepared from suprapure H₃BO₃, suprapure HCl, p.a. grade NiSO₄ and Milli-Q water. The Ag(111) crystal was chemically polished using a CrO₃-HCl mixture. Potentials are given versus Ag/AgCl (KCl sat.). STM images were obtained in constant current mode with the tip potential usually kept 50-100 mV below the sample potential. Ni coverages were evaluated directly from the STM images.

In-situ STM images of freshly polished Ag(111) crystals reveal atomically smooth, up to 2000 Å wide terraces separated by monoatomically high steps at potentials between –0.67 and –0.1 V. According to electrochemical and in-situ STM measurements noticeable Ni deposition commences at potentials < -0.67 V in form of a (111)-oriented Ni deposit [9], which grows at approximately constant rate. As shown in Fig. 1 the morphology of this deposit depends strongly on the deposition potential. At potentials slightly more negative than the onset of Ni deposition monolayer Ni islands nucleate at ascending Ag steps (Fig 1a inset), followed by step flow growth onto the lower Ag terrace (Fig. 1a). The Ni deposit can be easily distinguished from the Ag substrate by the presence of a Moiré pattern, resulting from the lattice mismatch of the substrate and admetal (see Fig. 1a, inset) [9]. This quasi two-dimensional growth behavior was observed from submonolayer up to multilayer Ni films of at least 3 to 4 layers in thickness [9]. An example of the smooth multilayer deposit formed in this potential regime is shown in Fig. 1b.

At only 40 mV more negative deposition potential a dramatically different deposit morphology is observed, where multilayer Ni islands are formed. These islands are located exclusively along the Ag steps (Fig. 1c and e). Hence, a crossover from quasi two-dimensional step-flow to three-dimensional step decoration occurs between these two deposition potentials. The multilayer islands seen in Fig. 1e have an average height á hñ of 8.0 Å (i.e., 4 Ni layers) and an average diameter á wñ of » 150 Å (see also Fig. 1g), independent of the size of the adjacent terraces. The latter can be seen particularly well on larger scale images (Fig. 1c) and can be rationalized only if material transport from the terraces to the steps does not contribute significantly to the growth of these islands. This is possible if the Ni²⁺ ions arriving at the steps are discharged and incorporated directly at the step sites in a one step process while those ions arriving on the terraces cannot be discharged. This effect, which is known for electrodeposition as "direct deposition" [10], can be rationalized, e.g., by different activation energies for the ion transfer reaction at the two different locations. A similar mechanism is not possible for MBE growth under ultrahigh vacuum (UHV) conditions. Accordingly, UHV-STM images for Ni deposition on Ag(111) at 300 K, where similar multilayer islands are observed at Ag steps and on terraces, show a clear influence of the terrace width on the island size [11].

Significant nucleation of Ni islands on the Ag terraces is observed only at even more negative deposition potentials, i.e., at higher overpotentials h (Fig. 1d). Even under these conditions, however, nucleation and growth of the three-dimensional islands occurs preferably at the Ag steps. The average island height á hñ and diameter á wñ are 15 Å and 100 Å, respectively, i.e., the aspect ratio á wñ /á hñ (at identical coverage) decreases with increasing h . Close inspection of the line scans (Fig. 1g and h) reveals mesalike island shapes with atomically flat top layers, very different from island shapes expected for simple kinetically controlled induced growth of multilayer islands. Possible mechanisms for this will be discussed in more detail below. The formation of Ni islands on terraces may result from homogeneous or heterogeneous nucleation, where the latter involves adlayer island nucleation on top of substitutional Ni atoms within the Ag surface layer. This process has recently been proposed for Ni growth on Ag(111) under UHV conditions [12] and can also occur in the electrochemical environment as indicated by in-situ STM [9].

The nucleation and growth process leading to step decoration by Ni multilayer islands can be visualized in time-resolved in-situ STM experiments (Fig. 2). After decreasing the potential to –0.73 V nucleation and lateral growth of small Ni monolayer islands commences at the lower edge of the Ag steps (Fig. 2a). Simultaneously, second layer Ni islands start to nucleate at the boundary between these Ni monolayer islands and the upper Ag terrace (see arrows). Also nucleation of the following 4 Ni layers (Fig. 2b-d), on top of existing Ni islands, occurs almost exclusively at the position of the one-dimensional Ag-Ni boundary (see arrows in Fig. 2d). The nucleation rate for the next Ni layer initially is high, but continuously decreases with increasing island height. This is concluded from the increasing size the top layer can reach before next-layer nucleation occurs (» 50 Å for 2^nd layer, » 200 Å for 5^th layer). A more quantitative evaluation of the average height á hñ and the aspect ratio á wñ /á hñ of the Ni islands as a function of the deposition time is shown in Fig. 3.

The above observations can be rationalized in a simple kinetic model where the (direct) deposition at different surface sites is assumed to depend differently on the deposition potential (illustrated schematically in the inset in Fig. 3). The rates for (direct) Ni deposition at the steps of the Ag substrate or existing Ni islands, k_Ni/Ag and k_Ni/Ni, most likely exhibit Tafel behavior (i.e., k_Ni/X µ  exp(a_Ni/X h ); X = Ag, Ni). Furthermore, the nucleation of second and higher layer admetal islands, required for multilayer growth, may exhibit a very different potential dependence. The observed crossover from quasi 2D step-flow to 3D island growth can be explained by assuming that the nucleation rate on top of the Ni-Ag boundary k_nucl is negligibly small as compared to k_Ni/Ag at the onset of Ni deposition, but increases much more rapidly with overpotential, and finally becomes comparable or even larger than the latter at sufficiently high h . In this case second and higher layer nucleation would be kinetically hindered for deposition at low h , resulting in a flat deposit, whereas at higher h a multilayer deposit would form. This model also explains the increase in the aspect ratio as well as in the Ni islands density along the Ag steps with overpotential.

The nucleation of second Ni layer islands proceeds, at least at not too high h , exclusively at the structural defect given by the admetal-substrate boundary. Since the lattice distortion induced by the underlying admetal-substrate boundary also continues into second and higher admetal layers, similar heterogeneous nucleation sites also exist for next layer nucleation, which allows true 3D growth. The small difference in apparent height associated with this defect as well as the preferred nucleation of the higher Ni layers at this location can be directly observed in the STM images. It is noteworthy, that a similar step decoration by multilayer islands was also observed for Cu electrodeposition on Au single crystal surfaces [13], indicating that this kind of nucleation behavior is not uncommon. With increasing island thickness the distortion in the admetal lattice on top of the admetal-substrate boundary is gradually removed via lattice relaxation, resulting in a decreasing nucleation probability. Consequently, with increasing island height a continuous crossover to a more lateral growth of the islands can be expected, as is indeed observed. In a simple two-dimensional model we assume the (layer-dependent) probability for heterogeneous nucleation to be proportional to the vertical growth rate dá hñ /dt. Assuming in addition that i) the binding energy and the energy for surface diffusion (i.e., the relevant energy term for a critical nucleus of size i = 0 and i = 1, respectively) depends linearly on lattice strain [14], and ii) the strain is decaying inversely proportional with the distance to the topological defect [15] (i.e., to the number of layers h), á hñ is given by dá hñ /dt  µ  exp (z × á hñ ^-1) [16]. This provides a good fit to the data for á hñ ³ 3 for an additional strain-induced energy for nucleation on top of an h layer thick Ni deposit of z  ×  kT ×  h^-1 = 130 meV ×  h^-1 (solid line in Fig.3).

The mesalike shape of the islands at high h can be rationalized by faster Ni growth on the Ni islands than on the Ag substrate, i.e., k_Ni/Ni >> k_Ni/Ag. Under these conditions the second and higher Ni layers would spread rapidly over the island surface, resulting in a limitation of the lateral size of the multilayer islands by the first Ni layer. On the other hand, at low h the few second layer islands grow rather slowly, i.e., k_Ni/Ni is equal or even smaller than k_Ni/Ag. Hence, k_Ni/Ni apparently increases more strongly with h than k_Ni/Ag (a_Ni/Ni > a_Ni/Ag). Following these ideas the relatively uniform width of the multilayer islands does not result from an intrinsic limitation such as strain effects but solely from the growth kinetics.

The strongly potential-dependent nucleation and growth behavior can be used to prepare multilayer admetal structures with well defined nanometer scale morphology, which may have potentially interesting magnetic properties. These structures may have three possible advantages in comparison to previous magnetic nanostructures [17]: First, the size of the resulting Ni multilayer islands is rather uniform on the m m scale, since the dimensions of the deposited islands do not depend on the size of the adjacent terraces, i.e., the substrate morphology (see Fig. 1c-d). Second, highly selective formation of Ni "nanowires" can be easily achieved at deposition potentials between –0.73 and –0.8 V, where the Ni islands are exclusively formed along the steps and even on terraces of ³  2000 Å in width no Ni islands are observed. Third, not only monolayer but also thicker deposits with (self-limiting) thicknesses in the range of 1 nm, i.e., comparable to the length scale of magnetic interactions, may be formed. Magnetooptical studies of these deposits are in preparation.

In summary, we have found a novel effect, the potential controlled transition from 2D step-flow to selective nucleation and growth of 3D islands at steps to 3D growth at steps and on terraces for Ni electrodeposition on Ag(111). We could rationalize this by a simple mechanistic model where the (layer dependent) nucleation rate increases more strongly with overpotential than the rate of (direct) deposition at descending steps. This effect affords new possibilities for controlled nanostructuring of metal electrodes.

We gratefully acknowledge a fellowship by the Alexander-von-Humbold foundation (S.M.) and financial support as well as a fellowship (O.M.M.) by the Deutsche Forschungsgemeinschaft.