Quantum size effects of small, size-selected and deposited CoRh clusters on Ni

The electronic, chemical and magnetic properties of in situ prepared, mass-selected and deposited CoxRhy alloy clusters with up to five atoms per cluster have been studied by means of x-ray absorption and x-ray magnetic circular dichroism spectroscopy. The clusters were supported on a remanently magnetized Ni film. The cluster properties exhibit strong size and composition dependent effects. By alloying of pure Co1,2 adatoms and dimers with Rh the chemical reactivity towards oxygen is significantly enhanced for specific compositions. The oxidation of the alloyed clusters strongly affects the magnetic properties which results in a reduction of the orbital and spin magnetic moments. The exchange coupling between the oxidized Co atoms and the Ni surface exhibits a size and a composition dependent oscillation between antiferromagnetic and ferromagnetic coupling.


Introduction
The size dependent investigation of nanoparticle and cluster properties is a very active research area. The closing gap between industrial application and fundamental research certainly is one reason of the strong interest. In divisions like electronics, magnetism, optics, mechanics and medical health care, devices and particles are scaled down and redesigned to the scale of several tens to sub-nanometers. In this region finite size effects become crucial and along with this size dependence unexpected material properties arise which can strongly differ compared to the bulk. For magnetic materials, it has been shown that reducing the dimensionality leads to enhanced magnetic moments compared to the bulk [1,2]. In the bulk, 4d transition metals like rhodium exhibit no magnetism, although being close to fulfill the Stoner criterion for ferromagnetism. Cox and coworkers showed that the reduction in size, down to a Rh cluster size of less then 60 atoms, led to superparamagnetism and enhanced magnetic moments of the clusters [3,4].
In modern high density magnetic recording applications alloys of ferromagnetic 3d transition metals with 4d or 5d transition metals are used. The aim is to overcome the superparamagnetic limit by increasing the orbital magnetic moment and the magnetic anisotropy energy with the help of the strong spin-orbit interaction of the 4d/5d elements. Since Rh clusters do exhibit a considerable magnetic moment, it is reasonable to consider 3d-Rh alloys as promising magnetic storage materials.
Alloying of Co with Rh has been studied experimentally and theoretically in a variety of systems, e.g. alloys [5,6], multilayers [7,8], nanoparticles [9][10][11][12][13][14] and clusters [15][16][17][18][19]. For these systems an induced magnetic moment on the Rh site was reported, strongly depending on the magnetic environment. Calculations on free neutral [15][16][17][18] as well as experiments on free ionic [19] CoRh clusters showed that the orbital and spin magnetic moments can be either enhanced or reduced if compared to the pure Co clusters. These non-monotonous trends are linked to the complex interplay between the cluster size, the geometry and its composition. However, so far no systematic studies on deposited, small and mass-selected CoRh clusters have been performed. The deposition is crucial from a technical point of view but typically changes the cluster properties with respect to the gas phase unless the cluster-substrate interaction is actively suppressed [20].
In this paper, experiments on mass-selected Co x Rh y alloy clusters deposited on a magnetized Ni/Cu(100) surface are presented. The clusters are systematically varied in size atom by atom up to the five atomic Co 2 Rh 3 cluster and investigated by x-ray absorption (XAS) spectroscopy techniques. The focus of the experiment is to study the evolution of the electronic, chemical and especially magnetic properties of the CoRh 3d/4d alloy clusters in a size and composition dependent manner on the Ni surface.

Experiment
The Co x Rh y cluster samples have been investigated at the BESSY II synchrotron light source using the soft x-ray UE52-SGM beamline [21]. The XAS and x-ray magnetic circular dichroism (XMCD) spectroscopy techniques are perfectly suited to study the electronic, chemical and magnetic sample characteristics in an element-selective way. With XAS, the unoccupied electronic states and the chemical environment of the cluster Co atoms were probed by exciting 2p 3d  transitions at the Co L 2,3 -edges. Using the circular polarized light of the beamline, the magnetic properties at the Co and Rh sites (Rh M 2,3 -edges, 3p 4d  ) were examined by XMCD.
All cluster samples were produced in situ using a UHV-cluster source [22] with mass-selecting dipole magnet. The cluster source was operated with 30 keV Xe + ions to sputter a Co 0.5 Rh 0.5 target. All positively charged sputter fragments are accelerated to an electrostatic lens by setting the target potential to +500 V. The lens guides the clusters to the dipole magnet offering a resolution of m m 50 D » . Behind the dipole magnet, the mass-selected clusters, shown in figure 1(a), are first decelerated to an energy of about 1 eV atom -1 and in the following deposited via a soft-landing scheme [23,24] on apseudomorphic 20-30 monolayer (ML) Ni/Cu (100) surface ( figure 1(b)). In this size regime, the Ni film shows a remanent out-of-plane magnetization if aligned by an external magnetic field [25,26]. The clusters are aligned parallel to the magnetic orientation of the film by exchange coupling. The Kr buffer gas (≈8-9 ML) adsorbed for the soft landing process was removed by annealing the sample from 35 K base temperature to about 100 K. With the used scheme, fragmentation of the deposited clusters is effectively suppressed [27]. In order to avoid agglomeration and inter cluster interaction, the cluster coverage was chosen to 0.03 ML. For all studied samples, the here described surface preparation process was carried out at a base pressure of about 3 10 10 mbar.
The XAS and XMCD measurements were carried out in the total electron yield (TEY) detection mode at a base pressure of 2 10 10 mbar. The resolving power E E D of the beamline at the Co L 2,3 -and the Rh M 2,3 -edges was around 2300 and 2500, respectively. Due to the lower cross section at the Rh M 2,3 -edges, the exit slit was opened from 100 to 150 μm with respect to the Co measurements. Moreover, due to time constraints, the step width was increased from 0.2 to 0.5 eV. Sufficient statistics were obtained with counting times of 4 s per data point for left and right circular photon polarization. In order to check the Ni film thickness and to normalize the magnetic cluster signal to the Ni magnetization, the Ni and Cu XAS signal at the L 2,3 -edges were recorded too.

Results and discussion
3.1. Chemical properties Figure 2 shows XAS spectra at the Co L 2,3 -edges of two Co 1 adatom as well as Co 1 Rh y and Co 2 Rh y cluster preparations. All spectra were normalized to the incoming photon flux through division by the TEY signal of the last refocussing mirror of the beamline. In a next step, the normalized spectra were subtracted by a fifth order polynomical fit representing a combination of the smooth substrate background and the continuum-steps at the L 2,3 -edges. Different signal to noise ratios in figure 2 correspond mainly to the change of the Co content of the individual cluster species. At the bottom, the measured and well known Co 1 adatom spectrum is presented [28,29]. The main peak of the Co L 3 -edge at 778.0 eV (left dashed line) exhibits no fine structure in contrast to the alloy clusters. There, a side peak at a constant energy of 779.5 eV is observed (right dashed line), independent of cluster size and composition. In contrast, a clear chemical shift of the main peak to higher excitation energies is present. The fine structure of the alloy clusters can be assigned to an increased chemical reactivity towards oxygen rather than resulting from alloying Co with Rh. This can be inferred from the shown Co 1 -O spectrum [28], representing the oxidized Co 1 adatom and thus, clearly correlating the side peak to the presence of oxygen. Even a broadening on the low photon energy side, in CoO bulk, layer and nanoparticles [30-32] seen by a small peak at 776.6 eV, is visible.Similar oxidation effects have been reported already for size selected CoPd and CoPt alloy clusters on Ni [28,29]. The fully oxidized Co 1 -O sample was prepared in [28] by introducing a waiting time of 60 min. between the magnetization of the Ni film and the deposition of the Kr buffer layer. This is in contrast to the here shown CoRh samples, exhibiting no waiting time. The missing CoRh oxides in the collected mass spectra ( figure 1(a)) show that the oxidation is surface mediated. Moreover, the process must be related to the soft-landing scheme since no changes of the XAS signal were present during the measurements. However, the mechanism of the oxidation process is still not fully understood and made it impossible to study the pure CoRh cluster properties. Nevertheless, the similar preparation of the clusters allows a size and composition dependent comparison among themselves.
The varying height of the side peak can be used to estimate the fraction of oxidized clusters in the preparations. Thus, the peak height of Co 1 -O has been normalized to 1 and to 0 for the non-oxidized Co 1 Figure 2. The XAS spectra of Co x Rh y clusters on Ni/Cu(100) are shown at the Co L 2,3 -edges. The L 3 -edge main peak shows two distinct features indicated by the dashed lines. While the left line does not fit for all species, the right perfectly matches. The Co 1 -O spectrum was already published in [28].
adatom, respectively. Figure 3(a) shows the resulting oxidation ratio for each preparation. Since all CoRh preparations have been performed under similar conditionschanges in the oxidation ratio directly reflect the size and composition dependent reactivity of the CoRh clusters. The Co 1 Rh y clusters exhibit a monotonous increase of the reactivity with increasing Rh content finally resulting in a nearly fully oxidized Co 1 Rh 2 cluster. For Co 2 Rh y the reactivity is generally decreased if compared to Co 1 Rh y , indicating that the second Co atom might not be oxidized. Moreover, size and composition dependent effects are observed. Adding one Rh atom to Co 2 Rh 1 reduces the reactivity by about 60%. In addition, with two additional Rh atoms it is enhanced by 30%.
The important role of the cluster geometry is demonstrated by the Co 1 Rh 1 and Co 2 Rh 2 cluster samples. Both species have the same alloying ratio but the reactivity of Co 2 Rh 2 is reduced by a factor 4. Unfortunately, the low cluster coverage does not allow to gain information about the structure, e.g. by the EXAFS absorption signal.

Electronic properties
Although being dominated by the characteristic CoO fine structure, the XAS spectra of figure 2 contain information about the hybridization of the Co electronic states by alloying with Rh. Figure 3(b) shows the oxidation ratio of the Co x Rh y clusters against the relative L 3 -peak shift with respect to the Co 1 adatom. The dashed line implies a linear energy shift of the main peak if the fraction of the oxidized Co 1 Rh y clusters in the ensemble increases up to the full oxidized Co 1 -O. All Co 1 Rh y samples clearly show a shift to higher photon excitation energies. On the one hand, this can be related to the increasing cluster size as has been shown already for free as well as deposited size-selected 3d transition metal clusters [33,34]. There, initial and final state effects contribute to the enhanced excitation energy by an increased delocalization of the valence electrons. On the other hand, the shift can be addressed to an increased electron density at the Rh site as has been reported also for theoretical calculations on small CoRh clusters in the gas phase [16][17][18]. Thus, the reduced electron density at the Co site increases the 2p core level binding energy due to the weakened Coulomb repulsion of the electrons.
A similar shift to higher excitation energies under alloying with 4d/5d transition metals was reported also for small CoPd and CoPt alloy clusters on Ni/Cu(100) [28,29]. Again, theoretical calculations on free alloy clusters confirm this trend for both systems and find an increased electron density on the 4d/5d site [35,36]. However, for the CoPt clusters an increased electron density at the Co atoms was proposed, leading to the observed catalytic oxidation reaction. The complex interplay between the intra-cluster and cluster-substrate interaction, leading on Ni with a high density of states at the Fermi level to a more complex hybridization of the electronic states as in the gas phase, together with the not yet fully understood oxidation process makes it difficult to address the driving catalytic atomic sites.
The Co 2 Rh y clusters show a similar but smaller increase of the Co binding energies upon alloying. In contrast to the Co 1 Rh y clusters, the contributions to the higher binding energy upon alloying and oxidation could not be determined since the Co 2 dimer on Ni does not oxidize in the used preparation scheme. Nevertheless, the energy shift of about 0.2 eV is in the order of the observed increase with cluster size [34].

Magnetic properties
While the XAS spectra in figure 2 only contain spin-averaged information, the XMCD spectra shown in figure 4 hold information about the spin-resolved unoccupied density of states. The spectra show the difference of the absorption signals for left and right circular polarized light. Using the so-called magneto-optical sum rules [37,38], the spectra can be related to the element-specific orbital (m l ) and effective spin (m m m 7 s s t eff = + ) magnetic moments. The effective spin magnetic moment represents the spin moment (m s ) together with the magnetic dipole moment (m t ) which accounts for the asphericity of the spin moment distribution. In the following, the effective spin moment m eff s will be denoted for simplicity as spin moment m s . In figure 4, the Co 1 adatom exhibits the strongest dichroism signal of all measurements. Its orbital and spin magnetic moment has been measured and reproduced before [28,29] and serves as reference for the accuracy of the preparations and measurements performed. The values in table 1 show that the evaluated magnetic moments are in good agreement to the former results.
A comparison between the Co 1 and the Co 1 -O spectra shows a change of sign in the dichroism signal as has been already reported earlier [28]. The change in sign is related to a different coupling to the underlying Ni substrate, meaning ferromagnetic coupling for Co 1 and antiferromagnetic coupling for Co 1 -O respectively. The different coupling of the oxidized adatom is important since alloying of Co 1 and Co 2 with Rh made it impossible to study non-oxidized clusters. Thus, the comparison of the calculated oxidation ratio with the evaluated magnetic moments and the observed coupling to the Ni substrate holds information how the oxidation of the Co x Rh y clusters effects their magnetic properties. Table 1 and figure 5(a) show that adding one Rh atom to the Co 1 adatom results in a 70% decrease of the total magnetic moment. The XMCD signal gives a small ferromagnetic coupling to Ni, and the enhanced activity due to Rh alloying results in half of the clusters being oxidized. Further alloying by one Rh atom to Co 1 Rh 2 flips the XMCD signal at the L 3 -edge to positive values and antiferromagnetic coupling. Note that the XMCD signal is shifted by 0.5 eV similar to Co 1 -O. However, compared to Co 1 -O the total moment is tripled with the spin moment m eff s increasing twice as much than the orbital moment. As a consequence the m l /m eff s ratio decreases. The high oxidation ratios, e.g. 80% for the Co 1 Rh 2 clusters, make it important to address whether the different coupling for Co 1 Rh 1 and Co 1 Rh 2 is mainly driven by Rh or oxygen. While the L 3 energy shift is most likely due to the oxygen, it was not possible to disentangle the different Co-O, Co-Rh contributions in the XMCD spectra, for example by a linear combination model between ferro-and antiferromagnetic coupled clusters as applied in [28]. Thus, XMCD measurements at the Rh M 2,3 -edges were performed for the antiferromagnetic coupled Co 1 Rh 2 clusters. In figure 6, the XAS and XMCD signals are shown respectively. No magnetic contribution of the Rh atoms is observed in contrast to induced moments in CoRh thin films [5], exchange coupled multilayers [8], nanoparticles [9,13] and the nearest neighbors of Co adatoms on Rh(111) [39].
The Co 2 dimer has a comparable orbital moment and an increased spin moment with respect to the adatom [29]. Thus, the total magnetic moment is increased while the ratio decreases by about 25%, as depicted in figures 5(a) and (b). Alloying the Co 2 dimer with a Rh atom again leads to an enhanced reactivity for the Co 2 Rh 1 cluster. The evaluated oxidation ratio is nearly half the value of Co 1 Rh 1 and could be a hint for a non-oxidized second Co atom. Comparing the magnetic moments of Co 2 Rh 1 and Co 1 Rh 1 , the observed increase of the Co 2 Rh 1 orbital and spin moment by a factor 2 and 2.5, respectively, strengthens this point. However, with respect to the pure dimer both magnetic moments are reduced by 35% and 55%. Thus, the total moment is halved.
Having seen already a distinct size dependent effect when adding a second Rh atom to Co 1 Rh 1 , a similar effect appears for the transition from Co 2 Rh 1 to Co 2 Rh 2 . The reactivity of the clusters decreases and only a small fraction is oxidized ( figure 3(a)). Compared with the observation for Co 1 Rh y clusters in figure 5(a), this Table 1. By sum rule analysis evaluated orbital, spin and total magnetic moments per number of d-holes and in units of B m are given for Co x Rh y alloy clusters on Ni/Cu(100). The estimated content of oxidized clusters per sample is given with an error of 10%  . The magnetic moments of the full oxidized Co 1 spectrum have been taken from [28] and for the Co 2 dimer from [29].  interestingly leads to an increase of the Co 2 Rh 2 m l and m s eff values, both increasing by 40% and 45% with respect to Co 2 Rh 1 . Exclusively for all Co x Rh y clusters investigated, the Co 2 Rh 2 orbital moment is enhanced by alloying. The magnetic moments are of comparable size to the pure Co 1 and Co 2 . Similar to the Co 1 Rh 2 clusters, no XMCD signal at the Rh M 2,3 -edges was observed.
By adding three Rh atoms to the Co 2 dimer the size-dependence of the Co 2 Rh y total magnetic moments is revealed beautifully, as can be seen in figure 5(a). As for the Co 1 Rh 2 clusters, having likewise an alloying ratio above 50%, an antiferromagnetic, positive Co L 3 -edge contribution is observed in the XMCD spectrum of figure 4 accompanied with a sign change for the magnetic moments. A detailed study of the Co 2 Rh 3 XMCD spectrum shows that there seems to be a small ferromagnetic contribution in addition. A similar effect is implied in the Co 1 Rh 2 XMCD signal. Surprisingly, the Co 2 Rh 3 XMCD spectrum at the Rh M 2,3 -edges, shown in figure 6, exhibits a dichroism signal. The negative signal implies an antiparallel coupling between the magnetic moments of the Rh and Co atoms (compare to figure 4) and a parallel coupling between the magnetic moments of Rh and Ni. Assuming that the clusters prefer a certain structure on the surface, this points to a complex, non-collinear spin structure [40,41]. The evaluated small orbital and spin magnetic moments strengthen this interpretation. The Co 2 Rh 3 cluster is showing the highest reactivity of all Co 2 Rh y clusters. Approximately half of the clusters are oxidized. Compared to the other antiferromagnetic coupled Co 1 Rh 2 cluster, the orbital and spin moments are reduced drastically by about 50%. The ratio of the magnetic moments stays unchanged.
A detailed study of the ratios of the magnetic moments in figure 5(b) reveals a certain dependency for Co 1 Rh y and Co 2 Rh y clusters. Compared to the pure Co 1 and Co 2 , the ratios are enhanced if the alloying ratio is 50  %.
Further alloying to 50 > % lowers the ratios to a minimum. For the smaller alloying ratios (Co 1 Rh 1 , Co 2 Rh 1 ) this is related to a faster decreasing spin moment while for the higher alloying ratios (Co 1 Rh 2 , Co 2 Rh 3 ) the orbital moment is suppressed. An exception makes the Co 2 Rh 2 cluster as described before.

Summary and conclusions
In summary, we have performed XAS and XMCD studies on mass-selected Co x Rh y clusters deposited on a magnetized Ni/Cu(100) surface. The clusters were systematically varied in size atom by atom up to the five atomic Co 2 Rh 3 . Alloying with Rh leads to an enhanced chemical reactivity towards oxygen for each Co x Rh y preparation. The photon excitation energy of the alloyed Co atoms is increased which corresponds to a reduced electron density at the Co site. Depending on their size and composition, the magnetic moments of all alloyed clusters are reduced by oxidation and exhibit an oscillatory ferro-/antiferromagnetic transition of the exchange coupling between the Co atoms and the Ni surface. The Co 1 Rh y cluster magnetic moments are of comparable size and with respect to Co 1 suppressed to about 20% of the orbital and 25% of the effective spin magnetic moment. For Co 2 Rh y , higher magnetic moments than for Co 1 Rh y are found together with a reduced oxidation ratio. Thus, the second Co atom does not seem to be oxidized. An exception is given by the Co 2 Rh 3 cluster with the lowest magnetic moments of all alloy clusters. The dichroism signals of the Co and Rh atoms show an antiferromagnetic coupling for Co and a ferromagnetic for Rh with respect to the Ni surface, leading to a possible non-collinear spin structure. The orbital to spin moment ratios show a similar dependency for Co 1 Rh y and Co 2 Rh y . Alloying ratios less than or equal to 50% lead to enhanced ratios compared to Co 1,2 , while further alloying to greater than 50% decreases the ratios. An exception is given by Co 2 Rh 2 , showing an oxidation ratio below 20% and magnetic moments in the order of Co 1,2 . Figure 6. The XAS and XMCD spectra at the Rh M 2,3 -edges of Co 1 Rh 2 and Co 2 Rh 3 on Ni/Cu(100) are presented. The Co 1 Rh 2 clusters exhibit no dichroism. In contrast, for Co 2 Rh 3 a magnetic contribution of the Rh atoms is found.