Pascal Ruffieux, EMPA, nanotech@surfaces Laboratory, Feuerwerkerstr. 39, 3602 Thun
A detailed knowledge of the local electronic properties of solids and their surfaces is crucial for an understanding of a large variety of processes, ranging from electron scattering in transport phenomena to catalytic reactions, which show pronounced site specificity. Recently, site-specific adsorption on inorganic surfaces has been reported for organic molecules on laterally inhomogeneous surfaces consisting mainly of varying stacking sequences of the outermost atomic layers. However, the understanding of the relevant local physical properties responsible for the site-specific substrate-adsorbate interactions remains very poor.
Among the first important parameters investigated for surfaces is the work function, which defines the minimum energy required for removing an electron from a metal to infinity at 0 K. However, a number of important surface-related phenomena, such as catalytic processes and electron emission, cannot be described with this macroscopic work function but require knowledge of local variations of the electrostatic potential close to the surface. This directly implies that local probes have to be used for the surface potential determination relevant for the above-mentioned phenomena. The pioneering work enabling the experimental determination of the local surface potential was based on photoemission of adsorbed xenon (PAX), which probes the surface potential at about 0.2 nm above the surface via the Xe electron states, which are pinned to the local vacuum level due to a weak coupling between the Xe atoms to the surface . Various techniques based on scanning probe microscopy, including Kelvin probe force microscopy and local barrier height measurements, are currently applied in order to overcome the missing imaging capability of PAX, which is crucial for complex nanostructured surfaces. However, both methods are either limited in spatial resolution or in the ability to quantitatively determine the local surface potential.
An alternative way to characterize the local surface potential is the analysis of the field emission resonances (FERs), which are detected with scanning tunnelling microscopy (STM) when applying voltages larger than the tip work function. Their sensitivity to variations of the surface potential has first been discussed by Binnig et al.  and has been applied for the qualitative description of surface potential modulations of thin ionic and oxide films grown on metal surfaces. Recently we showed  that the combination of the local detection of the FERs with STM and their modelling using a 1D model potential between STM tip and sample allows for a quantitative determination of the surface potential with a spatial resolution of ~1 nm. The method thus uses the marked proximity of the FERs to the surface and their sensitivity to local variations of the surface potential. It is applied for examining the site-specific interactions of C60 molecules deposited on a nanostructured template surface formed by deposition of two monolayers (MLs) of silver on Pt(111).
Experiments have been performed with a low-temperature STM (Omicron) working under ultrahigh vacuum conditions (base pressure 2·10-10 mbar). The Ag/Pt(111) strain relief pattern has been prepared by depositing 2 ML of silver on a clean Pt(111) surface and by subsequent annealing of the sample to 530°C. C60 fullerenes were deposited from a resistively heated quartz crucible where the deposition rate was determined with a quartz microbalance [4,5].
The annealing of two MLs of silver deposited on Pt(111) leads to the formation of a strain relief pattern exhibiting a high degree of long-range order. The apparent depressions in the STM topography image (Fig. 1) separate the three stacking domains labelled hcp1, hcp2, and fcc. They evolve from the relieving of strain built up by the lattice mismatch of ~4 % between Ag and Pt.
Regarding molecular deposition on the strain relief pattern we find a highly inhomogeneous immobilization of the molecules on the different surface domains. Deposition of C60 molecules at ~150 K followed by an annealing step to room temperature yields to the formation of stable molecular clusters, which are preferentially located in the hcp1 region [Fig. 1(b)]. This brings up the question of which local physical property determines the site-selective adsorption of C60 on the Ag/Pt strain relief pattern.
Local surface potential determination
The energy positions of the FERs are determined by locally acquiring z(V) spectra with the feedback loop closed, i.e. under constant current conditions. Accordingly, during the voltage ramp, the tip is further retracted from the surface when new states contribute to the tunnelling current. This results in a stair-shaped z(V) curve, which, by numerical differentiation, directly reveals the energy position of the lowest FERs (Fig. 2). Recording a large number of such spectra along a line crossing the different surface domains reveals a pronounced variation of the energy level of the lowest FERs. For the first state we find a difference of 0.23 V when comparing spectra recorded in the fcc and the hcp1 domain (Fig. 3).
In order to relate these energy shifts to variations in the local surface potential we numerically solve the one-dimensional Schrödinger equation in the direction perpendicular to the surface. The electrostatic potential V(z) between surface and tip takes into account the applied sample bias Vs, the varying tip-sample distance zt, the image plane zi at the sample surface and varying surface potential Δφ [Fig. 2(b)], leading to the expression
The unknown parameter such as tip work function and image plane position zi are determined by fitting the energy position of the simulated FERs to the four measured ones in the hcp2 region where the smallest lateral variation of the energy position is observed. These parameters are then kept constant and only the local surface potential Δφ is varied in the model potential in order to get the best agreement with the measured FERs.
Due to the increasing spatial extent of the FERs for higher order resonances we limit our analysis to the lowest two states. With this restriction a lateral resolution of the order of 1 nm is achieved for the FER-based surface potential determination. According to this analysis we can determine the surface potential at positions where spectra have been recorded (Fig. 3). We find a surface potential variation of 0.35 eV when comparing the fcc region with the hcp1 region. A similar analysis has been performed for a set of spectra acquired on a dense two-dimensional (2D) grid in the vicinity of the hcp1 region allowing the determination of the 2D surface potential landscape [Fig. 4(a)].
In order to further corroborate the appropriateness of the proposed method for a quantitative determination of the surface potential landscape, we have performed PAX experiments . This method is well established as a local work function probe for heterogeneous surfaces with the limitation of the missing simultaneous imaging of the surface. Figure 4(b) shows a Xe4d PAX spectrum for a 0.7(2) ML Xe coverage. In contrast to PAX spectra recorded on Ag(111), on the Ag/Pt strain relief pattern both spin-orbit split states reveal a broad distribution characterized by two main contributions (FWHM 0.30 eV). Since the binding energies in PAX are directly linked to local variations of the surface potential by ΔEFB(i,j) ≅ -Δφ(i,j) at two dissimilar surface sites i and j , this indicates a broad distribution of the local work function within the unit cell with two main contributions separated by 0.30 eV. These results can be compared to the FER-based analysis of the surface potential landscape by performing a histogram analysis of the local surface potential across the different surface domains. This analysis yields a broad distribution with two dominant contributions centred at -0.08 eV (mainly hcp1 and hcp2) and 0.2 eV (fcc). Their separation of 0.28 eV is in excellent agreement with PAX.
The observed variations of the local surface potential are firmly related to local changes of the in-plane lattice parameter a. From a careful analysis of atomic resolution STM images, we find average lattice parameters of 299(5), 289(4) and 180(4) pm for the hcp1, hcp2 and fcc domains, respectively. From an electron (and hence dipole) point of view this directly suggests a lowering of the work function for increasing lattice parameter, in agreement with the FER-based analysis of the local work function.
The local analysis of the FERs based on scanning tunnelling spectroscopy allows the determination of the surface potential landscape with simultaneous imaging of the nanostructures. The lateral resolution of the method depends on the spatial extent of the FERs and can thus be limited by choosing the lowest FERs. When restricting the analysis to the two lowest FERs this yields a resolution of the order of ~1nm. Regarding the understanding of site-selective interactions on nanostructured surfaces this gives access to an important surface property that is involved in adsorbate-substrate interactions with partial ionic bonding character Furthermore, it allows the determination of lateral electric fields, which is relevant for the description of site-specific interactions of polarisable adsorbates.
 K. Wandelt, Thin Metal Films and Gas Chemisorption (Elsevier, Amsterdam, 1987).
 G. Binnig et al., Phys. Rev. Lett. 55, 991 (1985).
 P. Ruffieux et al., Phys. Rev. Lett. 102, 086807 (2009).
 K. Aït-Mansour et al., Nano Lett. 8, 2035 (2008).
 K. Aït-Mansour et al., Phys. Chem. C, in press (2009), DOI:10.1021/jp901378v
[Released: May 2009]