J. Phys. Chem. C 2009, 113, 11301–11305
11301
Work Function Measurements of Thin Oxide Films on MetalssMgO on Ag(001) T. Ko¨nig, G. H. Simon, H.-P. Rust, and M. Heyde* Fritz-Haber-Institut der Max-Planck-Gesellschaft, Faradayweg 4-6, D-14195 Berlin, Germany ReceiVed: February 10, 2009; ReVised Manuscript ReceiVed: April 8, 2009
The metal work function is a suggested key parameter for charging adsorbates with high electron affinity on thin oxide films grown on metal single crystal surfaces. Here, the first experimental data for the work function shift of Ag(001) induced by 1, 3, and 8 monolayers of MgO is measured in situ by three independent scanning probe techniques on the same surface area, making the different approaches comparable. Furthermore, the results are compared to density functional theory calculations. The measurements are performed using a dualmode dynamic force microscope and scanning tunneling microscope working in ultrahigh vacuum at low temperature (5 K). The methods to detect the work function shift are based on Kelvin probe force microscopy measuring the contact potential difference, I(z) curves and field emission resonances. Introduction Much effort has been placed on the development of experimental and theoretical concepts in catalysis to understand the mechanisms responsible for activity and selectivity over the past decades.1,2 Still, concepts are far from being complete. The surprising catalytic activity of Au nanoparticles3,4 is attracting great attention in the research areas of catalysis and surface science. The studies of this effect have focused on the particle size, contribution of the support, and charging of the particles.5-7 The deposition and creation of Aun clusters (n ) 1-8) on supported ultrathin oxide films such as MgO and alumina have been intensively studied by various methods during the past years.8-16 Density functional theory (DFT) calculations predicted a charge transfer by tunneling from the metal oxide interface to the adsorbed Au species deposited on top of the supported ultrathin MgO film.8 Thus, these particles might become chemically active.17 The charging of the Au species has been experimentally confirmed by scanning tunneling microscope (STM) investigations.11,12 The transfer of charge from the metal oxide interface depends on a variety of different mechanisms and effects as recently discussed by different theoretical papers.18,19 Since the tunneling probability decreases with increasing MgO layers, the thickness must not exceed the tunneling length; otherwise, the charge transfer is shut off. Therefore, the film thickness might be used as a parameter to tune the catalytic properties of the adsorbed Au particles.20,21 Moreover, the electron affinity of the adsorbed species should be high as it is, e.g., for Au. A discussed factor for the charge transfer is a shift in the metal work function resulting in a reduction of the tunneling barrier and thus an increase of the tunneling probability. This work function shift results mainly from the metal/oxide bonding distance. The oxide film reduces the surface electron density overspill into the vacuum and thus the surface dipole that the electrons, tunneling from the substrate to the adsorbed particle, have to overcome. This is clearly resolved for MgO/Mo(001) having a 7-fold larger adhesion energy than MgO/Ag(001), thus resulting in a shorter oxide-metal bond followed by a higher surface electron density compression and consequently a larger work function shift than for MgO/ * E-mail:
[email protected].
Ag(001).22 Furthermore, the work function shift depends on metal-induced gap states. Such states arise from penetration of metal electron wave functions into the oxide overlayer and from chemical bonds between metal and oxide.23,24 The shift of the metal work function of MgO/Ag(001) has been calculated by DFT,25 and the first experimental data are now presented here. Experimental Section Beside integration techniques, e.g., photoemission spectroscopy and low-energy electron microscopy, there are three independent methods with lateral resolution on the nanometer scale and smaller that are in principle able to detect a work function shift: contact potential difference (CPD) based on Kelvin probe force microscopy (KPFM), the detection of the tunneling current and its exponential dependency, i.e., I(z) curves and field emission resonances (FER). These three methods have been used to detect the Ag(001) work function shift with respect to the number of adsorbed MgO layers. Therefore, we prepared 0.5 monolayer (ML) (effectively 1 ML high islands), 3 and 8 ML MgO on Ag(001) see Figure 1a-c. The preparation conditions were as follows: The Ag(001) has been sputtered with Ar+ ions at a current density of 10 µA/cm2 and 800 V for 30 min. Afterward, the Ag(001) was annealed to 690 K for 15 min. The sputtering and annealing cycle has been repeated several times. Figure 1a shows clean Ag(001) terraces as a basis for the MgO preparation. MgO was evaporated from a Knudsen cell in an oxygen atmosphere of 1 × 10-6 mbar, a substrate temperature of 560 K, and a deposition rate of about 1 ML of MgO/min. The desired number of MgO layers could be grown onto the Ag(001) (Figure 1b,c). Figure 1d shows a growth model of MgO on Ag(001). The setup employed is a dual-mode low-temperature dynamic force microscope (DFM) and scanning tunneling microscope (STM). The microscope operates in ultrahigh vacuum at 5 K. Details about the tuning fork setup and the electronics can be found elsewhere.26-28 The characteristic of the setup is the established separation of the frequency and the tunneling current signal.27 The resonance frequency shift of the tuning fork and the tunneling current, both with respect to the tip-sample separation, can be recorded simultaneously at the same surface area. Note that the tunneling current is detected via an additional wire attached to the Pt/Ir tip. From the frequency shift, the force
10.1021/jp901226q CCC: $40.75 2009 American Chemical Society Published on Web 06/05/2009
11302
J. Phys. Chem. C, Vol. 113, No. 26, 2009
Ko¨nig et al.
Figure 3. Contact potential measurements on clean Ag(001) (dashed parabola fit) and on 3 ML MgO/Ag(001) (dotted parabola fit). The dashed-dotted line indicates the shift of the work function. For each curve, the tip position was constant.
voltage between tip and sample ∆U can be written as
∆U ) UDC Figure 1. STM images: (a) bare Ag(001), 20 × 20 nm , UDC ) 10 mV, I ) 1 nA. (b) MgO islands (0.5 ML) grown on Ag(001), 50 × 50 nm2, UDC ) 3.5 V, I ) 400 pA. (c) MgO film of 3 ML on Ag(001), 30 × 30 nm2, UDC ) 3.5 V, I ) 200 pA. (d) Growth model of MgO on Ag(001), 1.5 × 1.5 nm2. The O-atoms occupy top sites, Mg-atoms the hollow sites of the Ag(001).
∆Φ e
(1)
2
where ∆Φ is the contact potential difference (difference in work function of tip and sample). The electrostatic energy Eel of a parallel plate capacitor with capacity C can be written as
Eel )
1 C∆U2 2
(2)
The force between tip and sample can now be derived as the gradient of Eel in the z direction as
Fel ) -
∂Eel 1 ∂C )∆U2 ∂z 2 ∂z
(3)
These considerations lead to Figure 2. Simultaneously recorded resonance frequency shift and tunneling current signals. (a,b) Signal vs distance curve at constant bias voltage. (c,d) Signal vs bias voltage at constant tip position.
acting between tip and the sample can in principle be calculated.29 The different independently detectable signals are shown in Figure 2. The switching between the two feedback signals can be done rapidly. Thus, this sensor setup is the ideal base for the in situ work function shift characterization using independent techniques investigating the same surface area. Therefore, this approach gives direct insights into the relation of DFM and STM investigations and make their results taken on the same surface area directly comparable. At this point, the three independent approaches (CPD, I(z), FER) to determine the shift of the metal work function will be discussed, followed by the data analysis. Contact Potential Measurements Based on KPFM. In the Kelvin probe force microscopy mode, the contact potential difference between tip and sample can be determined with a lateral resolution on the nanometer scale. Tip and sample are not in direct contact; therefore, KPFM is a nondestructive technique, but they are electrically connected via the electronics. As a consequence, their Fermi levels will align. This leads to an electric field between tip and sample resulting in an electrostatic force. As one applies a dc bias voltage UDC, the
Fel ) -
1 ∂C ∆Φ U 2 ∂z DC e
(
2
)
(4)
The change in ∆Φ when going from the clean Ag(001) to 8 ML MgO was determined at constant x, y, and z positions. The bias voltage was swept, and the parabolic behavior of the electrostatic force was detected as the resonance frequency shift (frequency modulation mode30). A typical curve taken on the clean Ag(001) and on 3 ML MgO is shown in Figure 3. Since the z position is constant, only the change of longrange electrostatic force is detected, while contributions to the frequency shift from other forces result in a frequency shift offset, which does not carry information on the work function shift and can therefore be subtracted. Since the maximum of the parabola in Figure 3 is given by ∆Φ/e ) UDC (compare equation 4), the shift of the maximum along the abscissa carries the information of interest. In a single measurement, ∆Φ ) Φsample - Φtip. By taking the difference ∆ΦMgO/Ag - ∆ΦAg, the effect of the tip can be canceled and the Ag(001) work function shift can be derivedsthis is also true for the other two techniques used here. Furthermore, the change in capacity established by tip and sample is given by the opening factor of the parabola. Within this model, it is assumed that the MgO structure lowers the electric field inside its layer but does not affect the electric field outside. Measurements
Thin Oxide Films on MetalssMgO on Ag(001)
J. Phys. Chem. C, Vol. 113, No. 26, 2009 11303
Figure 4. Tunneling current vs distance relation, i.e., I(z) curve (inset shows log plot). Dashed line indicates exponential fit.
TABLE 1: Comparison of Experimental and Theoretical Data25* theory
experiment
number of MgO layers on Ag(001)
∆Φ [eV]
CPD ∆Φ [eV]
I(z) ∆Φap [eV]
FER ∆Φ [eV]
MgO island 3 8
-0.94* -1.27
-0.5 -1.1 -1.1
-2.0 -1.4 -1.2
-1.2 -1.4 -1.3
* Calculated for 1 ML.
have been taken on the clean Ag(001), 0.5 ML (effectively 1 ML high), 3 and 8 ML MgO grown on Ag(001). The results are given in Table 1. I(z) Curves. The second method applied to characterize the shift in the metal work function utilizes the STM mode of our setup. The oscillation of the tip was disabled, and the tip was situated at a fixed lateral position. The tip was moved toward the sample surface while we simultaneously recorded the tunneling current and the z displacement of the tip. The tunneling current depends exponentially on the z displacement as described by Binnig and Rohrer31 by the well-known equation
I(z) ∝ exp(-2z√2mΦap /p2)
(5)
where m is the mass of the electron and Φap the apparent barrier height. A typical curve is shown in Figure 4. The graph shows the exponential dependence of I on z. The inset presents a logarithmic plot of the same data. The apparent barrier height can be calculated from an exponential fit or from the slope of the logarithmic plot.32 Nevertheless, it is not straightforward to extract the work function from the apparent barrier height. Controversial statements have been made in the past. Lang33 claims in his theoretical considerations a convergence of Φap to Φ for large z displacements, while Chen34 in his theoretical approach shows that Φap stays constant until point contact. The latter has been confirmed by Besenbacher et al.35 in an experiment. They point out that it is crucial to measure UDC simultaneously to I(z), due to the finite impedance of the tunneling current amplifier. We think that this effect is mainly a consequence of the deviation of the amplifier from an ideal amplification. Thus, a voltage drop occurs at the input of the amplifier. Furthermore, the resistance of the bias voltage supply can cause additional voltage deviations. Therefore, for performing these experiment a highly advanced current amplifier is necessary in combination with a stable bias voltage supply. Moreover, the bias voltage might deviate from the set value due to elastic deformations within the tunneling gap.35 These deformations reduce the tip sample distance in a nonlinear way, which can affect the tunneling current significantly. Depending on the stiffness of tip and sample, the relaxation effect occurs
Figure 5. Field emission resonance curve taken on MgO. Feedback on I(z).
with the tip a few angstro¨ms above the surface. The I(z) curves taken within these investigations do not show deviations from exponential behavior (see Figure 4). Field Emission Resonances. As a third method to identify the shift in the work function, we investigated field emission resonances, operating our setup in the STM mode. FER are a result of interfering electron waves. As soon as the applied bias voltage exceeds the work function, electrons can enter a regime between tip and sample. The electron waves are reflected at the potential walls formed by tip and sample leading to standing waves in the vacuum gap. These standing waves are the eigenstates to the sample-vacuum-tip potential. Gundlach36 in 1966 was the first to predict field emission resonances based on a theoretical approach. Therefore, FER are often referred to as Gundlach states. Binnig and Rohrer37 were the first to confirm FER experimentally. Since the detection of a large number of resonance states requires high voltages, tip changes are very likely in constant height mode due to the high electric field. Thus, the feedback loop was closed to kept the tunneling current constant. The tip was not oscillating and the FER were taken at constant lateral position while the bias voltage was swept (see Figure 5). For the analysis of the data, we assumed a 1D potential between tip and sample arising from the applied bias voltage.38 The condition for the FER to occur is then given by39
eVn ) Φ +
( ) 3πpe 2√2m
2/3
E2/3n2/3
(6)
where e is the charge of an electron, Vn is the voltage of the nth resonance, Φ is the work function, E is the electric field, and n is the number of the nth resonance. This equation does not account for the image potential, which rounds off the corners of the potential36 between tip and sample. Furthermore, the barrier is reduced, resulting in an increase of the tunneling current. The influence of the image potential mainly effects resonances with small n. Thus, the first resonance of the spectra was excluded from the data analysis. By fitting equation 6 to Vn vs n, we can determine the work function of the different sample systems and the shift by taking the difference from the bare Ag(001). Results and Discussion The results of the three independent techniques are listed in Table 1. The first column presents the number of MgO layers grown on the Ag(001) by the growth method previously stated. The MgO islands were effectively 1 ML high. The second column presents theoretical values calculated by Pacchioni et al.25 The calculations are based on density functional theory (DFT) using the generalized gradient approximation (PW-91 functional)40 implemented in the VASP program41,42 which uses
11304
J. Phys. Chem. C, Vol. 113, No. 26, 2009
Ko¨nig et al. N
Vts )
∑
N
Vi ) Vj +
i
∑ Vm
(7)
m*j
where Vj is the interaction of the tip atom with the surface atom directly beneath the tip. ΣmN * jVm describes the interaction of the tip atom with all other surface atoms except the one at position rj directly under the tip. The pair potential Vi is given by the Coulomb interaction
Vi )
q1q2 1 4πε0 ri
(8)
where ε0 is the dielectric constant, and q1 and q2 are the charges of the tip and a surface atom at distance ri, respectively. Thus, ri is given by Figure 6. (a) Work function shift vs z displacement. The crosses were taken on the MgO islands, while the circles indicate measurements between the islands on the Ag(001). For large z displacements, the measurements on the islands reveal the same values as on the Ag(001) between the islands. For small distances, the maximum positions diverge. (b) Simulation of the tip sample interaction (not all surface atoms considered in the simulation are shown). The projection of the Coulomb force onto the z axis is indicated. (c) The solid line represents the force Fj of the tip and the atom directly beneath the tip. The dashed line indicates the force Fts between the tip atom and all surface atoms. (d) Ratio Fj/Fts; see text.
√
ri ) (xq1 - xq2,i)2 + (yq1 - yq2,i)2 + (zq1 - zq2,i)2
(9)
Thus the force between the tip and the N surface atoms summarizes to
Fts(z) ) -
(
∂Vts ∂ ) - Vj + ∂z ∂z N
a plane wave basis set and a projector augmented wave (PAW) method43 for the treatment of core electrons. The third to fifth columns show the experimental data. Considering the CPD measurements, we find good agreement with theory for the work function shift expect for the MgO islands. The experimental value of -0.5 eV is too small compared to the theoretical value of -0.94 eV. This discrepancy arises from an averaging effect. Due to the long-range nature of the electrostatic force, the area of tip and sample involved in the CPD measurements is larger than the MgO islands; thus, measurements average over MgO covered areas and the bare Ag(001). This becomes obvious when we position the tip at different surface positions such that the relation of covered and uncovered areas varies. Furthermore, at tip sample distances of about 3 nm above the surface the measured work function shift on a MgO island (crosses) and on the bare Ag(001) (circles) between the islands are indistinguishable. When the tip gets closer to the surface, the measured shift on the island and the Ag diverge. The maximum difference is -0.5 eV taken at a tip position 200 pm closer to the surface than the set point. Note that the zero position of the z displacement belongs to the set point UDC ) -2 V, I ) 500 pA. At closer distances, tip changes occurred while sweeping the bias voltage, although the bias sweep was limited from -2 V to +2 V. The divergence of the shift at close distances shows clearly an increase in lateral resolution of the work function shift. To explain the increase in resolution, we simulated the attractive electrostatic interaction of tip and sample by a onedimensional model. The tip is represented by a single atom, while the surface is given by 25 atoms (Figure 6b). The interaction of the tip atom with the surface atoms has been calculated in analogy to ref 44 by the summation of pair potentials. For the pair potential, a Coulomb potential has been chosen with a negative and positive charge resulting in an attractive electrostatic interaction (compare Figure 3 and, e.g., eqs 2 and 3). Thus, the interaction of the tip atom with the surface atom results in
) Fj(z) +
N
m*j
q1q2 1 z 4πε0 m*j r2 rm
∑
)
∑ Vm
(10)
m
Fj(z) and Fts(z) have been calculated for different z displacements see Figure 6c. The solid line represents the force Fj between the tip atom and the surface atom directly beneath it. The dashed line indicates the force between the tip and all surface atoms Fts. Figure 6d presents the ratio of these interactions; therefore, it is defined by ratio ) Fj/Fts. This ratio clearly indicates that the contribution to the force of the atom directly beneath the tip increases with decreasing tip-sample separation. Therefore, the increase in resolution for the CPD measurements on MgO islands with decreasing z separations as shown in Figure 6a can be explained. Nevertheless, a further resolution enhancement meaning operating at closer distances, however, was not possible due to the mentioned tip instabilities. On the 3 and 8 ML MgO film, this averaging effect is not important, since the layers cover large areas. The detected values for the work function shift from CPD are close to theory. The standard deviation of the CPD measurements presented in Table 1 is about 11%. For the I(z) measurements, we found for the 0.5 ML MgO a value for ∆Φap ) -2.0 eV, which considerably larger than the calculated DFT value, despite the high lateral resolution due to the exponential tunneling dependence on z. The derived values on 3 and 8 ML MgO are in good agreement with theory. Nevertheless, we have to mention that the standard deviation of this method is the largest, about 20%. Considering the FER, we find good agreement between the experimental and theoretical data. The measurements on the 0.5 ML MgO film benefit from the previously discussed dependency of the tunneling current resulting in a value of ∆Φ ) -1.2 eV. The standard deviation is about 11%. Overall, the three techniques show no significant change in the metal work function when going from 3 to 8 ML MgO. This indicates that the work function shift is built up within the first
Thin Oxide Films on MetalssMgO on Ag(001) layers and is dominated by the MgO/Ag interface. Therefore, charge transfer to adsorbed particles with high electron affinity is possible from the first monolayer MgO and is limited by the tunneling length of the electrons. Conclusion In summary, we have investigated the metal work function shift of 0.5, 3, and 8 ML MgO grown on Ag(001). The work function shift is one of the key parameters for charging adsorbed particles having a high electron affinity like Au deposited on defect-poor MgO layers. Therefore, the work function shift is essential for the catalytic properties of Au/MgO/Ag(001) and makes these thin films a research area in their own right,20,21 since their behavior is different from Au on thick films or bulk MgO. The CPD measurements average over a surface area with a radius of 15-30 nm. On the 3 and 8 ML films, the averaging effect is not an issue and the derived values are in good agreement with theory. However, on 0.5 ML the averaging of the surface area does affect the detected shift. The resolution could be enhanced at close tip-sample proximities. This resolution enhancement is explained by a simulation. The contribution to the force signal of the atom directly under the tip increases with decreasing tip-sample distance. Nevertheless, the CPD value on 0.5 ML is too small and the increase in resolution was limited by tip instabilities. To derive the work function shift from I(z) curves is not straightforward. Moreover, this method is highly affected by the experimental conditions and relaxation effects within the tunneling gap. In consequence, it might not be surprising that this method shows the largest standard deviation for the methods used here. The FER analysis shows on all sample systems good agreement with theory and is featured by their stable detection conditions resulting in a small standard deviation. Besides the interesting research background of the work function and its important role in adsorption processes like catalysis, this work presents the in situ comparison of three independent methods on the same surface area gained by a dual-mode DFM/STM setup. Acknowledgment. The authors acknowledge fruitful discussions and financial support from H.-J. Freund and technical support from G. Thielsch. References and Notes (1) Ertl, G.; Kno¨zinger, H.; Schu¨th, F.; Weitkamp, J. Handbook of Heterogeneous Catalysis; Wiley-VCH: Weinheim, 2008. (2) Freund, H.-J. Top. Catal. 2008, 48, 137–144. (3) Haruta, M.; Kobayashi, T.; Sano, H.; Yamada, N. Chem. Lett. 1987, 16, 405–408. (4) Haruta, M. CATTECH 2002, 6, 102. (5) Haruta, M. Catal. Today 1997, 36, 153. (6) Valden, M.; Lai, X.; Goodman, D. W. Science 1998, 281, 1648. (7) Bongiorno, A.; Landman, U. Phys. ReV. Lett. 2005, 95, 106102.
J. Phys. Chem. C, Vol. 113, No. 26, 2009 11305 (8) Pacchioni, G.; Girodano, L.; Baistrocchi, M. Phys. ReV. Lett. 2005, 94, 226104. (9) Yulikov, M.; Sterrer, M.; Heyde, M.; Rust, H.-P.; Risse, T.; Freund, H.-J.; Scagnelli, A.; Pacchioni, G. Phys. ReV. Lett. 2006, 96, 146804. (10) Sterrer, M.; Risse, T.; Pozzoni, U. M.; Giordano, L.; Heyde, M.; Rust, H.-P.; Pacchioni, G.; Freund, H.-J. Phys. ReV. Lett. 2007, 98, 206103. (11) Sterrer, M.; Risse, T. M.; Heyde, H.; -P, R.; Freund, H.-J. Phys. ReV. Lett. 2007, 98, 096107. (12) Simic-Milosevic, V.; Heyde, M.; Nilius, N.; Ko¨nig, T.; Rust, H.-P.; Sterrer, M.; Risse, T.; Freund, H.-J.; Giordano, L.; Pacchioni, G. J. Am. Chem. Soc. 2008, 130, 7814–7815. (13) Benia, H. M.; Lin, X.; Gao, H. J.; Nilius, N.; Freund, H.-J. J. Phys. Chem. C 2007, 111, 10528. (14) Frondelius, P.; Ha¨kkinen, H.; Honkala, K. Phys. ReV. B 2007, 76, 073406. (15) Kulawik, M.; Nilius, N.; Freund, H.-J. Phys. ReV. Lett. 2006, 96, 036103. (16) Nilius, N.; Ganduglia-Pirovano, M. V.; Brazdova, V.; Kulawik, M.; Sauer, J.; Freund, H.-J. Phys. ReV. Lett. 2008, 100, 096802. (17) Yoon, B.; Ha¨kkinen, H.; Landman, U.; Wo¨rz, A. S.; Antonietti, J.-M.; Abbet, S.; Judai, K.; Heiz, U. Science 2005, 307, 403–407. (18) Frondelius, P.; Hellman, A.; Honkala, K.; Ha¨kkinen, H.; Gro¨nbeck, H. Phys. ReV. B 2008, 78, 085426. (19) Girodano, L.; Pacchioni, G.; Goniakowski, J.; Nilius, N.; Rienks, E. D. L.; Freund, H.-J. Phys. ReV. Lett. 2008, 101, 026102. (20) Freund, H.-J. Surf. Sci. 2007, 601, 1438–1442. (21) Freund, H.-J.; Pacchioni, G. Chem. Soc. ReV. 2008, 37, 2224–2242. (22) Giordano, L.; Cinquini, F.; Pacchioni, G. Phys. ReV. B 2005, 73, 045414. (23) Goniakowski, J.; Noguera, C. Interface Sci. 2004, 12, 93–103. (24) Yeo, Y.-C.; King, T.-J.; Hu, C. J. Appl. Phys. 2002, 92, 7266– 7271. (25) Prada, S.; Martinez, U.; Pacchioni, G. Phys. ReV. B 2008, 78, 235423. (26) Heyde, M.; Kulawik, M.; Rust, H.-P.; Freund, H.-J. ReV. Sci. Instrum. 2004, 75, 2446–2450. (27) Heyde, M.; Sterrer, M.; Rust, H.-R.; Freund, H.-J. Appl. Phys. Lett. 2005, 87, 083104. (28) Rust, H.-P.; Heyde, M.; Freund, H.-J. ReV. Sci. Instrum. 2006, 77, 043710. (29) Sader, J. E.; Jarvis, S. P. Appl. Phys. Lett. 2004, 84, 1801. (30) Albrecht, T. R.; Gru¨tter, P.; Horne, D.; Rugar, D. J. Appl. Phys. 1991, 69, 668–673. (31) Binnig, G.; Rohrer, H. Surf. Sci. 1983, 126, 236–244. (32) Repp, J.; Meyer, G.; Olsson, F. E.; Persson, M. Phys. ReV. B 2005, 71, 075419. (33) Lang, N. D. Phys. ReV. B 1988, 37, 10395. (34) Chen, C. J. Introduction to Scanning Tunneling Microscopy; Oxford University Press: New York, 1993. (35) Olesen, L.; Brandbyge, M.; Sorensen, M. R.; Jacobsen, K. W.; Laegsgaard, E.; Stensgaard, I.; Besenbacher, F. Phys. ReV. Lett. 1996, 76, 1485. (36) Gundlach, K. H. Solid-State Electron. 1966, 9, 949–957. (37) Binnig, G.; Rohrer, H. HelV. Phys. Acta 1982, 55, 726. (38) Rienks, E. D. L.; Nilius, N.; Rust, H.-P.; Heyde, M.; Freund, H.-J. Phys. ReV. B 2005, 71, R241404. (39) Kolesnychenko, O. Y.; Kolesnichenko, Y. A.; Shklyarevskii, O.; van Kempen, H. Physica B 2000, 291, 246. (40) Perdew, J. P.; Chevary, J. A.; Vosko, S. H.; Jackson, K. A. R.; Pederson, M.; Singh, D. J.; Fiolhais, C. Phys. ReV. B 1992, 46, 6671. (41) Kresse, G.; Hafner, J. Phys. ReV. B 1993, 47, R558. (42) Kresse, G.; Furthmu¨ller, J. Phys. ReV. B 1996, 54, 11169. (43) Blo¨chl, P. E. Phys. ReV. B 1994, 50, 17953. (44) Ho¨lscher, H.; Allers, W.; Schwarz, U. D.; Schwarz, A.; Wiesendanger, R. Phys. ReV. B 2000, 62, 6967.
JP901226Q
