The apparent height of the tunneling barrier in scanning tunneling microscopy measured on Au͑111͒, Ag͑111͒, and Cu͑111͒ surfaces is found to vary significantly with the bias voltage. In particular, the apparent barrier height a is asymmetric with respect to the bias polarity on all three surfaces, in contrast to simple interpretations of a in terms of an average work function of tip and sample. Model calculations of the tunneling current, which take band-structure effects into account, describe the experimental observations. A fundamental physical property of a metal surface is its work function , which is defined as the minimum work required to remove an electron from the metal at T =0 K. Knowledge about at the atomic scale can improve the understanding of chemical surface processes such as heterogeneous catalysis and adsorption. 1,2 Scanning tunneling microscopy ͑STM͒ offers the possibility to study of conductive samples at the atomic scale 3,4 using the exponential variation in the tunneling current I with the tip excursion z. 5,6 From I͑z͒ data an apparent barrier height a may be determined, which-within the WKB approximation of the tunneling current through a rectangular barrier-is related to the sample work function s by 7,8where t is the work function of the tip apex, V is the applied tunneling bias voltage, and −e is the electron charge. Hence, the effect of the finite voltage on a is assumed to be linear. A more realistic barrier shape, which takes into account a classical image potential, leads to minor modifications. 9 Atomistic calculations for tunneling between an Au͑100͒ sample and tip confirm this picture in the zero-bias limit. 6 The number of publications using measurements of a to characterize surfaces is considerable. Often a is evaluated at elevated sample voltages with ͉V͉ ӷ 0.1 V assuming that Eq. ͑1͒ is valid. 10-16 On the other hand, surprisingly few reports on voltage-resolved a measurements are available. [16][17][18][19][20][21] For example, a on the reconstructed Au͑111͒ surface was reported to show a voltage polarity dependence, which was suggested to result from the surface dipole layer originating from the reconstruction. 19 In a subsequent study no voltage polarity dependence was found. 20 Calculations predicted a voltage polarity dependence of a on Al͑100͒ and excluded the possibility that the polarity-induced difference may be due to the formation of an additional surface dipole layer. 22 Overall, more work appears to be required to clarify the origin of the voltage dependency of a on metal surfaces and its interpretation.Here, we report low-temperature STM results for voltageresolved apparent barrier heights a on the ͑111͒ surfaces of Au, Ag, and Cu. Probing the unoccupied states of defect-free surface areas we find similar behavior for all surfaces. a remains rather constant for sample voltages up to V Ϸ 3.5 V. At higher voltages, a decreases within some hundred millivolts by ⌬ a Ϸ 2 eV and then undergoes oscillations due to Gundlach resonances. 17 When probing the occupied sta...