Importance of thermovoltage on the interpretation of STM data
Homoth, J.; Wenderoth, M.; Druga, T.; Ulbrich, R. G.
Germany

We present Scanning Tunneling Microscopy data taken at 5K on Ag(111) surfaces in UHV. Standing-wave patterns (SWP) of the surface state were observed in constant current topography (CCT) data at low bias voltages (VBias). Similar to results reported on Be(0001) giant corrugations of 0.5 Å of the SWP were found [1]. To investigate the origin of the enhanced corrugation we studied the SWP with a variable temperature gradient between tip and sample. The temperature difference was controlled by means of a focused laser beam on the tip. An increase of amplitude and a lateral shift of the maxima with increasing temperature difference was observed. According to the theory of Tersoff and Hamann, data at low VBias are interpreted as surfaces of constant local density of states LDOS(VBias,x,y). We demonstrate that the temperature gradient across the vacuum gap produces a spatial varying thermovoltage VTh(x,y) that is hidden in CCT data. VTh(x,y) depends - in contrast to the true LDOS – on the gradient ∂LDOS(x,y)/∂E at the Fermi energy and increases with increasing temperature difference.
To study the crosstalk of thermovoltages on CCT images we used a Scanning Tunneling Potentiometry (STP) Microscope, that compensates VTh(x,y) by a second interlaced feedback loop at every scan point with an accuracy of 1 μV at 5 K [2]. Our approach allows a simultaneous mapping of i) the standard CCT ii) VTh(x,y) separately and iii) CCT with VTh(x,y) compensated. We demonstrate that only CCT data where VTh(x,y) is compensated represent the pure LDOS. The measured corrugation amplitude, phase shift and temperature dependence of the above data sets must be attributed to the crosstalk of VTh(x,y) on the CCT. Calculations in the framework of Ref. 3 are presented.
Since the tips of variable temperature STMs (VT STMs) are typically at room temperature while the sample is at temperatures down to 20 K the interpretation of CCT data at low VBias should be carefully reconsidered.
This work was supported by the SFB 602 TP A7.
[1] P. T. Sprunger et.al., Science 275, 1764 (1997) [2] K. Engel et. al., Phys. Rev. B 63, 165402 (2001) [3] J. A. Stövneng et. al., Phys. Rev. B 42, 9214 (1990)
back