Scanning Thermal Microscopy (SThM) allows the thermal characterization of materials with a submicrometric spatial resolution . While obtaining qualitative trends, e.g. by means of maps, is straightforward, determining reliable quantitative thermal data from the experiments is challenging because the probe-sample heat exchange depends strongly on parameters such as the size, geometry and surface states of probe and sample [2-3]. There is a need for a complete and accurate thermal measurement methodology, which is the aim of this work. We focus on heated probes dissipating heat flux into samples initially at ambient temperature. Simultaneous heating and measurement of the SThM probe temperature provide information on the sample effective thermal conductivity.
For three different electrically-resistive and Joule-heated SThM probes, involving different sizes and sensor materials (etched Wollaston-wire microprobe, palladium nanoprobe and doped silicon nanoprobe), an improved methodology for thermal conductivity local-point measurement was developed. Studies in ambient and vacuum conditions were considered. The methodology enables to eliminate the impact of thermal drifts occurring in such experiments, to deal with the variation in the laser irradiation on the cantilever and probe, and to detect any change at the tip apex due to a deformation or a contamination. As a consequence, results are reproducible with a temperature resolution of the order of few millikevins.
Specimens of well-known thermal conductivity spanning between 0.1 and 150 W.m-1.K-1 and surfaces controlled in terms of roughness and nanomechanical properties were used as reference materials in order to obtain a calibration curve for each probe. In the three cases, their applications allow us to clearly confirm that thermal-conductivity measurements with SThM are limited to low thermal conductivity materials (k < few W.m-1.K-1). This is due to the designs of the cantilever of the three probes and may not easily be improved. As expected, reducing the probe size and placing it in vacuum improves the spatial resolution, but it also induces unfortunately an increase of sensitivity to the tip-sample contact physical parameters (e.g. roughness, thermal boundary resistance, surface state), which is detrimental to the accuracy.
The impact of roughness on the thermal conductance at the contact was studied in detail. Samples consisting of several sets of silicon surfaces with out-of-plane Sq roughness parameters of ~0, 0.5, 4, 7 and 12 nm were prepared by anodic oxidation . Our results show that roughness induces a thermal conductance decrease at the contact up to 35% compared to a flat silicon sample, depending on the probe and environment. Precise knowledge of the surface state is therefore key to accurate measurements.
The calibration and methodology were applied to characterize thin films of oxides and materials involved in electronics. The experimental results are backed by finite-element simulations (FEM) including effective thermal conductivities when the size requires it, local heat sinks
representing thermal constrictions of typical sizes smaller than the energy carrier mean free paths (when FEM is not able to capture the correct physics), and thermal boundary conductances in the Acoustic or Diffuse Mismatch Models (AMM-DMM). Advantages and drawbacks of SThM are discussed in light of the results.
 S. Gomès et al. Phys. Status Solidi Appl. Mater. Sci., 212, 3, 477-494 (2015)
 Y. Ge et al. Nanotechnology, 27, 32, 325503 (2016)
 F. Menges et al. Rev. Sci. Instrum., 87, 7, 74902 (2016)
 A. Pic et al. THERMINIC 2017 – 23rd Int. Work. Therm. Investig. ICs Syst., Sept (2017)
 J. Zemek et al. Surface Science, 602, 7, 1440-1446 (2008)
The research leading to these results has received funding from the EU FP7 Programme under GA n°604668. and French ANR Programme under project TIPTOP.