Extracting Thermal Boundary Conductance of Arbitrarily Aligned Grain Boundaries with Beam Offset Hyperspectral Frequency Domain Thermoreflectance Imaging

Thermal transport at interfaces is often the limiting factor for performance in applications such as electronic chip assemblies, batteries, and energy generating devices. Understanding the role of interfacial thermal transport is of particular importance for thermoelectric devices, where low bulk thermal conductivity is desirable. Bulk thermal conductivity reduction can be accomplished through decreasing the material grain size to create more phonon scattering sites. While many models have been used to explain the thermal conductivity suppression, thermal transport suppression at grain boundaries is an assumed quantity because measuring an arbitrarily aligned interface between two grains is quite difficult experimentally. Both electrothermal and optical methods can be used to measure to angled or in-plane interfaces, but require extensive adaptations from traditional implementations to gain necessary sensitivity. In this talk we adapt hyperspectral frequency domain thermoreflectance (FDTR) imaging to measure the thermal boundary conductance between tin telluride (SnTe) grains.

The typical implementation of FDTR uses a pump laser beam to heat the sample, and a probe beam to measure the sample’s temperature response. A metal transducer layer is used to maximize both absorption of the pump beam and reflection of the probe beam. The temperature developed on the surface of the sample is typically fit to an analytical model which assumes radial symmetry. While computation time is low for the analytical approach, it is difficult to account for in-plane or angled interfaces such as grain boundaries. Here we present a new implementation of FDTR with the goal of increasing sensitivity to arbitrarily aligned interfaces. Our implementation of FDTR differs from typical measurements from the literature in 2 ways: (1) we offset beams scanned over the sample to increase sensitivity to grain boundary thermal boundary conductance (2) we use finite element analysis (FEA) combined with gradient descent optimization to fit for the thermal boundary conductance (TBC) of a SnTe grain boundary. Experimental results with this new technique allow us to image thermal transport at nanoscale grain boundaries via FDTR imaging. Using this data as inputs to our inverse solution results in extracted TBC values that are aligned well with previous literature. Through utilization of cross-sectional electron microscopy we verify that FDTR can visualize grain boundaries that are at arbitrary angles up to 60 degrees.

The varied alignment of boundaries present in typical multigrained samples present multiple challenges: (1) measurement sensitivity decreases as the angle with respect to the surface increases, and (2) the angle of the interface is unknown when the measurement is performed. These general problems are examined through sensitivity studies using the FEA model. Results show that there is an optimal offset distance for maximized measurement sensitivity compared to experimental noise. The results presented here demonstrate a beam-offset FDTR method and data analysis that can be used to gather new data on a wide variety of interfaces outside SnTe samples.

Sandia National Laboratories is a multimission laboratory managed and operated by National Technology & Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA0003525.