The 2013 Joseph W. Richards Summer Research Fellowship -- Summary Report: Microsegregation Effects on the Thermal Conductivity of Silicon-Germanium Alloys

Abstract

Due to increased need for renewable energy sources in recent years, a significant number of both experimental and theoretical efforts have been undertaken to find effective ways to enhance the performance of thermoelectric (TE) energy conversion. Since the 1960s, SiGe alloys have received much attention as one of the promising candidates for TE materials, due to their low thermal conductivity (κ), as compared to pure Si and Ge. Earlier studies1,2 demonstrated that the low κ is mainly attributed to phonon scattering as a result of the mass difference between Si and Ge atoms (the so-called alloy scattering). While the strength of alloy scattering is a strong function of the Si/Ge ratio, previous experiments3 also showed evidence that Si and Ge atoms often remain locally segregated in bulk SiGe samples prepared by mechanical alloying. However, no research has been reported regarding the microsegregation effect on κ. As a part of the ECS summer project, we performed a computational analysis to explore how the local atomic arrangement affects thermal transport in bulk SiGe; some results of this work are presented herein. To investigate the microsegregation effect, we prepared several Si0.8Ge0.2 samples by embedding spherical Ge particles of different sizes (ranging from 5 to 293 atoms) in the Si matrix. As illustrated in Fig. 1, embedded Ge particles were randomly positioned but not allowed to overlap each other. A nonequilibrium MD (NEMD) method4 with the Stillinger-Weber (SW) potential model5 was used to calculate the κ of SiGe alloys at 300 K, while the SW parameters were modified using the first-principles-based force-matching method.6 For each of the Si1-xGex systems considered, five independent NEMD simulations were performed with different atomic arrangements and initial velocity distributions. All NEMD simulations were performed using LAMMPS (Large-Scale Atomic and Molecular Massively Parallel Simulator)7 with a time step of 1 fs; a detailed description of the simulation steps can be found in Ref. 8 and 9. Figure 2 shows the variation of κ for the Si0.8Ge0.2 samples as a function of Ge particle diameter (De); here, De is approximated by (6NGeVGe/π) 1/3, where NGe is the number of Ge atoms in the particle and VGe is the volume per atom for Ge (= 0.0238 nm3 from our DFT-GGA calculation). The κ is predicted to monotonically increase with Fig 1. The various Si0.8Ge0.2 configurations show a random distribution of Si and Ge atoms (random) and embedded Ge particles with different diameters (De = 0.91, 1.58, and 2.37 nm) in the Si matrix. Green (black) balls and yellow lattices represent Ge and Si atoms, respectively.