Effect of the spin-orbit interaction on the thermodynamic properties of crystals: The specific heat of bismuth

TL;DR

This study investigates how spin-orbit interaction influences the specific heat of bismuth, showing that including this effect in calculations significantly improves agreement with experimental data.

Contribution

The paper demonstrates that accounting for spin-orbit interaction in ab initio calculations reduces discrepancies in specific heat predictions for heavy-element crystals, especially bismuth.

Findings

Including spin-orbit interaction decreases the discrepancy in specific heat maximum from 20% to 7%.

Reducing the spin-orbit Hamiltonian by 20% yields exact agreement with experimental data.

The effect is significant for heavy atoms like bismuth, impacting thermodynamic property calculations.

Abstract

In recent years, there has been increasing interest in the specific heat $C$ of insulators and semiconductors because of the availability of samples with different isotopic masses and the possibility of performing \textit{ab initio} calculations of its temperature dependence $C(T)$ using as a starting point the electronic band structure. Most of the crystals investigated are elemental (e.g., germanium) or binary (e.g., gallium nitride) semiconductors. The initial electronic calculations were performed in the local density approximation and did not include spin-orbit interaction. Agreement between experimental and calculated results was usually found to be good, except for crystals containing heavy atoms (e.g., PbS) for which discrepancies of the order of 20% existed at the low temperature maximum found for $C/T^3$. It has been conjectured that this discrepancies result from the neglect of spin-orbit interaction which is large for heavy atoms ($\Delta_0\sim$1.3eV for the $p$ valence electrons of atomic lead). Here we discuss measurements and \textit{ab initio} calculations of $C(T)$ for crystalline bismuth ($\Delta_0\sim$1.7 eV), strictly speaking a semimetal but in the temperature region accessible to us ($T >$ 2K) acting as a semiconductor. We extend experimental data available in the literature and notice that the \textit{ab initio} calculations without spin-orbit interaction exhibit a maximum at $\sim$8K, about 20% lower than the measured one. Inclusion of spin-orbit interaction decreases the discrepancy markedly: The maximum of $C(T)$ is now only 7% larger than the measured one. Exact agreement is obtained if the spin-orbit hamiltonian is reduced by a factor of $\sim$0.8.