Abstract
Within the framework of density functional theory (DFT), the total energy of crystal structures is calculated at zero temperature. Herein, we briefly discuss the DFT-based lattice-dynamics approach for computing crystal free energy, the quantity needed in various non-zero-temperature contexts. We illustrate this well-established approach by examining the temperature-dependent thermodynamic stability of several crystalline materials, including ZrO$_2$, HfO$_2$, KBH$_4$, and Zn(BH$_4$)$_2$.
Introduction
Standard Density Functional Theory (DFT) calculations provide a "snapshot" of materials at absolute zero ($T=0$). In this formalism, atoms are treated as static points in space, ignoring the constant vibrations present in real-world applications at finite temperatures. These vibrations contribute to the "vibrational free energy," a critical component for calculating Gibbs or Helmholtz free energy. Without accounting for these effects, researchers cannot accurately predict room-temperature stability, phase transition temperatures ($T_c$), or the spontaneity of chemical reactions. This paper outlines the use of lattice dynamics to bridge the gap between zero-temperature DFT and finite-temperature thermodynamics.
Methods
The evaluation of vibrational free energy typically employs the harmonic approximation. This model assumes that for small atomic displacements, the potential energy surface can be treated as a collection of quantum harmonic oscillators.
The computational workflow follows these steps:
- Structural Relaxation: DFT is used to identify the local potential energy minimum at $T=0$.
- Dynamical Matrix Calculation: The forces resulting from small atomic displacements are determined via the frozen-phonon (direct) method (using supercells) or the linear-response method (using perturbation theory).
- Phonon Density of States $g(\omega)$: The vibrational frequencies ($\omega$) are obtained by solving the eigenvalue problem of the dynamical matrix.
- Thermodynamic Integration: The phonon spectrum is integrated to calculate the vibrational free energy $F_{vib}(T)$, which includes the "zero-point energy" contribution. This term is then added to the static DFT energy to obtain the total free energy.
Software packages commonly used in this pipeline include VASP, PHONOPY, ABINIT, and Quantum Espresso.
Results and Discussion
The lattice-dynamics approach was applied to several benchmark systems to evaluate its predictive power.
Phase Transitions in Oxides and Borohydrides
For rigid structures like Zirconia ($ZrO_2$) and Hafnia ($HfO_2$), the method successfully predicted phase transition temperatures. By calculating the free energy of the high-temperature tetragonal phase ($P4_2/nmc$) relative to the low-temperature monoclinic phase ($P2_1/c$), the crossing point of the free energy curves identifies $T_c$.
As shown in the figure above, the calculated free energy of the high-temperature phases decreases with temperature, crossing the zero-stability line at approximately 1600 K for $ZrO_2$ and 2100 K for $HfO_2$. These results align closely with experimental values of 1500 K and 2200 K, respectively.
However, the model performed less accurately for Potassium borohydride ($KBH_4$). While the experimental transition occurs at 197 K, the harmonic model predicted a transition at 550 K. This discrepancy highlights a limitation of the harmonic approximation: it fails to capture "anharmonic" motions, such as the rotation of $BH_4$ groups, which are prevalent in complex molecular crystals. Furthermore, the study notes that extrapolating high-temperature phase properties from $T=0$ can be risky if the phase is inherently unstable at absolute zero.
Chemical Stability of Zinc Borohydride
The method was also used to assess the thermodynamic stability of $Zn(BH_4)_2$ during synthesis from $NaBH_4$ and $ZnCl_2$. The calculations confirmed that the reaction is energetically permissible and spontaneous up to at least 500 K, providing theoretical support for experimental synthesis routes.
Limitations and Future Outlook
The primary challenge in this field remains anharmonicity. While the harmonic approximation is robust for many rigid crystals, it breaks down in materials with low-energy rotational modes, large-scale diffusion, or near melting points. Future advancements involve more computationally intensive treatments, such as Ab Initio Molecular Dynamics (AIMD) or specialized anharmonic corrections, to better capture these complex dynamical behaviors.
Summary updated