ML LHEAT: Difference between revisions
No edit summary |
No edit summary |
||
| Line 23: | Line 23: | ||
<math> | <math> | ||
m_{i} \frac{d \mathbf{v}_{i}}{dt} = - \sum\limits_{j=1}{N_{a}} \nabla_{i} U_{j} | m_{i} \frac{d \mathbf{v}_{i}}{dt} = - \sum\limits_{j=1}{N_{a}} \nabla_{i} U_{j} | ||
</math> | |||
the heat flux can be simplified to | |||
<math> | |||
\mathbf{q}(t) = \sum\limits_{i=1}^{N_{a}} \mathbf{v}_{i} E_{i} - \sum\limits_{i=1}^{N_{a}} \sum\limits_{j=1}^{N_{a}} \mathbf{r}_{i} \left( \mathbf{v}_{i} \cdot \nabla_{i} U_{j} \right) + um\limits_{i=1}^{N_{a}} \sum\limits_{j=1}^{N_{a}} \mathbf{r}_{i} \left( \mathbf{v}_{j} \cdot \nabla_{j} U_{i} \right) = \sum\limits_{i=1}^{N_{a}} \mathbf{v}_{i} E_{i} + \sum\limits_{i=1}^{N_{a}} \sum\limits_{j=1}^{N_{a}} \left( \mathbf{r}_{i} - \mathbf{r}_{j} \right) \left( \mathbf{v}_{j} \cdot \nabla_{j} U_{i} \right). | |||
</math> | </math> | ||
Revision as of 09:48, 12 June 2021
ML_FF_LHEAT_MB = [logical]
Default: ML_FF_LHEAT_MB = .FALSE.
Description: This flag specifies whether the heat flux is calculated or not in the machine learning force field method.
The heat flux within machine learning force fields can is decomposed into atomic contributions written as
[math]\displaystyle{ \mathbf{q}(t) = \sum\limits_{i=1}^{N_{a}} \frac{d}{dt} \left( \mathbf{r}_{i} E_{i} \right), }[/math]
[math]\displaystyle{ E_{i}=\frac{m_{i} \left|\mathbf{v}_{i} \right|^{2}}{2} + U_{i} }[/math]
where [math]\displaystyle{ \mathbf{r}_{i} }[/math], [math]\displaystyle{ \mathbf{v}_{i} }[/math] and [math]\displaystyle{ E_{i} }[/math] denote the position vector, velocity and energy of atom [math]\displaystyle{ i }[/math], respectively. The number of atoms in the system is denoted by [math]\displaystyle{ N_{a} }[/math]. The heat flux can be further rewritten as
[math]\displaystyle{ \mathbf{q}(t) = \sum\limits_{i=1}^{N_{a}} \mathbf{v}_{i} E_{i} + \sum\limits_{i=1}^{N_{a}} \mathbf{r}_{i} \left( m_{i} \mathbf{v}_{i} \cdot \frac{d\mathbf{v}_{i}}{dt} + \sum\limits_{j=1}^{N_{a}} \mathbf{v}_{j} \cdot \nabla_{j} U_{i} \right). }[/math]
Using the equation of motions
[math]\displaystyle{ m_{i} \frac{d \mathbf{v}_{i}}{dt} = - \sum\limits_{j=1}{N_{a}} \nabla_{i} U_{j} }[/math]
the heat flux can be simplified to
[math]\displaystyle{ \mathbf{q}(t) = \sum\limits_{i=1}^{N_{a}} \mathbf{v}_{i} E_{i} - \sum\limits_{i=1}^{N_{a}} \sum\limits_{j=1}^{N_{a}} \mathbf{r}_{i} \left( \mathbf{v}_{i} \cdot \nabla_{i} U_{j} \right) + um\limits_{i=1}^{N_{a}} \sum\limits_{j=1}^{N_{a}} \mathbf{r}_{i} \left( \mathbf{v}_{j} \cdot \nabla_{j} U_{i} \right) = \sum\limits_{i=1}^{N_{a}} \mathbf{v}_{i} E_{i} + \sum\limits_{i=1}^{N_{a}} \sum\limits_{j=1}^{N_{a}} \left( \mathbf{r}_{i} - \mathbf{r}_{j} \right) \left( \mathbf{v}_{j} \cdot \nabla_{j} U_{i} \right). }[/math]
Finally (in a post-processing step), the thermal conductivity at temperature [math]\displaystyle{ T }[/math] in the Green-Kubo formalism can be calculated from the correlation of the heat flux [math]\displaystyle{ \mathbf{q} }[/math] as
[math]\displaystyle{ \kappa = \frac{1}{3Vk_{b}T^{2}} \int\limits_{0}^{\infty} \langle \mathbf{q}(t) \cdot \mathbf{q}(0) \rangle dt, }[/math]
where [math]\displaystyle{ V }[/math] and [math]\displaystyle{ k_{b} }[/math] denotes the volume of the system and the Boltzmann constant, respectively.
The heat flux is written to the file ML_HEAT.