Modified SIESTA for DeepH-DFPT

Developed several in-house versions of the SIESTA code to enable deep-learning density functional perturbation theory (DeepH-DFPT) calculations. These modifications support:

• Pulay correction for stress and forces
• Non-local correction terms
• Integration with the DeepH framework

Associated with the paper: Deep-learning density functional perturbation theory, Phys. Rev. Lett. 132, 096401 (2024).