# Preoptimization with DFTB and calculation of the Hessian
# We will reuse the geometry optimized at the DFTB level as a starting point for
# the DFT geometry optimization. We will also calculate the real Hessian with
# DFTB and use that as the initial Hessian for the Quasi-Newton based
# optimization at the DFT level. DFTB is so fast compared to DFT, that all of
# this is basically instantaneous. Our goal here is really just to reduce the
# number of steps in the DFT geometry optimization. If we save just a single
# step there, the initial DFTB calculation will already have paid for itself ...
AMS_JOBNAME=dftb_preopt $ADFBIN/ams << EOF
# Specify the system geometry: Aspirin
C 0.000000 0.000000 0.000000
C 1.402231 0.000000 0.000000
C 2.091015 1.220378 0.000000
C 1.373539 2.425321 0.004387
C -0.034554 2.451759 0.016301
C -0.711248 1.213529 0.005497
O -0.709522 3.637718 0.019949
C -2.141910 1.166077 -0.004384
O -2.727881 2.161939 -0.690916
C -0.730162 4.530447 1.037168
C -0.066705 4.031914 2.307663
H -0.531323 -0.967191 -0.007490
H 1.959047 -0.952181 -0.004252
H 3.194073 1.231720 -0.005862
H 1.933090 3.376356 -0.002746
O -2.795018 0.309504 0.548870
H -2.174822 2.832497 -1.125018
O -1.263773 5.613383 0.944221
H -0.337334 4.693941 3.161150
H 1.041646 4.053111 2.214199
H -0.405932 3.005321 2.572927
# Do a geometry optimization.
# Also compute the Hessian at the optimized geometry.
# Parallelize the calculation of the displacements used for the numerical
# calculation of the Hessian. Aspirin is much too small for the DFTB engine
# to parallelize efficiently internally, so parallelization at the driver
# level will give better performance.
# Settings for the DFTB engine:
# Geometry optimization with DFT
AMS_JOBNAME=dft_opt $ADFBIN/ams << EOF
# Start from the geometry that is already optimized at the DFTB level.
# (equivalent to loading the system from dftb_preopt.results/ams.rkf)
# Load the DFTB Hessian as the initial Hessian for the
# Quasi-Newton based optimizer.
# Settings for the BAND engine: