<div dir="ltr"><div>Dear all,</div>
<div>i would like to perform a MD simulation of a
large dinuclear ruthenium complex. Im am interested in a ligand exchange
process (Cl- <=> H2O) and would like to do a metadynamics
simulation to obtain the free energy surface for the ligand exchange. As
the system including water molecules in a cubic box of (35 Angstrom)^3
with ca 3100 atoms is quite large and a normal DFT calculation on our
cluster takes about 3 hours per MD step, i was looking into linear
scaling methods. Doing curvy steps with DFT (based on <a href="https://www.cp2k.org/exercises:2015_pitt:ls">this tutorial</a>), the time drops to 1.5 hours but this is still to long. <br></div>
<div><br></div>
<div>I was also trying ALMO-SCF and tried to adapt some of the settings found <a href="https://pc2.uni-paderborn.de/fileadmin/pc2/cp2k_workshop/almo.pdf">here</a>.
Unfortunately i am running into trouble in the first SCF cycle already,
as the "hotelling inversion did not converge". Is there something i
could tweak to get it running? Furthermore the informations about
ALMO-AIMD seem a little bit spare to me, so i guess i am on thin ice,
especially as the original paper recommends Langevin MD. According to
the tutorial slides i was trying to use ALMO-SCF without early stopping
and a larger cutoff radius to be able to use normal NPT_I ensemble, but i
am not sure if i did it correctly?</div>
<div><br>
</div>
<div>As an alternative i was also trying XTB and XTB with curvy steps but i assume the energies will not be that good.<br>
</div>
<div><br></div>
<div>I am grateful for any suggestion and insight. :)<br><br>Best wishes and thanks in advance<br>Phillip<br></div>
</div>