"Lost Atoms" error when using soft core lj potentials

Hey all,

I am trying to perform thermodynamic integration (TI) and for that I want to use the soft core potential. For electrostatic interaction, I would like to use linear scaling and use the soft core for Lennard-Jones interaction. From the documentation, I gather that \alpha_C=1 and \alpha_{LJ}=0.5 (I want Lennard-Jones interaction to have alpha as 0.5 as is in a paper I am following)

The system is 500 SPC\E water and 1 Methanol molecule. I have attatched the input files in this message.
I get the following error:

       300   411.08163     -4385.5682      1229.8492     -511.06892      0.96067272   
       310   409.77718     -4451.5803      1225.9467     -1191.1134      0.96067272   
       320   400.53116     -4477.4895      1198.285      -1762.8949      0.96067272   
       330   399.82864     -4533.9034      1196.1832     -1332.425       0.96067272   
ERROR: Lost atoms: original 1506 current 1503 (src/src/thermo.cpp:488)

With the scaling factor \lambda set to 0.9 (in ffnonbonded.settings), the simulation runs fine.

I don’t really understand why this is not working.


lammps_scaled_soft_test.zip (151.2 KB)

I am confused. In the first part you talk about \alpha and in the second part about \lambda.

Which parameter do you change?

Sorry for the confusion, I change \lambda.
After close inspection, I realised that I was not using the correct pair_coeff as were used in the publication. The electrostatic forces were much too high and that was leading to atoms lost error.

The simulation now runs, however, I get total garbage output for \frac{dU_s}{D\lambda}.

That is a different problem all together unless its related to some wrong way of simulating things.