On Tue, Feb 11, 2025, KLEIN Jordan wrote:
>
>I use a bonded model for the uranyl.
>
>The UO₂ is simulated in an ionic liquid, but I encounter the same problem in the organic phase and in water.
>There are periodic boundary conditions.
OK: makes sense, and thanks for the info.
>
>I haven't tried a 12-6 potential for uranyl because I only have a 12-6-4
>potential for this ion. I will try to simulate a "uranyl" with a 12-6
>potential and see if I encounter the same problem.
I recommend the following instead: prepare a short simulation (say 20 steps,
with ntpr=1.) Run that with pmemd, and then again with pmemd.cuda_DPFP,
making no other changes. Compare the outputs, particularly at step 0.
Do this for both 12-6 and 12-6-4; see if you can spot what is going on.
One possibility: CPU and GPU agree for 12-6, but not for 12-6-4. If so,
that would narrow down the problem a lot.
Also of interest: find out how quickly the U--O bond length changes from
1.78 to 1.72 in the GPU run. Is the length change pretty quick (a few
hundred steps) or is it some long term transition?
Debugging is generally accomplished by (a) changing only one thing (the
exectuable, or the presence/absence of 12-6-4) at at time; (b) running short
simulations and carefully studying the outputs.
...good luck...dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Feb 11 2025 - 07:30:02 PST
