problem when using PIGLET thermostat

[Xiaowei Zhang]

Hi all,

       

         I am new  to i-pi. 

 

        There was a problem with the PIGLET thermostat. It seems that the potential energy increases without an upper limit . At the same time, the centroid coordinates are far from the equilibrium positions. I have also tried more accurate functionals like SCAN, but things haven't changed.

         However, when I use the GLE thermostat, the potential energy can quickly reach equilibrium and the centroid coordinates don't  deviate too much from the equilibrium position. Does this indicate that PIGLET is not suitable for some systems?

  

         My input and output files are attached. Looking forward to your help!

        Thank you!

Best regards,

Xiaowei

------------------------------

Ph.D. Student Xiao-Wei Zhang

ICQM, School of Physics, Peking University

Email: taic...@gmail.com

------------------------------

[Michele Ceriotti]

Dear Xiaowey, 

   (1) I'd need to know the parameters with which you generated the GLE matrices. In general it's a good idea to paste into the input.xyz the comment section that is generated by GLE4MD

   (2) I don't know how you're restarting your simulations - if the simulation ends gracefully (wall-clock time limit elapsed, touch EXIT or SIGTRM) you can just use the RESTART file as input (i-pi RESTART &> log). that way the time should increment smoothly, unlike in your output files

   (3) The two simulations cannot be compared - the GLE thermostat you use is a classical thermo so the simulation is classical. PIGLET is quantized so there may be very large differences in behavior

   (4) Seems the quantum simulations are melting, hence the steady increase in enthalpy. Good news, you stumbled upon a system for which quantum effects matter!

So bottom line: no problem with i-PI, only minor aestetic problems with your setup. What you're seeing is the physics of this system as described by DFT without (GLE) and with (PIGLET) quantum effects. 

Michele

[Xiaowei Zhang]

Dear Michele,

        

         Thanks for your reply.

         (1)I have attached the input parameters. By the way, the maximum vibrational frequency of my system is 89.6 Thz (2988.7 cm-1).

         

         (2) I restart the simulation by reading chk file. The RESTART file cannot be written in my cluster.

          So if the input parameters are reasonable, do you think that my system is actually liquid?

        

Xiaowei

[Michele Ceriotti]

Hello.

  (1)   Input parameters are OK.

  (2)  To have i-PI print a RESTART just set total_time in the input.xml to be slightly lower than the wall-clock time of the job

  (3) Based on what I see, looks like you're doing no error in setting up the calculation, so either there is substantial zero-point energy leakage, or the quantum system is melting. Try to run with 16 beads and piglet and see if the potential energy goes up, down or stays where it is.

Michele