Langevin thermostat for Gaussian basis AIMD equilibrates higher than target temperature


Click here for full thread
Clicked A Few Times
I am trying to run a Gaussian basis AIMD NVT simulation at 300 K with the Langevin thermostat. When I don't set the target temperature (i.e., it is set to the default, 298.15 K), the average temperature is ~320 K (for up to 200 fs). I've also tried changing the friction parameter, but in my case it does not change the average temperature more than a couple K. Does this thermostat just take a really long time to equilibrate? Or is the temperature in the output file different from the target temperature?

I've also started a calculation with targ_temp set to 270 K and it's average temperature is also ~20 K higher at 290 K. I've also used a variety of XC functionals (B3LYP, BLYP, PBE0, PBE96) and this doesn't seem to be an issue with choice of functional. Below is a sample of my input file. Any help is appreciated!

basis
  * library 6-31G*
end

dft
 xc xpbe96 cpbe96
end

qmd
 nstep_nucl  300000
 targ_temp 270.0
 thermostat  langevin
 rand_seed   12345
 print_xyz    100
end
task dft qmd