LAMMPS-conversion of DPD temperature

Hi,
I am having an issue in converting the DPD temperature units into real temperature units as mentioned in the following two styles in order anneal my system-

pair_style dpd T cutoff seed
pair_style dpd 1.0 2.5 34387

Q1) This is an example of writing dpd unit of temperature (the bold one) for lj system where, T* = 1.0 means that T* = TkB/Epsilon = 1.0. In the lammps documentation, it was mentioned that here 1.0 is equivalent to 300K. Now, if I want to use increased temperature, for example 400K, what will be the equivalent non-dimensional number? I am really confused about this conversion. It will be much appreciated if anyone can please help me with an explanation.

Q2) In lammps if I want to mimic the experimental annealing ( e.g. raising temperature to reach equilibrium and then bringing the temperature back to room temperature) can just follow the following steps- a) run the simulation at room temp. b) take the last trajectory of atoms from step (a) and run the simulation at higher temp. and then c) take the last trajectory from step (b) to repeat the step (a) at room temperature. I am wondering if there might be any computational artifact generated due to such change of the temperature or it should mimic the experimental annealing system?

Q3) For the question (Q2), if I use it for systems having pair_coeff (= interation parameter obtained from Chi parameter),how this change of whole system temperature can potentially affect the interaction parameter. Or will it just change the total system temperature leaving the mutual interaction unaffected ?

Thanks a lot in advance !

Sincerely,

Tuhin

I don’t know why 1.0 in LJ units would necessarily be 300K. Where

does it say that in the documentation?

Regardless, if you want to increase the temperature

by 33% (300->400), then using 1.0->1.33 will do

that,

Steve

Hi Steve,
What I understood from your reply is that increasing percentage of the unit value will increase the temperature with equal percentage in Kelvin.

Now, for lammps, the normalized temperature , T* = T kB/Epsilon. Now, unless the value of this Epsilon is always system dependent, is it possible to know the set value of Epsilon for lammps DPD so that I can translate the dimensionless number into real temperature value in Kelvin?

And, this Epsilon is system dependent, it will also be much appreciated if you, please, shed some light on the possible mathematical conversion.

Thanks you very much for the help.

Regards,
Tuhin

Hi Steve,
What I understood from your reply is that increasing percentage of the
unit value will increase the temperature with equal percentage in Kelvin.

Now, for lammps, the normalized temperature , T* = T kB/Epsilon. Now,
unless the value of this Epsilon is always system dependent, is it possible
to know the set value of Epsilon for lammps DPD so that I can translate
the dimensionless number into real temperature value in Kelvin?

​you are not making much sense here. if you know epsilon (e.g. in kcal or
in eV) and want the temperature in kelvin, why on earth do you do your
simulation in reduced units, particularly if you are not familiar with how
to convert between those? just set units to "real" (or metal, respectively)
and be done with it.

And, this Epsilon is system dependent, it will also be much appreciated if
you, please, shed some light on the possible mathematical conversion.

​please look it up in your MD/Physics text book. if you want to work in
reduced units, you should understand properly how they work. a good
MD/Physics book should explain them.

axel.​

Hi Axel,
That’s a bit ‘Quinine’ way of swallowing the correct answer :slight_smile:
Fortunately, I am already a bit familiar with your style, and most importantly, it really helped !
Thank you very much.

Tuhin