Dear everyone,
Apologies for a confusion with some really basic ideas, however, I was wondering if someone could verify that what I am stating is correct:
Reduced temperature in LAMMPS is k_B/epsilon, and I am using the LJ potential, in which epsilon=1. Explicitly, I am using the langevin thermostat to set T=1. Does this mean, essentially that if I set epsilon to be 2.5 kJ/mol, this will give me T=300K (just the basic LJ units conversion)? Assuming this is correct, what happens if I set epsilon to 2: does this become 5 kJ/mol and T=601K in real units? Will changing epsilon also affect the time, as it is sigma sqrt(m/epsilon)? My thought is to perform simulations at varying epsilon of the lj potential, but such that the temperature stays the same for different epsilons. I am hopelessly confusing myself with these conversions
Thank you for any input!
With best wishes,
A

The epsilon you set in pair_coeff is not the unit epsilon in which energies are expressed. It is a bit confusing, but both the epsilon you set in pair_coeff and the value for k_B T you set in fix langevin are expressed in multiples of some arbitrary energy scale epsilon, which I will call epsilon*.
Therefore, if you set epsilon to 2 in pair_coeff but keep the temperature in fix langevin equal to 1, you have effectively made your interaction strength twice as large as the thermal energy. If you thought of your unit epsilon* as 2.5 kJ/mol then your interection strength becomes 5 kJ/mol but the temperature remains T = 1 epsilon*/k_B = 300 K.
For your case, it makes perfect sense to put fix langevin to 1, so that an energy of 1 epsilon* corresponds to k_B T. Then you can just put the epsilon in pair_coeff equal to the multiple of k_B T you are interested in.
Hope this helps.
Thank you a lot, Stefan. It is indeed confusing, but I think i understand it now: So when I set the epsilon as 1 in the input file, I can think of it as any value, say 2 kJ/mol, and the same for thermostatting the temperature, if I set T to 1 in the langevin thermostat, I can just think of it as 300K.
With best wishes,
A

Yes, indeed. You can arbitrarily fix one of them to 1 and express all the others in multiples of that. That also keeps interpreting the results the easiest.
At first it seems weird, but really epsilon* is just as arbitrary a reference for values of energy as the Joule or kcal/mole (unless perhaps if you are studying water).