Calculation of Allen-Feldman thermal conductivity


I want to calculate the contribution of diffusons to thermal conductivity using GULP, but I have two questions that I would like you to help with.

  1. As shown in the title of example 51 (GULP 5.2), the structure of amorphous silicon has been relaxed using the GULP Stillinger-Weber library. So I wonder if this relaxation refers to structural optimization at 0 K or molecular-dynamics relaxation at a specific temperature (e.g. 300 K) because example 51 was conducted at 300 K.

  2. In the command omega_af, the bulk modulus B in s/km* *2. But I am very confused about how to convert GPa to s/km**2.

Thanks very much!

Best regards,


  1. This is a lattice dynamics calculation and so relaxed means optimised (i.e. 0 K). MD at 300 K would be “equilibrated” rather than relaxed since there is an ensemble of structures rather than a single one.
  2. I think your confusion is that you think B is the bulk modulus, which it isn’t - it’s a parameter that controls the lifetime of the phonons in the acoustic regime.

Thanks very much for your reply. I got it.
Best regards

I found that the maximum number of steps is 1000 during the optimization. I wonder if we can increase it.

You can certainly increase it in one of 2 ways:

  1. Specify you want more cycles from the input file (see help text for command)
  2. Edit the source code
    That said, you may want to look at why your system is needing more than 1000 cycles to optimise and check that there isn’t an issue, or look at things like RFO as a better end game optimiser.

Thanks very much.

Hi Julian,
How can I find these values for Mg3N2 ; broaden scale and omega_af,
thanks a lot,

I’m afraid that’s not really a GULP question since it relates to the science of your project, rather than the code. The best thing is to study the literature on thermal conductivity to get an understanding of how to best choose these values for particular systems or to collaborate with someone who is an expert in this field who can advise.

1 Like