Trouble with argon

[jschindler89]

Hi,

I am playing around with LJ argon system at 10 K.

Using the files attached, I see that the max frequency for the DOS is about 4 THz which is twice as large as what we would expect for crystal LJ argon at low temperatures. Would you know how I can fix this?
Ar.zip

Jack

[brucefan1983]

Could you also provide the model.xyz file such that I can reproduce the results?

[jschindler89]

Here it is! Thank you.
model.xyz.zip

[jschindler89]

Also, the thermal conductivity seems to be ~ 2 times higher than literature. Thank you for your help. Jack

[brucefan1983]

Because you have set the mass to 12 in model.xyz. It should be 40. You can remove the mass item. GPUMD has a default mass for any element taken from a CRC handbook.

…

On Wed, Jun 5, 2024 at 7:27 PM jschindler89 ***@***.***> wrote: Also, the thermal conductivity seems to be ~ 2 times higher than literature. Thank you for your help. Jack — Reply to this email directly, view it on GitHub <#638 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AF546OLUALIJH6KWTOBDASDZF3YZJAVCNFSM6AAAAABIYNX6OGVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM4TMNZWHE4DS> . You are receiving this because you commented.Message ID: ***@***.***>

[jschindler89]

Thank you so much! Jack.

[jschindler89]

Hi Bruce,

Is there a way to shift up the LJ potential at a designated cutoff?

[brucefan1983]

You can modify the code by yourself :-) A couple of lines.

…

On Thu, Jun 6, 2024 at 5:15 AM jschindler89 ***@***.***> wrote: Hi Bruce, Is there a way to shift up the LJ potential at a designated cutoff? — Reply to this email directly, view it on GitHub <#638 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AF546OPXDO5GNVZAVCFTS6LZF55VZAVCNFSM6AAAAABIYNX6OGVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM4TMOBSGU3TQ> . You are receiving this because you commented.Message ID: ***@***.***>

[jschindler89]

Sounds good. I will try.

To know where to start, I need to change lj.cu in force folder, right?

[brucefan1983]

Yes, to modify this function (perhaps you need to pass the cutoff in and use it):

// get U_ij and (d U_ij / d r_ij) / r_ij (the LJ potential)
static __device__ void
find_p2_and_f2(double s6e4, double s12e4, double d12sq, double& p2, double& f2)
{
  double d12inv2 = 1.0 / d12sq;
  double d12inv6 = d12inv2 * d12inv2 * d12inv2;
  f2 = 6.0 * (s6e4 * d12inv6 - s12e4 * 2.0 * d12inv6 * d12inv6) * d12inv2;
  p2 = s12e4 * d12inv6 * d12inv6 - s6e4 * d12inv6;
}

[jschindler89]

Thank you!