`fmax` setting in `PhononCalc` and `QHACalc`

[Andrew-S-Rosen]

Currently, the fmax keyword argument in PhononCalc and QHACalc are set to fmax=0.1. This seems a bit dangerous to me, as it's almost certainly not close enough to the PES minimum and will lead to spurious vibrational modes. Perhaps we should consider reducing this to a much tighter threshold?

[shyuep]

Agree. Let’s set it to 0.05. Shyue Ping

…

On Mon, Dec 22, 2025 at 14:42 Andrew S. Rosen ***@***.***> wrote: *Andrew-S-Rosen* created an issue (materialyzeai/matcalc#127) <#127> Currently, the fmax keyword argument in PhononCalc and QHACalc are set to fmax=0.1. This seems a bit dangerous to me, as it's almost certainly not close enough to the PES minimum and will lead to spurious vibrational modes. Perhaps we should consider reducing this to a much tighter threshold? — Reply to this email directly, view it on GitHub <#127>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAEM4U7FAVPR7U2I2ZABSET4DBXWPAVCNFSM6AAAAACPZMMOWCVHI2DSMVQWIX3LMV43ASLTON2WKOZTG42TKNBVGQ4TENI> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[Andrew-S-Rosen]

Hmm, 0.05 eV/A might still be quite high for a phonon calculation. I tend to see 1e-3 as a reasonable default, but of course the flipside is that this might not be practical/trivial to achieve in high throughput.

Let's split the difference and call it 0.01?

[shyuep]

Sorry I meant 0.05. I think that is the usual safe default? Shyue Ping

…

On Mon, Dec 22, 2025 at 15:17 Andrew S. Rosen ***@***.***> wrote: *Andrew-S-Rosen* left a comment (materialyzeai/matcalc#127) <#127 (comment)> Hmm, 0.05 eV/A might still be quite high for a phonon calculation. I tend to see 1e-3 as a reasonable default, but of course the flipside is that this might not be practical/trivial to achieve in high throughput. Let's split the difference and call it 0.01? — Reply to this email directly, view it on GitHub <#127 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAEM4UZMW66TM55VGAH6B5L4DB3W5AVCNFSM6AAAAACPZMMOWCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTMOBUGUYDONRZGM> . You are receiving this because you commented.Message ID: ***@***.***>

[Andrew-S-Rosen]

Sorry I meant 0.05. I think that is the usual safe default?

Hmm. I have certainly seen people refer to 0.05 eV/A for a standard geometry optimization but not in the context of relaxing a structure for a subsequent phonon calculation. That said, I tried to find some definitive sources to back up my claim but came up empty handed.

I may be biased by often studying floppy systems where a max|F| of 0.05 eV/A will almost always yield imaginary modes in a phonon calculation because the PES is relatively shallow. In any case, I'll adjust it to 0.05 eV/A in #129 unless I find evidence that this would still be too high for general use.