statpt - geoired.f: lowest eigenvalue is too small

[jmkormanik]

Hello everyone,

I am working with short peptides (~400-500 atoms). When I am doing a minimization using jobex in TurboMole 7.7, usually in ired, sometimes I get this error during statpt: "geoired.f: lowest eigenvalue is too small". I was unable to find any information or what this means.

I am wondering why this is and if there is a way to fix it. Things we have tried and somewhat worked:
- small displacements of atoms before the start of the optimization (within 0.1 A)
- $redund_inp keyword in control set to "metric 2" (this works in majority of cases, but also occasionally fails)
- starting optimization in cartesian coordinates (without ired) for a few steps (tested 25) and then continuing in ired.

Despite all of these workarounds, which mostly work, statpt still occasionally fails in some cases (~1%) and I'm not sure whether this can be fixed or how to fix it. Can someone please elaborate what this error means and what I can do with it?

I'm attaching the control and coord files, as well as statpt output. cood.non_displaced.inp are the original coordinates of the structure, which was pre-optimized using xTB-GFN2 method. Upon visual inspection, I didn't notice any glaring issues with the structure itself. The structure is supposed to have certain restraints (to keep certain atoms together; a tetrahedron created by atoms 52, 118, 400 and 447), which might play some role, but in other similar calculations this didn't cause issues.

Thank you for your time and responses,

jmkormanik

[jmkormanik]

Hello, can anyone help, do you have any insights to offer? Thank you very much.

[uwe]

Hi,

easiest solution is to switch to cartesian coordinates.

Code: [Select]

$optimize
 internal   off
 redundant  off
 cartesian  on
If the internal redundant coordinates show very low eigenvalues (or if they are so low that statpt refuses to use them), they are most likely not more efficient than simply running the optimization in cartesian ones.

[jmkormanik]

Hi,

thank you for your response. Cartesian coordinates are usually, unfortunately, unreasonable for a system of this size, ~400 atoms is a lot and convergence easily takes 600+ steps in jobex even in internal redundant coordinates (these systems are sometimes quite far from equilibrium). Do you think that the optimization of Cartesian and internal redundant would then take roughly the same amount of time?

However, if there is no better solution, then this is what we'll have to do - either use Cartesian coordinates only or do several (~50) steps in Cartesian and then try to switch to ired and see if statpt catches on. If not, continue in Cartesian.

Thank you for the response! I was mostly wondering if there's anything else that we missed.

Have a nice day,

jmkormanik