Problematic DFTB geometry optimization

[Maxime Van den Bossche]

Hello!

I've been trying out the DFTB module in CP2K on a simple 
bulk TiO2 (rutile) structure. Strangely enough, it appears
impossible for the different optimizers to successfully
relax this structure. With BFGS, the atoms keep oscillating
along some kind of vibrational mode. With CG and LBFGS,
no such oscillations occur, but the forces cannot be brought
below circa 0.015 Hartree/Bohr.

I've attached a zip file with the input and output files for
the different optimizers. I've generated the Slater-Koster files
myself, and they work nicely with the DFTB+ code (without 
any geometry optimization issues). I'm using a development 
version of CP2K (6.0, git:8b033c3).

Could you help me understand what is going on? 

Best,
Maxime
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20180129/33baff1f/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: files.zip
Type: application/zip
Size: 740304 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20180129/33baff1f/attachment.zip>