I've got a very strange geometry oprimzation for EtSnCl3 when using LAND2DZdp ECP and valence shell for Sn and Cl and 6-31++G** for Et radical. It was my first experience in using ECP, so I expect that problem may be in wrong input data, but I cannot find the flaw, so I ask your help.
Namely, I picked up ECP and valence shell for Sn and Cl and 6-31 basis for C and H at https://bse.pnl.gov/bse/portal , put them into files *.ecp and *.bas and include them in the input via #libdir option. During geometry oprimization, hydrogens and methyl carbon come too close to Sn-C carbon, the same for this carbon at tin and tin itself, while chlorines go away from tin atom, and this is accompanied by tremendous energy decrease, almost 6000 kcal/mol. Finally, geoopt run crashs with "GRADIENT OUT OF RANGE".
This problem remains when I remove +* functions from light elements and when I use LAND2DZdp basis (I mean, the basis under this name in the basis set database) for them.
At the same time, SnCl4 is optimized without crash with LANL2DZdp ECP and valence shell on heavy atoms and has reasonable geometry and vibrations (I do not konow its exact IR spectrum, but tin chlorides are known to be transparent in 400-4000 cm-1 region and to have their bands below 400 cm-1, and vib. freqs are indeed below 400 cm-1).
Attachment to the post contains zip-archive with input file and used ECP and basis set.
Thank you in advance.
|This message contains the 4 kb attachment|
[ strange_SnEtCl3.zip ]