Hi all,
I encountered a problem when trying to run QM/MM minimisation with DFTB and
using *.skf files from the halorg-0-1 set. The error that occurs at the
very first step is:
QMMM SCC-DFTB: ***************************************************
QMMM SCC-DFTB: ERROR ON EWEVGE (Eigenvalue solver).
QMMM SCC-DFTB: ewevge: ier = 85 inner_scf_count= 1
QMMM SCC-DFTB: ***************************************************
I know this error has been discussed before, but I haven't seen a 'general'
solution.
I've been spending quite some time trying to narrow down where the error
comes from - and it appears to be related to the *.skf files (although
there may also be problems with how sqm reads them).
The amber-mailing list may therefore not be the appropriate forum, but I'm
hoping someone here may have some ideas - and at least it will be helpful
for others encountering the same problems.
Using single point calculations in the latest version of sqm (version 14,
as included in AmberTools15 and also in AmberTools14), I've run the
following tests:
1) When using the halorg-0-1 set, CH3F runs fine. (I used
the /AmberTools/test/sqm/ch3f/ch3f.pm3.sp.in file, with PM3 replaced by
DFTB)
2) When using the halorg-0-1 set with my QM region that contains a -CH2F
group (and also O, N, S atoms), I get the Eigenvalue solver error
3) When using the halorg-0-1 set with my QM region where the -CH2F group is
replaced by -CH3, I ALSO get the Eigenvalue solver error
4) When using the default mio-0-1 set (so no halogens) with my QM region
where the -CH2F group is replaced by -CH3, I DON'T get the Eigenvalue
solver error
At this point, it seems that the problem is with the halorg-0-1 *.skf files
that include O, N, or S.
5) When using the halorg-0-1 set and replacing S with O in my QM region
(and with the -CH2F group is replaced by -CH3), I ALSO get the Eigenvalue
solver error. (And still no error with the mio-0-1 set.)
The problem is now appears to be narrowed down to the halorg-0-1 *.skf
files that include O & N (although those with S may also have a problem).
BUT:
6) When running with an input where F is replaced with H in the
ch3f.dftb.sp.in file, so creating ch4.dftb.sp.in (again no problem with the
mio-0-1 set), I get:
I get:
Program received signal SIGSEGV: Segmentation fault - invalid memory
reference.
Backtrace for this error:
#0 0x2AAAAACE6297
#1 0x2AAAAACE689E
#2 0x2AAAAB70869F
#3 0x2AAAAADA5D9D
#4 0x2AAAAADA1BD3
#5 0x2AAAAADA38B6
#6 0x49BE34 in gettab_
#7 0x4A22FF in qm2_dftb_load_params_
#8 0x45B380 in qm2_load_params_and_allocate_
#9 0x40F886 in sqm_energy_
#10 0x4117D7 in MAIN__ at sqm.F90:0
Segmentation fault
(the output file is empty)
This appears to be a different issue, that I am rather lost on.
A further test shows that there is also a different problem with running
sqm with the halorg-0-1 *.skf files including O & N:
7) When running the $AMBERHOME/AmberTools/test/sqm/nma/nma.dftb.sp.in test
with the halorg-0-1 set (no error with the mio-0-1 set), I get:
*** glibc detected *** sqm: free(): invalid pointer: 0x00002aaaabb47010 ***
======= Backtrace: =========
/lib64/libc.so.6(+0x75e66)[0x2aaaab74be66]
sqm[0x49d185]
sqm[0x40b079]
sqm[0x40b0cd]
/lib64/libc.so.6(__libc_start_main+0xfd)[0x2aaaab6f4d5d]
sqm[0x401e09]
======= Memory map: ========
...
...
ffffffffff600000-ffffffffff601000 r-xp 00000000 00:00 0
[vsyscall]
Aborted
At this point, I am rather lost, apart from knowing that the halorg-0-1 set
does not work in sqm (leads to the Eigenvalue solver problem) when having
O, N or S in a molecule, even though it does work with CH3F.
In addition, I get "Program received signal SIGSEGV: Segmentation fault -
invalid memory reference." errors when using the halorg-0-1 set in
molecules containing only C & H or S & F, tested with c2h2.dftb.sp.in and
h2s.dftb.sp.in, respectively)
Sorry for the long email.
Any insight would be highly appreciated!
--Marc
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Feb 04 2016 - 05:00:03 PST
