[DFTB-Plus-User] Running the code successfully and several problems

[jsxz]

Dear Bálint,
Thank you very much for your  help patiently!
According to your guide, I have compiled the DFTB_MPI successfully, and runed the code with my system successfully.
The Computing speed for the parallel is very fast , but there are two problems in the running:
1) When I set the Eigensolver = DivideAndConquer {} or Standard {}, the program can be runed successfully.
But it is RelativelyRobust{}, the program refuses to run.  There are 288 carbon atoms in my system.
It is not the reason for the  Ewald parameter, because the  Eigensolver is DivideAndConquer {} or Standard {}, and EwaldParameter = 0.0,
the program can be runed successfully. The out.log file is as follows.

================================================================================
==   DFTB+ (Density Functional Tight Binding and much more)
==                          Unofficial release (r4465)
==                            (ParserVersion = 4)
================================================================================

********************************************************************************
** Parsing and initializing
********************************************************************************

Interpreting input file 'dftb_in.hsd'
--------------------------------------------------------------------------------
Reading SK-files:
  C-C.skf
Done.
Processed input in HSD format written to 'dftb_pin.hsd'

Starting initialization...
--------------------------------------------------------------------------------
PGRID:ORBITAL: 1 x 4 x 6
PGRID:ALLPROC: 4 x 6
PGRID:ATOM: 4 x 6
RECLATPOINTS: 115
WARNING!
-> Automatic Ewald parameter quite small, possibly resulting in
-> very long neighbor lists! You should consider to set the
-> Ewald parameter by hand.
Mode:                        MD without scaling of velocities
                             (a.k.a. NVE ensemble)
Self consistent charges:     Yes
SCC-tolerance:                 0.100000E-04
Max. scc iterations:                    100
Ewald alpha parameter:         0.279462E-01
Spin polarisation:           No
Nr. of up electrons:           576.000000
Nr. of down electrons:         576.000000
Periodic boundaries:         Yes
Diagonalizer:                Relatively robust (version 1)
Maximal SCC-cycles:                     100
Electronic temperature:        0.316682E-02
Time step:                     0.413414E+02
Random seed:                      102321536
Initial charges:             Set automatically (system chrg:   0.000E+00)
Included shells:              C:  s, p
K-points and weights:           1:  0.000000  0.000000  0.000000     0.200000
                                2:  0.000000  0.000000  0.200000     0.400000
                                3:  0.000000  0.000000  0.400000     0.400000
Using Lennard-Jones dispersion corrections
Extra options:

********************************************************************************
** Geometry step: 0
********************************************************************************

MAXNEIGHBORS: 2561
  iSCC Total electronic   Diff electronic      SCC error   
ERROR!
-> Not implemented

2) The bigger for the EwaldParameter, the longer for the  computer time.
      EwaldParameter = 0.0,   time=3 minutes;
      EwaldParameter = 0.101, time=15 minutes;
      EwaldParameter = 0.15, time=28 minutes;
      EwaldParameter = 0.2, time=50 minutes;
      EwaldParameter = 0.3, time=120 minutes;
      EwaldParameter = 0.4, time=285 minutes;
      EwaldParameter = 0.45, time=374 minutes;
   I want to know whether I should set the EwaldParameter smaller (such as 0.0) as soon as possible for any system?

3) I have tested the  RowBlockSize and ColBlockSize.
The default values  for dftb_pin.hsd is  RowBlockSize = 32,  ColBlockSize = 32 (not 64).
I find the bigger for the RowBlockSize and ColBlockSize, the longer for the  computer time.
I want to know whether I should set the RowBlockSize and ColBlockSize  smaller as soon as possible for any system?  Or only set them the default values.

4) How to set the parameter for group in the Parallel, only the default values?
 I set the Group=1 and 2 when I use the kpoints, there are no different phenomenon for the results, including the computer time.

Many thanks for your help!
Best regards
Xiaobao Zhang
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.zfn.uni-bremen.de/pipermail/dftb-plus-user/attachments/20141006/efd50b5b/attachment.html>