[Freeon-users] Core Hamiltonian guess fails

[Jose R. Valverde]

Dear friends,

        I had a look at the GIT repository and found the 'examples' directory, 
with plenty of examples to find my way around. Now it is evident how to use 
MPI etc.

        Obviously, the first thing I tried was to run FreeON on all the
examples and check the results. In doing so I have observed that FreeON
fails when the "Guess: core" option is used.

        The output produced is:

Invoke                           :: invoking QCTC
Invoke : Clump 1                 :: /u/jr/contrib/FreeON/bin/QCTC
HeH+_1935 GuessEqCore 0 1 1 0 1 1 /tmp/FreeON-scratch-HeH+_1935-OOSCWd/
Unknown error: </u/jr/contrib/FreeON/bin/QCTC HeH+_1935 GuessEqCore 0 1
1 0 1 1 /tmp/FreeON-scratch-HeH+_1935-OOSCWd/>
TimeStamp                        :: <<Entering QCTC 10/25/11 @ 10:20:07>>
RhoEcon                          :: called with NLink = 0
illegal value

        I understand the first numerical argument (NLink) specifies the
cycle number and is used to identify the Density matrix file to be used
as modifying it to a 1 gives

QCTC HeH+_1935 GuessEqCore 1 1 1 0 1 1 /tmp/FreeON-scratch-HeH+_1935-OOSCWd/
TimeStamp                        :: <<Entering QCTC 10/25/11 @ 09:49:49>>
Get_BCSR could not find 
/tmp/FreeON-scratch-HeH+_1935-OOSCWd/HeH+_1935_Geom#1_Base#1_Cycl#1_Clone#1.D

        Looking in the scratch directory I can find that file but with a
Cycle number of 0 instead of 1.

        This only happens when the Action argument is GuessEqCore, as NLink
is specified as 0 in other calculations with no adverse effects.

        Looking in RhoEcon in Density.f90 I see a check for Nlink < 1. But
changing that to accept an initial cycle number of 0 makes the program fail
as well, even modifying the loops that use NLink, this time with 

TimeStamp                        :: <<Entering QCTC 10/25/11 @ 10:19:34>>
 the number of values to be sorted is not positive.           0

        At this point, I have no good references to decide what is the actual
problem: it might be that changes in Density.f90:RhoEcon logic led to the
Core guess not working, or it might be (which sounds more likely) that the
initial cycle number should start at 1 for the Core Hamiltonian guess (maybe
a legacy from older versions using one-offset for calculations) and the problem
then should be fixed in the main FreeON driver executable instead.

        Looking at the repository, I see that HeH+.inp -for example- was
added in 2008-09-26, together with a bunch of other examples, so I guess it
did work at that time, maybe a look at that version will help.

        However, before going down that route, I thought it more appropriate
to ask if anybody else has noticed, and if there is any fix in the works.

        Or, if not, whether there is any suggestion on where should I start
looking to fix the problem myself.

        Thanks in advance and excuse me being such a nuisance.

                                j
-- 
                        EMBnet/CNB
                Scientific Computing Service
        Solving all your computer needs for Scientific
                        Research.

                http://bioportal.cnb.csic.es
                  http://www.es.embnet.org