The ABINIT Users Mailing List ( CLOSED )

["Eric J. Walter" <ewalter@particle.physics.wm.edu>]

Dear Abinit forum,

I think that I have encountered a minor bug in the 4.4.4 release.
Specifically, in Src_2psp/psp5in.F90:

------------------------------------------------------------
!Generate core charge function and derivatives, if needed
 if(n1xccc>0)then
! Use the revised expression of 5 Nov 1992, also used for format=1.
  call psp1cc(fchrg,n1xccc,xccc1d)
  xcccrc=3*rchrg
 end if
------------------------------------------------------------

It appears as though xcccrc is never initialized if one uses type 5
pseudopotentials WITHOUT a core correction (n1xccc=0).  I think this is
due to a possible error in pspini.F90:

------------------------------------------------------------
    do ipsp=1,npsp

    ekb(:)=zero;ffspl(:,:,:)=zero;vlspl(:,:)=zero
    xccc1d(:,:)=zero;xcccrc=zero                  
    indlmn(:,:)=0

    if(gencond0==0 .and. new_pspso(ipsp)==0) cycle

    write(message, '(a,i4,a,t38,a)' ) &
&    '- pspini: atom type',ipsp,'  psp file is',trim(psps%filpsp(ipsp))
    call wrtout(ab_out,message,'COLL')

!   Read atomic psp V(r) and wf(r) to get local and nonlocal psp:
    if(psps%usepaw==0)then ! cannot use the same call in case of bound 
checking, because of pawrad/pawtab
     call pspatm(psps%dimekb,dq,ekb,epsatm(ipsp),&
&     ffspl,psps%filpsp(ipsp),indlmn,ixc,&
&     
psps%lmnmax,psps%lnmax,psps%mpsang,psps%mpssoang,psps%mqgrid,psps%n1xccc,psps%optnlxccc,&
&     pawrad,pawtab,psps%pspso(ipsp),&
&     psps%qgrid,psps%usepaw,psps%useylm,vlspl,&
&     psps%xcccrc(ipsp),xccc1d,psps%znuclpsp(ipsp))
    else
     call pspatm(psps%dimekb,dq,ekb,epsa
.
.
.

------------------------------------------------------------

In the third line above, xcccrc is initialized to zero, but
psps%xcccrc is passed to pspatm.

Later, when doing a strain response calculation, this leads to a core
dump (on our alpha EV667s) due to code in Src_6response/eltfrxc3.F90:

------------------------------------------------------------
! Loop over atoms in unit cell
  do iatom=1,natom
   js=7+3*(iatom-1)
!  Set search range (density cuts off perfectly beyond range)
!  Cycle if no range.
   range=0.0d0
   range=xcccrc(typat(iatom))
   if(range<1.d-16) cycle
------------------------------------------------------------

Anyway, simply initializing explicitly in psp5in.F90 fixes this
problem.  But, most likely the initialization should occur in
pspini.F90

I have tried to following the link to:
http://www.abinit.org/ABINIT/bug_report.html to review the netiquette
for possible bug reports, but it seems that this link is broken.

I Hope this email is helpful to the developers.

Sincerely, 

Eric J. Walter
Department of Physics 
College of William and Mary