[Pw_forum] Problem in Variable-cell

[Varadharajan Srinivasan]

Dear users,

I have been performing some variable-cell relaxation calculations using
PWSCF on PbTiO3 and I come across an interesting problem. After having
relaxed the unit cell I take the relaxed structure and perform an scf
calculation on it to obtain again the forces, stress and energy that
vc-relax gives me anyway. (I made sure that I chose the structure
corresponding to the final energy and forces, ie. the structure reported
before the final estimate) However, I obtain very different answers. The
variable-cell gives me :

######################################################################################
  new lattice vectors (alat unit) : ..........(input alat =       7.3627 (a.u.))
   0.999624836   0.000000001   0.000000000
   0.000000001   0.999624847   0.000000000
   0.000000000   0.000000000   1.081152570
  new unit-cell volume =    431.1954 (a.u.)^3
  new positions in cryst coord
Ti       0.499999997   0.500000003   0.535979348
O        0.499999997   0.500000004   0.112901652
O        0.499999997   0.000000004   0.614268219
O       -0.000000003   0.500000004   0.614268218
Pb      -0.000000002   0.000000003   0.004230374

!    total energy              =  -332.37904437 ryd
     estimated scf accuracy    <     0.00000000 ryd

     convergence has been achieved

     Forces acting on atoms (Ry/au):

     atom   1 type  1   force =     0.00000000    0.00000000   -0.00000063
     atom   2 type  2   force =     0.00000000    0.00000000   -0.00000168
     atom   3 type  2   force =     0.00000000    0.00000000   -0.00000253
     atom   4 type  2   force =     0.00000000    0.00000000   -0.00000253
     atom   5 type  3   force =     0.00000000    0.00000000    0.00000738

     Total force =     0.000008     Total SCF correction =     0.000008


     entering subroutine stress ...

          total   stress  (ryd/bohr**3)                  (kbar)     P=    0.01
   0.00000008   0.00000000   0.00000000          0.01      0.00      0.00
   0.00000000   0.00000008   0.00000000          0.00      0.01      0.00
   0.00000000   0.00000000   0.00000002          0.00      0.00      0.00


     Parrinello-Rahman Damped Dynamics: convergence achieved, Efinal=  -332.37904437
#######################################################################################

While from the corresponding structure the scf calculation gives me :

#######################################################################################
ibrav=  6, celldm(1) =7.359938, celldm(3)=1.081558
Ti       0.499999997   0.500000003   0.535979348
O        0.499999997   0.500000004   0.112901652
O        0.499999997   0.000000004   0.614268219
O       -0.000000003   0.500000004   0.614268218
Pb      -0.000000002   0.000000003   0.004230374

!    total energy              =  -332.94499390 ryd
     estimated scf accuracy    <     0.00000000 ryd

     band energy sum           =   -12.68025211 ryd
     one-electron contribution =   -93.34743263 ryd
     hartree contribution      =    73.14958559 ryd
     xc contribution           =   -50.45891514 ryd
     ewald contribution        =  -262.28823173 ryd

     convergence has been achieved

     Forces acting on atoms (Ry/au):

     atom   1 type  1   force =     0.00000000    0.00000000   -0.00108980
     atom   2 type  2   force =     0.00000000    0.00000000    0.00054152
     atom   3 type  2   force =     0.00000000    0.00000000    0.00008334
     atom   4 type  2   force =     0.00000000    0.00000000    0.00008334
     atom   5 type  3   force =     0.00000000    0.00000000    0.00038160

     Total force =     0.001281     Total SCF correction =     0.000005


     entering subroutine stress ...

          total   stress  (ryd/bohr**3)                  (kbar)     P= -288.66
  -0.00195487   0.00000000   0.00000000       -287.57      0.00      0.00
   0.00000000  -0.00195487   0.00000000          0.00   -287.57      0.00
   0.00000000   0.00000000  -0.00197707          0.00      0.00   -290.84
##############################################################################


Not only is the total energy very different but also the forces and the
stress. I used a ecutwfc=40., ecutrho=240. in both calculations (and
ecfixed=35. Ry in variable-cell) to be consistent in the cutoff energies.

Is this discrepencancy to be expected or am I doing something wrong?
I would greatly appreciate any help/suggestions.

Thanks,
Vardha.