[Pw_forum] different energy values when toggling on nosym

[Ye Luo]

Dear QE users,

I just noticed that pw.x was giving me different energy values. I don't
know if it's just due to machine accuracy or something strange happens.

1) I did the scf calculation with k points generated automatically and got
energy_1 (-362.64960323 Ry). The input file is attached at the end of this
email.
2) I copy the k points and weights spit out by step 1) to the input file
and redo the calculation, I got exactly the same value as energy_1.
3) I add nosym=.true. in the input and redo the calculation. I got energy_3
(-362.64951123 Ry) which is different but not far from energy_1.
The difference is 0.00009200 Ry
Is this normal?

Then I tried to do a calculation with hybrid functional (HSE). q-mesh 1x1x1.
the same step 1) I got energy energy_hybrid_1 (-362.47024611 Ry)_
the same step 2) I got the same energy value as energy_hybrid_1
the same step 3) I got energy_hybrid_3 (-362.46997237 Ry)
Now the difference is 0.00027374 Ry about 3 times as the previous one.
Is this normal?

For my understanding, nosym affects the k point generation.
I checked the output of 1) 2) 3) and they have the same kpoints and weights.
In principle, I thought step 2) and 3) should yield the same energy. Only
machine accuracy plays a little bit. If nosym is doing more than K points,
please bear my limited knowledge and explain me where else symmetry is also
involved and which might be relevant to my calculation.

Let me know your thoughts. Thanks!

Ye

 &control
    calculation = 'scf'
    restart_mode='from_scratch',
    prefix='TiO2-rutile',
    tstress = .true.
    tprnfor = .true.
    pseudo_dir = '/home/yeluo/pw_pseudo/',
    outdir='./out',
 /
 &system
    ibrav     = 0,
    nat       = 6,
    ntyp      = 2,
    ecutwfc   = 300,
    input_dft = 'pbe'
    lda_plus_u = .true.,
    Hubbard_U(1) = 4.86041
 /
 &electrons
    mixing_beta = 0.7
    conv_thr =  1.0d-10
 /
CELL_PARAMETERS bohr
8.681023246 0 0
0 8.681023246 0
0 0 5.59094331
ATOMIC_SPECIES
 Ti  1.0  Ti.opt.upf
 O   1.0  O.ncpp
ATOMIC_POSITIONS crystal
 Ti  0.0000000   0.0000000   0.0000000
 Ti  0.5000000   0.5000000   0.5000000
 O   0.3050990   0.3050990   0.0000000
 O   0.6949010   0.6949010   0.0000000
 O   0.1949010   0.8050990   0.5000000
 O   0.8050990   0.1949010   0.5000000
K_POINTS automatic
8 8 8 1 1 1


===================
Ye Luo, Ph.D.
Leadership Computing Facility
Argonne National Laboratory
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20160327/0deaaed0/attachment.html>