[QE-users] Optimization stuck at bfgs step

[Barsha Pal]

Hello everyone,

I am trying to optimize this structure . I have run successful vc-relax
optimization with low criteria (etot_conv_thr = 1.0D-5, forc_conv_thr =
1.0D-4) ,but when I increase the criteria bfgs fails. This structure exists
in literature with buckling that I can't seem to get here.Due to bfgs
failure in vc-relax,  I had decided to run a relax followed by a vc-relax.
In my current relax calculation the optimization is stuck at bfgs 7 with
scf step 138. My question is if I want to successfully relax this
structure (I would be doing phonon afterwards ) what should I do here ?
Should I stop the current run and rerun ? Please let me know the
best solution.

Input:
&CONTROL
    calculation   = "relax"
    restart_mode  = "from_scratch"
    outdir="/tmp/"
    prefix="p"
    pseudo_dir="./"
    verbosity="high"
    etot_conv_thr = 1.0D-6
    forc_conv_thr = 1.0D-4
    nstep  = 1000
    tprnfor       = .true.
    disk_io       = "none"
/

&SYSTEM
  ibrav    = 0
  celldm(1) = 1
  nat    = 8
  ntyp    = 2
  ecutwfc   = 210
/

&ELECTRONS
    electron_maxstep = 600
    mixing_beta = 0.2
    diagonalization = "cg"
    diago_cg_maxiter = 500
    conv_thr =  1.0d-5
/

&IONS
    ion_dynamics = "bfgs"
/

ATOMIC_SPECIES
 B 00d0 B.pbe-n-rrkjus_psl.1.0.0.UPF
 A 00d0 A.pbe-dnl-kjpaw_psl.1.0.0.UPF

CELL_PARAMETERS (alat=  1.00000000)
  15.342766198   0.000000000   0.000000000
   0.000000000  15.342766198   0.000000000
   0.000000000   0.000000000  37.794630380

ATOMIC_POSITIONS (crystal)
A               0.6461706388        0.1462855988        0.5000000000
B               0.6495390386        0.8505502843        0.5000000000
B               0.3504609614        0.1494497157        0.5000000000
A               0.3538293612        0.8537144012        0.5000000000
A               0.1462855988        0.3538293612        0.5000000000
B               0.8505502843        0.3504609614        0.5000000000
B               0.1494497157        0.6495390386        0.5000000000
A               0.8537144012        0.6461706388        0.5000000000

K_POINTS {automatic}
  14 14 3 0 0 0




Current situation of output:

     Total force =     0.000970     Total SCF correction =     0.002258
     SCF correction compared to forces is large: reduce conv_thr to get
better values
     Energy error            =      7.3E-08 Ry
     Gradient error          =      3.9E-04 Ry/Bohr

     number of scf cycles    = 138
     number of bfgs steps    =   7

     energy             old  =   -1193.3574486014 Ry
     energy             new  =   -1193.3574485288 Ry

     CASE: energy            _new > energy            _old

     new trust radius        =       0.0000504644 bohr

     trust_radius < trust_radius_min

     resetting bfgs history

     new conv_thr            =       0.0000001000 Ry



Kind Regards,
Ashley Cooper
Research Scholar
Assam University Silchar
Assam, India
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20250828/2be51950/attachment.html>