Periodic optimization using hybrid functional

[rinatn]

Hello,

As a test, I am trying to optimize a small periodic structure from
`TURBOTEST/riper/short/POLYACETYLEN.RIPER.2D.COULOMBE/`
including the cell parameters at the B3LYP/pob-TZVP level with numerical gradients. During the statpt step I get the following error in `job.1`:

Code: [Select]

========================
 internal module stack:
------------------------
    statpt
========================

  Error reading lattice vectors/gradient
 statpt ended abnormally
statpt step ended abnormally
next step = statpt
Run command:

Code: [Select]

jobex -ri -c 300 -level NumGrad > jobex.out
Options from control:

Code: [Select]

$numgrad
  delta 0.001
$periodic 2
$optcell
$cell
4.74988535 7.55890454 90.0
If I remove `$optcell`, the optimization proceeds, but it does not appear to approach convergence:

Code: [Select]

cycle =    100 numerical gradient energy =     -595.4965886775 |dE/dxyz| = *********
cycle =    102 numerical gradient energy =     -624.4970699459 |dE/dxyz| = *********
cycle =    104 numerical gradient energy =     -634.5270019655 |dE/dxyz| = *********
cycle =    106 numerical gradient energy =     -668.8370431169 |dE/dxyz| = *********
cycle =    108 numerical gradient energy =     -679.3497986704 |dE/dxyz| = *********
cycle =    110 numerical gradient energy =     -664.5654587545 |dE/dxyz| = *********
cycle =    112 numerical gradient energy =     -596.8989551892 |dE/dxyz| = *********
cycle =    114 numerical gradient energy =     -615.0889735988 |dE/dxyz| = *********
cycle =    116 numerical gradient energy =     -610.0774804351 |dE/dxyz| = *********
cycle =    118 numerical gradient energy =     -601.1170209988 |dE/dxyz| = *********
cycle =    120 numerical gradient energy =     -626.1810090259 |dE/dxyz| = *********
cycle =    124 numerical gradient energy =     -638.8903650539 |dE/dxyz| = *********
cycle =    126 numerical gradient energy =     -587.5940619614 |dE/dxyz| = *********
cycle =    128 numerical gradient energy =     -640.8105645864 |dE/dxyz| = *********
cycle =    130 numerical gradient energy =     -585.1163768156 |dE/dxyz| = *********
Are the options in `control` set correctly for a 2D periodic optimization with numerical gradients including cell parameters?
Do you have a working example for optimization of periodic structure with a hybrid functional using numerical gradients?

I did not change the default convergence criteria.