ZBRENT: fatal error in bracketing during relaxation of La–FAU zeolite

[atb025]

Hi all,

I’m encountering a persistent “ZBRENT: fatal error in bracketing” during geometry optimization in VASP, and I’d really appreciate any insight. I am studying La³⁺ exchanged FAU zeolite.

My system consists of:

FAU framework (48 T atoms)
3 framework Al atoms (charge-balanced)
1 La³⁺ cation (in the supercage)

I have an initial run using the following INCAR settings:
SYSTEM = FAU48T_run1
NCORE = 4
LREAL = .TRUE.
ENCUT = 420
PREC = Normal
EDIFF = 1E-4
NELM = 60
ALGO = Fast
ISMEAR = 0
SIGMA = 0.05
IBRION = 2
NSW = 50
EDIFFG = -0.07
ISIF = 2
POTIM = 0.35
ISYM = 2
LWAVE = .FALSE.
LCHARG = .FALSE.

This initial run goes on just fine. But when I do my second run (copying the CONTCAR from the 1st run and using it as my new POSCAR) using the below settings, I encounter the bracketing error:

SYSTEM = FAU48T_run2
NCORE = 4
LREAL = .FALSE.
ENCUT = 520
PREC = Accurate
EDIFF = 1E-6
NELM = 120
NELMIN = 8
ALGO = Normal
ISMEAR = 0
SIGMA = 0.05
ADDGRID = .TRUE.
LASPH = .TRUE.
IBRION = 2
NSW = 120
EDIFFG = -0.03
ISIF = 2
POTIM = 0.05
ISYM = 0
LWAVE = .FALSE.
LCHARG = .FALSE.

I've tried:
- Restarting from CONTCAR several times, but I have the same error
- Switching IBRION = 2(conjugate gradient) to 1(RMM-DIIS), smaller POTIM (down to 0.05)
- Turning off symmetry (ISYM = 0)

Still encountering the same issue.

I want to obtain stable, relaxed La coordination structures in the FAU structure. Any suggestions would be very helpful.

Thanks in advance for your help!

[max_liebetreu]

Hello,

Welcome to the VASP Forums and thank you for reaching out to us!

Could you show us a minimal reproducible example, so at least POSCAR, INCAR, OUTCAR, so that we can get a better understanding of what is happening and try to reproduce the issue? You can attach files directly to your post under "Attachments" below the text window.
Which version of VASP are you using?

Best regards,
Max

[atb025]

Hello, Max.

Thanks for your reply! I've attached the requested input files below.

FAU48T-files.zip

Best,
Aaron

[max_liebetreu]

Hello Aaron,

Thank you for the files, that helps! The structure itself looks normal, but the OUTCAR shows that your electronic minimization does not converge, and neither do your forces.

I should also mention that the specific error you are seeing should be fixed in VASP 6.6, see the changelog:

- The ionic CG algorithm (IBRION = 2) for the Brent algorithm determined the bracketing interval improperly.

However, the error alone is probably not the main issue at play - but if you can, please check if you are encountering the same issue with VASP 6.6.

There's a few things in the settings that you can play around with to address the convergence issues, but this will likely be a bit of trial-and-error. For example, you may want to first try:

• Increase ENCUT (broader energy spectrum)
• Increase SIGMA or choose a different ISMEAR (your system looks like it does not have a pronounced bandgap)
• Increase NELM (allow more steps for electronic convergence)

I managed to get the electronic minimization to converge using VASP 6.5 (did not try your 2-step process, though) - using slightly different settings (higher ENCUT, higher SIGMA):

Code: Select all

SYSTEM = FAU48T_Al32_S6_run2

# Parallelization
NCORE  = 4
LREAL  = .FALSE.

# Electronic
ENCUT  = 700
PREC   = Accurate
EDIFF  = 1E-6
NELM   = 120
NELMIN = 8
ALGO   = Normal

# Smearing
ISMEAR = 0
SIGMA  = 0.2

# Force quality
ADDGRID = .TRUE.
LASPH   = .TRUE.

# Ionic relaxation
IBRION = 2
NSW    = 120
EDIFFG = -0.03
ISIF   = 2
POTIM  = 0.05

# Symmetry
ISYM   = 2

# Output
LWAVE  = .FALSE.
LCHARG = .FALSE.

Depending on your requirements, these settings might already be a step in the right direction?
Let me know how it goes!

Best,
Max

[atb025]

Hello Max,

Thank you for your help!

My electronic minimization has now converged. I used higher ENCUT (700) and SIGMA (0.2) values as you recommended while still using the same VASP version (6.3.2).

[max_liebetreu]

Hello Aaron,

Happy to help, and good luck with your calculations!

Best regards,
Max