Anuja Kumari

The error "internal error: some wfcs were lost" in Quantum Espresso usually arises due to issues related to wavefunction convergence or inappropriate input parameters when dealing with spin-orbit coupling (SOC). Here are several steps and considerations to help resolve this issue:

1. Check Pseudopotentials:

Ensure that the pseudopotentials you are using are suitable for SOC calculations. They should include spin-orbit coupling information. If the pseudopotentials are not SOC-compatible, this can lead to the loss of wavefunctions.

2. Increase ecutwfc and ecutrho:

SOC calculations can be more demanding. Try increasing the plane-wave cutoff energy (ecutwfc) and the charge density cutoff energy (ecutrho).

plaintextCopy codeecutwfc = 70.0 ecutrho = 560.0 (usually 8-10 times `ecutwfc`)

3. Reduce Smearing:

Smearing values that are too high can cause convergence issues. Reduce the value of degauss and see if it improves the situation.

plaintextCopy codedegauss = 0.01

4. Modify Mixing Parameters:

Adjusting the mixing parameters can help in achieving better convergence. Lowering mixing_beta can sometimes stabilise the calculation.

plaintextCopy codemixing_beta = 0.3 electron_maxstep = 100

5. Symmetry Considerations:

Ensure that the symmetry settings are correctly defined. If there is a possibility of symmetry breaking, try setting nosym = .true. to test if symmetry-related issues are causing the problem.

plaintextCopy codenosym = .true.

6. Initial Magnetisation:

Ensure that the starting magnetisation values are appropriate for your system. If these values are too large or too small, they can cause convergence problems.

7. Grid Density:

Ensure that the k-point grid density is sufficiently high. Sometimes a denser grid can help with convergence in SOC calculations.

plaintextCopy codeK_POINTS automatic 16 16 4 0 0 0

8. Convergence Thresholds:

While your current convergence thresholds are quite stringent, you might want to start with slightly looser thresholds and tighten them as the calculation progresses.

plaintextCopy codeconv_thr = 1.0d-8

Example Modified Input File:

plaintextCopy code&CONTROL calculation = 'scf' etot_conv_thr = 3.0d-5 forc_conv_thr = 1.0d-4 outdir = './tmp/' prefix = 'hsl' pseudo_dir = './pseudo/' tprnfor = .true. tstress = .true. verbosity = 'high' / &SYSTEM degauss = 0.01 ecutrho = 560.0 ecutwfc = 70.0 ibrav = 0 nat = 3 nosym = .false. nspin = 2 ntyp = 3 occupations = 'smearing' smearing = 'gaussian' starting_magnetization(1) = 0.1 starting_magnetization(2) = 0.1 starting_magnetization(3) = 0.45 noncolin = .true. angle1(1) = 90.0 angle2(1) = 0.0 angle3(1) = 0.0 lspinorb = .true. / &ELECTRONS conv_thr = 1.0d-8 electron_maxstep = 100 mixing_beta = 0.3 / ATOMIC_SPECIES Br 79.904 Br.rel-pbe-n-kjpaw_psl.1.0.0.UPF I 126.90447 I.rel-pbe-n-kjpaw_psl.1.0.0.UPF La 138.90547 La.rel-pbe-spfn-kjpaw_psl.1.0.0.UPF ATOMIC_POSITIONS crystal La 0.0000000000 0.0000000000 0.0006162566 Br 0.6666666700 0.3333332999 0.1111362558 I 0.6666666700 0.3333332999 0.8719820000 K_POINTS automatic 16 16 4 0 0 0 CELL_PARAMETERS angstrom 4.4924083940 0.0000000000 0.0000000000 -2.2462041970 3.8905397940 0.0000000000 0.0000000000 0.0000000000 15.9605549200

By making these adjustments, you should have a better chance of resolving the "internal error: some wfcs were lost" issue in your spin-orbit coupling calculations.

If you continue to experience problems, consider reaching out to the Quantum Espresso user forums or community for more specific advice based on your system's unique characteristics.

Good luck with your research!

Dear Anuja Kumari,

Besides reading the answer proposed by an AI, I believe your input parameters are inconsistent since you cannot simultaneously set:

nspin = 2

lspinorb = .true.

noncolin = .true.

nspin = 2 assumes the spins parallel to the z-axis. So they are collinear. The last parameter states they are not collinear. I suggest you revise this part and check again your input file (also be careful how you play with the initial angles). Is there a particular reason why you want to set those initial angles?

I hope this helps,

Roberto

PS: Since you are using a hexagonal lattice, use ibrav=4 and set a=4.4924083940

c=15.9605549200

This change helps the code in finding symmetries and reduces the computational cost.

Dear Sandeep Jaiswal

Thank you for detailed possibilities to check for this error. Although after trying each one of it the error remains same.

Thank

Dear Roberto D'Agosta

As per your precise suggestions, I have made respective changes but I am still getting the same error. I have attached both input and output files below. Kindly check for any other discrepancy that is causing this error.

You are setting

nbnd=44

but then asking the code to use smearing. During the construction of the Hamiltonian, the code finds 58 states, while you asked for 44. The code stops at this inconsistency.

My suggestion is that, unless very limited in memory, try to leave the code fix its nbnd. PW uses about 10% more bands, and minimum 4 more than the bare minimum. This is usually a reasonable overhead.

I hope this helps. The final suggestion is to try to impose less constraints as possible on each calculation.

Regards

Roberto

Do NOT set

nspin=4

and

noncolin=.true.

lspinorb=.true.