Problems met in constraining magnetic moments in non-collinear calculation with soc

[hongriTianqi]

Describe the bug

The calculation is ok when one wants to constrain magnetic moment values in x, y, and z directions.

Problem happens when constraining magnetic moment values in an arbitrary direction:

There are two problems:

• several calculations were not converged.
• with the converged two cases, the ground state energy and magnetic moments are inconsistent with that in x,y,z directions.

Expected behavior

The performance should be exactly the same when one wants to constrain magnetic moments in arbitrary direction compared to x,y,z directions.

To Reproduce

repeat.tar.gz arbitraryDirection.tar.gz

Environment

No response

Additional Context

No response

Task list for Issue attackers (only for developers)

• [x] Verify the issue is not a duplicate.
• [x] Describe the bug.
• [x] Steps to reproduce.
• [x] Expected behavior.
• [x] Error message.
• [ ] Environment details.
• [ ] Additional context.
• [x] Assign a priority level (low, medium, high, urgent).
• [x] Assign the issue to a team member.
• [x] Label the issue with relevant tags.
• [ ] Identify possible related issues.
• [ ] Create a unit test or automated test to reproduce the bug (if applicable).
• [ ] Fix the bug.
• [ ] Test the fix.
• [ ] Update documentation (if necessary).
• [ ] Close the issue and inform the reporter (if applicable).