A question about the code for reading pseudopotentials

[A-006]

Describe the bug

in this expression,the pp.lchi[nw] - pp.jchi[nw] - 0.5,The result of this expression is always false. Should we calculate the absolute value of pp.lchi[nw] - pp.jchi[nw] to be around 0.5

Expected behavior

No response

To Reproduce

No response

Environment

No response

Additional Context

No response

Task list for Issue attackers (only for developers)

• Verify the issue is not a duplicate.
• Describe the bug.
• Steps to reproduce.
• Expected behavior.
• Error message.
• Environment details.
• Additional context.
• Assign a priority level (low, medium, high, urgent).
• Assign the issue to a team member.
• 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).

[dyzheng]

@A-006 This code block try to dismiss SOC part of pseudopotential if the check of j == l + 0.5 or j == l - 0.5 fail, but there is a bug in it.
I'm interested that can you find a upf100 pseudopotential file with SOC but not comply with this j-l rule?

[A-006]

Thanks for your relpy,does it means that we can use the code as bellow
(abs(abs(pp.lchi[nw]-pp.jchi[nw])-0.5)<1e-7)
My understanding is that the rounding error range for single precision (float) is actually around 10e-7, while the rounding error range for double precision (double) is around 10e-14. Choosing a smaller range helps to reduce numerical errors？

I found it with TscanCode, which is a static code analysis tool. According to research on static code analysis, we can solve 30%-70% of the bugs in our program before we debug it when the problems happen. So, it's strongly recommended to use static code check tools. Actually, we can use three open-source code check tools: flawfinder, cppcheck, and TscanCode (most recommended!!!, [https://github.com/Tencent/TscanCode]

[mohanchen]

@A-006 This code block try to dismiss SOC part of pseudopotential if the check of j == l + 0.5 or j == l - 0.5 fail, but there is a bug in it. I'm interested that can you find a upf100 pseudopotential file with SOC but not comply with this j-l rule?

He has a background of computer science, he cannot understand your comment.

[AsTonyshment]

@A-006 This part of the code is related to spin-orbit coupling (SOC), where pp.lchi and pp.jchi represent the angular momentum quantum number $l$ (orbital) and $j$ (total) for the pseudo wavefunctions (PSWFC) defined in the pseudopotential file, while pp.lll and pp.jjj represent the angular momentum quantum number $l$ (orbital) and $j$ (total) for the nonlocal projectors (Beta Projectors).

The rule $j = l \pm 0.5$ is a fundamental requirement for SOC, as it ensures the correct coupling between spin and orbital angular momentum in quantum mechanics. The code is designed to check whether this rule is satisfied for both wavefunctions (pp.lchi and pp.jchi) and nonlocal projectors (pp.lll and pp.jjj). If the rule is not satisfied, the SOC part of the pseudopotential should be dismissed.

So, I believe the condition here should be modified to check whether $j$ is not equal to either $l + 0.5$ or $l - 0.5$, i.e.:

if (std::abs(pp.lchi[nw] - pp.jchi[nw] - 0.5) > 1e-7 && std::abs(pp.lchi[nw] - pp.jchi[nw] + 0.5) > 1e-7)

This ensures that the SOC part is dismissed only when $j$ does not match either $l + 0.5$ or $l - 0.5$. The same logic should also be applied to the nonlocal projectors (pp.lll and pp.jjj).

What do you think? @dyzheng