Preserve face orientation in intflux

[jlore]

Corrects the reconstruction of cell-centered fluxes in intflux by preserving the sign of the face projection factor fcQalf.

intflux derives an effective face velocity by dividing the signed face flux by the interpolated density and projected face area:

velocity = faceflux / (density * fcS * fcQalf)

The previous implementation used abs(fcQalf). On WG grids, face orientation can reverse across topology cuts, causing fcQalf to change sign. Taking its absolute value discarded that orientation information and could produce an incorrect sign in the reconstructed velocity and interpolated cell-centered flux.

Using the signed value of fcQalf ensures that equivalent physical flows retain a consistent direction across topology cuts. This also matches the signed-area convention already used by divide_by_area in b2us_geo.F

[hrdoktorx]

I do not review draft PRs. But you can encourage me by providing a detailed explanation of why this change is required.

[Copilot]

🟡 Changes recommended

AD-generated intflux_* implementations in src/differentiation/** still use abs(fcQalf) and will be inconsistent with the updated forward model until regenerated/updated.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

Updates the face-to-cell flux interpolation in intflux to preserve face-orientation sign (via fcQalf) when computing effective velocities, with accompanying in-code rationale.

Changes:

• Clarified the algorithm comment to explain why face orientation sign must be preserved across topology cuts.
• Changed effective-velocity computation to divide by the signed projected face area geo%fcS * geo%fcQalf (removing abs(...)).

File summaries

File	Description
src/utility/intflux.F	Preserve face orientation sign in effective-velocity calculation and document the rationale.

Review details

• Files reviewed: 1/1 changed files
• Comments generated: 2
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[hrdoktorx]

@copilot has requested changes. I agree with them.

Requested changes applied, thank you. Still need to see an example of why this is needed.

[jlore]

Ok, should be ready now.

[hrdoktorx]

I'll wait for the KUL review, since you ignored my request for a documented example of where this change matters.

[jlore]

Left corrected, right original

[jlore]

ITER case: Right original, left corrected

[jlore]

@hrdoktorx Here are some examples, the 2nd one is an ITER case

[hrdoktorx]

It does look ugly. Walk me through them, please.

[hrdoktorx]

@pshenovaa @shtirx Do you also see such things in your ITER runs?

[jlore]

I added a description to the PR, but here is another description:

intflux reconstructs a face velocity as
u_face = faceflux / (density_face * fcS * fcQalf)
Across a topology cut, the face orientation can reverse, so fcQalf changes sign. The old code used abs(fcQalf), which discarded that orientation and gave the reconstructed velocity the wrong sign on those faces. When those face values were averaged to cell centers, the error appeared as the discontinuity visible along the topology cut in the “before” plot.
The corrected code retains the sign of fcQalf, so the orientation in the stored face flux is properly canceled and the field remains continuous in the “after” plot.

[jlore]

And the ITER case I am plotting is from Andrei

[jlore]

Note this is just a b2plot error, it's already handled correctly elsewhere as I noted in the description

[hrdoktorx]

Is intflux only called by b2plot?

[jlore]

Yes, from what I can see. Only in b2plot and caltraj, which is only used by b2plot.

[hrdoktorx]

Since it is only a b2plot error, I am satisfied. Trust, but verify.