alsurfp.f

This file gives the outer boundary condition from the Allard model atmospheres: for a star's $\log T_\mathrm{eff}$ and $\log g$, it returns the pressure (and temperature) at which the interior is joined to the atmosphere. By default it returns $\log P$ at the point where $T = T_\mathrm{eff}$; with the option LALTPTau100, $\log P$ and $\log T$ at $\tau = 100$.

• Inputs: the arguments TEFFL ($\log T_\mathrm{eff}$), GL ($\log g$) and LPRT (print to the model file); the table in ALATM01_MOD from ALFILEIN and its limits in ALATM02_MOD from ALTABINIT; the option LALTPTau100 in NMLPAR_MOD, set in PARMIN.
• Who calls it: ENVINT (the envelope integration, for the surface boundary condition) and MASSLOSS.
• What it calls: LOCATE (bisection in a table) and POLINT (polynomial interpolation through 4 points).
• Outputs: AP ($\log P$) and AT ($\log T$) in ATMPRT_MOD, shared with ENVINT, SURFP, KCSURFP, QATM and MASSLOSS; the flag LAlFail, which tells the caller that the point is outside the table. Its comment says the caller should then switch to a gray (KTTAU = 0) or Kurucz (KTTAU = 3) atmosphere.
• What the callers do on failure: ENVINT switches to a gray atmosphere (KTTAU = 0, LTTAU = .TRUE.); MASSLOSS (marked "experimental code … llp 06/15/2009") tests LAlFail and skips the accreted matter's entropy, so the previous lookup's AT, AP are not used (massloss.f, lines 182–191↗; settled from the code, Nov 9, 2026).
• Text output: a line to ISHORT at every call, and to ISTOR as well when LPRT is true.

1. If $\log T_\mathrm{eff} \ge t_\mathrm{max}$, or $\log g$ is outside the overall range (GLXmin, GLXmax), set LAlFail and return. If $\log T_\mathrm{eff} \le t_\mathrm{min}$, stop the run.

2. Find 4 rows in $\log T_\mathrm{eff}$ around the point, shifted to stay inside the table. In each row, find 4 valid columns in $\log g$ around the point; if $\log g$ is outside that row's allowed range, set LAlFail and return. The header says this range extends one column's width beyond the table on each side (lines 13–16↗); ALTABINIT sets four widths below (altabinit.f, question 1).

3. In each row $i$, interpolate in $\log g$ with the polynomial through the 4 points: $$y_i(g) = \sum_{k=1}^{4} y(t_i, g_k) \prod_{l \ne k} \frac{g - g_l}{g_k - g_l},$$ for $y = \log P_\mathrm{eff}$, $\log P_{100}$ and $\log T_{100}$.

4. Interpolate the 4 row values in $\log T_\mathrm{eff}$ the same way. Without the $\tau = 100$ option, AP is $\log P$ at $T = T_\mathrm{eff}$ and AT $= \log T_\mathrm{eff}$; with it, AP $= \log P_{100}$ and AT $= \log T_{100}$.

Since rows can have different valid columns, each row uses its own 4 values of $\log g$: the grid need not be rectangular.

The routine stops the program below the table in $T_\mathrm{eff}$ but only flags the other failures; the stop is a bare STOP, exit status 0 (lines 312–319↗; KNOWN_ISSUES.md, issue 38, open). It writes to ISHORT on every call, which can make the short output long.