[victorkemp]

D. MITC4+: through-thickness temperature gradient

• Bug: MITC4 ignored a TEMPP1 gradient (TPRIME) entirely. A plate with a gradient didn't deflect at all, and the recovered moments were -D alpha TPRIME. MIN4 is exact.
• Fix: add [Bb]' SHELL_DALP TPRIME, the term MIN4's QPLT2 uses.
• Tests: free strips with TPRIME = 1000 (PSHELL isotropic, PSHELL with an orthotropic MID2 CTE, [0/90]s PCOMP) against k = D^-1 M_T, plus zero moments.

E. Shell CTEs: the xy row of ALPVEC (shared shell code)

• The convention: ALPVEC rows are (xx, yy, zz, xy, yz, zx). The rotations and the membrane CTEs use row 4 for xy.
• Bugs:
  • MATERIAL_PROPS_2D put a MAT2's A3 (alpha_12) in row 3, so it was ignored.
  • The bending CTEs (SHELL_DALP, recovered thermal curvature) read rows 1-3, losing any xy part.
• Effect on MITC4+: a free PSHELL strip with a MAT2 alpha_12 had no shear strain under DT and no twist under a gradient.
• Fix: MAT2 A3 goes to row 4, and the bending CTEs read rows 1, 2, 4.
• Side effect: MIN4/TRIA3 benefit too; e.g. TRIA3 with an orthotropic bending CTE under a gradient was 2.2x off.

F. PCOMP LAM=SYM with Z0 (shared bulk-data code)

• Bug: for LAM = SYM the reader mirrored the ply z about the reference plane (-z) instead of the laminate mid-plane (Z0 + T/2). The stiffness was right, because it integrates from Z0. But ZPLY users were wrong:
  • ply stresses
  • the buckling membrane forces
• Effect on MITC4+: a strip column with Z0 = -T buckled at exactly 1/4 of Euler.
• Fix: BD_PCOMP and BD_PCOMP1. Nothing changes for the default Z0.