r2SCAN forces and the derivative of the energy in CP2K 2025.2 (libxc 7.0.0): test files
=====================================================================================

System: one fluoroacetonitrile molecule (FCH2CN, 6 atoms; atom 1 is F) in a cubic cell of 12.76925 A,
GTH-SCAN pseudopotentials (POTENTIAL_UZH), DZVP-MOLOPT-SR-GTH, OT, 36 MPI x 8 threads.

geometries/fan_s0.xyz ... fan_s7.xyz
    The molecule shifted rigidly along x in 8 steps of 1/8 of the grid spacing at 500 Ry
    (180 points, h = 0.0709 A). F is on a grid point at s0. The C-F bond is along z.

shift_scan/<set>/s<k>/cp2k.inp, cp2k.out
    Single points with forces at the 8 shifts. For an isolated molecule the sum of the forces must be
    minus the derivative of the energy with respect to the shift.

debug_check/<set>/debug.inp, debug.out
    RUN_TYPE DEBUG (DX = 0.001 bohr) at shift 2: analytic forces against finite differences of the
    energy, for F (x, y, z) and for the x component of the other atoms.

Sets:
    r2scan_pw_500        r2SCAN, default XC_DERIV (PW), CUTOFF 500 Ry
    r2scan_spline3_500   r2SCAN, XC_DERIV SPLINE3, CUTOFF 500 Ry
    rscan_pw_500         rSCAN, default XC_DERIV, CUTOFF 500 Ry
    r2scan_pw_1500       r2SCAN, default XC_DERIV, CUTOFF 1500 Ry

    r2scan_pw_500_patched_libxc   (shift scan only) as r2scan_pw_500, run with a libxc 7.0.0 built with
                         patch/libxc-7.0.0-tauclamp.patch preloaded and LIBXC_TAUFIX=1

patch/libxc-7.0.0-tauclamp.patch
    Diagnostic change to src/work_mgga_inc.c of libxc 7.0.0 (not a proposed fix). At grid points where tau
    arrives equal to sigma/(8 rho), i.e. where CP2K has raised tau to that value, it (a) counts them and
    (b) if LIBXC_TAUFIX=1, returns the derivatives of the energy that is actually evaluated,
    e(rho, sigma, tau = sigma/(8 rho)):  vrho - vtau sigma/(8 rho^2),  vsigma + vtau/(8 rho),  vtau = 0.
    Without LIBXC_TAUFIX the results are identical to the unpatched library.

table.txt        the two tables; make_table.py rebuilds them from the outputs (python3 with numpy).

The inputs in shift_scan/ are those that were run, except for the header comment, the project name and
the path of the geometry file.
