int FSYM(xcgets) FCSYM(XCGETS)(fortran_int_t *fun, double *value, fortran_int_t *namelen, const char *name)
{
int i;
char buf[257];
if (*namelen>256)
xcint_die("In xcgets_(): name string too long",*namelen);
for (i=0;i<*namelen;i++)
buf[i] = name[i];
buf[*namelen] = 0;
return xc_get(fortran_functionals[*fun],buf,value);
}
// Compute the xc potential(s). This is simple for lda and more complicated for
// GGA's. For metaGAA's that depend on tau this is a non-local problem and
// cannot be solved by xcfun. Laplacian dependent metaGGA's could be implemented..
// Another option is to evaluate the divergence directly in cartesian coordinates,
// but one have to find a way to do this without introducing the gradient components
// explicitly (for niceness)
void xc_potential(xc_functional fun, const double *density, double *e_xc, double *v_xc)
{
if (fun->mode == XC_VARS_AB)
{
if (fun->type == XC_LDA)
{
double out[3];
xc_eval(fun,1,density,out);
e_xc[0] = out[0];
v_xc[0] = out[1];
v_xc[1] = out[2];
}
// Expecting lap_a and lap_b as element 5 and 6 of density
// Using that v_xc = dE/dn - div dE/dgradn
// When deriving the expressions it helps to have the operator
// (g_a).nabla act on the basic variables (and the same for g_b).
else if (fun->type == XC_GGA)
{
const int gaa = 2, gab = 3, gbb = 4, lapa = 5, lapb = 6;
double out[21];
xc_eval(fun,2,density,out);
e_xc[0] = out[XC_D00000];
v_xc[0] = out[XC_D10000];
v_xc[0] -= 2*density[lapa]*out[XC_D00100] + density[lapb]*out[XC_D00010];
v_xc[0] -= 2*(out[XC_D10100]*density[gaa] +
out[XC_D01100]*density[gab] +
out[XC_D00200]*(2*density[lapa]*density[gaa]) +
out[XC_D00110]*(density[lapa]*density[gab] + density[lapb]*density[gaa]) +
out[XC_D00101]*(2*density[lapb]*density[gab])
);
v_xc[0] -= (out[XC_D10010]*density[gab] +
out[XC_D01010]*density[gbb] +
out[XC_D00110]*(2*density[lapa]*density[gab]) +
out[XC_D00020]*(density[lapb]*density[gab] + density[lapa]*density[gbb]) +
out[XC_D00011]*(2*density[lapb]*density[gbb]));
v_xc[1] = out[XC_D01000];
v_xc[1] -= 2*density[lapb]*out[XC_D00001] + density[lapa]*out[XC_D00010];
v_xc[1] -= 2*(out[XC_D01001]*density[gbb] +
out[XC_D10001]*density[gab] +
out[XC_D00002]*(2*density[lapb]*density[gbb]) +
out[XC_D00011]*(density[lapb]*density[gab] + density[lapa]*density[gbb]) +
out[XC_D00101]*(2*density[lapa]*density[gab]) );
v_xc[1] -= (out[XC_D01010]*density[gab] +
out[XC_D10010]*density[gaa] +
out[XC_D00011]*(2*density[lapb]*density[gab]) +
out[XC_D00020]*(density[lapa]*density[gab] + density[lapb]*density[gaa]) +
out[XC_D00110]*(2*density[lapa]*density[gaa]));
}
else
{
xcint_die("xc_potential() not implemented for metaGGA's",fun->type);
}
}
else if (fun->mode == XC_VARS_N)
{
if (fun->type == XC_LDA)
{
double out[2];
xc_eval(fun,1,density,out);
e_xc[0] = out[0];
v_xc[0] = out[1];
}
else
{
xcint_die("xc_potential() GGA not implemented for XC_VARS_N",
fun->type);
}
}
else
{
xcint_die("xc_potential() GGA only implemented for AB mode",
fun->mode);
}
}
