# Restricted Hartree-Fock with a general McMurchie-Davidson integral engine,
# supporting s AND p Cartesian Gaussians (extends to higher angular momentum).
#
# The molecule/basis below is H2 (all s) for engine validation against the
# known s-only result -1.11671 hartree; swap the "MOLECULE / BASIS" block for
# water to do a real p-orbital calculation.

pi = 3.141592653589793
n_prim = 3

# =====================================================================
# McMurchie-Davidson integral engine (molecule-independent).
# =====================================================================

# Boys function F_n(T) = integral_0^1 t^(2n) exp(-T t^2) dt.
def boys_n(nn, T) {
    if T < 1.0e-12 {
        return 1.0 / (2.0 * nn + 1.0)
    }
    return simpson(fn(t) -> pow(t, 2.0 * nn) * exp(-T * t * t), 0.0, 1.0, 100)
}

# Hermite expansion coefficient E_t^{ij} (1D), exponents a,b, Qx = Ax - Bx.
def hermite_E(i, j, t, Qx, a, b) {
    if i < 0.0 { return 0.0 }
    if j < 0.0 { return 0.0 }
    if t < 0.0 { return 0.0 }
    if t > i + j { return 0.0 }
    p = a + b
    if i == 0.0 {
        if j == 0.0 {
            mu = a * b / p
            return exp(-mu * Qx * Qx)
        }
        xpb = a / p * Qx
        return (0.5 / p * hermite_E(i, j - 1.0, t - 1.0, Qx, a, b)
             + xpb * hermite_E(i, j - 1.0, t, Qx, a, b)
             + (t + 1.0) * hermite_E(i, j - 1.0, t + 1.0, Qx, a, b))
    }
    xpa = -b / p * Qx
    return (0.5 / p * hermite_E(i - 1.0, j, t - 1.0, Qx, a, b)
         + xpa * hermite_E(i - 1.0, j, t, Qx, a, b)
         + (t + 1.0) * hermite_E(i - 1.0, j, t + 1.0, Qx, a, b))
}

# Hermite Coulomb integral R_{tuv}^n.
def hermite_R(t, u, v, nn, p, PCx, PCy, PCz, RPC2) {
    if t == 0.0 {
        if u == 0.0 {
            if v == 0.0 {
                return pow(-2.0 * p, nn) * boys_n(nn, p * RPC2)
            }
            res = PCz * hermite_R(t, u, v - 1.0, nn + 1.0, p, PCx, PCy, PCz, RPC2)
            if v > 1.0 {
                res += (v - 1.0) * hermite_R(t, u, v - 2.0, nn + 1.0, p, PCx, PCy, PCz, RPC2)
            }
            return res
        }
        res = PCy * hermite_R(t, u - 1.0, v, nn + 1.0, p, PCx, PCy, PCz, RPC2)
        if u > 1.0 {
            res += (u - 1.0) * hermite_R(t, u - 2.0, v, nn + 1.0, p, PCx, PCy, PCz, RPC2)
        }
        return res
    }
    res = PCx * hermite_R(t - 1.0, u, v, nn + 1.0, p, PCx, PCy, PCz, RPC2)
    if t > 1.0 {
        res += (t - 1.0) * hermite_R(t - 2.0, u, v, nn + 1.0, p, PCx, PCy, PCz, RPC2)
    }
    return res
}

def overlap_1d(i, j, Qx, a, b) {
    return hermite_E(i, j, 0.0, Qx, a, b) * sqrt(pi / (a + b))
}

# 1D kinetic energy matrix element (differentiate the ket twice).
def kinetic_1d(i, j, Qx, a, b) {
    term = b * (2.0 * j + 1.0) * overlap_1d(i, j, Qx, a, b)
    term += -2.0 * b * b * overlap_1d(i, j + 2.0, Qx, a, b)
    term += -0.5 * j * (j - 1.0) * overlap_1d(i, j - 2.0, Qx, a, b)
    return term
}

# =====================================================================
# Primitive integrals over two/four basis functions (reading bf data
# by index from the globals defined in the MOLECULE/BASIS block).
# =====================================================================

def prim_overlap(a, b, mu, nu) {
    sx = overlap_1d(bfL[mu], bfL[nu], bfCx[mu] - bfCx[nu], a, b)
    sy = overlap_1d(bfM[mu], bfM[nu], bfCy[mu] - bfCy[nu], a, b)
    sz = overlap_1d(bfN[mu], bfN[nu], bfCz[mu] - bfCz[nu], a, b)
    return sx * sy * sz
}

def prim_kinetic(a, b, mu, nu) {
    sx = overlap_1d(bfL[mu], bfL[nu], bfCx[mu] - bfCx[nu], a, b)
    sy = overlap_1d(bfM[mu], bfM[nu], bfCy[mu] - bfCy[nu], a, b)
    sz = overlap_1d(bfN[mu], bfN[nu], bfCz[mu] - bfCz[nu], a, b)
    tx = kinetic_1d(bfL[mu], bfL[nu], bfCx[mu] - bfCx[nu], a, b)
    ty = kinetic_1d(bfM[mu], bfM[nu], bfCy[mu] - bfCy[nu], a, b)
    tz = kinetic_1d(bfN[mu], bfN[nu], bfCz[mu] - bfCz[nu], a, b)
    return tx * sy * sz + sx * ty * sz + sx * sy * tz
}

def prim_nuclear(a, b, mu, nu, Cx, Cy, Cz) {
    la = bfL[mu]
    ma = bfM[mu]
    na = bfN[mu]
    lb = bfL[nu]
    mb = bfM[nu]
    nb = bfN[nu]
    p = a + b
    Px = (a * bfCx[mu] + b * bfCx[nu]) / p
    Py = (a * bfCy[mu] + b * bfCy[nu]) / p
    Pz = (a * bfCz[mu] + b * bfCz[nu]) / p
    PCx = Px - Cx
    PCy = Py - Cy
    PCz = Pz - Cz
    RPC2 = PCx * PCx + PCy * PCy + PCz * PCz
    qx = bfCx[mu] - bfCx[nu]
    qy = bfCy[mu] - bfCy[nu]
    qz = bfCz[mu] - bfCz[nu]
    val = 0.0
    for t to la + lb + 1.0 {
        for u to ma + mb + 1.0 {
            for v to na + nb + 1.0 {
                e = hermite_E(la, lb, t, qx, a, b) * hermite_E(ma, mb, u, qy, a, b) * hermite_E(na, nb, v, qz, a, b)
                val += e * hermite_R(t, u, v, 0.0, p, PCx, PCy, PCz, RPC2)
            }
        }
    }
    return 2.0 * pi / p * val
}

def prim_eri(a, b, c, d, mu, nu, lam, sig) {
    p = a + b
    q = c + d
    Px = (a * bfCx[mu] + b * bfCx[nu]) / p
    Py = (a * bfCy[mu] + b * bfCy[nu]) / p
    Pz = (a * bfCz[mu] + b * bfCz[nu]) / p
    Qx = (c * bfCx[lam] + d * bfCx[sig]) / q
    Qy = (c * bfCy[lam] + d * bfCy[sig]) / q
    Qz = (c * bfCz[lam] + d * bfCz[sig]) / q
    alpha = p * q / (p + q)
    PQx = Px - Qx
    PQy = Py - Qy
    PQz = Pz - Qz
    RPQ2 = PQx * PQx + PQy * PQy + PQz * PQz
    qx1 = bfCx[mu] - bfCx[nu]
    qy1 = bfCy[mu] - bfCy[nu]
    qz1 = bfCz[mu] - bfCz[nu]
    qx2 = bfCx[lam] - bfCx[sig]
    qy2 = bfCy[lam] - bfCy[sig]
    qz2 = bfCz[lam] - bfCz[sig]
    val = 0.0
    for t1 to bfL[mu] + bfL[nu] + 1.0 {
        for u1 to bfM[mu] + bfM[nu] + 1.0 {
            for v1 to bfN[mu] + bfN[nu] + 1.0 {
                e1 = hermite_E(bfL[mu], bfL[nu], t1, qx1, a, b) * hermite_E(bfM[mu], bfM[nu], u1, qy1, a, b) * hermite_E(bfN[mu], bfN[nu], v1, qz1, a, b)
                for t2 to bfL[lam] + bfL[sig] + 1.0 {
                    for u2 to bfM[lam] + bfM[sig] + 1.0 {
                        for v2 to bfN[lam] + bfN[sig] + 1.0 {
                            e2 = hermite_E(bfL[lam], bfL[sig], t2, qx2, c, d) * hermite_E(bfM[lam], bfM[sig], u2, qy2, c, d) * hermite_E(bfN[lam], bfN[sig], v2, qz2, c, d)
                            sign = pow(-1.0, t2 + u2 + v2)
                            val += e1 * e2 * sign * hermite_R(t1 + t2, u1 + u2, v1 + v2, 0.0, alpha, PQx, PQy, PQz, RPQ2)
                        }
                    }
                }
            }
        }
    }
    return 2.0 * pow(pi, 2.5) / (p * q * sqrt(p + q)) * val
}

# Contracted integrals: sum primitive integrals weighted by contraction coeffs.
def cont_overlap(mu, nu) {
    s = 0.0
    for ka to n_prim {
        for kb to n_prim {
            s += bfC[mu, ka] * bfC[nu, kb] * prim_overlap(bfA[mu, ka], bfA[nu, kb], mu, nu)
        }
    }
    return s
}

def cont_kinetic(mu, nu) {
    s = 0.0
    for ka to n_prim {
        for kb to n_prim {
            s += bfC[mu, ka] * bfC[nu, kb] * prim_kinetic(bfA[mu, ka], bfA[nu, kb], mu, nu)
        }
    }
    return s
}

def cont_nuclear(mu, nu, Cx, Cy, Cz) {
    s = 0.0
    for ka to n_prim {
        for kb to n_prim {
            s += bfC[mu, ka] * bfC[nu, kb] * prim_nuclear(bfA[mu, ka], bfA[nu, kb], mu, nu, Cx, Cy, Cz)
        }
    }
    return s
}

def cont_eri(mu, nu, lam, sig) {
    s = 0.0
    for ka to n_prim {
        for kb to n_prim {
            for kc to n_prim {
                for kd to n_prim {
                    coef = bfC[mu, ka] * bfC[nu, kb] * bfC[lam, kc] * bfC[sig, kd]
                    s += coef * prim_eri(bfA[mu, ka], bfA[nu, kb], bfA[lam, kc], bfA[sig, kd], mu, nu, lam, sig)
                }
            }
        }
    }
    return s
}

# =====================================================================
# MOLECULE / BASIS: H2 in STO-3G (all s).  Validation target: -1.11671.
# =====================================================================
n_atoms = 2
n_basis = 2
n_occ = 1
R_HH = 1.4

atom_x = zeros(n_atoms)
atom_y = zeros(n_atoms)
atom_z = zeros(n_atoms)
atom_Z = zeros(n_atoms)
atom_z[1] = R_HH
atom_Z[0] = 1.0
atom_Z[1] = 1.0

bfL = zeros(n_basis)
bfM = zeros(n_basis)
bfN = zeros(n_basis)
bfCx = zeros(n_basis)
bfCy = zeros(n_basis)
bfCz = zeros(n_basis)
bfCz[1] = R_HH
bfA = zeros(n_basis, n_prim)
bfC = zeros(n_basis, n_prim)

h_a0 = 3.42525091
h_a1 = 0.62391373
h_a2 = 0.16885540
h_c0 = 0.15432897
h_c1 = 0.53532814
h_c2 = 0.44463454

for i to n_basis {
    bfA[i, 0] = h_a0
    bfA[i, 1] = h_a1
    bfA[i, 2] = h_a2
    # s primitive normalization (2a/pi)^{3/4}, folded into the coefficient.
    bfC[i, 0] = h_c0 * pow(2.0 * h_a0 / pi, 0.75)
    bfC[i, 1] = h_c1 * pow(2.0 * h_a1 / pi, 0.75)
    bfC[i, 2] = h_c2 * pow(2.0 * h_a2 / pi, 0.75)
}

# Normalize each contracted basis function so <mu|mu> = 1.
for i to n_basis {
    nrm = 1.0 / sqrt(cont_overlap(i, i))
    for k to n_prim {
        bfC[i, k] = bfC[i, k] * nrm
    }
}

# =====================================================================
# Build integrals and run the SCF.
# =====================================================================
S = zeros(n_basis, n_basis)
H_core = zeros(n_basis, n_basis)
for mu to n_basis {
    for nu to n_basis {
        S[mu, nu] = cont_overlap(mu, nu)
        v = 0.0
        for atom to n_atoms {
            v += -atom_Z[atom] * cont_nuclear(mu, nu, atom_x[atom], atom_y[atom], atom_z[atom])
        }
        H_core[mu, nu] = cont_kinetic(mu, nu) + v
    }
}

eri = zeros(n_basis, n_basis, n_basis, n_basis)
parallel for mu to n_basis {
    for nu to n_basis {
        for lam to n_basis {
            for sig to n_basis {
                eri[mu, nu, lam, sig] = cont_eri(mu, nu, lam, sig)
            }
        }
    }
}

# Symmetric orthogonalization X = S^{-1/2}.
s_vals = zeros(n_basis)
s_vecs = zeros(n_basis, n_basis)
eig_sym(S, s_vals, s_vecs)
D_inv_sqrt = zeros(n_basis, n_basis)
for i to n_basis {
    D_inv_sqrt[i, i] = 1.0 / sqrt(s_vals[i])
}
X = s_vecs @ D_inv_sqrt @ transpose(s_vecs)

# Nuclear repulsion.
E_nuc = 0.0
for i to n_atoms {
    for j to n_atoms {
        if j > i {
            dx = atom_x[i] - atom_x[j]
            dy = atom_y[i] - atom_y[j]
            dz = atom_z[i] - atom_z[j]
            E_nuc += atom_Z[i] * atom_Z[j] / sqrt(dx * dx + dy * dy + dz * dz)
        }
    }
}

# SCF.
P = zeros(n_basis, n_basis)
E_total = 0.0
for iter to 100 {
    G = zeros(n_basis, n_basis)
    for mu to n_basis {
        for nu to n_basis {
            g = 0.0
            for lam to n_basis {
                for sig to n_basis {
                    g += P[lam, sig] * (eri[mu, nu, lam, sig] - 0.5 * eri[mu, lam, nu, sig])
                }
            }
            G[mu, nu] = g
        }
    }
    F = H_core + G
    Fp = transpose(X) @ F @ X
    eps_vals = zeros(n_basis)
    Cp = zeros(n_basis, n_basis)
    eig_sym(Fp, eps_vals, Cp)
    C = X @ Cp
    P_new = zeros(n_basis, n_basis)
    for mu to n_basis {
        for nu to n_basis {
            psum = 0.0
            for a to n_occ {
                psum += C[mu, a] * C[nu, a]
            }
            P_new[mu, nu] = 2.0 * psum
        }
    }
    E_elec = 0.0
    for mu to n_basis {
        for nu to n_basis {
            E_elec += 0.5 * P_new[mu, nu] * (H_core[mu, nu] + F[mu, nu])
        }
    }
    E_total = E_elec + E_nuc
    delta = 0.0
    for mu to n_basis {
        for nu to n_basis {
            dd = P_new[mu, nu] - P[mu, nu]
            if dd < 0.0 { dd = -dd }
            if dd > delta { delta = dd }
        }
    }
    P = P_new
    if delta < 1.0e-9 { break }
}

show E_total

# SPDX-License-Identifier: MIT OR Apache-2.0
# SPDX-FileCopyrightText: 2026 The Viv authors