def2-SVP Energy mismatch between MESS and PySCF

[majhas]

Hey, MESS does not get the same energy as PySCF when using the def2-SVP basis set. For other basis sets (sto-3g; 6-31g) there is no issue. Here is the code to reproduce the problem:

import numpy as np
from mess.interop import to_pyscf
from mess import minimise, basisset, Hamiltonian, molecule
from pyscf import dft

basis_name = "6-31g"
xc_method = "pbe"


mol = molecule("water")
basis = basisset(mol, basis_name=basis_name)
H = Hamiltonian(basis, xc_method=xc_method)
E, C, sol = minimise(H)

scf_mol = to_pyscf(mol, basis_name)
s = dft.RKS(scf_mol, xc=xc_method)
s.kernel()

print(f"MESS Energy ({basis_name}/{xc_method}): {E:.6f} (hartree)")
print(f"PySCF Energy ({basis_name}/{xc_method}): {s.energy_tot():.6f} (hartree)")
print(f"MESS == PySCF ({basis_name}/{xc_method}): {np.allclose(E, s.energy_tot())}")


basis_name = "def2-SVP"

basis = basisset(mol, basis_name=basis_name)
H = Hamiltonian(basis, xc_method=xc_method)
E, C, sol = minimise(H)

scf_mol = to_pyscf(mol, basis_name)
s = dft.RKS(scf_mol, xc=xc_method)
s.kernel()


print(f"MESS Energy ({basis_name}/{xc_method}): {E:.6f} (hartree)")
print(f"PySCF Energy ({basis_name}/{xc_method}): {s.energy_tot():.6f} (hartree)")
print(f"MESS == PySCF ({basis_name}/{xc_method}): {np.allclose(E, s.energy_tot())}")

I get the following output:

MESS Energy (6-31g/pbe): -76.298987 (hartree)
PySCF Energy (6-31g/pbe): -76.298981 (hartree)
MESS == PySCF (6-31g/pbe): True

MESS Energy (def2-SVP/pbe): -75.505216 (hartree)
PySCF Energy (def2-SVP/pbe): -76.273886 (hartree)
MESS == PySCF (def2-SVP/pbe): False