Title: Dispersion without many-body density distortion: Assessment on atoms and small molecules

Abstract: We have implemented and tested the method we have recently proposed [J. Phys. Chem. Lett. 10, 1537 (2019)] to treat dispersion interactions, which is derived from a supramolecular wavefunction constrained to leave the diagonal of the many-body density matrix of each monomer unchanged. The corresponding variational optimization leads to expressions for the dispersion coefficients in terms of the ground-state pair densities of the isolated monomers only, which provides a framework to build new approximations without the need for polarizabilities or virtual orbitals. The question we want to answer here is how accurate this ``fixed diagonal matrices'' (FDM) method can be for isotropic and anisotropic $C_6$ dispersion coefficients when using monomer pair densities from different levels of theory, namely Hartree-Fock, MP2 and CCSD. For closed-shell systems, FDM with CCSD monomer pair densities yields the best results, with a mean average percent error for isotropic $C_6$ dispersion coefficients of about 7\% and a maximum absolute error within 18\%. The accuracy for anisotropic dispersion coefficients with FDM on top of CCSD ground states is found to be similar. The performance for open shell systems is less satisfactory, with CCSD pair densities not always providing the best result. In the present implementation, the computational cost on top of the monomer's ground-state calculations is $\mathcal{O}(N^4)$.