Research

New numerical strategies and algorithms for multiscale methods

  • Domain decomposition methods applied to continuum solvation models
  • Development of linear-scaling strategies for polarizable embedding models

Polarizable embedding models, either atomistic such as QM/MM or based on a continuum description of the environment, such as the polarizable continuum model (PCM) or the conductor-like screening model (COSMO) can become very demanding from a computational point of view when applied to large or very large systems – especially when they are both applied at the same time. This research line aims at creating new fast algorithm and state-of-the-art implementations of such methods in order to extend their range of applicability to large, complex systems.

The main goal of this research is to be able to compute molecular properties (spectroscopic observable, dynamical properties) of large, complex molecular systems, such as a chromophore in solution or embedded in a biological matrix, with particular interest to chiroptical spectroscopies.

Things done so far:

  • ddCOSMO, a domain-decomposition based linear scaling implementation of COSMO.
    • You can read more on ddCOSMO here
    • An open-source, freely available implementation of ddCOSMO can be found here
    • A software review paper describing the ddCOSMO implementation and how to interface it with existing QM or MD codes can be found here
  • A fully linear scaling polarizable embedding implementation based on the MMPol force field and on ddCOSMO

New things going on:

  • Domain decomposition for the polarizable continuum model
  • Domain decomposition models for metal nanoparticles
  • Higher order properties, including magnetic and mixed properties, with a QM/ddCOSMO and with a multiscale QM/MMPol/ddCOSMO approach
  • Linear scaling implementation of a polarizable embedding based on the AMOEBA force field
  • Ab-initio QM/MM Molecular Dynamics with a polarizable embeddng

Software development:

The multiscale models developed e been implemented as stand-alone, opensource libraries or as parts of existing academic or commercial codes.
The reference codes used for these developments are Gaussian 16 for QM calculations and Tinker (and Tinker-HP) for classical MD simulations. An interface between these two codes is under active development as a driver to perform ab-initio QM/MM MD simulations.

Quantum electronic structure theory

  • Development of efficient algorithms and implementations for the complete active space self-consistent field method (CASSCF)
  • Spin-free relativistic quantum chemistry
  • Reduced scale and compression techniques for electronic structure methods

Things done so far:

  • Quadratically convergent implementation of CASSCF for either the Coulomb or the Spin Free Dirac-Columb Hamiltonian
  • Coupling such an implementation with GeCCo, an internally-contracted Multireference Coupled Cluster code by M. Hanauer and A. K√∂hn
  • Cholesky Decomposition based, reduced scaling quadratically convergent SCF implementation

New things going on:

  • Quadratically convergent CASSCF (either state-specific and state-averaged) with Cholesky Decomposition (CD)
  • CD-CASSCF gradients
  • CD-CCSD(T) energy and gradients

Software development:

The reference code used for these implementations is the CFOUR suite of programs