Accurate force-fields for small molecules and ions

In this project accurate forcefields are being developed for small molecules, such as water, organic solvents, and ions. The primary advantage compared to DFT /CMPD calculations is the much larger speed, allowing for studies of much larger chemical systems and processes covering substantially larger timescales. Moreover, the energy can in principle be more accurately calculated than with DFT, albeit with substantially less generality. The speed difference between traditional force-fields and DFT calculations is of the order 10'000-1'000'000. Obviously, for more accurate force-fields, which need substantially more complicated functions than the traditional forcefields, the calculation speed is slower, although there is a large room for improvements, computational-speed-wise.


This project

In this project the strategy is basically as follows:

  1. Several ab initio (DFT, hybrid DFT, but typically WF-based, such as MP2, CC, using different basis sets) calculations are performed for small model systems to determine the computational level required to describe the basic interaction energy terms accurately enough for the problem at hand.

  2. Several interaction functions including various types and number of many-body terms are tested. This step is the most critical. The functional form will determine the possible accuracy and speed of the final forcefield calculations. The success of this step depends largely on how much is known about the constituent interactions in the system. When a successful function has been obtained it is also possible to draw conclusions about the relative importance of the various interactions, such as whether van der Waals interactions are very important, if electrostatics or covalent bonding are dominating.

  3. Using the knowledge in the previous steps, a large number (thousands to tens of thousands) of ab initio calculations are performed to determine the parameters required in the forcefield.

  4. The forcefield parameters are fitted to the results of the thousands of ab initio calculations.

  5. The forcefield is verified by comparison with more ab initio calculations, for geometries not included in the training set. If unsuccessful, steps 2-4 above are repeated.

  6. Classical MD-simulations can now be performed using the new forcefield.