Skip to main content

New methods for new materials

Periodic Reporting for period 1 - NEW4NEW (New methods for new materials)

Reporting period: 2015-04-01 to 2017-03-31

Computational modelling of molecules and materials has had a great impact on understanding experimental observations.
A prominent example are chemical reactions where modelling allows one to follow the motion of atoms and to obtain
a detailed insight into the process. However, to model chemical reactions, electrons need to be described using quantum mechanics.
Moreover, to simulate a given process reliably, both strong chemical bonds and weak van der Waals interactions need to be described
with a high accuracy. This is a great challenge for the methods that are currently being used, such as the density functional theory (DFT).
One class of systems where this issue is particularly important are porous materials, such as zeolites.
Zeolites are important industrial catalysts and also perspective materials for gas separation.
During the catalytic process, molecules interact first weakly with the zeolite before chemical reaction takes place.
Therefore, if we want to further improve the function of porous materials or develop new ones with desired chemical activity,
we need to be able to model reliably both strong and weak interactions.
The development of methods that could be used to obtain reliable data for studies of reactions in zeolites is the main goal
of the current project. This goal will be achieved by developing and implementing most promising DFT approximations for the
treatment of strong interactions and combining them with state-of-the-art methods for the treatment of weak forces.
Moreover, understanding the strong and weak points of various methods when applied to processes in zeolites will help us
to identify the scheme that could serve as the workhorse method for the next generation of studies.
During the project we implemented and tested different quantum chemistry methods
with the goal of identifying a promising scheme for the treatment of adsorption
and reactions of molecules in zeolites. Namely, we first added the necessary
computer routines to perform calculations with the so-called mid-range DFT
functionals into the VASP code. Unfortunately, tests with different such
functionals showed no clear winner and method that would lead to substantially
improved results.
In further work, we developed a test set for the adsorption of small molecules in
zeolites. For this dataset we obtained adsorption energies with state-of-the-art
dispersion corrected DFT functionals, the random phase approximation with singles
excitations scheme (RPA+SE), that we developed and implemented right before the
start of the MSC grant, and with the MP2 approach, which is usually considered to
offer reference quality results for adsorption in zeolites. We found that the
RPA+SE method gives results very similar to those obtained with MP2. To identify,
which one of these methods is actually better, we created an additional test
set for finite clusters cut out of the material. This allowed our collaborators
to obtain reference quality adsorption energies with the coupled cluster scheme.
These data show that the RPA with singles outperforms MP2 for the problem.
Its low cost makes it thus a very promising method for the treatment of adsorption
in zeolites. In fact, we applied the scheme to zeolite where recent experimental
data for adsorption of small molecules are available. We found a close
agreement with the reference and much improved performance compared to any other
scheme that is currently available.
Encouraged by the high quality of the results offered by the RPA with singles,
we applied it to several other problems. First, to molecular systems, where
accurate methods are also highly desirable. Moreover, they are highly relevant
for adsorption as the results reflect the accuracy of the description of the
interactions between the molecules adsorbing in the material. We again found
the best performance among all the methods that are currently available for the
treatment of molecular solids within periodic boundary conditions. Second, we
applied the scheme to systems relevant to catalysis. Namely, to the problem of
finding stable positions of extraframework ions in the zeolite matrix (siting).
Finally, we implemented a new version of the van der Waals density functional
scheme that takes explicitly into account the spin-polarisation of the material.
We also took part in the implementation of the low scaling version of the GW
approach for the calculation of electronic levels. This GW scheme can be applied
to systems with the typical cell size of zeolites.

So far, we have published two papers from the project, one more is close to being submitted.
We will publish at least one with the results for adsorption energies and at least one with the siting
results. The problems regarding the implementation of screened functionals will be also
published, once a clear conclusion can be made.

The results and project were presented at the following conferences:
* DPG Regensburg, March 2016 (oral presentation)
* JCP Frontiers, September 2016 (poster)
and following seminars:
* Faculty of Natural Sciences, Prague, April 2015
* Faculty of Mathematics and Physics, Prague, November 2015
* University College London, August 2016
* Institute of Physics, Prague, November 2016
The main step beyond the state of the art was made when we identified the RPA
with singles approach as a scheme that not only offers a high accuracy for
adsorption energies, but also gives more consistent results than MP2.
All this at a much lower computational cost. These factors mean that RPA
with singles is very likely to become widely used to obtain high quality
adsorption energies not only for zeolites, but also for other porous materials
or even for adsorption on solid surfaces. In fact, I will further explore this
route within a new project. If the scheme really fulfils its promise, it could
easily become the standard method to be used for studies in catalysis, surface
science, condensed matter, and related disciplines. The impact would be truly
large. Moreover, the larger reliability of the results would mean that we could
focus on understanding the science or experimental observations and wouldn't have
to waste our time on double-checking the correctness of the results.
Therefore, the methods can help to speed up the development of new catalysts
or new materials that would reduce our carbon footprint.
Propane adsorbed in zeolite.