Here is a list of research projects to choose from.
Project 1. Metal-peptide catalysts for CO2 activation (Heiðar)
Assoc. prof. Sigríður Suman and her group are creating metal-peptide complexes in the lab, some of which they expect to be active catalysts for CO2 activation and potentially polymerization. The peptides will consist of 3 linked amino acids.
The structural chemistry of such complexes has not been fully explored and it is not known how a metal ion will bind to different peptides which will undoubtedly be vital to the catalytic activity.
DFT methods with dispersion corrections and a continuum solvation model will be used to explore the chemistry and accurate binding energies will be calculated with hybrid functionals and/or post-HF methods.
Project aim:
Explore bindings of different peptides to first, second and third row metal ions. Find the lowest energy conformers and the lowest energy spin states.
CO2 binding will be explored with some of the complexes and it will be explored how the complex geometry affects binding energy of CO2.
Project 2. Bioinorganic modelling of an enzyme active site (Albert)
Activation and catalytic reduction of CO2 is of fundamental and practical interest. Studying the paths that nature took to perform such chemistry could reveal important insights that can be used in man-made catalysts.
Carbon monoxide dehydrogenases (CODH) are nature's biocatalysts for transforming CO to CO2. The enzymes uses H2O as a proton source and catalyse the reaction:
CO + H2O -> CO2 + 2e- + 2H+
The enzymes mainly work in this direction as CO is a highly poisonous compound that needs to be neutralized. However, the reverse reaction, transforming CO2 to CO is an important first step towards functionalization of CO2 or transformation into a fuel (e.g. MeOH) or perhaps a useful carbon product (polymer). Enzymes can work in both directions, depending on the condition.
A fascinating example of a CO dehydrogenase is the Nickel-Iron CODH. The enzyme utilises a 4-metal-sulfur cluster (1 Ni, 4 Fe) in its active site that binds CO2 or CO and transforms it with the help of H2O.
In 2007, crystal structures of the enzyme revealed an intermediate of the catalytic reaction where CO2 was bound between Ni and 1 Fe atom. Based on the crystallographic data, a simple structure-based reaction mechanism was proposed. In 2015 an even higher resolution structure became available.
There are still important unknown details about the mechanism, however.
Project aim:
Build up a simple cluster model of the enzyme active site from the published crystal structures and perform DFT calculations on the model to try and shed light on the reaction mechanism and how the enzyme performs this chemistry.
References:
http://www.sciencemag.org/content/318/5855/1461.short
http://onlinelibrary.wiley.com/doi/10.1002/anie.201501778/abstract
Project 3. Investigating DFT functional performance related to surface science (Halldór Darri, Anna)
This project is big enough to be split between 2-3 students.
Charles Campbell, experimental surface scientist at the University of Washington recently visited the department and gave a seminar.
He talked about recent experimental determinations of small molecule binding energies on clean surfaces and the comparison of the experimental binding energies to DFT-computed binding energies. Most popular DFT methods were found to perform rather poorly in this comparison, suggesting that density functional theory still has some room for improvement with regards to thermochemistry of surface science.
The work has recently been published.
http://www.sciencedirect.com/science/article/pii/S0039602815000837
One of the findings of the study is that a lot of the error of the DFT methods appears to be due to the difficulty to describe even the gas phase reactions (without metal surface) when the reactants or products contain double or triple bonds. This finding is helpful as it allows one to explore whether surface science DFT calculations can be improved by testing and studying smaller systems.
Subproject 1. Explore the gas-phase reaction energies. (Halldór Darri)
Perform basis set study and reproduce the PBE numbers from the article.
Then try different DFT methods such as hybrid functionals and double hybrid functionals to see whether they perform better for this test set.
Also try post-HF methods and see ones give better accuracy than DFT for this test set.
If time allows, we will also explore whether the reference values (theoretical) are accurate by performing CCSD(T) calculations.
Subproject 2. Modelling the simplest surfaces using cluster calculations. (Anna)
While ORCA does not support periodic boundary conditions that are usually ideal for modelling surfaces, one can instead build up larger clusters that ultimately should be large enough to give similar results as the infinite solid. This strategy is more difficult to use and require choices to be made regarding the cluster boundary and whether the neglect of long-range interactions can be corrected. The advantage of the approach is that one can employ any type of electronic structure method (while most periodic boundary conditions codes are limited to DFT).
Here we will choose one of the surfaces from the Campbell study and build up several cluster models of increasing size. We will study how large the cluster size needs to be and whether we can use point charge corrections to improve convergence. We will compare to the periodic DFT values from the study.
Single-point calculations will mostly be performed but we will also attempt geometry optimizations of one the clusters.
Project 4. Studies of a biomimetic N2-bound iron-sulfur-carbon complex
Nitrogenase is a fascinating enzyme that is able to catalytically reduce nitrogen (N2) into ammonia (NH3). The enzyme uses a complex iron-molybdenum-sulfur-carbon cofactor to perform this chemistry. Despite decades of research, the mechanism for this reaction remains unknown and even the binding site is controversial with both molybdenum and iron suggested.
In a recent 2015 article published in the journal Nature, Holland and coworkers have succeeded in creating an unusual iron-complex that features both sulfur and carbon ligands and is able to bind N2 and activate it. The existence of this complex may mean that Fe is the binding site in nitrogenase but regardless, the synthesized complex has fascinating properties.
An article published just days ago describes the synthesis of the first iron-sulfur complex with a bridging hydride. This complex may explain some of the H2 evolution chemistry of FeMoco.
Project aim:
Create a model of the iron-sulfur-carbon complex with and without N2 bound and explore its properties. Compare to the available spectroscopic data.
Create a model of the iron-sulfur-hydride complex and and explore its properties. Analyze the bonding by localized orbitals.
References:
http://www.nature.com/nature/journal/v526/n7571/full/nature15246.html
http://pubs.acs.org/doi/abs/10.1021/jacs.5b06841
Project 5. Calculating pKa and Redox potentials of nucleobases
Computing acid-base reaction or oxidation-reduction reactions in aqueous solutions is extraordinarily difficult to do. The reason is that the reactions involves charged species that will interact strongly with the surround solvent. To compute such reactions reliably thus requires both a reliably quantum mechanical method to describe the reaction energy and also a description of the solvent. Computing the reaction in the gas phase (or vacuum) is not good enough.
A recent paper by Schlegel et al. attempted to compute pKa values (related to deprotonation energies) and redox potentials (related to oxidation energies and electron affinities) and compare to available experimental data. DFT methods were used in combination with the SMD continuum solvation model. But the authors also tried to include several water molecules in the calculations (to better describe the short-range solvation effects) with some success.
Project aim:
Perform the same calculations as in the paper and reproduce the results. Check whether it is possible to improve the results by using a better electronic structure method. We will try either using a double-hybrid DFT method or the local post-HF method LPNO-pCCSD.
If a promising protocol is found it will be tried out on the phenol test set of Cramer and Truhlar (table 2 in this paper http://link.springer.com/article/10.1007/s00214-004-0577-0)
References:
http://pubs.acs.org/doi/pdfplus/10.1021/jp5088866
http://link.springer.com/article/10.1007/s00214-004-0577-0
Project 6. Peptide bond formation
The creation of peptide bonds is of fundamental interest. The peptide bond (an amide bond) is what links the amino acids together so that they form peptides and ultimate proteins. Forming the peptide bond, however, requires a catalyst. In the cell, the ribosome catalyzes this reaction. Recently peptides have been discovered that features intramolecular peptide bonds between sidechains of amino acids, so called isopeptide bonds. The interesting thing is that this new peptide bond is not made by the ribosome but is made after the formation of the peptide and is self-catalyzed by the peptide itself. Mutation studies revealed that this reaction can only occur between the amino acid lysine (Lys) and aspartate (Asp) in the peptide if another amino acid glutamate (Glu) is nearby. Elaborate QM/MM calculations were performed for this reaction of the whole peptide that revealed the detailed reaction mechanism for this reaction.
Later, a simplified gas phase mechanism was also calculated using simplified fragments but the general mechanism remained the same. These results suggest that an amide bond can spontaneously form between an amine and a carboxylic acid, if a nearby carboxylic acid acts as catalyst via a proton-shuttling mechanism. This mechanism has relevance for finding reliable catalysts for amide formation in general but a connection to origin-to-life questions also springs to mind: spontanteous formation of peptides in a soup of amino acids could be relevant for the formation of life as small peptides can act as catalysts for early biochemistry. Questions have also not been completely settled whether RNA/DNA or proteins came first although the RNA world theory remains the more popular one. See e.g. this recent discovery
Project aim:
Calculate the mechanism using different methods, including MP2, SCS-MP2, DLPNO-pCCSD and DLPNO-CCSD(T). Do a basis set convergence study at the MP2 level. See what DFT functionals give best agreement with DLPNO-CCSD(T) results.
It would be of much interest to model this reaction using molecular dynamics as only then can the entropic effects of the reaction be properly captured. Classical forcefields (that are typically used in molecular dynamics studies), however, can not describe bond-making, bond-breaking reactions. Quantum chemistry is thus necessary but most methods are too expensive to be used.
Calculate the same mechanism above using several low-cost methods (semiempirical HF and DFT as well as simplified Hartree-Fock) to see which low-cost method gives results that are the closest to the DLPNO-CCSD(T) results. Most methods are available in ORCA, some are only available in the DFTB+ or MNDO codes.
Try out AM1, PM3, HF-3c, OM3, DFTB3 methods.
References:
http://onlinelibrary.wiley.com/doi/10.1002/anie.201004340/abstract
Project 7. QM/MM calculation on molecular crystals
Most of the time, modelling molecular chemistry assumes that one can describe a molecule or a reaction well enough by calculating the single molecule in the vacuum (like we have been doing in ORCA so far). This vacuum model approximation works well when we compare to experiments that took place in the gas phase (ideally under low-pressure conditions) but it might not work as well when we compare to experiments that took place in solution or even a solid.
Newly synthesized organic or inorganic molecules are often structurally characterized using X-ray crystallography where the X-ray diffraction of molecular crystals is used to derive the 3D molecular structure of a molecule. It is a solid-state structural characterization technique and there are many other techniques where the experiment is performed on solid molecular crystals, e.g. solid-state NMR, IR spectroscopy, Mössbauer spectroscopy, single-crystal EPR spectroscopy and others.
When we compare our quantum chemistry computed structures or spectroscopic properties that have been computed in the vacuum to such solid-state data we should not expect the numbers to be in perfect agreement because we are actually not modelling the system perfectly. We know that crystal-packing effects can affect both structures and molecular properties.
A QM/MM based method (http://pubs.acs.org/doi/abs/10.1021/ct200824r) was developed by Ragnar Bjornsson a few years ago that allows one to model such solid-state effects directly. The method is implemented in Chemshell but uses ORCA to perform the quantum chemistry. The method generally works well but is still being developed and it requires further testing on more molecular systems.
Project aim:
Explore how well the QM/MM method works to describe the following molecular crystals.
1. The unusual triple/single to double/double transition in zirconium complex crystals:
http://pubs.acs.org/doi/abs/10.1021/om500678p
2. ?
Project 8. Benchmarking electronic structure methods (Hrólfur)
Many different approximations are available in density functional theory but also in wavefunction theory. While it is possible to compare methods based on basic physical constraints they describe (HF is for example exact for 1-electron systems while most DFT methods are not), typically the only practical way to compare different methods is to perform a benchmarking study. That involves testing how well methods perform for a certain class of molecules or a property type and compare either to experiment or highly accurate ab initio data. This allows one to compare the accuracy and computational cost of different methods and determine approximate error bars that one would expect in calculations of similar compounds as in the benchmarking test set.
Grimme et al. pointed out in 2009 that often benchmarking studies are too biased towards the goals of the researchers. They suggested the use of a "mindless" benchmarking set that removes bias as much as possible. http://pubs.acs.org/doi/abs/10.1021/ct800511q This test set is regarded as one of the most difficult test set for quantum chemistry methods and it can be seen that many DFT methods display large errors. http://www.thch.uni-bonn.de/tc/downloads/GMTKN/GMTKN30/MB08-165.html
Another very difficult test set for DFT is the self-interaction-error test set SIE11: http://www.thch.uni-bonn.de/tc/downloads/GMTKN/GMTKN30/SIE11.html
Reliable benchmarking data for transition metal chemistry is sadly considerably less. Donald Truhlar, however, has compiled a databases of 3d-metal metal-ligand/metal-metal bond energies: SRMBE13 and a database of multi-reference bond energies: MRBE10. There is also the DC9/12 database of difficult cases for DFT. These databases are very difficult for most DFT methods.
Project aim:
Test out recently developed post-HF methods and DFT methods that have not been tried out before for the Grimme and Truhlar databases. It will be especially interesting to see how SIC-DFT methods perform for these databases.
The Truhlar databases are quite small and all information is available on the websites.
The SIE test set is not too large and the inputfile can easily be created manually.
As the MB08-165 set is rather large, many inputfiles need to be created. This will require some simple scripting (programming). I will help with this part.
Try out the following methods in ORCA:
SCS-MP3, pCCSD/2a method using both HF and DFT orbitals, DSD-PBEP86, PBE with HF orbitals, HF with PBE orbitals
Methods only available in Erkale:
Self-interacted corrected PBE and PW91 by Hannes Jónsson et al.
Project 9. CO2 reduction on modified Nickel-Cyclam catalysts (Björk)
CO2 reduction is a hot topic. A simple but surprisingly active catalyst is the Ni(cyclam) complex. A recent theoretical study has explored the reaction mechanism in detail (http://pubs.acs.org/doi/abs/10.1021/ic500829p).
An even more recent experimental study immobilised the catalyst onto a metal oxide surface and also found that attaching a carboxylic acid group to the cyclam framework dramatically increased the activity in acidic solution and maintained CO2 selectivity. This further confirms the importance of a "local proton source" for enhanced electrocatalytic activity as seen before for proton reduction and CO2 reduction.
Project aim:
Using the same computational protocol, reproduce the main results of the original computational study and then use the same protocol to calculate the same steps but for the modified RCOOH-added cyclam ligand.
Can the increased activity of the modified ligand as found by experiment be explained by your calculations?
What reaction step is most affected by the ligand modification?
References:
http://pubs.acs.org/doi/abs/10.1021/ic500829p
http://pubs.rsc.org/en/content/articlelanding/2015/cp/c4cp04871g#!divAbstract