Perovskite Electrocatalyst for CO2 Reduction Design

Theoretical design of efficient perovskite electrocatalyst for CO₂ reduction

Summary

This project aims to engineer perovskite materials as efficient electrocatalysts for CO₂ to fuel conversion. Perovskite are appealing candidates because of their wide ranging and complex electronic structures. There is a scope to break some of the limitations of metallic catalysts to come up with new efficient electro catalyst. The endless varieties of electronic properties oxides’ show are truly astounding. We would like to search this large materials space systematically for possible catalyst with improved activity for CO₂ reduction. However, they should also be active, stable and conductive at relevant potentials to meet application targets. Identification and establishment of design principles for efficient oxide catalyst for CO₂ reduction will mark the scientific part of this project. Efforts to be made for predicting molecular pathway of CO₂ reduction reactions and develop unified search criterion like descriptors regarding these reactions. Then this knowledge to be applied for high throughput computational search for best perovskite electrocatalysts. In this project we plan to utilize the latest simulation methodologies developed based on density functional theory (DFT) towards understanding the molecular mechanism of CO₂ to fuel conversion on oxide surfaces. Further on, we will explore kinetic barriers using nudged elastic band (NEB) method to come up with possible efficient electro catalyst.

Get Help With Your Essay

If you need assistance with writing your essay, our professional essay writing service is here to help!
Find out more about our Essay Writing Service

Introduction and perspective on impact

Global energy consumption will increase manifold in a few decades as larger fraction of world population achieves higher quality of life. This demand could be met from fossil fuels, particularly coal. However, in recent time, carbon dioxide level in the air has reached the highest (>400 ppm) of the last 20 million years, causing radical and largely unpredictable changes in the environment. Thus to maintain sustainability for human kind, it will require invention, development, and deployment of carbon-neutral energy production at a scale larger than, the entire energy supply in modern civilization. To support high penetration of renewable energy sources such as solar and wind power it would require a commensurate increase in energy storage capacity to integrate them into the electrical power grid. This is to facilitate reliability in power delivery by smoothing out the large fluctuations.

Electrochemical conversion of CO₂ and H₂O into liquid fuels is the holy grail where high density renewable energy storage and CO₂ capture meet each other. However, no electro-catalyst known to man can catalyse this reaction efficiently. Copper (Cu) is the only metal with considerable activity, but its efficiency and selectivity for liquid fuels are far too low for practical use. Ruthenium dioxide can convert CO₂ to methanol at a low overpotential, however, the process is rather unselective and ruthenium is extremely scarce. It is of utmost importance to discover efficient electro catalyst with low over potential, high faraday efficiency and most importantly, made of earth abundant elements. Recent success obtained in photo electrocatalytic reduction of CO₂ on SrTiO₃ surfaces and electrocatalytic reduction of the same on Lanthanum Strontium Cuprate perovskite leads us towards selecting oxides especially perovskite as the most promising class of materials to study.

The potential phase space of (mixed metal at A and B site) perovskite materials is very large, thus experimentally testing all compounds is not practically feasible, but has to be narrowed down through computational screening. Simulation science has taken up a key role in development of new energy materials in the last couple of years, through computation of materials properties, which are difficult to measure experimentally. Development in computer power has enabled large-scale materials screening and design at atomistic scale. Within catalyst design, simulations can identify restrictions on catalyst activity and selectivity caused by scaling laws, and these laws enable efficient prediction of activity or selectivity for possible new catalysts.

Background

Hori did seminal work on electrochemical reduction of CO2 on pure metals. Cu is the only metal that does not desorb CO and can uniquely reduce CO2 to significant quantities of hydrocarbons (mainly CH4 and C2H4). Reduction of CO2 on Cu is accompanied with a very high over-potential that hinders this reaction from being energy efficient. Recently, a mechanism that explains copper’s unique ability in reducing CO2 to hydrocarbons and the origin of the high over-potential for the reduction of CO2 was identified by DFT calculations in conjunction with computational hydrogen electrode (CHE) model. Since electrochemical CO2 reduction to methane is an eight electron-proton transfer step that has seven intermediates, finding the best catalyst in principle demands understanding of a seven-dimensional molecule surface interaction space. Fortunately, the binding energies of carbon bounded species and oxygen bounded species scale with the binding energies of CO and OH, respectively. These correlations reduce the dimensionality to two binding energies but make it difficult to change the binding energies independently. Based on different reaction pathways and scaling relations Peterson and co-workers constructed volcano plots for different metals. It was shown that regardless of the reaction pathway, changing the metal surface marginally changes the over-potential. This helped to move the focus on other class of catalysts e,g, rutile oxides (Ru/Ir/Ti) can catalyze the conversion of CO2 to alcohols. However, very little is known about the reduction of CO2 to alcohols on oxide electrocatalysts. As the binding energies of OH/CO vary much widely on oxides than metals, it is possible to have different pathways and thermodynamic limiting steps on oxide surfaces than metallic ones. That makes template based computational search much more challenging, at the same time opening up possibilities of adsorbate-surface binding energies away from the established scaling laws.

In general, there are three criteria that should be fulfilled by a newly proposed catalyst material:

1. The catalyst should have high selectivity towards desired product
2. It should have high energy efficiency, i.e. low reduction over-potential
3. It should be stable at potentials of interest so that the activity does not degrade over time
4. It should have sufficient electronic/polaronic conductivity

Research plan

For the thermodynamic pathway of the reactions, computational hydrogen electrode model will be followed to calculate the potential dependent reaction free energies from density functional theory based calculations using BEEF-vdW functional and PAW method as implemented in VASP. Corrections for zero point energy, heat capacity, entropic contribution and other energy correction for free molecules will be taken into account. Usage of BEEF-vdW functional will enable the estimation of errors in first principles calculations and describe proper long range van der Waals interaction between adsorbates and surfaces. Statistical tools will be used to calculate corrections from vibrational modes of the adsorbates. In addition, to describe correctly the electronic structure of late transition and rare-earth metals, Hubbard U correction method will be employed as and when required. To know atomic structure of the catalyst surface, which is key to these calculations, potential dependent surface Pourbaix diagrams will be constructed. Kinetic barriers for individual reaction steps will be searched with the climbing image nudged elastic band (CI-NEB) method. This method lets us find saddle points and minimum energy paths between two atomic configurations and works by optimizing a number of intermediate images along the reaction path.

Using this methodology to study CO2 reduction over a handful of well-known perovskite materials, we will be able to establish activity descriptors for favorable catalysts. Through Brønsted–Evans–Polanyi relation between the activation energy and the reaction energy extends scaling laws to kinetic barriers as well. Using thermodynamic and kinetic scaling laws, it will be feasible to define most critical descriptors of the many electron reactions. These descriptors along with selectivity based parameters (e.g. suppressed hydrogen evolution) will be used to screen through a really large phase space of perovskite structures constructed in a 2x2x2 supercell (40 atoms) by using a large number of different elements in A/B or anion site as well as vacancies which are common in many perovskite materials. Significant fraction of these hypothetical structures will be discarded through simple rules like oxidation number sum, Goldstein’s rule and Valence Bond models. In the screening process, the stability of possible structures are assessed using an accurate scheme of comparing the total energy of each compound to a pool of reference systems using a linear programming algorithm, to determine whether the material is stable or not. The further level of screening will include looking for materials with small or no bandgap using GLLB-sc functional. This is a crude approximation for screening purpose. For few selected structures, other conduction mechanism such as quantum tunneling or polaron hopping will be studied in more detail using Marcus theory for polaron hopping and non-equilibrium Green function based transport modelling.

Even with the large reduction in search space through simple rule based screening, it will be impossible to perform DFT calculation for all of the possible structure. A genetic algorithm based search will enable us to effectively get the fittest candidates with existing computational resources. The parameters for the fit function will be similar as discussed above. Concepts of mutation and crossover will be used for quick searching.

The project will be carried out in collaboration with experimentalist from DTU Energy Conversion (Prof. Nini Pryds and his group) and DTU Physics (Prof. Ib Chorkendorff and group). This will enable rapid experimental validation of predicted materials as effective CO2 reduction electrocatalyst.

Work package and milestones

WP1: Establish reaction mechanism (Jan 2015 – Aug 2015)

1.1 Calculate atomistic structure of SrTiO3 and NaNbO3 (100) and (110) surfaces from surface Pourbaix diagram

1.2 Study wide variety adsorbates to confirm reaction pathway to alkane and alcohols

1.3 Estimate kinetic barriers for the reaction paths

WP2: Search for Descriptor (Sept 2015 – Feb 2016)

2.1 Calculate thermodynamic and kinetic barrier for CO2 reduction reactions for larger number (~50) of well-known perovskite.

2.2 Study these barriers for identifying best descriptors for the reactions

2.3 Do micro-kinetic modelling of the system considering different final products both carbonaceous and hydrogen based of the descriptors of reactions, to define region of selectivity and low over-potential requirement.

WP2: High throughput computing based catalyst search and validation (Mar 2015 – Dec 2016)

3.1 Setup Computational infrastructure (e.g. software framework working in unison) required for screening methodology over billions of structures. The layers in the screening (rule based and calculation based) as well as the genetic algorithm based evolutionary search tool has to work in tandem.

3.2 Perform the large scale search for optimum binding energies, kinetic barrier, conductivity and selectivity through GA based exploration of the phase space. The fit criterion for a specific product to be defined based on the results of the micro-kinetic modelling.

3.3 Synthesize and run experiments for measuring activity of a handful of selected candidates for different end products

Resource requirement

The scale of the computational search and complexity requires tier0 type supercomputing infrastructure. I expect to be able to use ~8 million cpu hours in the DTU HPC resource – NIFLHEIM. Applications have also been made for another 20 million cpu hours under the European supercomputing program – PRACE.

Scientific dissemination

The fundamental insight developed, catalyst predicted and validated throughout this project will generate utmost interest in the catalysis for sustainable energy field internationally. Thus findings will be published in peer-reviewed journals with a high visibility. Such publications can be expected during 4Q of 2015 and 2016. Preliminary results will be presented at relevant conferences within the fields of electrocatalysis, computational electrochemistry and surface science. Besides contributing to fundamental insight the project is focused on specific catalyst design and it is therefore an objective that one or more patents will be filed for at the end of the period.