Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

Computational High Throughput Searches for Efficient Catalysts

Saravanan, Karthikeyan (2019) Computational High Throughput Searches for Efficient Catalysts. Doctoral Dissertation, University of Pittsburgh. (Unpublished)

This is the latest version of this item.

[img] PDF
Restricted to University of Pittsburgh users only until 24 January 2021.

Download (12MB) | Request a Copy

Abstract

Several screening studies identifying new catalysts for different reactions have been reported over the past decade. Almost all of them employ Kohn-Sham density functional theory (KS-DFT) and thermodynamic descriptors to screen for new catalysts. Though usually considered reliable for descriptor-based analyses, KS-DFT calculations are computationally expensive and intractable for use when screening across the full chemical space of all possible alloy materials. In order to accelerate screening of catalysts, we employ a perturbation theory model, ”Computational Alchemy” to approximate KS-DFT energies at a fraction of the computational cost. In this thesis, we discuss about how computational alchemy and machine learning can be used to reliably screen for efficient catalysts for a wide range of electrochemical processes based on thermodynamic activity descriptors.

As a first step, we assess the promise of computational alchemy in predicting binding energies thousands of times faster than DFT. We identify distinct cases where alchemy performs significantly worse, indicating areas where modeling improvements are needed. We find that alchemical estimates yield binding energies within 0.1 eV of DFT values for a wide range of adsorbates. Largest errors (≈ 0.4 eV) were observed when Alchemy predicted BEs of the adsorbates on alloys that were obtained by changing large number of atoms and to a large change in nuclear charge. Using a Machine learning approach, the errors between Alchemy and DFT were corrected. Our results suggest that computational alchemy with machine learning is a very promising tool that warrants further consideration for high-throughput screening of heterogeneous catalysts.


Share

Citation/Export:
Social Networking:
Share |

Details

Item Type: University of Pittsburgh ETD
Status: Unpublished
Creators/Authors:
CreatorsEmailPitt UsernameORCID
Saravanan, Karthikeyankas389@pitt.edukas389
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairKeith, Johnjakeith@pitt.edujakeith
Committee MemberMcKone, Jamesjmckone@pitt.edujmckone
Committee MemberWang, Guofengguw8@pitt.eduguw8
Committee MemberSaidi, Wissamalsaidi@pitt.edualsaidi
Committee MemberHutchison, Geoffgeoffh@pitt.edugeoffh
Date: 24 January 2019
Date Type: Publication
Defense Date: 19 November 2018
Approval Date: 24 January 2019
Submission Date: 30 October 2018
Access Restriction: 2 year -- Restrict access to University of Pittsburgh for a period of 2 years.
Number of Pages: 123
Institution: University of Pittsburgh
Schools and Programs: Swanson School of Engineering > Chemical Engineering
Degree: PhD - Doctor of Philosophy
Thesis Type: Doctoral Dissertation
Refereed: Yes
Uncontrolled Keywords: Computational Chemistry, DFT, Machine Learning, Alchemy
Date Deposited: 24 Jan 2019 15:46
Last Modified: 24 Jan 2019 15:46
URI: http://d-scholarship.pitt.edu/id/eprint/35610

Available Versions of this Item

  • Computational High Throughput Searches for Efficient Catalysts. (deposited 24 Jan 2019 15:46) [Currently Displayed]

Metrics

Monthly Views for the past 3 years

Plum Analytics


Actions (login required)

View Item View Item