New publication in Nature Catalysis

| categories: news, publication | tags: | View Comments

Machine learning (ML) is impacting many fields, including catalysis. In this comment, I briefly discuss the major directions that ML is influencing the field of catalysis, along with some outlook on future directions. There were strict word and reference limits, so apologies in advance if I left out your work!

@article{kitchin-2018-machin-learn-catal,
  author =       {John R. Kitchin},
  title =        {Machine Learning in Catalysis},
  journal =      {Nature Catalysis},
  volume =       1,
  number =       4,
  pages =        {230-232},
  year =         2018,
  doi =          {10.1038/s41929-018-0056-y},
  url =          {https://doi.org/10.1038/s41929-018-0056-y},
  DATE_ADDED =   {Mon Apr 16 12:50:43 2018},
}

You can see a read-only version of the paper here: https://rdcu.be/LGrM

Copyright (C) 2018 by John Kitchin. See the License for information about copying.

org-mode source

Org-mode version = 9.1.6

Read and Post Comments

Zhitao Guo receives the 2017-2018 James C. Meade Fellowship in Chemical Engineering

| categories: news | tags: | View Comments

The James C. Meade Fellowship was made possible by a generous donation by James Meade. This will help support Zhitao during his research this year. Zhitao is a first year PhD student who is co-advised by Andy Gellman and myself (John Kitchin), and is working on segregation in ternary alloy thin films.

Zhitao joined us from Tsinghua University in Beijing, China, where he studied chemical engineering and double majored in economics.

Congratulations Zhitao!

Copyright (C) 2018 by John Kitchin. See the License for information about copying.

org-mode source

Org-mode version = 9.1.6

Read and Post Comments

New publication in Catalysis Today

| categories: news, publication | tags: | View Comments

In this paper we continue our exploration of using high-dimensional neural networks (NN) to model metal surface properties. Our first work started with modeling Au in a variety of structures using ReaxFF and a NN boes-2016-neural-networ. We then modeled atomic oxygen adsorbates on a Pd (111) surface boes-2017-neural-networ, and segregation of an Au-Pd alloy surface boes-2017-model-segreg. Our goal throughout this work has been to systematically build up complexity in the systems we are modeling, and to explore the limitations of these potentials for modeling surfaces. This current work happened in parallel with those works, and focused on modeling Pd adatom diffusion on Pd(111) surfaces. We show another example of how to train a neural network, and then to use it model the temperature dependent diffusion of adatoms on a metal surface using molecular dynamics and Arrhenius analysis.

@article{gao-2018-model-pallad,
  author =       {Tianyu Gao and John R. Kitchin},
  title =        {Modeling Palladium Surfaces With Density Functional Theory,
                  Neural Networks and Molecular Dynamics},
  journal =      {Catalysis Today},
  year =         2018,
  doi =          {10.1016/j.cattod.2018.03.045},
  url =          {https://doi.org/10.1016/j.cattod.2018.03.045},
  DATE_ADDED =   {Sun Apr 1 18:47:55 2018},
}

Bibliography

Copyright (C) 2018 by John Kitchin. See the License for information about copying.

org-mode source

Org-mode version = 9.1.6

Read and Post Comments

New publication in Topics in Catalysis

| categories: news, publication | tags: | View Comments

Single atom alloys are alloys in the extreme dilute limit, where single atoms of a reactive metal are surrounded by comparatively unreactive metals. This makes the single reactive atoms like single atom sites where reactions can occur. These sites are interesting because they are metallic, but their electronic structure is different than the atoms in more concentrated alloys. This means there is the opportunity for different, perhaps better catalytic performance for the single atom alloys. In this paper, we studied the electronic structure and some representative reaction pathways on a series of single atom alloy surfaces.

@article{Thirumalai2018,
  author =       "Thirumalai, Hari and Kitchin, John R.",
  title =        "Investigating the Reactivity of Single Atom Alloys Using
                  Density Functional Theory",
  journal =      "Topics in Catalysis",
  year =         "2018",
  month =        "Jan",
  day =          "25",
  abstract =     "Single atom alloys are gaining importance as atom-efficient
                  catalysts which can be extremely selective and active towards
                  the formation of desired products. They possess such desirable
                  characteristics because of the presence of a highly reactive
                  single atom in a less reactive host surface. In this work, we
                  calculated the electronic structure of several representative
                  single atom alloys. We examined single atom alloys of gold,
                  silver and copper doped with single atoms of platinum,
                  palladium, iridium, rhodium and nickel in the context of the
                  d-band model of Hammer and N{\o}rskov. The reactivity of these
                  alloys was probed through the dissociation of water and nitric
                  oxide and the hydrogenation of acetylene to ethylene. We
                  observed that these alloys exhibit a sharp peak in their atom
                  projected d-band density of states, which we hypothesize could
                  be the cause of high surface reactivity. We found that the
                  d-band centers and d-band widths of these systems correlated
                  linearly as with other alloys, but that the energy of
                  adsorption of a hydrogen atom on these surfaces could not be
                  correlated with the d-band center, or the average reactivity
                  of the surface. Finally, the single atom alloys, with the
                  exception of copper--palladium showed good catalytic behavior
                  by activating the reactant molecules more strongly than the
                  bulk atom behavior and showing favorable reaction pathways on
                  the free energy diagrams for the reactions investigated.",
  issn =         "1572-9028",
  doi =          "10.1007/s11244-018-0899-0",
  url =          "https://doi.org/10.1007/s11244-018-0899-0"
}

Copyright (C) 2018 by John Kitchin. See the License for information about copying.

org-mode source

Org-mode version = 9.1.6

Read and Post Comments

New publication in Molecular Simulation

| categories: news, publication | tags: | View Comments

This paper is our latest work using neural networks in molecular simulation. In this work, we build a Behler-Parinello neural network potential of bulk zirconia. The potential can describe several polymorphs of zirconia, as well as oxygen vacancy defect formation energies and diffusion barriers. We show that we can use the potential to model oxygen vacancy diffusion using molecular dynamics at different temperatures, and to use that data to estimate the effective diffusion activation energy. This is further evidence of the general utility of the neural network-based potential for molecular simulations with DFT accuracy.

@article{wang-2018-densit-funct,
  author =       {Chen Wang and Akshay Tharval and John R. Kitchin},
  title =        {A Density Functional Theory Parameterised Neural Network Model
                  of Zirconia},
  journal =      {Molecular Simulation},
  volume =       0,
  number =       0,
  pages =        {1-8},
  year =         2018,
  doi =          {10.1080/08927022.2017.1420185},
  url =          {https://doi.org/10.1080/08927022.2017.1420185},
  eprint =       { https://doi.org/10.1080/08927022.2017.1420185 },
  publisher =    {Taylor \& Francis},
}

Copyright (C) 2018 by John Kitchin. See the License for information about copying.

org-mode source

Org-mode version = 9.1.5

Read and Post Comments

Next Page ยป