We announced the Open Catalyst Project and released the first dataset from the effort! Check out opencatalystproject.org!
The Army Research Office awards a new project on ML for methanol chemistry on oxides. On to the science!
Our collaborative experiment/theory work with Ted Sargent (Toronto) on CO2 reduction catalysis is published in Nature!
Zack received a 3M Non-Tenured Faculty Award to support his research over the next three years.
Two new post-docs have joined: Javi will be working on ML methods applied to oxygen chemistry, and Rajesh will be working on our ARPA-E project on surface segregation. See their bio’s in the member section, and looking forward to the great science coming!
Our group was awarded a collaborative grant from the Sloan Foundation with Shoji Hall and Iryna Zenyuk on zinc air battery development. Pari will be leading!
Seoin Back, the first post-doc in the group, received an offer to join the Department of Chemical and Biomolecular Engineering at Sogang University in Korea. Congrats to Seoin, and best of luck in his new adventures!
Pari receives the Michael S. Phillips and Caroline B. Huang fellowship in energy and energy efficiency. Congrats Pari!
We are extremely grateful for a 28,000,000 service unit allocation on the Cori supercomputer at DOE NERSC. NERSC also hosts our high-throughput computational workflows and services, so it’s great to have computing resources too! This is up from 5.5M service units in 2018, and represents something like $1,000,000 in computing support if purchased on community cloud resources.
Our group is extraordinarily grateful to receive a grant of $900,000 from a foundation that wishes to remain anonymous to support our work in machine learning for catalysis. This grant will allow us to continue to invest in exploratory research for new ML methods and applications in this field, and means that we will be recruiting more PhD/post-docs for new projects in 2020!
Our group will work with Amir Barati Farimani (CMU MechE), Andy Gellman (CMU ChE) for a new ARPA-E project! We will use deep reinforcement learning and atomistic machine learning potentials to predict catalyst surface stability under reaction conditions. Current methods for determining the metastability of bifunctional and complex surfaces undergoing reaction are difficult and expensive. Carnegie Mellon’s technology will enable stability analysis in both traditional catalysts and new classes of materials, including those used in tribology, corrosion-resistant alloys, additive manufacturing, and battery materials. https://arpa-e.energy.gov/?q=news-item/were-not-saying-were-the-avengers-but
We are very fortunate to receive funding from the ACS Petroleum Research Fund (PRF), to support a new project starting in 2020 on the accelerated discovery of direct-conversion catalysts.
Jun and Pari passed their third year proposals. Jun will be working on machine learning potentials and deep reinforcement learning, while Pari will be applying machine learning methods to the oxygen reduction reaction and nanoparticle stability.
Zack participated in the 2020 Advanced Energy Storage Scialog conference run by RCSA. His team with Shoji Hall (JHU) and Iryna Zenyuk (UCI) was selected for 1-year of funding to pursue bifunctional intermetallics for zinc-air batteries.
We are part of two new projects funded by the DOE Basic Energy Sciences office on the use of data science and machine learning in catalysis. See the full list of projects here: https://www.energy.gov/articles/department-energy-provide-276-million-data-science-research-chemical-and-materials-sciences
We will be working with Andy Gellman and John Kitchin (CMU) as well as Jingguang Chen (Columbia / Brookhaven National Lab) on a new project for designing compositionally-diverse propylene epoxidation catalysts sponsored by the NSF ‘Designing Materials to Revolutionize our Engineering Future (DMREF)’ program. On to the science! See more details about the project here: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1921946&HistoricalAwards=false
Muhammed Shuaibi, the fourth PhD student in the group, passed his first-year qualifying exam, discussing his work on augmenting machine learning potentials with physics-based potentials as part of our DOE CCS project on exascale computing. Next step Thesis Proposal!
In July two papers demonstrating graph convolution methods in catalysis, one for materials for the oxygen evolution reaction, and a methods paper for small molecule adsorbates were published in ACS Catalysis and JPC Letters respectively! https://pubs.acs.org/doi/abs/10.1021/acs.jpclett.9b01428 https://pubs.acs.org/doi/10.1021/acscatal.9b02416 All data/code is on github; get in touch if you’re interested!
Excited to be part of an academic/industrial partnership led by Shawn Litster (CMU MechE) to design new oxygen-permeable ionomers for platinum interfaces! https://www.energy.gov/articles/department-energy-announces-50-million-commercial-truck-road-vehicle-and-gaseous-fuels-0
To enjoy the sunny summer weather in Pittsburgh, we did a group trip to go Kayaking in North Park in Pittsburgh. Group mascot Laika also joined!
Great to celebrate another year of graduation at CMU. Gorgeous weather, and first group UG Kaylee Tian was walking.
Congratulations to one of our newest undergraduate researchers, Rui Qi (Richie) Chen, for winnig both a ChESS fellowship and a SURF fellowship! He will be working on Chenru Duan’s and Heather Kulik’s machine learning method for predicting whether or not our DFT calculations will converge prior to even running them. His work will speed up our GASpy framework by allowing us to skip DFT calculations that are more likely to fail.
Kevin just accepted a CCMS internship to work with Dr. Joel Varley and the Quantum Simulations Group at Lawrence Livermore National Lab (LLNL) this summer. They will be refining hybrid DFT simulation methods to model the effects of solvents, electrolytles, pH, and electric fields during electrochemical conversion reactions. This collaboration will lead to a more holistic and robust active optimization of electrochemical processes.
Kevin was interviewed on This Week in Machine Learning & AI for our research in active optimization of electrocatalysts. Check out the podcast on their site or on Spotify!
Our supercomputing allocation on Cori at the DOE National Energy Research Scientific Computing Center was renewed and increased to 5.5 million service units. Cori is currently the 12th fastest supercomputer in the world, and the equivalent cost on commodity cloud resources would be something like $200,000. We are excited to use these resources to develop the datasets for our machine learning methods and materials discovery efforts.
Kevin Tran was awarded the Neil and Jo Bushnell Fellowship from CIT for his research in nanotechnology and electronic materials. Congratulations Kevin!
We are working with Prof. Andrew Peterson and team of four others to develop the open-source machine learning simulation AMP code for exascale machines. The project is sponsored by the Computational Chemical Sciences program in the Department of Energy Basic Energy Sciences division. This project builds and complements the other efforts in our group to develop the datasets and methods necessary to train machine learning models, develop the machine learning methods themselves, and apply these tools to solve engineering design problems. You can read more about the project here: https://news.brown.edu/articles/2018/09/simulations
Congratulations to PhD student Kevin Tran for his first paper published (and group’s first paper on our own!) in Nature Catalysis! Read more about it here: https://www.nature.com/articles/s41929-018-0142-1 We’re really excited about this work, so don’t hesitate to get in touch if you have questions or are interested in follow-up work for other systems!
<iframe width='640' height='360' src='https://www.youtube.com/embed/R5AaK4kqeDg' frameborder='0' allow='autoplay; encrypted-media' allowfullscreen></iframe> <p>
‘Theoretical Investigations of Transition-Metal Surface Energies Under Lattice Strain and CO Environment’ published with Michael Tang and Karen Chan from Stanford in JPC-C. Full abstract below! https://pubs.acs.org/doi/pdf/10.1021/acs.jpcc.8b02094 An understanding of the relative stability of surface facets is crucial in order to develop predictive models of catalyst activity and to fabricate catalysts with controlled morphology. In this work, we present a systematic density functional theory (DFT) study of the effect of lattice strain and of a CO environment on the surface formation energies of Cu, Pt, and Ni. First, we show both compressive and tensile lattice strain favors the formation of stepped vs. low-index terraces such as (111) and (100). Then, we investigate the effect of a CO environment using configurations of CO at various coverages, determined using a greedy, systematic approach, inspired by forward stepwise feature selection. We find that a CO environment favors stepped facets on Ni, Cu and Pt. These trends are illustrated with the corresponding equilibrium Wulff shapes at various strain and CO pressures. In general, the surface energies of the studied transition metals are highly sensitive to strain and CO coverage, which should be considered when rationalizing trends in catalytic activity.
We went hiking in Ohiopyle (~1 hour outside of Pittsburgh) with Kevin, Pari, Jun, and Kaylee. Wen and Zong Qian were working hard on their systems final project so couldn’t make it.
Joint experiment/theory work from time at Stanford, full article here: full article link Title: Copper Silver Thin Films with Metastable Miscibility for Oxygen Reduction Electrocatalysis in Alkaline Electrolytes Abstract: Increasing the activity of Ag-based catalysts for the oxygen reduction reaction (ORR) is important for improving the performance and economic outlook of alkaline-based fuel cell and metal–air battery technologies. In this work, we prepare CuAg thin films with controllable compositions using electron beam physical vapor deposition. X-ray diffraction analysis indicates that this fabrication route yields metastable miscibility between these two thermodynamically immiscible metals, with the thin films consisting of a Ag-rich and a Cu-rich phase. Electrochemical testing in 0.1 M potassium hydroxide showed significant ORR activity improvements for the CuAg films. On a geometric basis, the most active thin film (Cu70Ag30) demonstrated a 4-fold activity improvement vs pure Ag at 0.8 V vs the reversible hydrogen electrode. Furthermore, enhanced ORR kinetics for Cu-rich (>50 at. % Cu) thin films was demonstrated by a decrease in Tafel slope from 90 mV/dec, a commonly observed value for Ag catalysts, to 45 mV/dec. Surface enrichment of the Ag-rich phase after ORR testing was indicated by X-ray photoelectron spectroscopy and grazing incidence synchrotron X-ray diffraction measurements. By correlating density functional theory with experimental measurements, we postulate that the activity enhancement of the Cu-rich CuAg thin films arises due to the non-equilibrium miscibility of Cu atoms in the Ag-rich phase, which favorably tunes the surface electronic structure and binding energies of reaction species.
Great turnout (packed rooms) for a new AI in materials symposium at the MRS spring meeting in Phoenix. Prof. Ulissi talked about practical benefits and approaches to integrating these into the search for new catalyst materials.
Prof Ulissi co-chaired a session on machine learning methods in catalysis, organized by Hongliang Xin, Andy Peterson and CMU’s John Kitchin. He also gave a talk on the group’s perspective of how ML can fit into a day-to-day computational catalysis workflow.
Congratulations to Kaylee for winning a ChESS fellowship to support her research in the group this summer. Her project is titled ‘Variational Autoencoders to Learn Efficient Representations of Catalyst Surfaces’.
Pari received a departmental PPG fellowship for 2017-2018, courtesy of the PPG Foundation. Congratulations!
We visited the Pittsburgh Supercomputing Center (PSC) co-location facility to do some hands-on maintenance with our machines hosted there. Luckily, this was the most labor intensive part of starting the group and building the ‘lab’! We share about 100 high performance CPU and GPU-accelerated nodes with other groups in the college of engineering. Understanding High Performance Computing (HPC) systems and architectures is a valuable skill; if you like combining these challenges with chemical engineering get in touch!
We received 4 million service unit allocation for 2018 on the Cori supercomputer at the Department of Energy National Energy Research Supercomputing Center (NERSC). This is a boost for our high throughput calculation and NERSC also supports our workflow, machine learning, and database tasks. This is roughly equivalent to about $200,000 in usage costs at a commmercial facility like Amazon EC2. Thanks!
Group dinner for the 2017 holidays! We played with our dog Kepler, watched Kevin beat everyone in ping pong, and made and ate Osso Buco and homemade lemon/speculoos ice cream.
Wen and Zong Qian will be completing research projects from January through the end of next year. Zong Qian will be working on predictive methods for materials discovery, and Wen will be working on predictive and high-throughput methods for patterned surfaces!
A commentary that was written from the proceedings of a workshop we attended in May 2017 was published in Nature, with Zachary Ulissi as a co-signatory. See the article for a high-level description of challenges in applying predictive methods to the discovery of new energy materials: Use machine learning to find energy materials
Pari Palizhati and Junwoong Yoon joined the group in November 2017. Welcome!
The group’s first PhD student, Kevin Tran, passed his quals with a presentation on his first project ‘Automated materials simulation: DFT-calculated adsorption energies’. Congratulations Kevin!
Recent work on using machine learning techniques to accelerate the search for bimetallic catalyst reactivity has been published in ACS Catalysis! See below for more details: ACS catalysis Paper
Reaction mechanism reduction in hydrocarbon networks is a persistent challenge due to the complexity involved. We have shown that this complexity can be treated with a small number of DFT calculations and predictive machine-learning techniques. Read more here: http://www.nature.com/articles/ncomms14621
We are actively recruiting for a senior post-doctoral or staff position in the fields of computational chemistry, catalysis, machine learning, and high performance computing. Application include energy storage, hydrogen production, fuel cells, water quality, and more. The ideal applicant would have a strong background in computational chemistry for material science, chemistry, or catalysis, as well is high-throughput computing systems. Excellent programming skills and knowledge of HPC systems is important. Knowledge with at least one machine learning platform (pytorch/tensorflow/etc) is a plus. You would be expected to mentor students in the group and actively develop existing and new collaborations and initiatives. Applicants at the PhD or Post-doctoral level are possible, and the position could either be a post-doc or staff scientist position for the right candidate. If you’re interested email (zulissi at andrew.cmu.edu) with your background and CV.
Kevin Tran joined in December 2016, and will be working on systems engineering approaches applied to materials discovery. Welcome to the group!