John Dagdelen

John Dagdelen

Computational Materials and Machine Learning

Talks and presentations

Driving Materials Innovation with Natural Language Processing

October 18, 2019

Keynote, National Research Council Workshop on AI for Design, Ottawa, Canada

The majority of all materials data is currently scattered across the text, tables, and figures of millions of scientific publications. In my talk, I will present the work of our team at Lawrence Berkeley National Laboratory on the use of natural language processing (NLP) and machine learning techniques to extract and discover materials knowledge through textual analysis of the abstracts of several million journal articles. With this data we are exploring new avenues for materials discovery and design, such as how functional materials like thermoelectrics can be identified by using only unsupervised word embeddings for materials. To date, we have used advanced techniques for named entity recognition to extract more than 100 million mentions of materials, structures, properties, applications, synthesis methods, and characterization techniques from our database of over 3 million materials science abstracts. With this data, we are developing machine learning tools for autonomously building databases of materials-properties data extracted from unstructured materials text. Finally, my talk will also feature a sneak peek into the public-facing website and API we have developed to make this data freely available to the materials research community.

Natural Language Processing for Materials Discovery

November 27, 2018

Talk, Materials Research Society Fall Meeting, Boston, MA

The majority of all materials data is currently scattered across the text, tables and figures of millions of scientific publications. We present recently developed natural language processing and machine learning techniques to extract materials knowledge by textual analysis of the abstracts of several million journal articles. We describe our use of Word2Vec to map words in our corpus to vector representations, which we then use as inputs to named entity recognition (NER) classifiers to extract materials, structures, properties, applications, synthesis methods, and characterization techniques from the abstracts in our database. With this information, we have created new tools for materials literature review such as: searching within chemical systems, filtering articles by experiment/theory, summarizing the known attributes of a material, or finding similar materials to a target. Furthermore, we report how these techniques can be used not only to automatically summarize existing knowledge, but enable new ways of discovering novel materials such as thermoelectrics or ion-conductors by revealing previously undiscovered relationships between materials and their properties.

Computational Investigation of Poisson’s Ratio and its Relationship to Crystal Structure.

July 20, 2018

Talk, American Crystallographic Association Annual Meeting, Toronto, Canada

A material’s Poisson’s ratio describes the magnitude of transverse strain that results when it undergoes a tensile strain. However, Poisson’s ratio can be generalized in anisotropic crystals though the elastic tensor to give the single-direction Poisson’s ratio for a strain in any specific crystallographic direction. In a few rare cases, a crystalline material can possess extreme directional Poisson’s ratios and exhibit surprising properties as a result. One example of this phenomenon is the appearance of an overall negative average Poisson’s ratio in the polycrystalline bulk. Such materials have been shown to have exceptional mechanical properties and are enabling advancements in technologies like high-precision sensors, tougher ceramics, and impact-resistant composites. While only a few hundred experimental elastic tensors have been measured to date, new computational methods are now enabling researchers to calculate the elastic tensors of thousands of materials at a time. In this talk (or poster), I discuss how we can use high-throughput computational screening methods based on new descriptors for similarity between crystal structures in combination with open materials databases like the Materials Project database to identify materials that are likely to exhibit unusual mechanical properties (such as negative Poisson’s ratio). I then show how the mechanical properties of these materials can be further investigated through computation of the elastic tensor via density functional theory calculations to yield a complete topographical picture of Poisson’s ratio for a given crystalline material.

Review of Energy Generation Trends in the United States.

September 30, 2017

Keynote, International Energy Raw Materials and Energy Summit., Istanbul Technical University. Istanbul, Turkey

As part of the 2015 Paris Climate Agreement, the Obama Administration pledged that the United States would reduce its carbon emissions by 25% before 2025. To meet this target, Obama and the US Environmental Protection Agency (EPA) proposed a set of regulations known as the Clean Power Plan (CPP), which laid out new efficiency standards for fossil-fuel power plants and policies designed to expand the country’s low/zero-carbon power generation capacity. As a result of the combination of regulations like the CPP and market pressures caused by the availability of inexpensive natural gas, the role of coal in electricity generation in the USA has significantly declined over the last few years. While coal accounted for almost half of all electricity generated in the United States in 2007, less than one-third of electricity in US is generated from the burning of coal today. As the number of US coal power plants decreased from 639 in 2005 to 416 in 2016, the number of people employed in the coal industry also shrank. Whereas about 250,000 people were employed in the coal industry in the 1980’s, that figure currently stands at 100,000. It was argued that any jobs lost due to closed coal mines, coal-fired power plants, and frozen construction of new plants would be replaced with those created by the expansion of new wind and solar projects. The EPA estimated that the CPP would create more than 80,000 new jobs to replace those lost in the coal sector due to the transition to cleaner power sources, but many people remain skeptical whether this will in fact be the case. This presentation will look at the role coal, wind, and solar power as sources of energy and their impact on carbon emissions and the American jobs landscape in the coming decades.