diff --git a/introduction.tex b/introduction.tex index 7eb057d..69559e8 100644 --- a/introduction.tex +++ b/introduction.tex @@ -2,7 +2,7 @@ \chapter{Introduction} \label{sec:Introduction} \section{Big Picture} \label{intro:sec:bigpicture} -The discovery of novel materials that solve societal challenges or otherwise improve human lives is arguably one of the most critical components of building a modern world. Starting from the bronze age, humans were able to reliably combine raw materials in a structured fashion to achieve desired results, even though, at the time, there was no mechanistic understanding of \emph{why} things happen. This has changed with gradual introduction of the scientific method, which standardized and systematized the discovery approach, with revolutionary advancements in materials happening every time a new technology for sharing and combining knowledge, such as the propagation of the Greek language or the prinitng press, has been introduced and widely adopted. +The discovery of novel materials that solve societal challenges or otherwise improve human lives is arguably one of the most critical components of building a modern world. Starting from the bronze age, humans were able to reliably combine raw materials in a structured fashion to achieve desired results, even though, at the time, there was no mechanistic understanding of \emph{why} things happen. This has changed with gradual introduction of the scientific method, which standardized and systematized the discovery approach, with revolutionary advancements in materials happening every time a new technology for sharing and combining knowledge, such as propagation of the Greek language, printing press, or computer aided design (CAD), has been introduced and widely adopted. In the current world, which went through the Internet revolution around 2000 and is currently going through the artificial intelligence (AI) revolution of the 2020s, one can point to the informatiztion of materials science as one such communication technology with a potential to revolutionize materials discovery by combining vast amounts of multidimensional data, intricate multidisciplinary domain knowledge, and ability to guide experiments beyond level achievable by a human being. In order to achieve this, one has to consider how to combine these \emph{efficiently}, mitigating problems such as inhomogenities between data sources, computational challenges related to vast design spaces, hidden uncertainties in the reported values, and many flavors of errors, unavoidably present in the complex datasets involved. @@ -37,7 +37,7 @@ \section{Flow of Material Discovery and This Work} \label{intro:sec:flow} All of \texttt{MPDD} is then harvested to model materials at the physical scale by (1) serving as inputs to thermodynamic model generation using \texttt{pycalphad} \cite{Otis2017Pycalphad:Python} and \texttt{ESPEI} \cite{Bocklund2019ESPEICuMg} or training of \texttt{pySIPFENN} ML models generating needed data, and (2) informing experimental observations by, for instance, automatically compiling a set of carbides stable in an alloy system at 0K. At the same time, the largest experimental HEA data infrastructure, called \texttt{ULTERA}, is compiled joining together over 6,800 property datapoints manually extracted from 555 literature publications. -The experimental database is curated through novel \texttt{PyQAlloy} package created to detect abnormalities and dramatically reduce fraction of erroneous data relative to other similar ones in the literature. Once curated, the \texttt{nimCSO} package can guide ML efforts in terms of which components of the data (chemical elements) should be considered when modeling to optimize trade-off between applicability and data density available to the models +The experimental database is curated through novel \texttt{PyQAlloy} package created to detect abnormalities and dramatically reduce fraction of erroneous data relative to other similar ones in the literature. Once curated, the \texttt{nimCSO} package can guide ML efforts in terms of which components of the data (chemical elements) should be considered when modeling to optimize trade-off between applicability and data density available to the models. Lastly, compositional space representations generated through \texttt{nimplex} and inverse design workflows serve as deployment vehicles for the trained methods. diff --git a/main.tex b/main.tex index 6bd2f5b..a91c657 100644 --- a/main.tex +++ b/main.tex @@ -26,7 +26,7 @@ \renewcommand{\familydefault}{\sfdefault} \definecolor{darkgreen}{rgb}{0.05, 0.3, 0.1} -%\usepackage[htt]{hyphenat} %texttt hyphenation breaks +\usepackage[htt]{hyphenat} %texttt hyphenation breaks \let\oldtexttt\texttt \renewcommand{\texttt}[1]{\oldtexttt{\textcolor{darkgreen}{#1}}} @@ -245,15 +245,15 @@ \chapter*{Acknowledgments} I would like to thank all of my family, friends, and collaborators who supported me along the way, with the spotlight given to my parents, \textbf{MikoĊ‚aj Krajewski} and \textbf{Izabella Krajewska}, without whom I would not (statistically) become a scientist, let alone become a \emph{doctor}. However, the degree is just a classifier in a database somewhere without being backed by science created while completing it. Thus, I would like to thank my advisor, \textbf{Zi-Kui Liu}, for guiding me over the last five years of exceptionally productive research that pushed me to \emph{do better than my best}. -I want to thank my research group colleagues, whom I worked with over the years, including \textbf{Zi-Kui Liu}, who was a great colleague, in addition to being a great advisor, ShunLi Shang, Yi Wang, Brandon Bocklund, Jorge Paz Soldan Palma, Hongyeun Kim, John Shimanek, Hui Sun, Rushi Gong, Shuang Lin, Alexander Richter, Luke Myers, and Ricardo Amaral. +I want to thank my research group colleagues, whom I worked with over the years, including \textbf{Zi-Kui Liu}, who was a great colleague, in addition to being a great advisor, \textbf{ShunLi Shang, Yi Wang, Brandon Bocklund, Jorge Paz Soldan Palma, Hongyeun Kim, John Shimanek, Hui Sun, Rushi Gong, Shuang Lin, Alexander Richter, Luke Myers}, and \textbf{Ricardo Amaral}. -I would like to thank my colleagues who, to the best of their abilities, kept me from falling into an abyss of scientific insanity through hours spent on less technical conversations. In particular, but in no particular order, I would like to acknowledge several of them who regularly attended my weekly office hours over the years: Stephen Holoviak, Alexander Richter, Luke Myers, Cooper Pan, Curtis Warner, Ian Cunningham, James Ricardo, Ellie Franklin, Hamdan Almarzooqi, Jorge Paz Soldan Palma, Brandon Bocklund, and Stephanie Castro Baldivieso. +I would like to thank my colleagues who, to the best of their abilities, kept me from falling into an abyss of scientific insanity through hours spent on less technical conversations. In particular, but in no particular order, I would like to acknowledge several of them who regularly attended my weekly office hours over the years: \textbf{Stephen Holoviak, Alexander Richter, Luke Myers, Cooper Pan, Curtis Warner, Ian Cunningham, James Ricardo, Ellie Franklin, Hamdan Almarzooqi, Jorge Paz Soldan Palma, Brandon Bocklund}, and \textbf{Stephanie Castro Baldivieso.} -I would like to thank my Lawrence Livermore National Lab colleagues Aurelien Perron, Brandon Bocklund, Kate Elder, Joseph McKeown, and other amazing colleagues from the Materials Science Division at Lawrence Livermore National Lab (LLNL) for having the opportunity of working with them on solving challenging problems with great implementation flexibility which prompted me to deepen my understanding of highly dimensional design spaces, without which some of the work in this dissertation would never happen. +I would like to thank my Lawrence Livermore National Lab colleagues \textbf{Aurelien Perron, Brandon Bocklund, Kate Elder, Joseph McKeown}, and other amazing colleagues from the Materials Science Division at Lawrence Livermore National Lab (LLNL) for having the opportunity of working with them on solving challenging problems with great implementation flexibility which prompted me to deepen my understanding of highly dimensional design spaces, without which some of the work in this dissertation would never happen. -On the software side, I would like to thank (1) Jinchao Xu from PSU/KAUST for his contribution to the development of SIPFENN; (2) Richard Otis and Brandon Bocklund from Materials Genome Foundation for supporting this work since 2019 in a variety of ways, including invaluable guidance in organizing community workshops; (3) Ricardo Amaral and Luke Myers for testing a lot of my work; (4) +On the software side, I would like to thank (1) \textbf{Jinchao Xu} from PSU/KAUST for his contribution to the development of SIPFENN; (2) \textbf{Richard Otis} and \textbf{Brandon Bocklund} from \textbf{Materials Genome Foundation} for supporting my work since 2019 in many ways, including invaluable guidance in organizing community workshops. -In the Fall of 2023, I had an opportunity to be a Visiting PhD Student at the University of Cambridge, for which I am very grateful to Gonville \& Caius College, invited me and Dr. Gareth Conduit for generously sponsoring said invitation, as well as Peter and Carol Thrower for sponsoring the fellowship enabling this travel. +In the Fall of 2023, I had an opportunity to be a Visiting PhD Student at the University of Cambridge, for which I am very grateful to \textbf{Gonville \& Caius College}, invited me and \textbf{Gareth Conduit} for generously sponsoring said invitation, as well as \textbf{Peter and Carol Thrower} for sponsoring the fellowship enabling this travel. This work was made possible by the financial support and training provided by US Department of Energy (DOE) via Awards DE-FE0031553 and DE-EE0008456, DOE Advanced Research Projects Agency-Energy (ARPA-E) via DE-AR0001435, the DOE BES (Theoretical Condensed Matter Physics) via DE-SC0023185, US Office of Naval Research (ONR) via N00014-17-1-2567 and N00014-23-2721, The Pennsylvania State University via ICDS Seed Grant, US National Science Foundation (NSF) via CMMI-1825538, and Pathways to Enable Open-Source Ecosystems (POSE) via FAIN-2229690. Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the author and do not necessarily reflect the views of the funding agencies.