Exploring emerging scientific fields: Big data-driven materials science Developments in methods to extract knowledge from data provide unprecedented opportunities for novel materials discovery and design. Nevertheless, the emerging field of data-driven materials science is still in its infancy. To identify what it takes to quickly advance this emerging field, Prof. Rinke from Aalto University s School of Science joined efforts with Prof. Granqvist from Aalto s School of Business to initiate a unique, interdisciplinary and real-time study. Dr. Amber Geurts, postdoctoral researcher at both Schools, has taken up this challenging research project in which the pioneering scientists working on this frontier are the subject of her research. Mission: Materials innovation Materials are perhaps one of the most underappreciated aspects of our everyday lives. Materials are the foundation of our modern society and largely determine the way we live. Throughout pre-history, new materials enabled rapid development of civilizations. It is for this reason that we have named our ages after materials think of the stone age, the iron age, and the bronze age. Today, the dependence on materials to promote human development persists. The current age is often referred to as the silicon age, because it enabled computers, long distance communication and the internet. To address current societal challenges in clean energy production and storage, mobility, and health, there is a need to continue the materials revolution because societal challenges are materials challenges. However, the development of novel materials and their application in products is time consuming, expensive, and often relies on human intuition. Using an Edisonian approach to innovation, large sets of humanly-selected material candidates are tried and tested in experiments or in computational simulations in materials research. Even when such methodologies uncover new materials and their functionalities and properties, this data is rarely shared among scientists or with industry and is often as quickly forgotten as it was generated. What is more, even if such new materials are noticed or reactivated, it still takes about 20 years from materials discovery to commercial success. And, any intellectual property rights that might protect the commercial success of the new material are by then outdated. The increasing societal demand for novel materials, therefore, requires a new, costefficient method that requires less time investment and encourages data sharing. 1
Moogle : A Google for materials? In Iron Man 2 (2010), protagonist Tony Stark uses advanced technologies to identify and design a new, stable material for his power suit in a matter of minutes. What is more, he only relied on 485 simulations to identify the correct combination to design such a material 1. In a somewhat similar way, around 10 years ago, the technologies and software of our silicon age inspired a group of visionary scientists to use these new, technological possibilities to address current bottlenecks and research needs to help design and make novel materials much better, faster, and cheaper. These scientists launched something that can be considered as the beginning of a Google for materials (a Moogle perhaps), in which calculated data was put in an open access database to enable future research to find appropriate materials for a better tomorrow 2. Big data-driven materials science became an emerging scientific field of research, and more and more centers opened over time 3. Years after this database was initiated, in a speech on 24 June 2011, former US President Obama announced the Materials Genome Initiative as a major national research priority: To help businesses discover, develop, and deploy new materials twice as fast, we are launching what we call the Materials Genome Initiative 4. By using the term genome, Obama made the analogy to the human genome. Unlike the human genome, the materials genome has not yet been decoded, even though it is fundamental to understanding and developing materials. Big data-driven materials science is ideally suited to decode the materials genome. Although the idea of creating databases is nothing new, the idea to collect experimental or computational data from various research groups and to make that data available in a shared, central location is. Inspired by increasing demands from various governments legislation for open science, these databases are developed using open science formats that allow data sharing and develop software based on open technological regimes. Most recently, the hype around artificial intelligence (AI) has inspired materials scientists to combine these databases with new AI tools to provide unprecedented intelligence for materials discovery and design in ways traditional methods cannot. Such big data-driven discoveries might fundamentally 1 See: https://vimeo.com/148568736. 2 The first center has been the Materials Project at Lawrence Berkeley National Laboratory in the USA (2008). 3 Soon after, others followed, most notably the Swiss Computational Design and Discovery of Novel Materials (MARVEL), the European Novel Materials Discovery (NoMaD) Laboratory, and the Japanese Materials Database (MatNavi) of the National Institute for Materials Science (NIMS) in Japan. 4 Materials Genome Initiative: https://www.mgi.gov/. 2
advance the field of materials science research and change the way material scientists do science. Figure 1 illustrates big data-driven materials research as a new scientific paradigm. Figure 1: The new big data-driven materials research paradigm. The traditional approach (left) relies on experimentation and theory to discover materials one at a time. In the big datadriven approach (right), data from current and previous studies is collected and machine learning methods discover new materials from large amounts of data. Studying emerging scientific fields from a social science perspective Despite these recent technological and scientific developments, the field of big data-driven materials science is still in its infancy. However, strong momentum is building. The number of large-scale initiatives has grown during the last couple of years, and this trend is increasingly recognized by the physics community, funding agencies, and publishing houses. Yet, the institutions and culture of science remain rooted in the pre-digitization era. Big data in materials research thus requires a paradigm shift in the institutions and practices of science. On the one hand, science relies on knowledge developed by systematic study and application of scientific principles. On the other hand, science and scientific fields like any other social institution is continuously determined by its human participants and the shared culture, 3
including scientific practices, ethical values, and specialist languages. Hence, what we understand as science in various disciplines is subject to gradual change. Science and scientific practice is also impacted by hypes and trends, which are particularly visible in science policy and funding, for instance. Not all changes are lasting, however, as indicated in Figure 2. If we, social scientists, continue to conduct only retrospective studies that take a historical perspective, we might miss important discoveries of what determines the shape, boundaries and content of what is considered science. In 2017, Prof. Rinke, Prof. Granqvist and Dr. Amber Geurts teamed up to explore the trajectory of Big Data in Materials Research from the utopian margins to mainstream science. Tracing the development of this ongoing paradigm shift in material sciences provides a unique, real-time research setting. We study how actors engage in efforts to alter frames, practices, and cultures to adjust and negotiate the boundaries of this emerging scientific community. At Aalto University, we are in a unique position to conduct such research. Profs. Adam Foster and Rinke participate in Aalto University s interdisciplinary thinktank, the Materials Platform 5, and partners in the Novel Materials Discovery (NOMAD) Laboratory 6 a prevalent, open access materials repository and materials analytics platform. These connections enable access to the emerging field of big data-driven materials science and its main stakeholders. Using social scientific methods, like in-depth semi-structured interviews, the analysis of such archival data as reports and news articles, and ethnographic fieldwork among the various members of this emerging field, we aim to determine where this emerging science is now and how it is developing. The aim is also to understand how to quickly advance this emerging field of science. By doing so, our findings will have implications for both materials science and social scientific managerial research on field emergence. 5 Aalto University s Materials Platform: http://materials.aalto.fi/en 6 Novel Materials Discovery (NoMaD) Laboratory: https://nomad-coe.eu 4
Figure 2: Potential development of emerging scientific fields over time by interest. In our research, we are trying to establish which curve database driven materials science follows and where we are on it. Outlook: where is this emerging scientific field now, and where is it going? Although our socio-economic inquiry into the emerging field of big data-driven materials science has just begun, we have already observed that an increasing number of researchers and research groups globally are joining the bandwagon that was initially led by a few visionary star scientists advocating a new paradigm in scientific materials research. These pioneers are largely supported by a growing group of enthusiastic young scholars, eager to capitalize on the hype in relating fields most specifically AI (i.e., machine learning, pattern recognition, data mining, database technologies) to transform the perhaps utopian ideas of the early frontrunners into reality. Despite its promise and growing membership numbers, however, the community is still small and members have to negotiate their positions within the established scientific community: I still have to convince colleagues that this is possible now (Note from fieldwork in Switzerland). While the field boundaries are still permeable and the application of big datadriven materials science remains utopian for most material scientists, getting the necessary research funding to further develop the field is challenging. Funding agencies not only have to evaluate the novelty of proposed projects, but they also have to guess the future impact and sustainability of the emerging field. Easier said than done. 5
In Finland, for example, only recent grant calls have started to pay attention to the needs of the emerging scientific community of big data-driven materials science 7 : An emerging field of science struggles even more [than established scientific fields] to receive governmental funding (Note from fieldwork in Finland). Perhaps as a consequence of governments hesitation to fund such projects, Finnish industry has also remained reluctant to invest in big data and AI-driven (scientific) applications: [Finnish industry] People are talking about this [artificial intelligence] a lot, but they are not doing anything. yet. (Note from fieldwork in Finland). At the same time, however, industry elsewhere is quick to scoop up the talent from this emerging field of science, resulting in worries over a potential brain drain in the academy: Did you hear; we lost another one [PhD researcher]. My student got an offer from Google [tech industry giant] and he accepted (Conversation among research group leaders during fieldwork in Switzerland). Conclusion Novel materials discovery and design, powered by big data-driven methods, provide unprecedented opportunities to find materials that can address current societal challenges from fundamental scientific discovery to product release. Nevertheless, the emerging field of data-driven materials science is still in its infancy, threatened by ongoing community boundary negotiations, limited governmental funding, and the talent scout role played by industry that shrinks the pool of skilled research in the academy. To establish fully autonomous platforms for materials discovery, multidisciplinary and multibody joint efforts between research, industry, and public and governmental organizations are necessary. 7 See the Academy of Finland s new program for novel applications of artificial intelligence in physical sciences and engineering research (2017) https://www.aka.fi/en/research-and-science-policy/academyprogrammes/current-programmes/ai2/ or the recent Future Makers Funding Program at Technology Industries of Finland (2018) http://techfinland100.fi/future-makers-funding-program-2018/. 6