DATA, ARTIFICIAL INTELLIGENCE, AND MACHINE LEARNING AT DOE SCIENTIFIC USER FACILITIESThe U.S. Department of Energy’s (DOE’s) scientific user facilities provide access to the world’s most advanced research instruments and produce increasingly larger quantities of data. While impressive, reaching the full potential of the rapidly growing data streams will require new innovations to solve a variety of technical challenges in data acquisition, control, modeling, and analysis. Artificial intelligence and machine learning (AI/ML) have opened corresponding new avenues in optimization, efficient surrogate models, data analytics, and inverse problems. These intriguing capabilities suggest that AI/ML can greatly accelerate the quest to probe and understand fundamental phenomena across a vast range of lengths and timescales, potentially leading to transformative advances across scientific disciplines.We envision a future of AI/ML-enabled facilities that maximize the DOE’s scientific impact. As a first step towards this vision, DOE announces its interest in receiving National Laboratory proposals from the twelve Basic Energy Sciences (BES), three High Energy Physics (HEP), and three Nuclear Physics (NP) Scientific User Facilities to support data, artificial intelligence (AI) and machine learning (ML) research and development of tools addressing the priorities identified in the Supplementary Information below. It is expected that the proposed research will impact other Office of Science facilities and programs and envision future partnering with those facilities and programs to further evolve and strengthen the advances.DEPARTMENT OF ENERGY OFFICE OF SCIENCEDOESC_8096bb6c-647f-11ea-a101-f4400183ea00The Department of Energy (DOE) Office of Science (SC) performs many functions for DOE national laboratory proposals in the Portfolio Analysis and Management System (PAMS), which is available at https://pamspublic.science.energy.gov.Technical/Scientific Program ContactsDr. Eliane LessnerBasic Energy SciencesDr. John BogerHigh Energy PhysicsDr. Manouchehr FarkhondehIndustryBoth industry and science already use AI/ML approaches for data analysis.ScienceUser facilities, however, crucially require AI/ML tools throughout the lifetime of an experiment: not just for data analysis, but also for data creation, acquisition, and storage. In the next 10 years, AI/ML are expected to go beyond traditional data analysis to aid the design and control of complex facilities, enable real-time capabilities to acquire and analyze large data volumes, automatically steer data collection for in-the-loop experiments, and support experimentalists’ use of exascale computing. These advances will in turn open new avenues of scientific research in energy sciences and beyond.Scientific User FacilitiesIn addition, AI/ML approaches can have a significant impact on increasing the operational efficiencies of large, complex scientific user facilities and scientific instrumentation. AI/ML approaches, for example, can be used to predict detector and accelerator component performance which can result in improved performance and higher beam availability for research... The BES, HEP, and NP scientific user facilities (SUFs) annually serve over 21,000 users. Efficient management of the rapidly increasing quantity of data in these complex systems demands increasing human and instrumentation resources. Current and upgraded facilities face a variety of technical challenges in simulations, control, data acquisition, and analysis. AI/ML methods and techniques promise to address these challenges and impart a drastic acceleration of experimental and computational discovery. AI/ML approaches can also facilitate and improve the operations of these complex machines and their instrumentation.Basic Energy Sciences (BES) Scientific User FacilitiesHigh Energy Physics (HEP) Scientific User FacilitiesNuclear Physics (NP) Scientific User FacilitiesEligible ApplicantsThis is a DOE National Laboratory-only Announcement. FFRDCs from other Federal agencies are not eligible to submit in response to this Program Announcement. Proposals will be accepted only from:BES Scientific User FacilitiesTwelve BES Scientific User Facilities (https://science.osti.gov/bes/suf/User-Facilities)Advanced Light Sourceat Lawrence Berkeley National LaboratoryAdvanced Photon Sourceat Argonne National LaboratoryCenter for Functional Nanomaterialsat Brookhaven National LaboratoryCenter for Integrated Nanotechnologiesjointly managed by Sandia National Laboratory and Los Alamos Laboratory, with locations in Albuquerque and Los Alamos, New MexicoCenter for Nanophase Materials Sciencesat Oak Ridge National LaboratoryCenter for Nanoscale Materialsat Argonne National LaboratoryHigh Flux Isotope Reactorat Oak Ridge National LaboratoryLinac Coherent Light Sourceat SLAC National Accelerator LaboratoryMolecular Foundryat Lawrence Berkeley National LaboratoryNational Synchrotron Light Source IIat Brookhaven National LaboratorySpallation Neutron Sourceat Oak Ridge National LaboratoryStanford Synchrotron Radiation Light Sourceat SLAC National Accelerator LaboratoryHEP Scientific User FacilitiesThree HEP scientific user facilities (https://science.osti.gov/hep/Facilities/User-Facilities )Facility for Advanced Accelerator Experimental Tests (FACET)at SLAC National Accelerator LaboratoryFermilab Accelerator Complexat Fermi National Accelerator LaboratoryAccelerator Test Facilityat Brookhaven National LaboratoryNP Scientific User FacilitiesThree NP scientific user facilities (https://science.osti.gov/np/Facilities/User-Facilities )Argonne Tandem Linac Accelerator Systemat Argonne National LaboratoryContinuous Electron Beam Accelerator Facilityat Thomas Jefferson National Accelerator LaboratoryRelativistic Heavy Ion Colliderat Brookhaven National LaboratoryA future of AI/ML-enabled facilities that maximize the DOE’s scientific impact._8096bc70-647f-11ea-a101-f4400183ea00To support data, artificial intelligence (AI) and machine learning (ML) research and development of tools_8096bd42-647f-11ea-a101-f4400183ea00Open ScienceSC is dedicated to promoting the values of openness in Federally supported scientific research, including, but not limited to, ensuring that research may be reproduced and that the results of Federally-supported research are made available to other researchers. These objectives may be met through any number of mechanisms including, but not limited to, data access plans, data sharing agreements, the use of archives and repositories, and the use of various licensing schemes. The use of the phrase “open-source” does not refer to any particular licensing arrangement but is to be understood as encompassing any arrangement that furthers the objective of openness.CollaborationMulti-institutional and multi-facility partnerships to leverage capabilities are encouraged. Collaborative applications, e.g. submissions of identical proposals by different institutions, will not be accepted under this National Laboratory Program Announcement. Group/team efforts must be submitted by the lead institution.Strategic InformationEfficiently extract critical and strategic information from large complex data sets._8096be00-647f-11ea-a101-f4400183ea001MeaningExtract robust and meaningful information from vast and complex data._8096bebe-647f-11ea-a101-f4400183ea001.1Address how AI/ML can extract robust and meaningful information from the increasingly vast and complex data now being produced at the SUFs. AI/ML techniques have the potential to significantly reduce the effort to process and analyze the data to obtain the desired physical information.ComplexityFind connections elusive to human observations._8096bf86-647f-11ea-a101-f4400183ea001.2In addition, AI/ML can help unmask the complexity hidden in problems in high-dimensional spaces (multi-modal measurements, many experimental variables, etc.) by finding connections elusive to human observations.AutonomyAddress the challenges of autonomous control and experimentation._8096c08a-647f-11ea-a101-f4400183ea002Incorporate use of AI/ML to address the challenges in the real-time operation of large, complex scientific user facilities.SearchingEfficiently search large, complex parameter spaces in real time._8096c152-647f-11ea-a101-f4400183ea002.1AI/ML based methods are needed to efficiently search large, complex parameter spaces in real time, and ...PredictionPredict the health and failure of instruments._8096c210-647f-11ea-a101-f4400183ea002.2to predict the health and failure of instruments that operate at high power sources and experiments that run on these instruments. Performance & ProductivityImprove facility performance and maximize productivity._8096c2d8-647f-11ea-a101-f4400183ea002.3Such capabilities could dramatically reduce facility tuning-time and downtime, improve facility performance, and maximize the productivity of the SUFs.Facilities & ExperimentsEnable offline design and optimization of facilities and experiments._8096c3a0-647f-11ea-a101-f4400183ea003Research to enable offline design and optimization of the facilities and experiments to achieve new scientific goals.GuidanceGuide in-silico experiments from conception to synthesis and measurements._8096c468-647f-11ea-a101-f4400183ea003.1Physically accurate virtual laboratory environments of experimental facilities (a lab in the computational cloud) will help in guiding in-silico experiments from conception to synthesis and measurements.Capabilities & StrategiesEnable the design of new facility capabilities and execution of optimal experimental strategies._8096c58a-647f-11ea-a101-f4400183ea003.2Digital TwinsDigital twins that faithfully mimic facilities, including shared workflows and continuous updates from real experiments, can enable the design of new facility capabilities and execution of optimal experimental strategies to drive physics knowledge acquisition at the SUFs.DataUse shared scientific data for machine learning driven discovery._8096c666-647f-11ea-a101-f4400183ea004Address how to catalyze scientific discovery by leveraging the wealth of diverse and complementary data recorded across the SUFs.Sharing, Curation & AnalysisRadically improve data sharing, curation, and analysis._8096c738-647f-11ea-a101-f4400183ea004.1A radical improvement in data sharing, curation, and analysis is needed to catalyze scientific discovery across all facilities.IntegrationIntegrate diverse scientific data resources._8096c81e-647f-11ea-a101-f4400183ea004.2Through the application of new AI/ML platforms to integrate diverse scientific data resources, the SUFs could create extensive new datasets from heterogeneous experimental and simulated data, leading to new opportunities for scientific discovery.Standards, Formats & PrioritiesCatalyze development of data standards, formats, and priorities._8096c8f0-647f-11ea-a101-f4400183ea004.3Coordinated development of workflows on a shared facility-based data repository could catalyze development of data standards, formats, and priorities across the SUFs.AI/ML MethodsDevelop new AI/ML methods._8096c9cc-647f-11ea-a101-f4400183ea004.4As a byproduct, these curated datasets can serve as training sets for developing new AI/ML methods.Accelerator ComplexesImprove operations of large accelerator complexes._8096cab2-647f-11ea-a101-f4400183ea005FermilabWork conducted on large HEP facilities, such as the Fermilab accelerator complex, are of particular interest, but work may be conducted on other accelerator facilities that have analogous properties.In addition to the above topics, specific interest areas for HEP are implementation of data science techniques such as AI/ML to improve operations of large accelerator complexes, including development of tools and techniques that enable guided optimization, semi-autonomous operations, de-noising and data mining, digital twinning (e.g., virtual laboratories), and failure anticipation... For additional information, go to “Basic Research Needs Workshop on Compact Accelerators for Security and Medicine” https://science.osti.gov/- /media/hep/pdf/Reports/2020/CASM_WorkshopReport.pdf?la=en&hash=AEB0B318ED0436B1 C5FF4EE0FDD6DEB84C2F15B2, May 6-8, 2019. See sections 2.2 and 5.2.1-5.2.3. OptimizationDevelop tools and techniques enabling guided optimization._8096cb8e-647f-11ea-a101-f4400183ea005.1Semi-AutonomyDevelop tools and techniques enabling semi-autonomous operations._8096cc6a-647f-11ea-a101-f4400183ea005.2De-noising & Data MiningDevelop tools and techniques enabling de-noising and data mining._8096cd5a-647f-11ea-a101-f4400183ea005.3Digital TwinsDevelop tools and techniques enabling digital twinning (e.g., virtual laboratories)._8096ce40-647f-11ea-a101-f4400183ea005.4Virtual LaboratoriesFailureDevelop tools and techniques enabling failure anticipation._8096cf26-647f-11ea-a101-f4400183ea005.5Accelerator Systems & DetectorsSupport technical development at the intersections between real-time ML and the control and optimization of accelerator systems operation and detector design._8096d016-647f-11ea-a101-f4400183ea006In addition to the above topics, NP is interested in supporting technical development at the intersections between real-time ML and the control and optimization of accelerator systems operation and detector design using AI models.Beams & Nuclear PhysicsAddress impact beam availability and nuclear physics data collection._8096d106-647f-11ea-a101-f4400183ea006.1NP Scientific User FacilitiesWork conducted at NP scientific user facilities are of particular interest. NP held a one-day roundtable on “Machine Learning and Artificial Intelligence (ML/AI) for NP Accelerator Facilities” on January 30, 2020, with a focus on discussing opportunities in AI/ML for improving efficiencies of accelerator operations of NP facilities. Additional information on the workshop and copies of presentations can be found at: https://science.osti.gov/np/Research#ac.Specific AI/ML applications are sought in accelerator and detector operations that impact beam availability and nuclear physics data collection. Examples include: predicting and minimizing superconducting radio frequency (SRF) faults, SRF cavity instability detection using trained AI models, and development of data driven models to predict machine and detector behaviors for increased performance.SRF FaultsPredict and minimize superconducting radio frequency (SRF) faults._8096d1f6-647f-11ea-a101-f4400183ea006.1.1SRF CavitiesAddress SRF cavity instability detection using trained AI models._8096d2f0-647f-11ea-a101-f4400183ea006.1.2Machine & Detector behaviorsDevelop data driven models to predict machine and detector behaviors for increased performance._8096d3f4-647f-11ea-a101-f4400183ea006.1.3CapabilitiesApply and advance state-of-the-art capabilities._8096d4e4-647f-11ea-a101-f4400183ea007Advanced Scientific Computing Research (ASCR) ProgramAddressing the above challenges may require advances in, or applications of state-of-the-art capabilities in, for example, storage systems, I/O, data and metadata management, and advanced computing hardware. Proposed teams should include expertise to address these challenges, as needed. Opportunities for collaboration with the Advanced Scientific Computing Research (ASCR) program may be identified in order to address these challenges.Storage SystemsAddress challenges, advances, and applications of storage systems._8096d5f2-647f-11ea-a101-f4400183ea007.1I/OAddress challenges, advances, and applications of I/O._8096d6ec-647f-11ea-a101-f4400183ea007.2Data & MetadataAddress challenges, advances, and applications of data and metadata management._8096d7e6-647f-11ea-a101-f4400183ea007.3Computing HardwareAddress challenges, advances, and applications of advanced computing hardware._8096d944-647f-11ea-a101-f4400183ea007.42020-03-092020-05-012020-03-12https://science.osti.gov/-/media/grants/pdf/lab-announcements/2020/LAB_20-2261.pdfOwenAmburOwen.Ambur@verizon.net