End-to-End Virtual Accelerators (EVA) (Contribution to Snowmass21) Jean-Luc Vay *1 , David Sagan 2 , Axel Huebl 1 , Maxence Th´ evenet 3 , R´ emi Lehe 1 , Zhirong Huang 4 , Cho-Kuen Ng 4 , Henri Vincenti 5 , Michael Bussmann 6,7 , Alexander Debus 7 , Richard Pausch 7 , Ji Qiang 1 , Adi Hanuka 4 , Brigitte Cros 8 , Daniel Winklehner 9 , David Grote 10 , and Auralee Edelen 4 1 Lawrence Berkeley National Laboratory, Berkeley, CA, USA 2 Cornell University, Ithaca, NY, USA 3 DESY, Hamburg, Germany 4 SLAC National Laboratory, Menlo Park, CA, USA 5 LIDYL, CEA-Universit´ e Paris-Saclay, CEA Saclay, Gif-sur-Yvette, France 6 CASUS – Center for Advanced Systems Understanding, G¨orlitz, Germany 7 Helmholtz-Zentrum Dresden - Rossendorf (HZDR), Institute for Radiation Physics, Dresden, Germany 8 CNRS, Universit´ e Paris Saclay, Orsay, France 9 MIT, Cambridge, MA, USA 10 Lawrence Livermore National Laboratory, Livermore, CA, USA August 2020 Abstract The growing importance of particle accelerators to society, together with their increasing complexity and high cost, demands that we bring the most advanced computing tools to bear on their design. Modeling of beams at extreme intensities and densities (toward the quantum degeneracy limit), and with ultra-fine control (down to the level of individual particles), calls for integrated predictive tools that can take advantage of the largest supercomputers, with an ultimate goal of creating End-to-end Virtual Accelerators (EVA) that model particle accelerators from start to end. Ultimately, with the help of surrogate models, such virtual accelerators could have many of the characteristics of a Virtual Reality (VR) simulator, allowing the operator to quickly and efficiently model operation of the accelerator for its intended application in real time. * jlvay@lbl.gov 1