Experimental science is evolving. With the advent of new technology, scientific facilities are collecting data at increasing rates and higher resolution. However, making sense of this data is becoming a major bottleneck. New mathematics and algorithms are needed to extract useful information from these experiments.
To address these growing needs, the Department of Energy has announced approval of a joint ASCR-BES-supported grant of $10.5 million over three years to expand the Center for Advanced Mathematics for Energy Research Applications (CAMERA). CAMERA’s mission is to develop fundamental mathematics and algorithms, delivered as data analysis software that can accelerate scientific discovery. Originally focused on challenges facing the Advanced Light Source (ALS) and Molecular Foundry (MF) at the Lawrence Berkeley National Laboratory (Berkeley Lab), this new round of support will help transform CAMERA into a national resource across DOE computational, network, and light source facilities. ALS and MF are DOE Office of Science User Facilities.
Reaching across traditional boundaries
“To tackle the emerging challenges at scientific user facilities across DOE, we need to invent and exploit new mathematics from a variety of different fields, and couple that with experimental expertise to target scientific problems, ” says James Sethian, CAMERA Director, Head of Berkeley Lab’s Mathematics Group, and Professor of Mathematics at UC Berkeley.
CAMERA’s approach is to assemble teams of applied mathematicians, statisticians, experimental scientists, computational physicists, computer scientists and software engineers. By having a broad range of expertise focused together in one place, CAMERA teams are able to reach across traditional boundaries, find a common language, frame experimental data needs in mathematical terms and package mathematical solutions as software usable by the external community.
CAMERA launched in 2009 as a Laboratory Directed Research and Development (LDRD)program at Berkeley Lab. Based on this success, DOE’s Office of Advanced Scientific Computing Research (ASCR) and Basic Energy Science (BES) jointly invested in CAMERA as a pilot project. This new round of funding represents a major expansion in CAMERA’s mission and goals.
“The ultimate goal of the LDRD program is to enable bold, innovative projects with the potential for high scientific impact. In this sense, CAMERA is a true success story,” says Horst Simon, Berkeley Lab Deputy Director. “Years ago, we funded CAMERA because we saw its potential. Since then, it has built on work performed and supported throughout DOE and has helped develop and deliver mathematical analysis software across a range of science fields.”
CAMERA products include new algorithms and software for ptychography (SHARP), grazing incidence shallow angle x-ray scattering (HipGISAXS), reconstruction and analysis of imaged materials (QuantCT and F3D), chemical informatics for analysis of crystalline porous materials (Zeo++), fast methods for electronic structure calculations (PEXSI), nanocrystallography, and fluctuation X-ray scattering.
“With this new round of funding, CAMERA is now poised to expand on its initial efforts, tackle new mathematical challenges that naturally emerge from evolving experimental needs, support visiting scientists and postdoctoral fellows from other Labs and facilities, and provide a curated software portal for advanced algorithms built by the community to meet these data challenges,” says David Brown, Director of Berkeley Lab’s Computational Research Division.
Tailoring Mathematics to Meet Experimental Needs
CAMERA tailors mathematics, algorithms and software specifically to the experimental requirements and computational resources at hand.
“With detectors being able to collect increasingly vast amounts of data at accelerating rates, advances in computing architectures offer new opportunities to keep up with the data-processing demands,” says Sethian, who is also a Co-Investigator of the Berkeley Institute of Data Sciences (BIDS). “And, rethinking the mathematics often produces unprecedented advances in accuracy and speed, independent of computing power.”
Here are some examples of how CAMERA has tailored their approach to different experimental needs:
- To meet in-situ real-time ptychographic imaging needs, CAMERA has recently embedded part of its core highly accelerated ptychography code “SHARP” (Scalable Heterogeneous Adaptive Robust Ptychography) directly with CCD detector systems, producing real-time reconstruction of combined diffraction and microscopy as the scanner moves over the sample.
- To address the image analysis needs of micro-CT users “on the shop floor,” CAMERA scientists have recently introduced Quant-CT and F3D, which are image filtering and reconstruction algorithms based on combinations of non-linear filters, PDE-based and template matching image segmentation, and statistical classification methods to analyze material porosity and topologies, deployed as user friendly Fiji plugins and running on workstations directly connected to ALS beamline 8.3.2.
- As an example of using remote supercomputing to meet data analysis needs, the SPOT Suite workflow management system running at the National Energy Research Scientific Computing Center (NERSC) was used to create a prototype data pipeline: as data was collected from an ALS GISAXS experiment, it was sent via DOE’s Energy Sciences Network (ESnet) to the Titan Supercomputer at the Oak Ridge Leadership Computing Facility (OLCF) for analysis on 8000 nodes using CAMERA’s HipGISAXS code, a customized high performance code that exploits advanced graphics processors and particle swarm optimization to quickly reverse engineer the sample from simulated scattering patterns based on distorted wave Born approximations. NERSC and OLCF are DOE Office of Science User Facilities.
- An example of mathematics fundamentally altering imaging capabilities comes from an emerging imaging technique known as fluctuation X-ray scattering (FXS). FXS offers the possibility to increase the extractable information from solution scattering experiments by several orders of magnitude by exploiting powerful new light sources including DOE’s free electron lasers, such as the Linac Coherent Light Source (LCLS) at SLAC National Accelerator Laboratory. However, a major bottleneck has been a lack of techniques to determine general molecular structure from the data. To address this issue and solve the reconstruction problem of FXS, CAMERA scientists introduced a powerful new mathematical theory and algorithm, called multi-tiered iterative phasing (M-TIP). CAMERA’s new code can quickly determine general structure in only a few minutes on a desktop computer.
CAMERA is jointly supported by the ASCR and BES in the U.S. Department of Energy’s Office of Science. Further information about CAMERA, current projects, future expansion, and engagement opportunities may be found at camera.lbl.gov.