High Energy and Nuclear Physics are concerned with understanding of matter and space-time at the fundamental level. There are many unanswered questions:
The standard model of elementary particle interactions provides a detailed description of the components of matter and how they interact. Strongly interacting particles such as protons and neutrons are bound states of quarks and gluons. The dynamics of this binding is described by quantum chromodynamics. The only technique available for computing with this theory on the scales of nuclear binding is lattice gauge theory.
In the search for new particles or symmetries, high energy physics experiments study the fragments emitted from extremely energetic collisions of relatively simple particles (electrons, protons, anti-protons). Planned experiments of this type that will be in operation in the early years of the next decade at several facilities include BaBar, using electron positron collisions at the Stanford Linear Accelerator Center, and CDF and D0 using the proton-antiproton collisions at Fermi National Accelerator Laboratory in Illinois. In 2005, ATLAS and CMS are expected to commence operation at CERN in Geneva using proton-proton collisions. The BaBar experiment is designed to study in detail the properties of b-mesons with the aim of understanding the breaking of matter-antimatter symmetry. The others are aimed at understanding the mechanisms by which particle masses are generated, and exposing the fundamental symmetries of nature.
In the search for new states of strongly interacting matter under extreme conditions of energy density or angular momentum, nuclear physics experiments are carried out at a variety of both high and low energy accelerators. A new nuclear physics facility, the Relativistic Heavy Ion Collider (RHIC), is scheduled to start operation at Brookhaven National Lab in 1999. Here, experiments including PHENIX and STAR will use very high energy colliding beams of heavy nuclei to study the properties of nuclear matter under extreme energy density, replicating the environment of the early universe. A new detector facility, gamma ray energy tracking array (GRETA), is being developed to operate at various low energy nuclear physics accelerators, to investigate the structure of cold nuclei at very high spin states.
The last few years have witnessed an explosion in the complexity of the experiments being carried out in high energy and nuclear physics, both in the volume of data produced and in the complexity of the analysis required to extract the physics from them. The next generation of experiments will produce data sets up to 1000 times that or previous experiments.
ATLAS at LHC
ATLAS is one of the two multipurpose detectors under construction to operate at the LHC at CERN. Proton-proton collisions will be used at 14 TeV center of mass energy. The goal of ATLAS is to provide data that will enable us to determine mechanism responsible for the generation of mass. This hinges on the ability to extract rare events and subtle effects in huge data volumes. Understanding the underlying physics requires detailed simulations. Hence, the simulated datasets will be comparable in size to the experimental ones. Although the amount of CPU required to perform those simulation is large (for example of the order of 2 million MIPS for ATLAS alone), the most pressing problem is data management. We need a distributed Peta-byte-scale data storage facility and sufficient networking bandwidth to guarantee the effective use of this data by the physicists distributed globally. We need data transfer rates of the order of 100 Mbits/s at the university level and tens of Gbits/s in and from the data-store.
The PDSF facility is being used by the Atlas collaboration to simulate
the response of the detector to jets. So far 321000 events have been
generated using the PDSF system.![[*]](http://www.emsl.pnl.gov/images/latex2html/foot_motif.gif)
The data are sent back to CERN for analysis.
STAR (Solenoidal Tracker at the Relativistic Heavy Ion Collider)
The STAR experiment is one of two major experimental facilities being constructed at the Relativistic Heavy Ion Collider (RHIC), which will begin operation in October 1999 with the collision of counter-circulating beams of gold ions, each having an energy of 100 GeV/nucleon. Head-on (``central'') collisions of these gold ions will produce a system with a total energy in the center of mass of 3.9 TeV, spread over the volume of the colliding system. Such violent collisions will produce final states of a complexity unprecedented in accelerator-based nuclear and high energy physics. The final state will consist of about 10,000 particles, mostly pions of energy below 1 GeV. Analysis of this enormously complex final state will yield signals of the conditions at the hottest and densest moment of the collision and the system's subsequent decay, and possibly signs of the fleeting phase transitions from confined to de-confined matter and back again. STAR aims to track and measure a large fraction of these particles, perhaps up to 4000 per event, which will be recorded at a rate of one collision per second. Buried among this huge number of ``soft'' pions are other, rarer signals, such as high transverse momentum jets and electron-positron pairs, which are uniquely sensitive to the hot, dense early phase of the collision, and which STAR will also detect and analyze.
The data from a central gold-gold collision at STAR is expected to have a volume between 12 and 20 MB (in the case of a strong first order phase transition, events could be much larger), leading to a total data volume (after additional processing) of about 300 TByte produced per year. Estimates of the CPU power needed to analyze these data, together with the simulations to understand instrumental effects, range between 10 and 15 kSPECint95. Efficient and organized access to these data sets requires new developments in data handling and databases.
The required computing power and infrastructure to analyze this enormous amount of data is not expected to be fully satisfied by the RHIC Computing Center. NERSC plays an essential role in the analysis of STAR data, both in terms of capacity and capability, the latter being perhaps especially important, since NERSC brings many years of experience in the handling and analysis of very large data sets.
GRETA
GAMMASPHERE is one of the world's two advanced gamma-ray spectrometers which have continued the marked advance in nuclear structure physics seen in the last decade. Detector arrays of greater sensitivity such as GRETA, which is currently being prototyped at Berkeley Lab, promise to open exciting new vistas in nuclear physics. However to take full advantage of current and proposed spectrometers, new computing capabilities are required.
A major focus of GAMMASPHERE has been to investigate the behavior and response of nuclear systems when they are subjected to extremely high angular momentum values. This nuclear rotation can be as fast as a hundred billion billion revolutions per second. The crucial nuclear structure information is contained in the details of the gamma-ray de-excitation flash when 30 or more gamma-rays are emitted. A number of preferred pathways in the de-excitation process can occur. They relate to favorable arrangements of protons and neutrons which often can be associated with specific symmetries or nuclear shapes.
The increase in sensitivity to weak pathways in current and planned detectors is realized by spreading the less correlated background events over a space of increased dimensionality. For instruments such as GRETA, the optimal dimensionality for data analysis is 7-8. Large, specialized databases are required to store and query the data from these experiments. Given a typical datasets consists of 10**9 to 10**10 events with 50 such experiments being conducted each year, a data store of approximately 2TByte is required. NERSC is in a unique position to provide such a super-computing level data store with the computational resources to query such databases efficiently. With the inclusion of proper network resources, such a facility would be a valuable resource to the low-energy nuclear physics community throughout DOE laboratories and the university community. Furthermore, access to high quality scientific visualization services (as described in 4) will be important in the analysis of these high-dimensional datasets.