Dec. 17, 2001
Data analysis during proton-proton collisions at RHIC
Proton-proton collisions are detected by three experiments: STAR,
PHENIX and the proton-proton elastic scattering experiment. During
each collision, about 50 particles are emitted mainly in the backward
and forward directions. While crossing the experiments, the particles
leave impacts that are electronically recorded, stored on tapes, and
later analyzed by physicists.
Event Selection And Data Acquisition
Among the 30,000 collisions produced every second, STAR and PHENIX can only read 100 and 1,000 events per second, respectively. The scientists select the most interesting events using a specifically designed trigger system.
“To keep up with the collision rate, trigger decisions have to be fast,” says physicist Tonko Ljubicic, responsible for the trigger and data acquisition at STAR. “In less than 100 billionths of a second, the basic trigger selects interesting events, and in 20 thousandths of a second, a more sophisticated trigger system decides whether to keep the event.”
Unlike STAR and PHENIX, which are looking for various types of reactions between proton constituents, the proton-proton elastic scattering experiment selects only reactions producing two protons at the end of the reaction. These reactions are easier to select than other types of reactions; millions of them can be detected in a few days.
Data Storage and Analysis
The data coming from any of the three experiments are sent via an optical fiber to a computer room, the RHIC Computing Facility (RCF). There, the data are recorded on tapes enclosed in four big silos.
“Each silo can store about 300 terabytes of data, which is about 100,000 times the storage capacity of a desktop computer,” says physicist David Morrison, PHENIX software and computing leader. “When either PHENIX or STAR is running, we collect about a terabyte per day, so one silo is about a year’s worth of data from either experiment.”
The data is later read back from the tapes and sent to computers in another room, where the entire collision is reconstructed. The “reconstructed” data, which consists of track hit positions, track momenta, and particle decay positions, is put either on tapes or on disk, and is ready to be analyzed by physicists.
Information about the event can be analyzed simultaneously from several hundred other computers at Brookhaven Lab. PHENIX data is also analyzed at the RIKEN laboratory in Saitama, Japan, and part of the STAR data is transferred to DOE’s National Energy Research Scientific Computing Center (NERSC) in Oakland, California, where data can be accessed in a complementary way to RCF data.
STAR and PHENIX will be producing data for about 15 years. Preparing for the important flow of data, Brookhaven is now considering an expansion of RCF’s computing capabilities.
“RCF is quite big now,” Morrison says. “But over the next few years, it is going to get even bigger and more capable with faster and more powerful computers.”
The U.S. Department of Energy's Brookhaven National Laboratory conducts research in the physical, biomedical, and environmental sciences, as well as in energy technologies. Brookhaven also builds and operates major facilities available to university, industrial, and government scientists. The Laboratory is managed by Brookhaven Science Associates, a limited liability company founded by Stony Brook University and Battelle, a nonprofit applied science and technology organization.