SLAC Today logo

Triggering ATLAS

When the Large Hadron Collider (LHC) begins scientific experiments in Geneva next year, bunches of 100 billion protons will cross paths 40 million times every second in the A Toroidal LHC ApparatuS (ATLAS) experiment. Although not every proton will collide with another, roughly 2 billion of them will. If each collision were recorded, the data would fill 6 million CDs every minute. There is no physical way that scientists could record that much information, but that's okay. They don't want to.

Not every event at ATLAS will contain "new" physics that scientists want to record, so there should be enough computer storage space to save these. But choosing which events to keep and which to ignore is a major challenge. "Triggering" software, which separates the best from the rest, has been around for decades. Not only does the software have to be fast, making decisions of whether or not to record an event in 40 microseconds, it has to be accurate. Throwing away events showing new particles being created would undermine the entire experiment. These are the challenges and pressures that SLAC physicists Sarah Demers and Ignacio Aracena face on a daily basis as they program triggers for the ATLAS experiment.

"It's a major challenge to balance the speed of the algorithms with their accuracy," said Aracena. "But it's a lot of fun."

The specific fun Aracena speaks of is his contribution to the project: writing code for the missing transverse energy trigger. If a new particle is created that the detector cannot directly "see," its presence can still be inferred by counting up the energy perpendicular to the beam of all the other particles. Due to the principle of conservation of energy, if an imbalance appears, it indicates there is a new particle the detectors can't see. But detecting this imbalance is very difficult. Most triggers can focus on a single decay in a relatively small area. But the transverse energy software must analyze data from the entire detector, which takes time that Aracena must keep to a minimum. Simultaneously, up to 25 other protons are partially colliding, creating background noise that can give false signals. False signals can also be produced by minor fluxuations in the detector.

"The detector has to be working perfectly or you have to know exactly how it isn't working perfectly," said Demers.

While Aracena is writing code for the missing transverse energy trigger, Demers is working on a tau particle trigger. A tau particle is a lepton—a tiny, elementary charged particle like an electron—that decays almost immediately, usually into a group of pions. Unfortunately, pions are commonly created in other processes that have nothing to do with tau particles. This makes taus very difficult to detect.

"To detect a tau, you basically have to figure out how to tell pions from pions," laughed Demers. "It's tricky, but you can do it."

And to make matters even more difficult, some triggers are based on simulations of untested theories. Because the LHC will operate at energy levels never before reached, nobody knows for sure what new particles will be created, what they will look like, or how to find them. Because of this, the ATLAS project is taking few risks with the triggers at first so important events aren't overlooked.

"The triggers are based on simulations that we hope are reasonable," said Demers. "But we won't really know until the data start coming in. Once we get glimpses of new physics, we can adapt the triggers to better capture the events we want."

"There has been a lot of effort and a lot of progress on triggers in the past couple of years," added Aracena. "I think we're in good shape."

Ken Kingery, SLAC Today, September 18, 2007

Above image: The ATLAS detector. (Image courtesy of Eric Doyle.)