A lot has been written about the new LHC schedule. Like everyone here, I was deeply disappointed by the “incident” last year and I’m eagerly anticipating first collisions this year. Because of the delay in the collider schedule, my family thinks I’m on an extended vacation, but somehow I’m almost busier now than if we really were taking data.
As we wait for the accelerator repairs, my major focus is on managing one of the data analysis working groups. Actually maybe “facilitating” is the right word. My job is to make sure that the folks in our working group are ready to look at the first data with the CMS detector with all their software (and youthful enthusiasm!) revved up and ready to go. We do this using what we call “Monte Carlo Data” — which is a bit oxymoronic considering “Monte Carlo” means simulations of what we think LHC collisions will look like in our detector and “Data” implies the real signals from real LHC collisions.
The Monte Carlo “Data” is made in several steps:
Step 1 – Event Generation: In this step we generate a list of particles that might be created from a particular proton-proton collision. To do this we use the theory of Quantum Chromo-Dynamics (or QCD) combined with results from all previous High Energy Physics experiments to predict all the end products (particles) in an LHC proton-proton collision. This is (only!) our best guess of what the collisions will look like. There are many different “generators” for generating the final particles in a collision as we can only approximate QCD predictions and there are different models for how and what to approximate. Also there are a lot of input parameters to QCD (like particle masses for example) that have to be put in by hand. The output from the generation step is a list of all the particles from the collision, their charge, their direction and how fast they are traveling (i.e. momentum). Because in each collision there is only a probability for a particular set of final particles, we “roll the dice” and repeat the collision millions of time in order to get, on average, an idea of what the collisions look like. Hence the origin of the name “Monte Carlo”. Unfortunately knowing how to use a QCD Monte Carlo generator does not help you win at the gambling tables.
Step 2 – Simulation: Once we have a set of particles from Step 1, we simulate how the particles would interact in our detector. To do this we have to have a very complete implementation in software of our detector, including the positions of all the components and exactly what the signal from each type of particle would look like in each component. Each part of the detector is designed to collect complementary signals from a particle. Even parts like the cables that bring signals from the inside of the detector out to the electronics that register the data have to be in the simulation since there is some probability that a particle will interact in the cables!
Step 3 – Reconstruction: In this step, we now forget we ever knew anything about the original generated particles and we try to reconstruct them given the signals we simulated in Step 2. This step we also perform on real data once we get it.
Step 4 – Calibration: From the reconstructed Monte Carlo events of Step 3, we can ask ourselves how well the reconstructed particle matches the particle we generated in Step 1. This gives us an idea of how well our detector will perform when we reconstruct real data (where we don’t know what the generated particles).
The better our simulation and reconstruction steps (steps 2 and 3) the more like real data our Monte Carlo will look like. The better our calibration step (Step 4) the better our understanding of the particles coming out of our real proton-proton collision will be.
Step 5 – Analysis: this is where we take the final calibrated collisions and look for all the cool stuff we hope to find like the Higgs boson or mini black holes. But most of the data will come from more mundane QCD processes, so in order to claim we see something new and different, we will have to work hard to make sure we understand the data in terms of our QCD generators (from Step 1). Likely we will have to do some tweaking of the generators since will be looking at hadron collisions at a much higher energy than anyone has ever before studied and we will have to modify some of our assumptions and approximations.
Right now, as we wait for the repairs to the LHC, we are busy creating millions of Monte Carlo collisions using different models for Step 1 and ever more precise detector information for Steps 2, 3 and 4. It’s a lot of work, but it will pay off in faster data analysis time once the real collisions begin. And I should mention that even though we think we are planning for all the eventualities in how the detector will react to particles from collisions, there are almost always surprises. So while we will see nice pictures of collisions at the instant they happen, it will take us longer to understand and analyze them.
So, in short, this is not much of a winter vacation. And I left out all the work we are doing repairing andimproving our own CMS detector while we have the opportunity during this down time!
P.S. I’m new to this whole blog thing, so please feel free to critique/ask questions/boo me off the stage…