MC production takes place in 6 steps:
-
Matrix element calculation: This samples individual events from the matrix element for the hard scatter process. This is based on “exact” QCD/QED calculations for the requested process, e.g., a leptoquark calculation at leading order in QCD. The outcome of this step is a (very short) list of “hard” (high-momentum) partons, usually in a Les Houches Event format. Depending on the complexity of the calculations and the filtering scheme used, this can be reasonably fast or one of the slowest steps in the MC production process.
-
Parton shower/hadronization: This takes the partons from the hard scatter and simulates the QCD processes that happen at lower energy scales all the way down to hadronization (the formation of color-singlet bound states). This is generally done using phenomenological models and the outcome is a list of “stable” particles that enter the detector. Particles are defined as stable if their lifetime is sufficient for them to enter the detector, even if they decay quickly compared to everyday timescales. Up to this point, the result is a “truth record” and is not ATLAS-specific as it depends entirely on theory predictions. The matrix element, parton shower, and hadronization steps are collectively known as “event generation”, and are sometimes done with a single program.
-
Detector simulation: This simulates the path of “stable” particles through the detector as they interact with the detector material and magnetic fields, produce new particles in those interactions, or potentially decay. Computationally, this is often the most expensive part of MC production, though fast-simulation techniques are often used to more quickly estimate the total energy deposited in the calorimeters instead of fully simulating the electromagnetic and hadronic showers. The outcome of this step is a list of energy deposits in the various subdetectors of ATLAS.
-
Digitization: This converts the simulated energy deposits into digitized signals. The outcome of this is equivalent to the raw data that is read from the ATLAS detector, though with additional “truth” information about what really happened in the event.
-
Trigger Simulation: This provides trigger decisions, like those available in the data. A different menu might be used for MC production compared to a specific data-taking run, so you might need to check that a trigger you wish to use is available in both places.
-
Reconstruction: This reconstructs the digital signals into physics objects (such as electrons and hadronic jets), producing the same kind of files as actual data that can be used for analysis, called Analysis Object Data or AOD files. MC files of this type typically store some version of the truth record that can be useful for cross-checks in analysis.
All of these steps are performed in the central MC production system when a request is submitted. This tutorial introduces you to the necessary steps that lead to the submission of MC requests for the central production system.
Only MC produced centrally is allowed to be used in approved ATLAS results. It is typically fine to use privately produced MC samples for studies related to designing your analysis, but these will ultimately need to be replaced before your analysis enters the review process.
In order to minimize wasting computing resources, MC samples need to be efficiently designed and properly validated before a request can be submitted. This procedure consists of 4 steps:
- Writing jobOption files for your samples and running the matrix element calculation and parton shower/hadronization steps
- Producing TRUTH derivations from the output
- Navigating the truth record and plotting variables to confirm that the desired process is being correctly simulated
- Getting group approval for the request and submitting it to PMG