1. Neutron Background Simulation Long-lived neutrons created, diffuse around collision hall They get captured by nuclei, emitting a photon Compton scattering or photoelectric effect makes MeV electrons, which cause hits in muon chambers

2. Differences GEANT3 and GEANT4 GEANT3→ no ion simulation GEANT4→ ion simulation → perfect elastic scattering of neutron → additional physics package for thermal neutrons CMSIM 121: 1.1 hits per event in ME t>250 ns OSCAR_3_3_2+QGSP_BERT_HP: 0.46 hits per event in ME t>250 ns

3. Results QGSP_BERT_HP, 2k events, Endcap detectors

4. Because neutrons can live up to a second before making a signal They can’t be treated like ordinary minimum-bias pileup, because millions of collisions in the past can contribute Why is neutron background hard to simulate?

5. The Way It Was Done Before: Parametrization ~9 years ago, UC Davis group (Hessian, Fisyak, Breedon) Based on 2000 simulated min-bias events, simulated down to low energies and long times Start with “mother” hits with some distribution in energy, position, and direction Add some number additional hits in same layer Propagate each hit to next layer and repeat

6. Disadvantages of Parametrization Hard to maintain Hard to maintain r Many parameters r Needs to be done separately for each detector type for CSC, DT, and RPC r Needs to be updated when geometry or shielding changes Can we use the original events rather than a parametrization of them? Can we use the original events rather than a parametrization of them?

7. What I’ve Done: Database of Neutron Hit Patterns Start with a sample of simulated min-bias events Take the events apart. Treat each chamber with hits as an independent event. Zero out the time.

8. Database of Chamber SimHit Patterns Store these patterns of neutron hits in a ROOT file, grouped by chamber type: ME1/A ME 1/1 ME1/2 ME1/3 ME2/1 ME2/2 …. When I need to add neutron background to the simulation, I just read in some number of these patterns and superimpose them Done before electronics simulation, of course, so things pile up correctly.Done before electronics simulation, of course, so things pile up correctly. Only done for chambers that already have signal!Only done for chambers that already have signal!

9. How Many Patterns to Superimpose? Say we simulate a window of 24 bunch crossings around the event. That’s ~330 min bias events (at 10 34 ) that may create signals in future crossings. I think we can assume that the amount of neutron signal in our 24-bx window is the total amount that would come from 330 minimum bias events in the past. If ME2/1 chambers have a neutron-induced occupancy of 0.24% per min-bias event per chamber, we should superimpose a Poisson mean of 330*0.24% = 0.8 hit patterns per chamber.

10. Adding neutron events

11. Effects of adding neutrons Other subdetectors: è Endcap RPCs may want to use this CSC RecHit occupancy: è Not much. A few percent. è We’re simulating 24 bx. Overlapping with existing RecHits DAQ rate: è How often will they make fake segments in empty chambers? è Needs to be implemented separately r Database of only neutrons which make real segments

12. Other issues RecHit occupancy: è Not much. A few percent. è We’re simulating 24 bx. Overlapping with existing RecHits DAQ rate: è How often will they make fake segments in empty chambers? è Needs to be implemented separately r Database of only neutrons which make real segments