PhD Thesis -- Simulation and Blind Analysis -RPM 07.11.29 CHAPTER COMMENTS: This chapter will include everything to do with the simulation, including the Michel spectrum generation and the Blind Analysis technique. Geant validation is the bulk of the work I did personally, so I should probably go into some detail. It's probably best to sort this by which MC aspect is being verified, rather than by which type of data was used, since the verification is what we're ultimately interested in; that would allow me to emphasize when results are verified by different methods. Alternatively I could treat each type of study as a separate "experiment", with method and results; this might actually be better, as some of these studies take some effort to explain, and too much intermingling might make it cluttered. Maybe it's best to sort by method, with lots of cross-referencing. - Look to whatever definitions I come up with for "soft interactions" and "hard interactions". That will imply different studies. Should still sort this chapter by study, this might add a study or two. - If I do get most of my "positron interactions" validations from US Stops, I could probably have a Positron Interactions section, and describe the US Stops as a subsection of that. - Actually, that's probably a good idea. This chapter is about validating Geant, so from an external POV it makes the most sense to describe the sections in terms of what part of Geant is being validated. Most of the other sections I've got listed are already labelled by Geant aspect. Emphasize that Geant3 matches data as well as it does with no tuning on our part; everything looks "really good", certainly good enough for our current needs. Note that the actual systematics studies belong in the Systematics chapter. CHAPTER OUTLINE: Overview - Start the chapter with a brief statement about how TWIST extracts Michel parameters from data by comparing the data spectrum to MC. (I think this is mainly a reminder from the Analysis chapter.) Obviously, then, validation of MC is vital. - Point out how important it is that the validation is done using methods that are independent of the Michel parameters. - Mention in brief the use of the standalone spectrum generation as an input to the MC, and that this facilitates the TWIST blind analysis technique. Don't go into details; just forward ref. Michel spectrum generation - No notable improvements since previous round. - Standalone decay spectrum generator (micheld); produces (p,costh) pairs, with weights for use with derivs (forward ref the mcfitter section). Pairs are used by geant during muon decay, to determine the initial (p,costh) of the decay positron. - Includes nearly all available RCs. - Accept/reject method, keeping track of nthrown for normalization. - Assumes some values for the Michel parameters; these values are stored in an encrypted database. Result is the Black Box. Opening the box means passing the encryption key to the database and revealing the hidden Michel parameters. - Note that the database can have multiple keys for multiple sets of parameters, so that I can open my box while leaving another one blind. - Can also generate derivative spectra. - Absolute value of derivative formula is taken as the probability distribution. - Sign of the derivative at a selected (p,costh) pair is stored along with the pair. Sign is used at treesumming to add or subtract the event from the final decay spectrum histogram. - Use resulting probability distributions in standard simulation, run through standard analysis and treesum. Geant description - Include improvements since previous rounds. (Highlight this separately, with forward references.) - Asymmetric cell geometry and STRs (shift, "bulge") - Improved dead zone simulation - Beam profiles from TEC; results of improved beam studies - Basic simulation technique. (Describe the particle Monte Carlo concept in broad strokes.) - The geometry and properties of the detector are described. Most of the detector is hard-coded, but some parameters are read in from the same geometry file used for analysis. - Particles are tracked through the mapped magnetic field, using a fourth-order Runge-Kutta integration technique. - Energy loss and scattering below certain cutoffs are included (randomly) in the tracking. Other processes, such as secondary particle production (Brem and deltas), are performed at appropriate random intervals. - I do need to get into the "free path length" business, as that's what I'm using to tune Brem and Deltas. - Ionization clustering and digitization. - Geant determines locations where ion clusters are produced along the particle track. - STR lookup gives drift times for the ion clusters. - From drift times, and preset cluster signal durations, Geant determines when ion cluster signals overlap. - Signals add when overlapping. Hit is produced when the total signal goes over some threshold. - Signals are written out in data files using the same format as the real data aquisition produces (\ref{sec:hardware_chambers}) Additionally, optional data banks can be included in the data files. - Describe optional data banks. Don't need a lot of detail; basically explain what "vertex banks" and "space points banks" are, broadly. - Cite the TWIST Chambers paper regarding the tuning of the clusterization. (Make sure it's in there first! Might not be; in that case see if there's a technote or something.) - Rough description of the muon and positron beams, with references to how they're measured. - Include mention of pileup, etc. - Muon beam is sampled from correlated position and angle distributions, produced from field-on TEC measurements. Vary between sets. (Ref?) - Positron beam is sampled from correlated position and angle distributions (though simpler than the muon dists?), produced from field-off DC measurements. Same for all sets. (Ref?) - Muon decay and micheld sampling. - Use of calibration files (alignment, STRs, beam profiles, etc). - Uses the same calibration files as the analysis will use. - Stores list of calibration files in the output data file for analysis software to read. - Probably useful to look through the FFCARDS and pick out any notable ones used. Geant Validation - Overview - Positron Interactions (US Stops) - Description of data, specifically this round. (Could compare to versions used in previous rounds.) Include quantity of data taken, and quantity of accepted events. Show acceptance plots. - List the cuts and track selections used. - Explain the beam positron issue. - Quantities of interest: dp, dth, pdth (and/or other normalization) (should probably start using projected scattering angle instead of dth, though!), particularly vs various things (like 1/costh). Justify these -- why are they interesting, why the funny normalizations if any. - Compare distribution shapes -- peak positions and FWHM. - Compare relative tail counts, and show log plots of the tails and their excellent agreement between MC and data. - Show PeakDP vs 1/costh for 30 and 40 MeV/c. Discuss slopes. - Specifically this is the motivation for applying energy calibration as a momentum-independent shift. Point this out and back-ref. - Compare appropriate results here against what Energy Calibration tells us. Show where we should get the same answer, and where we shouldn't. - Actually, this may not be an appropriate thing to discuss here; this chapter is about Geant Verification, after all. Probably better to discuss this comparison in the Energy Calibration section (TWIST Overview chapter), listing numbers there with forward references to this section to say where the numbers come from. Exactly how I do this will of course depend on what we ultimately come up with as we grow in our understanding of the ecal... - Reconstruction Efficiency - Backref to US Stops, natch. - 2D inefficiency plots for MC and Data, US and DS, and p and costh projections for US ("DS is similar"). - Explain beam positron stripe. - 2D ineffdiff plots for MC and Data, US only ("DS is similar"), and P and costh projections for US ("DS is similar"). - 2D DS-US diffs for MC and Data (since this was, to some degree, a source of a possible systematic), and probably the "fiducial average" effdiff (see systematics report). - Outside Material (DS Aluminum) - Description of data. Explanation of what it's supposed to represent. - Data vs MC comparison, and forward reference to calculation of resulting systematic error. - TDC Spectra - Show the matched leading edge, and describe the tuning of MC parameters that was needed to get it to look good. (Not so much a verification as a tuning, of course, but it's something in MC that has to match data, so there it is.)