PhD Thesis -- TWIST Software CHAPTER COMMENTS: This chapter needs to describe the software in enough detail to understand how the experment is done, with an emphasis on improvements since the previous round. Emphasize these improvements at the front of each section; they'll act as a sort of "things to watch for" summary. Need to include some details about how MC Simulation works, for the sake of my external(s). Refer to the appendix containing the full list of event types. Include references to appropriate systematics sections, particularly for Geant. I'll probaly want to try to emphasize, or at least point out, any stuff I worked on specifically. Should the McFitter be here instead of getting its own chapter? Andrei had a separate chapter because he wrote the thing. For me it's basically an established program. Yeah, move it here for now, see what people say. CHAPTER OUTLINE: - Michel Fitter technique - Point to the full Michel spectrum formula (reference the page in the Theory chapter where it occurs). Mention the products PmuXi and PmuXiDelta. - Since the decay spectrum is linear in the decay parameters, the derivatives wrt those parameters are independent. Thus we can write data in terms of MC + derivs. Show formula. Deriv coeff's are the differences in decay parameters. - List the derivatives. - Include plot of full spectrum and derivatives vs momentum, for costh=1, to show relative shapes and sizes. Should probably also show the 2D derivative spectra; could do a single figure with the base and 3 derivs, each in a square aspect ratio. - Describe fit technique. - Use standard full-spectrum simulation for normalization. - Compare counts/nthrown in std vs deriv. - Fit determines best coeff's for derivs. - Convert to xi,delta notation by dividing, and accounting for the errors with a Jacobian. (See Andrei's thesis?) - TWIST results are quoted using eta fixed to the SM value (or to Black Box value before box is opened). Quote correlations of other derivs with eta and point out resulting systematic on rho. - Actually, this really belongs in the Systematics chapter, with a quick mention here and a forward ref... - Show residuals and residuals vs p. Also show the residual distribution w/ Gaussian fit. - Mofia - Include improvements since previous rounds. (Highlight this separately, with forward references.) - Asymmetric STRs (shift, bulge) - Improved cell hit calculations (finding the point of closest approach) - Improved first guess at small angles - Outline of program flow - Broad strokes list of the various steps of the program: - Initialization, selecting and reading calibration files - Unpacking each event, including optional data banks - Crosstalk removal (data only) - Window definitions, window classification and particle ID, event classification - First Guess (with some details) - Two-stage helix fit (with some details) - Tree output - Treesum - Include improvements since previous rounds. (Highlight this separately, with forward references.) - Neither Blair's nor Andrei's theses compare their treesums against previous versions. Should check Jim's, since some treesum questions resulted in an "analysis systematic". - Differences between my config file and Andrei's treesum description: - Decay time cut: now using track time instead of window time. - Track pair matching: I don't know of any changes since Andrei, and Blair says he didn't touch this. - Idea and method. - Include list of cuts, with histograms. - Blair has a good example of this in his thesis (fig. 3.10). - Direct (classic) Energy Calibration - This ecal is used to determine endpoint resolution, at least as a reference, so I should describe it in brief. Don't need much detail. - Comparative Energy Calibration - Purpose of the ecal, and what sorts of effects it can be correcting for. - Describe the new ECal technique. - Initially, a McFit is done between the raw data and MC spectra (or whatever two spectra are being calibrated), to compensate for differences in polarization. The fitted derivatives are applied to one of the endpoint spectra as appropriate before the relative ecal is done. - Look at each costh bin of the endpoint region (p,costh) histograms to be compared. Each angle bin contains a momentum spectrum of the endpoint. - Bin widths are constant in 1/costh, to improve the number of bins near 1/costh=0. The endpoint shape in that region is of interest in testing for helix fitter biases etc. - Select a pre-defined region of the data momentum spectrum which contains the edge. Select the same size region of the MC spectrum, but starting at the low momentum end. Normalize the selected parts of each spectrum by the total counts in that region, then compare these regions of the data and MC. A chi2-like goodness of fit is calculated. Then the selection region of the MC is incremented by one momentum bin and the process repeated, etc. The result will be a plot of chi2 vs MC spectrum offset, in steps of the momentum bin widths. - Find the minimum of the chi2 plot, and fit a parabola to a region around this point. The momentum at which the parabola's minimum occurs is the difference in edge positions between the two spectra for this angle bin. - Plot the edge position vs 1/costh. For a "fiducial region" (different from the TWIST fiducial), fit straight lines to the upstream and downstream edges. These fits represent the energy calibration as a function of momentum. - Show equation of how ecal gets applied during second treesum. - Include our best understanding of the results. Forward reference to US Stops for comparison. - Fiducial volume (with forward references for justification) - Compare current fiducial with that from previous rounds. List relative increase in statistics, and the increase in sensitivity (lower McFit errors and correlations). - Forward reference to US Stops inefficiency study.