h1. 2016-11-22-straight-tracks-paper

h2. Details

Tuesday 22nd November at 15:00 and continued on Wednesday 23rd November at 14:00.

http://mice.iit.edu/phonebridge.html

Note:

http://mice.iit.edu/micenotes/public/pdf/MICE0497/MICE0497.pdf

h2. Notes

Present:

John Cobb

Paul Soler

Mariyan Bogomilov

Chris Rogers

Alan Young

Ryan Bayes

h3. Page 1

First three comments may be more appropriate for the paper rather than the MICE note

* Should defend choice of Xenon

* Should define symbols in emittance formula

* Should define equilibrium emittance

# "Thus, despite the limited angular range it is extremely attractive as a direct measure of the

scattering of muons" Need to add some more quantitative discussion of angular acceptance. See below.

h3. Page 2

# Table 1: Row for Al numbers are suspect - correct

# Table 1: General comment that some description of the material budget is required; how much material and where?

# Table 1: In similar vein, some description of absorber is necessary

# Scattering angle definitions eqn (3) and (4) are not correct - see supplementary note from JHC attachment:Projected-angles.pdf

# Correction @theta = \sqrt(<theta_scatt>)@ should be @theta = \sqrt(<theta^2>)@

# There was some discussion over the use of a Gaussian fit, as mentioned in final sentence of 1.1. Comment that the description would be better in the "analysis" section where the relevant measurements and plots are listed.

# 0.0038 should be 0.038

# Comment that section 1.2 needs a lot of work; it was not reviewed in detail.

h3. Page 3

# "deviation between data and simulation": some discussion over this comment. Which direction is "South"? Why does beam offset introduce bias?

** Beam offset causes clipping, which is a bias

** the 4 cm shift is in MC

** It is a bias in the simulated measurement, not in the data itself

h3. Page 4 (Fig. 1)

# Be explicit that this is MC through a "bare absorber"

# Why is the Geant4 MC of Xenon so weird - e.g. asymmetric?

** Ryan is not sure and will (continue to) dig

# Why is the Cobb-Carlisle MC of Xenon so weird (e.g. thick)?

** JHC will dig

h3. Page 5

No comment

h3. Page 6 and 7

# Some discussion of the cuts and rate accounting

** Timing selection is a very small time bin, e.g. 200 ps over the few ns width of the muon peak; so 10 % acceptance in the cut is okay

** Fiducial selection is a 300 mm diameter tracker over a 5 m drift length; so 10 % transmission in the cut is okay

** John asked whether the various scattering plots were really for this very narrow TOF bin. Ryan said "yes"

** Need to add some description of the basic geometry

# Need to describe handling of the overflows; e.g. what angle is assigned

# Line 95; worth referencing table 4

# table 4 please add the momentum range (as well as the mean). The issue of TOF range vs momentum range came up quite a few times in the discussion.

# Fig. 2b - where does the a/p+b fit come from? where does the momentum calculation come from? Ryan: a/p +b fit is empirical, fit is done to all points (red and black).

# Fig. 2c/2d momentum @p_mc@ is defined at upstream face of LiH. There was a lot of discussion over sources of bias in the figures. E.g. why are there different trends for different data sets? Presumed that the trend comes from some aspect of the transverse distribution. Two guesses:- either handling of non-straight path between TOF1 and TOF2 or asymmetric scraping of the distributions. Should write down the formula/algorithm used for calculating momentum in more detail. Need to at least explain the source of the bias (this yields main systematic so it is important). John suggested a data-driven comparison of TOF01 with TOF12, for example using Rayner method in TOF01.

# Downstream tracker efficiency was mentioned. It was covered in mroe detail during systematics discussion.

# Fig. 3/4. Asked why are there events with r > 150 mm? It is the extrapolated tracks, so tracks with space points < 150 mm can be extrapolated to Station 1 with r > 150 mm. Should look at 4 point compared with 5 point tracks to understand the effects. There was discussion of tracker resolution. Should investigate a chi2 cut on the tracker tracks. Also should look at the error matrix.

h3. Page 8

# Reference to fig. 5 should be reference to fig. 1

# Equation 8. Quite a lot of discussion of @n_data@ in the denominator. Ryan and John agreed it should not be there.

# Should have i,j in the subscript on theta sys (it is done bin by bin)

# John asks what is the  convergence criterion for the deconvolution? How sensitive are we to this? Ryan watches for chi^2 < 2. Paul is interested in the change in chi^2 per iteration. A plot of chi^2 vs iteration number would be helpful.

# Fig. 6 there was quite some discussion of the "flat top" in empty AFC data set. I missed most of it as I had to call the MLCR. I don't think there was a conclusion apart from "it looks strange".

h3. Page 13

# General remark - error is wrong term, please use uncertainty

# Ryan noted equation 11 is missing @sigma_alpha@ in middle bit. Also note that @sigma_sys@ is not the same as that of equation 8.

# Discussion of what Ryan is doing - he is taking a big change in <parameter alpha>, looking at the change in measured width, and then scaling by the estimated error on <parameter alpha>.

# John noted that in general some errors should be taken out in quadrature, i.e those that broaden the distribution, while some can change the distribution linearly (e.g. fibre pitch).

h3. TOF systematic

# this is an uncertainty on the momentum, not on the measured width. Lots of discussion. Settled on a concept like "quote systematic error on the momentum and then explain that this is tied in to the uncertainty on the model comparison by the process in the paper".

# Fig. 7 needs to be described in greater detail. E.g. what is plotted, what momentum beam, how is calculated (plot in +/- 400 ps bins are made, then normalised, then the difference is calculated). It is not clear that there is a systematic error - looks quite statistical in nature. Ryan to check (e.g. chi2 between the two distributions).

# The quoted 129 ps error arises as a quadrature sum between the 70 ps TOF01 resolution and the systematic error estimated from Fig. 2d

h3. Absorber material sensitivity

# Add a reference to the LiH note

# The 0.254 resolution arises due to non-flatness of the LiH. It is an uncertainty on both diameter and thickness (not quite sure if those two statements are consistent).

# Again, comment that this is an uncertainty on the model, not the measurement.

h3. Alignment

# Agreed that displacement does not matter? Surely only @theta_x@ and @theta_y@ is significant?

# Discussion of roll. It is probably second order, but not so well constrained by measurements. Should check whether it is a significant contribution to systematics.

# Table 5 through 8 - quote units.

h3. Fiducial

# Gradient is a0 as quoted in Table 3

# "The grid" is anachronistic and should be replaced

# 0.56 mrad for angular resolution of the tracker is wrong. The scattering is dominant and has not been accounted for. Back of envelope suggests about 9 mrad.

# Proposed studying probability of measuring downstream track as a function of position and angle of incoming tracker (i.e. what is the tracker efficiency?)

# Discussed projecting track from upstream to downstream and calculating probability of reconstructing a track - i.e. what is the efficiency and acceptance of the downstream system

# Also look at MC acceptance of downstream system (i.e. independent of detector effects). Pen and paper calculations should also be presented.

h3. Page 17

# various quoted values of dof are inconsistent and should be made consistent. Ryan: table is right and text is wrong

# Table 9+10. Discussion of what the different columns and numbers mean (which error is statistical vs systematic). They should be described more carefully in the text.

# Fig. 9 "normalise residual" means chi in a given bin

# It was noted that the model is not equally above and below the data - i.e. the data width is systematically larger than the model. Note data are normalised to total number of events that pass cuts.

# Need to examine distributions for data and convoluted models - do they agree (e.g. x and y)

# Xenon - we agree the MC is wrong and recommend it is moved to a separate note, to be pursued independently

h3. Page 22

# Why does PDG agree with theta scatt but not projections? Wonder if there is bias due to error in eqn (3) and (4)? Note it is complicated - PDG is RMS of central 98 % of distribution cf Lynch and Dahl. Gaussian is fit to core. Need to check what PDG is doing and try to reproduce. But in fact, the data is not self consistent. Try using a fit range for data that is consistent between theta scatt and projections.

# Better to be consistent throughout the paper and use RMS theta scatt (not RMS theta scatt/sqrt(2))

h3. Page 23

# Conclusions will be to be redrafted following other corrections

# There was some discussion of implementation of corrections - who would do what...