| next >

2021-06-23 solenoid-mode

Discussion of solenoid mode paper

14:00-16:00 BST Friday 25th June 2021

Zoom details:
  • [[computing-software:MICE_zoom_info|MICE Zoom Connection]] (login required)
  • Status of the analysis, including tasks left to complete (Tom Lord)


Referee's Meeting

Tom Lord, Steve Boyd, Maurizio Bonesini, Chris Rogers

Tom gave an overview of the analysis. Highlighted a number of issues still to resolve.

Jaroslaw asked what is the TOF01 cut? Tom said it is normalised with respect to the electron peak.

Tom showed MC vs data during cut. Jaroslaw questioned the missing data sets. Tom responded that the 4-140 was not invented for flip mode; and that 10-140 for LiH is missing due to a misunderstanding during data taking. Chris added that one could potentially finesse the presentation for the paper (but for now we just show the plots).

Tom noted some differences in some of the plots between data and MC. In general agreement was pretty good. Noted TOF peak and momentum offset data vs MC. Chris noted that the TOF and momentum offsets are consistent (i.e. data is always faster than TOF). Tom noted that the MC has not been so well-tuned. More MC is coming. ACTION: review when more MC comes.

Maurizio questioned the diffuser in 10-140. What is the origin of the double peak? Chris commented that we don't have a good model at the diffuser edge - for 10-140 it is dominated by the diffuser support structure which is not well-modelled in MAUS.

Jaroslaw asked what is the efficiency in TKD. ACTION: Tom said he will check. Chris asked if this is including particles which are lost in the cooling channel. Tom was not sure, but seemed to agree.

Tom went on to show TKU beam distributions. Tom noted there were some missing events at low px. Noted some apparent misalignments in the MC y.

Tom noted that the 3-140 distribution was quite clearly different between lH2 and LiH data sets, upstream. Jaroslaw asked if Tom really checked magnet by magnet the currents are the same. ACTION: Tom will check again the settings from the CDB. Chris suggested Tom could also check EPICS. Look at the alignment parameters as well. Jaroslaw was concerned that it was too big a change for small MC variations.

Tom noted that the axes ranges MC vs data in 2D px-py plots are different which makes it hard to compare. ACTION: fix the axes. Also remove the spurious vertical line in MC.

Tom explained the amplitude calculation and statistical error estimation.

Tom went on to explain the systematic correction mechanism and error estimation. Jaroslaw asked whether there should be a systematic uncertainty associated with the vertical displacement. Tom said he had not done one. Chris and Steve noted that as a cylindrically symmetric system it should not matter?

Tom described the correction effect. He showed absolute numbers of the magnitude of the correction on the hybrid MC. Jaroslaw asked what is the normalisation (how many events were there in the hybrid MC?) ACTION: Tom said he will check. There are approx 50,000 to 500,000 events in each sample.

Tom showed the impact of systematic uncertainty in the detector arrangement on the reconstruction efficiency. Jaroslaw asked whether Tom understood the behaviour of the systematics correction. Tom said not really. Steve pointed out that a lot of it is statistical. Chris asked why the different systematic effects all yield offsets in same direction and magnitude. Tom said he wasn't sure. Tom noted that the sample sizes were quite different. ACTION: add labels indicating which beam line setting corresponds to which plot. ACTION: seek to understand to what extent the systematics are dominated by statistics.

Chris asked whether Tom did the errors for different absorber settings. He indicated not at the moment. He did look at empty and didn't see a huge difference. Chris asked if there were separate systematics for TKU and TKD. Tom said yes he did. He hasn't plotted them separately.

Tom showed effect on the migration matrix. Variations are small.

Tom showed the amplitude distributions. He noted the difference between the MC and data. Chris noted that the systematic uncertainty looked a bit smaller. Jaroslaw pointed out for example the upstream amplitude distribution had narrower uncertainty bands compared to the Nature paper. ACTION: check the systematic calculation. It seems too narrow.

Tom showed the PDF ratios. He noted that there was a small discrepancy in 6-140. Noted that the heating in 3-140 makes it hard to show the plots. Also noted that the MC vs data comparison is pretty poor for lH2 full/empty plots. Needs a different vertical scale which is a little uncomfortable, or we do something like let the data fall off the top of the plot. Jaroslaw felt that the 3-140 plot is not suitable for publication. ACTION: review the analysis following new MC.

Tom showed CDF and CDF ratios. Noted that the plots were more presentable, but the MC vs data is still quite different for empty/full lH2.

Tom moved on to the canonical angular momentum. He showed us plots of kinetic angular momentum, field angular momentum and canonical (combined) angular momentum. He went on to show the differences. Jaroslaw asked if the increase in empty is consistent with windows. Steve replied that it has to be because there is no other material.

We discussed different ways of expressing the angular momentum. Tom considered the possibility to do either a plot of mean angular momentum vs z or 1D histogram of change in angular momentum. ACTION: remove the line joining reco upstream to downstream. The line joining MC upstream to downstream should show the effect of the absorber.

We discussed how to do the errors on the 1D histogram of delta angular momentum. Considered looking at the effect of uncertainty on each bin, doing a bin-by-bin error. The bins are correlated so this is a bit challenging. Might be better to show the uncertainty on the mean. ACTION: blow the axes up. Consider adding a stats box. Consider a vertical line displaying mean, with width indicating uncertainty on the mean. We discussed how to do the error. Felt that - plot the mean on the 1D histogram. The systematic error is the difference between MC reco and MC truth. Variations

Updated by Rogers, Chris 10 months ago ยท 8 revisions