Section 3 - Muon reco, ID, and samples (ll.110-272)

Q3a (006): L112: “, which induces signals on electrified strips or wires.”. It’s the mouvement of the charges produced by the the primary ionizations that induces an electric signal. Moreover, I wonder whether the CSC or RPC strips are “electrified” (if that means keeping them at a certain fixed potential different than ground). Minimal proposal: “which eventually cause electric signals to be induced on the wires and strips of the detectors”.

A3a (006): Comment accepted: Muons and other charged particles that traverse a muon subdetector (CSC, DT, or RPC) ionize the gas in the chambers, which eventually causes electric signals to be induced on the wires and strips of the detector.

(Boris) - "The RPC strips are grounded via resistors."

(Tim) - I have some sympathy with not liking 'electrified strips or wires' but for the CSCs the charge on the strips is definitely not from the movement of charge - it is the standard MWPC operating principle that the cathode charge is 'induced' by the avalanche charge around the anode. Maybe somebody has since demonstrated Charpak was wrong? However, I think the suggestion is good: “[...] ionize the gas in the chambers, which eventually causes electric signals to be induced on the wires and strips of the detectors”. We could argue for ever whether 'induce' is being used in a non-technical or technical electromagnetiic sense here, but it seems to me to work.

Q3b (006): L114: the definition of a “hit” is not clear. Is it a reconstructed particle crossing position, i.e. a measurement with associated uncertainty? Or is it rather a generic short for the particle crossing at a given position in a detector ? Need to clarify.

A3b (006): Try to get the wording clear, and to begin to make a consistent use of "reconstructed" hits: These signals are read out by electronics and are associated with well-defined locations, generically called "hits", in the detector. The precise location of this hit is reconstructed from the electronic signals using different algorithms depending on the detector technology. The CSC and DT chambers are multi-layer detectors, in which hits are reconstructed at each layer. From the reconstructed hits, straight-line track ``segments'' are built within each CSC or DT chamber.

(Tim) I'm a bit upset about this comment considering it was my idea to try to define 'hit' in the first place - a term people in CMS bandy about without ever worrying much what it means. And, as far as I can see, I already wrote EXACTLY what this commentator asks for! "These signals are read out by electronics and are associated with well-defined locations, generically called “hits”, in the detector." It says 'generically' and doesn't mention 'measurement' or 'uncertainty', which of course it shouldn't!

Q3c (006): L115: Segments are not built in RPC. In addition it may be worth recalling that the straight line model is appropriate for DT and CSC as the residual magnetic field is negligible with the volume of the chambers.

A3c (006): Split the hit and segment sections. Describe DT hits, CSC hits, then RPC hits, and introduce segment part as: While the RPC chambers are single-layer detectors, the CSC and DT chambers are multi-layer detectors in which hits are reconstructed at each layer. From the reconstructed hits, straight-line track ``segments'' are built within each CSC or DT chamber.

Boris: We agree that segments are not built in RPC. The sentence on L115 is connected to sentence L114 where only DTs and CSCs are mentioned. So, the segments are specific for DT and CSC. The hits in the RPCs are a group of adjacent simultaneously fired strips. The RPC strips is characterized by it's position, dimension and time.

(Tim) Seems fine to me already. We only mention segments for CSC and DT. I wouldn't want to make a statement about the residual magnetic field being negligible in the ME1/1 chambers! It certainly isn't, otherwise we'd have no reason to use tilted anode wires (which DO improve the resolution, as can can easily be confirmed by doing as I have done accidentally, running the simulation with the local magnetic field in the wrong direction and finding that the resolution degrades a lot.) Despite that, it is clear that segments in ME1/1 are generally pretty straight!

Q3d (006): L118: “chambers/stations” the confusion continues…. ; also remove “often together”

(Tim) I agree with this comment. The sentence is "Combining information from multiple chambers/stations of the CSC, DT, and RPC detectors, and often together with data from the silicon inner tracker, muons are reconstructed and identified for CMS physics analyses." I personally try to avoid structures like 'chambers/stations' because I rarely know exactly what the intended meaning is. Maybe we can come up with a more precise sentence together?

Q3a (001): L118 “and often together with data from the silicon tracker”: a global muon “always” has information from the silicon tracker otherwise is called Stand Alone Muon, see also ll.159-161

Q3e (006): L120: The sentence “A detailed description…” reads odd when one notices that the following subsection describes local reconstruction. If the description given in the cited paper is more detailed than the one provided here, then please keep this sentence (changed into “ A more detailed description…” for the end of section 3.1. Same thing for the inner tracker and global muon reco.

A3e (006): Done

Q3b (001) and Q3a (003) and Q3h (006): LL120-134: Can you give more information on calibration procedure. Is this done using data or simulation? The words “very close to the wire” is vague. How is this determined ? Please specify in the text. “The system”. Is that the local readout electronics? To the Muon Trigger? CMS? How is the reader supposed to know what “the system” is.

A3b (001): Rewrite DT reconstructed hit description as: Hit reconstruction in a DT drift cell specifies the transverse distance between the wire and the intersection of the muon trajectory with the plane containing the wires in the layer. The electrons produced through gas ionization by a muon crossing the cell are collected at the anode wire. Their arrival time is registered by a Time-to-Digital Converter (TDC). The position of a DT hit is reconstructed using the trigger time (T_trig) and the electron drift velocity (v):
position = (T_TDC - T_trig) x v.
The T_trig accounts for the time it takes for a trigger decision to arrive at the chamber electronics after the passage of the muon through the chamber volume. It includes the generation of trigger primitives, the processing by the L1 trigger electronics, the distribution of the L1A, and the receipt of the L1A back at the readout electronics on the chamber. Calibration of T_trig is done heuristically, by centering the data distributions of residuals at zero (see Section 4). This effectively makes the drift time (T_TDC - T_trig)=0 for muons which cross the chamber negligibly close to the wire. The calibration is redone whenever something changes in the detector conditions, such as gas conditions or time shifts in the electronics to optimize trigger efficiency.

Francesca:
- Re-reading the paragraph, I believe that "distance of closest approach" is not correct. The distance is measured in the plane transverse to the wire and is the distance between the wire and the intersection of the muon trajectory with the wire layer plane (i.e. the plane containing all the wires of the layer)
- Calibration is done using real data and redone whenever something changes in the detector conditions (e.g. gas conditions, time shifts tuned in hardware to optimize trigger efficiency)
- tTrig is such that T_TDC-tTrig=0 when the drift path is 0, so when the muon crosses the cell in the wire position. Technically this cannot be evaluated directly from the time distributions as 0 drift path yields 0 ionized charge. The present algorithm makes use of residual distributions: tTrig is optimized to have them centered at 0.
- tTrig is the time between the muon crossing and the arrival of the L1A signal to the superlayer, triggering the TDC readout. Therefore tTrig comprises all the trigger chain: from the generation of trigger primitives, to the L1A and back to the readout electronics sitting on the chamber.
- the time taken by signal propagation along the wire cannot be taken into account by the calibration, being the tTrig an average value for the whole superlayer. (nonetheless it is taken into account by the local reconstruction but this is another story - see A3g (006))

Q3f (006): LL123+1 “in a DT cell” to be consistent with previous definitions. Also, the local reco in DT does not aim to estimate “the distance of closest approach to the wire” (if it did, the segment reconstruction would be wrong as it assumes that hts are on the plane of the wires) , but rather the distance between the wire and the point of intersection between the particle trajectory and the plane containing the wire and parallel to the largest sides of the station.

A3f (006): Please see above: A3b (001)

Q3g (006): LL124-137 This whole simplified description lacks the information that a correction for the propagation of the signal along the wire can also be applied once information from other SL is available; you should discuss this point as I believ that segment reconstruction is actually performed iteratively (later on you should then specify which segments are considered for the resolution study, and which for the track property study). “to propagate through the system” is very vague; maybe “to become available after the bunch crossing has occurred”

(Francesca) In the resolution studies we use segments produced by 2-parameter fit at the end of the iterative procedure: so corrections for signal propagation along the wire are applied for MB1, MB2, MB3 where the information on the position along the wire is provided by the orthogonal plane of wires, not for MB4 that only has two superlayers of parallel wires. This explains the poorer resolution observed in fig.3 for MB4, as mentioned in line 302.

Q3b (003) and (007): L130:

• “New” as in “new physics”? Probably not, but “new particles” is a loaded term.
• heavy particles that travel

Q3i (006): L132: “within a chamber” a DT chamber ? Please specify in the text

Q3a (002): LL132-137: is this new compared with Run 1? If yes, state it clearly.

A3g (006) and A3b (003) and (007) and A3i (006) and A3a (002): Modify DT segment reconstruction to read: Segment reconstruction in the DTs was modified prior to Run~2~\cite{REF-DT-MEANTIMER}. The calibration of T_trig in the DT reconstructed hit position (Eq.~\ref{eq:DTHitPosition}) implicitly assumes that all muons take exactly the same time, traveling at the speed of light, to reach a given layer from the interaction point. However, this assumption is not exactly true since there is an intrinsic time-of-flight spread related to the muon momentum, the exact position in the layer, and the propagation of the signal time in the wire. Moreover, hits could come from muons originating from other bunch crossings ("out-of-time muons"), or they could be produced by heavy particles that travel at a reduced speed. Any such shift in the muon crossing time would cause all hits produced within a chamber to be shifted in space by the same quantity. In order to take these time shifts into account, segment reconstruction in the DTs is performed as a three-parameter fit, which includes the muon crossing time as a third parameter in addition to the standard two-dimensional straight-line fit of position (transverse to the wire direction) and slope. Spurious early hits, produced by delta rays, are removed from the segment reconstruction. This procedure yields an optimum hit time alignment and thus improves the spatial resolution. Moreover, it also provides the best estimate of the muon crossing time to be used in physics analyses.

Q3j (006): L143: add something like “along the muon path” after weakens

A3j (006): Rewrite as: "In the endcaps the solenoidal field is first parallel to the z direction but then diverges radially, so a muon is first deflected in one azimuthal direction and then deflected in the other, with the maximum deflection occurring in the first station."

(Tim) The sentence is "In the endcaps the magnetic field changes direction and weakens so the maximum bending of a muon track occurs in the first station, and subsequent stations see a smaller bending." I tend to agree that we're not stating things very clearly. Perhaps something like:
"In the endcaps the solenoidal field is first parallel to the z direction but then diverges radially, so a muon is first deflected in one azimuthal direction and then deflected in the other, with the maximum deflection occurring in the first station."
(I aim to suppress 'track' - it's a muon - and 'bending' - which seems a vague concept related to a visualized track rather than an actual muon : ) )

Q3k (006): L144: “typically” ?? Why not being accurate and state that they are othogonal except in ME11 ?

A3k (006): Take Tim's suggestion: "The wires are orthogonal to the strips, except in ME1/1 where they are tilted to compensate for the Lorentz drift of ionization electrons in the non-negligible magnetic field in this region. They are ganged into wire groups of..."

Q3l (006): L148: “promptly in a pp collision”-->”promptly in the triggering bunch crossing”

A3l (006) Done

(Tim) I think I agree with this. In the context of pileup and backgrounds we have pp collisions which are not at BX0 and those are not calibrated to T=0!

Q3m (006): L157: “An RPC hit is the strip cluster centroid” from this sentence, it may seem that the previously given vague definition of hit is the central value of the position measurement (so without uncertainty…). Please, make everything consistent

A3m (006): attempt to call "hits" as "reconstructed hits" and otherwise, if the context is clear.

(Tim) I guess the issue now is that we should avoid talking of 'reconstructed hits' as 'hits', if we want to use 'hits' to refer to generic, detector-independent points where a muon has crossed a detector element. We tend to use 'hits' colloquially to cover many different contexts and situations. CMS has solved that in principle by using 'simhits', 'rechits', etc. with specific formal definitions for different subdetectors. SO, in this particular case, I guess we need to say 'An RPC reconstructed hit...' (I presume we want to avoid 'rechit' as an abbreviation.) We could also go into a long rigmarole about all the different contexts for the use of 'hit' but it would be mind numbing.

Q3o (006): L160 “inner tracks” not defined; you should replace it with “position measurements from the inner silicon tracker detector”. Moreover the whole paragraph 167-170 comes out of the blue and gives a lot of (unnecessary) details without specifying that they only concern the seeding step, while the actual track finding and fitting are always the same). It could make sense to merge it with the following one, where tracker muons are explained.

Q3c (003): L163: "Muon candidates” or something similar to make clear that these are requirements on the evidence of the muon from the detector?

Q3a (005): L164: Make clear that the stand alone muon reconstruction requires the muons to have penetrated at least two chambers (so that it is clear afterwards why the tracker muon is more efficient at low momentum)

Q3c (001): L166: Is the geometrical matching done using a fixed dR? With which value? Otherwise, how is the matching done?

A3c (001): POG - can you please clarify how the geometrical matching is done?

Q3b (005): LL172-176: Related to comment at L164. It is not clear from the text why the tracker muon reconstruction performs better for low pT muons or critical detector regions with respect to the tracker or global muon reconstruction. In fact, you haven't specified that the algorithm just require one matched station in the muon detector. Hence, it is not clear with respect to what the tracker muon reconstruction increases the probability of misidentification.

Q3b (002): LL174-76: the sentence is misleading: most of the tracker muons are matched not only to a segment in the innermost station, but to segments in other stations as well. Are you talking of tracker muons that are not global muons? Please reword.

A3b (002): POG - can you please clarify the description of Tracker Muons matching to one (or more) stations?

Q3p (006): L173 “cracks” would probably need to be defined as “non-instrumented regions” ... or ... perhaps “gaps” ... or ... may want to explain chimneys in half a sentence, in order not to give an impression that cracks were left in the detector for no good reasons

Q3q (006) and Q3c (005): L176: "punchthrough" is explained nowhere. Also, there is some notation inconsistency in the text ("punchthrough" or "punch-through", see e.g. L221, 233).

Q3c (001): L177: Is the geometrical matching done using a fixed dR? With which value? Otherwise, how is the matching done?

A3c (001): POG - please specify how the geometrical matching is done to combine standalone muons with inner tracks to reconstruct global muons? (geometrically matched on common surface or some such)

Q3r (006): L180 “identification purity”; purity is not defined

Q3s (006): L186: Are the tracks rebuilt or only refitted (by the way note that you need to write “refitted” and not “refit”) ? It is unclear how a simple refit would change the hit efficiency.

A3s (006): POG - The inside-out muon-specific tracking iteration actually "rebuilds" tracks.

Q3e (005) and Q3g (005) and Q3f (005) and Q3h (005) and Q3i (005): description of PF

• L191: ‘Muons are included as part as the CMS particle-flow algorithm' is an ambiguous sentence, as it leads the reader to believe that all muon candidates are included in the PF particle list. This is not the case, as not all standalone, tracker, and global find their way to the PF algorithm. They are submitted to a selection that maximises the efficiency for prompt muons, HF muons and light-hadron decay-in-flight muons, while minimising the contamination from other particles (esp. punch through). In addition, and FYI, in Run1, particles that were not identified as tracker or global muons (typically standalone muons) may be salvaged in the PF algorithm, either because a local deficit on calorimetric energy is found in a PF block, or because the global event properties (esp. the MET) would behave better if a particle is called post-identified as a muon rather than as a charged hadron. See Sections 4.4 and 4.6 of the PF paper.
• L192: ‘The PF reconstruction exploits information from different sub detectors to identify individual particles (e.g., electrons, photons, charged or neutral hadrons, or muons)’ is inaccurate. First, "identify" should be replaced by "identify and reconstruct" (in this order). Second "individual particles" gives the impression that the PF identifies (and reconstructs) only a subset of the particles in a given event. It actually aims at a global event description, with the identification and the reconstruction of all individual particles in a given event. Third, the PF reconstruction provides only a list of electrons, photons, charged and neutral hadrons, and muons", so "e.g." gives the wrong impression that other particles may be identified and reconstructed.
• L193: ‘The PF reconstruction ... builds high level objects (e.g., jets or missing transverse momentum)’ is a wrong statement. The PF reconstruction "only" builds an inclusive list of particles (aka global event description) for each collisions, later digested by the POGs to provide "high-level" objects.
• L194: ‘by means of a global event description paradigm’ is a meaningless succession of words, which brings only confusion to the already wrong and inaccurate preceding statements. If "global event description" is to be kept, it has to be introduced as simply explained above.
• L195: ‘Muons are identified by PF means of track quality and isolation requirements’ is at best very incomplete, and actually wrong, a statement. The PF develops a whole strategy to improve the efficiency of non-isolated, non-prompt, muons, without increasing (and often decreasing) the fake rate, in view of providing CMS with the best physics objects afterwards.

Q3d (001): LL196 - 212 would fit better after the paragraph from LL177-182

Q3u (006): LL196-211: 1) it is not clear which are the candidates that are refitted. 2) a number of algorithms are listed, but the way they work and above all how the Tune-P chooses among them (or combine them?) remains very vague: how goodness-of-fit and sigma/pT are combined ? What is the relationship between these refits and the previously described fits (tracker muons, global muons, stand-alone) ? Are the previously described fits used at all apart from providing their individual measurements to these refits ? sigma(pT) is not defined

Q3v (006): L202 “showering” is not defined

Q3w (006): L206 and 244 - are delta(pt) and sigma(pt) the same thing ?

Q3x (006): L206 “goodness-of-fit” is not defined.

Q3d (003): L207: As is, this seems like jargon. To “reduce tails” of what distribution?

Q3y (006): L215 chi^2 of what ?? “Track fit chi^2”

Q3z (006): L217 “primary vertex” should be better defined than with “i.e. the hard interaction”

Q3aa (006): L218: “sum of energies from tracker and calos” are you double counting ? Is this intended for an expressly vague introduction or is it supposed to be truly informative ?

Q3e (008): L219, (a) since the "delta eta**2" (the "delta phi**2" is the same) may be looked ambiguous as either "delta((eta)**2)" or "(delta(eta))**2", in order to be less confusing, it's better to add two pairs of bracket in the formula of delta(R) as many other CMS papers have used; (b) as the variable "phi" has taken the unit of degree on L76, etc., but in delta(R), it should have the unit of radians, so this should be specified explicitly here, i.e. "delta(R) = sqrt (delta eta**2 + delta phi**2)" --> "delta(R) = sqrt ((delta eta)**2 + (delta phi)**2), where phi is in radians"

Q3c (002): LL220-221, "The matching of inner tracks with muon chamber segments is a useful method... to identify muons from light hadron decays." This sentence may be misinterpreted as if we can distinguish muons from decays in flight from muons from other sources, which is not true. Are you trying to say that muons from decays in flight are more often reconstructed as tracker muons than global muons? Please explain better.

Q3bb (006): L223 “tightness of the matching” is not defined

Q3k (005) and Q3cc (006): L226:

• We propose to call this ID: ‘PF Muon’, and insert here some reference to the global scope of the PF reconstruction to explain which categories of muons are retained by this ID and how (see previous comments on chapter 3.2). In the rest of the paper you use PF Muons or Loose Muons, we propose to stick to PF Muons and explain that are also called Loose Muons in CMS physics analyses.
• “prompt muons” are not defined. It would also be useful to specify what we define heavy flavour decays. Does Loose mean: PF AND (GlobalMuon OR TrackerMuon) ? Making this crystal clear would help the reader; also clarifying that PF is another identification algorithm on the same footing as Global and TrackerMuon would increase clarity.

A3k (005) and A3cc (006): PF muon ID aims for high efficiency... PF muons are all global or tracker muons identified by the PF algorithm [PF reference].

Q3e (003) and Q3e (001): LL226-237:

• Can the traits of the muon classes be described in terms of inheritance from previously mentioned classes? E.g., can it simply be said “A [type 2] muon is a [type 1] muon with the following extra criteria:…”?
• How tight are thece criteria relative to each other. Please make clear if for example medium is a strict subset of loose or not

A3e (003) and A3e (001): POG - Both Medium and Tight are a subset of Loose, but Tight is not a subset of Medium. What can be made more explicit in the text is that Tight is a Subset of Loose (e.g. by replacing part of LL333-334 with: Tight muons are loose muons identified both as global and tracker muons that have constraints ...), for Medium this is actually already mentioned.

Q3dd (006): L230: “well reconstructed” ??? “matching” ??? “trajectory kinks” ? “segment compatibility” (which segments,? Compatible with what ?)

Q3ee (006) and Q3l (005): L232:

• “decays in flight” should correspond to what was previously referred to as “light flavour decays” isn’t it ? Maybe increase consistency ?
• This is true also for the Medium Muon ID, indeed both the Tight and the Medium have the same 'fake-rate'. The only difference is that the Tight is 2% less efficient both on the prompt and the muons from heavy-flavour decays. We understand several CMS analyses still uses the Tight ID and therefore you want to document it, but we think there's actually not a 'conceptual' difference between Tight and Medium.

Q3gg (006): L237: purity with respect to what ? Contamination from non muons or from muons that are not from heavy flavour decays ?

Q3hh (006): L241: to be consistent with what was written previously you must aim at something for these muons with pT>200 GeV: efficiency, purity, resolution, reduction of momentum measurement tails, all ???

Q3ii (006): L244 relative uncertainty in pT had been already defined with a different symbol in l206. Pick-up one and consistently use it throughout the text

Q3jj (006): L246: “relative isolation” not defined

Q3kk (006): LL246-252: it is not clear if this isolation is part of the previous identification algorithms or not. “particle flow candidates” has not been defined. The reason for considering two strategies is probably to take into accounts neutral particles too. You should specify that.

Q3m (005): LL248-252: Details like efficiencies are provided for the isolation working points but not for the IDs, why? Also the definition of the samples being used to tune their working points seems not to belong to this chapter. Instead the description of the isolation variable could be extended quite a bit (which dR, which PF candidates, how to deal with PU subtraction..) or put a reference for it.

Q3f (001): L249: This line indicated that the isolation was not tested on data?

Q3c (007): L252: not clear what W+jets events you use: do you require the W to decay into a muon? Please, expand.

Q3ll (006) and Q3n (005) and Q3pp (006): pileup

• L254: it is of fundamental importance to specify the average pile-up expected in these two periods
• LL270-272: Performances are studied separately for 2015 and 2016. I guess therefore the samples are reweighted to the PU distribution of 2015 and 2016 separately according to which data are they compared to.
• L272: is this 11 the pileup in 2015 and 2016 (see previous comment – if the answer to this previous comment is a different pileup value then you should explain here how you reproduce the conditions in the two data periods) ?
• L272: "with an average pileup of about 11 interactions per bunch crossing": this number was 15 in L98.

Q3d (002) and Q3mm (006) and Q3e (002): physics in samples

• LL256-257: how low "low pT threshold" is -- do these samples include muons from J/psi and Upsilon?
• LL261-272: Add here also the description of QCD sample and remove it later from L528
• Section 3.4: did not the set of samples include other usual sources of dimuons like dibosons (cf. L420) and QCD?
• L262: why only 2015 data is mentioned here? Add 2016.

A3d (002) and A3mm (006) and A3e (002): POG, DT, CSC, RPC - in the analyzed data samples, how low was the "low pT thresholds"? Were they low enough to including J/psi, etc (see Fig 16)?

(Francesca) In the sample I used (/SingleMuon/Run2016G-ZMu-23Sep2016-v1/RAW-RECO) I see 1 muon has pt>20 GeV, the other one pt> 10 GeV

Q3f (007) and Q3oo (006): LL265-267: sample citations

• L263: are generated at next-to-leading order (NLO) with {\sc MadGraph5_aMC@NLO} 2.3.3 [13].
• background from single top quark tW production is generated at NLO with {\sc powheg} v1.0 [14] (I believe the powheg version used is 2.0 and not 1.0)
• The {\sc pythia} 8.2xy [15,16] package ... of the underlying event via tune CUETP8M1 [xx]. (please specify the exact pythia version) ; Overall, please use \PYTHIA, etc. for all generator names. Also, the version of the generators should be put in parenthesis e.g. \PYTHIA (v8.0) etc.
• NNPDF2.3LO [17] set [no space] (which sample you are referring to ? Please cross check. I believe the NLO version of the NNPDF is used for the NLO sample mentioned on line 265)
• {\sc geant4} package [18];

-- CarloBattilana - 2017-12-18