D. Abbaneo, M. Abbrescia, E. Barberis, C.
Bedoya, M. Dallavalle, J. Hauser, K.
Hoepfner, V. Khotilovich, S. Krutelyov, D.
Nash, P. Paolucci, A. Safonov, A. Sharma, D.
Trocino

• Near tagger ME-0 at the back of present HE
• Coverage: 2.1<| h
|<4.0
m new HE
• Upper portion of 2.1<| h
|<2.5 has trigger capabilities
• Lower portion is only used in the offline
2

• Scenario A: recover the high eta muon from H->ZZ
– Look for the 4th muon in events with 3 reconstructed leptons
– Increasing muon coverage to h
=4.0 adds about 50% to the current CMS acceptance
– The vertex in this scenario is known (tagged by 3 leptons)
• Scenario B: a search for exotic two photons plus a muon events
– Will also see a large increase in acceptance
– More difficult as the vertex is not known (photons don’t tell you which vertex is yours)
• Question: can we find the 4 th muon effectively?
– If the new system reconstructs a lot of “muons” (fakes) in every event, we will conclude that the answer to above is “No”
• Qualification of “a lot” is analysis dependent (having a few fakes that don’t fit the kinematics of your search signature can often be okay)

• Search for H->ZZ->4mu as 3mu+forward pixel extension track in events with 3 reconstructed leptons
• Scenario A requirements: we don’t want many reconstructed “4 th muons” in either signal or background
– Tracks of interest: p
T
>5 GeV (p~25-100 GeV at this h
)
• Simulation at generator level (@13 TeV): 𝜂=4 𝜂=2
∞ 𝑝
𝑇
=3 𝑑
2
𝑁 𝑑𝜂𝑑𝑝
𝑇 charged particles/side per minbias event 𝑑𝜂𝑑𝑝
𝑇
= 0.06
• Start at lower p
T
(3 instead of 5) to be conservative, double the number to account for photon conversions
• N trks
<0.25 “muons” (fakes) in 2.0<h<4.0 per BX (we know the vertex)
• Scenario B: vertex is not known, need to multiply by N
PU
– N trks
~50 “muons” (fakes) for N
PU
=200 per BX
• Summary: one can possibly get away with a tracker-only solution for such overconstrained signatures as HZZ, but in general the fake rate is way too high

• A new muon system is used to look for the 4 th muon in events with 3 reconstructed leptons
– Assume no tracking, so a muon is just a hit and no momentum information
• Scenario A: we don’t want many reconstructed “4 th muons” fitting reasonable analysis selections in either signal or background
– Assumes a well-shielded and multi-layer new system, we use CSC LCT numbers to extrapolate to higher eta coverage (1.3 LCTs/BX at PU=25)
– Number of LCTs in CSC: N m
(CSC)~ 1.3/ BX at PU=25
• Dominated by soft muons (neutron contribution is suppressed) from all sources (b-jets, decays in flight, decays in the calorimeter/punch-through)
– The threshold to get through the material is over 3 GeV in p (not p
T
)
• Simulation: the mean number of tracks with p>3 GeV in 1<| h
|<2 (CSC coverage): N~1.2 per min-bias event
• The mean number of tracks with p>3 GeV in 2<| h
|<4 (ME-0 coverage): N~13 per min-bias event
– Final estimate for the total number of “muons” in ME-0:
– N m
(ME-0, PU=200) = N m
(CSC, PU=25) x 200 / 25 x 13 / 1.2 ~100 muons per BX at PU=200
• Conclusion: this will not work even for H->ZZ->4mu (equally bad for either physics scenario)

• Minimal requirements for muon extension:
– What is required detector segmentation?
• Should be less than multiple scattering for target range of muon pT
– At these h
, p
T
~5 GeV corresponds to p~20-100 GeV (“Higgs range”)
– Can it work for offline?
• Do we get so many reco tracks matching a single muon hit within the matching window that such tagging stops being useful?
• Do we get too many muon hits so that an event has too high probability of having a high momentum muon reconstructed given the matching windows used?
– Assume adequate shielding and multiple redundancy so that muon hits are primarily due to muons (primary or secondary) like in the current system.
– Can it improve momentum measurement using larger lever arm?
• What segmentation (spatial resolution) are needed to improve track fit momentum measurement?
– Is the bending of the track in the magnetic field at the muon tagger z exceeding what is expected deviation due to multiple scattering at the targeted pt? Is segmentation sufficient to measure the difference?

IP
Daniele Trocino et al (NEU) gen track
“sim-hit”
P
SIM
P
GEN
Δ
M
Δ
P
P
REC
Δ
MP reco pixel track
560 cm
• Estimate multiple scattering:
– Use FastSim muon gun with p and current muon propagator
T
=20 GeV
• Assumes the same amount of material as in the current detector
– New muon detector at z=560 cm
• For p
T
=5 GeV, multiply RMS by ~4 p
T
=20 GeV

Mean bend in magnetic field (cm) p
T
=20 GeV h
~2.5 2.5 cm p
T
=20 GeV h ~4.0 0.1 cm p
T
=5 GeV h
~2.5
10 cm p
T
=5 GeV h
~4.0
0.4 cm
Very high p
T track 0
RMS due to multiple scattering cm Dh
Propagated pixel reco uncertainty at muon detector surface
2.7 mm 0.003
?
1 mm 0.005
?
10 mm 0.012
?
4 mm
0
0.02
0
?
0.2 mm
• Bend is currently back of envelope (can be off by up to a factor of 2-3)
• Accuracy in propagating a straight pixel track: assume 100 m m precision at Z=250 cm (halfway between IP and muon system) and perfect vertex finding (we are in offline!)
• A 2-sigma matching window size (determined by multiple scattering) range
Dh x
Df
~0.05x0.04=0.002 for p
T
=5 GeV
– Assume we are not constrained by the too crude detector segmentation here

• We are matching tracks and muon hits
– The aim is to minimize fakes from accidental overlaps of muon hits and
“energetic” pixel tracks (pT=t qualifies as energetic)
– The matching window is about
Dh x
Df
=0.002 (p
– Total area of the detector:
Dh x
Df
=2x2x p
=12
T
=5, for higher p
T it’s smaller)
– Total area fits 6000 matching windows
• Scenario A (event vertex is known):
– About 0.25 pixel tracks and about 100 muon hits per BX
• Crudely, an average number of accidental matches with p
T
>5 GeV: N~ 0.25 tracks
*100 muon hits / 6000 windows ~ 0.004 fake muons per BX
• Scenario B (event vertex is not known):
– The number has to be multiplied by N
• For PU=200, N~0.8 fake muons/BX for p
T
• For p
T
>10 GeV N~0.08 fakes
PU
=5 GeV
• Very crude estimate, additional muon-tracker compatibility checks can presumably reduce this figure further down, detailed studies are needed
• However, the conclusion is clear: such system will work and will be a general purpose system

• The new system can potentially contribute to momentum measurement
– At h
=4.0, the bend is equal to the amount of multiple scattering
• Seems like a showstopper
– But at h
=2.5 the bend is x10 larger than multiple scattering
• For p
T
=100, it is 5 mm at h
=2.4, should be easily measurable
• This need to be studied
• Tracker-only resolution for tracks with p
T
=2, 10 and
100 GeV

• Neither tracker alone nor muon system alone can provide a general purpose muon tagging in the very forward region
• A combined system of forward pixel extension and a new multi-layered muon detector meet the requirements
– Redundancy is important, else we will be swamped by muon backgrounds!
• The best place for the muon system is in front of ME-1/1:
– At ME-2/1 the area of the matching window is twice larger doubling the fraction of overlaps
– If we hope for any kind of momentum measurement improvement (or confirmation) from the muon system, it cannot be further away than ME-1/1
• Else magnetic field bend is too small due to radial field while multiple scattering is larger

• L1 muon momentum resolution can scale be improved with a second detector:
– Inner tracker tracks at L1 – best solution, but not available until LS3
• A second muon system could improve momentum resolution if one can measure the “bending angle”
• Used in the Barrel; CSCs are too thin (~11 cm) to see the bend
• A new detector in YE-1/1 (least affected by scattering, largest B)
– Increase “lever arm” from 10 cm to
D z=30-50 cm (physical constraints)
– Need ~2 mm or better trigger resolution to effectively discriminate 5 GeV
YE-1/1 muons from 20+ GeV ones
D x
D z=30 cm p
T
=5 GeV p
T
=20 GeV
12±3mm 3±1mm
13

“Close” chamber pairs
“Far” chamber pairs
Df
(CSC-GEM)~ 4 mrad
D x ~8 mm p
T
=20 GeV p
T
=5 GeV
V. Khotilovich et al. (TAMU)
• GEM-CSC bending angle measurement using full GEANT simulation:
– Muons with PT=5 and PT=20 GeV
– GEM L1 Trigger pads (4 strip OR)
– Good discrimination and a powerful new handle on p
T resolution
• An “OR” of two GEM chambers within a super-chamber is ~100% efficient
14

V. Khotilovich et al. (TAMU)

• Muon Level-1 Trigger will rely on tracking trigger and Muon matching
• CMS is in a real danger to lose triggering capabilities in 2.1<| h
|< 2.4
• Efficiency losses in L1 Track Trigger are due to tighter selections
– Can loosen, but fakes will shoot way up
• This is the exact same region where muon trigger rates shoot up:
– Weakness in momentum measurement causes trigger rate to shoot up by x5
Reco’ed stubs p
T
>2
Stubs from true particles w/ p
T
>2
Muon gun p
T
>5 GeV
Efficiency includes track finding only. No muon system inefficiencies incorporated.
17

h
•
h
•
– Create a category of “loose” L1 tracking trigger track candidates for forward muon triggers
• Less hits means less accurate momentum measurement and allowing fake rate to shoot up
– Match track to an improved muon system
• Let muon requirements suppress the rate increase associated with loosening tracking selections
– Optimize combined selections to keep efficiency high and trigger rate low
• Assume that between the two systems we will have enough information to make a sensible decision and keep the L1 output rate at the acceptable level

• New systems: a near tagger ME-0 and redundant systems GE-2/1, RE-3/1,4/1
– GE-1/1 is assumed to be in, but it is not relevant for this discussion (beyond h
=2.1)
• Purpose: attempt to re-utilize the bending angle idea and the additional redundancy to reduce trigger rate
GE-1/1
• Evaluate gains of a combined system and separate impact of each of the two parts
• Evaluate required parameters of each system

S. Krutelyov et al. (TAMU)
• Assume detectors with perfect spatial resolution
– But faithfully simulate multiple scattering
• Measure bending angle in YE-
1/1, 2/1, 3/1, 4/1 for pT=10
GeV
– Assume muon trigger threshold of ~20-25 GeV: the rate is dominated by mismeasured softer muons
Width is driven by multiple scattering and by dependence of bend on eta within a chamber
– Important to have a handle on momentum measurement for softer muons
• YE-3/1 and 4/1 have almost no power:
– Tracks bent back by radial field, any residual bend is smeared by multiple scattering

ME-0 only
GE-2/1 only
Combined
Demand presence of combined ME0-
ME1/1 stub
Demand combined
ME0-ME11 stubs and stubs in YE-2, 3, 4
Combined + 4/4 stubs
Signal:
Muons w/ p
T
=30 GeV
Ideal trigger rate reduction
95-98%
95-98%
98+% (?)
98+% (?)
95% (?) x3-4 x1.9-2 x5-6 (?)
N/A
N/A
N/A
Assumed realistic resolution
Realistic rate reductio n
Limitation
~0.5-1 mm x2.8-3.2
1-2mm (?) x1.4-1.5
Resolution of
ME-2/1 x3-4 (?) x1.2
x1.5
x5-6
Start using new additional redundancy
Assume full redundancy
(GEMs and
GRPCs)
ME-0

•
– The only other alternative is to replace ME-1/1 with a
“thick” high precision detector
– ME-0 alone will provide about x3-4 in rate reduction through the use of bending angle and additional redundancy
•
– GE-2/1 can have lower segmentation compared to GE-1/1, thus lower electronics cost
– GRPCs can also have lower segmentation (if we only need a confirmation)

– By far the most effective solution for extending CMS muon coverage is the near tagger
• Implies forward pixel extension
• Moving the tagger further out causes a fast increase in the rate of coincidences (many fakes per event) and drastically reduces ability to do any kind of momentum measurements above 10 GeV