1. Julie Kirk Rutherford Appleton Laboratory
ROI based B-triggers
Julie Kirk
Rutherford Appleton Laboratory
Julie Kirk, RAL
ATLAS UK Physics Workshop, Durham (20/9/06)

2. RoI guided approach
At low luminosity (< 2x1033) introduce a single muon trigger at LVL1 in addition to the di-muon trigger used at high lumi.
 tags one b – then try to reconstruct other b (the one of interest)
2 approaches for track reconstruction at LVL2 : Fullscan of ID or RoI-guided (use secondary LVL1 Regions of Interest (RoIs) to guide LVL2 reconstruction).
RoI based approach :
lower execution time than full-scan: depends on RoI size and mean multiplicity of RoIs
lower efficiency for low pT B (slow turn-on threshold)
Outline :
LVL1 JET and EM RoI multiplicities for bb→X
EM RoI triggers (J/ψ→ee and rare radiative decays ((e.g. Bd→K*0γ, Bs→γ):))
Muon RoI triggers (J/ψ→ and rare decays b→(X))
For Jet RoI - comparison of fullscan and RoI guided approach for Bs→Ds((KK)π)π
Julie Kirk, RAL
ATLAS UK Physics Workshop, Durham (20/9/06)

3. LVL1 RoI multiplicities
RoI multiplicity bb→(6)X
Limits on cpu and bandwidth for B-triggers => need to be fast, efficient and selective.
If retrieve information for smaller region of detector => faster execution times
Overall speed depends on RoI size and mean LVL1 RoI multiplicity per event
RoI multiplicity required to be about 1-2 to keep resource needs reasonable => determines thresholds chosen.
Higher ET threshold → lower rate but also reduce efficiency
With required multiplicity efficient for:
EM RoI => electron pT>~5GeV
Jet RoI => B-hadron pT>~10GeV
Jet RoI
EM RoI
Julie Kirk, RAL
ATLAS UK Physics Workshop, Durham (20/9/06)

4. EM RoI Bd J/y (e+e-) Ks(p+p-) J/ψ->e+e- e+ e- TRT EM Calorimeter
LVL1: 1 μ (> 6GeV) + ≥1 EM RoI
LVL2:
confirm LVL1 muon and EM cluster
reconstruct tracks in enlarged area around EM RoI (ΔxΔη=1.2x1.2)
search for J/ψ, combine opposite sign tracks apply mass cuts
Efficiency after LVL2 ~ 68% (both e+/e- pT>5GeV)
background ~170 Hz
Julie Kirk, RAL
ATLAS UK Physics Workshop, Durham (20/9/06)

5. EM RoI (2) Rare radiative decays (e.g. Bd→K*0γ, Bs→γ):
Possible with fullscan, investigating possibility of using RoI based trigger
Previous “trigger-like” analysis using offline code predicted for 30fb-1 :
15,000 Bd  K*0γ
4,800 Bs γ
Rate after EF: 0.6 Hz Bd  K*0 γ
0.5 Hz Bs  γ
Need large RoI (half-width ~1)
Δφ (γ - daughter)
Now repeating analysis using trigger code
Δη (γ - daughter)
Julie Kirk, RAL
ATLAS UK Physics Workshop, Durham (20/9/06)

6. J/ψ→+- LVL2: Efficiency 68-77% (1st  pT>6Gev, 2nd  pT>3GeV)
Single muon at LVL1 (22kHz)
LVL2
confirm muon (first in muon detector and then combined with ID) => rate ~ 5 kHz
Open region around muon and search for J/psi in inner detector.
Mass cut (M()>2.8GeV)
Extrapolate tracks to muon system.
EF : Refit tracks in RoI
Vertex reconstruction
Efficiency vs. Δη (RoI half-width)
1.1 1
Efficiency
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Delta eta
J/y  m+ m- efficiency vs. background rate 78
LVL2: Efficiency 68-77%
(1st  pT>6Gev, 2nd  pT>3GeV)
Background Hz 76 74
J/psi Efficiency (%) 72 70 68
Use similar method to improve efficiency for other di-muon channels at low luminosity 66
0.2
0.25
0.3
0.35
0.4
0.45
Background rate, (kHz)
Julie Kirk, RAL
ATLAS UK Physics Workshop, Durham (20/9/06)

7. Bs→Ds(φπ)π : LVL1 Efficiency pT(Bs) (GeV)
Efficiency for B to be within LVL1 RoI
LVL1: 1μ (> 6GeV) + ≥1 Jet RoI
Jet ET threshold chosen such that RoI multiplicity ~1-2
Efficiency for B (pT>10GeV) to be within RoI is 78%
Efficiency
pT(Bs) (GeV)
Julie Kirk, RAL
ATLAS UK Physics Workshop, Durham (20/9/06)

8. Bs→Ds(φπ)π : LVL2 Confirm LVL1 muon (muFast and muComb) True KK(π)
combinations
Confirm LVL1 muon (muFast and muComb)
Reconstruct tracks within a JET RoI (ΔФxΔη=1.5x1.5)
Combine pairs of tracks with K+K- mass hypothesis to form φ candidates. Mass cuts:
|M(KK) - M(φ)| < 3σ
If pass then add third track with π mass hypothesis to form Ds candidates.
Apply +/- 3σ mass cuts: |M(KKπ) - M(Ds)| < 3σ
Ds →φπ
signal
Mean = GeV
σ = 5 MeV
bb→µX
background
M(KK) (GeV)
M(KK) (GeV)
Ds →φπ
signal
bb→µX
background
Mean = GeV
σ = 18 MeV
M(KKπ) (GeV)
M(KKπ) (GeV)
Julie Kirk, RAL
ATLAS UK Physics Workshop, Durham (20/9/06)

9. Bs→Ds(π)π : RoI vs Full Scan
For fullscan method:
LVL1 muon confirmed at LVL2
reconstruct tracks in whole inner detector
Lose less low pT B’s
LVL2 efficiencies pT(B)>10 GeV:
RoI: 60%
Fullscan: 68%
Background rate (bb→X) ~ 175Hz (1x1033)
Efficiency to select Ds after LVL2
RoI guided
Fullscan
Julie Kirk, RAL
ATLAS UK Physics Workshop, Durham (20/9/06)

10. Timings for track reconstruction
LVL2 muon confirmation reduces rate by factor ~4 (21kHz → 5kHz)
Time available for track reconstruction x4 (10ms → 40ms)
<Time/RoI> = 23ms
(<Time/event> = 44 ms)
<Time/event>
(Fullscan) = 160ms
400 ms
100 ms
RoI guided approach ~4 x faster than full scan.
Once include EM and muon RoIs as well times may be comparable.
Fullscan looks possible for early running or if use a higher muon pT threshold (pT>8GeV => ~2x reduction in rate)
Julie Kirk, RAL
ATLAS UK Physics Workshop, Durham (20/9/06)

11. Summary
Can use single LVL1 muon trigger with reconstruction in secondary RoIs at HLT to increase B-physics programme for low luminosity running
Illustrated triggers using Muon, EM and Jet RoIs.
Comparison of fullscan vs. RoI guided. Allows flexibility depending on running conditions.
Expected efficiencies/rates (including eff~85% for LVL1 muon):
channel
Efficiency
Rate
(L=1x1033)
Bd J/y (m+m-) Ks
~65%
Events with one m pT>6 GeV and other m pT > 3 GeV
~ Hz
Bd  J/y (e+e-) Ks
~60% (e+ & e- pT > 5 GeV)
~40% (1st e pT>5 GeV, 2nd e pT>2 GeV)
~170 Hz
Bs  Ds p
Ds  f (K+K-) p
~50%
Events with Bs pT>10GeV
and K, K, p pT > 1.5 GeV
~175 Hz
Julie Kirk, RAL
ATLAS UK Physics Workshop, Durham (20/9/06)

12. Backup slides Julie Kirk, RAL
ATLAS UK Physics Workshop, Durham (20/9/06)

13. B→μX: LVL1 JET RoI multiplicities
4x4 Jet RoI
Aiming for a multiplicity of 1-2.
Implies ET threshold of about 4-5GeV.
Varied tower thresholds (EM and hadronic) for individual elements before cluster is formed. Some effect for hadronic case – increased threshold → reduced multiplicity.
Investigated different RoI sizes (4x4 and 8x8)
Default(EM=500MeV,hadronic=750MeV)
EMTowerthresh=750MeV
EMTowerthresh=1000MeV
HadTowerthresh=1000MeV
RoI Multiplicity
DC1 and Rome data without pu agree, but won’t necessarily be the case for pile-up data.
Need to keep mean multiplicity low to minimise data transfer and cpu usage at LVL2
Aiming for a multiplicity of 1-2
LVL1 Threshold Energy (GeV)
Julie Kirk, RAL
ATLAS UK Physics Workshop, Durham (20/9/06)

14. Increase hadronic tower threshold (ET>4GeV) reduces efficiency.
Bs →Ds(φπ)π:LVL1 efficiencies for B-hadron to be within the RoI (4X4 Jet RoIs)
Efficiency
ET > 4GeV
ET > 5GeV
pT(Bs) (GeV)
low tower thresh
high tower thresh
EFFECT OF ET THRESHOLD
EFFECT OF HADRONIC TOWER THRESHOLD
ET > 4GeV
Increase hadronic tower threshold (ET>4GeV) reduces efficiency.
BUT reduced ROI multiplicity means can reduce ET threshold to recover similar efficiency.
Vary ET threshold
For pT(B)>10GeV (offline cut)
ET>5GeV => 69% eff.
ET>4GeV => 76% eff.
Julie Kirk, RAL
ATLAS UK Physics Workshop, Durham (20/9/06)

15. LVL1 efficiencies for B to be in ROI
RoI size / tower thresholds
ET threshold
<nRoI>
Efficiency
All events
pT(B)>10 GeV
pT(B)>20 GeV
4x4 jet
4 GeV
5 Gev
1.8
1.1
58%
51%
76%
69%
98%
96%
high had TT
3.5 GeV
4.5 GeV
1.5
1.0
50%
68%
97%
8x8 jet
6 GeV
7 GeV
2.0
1.3
55%
49%
72%
66%
5 GeV
59%
53%
78%
71%
Studied the effect of varying tower thresholds/RoI sizes
→ defaults are OK
→ similar performance achievable with different setups
Julie Kirk, RAL
ATLAS UK Physics Workshop, Durham (20/9/06)