From the ATLAS electromagnetic calorimeter to SUSY Freiburg, 15/06/05 Dirk Zerwas LAL Orsay • Introduction • ATLAS EM-LARG • Electrons and Photons • SUSY measurements • Reconstruction of the fundamental parameters • Conclusions
Dec 21, 2015
From the ATLAS electromagnetic calorimeter to SUSY
Freiburg, 15/06/05Dirk ZerwasLAL Orsay
• Introduction• ATLAS EM-LARG• Electrons and Photons• SUSY measurements• Reconstruction of the fundamental parameters• Conclusions
Introduction• LHC: CERN’s proton-proton collider at 14TeV• 2800 bunches of 1011 protons• bunch crossing frequency: 40.08MHz • Low Luminosity: 1033cm-2/s meaning 10fb-1 per experiment (3 years)• High Luminosity 1034cm-2/s meaning 100fb-1 per experiment (n years)• SLHC: most likely 1035cm-2/s meaning 1000fb-1 per experiment (2015+)• startup for physics: late 2007
Two multipurpose detectors: ATLAS, CMS
The experimental challenges of the LHC environment:• bunch crossing every 25ns • 22 events par BX (fast readout, 40MHz 200Hz, event-size 1.6MB)• High radiation FE electronics difficult (military and/or space technology)and with that do precision physics!
Physics at the LHC
Process Events/s Events/year other machines
Weν 15 108 104 LEP/ 107 TeV.Z ee 1.5 107 107 LEPtt 0.8 107 104 TeVatronbb 105 1012 108 Belle/BabarQCD jets 102 109 107
pT>200GeV
If the machine works well:
Factory of Z, W, top and QCD jets. Will be limited quickly by systematics!
Measurements:• W mass to 20MeV needs control of the linearity/energy scale (0.02% energy scale)• Higgs mass measurement (if etc) in γγ• SUSY precision measurements with leptonsStringent requirements on the energy scale, uniformity and linearity of the ATLAS-EM Calorimeter response! Startup date getting closer, need to prove that we understand and are prepared Calo!
You have heard already much about the physics from Sven Heinemeyer, Tilman Plehn, Christian Weiser,… plus in-house expertise on ATLAS-Tracker, ATLAS-Muons, Higgs physics,…..so try to find things of added value not covered so far: Calo+SUSYreco
The ATLAS Electromagnetic Calorimeter
Liquid Argon Sampling Calorimeter:• lead (+s.s.) absorbers (1.1, 1.5mm Barrel)• liquid Argon gap 2.2mm 2kV (barrel)• varying gap and HV in the endcap• accordion structure no dead area in φ • “easy” to calibrate
R 4m
φ
Z 3.2m
Z 0mR 2.8m
Granularity (typical Δη X Δφ ):Presampler = 0.025x0.1 (up to η=1.8)Strips = 0.003x0.1 (EC )Middle = 0.025x0.025 (main energy dep)Back = 0.05 x0.025
The barrel and endcap EM-calorimeters!Some numbers:2048 barrel absorbers2048 barrel electrodesgiving 32 barrel modules (4years of production and assembly)16 endcap modulesAll assembled and inserted in their cryostatsBarrel cryostat in pit waiting for electronics
Thickness: 24-30X0
Barrel
Endcaps
Calibration of the ATLAS EM calorimeter
General Strategy and Sequence for electrons and photons:• Calibration of Electronics
• necessitates a good understanding of the physics and calibration signal• Corrections at the cluster level:
• position corrections • correction of local response variations• corrections for losses in upstream (Inner detector) material and longitudinal leakage
• Refinement of corrections depending on the particle type (e/γ)• uniformity 0.7% with a local uniformity in ΔηXΔφ=0.2x0.4 better than 0.5% • inter-calibrate region with Zee
What can be studied where?• Calibration of electronics studied in testbeam• Corrections at cluster level: testbeam and ATLAS simulations• uniformity: testbeam• Zee: simulation
The best Monte Carlo is the DATA! For ATLAS:Testbeam TestbeamMC ATLASMC ATLAS
ATLAS series modules in testbeam
1998-2002: prototype and single module tests at CERN:4 ATLAS barrel modules3 ATLAS endcap modulesSingle electron beams20-245GeV
Studies of:• energy resolution• linearity• uniformity• particle ID
2004: combined testbeam endcap and barrel including tracking and muons
FE electronicsSitting directly on the feedthroughas in ATLAS
Electronics:• bipolar signal • time to peak 50ns (variable)• 40MHz sampling of 5 samples (125ns)• three gain system 1/9.5/10(automatic choice)
From 5 samples in time to one “energy”:Optimal Filtering coefficients:• exponential versus linear• different entry points • inductance effect: parallel versus serial• electronic gains
The Signal/Electronics Calibration
Calibration signal : ~0.2%
Physics signal
0 1.4
L non-uniform:2-3 % effecton E along
Preamp + shaper (3gains) + SCA
60
30
10
L (nH)
Hamac SCA: Atlas Calorimeter Electronics.
Sampling of 3x4 signals at 40MHz, 13.5 bits of dynamic range with simultaneous write and read in rad-hard technology (DMILL).
Same type of chip used in digital oscilloscope: keep the high dynamic range and increase the sampling rate and bandwidth while using the cheapest technology on the market: 0.8µ pure CMOS (patent filed in April 2001).
Instruments are based on the MATACQ chip which is a sampling matrix able to sample data at 2GS/s over 2560 points and 12 bits of dynamic range with a very low power consumption compared to standard systems.
This structure has first been used in the design of the new digital oscilloscope family of Metrix (0X7000). This product is the first autonomous 12-bit scope on the market. Award for technology transfer to industry of the SFP (DPG)
Also used in a 4-channel VME and GPIB board. The latter offers the 2GS/s – 12bits facility with low power at low cost. It’s perfectly suited for high dynamic range precise measurements in harsh environments (CAEN).
Digression:From physics to industry.
Dominique Breton LAL-Orsay, Eric Delagnes CEA-Saclay
Cluster Corrections
Clustering with fixed size• Correct position S-shape in eta• Correct phi offset• S shape eta in strips• local energy variations phi (accordion)• local energy variations eta
Testbeam: phi modulation Endcap:
ATLAS simulation: S-shape
Variation of correction as function of η under control (smooth behaviour)
Cluster Corrections: longitudinal weighting
Non-negligeable amount of material before the calorimeterReconstruction needs to optimise simultaneously energy resolution and linearity.Method based on Monte Carlo and tested with data in one point η= 0.68:
impactcellbremleaki
caloi
visVisPS
VisPS
rec fEfdepthfEEdEEEcEEbEaE 3,1
5.01 ).()).(1).().().)(().()((
Correct for energyloss upstream ofPresampler(cryostat+beam linematerial)
Energy lost between PS and calo(Cable/board)
Small dependence of calo sampling fraction+ lateral leakage with energy
Longitudinalleakage depthfunction of depth only
fbrem extracted from simulation and beam transport of H8 beam line, not present in ATLAS
1.5 X0, 3.6 %@10 GeV 0.9 X0, 4.1 %@10 GeV > 30 X0, 0.3 %@10 GeV
EPS = energy in presamplerEi =energy in calorimeter compartments
Linearity
Achieved better than 0.1 % over 20-180 GeV but : - done only at one position in a setup with less material than in ATLAS and no B field-No Presampler in Endcap (ATLAS) for >1.8
Systematics at low energy ~0.1 %
Dedicated setup was used in 2002 to have a very precise beam energy measurement : - Degaussing cycle for the magnet to ensure B field reproducibility ateach energy (same hysteresis) - Use a precise Direct Current-Current Transformer with a precision of 0.01 % - Hall probe from ATLAS-Muon in magnet to cross-check magnet calibration lots of help from EA-team (I. Efthymiopoulos) Limitation of calorimeter linearity measurement is 0.03 % from beam energy knowledge - Absolute energy scale is not known in beam test to better than ~1 %- Relative variation is important
Energy resolution
Resolution at =0.68 Local energy resolution well understood since Module 0 beam tests and well reproduced by simulation : Uniformity is at 1% level quasi onlinebut achieving ATLAS goal (0.7 %) difficult
Good agreement for longitudinal shower development between data and testbeam MC
Cluster Energy CorrectionsIn ATLAS: use a simplified formula:E(corr) = Scale(eta)*(Offset(eta)+W0(eta)*EPS+E1+E2+W3(eta)*E3)
3x7
0.1%-0.2% spread from 10GeV to 1TeV over all eta remember testbeam was 1point: proof that the method works!
10GeV
50GeV
100GeV
Energy resolution in ATLAS Simulation
100GeVresolution
X0 in frontof strips
Energy resolution in ATLAS wrt testbeam 20% worseTypically 2-4 X0 in front of calorimeterGood correlation with resolution
Current method at the limit of its sensitivityFor historians: wrt TDR 25% degradation, but in TDR simulation Inner Detector Material description incomplete
Barrel uniformity @ 245 GeV in testbeam
In beam setup, one feedthrough had qualityproblem ( open symbols) due to largeresistive cross talk (non-ATLAS FT). > 7 is ATLAS like and can be used as reference : uniformity better than 0.5 %Energy scale differs by 0.13 %
quality of module construction is excellent
rms0.62%
0.45%
4.5‰0.49%
Module P13
Module P13 > 7
Module P15 > 7
Module P13Energy resolution
Similar results for endcap modules
Position/Direction measurements in TB
245 GeV Electrons ~550 μmat =0
~250 μmat =0
mid
strip
H γγ vertex reconstructed with 2-3 cm accuracy in ATLAS in z Precision of theta measurement 50mrad/sqrt(E)
Good agreement of data and simulation
Z~5mm
Z~20mm
Zee
• uniformity 0.2x0.4 ok in testbeam• description of testbeam data by Monte Carlo satisfactory• make use of Zee Monte Carlo and Data in ATLAS for intercalibration of regions• 448 regions in ATLAS (denoted by i)• mass of Z know precisely• Ei
reco = Eitrue(1+αi)
• Mijreco =Mij
true(1+(αi+αj)/2)• fit to reference distribution (Monte Carlo!!!)• beware of correlations, biases etc…
At low (but nominal) luminosity, 0.3% of intercalibration can be achieved in a week (plus E/P later on)! Global constant term of 0.7% achievable!
Mass resolution of Higgs bosons
Hγγ120.96GeVσ= 1.5GeVH γγ
Note that the generated Higgs mass is 120GeV:Effect: calibration with electrons, so the photon calibration is off by 1-2%Getting from Electron to photon in ATLAS will require MC!
H ZZ 4e:Mass scale correct within 0.1GeV σ=2.2GeV
Particle Identification/jet rejection
Dijet cross section ~1mbZ ee 1.5 10-6 mbW eν 1.5 10-5 mbNeed a rejection factor of 105 for electronsUse the shower shape in the calorimeter
Use the trackerUse the combination of the calo+tracker
Cut based analysis gives for electrons an efficiency of about 75-80% with a rejection factor of 105
Multivariate techniques are being studied for possible improvements(likelihood, neural net, boost decision tree)
Soft electrons
Two possibilities for seeded electron reconstruction• calo• trackerReconstruction of electrons close to jets difficult, and interesting (b-tagging) especially for soft electrons. Dedicated algorithm:• builds clusters around extrapolated impact point of the tracks• calculates properties of the clusters• PDF and neural net for ID• useful per se as well as for b-tagging
Hbbpions
e id efficiency = 80% Pion rejection in: J/Psi : 1050±50 WH(bb) : 245±17 ttH : 166 ±6
J/Psi
WH
ttH
What can we do now with all that?
Supersymmetry
3 neutral Higgs bosons: h, A, H1 charged Higgs boson: H±
and supersymmetric particles:
spin-0 spin-1/2 spin-1
Squarks:
qR, qL
q
Gluino: g g
Sleptons:
ℓR, ℓL
ℓ
h,H,A Neutralino χi=1-4
Z, γ
H± Charginos:χ±
i=1-2 W±
~~
~~
The parameters of the Higgs sector:• mA : mass of the pseudoscalar Higgs boson• tanβ: ratio of vacuum expectation values• mass of the top quark• stop (tR, tL) sector: masses and mixing
~
~ ~
Theoretical limit:mh 140GeV/c2
Many different models:• MSSM (minimal supersymmetric extension SM)• mSUGRA (minimal supergravity)• GMSB• AMB• NMSSM
Conservation of R-parity• production of SUSY particules in pairs• (cascade) decays to the lightest sparticle • LSP stable and neutral: neutralino (χ1)• signature: missing ET
See talks by Sven and Tilman: Here only a reminder for completeness sake
At the LHC
Large cross section for squarks and gluinos of several pb, i.e. several kEventssum jet-PT and ET effective massSquarks and gluinos up to 2.5TeV “straight forward”Largest background for SUSY is SUSY (but…)
Large masses means long decay chainsSelection: multijet with large PT (typically 150,100,50 GeV) and OS-SF leptonsInvariant masses jet-lepton, lepton-lepton, lepton-lepton-jet related to masses
SM
SUSY
SUSY at the LHC (and ILC)
Moderately heavy gluinos and squarks
light sleptons
Heavy and light gauginos
Higgs at the limitof LEP reach
τ1 lighter than lightest χ± :• χ± BR 100% τν• χ2 BR 90% ττ • cascade:qL χ2 q ℓR ℓ q ℓ ℓ qχ1
visible
m0 = 100GeV m1/2 = 250GeV A0 = -100GeV tanβ =10 sign(μ)=+favourable for LHC and ILC (Complementarity)
~
~
~~
~
Examples of measurements at LHC
Gjelsten et al: ATLAS-PHYS-2004-007/29
From edges to masses: System overconstrainedplus other mass differences and edges…
Using the kinematical formula (no use of model) and a toy MC for the correlated energy scale error: • energy scale leptons 0.1%• energy scale jets 1%Mass determination for 300fb-1 (thus 2014):
Coherent set of “measurements”for LHC (and ILC) “Physics Interplay of the LHC and ILC”Editor G. Weiglein hep-ph/0410364
Polesello et al: use of χ1 from ILC (high precision) in LHC analyses improves the mass determination
From Mass measurements to Parameters
SFITTER (R. Lafaye, T. Plehn, D. Z.): tool to determine supersymmetric parameters from measurementsModels: MSUGRA, MSSM, GMSB, AMB
The workhorses:• Mass spectrum generated by SUSPECT (new version interfaced) or SOFTSUSY• Branching ratios by MSMLIB• NLO cross sections by Prospino2.0• MINUIT
The Technique:• GRID (multidimensional to find a non-biased seed, configurable)• subsequent FIT
Other approaches:• Fittino (P. Bechtle, K. Desch, P. Wienemann)• Interpolation (Polesello)• Analytical calculations (Kneur et al, Kalinowski et al)• Hybrid (Porod)
Beenakker et al
SPS1a ΔLHC ΔILC ΔLHC+ILC
m0 100 3.9 0.09 0.08
m1/2 250 1.7 0.13 0.11
tanβ 10 1.1 0.12 0.12
A0 -100 33 4.8 4.3
Results for MSUGRA
Start SPS1a LHC ILC LHC+ILC
m0 100 1TeV 1TeV 1TeV
m1/2 250 1TeV 1TeV 1TeV
tanβ 10 50 50 50
A0 -100 0GeV 0GeV 0GeV
• Convergence to central point• errors from LHC %• errors from ILC 0.1%• LHC+ILC: slight improvement• low mass scalars dominate m0
Two separate questions:• do we find the right point?
• need and unbiased starting point• what are the errors?
Once a certain number of measurements are available, start with the most constrained model
Sign(μ) fixed
Masses versus Edges
need correlations to obtain the ultimate precision from masses….
SPS1a ΔLHC masses
ΔLHCedges
m0 100 3.9 1.2
m1/2 250 1.7 1.0
tanβ 10 1.1 0.9
A0 -100 33 20
Δm0 Effect on mℓR Effect on mℓℓ
1GeV 0.7/5=0.14 0.4/0.08=5
• use of masses improves parameter determination!• edges to masses is not a simple “coordinate” transformation:
Similar effect for m1/2
Sign(μ) fixed
Total Error and down/up effect
Higgs sleptons Squarks,gluinos Neutralinos, charginos
3GeV 1% 3% 1%
Theoretical errors (mixture of c2c and educated guess):
Running down/up• spectrum generated by SUSPECT• fit with SOFTSUSY (B. Allanach)• central values shifted (natural)• m0 not compatible
SPS1a ΔLHC+ILCexp
ΔLH+ ILCth
m0 100 0.08 1.2
m1/2 250 0.11 0.7
tanβ 10 0.12 0.7
A0 -100 4.3 17
Including theory errors reducessensitivity by an order of magnitude
SPS1a SoftSUSYup ΔLHC+LC
m0 100 95.2 1.1
m1/2 250 249.8 0.5
tanβ 10 9.82 0.5
A0 -100 -97 10
Higgs error: Sven Heinemeyer et al.
Between MSUGRA and the MSSM
Start with MSUGRA, then loosen the unification criteria,less restricted model defined at the GUT scale:• tanβ, A0, m1/2 , m0
sleptons, m0squarks, mH
2 , μ• experimental errors only
SPS1a LHC ΔLHC
m0sleptons 100 100 4.6
m0squarks 100 100 50
mH2 10000 9932 42000
m1/2 250 250 3.5
tanβ 10 9.82 4.3
A0 -100 -100 181
• Higgs sector undetermined • only h (mZ) seen
• slepton sector the same as MSUGRA• light scalars dominate determination of m0
• smaller degradation in other parameters, but still % precision
The highest mass states do not contain the maximum information in the scalar sector, but they do in the Higgs sector!
Sfitter-team and Sabine Kraml
MSSM
MSSM fit:bottom-up approach24 parameters at the EW scale
LHC or ILC alone:• certains parameters must be fixedLHC+ILC:• all parameters fitted• several parameters improved
Caveat:• LHC errors ~ theory errors• ILC errors << theory errorsSPA project: improvement oftheory predictions and standardisation
LHC ILC LHC+ILC
With more measurements available: fit the low energy parameters
Impact of TeVatron Data?
With Volker Buescher (Uni Freiburg): • 2008 too early for Higgs to γγ with 10fb-1 at LHC• only central cascade SUSY measurements are available: χ1, χ2, qL, ℓR
• Higgs is sitting on the edge of LEP exclusion • WH+ZH 6 events per fb-1 and experiment at TeVatron• end of Run: Δmh = ± 2GeV• adding background: Δmh = ± 4-5GeV
•A hint of Higgs from the TeVatron would help the LHC at least the first year! • mtop from TeVatron with 2GeV precision makes impact on fit negligible
Higgs mass from γγ
~~
No Higgs, edges from the LHC:m0 = 100 ± 14 GeV m1/2 = 250 ± 10 GeVtanβ = 10 ± 144 A0 = -100.37 ± 2400 GeV
Higgs hint plus edges from the LHC:m0 = 100 ± 9 GeVm1/2 = 250 ± 9 GeVtanβ = 10 ± 31 A0 = -100 ± 685 GeV
And the Egret point?Wim de Boer: astro-ph/0408272EGRET: on Compton gamma ray observatory, measured high energy gamma ray flux.Compatible with Standard Model, but also SUSY:m0 =1400 GeV m1/2 = 180 GeV A0=700 GeV tanβ = 51 μ > 0
0
WMAP
EGRET
Stau coannihilation
mA resonance
Bulk
Incomp. withEGRET data
StauLSP
No EWSB
Dominant Processes at the LHC:
m0 =1400 ± (50 – 530)GeV m1/2 = 180 ± (2-12) GeV A0 =700 ± (181-350) GeV tanβ= 51 ± (0.33-2)
Measurements:• Higgs masses h,H,A• mass difference χ2-χ1
• mass difference g- χ2
Sufficient for MSUGRA
~
Uncertainties:• b quark mass• t quark mass• Higgs mass prediction
Les Houches 2005: P. Gris, L. Serin, L.Tompkins, D.Z.
Tri-lepton signal promissing
Conclusions
• Construction of ATLAS-EM calorimeter modules finished• Testbeam studies have driven the improvement of the understandingof the combined optimisation of linearity and resolution of the calorimeter • EM calibration under control• electron (and photon identification) are at the required levelwith multivariate approaches under study • SFitter (and Fittino) will be essential to determine SUSY’s fundamentalparameters
• mass differences, edges and thresholds are more sensitive than masses• the LHC will be able to measure the parameters at the level %• LC will improve by a factor 10• LHC+LC reduces the model dependence• EGRET: in MSUGRA, LHC has enough potential measurements to confront the hypothesis
Many thanks to Laurent Serin for his help in the preparation of the talk!