Search for new dilepton resonances using ATLAS open data Magnar K. Bugge University of Oslo May 20th 2017 Magnar K. Bugge New resonance search using ATLAS open data 1 / 19
Goal for the day You will perform a search for new dilepton resonances using real ATLAS data! Final results similar to official ATLAS s = 8 TeV results below, but with only a subset of the data Events 7 6 5 4 3 ATLAS Z ee s -1 L dt = 20.3 fb = 8 TeV Data 2012 Z/γ* Top uark Dijet & W+Jets Diboson Z SSM (1.5 TeV) Z SSM (2.5 TeV) Events 7 6 5 4 3 ATLAS Z µµ s -1 L dt = 20.5 fb = 8 TeV Data 2012 Z/γ* Top uark Diboson Z SSM (1.5 TeV) Z SSM (2.5 TeV) 2 2 1 1-1 -1 Data/Expected 1.4 1.2 1 0.8 0.6 0.08 0.1 0.2 0.3 0.4 0.5 1 2 3 4 m ee [TeV] Data/Expected 1.4 1.2 1 0.8 0.6 0.08 0.1 0.2 0.3 0.4 0.5 1 2 3 4 m µµ [TeV] Magnar K. Bugge New resonance search using ATLAS open data 2 / 19
Searching for new physics ATLAS Exotics Searches* - 95% CL Exclusion Status: August 2016 ATLAS Preliminary L dt = (3.2-20.3) fb 1 s = 8, 13 TeV Model l, γ Jets E miss T L dt[fb 1 ] Limit Reference LQ DM CI Gauge bosons Extra dimensions ADD GKK + g/ 1 j Yes 3.2 MD 6.58 TeV n = 2 1604.07773 ADD non-resonant ll 2 e, µ 20.3 MS 4.7 TeV n = 3 HLZ 1407.24 ADD QBH l 1 e, µ 1 j 20.3 Mth 5.2 TeV n = 6 1311.2006 ADD QBH 2 j 15.7 Mth 8.7 TeV n = 6 ATLAS-CONF-2016-069 ADD BH high pt 1 e, µ 2 j 3.2 Mth 8.2 TeV n = 6, MD = 3 TeV, rot BH 1606.02265 ADD BH multijet 3 j 3.6 Mth 9.55 TeV n = 6, MD = 3 TeV, rot BH 1512.02586 RS1 GKK ll 2 e, µ 20.3 GKK mass 2.68 TeV k/mpl = 0.1 1405.4123 RS1 GKK γγ 2 γ 3.2 GKK mass 3.2 TeV k/mpl = 0.1 1606.03833 Bulk RS GKK WW lν 1 e, µ 1 J Yes 13.2 GKK mass 1.24 TeV k/mpl = 1.0 ATLAS-CONF-2016-062 Bulk RS GKK HH bbbb 4 b 13.3 GKK mass 360-860 GeV k/mpl = 1.0 ATLAS-CONF-2016-049 Bulk RS gkk tt 1 e, µ 1 b, 1J/2j Yes 20.3 gkk mass 2.2 TeV BR = 0.925 1505.07018 2UED / RPP 1 e, µ 2 b, 4 j Yes 3.2 KK mass 1.46 TeV Tier (1,1), BR(A (1,1) tt) = 1 ATLAS-CONF-2016-013 SSM Z ll 2 e, µ 13.3 Z mass 4.05 TeV ATLAS-CONF-2016-045 SSM Z ττ 2 τ 19.5 Z mass 2.02 TeV 1502.07177 Leptophobic Z bb 2 b 3.2 Z mass 1.5 TeV 1603.08791 SSM W lν 1 e, µ Yes 13.3 W mass 4.74 TeV ATLAS-CONF-2016-061 HVT W WZ νν model A 0 e, µ 1 J Yes 13.2 W mass 2.4 TeV gv = 1 ATLAS-CONF-2016-082 HVT W WZ model B 2 J 15.5 W mass 3.0 TeV gv = 3 ATLAS-CONF-2016-055 HVT V WH/ZH model B multi-channel 3.2 V mass 2.31 TeV gv = 3 1607.05621 LRSM W R tb 1 e, µ 2 b, 0-1 j Yes 20.3 W mass 1.92 TeV 14.43 LRSM W R tb 0 e, µ 1 b, 1 J 20.3 W mass 1.76 TeV 1408.0886 CI 2 j 15.7 Λ 19.9 TeV ηll = 1 ATLAS-CONF-2016-069 CI ll 2 e, µ 3.2 Λ 25.2 TeV ηll = 1 1607.03669 CI uutt 2(SS)/ 3 e,µ 1 b, 1 j Yes 20.3 Λ 4.9 TeV CRR = 1 1504.04605 Axial-vector mediator (Dirac DM) 0 e, µ 1 j Yes 3.2 ma 1.0 TeV g=0.25, gχ=1.0, m(χ) < 250 GeV 1604.07773 Axial-vector mediator (Dirac DM) 0 e, µ, 1 γ 1 j Yes 3.2 ma 7 GeV g=0.25, gχ=1.0, m(χ) < 150 GeV 1604.01306 ZZχχ EFT (Dirac DM) 0 e,µ 1 J, 1 j Yes 3.2 M 550 GeV m(χ) < 150 GeV ATLAS-CONF-2015-080 Scalar LQ 1 st gen 2 e 2 j 3.2 LQ mass 1.1 TeV β = 1 1605.06035 Scalar LQ 2 nd gen 2 µ 2 j 3.2 LQ mass 1.05 TeV β = 1 1605.06035 Scalar LQ 3 rd gen 1 e, µ 1 b, 3 j Yes 20.3 LQ mass 640 GeV β = 0 1508.04735 Heavy uarks Excited fermions VLQ TT Ht + X 1 e, µ 2 b, 3 j Yes 20.3 T mass 855 GeV T in (T,B) doublet 1505.04306 VLQ YY Wb + X 1 e, µ 1 b, 3 j Yes 20.3 Y mass 770 GeV Y in (B,Y) doublet 1505.04306 VLQ BB Hb + X 1 e, µ 2 b, 3 j Yes 20.3 B mass 735 GeV isospin singlet 1505.04306 VLQ BB Zb + X 2/ 3 e, µ 2/ 1 b 20.3 B mass 755 GeV B in (B,Y) doublet 1409.5500 VLQ QQ WW 1 e, µ 4 j Yes 20.3 Q mass 690 GeV 1509.04261 VLQ T5/3T5/3 WtWt 2(SS)/ 3 e,µ 1 b, 1 j Yes 3.2 T5/3 mass 990 GeV ATLAS-CONF-2016-032 Excited uark γ 1 γ 1 j 3.2 mass 4.4 TeV only u and d, Λ = m( ) 1512.059 Excited uark g 2 j 15.7 mass 5.6 TeV only u and d, Λ = m( ) ATLAS-CONF-2016-069 Excited uark b bg 1 b, 1 j 8.8 b mass 2.3 TeV ATLAS-CONF-2016-060 Excited uark b Wt 1 or 2 e, µ 1 b, 2-0 j Yes 20.3 b mass 1.5 TeV fg = fl = fr = 1 15.02664 Excited lepton l 3 e, µ 20.3 l mass 3.0 TeV Λ = 3.0 TeV 1411.2921 Excited lepton ν 3 e, µ, τ 20.3 ν mass 1.6 TeV Λ = 1.6 TeV 1411.2921 Other LSTC at W γ 1 e, µ, 1 γ Yes 20.3 at mass 960 GeV 1407.8150 LRSM Majorana ν 2 e,µ 2 j 20.3 N 0 mass 2.0 TeV m(wr ) = 2.4 TeV, no mixing 1506.06020 Higgs triplet H ±± ee 2 e (SS) 13.9 H ±± mass 570 GeV DY production, BR(H ±± L ee)=1 ATLAS-CONF-2016-051 Higgs triplet H ±± lτ 3 e, µ, τ 20.3 H ±± mass 400 GeV DY production, BR(H ±± lτ)=1 L 1411.2921 Monotop (non-res prod) 1 e, µ 1 b Yes 20.3 spin-1 invisible particle mass 657 GeV anon res = 0.2 14.5404 Multi-charged particles 20.3 multi-charged particle mass 785 GeV DY production, = 5e 1504.04188 Magnetic monopoles 7.0 monopole mass 1.34 TeV DY production, g = 1gD, spin 1/2 1509.08059 s = 8 TeV s = 13 TeV 1 1 *Only a selection of the available mass limits on new states or phenomena is shown. Lower bounds are specified only when explicitly not excluded. Small-radius (large-radius) jets are denoted by the letter j (J). Mass scale [TeV] Magnar K. Bugge New resonance search using ATLAS open data 3 / 19
Searching for new physics why The Standard Model (SM) has been rigorously tested in experiment to great precision, and provides an excellent description of nature down to approximately 18 m Still, there are reasons to look beyond the SM: Gravity Dark matter and dark energy Neutrino mass Grand unification(?) The hierarchy problem(?) Magnar K. Bugge New resonance search using ATLAS open data 4 / 19
Searching for new physics why The Standard Model (SM) has been rigorously tested in experiment to great precision, and provides an excellent description of nature down to approximately 18 m Still, there are reasons to look beyond the SM: Gravity Dark matter and dark energy Neutrino mass Grand unification(?) The hierarchy problem(?) Magnar K. Bugge New resonance search using ATLAS open data 4 / 19
Searching for new physics why The Standard Model (SM) has been rigorously tested in experiment to great precision, and provides an excellent description of nature down to approximately 18 m Still, there are reasons to look beyond the SM: Gravity Dark matter and dark energy Neutrino mass Grand unification(?) The hierarchy problem(?) Magnar K. Bugge New resonance search using ATLAS open data 4 / 19
Searching for new physics why The Standard Model (SM) has been rigorously tested in experiment to great precision, and provides an excellent description of nature down to approximately 18 m Still, there are reasons to look beyond the SM: Gravity Dark matter and dark energy Neutrino mass Grand unification(?) The hierarchy problem(?) Magnar K. Bugge New resonance search using ATLAS open data 4 / 19
Searching for new physics why The Standard Model (SM) has been rigorously tested in experiment to great precision, and provides an excellent description of nature down to approximately 18 m Still, there are reasons to look beyond the SM: Gravity Dark matter and dark energy Neutrino mass Grand unification(?) The hierarchy problem(?) Magnar K. Bugge New resonance search using ATLAS open data 4 / 19
Searching for new physics why The Standard Model (SM) has been rigorously tested in experiment to great precision, and provides an excellent description of nature down to approximately 18 m Still, there are reasons to look beyond the SM: Gravity Dark matter and dark energy Neutrino mass Grand unification(?) The hierarchy problem(?) f H H Magnar K. Bugge New resonance search using ATLAS open data 4 / 19
Searching for new physics why Z? Gauge structure of SM: SU(3) C SU(2) L U(1) Y A set of gauge bosons for each symmetry group Models beyond the SM often extend the gauge symmetry: Adding additional symmetries (e.g. an SU(2) R in left-right symmetric models) Embedding SM gauge groups into higher symmetry (i.e. grand unification) Results in new gauge bosons! Magnar K. Bugge New resonance search using ATLAS open data 5 / 19
Searching for new physics how Testable theories of physics beyond the SM predict deviations from SM predictions Need excellent understanding of the SM Need to go to the frontier of precision... γ γ γ γ W W γ µ µ Z µ µ νµ µ µ a exp µ = 11 659 209.1(5.4)(3.3) γ γ had µ µ... or energy l Events 7 6 5 4 3 ATLAS Z ee s -1 L dt = 20.3 fb = 8 TeV Data 2012 Z/γ* Top uark Dijet & W+Jets Diboson Z SSM (1.5 TeV) Z SSM (2.5 TeV) 2 Z 1-1 l + Data/Expected 1.4 1.2 1 0.8 0.6 0.08 0.1 0.2 0.3 0.4 0.5 1 2 3 4 m ee [TeV] Magnar K. Bugge New resonance search using ATLAS open data 6 / 19
Searching for new physics at the LHC 27 km circumference proton-proton collisions at s = 13 TeV (previously s = 7 TeV and s = 8 TeV) Magnar K. Bugge New resonance search using ATLAS open data 7 / 19
Searching for new physics at the LHC ATLAS detector detecting and measuring final state particles emerging from the collision Magnar K. Bugge New resonance search using ATLAS open data 8 / 19
Searching for new physics at the LHC After reconstruction of the event, it can be analysed in terms of final state particles/objects such as electrons, muons, jets,... Magnar K. Bugge New resonance search using ATLAS open data 9 / 19
Searching for new physics at the LHC Main ingredients in any search for new physics at the LHC: Identify experimental signature which can be used to identify signal events (and which is not easily mimicked by background events) The experimental signature typically consists of a final state (e.g. two leptons) and expected kinematical properties (e.g. high p T, high invariant mass) Estimate backgrounds and hypothetical signal contribution (a background is a SM process which produces the experimental signature of interest) Compare data to background and signal estimates Events 7 6 5 4 3 ATLAS Z ee s -1 L dt = 20.3 fb = 8 TeV Data 2012 Z/γ* Top uark Dijet & W+Jets Diboson Z SSM (1.5 TeV) Z SSM (2.5 TeV) 2 1-1 Data/Expected 1.4 1.2 1 0.8 0.6 0.08 0.1 0.2 0.3 0.4 0.5 1 2 3 4 m ee [TeV] Magnar K. Bugge New resonance search using ATLAS open data / 19
Search for heavy Z bosons experimental signature Need to choose a final state in which to search for the Z : Z νν: very hard to detect Z : very large background Z l + l, l = e, µ: easy to detect and low background! l + l final state l final state Z Z l + l Z/γ Z/γ l + g g g g Magnar K. Bugge New resonance search using ATLAS open data 11 / 19
Search for heavy Z bosons backgrounds Experimental signature: two oppositely charged leptons with high invariant mass Background processes include: Z l + l : invariant mass peaked at 91 GeV, but tail extends to high invariant mass t t l + ν l l ν l b b W + W l + ν l l ν l Fake leptons, i.e. jets misidentified as leptons largely suppressed by isolation cut νl W + g g t b l + W + νl l + g t b l l W νl W νl Magnar K. Bugge New resonance search using ATLAS open data 12 / 19
Search for heavy Z bosons modeling Estimate signal and background process contributions primarily by MC simulation Z l l + Event generation Simulated final state Detector simulation Simulated raw data Reconstruction Simulated events for analysis Z/γ l l + Event generation Detector simulation Reconstruction Simulated final state Simulated raw data Simulated events for analysis νl g g t t b b l + Event generation Detector simulation Reconstruction Simulated final state Simulated raw data l Simulated events for analysis νl Real raw data Reconstruction Real events for analysis Analysis code selects events of interest in the same way from both real and simulated data Magnar K. Bugge New resonance search using ATLAS open data 13 / 19
Search for heavy Z bosons data/mc comparison Histograms from real data finally compared to sum of simulated contributions Agreement between data and signal/background predictions can be judged roughly by eye Further statistical analysis reuired to uantify presence (significance) or absence (exclusion limit) of signal respectively Events 7 6 5 4 3 ATLAS Z ee s -1 L dt = 20.3 fb = 8 TeV Data 2012 Z/γ* Top uark Dijet & W+Jets Diboson Z SSM (1.5 TeV) Z SSM (2.5 TeV) Events 7 6 5 4 3 ATLAS Z µµ s -1 L dt = 20.5 fb = 8 TeV Data 2012 Z/γ* Top uark Diboson Z SSM (1.5 TeV) Z SSM (2.5 TeV) 2 2 1 1-1 -1 Data/Expected 1.4 1.2 1 0.8 0.6 0.08 0.1 0.2 0.3 0.4 0.5 1 2 3 4 m ee [TeV] Data/Expected 1.4 1.2 1 0.8 0.6 0.08 0.1 0.2 0.3 0.4 0.5 1 2 3 4 m µµ [TeV] Magnar K. Bugge New resonance search using ATLAS open data 14 / 19
Statistical analysis basics single bin approach Create a signal region by making a suitable cut on the dilepton invariant mass Cut depends on the signal mass considered, e.g. m l + l > 2 TeV may be a suitable cut for m Z = 2.5 TeV Count the number of events in signal region data condensed down to one integer The data count is the realization of a Poisson distributed stochastic variable, the expectation value of which is given by the sum of signal and background contributions Notation: n: number of events (stochastic variable) with observed value n obs in data b: number of expected background events s: number of expected signal events n Poisson(s + b) P(n = k) = (s + b)k e (s+b) Magnar K. Bugge New resonance search using ATLAS open data 15 / 19 k!
Single bin statistical analysis significance If an excess is observed over the background prediction, calculate p-value which uantifies how unlikely it is to get a corresponding excess as a fluctuation under the background-only hypothesis: p = P(n n obs s = 0) = k=n obs P(n = k s = 0) = k=n obs The smaller the p-value the stronger the evidence that s 0 Usually uote significance Z ( euivalent number of standard deviations for Gaussian fluctuation ): p = Z e x 2 /2 2π dx Z = 3 ( 3σ ) corresponds to p = 1.3 3 Z = 5 ( 5σ ) corresponds to p = 2.9 7 bk e b k! Magnar K. Bugge New resonance search using ATLAS open data 16 / 19
Single bin statistical analysis exclusion limit If no significant excess is observed, set exclusion limit on s Consider p-value corresponding to downwards fluctuations under the signal+background hypothesis: n obs p(s) = P(n n obs s) = P(n = k s) = k=0 n obs k=0 (s + b) k e (s+b) Classical freuentist: exclude s s up at 95% CL where p(s up ) = 5% Problem: Can exclude arbitrarily small s if data is low compared to b, e.g. consider n obs = 0: p(s) = P(n 0 s) = P(n = 0 s) = e (s+b) p(s up ) = e (sup+b) = 0.05 s up = ln(0.05) b 3 b Can exclude any signal as b 3 Magnar K. Bugge New resonance search using ATLAS open data 17 / 19 k!
Single bin statistical analysis exclusion limit Arbitrarily strong exclusion happens when the background hypothesis itself can be excluded as much as the signal+background hypothesis Can be avoided by dividing signal+background p-value by background-only euivalent: p mod (s) = p(s) p(0) CL s+b CL b CL s Exclusion at 95% CL for s s up where p mod (s up ) = 5% Revisit n obs = 0: p mod (s) = p(s) p(0) = P(n = 0 s) P(n = 0 s = 0) = e (s+b) e b = e s p mod (s up ) = e sup = 0.05 s up = ln(0.05) 3 Simple rule of thumb: always exclude a signal which predicts 3 expected events when n obs = 0 Magnar K. Bugge New resonance search using ATLAS open data 18 / 19
ATLAS open data 1 fb 1 of real s = 8 TeV ATLAS data released to the public! Comes with analysis and plotting tools, complete and ready for you to do your own analysis Magnar K. Bugge New resonance search using ATLAS open data 19 / 19