Searching for Higgs boson decays to charm quark pairs with charm jet tagging at ATLAS

Transcription

1 Searching for Higgs boson decays to charm quark pairs with charm jet tagging at ALAS Birmingham HEP Seminar 6th December 217 Andy Chisholm (CERN)

2 Introduction - Yukawa Couplings 1 Yukawa couplings between the Higgs (φ) and fermion (ψ) fields are possible: L fermion = y f [ ψl φψ R + ψ R φψl ] If φ has a non-zero VEV, expansion leads to: where h is the physical Higgs boson field... L fermion = y f v y f ψψ h ψψ 2 2 }{{}}{{} mass term Yukawa coupling term he End Result: Gauge invariant fermion mass terms Higgs fermion coupling proportional to the fermion mass (g Hf f = m f /v) H f f g Hf f While y f are still free parameters in the model, v 246 GeV is known from electroweak measurements and we know the fermion masses... We can predict the couplings in the SM!

3 Introduction - Charm Yukawa Coupling 2 Why is the charm quark Yukawa coupling important? he smallness of the charm (c) quark coupling 2mc (m H ) (y c = 4 3 ) make it highly v susceptible to modifications from potential new physics H c c decays constitute the largest part of the SM prediction for Γ H for which we have no experimental evidence o date, we only have experimental evidence for 3rd generation Yukawa couplings! What are the existing indirect constraints? 38.9% 38.9%.1% 2.9% 58.2%.1% 2.9% 58.2% H b b H c c H s s H other Cartoon of SM 125 GeV H q q branching fractions, H uū/d d too small to show! Constraints on unobserved Higgs decays impose around B(H c c) < 2%, global fits to LHC data indirectly bound Γ H leading to y c/yc SM < 6, assuming SM Higgs production and no BSM decays (arxiv:13.729, arxiv:153.29) Direct bound of around Γ H < 1 GeV from H γγ and H 4l lineshapes impose around y c/yc SM < 12, but this is model independent (arxiv:153.29) How can we constrain these couplings in a more direct way?

4 Direct probes of the charm quark Yukawa coupling at the LHC 3 Several methods to study the charm quark Yukawa couplings at the LHC have been proposed in the literature, the most promising (in my opinion) are: Idea 1 - Exclusive H J/ψ γ decays Rare exclusive radiative Higgs boson decays to vector mesons are sensitive to the Hq q couplings (arxiv:153.29) he H J/ψ γ decay has been proposed as a clean probe of the charm quark to Yukawa coupling, though decay width only evolves as (const. + y c) 2 (const. y c) Both ALAS and CMS have already begun to search for such decays in LHC Run 1... Idea 2 - Associated production of a Higgs boson and charm quark ree level sensitivity to charm quark to Yukawa coupling (arxiv: , arxiv: ) Use jet c-tagging to identify charm quark signature and a suitably clean Higgs decay (e.g. H γγ) Alternatively, study p H distribution to look for potential shape modifications... Idea 3 - Inclusive H c c decays (he focus of this seminar...) Inclusive H c c decays are directly sensitive to the charm quark to Yukawa coupling, with the decay width evolving as Γ H c c y 2 c Use double jet c-tagging and focus on VH (V = W, Z) production with leptonic V decays to mitigate the large multi-jet background

5 Idea 1 - H J/ψ γ Decays 4 he radiative decay H J/ψ γ could provide a clean probe of charm quark Yukawa coupling at the LHC Interference between direct (H c c) and indirect (H γγ ) contributions Direct (upper diagram) amplitude provides sensitivity to the magnitude and sign of the Hc c coupling Indirect (lower diagram) amplitude provides dominant contribution to the width, not sensitive to Yukawa couplings Very rare decays in the SM, but rate dominated by indirect component, sensitivity to Yukawa coupling somewhat diluted Γ = C I C D y c yc SM 2 7 MeV (C I,C D 1) B (H J/ψ γ) = (2.8 ±.2) 6 More details: JHEP 158 (215) 12 (arxiv: ) and Phys. Rev. D 9, 113 (214) (arxiv: )

6 H J/ψ γ - Run 1 Results (Phys. Rev.Lett. 114 (215), (arxiv: )) 5 First search for such rare Higgs decays was performed by ALAS with Run 1 dataset PRL 114, (215) P H Y S I C A L R E V I E W L E E R S Search for Higgs and Z Boson Decays to J=ψγ and ϒðnSÞγ with the ALAS Detector G. Aad et al. * (ALAS Collaboration) (Received 15 January 215; published 26 March 215) A search for the decays of the Higgs and Z bosons to J=ψγ and ϒðnSÞγ (n ¼ 1; 2; 3) is performed with pp collision data samples corresponding to integrated luminosities of up to 2.3 fb 1 collected at pffiffi s ¼ 8 ev with the ALAS detector at the CERN Large Hadron Collider. No significant excess of events is observed above expected backgrounds and 95% C.L. upper limits are placed on the branching fractions. In the J=ψγ final state the limits are and for the Higgs and Z boson decays, respectively, while in the ϒð1S; 2S; 3SÞγ final states the limits are ð1.3; 1.9; 1.3Þ 3 and ð3.4; 6.5; 5.4Þ 6, respectively. DOI:.13/PhysRevLett PACS numbers: 14.8.Bn, Dg, 14.7.Hp, 14.8.Ec week ending 27 MARCH 215 Studied H J/ψ γ with J/ψ µ + µ First direct information on decay modes sensitive to the Hc c coupling Similar limit subsequently found by CMS Interpreted as Hc c coupling limit of y c/yc SM < 22 at 95% CL (assuming dependence on σ(pp H)/Γ H is removed by considering ratio with H 4l rate) Phys. Lett. B753 (216) 341 (arxiv: ) Phys. Rev. D92, 3316 (215) (arxiv:153.29) Events / 4 GeV ALAS s=8 ev Ldt = 19.2 fb Data S+B Fit Background -3 H [B= ] -6 Z [B= ] m µµγ [GeV] Branching fraction limit (95% CL): B (H J/ψ γ) < Around 5 the SM expectation -1

7 Prospects for H J/ψ γ in a HL-LHC scenario (AL-PHYS-PUB ) 6 Run 1 H J/ψ γ analysis projected to s = 14 ev scenario with 3() fb 1 Events / ( 2 GeV ) 9 ALAS Simulation Preliminary 8 s=14 ev, 3 fb Data (Bkg. Only) S+B Fit Background H Signal Z Signal Expected branching ratio limit at 95% CL B (H J/ψγ) [ 6 ] B (Z J/ψγ) [ 7 ] Cut Based Multivariate Analysis Cut Based 3 fb fb Standard Model expectation B (H J/ψγ) [ 6 ] B (Z J/ψγ) [ 7 ] 2.9 ±.2.8 ± m µµγ (GeV) Events / ( 2 GeV ) 16 ALAS Simulation Preliminary 14 s=14 ev, 3 fb 12 8 Optimistic scenario with MVA analysis still only sensitive to B (H J/ψ γ) at 15 SM value with 3 fb 1 Data (Bkg. Only) S+B Fit Background H Signal Z Signal New ideas likely required to reach SM sensitivity in a HL-LHC scenario with this channel!

8 Idea 2 - Associated Higgs boson + charm quark production 7 he production of Higgs boson in association with a charm quark is directly sensitive to the charm quark Yukawa coupling Examples of direct (left and centre) and indirect (right) cg Hc diagrams (from arxiv: ) Expected p-value as a function of κ c = y c /y SM c (from arxiv: ) t-channel diagram (left) is expected to dominate the cross-section and is sensitive to the Yukawa coupling, highly sensitive channel! No experimental measurements yet, though the sensitivity at the HL-LHC has been surveyed in the literature (arxiv: ) Assuming a data sample of 3 ab 1 at s = 14 ev, O(1) constraints on y c/yc SM are expected to be obtained...

9 Idea 2 - Associated Higgs boson + charm quark production 8 Alternatively, don t c-tag at all... (1/σ dσ/dp,h)/(1/σ dσ/dp,h )SM κ c = - κ c = -5 κ c = κ c = 5 κb SM Δχ 2 = 2.3 Δχ 2 = 5.99 LHC Run I p,h [GeV] κ c Left: Effect of modified κ c on p H from cg Hc diagrams Right: bounds from Run 1 data (both from arxiv: ) In the case of a modified heavy quark Q = c, b Yukawa coupling, the shape of the inclusive p H spectrum would change due to the modified gq HQ contribution p H can be measured in the H γγ and H 4l channels, which imposes a 95% CL bound of 16 < y c/yc SM < 18 (arxiv: , based on ALAS+CMS Run 1) Projecting to HL-LHC scenario with 3 ab 1, bound evolves to.6 < y c/yc SM < 3.

10 Idea 3 - Inclusive H c c decays 9 Motivation Strategy he branching fraction for H c c decays is around 2.9% for a SM Higgs boson with m H = 125 GeV In comparison to the H J/ψ γ decay, this is a huge rate! Furthermore, it scales directly with y 2 c... In s = 13 ev pp collisions, one expects around 16 H c c decays in every 1 fb 1 of data! But, how can we hope to separate H c c from the HUGE jet background at the LHC? H c c Charm quark initiated jets (c-jet) will typically contain a c-hadron, though most of the jets produced in LHC pp collisions will not... If we can exploit the presence of a c-hadron within the jet, we can hope to separate c-jets from light flavour (u, d, s, g) and b-jets (which also have a unique signature) Focus on production channels involving leptons or large E miss and/or W (lν)h), to reduce the jet backgrond (e.g. Z(ll, νν)h

11 Part I - Charm jet tagging with ALAS Introduction Jets containing either c- or b-hadrons can be tagged by virtue of the unique properties of the heavy flavour hadrons hese techniques are collectively known as jet flavour tagging and only differ in the fine details if one is interested to tag c-jets or b-jets I will describe how these techniques are implemented within the ALAS experiement ( flavour tagging can mean different things to different collider experiments) b-jet: Jets containing a b-hadron Jet Labelling Conventions c-jet: Jets containing a c-hadron but no b-hadron Light flavour jet: Jets containing no b or c-hadrons (originating from u, d, s quark and gluon fragmentation)

12 he ALAS Detector at the LHC 11 General purpose detector, well suited to studying heavy flavour jets Inner Detector (ID): Silicon Pixels and Strips (SC) with ransition Radiation racker (R) η < 2.5 and (new for Run 2) Insertable B-Layer (IBL) LAr EM Calorimeter: Highly granular + longitudinally segmented (3-4 layers) Had. Calorimeter: Plastic scintillator tiles with iron absorber (LAr in fwd. region) Muon Spectrometer (MS): riggering η < 2.4 and Precision racking η < 2.7 Jet Energy Resolution: ypically σ E /E 5%/ E( GeV) 3% rack IP Resolution: σ d 6 µm and σ z 14 µm for p = 1 GeV (with IBL)

13 Properties of b-hadrons 12 Lifetime: Long enough to lead to a measureable decay length (around 5mm for a 5 GeV boost) Mass: Weakly decaying b-hadrons have masses around 5 GeV, leading to high decay product multiplicities (average of 5 charged particles per decay) Fragmentation: Much harder than jets initiated by other species (b-hadrons carry around 75% of jet energy, on average) Fraction of decays ALAS Simulation Preliminary Pythia8 PYHIA6 Herwig++ HERWIG Pythia8+EvtGen 1/N dn/dz ALAS Simulation Preliminary POWHEG pp tt R=.4 b-jets Anti-k t Pythia8 PYHIA6 Herwig++ HERWIG Multiplicity of B charged stable decay products Left: Mean charged multiplicity in B + mesons decays z p p / p jet C jet Right: b-quark fragmentation function

14 Properties of c-hadrons 13 Lifetime: Shorter than the b-hadrons by around a factor of 2-3, still enough for measureable decay length (around 1-3mm for a 5 GeV boost) Mass: Weakly decaying c-hadrons have masses around 2 GeV, around 2 3 lower than b-hadrons (mean of 2 charged particles per decay) Fragmentation: Softer than b-jets, but still harder than jets initiated by light species (c-hadrons carry around 55% of jet energy, on average) Fraction of decays ALAS Simulation Preliminary Pythia8 PYHIA6 Herwig++ HERWIG Pythia8+EvtGen Multiplicity of D charged stable decay products Left: Mean charged multiplicity in D + mesons decays 1/N dn/dz ALAS Simulation Preliminary POWHEG pp tt R=.4 c-jets Anti-k t Pythia8 PYHIA6 Herwig++ HERWIG z p p / p jet C jet Right: c-quark fragmentation function

15 Anatomy of a light flavour (u, d, s) jet 14 ypical Experimental Signature Light-quarks hadronise into many light hadrons which share the jet energy racks from this vertex often have impact parameters consistent with zero Long-lived light hadrons (e.g. K S, Λ ) can be produced, though they are more likely to decay very far (many cm) from the primary pp vertex

16 Anatomy of a c-jet 15 ypical Experimental Signature c-quark fragments into a c-hadron which carries around half of the jet energy c-hadron decay vertex often displaced from the primary pp vertex by a few mm racks from this vertex can often have large impact parameters

17 Anatomy of a b-jet 16 ypical Experimental Signature b-quark fragments into a b-hadron which carries most of the jet energy Most b-hadrons ( 9%) decay into c-hadrons b-hadron decay vertex often displaced from the primary pp vertex by a few mm Subsequent c-hadron decay vertex often displaced by a further few mm racks from both of these vertices often have large impact parameters

18 Introduction to charm jet tagging 17 Charm tagging is not new, many experiments at high energy ( s m B B) colliders (e.g. Sp ps, evatron, SLD, LEP, HERA) have built charm taggers which tend to fall within the following classes: Exclusive charm jet tagging Focus on the full reconstruction of exclusive c-hadron decay chains (e.g. D ± D (K π + )π ± ) or leptons from semi-leptonic c-hadron decays Can often provide a very pure sample of jets containing c-hadrons he efficiency is typically low O(1%), limited by the c-hadron branching fractions of interest Inclusive charm jet tagging An alternative approach is to to exploit more inclusive observables, such as track impact parameters or secondary vertices he efficiency of this approach is typically very high O(%)) he c-jet purity is often lower than these traditional approaches More suited for use with machine learning (ML) techniques ALAS have developed an inclusive c-tagging algorithm based on several low level taggers combined into a high level tagger using ML techniques

19 ALAS Low Level aggers: 1 - rack Impact Parameters (IP) 18 he signed IPs of tracks associated to jets are powerful jet flavour distriminants: Exploit sign of impact parameter: positive if track point of closest approach to PV is downstream of plane defined by the PV and jet axis racks from b-hadrons tend to have highly significant (IP/σ IP ) positive IPs, while most tracks from the PV have a narrow, symmetric distribution Very inclusive and highly efficient Relies upon accurate measurement of jet axis, sensitive to mis-tag high IP tracks from V decays or material interactions, IP/σ IP difficult to model in detector simulation Arbitrary units 2 3 ALAS Simulation Preliminary s=13 ev, tt b jets c jets Light flavour jets Arbitrary units 1 2 ALAS Simulation Preliminary s=13 ev, tt b jets c jets Light flavour jets rack signed d significance (Good) IP3D log(p /P ) b u Left: ransverse IP significance distribution Right: likelihood ratio discriminant based on 3D IPs of tracks

20 ALAS Low Level aggers: 2 - Secondary Vertices (SV) 19 Exploit expectation of a secondary vertex from either b or c-hadron decays: Attempt to reconstruct a secondary vertex from high IP tracks associated with jet Use invariant mass of tracks at SV to discriminate b or c-hadron decay vertices from V decays or material interations Exploit hard c/b-jet fragmentation, SV should carry a large fraction of jet energy SV found in up to 8% of b-jets but only a few % of light flavour jets Degraded light jet rejection as jet p increases, careful considerations to mitigate tagging of material interactions required Arbitrary units ALAS Simulation Preliminary s=13 ev, tt Arbitrary units ALAS Simulation Preliminary s=13 ev, tt b jets c jets Light flavour jets Arbitrary units ALAS Simulation Preliminary s=13 ev, tt b jets c jets Light flavour jets b jets c jets Light flavour jets m(sv) [GeV] S (SV) xyz f (SV) E Left: Inv. mass of tracks at SV Centre: 3D SV decay length significance Right: Energy fraction of SV tracks

21 ALAS Low Level aggers: 3 - Decay Chain (JetFitter algorithm) 2 Exploit common occurance of cascade decay chain; b-hadron c-hadron: Use Kalman filter to search for common axis on which three vertices lie: primary (pp) secondary (b-hadron) tertiary (c-hadron) Can then look for 1 track vertices with decay chain axis Addition of 1 track vertices improves efficiency, constraint to decay chain axis improves separation power of SV based discriminants Degraded performance for c/b-hadron vertices as jet p increases, high fake rate for 1 track vertices (increases light jet mis-tag rate) Arbitrary units ALAS Simulation Preliminary s=13 ev, tt b jets c jets Light flavour jets Arbitrary units ALAS Simulation Preliminary s=13 ev, tt b jets c jets Light flavour jets JetFitter efficiency ALAS Simulation Preliminary s=13 ev, tt b jets b jets 2 tracks c jets c jets 2 tracks Light jets Light jets 2 tracks N (JF) 1 trk vertices N (JF) 2 trk vertices Jet p [GeV] Left: Multiplicity of 1 track vertices Centre: Multiplicity of 2+ track vertices Right: Chain reco. efficiency vs. jet p

22 ALAS High Level c-tagger - Bringing Everything ogether 21 Combine approaches to exploit all features of c/b-jets and mitigate the shortcomings of the individual methods: Benefit from the advantages of all basic techniques/algorithms Complex sensitivity to convolution of all detector and physics modelling issues relies strongly on calibration in data (see next slide) Use the output of the three basic approaches as input to a boosted decision tree (BD) to build two discriminants, one trained to separate c-jets from b-jets (x-axis), another to separate c-jets from light-jets (y-axis) c-jets b-jets light flavour (u, d, s, g) jets vs light 1 3 Density vs light 1 3 Density vs light 1 3 Density c.5 41% efficiency c.5 41% efficiency c.5 41% efficiency c jets s = 13 ev, tt ALAS Simulation Preliminary b vs c b jets s = 13 ev, tt ALAS Simulation Preliminary b vs c light jets s = 13 ev, tt ALAS Simulation Preliminary c-tag jets by making a cut in the 2D discriminant space, working point optimised for H c c is shown in the rectangular selection (shaded region rejected) b vs c 5

23 Performance of the ALAS c-tagger 22 c tagging efficiency ALAS Preliminary s = 13 ev, 36.1 fb c jets, η < % c tagging efficiency working point Data Efficiency otal Uncertainty c tagging efficiency ALAS Preliminary s = 13 ev, 36.1 fb b jets, η < % c tagging efficiency working point Data Efficiency otal Uncertainty c tagging efficiency ALAS Preliminary s = 13 ev, 36.1 fb light jets, η < % c tagging efficiency working point Data Efficiency otal Uncertainty Fractional Uncertainty p [GeV] c-jets Fractional Uncertainty p [GeV] b-jets Fractional Uncertainty p [GeV] light flavour (u, d, s, g) jets Efficiency of c-tagging algorithm for b-, c- and light flavour jets measured in data Working point for H c c exhibits a c-jet tagging efficiency of around 4% Rejects b-jets by around a factor 4 and light jets by around a factor Efficiency calibrated in data with samples of b-jets from t Wb decays and c-jets from W cs, cd decays (in t t events) ypical total relative uncertainties of around 25%, 5% and 2% for c-, b- and light jets, respectively

24 Part II - Search for H c c decays with ALAS How can we use the charm tagger to search for H c c decays?

25 Search for H c c with pp ZH production (ALAS-CONF ) Given the success of the W /Z associated production channel in providing evidence for H b b decays, this channel is an obvious first candidate for a H c c search l + Z H p c 24 l p c Focus on ZH production with Z e + e and Z µ + µ decays for first ALAS analysis (ALAS-CONF ) Low exposure to experimental uncertainties, main backgrounds from Z + jets, Z(W /Z) and t t Pioneer use of new c-tagging algorithm developed by ALAS for Run 2 to identify the experimental signature of an inclusive H c c decay ALAS: arxiv: CMS: arxiv:

26 Introduction to pp ZH production at the LHC In s = 13 ev pp collisions, Higgs boson production in association with a Z boson represents around 1.6% of the inclusive Higgs boson production rate q 25 H he cross-section is dominated by the q q ZH process, with total cross-section σ q q.76 pb Smaller contributions from gg ZH, with total cross-section σ gg.12 pb, though it exhibits a harder p H spectrum below 15 GeV q g q q ZH diagram Z H - 1 gg ZH riangle Box q q ZH (NLO) Flipped Yukawa dσ/dp H [fb/bin] MSW28LO.1 LHC14eV p H [GeV] p H distribution for q q and gg initiated ZH production (from arxiv: ) Representative Feynman diagrams for q q/gg ZH processes MadGraph5 amc@nlo g Z gg ZH box diagram g H g Z gg ZH triangle diagram

27 Data Sample and Event Selection 26 Use a s = 13 ev pp collision sample collected during 215 and 216 corresponding to an integrated luminosity of 36.1 fb 1 Z l + l Selection rigger with lowest available p single electron or muon triggers Exactly two same flavour reconstructed leptons (e or µ) Both leptons p > 7 GeV and at least one with p > 27 GeV Require opposite charges (dimuons only) 81 < m ll < 1 GeV p Z > 75 GeV H c c Selection Consider anti-k R =.4 calorimeter jets with η < 2.5 and p > 2 GeV At least two jets with leading jet p > 45 GeV Form H c c candidate from the two highest p jets in an event At least one c-tagged jet from H c c candidate Dijet angular separation R jj requirement which varies with p Z Split events into 4 categories (with varying S/B) based on H c c candidates with 1 or 2 c-tags and p Z above/below 15 GeV

28 Signal and Background Modelling 27 Background Modelling Background dominated by Z + jets (enriched in heavy flavour jets) Smaller contributions from ZZ(q q), ZW (q q ) and t t Negligible (<.5%) contributions from W + jets, WW, single-top and multi-jet Simulation of ZH(c c/b b) Normalised with LHC Higgs XS WG YR4 recommendations (arxiv: ) ZH(b b) treated as background normalised to SM expectation (with σ B uncertainty) Events /.1 Data/Sim Data Stat. Uncertainty ZH(cc) 2 ZH(bb) Z + ll Z + cl Z + cc Z + bb,bc,bl ZW ZZ tt ALAS Preliminary s = 13 ev, 36.1 fb Z 2 c tags, 75 < p < 15 GeV 5 < m cc < 2 GeV Simulation normalised to data R cc Process MC Generator Normalisation Cross section q q ZH(c c/b b) Powheg+GoSaM+MiNLO+Pythia8 NNLO (QCD) NLO (EW) gg ZH(c c/b b) Powheg+Pythia8 NLO+NLL (QCD) Z + jets Sherpa NNLO ZZ and ZW Sherpa NLO t t Powheg+Pythia8 NNLO+NNLL he nominal MC generators used to model the signal and backgrounds

29 Background composition after c-tagging 28 Left: 1 c-tag events Events / GeV Data/Sim. Events / GeV Data/Sim ALAS Preliminary s = 13 ev, 36.1 fb Z 1 c tag, 75 < p < 15 GeV 5 < m cc < 2 GeV Simulation normalised to data Data Stat. Uncertainty ZH(cc) 2 ZH(bb) Z + ll Z + cl Z + cc Z + bb,bc,bl ZW ZZ tt Leading jet p [GeV] ALAS Preliminary 2. s = 13 ev, 36.1 fb Z c tag, p > 15 GeV 5 < m < 2 GeV 1.6 cc Simulation normalised to data Data Stat. Uncertainty ZH(cc) 2 ZH(bb) Z + ll Z + cl Z + cc Z + bb,bc,bl ZW ZZ tt Leading jet p [GeV] Events / GeV Data/Sim. Events / GeV Data/Sim ALAS Preliminary s = 13 ev, 36.1 fb Z 2 c tags, 75 < p < 15 GeV 5 < m cc < 2 GeV Simulation normalised to data Data Stat. Uncertainty ZH(cc) 2 ZH(bb) Z + ll Z + cl Z + cc Z + bb,bc,bl ZW ZZ tt Leading jet p [GeV] ALAS Preliminary s = 13 ev, 36.1 fb Z 2 c tags, p > 15 GeV 5 < m cc < 2 GeV Simulation normalised to data Data Stat. Uncertainty ZH(cc) 2 ZH(bb) Z + ll Z + cl Z + cc Z + bb,bc,bl ZW ZZ tt Leading jet p [GeV] Right: 2 c-tag events

30 Z + jets flavour composition after c-tagging 29 Left: 1 c-tag events Events / GeV Events / GeV Flavour composition of the Z + jets sample enriched with c-jets ALAS Preliminary s = 13 ev, 36.1 fb Z 1 c tag, 75 < p < 15 GeV Z + jets Simulation Dijet Flavour Composition ll cl cc bl bc bb ALAS Preliminary s = 13 ev, 36.1 fb Z 1 c tag, p > 15 GeV [GeV] m cc Z + jets Simulation Dijet Flavour Composition ll cl cc bl bc bb [GeV] m cc Events / GeV Events / GeV 3 1. ALAS Preliminary s = 13 ev, 36.1 fb Z.8 2 c tags, 75 < p < 15 GeV Z + jets Simulation Dijet Flavour Composition ll cl cc bl bc bb [GeV] m cc 25 ALAS Preliminary Z + jets Simulation s = 13 ev, 36.1 fb Z 2 c tags, p > 15 GeV Dijet Flavour Composition 2 ll cl cc bl bc bb [GeV] m cc Right: 2 c-tag events

31 ZZ and ZW flavour composition after c-tagging 3 c-tagged ZZ and ZW production enriched in Z c c and W cs, cd decays Left: 1 c-tag events Events / GeV Events / GeV ALAS Preliminary s = 13 ev, 36.1 fb Z 1 c tag, 75 < p < 15 GeV Z(W/Z) Simulation Dijet Flavour Composition ll cl cc bl bc bb [GeV] m cc 22 ALAS Preliminary Z(W/Z) Simulation 2 s = 13 ev, 36.1 fb Dijet Flavour Composition 18 Z 1 c tag, p > 15 GeV 16 ll cl cc 14 bl bc bb [GeV] m cc Events / GeV Events / GeV ALAS Preliminary s = 13 ev, 36.1 fb Z 2 c tags, 75 < p < 15 GeV Z(W/Z) Simulation Dijet Flavour Composition ll cl cc bl bc bb [GeV] m cc 25 ALAS Preliminary Z(W/Z) Simulation s = 13 ev, 36.1 fb Z 2 c tags, p > 15 GeV Dijet Flavour Composition 2 ll cl cc bl bc bb [GeV] m cc Right: 2 c-tag events

32 Quantifying the presence/absence of ZH(c c) production 31 Statistical Model Use the H c c candidate invariant mass m c c as S/B discriminant Perform simultaneous binned likelihood fit to 4 categories within region 5 < m c c < 2 GeV ZH(c c) signal parameterised with free signal strength parameter, µ, common to all categories Z + jets background determined directly from data with separate free normalisation parameter for each of the four categories Systematic Uncertainties Included in the fit model as constrained nuisance parameters which parametrize the constraints from auxiliary measurements (e.g. lepton/jet calibrations) Experimental uncertainties associated with luminosity, c-tagging, lepton and jet performance are all included in the model Normalisation, acceptance and m c c shape uncertainties associated with signal and background simulation are also included

33 Understanding the Sensitivity 32 Sensitivity dominated by systematic uncertainties, clear that these uncertainties should be reduced in order to fully exploit a larger dataset in the future Source σ/σ tot Statistical 49% Floating Z + jets Normalisation 31% Systematic 87% Flavour agging 73% Background Modeling 47% Lepton, Jet and Luminosity 28% Signal Modeling 28% MC statistical 6% Note: correlations between nuisance parameters within groups leads to i σ2 i σ 2 Syst. Background modelling (particularly Z + jets shape uncertainties) followed by c-tagging uncertainties have the dominant impact However, we can expect many of these uncertainties (particularly effect of the Z + jets normalisation) to reduce with a larger dataset

34 Fit Result 33 p Z > 15 GeV 75 < p Z < 15 GeV Events / ( GeV ) Data/Bkgd. Events / ( GeV ) Data/Bkgd ALAS Preliminary s = 13 ev, 36.1 fb Z 1 c tag, p > 15 GeV 1 c-tag 2 c-tags Data Pre fit Fit Result Z + jets tt ZZ ZW ZH(bb) ZH(cc) ( SM) obs_x_chan_hi_1t2pj_2l [GeV] 3 6 ALAS Preliminary s = 13 ev, 36.1 fb Z 5 1 c tag, 75 < p < 15 GeV m cc Data Pre fit Fit Result Z + jets tt ZZ ZW ZH(bb) ZH(cc) ( SM) obs_x_chan_mi_1t2pj_2l [GeV] m cc Events / ( GeV ) Data/Bkgd. Events / ( GeV ) Data/Bkgd. 224 ALAS Preliminary Data Pre fit s = 13 ev, 36.1 fb 2 Fit Result Z 2 c tags, p > 15 GeV Z + jets 183 tt ZZ 16 ZW ZH(bb) 142 ZH(cc) ( SM) obs_x_chan_hi_2t2pj_2l [GeV] ALAS Preliminary s = 13 ev, 36.1 fb Z 2 c tags, 75 < p < 15 GeV m cc Data Pre fit Fit Result Z + jets tt ZZ ZW ZH(bb) ZH(cc) ( SM) obs_x_chan_mi_2t2pj_2l [GeV] m cc No significant evidence for ZH(c c) production Data consistent with background only hypothesis SM expected number of ZH(c c) events 1 c-tag 75 < p Z < 15 GeV c-tag p Z > 15 GeV c-tags 75 < p Z < 15 GeV.5 2 c-tags p Z > 15 GeV.3

35 Fit Result 34 p Z > 15 GeV 75 < p Z < 15 GeV Events / ( GeV ) Data/Bkgd. Events / ( GeV ) Data/Bkgd ALAS Preliminary s = 13 ev, 36.1 fb Z 1 c tag, p > 15 GeV 1 c-tag 2 c-tags Data Pre fit Fit Result ZH(bb) ZZ ZW Z + jets tt ZH(cc) ( SM) obs_x_chan_hi_1t2pj_2l [GeV] ALAS Preliminary s = 13 ev, 36.1 fb Z 1 c tag, 75 < p < 15 GeV m cc Data Pre fit Fit Result ZH(bb) ZZ ZW Z + jets tt ZH(cc) ( SM) obs_x_chan_mi_1t2pj_2l [GeV] m cc Events / ( GeV ) Data/Bkgd. Events / ( GeV ) Data/Bkgd. 22 ALAS Preliminary Data 3 Pre fit s = 13 ev, 36.1 fb 2 Fit Result Z 2 c tags, p > 15 GeV ZH(bb) ZZ ZW 16 Z + jets 2 tt ZH(cc) ( SM) obs_x_chan_hi_2t2pj_2l [GeV] ALAS Preliminary s = 13 ev, 36.1 fb Z 2 c tags, 75 < p < 15 GeV m cc Data Pre fit Fit Result ZH(bb) ZZ ZW Z + jets tt ZH(cc) ( SM) obs_x_chan_mi_2t2pj_2l [GeV] m cc No significant evidence for ZH(c c) production Data consistent with background only hypothesis SM expected number of ZH(c c) events 1 c-tag 75 < p Z < 15 GeV c-tag p Z > 15 GeV c-tags 75 < p Z < 15 GeV.5 2 c-tags p Z > 15 GeV.3

36 Results 35 Cross check with ZV production o validate background modelling and uncertainty prescriptions, measure production rate of the sum of ZZ and ZW relative to the SM expectation Observe (expect) ZV production with significance of 1.4σ (2.2σ) Measure ZV signal strength of , consistent with SM expectation Limits on ZH(c c) production 95% CL CL s upper limit on σ(pp ZH) B(H c c) [pb] Observed Median Expected Expected +1σ Expected 1σ No evidence for ZH(c c) production with current dataset (as expected) Upper limit of σ(pp ZH) B(H c c) < 2.7 pb set at 95% CL, to be compared to an SM value of pb Corresponds to 1 the SM expectation World s most stringent direct constraint on H c c decays!

37 Interpreting the limit in terms of a constraint on y c - I 36 Warning: None of the following interpretation is sanctioned by ALAS, responsibility lies solely with me! Use the leading order motivated kappa framework to study how a potential modifications to the Higgs-charm coupling would affect σ(pp ZH) B(H c c) σ i B j = σ i( κ) Γ j ( κ) Γ H As described in arxiv: , assume the factorisation of production and decay shown above, afforded by the narrow width approximation Define set of kappa coupling modifiers κ such that LO production or decay modes (e.g. H c c) change as κ 2 i = σ i /σi SM or κ 2 i = Γ i /Γ SM i Production modes or decays involving loops (e.g. H γγ, gg H) can also be studied by resolving the loop in terms of their tree level couplings (e.g. t th) Can approximate modifications to pp ZH cross section and B(H c c) with: σ pp ZH (κ Z, κ t) = κ 2 Z σ q q ZH + (2.27 κ 2 Z +.37 κ 2 t 1.64 κ tκ Z ) σ gg ZH B(H c c)(κ c) = κ 2 c B(H c c) SM 1 + (κ 2 c 1) B(H c c) SM (where the gg c c/b b ZH loops have not been included (very small effect) and evolution of Γ H varies only with κ c )

38 Interpreting the limit in terms of a constraint on y c - II 37 For SM pp ZH production, the rate vs. κ c saturates at around 33 the SM value when B(H c c) 1 (far below the limit)... However, in a general BSM scenario, one could also expect the other Higgs couplings to be modified! SM B(H cc) B(H cc)] σ(pp ZH) [σ(pp ZH) κ t κ t κ t κ t κ t = 1, κ Z = 1 = 5, κ Z = 1 =, κ Z = 1 = 1, κ Z = 1.5 = 1, κ Z = 2 Observed Limit (ALAS CONF ) In a scenario where the ZH coupling is modified (e.g. κ Z 2), strong bounds of around κ c < can be obtained (assuming the predicted Γ H, i.e. no new particles) Similarly, if one modifies the t th coupling (e.g. κ t ) bounds of around κ c < 4 are also possible, BU both scenarios are strongly disfavoured by LHC data... κ c

39 Interpreting the limit in terms of a constraint on y c - III 38 For very large values of κ c, the tree level c c ZH process (i.e. two c quarks from the protons) becomes important! (see arxiv: for more details) SM B(H cc) B(H cc)] σ(pp ZH) [σ(pp ZH) = 1 = 1, κ Z κ t Observed Limit (ALAS CONF ) c c H Z his additonal production mechanism allows a bound of around κ c < 3 to be obtained, without modifying any other Higgs boson couplings However, by the time this becomes relevent, Γ H would be saturated by H c c decays κ c

40 Summary and Prospects Summary Search for ZH(c c) production exploiting new c-tagging techniques provides limit of σ(pp ZH) B(H c c) < 2.7 pb excluding 1 SM expectation Demonstrates that this inclusive channel is likely more sensitive to the charm quark Yukawa coupling than the exclusive H J/ψ γ channel Not yet able to compete with constraints obtained from interpreting measurements of Higgs boson kinematic distributions in terms of modified gc Hc production Clear that no single approach can yet claim it will manage to probe the charm quark Yukawa coupling down to the SM prediction by the end of the LHC era Likely that multiple approaches will be required, this channel will become ever more important as larger datasets are collected! What next for inclusive H c c decays? Large gains in sensitivity possible with multivariate techniques and other VH channels (e.g. W (lν)/z(νν)) or a dedicated search/category in the high p H boosted regime If future c-tagging algorithms can reach the performance of today s b-tagging, one could expect to observe H c c decays at the LHC! Performance of c-tagging is developing rapidly, next generation algorithms already exploit advanced ML techniques (AL-PHYS-PUB ), huge scope for innovation!

41 Additional Slides

42 Examples of c-tagging input variables 41 Arbitrary units ALAS Simulation Preliminary s = 13 ev, tt b jets c jets Light flavour jets Arbitrary units ALAS Simulation Preliminary s = 13 ev, tt b jets c jets Light flavour jets Arbitrary units 1 1 ALAS Simulation Preliminary s = 13 ev, tt b jets c jets Light flavour jets rack signed d significance (Good) IP3D log(p b /P u ) L xyz/σ Lxyz (SV) Arbitrary units 1 ALAS Simulation Preliminary s = 13 ev, tt b jets c jets Arbitrary units 1 ALAS Simulation Preliminary s = 13 ev, tt b jets c jets Arbitrary units ALAS Simulation Preliminary s = 13 ev, tt b jets 1 c jets 1 Light flavour jets 1 Light flavour jets 1 Light flavour jets f E (SV) m(sv) [GeV] 5 More details in AL-PHYS-PUB N rkatvtx (SV)

43 ALAS Low Level aggers: Using muons (Soft Muon agger) 42 Exploit the large branching fractions for the semi-leptonic c/b hadron decays and the clean muon-in-jet experimental signature: Expect much higher rate of muons within b/c-jets, relative to light flavour jets, due to the decays B µνx and B DX µνx (B of around % each) Complementary to SV and IP based taggers, different c/b hadron properties exploited and ALAS detector components employed Light flavour jet backgrounds from muons produced in π/k decays in flight difficult to model in simulation Entries / b jets c jets Light flavour jets ALAS Simulation Preliminary s = 13 ev, t t R Entries / 2 GeV.6 b jets c jets Light flavour jets ALAS Simulation Preliminary s = 13 ev, t t rel p [GeV] Entries / b jets c jets Light flavour jets ALAS Simulation Preliminary s = 13 ev, t t SM Left: R of muon w.r.t. jet axis Centre: p of muon relative to the jet axis Right: BD built from muon observables