AtlasPublic Web 
Internal TWiki
|
Evidence for Higgs boson decays to a low-mass dilepton system and a photon in p p collisions at √s = 13 TeV with the ATLAS detector ATLAS-CONF-2021-002 2 February 2021
|
||
| Content | Preview | |
|---|---|---|
| Main document (CDS record), Physics Briefing - internal | pdf from CDS | |
| Figures Tables | - | |
| Abstract | ||
| A search for the Higgs boson decaying into a photon and a pair of electrons or muons with an invariant mass $m_{\ell\ell} < 30$ GeV is presented. The analysis is performed using 139 fb$^{-1}$ of proton-proton collision data, produced by the LHC at a centre-of-mass energy of 13 TeV and collected by the ATLAS experiment. Evidence for the $H\to\ell\ell\gamma$ process is found with a significance of $3.2\sigma$ over the background-only hypothesis, compared to an expected significance of $2.1\sigma$. The best-fit value of the signal strength parameter, defined as the ratio of the observed signal yield to the one expected in the Standard Model, is $\mu = 1.5 \pm 0.5$. The Higgs boson production cross-section times the $H \rightarrow\ell\ell\gamma$ branching ratio for $m_{\ell\ell} < 30$ GeV is determined to be $8.7 ^{+2.8}_{-2.7} \text{fb}$. | ||
| Figures | ||
|
Figure 01a: Representative Feynman diagrams of the H → ℓℓγ process. The H→ ℓℓ decay with a photon from final state radiation is shown in diagram (d). png (17kB) pdf (77kB) |
 |
|
|
Figure 01b: Representative Feynman diagrams of the H → ℓℓγ process. The H→ ℓℓ decay with a photon from final state radiation is shown in diagram (d). png (24kB) pdf (76kB) |
 |
|
|
Figure 01c: Representative Feynman diagrams of the H → ℓℓγ process. The H→ ℓℓ decay with a photon from final state radiation is shown in diagram (d). png (25kB) pdf (67kB) |
 |
|
|
Figure 01d: Representative Feynman diagrams of the H → ℓℓγ process. The H→ ℓℓ decay with a photon from final state radiation is shown in diagram (d). png (13kB) pdf (43kB) |
 |
|
|
Figure 02a: (a) Ratio of reconstructed to true merged-ee energy in simulated H → γ*γ → eeγ events as a function of the true merged-ee pT for several energy calibration techniques. The merged-ee object is calibrated as a photon with a conversion radius of 30 mm (black, analysis choice), 100 mm (red), and 400 mm (blue) or as an electron (purple). (b) Combined merged-ee identification and isolation efficiency extracted from Z→ℓℓγ events with a photon that coverts within a radius of rconv<160 mm. The efficiencies are shown for photons with |η|<0.8 as a function of pT. Data (black) are compared with simulated Z→ℓℓγ events (red). The resulting efficiencies are also compared with efficiencies in simulated H → γ*γ → eeγ events (blue). On all points, the vertical error bars indicate the statistical uncertainties. png (67kB) pdf (22kB) |
 |
|
|
Figure 02b: (a) Ratio of reconstructed to true merged-ee energy in simulated H → γ*γ → eeγ events as a function of the true merged-ee pT for several energy calibration techniques. The merged-ee object is calibrated as a photon with a conversion radius of 30 mm (black, analysis choice), 100 mm (red), and 400 mm (blue) or as an electron (purple). (b) Combined merged-ee identification and isolation efficiency extracted from Z→ℓℓγ events with a photon that coverts within a radius of rconv<160 mm. The efficiencies are shown for photons with |η|<0.8 as a function of pT. Data (black) are compared with simulated Z→ℓℓγ events (red). The resulting efficiencies are also compared with efficiencies in simulated H → γ*γ → eeγ events (blue). On all points, the vertical error bars indicate the statistical uncertainties. png (60kB) pdf (16kB) |
 |
|
|
Figure 03a: mℓℓγ distributions of the selected events and the results of the global fit, for the VBF-enriched categories (a, b, c), the high-pTt categories (d, e, f), and the low-pTt categories (g, h, i). The ee-resolved categories are shown in the left column, the ee-merged categories in the middle and the μμ categories in the right column. The data are shown as the black points with statistical uncertainties. The red curve shows the combined signal-plus-background model, the dashed black line shows the model of the non-resonant background component and the dotted blue line denotes the sum of the non-resonant background and the resonant H → γγ background. The curves are obtained from the fit, i.e. they include the best-fit values of the parameter of interest and the nuisance parameters, including the spurious signal. The bottom panel shows the residuals of the data with respect to the non-resonant background component of the signal-plus-background fit. png (108kB) pdf (21kB) |
 |
|
|
Figure 03b: mℓℓγ distributions of the selected events and the results of the global fit, for the VBF-enriched categories (a, b, c), the high-pTt categories (d, e, f), and the low-pTt categories (g, h, i). The ee-resolved categories are shown in the left column, the ee-merged categories in the middle and the μμ categories in the right column. The data are shown as the black points with statistical uncertainties. The red curve shows the combined signal-plus-background model, the dashed black line shows the model of the non-resonant background component and the dotted blue line denotes the sum of the non-resonant background and the resonant H → γγ background. The curves are obtained from the fit, i.e. they include the best-fit values of the parameter of interest and the nuisance parameters, including the spurious signal. The bottom panel shows the residuals of the data with respect to the non-resonant background component of the signal-plus-background fit. png (113kB) pdf (22kB) |
 |
|
|
Figure 03c: mℓℓγ distributions of the selected events and the results of the global fit, for the VBF-enriched categories (a, b, c), the high-pTt categories (d, e, f), and the low-pTt categories (g, h, i). The ee-resolved categories are shown in the left column, the ee-merged categories in the middle and the μμ categories in the right column. The data are shown as the black points with statistical uncertainties. The red curve shows the combined signal-plus-background model, the dashed black line shows the model of the non-resonant background component and the dotted blue line denotes the sum of the non-resonant background and the resonant H → γγ background. The curves are obtained from the fit, i.e. they include the best-fit values of the parameter of interest and the nuisance parameters, including the spurious signal. The bottom panel shows the residuals of the data with respect to the non-resonant background component of the signal-plus-background fit. png (69kB) pdf (21kB) |
 |
|
|
Figure 03d: mℓℓγ distributions of the selected events and the results of the global fit, for the VBF-enriched categories (a, b, c), the high-pTt categories (d, e, f), and the low-pTt categories (g, h, i). The ee-resolved categories are shown in the left column, the ee-merged categories in the middle and the μμ categories in the right column. The data are shown as the black points with statistical uncertainties. The red curve shows the combined signal-plus-background model, the dashed black line shows the model of the non-resonant background component and the dotted blue line denotes the sum of the non-resonant background and the resonant H → γγ background. The curves are obtained from the fit, i.e. they include the best-fit values of the parameter of interest and the nuisance parameters, including the spurious signal. The bottom panel shows the residuals of the data with respect to the non-resonant background component of the signal-plus-background fit. png (131kB) pdf (23kB) |
 |
|
|
Figure 03e: mℓℓγ distributions of the selected events and the results of the global fit, for the VBF-enriched categories (a, b, c), the high-pTt categories (d, e, f), and the low-pTt categories (g, h, i). The ee-resolved categories are shown in the left column, the ee-merged categories in the middle and the μμ categories in the right column. The data are shown as the black points with statistical uncertainties. The red curve shows the combined signal-plus-background model, the dashed black line shows the model of the non-resonant background component and the dotted blue line denotes the sum of the non-resonant background and the resonant H → γγ background. The curves are obtained from the fit, i.e. they include the best-fit values of the parameter of interest and the nuisance parameters, including the spurious signal. The bottom panel shows the residuals of the data with respect to the non-resonant background component of the signal-plus-background fit. png (127kB) pdf (23kB) |
 |
|
|
Figure 03f: mℓℓγ distributions of the selected events and the results of the global fit, for the VBF-enriched categories (a, b, c), the high-pTt categories (d, e, f), and the low-pTt categories (g, h, i). The ee-resolved categories are shown in the left column, the ee-merged categories in the middle and the μμ categories in the right column. The data are shown as the black points with statistical uncertainties. The red curve shows the combined signal-plus-background model, the dashed black line shows the model of the non-resonant background component and the dotted blue line denotes the sum of the non-resonant background and the resonant H → γγ background. The curves are obtained from the fit, i.e. they include the best-fit values of the parameter of interest and the nuisance parameters, including the spurious signal. The bottom panel shows the residuals of the data with respect to the non-resonant background component of the signal-plus-background fit. png (73kB) pdf (22kB) |
 |
|
|
Figure 03g: mℓℓγ distributions of the selected events and the results of the global fit, for the VBF-enriched categories (a, b, c), the high-pTt categories (d, e, f), and the low-pTt categories (g, h, i). The ee-resolved categories are shown in the left column, the ee-merged categories in the middle and the μμ categories in the right column. The data are shown as the black points with statistical uncertainties. The red curve shows the combined signal-plus-background model, the dashed black line shows the model of the non-resonant background component and the dotted blue line denotes the sum of the non-resonant background and the resonant H → γγ background. The curves are obtained from the fit, i.e. they include the best-fit values of the parameter of interest and the nuisance parameters, including the spurious signal. The bottom panel shows the residuals of the data with respect to the non-resonant background component of the signal-plus-background fit. png (142kB) pdf (24kB) |
 |
|
|
Figure 03h: mℓℓγ distributions of the selected events and the results of the global fit, for the VBF-enriched categories (a, b, c), the high-pTt categories (d, e, f), and the low-pTt categories (g, h, i). The ee-resolved categories are shown in the left column, the ee-merged categories in the middle and the μμ categories in the right column. The data are shown as the black points with statistical uncertainties. The red curve shows the combined signal-plus-background model, the dashed black line shows the model of the non-resonant background component and the dotted blue line denotes the sum of the non-resonant background and the resonant H → γγ background. The curves are obtained from the fit, i.e. they include the best-fit values of the parameter of interest and the nuisance parameters, including the spurious signal. The bottom panel shows the residuals of the data with respect to the non-resonant background component of the signal-plus-background fit. png (142kB) pdf (24kB) |
 |
|
|
Figure 03i: mℓℓγ distributions of the selected events and the results of the global fit, for the VBF-enriched categories (a, b, c), the high-pTt categories (d, e, f), and the low-pTt categories (g, h, i). The ee-resolved categories are shown in the left column, the ee-merged categories in the middle and the μμ categories in the right column. The data are shown as the black points with statistical uncertainties. The red curve shows the combined signal-plus-background model, the dashed black line shows the model of the non-resonant background component and the dotted blue line denotes the sum of the non-resonant background and the resonant H → γγ background. The curves are obtained from the fit, i.e. they include the best-fit values of the parameter of interest and the nuisance parameters, including the spurious signal. The bottom panel shows the residuals of the data with respect to the non-resonant background component of the signal-plus-background fit. png (136kB) pdf (22kB) |
 |
|
|
Figure 04: Best-fit values of the signal strength parameters for all event categories, in a fit where the signal strength in each category is allowed to float independently, compared with the result of the global fit. png (61kB) pdf (15kB) |
 |
|
|
Figure 05: mℓℓγ distribution, with every data event reweighted by a category-dependent weight, ln(1+S90/B90), where S90 is the number of signal events in the smallest window containing 90% of the expected signal, and B90 is the expected number of background events in the same window, estimated from fits to the data sidebands using the background models. The data are shown as the black points with statistical uncertainties. The parameterised signal and backgrounds are also added up with the category-dependent weight. The red curve shows the combined signal-plus-background model when fitting all analysis categories simultaneously, the dashed black line shows the model of the non-resonant background component and the dotted blue line denotes the sum of the non-resonant background and the resonant H → γγ background. The curves are obtained from the fit, i.e. they include the best-fit values of the parameter of interest and the nuisance parameters, including the spurious signal. The bottom panel shows the residuals of the data with respect to the non-resonant background component of the signal-plus-background fit. png (173kB) pdf (30kB) |
 |
|
|
Figure 06: Best-fit values of the signal strength parameters, in different fits with common signal strength parameters defined per event category or lepton flavour. In the first fit, there is a common signal signal strength for all ee-resolved categories, a separate one for all ee-merged categories, and one for all μμ categories. In the second fit, there are three independent signal strenth parameters for VBF-enriched, high-pTt, and low-pTt categories. In the final fit, there are two signal strengh parameters, one shared by all ee categories and one for all μμ categories. These results are compared with the result of the global fit. png (54kB) pdf (15kB) |
 |
|
|
Figure 07a: Combined merged-ee identification and isolation efficiency extracted from Z→ℓℓγ events with a converted photon with rconv<160 mm as a function of merged-ee pT in the (top-left) 0.8<|η|<1.37, (top-right) 1.52<|η|<2.01 and (bottom) 2.01<|η|<2.37 range. Data efficiencies (black points) are compared with efficiencies in simulated Z→ℓℓγ events (blue markers). Additionally, the results are compared with those in simulated H → γ*γ → eeγ events. png (60kB) pdf (16kB) |
 |
|
|
Figure 07b: Combined merged-ee identification and isolation efficiency extracted from Z→ℓℓγ events with a converted photon with rconv<160 mm as a function of merged-ee pT in the (top-left) 0.8<|η|<1.37, (top-right) 1.52<|η|<2.01 and (bottom) 2.01<|η|<2.37 range. Data efficiencies (black points) are compared with efficiencies in simulated Z→ℓℓγ events (blue markers). Additionally, the results are compared with those in simulated H → γ*γ → eeγ events. png (60kB) pdf (16kB) |
 |
|
|
Figure 07c: Combined merged-ee identification and isolation efficiency extracted from Z→ℓℓγ events with a converted photon with rconv<160 mm as a function of merged-ee pT in the (top-left) 0.8<|η|<1.37, (top-right) 1.52<|η|<2.01 and (bottom) 2.01<|η|<2.37 range. Data efficiencies (black points) are compared with efficiencies in simulated Z→ℓℓγ events (blue markers). Additionally, the results are compared with those in simulated H → γ*γ → eeγ events. png (61kB) pdf (16kB) |
 |
|
|
Figure 08a: mℓℓγ distributions for events satisfying the H→ γ*γ→ℓℓγ selection in data for several event categories: (top-left) ee resolved low-pTt, (top-right) ee merged low-pTt, and (bottom) μμ low-pTt. The black points show the data with error bars for the statistical uncertainty. The contribution from non-resonant ℓℓγ process (green, obtained from simulation), and the contributions from the ℓℓ+jets and γ+jets processes (red and blue/purple, obtained from data control regions) are shown as stacked histograms; the total background estimate is normalised to the number of data events. Two γ+jets populations with different fake rates and mass distributions are found in the categories involving resolved electrons, separated according to the angular distance between the electrons ΔR. The bottom panels show the normalised residuals (pulls) of the data with respect to the sum of background components. png (153kB) pdf (27kB) |
 |
|
|
Figure 08b: mℓℓγ distributions for events satisfying the H→ γ*γ→ℓℓγ selection in data for several event categories: (top-left) ee resolved low-pTt, (top-right) ee merged low-pTt, and (bottom) μμ low-pTt. The black points show the data with error bars for the statistical uncertainty. The contribution from non-resonant ℓℓγ process (green, obtained from simulation), and the contributions from the ℓℓ+jets and γ+jets processes (red and blue/purple, obtained from data control regions) are shown as stacked histograms; the total background estimate is normalised to the number of data events. Two γ+jets populations with different fake rates and mass distributions are found in the categories involving resolved electrons, separated according to the angular distance between the electrons ΔR. The bottom panels show the normalised residuals (pulls) of the data with respect to the sum of background components. png (81kB) pdf (25kB) |
 |
|
|
Figure 08c: mℓℓγ distributions for events satisfying the H→ γ*γ→ℓℓγ selection in data for several event categories: (top-left) ee resolved low-pTt, (top-right) ee merged low-pTt, and (bottom) μμ low-pTt. The black points show the data with error bars for the statistical uncertainty. The contribution from non-resonant ℓℓγ process (green, obtained from simulation), and the contributions from the ℓℓ+jets and γ+jets processes (red and blue/purple, obtained from data control regions) are shown as stacked histograms; the total background estimate is normalised to the number of data events. Two γ+jets populations with different fake rates and mass distributions are found in the categories involving resolved electrons, separated according to the angular distance between the electrons ΔR. The bottom panels show the normalised residuals (pulls) of the data with respect to the sum of background components. png (142kB) pdf (25kB) |
 |
|
|
Figure 09: Simulated mℓℓγ invariant mass distribution for mH = 125.09 GeV in the ee-merged high-pTt (red open circles) and ee-resolved low-pTt (blue closed circles) categories. Of the nine considered categories, the ee-merged high-pTt category has the best resolution, and the ee-resolved low-pTt category the worst. The parameterised signal models, obtained by fits with a Double-Sided Crystal-Ball function, are also shown. png (141kB) pdf (29kB) |
 |
|
|
Figure 10: Merged-ee identification efficiency as a function of the true ΔR between the electrons (ΔRee) for simulated H→ γ*γ → eeγ events. The denominator of the efficiency includes objects with an EM cluster matched to two opposite-charge tracks, each having pT > 5 GeV, |η| < 2.5 and at least seven hits in the pixel and microstrip detectors combined. These preselected merged-ee objects are also required to have pT > 20 GeV and |η| < 2.37. In addition, only events where both merged-ee tracks are matched to a generator-level electron from the γ*→ee process are considered. Approximately 80% of events in the denominator are in the region ΔRee < 0.1, where the standard electron identification algorithms are less efficient. The error bars represent the statistical uncertainty. png (48kB) pdf (15kB) |
 |
|
|
Figure 11: Event display of a candidate H→ eeγ event from the ee-merged VBF-enriched category. The invariant mass of the ee system is 0.05 GeV and the invariant mass of the eeγ system is 125.8 GeV. The pT of the photon is 72 GeV and the pT of the merged-ee candidate is 103 GeV. The two visible jets have pTj1=124 GeV and pTj2=64 GeV, respectively, and mjj=664 GeV and |Δηjj|=4.0. png (2MB) |
 |
|
|
Figure 12: Event display of a candidate H→μμγ event from the μμ low-pTt category. The invariant mass of the μμ system is 0.6 GeV and the invariant mass of the μμγ system is 124.9 GeV. The pT of the photon is 60 GeV, the pT of the leading muon is 60 GeV and the pT of the subleading muon is 28 GeV. png (1MB) |
 |
|
| Tables | ||
|
Table 01: Number of data events selected in each analysis category in the mℓℓγ mass range of 110-160 GeV. In addition, the following numbers are given: number of H → γ*γ → ℓℓγ events in the smallest mℓℓγ window containing 90% of the expected signal (S90), the non-resonant background in the same interval (B90N) as estimated from fits to the data sidebands using the background models described in Section 6, the resonant background in the same interval (BH → γγ), the expected signal purity f90 = S90/(S90+B90), and the expected significance estimate defined as Z90 = √ 2( (S90+B90) ln(1+S90/B90) - S90) where B90 = B90N+BH → γγ. png (25kB) pdf (49kB) |
 |
|
|
Table 02: Best-fit values of the signal strength parameters, in different fits with common signal strength parameters defined per event category or lepton flavour. In the first fit, there is a common signal strength for all ee resolved categories, a separate one for all ee merged categories, and one for all μμ categories. In the second fit, there are three independent signal strength parameters for VBF-enriched, high-pTt, and low-pTt categories. In the final fit, there are two signal strength parameters, one shared by all ee categories and one for all μμ categories. These results are compared with the result of the global fit. png (34kB) pdf (67kB) |
 |
|
|
Table 03: Best-fit values of the signal strength parameters for all event categories, in a fit where the signal strength in each category is allowed to float independently, compared with the result of the global fit. png (36kB) pdf (66kB) |
 |
|
|
Table 04: Systematic uncertainties on the measured signal strength and the measured cross section times branching ratio, in percent. The uncertainties are symmetrized and ordered by impact. The statistical and total uncertainties are shown for comparison. png (17kB) pdf (35kB) |
 |
|
