Chapter 5: First evidence of a Higgs boson decay to a pair of muons
5.6 VBF category
5.6.6 Results
We perform independent fits to the signal strength modifierπin all three years, and take in to account the signal region and the side-band simultaneously per fit. These same fits are also used to derive the signal significance. A combined fit is then performed to the globalπdefined as(πΓ B (π» β π π))π π π /(πΓ B (π» β π π))π π. The expected and observed significance for each of the fits are shown in table5.3.
The data and simulation distributions for the fitted DNN is shown in Figures5.29, 5.30,5.31and5.32. The best fit signal strength forππ» 125.38 GeV isπ=1.36+β00..6961
corresponding to an observed significance of 2.40π.
Year Significance
Observed Expected
2016 2.01 1.04
2017 0.00 0.88
2018 2.21 1.21
Combined 2.40 1.77
Table 5.3: The observed and expected significance in the VBF category for excluding the background-only null hypothesis forππ»=125.38 GeV, for each year and for the combined data-taking period.
β1 10 1 10 102 103 104 105 106
Events
Data Zjj-EW
DY Top quark
Diboson
(13 TeV) 35.9 fb-1
CMS Post-fit VBF-SB 2016
= 125.38 GeV mH
0 0.5 1 1.5 2
Data/Bkg.
Pre-fit
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
VBF DNN output
0 0.5 1 1.5 2
Data/Bkg.
Post-fit
β2 10
β1 10 1 10 102 103 104 105 106 107 108
Events
Data Hβ¡¡
Zjj-EW DY
Top quark Diboson
VBF ggH
(13 TeV) 35.9 fb-1
CMS Post-fit VBF-SR 2016
= 125.38 GeV mH
0 0.5 1 1.5 2 2.5
Data/Bkg.
Pre-fit
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
VBF DNN output
0 0.5 1 1.5 2 2.5
Data/Bkg.
Post-fit
Figure 5.29: The observed DNN output distribution in the VBF-SB (left) and VBF- SR (right) regions compared to the post-fit background estimate for various SM processes. The predicted backgrounds are obtained from a combined signal-plus- background fit performed across analysis regions and eras. Distributions reported are related to the 2016 data-taking period. In the second panel, the ratio between data and the pre-fit background prediction is shown. The gray band indicates the total pre-fit uncertainty obtained from the systematic sources previously described.
The third panel reports the ratio between data and the post-fit background prediction from the signal-plus-background fit. The gray band indicates the total background uncertainty after performing the fit, while the blue histogram refers to the total signal extracted from the fit.
An unbiased mass distribution representative of the fit result in the VBF category is obtained by weighting both simulated and data events from the VBF-SR and VBF-SB regions by the per-event S/(S+B) ratio, computed as a function of the mass-decorrelated DNN output, defined in Sec. 5.6.5, for events within ππ π = 125.38 GeVΒ±hwhm. The best-fit estimates for the nuisance parameters and signal strength are propagated to the ππ π distribution. Figure 5.33 shows the observed and predicted weighted ππ π distributions for events in the VBF-SB and VBF-SR regions, combining 2016, 2017, and 2018 data. The lower panel shows the residuals between the data and the post-fit background prediction, along with the post-fit uncertainty obtained from the background-only fit. The best-fit signal contribution with ππ» =125.38 GeV is indicated by the blue line. An excess is observed in the weighted data distribution that is consistent with the expected resonant mass distribution for the signal with ππ» near 125 GeV and compatible with the excess observed at high DNN score in Fig.5.32.
β1 10 1 10 102 103 104 105 106 107
Events
Data Zjj-EW
DY Top quark
Diboson
(13 TeV) 41.5 fb-1
CMS Post-fit VBF-SB 2017
= 125.38 GeV mH
0 0.5 1 1.5 2
Data/Bkg.
Pre-fit
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
VBF DNN output
0 0.5 1 1.5 2
Data/Bkg.
Post-fit
β2 10
β1 10 1 10 102 103 104 105 106 107 108
Events
Data Hβ¡¡
Zjj-EW DY
Top quark Diboson
VBF ggH
(13 TeV) 41.5 fb-1
CMS Post-fit VBF-SR 2017
= 125.38 GeV mH
0 0.5 1 1.5 2 2.5
Data/Bkg.
Pre-fit
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
VBF DNN output
0 0.5 1 1.5 2 2.5
Data/Bkg.
Post-fit
Figure 5.30: The observed DNN output distribution in the VBF-SB (left) and VBF- SR (right) regions compared to the post-fit background estimate for various SM processes. The predicted backgrounds are obtained from a combined signal-plus- background fit performed across analysis regions and eras. Distributions reported are related to the 2017 data-taking period. The description of ratio panels is the same as in Fig.5.29
β1 10 1 10 102 103 104 105 106 107
Events
Data Zjj-EW
DY Top quark
Diboson
(13 TeV) 59.8 fb-1
CMS Post-fit VBF-SB 2018
= 125.38 GeV mH
0 0.5 1 1.5 2
Data/Bkg.
Pre-fit
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
VBF DNN output
0 0.5 1 1.5 2
Data/Bkg.
Post-fit
β2 10
β1 10 1 10 102 103 104 105 106 107 108
Events
Data Hβ¡¡
Zjj-EW DY
Top quark Diboson
VBF ggH
(13 TeV) 59.8 fb-1
CMS Post-fit VBF-SR 2018
= 125.38 GeV mH
0 0.5 1 1.5 2 2.5
Data/Bkg.
Pre-fit
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
VBF DNN output
0 0.5 1 1.5 2 2.5
Data/Bkg.
Post-fit
Figure 5.31: The observed DNN output distribution in the VBF-SB (left) and VBF- SR (right) regions compared to the post-fit background estimate for various SM processes. The predicted backgrounds are obtained from a combined signal-plus- background fit performed across analysis regions and eras. Distributions reported are related to the 2018 data-taking period. The description of ratio panels is the same as in Fig.5.29.
β1
10 1 10 102
103
104
105
106
107
Events
Data Zjj-EW
DY Top quark
Diboson
(13 TeV) 137 fb-1
CMS
Post-fit VBF-SB Run2
= 125.38 GeV mH
0 2 4 6 8 10 12
VBF DNN bin
0 0.5 1 1.5 2
Data/Bkg.
β2
10
β1
10 1 10 102
103
104
105
106
107
108
Events
Data Hβ¡¡
Zjj-EW DY
Top quark Diboson
VBF ggH
(13 TeV) 137 fb-1
CMS
Post-fit VBF-SR Run2
= 125.38 GeV mH
0 2 4 6 8 10 12
VBF DNN bin
0 0.5 1 1.5 2 2.5
Data/Bkg.
Figure 5.32: The observed DNN output distribution in the VBF-SB (left) and VBF- SR (right) regions compared to the post-fit background estimate for various SM processes. The predicted backgrounds are obtained from a combined signal-plus- background fit performed across analysis regions and eras. Distributions reported are related to the full Run-2 data-taking period. The lower panel shows the ratio between data and the post-fit background prediction from the S+B fit. The best fit π» β π πsignal contribution forππ»= 125.38 GeV is indicated by the blue solid line, while the grey band indicates the total background uncertainty.