Search for invisible Higgs boson decays in vector boson fusion at $\sqrt{s}=13$ TeV with the ATLAS detector

Phys. Lett. B 793 (2019) 499

18 September 2018

Contact: ATLAS Exotics conveners
ContentPreview
e-print arXiv:1809.06682, Physics Briefing - internal pdf from arXiv
Inspire record -
Figures Tables Auxiliary Material -
Abstract
We report a search for Higgs bosons that are produced via vector boson fusion and subsequently decay into invisible particles. The experimental signature is two energetic jets with $\mathcal{O}(1)$ TeV invariant mass and $\mathcal{O}(100)$ GeV missing transverse momentum. The analysis uses 36.1 fb$^{-1}$ of $pp$ collision data at 13 TeV recorded by the ATLAS detector at the LHC. In the signal region the 2252 observed events are consistent with the background estimation. Assuming a 125 GeV particle with Standard Model cross sections, the upper limit on the branching fraction of the Higgs boson decay into invisible particles is 0.37 at 95% confidence level where 0.28 was expected. This limit is interpreted in Higgs portal models to set limits on the WIMP-nucleon scattering cross section. We also consider invisible decays of additional scalar bosons with masses up to 3 TeV for which the upper limits on the cross section times branching fraction are in the range of $0.3-2.0$ pb.
Figures
Figure 01:
Data-to-MC yield comparisons in the 27 subsamples used in the statistical fit. The observed data N (dots) are superimposed on the prefit backgrounds B (stacked histogram with shaded systematic uncertainty bands). The hypothetical signal S (empty blue histogram) is shown on top of B for ÔC;inv = 1. The bottom panels show the ratios of N (dots) and B+S (blue line) to B with the systematic uncertainty band shown on the line at 1. The 1, 2, and 3 bin label corresponds to the three mjj bins with [1, 1.5, 2, -] TeV boundaries, respectively. The "e fakes" refers to SMET < 4 √GeV selection and is determined by the fit, so postfit values are shown for the purposes of illustration. The diboson contribution is included in the electroweak (EW) W and Z bosons.

png (40kB)  pdf (24kB) 
Figure 02:
Distribution of event yields in the Z (top) and W (bottom) control regions. The postfit normalizations for mjj (left) and ETmiss (right) are summed over the subsamples. The distributions start at 180 GeV as indicated. The observed data N (dots) are superimposed on the sum of the backgrounds B (stacked histogram with shaded systematic uncertainty bands). The breakdown of the B is given in the lower left box in each panel. The bottom panels show the ratios of N to B with the systematic uncertainty band shown on the line at 1. The "other," as listed in Table 1 contribute a few events at low values of mjj and ETmiss, and are omitted. The last bin in each plot contains the overflow.

png (33kB)  pdf (71kB) 
Figure 03:
Distribution of event yields in the multijet validation region for mjj (left) and ETmiss (right). The mjj plot shows the 100 < ETmiss < 120 GeV subset of the right plot as indicated by the arrow. The N observed data (dots) are superimposed on the sum of the B backgrounds (stacked histogram). The systematic uncertainty band applies only to the multijet component. The statistical uncertainties are relatively large because of the weighting of the trigger samples with large prescale values. See the caption of Fig. 2 for other plotting details.

png (29kB)  pdf (21kB) 
Figure 04:
Contributions to the relative uncertainty in the transfer factors αZ (left) and αW (right) in the three mjj bins of the SR. The theoretical uncertainties from the sources noted in the legend are combined in quadrature.

png (9kB)  pdf (14kB) 
Figure 05:
Distribution of event yields in the signal region for mjj (left) and ETmiss (right). The ETmiss distributions start at 180 GeV and shows the most sensitive mjj > 2 TeV subset of the SR as indicated by the arrow. The postfit normalizations for mjj (ETmiss) distributions use separate background, B, normalizations in the mjj range of [1, 1.5, 2, -] ([2, -]) TeV, and sum the contributions from W and Z bosons (electroweak and strong production modes). The hypothetical signal S (empty blue histogram) is shown on top of B for ℬinv = 1. The bottom panels show the ratios of N (dots) and B+S (blue line) to B with the systematic uncertainty band shown on the line at 1. The bin width in the mjj plots (ETmiss) is 500 GeV (50 GeV except for the first bin with the non-zero entry, which is 20 GeV). See the caption of Fig. 2 for other plotting details.

png (34kB)  pdf (25kB) 
Figure 06a:
Upper limits on (a) the spin-independent WIMP-nucleon cross section using Higgs portal interpretations of ℬinv at 90% CL vs. mWIMP and (b) the VBF cross section times the branching fraction to invisible decays at 95% CL vs. mscalar. The top plot shows results from Ref. [77, 78, 79].

png (37kB)  pdf (25kB) 
Figure 06b:
Upper limits on (a) the spin-independent WIMP-nucleon cross section using Higgs portal interpretations of ℬinv at 90% CL vs. mWIMP and (b) the VBF cross section times the branching fraction to invisible decays at 95% CL vs. mscalar. The top plot shows results from Ref. [77, 78, 79].

png (11kB)  pdf (14kB) 
Tables
Table 01:
Event yields in the signal region (SR) and control regions (CR) summed over lepton charge and flavor. The yields are the prefit values for mjj>1 TeV. The observed data (N), the background estimate (B), and the signal (S for mH = 125 GeV with Binv = 1) are given. The B and S values for individual processes are rounded to a precision commensurate with the sampling uncertainty associated with the finite MC sample size. For all processes the fractions of electroweak production [EW] are given. "Other" is defined in the text.

png (4kB)  pdf (294kB) 
Table 02:
Sources of uncertainty. The first set shows Δ, the relative improvement of the 95% CL upper limit on Binv when the source of uncertainty is "removed" by fixing it to its best-fit value. Combined rows are not simple sums of the rows above because of the Δ metric; the symbols (†, ‡, ★) are parenthetically defined in the table. The column labeled "visual" shows bars whose lengths from the center tick are proportional to Δ. The second set of columns shows the effect on the yields and the α transfer factors; values in the 1 < mjj < 1.5 TeV bin are shown. The yields are for the signal process in the SR (S), Z MC in the SR (BZSR), and Z MC in the CR (BZCR). The α is given to demonstrate the reduction in the uncertainty in the ratio BZSR/BZCR. The individual yields for the W are not shown because the cancellation effects are similar to the Z counterparts. The abbreviations for the theoretical sources are described in the text. The '-' indicates that the quantity is not applicable. The penultimate (last) row shows the summary impact of removing the systematic uncertainties due to the experimental and theoretical sources (as well as statistical uncertainties of the MC samples).

png (16kB)  pdf (253kB) 
Auxiliary figures and tables
Figure 01:
Example diagrams for the signal (top) and the leading background processes (bottom).

png (8kB)  pdf (38kB) 
Figure 02:
Data-to-MC yield comparisons in the 27 subsamples used in the statistical fit. The observed data N (dots) are superimposed on the postfit backgrounds B (stacked histogram with shaded systematic uncertainty bands). The hypothetical signal S (empty blue histogram) is shown on top of B for ℬinv = 1. The bottom panels show the ratios of N (dots) and B+S (blue line) to B with the systematic uncertainty band shown on the line at 1. The 1, 2, and 3 bin label corresponds to the three mjj bins with [1, 1.5, 2, -] TeV boundaries, respectively. The "e fakes" refers to SMET < 4 √GeV selection and is determined by the fit, so postfit values are shown for the purposes of illustration. The diboson contribution is included in the electroweak (EW) W and Z bosons.

png (36kB)  pdf (23kB) 
Figure 03:
Upper limits on the spin-independent WIMP-nucleon cross section vs. mWIMP The plot shows the Higgs portal interpretations of ℬinv at 90% CL with other results [77, 78, 79, 1611.01499, 1711.07692].

png (38kB)  pdf (26kB) 
Figure 04a:
Event display of a candidate in the signal region. The image on the top (bottom) shows calorimeter cell energy deposits for ET > 0.25 (1) GeV and charged tracks for pT > 1 (2) GeV. For each image, the left (right) shows a longitudinal (transverse x-y) view. In the transverse view, the concentric circles on the right correspond to the perspective view of the TRT subdetector system shown in the longitudinal view; the leading (subleading) jet j1 (j2) is coming out of (going into) the page. The two jets shown have the properties (pT,j1 = 450 GeV, (pT,j2 = 110 GeV, ηj1 = -1.8, ηj2 = 3.7) that result in the dijet properties (mjj = 3.6 TeV, Δηjj = 5.5, Δφjj = 0.1) and ⃗ETmiss quantities (ETmiss = 564 GeV, = 2.8). During the run in which this event was recorded, the peak averaged number of pp interactions per event was around 17.

png (601kB)  pdf (1MB) 
Figure 04b:
Event display of a candidate in the signal region. The image on the top (bottom) shows calorimeter cell energy deposits for ET > 0.25 (1) GeV and charged tracks for pT > 1 (2) GeV. For each image, the left (right) shows a longitudinal (transverse x-y) view. In the transverse view, the concentric circles on the right correspond to the perspective view of the TRT subdetector system shown in the longitudinal view; the leading (subleading) jet j1 (j2) is coming out of (going into) the page. The two jets shown have the properties (pT,j1 = 450 GeV, (pT,j2 = 110 GeV, ηj1 = -1.8, ηj2 = 3.7) that result in the dijet properties (mjj = 3.6 TeV, Δηjj = 5.5, Δφjj = 0.1) and ⃗ETmiss quantities (ETmiss = 564 GeV, = 2.8). During the run in which this event was recorded, the peak averaged number of pp interactions per event was around 17.

png (424kB)  pdf (927kB) 
Figure 05:
Event display of a candidate in the signal region with the highest value of mjj in the selected sample. This image shows calorimeter cell energy deposits for ET > 1 GeV and charged tracks for pT > 2 GeV. The left (right) shows a longitudinal (transverse x-y) view. In the transverse view, the leading (subleading) jet j1 (j2) is going into (coming out of) the page. The two jets shown have the properties (pT,j1 = 520 GeV, pT,j2 = 190 GeV, ηj1 = 2.4, ηj2 = -3.1) that result in the dijet properties (mjj = 5.0 TeV, Δηjj = 5.5, Δφjj = 0.7) and ⃗ETmiss quantities (ETmiss = 674 GeV, φ = 4.4). During the run in which this event was recorded, the peak averaged number of pp interactions per event was around 36.

png (359kB)  pdf (878kB) 
Table 01:
Number of VBF invisible signal MC events SVBF normalized to the Higgs cross section and 36 fb-1 at 13 TeV for mH = 125 GeV. The SVBF are rounded to a precision commensurate with the sampling uncertainty associated with the finite MC sample size. The selections are applied sequentially. The selection efficiency is with respect to the selection on the previous line.

png (4kB)  pdf (189kB) 
Table 02:
Plot content of Fig. 1 of the paper, which is prefit The B values are given to the nearest digit. The "rest" in this table is "other" defined in the text without the dibosons.

png (177kB)  pdf (102kB) 
Table 03:
Plot content of Fig. 2. The B values are given to the nearest digit or the nearest tenth, depending on the entry. The bin numbers i correspond to the bins in the plot starting with the first non-zero entry. The i does not refer to the SR mjj bins for the statistical fit. The first bin for the ETmiss is 20 GeV wide and the rest are 50 GeV wide.

png (8kB)  pdf (105kB) 
Table 04:
Plot content of Fig. 5. The B and S values are given to the nearest digit or the nearest tenth, depending on the entry. The bin numbers i correspond to the bins in the plot starting with the first non-zero entry. The i does not refer to the SR mjj bins for the statistical fit. The first bin for the ETmiss is 20 GeV wide and the rest are 50 GeV wide.

png (5kB)  pdf (118kB) 
Table 05:
Plot content of Fig. 6(b). Numbers are given to two decimal places.

png (4kB)  pdf (118kB) 
Table 06:
Analysis updates. Notable changes to the analysis with respect to Ref. [28] are listed.

png (9kB)  pdf (182kB) 
Table 07:
Plot content of Fig. 2 of the Auxiliary figures and tables (online), which is postfit. The B values are given to the nearest digit. The "rest" in this table is "other" defined in the text without the dibosons.

png (175kB)  pdf (102kB) 

© 2018 CERN for the benefit of the ATLAS Collaboration.
Reproduction of the article, figures and tables on this page is allowed as specified in the CC-BY-4.0 license.

2024-03-29 00:39:18