Search for heavy resonances decaying into two Higgs bosons in the b b ¯ τ + τ − \mathrm{b}\bar{\mathrm{b}}\,\tau^{+}\tau^{-} final state in proton-proton collisions at s = 13 TeV \sqrt{s}=13~\text{TeV} with the CMS detector
\firstname Ganesh \lastname Parida on behalf of the CMS collaboration \fnsep · 2026
Read Search for heavy resonances decaying into two Higgs bosons in the b b ¯ τ + τ − \mathrm{b}\bar{\mathrm{b}}\,\tau^{+}\tau^{-} final state in proton-proton collisions at s = 13 TeV \sqrt{s}=13~\text{TeV} with the CMS detector by \firstname Ganesh \lastname Parida on behalf of the CMS collaboration \fnsep (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-23502.txt.
Introduction
Parida searches CMS collision data for heavy resonances that decay into pairs of Higgs bosons, with one Higgs giving bottom quarks and the other tau leptons. The observed events match standard background expectations, so no new particle appears. The analysis instead sets the tightest LHC limits yet on this decay channel across most of the probed mass range.
Catalog record
| CREATOR | \firstname Ganesh \lastname Parida on behalf of the CMS collaboration \fnsep |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 10781 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.23502
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Search for heavy resonances decaying into two Higgs bosons in the b b ¯ τ + τ − \mathrm{b}\bar{\mathrm{b}}\,\tau^{+}\tau^{-} final state in proton-proton collisions at s = 13 TeV \sqrt{s}=13~\text{TeV} with the CMS detector \firstname Ganesh \lastname Parida on behalf of the CMS collaboration \fnsep Abstract A search is presented for massive narrow-width resonances in the mass range of 1 - 4.5 TeV 1\text{-}4.5\,\text{TeV} decaying into pairs of Higgs bosons (HH), using proton-proton collision data at a center-of-mass energy of 13 TeV 13\,\text{TeV} collected with the CMS detector at the LHC during the 2016 - 2018 2016\text{-}2018 data-taking. The data correspond to an integrated luminosity of 138 fb − 1 138~\mathrm{fb}^{-1} . The analysis targets final states where one Higgs boson decays into a pair of bottom quarks and the other into a pair of tau leptons, X → HH → b b ¯ τ + τ − \mathrm{X}\rightarrow\mathrm{HH}\rightarrow\mathrm{b}\bar{\mathrm{b}}\,\tau^{+}\tau^{-} . The observed data are found to be consistent with standard model background expectations. Upper limits at 95 % 95\% confidence level (CL) are set on the production cross section for resonant HH production for masses between 1 1 and 4.5 TeV 4.5\,\text{TeV} . This analysis sets the most sensitive LHC limits to date on X → HH → b b ¯ τ + τ − \mathrm{X}\rightarrow\mathrm{HH}\rightarrow\mathrm{b}\bar{\mathrm{b}}\,\tau^{+}\tau^{-} decays in the mass range of 1.4 1.4 to 4.5 TeV 4.5\,\text{TeV} . 1 Introduction The discovery of the Higgs boson by the ATLAS ATLAS:2013xga and CMS CMS:2012qbp experiments established the standard model (SM) mechanism of electroweak symmetry breaking, yet several open questions motivate new physics at the TeV scale. Many beyond the standard model (BSM) scenarios predict heavy resonances (X) decaying to Higgs boson pairs (HH), such as, warped extra-dimension models proposed by Randall and Sundrum randall1999large , both spin-0 scalars goldberger1999modulus and spin-2 Kaluza–Klein gravitons davoudiasl2000phenomenology . These processes, if they exist, can significantly enhance the HH production rate. This analysis adopts a unified, model-independent strategy for both spin-0 and spin-2 resonances, targeting \mathrm{X}\to\mathrm{HH}\to\mathrm{b\bar{b}}\,\mathrm{\tau\tau} in the mass range 1–4.5 TeV using the full Run 2 CMS dataset (2016–2018). Previous CMS searches in this final state, based on partial Run 2 data (35.9 \mathrm{fb^{-1}} ) 2016cmsheavybbtautau , observed no significant deviations from SM expectations. The corresponding full Run 2 ATLAS searches aad2020reconstruction ; Run2atlasheavybbtautau also reported results consistent with the hypothesis of background-only. 2 Signal topology and physics object reconstruction and identification This analysis assumes a narrow resonance width, i.e., the natural width of the resonance is significantly smaller than the experimental resolution. In the mass range considered, the Higgs bosons are frequently produced with large transverse momenta (Lorentz boosted regime). In the \text{H}\to\text{b}\bar{\text{b}} decay, the two bottom quarks are then emitted with a small angular separation and, after hadronization, often merge into a single large-radius jet ( R=0.8 ). These AK8 jets are identified with the ParticleNet jet tagger qu2020jet . The mass regression ( M_{\mathrm{H(b\bar{b})}} ) branch of ParticleNet is also used in the definition of the signal region (SR) and sidebands (SB), as described later in Sec. 3 . Simillarly, the decay products of the \tau -lepton pair, whether fully hadronic ( \tau_{h}\tau_{h} ) or semileptonic ( \ell\tau_{h} ), also tend to overlap spatially, making di- \tau reconstruction particularly challenging. The reconstruction of such Lorentz-boosted \tau leptons was first addressed in a similar CMS search using 8~\mathrm{TeV} proton–proton collision data CMS:2015fjg . Higgs bosons decaying to one or two hadronically decaying \tau leptons are reconstructed using a dedicated boosted- \tau algorithm that considers subjets of the jets clustered with the Cambridge–Aachen algorithm and applies a mass-drop criterion to seed the standard hadron-plus-strips (HPS) procedure. A key element of this analysis is the new BoostedDeepTau CMS-DP-2025-047 convolutional-neural-network tagger, developed specifically to identify highly Lorentz-boosted \tau leptons. Built upon the DeepTau CMS:2022prd architecture, it incorporates both low-level detector information and a set of 42 high-level variables describing kinematics, isolation, and lifetime features. Trained on simulated Run 2 samples of signal and major backgrounds, BoostedDeepTau improves background rejection by a factor 2–4 at p_{\mathrm{T}}<100~\mathrm{GeV} and by more than an order of magnitude at higher momenta. Performance summaries are shown in Figs. 1 and 2 . In addition, this analysis also considers AK4-seeded \tau candidates identified with the DeepTau algorithm. The SM processes with the largest contributions to the background are \mathrm{t\bar{t}} (top quark–antiquark production), \mathrm{W+jets} , and \mathrm{Z+jets} production, although the precise composition depends on the specific channel under consideration. Figure 1: BoostedDeepTau shows a discrimination power that is a factor of 2–4 ( >10 ) better than the MVA Iso discriminator evaluated on individual reconstructed \tau_{\mathrm{h}} from a boosted ditau system for p_{\mathrm{T}}^{\tau_{\mathrm{h}}}<100~\text{GeV} ( >100~\text{GeV} ). Here, p_{\mathrm{T}}^{\tau_{\mathrm{h}}} refers to the transverse momentum of the individual reconstructed \tau_{\mathrm{h}} from a boosted ditau system. The Boosted DeepTau ID efficiency is estimated from a spin-0 \mathrm{X\rightarrow HH\rightarrow bb\tau\tau} MC sample (with M_{X}=3.5~\text{TeV} ). The jet misidentification probability is estimated from \mathrm{t}\bar{\mathrm{t}} +jets MC samples. Figure 2: H \rightarrow\tau\tau reconstruction and identification efficiency as a function of generator–level Higgs boson p_{\mathrm{T}} is evaluated using spin-0 \mathrm{X\rightarrow HH\rightarrow bb\tau\tau} MC samples (mass of X=1.6~\mathrm{TeV} ). The reconstructed \tau_{\mathrm{h}} are required to have p_{\mathrm{T}}>20~\mathrm{GeV} , |\eta|<2.5 , and to pass the “Loose” working point of the respective identification algorithms, which corresponds to a per- \tau_{\mathrm{h}} signal efficiency exceeding 95%. 3 Analysis design The vector \,\vec{p}_{\mathrm{T}}^{\mathrm{miss}} is defined as the negative of the vector sum of the transverse momenta of all reconstructed particles associated with the primary vertex, and p_{\mathrm{T}}^{\mathrm{miss}} is its magnitude. Events are selected using missing transverse momentum triggers with thresholds above 110~\mathrm{GeV} , together with an offline requirement of p_{\mathrm{T}}^{\mathrm{miss}}>180~\mathrm{GeV} to ensure high signal efficiency. Each event must contain one Higgs boson candidate decaying to \ell\tau or \tau\tau , and a second Higgs boson reconstructed as a large radius jet ( R=0.8 ). The four-momentum of the di- \tau system is reconstructed with the FastMTT Matyszkiewicz:2025gal algorithm, which uses a dynamical-likelihood approach with a collinear approximation. The resonance mass is defined as the invariant mass of this di- \tau system and the Higgs-jet candidate, and only events with M_{\mathrm{X}}>750~\mathrm{GeV} are kept. The overall selection efficiency for simulated spin-0 resonances ranges from 2–14% in the \tau\tau channel and 4–11% in the \ell\tau channel. Figure 3: Post-fit reconstructed mass distribution of resonance X in the SR (left) and SB (right) after applying all selection criteria, for the sum of the \tau_{\mathrm{h}}\tau_{\mathrm{h}} and \ell\tau_{\mathrm{h}} channels. Minor background contributions are grouped into a single category labeled “others”. The lower panels show the “Pull” defined as (Data - Prediction)/Total uncertainty . Following the full event selection, \mathrm{t\bar{t}} becomes the dominant background in both channels. To improve the modeling of \mathrm{t\bar{t}} , an unconstrained bin-by-bin multiplicative parameter is introduced for each bin of the reconstructed resonance mass ( M_{X} ) distribution. These parameters adjust the \mathrm{t\bar{t}} yield independently in every ( M_{X} ) bin, simultaneously in the SR and the SB, thereby constraining both the normalization and shape of \mathrm{t\bar{t}} using data in the SB for a given channel. A simultaneous binned likelihood fit is performed for signal extraction. Systematic uncertainties affecting