Search for dark matter produced in association with a Higgs boson decaying to bottom quarks in proton-proton collisions at $\sqrt{s}$ = 13 TeV

Using 101 fb$^{-1}$ of 13 TeV proton-proton collision data from the CMS detector, this study searches for dark matter produced in association with a Higgs boson decaying to bottom quarks, finding no evidence of new physics and setting 95% confidence level exclusion limits on baryonic-Z' and 2HDM+a models.

The Gist

The Big Picture: Hunting the Invisible Ghost

Imagine the universe is a giant, bustling party. We can see most of the guests (stars, planets, you, me), but we know there's a massive crowd of invisible guests we can't see. We call this Dark Matter. We know they are there because they have gravity—they pull on the visible guests—but they don't talk to us (they don't emit light or interact with our usual senses).

Scientists at CERN's Large Hadron Collider (LHC) are like detectives trying to catch these invisible guests. They smash protons together at incredible speeds to create a chaotic "party" where new particles might be born. If Dark Matter is created, it will zip right out of the detector without leaving a trace.

How do they know it's there?
They look for a "missing person" report. If they see a visible object (like a Higgs boson) flying off in one direction, but the math says there should be more momentum in the opposite direction, they know something invisible must have kicked it. This is called Missing Transverse Momentum.

The Specific Hunt: The "Higgs" Bouncer

This paper focuses on a specific scenario: looking for Dark Matter that is produced alongside a Higgs boson.

Think of the Higgs boson as a very famous, heavy celebrity at the party. Usually, when this celebrity is created, they decay (break apart) into two bottom quarks (which are like heavy, short-lived particles).

• The Goal: Find a Higgs boson that broke apart into two bottom quarks, while simultaneously seeing a huge "kick" of missing energy (the Dark Matter) flying the other way.
• The Challenge: The Higgs boson is hard to spot, and the "kick" from Dark Matter can be subtle.

The Strategy: Two Different Ways to Look

The scientists split their search into two categories, depending on how fast the Higgs boson is moving:

1. The "Resolved" Category (The Slow Walk):
   • Analogy: Imagine the Higgs boson is walking slowly. When it breaks apart, the two bottom quarks separate and walk in different directions.
   • The Method: The detectors look for two distinct, separate jets of particles (like seeing two separate people walking away from each other).
2. The "Merged" Category (The Sprint):
   • Analogy: Imagine the Higgs boson is sprinting at near light speed. When it breaks apart, the two bottom quarks are so close together they look like a single, blurry blob.
   • The Method: The detectors look for one giant, wide jet of particles. To find the Higgs inside this blob, they use a sophisticated "AI camera" (a deep neural network) that can see the internal structure of the blob and say, "Ah, this looks like two bottom quarks squished together!"

The Data: A Bigger Sample Size

This paper combines data from two time periods:

• 2016: A previous search (about 36 units of data).
• 2017 & 2018: New data (about 101 units of data).
• Total: They now have a massive dataset (138 units), which is like looking at a much larger crowd to find the invisible guests. They also improved their "AI camera" to spot the sprinting Higgs bosons much better than before.

The Results: No Ghosts Found (Yet)

After sifting through all this data, the scientists compared what they saw against what the "Standard Model" (our current best theory of physics) predicts.

• The Verdict: The data matched the predictions perfectly. There were no unexpected spikes or "ghosts" hiding in the crowd.
• What this means: They didn't find Dark Matter in this specific scenario. However, in science, "not finding it" is still a victory because it tells us where not to look next.

The "Exclusion Zones": Drawing the Map

Since they didn't find the particles, they drew a map of where the particles cannot be. They set limits on two specific theoretical models:

1. The "Baryonic-Z'" Model:

   • Imagine a heavy mediator particle (a Z' boson) that acts as a bridge between normal matter and Dark Matter.
   • The Result: If this Z' particle exists, it must be heavier than 2.25 TeV (a very heavy weight) if the Dark Matter is very light. If the Z' is lighter (around 1.25 TeV), the Dark Matter particle must be heavier than 550 GeV.
   • Analogy: They checked the "lightweight" section of the Z' shelf and found it empty. The Z' must be in the "heavyweight" section.

2. The "2HDM+a" Model:

   • This model suggests there are extra types of Higgs-like particles (heavy and light pseudoscalars).
   • The Result: They ruled out specific combinations of masses. For example, if the light particle is 350 GeV, the heavy one cannot be between 850 and 1300 GeV.
   • Analogy: They tried to fit specific puzzle pieces together and found that these particular shapes don't fit the picture of the universe we see.

Summary

The CMS collaboration used a massive amount of new data and smarter AI tools to hunt for Dark Matter appearing alongside a Higgs boson. They looked for it in two ways: when the Higgs is moving slowly (two separate pieces) and when it's moving fast (one merged blob).

The outcome: They found no evidence of Dark Matter. However, they successfully tightened the net, proving that if these specific types of Dark Matter exist, they must be heavier or have different properties than previously thought. The search continues, but the "no-go" zones on the map have grown larger.

Technical Summary

Technical Summary: Search for Dark Matter Produced in Association with a Higgs Boson Decaying to Bottom Quarks

Problem and Motivation
This paper presents a search for dark matter (DM) particles produced in association with a Standard Model (SM) Higgs boson ($h$) decaying to a bottom quark-antiquark pair ($b\bar{b}$) in proton-proton collisions at a center-of-mass energy of $\sqrt{s} = 13$ TeV. While astrophysical evidence strongly supports the existence of DM, its particle nature and potential interactions with SM particles remain unknown. The "mono-Higgs" channel offers a promising signature for probing DM, particularly in scenarios where the DM-SM coupling is enhanced or mediated exclusively through the Higgs sector. The analysis targets the final state characterized by large missing transverse momentum ($p^{\text{miss}}_T$), recoiling against a Higgs boson candidate.

Methodology
The analysis utilizes data collected by the CMS detector during 2017 and 2018, corresponding to an integrated luminosity of 101 fb$^{-1}$. To cover a wide range of Higgs boson transverse momenta ($p^h_T$), the event selection is divided into two exclusive categories:

1. Merged Category: Targets Lorentz-boosted Higgs bosons where the $b\bar{b}$ decay products are collimated and reconstructed as a single large-radius (AK8) jet.
2. Resolved Category: Targets lower to moderate $p^h_T$ where the $b$-quarks are reconstructed as two separate small-radius (AK4) jets.

Key Technical Features:

• Object Identification: The analysis employs advanced deep learning algorithms. A deep neural network (DNN) based tagger, PARTICLENET-MD, is used to identify boosted $h \to b\bar{b}$ decays in the merged category, replacing previous N-subjettiness criteria. In the resolved category, the DEEPJET algorithm identifies $b$-tagged AK4 jets.
• Signal Region (SR): Events are selected requiring high $p^{\text{miss}}_T$ (>250 GeV for merged, >200 GeV for resolved), a Higgs candidate mass within 70–160 GeV, and the absence of isolated leptons to suppress $t\bar{t}$ and $W$+jets backgrounds.
• Background Estimation: The dominant backgrounds, $Z(\nu\nu)$+jets and $t\bar{t}$, are estimated using data-driven methods. Dedicated control regions (CRs) enriched in single-lepton ($t(\ell)$) and dilepton ($Z(\ell\ell)$) events are used to constrain the normalization and shape of these backgrounds. The hadronic recoil ($U$) in CRs serves as a proxy for $p^{\text{miss}}_T$ in SRs.
• Statistical Analysis: A simultaneous binned maximum likelihood fit is performed across the SRs and CRs in both categories. The fit extracts the signal strength modifier ($\mu$) while profiling nuisance parameters associated with experimental and theoretical systematic uncertainties.

Key Contributions and Improvements
This work represents a significant advancement over previous CMS searches in the same channel (specifically Ref. [19], based on 2016 data) through:

• Increased Statistics: The dataset size has nearly tripled (101 fb$^{-1}$ vs. 35.9 fb$^{-1}$), and when combined with the 2016 data, the total integrated luminosity reaches 138 fb$^{-1}$.
• Enhanced Sensitivity: The adoption of the PARTICLENET-MD tagger for boosted topologies has improved the identification efficiency of merged $h \to b\bar{b}$ signals compared to previous N-subjettiness-based selections.
• Comprehensive Combination: The results are statistically combined with the 2016 analysis, providing a unified constraint on the parameter space.

Results
The observed data show good agreement with the SM background predictions across all signal regions and control regions. No significant excess of events is observed. Consequently, upper limits at the 95% confidence level (CL) are set on the production cross section for DM associated with a Higgs boson. The results are interpreted within two simplified benchmark models:

1. Baryonic-$Z'$ Model: A vector mediator ($Z'$) couples DM to SM quarks.

   • For a DM candidate mass ($m_\chi$) of 1 GeV, mediator masses ($m_{Z'}$) below 2.25 TeV are excluded (expected limit: 2.35 TeV).
   • For a mediator mass of 1.25 TeV, DM masses up to 550 GeV are excluded (expected limit: 600 GeV).
   • This represents an improvement of 650 GeV in the excluded $m_{Z'}$ range compared to the 2016 result.

2. 2HDM+a Model: A Type-II Two-Higgs-Doublet Model extended with a light pseudoscalar ($a$) and a heavy pseudoscalar ($A$).

   • For a heavy pseudoscalar mass ($m_A$) of 1000 GeV, light pseudoscalar masses ($m_a$) below 360 GeV are excluded.
   • For a light pseudoscalar mass of 350 GeV, heavy pseudoscalar masses between 850 and 1300 GeV are excluded.
   • Values of the mixing angle $\sin \theta$ between 0.15 and 0.95 are excluded, and $\tan \beta$ values below 4.2 are excluded for specific benchmark points.

Significance
The paper claims that these results improve upon existing CMS limits primarily due to the larger integrated luminosity and the enhanced identification of $h \to b\bar{b}$ decays using modern machine learning techniques. The analysis demonstrates the effectiveness of combining merged and resolved topologies to probe a broad kinematic range of DM-Higgs interactions. The absence of a signal places stringent constraints on simplified models of DM production involving a Higgs portal, specifically ruling out significant portions of the parameter space for baryonic-$Z'$ and 2HDM+a scenarios. The results are consistent with the Standard Model background, providing no evidence for new physics in this specific channel.