Searching for new physics with the Higgs boson and dark matter
There are a number of shortcomings of the Standard Model, indicating the existence of new 'Beyond Standard Model' physics, including the lack of a satisfactory dark matter candidate.
The discovery of the Higgs boson gives us a powerful tool to search for new physics, dark matter, additional Higgs bosons and unexpectedly frequent Higgs boson pair/triple ("di-Higgs"/"tri-Higgs") production which would imply new phenomena.
We use the ATLAS collision dataset from LHC Run-2 and Run-3 to search for new particles and phenomena. These powerful datasets are very sensitive, opening up new channels and signatures for discovery.
Di-Higgs and Tri-Higgs
Last year the group published the first ever limits on triple Higgs production. The search was for 3 Higgs bosons all decaying into b-quarks (or HHH->6b). A four-Higgs coupling is predicted in the SM and we set the first ever constraints on this, but it has a very low cross section, so the search also set limits on heavy new Higgs-like particles that could act as intermediary states and could occur with higher cross sections.
Our explorations of Higgs pair production have focussed on the '4b' channel, in which both Higgs bosons decay to a pair of b-quarks.
Alongside the triple Higgs project, recent student thesis have centred around searches for new particles decaying to two Higgs bosons in the resolved, and higher-momentum 'boosted' channel; the first use of the vector boson fusion (VBF) process to constrain the non-resonant Higgs pair production; and the full Run-2 result and first exploration of effective field theory in this signature.
The group is about to embark on another search for HHH (triple Higgs) with others at CERN which will include all the data that will be collected at the LHC prior to the long shut-down. This search would go beyond our previous result by considering more new physics models, boosted signatures to reach heavier potential new particles, consider more Higgs decay modes, and more sophisticated ML analysis approaches. Our newest student is working on this search but there is potential for future students to join and expand the scope.
The group is also involved in feasibility studies for the FCC-hh project, looking at how well we'd be able to constrain or even measure the Higgs quartic self-coupling at this O(100 TeV) collider.
Dark Matter
There are a wide variety of well-motivated models for dark matter and how it might interact with the Standard Model particles we know well. Weakly interacting dark matter will be invisible to the detector, and its presence must be inferred from momentum imbalance ("missing transverse momentum", or "MET").
One elegant means for dark matter to interact is through a Higgs boson, or new Higgs-like particles, which are predicted in many well-motivated extended Higgs-sector models that predict dark matter. We have been exploring this possibility through searches for 1 or 2 Higgs(-like) particles being produced alongside dark matter.
Our recent students produced the first ever search for dark matter with a resonant di-Higgs + MET signature, constraining the 'dark higgs' model with the Run-2 dataset. Our current student is working on upgrading this for the full Run-2+3 datasets using smarter ML-driven methods, broader interpretations, and potential expansion into non-resonant topologies sensitive to other models like the 2HDM+a and supersymmetry. An incoming student could join this search and expand it's scope further to consider boosted signatures and low-MET signatures to target particularly heavy or light dark matter particles/mediators. Alternatively an incoming student could search for single Higgs/scalar + MET, or ZZ/WW + MET signatures which are also very well motivated and would give sensitivity to a complementary range of models.
Machine Learning, Missing Transverse momentum & b-tagging
Our signatures involve invisible particles and Higgs boson decays to a b-quark jet pair, and the quality of our results depends strongly on how well we can reconstruct these from detector outputs.
We have historically worked on enhancing our ability to identify b-quark jets in the detector, and currently work on improving ATLAS's ability to determine the amount of missing energy in our events and trigger on it. We are designing ATLAS's first full MET calibration and reconstruction to use in our current dark matter searches, and new ML-based MET predictions to allow us to trigger on dark-matter-like signals at HL-LHC. Future students could continue these kinds of projects to improve our current/future results ATLAS-wide.
There are also large backgrounds to our signal processes, and so advanced methods to reduce them are a priority. We employ a wide range of machine learning techniques, such as adversarial and graph neural networks, anomaly detection, deep sets, transformers and boosted decision trees to help us identify our signals.
Doctoral thesis topics are available searching for tri-Higgs production or dark matter production alongside Higgs-like particles or vector bosons.
Contact: Todd Huffman and Holly Pacey. We also collaborate strongly with James Frost, who is now a visitor to the group based at RAL.
Current Members
Former Members
Cigdem Issever
Bill Balunas
Lydia Beresford
Nufikri bin Norjoharuddeen
Beojan Stanislaus
Santiago Parades
Migle Stankaityte
James Grundy
Federico Celli
Iza Veliscek
Tim Bruckler
Maggie Chen
Tom Dingley