This paper studies whether a specific collider signature can tell different kinds of dark matter apart. The authors focus on vector boson fusion (VBF) events at the Large Hadron Collider (LHC) that produce two forward jets plus large missing transverse energy (MET). They show, as a proof of principle, that the shapes of the jet and angular distributions in these events differ between Higgs‑portal dark matter and neutralino dark matter. Those differences can be used to distinguish the two hypotheses at the High‑Luminosity LHC (HL‑LHC).

The team simulated VBF production of dark matter for two concrete model classes. In the Higgs‑portal case the dark matter couples to ordinary matter through the Higgs boson; they consider both a fermion+scalar mediator setup and a scalar dark matter benchmark with mass 130 GeV and strong portal couplings (examples: fermion case with mχ = 130 GeV, heavy scalar mass mh2 = 275 GeV, cosα = 0.95, coupling λ = 3, and a scalar case with mS = 130 GeV and λHS = 3). In the neutralino case they use the Minimal Supersymmetric Standard Model (MSSM) with a compressed spectrum of electroweakinos, so charginos decay to the neutralino plus very soft particles and all electroweakino channels appear experimentally as the same two‑jet plus MET final state. They follow the HL‑LHC invisible Higgs search strategy and apply linear discriminant analysis plus a Kolmogorov–Smirnov statistical test to tell the signals apart.

At a qualitative level the discrimination comes from the polarization of the exchanged weak bosons (W and Z) in VBF. That polarization affects how much transverse momentum the tagged forward jets carry. For the Higgs‑portal scenarios the tagged jets tend to be less energetic in the transverse direction than for the neutralino scenarios. Two angular variables also behave differently: Δη, the separation of the two jets along the beam axis (difference in pseudorapidity), and Δφ, the angle between the jets around the beam axis (difference in azimuth). The authors report that these kinematic differences provide non‑trivial power to separate the models.