Recursive Jigsaw Reconstruction (RJR) Technique

• Recursive Jigsaw Reconstruction (RJR) is a hypothesis-driven technique that reconstructs complex event topologies with invisible particles using user-defined decay trees and kinematic constraints.
• It employs modular 'jigsaw rules' to resolve combinatorial and kinematic ambiguities, enhancing the accuracy of decay chain reconstruction in challenging high energy physics scenarios.
• RJR has demonstrated improved signal sensitivity and background rejection in LHC analyses, notably in compressed supersymmetry and exotic multi-body decay events.

Recursive Jigsaw Reconstruction (RJR) is a systematic, hypothesis-driven event analysis technique for high energy physics that enables the reconstruction of complex topology events in the presence of both invisible particles and combinatorial ambiguities. RJR employs user-defined decay trees and a flexible set of kinematic rules—"jigsaw rules"—to recursively decompose events into rest frames and resolve unknowns such as invisible particle momenta and object assignments. Originally developed to optimize searches for new physics scenarios with compressed spectra and ambiguous final states, RJR has become a central methodology in LHC analyses, particularly for difficult scenarios involving multiple invisibles, initial-state radiation (ISR), and combinatoric partitions among indistinguishable objects (Jackson et al., 2017, Santoni, 2017, Jackson et al., 2016, Desai et al., 27 Jan 2026).

1. Core Principles and Decay Tree Construction

RJR is anchored in the concept of a hypothesis-motivated "decay tree" that represents the underlying event topology and its successive decays to measured "visible" (V) and unmeasured "invisible" (I) objects. Each decay node defines a unique parent rest frame. The kinematics of the entire event are parameterized using measured four-vectors in the laboratory frame, and the required set of missing degrees of freedom: unknown boosts between different rest frames and invisible momenta, as well as combinatoric ambiguities in object assignment to parent branches (Jackson et al., 2017).

The procedure begins by assigning reconstructed objects (leptons, jets) and the total missing transverse momentum vector to the leaves of the decay tree. For compressed supersymmetric or multi-body decay topologies, the tree typically includes:

• The lab frame as the root;
• Center-of-mass (CM/PP) frame of the full signal + ISR system;
• ISR and signal ("S") systems recoiling against each other in the CM frame;
• Signal sub-trees, each further splitting into visible and invisible arms at each decay level (e.g., parent → ℓ + ν + χ⁰).

RJR's central strategy is to solve for unknowns via local decisions at each split using kinematic constraints ("jigsaw rules") that are modular and context-dependent, rather than attempting a global reconstruction (Jackson et al., 2017, Santoni, 2017).

2. Jigsaw Rules: Kinematic and Combinatoric Resolution

Jigsaw rules ("JR"s) are recursive, interchangeable algorithms that resolve kinematic and combinatoric ambiguities at each node of the decay tree (Jackson et al., 2017, Desai et al., 27 Jan 2026). Key classes include:

• Invisible Rapidity Matching: Sets the unknown longitudinal momentum components of an invisible system by matching its rapidity to that of an associated visible system, or by minimizing invariant masses (Jackson et al., 2017).
• Contra-boost Invariant Split: Divides an invisible system between daughter subsystems by imposing constraints such as equal reconstructed parent masses:

$$M_{A}^2 = (p_{V_A} + p_{I_A})^2 = M_{B}^2 = (p_{V_B} + p_{I_B})^2$$

This yields analytic solutions for the invisible momenta consistent with event kinematics (Santoni, 2017, Desai et al., 27 Jan 2026).

• Combinatoric Metric Minimization: Selects the best object assignment among all partitions using metrics such as minimizing the sum of squared invariant masses of candidate decays (Jackson et al., 2017).
• N-body Minimization: For cases with multiple indistinguishable invisible particles, the method seeks the partition of invisible momentum that minimizes a global function (e.g., the sum of parent masses).

Each rule is applied only to the subset of degrees of freedom relevant at that node, preserving kinematic invariances and reducing correlations between reconstructed observables.

3. Key Kinematic Observables and Discriminants

RJR yields a basis of kinematic observables that exploit the recursive event decomposition and frame hierarchy. The following classes are utilized extensively for signal-background separation, particularly in compressed BSM scenarios (Jackson et al., 2016, Santoni, 2017):

• ISR-based Observables (Topology-Independent):
  • $p_{\text{ISR},T}^{\text{CM}}$: The total transverse momentum of ISR jets in the CM frame.
  • $$R_{\text{ISR}} = |\vec{p}_{I,T}^{\text{CM}} \cdot \hat p_{\text{ISR},T}^{\text{CM}}| / p_{\text{ISR},T}^{\text{CM}}$$: Projects the invisible system's momentum onto the ISR axis; for compressed spectra, $$R_{\text{ISR}} \approx m_{\text{LSP}} / m_{\text{parent}}$$.
  • $\Delta\phi_{\text{ISR},I}$: The opening angle between ISR and invisible systems.
• Signal System Observables (Topology-Dependent):
  • $M_T^V$: Transverse mass of the visible decay products.
  • $M_{\ell^+ \ell^-}$, $M^{\tilde \chi^\pm}$: Invariant and reconstructed parent masses.
  • Angular variables: $\Delta\phi_{\ell, I}$, $\cos\theta$ (alignment of invisible momentum with system boosts).
• Multi-level Frame-Dependent Variables: For example, the scalar product of lepton momenta in specific rest frames provides sensitivity to spin correlations and entangled kinematic patterns (Desai et al., 27 Jan 2026, Desai et al., 26 Jan 2026).

The modularity of RJR allows for tailored observable sets in each analysis, optimizing background rejection and signal sensitivity for the target process.

4. Applications in Compressed Supersymmetry and Exotic Topologies

RJR was originally motivated by the need to probe compressed SUSY spectra, where small mass splittings between parent sparticles and LSPs yield low $p_{\text{ISR},T}^{\text{CM}}$0 visible particles and large missing energy, making traditional high-$p_{\text{ISR},T}^{\text{CM}}$1 analyses ineffective (Santoni, 2017, Jackson et al., 2016). Specific applications include:

• Chargino–Neutralino (3ℓ+$p_{\text{ISR},T}^{\text{CM}}$2) and Chargino–Chargino (2ℓ+$p_{\text{ISR},T}^{\text{CM}}$3): The decay tree structure with explicit ISR assignment allows for the construction of $p_{\text{ISR},T}^{\text{CM}}$4, $p_{\text{ISR},T}^{\text{CM}}$5, $p_{\text{ISR},T}^{\text{CM}}$6, and topology-selective angles to isolate SUSY from SM diboson backgrounds even in the compressed regime. Performance benchmarks show $p_{\text{ISR},T}^{\text{CM}}$7 discovery reach for $p_{\text{ISR},T}^{\text{CM}}$8–$p_{\text{ISR},T}^{\text{CM}}$9 GeV in typical $$R_{\text{ISR}} = |\vec{p}_{I,T}^{\text{CM}} \cdot \hat p_{\text{ISR},T}^{\text{CM}}| / p_{\text{ISR},T}^{\text{CM}}$$0 windows (Santoni, 2017).
• Squark/Gluino Hadronic Signatures: Implementation of jet-based compressed trees and jet-assignment jigsaw rules allows sensitivity to mass splittings as low as 25 GeV and parent masses up to 1 TeV, using $$R_{\text{ISR}} = |\vec{p}_{I,T}^{\text{CM}} \cdot \hat p_{\text{ISR},T}^{\text{CM}}| / p_{\text{ISR},T}^{\text{CM}}$$1, jet multiplicities, $$R_{\text{ISR}} = |\vec{p}_{I,T}^{\text{CM}} \cdot \hat p_{\text{ISR},T}^{\text{CM}}| / p_{\text{ISR},T}^{\text{CM}}$$2, and $$R_{\text{ISR}} = |\vec{p}_{I,T}^{\text{CM}} \cdot \hat p_{\text{ISR},T}^{\text{CM}}| / p_{\text{ISR},T}^{\text{CM}}$$3 (Jackson et al., 2016).
• Toponium Reconstruction: In $$R_{\text{ISR}} = |\vec{p}_{I,T}^{\text{CM}} \cdot \hat p_{\text{ISR},T}^{\text{CM}}| / p_{\text{ISR},T}^{\text{CM}}$$4 threshold studies, RJR enables full kinematic reconstruction of toponium decay chains and the introduction of novel frame-dependent angular variables that enhance discrimination between signal and SM background by up to $$R_{\text{ISR}} = |\vec{p}_{I,T}^{\text{CM}} \cdot \hat p_{\text{ISR},T}^{\text{CM}}| / p_{\text{ISR},T}^{\text{CM}}$$5 in significance (Desai et al., 27 Jan 2026, Desai et al., 26 Jan 2026).

5. Implementation in LHC Analyses and Emulation Techniques

RJR has been integrated into major LHC searches and is supported by the RestFrames software package. In practice, the analysis workflow involves:

• Decay tree specification,
• Systematic partition of visible/invisible and ISR objects,
• Application of jigsaw rules at each recursion step,
• Computation of a comprehensive set of rest-frame and lab-frame observables (Santoni, 2017, Jackson et al., 2017).

For large datasets where recursive event reconstruction may be computationally intensive, emulated RJR (eRJR) strategies have mapped frame-dependent observables onto simpler laboratory proxies (e.g., MET, scalar sums of $$R_{\text{ISR}} = |\vec{p}_{I,T}^{\text{CM}} \cdot \hat p_{\text{ISR},T}^{\text{CM}}| / p_{\text{ISR},T}^{\text{CM}}$$6), achieving high correlation ($$R_{\text{ISR}} = |\vec{p}_{I,T}^{\text{CM}} \cdot \hat p_{\text{ISR},T}^{\text{CM}}| / p_{\text{ISR},T}^{\text{CM}}$$7) with full RJR results in both signal and control regions (Resseguie, 2019, Collaboration, 2018). This approach allows rapid deployment of RJR principles while preserving much of the analysis power.

6. Performance Metrics, Best Practices, and Limitations

RJR analyses have reported high efficiency and purity in reconstruction (O(80–90%)), robust significance gains over traditional approaches, and successful background suppression in compressed topology searches (Santoni, 2017, Desai et al., 27 Jan 2026). Best practices distilled from multiple analyses include:

• Use event-specific decay trees, reflecting all distinguishable decay substructures.
• Assign visible branches by flavor/charge where possible, maximize use of topology-dependent observables.
• Exploit the proportionality $$R_{\text{ISR}} = |\vec{p}_{I,T}^{\text{CM}} \cdot \hat p_{\text{ISR},T}^{\text{CM}}| / p_{\text{ISR},T}^{\text{CM}}$$8 for compressed-mass discriminants, tuning cuts tightly at low $$R_{\text{ISR}} = |\vec{p}_{I,T}^{\text{CM}} \cdot \hat p_{\text{ISR},T}^{\text{CM}}| / p_{\text{ISR},T}^{\text{CM}}$$9.
• Moderate ISR boosting suffices if combined with multi-frame mass and angular observables.
• Leverage 3D frame reconstructions when multiple invisible particles are present, and transverse-only variables are degraded.
• Tune cuts per $$R_{\text{ISR}} \approx m_{\text{LSP}} / m_{\text{parent}}$$0 bin to capture endpoint shifts and maximize signal yield.
• Validate analysis across $$R_{\text{ISR}} \approx m_{\text{LSP}} / m_{\text{parent}}$$1 space and rigorously consider detector effects at low $$R_{\text{ISR}} \approx m_{\text{LSP}} / m_{\text{parent}}$$2 (Santoni, 2017).

Limitations arise from detector-level uncertainties, combinatorics in object assignments, and the idealized treatment of invisible system longitudinal momenta in some implementations (notably eRJR); systematic errors must be carefully propagated, and performance can decrease with pronounced combinatorial backgrounds or degraded MET resolution (Resseguie, 2019, Desai et al., 27 Jan 2026).

7. Extensions and Outlook

RJR's design is fully generalizable, supporting arbitrary decay trees, an extensible set of kinematic rules, and the construction of both standard and novel discriminants (e.g., kinematic edges, frame-aligned angles, entanglement-sensitive variables) for any new physics search including those with significant missing energy or multi-body final states (Jackson et al., 2017, Desai et al., 27 Jan 2026). Its recursive approach fosters the creation of powerful, uncorrelated observables and enables the full exploitation of substructure in LHC event data. Continued development and application of RJR and its emulated forms are expected to drive advances in the search for BSM signatures, especially where conventional techniques fail due to kinematic or combinatorial degeneracies.