MadGraph5_aMC@NLO + Pythia 8.3 event samples for cell resampling studies (13 TeV, Z+jets and top quark pair production)

MadGraph5_aMC@NLO + Pythia 8.3 event samples for cell resampling studies (13 TeV, Z+jets and top quark pair production) Overview This record contains two simulated proton–proton collision samples at a center-of-mass energy of √s = 13 TeV, each of 100,000 events, stored as complete truth-level particle records: * Z+jets — production of a Z boson in association with jets (ppzjj_NLO_100k_slim.root)* ttbar — top quark pair production (ttbar_NLO_100k_slim.root) The samples were produced for a study of optimal-transport-based cell resampling of negatively weighted Monte Carlo events, and are released so that the study can be reproduced and to support further work on negative-weight reduction. Event generation Both samples were generated identically apart from the hard process. Hard-scattering matrix elements are calculated at next-to-leading order (NLO) in QCD with MadGraph5_aMC@NLO 3.5.6, using the NNPDF2.3 NLO parton distribution function set (LHAPDF ID 244600). Renormalization and factorization scales are set dynamically to half the sum of the transverse masses of the final-state particles (the MadGraph5_aMC@NLO default). The NLO matrix elements are matched to the parton shower with the MC@NLO method; no multileg merging is applied (ickkw = 0). Parton showering, hadronization, multi-parton interactions and the underlying event are modeled with Pythia 8.3 using the default Monash set of tuned parameters. Hadronization uses Pythia's default Lund string fragmentation model. No folding of the NLO subtraction terms is applied, so both samples retain the full negative weight fraction produced by the MC@NLO matching procedure. Both leptonic and hadronic decays of the Z boson and of the top quarks are allowed. The complete run cards are included (run_card_zjj.dat, run_card_ttbar.dat). They are identical except for a job-splitting parameter (nevt_job). Generator-level selection A loose generation-level selection keeps event generation efficient while remaining inclusive. It is the same for both samples: - Jets clustered with the kt algorithm, R = 0.7, pT > 10 GeV (FastJet, via MadGraph5_aMC@NLO)- Same-flavor opposite-sign lepton pairs for Z+jets events: m(ll) > 30 GeV No further selection is applied to the stored particle record. Analysis-level cuts (e.g. pT > 0.1 GeV and |η| < 4.9 for final-state particles) are left to the user. File contents Both ROOT files contain a single TTree named Events with an identical branch schema. Each entry is one event holding the complete Pythia 8 particle record, flattened from the HepMC output. Event-level branches (one scalar per event):Event_number, Event_numberV, Event_numberP, Event_weight (per-event weight, pb), Xsection_value (running sum of event weights, pb), Xsection_error, Event_barcodeBP1, Event_barcodeBP2, nParticle (storage counter, equal to Event_numberP). Particle-level branches (variable-length arrays, length Event_numberP):Particle_pid (PDG ID), Particle_status (Pythia 8 status code), Particle_px, Particle_py, Particle_pz, Particle_energy (float32, GeV), Particle_mass (float64, GeV), Particle_x, Particle_y, Particle_z, Particle_ctau (production vertex, mm), Particle_barcode (HepMC barcode; integer-valued but stored as double), Particle_d1, Particle_d2 (first and last daughter, as 0-based indices into these arrays; -1 means no daughter / single daughter respectively). Only daughter links are stored — there are no mother branches. The files are ZSTD-compressed, which requires ROOT 6.20 or later (2020). Simulation stages Because the full Pythia 8 event record is retained, events can be reconstructed at three successive stages of the simulation by selecting on Particle_status: HS — hard scattering, immediately after the matrix element (status 23 outgoing) PS — after the parton shower (status 41–44 ISR/FSR, 51–53, 62, 71) HAD — after hadronization (status 1, final-state) This makes the samples suitable for studies that compare observables or distance metrics across simulation stages using infrared- and collinear-safe quantities applied directly to particle-level information. The parton shower and hadronization introduce no additional negatively weighted events, so comparisons across stages use the same event samples. Numerical precision Particle momenta and energies (Particle_px, Particle_py, Particle_pz, Particle_energy) are stored as float32, giving a relative precision of ~6e-8. All other branches, including Particle_mass and the vertex positions, are float64. This is a reduced precision relative to the published results; however, the effect of the reduced precision is far below any physically meaningful scale: anti-kt R = 0.4 jet pT (pT > 20 GeV) changes by at most a relative factor of ~5e-8, the jet multiplicity is unaffected, and the optimal-transport event distance changes by at most a relative factor of ~3e-8. Fourteen branches that the HepMC-to-ROOT converter declared but never populated (Event_scale, Event_alphaQCD, Event_alphaQED, Event_numberMP, Event_barcodeSPV and the nine PDF_* branches) have been removed. Usage Event weights are in the Event_weight branch, in pb. The magnitude is constant within each sample and only the sign varies, as expected for an MC@NLO sample with no folding applied. import numpy as np import uproot tree = uproot.open("ttbar_NLO_100k_slim.root")["Events"] weights = np.asarray(tree["Event_weight"].array()) print(weights.sum()) # 678.62 pb print((weights < 0).mean()) # 0.22837 The Xsection_value branch is also retained: it holds the running sum of event weights as written by the generator, from which Event_weight was reconstructed. Selecting the final-state particles of the first event: arrays = tree.arrays( ["Particle_pid", "Particle_status", "Particle_px", "Particle_py", "Particle_pz", "Particle_energy"], entry_stop=1, ) final_state = arrays["Particle_status"] == 1 pid = arrays["Particle_pid"][final_state] The final-state record is fully inclusive: it contains charged leptons and neutrinos as well as hadrons. In the Z+jets sample, hadronic and invisible Z decays are both allowed, so Z → νν events contribute a population of high-pT neutrinos. Citation If you use these datasets, please cite arXiv:2607.08723 as well as this record.