Data
(last update of November 2026 — campaign 2026-11)
Timeframe data (2026-11)
Campaign 2026-11 is the streaming-readout study set: 40k K⁺Λ events per energy (9x100, 9x130, 9x275; 5x41 signal only) simulated alone through Geant4 and mixed after Geant4 by TimeframeBuilder with the official ePIC background cocktails. It stops at EDM4hep (no reco, no CSV). Image eic_xl:26.11-stable.
On JLab ifarm:
/work/eic3/users/romanov/meson-structure-2026-11
dd4hep/<E>/ # signal only, 80 files x 500 events per energy (*.edm4hep.root)
dd4hep-tfb2us/<E>/ # 2 us frames: one signal at a random crossing + background, 500 frames per file
dd4hep-tfb200ns/<E>/ # 200 ns "found event" frames: signal pinned at t = 0 + background, 500 frames per file
reco/<E>/ reco-tfb200ns/<E>/ reco-tfb2us/<E>/ # EICrecon 1.43.0 output of the three above (in production)
csv-reco/<E>/ csv-tfb200ns/<E>/ csv-tfb2us/<E>/ # eic2ai CSV zips, one per role (in production)
dd4hep-merger/<E>/ # control sample: merger-path 2 us frames (9x100, 9x275; 40 files)
/work/eic3/users/romanov/background-only-2026-11
frames-2us/<E>/ # background-only 2 us frames, 80 files x 500 frames per energy
reco-2us/<E>/ csv-2us/<E>/ # their EICrecon output and eic2ai CSV zips (in production)
frames-2us-merger/<E>/ # control sample: merger-path background-only frames (9x100, 9x275; 40 files)
pools/<E>/<source>/ # per-source Geant4 pools the frames are built fromFrame files carry the ePIC status offsets per background source (2000 SynRad, 3000 e-gas, 4000 Coulomb, 5000 Touschek, 6000 p-gas) and a SubEventHeaders collection listing every sub-event with its time offset; see Backgrounds in ePIC simulation.
Production finished 2026-11-05: 80 files × 500 events or frames per energy in every dataset. Sizes per energy (9x100 / 9x130 / 9x275): signal 60 / 72 / 62 GB (5x41: 28 GB), 200 ns frames 87 / 99 / 89 GB, 2 µs frames 325 / 338 / 335 GB, background-only 2 µs frames 265 / 265 / 271 GB, pools 98 / 98 / 116 GB. EICrecon output (reco/<E>, reco-tfb200ns/<E>, reco-tfb2us/<E>, background-only-2026-11/reco-2us/<E>) and eic2ai CSV zips (csv-reco/<E>, csv-tfb200ns/<E>, csv-tfb2us/<E>, background-only-2026-11/csv-2us/<E>) are in production (2026-11-05); a merger-path control sample (20,000 frames at 9x100 and 9x275, dd4hep-merger/<E>, frames-2us-merger/<E>) is being produced for the comparison of the two mixing methods. See the campaign page for the details.
Location (2026-07)
Campaign 2026-07 introduces the new 9xN energy scheme (5x41, 9x100, 9x130, 9x275 GeV) for the EIC early-science program.
For machine-background-mixed data, see the Machine background (2026-06) section below.
Reconstruction-based analyses can use reco / csv_reco directly.
On JLab ifarm:
/work/eic3/users/romanov/meson-structure-2026-07XRootD: not exported from the new public door (see Accessing Data with XRootD).
Subdirectories (per-energy dirs 5x41, 9x100, 9x130, 9x275; files named msf_ev1000_NNNN.*) :
afterburner— eg output with crossing-angle + beam effectsbg_merged— Afterburner + background mergeddd4hep— DD4hep full simulation (*.edm4hep.root; ~5.0 TB, ~1000 files/energy)reco— EICrecon reconstruction ofdd4hep(*.edm4eic.root; ~2.2 TB, ~1000 files/energy)csv_dd4hep— acceptance / MC-truth CSVs fromdd4hep(~21 GB)csv_reco— reconstruction CSVs fromreco(*.mc_dis.csv,*.reco_dis.csv,*.mcpart_lambda.csv,*.reco_ff_lambda.csv; ~1.2 GB)eg-split,eg-hepmc,eg-original-2026-06— event-generator inputs
1000 events per file (~1000 files per energy)
# on ifarm (not exported through XRootD, see below)
/work/eic/users/romanov/meson-structure-2026-07/eg-hepmc
/work/eic/users/romanov/meson-structure-2026-07/eg-original-2026-06
/work/eic/users/romanov/meson-structure-2026-07/eg-split
/work/eic/users/romanov/meson-structure-2026-07/afterburner
/work/eic/users/romanov/meson-structure-2026-07/bg_merged
/work/eic/users/romanov/meson-structure-2026-07/dd4hep
/work/eic/users/romanov/meson-structure-2026-07/reco
/work/eic/users/romanov/meson-structure-2026-07/csv_dd4hep
/work/eic/users/romanov/meson-structure-2026-07/csv_for_ai
/work/eic/users/romanov/meson-structure-2026-07/csv_reco
/work/eic/users/romanov/meson-structure-2026-07/analysis
/work/eic/users/romanov/meson-structure-2026-07/datasets-no-backgroundPer-energy .root counts / sizes:
| Energy | dd4hep (.edm4hep.root) | reco (.edm4eic.root) |
|---|---|---|
| 5x41 | 998 (647 GB) | 998 (472 GB) |
| 9x100 | 1000 (1.4 TB) | 999 (718 GB) |
| 9x130 | 1000 (1.7 TB) | 1000 (421 GB) |
| 9x275 | 1000 (1.4 TB) | 1000 (630 GB) |
Machine background (2026-06)
Machine-background-mixed data comes from campaign 2026-06, where the afterburned signal is mixed with the official ePIC background cocktails (synrad, ebrems, ecoulomb, etouschek) via SignalBackgroundMerger into 2 µs timeframes (one signal per frame), then run through npsim with status-offset flags. See Backgrounds in ePIC simulation for the full mechanism.
(!) The background datasets use the old energy scheme
10x100, 10x130, 18x275(5x41 is skipped) and 5000 events per fil, namedk_lambda_{beam}_5000evt_NNNN.*.
On JLab ifarm, under /work/eic3/users/romanov/meson-structure-2026-06:
bg_merged—SignalBackgroundMergeroutput (*.bg.hepmc3.tree.root; the merged input to npsim; ~5.0 TB, 200 files/energy)dd4hep-background— DD4hep on the background-merged input (*.edm4hep.root; ~12 TB)reco-background— EICrecon reconstruction ofdd4hep-background(*.edm4eic.root; ~4.1 TB, partial)csv_dd4hep-background— acceptance / combinatorics CSVs fromdd4hep-background(~9.1 GB)
Per-energy .root file counts (background):
| Energy | bg_merged | dd4hep-background | reco-background |
|---|---|---|---|
| 10x100 | 200 | 133 | 68 |
| 10x130 | 200 | 200 | 43 |
| 18x275 | 200 | 200 | 107 |
Campaign 2026-06 also produced non-background flavors (
dd4hep-official,dd4hep-saveall,dd4hep-stv, and theirreco/csv) — see the campaign 2026-06 page.
Accessing Data with XRootD
(!)
dtn-eic.jlab.orgis retired (it no longer resolves, checked 2026-10-14). The public JLab door is nowroot://dtn2304.jlab.org:8443/and it exports/jlab-osdf-ro/eic/EPIC/{volatile,cache,work}only — the user campaign directories under/work/eic3/users/romanovare not reachable through it. Use the ifarm paths above on the JLab farm, or ask for a copy. Thexrdfs/xrdcpexamples below are kept for the exported ePIC directories. The door needs an XRootD 5.9+ client (theeic_xlimages have it).
To browse the exported files one can use xrdfs command:
xrdfs root://dtn2304.jlab.org:8443/
ls /jlab-osdf-ro/eic/EPIC/volatile/EVGEN/BACKGROUNDSTo download files:
xrdcp root://dtn2304.jlab.org:8443//jlab-osdf-ro/eic/EPIC/volatile/EVGEN/BACKGROUNDS/BEAMGAS/electron/coulomb/EIC_ESR_Xsuite/dataprod_rel_1.0.1/10x275/10000Ahr/MachineRuntime50s/dataprod_rel_1.0.1_electron_coulomb_10x275_10000Ahr.hepmc3.tree.root ./To use directly in scripts:
# Both uproot and pyroot can work with links directly
# if XRootD is installed in the system
import uproot
file = uproot.open("root://dtn2304.jlab.org:8443//jlab-osdf-ro/eic/EPIC/volatile/EVGEN/BACKGROUNDS/BEAMGAS/electron/coulomb/EIC_ESR_Xsuite/dataprod_rel_1.0.1/10x275/10000Ahr/MachineRuntime50s/dataprod_rel_1.0.1_electron_coulomb_10x275_10000Ahr.hepmc3.tree.root")