The REC:: Bank Family¶

Every reconstructed CLAS12 event is a collection of REC::* banks. They fall into three groups:

1. Event-level metadata — one row per event, global to the trigger.
2. The particle list — one row per reconstructed particle.
3. Detector-level banks — one row per detector response, linked back to a particle via pindex. See Cross-bank linking via pindex.

The column listings below come from the CLAS12 reconstruction schema definitions. The descriptions match the per-column documentation in the CLAS12 cook pipeline. At runtime you can always introspect a schema with dict.getSchema("REC::Particle").show() to confirm the layout in the file you're reading.

REC::Event — event-level metadata¶

One row per event, so always read row 0.

Typical uses: cut on liveTime, split by helicity for asymmetry measurements, use startTime as the time origin for TOF calculations.

auto list = reader.getBanks({"REC::Event"});
auto& evt = list[0];
const int helicity_col = evt.getSchema().getEntryOrder("helicity");

while (reader.next(list)) {
    if (evt.getRows() == 0) continue;
    int h = evt.getInt(helicity_col, 0);
    if (h == 1) /* … */;
}

REC::Particle — the particle list¶

One row per reconstructed particle. This is the bank every analysis touches.

status — which detectors saw the particle¶

The status field encodes the detector combination that produced the particle. The magnitude tells you the primary detector system:

Combinations (e.g., a FD particle with a FT match) raise the value — consult the CLAS12 analysis note for your pass if the exact decoding matters. For most cuts, the range-based test is enough:

const int status_col = parts.getSchema().getEntryOrder("status");
int s = std::abs(parts.getInt(status_col, row));
bool in_fd = s >= 2000 && s < 4000;

pid vs. chi2pid¶

pid is a discrete label; chi2pid is the underlying PID confidence. A common high-purity cut is:

int   pid     = parts.getInt(pid_col, row);
float chi2pid = parts.getFloat(chi2pid_col, row);
if (std::abs(pid) == 211 && std::abs(chi2pid) < 3.0f) { /* good π */ }

Detector-level banks — one row per response¶

Every detector bank has the same preamble: a pindex column pointing to a row in REC::Particle, plus detector, sector, and (usually) layer, plus detector-specific measurements. One particle can produce multiple rows in a detector bank (e.g., one track can light up PCAL, EC inner, and EC outer — three rows in REC::Calorimeter, all with the same pindex).

REC::Calorimeter¶

Responses in the electromagnetic calorimeter (PCAL, ECin, ECout). Key columns:

Typical use: sum energy across all PCAL+EC rows linked to the electron to build its total calorimeter energy (see pindex linking).

REC::Scintillator¶

Responses in the FTOF (Forward) and CTOF (Central) scintillators.

Typical use: the source of the beta column in REC::Particle, and used for more refined hadron PID than chi2pid.

REC::Cherenkov¶

Responses in HTCC / LTCC / RICH.

Typical use: nphe > 2 is the standard electron-ID threshold against pion contamination.

REC::Track¶

The reconstructed track-level summary. Tracking produces one row per track; that track may or may not match a particle in REC::Particle.

REC::Traj¶

Sampled trajectory points along each track — positions and direction cosines at every detector plane the track crossed. Use it for fiducial cuts that need the track's entry position into a specific detector, not just the hit position.

edge is the standard input to fiducial cuts — e.g., edge > 10 cm rejects particles near the detector boundaries where acceptance is poorly known.

Pass-specific bank variants¶

CLAS12 reconstruction writes three parallel families of banks into the same file:

• REC::* — time-based reconstruction (the default output).
• RECHB::* — hit-based reconstruction (intermediate output, sometimes used for cross-checks or calibration).
• RECFT::* — Forward-Tagger reconstruction.

The column layouts are mostly shared but differ slightly. Pick the family that matches the bank names in the file you're reading — reader.readDictionary(dict) followed by dict.getSchemaList() will print every schema present.

Next¶

• Link detector responses back to particles → Cross-bank linking via pindex
• Filter on PID, charge, status → Recipes — Row filtering
• Introspect a file's schemas at runtime → Recipes — Listing Available Banks