ATLAS dataset nomenclature

Reference: ATL-COM-GEN-2007-003 "ATLAS Dataset Nomenclature" (2024 edition).

Logical Dataset Names (LDNs)

AMI identifies all datasets via Logical Dataset Names (LDNs). For centrally produced ATLAS data the project field doubles as the Rucio scope (e.g. mc20_13TeV, data18_13TeV).

Data hierarchy:

events ∈ files ∈ datasets ∈ dataset containers ∈ campaigns

Scopes and AMI catalogs

Scope pattern	AMI catalog	Description
`mc16_13TeV`	`mc15_001:production`	Run-2 MC, 13 TeV (original)
`mc20_13TeV`	`mc15_001:production`	Run-2 MC, 13 TeV (reprocessing)
`mc21_13p6TeV`	`mc21_001:production`	Run-3 MC pilot, 13.6 TeV
`mc23_13p6TeV`	`mc23_001:production`	Run-3 MC, 13.6 TeV
`data15_13TeV` – `data24_13p6TeV`	—	Real collision data by year
`user.<cern_username>`	—	Personal user datasets
`group.<atlas_group>`	—	Group-owned datasets

Note

mc16 and mc20 evgen datasets live in the mc15_001:production AMI catalog because they were generated with mc15-era job options.

Monte Carlo format

project.datasetNumber.physicsShort.prodStep.dataType.AMITags

Example:

mc20_13TeV.700320.Sh_2211_Zee_maxHTpTV2_BFilter.deriv.DAOD_PHYS.e8351_s3681_r13144_r13146_p5855

Field	Value	Meaning
`project`	`mc20_13TeV`	MC campaign, Run-2, 13 TeV
`datasetNumber`	`700320`	DSID — identifies the physics sample
`physicsShort`	`Sh_2211_Zee_maxHTpTV2_BFilter`	Generator, tune, process
`prodStep`	`deriv`	Production step
`dataType`	`DAOD_PHYS`	Output format
`AMITags`	`e8351_s3681_r13144_r13146_p5855`	Processing history chain

Real data formats

Primary datasets (individual runs):

project.runNumber.streamName.prodStep.dataType.AMITags

data18_13TeV.00348885.physics_Main.deriv.DAOD_PHYS.r13286_p4910_p5855

Physics containers (preferred for analysis — groups all runs in a period):

project.periodName.streamName.PhysCont.dataType.contVersion

data15_13TeV.periodAllYear.physics_Main.PhysCont.DAOD_PHYSLITE.grp15_v01_p5631

Use periodAllYear to get the full dataset for a given year, or specific period letters (periodA, periodB, …) for subsets.

AMI tags

AMI tags record the processing history of a dataset as a chain of single-letter codes followed by a version number:

Tier-0 (online/express processing):

Letter	Step
`f`	Reconstruction (bulk)
`c`	Reconstruction (commissioning)
`x`	Reconstruction (express)
`m`	Merging

ProdSys (offline processing):

Letter	Step
`e`	Event generation (evgen)
`s`	Detector simulation (simul)
`d`	Digitisation
`r`	Reconstruction
`p`	Group production / derivation
`t`	Merging
`n`	Event picking

Common data types

Type	Description
`DAOD_PHYS`	Full derivation — most complete, largest
`DAOD_PHYSLITE`	Lightweight derivation — preferred for analysis
`DAOD_EXOT*`	Exotic physics group derivations
`DAOD_SUSY*`	SUSY group derivations
`AOD`	Analysis Object Data — pre-derivation
`ESD`	Event Summary Data — large, rarely used directly
`EVNT`	Generator-level events
`HITS`	Simulated detector hits
`RDO`	Raw detector output

Other dataset types

Beyond MC and collision data, you may encounter:

Type	Format
Calibration	`dataNNN_calib.xxxxxxxx.calibration_<Det>-<info>.daq.RAW`
User datasets	`user.<cern_username>.<project>.<fields>.<dataType>.<versionTag>`
Group datasets	`group.<atlas_group>.<project>.<fields>.<dataType>.<versionTag>`
Conditions	`project.internalCondNumber.datasetUsage.COND`

Field length limits

Field	Limit
`scope`	≤ 25 chars
`name` (dataset)	≤ 255 chars
`name` (container)	≤ 150 chars
`project`	≤ 15 chars
`datasetNumber`	6–8 digits
`physicsShort`	≤ 50 chars
`dataType`	≤ 15 chars
AMITag (one tag)	≤ 32 chars

Finding MC job options

MC samples are defined by their DSID. Job options live in:

https://gitlab.cern.ch/atlas-physics/pmg/mcjoboptions/-/tree/master/<DSIDxxx>/<DSID>

where <DSIDxxx> is the first three digits of the DSID followed by xxx (e.g. DSID 700320 → 700xxx/700320).