AIMS Provenance: Rubin Observatory / LSST Alerts

Publisher source

Rubin Observatory describes LSST alerts as world-public products of Prompt Processing. Alert packets contain measurements associated with sources detected in difference images.

The LSST Alert Production Database schema defines DiaSource fields used by AIMS, including diaSourceId, midpointMjdTai, ra, dec, raErr, decErr, psfFlux, psfFluxErr, band, visit and detector identifiers, quality flags, and optional Solar System association.

What AIMS selects

AIMS receives alert data through broker-backed BigQuery tables and selects compact alert fields into daily source tables. Production uses the Pitt-Google broker stream; development environments may use simulated alert tables.

A candidate must pass the PSF measurement flags:

psfFlux_flag is not set;
psfFlux_flag_edge is not set;
psfFlux_flag_noGoodPixels is not set.

It must then have either a non-null ssObjectId or psfFlux / psfFluxErr >= 5. Publication into AIMS additionally requires non-null midpointMjdTai, RA, and Dec.

These cuts select usable prompt difference-image measurements for moving-source search. They do not replace the full alert packet or downstream broker classification.

Coverage and observatory

AIMS coverage defaults to 2025-09-01 onward and is continuing.
AIMS observatory code: X05.

Normalization into AIMS

AIMS field	LSST source or derivation
`observation_id`	`lsst:<diaSourceId>`.
`exposure_id`	`lsst:<visit>-<detector>`; missing components are represented deterministically.
`dataset_id`	Constant `lsst`.
`observatory_code`	Constant `X05`.
Observation time	`midpointMjdTai` converted from TAI MJD to UTC MJD and timestamp.
`exposure_duration_s`	Operational default of 30 seconds.
`ra_deg`, `dec_deg`	`ra`, `dec`.
`ra_sigma_deg`, `dec_sigma_deg`	`raErr`, `decErr`, mapped to the AIMS east/north per-axis fields.
`mag`	AB magnitude derived from calibrated positive `psfFlux` in nJy.
`mag_sigma`	First-order propagated AB magnitude uncertainty from `psfFluxErr`.
`filter`	`band`.
`healpixel`	Downsampled from NESTED `healpix19` to nside=128 NESTED.

Time, photometry, and spatial indexing

AIMS uses a leap-second-aware TAI-to-UTC conversion for midpointMjdTai.

For positive flux, the photometric conversion is:

m_AB = -2.5 * log10(psfFlux_nJy) + 31.4
sigma_mag = (2.5 / ln(10)) * psfFluxErr / psfFlux

AIMS stores null magnitude or magnitude uncertainty when the flux values cannot support these conversions.

The NESTED spatial index is downsampled with:

healpixel = floor(healpix19 / 16,777,216)

Published AIMS products

In the AIMS BigQuery dataset, normalized LSST alert rows are written to aims_lsst, copied into aims_detections_hot, and exposed through the aims_detections_unified view with dataset_id = lsst.

The same normalized rows are exported to Google Cloud Storage through the daily aims_detections_parquet product. AIMS also derives aims_frames_unified BigQuery rows and aims_frames_parquet search indexes grouped by visit/detector exposure identity and spatial cell. Complete detection/frame pairs are listed in aims_unified_parquet_catalog/catalog.json for consumers such as Precovery. These products contain selected normalized DiaSource measurements, not the complete alert packets.

Scientific validity and limitations

Source position, time, band, flux, and uncertainty fields retain Rubin’s published DiaSource semantics apart from documented conversions. AB magnitudes use the standard nJy conversion, but negative difference flux correctly has no magnitude representation.
The quality and signal-to-noise cuts create a search-oriented alert subset, not a full LSST data-release catalog or an unbiased transient sample.
The 30-second exposure duration is an AIMS operational default. Observation time itself comes from the publisher’s midpoint field.
Rubin defines raErr and decErr as local tangent-plane components, but the current AIMS row omits the source cross-axis covariance.
Some pre-DP2 AIMS alert generations contain an additional cos(dec) factor in the stored eastward uncertainty. The normalized row lacks sufficient version information for reliable read-time repair, so Rubin uncertainty values are generation-dependent and should not be treated as a uniform calibrated covariance.
Alert schemas and broker delivery can evolve. Any source-version change that alters field semantics requires renewed validation of the normalized product.