Suffix Explanation

Column Name Suffix	Description
_merged	Best single value for this source, computed by weighted averaging (photometry) or priority selection (redshift, morphology). This is the column you should use for science.
_all	JSON array of ALL values from every matched catalog for this source, e.g. [1.234, 1.237, 1.240]. Preserves full provenance — useful for checking agreement across surveys. Values are homogenized to the AGN-DB unit system, not reproduced in the units each catalog published: magnitudes are converted to AB, and the X-ray (xf) and IR/radio (rf) flux families are stored as log10, in the same units as their _merged counterparts (sections 5 and 10). The per-catalog decomposition is what _all preserves; _origin identifies which catalog each entry came from.
_origin	JSON array of integer catalog IDs (matching _all order), identifying which catalog contributed each value. Catalog ID mapping is in index2ds.pickle.
_sub_all	JSON array of sub-classifications (morphology only), providing finer detail than the main class label.
e_	Prefix indicating measurement uncertainty (error) on the quantity that follows. E.g., e_U_mag_merged is the error on U_mag_merged.

Missing Data

NaN (Not a Number) indicates no measurement is available.
Inside an e_*_all error-provenance array, the bare JSON token `null` appears in slots where the contributing catalog reported the value but no error for it (e.g., e_W1_mag_all = [null, null, null, 2.718]). Index alignment with the matching *_all and *_origin arrays is preserved. Parse with json.loads().

Identification & Coordinates

Column Name

Description

AGN-DB_ID

Stable integer source identifier

AGN-DB_name

Unique coordinate-based source identifier following the IAU designation convention (1AGNDB JHHMMSS.sss±DDMMSS.sss, e.g. 1AGNDB J095608.945+024220.318)

Right Ascension (J2000) [degrees, 0–360]

DEC

Declination (J2000) [degrees, -90 to +90]

pos_error

Combined positional uncertainty of the error-weighted mean position (RA, DEC), in arcseconds. For matched sources this is (sum of 1/sigma_i^2)^(-1/2) over the member entries; for single-catalog sources it is the input catalog's positional error. Input uncertainties that are missing or unphysical (NaN/inf/<=0) are replaced at ingestion with the median positional uncertainty of the valid entries of the same catalog (or, if a catalog has no valid entries, with its angular resolution).

origin

Python-tuple-repr string of catalog IDs that contributed to this source after LYRA cross-matching, e.g. "(21, 32, 81)". Parse with ast.literal_eval().

ra_all

JSON array of the original per-catalog Right Ascensions (J2000, degrees) of the constituent entries, ordered by ascending catalog ID (same ordering as `origin`), e.g. [149.037269, 149.039382]. Together with dec_all this allows any association to be decomposed into its constituent catalog entries. Values rounded to 1e-6 deg (3.6 mas).

dec_all

JSON array of the original per-catalog Declinations (J2000, degrees), same ordering and conventions as ra_all.

in_galactic_plane

Boolean flag: True if the source lies within |b| < 10 degrees of the Galactic plane. Sources near the plane are more likely to suffer from stellar contamination, high extinction, and source confusion. Users building clean extragalactic samples should filter on in_galactic_plane == False.

galaxy_only_flag

Boolean flag (stage 7.52): True when the source has at least one "galaxy" classification and NO AGN-type classification (type1, type2, generic_agn, blazar, lens_system, dual/merger) in any tier. These may be obscured/buried AGN where nuclear activity is undetected.

unknown_only_flag

Boolean flag (stage 7.52): True when the source carries no usable classification — either every reported value across all five tiers is "unknown", or no tier reports any classification at all (2026-07-11 change; before that date the flag covered only the all-"unknown" case). The raw class arrays preserve the distinction: empty arrays mean nothing was reported, explicit "unknown" entries mean a catalog attempted a classification without reaching a conclusion.

star_flag

Categorical (stage 7.5). One of:

'confirmed' — spec_class contains 'star' and no non-star spec_class. Spectroscopic star call; treat as definitive.
'likely' — no spec_class, all non-spec classifications (SED/xray/image/gen) say 'star', and no X-ray detection in any xf*_all column. Photometric star call with no rescue signal. Caveat: a null X-ray flux can also mean no X-ray coverage at that position, not just a non-detection.
'unlikely' — non-spec classifications all say 'star' BUT at least one xf*_all is non-null. The X-ray detection overrides the photometric star call, suggesting a misclassified AGN.
'none' — not flagged as a star (any non-star classification in any tier, no classification at all, or every classification is 'unknown'). No rows are removed by stage 7.5.

Typical filters:

df.star_flag == 'none'	# strict
df.star_flag != 'confirmed'	# permissive
df.star_flag.isin(['none','unlikely'])	# matches legacy behaviour

close_intra_dup_arcsec

Float (stage 7.53): separation in arcseconds to the nearest *intra-catalog* neighbor within 1.0", or NaN when none exists. Two merged sources are intra-catalog neighbors when at least one source-catalog contributed to both (set(origin_A) & set(origin_B) ≠ ∅). LYRA forbids merging two detections from the same catalog, so a small value here reflects a catalog-side data quality artifact, NOT a LYRA assignment failure. Typical filter for "trust this source is well isolated": df.close_intra_dup_arcsec.isna() | (df.close_intra_dup_arcsec > 0.1).

close_intra_dup_partner

Int64 nullable (stage 7.53): row index in this same table of the intra-catalog neighbor recorded in close_intra_dup_arcsec, or "NA". Use to inspect the counterpart directly: df.iloc[partner_idx].

Crossmatch Quality

Column Name	Description
match_quality_tier	Matched sources (>= 2 catalogs) are tiered on match_probability_relative (P_rel), tested in order: 'secure' — P_rel >= 0.9, or >= 0.5 for >= 3 catalogs; 'reliable' — P_rel >= 0.5, or >= 0.1 for >= 3 catalogs; 'possible' — all remaining accepted matches. Unmatched, single-catalog sources get 'unsure' (a candidate association was evaluated and rejected) or 'isolated' (no counterpart was ever found within the search radius).
match_probability	Float in (0, 1]. P_match = O/(1+O), where O is the posterior odds of the accepted hypothesis. NaN for single-catalog sources.
match_probability_relative	Float in (0, 1]. Relative probability P_rel of the accepted hypothesis, normalized over the conflicting hypotheses (those sharing at least one source) within its spatial group; the quantity on which the quality tiers are based. NaN for single-catalog sources.
astrometry_attached	Boolean. True when the match includes entries added by the post-greedy astrometric reattachment step (unmatched source within 0.1″ of a winning match, normalized separation within ~1σ).

Redshift

Redshift Types

Column Name	Description
spec_Z_all	Spectroscopic redshift — all values [JSON array]
spec_Z_origin	Catalog IDs for spec_Z values [JSON array]
e_spec_Z_all	Errors on spectroscopic redshifts [JSON array]
phot_Z_all	Photometric redshift — all values [JSON array]
phot_Z_origin	Catalog IDs for phot_Z values [JSON array]
e_phot_Z_all	Errors on photometric redshifts [JSON array]
min_phot_Z_all	Lower bound of photo-z confidence interval [JSON]
max_phot_Z_all	Upper bound of photo-z confidence interval [JSON]
Z_all	Generic redshift (type unspecified) — all values
Z_origin	Catalog IDs for Z values [JSON array]

Best redshift values

Column Name	Description
best_Z_all	Best redshift, selected by priority: 1st = spectroscopic (most reliable), 2nd = photometric, 3rd = generic. Stored as JSON array (first element is the best).
best_Z_merged	Scalar best redshift value [float]: the MEAN of the values stored in best_Z_all. Averaging is always confined to the winning category — a spectroscopic redshift is never averaged with a photometric or generic one. For 77.6% of sources the winning category holds a single value and best_Z_merged is simply that value; for the remaining 22.4% it is the mean over the catalogs contributing to that one category. This is the column you should use for science when you want a single numeric redshift per source; use best_Z_all if you need the individual measurements.
best_Z_origin	Catalog ID of the best redshift [JSON array]
f_best_Z	Flag indicating best redshift type [float]: 1.0 = spectroscopic, 2.0 = photometric, 3.0 = generic

Classification

Label scheme

type1: Broad-line AGN (Seyfert 1, QSO Type 1, BLAGN)
type2: Narrow-line AGN (Seyfert 2, NLAGN, Sy1.5–1.9)
generic_agn: AGN without type specification
blazar: Blazars (BL Lac, FSRQ)
galaxy: Non-AGN galaxies (elliptical, spiral, starburst, etc.)
star: Stars, white dwarfs, CVs, X-ray binaries
lens_system: Gravitational lens systems
dual/merger: Dual AGN, merging pairs, interactions
cluster: Galaxy clusters and groups
unknown: Unclassifiable or insufficient data

Column Name	Description
spec_class_all	Spectroscopic classification — all values
spec_class_origin	Catalog IDs [JSON array]
spec_class_sub_all	Sub-classification detail (e.g. "sy1.5", "bllac")
gen_class_all	Generic classification — all values
gen_class_origin	Catalog IDs [JSON array] (Note: gen_class has no _sub_all column; the other four tiers do.)
SED_class_all	SED-fitting classification — all values
SED_class_origin	Catalog IDs [JSON array]
SED_class_sub_all	Sub-classification detail
xray_class_all	X-ray classification (point-like vs extended)
xray_class_origin	Catalog IDs [JSON array]
xray_class_sub_all	Sub-classification detail
image_class_all	Visual/image-based classification — all values
image_class_origin	Catalog IDs [JSON array]
image_class_sub_all	Sub-classification detail
best_class_all	Best classification, selected by priority: spec_class > SED_class > xray_class > image_class > gen_class
best_class_origin	Catalog ID of the best classification
best_class_sub_all	Sub-classification of the best class

Photometry — UV/Optical/Near-IR Magnitudes

All magnitudes are in the AB system (after unit conversion). Each band has 5 columns: {band}_merged, e_{band}_merged, {band}_all, e_{band}_all, and {band}_origin.

Band	Wavelength	System / Instrument
FUV_mag	1539 Å	GALEX Far Ultraviolet
NUV_mag	2316 Å	GALEX Near Ultraviolet
1450Mag	1450 Å	1450 Å absolute magnitude
1450mag	1450 Å	1450 Å apparent magnitude
u_mag	3557 Å	SDSS u
U_mag	3652 Å	Johnson U
B_mag	4448 Å	Johnson B
g_mag	4825 Å	SDSS g
GBP_mag	5320 Å	Gaia Blue BP
V_mag	5505 Å	Johnson V
r_mag	6261 Å	SDSS r
R_mag	6581 Å	Cousins R
G_mag	6730 Å	Gaia broad G
i_mag	7672 Å	SDSS i
GRP_mag	7970 Å	Gaia Red (RP)
I_mag	8059 Å	Cousins I
z_mag	9097 Å	SDSS z
Y_mag	1.02 µm	Y band
J_mag	1.24 µm	J band
H_mag	1.65 µm	H band
K_mag	2.16 µm	K band
W1_mag	3.4 µm	WISE band 1
W2_mag	4.6 µm	WISE band 2
W3_mag	12 µm	WISE band 3
W4_mag	22 µm	WISE band 4

Photometry — Infrared & Radio Fluxes [log10 mJy]

All IR and radio flux densities are stored as log10(flux density in mJy) in the merged columns. Each band has 5 columns: {band}_merged, e_{band}_merged, {band}_all, e_{band}_all, and {band}_origin.

Note:

rf1–rf4, rf9, rf13, and rf17 are defined in the schema but may not appear in the final table if no data were merged for those bands.

Name	Wavelength / Frequency	System/instrument
rf5	20 cm / 1.4 GHz	Radio (e.g., VLA FIRST)
rf7	24 µm	Mid-IR (e.g., Spitzer MIPS)
rf8	250 µm	Far-IR (e.g., Herschel SPIRE)
rf10	3.6 µm	Mid-IR (e.g., Spitzer IRAC)
rf11	350 µm	Sub-mm (e.g., Herschel SPIRE)
rf12	4.5 µm	Mid-IR (e.g., Spitzer IRAC)
rf14	500 µm	Sub-mm (e.g., Herschel SPIRE)
rf15	8.0 µm	Mid-IR (e.g., Spitzer IRAC)
rf16	5.8 µm	Mid-IR (e.g., Spitzer IRAC)
rf18	20 GHz / 1.5 cm	Radio (e.g., ATCA)
rf19	8 GHz / 3.7 cm	Radio (e.g., ATCA)
rf20	5 GHz / 6 cm	Radio (e.g., VLA)
rf21	2.1 m / 144 MHz	Radio (e.g., LOFAR LoTSS)

Photometry — X-ray Fluxes [erg/s/cm²]

All X-ray fluxes are stored as log10(flux) in the merged columns. Each band has 5 columns: {band}_merged, e_{band}_merged, {band}_all, e_{band}_all, and {band}_origin.

X-ray count rates (xcr*) and photon counts (counts*) are also released where available, in the same energy bands as the fluxes above — e.g. xcr2 and counts2 correspond to the same band as xf2.

Band	Energy Range	System/instrument
xf1	0.1–2.4 keV	e.g., ROSAT
xf2	0.2–12 keV	e.g., XMM-Newton
xf3	2–12 keV	XMM-Newton Slew (hard)
xf4	0.5–10 keV	e.g., XMM-Newton, Chandra
xf5	0.5–2 keV	e.g., XMM-Newton, eROSITA
xf6	0.5–4.5 keV	e.g., XMM-Newton
xf7	0.5–7 keV	e.g., Chandra
xf8	14–195 keV	e.g., Swift
xf9	2–10 keV	e.g., XMM-Newton, Chandra
xf10	0.2–2 keV	XMM-Newton Slew (soft)
xf11	2–7 keV	e.g., Chandra
xf12	100 MeV – 100 GeV	e.g., Fermi
xf13	4.5–7.5 keV	e.g., XMM-Newton
xf14	0.2–0.5 keV	e.g., XMM-Newton
xf15	0.5–1 keV	e.g., eROSITA, XMM-Newton
xf16	1–2 keV	e.g., eROSITA, XMM-Newton
xf17	2–4.5 keV	e.g., XMM-Newton
xf18	4.5–12 keV	e.g., XMM-Newton
xf20	1.2–2.0 keV	Chandra CSC medium band
xf21	0.5–1.2 keV	Chandra CSC soft band
xf22	0.2–0.5 keV	Chandra CSC ultrasoft band
xf23	~0.1–10.0 keV	Chandra CSC wide band
xf24	18–55 keV	e.g., Swift, INTEGRAL
xf25	2.3–5 keV	e.g., eROSITA
xf27	0.2–2.3 keV	e.g., eROSITA
xf28	2–5 keV	e.g., eROSITA
xf29	5–8 keV	e.g., eROSITA
xf30	3–8 keV	e.g., NuSTAR
xf31	8–24 keV	e.g., NuSTAR
xf32	3–24 keV	e.g., NuSTAR

Hardness Ratios [dimensionless]

Hardness ratios quantify the spectral shape: higher values indicate harder (more energetic) X-ray spectra.

Defined as (H-S)/(H+S) where H and S are counts in hard and soft bands respectively.

Name	Ratio Name	Instrument / Catalog
HR1	Hardness ratio 1	XMM-Newton source catalog
HR2	Hardness ratio 2	XMM-Newton source catalog
HR3	Hardness ratio 3	XMM-Newton source catalog
HR4	Hardness ratio 4	XMM-Newton source catalog
HR5	Hard/Medium counts ratio (HRhm)	Chandra source catalog
HR6	Hard/Soft counts ratio (HRhs)	Chandra source catalog
HR7	Medium/Soft counts ratio (HRms)	Chandra source catalog
HR8	(2–4.5 keV) / (0.5–2 keV)	Mixed
HR9	(2–7 keV) / (0.5–2 keV)	Mixed