Gravitational-Wave Detectors and Open Data
Scrollable formula. Use Left and Right arrow keys to reveal hidden content.
Scrollable table. Use Left and Right arrow keys to reveal additional columns.
Scrollable figure. Use Left and Right arrow keys to inspect hidden content.
The Gravitational Wave Open Science Center’s O4a and O4b bulk releases provide calibrated detector strain together with one-second data-quality and hardware-injection masks. They do not provide an automatically valid analysis segment, an event catalog, or a posterior. This page fixes the two accepted release identities and the choices that must accompany any use. Product and documentation links were checked on 2026-08-25.
Helpful background. Probability and Inference for Gravity explains uncertainty propagation, likelihood construction, and selection effects used downstream of the data choices recorded here.
O4a identifier status — rechecked 2026-08-25. The official O4a release page declares DOI
10.7935/kt51-6n86. A normal DOI request now returns HTTP 302 to the official O4a release alias, so the identifier resolves. Machine-readable citation metadata remain unavailable: CSL JSON negotiation ends at Crosscite with HTTP 404, the exact DataCite record returns 404, and an exact DataCite query returns no records. Treat this as a resolving DOI with unavailable registry citation metadata. Do not invent missing metadata or substitute another identifier; bind the DOI to the official release page, exact R1 archive IDs, checksum manifests, and retrieval date.
Two bulk-strain releases
Section titled “Two bulk-strain releases”| Field | O4a | O4b |
|---|---|---|
| Persistent identity | Resolving DOI 10.7935/kt51-6n86; registry citation metadata unavailable | Resolving DOI 10.7935/8emv-ag54 |
| Archive IDs | O4a_4KHZ_R1, O4a_16KHZ_R1 | O4b_4KHZ_R1, O4b_16KHZ_R1 |
| First public release | 2025-08-26 | 2026-05-26 |
| Nominal run interval | 2023-05-24 15:00 UTC, GPS 1368975618, through 2024-01-16 16:00 UTC, GPS 1389456018 | 2024-04-10 15:00 UTC, GPS 1396796418, through 2025-01-28 17:00 UTC, GPS 1422118818 |
| Actual bulk-release start | 2023-05-15 14:13:22 UTC, GPS 1368195220; early segments precede the nominal run start | 2024-04-06 05:37:12 UTC, GPS 1396417050; early ER16 segments precede the nominal run start |
| Detectors | LIGO Hanford H1; LIGO Livingston L1 | H1; L1; Virgo V1 |
| Final source channels | H1:GDS-CALIB_STRAIN_CLEAN_AR; L1:GDS-CALIB_STRAIN_CLEAN_AR | H1:DCS-CALIB_STRAIN_CLEAN_AR01; L1:DCS-CALIB_STRAIN_CLEAN_AR01; V1:Hrec_hoftRepro1AR_16384Hz |
| Pre-run source channels | Same LIGO channels as the final release | H1:GDS-CALIB_STRAIN_CLEAN_AR; L1:GDS-CALIB_STRAIN_CLEAN_AR; V1:Hrec_hoftRepro1AR_16384Hz |
The interval stored in a fixed-boundary file is not necessarily observing live
time. Use the DATA bit and the analysis-specific quality category at every
step. The absence of Virgo from the O4a bulk product and its presence in O4b
are product facts, not statements about every instrument activity during the
broader O4 campaign.
The accepted O4a detector set is exactly H1 and L1; it contains no V1
strain. The accepted O4b detector set is exactly H1, L1, and V1. This
explicit set membership matters when constructing coincident segments: adding
a detector because it participated elsewhere in the campaign, or dropping one
without recording the selection, changes the data product used by the
calculation.
Files, rates, and calibrated strain
Section titled “Files, rates, and calibrated strain”Both releases provide HDF5 and GWF files on fixed 4096-second boundaries at 4096 Hz and 16384 Hz. A 4 kHz file therefore has 16,777,216 strain samples; a 16 kHz file has 67,108,864. Both carry 4,096 one-second points for each quality mask. The official O4 technical description gives the structural names:
| Quantity | HDF5 path | GWF channel suffix |
|---|---|---|
| Strain | /strain/Strain | GWOSC-4KHZ_R1_STRAIN or GWOSC-16KHZ_R1_STRAIN |
| Data quality | /quality/simple/DQmask | GWOSC-4KHZ_R1_DQMASK or GWOSC-16KHZ_R1_DQMASK |
| Injection state | /quality/injections/Injmask | GWOSC-4KHZ_R1_INJMASK or GWOSC-16KHZ_R1_INJMASK |
Prefix a GWF name with the detector code, for example
H1:GWOSC-16KHZ_R1_STRAIN. R1 identifies the released strain-channel
revision; it is not an event-catalog or parameter-inference revision.
The strain calibration is specified from 10 Hz through 5000 Hz, with data
quality less well studied above 2 kHz. The 4096 Hz stream was downsampled with
scipy.signal.decimate. Its anti-alias filter distorts the band near Nyquist,
so work around or above about 1700 Hz must use the 16384 Hz product. GWOSC
requested eight seconds of padding before downsampling each file, but padding
was not always available; small file-boundary effects can remain in the 4 kHz
stream.
The public strain has independently measured noise contributions removed, including calibration lines and power-mains harmonics. It is processed strain, not raw photodetector output and not necessarily the online channel used at the time of observation.
Calibration uncertainty is a separate input. The official
LIGO-T2500288-v5 record supplies
LIGO_O4a_cal_uncertainty.tgz and LIGO_O4b_cal_uncertainty.tgz. O4b LIGO
AR01 incorporates offline C01 recalibration from GPS 1409425408 through
1415647232 and otherwise uses C00. Virgo AR1U02 includes C01 where available
during the first month and otherwise C00; Virgo uncertainty products accompany
the O4b alternate-strain release. A
strain file without the applicable calibration-uncertainty product is not a
complete inference input.
Data-quality and injection selections
Section titled “Data-quality and injection selections”Each set bit in the nine-bit data-quality mask means that its named condition passes. Categories are analysis-specific: a passing Burst category does not establish a passing CBC category.
| Bit family | Exact bits | Meaning of a set bit |
|---|---|---|
| Availability | DATA | Calibrated strain is available for that second |
| Compact-binary coalescence | CBC_CAT1, CBC_CAT2, CBC_CAT3 | The corresponding CBC quality category passes |
| Bursts | BURST_CAT1, BURST_CAT2, BURST_CAT3 | The corresponding Burst quality category passes |
| Stochastic searches | STOCH_CAT1 | The stochastic category passes |
| Continuous waves | CW_CAT1 | The continuous-wave category passes |
Each set bit in the five-bit injection mask means that the named injection
class is absent: NO_CBC_HW_INJ, NO_BURST_HW_INJ,
NO_DETCHAR_HW_INJ, NO_CW_HW_INJ, and NO_STOCH_HW_INJ. GWOSC does not
subtract injected signals from strain. The official O4 injection
record reports no transient CBC, Burst,
Stochastic, or detector-characterization injections during the released O4a
and O4b observing times, but continuous-wave injections were active during
parts of the LIGO data. Select on NO_CW_HW_INJ when the calculation requires
their absence; do not infer injection state from the run name.
For a worked analysis, record the detector, GPS interval, archive ID, file
names, source and exported channels, sample rate, DATA selection,
analysis-specific category, injection-bit selection, calibration-uncertainty
file, and any boundary padding or conditioning. A generic statement such as
“we used O4 data” is not enough to reconstruct the input.
Verify retrieval against finite inventories
Section titled “Verify retrieval against finite inventories”GWOSC publishes per-file MD5 manifests. The table also records locally computed SHA-256 fingerprints of the complete manifest bytes retrieved on 2026-08-25. These local hashes are not publisher signatures and do not replace the MD5 values inside each manifest.
| Product and format | Official manifest | Entries | Detector counts | Retrieval-local manifest SHA-256 |
|---|---|---|---|---|
| O4a 4 kHz HDF5 | MD5 inventory | 7,774 | H1 3,901; L1 3,873 | 531912260ea45cd5265198f7a1990a8a89f69ffa98abca8a21930140e332b7ba |
| O4a 4 kHz GWF | MD5 inventory | 7,774 | H1 3,901; L1 3,873 | e47eeeed311125ae2ff577721a381cfaabe668f8844bce8e896e81c9228d17cf |
| O4a 16 kHz HDF5 | MD5 inventory | 7,774 | H1 3,901; L1 3,873 | 78dad17b8ae69cca0d312a2d8a410b38ede8b2954ed68d19a3ae606c948f91ba |
| O4a 16 kHz GWF | MD5 inventory | 7,774 | H1 3,901; L1 3,873 | ee8b0f26146ce7b662529f38261b73194c1e1f231318203c00f442e86299bc0b |
| O4b 4 kHz HDF5 | MD5 inventory | 13,432 | H1 3,735; L1 4,755; V1 4,942 | b3d6d9fa4f176df7b796995a28e70d7cb910a68d1785c602405180551a7eb232 |
| O4b 4 kHz GWF | MD5 inventory | 13,432 | H1 3,735; L1 4,755; V1 4,942 | 02cb70576a608bd22e35c7c3e26252f41e655ac9db80d36cb05888f46269d2d7 |
| O4b 16 kHz HDF5 | MD5 inventory | 13,432 | H1 3,735; L1 4,755; V1 4,942 | 0c914bd32a8674cace092d83d2b73b0493fce471cab202a0488d9ca9c924b24c |
| O4b 16 kHz GWF | MD5 inventory | 13,432 | H1 3,735; L1 4,755; V1 4,942 | 4cf2c0fc7dc8ee9fe9f93119466b37224ef5aa83eef2dcabadb8ecbc8755fb43 |
Manifest entry counts are file counts, not detector duty cycle, continuous coverage, or analysis-ready live time. Verify each downloaded file against its published MD5, then use the masks rather than treating the inventory as a science segment list.
Citing the data and using them in an analysis
Section titled “Citing the data and using them in an analysis”Both bulk releases are licensed under Creative Commons Attribution 4.0. Follow the official GWOSC acknowledgement instructions, cite the release paper, and cite each dataset DOI. For O4a, obtain descriptive citation metadata from the official release page or paper until DataCite/Crosscite supplies a record; DOI resolution alone does not provide the missing metadata.
GWTC-4.0 and GWTC-5.0 event, candidate, posterior, and population products are separate identities. Bulk strain availability neither establishes a detection nor supplies the likelihood, waveform model, calibration marginalization, priors, or selection function for an inference. Current event and population conclusions therefore require those additional analysis inputs.
Continue to Open Gravitational-Wave Data Products for segment selection, conditioning, calibration propagation, event-product distinctions, and reproducible workflows.
References
Section titled “References”- Gravitational Wave Open Science Center. O4 Data Set Technical Details, Hardware Injections, and Acknowledgement Policy. Official GWOSC documentation (2026), technical details, hardware injections, and acknowledgement policy, checked 2026-08-25.
- Gravitational Wave Open Science Center. O4a Data Release. Official GWOSC release record (2025), release page and dataset DOI, first released 2025-08-26; DOI resolution verified 2026-08-25, while DataCite/Crosscite citation metadata remained unavailable.
- Gravitational Wave Open Science Center. O4b Data Release. Official GWOSC release record (2026), release page, first released 2026-05-26; DOI.
- LIGO Scientific Collaboration, Virgo Collaboration, and KAGRA Collaboration. “Open Data from LIGO, Virgo, and KAGRA through the First Part of the Fourth Observing Run,” The Astrophysical Journal (2026), DOI.
- LIGO Scientific Collaboration, Virgo Collaboration, and KAGRA Collaboration. “Open Data from LIGO, Virgo, and KAGRA through the Second Part of the Fourth Observing Run,” arXiv:2605.27090v2 (2026), open preprint.
- LIGO Scientific Collaboration, Virgo Collaboration, and KAGRA Collaboration. O4 Calibration Uncertainty. LIGO Document Control Center, version 5 (2026), LIGO-T2500288-v5, checked 2026-08-25.