---
license: cc-by-4.0
pretty_name: "ICRF3 Celestial Reference Frame"
language:
  - en
description: "The third International Celestial Reference Frame (ICRF3) is the fundamental coordinate reference frame for astronomy, adopted by the International Astronomical Union in 2018. It is defined by extraga"
task_categories:
  - tabular-classification
tags:
  - space
  - icrf
  - reference-frame
  - astrometry
  - quasar
  - vlbi
  - open-data
  - tabular-data
  - parquet
size_categories:
  - 1K<n<10K
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/icrf3_reference_frame.parquet
    default: true
---

# ICRF3 Celestial Reference Frame


<div align="center">
  <img src="banner.jpg" alt="The gamma-ray sky as seen by NASA's Fermi telescope" width="400">
  <p><em>Credit: NASA/DOE/Fermi LAT Collaboration</em></p>
</div>


*Part of a [dataset collection](https://huggingface.co/collections/juliensimon/astronomy-datasets-69c24caf2f17e36128946743) on Hugging Face.*

## Dataset description

The third International Celestial Reference Frame (ICRF3) is the fundamental coordinate reference frame for astronomy, adopted by the International Astronomical Union in 2018. It is defined by extragalactic radio sources (primarily quasars) observed by Very Long Baseline Interferometry (VLBI). This dataset contains structure parameters from Xu et al. (2019) that quantify each source's compactness and positional stability.

The ICRF is the realization of the International Celestial Reference System (ICRS) at radio wavelengths. ICRF3 is based on nearly 40 years of VLBI observations and provides the most accurate positions of extragalactic objects, with median positional uncertainties of ~30 microarcseconds for the defining sources.

The ICRF is conceptually the modern replacement for the FK5 optical fundamental star catalog. While FK5 was limited by the proper motions and parallaxes of nearby stars, the ICRF uses extremely distant quasars whose apparent motions are negligible, providing a quasi-inertial reference frame tied to the large-scale structure of the universe. Source structure — the extended radio jets that cause apparent position shifts — is the dominant systematic error in VLBI astrometry. This dataset includes structure index parameters that quantify each source's compactness, which is critical for selecting calibrators for VLBI observations and for space geodesy applications.


## Schema

| Column | Type | Description | Sample | Null % |
|--------|------|-------------|--------|--------|
| `iers_name` | object | Official IERS source designation in B1950 sexagesimal format (HHMM+DDd, e.g. '0002-478'); this is the authoritative name used in geodetic VLBI scheduling, Earth orientation monitoring, and spacecraft navigation | 0305+039 | 0.0% |
| `closure_amplitude_rms_n` | float64 | RMS of closure amplitude residuals for narrow-field VLBI imaging (dimensionless); measures source compactness — lower values indicate more point-like sources with more stable astrometric positions | 0.41 | 0.0% |
| `closure_amplitude_rms_b` | float64 | RMS of closure amplitude residuals for broad-field imaging (dimensionless); sensitive to extended jet structure on longer baselines | 0.44 | 0.0% |
| `closure_amplitude_rms_u` | float64 | RMS of closure amplitude residuals for uniform weighting (dimensionless); intermediate sensitivity between narrow and broad-field | 0.47 | 0.0% |
| `n_closure_amplitude` | Int64 | Number of closure amplitude measurements used in the structure analysis; more measurements yield more reliable structure diagnostics | 8110 | 0.0% |
| `closure_phase_rms_n_deg` | float64 | RMS of closure phase residuals for narrow-field imaging in degrees; closure phase is immune to antenna-based errors, making it a robust measure of source asymmetry | 18.8 | 0.0% |
| `closure_phase_rms_b_deg` | float64 | RMS of closure phase residuals for broad-field imaging in degrees; elevated values indicate resolved jet structure | 19.4 | 0.0% |
| `closure_phase_rms_u_deg` | float64 | RMS of closure phase residuals for uniform weighting in degrees | 22.8 | 0.0% |
| `n_closure_phase` | Int64 | Number of closure phase measurements used in the structure analysis | 1639 | 0.0% |
| `n_sessions` | Int64 | Number of VLBI observing sessions in which this source was observed; well-observed sources have hundreds of sessions over decades | 36 | 0.0% |
| `n_observations` | Int64 | Total number of individual VLBI observations (delay measurements) across all sessions; high counts indicate heavily-used calibrator sources | 1072 | 0.0% |
| `flag_icrf2` | object | Source classification flag in ICRF2: 'D' = defining, 'N' = non-defining, 'S' = special handling; null if not in ICRF2 | O | 0.0% |
| `flag_icrf3` | object | Source classification flag in ICRF3: 'D' = defining (303 sources with the most stable positions), 'S' = special handling (688 sources with extended structure); null for non-defining sources | D | 0.0% |
| `icrf2_structure_index` | Int64 | Source structure index from ICRF2 (1-4 scale); 1 = very compact, 4 = highly extended; used for calibrator selection in VLBI observations | 1 | 0.0% |
| `ra_deg` | float64 | Right ascension of the extragalactic radio source in degrees, ICRS J2000.0 epoch; range 0-360; defining sources carry positional accuracies of ~30 microarcseconds | 47.109265834 | 0.0% |
| `dec_deg` | float64 | Declination of the extragalactic radio source in degrees, ICRS J2000.0; range -90 to +90; southern hemisphere coverage is sparser due to fewer southern VLBI stations | 4.110916884 | 0.0% |

## Quick stats

- **3,417** ICRF3 radio sources
- **283** defining sources (highest positional stability)
- **0** special-handling sources (extended jet structure)
- Median observing sessions per source: **4**
- Most-observed source: **439,113** individual VLBI measurements

## Usage

```python
from datasets import load_dataset

ds = load_dataset("juliensimon/icrf3-reference-frame", split="train")
df = ds.to_pandas()

# Defining sources — the foundation of the celestial reference frame
defining = df[df["flag_icrf3"] == "D"]
print(f"{len(defining)} ICRF3 defining sources")

# Sky distribution of defining vs. non-defining sources
import matplotlib.pyplot as plt
fig, ax = plt.subplots(figsize=(12, 6))
other = df[df["flag_icrf3"] != "D"]
ax.scatter(other["ra_deg"], other["dec_deg"], s=1, alpha=0.3, label="Non-defining")
ax.scatter(defining["ra_deg"], defining["dec_deg"], s=5, c="red", label="Defining")
ax.set_xlabel("RA (deg)")
ax.set_ylabel("Dec (deg)")
ax.invert_xaxis()
ax.legend()
ax.set_title("ICRF3 All-Sky Distribution")
plt.tight_layout()
plt.show()
```

## Data source

https://vizier.cds.unistra.fr/viz-bin/VizieR-3?-source=J/ApJS/242/5

## Related datasets

- [juliensimon/pulsar-catalog](https://huggingface.co/datasets/juliensimon/pulsar-catalog)

- [juliensimon/open-star-clusters](https://huggingface.co/datasets/juliensimon/open-star-clusters)

## Citation

```bibtex
@dataset{icrf3_reference_frame,
  title = {ICRF3 Celestial Reference Frame},
  author = {juliensimon},
  year = {2026},
  url = {https://huggingface.co/datasets/juliensimon/icrf3-reference-frame},
  publisher = {Hugging Face}
}
```

## License

[CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/)