Independent validation of the RBH-1 hypervelocity SMBH candidate — testing whether the 62 kpc supersonic bow shock interpretation survives expanded methodological freedom.
RBH-1 Validation & Reanalysis is an independent validation study of the extraordinary RBH-1 linear feature (van Dokkum et al. 2025), interpreted as a 62 kpc supersonic bow shock from a hypervelocity supermassive black hole ejected at ~1000 km/s. This project contributes open tools, reproducible methods, and a complete Analysis-Ready Dataset (ARD) to the community discussion around this remarkable object.
- 📖 Validation Methodology - Complete methodology documentation
- 📊 Data Sources - HST and JWST observation details
- 🔬 Validation Criteria - What "validation" means for this project
- 🏗️ Repository Structure - Navigate the codebase
- 💻 Scripts - Phase-organized execution scripts
- 📦 Source Code - Reusable library modules
- 📝 Work Logs - Complete development history
- ✅ Validation Outputs - Data integrity and QA results
- 📓 Notebooks - Exploratory analysis
RBH-1 represents one of the most extraordinary astronomical claims in recent years: a supermassive black hole ejected from its host galaxy, now plowing through the circumgalactic medium at supersonic velocities and leaving a 62 kiloparsec ionized wake. If confirmed, this would be the first direct observation of gravitational wave recoil in action.
This project provides independent validation through two inseparable deliverables:
| Deliverable | Description |
|---|---|
| Validation Paper | Bayesian reanalysis testing whether the bow shock interpretation survives expanded methodological freedom |
| Analysis-Ready Dataset | Complete ARD on Zenodo enabling community reanalysis without climbing the compute wall |
Our team has a deep interest in extreme objects. Contributing to the scientific discussion around RBH-1—through open methods, transparent analysis, and reusable data products—is not just scientifically valuable, it's an honor. We approach this as neutral validation: either strengthening the extraordinary claim or identifying where additional work is needed. Both outcomes advance the field.
RBH-1 is a highly linear feature extending ~62 kpc from a compact source at redshift z ≈ 0.96. JWST NIRSpec IFU spectroscopy reveals:
- Kinematic discontinuity: ~600 km/s velocity change across ~1 kpc at the "tip"
- Shock-consistent line ratios: Emission diagnostics matching fast radiative shock models
- Surface brightness truncation: Factor >40× intensity drop at the tip over one resolution element
Van Dokkum et al. (2025) interpret these observations as a supersonic bow shock from an ejected SMBH. The competing "edge-on galaxy" hypothesis (Sánchez Almeida et al. 2023) proposes the feature is a serendipitously aligned, extremely thin bulgeless dwarf galaxy.
We validate the bow shock interpretation through:
- Expanded Bayesian inference on shock parameters with proper marginalization
- Stress-testing the background subtraction methodology
- Quantitative falsification of the edge-on galaxy alternative through forward modeling
Confirmation conditions — the interpretation is validated if:
- 95% CI of shock velocity from line ratios overlaps spatial gradient measurement
- Difference-of-pointings artifacts account for <20% of wake flux
- Inferred metallicity consistent with CGM (Z < 0.5 Z☉)
- Velocity discontinuity persists across all jackknife subsets
Tension conditions — tension is declared if:
- Marginalizing over magnetic parameter broadens v_s posterior to include virial velocities
- Standard Level 3 extraction significantly diminishes wake signal
- Jackknife analysis shows exposure/region dependence
Astronomical reanalysis has a hidden cost: the weeks of CPU time between downloading calibrated data and having usable posteriors. Every researcher who wants to revisit a result faces this compute wall, often regenerating artifacts that previous researchers already produced and discarded. This is institutionalized waste.
So, we're keeping our dataset, and we're sharing it.
The effort to package computed artifacts—chains, synthetic cubes, likelihood interfaces—is negligible compared to the effort to generate them. If sharing them saves even one researcher from repeating weeks of work, the cost-benefit is obvious.
This isn't a methodology innovation. It's just not throwing away what we already computed.
| Layer | Contents | Purpose |
|---|---|---|
| Likelihood Interface | Observed cube, weights, masks, instrument kernels | Drop-in forward model testing |
| Inference Layer | Full MCMC chains (~10M samples) | Statistical reanalysis, prior reweighting |
| Representative Sample | ~30 synthetic cubes + intrinsic velocity fields | Visual inspection, residual analysis |
| Validation Layer | Jackknife distributions, rejection surfaces | Robustness verification |
A researcher receiving the RBH-1 ARD can:
- Skip weeks of compute — Posterior samples already exist
- Test alternative priors — Reweight existing chains instead of rerunning
- Run their own forward model — Likelihood interface provides the contract
- Validate against new data — Future observations plug directly in
- Reproduce exactly — Full provenance chain from MAST to final measurement
| Product | Type | Purpose |
|---|---|---|
*_flc.fits |
CTE-corrected exposures | Custom error analysis |
*_drc.fits |
Drizzled mosaics | Astrometric alignment |
Filters: F200LP (~15ks), F350LP (~15ks) — 6 visits, 146 products, ~44 GB
| Product | Type | Purpose |
|---|---|---|
*_cal.fits |
Level 2b calibrated | Sensitivity testing |
*_s3d.fits |
Level 3 cubes | Primary extraction |
*_x1d.fits |
1D extracted | Sanity checks |
Configuration: G140M/F100LP, ~7ks total — 52 products, ~5 GB
Pre-computed grids from 3MdB database for radiative shock parameter inference.
rbh1-validation-reanalysis/
├── 🎨 assets/ # Hero images and diagrams
├── 💾 data/ # Data manifest and staging
│ ├── 01_raw/ # Raw observation pointers
│ ├── 02_reduced/ # Extracted spectra
│ ├── 03_inference/ # MCMC outputs
│ └── 04_ard/ # ARD package staging
├── 📚 docs/ # Methodology documentation
│ └── documentation-standards/ # Templates and conventions
├── 📓 notebooks/ # Exploratory analysis
├── 📦 rbh1-ard-v1/ # ARD release package
├── 💻 scripts/ # Phase-organized execution
│ ├── 00-ideation-and-setup/
│ ├── 01-data-acquisition/
│ ├── 02-standard-extraction/
│ ├── 03-cube-differencing/
│ ├── 04-noise-model/
│ ├── 05-kinematic-fitting/
│ ├── 06-mappings-inference/
│ ├── 07-robustness-tests/
│ ├── 08-galaxy-falsification/
│ └── 09-ard-materialization/
├── 🔧 src/ # Reusable library code
│ ├── extraction/ # Spectral extraction pipeline
│ ├── inference/ # MCMC and likelihood
│ ├── falsification/ # Galaxy hypothesis testing
│ └── visualization/ # Figure generation
├── ✅ validation/ # Data validation outputs
└── 📝 work-logs/ # Phase documentation
├── 00-ideation-and-setup/
├── 01-data-acquisition/
└── ...| Directory | Purpose | README |
|---|---|---|
assets/ |
Hero images and diagrams | — |
data/ |
Data manifest and staging areas | README |
docs/ |
Methodology and specifications | README |
notebooks/ |
Exploratory Jupyter analysis | README |
rbh1-ard-v1/ |
ARD release package | README |
scripts/ |
Phase-organized execution scripts | README |
src/ |
Reusable library modules | README |
validation/ |
Data integrity and QA outputs | README |
work-logs/ |
Phase worklogs and development history | README |
This project uses dedicated VMs from the Proxmox Astronomy Lab cluster.
| Node | CPU | RAM | GPU | Role |
|---|---|---|---|---|
| proj-gpu01 | AMD 5950X (16c) | 48 GB | NVIDIA A4000 16GB | GPU inference, task broker |
| proj-cpu01 | Intel 12900K (12c) | 48 GB | — | MCMC walker pool |
| proj-cpu02 | Intel 12900K (12c) | 48 GB | — | MCMC walker pool |
| proj-cpu03 | Intel 12900K (12c) | 48 GB | — | MCMC walker pool |
The falsification campaign uses a Producer-Consumer pattern optimized for heterogeneous hardware:
- CPU Nodes (Producers): Propose parameters, compute galactic dynamics via
galpy, apply Toomre Q stability filter - GPU Node (Consumer): JAX-based 3D cube generation, PSF/LSF convolution, likelihood evaluation
- DragonFlyDB: High-throughput task broker (ephemeral)
- PostgreSQL: ACID-compliant chain storage (durable)
| Phase | Description | Status |
|---|---|---|
| Phase 0 | Ideation and Setup | ✅ Complete |
| Phase 1 | Data Acquisition + Validation | ✅ Complete |
| Phase 2 | Standard Extraction Baseline | ⏳ Pending |
| Phase 3 | Cube-Space Differencing | ⏳ Pending |
| Phase 4 | Empirical Noise Model | ⏳ Pending |
| Phase 5 | Tied Kinematic Fitting | ⏳ Pending |
| Phase 6 | MAPPINGS V Inference | ⏳ Pending |
| Phase 7 | Robustness Tests | ⏳ Pending |
| Phase 8 | Edge-on Galaxy Falsification | ⏳ Pending |
| ARD | Dataset Materialization | ⏳ Pending |
Current Milestone: Phase 01 complete. HST and JWST data acquired (198 files, ~45 GB), validated against van Dokkum et al. (2025) claims, and QA visualizations generated. All validation checks pass. See Phase 01 Worklog for details.
Next: Phase 02 — Standard extraction baseline from JWST S3D cubes and HST DRC photometry.
ARD Target: v1.0 release alongside paper submission
This project is part of the radioastronomyio research program:
| Project | Description |
|---|---|
| proxmox-astronomy-lab | Platform documentation, VM inventory, network architecture |
| desi-cosmic-void-galaxies | Galaxy populations in cosmic voids using DESI DR1 |
| desi-quasar-outflows | AGN outflow spectral fitting and Cloudy modeling |
| desi-qso-anomaly-detection | ML anomaly detection for quasar spectra |
| Resource | Link |
|---|---|
| van Dokkum et al. (2025) | arXiv preprint |
| HST GO-17301 | MAST Archive |
| JWST GO-3149 | MAST Archive |
| MAPPINGS V | 3MdB Database |
This project is committed to full transparency and reproducibility:
| Artifact | Platform | Contents |
|---|---|---|
| Code | GitHub | All analysis scripts, pipeline code |
| Work Logs | GitHub | Decision documentation, methodology notes |
| ARD | Zenodo | Complete Analysis-Ready Dataset with DOI |
| Paper | arXiv | Preprint with full methodology |
The ARD will include everything needed to reproduce our results or test alternative hypotheses without access to HPC resources.
This project implements systematic AI-human collaboration through transparent documentation and open science practices.
| Principle | Implementation |
|---|---|
| Documentation First | All methodology documented before execution |
| Reproducibility | Full provenance chain, versioned artifacts |
| Open Science | Code, logs, and data publicly released |
| Neutral Stance | Validation, not advocacy |
- Review the Validation Specification
- Check the Work Logs for current status
- See phase-specific READMEs for implementation details
This project is licensed under the MIT License — see the LICENSE file for details.
- van Dokkum et al. — Original RBH-1 discovery and interpretation
- STScI/MAST — HST and JWST data archive
- MAPPINGS V Team — Shock model grids
- Proxmox Astronomy Lab — Compute cluster
- JAX/Google — GPU-accelerated inference
- ArviZ — Bayesian analysis ecosystem
Contributing to the discussion of extraordinary objects | Part of radioastronomyio
Last Updated: December 23, 2025 | Current Phase: Phase 02 — Standard Extraction