🔭 astrophysics

Multifrequency Synthesis via CHIBI: Colorful Hierarchical Interferometric Bayesian Imaging

This paper introduces CHIBI, a hierarchical Bayesian imaging framework that leverages multifrequency synthesis and synchrotron radiation models to reconstruct high-fidelity images and spectral maps of radio sources, demonstrating its effectiveness on MOJAVE VLBA data and future Event Horizon Telescope observations.

Original authors: Erandi Chavez, Paul Tiede, Sara Issaoun, Michael D. Johnson, Dominic Pesce, Yuh Tsunetoe, Daniel C. M. Palumbo

Published 2026-06-04

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Erandi Chavez, Paul Tiede, Sara Issaoun, Michael D. Johnson, Dominic Pesce, Yuh Tsunetoe, Daniel C. M. Palumbo

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine trying to understand a complex, colorful painting, but you can only see it through a series of foggy, single-colored glasses. Sometimes you look through red glasses, sometimes blue, sometimes green. Each time, the picture looks a bit different, and the details are blurry. If you try to piece together the "true" picture by looking at each color separately and then gluing the results together, you might end up with a mismatched, confusing mess.

This is the challenge astronomers face when studying radio waves from space. Radio telescopes don't take a single "photo" like a camera; they collect scattered data points (called "visibilities") that need to be pieced together to form an image. Traditionally, astronomers have looked at different radio frequencies (colors) one by one, reconstructed an image for each, and then tried to compare them.

Enter CHIBI: The "Colorful" Detective

This paper introduces a new method called CHIBI (Colorful Hierarchical Interferometric Bayesian Imaging). Think of CHIBI not as a photographer taking separate snapshots, but as a detective who looks at the entire crime scene all at once, using clues from every color simultaneously to solve the mystery.

Here is how it works, using simple analogies:

1. The "Recipe" for Radio Light

Most of the radio light from black holes and jets comes from a process called synchrotron radiation. The authors realized that this light follows a predictable "recipe" across different colors (frequencies). It's like a song that changes pitch but keeps the same melody.

The Old Way: Listen to the song at 800Hz, write down the notes. Then listen at 1200Hz, write down the notes. Then try to guess how they relate.
The CHIBI Way: Assume the song follows a specific mathematical rule (a power law). Listen to all the frequencies at once and reconstruct the entire song in one go, ensuring the melody makes sense across all pitches.

2. The "Foggy Window" Problem

Radio telescopes are like looking at a painting through a very foggy window. The "fog" is caused by the atmosphere and the telescope's own imperfections.

The Old Way: You try to clean the window for the red picture, then clean it again for the blue picture. Sometimes you clean it too much, sometimes not enough, and the pictures don't line up perfectly.
The CHIBI Way: CHIBI treats the "fog" (instrument errors) and the "painting" (the actual image) as a single team. It figures out the fog and the picture together. Because it uses data from all colors at once, it has more clues to figure out what the fog is doing, leading to a much clearer picture.

3. Filling in the Blanks (Super-Resolution)

Imagine you are trying to guess the shape of a hidden object based on a few scattered shadows. If you only have shadows from one angle, you might guess it's a circle. If you have shadows from many angles, you can see it's actually a star.

Radio telescopes often have "holes" in their data (missing angles).
CHIBI uses the fact that the object looks similar (but slightly different) across all colors to fill in those missing holes. It effectively "super-resolves" the image, seeing details that are smaller than the telescope's usual limit, but only if the data strongly supports it. It doesn't just guess; it calculates the probability of what is really there.

4. The "Map" of Colors (Spectral Index)

One of the most exciting results of this method is the creation of Spectral Index Maps.

Imagine a weather map, but instead of showing temperature, it shows the "color" of the radio light at every single point on the object.
In the paper, they applied this to OJ287 (a black hole with a jet) and M87* (the famous black hole imaged by the Event Horizon Telescope).
The Result: They could clearly see that the "core" of the jet was one color (positive spectral index) and the "tail" was a different color (negative spectral index). This helps scientists understand the physics of how the jet is launched and how the particles are moving.

Real-World Tests in the Paper

The authors tested CHIBI in four different scenarios:

Real Data (OJ287 & PKS 1424+240): They looked at real observations from the VLBA (a giant network of radio telescopes). CHIBI produced sharper images and clearer "color maps" than the standard methods used for decades.
Current EHT (M87):* They simulated data for the current Event Horizon Telescope. They showed that for the highest frequencies (345 GHz), the data is so sparse that you cannot make an image using old methods. CHIBI, by combining it with lower frequencies, successfully created an image and a color map.
Future Space Telescopes (BHEX): They simulated a future mission where a telescope in space works with ground telescopes. CHIBI showed that this setup would allow for incredibly detailed maps of the black hole's jet.
Next-Gen EHT (ngEHT): They simulated a future, denser array of telescopes. Here, the biggest win wasn't just a sharper picture, but a much more accurate "color map," revealing the structure of the jet with high precision.

The Catch (Degeneracies)

The paper also admits a limitation. Sometimes, the math allows for a "shift" in the picture. It's like if you have a map of a city, but you don't know exactly where "North" is, or you don't know the exact total size of the city.

CHIBI can tell you the shape of the city and the relative colors perfectly.
However, without extra information (like knowing the exact total brightness of the object beforehand), the "color map" might be shifted up or down by a constant amount. The authors explain this clearly and show how to fix it if you have that extra data.

Summary

In short, CHIBI is a new, smarter way to build radio images. Instead of building a puzzle piece by piece for each color, it builds the whole puzzle at once, using the rules of physics to guide the pieces into place. This results in sharper images, better "color maps" of black hole jets, and the ability to see details that were previously hidden in the noise. It turns a blurry, multi-colored mess into a clear, scientifically rich picture of the universe's most extreme objects.

Technical Summary: Multifrequency Synthesis via CHIBI

Problem Statement
Astronomical radio sources are inherently chromatic; understanding their physical origins requires analyzing their spectral evolution rather than relying on single-frequency measurements. While multifrequency observing has become standard for modern and next-generation interferometers (spanning orders of magnitude in frequency), standard imaging techniques often reconstruct images at individual frequencies separately before comparing them. This approach faces challenges including inhomogeneous baseline coverage, varying angular resolution across frequencies, and the inability to enforce physical relationships between frequencies during the reconstruction process. Furthermore, traditional methods like CLEAN rely on deconvolution and blurring, which can limit resolution and obscure fine spectral structures. There is a need for a unified framework that simultaneously reconstructs images across multiple frequencies, leveraging spectral priors to improve imaging fidelity and constrain spectral parameters.

Methodology
The authors introduce CHIBI (Colorful Hierarchical Interferometric Bayesian Imaging), a new method for Multifrequency Synthesis (MFS) implemented within the Comrade.jl software repository. CHIBI operates within a hierarchical Bayesian inference framework using Monte Carlo methods (specifically the No-U-Turn Sampler, NUTS) to sample the full posterior distribution of image parameters.

Spectral Model: The core of the method is a spectral parametrization grounded in synchrotron radiation physics. The image intensity $I(x, y; \nu)$ at any frequency $\nu$ is modeled as a power-law expansion in log-log space relative to a reference image $I_0$ at frequency $\nu_0$ :
$\log I(x, y; \nu) = \log I_0(x, y; \nu_0) + \sum_{n=1}^{N} \alpha_n(x, y) \log^n \left( \frac{\nu}{\nu_0} \right)$
Here, $\alpha_1$ represents the spectral index map and $\alpha_2$ the spectral curvature. These coefficients are treated as spatially varying pixel values.
Image Priors: The model utilizes a zero-mean Gaussian Markov Random Field (GMRF) prior for image fluctuations and spectral maps. This prior assumes a power-law power spectrum that suppresses structures smaller than a correlation length $\rho$ , effectively acting as a regularization that allows for "super-resolution" (recovering structure below the nominal beam size) only when strongly supported by the data.
Instrumental Modeling: The framework can jointly model station-based gain factors (amplitude and phase) alongside the image structure. This accounts for residual calibration errors and atmospheric corruption, folding calibration uncertainties into the posterior.
Degeneracy Handling: The authors address fundamental degeneracies in VLBI imaging, such as unconstrained total flux and image centroids. In multifrequency synthesis, these manifest as unknown offsets in spectral maps and relative centroids between frequencies. The method mitigates these by:
1. Fixing total flux for the reference image (or using priors).
2. Using a GMRF prior with a flat mean for spectral maps to prefer aligned structures.
3. Explicitly parameterizing spectral offsets in specific tests to allow the sampler to explore these values.
4. Centering posterior samples based on the brightest pixel or centroid to correct for drift.

Key Contributions

Hierarchical MFS Framework: The extension of the HIBI (Hierarchical Interferometric Bayesian Imaging) approach to multifrequency synthesis, enabling the simultaneous reconstruction of total intensity and spectral index maps.
Physical Spectral Parametrization: The implementation of a spectral model based on synchrotron power laws, allowing the spectral index and curvature to vary spatially across the image.
Validation and Application: The method is validated against synthetic Gaussian data and applied to four distinct regimes:
- Real Data: MOJAVE VLBA observations of blazars OJ287 (8.1/12.1 GHz) and PKS 1424+240 (15/23/43 GHz).
- Synthetic Data: Simulations of M87* using the current Event Horizon Telescope (EHT), the next-generation EHT (ngEHT), and the proposed Black Hole Explorer (BHEX) space-ground interferometer.

Results

Validation: CHIBI successfully recovered ground-truth total intensity and spectral index maps from synthetic data, outperforming single-frequency imaging in both accuracy and precision. The method demonstrated that MFS transfers information between frequencies, improving constraints on high-frequency structures even where sampling is sparse.
MOJAVE Observations: For OJ287 and PKS 1424+240, CHIBI produced images consistent with published CLEAN results but with significantly higher effective resolution (super-resolution). The spectral index maps were markedly improved, clearly resolving the positive spectral index core and negative spectral index jet structure predicted by the Blandford & Königl model. The method demonstrated that MFS aligns structure across frequencies, reducing artifacts in spectral maps.
EHT (Current): For the current EHT, the 345 GHz coverage is too sparse for single-frequency imaging. MFS enabled the reconstruction of a 345 GHz image and a 230–345 GHz spectral index map, which would be impossible otherwise.
BHEX and ngEHT: Simulations for BHEX (space VLBI) and ngEHT showed that MFS significantly improves the recovery of jet structures and spectral index maps. For ngEHT, the dense coverage combined with MFS alignment produced exceptionally clean spectral index maps, highlighting the synergy between instrument design and Bayesian synthesis.

Significance and Claims
The paper claims that CHIBI provides a statistically grounded framework for multifrequency synthesis that yields scientifically richer results than traditional approaches. By enforcing physical relationships between frequencies, the method produces higher-fidelity images and more precise spectral index maps. The authors emphasize that the improved results are data-driven, with the GMRF prior allowing super-resolution only when justified by the likelihood.

The work highlights the critical role of multifrequency synthesis for future instruments like the ngEHT and BHEX, where simultaneous or quasi-simultaneous wide-bandwidth observations will be standard. The authors note that while current implementations address flux and centroid degeneracies through priors and post-processing, future developments involving correlated gain modeling (e.g., via Frequency Phase Transfer) could further break degeneracies and enable Bayesian multifrequency astrometry. The method is presented as a modular tool capable of handling various VLBI arrays and is implemented in the open-source Comrade.jl repository.