⚛️ phenomenology

New bounds on Axion-Like Particles in the Ultraviolet from Legacy Data

This paper utilizes legacy Hubble Space Telescope and International Ultraviolet Explorer data to establish new, significantly improved upper limits on axion-like particle-photon couplings, particularly ruling out values above $2.3 \times 10^{-12}~\mathrm{GeV}^{-1}$ for masses between 12.4 and 14.5 eV.

Original authors: Elisa Todarello

Published 2026-02-26

📖 5 min read🧠 Deep dive

CC0 1.0

Original authors: Elisa Todarello

Original paper dedicated to the public domain under CC0 1.0 (http://creativecommons.org/publicdomain/zero/1.0/). ✨ 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

The Big Picture: Hunting for Invisible Ghosts

Imagine the universe is filled with a mysterious, invisible fog called Dark Matter. Scientists have known this fog exists because it has gravity (it holds galaxies together), but they have never seen a single particle of it.

One leading theory suggests this fog is made of tiny, ghostly particles called Axion-Like Particles (ALPs). These particles are so light and shy that they usually just float around. However, this paper proposes a way to catch them: if an ALP is old enough, it might spontaneously "decay" (break apart) into two flashes of light (photons).

The author, Elisa Todarello, went on a detective hunt to find these specific flashes of light in the ultraviolet spectrum. She didn't build a new telescope; instead, she dug through the "attic" of space history, using old data from the Hubble Space Telescope and the International Ultraviolet Explorer (IUE) from the 1990s and 70s.

The Analogy: Listening for a Whisper in a Storm

Think of the universe as a very loud, busy concert hall.

The Background Noise: The "stars," "galaxies," and "sky glow" are like the loud music and chatter of the crowd.
The ALP Signal: The decay of an ALP would be a single, very specific musical note (a spectral line) played by a tiny violin in the back of the room. It's a very faint, specific frequency.

The challenge is that the "violin" is playing so quietly that it gets drowned out by the "crowd." The job of this paper was to listen to the old recordings of the concert hall, filter out the crowd noise, and see if that specific violin note was ever played.

The Detective Work: Two Different Tools

The author used two different "ears" (telescopes) to listen in different parts of the room:

The Hubble Space Telescope (HST):
- The Job: She looked at "blank sky" (empty patches of space far from stars).
- The Analogy: Imagine looking at a blank white wall to see if there's a faint, specific stain on it. If you see a stain, it must be coming from the wall itself (our galaxy's dark matter fog), not from a picture hanging on it.
- The Result: She found no stains. This set a limit on how "loud" the ALPs could be.
The International Ultraviolet Explorer (IUE):
- The Job: She looked at the Virgo Cluster, a massive group of galaxies, specifically focusing on a giant galaxy called M87.
- The Analogy: This is like looking at a giant, dense cloud of fog. If ALPs are the fog, looking at a huge cloud should make the "violin note" much louder because there are more ALPs to decay.
- The Result: This was the most sensitive test. She found no note here either, but because the "cloud" was so dense, she could rule out a much wider range of possibilities.

The "Smoking Gun" Discovery (and Correction)

One of the most exciting parts of the paper isn't just what she found, but what she fixed.

Recently, other scientists claimed to have found strong evidence for these particles using data from different telescopes (Swift and AstroSat). They thought they had found the "violin note."

Elisa showed that these scientists made a mistake in how they listened.

The Mistake: They used a "net" with very wide holes (broad filters) to catch the sound. They assumed that if they caught any sound in that net, it was the violin.
The Reality: Because the net was so wide, it caught a lot of background noise that looked like the violin note but wasn't.
The Fix: Elisa showed that when you use a "net" with the right size (accounting for the specific width of the ALP signal), the "evidence" disappears. The previous claims were actually just noise. It's like realizing you thought you heard a ghost whispering, but it was just the wind blowing through a wide-open window.

The Verdict: What Did We Learn?

The Limit: The author set a new, stricter rule for these particles. If ALPs exist in the mass range she tested (between 12.4 and 14.5 electron-volts), they must be even more "shy" than we thought. Their interaction with light must be at least 7 times weaker than previous limits allowed.
The Mass Range: She focused on a specific "weight" for these particles that hadn't been explored much before. It's like checking a specific shelf in a library that no one had looked at in years.
Old Data is Gold: This paper proves that you don't always need a brand-new, billion-dollar telescope to make discoveries. Sometimes, the best way to find new physics is to re-examine old data with better math and a fresh perspective.

Summary

Elisa Todarello used old telescope data to listen for a specific "ghostly" signal from Dark Matter. She didn't find the ghost, but she proved that if the ghost exists, it is much quieter than we thought. Along the way, she corrected a recent scientific claim, showing that what looked like a discovery was actually just a misunderstanding of how the data was processed. It's a victory for careful listening and precise math.

1. Problem Statement

Axion-Like Particles (ALPs) are compelling candidates for dark matter (DM). If they constitute the DM halo, they can spontaneously decay into two photons ( $a \to \gamma\gamma$ ), producing a narrow spectral line with a frequency $\nu \approx m_a/2$ . The paper focuses on the 12.4–14.5 eV mass range, a region that has remained relatively unexplored. Previous constraints in this window were limited, with the strongest existing bound ( $g_{a\gamma} \lesssim 1.6 \times 10^{-11} \text{ GeV}^{-1}$ ) derived from cosmic optical background anisotropies. The authors aim to significantly improve these limits by utilizing legacy ultraviolet (UV) spectroscopic data to search for the faint, narrow emission lines characteristic of ALP decay.

2. Methodology

Data Sources

The study utilizes legacy data from two space-based UV observatories:

Hubble Space Telescope (HST) Faint Object Spectrograph (FOS):
- Data: Blank sky observations (1991–1992) taken during the science verification phase (pre-COSTAR) using the G160L grating (red detector, 165–240 nm).
- Selection: 113 files covering 37 high-galactic-latitude pointings.
- Processing: Data were reduced using the calfos pipeline. The authors constructed a forward model to convert predicted ALP flux into detector counts, accounting for the instrument's Line Spread Function (LSF) and extended-source sensitivity. They stacked spectra by line of sight to maximize exposure (totaling ~8,600 seconds per pixel).
International Ultraviolet Explorer (IUE):
- Data: Observations of the Virgo Cluster (specifically M87) from 1978, 1981, and 1988, using both short-wavelength (SWP, 123–200 nm) and long-wavelength (LWP/LWR, 195–325 nm) spectrographs.
- Selection: 11 spectra (6 SWP, 5 LWP/LWR) extracted from the INES system (Newly Extracted Spectra) to ensure improved noise modeling and background subtraction compared to older NEWSIPS calibrations.
- Processing: The authors modeled the continuum emission of M87 using the asymmetrically reweighted Penalized Least Squares (arPLS) algorithm to isolate potential narrow spectral lines.

Signal Modeling

Decay Rate: The authors derived the spectral intensity ( $I_\nu$ ) from the Boltzmann equation, explicitly accounting for the ALP momentum distribution (Maxwell-Boltzmann) within the DM halo.
Flux Calculation: The expected flux density depends on the ALP-photon coupling ( $g_{a\gamma}$ ), the ALP mass ( $m_a$ ), and the "D-factor" (integrated DM density along the line of sight).
Halo Profiles:
- For HST (Milky Way halo): Tested Navarro-Frenk-White (NFW) profiles (including MilkyWayPotential2022 and variations from literature) and an Einasto profile.
- For IUE (Virgo halo): Used an NFW profile with parameters fitted to Gaia rotation curves, varying scale density and radius to estimate systematic uncertainties.
Statistical Analysis:
- HST: Used a Poissonian likelihood for low-count data. A second-order polynomial was used to model the smooth background. Limits were set using a profile likelihood ratio test statistic ( $q$ ).
- IUE: Used a Gaussian likelihood (appropriate for high-count, integrated video tube data) to compare observed flux density against the theoretical model.
- Calibration: Monte Carlo simulations were performed to calibrate the test statistic threshold ( $\Delta$ ) for the HST data, ensuring correct 95% Confidence Level (C.L.) coverage despite low photon counts (finding $\Delta = 2.93$ rather than the asymptotic 2.71).

3. Key Contributions

Significant Improvement in Bounds: The study establishes the most stringent limits to date for ALPs in the 12.4–14.5 eV mass range, improving upon previous constraints by a factor of seven.
Re-evaluation of Photometric Bounds: The authors critically re-analyzed recent claims based on photometric data from Swift-UVOT and AstroSat-UVIT (Ref. [22]). They demonstrated that those bounds were significantly overestimated (by factors of 20–60) due to:
- Incorrect interpretation of aperture sizes (using FWHM as radius instead of diameter).
- Failure to account for the fact that photometric bands are much broader than the intrinsic ALP line width, leading to a dilution of the signal flux density.
Legacy Data Utilization: The work demonstrates that "old" data from pre-COSTAR HST and IUE, often considered obsolete, remains highly valuable for searching for weak, narrow spectral lines when processed with modern statistical techniques and forward modeling.
Systematic Uncertainty Quantification: The paper provides a rigorous treatment of systematic errors, including DM halo profile variations, halo center position, velocity dispersion modeling, and continuum subtraction methods.

4. Results

HST FOS Results:
- Derived an upper limit of $g_{a\gamma} < 4.6 \times 10^{-12} \text{ GeV}^{-1}$ in the 12.4–14.5 eV range.
- The bound is robust across different DM halo profiles (variations < 5%).
IUE Results:
- Derived the most stringent limit: $g_{a\gamma} < 2.3 \times 10^{-12} \text{ GeV}^{-1}$ in the 12.4–14.5 eV range.
- This corresponds to excluding ALP lifetimes below $10^{25}$ seconds.
- The IUE limits are tighter than HST's despite lower sensitivity because the IUE spectral resolution better matches the intrinsic width of the ALP decay line in the Virgo cluster, resulting in a higher signal-to-noise ratio.
Comparison with Literature:
- The new IUE bound improves the previous best limit ( $1.6 \times 10^{-11} \text{ GeV}^{-1}$ ) by a factor of ~7.
- The re-analysis of Swift and AstroSat data shows that their reported constraints are invalid for narrow spectral line searches due to bandwidth mismatch; the corrected bounds are much weaker than previously claimed.

5. Significance

This paper represents a major advancement in the search for Axion-Like Particles in the ultraviolet regime. By leveraging legacy data and applying sophisticated forward modeling and statistical calibration, the authors have closed a significant gap in the parameter space for ALP dark matter.

The work highlights two critical lessons for future searches:

Spectral Resolution is Key: For narrow line searches, instruments with resolution matching the signal width (like IUE in this context) can outperform more sensitive instruments with broader resolution (like HST FOS in low-dispersion mode).
Modeling Precision: Accurate modeling of instrumental effects (aperture definitions, line-spread functions) and signal averaging over broad photometric bands is essential to avoid false exclusions or overly optimistic constraints.

These results set a new benchmark for ALP searches in the 10–20 eV mass range and provide a template for analyzing archival data from current and future UV observatories.