🔭 astrophysics

Significant or Not? The Impact of Randomisation During Data Reduction on Confirming a New Pulsating Ultraluminous X-ray Source Candidate in Centaurus A

This paper reports the discovery of a new candidate pulsating ultraluminous X-ray source in Centaurus A with a unique soft spectrum, while critically highlighting how randomization in XMM-Newton's data reduction software introduces significant uncertainty in marginal pulsation signals, potentially leading to false positives or negatives in future PULX searches.

Original authors: Amy H. Knight, Timothy P. Roberts, Callum Potter, Alistair T. Pagan, Dominic J. Walton

Published 2026-05-15

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Amy H. Knight, Timothy P. Roberts, Callum Potter, Alistair T. Pagan, Dominic J. Walton

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

The Big Picture: Finding a Cosmic Heartbeat

Imagine the universe is a giant, noisy concert hall. Most of the time, it's just a chaotic roar of sound. But every now and then, you hear a distinct, rhythmic drumbeat: thump-thump, thump-thump. In astronomy, this "drumbeat" is called a pulsation, and it usually comes from a dead star (a neutron star) spinning incredibly fast.

The scientists in this paper were looking for a new "drummer" in a galaxy called Centaurus A. They found a candidate that seemed to be beating at a steady rhythm, but they had a major problem: they weren't sure if the beat was real or just a trick of the light caused by their own tools.

The Discovery: A Faint, Flickering Star

The team found a new object, let's call it J1325.

The Behavior: J1325 is like a shy celebrity. It usually hides in the dark (quiescence), but every few years, it bursts onto the stage for a short time (an outburst). In 2014, it was very bright for about 8 months.
The Mystery: When they looked at the data from that 2014 burst, they saw a signal suggesting the star was spinning 1.27 times every second. This would make it a PULX (Pulsating Ultraluminous X-ray source)—a super-bright star powered by a neutron star.
The Oddity: Usually, these super-bright pulsating stars are "hard" and energetic (like a high-pitched, sharp whistle). J1325, however, sounded "soft" (like a low, muffled hum). This made the scientists suspicious. Was it really a pulsating star, or just a regular bright star that looked weird?

The Problem: The "Random Shuffle" Glitch

This is where the paper gets really interesting. The scientists used a standard computer program (called XMM-SAS) to clean up the raw data from the telescope, much like using a photo editor to remove noise from a blurry picture.

However, they discovered a hidden quirk in the software: it has a "random shuffle" button.

The Analogy: Imagine you are trying to count the number of people in a crowded room by looking at a video. To make the video clearer, the software randomly shifts the position of every person by a tiny, invisible amount (less than a pixel) to smooth out the image.
The Glitch: Because the software uses the current time of day to decide how to shuffle the pixels, if you run the same video through the software twice (even just minutes apart), the "shuffled" result is slightly different.
The Consequence: When the scientists ran their search for the heartbeat once, they found a signal. When they ran it again with a slightly different "shuffle," the signal got weaker or disappeared. When they ran it a third time, it got stronger again.

It was like trying to hear a whisper in a windstorm. Sometimes the wind died down, and you heard the whisper clearly. Other times, the wind blew, and the whisper vanished. The "wind" was the random shuffling of the computer code.

The Solution: The "2,500 Times" Test

To solve this, the scientists didn't just look at the data once. They ran the entire process 2,500 times, letting the computer shuffle the data randomly each time.

The Result: They found that the 1.27 Hz heartbeat appeared consistently across these thousands of attempts. It wasn't just a fluke of one specific "shuffle."
The Catch: Even though the heartbeat was there, its "loudness" (statistical significance) varied wildly. Sometimes it was a clear shout; other times, it was a faint murmur that barely passed the threshold of detection.

The Conclusion: "Probably Real, But We Need Proof"

The paper concludes with a cautious "Yes, but..."

It's likely a real pulsating star: The signal has the right shape (a smooth wave), it appears consistently when you look hard enough, and it has "harmonics" (echoes of the beat) that real stars produce.
It's weird: It's much softer (cooler) than other known pulsating stars, which breaks the usual rules of how these stars behave.
It's not 100% confirmed: Because the signal is "marginal" (on the edge of being detectable) and they haven't found the same heartbeat in a second observation, they can't say for certain it's a new type of star yet. It might be a false alarm caused by the random noise of the universe.

The Bigger Lesson for Science

The most important part of this paper isn't just about this one star. It's a warning to all astronomers.

The authors realized that many past discoveries might have been affected by this same "random shuffle" glitch. If a scientist runs a search once and gets a "lucky" strong signal, they might publish it as a discovery. If they run it again and get a weak signal, they might ignore it.

The Takeaway: When looking for faint signals in space, you can't just look once. You have to run the test thousands of times to see if the signal is a real heartbeat or just a random glitch in the software. This paper introduces a new, more rigorous way to listen to the universe, ensuring we don't mistake static for music.

Problem
The identification of Neutron Star (NS) accretors within the population of Ultraluminous X-ray sources (ULXs) is a critical challenge in high-energy astrophysics. While the discovery of coherent X-ray pulsations in several ULXs has confirmed the existence of Pulsating ULXs (PULXs), confident identification remains difficult. It typically requires datasets with high photon counts to overcome low pulse fractions, and pulsations are often transient. Furthermore, existing PULX candidates generally exhibit hard X-ray spectra below 10 keV, potentially biasing searches against softer sources. A significant, often overlooked technical issue complicates these searches: the intrinsic randomisation employed by the XMM-Newton Science Analysis Software (XMM-SAS) during data reduction. This randomisation, driven by a pseudo-random seed based on local clock time, introduces variability in event lists between reduction iterations, potentially causing marginal pulsation signals to fluctuate in significance or disappear entirely, leading to false negatives or inconsistent results.

Methodology
The authors report the discovery of a new PULX candidate, 4XMM J132542.2–425943 (J1325), in NGC 5128 (Centaurus A), utilizing a novel, rigorous two-stage search procedure designed to account for data reduction variability.

Source Selection and Data Reduction: The authors selected J1325 from the Walton et al. (2022) ULX catalogue based on its transient nature and count-rich datasets. They re-analysed archival XMM-Newton and Chandra observations. Crucially, instead of a single data reduction, they performed $n=2500$ independent reduction iterations of the XMM-Newton Observation Data Files (ODFs). This approach was necessitated by the discovery that XMM-SAS's epproc task introduces randomisation in sub-pixel event positions, causing events to shift in or out of extraction regions between runs.
Two-Stage Pulsation Search:
- Stage One (Broad Search): Using the hendrics package, they performed Fourier domain acceleration searches (both standard and with interbinning) across a wide frequency band (0.01–5 Hz) for all 2500 iterations. This generated distributions of candidate frequencies and powers to identify consistent signals and filter out false positives arising from randomisation.
- Stage Two (Targeted Search): Promising candidates identified in Stage One were subjected to targeted $Z^2_1$ , $Z^2_2$ , and $Z^2_3$ searches (Rayleigh tests with 1, 2, and 3 harmonics) in progressively narrower frequency bands (0.5 Hz, 0.1 Hz, and 0.02 Hz). These searches were conducted on subsets of the 2500 iterations to assess the stability of the signal's power and occurrence rate.
Spectral Analysis: The authors performed spectral fitting on the XMM-Newton and Chandra data using absorbed power-law and multi-component thermal models to characterise the source's spectral state during its outburst.

Key Contributions

Discovery of a Soft PULX Candidate: The identification of J1325 as a candidate PULX with a soft X-ray spectrum ( $\Gamma \sim 2.7$ ), contrasting with the typically hard spectra of known PULXs.
Methodological Innovation: The development and demonstration of a robust search methodology that explicitly accounts for the intrinsic randomisation in XMM-SAS data reduction. The authors show that relying on a single reduction iteration can lead to misleading conclusions regarding the significance of marginal signals.
Quantification of Uncertainty: The paper quantifies how the random seed variable affects the measured power of a marginal signal, demonstrating that a single candidate frequency can yield powers ranging from $\sim 15$ to $\sim 50$ depending on the reduction iteration and search parameters.

Results

Pulsation Detection: In the XMM-Newton observation ObsID 0724060801 (February 2014), the authors detected coherent X-ray pulsations at a frequency of $1.269$ Hz (approx. $1.27$ Hz). This signal was consistent across the 2500 reduction iterations, appearing in all search types (acceleration and $Z^2$ ) and energy bands where the source was bright.
Signal Characteristics: The candidate exhibits a sinusoidal folded pulse profile with a pulsed fraction (PF) of approximately $15-17\%$ and a spin-up rate ( $\dot{f}$ ) of $\sim 4 \times 10^{-9}$ Hz s $^{-1}$ . The signal also shows the presence of second and third harmonics.
Spectral Properties: Unlike typical PULXs, J1325 displays a soft spectrum during the pulsating state, best described by an absorbed power law with a photon index $\Gamma \sim 2.7$ or a two-component thermal model. The source flux exceeded $10^{-12}$ erg cm $^{-2}$ s $^{-1}$ at peak, corresponding to a luminosity of $\sim 1.7 \times 10^{39}$ erg s $^{-1}$ .
Light Curve Behavior: J1325 is a transient source with a major $\sim 8$ -month outburst in 2014 and two subsequent, less luminous detections in 2017 and 2024. No corroborating pulsation signal was found in the earlier XMM-Newton observation (ObsID 0724060701) or in other available data, likely due to the transient nature of the pulsations or the source being fainter/softer at that time.
Randomisation Impact: The study demonstrates that the randomisation in XMM-SAS can generate widespread false positives and false negatives. A signal that appears significant in one reduction iteration may fall below detection thresholds in another, solely due to the random seed value.

Significance
The paper claims that J1325 is a "good but unconfirmed" PULX candidate. While the pulsation signal is robust against variations in search parameters and data reduction iterations, the lack of a second, independent detection prevents a definitive claim of a pulsating nature. The authors suggest two interpretations: either J1325 is a genuine, spectrally soft PULX that challenges current models of PULX emission (which typically assume hard spectra), or the signal is a false positive induced by a random alignment of photon arrival times.

The primary significance of the work lies in its methodological warning: the intrinsic randomisation in XMM-Newton data reduction introduces considerable uncertainty in the strength of marginal pulsation signals. The authors argue that previous searches relying on single data reduction iterations may have incorrectly discarded viable candidates (false negatives) or reported false positives. They conclude that future PULX searches must employ repetitive reduction strategies to establish the reproducibility of a signal, thereby improving the reliability of identifying NS accretors in the ULX population.