The V30 QUICK4 analysis of GW230927_153832 found credible post-merger structure in the public LIGO data, but it did not produce a reliable ringdown-frequency measurement.

That distinction matters. The event contains enough information to justify a deeper investigation, yet several competing interpretations remain possible. Rather than choosing the most impressive-looking number, the pipeline classified the result as DEEP NEXT.

This is exactly what QUICK4 is intended to do: identify promising signals, preserve their waveform structure and decide which events deserve the much heavier DEEP analysis.

The event

GW230927_153832 is catalogued as a binary black-hole merger at an estimated distance of 1,190 megaparsecs, approximately 3.9 billion light-years. The network signal-to-noise ratio is 20.3, with usable strain data from the Hanford and Livingston detectors.

The analysis used verified public GWOSC strain. No theoretical ringdown frequency was supplied to the search, and no Kerr or Differential Expansion Framework prediction was used to select the result.

The main summary scores were:

• Scientific Information Score: 44/100
• Signal Structure Score: 49/100
• Pipeline decision: DEEP NEXT
• Primary frequency: not measurable
• Early two-decay structure: candidate
• Residual follow-up peaks: none

These scores place the event in the middle ground: it is not a clean benchmark-quality ringdown, but it contains more structure than should be dismissed as ordinary background noise.

The main candidate near 386 Hz

The joint-detector search selected a candidate frequency basin centred on approximately 386.1 Hz.

The diagnostic fit gave:

• Candidate frequency: 386.1 Hz
• Guarded uncertainty: 33.9 Hz
• Candidate damping time: 3.39 milliseconds
• Fit start: 18 milliseconds after the merger peak
• Analysed duration: 64 milliseconds
• Approximate cycles in the interval: 24.7

These values are useful for locating the structure, but they are not reported as physical measurements. Six of the 36 tested fits contributed to the selected cluster, and they came from only one start-time position. The frequency did not form a convincing plateau across the start-time grid.

The candidate therefore depends too strongly on the choice of analysis window.

The time-frequency ridge test also found only one independently supported point. That is not enough to trace a reliable frequency trajectory or determine whether the signal is genuinely stationary, drifting or settling with time.

For these reasons, the report correctly labels 386.1 Hz as a candidate basin, not a measured ringdown frequency.

What the two detectors saw

The joint fit searched for one frequency and damping time shared by the detector network, while allowing Hanford and Livingston to have their own amplitudes, phases and offsets.

Within the selected basin, the local detector results were:

Detector	Independent strongest peak	Best local fit near joint basin	Evidence at joint basin
Hanford	182.4 Hz	355.0 Hz	ΔBIC 7.2
Livingston	397.3 Hz	397.5 Hz	ΔBIC 51.5

The local joint-basin frequencies have a spread of about 8%, which is reasonably encouraging. Both detectors contain some support near the selected network basin.

The strength of that support is not equal, however. Livingston provides much stronger evidence for the high-frequency candidate than Hanford. Hanford’s strongest independent feature is instead near 182 Hz.

This does not mean the black hole produced different physical frequencies in the two detectors. Independent searches are free to select whichever feature is strongest in each noisy strain record. If several weak signal or transient basins are present, the detectors may select different maxima.

The important question is whether both detectors contain local support for the same network feature. Here they do, but Hanford’s contribution is comparatively weak. With only two usable detectors, the pipeline cannot perform a leave-one-detector-out jackknife to show how far the solution moves when either detector is removed.

The 386 Hz result may therefore be influenced heavily by Livingston. DEEP analysis will need to test that directly.

The competing structure near 185 Hz

The most interesting part of the report may not be the selected 386 Hz basin. A second frequency family appears repeatedly around 185 to 187 Hz.

The network cluster search found:

• A 187.4 Hz short-lifetime competitor with ΔBIC 53.9
• A 205.2 Hz short-lifetime cluster with ΔBIC 37.3
• A 174.0 Hz cluster with ΔBIC 40.8
• Hanford’s strongest independent peak at 182.4 Hz

A separate publication-style free damped-mode fit also found a supported component at:

• Frequency: 185.0 Hz
• Damping time: 7.97 milliseconds
• ΔBIC: 53.2
• Recurring start positions: three

This is not a minor secondary bump. It is a coherent family of results appearing through several parts of the analysis.

The pipeline did not automatically replace the 386 Hz selection with this lower-frequency feature because the two searches answer slightly different questions and use different selection rules. The lower-frequency component is also classified as short-lived, which makes it more vulnerable to merger contamination and window-selection effects.

Nevertheless, it creates a genuine ambiguity. There are at least two plausible post-peak structures:

1. A higher-frequency basin near 386 Hz, strongly supported by Livingston and locally supported by Hanford.
2. A shorter-lived component near 185 Hz, recurring across several starts and matching Hanford’s strongest independent basin.

This competition is one of the principal reasons the frequency gate failed. Choosing either frequency as the event’s ringdown measurement at the QUICK stage would be premature.

Is the signal localised to the merger?

The localisation tests are one of the stronger parts of the result.

The blind joint search produced an event ΔBIC of 48.7. None of the 16 off-source trials matched it, giving an empirical bound of p ≤ 0.0588.

After selecting the candidate mode, the fixed-frequency localisation test produced an event ΔBIC of 57.2. None of 30 recurrence trials matched the event result, giving p ≤ 0.0323.

The pre-merger maximum was ΔBIC -15.5, so the selected structure was not already prominent before coalescence.

Together, these tests suggest that the recovered structure is associated with the merger time rather than being a permanently recurring line in the detector. The report therefore classifies it as transient-like.

These results are encouraging, but the limited number of null trials must be kept in mind. Zero matches out of 16 or 30 trials provides a useful screening result, not discovery-level statistical evidence. DEEP mode needs a much larger background population before a strong significance claim can be considered.

There is also no contradiction between the report calling the primary channel “stationary” and the signal “transient-like”. Stationary describes the frequency model used within the selected post-merger interval. Transient-like means that the feature is localised in time and does not recur strongly elsewhere in the surrounding strain.

Possible evolution and two-decay structure

The pipeline found evidence for additional early-time structure, with an improvement of ΔBIC 34.0 and support from both detectors. It did not pass the full support rule because the feature recurred in only one tested window rather than the required two or more.

The result is therefore recorded as CANDIDATE.

An evolving-frequency fit was also tested, but it did not outperform the stationary description in the selected window. The fitted frequency derivative of 3,000 Hz per second should not be treated as a physical measurement because the model comparison did not favour it.

The report also identifies a possible fast-plus-slow decay pattern. At this stage, that should be described only as two-decay morphology. It cannot yet be called a fundamental mode and overtone, or assigned labels such as 220 and 221. Those labels require a successful secondary physical interpretation after the waveform structure itself has been established.

This separation between observation and interpretation is important. A waveform can prefer two decay scales without proving which black-hole modes produced them.

What was not resolved

The blind inspiral reconstruction returned a diagnostic detector-frame chirp mass of 34.00 ± 0.37 solar masses, but the result was not promoted.

The detector-to-detector spread was 32.3%, and the search did not obtain a generic chirp-track lock. The quoted numerical uncertainty describes the local fit and does not include the much larger disagreement between detectors. The correct result is therefore UNRESOLVED.

No numerical damping time or quality factor was reported either. Although the candidate fit contains a damping estimate, the pipeline requires an accepted frequency measurement before promoting damping time and Q. Reporting Q without a stable frequency would give the result more certainty than the data support.

No separated residual peak passed the follow-up threshold after subtraction of the primary candidate. This means the current residual search found no additional joint-detector target. It does not prove that the remaining waveform contains no weaker structure.

Why this event needs DEEP analysis

The report lists six failed measurement gates:

• Off-source detection requires stronger validation
• Data-driven mode activation was incomplete
• The uncertainty assessment did not reach measurement quality
• Start-time stability was not demonstrated
• A strong short-lifetime competing basin remains
• Publication-level validation requires DEEP mode

The DEEP analysis should concentrate on the competition between the 386 Hz and 185 Hz families.

In particular, it needs to establish:

• Whether the 386 Hz basin survives a denser start-time and duration grid
• How strongly the result depends on Livingston
• Whether the 185 Hz feature remains coherent across detectors
• Whether the lower-frequency component is merger leakage, a short-lived physical component or a separate noise basin
• Whether either frequency produces a stable plateau
• Whether the possible two-decay structure recurs across neighbouring windows
• How often comparable results appear in a much larger off-source background
• Whether alternative whitening and conditioning choices preserve the same morphology

Only after those tests should either candidate be considered for a numerical ringdown measurement.

What this run tells us about QUICK4

GW230927_153832 is a useful test of the pipeline because it presents exactly the kind of ambiguous signal that can produce misleading certainty.

A less guarded analysis could easily have reported 386 Hz as the ringdown frequency. Another method might have selected the apparently persistent 185 Hz component. QUICK4 retained both structures, measured their behaviour and declined to force either into the final result.

That is a scientific success, even though the primary frequency remains unresolved.

The report also exposes areas where the pipeline can improve. The background trials are still too limited for strong significance estimates. The two-detector network prevents a proper jackknife. Some diagnostics favour the 386 Hz basin while the publication-style fit favours the 185 Hz family. The four-event calibration suite is not yet complete, and no event-specific reference comparison was available for this run.

These are not reasons to discard the event. They define the work that needs to be done next.

Conclusion

GW230927_153832 contains merger-localised post-peak structure supported across Hanford and Livingston. The strongest network-selected basin lies near 386 Hz, but the evidence is uneven between detectors and unstable across start times. A second, shorter-lived frequency family near 185 Hz appears repeatedly and may be at least as important.

The current data do not justify choosing between them.

The correct QUICK4 result is therefore:

Interesting waveform structure, no promoted ringdown measurement, and a clear recommendation for DEEP analysis.

The value of this event lies in the unresolved morphology. It gives the pipeline a demanding case in which detector agreement, start-time dependence, competing decay components and off-source recurrence all matter. Preserving that ambiguity is more useful than turning one attractive candidate into a number that the data have not yet earned.