Detecting Gravitational Wave Memory in the Next Galactic Core-Collapse Supernova

This paper presents a detection method combining Linear Prediction Filtering and Matched-Filtering to identify gravitational wave memory from Galactic core-collapse supernovae, demonstrating that while the approach can detect signals from progenitors like D9.6-3D at 1 kpc, it remains insufficient for distances of 10–100 kpc, a conclusion reinforced after correcting an error that had previously underestimated neutrino-induced waveforms.

The Gist

Imagine the universe as a giant, quiet ocean. For a long time, scientists have been listening to this ocean for "waves" caused by massive events, like two black holes crashing together. These waves are called Gravitational Waves.

However, there is a specific type of wave that has been predicted by Einstein's theory of gravity but never actually caught in the act. The authors of this paper call it "Gravitational Wave Memory."

Here is a simple breakdown of what they did and what they found, using everyday analogies.

The Problem: The "Whisper" vs. The "Roar"

When a massive star dies in a supernova explosion, it creates two types of gravitational signals:

1. The Roar: A chaotic, loud burst of energy happening quickly (like a thunderclap). This is what current detectors usually look for.
2. The Whisper (Memory): A slow, steady "push" that happens as the explosion settles. Imagine a heavy door being slowly pushed open and then staying open. The door doesn't slam; it just stays in a new position. That permanent shift is the "memory."

The Challenge: Current gravitational wave detectors are like very sensitive microphones that are terrible at hearing low, slow sounds. They are great at hearing the "Roar" (high frequencies) but struggle to hear the "Whisper" (low frequencies) because the ocean is naturally noisy down there.

The Solution: A Smart Noise-Canceling Headphone

The authors realized that while the "Whisper" is hard to hear, it is also very predictable. It doesn't jump around randomly; it rises slowly and smoothly, like a ramp.

They developed a two-step trick to find it:

1. The "Noise-Canceling" Filter (Linear Prediction):
   Imagine you are trying to hear a friend speak in a crowded, noisy room. Instead of just turning up the volume, you use a smart system that learns the pattern of the background chatter (the noise) and subtracts it out.
   The authors used a computer algorithm (Linear Prediction Filter) to learn the "chatter" of the detector's noise and remove it. This made the quiet "Whisper" stand out much more clearly.

2. The "Template" Match (Matched Filtering):
   Once the noise was quieted down, they used a "template." Think of this like having a specific shape of a key. They knew exactly what the "Whisper" from a supernova should look like (a smooth ramp). They slid this "key" over the cleaned-up data to see if it fit perfectly.

What They Did

They didn't wait for a real explosion to happen. Instead, they used computer simulations of three different types of dying stars (small, medium, and large). They took the "sound" these simulations made and injected it into real data recorded by the LIGO detectors (the actual gravitational wave observatories).

They asked: If a supernova happened right now, could our new trick find the "Whisper" in the noise?

The Results

• The Big and Medium Stars: For the larger simulated stars, the answer was a resounding YES. Even with the noise of the current detectors, their method could clearly spot the "Whisper" if the explosion happened within our own galaxy (about 10,000 light-years away).
• The Small Star: For the smallest simulated star, the signal was too weak to be seen above the noise with current technology.
• The "False Alarm" Check: They tested how often their method might mistake random noise for a signal. They found that if they combined the data from two detectors (like having two ears), the chance of a false alarm was incredibly low.

Why This Matters

The paper claims that this is the first time anyone has shown a practical way to detect this specific "Memory" effect using current technology.

• The "Door" Analogy: If they succeed, they will have proven that gravity leaves a permanent "scar" or "memory" on space-time after an event, just like a door that stays open after being pushed. This confirms a major prediction of Einstein's General Relativity that has never been seen before.
• The Reach: They can currently "hear" this memory if it happens in our own galaxy. However, they note that with future, more sensitive detectors (like the Einstein Telescope), they might be able to hear this "Whisper" from millions of light-years away, potentially without needing help from other types of telescopes (like neutrino detectors) to tell them when to listen.

In short: The authors built a special "noise-canceling" and "pattern-matching" system that allows us to finally hear the slow, quiet "memory" left behind by exploding stars, confirming a long-held theory about how gravity works.

Technical Summary

Technical Summary: Detecting Gravitational Wave Memory in the Next Galactic Core-Collapse Supernova

Problem Statement
Gravitational wave (GW) memory is a fundamental prediction of general relativity arising from the non-linear nature of the theory, specifically the permanent displacement of test masses following the passage of a GW pulse. In the context of core-collapse supernovae (CCSNe), this memory is generated by asymmetric neutrino emission and the non-spherical expansion of the blast wave. While previous detection strategies for CCSN GWs have relied on excess-energy methods due to the stochastic nature of high-frequency signals (>50 Hz), the memory component represents a distinct, slowly evolving signal below a few tens of Hz. This low-frequency component has been largely overlooked by current ground-based interferometers (such as LIGO, Virgo, and KAGRA) due to their limited sensitivity below 10 Hz and the lack of established detection pipelines for such secular signals. The challenge lies in distinguishing this weak, slowly ramping signal from detector noise and instrumental artifacts without the benefit of precise waveform templates for the stochastic high-frequency components.

Methodology
The authors propose a detection framework combining Linear Prediction Filtering (LPF) and Matched-Filtering to isolate and detect GW memory in current interferometer data. The methodology proceeds as follows:

1. Signal Modeling: The study utilizes three state-of-the-art, three-dimensional CCSN simulations (D9.6, D15, and D25) generated with the Chimera code, representing progenitor masses of 9.6, 15, and 25 solar masses with varying metallicities. The GW strain signals ($h_+$) from these models exhibit a characteristic "secular ramp-up" to a non-zero saturation value.
2. Analytical Template Construction: To facilitate matched filtering, the ramp-up and memory phases are fitted to a tapered logistic function. This function models the rise time, saturation value, and a tapering window to suppress high-frequency noise induced by the abrupt end of simulations. The tapering frequency is set to 0.1 Hz to minimize energy injection into the sensitive band (>10 Hz).
3. Noise Mitigation via LPF: Recognizing that the memory signal is regular and slow, the authors employ a Linear Prediction Filter trained on signal-free segments of GWOSC (Gravitational Wave Open Science Center) data. The LPF predicts and subtracts the correlated noise component from the detector strain data. This process significantly improves the signal-to-noise ratio (SNR) and cross-correlation metrics for the memory component by several orders of magnitude, effectively "whitening" the data for the specific memory template.
4. Matched Filtering and Coincidence: The filtered strain data ($\hat{S}$) is cross-correlated with the tapered logistic templates. A two-detector network (LIGO Hanford and LIGO Livingston) is simulated by multiplying the correlation outputs of both detectors to reduce false alarms.
5. Search Window: The analysis assumes an "On-Source Window" compatible with a neutrino detection, restricting the search to 2-second segments. This duration is chosen because CCSN memory ramp-ups are predicted to occur within this timeframe, and longer durations would be buried in noise or filtered out by the high-pass filters applied to the data.

Key Results
The study injects the tapered memory signals into 4096-second segments of O3b run data from the Livingston and Hanford detectors. The results indicate:

• Detection Range: The approach successfully detects GW memory from CCSNe out to a distance of 10 kpc (kiloparsecs) for progenitor models D15 and D25.
• False Alarm Probabilities (FAP):
  • For the D15 model at 10 kpc, a match threshold of 0.8 yields a FAP of 0.097% (2 triggers out of 2048 segments).
  • For the D25 model at 10 kpc, the same threshold yields a FAP of $\le$ 0.05% (0 false triggers).
  • The D9.6 model (low-mass progenitor with low ejecta asymmetry) produces a signal too weak to be clearly distinguished from noise at 10 kpc, with an SNR of approximately 1.2.
• Signal Characteristics: The D15 and D25 signals are identifiable even in the presence of noise glitches, provided the correlation threshold is appropriately set. The correlation peaks for the actual signal are distinct from noise events, although some noise clusters (glitch-like features) can mimic signals for specific templates (e.g., a glitch at ~750s correlated with the D15 template).
• Future Detectors: The authors note that for next-generation detectors like the Einstein Telescope (ET) and Cosmic Explorer (CE), the projected noise reduction (particularly below 10 Hz) would increase the SNR by at least two orders of magnitude, potentially enabling detection at Megaparsec (Mpc) distances without requiring a coincident neutrino detection.

Significance and Claims
The paper claims to demonstrate, for the first time, that matched-filtering techniques can be effectively utilized to detect GW memory from CCSNe using current interferometers, provided that Linear Prediction Filtering is used to mitigate noise. The primary significance of this work is the potential to confirm an important prediction of general relativity that has remained unverified.

The authors emphasize that while the detection range for current detectors is limited to the Galactic center (approx. 10 kpc), this is a critical regime where multimessenger detection (combining GWs with neutrino observations) is expected. The successful detection of memory would validate the theoretical understanding of asymmetric neutrino emission and hydrodynamic instabilities in supernovae. The paper maintains a modest tone regarding the D9.6 model, acknowledging its lower detectability, and explicitly states that the detection of memory at extragalactic distances (Mpc scales) with current detectors is not feasible without significantly larger saturation values or next-generation instrumentation. The work serves as a proof-of-concept for a new detection strategy that bridges the gap between the stochastic high-frequency signals and the secular low-frequency memory component.