Detecting Gravitational Wave Memory in the Next Galactic Core-Collapse Supernova
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Astrophysics Radio. Generated commentary on the latest astrophysics papers.
Vera: We're looking at the paper: "Detecting Gravitational Wave Memory in the Next Galactic Core-Collapse Supernova".
Jocelyn: Here's the gist of the paper. Detecting gravitational wave memory in a Galactic core-collapse supernova using linear prediction filtering and matched-filtering techniques is presented to confirm a prediction of general relativity The authors demonstrate an approach to detecting (linear) gravitational wave memory in a Galactic core-collapse supernova using current interferometers.
Vera: They show that the slow and regular time evolution of the memory is uniquely suited to matched-filter techniques.
Jocelyn: Observing the memory, or signs of it, would confirm an important prediction of general relativity that has yet to be confirmed.
Vera: Pruning redundant reasoning tokens The deaths of massive stars in corecollapse supernovae (CCSNe) are promising sources of gravitational waves (GWs).
Paper summary: Vera: So, to recap what we just covered, this paper is about looking for gravitational wave memory in core-collapse supernovae using linear prediction filtering and matched-filtering techniques >
Jocelyn: They are trying to confirm a prediction of general relativity that hasn't been confirmed yet by finding this specific slow signal from these explosions >
Subrahmanyan: That slow signal is tied to the asymmetric neutrino emission and the non-spherical expansion of the supernova blast wave, which is what drives this memory >
Vera: The authors set up their search by using a tapered logistic function to isolate that memory, which they call h fit(t) = L one plus e minus k of t minus t0 (one minus H of t minus ts) plus L two one plus cos two pi f t of t minus ts H of t > <ref:2404.02131#pg1>
Jocelyn: They then inject these signals into simulated LIGO-Virgo-Kagra data and apply a high-pass filter with a cut-off of eight hertz to match the detector's filtering >
Subrahmanyan: The next step they take is training a linear prediction filter, or LPF, with sixteen thousand three hundred eighty-four trained parameters on two thousand forty-eight seconds of data that doesn't have the signal >
Vera: They then subtract what this LPF predicts from the actual detector strain to get S hat equals S minus SLPF, which helps them isolate the memory by several orders of magnitude >
Jocelyn: After all that filtering and training, they present matched-filter results for three different supernova models at a source distance of ten kiloparsecs >
Subrahmanyan: Those results showed signal-to-noise ratios around one point two for the D9 point 6 signal, and about twenty-one point zero for the D15 model and about twenty-one point nine for the D25 model >
Vera: And they've also done some analysis on how different match thresholds and distances change the false alarm probability, showing that at ten kiloparsecs, both D15 and D25 models have an FAP less than five percent for a threshold of zero point three >
Jocelyn: It really highlights how sensitive this technique is when looking for these signals compared to what we see in other searches like gamma-ray burst searches >
Subrahmanyan: This work connects the specific physical characteristics of core-collapse supernovae to the bigger picture of general relativity and structure formation >
Conclusion: Vera: So, looking at this paper, "Detecting Gravitational Wave Memory in the Next Galactic Core-Collapse Supernova," we see they’ve successfully demonstrated an approach using current interferometers >
Jocelyn: They've shown that by combining linear prediction filtering and matched-filtering, you can search for this specific gravitational wave memory signal >
Subrahmanyan: And they’ve confirmed a prediction of general relativity that has been waiting for confirmation since the theory was first proposed >
Vera: The implication is that this method could be effective out to a distance of ten kiloparsecs with current detectors, and it opens the door for future work in next-generation interferometers >
Jocelyn: It suggests that we might be able to detect these supernovae memory signals even at greater distances if we improve our detector noise floor significantly >
Subrahmanyan: They mention that if neutrino-induced kicks are large, those could imply a substantial gravitational wave memory detectable at far greater distances than what their current signals suggest >
Colter J. Richardson, *Haakon Andresen, Anthony Mezzacappa, Michele Zanolin, Michael G. Benjamin, Pedro Marronetti, Eric J. Lentz, *Marek J. Szczepa´nczyk
Department of Physics and Astronomy, University of Tennessee, Knoxville, TN 37996, USA · The Oskar Klein Centre, Department of Astronomy, AlbaNova, SE-106 91 Stockholm, Sweden · Embry-Riddle Aeronautical University · Physics Division, National Science Foundation · Physics Division Oak Ridge National Laboratory · Faculty of Physics University of Warsaw
astro-ph.HE, gr-qc
Submitted: 2024-04-02
Updated: 2026-06-12
Comments: 8 pages, 5 figures
DOI: 10.1103/PhysRevLett.133.231401
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 90/100
The gist: Detecting gravitational wave memory in a Galactic core-collapse supernova using linear prediction filtering and matched-filtering techniques is presented to confirm a prediction of general relativity…
Key concepts
- Gravitational Wave Memory
- This is a slow, regular time evolution in gravitational waves caused by the asymmetric emission of neutrinos during a core-collapse supernova. It represents a permanent distortion or 'memory' left in spacetime after the explosion, which is predicted by general relativity.
- Linear Prediction Filter (LPF)
- The LPF is a specialized filter trained on data that does not contain the signal. It helps isolate the slow, regular evolution of the gravitational wave memory from background noise by predicting and subtracting what it expects to see in the noise.
- Matched-Filter Technique
- This technique involves comparing observed data against pre-calculated templates of expected signals. Because gravitational wave memory has a specific shape, matching it against these templates significantly improves the ability to detect faint signals buried in detector noise.
Terminology
Summary
Detecting gravitational wave memory in a Galactic core-collapse supernova using linear prediction filtering and matched-filtering techniques is presented to confirm a prediction of general relativity
The authors demonstrate an approach to detecting (linear) gravitational wave memory in a Galactic core-collapse supernova using current interferometers. They show that the slow and regular time evolution of the memory is uniquely suited to matched-filter techniques. p. 0 In this Letter, we demonstrate that the slow and regular time evolution of the memory is uniquely suited
to matched-filter techniques. We show how matchedfiltering can be utilized to detect the GW memory from
CCSNe in current interferometers. We show how matchedfiltering can be utilized to detect the GW memory from
CCSNe in current interferometers. Observing the memory, or signs of it, would confirm an important prediction
of general relativity that has yet to be confirmed.Where the paper says it
Where the paper says it
Pruning redundant reasoning tokens
The deaths of massive stars in corecollapse supernovae (CCSNe) are promising sources of gravitational waves (GWs). Stellar core collapse, core bounce at super-nuclear densities, fluid instabilities in the newly-formed proto-neutron star and in the cavity between the proto-neutron star surface and the supernova shock wave, believed to be vital to the explosive central engine, as well as anisotropic neutrino emission, are all expected to generate GWs [1–3]. p. 0 Introduction The deaths of massive stars in corecollapse supernovae (CCSNe) are promising sources of
gravitational waves (GWs). Stellar core collapse, core
bounce at super-nuclear densities, fluid instabilities in
the newly-formed proto-neutron star and in the cavity
between the proto-neutron star surface and the supernova shock wave, believed to be vital to the explosive
central engine, as well as explosion itself and anisotropic
neutrino emission, are all expected to generate GWs [1–
3]. Stellar core collapse, core
bounce at super-nuclear densities, fluid instabilities in
the newly-formed proto-neutron star and in the cavity
between the proto-neutron star surface and the supernova shock wave, believed to be vital to the explosive
central engine, as well as explosion itself and anisotropic
neutrino emission, are all expected to generate GWs [1–
3]. The fluid instabilities, as well as the turbulence they
induce, are expected to excite GW emission at frequencies between 50 Hz and a few kHz [4–64]. However, it has been pointed out recently that matched filtering alongside multi-messenger observations can improve
the detection efficiency of nearby events [55]. Besides
emissions above 50 Hz, a slowly evolving signal component, associated with the GW (linear) memory, is expected below a few 10’s of Hz [6, 10, 11, 13, 15, 16, 18,
21, 28, 36, 41, 58–60, 65–75]. Besides
emissions above 50 Hz, a slowly evolving signal component, associated with the GW (linear) memory, is expected below a few 10’s of Hz [6, 10, 11, 13, 15, 16, 18,
21, 28, 36, 41, 58–60, 65–75]. The memory in a CCSN
stems from asymmetric emission of neutrinos during the
explosion and the non-spherical expansion of the supernova blast wave.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
How the training signal is built
The analytical fit for isolating the memory uses a tapered logistic function defined as h fit(t) = L 1 + e−k(t−t0) (1 − H(t − ts)) + L 2 1 + cos (2πft(t − ts)) H(t − ts). The tapering is characterized by the tapering frequency, ft, which they chose based on the noise characteristics of the LIGO-Virgo-Kagra (LVK) detectors. p. 1 (1)
Here t0 is the center of the rise time, k is the inverse of
the typical rise time, L is the memory saturation value,
ts is the time of saturation (or the end time of the simulation), and H(t) is the Heaviside step function. The
tapering is characterized by the tapering frequency, ft,
which we chose based on the noise characteristics of the
LIGO-Virgo-Kagra (LVK) detectors. (1)
Here t0 is the center of the rise time, k is the inverse of
the typical rise time, L is the memory saturation value,
ts is the time of saturation (or the end time of the simulation), and H(t) is the Heaviside step function. The
tapering is characterized by the tapering frequency, ft,
which we chose based on the noise characteristics of the
LIGO-Virgo-Kagra (LVK) detectors. 6
signal is scaled up by a factor of 10. using simple logistic functions, making them templatable. 6
signal is scaled up by a factor of 10. using simple logistic functions, making them templatable.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the gains come from
The procedure implemented involves injecting the tapered signals into a sample of LVK data obtained from the Gravitational Wave Open Science Center (GWOSC). After injection, they apply a high-pass Butterworth filter with a cut-off of 8 Hz to mimic the filter applied by the LVK Scientific Collaboration. p. 1 Due to the nature of the publicly available data from GWOSC, which has a high-pass
filter already applied to it, after injection we apply to
the strain a high-pass Butterworth filter with a cut-off
of 8 Hz. This high-pass Butterworth filter was used to
mimic the filter applied by the LVK Scientific Collaboration to remove everything below 10 Hz. Due to the nature of the publicly available data from GWOSC, which has a high-pass
filter already applied to it, after injection we apply to
the strain a high-pass Butterworth filter with a cut-off
of 8 Hz. This high-pass Butterworth filter was used to
mimic the filter applied by the LVK Scientific Collaboration to remove everything below 10 Hz. For all of the models,
the secular ramp-up is reduced, but not erased. After injecting the signal, we train a Linear Prediction Filter (LPF) [78] with 16384 trained parameters on
a 2048 s segment of the data that does not contain the
signal. For all of the models,
the secular ramp-up is reduced, but not erased. After injecting the signal, we train a Linear Prediction Filter (LPF) [78] with 16384 trained parameters on
a 2048 s segment of the data that does not contain the
signal. The training converged for training segments shorter than the 2048 seconds used here. We then
subtract the portion of the signal predicted by the LPF
and define
Sˆ = S − SLPF, (2)
where S is the strain from the detector, including the
injected signal, and SLPF is the output of the LPF. The training converged for training segments shorter than the 2048 seconds used here. We then
subtract the portion of the signal predicted by the LPF
and define
Sˆ = S − SLPF, (2)
where S is the strain from the detector, including the
injected signal, and SLPF is the output of the LPF. The
filtered data do not represent the actual detector strain,
but by predicting, and then removing the predicted portion of the noise, we are able to better locate the memory
component. The LPF below 200 Hz improves detectability metrics (like the SNR or cross-correlations with our
templates) for the memory by several orders of magnitude. The
filtered data do not represent the actual detector strain,
but by predicting, and then removing the predicted portion of the noise, we are able to better locate the memory
component. The LPF below 200 Hz improves detectability metrics (like the SNR or cross-correlations with our
templates) for the memory by several orders of magnitude.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Results
The matched-filter results for all three models, with a source distance of 10 kpc and using a twodetector network at O3b sensitivity, are presented in Fig. 4. For the D9.6 signal injected at 10 kpc, the SNR is approximately 1. p. 3 At a distance of 10 kpc,
the SNR of the D9. 6, D15, and D25 models are approximately 1. p. 3 At a distance of 10 kpc,
the SNR of the D9. 6, D15, and D25 models are approximately 1. At a distance of 10 kpc,
the SNR of the D9. 6, D15, and D25 models are approximately 1. At a distance of 10 kpc,
the SNR of the D9. 6, D15, and D25 models are approximately 1. At a distance of 10 kpc,
the SNR of the D9. 6, D15, and D25 models are approximately 1. At a distance of 10 kpc,
the SNR of the D9. 6, D15, and D25 models are approximately 1. At a distance of 10 kpc,
the SNR of the D9. 6, D15, and D25 models are approximately 1. 097%, respectively. The
FAPs reported here correspond to detections with a significance level above 3 sigma. 097%, respectively. The
FAPs reported here correspond to detections with a significance level above 3 sigma.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Conclusions
In this Letter, they have shown that, given the secular ramp-up of the linear GW memory in a CCSN and given the use of a Linear Prediction Filter, a matched-template search (similar to the current detection strategy for binary mergers) can be performed to detect the memory using current interferometers and for the first time confirm an important prediction of general relativity. With a focus on detecting CCSNe memory, their approach would be effective out to a distance of 10 kpc. p. 4 This is our primary finding. With a focus on detecting CCSNe memory, we have
shown that for current detectors our approach would
be effective out to a distance of 10 kpc. Of course, at
these distances a multimessenger detection is expected. The detection range afforded by our approach could also
be considered in the context of next-generation interferometers.Where the paper says it
Where the paper says it
The authors note that the signals are injected at a random time and that choosing a different injection time does not change the general conclusions of this work. Given a coincident neutrino detection (or a search with a two-second temporal window), Fig. p. 3 4 are
consistent even though the ramp-up templates differ. )
Given a coincident neutrino detection (or a search with
a two-second temporal window), Fig. 4 are
consistent even though the ramp-up templates differ. )
Given a coincident neutrino detection (or a search with
a two-second temporal window), Fig. )
Given a coincident neutrino detection (or a search with
a two-second temporal window), Fig. 5 shows how the
FAP depends on the chosen match threshold and the
distance to the source. )
Given a coincident neutrino detection (or a search with
a two-second temporal window), Fig. 5 shows how the
FAP depends on the chosen match threshold and the
distance to the source. Different markers indicate different distances:
1 kpc (dots), 10 kpc (triangles), and 100 kpc (squares). At a distance of 1 kpc, the memory in the D15 and D25
models is detectable with a FAP less than 0. Different markers indicate different distances:
1 kpc (dots), 10 kpc (triangles), and 100 kpc (squares). At a distance of 1 kpc, the memory in the D15 and D25
models is detectable with a FAP less than 0. 6 and a particularly loud
noise event for D15). To calculate a FAP, we select a
match threshold and only consider events with a correlation, ⟨S, h ˆ fit⟩
N(tn)/⟨S, h ˆ fit⟩
N(tinj ), higher than the
chosen threshold as potential detection candidates. 6 and a particularly loud
noise event for D15). To calculate a FAP, we select a
match threshold and only consider events with a correlation, ⟨S, h ˆ fit⟩
N(tn)/⟨S, h ˆ fit⟩
N(tinj ), higher than the
chosen threshold as potential detection candidates. At a distance of 1 kpc, the memory in the D15 and D25
models is detectable with a FAP less than 0. 05% for any
match threshold, while for the D9. At a distance of 1 kpc, the memory in the D15 and D25
models is detectable with a FAP less than 0. 05% for any
match threshold, while for the D9. 100 kpc) will vary with injection times, and will require
potentially larger match thresholds at those distances. The FAP presented in this Letter for a threshold of 0. 100 kpc) will vary with injection times, and will require
potentially larger match thresholds at those distances. The FAP presented in this Letter for a threshold of 0. 100 kpc) will vary with injection times, and will require
potentially larger match thresholds at those distances. The FAP presented in this Letter for a threshold of 0. The FAP presented in this Letter for a threshold of 0. 8
shows that the D25 model can be identified at 10 kpc
with a FAP < 0. 8
shows that the D25 model can be identified at 10 kpc
with a FAP < 0. 05%, which is two orders of magnitude
smaller than the 5% threshold of interest currently used
for gamma-ray-burst searches [81]. 8
shows that the D25 model can be identified at 10 kpc
with a FAP < 0. 05%, which is two orders of magnitude
smaller than the 5% threshold of interest currently used
for gamma-ray-burst searches [81]. 8
shows that the D25 model can be identified at 10 kpc
with a FAP < 0. 05%, which is two orders of magnitude
smaller than the 5% threshold of interest currently used
for gamma-ray-burst searches [81]. For our D25 model, assuming an optimally-orientated
detector, a source distance of 10 kpc, without any data
whitening, and using the LVK-O3 noise curve between
10 and 20 Hz, we obtain an SNR of 0. With the
Einstein Telescope (ET) and Cosmic Explorer (CE) projected noise curves integrated from 5 Hz to 20 Hz and 1
Hz to 20 Hz, respectively, we obtain SNRs of 66. With the
Einstein Telescope (ET) and Cosmic Explorer (CE) projected noise curves integrated from 5 Hz to 20 Hz and 1
Hz to 20 Hz, respectively, we obtain SNRs of 66. 324 and
80. 324 and
80. 545, respectively, which indicate an improvement by
at least two orders of magnitude. This, coupled with
the noise floor reductions expected in next-generation detectors [82, 83], shows the potential for detectability at
Mpc distances with the same FAP thresholds. We plan
to explore this direction in future publications. This, coupled with
the noise floor reductions expected in next-generation detectors [82, 83], shows the potential for detectability at
Mpc distances with the same FAP thresholds. We plan
to explore this direction in future publications. This, coupled with
the noise floor reductions expected in next-generation detectors [82, 83], shows the potential for detectability at
Mpc distances with the same FAP thresholds. We plan
to explore this direction in future publications.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
The paper presents results on three state-of-the-art, three-dimensional core-collapse supernova simulations. The simulations were carried out with the Chimera [76] code, initiated from three non-rotating progenitors with zero-age main sequence masses of 9. p. 0 Models We study the memory from three state-ofthe-art, three-dimensional core-collapse supernova simulations. The simulations were carried out with the
Chimera [76] code, initiated from three non-rotating
progenitors with zero-age main sequence masses of 9. The simulations were carried out with the
Chimera [76] code, initiated from three non-rotating
progenitors with zero-age main sequence masses of 9. 6,
15, and 25 Solar masses, and zero and Solar metallicity
[58]. The models are labeled by a “D” (for Chimera
D-series simulations) followed by the mass of the progenitor from which the simulation in the series was initiated. Rapid shock expansion sets in at ∼125 ms, ∼250 ms, and
∼500 ms for D9. Rapid shock expansion sets in at ∼125 ms, ∼250 ms, and
∼500 ms for D9. 6, D25, and D15, respectively. 6, D25, and D15, respectively. Gravitational Wave Signals The solid lines in the top
panel of Fig. Gravitational Wave Signals The solid lines in the top
panel of Fig. 1 show the plus polarization mode of the
combined (matter and neutrino) GW strains (h+) from
our three models: the blue, orange, and green curves
represent D25, D15, and D9. 6, respectively. Within a
spherical coordinate system centered on the simulations,
the models are observed at randomly-chosen directions:
D9. Within a
spherical coordinate system centered on the simulations,
the models are observed at randomly-chosen directions:
D9. 6 at ϕ = −35◦
, θ = 90◦
; D15 at ϕ = 60◦
, θ = 70◦
;
and D25 at ϕ = 35◦
, θ = 0◦
. 6 at ϕ = −35◦
, θ = 90◦
; D15 at ϕ = 60◦
, θ = 70◦
;
and D25 at ϕ = 35◦
, θ = 0◦
. The GW signals of all
three models show the slow ramp-up to a non-zero strain
value that is characteristic of the memory.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
The results show that for the D15 signal injected at 10 kpc, the correlation has two distinct peaks, one at the injection site and one at the time of a glitch. This glitch is not present in either the D9. p. 3 For the D15 signal injected at 10 kpc, the correlation
has two distinct peaks, one at the injection site and one
at the time of a glitch. This glitch is not present in either
the D9. This glitch is not present in either
the D9. 6 or the D25 case, indicating that this noise event
only correlates with the D15 template. We leave a more complete exploration of the
impact of detector networks to future work. We split the 4096 s window into two-second segments
and define the FAP as the ratio between the number of
segments with triggers above some threshold and the total number of segments. Note, the FAP curves for the D9. 6 model
and at large distances for models D15 and D25 (e. Different line
markers indicate different source distances, and the colors correspond to different models. 100 kpc) will vary with injection times, and will require
potentially larger match thresholds at those distances. We taper with a frequency of 1/10 Hz. The top panel displays
unfiltered signals; the bottom panel shows the same signals
with an 8 Hz high-pass Butterworth filter applied. We taper with a frequency of 1/10 Hz. The top panel displays
unfiltered signals; the bottom panel shows the same signals
with an 8 Hz high-pass Butterworth filter applied. The FAP presented in this Letter for a threshold of 0. 8
shows that the D25 model can be identified at 10 kpc
with a FAP < 0. The FAP presented in this Letter for a threshold of 0. 8
shows that the D25 model can be identified at 10 kpc
with a FAP < 0. The
filtered data do not represent the actual detector strain,
but by predicting, and then removing the predicted portion of the noise, we are able to better locate the memory
component. The LPF below 200 Hz improves detectability metrics (like the SNR or cross-correlations with our
templates) for the memory by several orders of magnitude. The
tapering is characterized by the tapering frequency, ft,
which we chose based on the noise characteristics of the
LIGO-Virgo-Kagra (LVK) detectors. Current groundbased detectors have a sharp increase in their characteristic noise at approximately 10 Hz. The
tapering is characterized by the tapering frequency, ft,
which we chose based on the noise characteristics of the
LIGO-Virgo-Kagra (LVK) detectors. Current groundbased detectors have a sharp increase in their characteristic noise at approximately 10 Hz. Rapid shock expansion sets in at ∼125 ms, ∼250 ms, and
∼500 ms for D9. 6, D25, and D15, respectively.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
The authors acknowledge that the results at O(1) kpc reflect the results they can expect at O(10) kpc with next-generation detectors. The impact on the FAP of using a template bank for the memory is expected to be small because, when a different ramp-up time is used, the matched-filter peaks will be of smaller amplitude and the new template would not be chosen as the representative template. p. 3 Our results suggest that a specialized detector characterization for a memory search is warranted. The impact on the FAP of using a template bank for
the memory (as opposed to the three templates used
here) is expected to be small because, when a different
ramp-up time is used, the matched-filter peaks will be of
smaller amplitude and the new template would not be
chosen as the representative template. Recent findings suggest that we can have saturation values up to 60 times larger than our signals [60], and even
larger for GRBs [84], potentially extending the detection
range to several Mpc in current detectors. Additionally,
it has been found that asymmetric neutrino emission can
lead to large neutron star kicks, with potential amplification through neutrino flavor conversion [85, 86]. Additionally,
it has been found that asymmetric neutrino emission can
lead to large neutron star kicks, with potential amplification through neutrino flavor conversion [85, 86]. Such
large, neutrino-induced kicks would imply a substantial
GW memory, detectable at far greater distances than our
current signals suggest. The opinions and conclusions expressed herein are those of the authors and do
not represent the National Science Foundation. This research has made use of data or software obtained from the Gravitational Wave Open Science Center
(gwosc. org), a service of the LIGO Scientific Collaboration, the Virgo Collaboration, and KAGRA. This material is based upon work supported by NSF’s LIGO
Laboratory, which is a major facility fully funded by
the National Science Foundation, as well as the Science and Technology Facilities Council (STFC) of the
United Kingdom, the Max-Planck-Society (MPS), and
the State of Niedersachsen/Germany for support of the
construction of Advanced LIGO and construction and
operation of the GEO600 detector. org), a service of the LIGO Scientific Collaboration, the Virgo Collaboration, and KAGRA. This material is based upon work supported by NSF’s LIGO
Laboratory, which is a major facility fully funded by
the National Science Foundation, as well as the Science and Technology Facilities Council (STFC) of the
United Kingdom, the Max-Planck-Society (MPS), and
the State of Niedersachsen/Germany for support of the
construction of Advanced LIGO and construction and
operation of the GEO600 detector. Additional support
for Advanced LIGO was provided by the Australian Research Council. Virgo is funded, through the European
Gravitational Observatory (EGO), by the French Centre National de Recherche Scientifique (CNRS), the Italian Istituto Nazionale di Fisica Nucleare (INFN) and the
Dutch Nikhef, with contributions by institutions from
Belgium, Germany, Greece, Hungary, Ireland, Japan,
Monaco, Poland, Portugal, and Spain. Additional support
for Advanced LIGO was provided by the Australian Research Council. Virgo is funded, through the European
Gravitational Observatory (EGO), by the French Centre National de Recherche Scientifique (CNRS), the Italian Istituto Nazionale di Fisica Nucleare (INFN) and the
Dutch Nikhef, with contributions by institutions from
Belgium, Germany, Greece, Hungary, Ireland, Japan,
Monaco, Poland, Portugal, and Spain. Virgo is funded, through the European
Gravitational Observatory (EGO), by the French Centre National de Recherche Scientifique (CNRS), the Italian Istituto Nazionale di Fisica Nucleare (INFN) and the
Dutch Nikhef, with contributions by institutions from
Belgium, Germany, Greece, Hungary, Ireland, Japan,
Monaco, Poland, Portugal, and Spain. KAGRA is supported by the Ministry of Education, Culture, Sports,
Science, and Technology (MEXT), the Japan Society for
the Promotion of Science (JSPS) in Japan, the National
Research Foundation (NRF) and Ministry of Science and
ICT (MSIT) in Korea, and the Academia Sinica (AS)
and National Science and Technology Council (NSTC)
in Taiwan. Virgo is funded, through the European
Gravitational Observatory (EGO), by the French Centre National de Recherche Scientifique (CNRS), the Italian Istituto Nazionale di Fisica Nucleare (INFN) and the
Dutch Nikhef, with contributions by institutions from
Belgium, Germany, Greece, Hungary, Ireland, Japan,
Monaco, Poland, Portugal, and Spain. KAGRA is supported by the Ministry of Education, Culture, Sports,
Science, and Technology (MEXT), the Japan Society for
the Promotion of Science (JSPS) in Japan, the National
Research Foundation (NRF) and Ministry of Science and
ICT (MSIT) in Korea, and the Academia Sinica (AS)
and National Science and Technology Council (NSTC)
in Taiwan. KAGRA is supported by the Ministry of Education, Culture, Sports,
Science, and Technology (MEXT), the Japan Society for
the Promotion of Science (JSPS) in Japan, the National
Research Foundation (NRF) and Ministry of Science and
ICT (MSIT) in Korea, and the Academia Sinica (AS)
and National Science and Technology Council (NSTC)
in Taiwan. The authors would like to acknowledge Lella et al. KAGRA is supported by the Ministry of Education, Culture, Sports,
Science, and Technology (MEXT), the Japan Society for
the Promotion of Science (JSPS) in Japan, the National
Research Foundation (NRF) and Ministry of Science and
ICT (MSIT) in Korea, and the Academia Sinica (AS)
and National Science and Technology Council (NSTC)
in Taiwan. The authors would like to acknowledge Lella et al. The authors would like to acknowledge Lella et al. [87]
for pointing out the discrepancy between our computed
neutrino-induced waveforms and theirs, which prompted
us to investigate the discrepancy and which led to the
discovery of our error and corrections herein. The authors would like to acknowledge Lella et al. [87]
for pointing out the discrepancy between our computed
neutrino-induced waveforms and theirs, which prompted
us to investigate the discrepancy and which led to the
discovery of our error and corrections herein.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
The authors have shown that for current detectors their approach would be effective out to a distance of 10 kpc. p. 4 This is our primary finding. With a focus on detecting CCSNe memory, we have
shown that for current detectors our approach would
be effective out to a distance of 10 kpc. Of course, at
these distances a multimessenger detection is expected. The detection range afforded by our approach could also
be considered in the context of next-generation interferometers. The detection range afforded by our approach could also
be considered in the context of next-generation interferometers. For the ET and CE, the combination of their
reduced noise floor across all frequencies, which is projected to be approximately one order of magnitude in
magnitude across the sensitivity band, and the reduction of the low-frequency wall from 10 Hz to below 20
Hz with improved controls noise, may enable the detection range of CCSNe—specifically, through the detection
of GW memory—out to Mpc distance scales, necessarily
without a concurrent neutrino detection, if a sufficiently
low FAP could be achieved with OSWs of the order of
days. 4, are to be expected. A systematic exploration of this direction, including customized detector
characterization, is left for a future publication.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
The paper has made use of data or software obtained from the Gravitational Wave Open Science Center (gwosc.org). This material is based upon work supported by NSF’s LIGO Laboratory, which is a major facility fully funded by the National Science Foundation. Virgo is funded, through the European Gravitational Observatory (EGO), by the French Centre National de Recherche Scientifique (CNRS), the Italian Istituto Nazionale di Fisica Nucleare (INFN) and the Dutch Nikhef. KAGRA is supported by the Ministry of Education, Culture, Sports, Science, and Technology (MEXT), the Japan Society for the Promotion of Science (JSPS) in Japan, the National Research Foundation (NRF) and Ministry of Science and ICT (MSIT) in Korea. The authors would like to acknowledge Lella et al. [87] for pointing out the discrepancy between their computed neutrino-induced waveforms and theirs, which prompted them to investigate the discrepancy and which led to the discovery of their error and corrections herein.
The authors have shown that given the secular ramp-up of the linear GW memory in a CCSN and given the use of a Linear Prediction Filter, a matched-template search can be performed to detect the memory using current interferometers. This is their primary finding. p. 4 Conclusions In this Letter, we have shown that, given
the secular ramp-up of the linear GW memory in a
CCSN and given the use of a Linear Prediction Filter,
a matched-template search (similar to the current detection strategy for binary mergers) can be performed to
detect the memory using current interferometers and for
the first time confirm an important prediction of general
relativity. This is our primary finding.Where the paper says it
The analytical fit for isolating the memory uses a tapered logistic function defined as h fit(t) = L 1 + e−k(t−t0) (1 − H(t − ts)) + L 2 1 + cos (2πft(t − ts)) H(t − ts). The tapering is characterized by the tapering frequency, ft, which they chose based on the noise characteristics of the LIGO-Virgo-Kagra (LVK) detectors. They use simple logistic functions to make them templatable.
The procedure implemented involves injecting the tapered signals into a sample of LVK data obtained from the Gravitational Wave Open Science Center (GWOSC). After injection, they apply a high-pass Butterworth filter with a cut-off of 8 Hz to mimic the filter applied by the LVK Scientific Collaboration. The Linear Prediction Filter (LPF) is trained with 16384 trained parameters on a 2048 s segment of data that does not contain the signal. Subtracting the portion of the signal predicted by the LPF defines Sˆ = S − SLPF. The LPF below 200 Hz improves detectability metrics (like the SNR or cross-correlations with our templates) for the memory by several orders of magnitude.
The matched-filter results for all three models, with a source distance of 10 kpc and using a twodetector network at O3b sensitivity, are presented in Fig. 4. For the D9.6 signal injected at 10 kpc, the SNR is approximately 1.2. For the D15 signal injected at 10 kpc, the SNR is approximately 21.0. For the D25 signal injected at 10 kpc, the SNR is approximately 21.9. The FAP reported here correspond to detections with a significance level above 3 sigma.
The authors note that the signals are injected at a random time and that choosing a different injection time does not change the general conclusions of this work. Given a coincident neutrino detection (or a search with a two-second temporal window), Fig. 5 shows how the FAP depends on the chosen match threshold and the distance to the source. At a distance of 1 kpc, the memory in the D15 and D25 models is detectable with a FAP less than 0.05% for any match threshold. At 10 kpc, both D15 and D25 have a FAP of less than 5% for a match threshold of 0.3. The FAP presented in this Letter for a threshold of 0.8 shows that the D9.6 model can be identified at 10 kpc with a FAP < 0.05%, which is two orders of magnitude smaller than the 5% threshold of interest currently used for gamma-ray-burst searches [81].
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Enhanced detection capabilities for gravitational wave signals in core-collapse supernovae by implementing a matched-filter search strategy utilizing Linear Prediction Filtering. This improved system can detect linear gravitational wave memory from core-collapse supernovae in current interferometers out to 10 kpc.
-
Ability to distinguish the secular ramp-up of gravitational wave memory from stochastic background noise by fitting signals to a tapered logistic function defined by equation 1. The AI system will perform this template matching against simulated and real detector data.
-
Improved false alarm probability estimation for core-collapse supernova searches by incorporating On-Source Windows compatible with neutrino detection. This allows the system to assess the significance of potential gravitational wave detections with greater accuracy, as stated in
We also present the False Alarm Probabilities assuming an On-Source Window compatible with the presence of a neutrino detection.
-
Detection range extension for future gravitational wave observatories like the Einstein Telescope and Cosmic Explorer by projecting performance based on noise floor reductions. The system can estimate potential detection ranges up to Mpc distances under specific conditions, as suggested by
With the Einstein Telescope (ET) and Cosmic Explorer (CE) projected noise curves integrated from 5 Hz to 20 Hz and 1 Hz to 20 Hz, respectively, we obtain SNRs of 66.324 and 80.545
. -
Automated analysis pipeline for signal extraction from noisy detector data by subtracting the predicted portion of the noise using a Linear Prediction Filter (LPF). This process yields S-hat = S - SLPF, which serves as a better starting point for matched filtering than actual strain data.
Abstract
We present an approach to detecting (linear) gravitational wave memory in a Galactic core-collapse supernova using current interferometers. Gravitational wave memory is an important prediction of general relativity that has yet to be confirmed. Our approach uses a combination of Linear Prediction Filtering and Matched-Filtering. We present the results of our approach on data from core-collapse supernova simulations that span a range of progenitor mass and metallicity. We are able to detect gravitational wave memory out to 10 kpc. We also present the False Alarm Probabilities assuming an On-Source Window compatible with the presence of a neutrino detection. Errata: The neutrino-induced gravitational waveforms, a component of the total waveforms used to demonstrate the efficacy of the proposed detection method for gravitational wave memory, were computed incorrectly. They were underestimated by a factor of 4π. As a result, the detection results improve and, most important, our primary conclusions are reinforced. In particular, with the corrected waveforms, see Fig. 1 and Fig. 3, the signal for D9.6-3D is in fact detectable at 1 kiloparsec, see Fig. 6 and Fig. 4. The corrected fit parameters are presented in Table 1. The false alarm probability for all three signals is impacted, as shown in Fig. 5. Fig. 5 demonstrates that our proposed method can reliably detect the D9.6-3D signal at 1 kpc for all cutoff values, but our conclusions regarding the inability to detect this signal at 10 and 100 kpc remain the same. We see a boost in the detectability of the D15-3D and D25-3D signals. We now obtain a lower false alarm probability at 10 and 100 kpc. The authors would like to acknowledge Lella et al. [87] for pointing out the discrepancy between our computed neutrino-induced waveforms and theirs, which prompted us to investigate the discrepancy and which led to the discovery of our error.
Sources
- Gravitational Waves from Neutrino-Driven Core Collapse Supernovae: Predictions, Detection, and Parameter Estimation
- Gravitational waves from relativistic rotational core collapse
- Gravitational Radiation from Axisymmetric Rotational Core Collapse
- Gravitational Radiation from Rotational Core Collapse: Effects of Magnetic Fields and Realistic Equation of States
- Towards Gravitational Wave Signals from Realistic Core Collapse Supernova Models
- Stochastic Nature of Gravitational Waves from Supernova Explosions with Standing Accretion Shock Instability
- Equation-of-State Dependent Features in Shock-Oscillation Modulated Neutrino and Gravitational-Wave Signals from Supernovae
- A Model for Gravitational Wave Emission from Neutrino-Driven Core-Collapse Supernovae
- Effects of Rotation on Stochasticity of Gravitational Waves in Nonlinear Phase of Core-Collapse Supernovae
- Parametrized 3D models of neutrino-driven supernova explosions: Neutrino emission asymmetries and gravitational-wave signals
- Gravitational wave signatures in black-hole-forming core collapse
- A New Multi-Dimensional General Relativistic Neutrino Hydrodynamics Code of Core-Collapse Supernovae III. Gravitational Wave Signals from Supernova Explosion Models
- Gravitational Wave Signatures from Low-mode Spiral Instabilities in Rapidly Rotating Supernova Cores
- Gravitational Wave Signatures of Ab Initio Two-Dimensional Core Collapse Supernova Explosion Models for 12-25 Solar Masses Stars
- Coherent Network Analysis of Gravitational Waves from Three-Dimensional Core-Collapse Supernova Models
- Circular polarizations of gravitational waves from core-collapse supernovae: a clear indication of rapid rotation
- A New Gravitational-Wave Signature from Standing Accretion Shock Instabilities in Supernovae
- Correlated Signatures of Gravitational-Wave and Neutrino Emission in Three-Dimensional General-Relativistic Core-Collapse Supernova Simulations
- Equation of State Effects on Gravitational Waves from Rotating Core Collapse
- The gravitational wave signal from core-collapse supernovae
Related papers
- Low- T/W instabilities in differentially rotating neutron stars resembling merger remnants
- The imprint of individual neutron star spins on ejecta, r-process nucleosynthesis, and kilonovae in binary neutron star mergers
- Early panchromatic spectra reveal helium in the thermonuclear supernova 2024kce
- The Double-Peaked Calcium-Strong SN 2025coe: Progenitor Constraints from Early Interaction and Ejecta Asymmetries
- The pulsar population of Omega Centauri: New TRAPUM searches, timing, and cluster dynamics with MeerKAT and Parkes
- One year of broadband radio monitoring of the enigmatic transient GRB 250702B reveals the evolution of the relativistic jet