Logo

GW Open Data Workshop #3, Day 2 - Shared screen with speaker view - Recording 1/2
Naresh Adhikari
42:03
sure
Petra Tang
56:32
in a bounded binary system, does kick velocity still play a role during merging
jkanner@caltech.edu
57:08
yes. the “kick” comes from momentum lost to gravitational waves
Petra Tang
59:51
is the "kick" you meant the same kick during the pre-supernova?
Sylvia Biscoveanu
01:00:35
No, the kick that is imparted by the supernova is the “natal kick”. The final merger remnant also gets a kick from the conversation of angular momentum after the merger
jkanner@caltech.edu
01:01:35
for example: https://arxiv.org/abs/1802.04276
Petra Tang
01:02:06
cool thank you :D
rcayuso
01:07:26
I was wondering if you could expand on how deviations from GR searches are performed in LIGO.
Abhishek MS
01:07:29
why won't there is a head on collapseing,suppose a small BH is sucked to a galactic center..would there be a prturbation or wobling make GWS
Felipe Meza
01:08:08
Thanks for sharing the presentation Prof. Alan…It will be a good idea to have a folder in GitHub with all presentations, lots of details and additional resources sources in there. Thanks!
Sylvia Biscoveanu
01:09:14
@Abhishek Yes, when a small black hole merges with a galactic center, this is called an extreme mass ratio inspiral. These types of systems emit GW’s at much lower frequencies than those probed by LIGO, and will be detectable by LISA, the space-based detector
rcayuso
01:11:36
Could you expand on how LIGO performs deviations from GR searches?
Sylvia Biscoveanu
01:14:35
LIGO performs a number of tests of General relativity including looking for the alternative polarizations that Alan mentioned, calculating the speed of gravitational waves since they should travel at the speed of light if GR is correct, consistency tests looking at the estimated source parameters using different parts of the waveform in isolation like the inspiral and merger/ringdown, and also looking for deviations from the post-newtonian parameters used the describe the inspiral part of the waveform
Plamen Fiziev
01:14:38
And what are present days observational .results, even in low confidence level
Sylvia Biscoveanu
01:15:20
Here is a good reference for the recent tests of GR that LIGO has performed https://arxiv.org/abs/1903.04467
Gregory Harry
01:15:58
Peter Saulson answers this question here
Gregory Harry
01:16:00
https://aapt.scitation.org/doi/10.1119/1.18578
Plamen Fiziev
01:16:38
I know this paper. It seems to me that in it we already see deviation at low confidence level, but there are no such comments at all...
Abhishek MS
01:17:18
stationary on earth but still moving with a enormous around milkyway
Gregory Harry
01:17:28
pdf version: https://pdfs.semanticscholar.org/393a/af6b1ced305ee40d175d5f3c3a2b6020348d.pdf
Raymond Yeung
01:19:37
I have a question: if derivation of waveforms is supposed to be hard (e.g. needs numerical relativity), how could we easily generate them with libraries like SEOBNR, Phenom etc.?
Raymond Yeung
01:20:06
IMRPhenom*
rcayuso
01:20:07
Thank you Sylvia and Gregory
Sylvia Biscoveanu
01:22:42
The libraries like SEOB and Phenom are approximations based on the results of numerical relativity. Even with these approximations, the waveforms are extremely expensive to compute.
Alan Weinstein
01:25:11
Phenom is much less expensive than SEOB. The SEOB waveforms are computed in the time domain by solving coupled differential eons, then you FFT them to get the freq domain waveform. Phenom, and NRSur, are computed quickly in the frequency domain
Santanu Ganguly
01:26:17
Many thanks Prof Weinstein!
Alan Weinstein
01:26:28
The difficulty comes when “walking” through parameter space in an MCMC. There, you need to compute the waveforms maybe a million times. It can take hours or days.
Raymond Yeung
01:26:54
Thank you all for answering. Just curious how long the developers of these libraries spent to develop them…
Alan Weinstein
01:27:17
Many, many person years of effort. Hundreds of papers…
Raymond Yeung
01:29:58
I see. Sometimes it feels complicated to use these products right away, without knowing much about their inner workings… ^^”
Petra Tang
01:42:17
is applying matched filter limited to LIGO events?
Abhishek MS
01:43:41
can we take time corresponds to the peak in SNR as merger time
Alan Weinstein
01:44:16
@Petra Matched filtering was developed in the 40’s for radar and (many, many) other applications for finding weak signals in noisy data. Your brain uses it to pick up “Hey Petra!” in a crowded noisy party (pre-covid).
rcayuso
01:44:37
How does LIGO deal with the "Look elsewhere effect" when considering such a big parameter spce for the search?
Alan Weinstein
01:45:52
@Abhishek our matched filtering algorithm reports the time in terms of the “coalescence time” of the waveform template, which we *conventionally” define as the peak of the waveform template.
Gregory Harry
01:45:53
Sorry, I need to drop off for a bit
Petra Tang
01:46:28
@Alan, thank you, so can we use matched filter for the stochastic GWBG?
Alan Weinstein
01:46:35
@rcayruso Background estimation, as Derek will describe soon
rcayuso
01:47:08
Thank you Alan
Alan Weinstein
01:47:57
@Petra for stochastic, we don’t have a “template”. Instead, we cross-correlate data from a pair of detectors (eg, LIGO Hanford and LIGO Livingston). The data from one detector is the stochastic “template” for the data from the other detector. So yes, it is a form of matched filtering.
Alan Weinstein
01:51:25
@Petra All of the math for our (all sky) stochastic searches are in our many papers on the subject. Eg, our very first stochastic paper from 2003 - https://arxiv.org/pdf/gr-qc/0312088.pdf
Petra Tang
01:52:11
@Alan perfect thank you
Alan Weinstein
01:52:12
All developed by Joe Romano and Bruce Allen, more than 20 years ago
Alan Weinstein
01:56:47
@rcayuso the plots Derek is showing now take into account the “look elsewhere” effect associated with 500,000 matched filter templates (also known as “trials factor”).
Abhishek MS
01:56:48
..thank you:)
Ezequiel
02:03:56
I’ve a question here in regards of GAN networks.
Ezequiel
02:04:15
What is the research done in this regard?
Ezequiel
02:04:22
In detecting GW?
Ezequiel
02:04:50
“Many studies”. I just want to double click here….haha.
Alan Weinstein
02:05:47
@Derek clap clap!!
Abhishek MS
02:05:47
can we use matched filter the strain data from one detector and use templet as strain data from another detector neglecting some known noice
Ezequiel
02:06:15
How may I raise my hand?
jkanner@caltech.edu
02:06:36
in “participant window”, press raise hand
Alan Weinstein
02:06:38
@Abhishek that’s essentially what unmodeled burst searches do.
Ezequiel
02:06:40
(done)
Ezequiel
02:06:57
Thank you @Jonnah.
Alan Weinstein
02:07:01
And stochastic searches
Alan Weinstein
02:08:31
@Ezekiel - MANY groups are working on this - https://phys.org/news/2018-04-machine-gravitational.html
Abhishek MS
02:08:52
if stochastic signal are in high frequency region
Abhishek MS
02:09:18
with less amplitude
Agata Trovato
02:10:28
@Ezekiel This is a recent summary of machine learning applications to GW data analysis: https://arxiv.org/pdf/2005.03745.pdf
Ezequiel
02:10:44
Thank you @Alan! I’m just working into a postgraduate thesis that’s looking to improve on some of those methods. (Is not a PhD. something simpler…)
Ezequiel
02:10:54
So, this link could be of huge help.
Alan Weinstein
02:11:13
@Ezekiel There are many papers like this one - https://inspirehep.net/literature/1792015
Ezequiel
02:11:15
(If you guys knows of any other, that could be really help full) thanks a ton!