GWPAW Workshop

Cross-Correlation Searches for Periodic Gravitational Waves (PDF)

Prabath Peiris, John Whelan, Rochester Institute of Technology

Cross-correlation of gravitational-wave (GW) data streams has been used to search for stochastic backgrounds, and the same technique was applied to look for periodic GWs from the low-mass X-ray binary (LMXB) Sco X-1. A technique has been developed which refines the cross-correlation scheme to take full advantage of the signal model for periodic gravitational waves from rotating neutron stars. By varying the time window over which data streams are correlated, the search can "trade off" between parameter sensitivity and computational cost. We describe prospects for using this cross-correlation method to search LIGO and Virgo data for periodic GWs from systems with partially-known parameters, such as supernova remnants without an associated known pulsar, the center of the Milky Way Galaxy, and LMXBs.

Coherent Follow-up of Continuous Gravitational Wave Candidates (PDF)

Miroslav Shaltev, AEI Hannover

Direct detection of continuous emission of gravitational waves from spinning neutron stars may become an important tool for neutron-star astrophysics. The weakness of the expected signal compared to the noise level of present instruments requires very long observation times. As a consequence, searches for previously unknown continuous gravitational wave sources must cover an enormous parameter space. Searching with fully coherent matched-filtering is computationally prohibitive, and therefore advanced semicoherent techniques, such as those developed for the distributed computing project Einstein@Home, are used. Candidates from these searches require follow-up in only a small region of parameter space. Here we discuss two different methods to determine the required integration time and related computing cost for a fully coherent follow-up of such candidates.

Modeling the Chi-square veto in the Inspiral searches. (PDF)

Rahul Biswas, University of Texas at Brownsville, and Sarah Caudill, Louisiana State

University

Chi-square discriminatory veto is an computationally expensive signal based veto as it requires lots of computing time. This veto is applied to the triggers surviving the second stage of the pipeline in the inspiral searches. We shall make an attempt to understand the nature of this veto and whether we can predict the chi-square values of triggers surviving this computationally expensive veto. This veto helps us to discriminate real signals from those arising due to noise transients by measuring the goodness of fit quantity. We discuss here two separate approaches. One is purely analytical which is based on the power accumulated in a given frequency bin due to the presence of non-stationary and non-Gaussian noise. We know that sum of squares of Gaussian random variables follow a central chi-square distribution whereas in case on non-Gaussian noise they obey a non-central chi-square distribution and hence the mean value has a contribution from central part non-central parameter lambda. We shall show that this non-central parameter lambda is a function of signal to noise ratio (SNR) of the triggers. Thus enabling us to model the expected chi-square values of the triggers using the SNR. The second approach we will talk involves trying to obtain a parabolic fit to the measured chi-squared distribution for a range of values as a function of signal to noise ratio.

Status of the search for gravitational wave ringdowns from perturbed black holes (PDF)

Sarah Caudill, Louisiana State University

We report on the status of the search for gravitational wave ringdowns from perturbed black holes with masses between 25-100 M_sun in LIGO's fifth science run. This mass range is the first regime explored as part of the ringdown search's participation in the Inspiral-Merger-Ringdown (IMR) project, a joint effort between LIGO's Burst group and Compact Binary Coalescence group to study the efficiency with which we detect high mass binary black holes with masses between 25-500 M_sun. We also review the recent changes made to the ringdown analysis pipeline for LIGO's fifth science run. These changes include the implementation of a new 3D coincidence test to compute the distance between pipeline triggers in frequency, quality, and time space as well as a number of improvements to increase efficiency and automation in the pipeline.

Statistical constraints on binary black hole inspiral dynamics (PDF)

Chad Galley, Jet Propulsion Laboratory, California Institute of Technology

We perform a statistical analysis of binary black hole inspirals in the post-Newtonian (PN) approximation. We sample the parameter space of spinning quasi-circular inspirals and identify statistically conserved quantities. We compare these quantities to known PN expressions and find that they perform better in a statistical sense.

Reduced Basis Catalogs for Gravitational Wave Templates (PDF)

Frank Herrmann, University of Maryland

We introduce Reduced Basis as a method for gravitational wave representation and analysis. We comment on computational aspects and compare template catalogs for different detectors to other methods.

Singular Value Decomposition Applied to Binary Black Hole Waveforms (PDF)

Drew Keppel, Albert Einstein Institute

Singular Value Decomposition has been shown to eliminate redundancy inherent in templates banks of post-Newtonian waveforms used to search for gravitational-wave signals from binary neutron stars. Here we present results a similar analysis applied to binary black hole waveforms in the total mass range of 30 to 200 solar masses.

The eccentricity distribution of compact binaries (PDF)

Izabela Kowalska, Univerity of Warsaw

We discuss the eccentricity distribution of inspiral compact object binaries during they inspiral phase. We analyze the expected distributions of eccentricities at three frequencies that are characteristic of three future detectors: Advanced LIGO/VIRGO (30 Hz), Einstein Telescope (3 Hz), and DECIGO (0.3 Hz). We use the StarTrack binary population code to investigate the properties of the population of compact binaries in formation. We evolve their orbits until the point that they enter a given detector sensitivity window and analyze the eccentricity distribution at that time. We find that the eccentricities of BH-BH and BH-NS binaries are quite small when entering the Advanced LIGO/VIRGO detector window for all considered models of binary evolution. Even in the case of the DECIGO detector, the typical eccentricities of BH-BH binaries are below 10^{-4}, and the BH-NS eccentricities are smaller than 10^{-3}. Some fraction of NS-NS binaries may have significant eccentricities. Within the range of considered models, we found that a fraction of between 0.2% and 2% NS-NS binaries will have an eccentricity above 0.01 for the Advanced LIGO/VIRGO detectors. For the ET detector, this fraction is between 0.4% and 4%, and for the DECIGO detector it lies between 2% and 27%.

Neutron star and black hole binaries in dense star clusters (PDF)

Hyung Mok Lee, Seoul National University

Dense stellar systems such as globular clusters are likely to possess black holes and neutron stars as a result of the stellar evolution. These objects become the most massive components as most of the high mass stars evolve away from the main-sequence. The mass segregation process brings the high mass components to the central parts and binaries can form through three-body processes. The hard binaries become harder through the super-elastic encounters with other stars. The hardening proceeds until either the binaries are kicked out of the cluster or inspiral time scale becomes shorter than the typical encounter time scale. Under typical physical conditions of the globular clusters, the black hole binaries are likely to be completely depleted because the pumping is efficient. The black hole binaries escaped from the clusters have inpiral time scale much longer than the Hubble time. Neutron star binaries will eventually escape from the cluster, but the inspiral time scale for some of them is significantly shorter than the Hubble time. The event rate is very uncertain because of the uncertainties in the initial conditions of the clusters, but could be about 1 per one million years in our Galaxy.

Evaluating errors in hybrid gravitational waveforms (PDF)

Ilana MacDonald, University of Toronto

The most accurate models of coalescing compact objects are obtained through numerical relativity simulations. Gravitational wave detectors, such as LIGO, need gravitational waveform templates for detection, and ideally, these would be produced entirely by the afore-mentioned method. Unfortunately, this would be extremely computationally expensive, and thus it is necessary to create hybrid waveforms comprised of a long post-Newtonian inspiral grafted to a numerical waveform covering late inspiral, merger and ringdown. This talk presents an error analysis of various hybrid waveforms. Specifically, we determine the necessary accuracy of the numerical waveform, the best type of post-Newtonian waveform, and at what point during inspiral the two waveforms should be matched.

Accuracy of hybrid waveform families (PDF)

Frank Ohme, Albert Einstein Institute

The detection and interpretation of gravitational waves from black hole binaries requires a bank of template signals. Advances in numerical relativity (NR) have made it possible to extend analytical post-Newtonian (PN) models of the early inspiral by including information from the binary's last few orbits, merger and ringdown. Here we focus on straight-forward combinations of PN and NR data in hybrid waveforms, which successfully yield wave signatures of the entire coalescence process. It is important to quantify the reliability of such models, and we discuss various aspects of error estimates including the important question of how long NR waveforms have to be to meet reasonable accuracy requirements. We present a novel scheme that makes it possible to estimate a priori the length requirements of numerical simulations, i.e., before thousands of CPU-hours are used to actually calculate the needed data.

Searching For Pulsars To Detect Gravitational Waves (P21)

Xavier Siemens, University of Wisconsin-Milwaukee

A low frequency stochastic background of gravitational waves could be detected by pulsar timing experiments in the next five to ten years. Increasing the number of time of arrival data sets available for gravitational wave searches will improve the sensitivity of a pulsar based gravitational wave detector. To achieve this goal, a group of faculty, staff, postdoctoral researchers, and a graduate student at the University of Wisconsin--Milwaukee are participating in a broad effort to increase the number of known stable pulsars collecting and analyzing the pulsar Arecibo L-band Feed Array (P-ALFA) survey data, and the Green Bank Northern Celestial Cap survey data. We have followed the pioneering model started at the University of Texas-Brownsville (UTB) to involve undergraduate and high school students in this research. In close collaboration with the group at UTB we have engaged two local high school teachers, several high school students, and about 15 UWM undergraduates in remotely commanding and observing using the Arecibo radio telescope and the Green Bank telescope, in searches in the collected data for new candidate pulsars, and follow-up observations of of potential pulsar candidates. In addition, the group is using its expertise in LIGO data analysis to improve gravitational wave searches in pulsar timing data. Other presenters will be: Christopher Biwer, David Day, Robin karr, & Matthew Rohr

A Parameter Space Investigation of Junk Radiation in Binary Black Hole Simulations (PDF)

Nick Tacik, University of Toronto

We investigate the parameter space (spin and initial separation) dependence of junk radiation in simulations of equal-mass, equal-spin, non-precessing binary black holes. We look at six different spins (from 0 to 0.5), and five differential initial separation (from 12M to 30M). Results are quantified using three different parameters: The energy carried away from the system by junk radiation, the fractional mass change of a black hole due to junk radiation, and the fractional spin decrease of a black hole due to junk radiation.

Coherently searching for perturbed black-hole ringdown signals with a network of gravitational-wave detectors (PDF)

Dipongkar Talukder, Sukanta Bose, Washington State University

We present results in Gaussian data from a template-based multi-detector coherent search for perturbed-black-hole ringdown signals. Like the previous "coincidence" ringdown searches in LIGO data, our method incorporates knowledge of the ringdown waveform in constructing the search templates. Additionally, it checks for consistency of signal amplitude and phase with the signals' times-of-arrival at the detectors. The latter feature is common to both of our method and the "coherent WaveBurst" algorithm, and can help bridge the gap in performance between the coincidence search and the coherent WaveBurst search for ringdown signals.

Compact binary coalescence searches with low latency: why and how (PDF)

Nick Fotopoulos, California Institute of Technology

Low-latency gravitational-wave (GW) detection of a compact binary coalescence (CBC) will allow electromagnetic (EM) followups to observe earlier parts of the corresponding lightcurves, which are brighter, convey more information about the progenitor system, and allow a more confident association of GW and EM transients. Conventional matched filter banks, common in CBC searches, are computationally efficient, but incur a latency of many minutes. Searches with latencies of seconds and significantly increased throughput are achievable with techniques such as principal component analysis, to reduce the number of filtered templates, hierarchical detection with singular value decomposition by-products, and exploitation of the quasi-monochromatic structure of chirps to filter time-slices at different sample rates. We present an implementation of these ideas called LLOID, based on the LSC Algorithm Library and the gstreamer multimedia framework.

Gstlal live demo

Nick Fotopoulos, California Institute of Technology

Taming Instrument Glitches and Detecting Burst Signals (PDF)

Paul T Baker, Montana State University

With a network of gravitational wave detectors it is possible to distinguish between instrumental artifacts, or glitches, and un-modeled gravitational wave signals. The LIGO-Virgo Burst group has developed several effective algorithms for detecting un-modeled, short duration burst signals such as might be generated by core collapse supernovae or associated with gamma ray burst. Here we present a new algorithm that uses Bayesian model selection to decide if features in the data are better described as gravitational wave signals or instrument glitches. As a by-product, even when no burst signals are detected, this procedure produces cleaned data streams that are free of loud glitches. The cleaned data can then be used by standard template based searches for modeled signals such as binary inspirals, but now with significantly reduced backgrounds, making it possible to detect weaker signals.

NASA EM follow-up of LIGO-Virgo candidate events (PDF)

Lindy Blackburn, GSFC

This poster presents a methodology for a follow-up of LIGO-Virgo candidate gravitational-wave events using offline survey data from several NASA high-energy photon instruments aboard RXTE, Swift, and Fermi. Time and sky-location information provided by the GW trigger allows for a targeted search for prompt and afterglow electromagnetic signals. In doing so, we expect to be sensitive to signals which are too weak to be publicly reported as astrophysical EM events.

Localization of gravitational wave sources with network of advanced detectors (PDF)

Marco Drago, Trento University, INFN Padova

The localization of a gravitational wave (GW) sources is extremely valuable for the future GW astronomy. It will anable joint observations between the GW detectors and other astronomical instruments. In the talk we present studies of the source localization capabilities with the future networks of advanced GW detectors and describe their fundamental properties. We estimate the accuracy of the source localization by injecting ad hoc signals into simulated detector noise, and study its dependence on the strength and waveform morphology of the injected signals. We consider different network configurations including the advanced LIGO and Virgo detectors, the LCGT detector and a possible detector in Australia (LIGO-Austrialia)

A Bayesian Approach to Low-Latency Sky Localization (PDF)

Benjamin Farr, Northwestern University

The electromagnetic followup of a gravitational wave event would not only increase confidence in the first detection, but also allow us to extract substantially more astrophysical information from the source. For successful followups, the sky position is needed as quickly and accurately as possible. This is typically done using triangulation methods, which are capable of producing results with very low latency through a frequentist approach. We present the use of a Markov-Chain Monte Carlo parameter estimation code designed to produce the full 15 parameter PDF for gravitational waves emitted by coalescing compact binary objects with spin. By fixing intrinsic parameters to values output by the detection pipeline, an extremely fast MCMC search can be done over extrinsic parameters (e.g. sky position, time of coalescence, etc.), involving only waveform projections rather than full template generation. For non-spinning binaries, this approach allows for skymaps to be produced on the order of minutes rather than weeks as required for full PDF convergence, showing MCMC methods to be a viable approach to low latency sky localization.

Uncovering the Progenitor of Short-duration Gamma-ray Bursts (PDF)

Wen-fai Fong, Harvard University

The origin of short-duration gamma-ray bursts (GRBs) is unclear at present. One way to gain insight on the nature of the progenitor is to study their local and galactic environments. Here, I present HST observations of several SHB hosts along with ground-based follow-up. These observations allow us to place powerful constraints on the short GRB progenitor. Using a variety of techniques, we determine the hosts' morphological properties, measure the physical and host-normalized offsets relative to the galaxy centers, and study the locations of short GRBs relative to their host light distributions. We also compare our results to those for long GRBs. We find our results to be consistent with a progenitor population of NS-NS binary mergers, but they do not rule out other potential progenitor models.

Coherent compact binary coalescence searches for external triggers with large sky-position errors (PDF)

Shaon Ghosh, Sukanta Bose, Washington State University

Short hard gamma-ray bursts (SGRBs) are conjectured to have compact binary coalescences (CBCs) as progenitors. Therefore, SGRBs provide external triggers for searching signals from CBCs in gravitational-wave (GW) detectors. Whereas for many SGRBs the sky-position is known in advance to high accuracy, for some others it can be off by several degrees. Here we develop a method for coherently searching a patch of the sky, several degrees wide, for CBC signals in multiple baselines of GW detectors. We compare its performance in Gaussian noise with that of an all-sky (or "blind") search and a targeted search and show where it can perform better than the latter two.

Searching for transient gravitational waves from isolated neutron stars using STAMP (PDF)

Stefanos Giampanis, UWM

Isolated (non-accreting) neutron stars (NSs) are expected to emit continuous gravitational waves (GWs) via a number or mechanisms (non-axisymmetric distortions, free precession, velocity perturbations in the star's fluid). Transient GWs from isolated NSs are also plausible via similar mechanisms. These include GWs due to instabilities occurring at early stages in the NSs evolution, such as r-modes and bar-modes; GWs due to magnetically induced deformations in young fast rotating magnetars; and GWs associated with electromagnetic glitches. Depending on the process the duration of the GWs can vary from minutes to weeks and possibly months. STAMP, the Stochastic Transient Analysis Multi-detector Pipeline, is a unique tool for searching for such signals in data from current and future generation gravitational wave detectors. We discuss how STAMP can be applied and optimized for these signal types and present preliminary results using synthetic GW data.

X-pipeline Analysis Methods for the A5 GRB Search (PDF)

Daniel Hoak, UMass-Amherst

The central engines of GRBs are thought to be compact, energetic, and asymmetric, and are expected to be powerful sources of gravitational waves. Using the time and sky location of the GRB observed by satellite experiments, we are able to perform targeted searches for GW emission. This allows for both improved sensitivity to real astrophysical signals and strong consistency tests to suppress background noise. We discuss the methods used to search for GWs from observed GRBs during the fifth 'Astrowatch' period of the LIGO-Virgo detectors, and consider the model parameters applied in the consistency cuts and in the estimation of upper limits.

Summed Parallel Infinite Impulse Response (SPIIR) Filters For Low-Latency and Efficient Gravitational Wave Detection (PDF)

Shaun Hooper, University of Western Australia

Advanced gravitational wave detectors are expected to make the first detection in the next decades. Prompt electromagnetic follow-up observations of gravitational wave (GW) events will require low-latent GW triggers. I will present a new time-domain low-latency algorithm for detecting GWs from coalescing binaries of compact objects using infinite impulse response (IIR) filters. I will show that with a good choice of filter coefficients, the summation of a bank of IIR filters can approximate the value of the signal to noise ratio up to 99% that from the optimal matched filter method, whilst maintaining similar the detection efficiency. I will also compare the computational efficiency of this method to the frequency domain matched filtering approach, and discuss the implication for advanced detectors.

Search for long gravitational-wave bursts and high-energy neutrino coincidences (PDF)

Shivaraj Kandhasamy, University of Minnesota

Gamma-ray bursts (GRBs) are predicted to be sources of both gravitational waves (GWs) and high energy neutrinos (HENs). The duration of gravitational waves can vary from a few seconds to hundreds of seconds depending on the source mechanism. A coincident search for GWs and HENs can provide us useful information about GRB events, which are currently limited to photonic observations. Also such coincidences will improve the statistical confidence of observed GW candidates. I present here a 'box search' analysis, method and simulation results, using Stochastic Transient Algorithm Multi-detector Pipeline (STAMP) developed to look for long duration (a few seconds to weeks) GWs in LIGO data.

Searching for Radio Transients with MWA and ASKAP (PDF)

David Kaplan, UWM

We will discuss ongoing and future efforts to find radio transients across a range of timescales with two new radio telescopes: the Murchison Widefield Array and the Australian Square Kilometer Array Pathfinder, both of which are being built in Western Australia.

The prospect of detecting ultra-low frequency GWs using kHz continuous GW sources with 3rd generation detectors (P54)

Chris Messenger, Cardiff University

We present a study of the possibility of detecting the ultra-low frequency (nanohertz) GW signals from a stochastic background and/or a super-massive blackhole binary system using a network of continuous GW sources as the signal carrier. The parallels with EM pulsar timing arrays are clear. In the standard PTA scenario the correlated timing delays measured in the time of arrival of individual pulses can be used to determine the nature of a variety of possible ultra-low frequency GW sources. Here we replace the EM pulsars with continuously emitting GW sources and investigate primarily the feasibility of such a detection strategy in the 3rd generation GW detection era in which we assume that CW signals will be detected a moderate SNR.

Estimation of binary black hole coalescence event rate exclusion plots with mass and spin parameters from burst search results. (PDF)

Satya Mohapatra, University of Massachusetts Amherst

The sensitivity of burst searches have been generally expressed in terms of the root-sum-squared strain amplitudes on a variety of waveforms such as Sine-Gaussians, Gaussians and Harmonic Ringdowns [1]. Upper limits have been expressed as event rate versus root-sum-squared strain amplitudes exclusion plots for several of these waveforms. Binary black hole coalescence Gravitational Wave (GW) signal properties depend on the mass and spin parameters of the binary system. The resulting signal in a ground-based detector’s sensitive band could be a long duration (few seconds) chirp, for small mass binary system, e.g. total mass = 6 M_solar; or could resemble a short duration (10 ms to < 1 s) GW burst for high mass binary system, e.g. total mass = 150 M_solar;. The root-sum-squared strain amplitudes of these signals in a ground-based detector’s sensitive band also depends on the source parameters (mass, spins and distance). This poster focuses on these high mass binary systems (> 150 M_solar;) and the corresponding 'bursty' GWs. In this poster exclusion curves of event rate versus mass and spin parameters of the binary black holes are estimated instead of as a function of a signal amplitude using published burst search results. For the binary black hole coalescence model spin-aligned phenomenological waveform [2] is used. [1] J. Abadie et al. (LIGO Scientific Collaboration and Virgo Collaboration), All-sky search for gravitational-wave bursts in the first joint LIGO-GEO-Virgo run, Physical Review D 81, 102001 (2010). [2] Ajith, P., et al. “Complete” gravitational waveforms for black-hole binaries with non-precessing spins, 2009. arXiv:0909.2867

Sky localization for inspiralling compact binaries using a network of ground based gravitational wave interferometers. (P28)

Samaya Nissanke, Caltech/JPL

I will discuss our recent work on sky localization for inspiralling compact binaries using gravitational wave measurements only with a network of advanced ground-based interferometers. We explicitly compute sky localization errors for non-spinning neutron star binary populations using Markov Chain Monte Carlo (MCMC) techniques. Localizing the source on the sky is of particular relevance for any follow-up observation using conventional electromagnetic telescopes.

Bayesian Approach to Multi-messenger Astronomy: Population Study of Gamma-ray Bursts (PDF)

Marc Normandin, University of Texas at Brownsville

Several astrophysical models predict gravitational wave (GW) emission from gamma-ray bursts (GRBs) progenitors, although the GW signal to noise ratio associated with single events is expected to be too low for detection in the majority of cases given the large distances involved. It is possible, however, to combine data from several GRB triggers to look for the GW signature of the progenitor population as a whole. We present an improved algorithm for population study which optimally uses information obtained from the electromagnetic counterpart of events. Existing methods do not associate suitable weight factors to the GRB triggers besides trivial ones like a hard cutoff on flux. The assignment of weights needs to take the background astrophysical model to be estimated into account. This is done using the Bayesian approach where electromagnetic data is explicitly incorporated into the likelihood functional of GW data before the parameters of a given population distribution are estimated. We present a preliminary version of this method using simulated data corresponding to simplified models of GWs and electromagnetic detectors along with quasi-realistic astrophysical population models.

Search for long gravitational-wave transients from gamma-ray bursts during LIGO S5 and S6 runs (PDF)

Peter Raffai, Lorand Eotvos University

We propose to analyze available LIGO-Virgo data coincident with gamma-ray bursts (GRBs) detected during the S5 and S6 LIGO runs. The analysis is to be carried out by applying two independent pattern-recognition techniques on frequency-time (ft-)maps of gravitational-wave (GW) strain cross-power. The ft-maps are produced by a cross-correlation-based pipeline using data of different pairs of GW detectors. The aim of the search is to set upper limits on the total cross-power integrated along quasi-monochromatic tracks in the ft-maps. We have used a catalog of galaxies to associate possible host galaxies to the GRB triggers. Using future search results and the known distances of the associated galaxies, we propose to derive upper limits on the total energy emitted by the GRB sources in the form of O(1-100 sec) long GW transients. The resulting empirical upper limits on GW emission energies will be used to derive implications on the distances of the GRB sources, and to 'a posteriori' test theoretical models of long-duration GW emission from GRB progenitors.

Human Vetting of Candidate Gravitational Wave Events for Wide-Field Optical Follow-Up Observation (PDF)

Amber Stuver, LIGO Livingston

Interferometric gravitational wave (GW) detectors have reached the sensitivity and refinement in data analysis to begin to participate in the multi-messenger astronomy community as an event generator. The LIGO-Virgo Collaboration (LVC) has entered into MOUs with wide-field optical telescopes and developed an infrastructure to implement low-latency Target-of-Opportunity (ToO) requests in search of optical transients accompanying a candidate GW event. Near real-time burst and binary inspiral data analysis methods were used to produce candidates for follow-up. Before any ToO requests are submitted, human vetting is necessary to insure reasonable requests, at least at this early stage of infrastructure development. Vetting focuses on determining candidate GW event significance with respect to trigger significance and interferometer performance. A group of about 25 LVC wide-field optical follow-up experts provided remote 24/7 support to scientific monitors and operators who were located at the detectors during a winter 2009-’10 and summer ’10 follow-up run. Presented here is a summary of the wide-field optical follow-up ToO infrastructure with emphasis on the role of human vetting procedures.

GPU-Accelerated Searches for Gravitational Waves from Coalescing Binaries of Compact Objects (PDF)

Linqing Wen, UWA, Caltech

Technique for fast detection and localisation of gravitational waves is in demand to allow prompt electromagnetic follow-up observations of gravitational wave events. Graphics Processing Units (GPUs) are now well known as cost-effective multi-thread devices that are promising to achieve such task. I will present our work on using the GPUs to speed up the searches of gravitational waves from coalescing binaries of compact objects on both the existing search pipeline and on a new low-latency time-domain search method. The implications to real-time low latency detection of gravitational waves in the era of advanced detectors will be discussed.

Burst search with gstreamer pipeline

Satya Mohapatra, University of Massachusetts Amherst

We present a live demonstration of a burst search based on gstlal [1], an interface to the LSC Algorithm Library based on the gstreamer multimedia framework. This demonstration algorithm uses sine-Gaussians as template basis and shows time-frequency morphology of binary black hole coalescence signals injected into simulated advanced LIGO detector noise.

Detection with non-Gaussian noise and optimal strategies with likelihood ratios (PDF)

Jordi Burguet-Castell, University of Wisconsin - Milwaukee

The detection of gravitational waves requires the analysis of time-series data for signals of various types. If the detector noise is stationary and Gaussian, a matched-filter search is optimal. But interferometric detectors are affected by different kind of disturbances resulting in a significant non-Gaussian noise component. A strategy based on likelihood ratios provides a unified approach to data analysis which combines different kinds of information to better rank gravitational-wave detection candidates.

Bayesian parameter estimation of EMRI signals in LISA data using parallel tempering MCMC

algorithm (PDF)

Nelson Christensen, Carleton College

The planned space-based interferometric gravitational-wave detector LISA will observe numerous extreme mass-ratio inspiral (EMRI) signals during its mission. EMRI waveforms have a very complicated spectrum. Furthermore a typical EMRI signal will be buried in the background of detector noise, as well as other astrophysical and cosmological signals. This makes the detection of EMRI signals and the parameter estimation for them a challenging technical and statistical problem. This poster presents an efficient Bayesian framework for the detection and parameter estimation of EMRI signals using parallel tempering MCMC algorithm; this is a very effective technique when dealing with multi-modality problems.

Critical Coupling Likelihood: A new approach to integrate LIGO data quality with searches for gravitational waves from compact binary coalescence (PDF)

Cesar Costa, Louisiana State University

As part of the current LIGO search for compact binary coalescence (CBC) gravitational waves (GW) we find ourselves trying to determine if noise is coupling into the instrument indirectly using our data quality knowledge. The Critical Coupling Likelihood (CCL) method will allow us to directly fold information about potential GW triggers directly into our search efforts. Using this technique we can quantitatively inspect the data quality of each individual CBC candidate, and make more rigorous cuts improving the quality of our search. The CCL method will give the CBC search the potential to integrate all of LIGO's physical and environmental monitors (PEM) into a search for gravitational waves. Using the CCL approach and assuming required models of uncoupled (background) and coupled (foreground) instrument states, we should be able to increase our ability to discriminate between noise sources, that can hamper our current search efforts, and, high quality gravitational wave signal candidates. We illustrate how the method works by demonstrating environmental coupling in LIGO S6 data. An approach like CCL will become increasingly important as we move into the Advanced LIGO era, as we go from the first GW detection to gravitational wave astronomy.

Noise Identification in LIGO S6 and Virgo VSR3 Using STAMP (PDF)

Michael Coughlin, Carleton College

The LIGO and Virgo detectors are sensitive to a variety of noise, such as instrumental artifacts and environmental disturbances. The Stochastic Transient Analysis Multi-detector Pipeline (STAMP) has been developed to search for long-duration (t>1s) gravitational-wave (GW) bursts. This pipeline can also be used to identify environmental noise transients. Here we present an algorithm to determine periods of time when long-duration noise sources couple into the interferometer and what these noise sources are. We analyze the cross power between a GW channel and an environmental sensor, using pattern recognition tools to identify statistically significant structure in cross-power spectrograms. We identify GW noise induced from airplanes, helicopters, thunderstorms and other sources. Examples from LIGO’s sixth science run, S6, and Virgo's third scientific run, VSR3, are presented.

Characterization Study of the Virgo Seismic Environment (PDF)

Michael Coughlin, Carleton College

The Virgo gravitational wave detector is an interferometer with 3km arms located in Pisa, Italy. From July to October 2010, Virgo performed its third science run (VSR3) in coincidence with the LIGO detectors. Despite several techniques adopted to isolate the interferometer from the environment, seismic noise is an important issue for Virgo. Vibrations produced by the detector infrastructure (such as air conditioning units, water chillers/heaters, pumps) are found to affect the Virgo sensitivity, with the main coupling mechanisms being through beam jitter and scattered light processes. The Advanced Virgo (AdV) design seeks to reduce ITF couplings to environmental noise by having most vibration-sensitive components suspended and in-vacuum as well to muffle and relocate loud machines. During the months of June and July 2010, a Guralp-3TD seismometer was stationed at various locations around the Virgo site hosting major infrastructure machines. Seismic data were taken and examined by means of spectral and coherence analysis with seismic probes close to the detector. The primarily aim of this study was to identify noisy machines which seismically affect the ITF environment and thus require mitigation attention (either immediate or for AdV). Analyzed machines are located at various distances from the experimental halls, ranging from 10m to 100m. An attempt is made to measure the attenuation of emitted noise at the ITF and correlate it to the distance from the source and to seismic attenuation models in soil. This indicates the existence of shortcuts (to be removed) in the present installations and gives a rough estimate of the upper limit attenuation we might expect in AdV from relocating some machines at a further distance.

Multi-baseline signal consistency tests for searching gravitational-wave signals in LIGO and Virgo detectors from compact binary coalescences (PDF)

Thilina Dayanga, Sukanta Bose, Washington State University, Pullman

The non-Gaussian and non-stationary nature of real data is known to hurt the performance of the gravitational wave signal searches. Incorporating signal-based discriminators that exploit the differences between the time-frequency structure of signals and noise artifacts has been shown to improve their performance for modelled sources. However, the power of these discriminators varies across the signal parameter space. Here we study how the performance of the null-stream statistic for detecting compact binary coalescence signals in a multi-baseline network varies as a function of the sky. We report results on simulated Gaussian data with LIGO sensitivities, with and without signal injections. We compare them with those expected theoretically. These results serve as benchmarks for subsequent studies in real data, and can help in formulating data-analysis strategies for reducing the gap in performance of search pipelines in real and Gaussian data.

Insidious inter-detector correlations and stochastic backgrounds: How likely is an accidental cancellation? (PDF)

Nickolas Fotopoulos, California Institute of Technology

Searches for the stochastic gravitational-wave background rely on cross-correlating gravitational-wave strain between pairs of detectors. Using co-located detectors would provide a large sensitivity advantage over separated and misaligned detectors if their correlated environmental and instrumental noise could be measured or eliminated. All foreseeable methods for measuring and eliminating such contamination are necessarily imperfect, leaving some unknown remainder. We present the Bayesian method for quantifying the information that can be extracted from a measurement about the stochastic background in the presence of unknown non-gravitational correlations.

Multivariate classification of signal versus background in the LIGO-Virgo search for high-mass compact binary coalescences (P46)

Kari Hodge, Caltech

The LIGO-Virgo collaboration searches for gravitational waves (GWs) from astrophysical sources such as compact binary coalescences (CBCs). Unfortunately, the expected GW signals from high-mass CBCs have waveforms similar to those from glitches in the detectors. Thus, our goal is to develop robust methods for separating rare GW signals from the background of glitches. We will discuss why multivariate statistical classification methods are not only naturally suited for this problem, but also can give insight into our understanding of the background.

Analysis in cross-correlating statistically independent noise channels (PDF)

Antonis Mytidis, University of Florida

We consider using cross-correlated data in the search for long-transients. To create frequency-time (f-t) maps for this search, in our processing of the data, we use coarse-graining and zero padding in a way that achieves the desired 0.25Hz resolution from 52s data segments. In naïve processing, without windowing or zero padding, cross- and power-spectral density (csd and psd) are well defined distributions but suffer from spectral leakage. In order to minimize this effect the data is also Hann-windowed, with 50% overlap. Using Monte Carlo data we applied our processing to study the departure of the csd and psd distributions from their well defined counterparts in naïve processing. We found that both the csd and psd means are unaffected. However, the psd variance is such that the psd distribution can be characterized by a rescaled chi-square of 47.467 degrees of freedom, rather than nominal 50. For the csd distribution, we found that the variance is increased by a factor of 1.922. These results enabled us to demonstrate qualitative agreement between the SNR distributions obtained from Monte Carlo data and instrumental data from real interferometers, where an unphysical time shift has been introduced between the data streams in order to remove all astrophysical content. This analysis is part of an extensive investigation of the statistical properties of a stochastic intermediate data (SID) product, based on using real data, semi-analytic techniques and Monte Carlo simulations.

Glitch or gravitational wave? Removing noise events in cross-correlation analyses (PDF)

Tanner Prestegard, University of Minnesota

Gravitational-wave detectors are often affected by noise transients called "glitches." By causing sudden jumps in strain power, glitches can reduce the usefulness of a simple model of slowly varying, nearly stationary Gaussian noise by mimicking the effect of a gravitational wave (GW) signal. However, glitches are not likely to imitate all characteristics of a true GW event. To this end, we propose a maximum likelihood algorithm for evaluating whether or not a GW candidate is consistent with a physical signal. Our likelihood function is based on the coherence (of lack thereof) between multiple detectors, as well as the relative signal amplitude in each detector. We demonstrate a simple model of this algorithm with time-shifted GW data.

Multi-Detector Bayesian Line Veto for Continuous Gravitational-Wave Searches (PDF)

Reinhard Prix, Albert-Einstein-Institut Hannover

Coherently combining data from different detectors increases the sensitivity of searches for continuous gravitational waves (CW) in pure Gaussian stationary noise. It does not, however, discriminate against commonly present single-detector artifacts that resemble CW signals ("lines"). We present an extended multi-detector statistic that incorporates a coincidence veto against such line artifacts, using Bayesian model selection. We discuss different ways to apply this method (coherently and incoherently), and we re-derive a Frequentist ad-hoc coincidence veto as a special case. We show early Monte-Carlo results testing the efficiency of this method, using simulated Gaussian noise with injected line-artifacts.

Application of a novel clustering technique for glitches in gravitational wave data. (PDF)

Papia Rizwan, Soma Mukherjee, University of Texas at Brownsville

Data from the current generation of gravitational wave detectors such as LIGO (Laser Interferometric Gravitational Wave Observatory) contain a large number of spurious signals from instrumental and environmental sources that need to be distinguished from astrophysical gravitational wave signals. Clustering (or grouping) of such data might play a key role in solving the above problem. K-means, known to be a fast clustering algorithm, takes an inconveniently long time to produce output when faced with LIGO data. Longest Common Subsequence (LCSS) has applications in data mining whenever pairwise sequence matching is necessary. Some thoughts have been given to compute pairwise similarities between glitch signals and to run K-means on the similarity values to make the clustering faster. As predicted, LCSS proves to be a useful tool to classify these glitch signals. However, the process is still slow and nonviable when a huge amount of data such as a large set of lengthy time series is involved. The main idea of this work is to implement LCSS with Fast Time Series Evaluation (FTSE) algorithm on simulated data to test if the algorithm is more efficient and accurate compared to the general LCSS algorithm. The efficiency of LCSS with FTSE algorithm is compared to that of general LCSS in terms of time complexity. The accuracy is determined by finding out the number of mismatching points between the outputs of FTSE with LCSS and general LCSS.

Search for a Stochastic Gravitational Wave Background with Torsion-bar Antennas (PDF)

Ayaka Shoda, University of Tokyo

Recently, large interferometers have been working and developed to detect gravitational waves. They have good sensitivity roughly above 10 Hz. However, gravitational waves from scientifically interesting targets such as a gravitational wave background and massive black hole binaries are predicted to have bigger amplitude at lower frequencies. Therefore, it is difficult to observe with ground based interferometers. Gravitational wave antennas in space such as LISA and DECIGO will be able to detect low-frequency gravitational waves. Nevertheless ground based antennas have advantages over space antennas in that it is easier for us to take measures against noise, perform continuous runs, upgrade them and so on. Therefore, we developed a ground based torsion-bar antenna which has a good sensitivity below 1 Hz. The torsion bar is levitated by the flux pinning effect of a superconductor allowing it to rotate freely by the force from gravitational waves and has good sensitivity at low frequencies. Now two torsion-bar antennas are set up at Tokyo and Kyoto in Japan. We performed about 5 hours observational run simultaneously at the both sites. In this poster, we present the cross correlation analysis to search for or establish an upper limit on a stochastic gravitational wave background. This result is expected to set a new upper limit on a stochastic gravitational wave background around 0.1-1 Hz.

Mining Gravitational Wave Glitches via Deep Learning - A New Approach to Realtime Classification in Time Domain (PDF)

Lappoon Tang, Soma Mukherjee, Ting Zhang, University of Texas at Brownsville

Gravitational wave data from the current generation of instruments are dominated by instrumental and environmental glitches of a wide range of physical properties. The high rate of data arrival necessitates application of automated methods for efficient and reliable interpretation of the information content in these glitches so as to facilitate tracking of signal sources. The study proposes an original work on realtime classification of glitches seen in the gravitational wave data in time domain. Glitches present in the noisy data are discretized hierarchically and classified with particular emphasis on the possibilities of having clusters with arbitrary shapes and presence of very small groups. The latter addresses the issue of how to distinguish between large noisy clusters and possible signals that may be present in the data. Data are processed via formation and re-organization of internal representation of the data in successive layers where discovery of clusters is achieved through emergent computation carried out in the deep learning architecture. In particular, one end result of the pipeline in this study is creating visualization of possible cluster structures in the data using self organizing maps, which can be of particular interest for realtime data analysis and detector characterization.

Robustness of Broadband Searches for Continuous-Wave Gravitation Radiation to Deviations from Ideal Signal Model (PDF)

Vladimir Dergachev, California Institute of Technology

Isolated rotating neutron stars are expected to emit gravitational radiation of nearly constant frequency and amplitude. Searches for such continuous waves (CW) are under way in data collected by the LIGO and Virgo Detectors over the last several years. Because CW signal amplitudes are thought to be extremely weak, long time integrations must be carried out to detect a signal. Integration is complicated by the motion of the Earth (daily rotation and orbital motion) which induces substantial modulations of detected frequency and amplitude that are highly dependent on source location. Large volumes of acquired data make this search computationally difficult. We will present the PowerFlux and "Loosely coherent" analysis pipelines, which account for these modulations, and discuss robustness to deviations from the ideal model of a monochromatic source, which could arise from companion bodies, instrument issues or alternative gravity theories. Results using data from the S5 run will be shown as well.

Deriving upper limits on signal strength in case of non-detection (PDF)

Christian Roever, AEI Hannover

While data analysis from gravitational-wave detectors has not yet yielded a detection claim, the data are commonly utilized for constraining the strength of certain signals. We study the philosophical and technical differences between the Bayesian and frequentist approaches to this problem, and illustrate these using an example.