Table of contents

It is now almost two decades since Bernard Schutz organized a landmark meeting on data analysis for gravitational wave detectors at the University of Cardiff, UK [1]. The proceedings of that meeting make interesting reading. Among the issues discussed were optimal ways to carry out searches for binary inspiral signals, and ways in which the projected growth in computer speed, memory and networking bandwidth would influence searches for gravitational wave signals.

The Gravitational Wave Data Analysis Workshop traces its history to the mid-1990s. With the construction of the US LIGO detectors and the European GEO and VIRGO detectors already underway, Kip Thorne and Sam Finn realized that it was important for the world-wide data analysis community to start discussing some of the big unsettled issues in analysis. What was the optimal way to perform a pulsar search? To ensure confident detection, how accurately did binary inspiral waveforms have to be calculated? It was largely Kip and Sam's initiative that got the GWDAW started.

The first (official) GWDAW was hosted by Rai Weiss at Massachusetts Institute of Technology, USA in 1996, as a follow-on to an informal meeting organized in the previous year by Sam Finn. I have pleasant memories of this first MIT GWDAW. I was new to the field and remember my excitement at learning that I had many colleagues interested in (and working on) the important issues. I also remember how refreshing it was to hear a pair of talks by Pia Astone and Marialessandra Papa who were not only studying methods but had actually carried out serious pulsar and burst searches using data from the Rome resonant bar detectors.

A lot has changed since then. This issue is the Proceedings of the 8th Annual Gravitational Wave Data Analysis Workshop, held on 17-20 December 2003 at the University of Wisconsin-Milwaukee, USA. Many of the contributions concern technical details of the analysis of real data from resonant mass and interferometric detectors, setting upper limits on known pulsars, the gravitational wave stochastic background, and rates of burst and inspiral signals.

Barring something unforeseen, the next decade of the GWDAW may see the launch of the LISA space-based detector, and should see the definitive detection of gravitational waves with terrestrial detectors. The scientific significance of this discovery is great enough that it probably will not be announced at a future GWDAW, but I am sure that the technical details of the analysis will get a great deal of attention there!

References

[1] Schutz B F (ed) 1989 Gravitational Wave Data Analysis, Proc. NATO Advanced Research Workshop (Cardiff, UK) (Amsterdam: Kluwer)

The LIGO Scientific Collaboration has completed the analysis of the data from its first science run (Aug–Sep 2002), is far along in the analysis of the second science run (Feb–April 2003), and is beginning the analysis of data taken during its third science run (Nov 2003–Jan 2004). We present in this paper an overview of the performance of the instruments during the science runs, of the data analysis efforts completed so far and of the plans for the future.

The two gravitational wave detectors Explorer (located at CERN) and Nautilus (in Frascati, LNF) have been operating for many years. These detectors allow us to investigate various classes of signals, such as bursts, continuous waves and the stochastic background. They operated in the years 2001 and 2003 with unprecedented sensitivities. In this paper we will recall some of the results obtained by the collaboration and summarize our plans for the near future.

One of the most exciting prospects for the LISA gravitational wave observatory is the detection of gravitational radiation from the inspiral of a compact object into a supermassive black hole. The large inspiral parameter space and low amplitude of the signal make detection of these sources computationally challenging. We outline here a first-cut data analysis scheme that assumes realistic computational resources. In the context of this scheme, we estimate the signal-to-noise ratio that a source requires to pass our thresholds and be detected. Combining this with an estimate of the population of sources in the universe, we estimate the number of inspiral events that LISA could detect. The preliminary results are very encouraging—with the baseline design, LISA can see inspirals out to a redshift z = 1 and should detect over a thousand events during the mission lifetime.

A binary compact object early in its inspiral phase will be picked up by its nearly monochromatic gravitational radiation by LISA. But even this innocuous appearing candidate poses interesting detection challenges. The data that will be scanned for such sources will be a set of three functions of LISA's twelve data streams obtained through time-delay interferometry, which is necessary to cancel the noise contributions from laser-frequency fluctuations and optical-bench motions to these data streams. We call these three functions pseudo-detectors. The sensitivity of any pseudo-detector to a given sky position is a function of LISA's orbital position. Moreover, at a given point in LISA's orbit, each pseudo-detector has a different sensitivity to the same sky position. In this work, we obtain the optimal statistic for detecting gravitational wave signals, such as from compact binaries early in their inspiral stage, in LISA data. We also present how the sensitivity of LISA, defined by this optimal statistic, varies as a function of sky position and LISA's orbital location. Finally, we show how a real-time search for inspiral signals can be implemented on the LISA data by constructing a bank of templates in the sky positions.

We describe the current status of the search for gravitational waves from inspiralling compact binary systems in LIGO data. We review the result from the first scientific run of LIGO (S1). We present the goals of the search of data taken in the second scientific run (S2) and describe the differences between the methods used in S1 and S2.

In the framework of matched filtering theory, which is the most promising method for the detection of gravitational waves emitted by coalescing binaries, we report on the ability of a template bank to catch a simulated binary black-hole gravitational wave signal. If we suppose that the incoming signal waveform is known a priori, then both the (simulated) signal and the templates can be based on the same physical model and therefore the template bank can be optimal in the sense of Wiener filtering. This turns out to be true for the case of neutron star binaries but not necessarily for the black-hole case. When the templates and the signal are based on different physical models the detection bank may still remain efficient. Nonetheless, it might be a judicious choice to use a phenomenological template family such as the so-called BCV templates to catch all the different physical models. In the first part of that report, we illustrate in a non-exhaustive study, by using Monte Carlo simulations, the efficiency of a template bank based on the stationary phase approximation and show how it catches simulated signals based on the same physical model but fails to catch signals built using other models (Padé, EOB, ...) especially in the case of high mass binaries. In the second part, we construct a BCV-template bank and test its validity by injecting simulated signals based on different physical models such as the PN-approximants, Padé-approximant and the effective one-body method. We show that it is suitable for a search pipeline since it gives a match higher than 95% for all the different physical models. The range of individual mass which has been used is [3–20]M_⊙.

The search for periodic sources of gravitational waves is a heavy task from the computational point of view, and optimal methods cannot be applied for wide-area searches. Non-optimal, hierarchical methods have been developed and are based on alternating coherent and incoherent steps. In this paper we discuss the spread of the spectral signal power due to the Earth's rotation. It affects the signal-to-noise ratio in the second, coherent, step of the analysis and a way to take this into account is outlined. Moreover, a new method for the hierarchical data analysis procedure is described.

We present a Bayesian Markov chain Monte Carlo technique for estimating the astrophysical parameters of gravitational radiation signals from a neutron star in laser interferometer data. This computational algorithm can estimate up to six unknown parameters of the target, including the rotation frequency and frequency derivative, using reparametrization, delayed rejection and simulated annealing. We highlight how a simple extension of the method, distributed over multiple computer processors, will allow for a search over a narrow frequency band. The ultimate goal of this research is to search for sources at known locations, but uncertain spin parameters; an example would be SN1987A.

In a blind search for continuous gravitational wave signals scanning a wide frequency band one looks for candidate events with significantly large values of the detection statistic. Unfortunately, a noise line in the data may also produce a moderately large detection statistic. In this paper, we describe how we can distinguish between noise line events and actual continuous wave (CW) signals, based on the shape of the detection statistic as a function of the signal's frequency. We will analyse the case of a particular detection statistic, the F statistic, proposed by Jaranowski, Królak and Schutz. We will show that for a broad-band 10 h search, with a false dismissal rate smaller than 10⁻⁶, our method rejects about 70% of the large candidate events found in a typical data set from the second science run of the Hanford LIGO interferometer.

We describe analysis methods and results for burst gravitational waves with data obtained in the eighth observation run by the TAMA300 detector. In this analysis, we used an excess-power filter for signal detection, and two types of veto for fake-event rejection; one is a time-scale selection of events and the other is a veto with auxiliary information recorded together with the main signal. We generated an event-candidate list with this analysis procedure, which will be used for coincidence analysis with the other detectors.

In this paper we describe the performance of the WaveBurst algorithm which was designed for detection of gravitational wave bursts in interferometric data. The performance of the algorithm was evaluated on the test dataset collected during the second LIGO Scientific run. We have measured the false alarm rate of the algorithm as a function of the threshold and estimated its detection efficiency for simulated burst waveforms.

The burst search in LIGO relies on the coincident detection of transient signals in multiple interferometers. As only minimal assumptions are made about the event waveform or duration, the analysis pipeline requires loose coincidence in time, frequency and amplitude. Confidence in the resulting events and their waveform consistency is established through a time-domain coherent analysis: the r-statistic test. This paper presents a performance study of the r-statistic test for triple coincidence events in the second LIGO Science Run (S2), with emphasis on its ability to suppress the background false rate and its efficiency at detecting simulated bursts of different waveforms close to the S2 sensitivity curve.

In the search for unmodelled gravitational wave bursts, there are a variety of methods that have been proposed to generate candidate events from time series data. BlockNormal is a method of identifying candidate events by searching for places in the data stream where the characteristic statistics of the data change. These change points divide the data into blocks in which the characteristics of the block are stationary. Blocks in which these characteristics are inconsistent with the long term characteristic statistics are marked as event triggers, which can then be investigated by a more computationally demanding multi-detector analysis.

The GEO 600 interferometric gravitational detector took part in an extended coincident science run of the LIGO Scientific Collaboration (S3) that started in November 2003. GEO had recently been upgraded to be the first large-scale fully suspended dual-recycled interferometer in the world and was in the early stages of commissioning in this configuration. In order to prepare the GEO 600 data for the possible extraction of science results and for exchange between analysis groups of the gravitational wave community, the data need to be accurately calibrated. An online, time-domain calibration scheme that was initially developed to calibrate the power-recycled GEO 600 configuration, has been extended to cover the significantly more complicated case of calibrating the dual-recycled interferometer, where the optical response of the instrument is much more difficult to measure and calibrate out online. This report presents an overview of this calibration scheme as it was applied to calibrating the GEO S3 science run data. In addition, results of the calibration process are presented together with some discussion of the accuracy achieved.

The conversion of the read-out from the anti-symmetric port of the LIGO interferometers into gravitational strain has thus far been performed in the frequency domain. Here we describe a conversion in the time domain which is based on the method developed by GEO. We illustrate the method using the Hanford 4 km interferometer during the second LIGO science run (S2).

The German–British laser-interferometric gravitational-wave detector GEO 600 is currently being commissioned as part of a worldwide network of gravitational-wave detectors. GEO 600 recently became the first kilometre-scale interferometer to employ dual recycling—an optical configuration that combines power recycling and signal recycling. We present a brief overview of the commissioning of this dual-recycled interferometer, the performance results achieved during a subsequent extended data-taking period, and the plans intended to bring GEO 600 to its final configuration.

Presented is a summary of studies by the LIGO Scientific Collaboration's Inspiral Analysis Group on the development of possible vetoes to be used in the evaluation of data from the first two LIGO science data runs. Numerous environmental monitor signals and interferometer control channels have been analysed in order to characterize the interferometers' performance. The results of studies on selected data segments are provided in this paper. The vetoes used in the compact binary inspiral analyses of LIGO's S1 and S2 science data runs are presented and discussed.

Searches for binary inspiral signals in data collected by interferometric gravitational wave detectors utilize matched filtering techniques. Although matched filtering is optimal in the case of stationary Gaussian noise, data from real detectors often contain 'glitches' and episodes of excess noise which cause filter outputs to ring strongly. We review the standard χ² statistic which is used to test whether the filter output has appropriate contributions from several different frequency bands. We then propose a new type of waveform consistency test which is based on the time history of the filter output. We apply one such test to the data from the first LIGO science run and show that it cleanly distinguishes between true inspiral waveforms and large-amplitude false signals which managed to pass the standard χ² test. Future searches may benefit significantly from applying this new type of waveform consistency test in addition to the standard χ² test.

We describe a test to distinguish between actual gravitational waves from binary inspiral and false noise triggers. The test operates in the time domain, and considers the time evolution of the correlator and its statistical distribution. It should distinguish true versus noisy events with the same signal-to-noise ratio and chi-square frequency distribution. A similar test has been applied to S1 LIGO data.

Searches for gravitational-wave bursts have often focused on the loudest event(s) in searching for detections and in determining upper limits on astrophysical populations. Typical upper limits have been reported on event rates and event amplitudes which can then be translated into constraints on astrophysical populations. We describe the mathematical construction of such upper limits.

This study describes algorithms developed for modelling interferometric noise in a realistic manner, i.e. incorporating non-stationarity that can be seen in the data from the present generation of interferometers. The noise model is based on individual component models (ICM) with the application of auto regressive moving average (ARMA) models. The data obtained from the model are vindicated by standard statistical tests, e.g. the KS test and Akaike minimum criterion. The results indicate a very good fit. The advantage of using ARMA for ICMs is that the model parameters can be controlled and hence injection and efficiency studies can be conducted in a more controlled environment. This realistic non-stationary noise generator is intended to be integrated within the data monitoring tool framework.

We compare a few different strategies for operating a network of interferometric detectors, using as a case study the search for coalescing binaries events with definite intrinsic parameters but unknown source direction and polarization. The strategies considered include the fully coherent one and the detection in twofold or threefold coincidence. Although in Gaussian stationary noise the coherent strategy is optimal, we find that if the SNR available to the network is larger than about 7 the coincidence strategies are competitive. Further we show that a preselection of the events performed on the single detectors may allow us to perform the coherent analysis only over a manageable subset of the data; the preselection introduces a loss in the detection efficiency, which appears to be negligible for those events having a network SNR larger than 6.

We describe the plans for a joint search for unmodelled gravitational wave bursts being carried out by the LIGO and TAMA Collaborations using data collected during February–April 2003. We take a conservative approach to detection, requiring candidate gravitational wave bursts to be seen in coincidence by all four interferometers. We focus on some of the complications of performing this coincidence analysis, in particular the effects of the different alignments and noise spectra of the interferometers.

We present two search algorithms that implement logarithmic tiling of the time–frequency plane in order to efficiently detect astrophysically unmodelled bursts of gravitational radiation. The first is a straightforward application of the dyadic wavelet transform. The second is a modification of the windowed Fourier transform which tiles the time–frequency plane for a specific Q. In addition, we also demonstrate adaptive whitening by linear prediction, which greatly simplifies our statistical analysis. This is a methodology paper that aims to describe the techniques for identifying significant events as well as the necessary pre-processing that is required in order to improve their performance. For this reason we use simulated LIGO noise in order to illustrate the methods and to present their preliminary performance.

In this paper, we describe a method for detection of gravitational wave bursts in data from gravitational wave interferometers. The method is based on wavelet transformations, which give a time–frequency representation of data. The bursts are identified by looking for regions in the wavelet domain with an excess of power inconsistent with stationary detector noise. The method is applied simultaneously for two or more gravitational wave detectors, resulting in the production of coincident burst triggers. To describe the statistical properties of the triggers, non-parametric statistics are used. This makes the method more robust against uncertainties in the detector noise. The excess power approach does not require an accurate knowledge of the burst waveforms allowing detection of signals from poorly modelled or unmodelled gravitational wave sources.

One of the brightest gamma ray bursts ever recorded, GRB030329, occurred during the second science run of the LIGO detectors. At that time, both interferometers at the Hanford, WA LIGO site were in lock and were acquiring data. The data collected from the two Hanford detectors were analysed for the presence of a gravitational wave signal associated with this GRB. This paper presents a detailed description of the search algorithm implemented in the current analysis.

The detection of gravitational waves from astrophysical sources of gravitational waves is a realistic goal for the current generation of interferometric gravitational-wave detectors. Short duration bursts of gravitational waves from core-collapse supernovae or mergers of binary black holes may bring a wealth of astronomical and astrophysical information. The weakness of the waves and the rarity of the events urges the development of optimal methods to detect the waves. The waves from these sources are not generally known well enough to use matched filtering however; this drives the need to develop new ways to exploit source simulation information in both detection and information extraction. We present an algorithmic approach to using catalogues of gravitational-wave signals developed through numerical simulation, or otherwise, to enhance our ability to detect these waves. As more detailed simulations become available, it is straightforward to incorporate the new information into the search method. This approach may also be useful when trying to extract information from a gravitational-wave observation by allowing direct comparison between the observation and simulations.

The algorithms for the detection of gravitational waves are usually very complex due to the low signal-to-noise ratio. In particular, the search for signals coming from coalescing binary systems can be very demanding in terms of computing power, as in the case of the well-known standard matched filter technique. To overcome this problem, we tested a dynamic matched filter technique, based on matched filters, whose main advantage is the requirement of a lower computing power. In this work this technique is described, together with its possible application, as a pre-data analysis algorithm. The results on simulated data are also reported.