The LIGO Web Newsletter

Front Page

Livingston

Caltech

LIGO MIT News

Preparing to Search for Gravitational Wave Bursts

Preparing to Search for Gravitational Wave Bursts

- Contributed by Erik Katsavounidis

One type of gravitational radiation LIGO is expecting to detect is that attributed to bursting episodes in the universe. In preparation of LIGO's upcoming first science run, and armed with the data collected when all three LIGO detectors ran simultaneously for the first time during the January 2002 engineering run (E7), we will shortly begin pursuing the end-to-end analysis exercise of that data in search for sources of burst-type gravitational radiation. The LIGO Data Analysis System (LDAS) at MIT will be the primary venue where this exercise will take place.

Bursts of Gravitational Waves: Burst sources emit gravitational radiation lasting just a few cycles within the characteristic frequency band of LIGO, and they generally come with little or no indication of their waveforms. This last aspect has become, to good extent, part of the "bursts" definition at least in the context of the initial LIGO searches (unmodeled bursts). Among such anticipated sources are supernovae explosions, as long as they are asymmetric. Depending on the degree of asymmetry during the explosion, a fraction of the total gravitational binding energy is emitted in the form of gravitational waves. It will be hard to miss such a star collapse if it were to happen somewhere in our galaxy (a collapse emitting 10^-3 of the total energy into gravitational waves will result to an rms strain at the level of 10^-18 in LIGO's band). Unfortunately, such an event has a mean rate of the order of one in 35 years, thus in order to be able to see a few events in a year's worth of observation time we should be able either to survey a large volume of the universe or equivalently maintain high sensitivity. The search for bursts may achieve such high sensitivity by implementing correlation methods between the burst triggers coming from the three LIGO detectors (in the so-called internal or self-trigger mode) as well as by exploiting the triggers coming from the gamma-ray burst and neutrino detectors (external trigger mode).

LDAS at MIT: Since August 2001, the LIGO Data Analysis System (LDAS) has been installed and running at the LIGO Laboratory at MIT awaiting the moment to start looking at interferometer data of significant sensitivity. The LDAS-MIT configuration is comprised of 16 PC's each with 1.1GHz AMD Athlon processor and 1GB RAM. This clustered configuration of PC's running linux is often referred to as a "beowulf" cluster. The system also includes a number of multi-processor PC's serving as "master" and data-conditioning nodes in the cluster, as well as two quad-processor SUN^TM units that act as the frame (data) and database servers. In its current configuration, the LDAS-MIT system offers a bit over 1TB of disk storage while it is equipped with a stand-alone AIT-2 tape unit. Here is a recent picture of the MIT cluster, celebrating its ninth month:

Burst Search at LDAS-MIT: The Bursts Working Group of the LIGO Scientific Collaboration has undertaken the task of implementing the end-to-end Bursts analysis pipeline. The group's main challenge is to prepare a pipeline able to detect an unmodeled, possibly very weak signal, to distinguish it from detector noise and to interpret the result in an astrophysically meaningful way. There are several pieces that make up this challenge, some highlights of which may be itemized in the following.

• Data are brought in from the sites to the local LDAS cluster.

• Bursts algorithms are then executed on the LDAS system for the identification of possible candidate events. The algorithms currently implemented in LDAS are based on simple time-domain filters (Arnaud et al, Daw) or time-frequency analysis starting with Fourier transforms of the time domain detector data (Anderson et al, Sylvestre). Events are identified whenever the total power of the data in a frequency band of interest exceeds by some statistical significance the noise power or by looking in a more detailed way at the clustering of "loud" events in the time-frequency plane.

  

  Figure 2: A time-frequency plane obtained from stacked-up periodograms of the gravitation wave channel from the Hanford 2K detector during E7. Two clusters of "loud" events in this plane are identified with the two thin, black rectangular boxes lying vertically at left and right in the figure. Several line features (horizontal stripes) can be seen, with most of them identified as harmonics of the 60Hz power lines. This is the time-frequency cluster method developed by Sylvestre.

  Once identified, the events' time-frequency features (time stamp, duration, central frequency, bandwidth), as well as their significance, are written to the database.

• At the same time, transient algorithms are executed on auxiliary and environmental channels on the DMT system. These channels are expected to have very much reduced (with respect to the gravitational wave channel) coupling or no coupling at all to gravitational waves of astrophysical origin. Thus they define the so-called "veto" events and, just like with the bursts candidate events, their time-frequency profile and significance are written to the database.

• An event list for each detector is then composed by requiring the anti-coincidence of the candidate and "veto" events.

  

  Figure 3: Burst candidate events as identified by the time-frequency cluster method before and after the application of a veto based on glitches from the Pre-Stabilized Laser.

• Apply the above single-detector pipeline to all instruments; perform coincidence analyses with multi-detector outputs as well as with triggers from gamma-ray burst (for E7) detectors.

It's an exciting moment for the burst search with the LIGO detectors, and a moment of intense activity across the LIGO Scientific Collaboration's Bursts Working Group in the fine-tuning and preparation of its pipeline for the imminent science run. The E7 analysis reflects a milestone in this effort. Expect the unexpected and stay tuned!

Onwards!

to the LIGO MIT Home Page

Or  Return!

to the Newsletter Front Page