Summary of recent preliminary LIGO results

Alan Wiseman, University of Wisconsin-Milwaukee for the LIGO Science Collaboration

The LIGO Lab and the LIGO Scientific Collaboration (LSC) continue to interweave detector commissioning and data taking with data analysis and the presentation and publication of scientific results. In the Spring 2003 issue of Matters of Gravity, Gary Sanders summarized the status of the LIGO detector commissioning effort leading up to the first Science Run (S1) and recapped the preliminary ``upper-limits" results that Albert Lazzarini had reported at the AAAS meeting in Denver. In last Fall's issue, Stan Whitcomb reported on the completion of a second LIGO data taking run, S2. In the intervening year, the LIGO Scientific Collaboration has published a sequence of five major articles that culminate the work on the S1 data set [1-5], as well as numerous conference proceedings and technical reports [6]. We also completed another two-month science run (S3) in early January of 2004. In this note, I would like to briefly recap the published S1 results, as well as summarize the preliminary S2 and S3 results that were presented at the Denver APS meeting in May and the Dublin GR17 meeting in July.

The data analysis effort within the LSC is currently divided among four groups reflecting four distinct source types: the Inspiral Upper Limits Group, the Stochastic Background Upper Limits Group, the Pulsar (continuous waves) Upper Limits Group and the ``Burst" Upper Limits Group. When the groups were formed some years ago, the qualifier ``upper-limits" was included in the group name to reflect the fact that the sensitivity of the instrument during the early running would likely lend itself to only setting upper limits on flux strength and population models. However, as the sensitivity of the detectors has improved, each group has begun to set their sites on true detections, and thus the use of the qualifier is falling by the way-side.

However, the first article [1] to wind its way through both the internal LSC review1 and the external peer-review process was not an astrophysical paper originating from within the analysis groups. Rather, the first paper gave a detailed description of the configuration and performance of the LIGO detectors and the British-German GEO detector during the 17 day S1 data run in August and September of 2002. This ``detector" paper was then followed by ``upper-limits" papers from each of the four search groups. Although the final astrophysical results in these four papers do not challenge any existing theories, they do present complete analyses which show how to search real data for small gravitational wave signals and how to translate those searches into astrophysical limits.

The presentation of the preliminary S2 and S3 results at the APS and GR meetings followed a pattern similar to the S1 publications: a summary presentation describing the status of the detectors was followed by a talk from each of the four analysis groups. Although the results are summarized below, we invite everyone to take a look at the vu-graphs that were presented [7]. The central feature of the summary talk was to show the dramatic improvement in detector sensitivity over the last few years. Figure 1 shows that the commissioning efforts are paying off, and the detector is nearing the design sensitivity.

Figure 1: Best Strain Sensitivities for the LIGO Interferometers. LHO refers to the Hanford Observatory. LLO refers to the Livingston Observatory.
In the published S1 analysis [2], the Burst Group took on the daunting task of limiting the rate and strength of poorly modeled burst sources of gravitational waves. [It is hard to look for something when you don't know what it looks like.] The primary result of this analysis was to quantify an excluded (or low probability) region in the rate versus signal strength plane. Although a similar analysis is being repeated with the more sensitive S2 and S3 data, the Burst Group has also added a new type of search to the mix: a ``triggered" burst search. Some preliminary results from this search were presented at the APS and the GR17 meetings. During S2 an especially strong $\gamma$-ray burst (GRB 030329) popped off nearby ($z=0.1685$). The data from the two Hanford detectors were cross correlated in a 180 second interval surrounding the arrival time of the burst. [Unfortunately the Livingston and GEO detectors were off-line during the burst.] This was compared to a threshold set by a similar analysis conducted on data taken well away from the burst arrival time. Although no gravitational wave burst was detected, a strain upper limit of $h_{rss} \approx 6\times10^{-21} {\rm Hz}^{-1/2}$ was set. The Inspiral Group searched for non-spinning binary neutron star inspirals in 236 hours of S1 data [3]. Unfortunately, the sensitivity of the instruments only allowed them to see inspirals within a portion of the Milky Way Galaxy. The group used a ``loudest event" statistic to determine the upper limit on the event rate of neutron star coalescences. The published S1 paper gives a ninety-percent confidence limit on the event rate of binary neutron star coalescences as

$\displaystyle R_{S1} < 1.7 \times 10^2 \; {\rm per \;year \; per \; MWEG}
\; .$      

Here, MWEG means Milky Way Equivalent Galaxy. During the S2 run the instrument was considerably more sensitive and could see beyond the Galaxy. For example, the Livingston Detector, could reach M31. However, because the group is beginning to look beyond simply setting upper limits and toward a possible detection, they have modified their analysis pipeline accordingly. Only coincident data (when two or more interferometers were operating) was included in their upper limit. This reduced the ``live time" to about 355 hours and therefore reduced the upper limit that might have been attained had they analyzed the ``singles" S2 data in the same way they analyzed the S1 data. Nevertheless, an improved preliminary upper limit of
$\displaystyle R_{S2} < 50 \; {\rm per \;year \; per \; MWEG}$      

was obtained. As the Inspiral Analysis Group moves forward, they will be casting a broader net and include searches for binary black holes with masses greater than $3 M_{\rm Sun}$. Searching for these signals presents a special challenge as neither the waveforms or the population models are well known. They will also be looking for inspiraling massive halo objects (MACHOs) with masses in the range of $0.2 - 1.0M_{\rm Sun}$. The MACHO search presents a different challenge as the template waveforms for these low mass systems spend much longer in the LIGO sensitivity band than do the neutron star binaries.

In analyzing the S1 data [4], the Pulsar Analysis Group used two very different techniques to perform a search aimed at setting an upper limit on the ellipticity of PSR 1939+2134: a Bayesian time-domain search and a frequency-domain search. Several features of this search are worth noting. First, a directed search at a pulsar with known frequency and sky location is less computationally intensive than a search for unknown pulsars in a broad region of the sky or in a broad frequency band. Second, during S1, both GEO and LIGO were operating, and therefore the results combine information from both detectors. Third, the fact that two different analysis methods arrived at essentially the same results instills confidence in the implementation of both methods. The result is a bound on the ellipticity of this pulsar of $\epsilon_{S1} < 2.9\times 10^{-4}$. Using the S2 data, the Pulsar Analysis Group has now applied the ``time domain" search method to 28 known isolated pulsars with good timing data. In the case of PSR 1939+2134, this analysis has improved the upper limit on the ellipticity by about an order of magnitude. A full list of the ellipticity bounds should be published soon. The Pulsar Analysis Group also has ambitious plans for doing wide parameter space searches to look for unknown pulsars.

A search through the S1 data for a stochastic background of gravitational waves has also been completed and published [5]. The basic idea is to cross-correlate data from two detectors with the appropriate range of time lags. The result for the Hanford-Livingston cross correlation is a bound on $\Omega_{gw}h_{100}^2 < 23 \pm 4.6$ in the frequency band $64-265$Hz, where $\Omega_{gw}$ is the energy density per logarithmic frequency interval divided by the energy density required to close the universe, and $h_{100}$ is the Hubble constant in units of $100{\rm km/sec \; Mpc}$. At the GR17 meeting, a preliminary S2 result of $\Omega_{gw}h_{100}^2 < 0.018 (+0.007,-0.003)$ in the frequency band $50-300$Hz was presented. Based on sensitivity estimates from the S3 noise curves, we expect to be able to place a bound of roughly $5\times 10^{-4}$ in the same frequency band.


Jorge Pullin 2004-09-10