Welcome to the EGRET Pulsar Search Homepage
The EGRET Pulsar Search Collaboration consists of:
- Danny Koh (Caltech)
- Adam Chandler (Caltech)
- Dick Lamb (Caltech)
- Daryl Macomb (Goddard Space Flight Center)
- John Mattox (Boston University)
- Tom Prince (Caltech)
- Paul Ray (Naval Research Laboratory)
The gamma-ray source Geminga was first discovered in 1972 by the SAS-2
satellite. Despite being a very bright source
at energies above 100 MeV, it has no counterparts at any other
wavelengths, except for the X-ray band. Consequently, its true
identity was largely a mystery for the 20 years following its discovery.
A breakthrough came in 1992 when observations by the ROSAT satellite
discovered soft X-ray pulsations from Geminga, thus revealing it
to be a radio-quiet gamma-ray pulsar . This raises
the tantalizing possibility that many of the unidentified sources
detected by the EGRET instrument on-board the Comption Gamma Ray
Observatory could be Geminga-like pulsars. We have therefore
embarked on a systematic search for pulsations from these unidentified
Pulsation Search Technique and Strategy
Due to the extremely low event rate even from bright gamma-ray
sources (about 150 photons per day), long observation
intervals of several weeks are typically required to attain
a sufficiently high signal-to-noise ratio for detection.
In 1987, Buccheri, Ozel and Sacco wrote that
although the Fast Fourier Transform (FFT) could reduce computation
time by about orders of magnitude over epoch folding in the search for
high-energy periodicity, ``the enormous problems of computer
memory to face (arrays of 10^10 locations) make the search
unpracticable'. Fortunately, the exponential rate at
which computational power evolved over the past decade has
now made this search strategy highly feasible.
The Center for Advanced Computing Research (CACR) at Caltech
owns the 512-node Intel Touchstone Delta which has a
memory of 8 Gigabytes and an attained operating speed of 10 GFLOPS.
The parallel FFT and the associated
pulsation detection software which we have implemented on
the Delta are capable of processing a billion-point time series
in less than 5 minutes. In view of the spin-down expected
of most rotation-powered pulsars, a wide range of frequency
derivatives must be searched. For each trial frequency derivative,
the search code performs the following sequence of procedures:
Candidate events are then scrutinized using more sensitive
techniques like epoch-folding.
The following is the power spectrum from an FFT of 2^28 points
of data for the EGRET viewing period 1.0 observation of Geminga.
The double-peaked nature of the Geminga pulse profile results
in most of the power appearing at 8.4 Hz, which is the 2nd harmonic
of the pulsar rotational frequency. Unfortunately, our Geminga
search results are being described here instead of Nature because
of the Rossi Prize work done by Jules Halpern and Steve Holt
with the ROSAT data in the interval between the EGRET data acquisition
and its public availability.
- Accelerate the arrival times of the photons corresponding
to the expected frequency derivative
- Perform the FFT
- Calculate the power spectrum
- Normalize the power spectrum by the global mean power
- Sum the normalized powers over a range of harmonics corresponding
to each fundamental frequency
- Save all candidate frequencies and frequency derivatives at which
the powers exceed pre-determined thresholds.
We have begun extensive searches on archival data for GRO J1744-28,
GRO J2019+40 and GRO J0617+22. No Geminga-like pulsars are present in
these data to the limits of our sensitivity, and we can also rule out
strong Vela-like pulsars.
To date, we have consumed more than 80,000 CPU node hours,
corresponding to about 561 teraflops, on the Delta.
We are carrying out searches for pulsations
at higher frequency derivatives, making use of all available
supercomputer time. In addition, searches on archival data for other
promising candidates are in progress.