The Net Advance of Physics: The Nature of Dark Matter, by Kim Griest -- Section 7C.

The MACHO Collaboration Experiment

The MACHO experiment is led by Charles Alcock and is a

collaboration of Physicists and Astronomers from Lawrence

Livermore National Lab, The UC Berkeley Center for Particle

Astrophysics, Mount Stromlo and Siding Spring Observatory, The

University of Washington, Oxford, McMaster, and UC San Diego.

We have essentially full-time use of the 1.27-meter telescope at

Mount Stromlo Observatory, Australia, for a period of about 8

years from July 1992. In order to maximize throughput a dichroic

beamsplitter and filters provide simultaneous images in two

passbands, a `red' band (approx. 5900-7800 Å) and a `blue' band

(approx. 4500-5900 Å). Two very large CCD cameras [42] are

employed at the two foci; each contains a mosaic of

pixel Loral CCD imagers, giving us a sky coverage of 0.5

square degrees. Observations are obtained during all clear nights

and part nights, except for occasional gaps for telescope

maintenance. The default exposure times are 300 seconds for LMC

images, 600 sec for the SMC and 150 seconds for the bulge, so over

70 exposures are taken per clear night.

Photometric measurements from these images are made with a

special-purpose code known as SoDoPHOT [43], derived from

DoPHOT [44]. For each star, the estimated magnitude and error

are determined, along with 6 other parameters (quality flags)

measuring, for example, the crowding, and the of the

point-spread-function fit. It takes about an hour on a Sparc-10 to

process a field with 500,000 stars, and so with the computer

equipment available to us we manage to keep up. The set of

photometric data points for each field are re-arranged into a

time-series for each star, combined with other relevant

information including the seeing and sky brightness, and then

passed to an automated analysis to search for variable stars and

microlensing candidates [45]. The total amount of data collected to

date is more than two Terabytes, but the time-series database used

for analysis is only about 100 Gbytes.

BIBLIOGRAPHY