Astronomical
imaging with the full diffraction-limited resolution of large optical telescopes
has been an attractive though largely unfulfilled prospect ever since Michelson's
pioneering work in the 1920s with the Mt. Wilson stellar interferometer.
The principle of interferometric imaging is well known and straightforward
- take light from a star incident on 2 small apertures; form interference
fringes by combining the two beams; and measure the position (phase) and
modulation (visibility) of the fringes for a range of interferometer baselines.
A Fourier transform of the results the gives the brightness distribution
of the source. The experimental problem is that, whereas the visibility
can be measured with sensitive modern detectors, the fringe phase is entirely
corrupted due to the different atmospherically disturbed optical paths
from the source to the two apertures. For Michelson this was unimportant
since stellar discs could be assumed to be symmetric and the expected fringe
phase was either 0° or 180°; for imaging of objects of arbitrary
structure it is a crucial difficulty.
The revival of interest in optical
Michelson interferometry stems from the remarkable success it has had in
radio astronomy. Aperture synthesis has been providing diffraction-limited
images of ever increasing resolution over the past 30 years, but only because
the radio interferometer phases are relatively undisturbed by atmospheric
fluctuations. In VLBI, however, where the telescopes are separated by as
much as 10,000 km, large and varying phase uncertainties do become important.
The solution adopted by radio astronomers has been not to consider the
corrupted interferometer phases themselves rather their sums round closed
loops of baselines - the so-called closure phases. Such sums of phases
are completely independent of the atmospheric contributions above each
telescope; they neither represent the whole of the phase information nor
do they present it in a convenient form, but fortunately they almost always
uniquely constrain the images that fit the visibility amplitudes. Radio
images, with dynamic ranges of several hundred to one, of complex jets
in galactic nuclei have been made with arrays of fewer than 10 antennae.
The high quality of these images encourages one to believe that the closure
phase technique may be equally succesful for optical imaging.
The attainment of diffraction-limited
images with large ground-based optical telescopes is an important objective
for many astronomical programmes. The limited resolution set by atmospheric
fluctuations in refractive index can be overcome by applying aperture symthesis
and phase-closure techniques to short-exposure images taken through non-redundant
aperture masks.
Most previous attempts to obtain
high-resolution astronomical images from the ground used short-exposure
'speckle' images obtained by using the whole telescope aperture. Some of
these methods enable the amplitudes, but not the pases, of the spatial
coherence function to be measured and hence does not in general permit
unambiguous image reconstruction. Later developments have enabled some
phase information to be retrieved and true images have been made in a few
cases. The alternative approach is to mask the telescope mirror with an
array of apertures each no larger than the scale size r0 of
the atmospheric fluctuations. Short-exposure images can then be analysed
to give both amplitudes and phases.
The discovery team had already
demostrated that closure phases can be obtained at high light levels. They
have made systematic observations to measure the spatial-coherence function
with this technique at low photon rates and have derived high-resolution
images from the data.
First observing run on the Isaac
Newton Telescope, in November 1985, used the Image Photon Counting System
(IPCS) together with the empty TAURUS box as a bench for the optical components.
A 136-mm focal length lens behind the f/15 Cassegrain focus re-imaged the
pupil at a diameter of 9.0 mm. The aperture mask consisted of four holes
whose separations gave six uniformly space non-redundant baselines in one
dimension: the hole diameter corresponded to 5.6 cm at the pupil and the
unit baseline separation to 16 cm. After passing through the aperture mask
the collimated beam was refocused onto the IPCS at an image scale of 1.24
arcsec/mm, giving 3.8 pixels per fringe for the finest fringes. An area
of 512 pixels × 128 pixels was scanned every 16 ms and the coordinates
of each photon recorded on magnetic tape. A 12-nm (FWHM) interference filter
centred on 512 nm defined the bandwidth and no atmospheric dispersion corrector
was used. Stellar profiles, observed with the whole of the telescope pupil,
had widths of 1.2 arcsec, which corresponds to a value of r0
of approximately 9 cm.
Two of the stars observed were
Lambda Peg, a 3.95-m single star and Phi And, a binary system with a separation
of 0.45 arcsec and magnitude difference of 1.2 m. Lambda is unresolved
at a 1-m baseline, providing a good calibration source, whereas the wide
double is well resolved at the maximum baseline.
Subsequent observing runs have
taken advantage of the new RGO/RSRE imaging box. This contains two easily
accessible optical benches for the reimaging and magnifying optics. Two
more bunary stars were observed in July 1987. A linear 4-hole mask, similar
to the one employed in the first experiments, was used in an almost identical
instrumental configuration. Again both images are diffraction-limited and
have good dynamic ranges. One of the major successes of the run is the
image of Delta Equ. This is the first reconstruction from data obtained
with a 2 m maximum baseline and has a resolution of 50 milliarcseconds.
The images produced show that
good results can be obtained even at low photon rates. For the simplest
image-plane interferometer, such as the one described here, a limiting
magnitude as faint as +11 is expected to be achievable if larger apertures
and wider bandwidths are used. By working at longer wavelengths, where
both the seeing scale size and the atmospheric coherence time increase
favourably, and by correcting for image motions in the focal plane, high-quality
images should be obtained at much fainter limiting magnitudes .
|