HST Coronagraph

Indirect techniques hold the greatest promise for the first detection of extra-solar planets. However, if Jupiter sized planets are common, the first direct detection could come from HST. Unfortunately, HST with its present and planned instrumental complement is not able to detect extra-solar planets because of the high levels of scattered light in the image wings around a star. Techniques for the suppression of this light exist, and can be expected to give HST the capability to directly image Jupiter sized planets out to a distance of about 5 parsec. Such an instrument, could be flown in the scheduled HST servicing mission in 2003.

If such an instrument is flown, several of its other scientific applications have also long been specifically identified as high priority endeavors. Furthermore, several of these applications also bear directly on our desire to understand and characterize the formation and evolution of planetary systems. In particular, in addition to allowing extra-solar planet detection, and some limited spectral investigation, the coronagraph will let us study star formation and proto-planetary disks and search for sub-stellar brown dwarf companion stars. All of these studies are presently limited by the achievable contrast in both HST and ground based telescopes.

HST with a coronagraph would combine the high spatial resolution essential to achieving sufficient signal to noise on the planet, with low scattered light levels. In principle a system with active optics on the ground might achieve such a combination, however, we feel that the space based option is much easier to implement because the temporal bandwidth of the active optics control system is five orders of magnitude lower, and the stroke required is some two orders of magnitude smaller. In addition scintillation (time dependent amplitude variations) noise is absent in space. The correction of scintillation noise from the ground is yet to be demonstrated.

The HST coronagraph would reduce or eliminate background light from the central source by a variety of techniques. In the visible on HST we are presently limited by detector defects, filter ghosts, diffraction and mirror polishing defects. All of these problems can be greatly reduced or eliminated in a design that concentrates on them. The elements of a workable design are clear. Detector problems, and filter ghosts would be eliminated as problems by ensuring that the light from the bright central source is largely eliminated before reaching the filters or detector.

Light would enter the instrument after reflecting from the HST primary and secondary mirrors, possibly after an additional reflection from a super-smooth and clean pick-off mirror. The pick-off mirror is necessary if a radial instrument (to replace the present Wide Field Planetary Camera) is chosen. The beam would be corrected for HST's spherical aberration, possible off axis astigmatism, and for time dependent pointing and focus variations.

The HST is known to suffer from pointing variations at the level of 5 milli-arcseconds rms, with a dominant period of 12 seconds, and focus changes with a peak to peak amplitude of 1/20 wave and a period of 95 minutes. Low spatial frequency and low temporal bandwidth wave-front sensing is therefore necessary, with the bright central star providing an ideal phase reference. Electro-strictive actuators similar to those flown on WFPC2 can be used to steer the corrected beam. If placed on a mirror with sufficient optical power, they will also provide sufficient stroke for focus correction (about 5 microns throw at a magnification of 10 is needed). In addition, a higher frequency fixed correction would be applied to correct for the fixed pattern of phase errors introduced by the primary and secondary mirrors. This correction would only need to work at the field position of the central star, and would need to have approximately 40 actuators across the pupil diameter, in order to suppress the scattered light due to mirror imperfections within the central arcsecond or so.

With this optical correction in place, the unattenuated image reaching the first focal plane would be very close to diffraction limited, and a coronagraph can then further reduce the wing intensity by factors of up to 1000. We have modeled this system in detail with the measured phase errors that are present on the HST being corrected at various levels of accuracy. A graded Gaussian focal plane mask with a width of 0.5 arcseconds was used. The pupil plane mask used in the simulations was not graded, and no doubt performance improvements in the coronagraph would be possible if it were. One immediate conclusion from this modeling is that one is always limited in practice by residual wave-front errors. For any level of pupil correction, it is possible to design a coronagraph to take advantage of it.

The filter wheel would be positioned after this camera. It would contain both broad band filters (for sensitivity and wavelength coverage) and narrow band filters (for searches for point or extended sources "between the airy rings" at still higher contrast). It should also contain a number of zero focal shift polarizing filters both for circumstellar scattering studies, and for additional suppression of instrumental scattered light. A selection of polarizers which are anti- reflection coated to match the various visible filters would be ideal. The double bounce ghost which arises between the filter pairs can be moved to one side of the central peak by tilting the polarizer wheel axis by a few degrees. Then spacecraft roll can be used to cover the region of lower contrast

The detector should be backside illuminated to eliminate scatter from front-side electrode structures. It should have a blackened spot deposited over the central 0.2 arcsecond region covered by the central bright object. This will eliminate scatter and bleeding effects caused by the brightest part of the residual stellar image. The detector itself should be tilted to shift detector-filter ghosts to one side. (this can be done in the slow beams we need for adequate image sampling without significant focus errors) The single detector should critically sample the beam at blue wavelengths, which makes resampling (for small image shifts in data reduction) a mathematical possibility. It is very important to control Charge Transfer Efficiency in the detector. This can be done at least partially by operating at very low temperatures (<100C), and by selecting devices made with very pure silicone - eliminating a population of traps (CTE is the major source of contrast loss for WFPC2 beyond about 5 arcseconds from the image core). Condensation on the detector might be eliminated by fitting a warm window below the filter wheel but is not expected to be a major problem at visible wavelengths. We would rather operate the detector without a window if this is technically possible.

There are two requirements that must be met if a planet is to be detected. First, the ratio of its brightness to the residual star light at its location in the image must be big enough. A point-like source of 1/10 the local wing intensity in a broad-band image is easy to detect for example by rolling the spacecraft and comparing the images, or by local profile fitting. Conversely, a source with a signal only 1/1000 of the average wing intensity is in practice very hard to detect, because it is unlikely that the instrument response will be stable at this level. We feel that a contrast of 1/100 represents a reasonable detectability threshold.

The second requirement for detectability is that imposed by photon statistics. The signal in an Airy disk from the background must fluctuate much less in the integration time used than the signal from the planet. Clearly, a reasonable requirement is that the survey of candidate stars which have sufficient contrast from the first criterion, be completed in a year. It turns out that this is not a problem, given that the first criterion limits one to about 30 candidates, and the integration time needed for a zeroth magnitude star at 5 parsecs would be about 40 minutes for a signal to noise of 5. Note that the signal to noise goes as the star brightness (not the square root), because the background decreases as the signal gets weaker.