Extrasolar planets orbiting nearby stars induced small perturbations to the apparent position of the star against a reference. The same perturbations, however, can be induced by the propagation of light through the Earth's turbulent atmosphere. The differential centroid motion between a target star and a reference star of small angular separation, a double star system, has been investigated. We have verified the theory of differential centroid motion in the 2-20 arcsec separation regime as a function of aperture diameter and integration time. We have conducted a comparison of the traditional centroid differencing technique of planet detection and an alternate technique using the cross-correlation of the two short-exposure speckle-grams to form a separation estimate. The speckle cross-correlation technique can exceed the differential centroid technique in precision, but is a strong function of the effective thickness of the turbulent atmosphere. Even so, we have determined that the atmosphere will allow an 8 meter diameter telescope to achieve standard errors sufficient to detect the presence of Jupiter-sized planets in orbit about Sun-sized stars at 10 pc at a 5 sigma confidence with regular observations of 20 minutes in length when r₀ = 20 cm at λ = 0.9 μm. Achieving this precision in practice will require unprecedented control of systematic errors. Also with the motivation of unprecedented angular resolution astrometry, we have developed a new theory and experimentally verified a piston-phasing technique for array telescopes that resulted in a diffraction-limited image from a 6.87 meter baseline imaging array at a wavelength of 2.2 microns, the highest resolution image ever obtained to date of a star from the ground. Finally, we have demonstrated the feasibility and utility of predicting centroid motion over a 6.87 meter baseline imaging array. We have shown how spatial information improves the prediction compared to temporal information alone, particularly in poor seeing conditions. We have verified that for moderate conditions at the MMT site, simple delta rule training of a linear predictor yields excellent results and that the complexity of a two-layer perceptron neural network based predictor is not necessary.
展开▼
