Imaging Circumstellar Environments with a Nulling ... - CiteSeerX

Imaging Circumstellar Environments with a Nulling Interferometer Philip M. Hinz , J. Roger P. Angel , William F. Homann, Donald W. McCarthy Jr. , Patrick C. McGuire , Matt Cheselka , Joseph L. Horay , & Neville J. Woolf  Steward Observatory, University of Arizona, Tucson, AZ ySmithsonian Astrophysical Observatory, Cambridge, MA

Extrasolar planets must be imaged directly if their nature is to be better understood. But this will be dicult, as the bright light from the parent star (or rather its diracted halo in the imaging apparatus) can easily overwhelm nearby faint sources. Bracewell has proposed1 a way of selectively removing starlight before detection, by superposing the light from two telescopes so that the stellar wavefronts interfere destructively. Such a 'nulling' interferometer could be used in space to search for extrasolar Earth-like planets through their thermal emission and to determine through spectroscopic analysis if they possess the atmospheric signatures of life2?4. Here we report mid-infrared observations using two co-mounted telescopes of the Multiple Mirror Telescope that demonstrate the viability of this technique. Images of unresolved stars are seen to disappear almost completely, while light from a nearby source as close as 0.2 arcsec remains, as shown by images of Betelgeuse. With this star canceled, there remains the thermal image of its surrounding, small dust nebula.In the future, larger ground-based interferometers that correct for atmospheric distortions (using adaptive optics), should achieve better cancellation, allowing direct detection of warm, Jupiter-size planets and faint zodiacal dust around other nearby stars5. With a conventional telescope, the very high contrast ratio needed to resolve a planet from a star requires diraction-limited resolution several times sharper than the angular separation. At the longer thermal wavelengths, of most interest for imaging warm planets, telescope apertures of tens of meters would be needed. In Bracewell's interferometric method two small apertures suce, provided the stellar wavefronts are exactly superposed out of phase with no relative tilt, so the cancellation is in detail across the whole pupil. Then the transmission pattern on the sky, for monochromatic light, is one of nely spaced fringes given by d (1) T () = sin2 ( );  where d is the element spacing,  is the wavelength of observation, and  is the angular distance from the line on the sky through the star perpendicular to the interferometer baseline. Our test of Bracewell's concept at the MMT used two telescopes spaced by 5 m with detection at 10 m wavelength. Then the fringe spacing on the sky is 0.4 arcsec, and full constructive interference for a source is only 0.2 arcsec from the nulled star (Equation 1). The optical con guration (Fig. 1), is based on previously proposed designs with a single pass beamsplitter6 7 for a single nulled output. The wavefronts are translated for superposition without relative rotation or tilt. In operation, the 10 m image is observed live on a computer monitor while the beamsplitter is translated to adjust path length. The star intensity remains steady until the path-lengths are nearly equal, when it is seen to icker. This is caused by atmospheric turbulence which induces path-length changes of 5 m, enough to shift the interference randomly between constructive and destructive states relative to its non-interfered ux. The minima are deepest when the average phase dierence between the wavefronts is 180 degrees and the average slopes are the same (that is, the individual images are coincident). ;

{2{ Figure 2a shows the brightest and faintest snapshot images of the star  Tauri, which is unresolved at our baseline. The characteristic of the interferometer to cause the entire stellar Airy pattern to disappear is clearly shown. The right hand nulled image has a peak intensity 4.0% and a total integrated ux of 6.0% that of the left-hand constructive image. This small residue arises because phase and tilt dierences between the beams are not ideal or stable through even the best nulled exposure, and because higher-order atmospheric aberrations remained uncorrected. But already from this single, short exposure we can deduce that there are no companions lying in the constructive fringes as close as 0.2 arcsec from the star, to a level 3.5 magnitudes (25 times) fainter than the star. Weather did not allow observations through the night for rotated baseline orientation, as prescribed by Bracewell; combination and comparison of such data would have allowed the detection of any companion more than 1% of the star's brightness with separation  0.2 arcsec. The ability of the interferometer to directly reveal emission sources close to a nulled star is illustrated by images of late-type giant stars  Orionis (Fig. 2b) and R Leonis. The faintest nulled images are still 36% and 24% respectively of the integrated ux of the brightest image. The large residual uxes arise because both stars have surrounding warm dust nebulae with signi cantly more than a 0.2 arcsec diameter, and thus lying mostly outside the central null fringe. The stellar discs themselves are small enough (0.04 arcsec diameter) that only 0.4% of their ux is transmitted at the limb. A feature in our implementation of the nulling interferometer, dierent from Bracewell's original conception, is that the focal plane shows an image of the eld about the nulled star. The image formed is the object ux multiplied by the transmission pattern of equation (1) and convolved with the point-spread function (PSF) of the individual elements. As the PSF is broader than the transmission pattern, the interference fringes are not visible in the focal plane. Thus true images are formed of smooth, extended features such as circumstellar dust, whose structure is on a scale larger than the fringe spacing.  Ori and R Leo are shown in Figure 3 with increased signal-to-noise in composite images, obtained from the 12 best nulled images by shifting to a common center and adding, after repixelization. In the case of R Leo, the residual image with 1.7 arcsec FWHM is barely wider than the measured constructive image width of 1.5 arcsec, indicating a nebular width  1 arcsec. But Betelgeuse's dust nebula is clearly resolved, showing a FWHM of 2.4 arcsec and signi cant asymmetry. At 1% of the stellar intensity it extends radially 4 arcsec from the star. Our image with the star interferometrically removed is the rst to show the nebula directly. Previous work has established the amount of dust and its spatial extent from spectral analysis and interferometric synthesis. Thus Danchi et al.9 have created visibility curves for  Orionis and R Leonis which have values of approximately 60% and 70% respectively for a 5 m baseline at 11 m. Calculating our visibility in the same manner, using  Tauri as our calibration for unit visibility we obtain good agreement: 534% for  Orionis and 695% for R Leonis. Figure 4 shows dierence images formed by subtracting destructive from constructive images, each formed by recentering and co-adding the best individual exposures. The result for R Leonis (Fig. 4a) shows a well de ned diraction pattern, out to the third Airy ring which is accentuated to 0.4% of the peak by the central obscuration of 10% of the diameter. When a dierence image is formed for  Ori, a skewed Airy pattern results; however, by introducing a shift of 0.3 arcsec between the images, the nearly symmetric pattern shown (Fig. 4b) is recovered. The positions of the subtracted stars are shown by the crosses in Fig. 3, as determined by the osets used to produce the symmetric patterns of Fig. 4. The oset for  Orionis does not seem to be an eect of our instrument since it tracks eld orientation in dierent exposures. We thus conclude that the center of gravity of the nebular emission is displaced from the star. The oset angle is approximately 25 degrees East of North. Asymmetries in the photosphere and immediate surroundings

{3{ of  Orionis have been observed in H emission10, the UV continuum11 and 7 mm radio observations12. Bester et al.13 speculate the component of 10 m emission they saw from newly forming dust in 1994 was most likely asymmetric. However, the oset in center of gravity of the extended 10 m emission has not been apparent in previous observations which were limited in their baseline orientation. In the present system the best-nulled short exposures simply re ect the best destructive interference found among many images formed from atmospherically perturbed wavefronts. Change in phase and dierential image motion over the integration time of the exposure prevent perfect starlight cancelation. From the measured rate of phase change, we expect these eects alone to result in a residual stellar intensity of 3% for  Tauri14. Uncorrected higher order errors across each aperture such as astigmatism and focus also contribute to the residual intensity. Based on Noll's analysis15 a dierential phase error  = 0.5(d/r0)5 6 is expected, adding a further 2% transmission of  Tauri (Fried's length, r0 = 3.7 m), for a total residual ux of 5%, in agreement with the measured value of 6%. Future systems with phase and wavefront errors corrected in real time by adaptive optics will largely eliminate such residual transmission. Correction of path length to 10 nm rms and high order phase dierences to 100 nm are projected, for star nulling at the 10?4 level. This development cannot be undertaken with the present instrument, since the MMT has now been been dismantled, but a similar and much more powerful nulling con guration is planned for the Large Binocular Telescope (LBT), with 8.4 m primary mirrors also on a common mount16. With still a relatively short baseline, (14.4 m center to center), the null fringe will be broad enough to suppress the  1 marcsec discs of nearby stars by a factor of 10,000 at 10 m, so full advantage can be taken of the deep null of well corrected wavefronts. Sensitivity to faint sources is determined by thermal background shot noise, which will be kept low in the LBT by making the adaptive correction at deformable secondary mirrors, with no additional optics. Thus there will be, as in Fig. 1, only three warm re ecting surfaces contributing to the background before the beam combination is made with cooled optics in a large cryostat. In this way, Bracewell interferometry from the ground should have the sensitivity, at 5 m wavelength, to nd warm giant extra-solar planets to a 10 pc distance, to analyze them spectroscopically, and, at 10 m, to detect zodiacal dust clouds at solar level5. Zodiacal cloud measurements are crucial to the design of space systems to detect extra-solar Earth-like planets, because bright dust could overwhelm the much fainter planet signal 17. RM S

=

{4{

REFERENCES 1. Bracewell, R.N. Detecting nonsolar planets by spinning infrared interferometer, Nature 274, 780-781 (1978). 2. Angel, J.R.P., Cheng, A.F., & Woolf, N.J. A space telescope for infrared spectroscopy of earthlike planets, Nature 322, 341-343 (1986). 3. Leger, A., Mariotti, J.M., Menesson, B., Puget, J.L., Rouan, D., & Schneider,J. Could we search for primitive life on extrasolar planets in the near future? Icarus, 123, 249-255 (1996). 4. Woolf, N.J., & Angel, J.R.P. Planet Finder options I: new linear nulling array con gurations, in Planets Beyond the Solar System and the Next Generation of Space Missions, D. Soderblom, ASP Conf. Series, 119, 285-293 (1997). 5. Angel, J.R.P. & Woolf, N.J. The LBT: A unique scienti c and technology precursor to Planet Finder, in Planets Beyond the Solar System and the Next Generation of Space Missions, D. Soderblom, ASP Conf. Series, 119, 207 (1997). 6. Angel, J.R.P. Use of a 16m telescope to detect Earthlike planets, in The Next Generation Space Telescope, Ed. P.Y. Bely, C.J. Burrows, G.D. Illingworth, p. 81-94. Baltimore: Space Telescope Science Institute (1990). 7. Angel, J.R.P., Burge, J.H., & Woolf, N.J., Detection and spectroscopy of exo-planets like Earth Proc. SPIE 2871, 516-519 (1996). 8. Homann, W.F., Hora, J.L., Fazio, G.G., Deutsch, L.K., & Dayal, A., MIRAC2, a mid-infrared camera for astronomy Proc. SPIE 3354, in press (1998). 9. Danchi, W.C., Bester, M., Degiacomi, C.G., Greenhill, L.J., & Townes, C.H. Characteristics of dust shells around 13 late-type stars Astron.J. 107, 1469-1513 (1994). 10. Hebden, J.C., Christou, J.C., Cheng, A.Y.S., Hege, E.K., Strittmatter, P.A., Beckers, J.M., & Murphy, H.P. Two-dimensional images of alpha Orionis Astrophys.J. 309, 745-754 (1987). 11. Gilliland, R.L., & Dupree, A.K., First images of the surface of a star with the Hubble Space Telescope Astrophys.J. 463, L29-L32 (1996). 12. Lim, J., Carilli, C.L., White, S.M., Beasley, A.J., & Marson, R.G. Large Convection cells as the source of Betelgeuse's extended atmosphere Nature 392, 575 (1998). 13. Bester, M., Danchi, W.C., Hale, D., Townes, C.H., Degiacomi, C.G., Mekarnia, D., Geballe, T.R. Measurement at 11 micron wavelengths of the diameter of  Orionis and  Scorpii, and changes in eective temperature of  Orionis and very recent dust emission Astrophys.J. 463, 336-343 (1996). 14. Hinz, P.M, Angel, J.R.P., Homann, W.F., McCarthy, D.W., McGuire, P.C., Cheselka, M., Hora, J.L., Woolf, N.J. First results of nulling interferometry with the Multiple Mirror Telescope Proc. SPIE 3350 in press. (1998). 15. Noll, R.J. Zernike polynomial and atmospheric turbulence J. Opt. Soc. Am. 66, 207-211 (1976). 16. Hill, J.M. & Salinari, P. Large Binocular Telescope project Proc. SPIE 3352 in press. (1998). 17. Angel, J.R.P. & Woolf, N.J. An imaging nulling interferometer to study extra-solar planets Astrophys.J. 475, 373-379 (1996). This preprint was prepared with the AAS LATEX macros v4.0.

{5{

1. Acknowledgements This work is supported by AFOSR, NASA, and NSF. We would like to thank Craig Foltz and the MMT sta for their essential support of the project.

Please address correspondence and requests for materials to Philip Hinz, 933 N. Cherry Ave., Tucson, AZ, 85721 tel:520-621-7866 email:[email protected].

{6{

5.0 m Telescope Axis

M1

1.83 m A

M2 Z2

Z1 B M3

Infrared Detector

Fig. 1.| Schematic of our nulling interferometer at the MMT.The telescopes are co-mounted on a rigid frame pointed at the star, with no need for the variable delay lines for path equalization found in interferometers using separately mounted telescopes. The small number of re ections results in low thermal emission. Unmatched re ections are made at nearly normal incidence to minimize polarization dierences. The f/31 folded Cassegrain beams are combined at the zinc selenide beamsplitter Z1, coated for equal transmission and re ection in the 8-13 m band. Beam A, after crossing the axis of the telescope, is folded down at M1, and is transmitted by the beamsplitter before coming to a focus. Beam B is folded downward at M2 before reaching the axis of the telescope, through zinc selenide plate Z2 and back up at M3 to equalize the path-lengths before being re ected from the underside of the beamsplitter. Achromatic 180 degree phase dierence is realized by balancing a slight dierence in air path with a 42 m path dierence between the two zinc selenide elements, obtained by slight rotation of Z2. The residual chromatic error results in transmission ideally 0.008% over the band 8-12 m. In practice, the achromatic null is found initially by translating the beamsplitter vertically, with occasional subsequent small adjustments to correct for exure. Image detection was made with the Rockwell (now Boeing) 128  128 arsenic-doped silicon BIB array of the Mid-infrared Array Camera, MIRAC28. An internal cold stop limited the eective apertures of the telescopes to 1.6 m, for a diraction limited image width of 1.5 arcsec.

{7{

α Tau

a

1 arcsec

FWHM 1.5"

constructive

destructive

α Ori

b

1 arcsec

FWHM 1.6"

FWHM 2.4"

Fig. 2.| Single short-exposure frames at 10.3 m showing constructive and destructive interference. The second image of  Ori with the start nulled shows directly its surrounding dust nebula. The one-dimensional plots under the images are smoothed cuts through the central row of pixels These images were selected for maximal and minimal ux from a series of 1000 50 msec frames of  Tauri and 500 100 msec exposures of  Orionis, taken at 10% bandwidth by MIRAC2 operating in a continuous recording mode with no \dead time".  Tauri which is unresolved was observed on February 17, 1998 under seeing conditions characterized by Fried's length r0 (10 m) = 3.7 m.  Orionis was observed on January 17-18, 1998, under better seeing conditions (r0 = 4.4 m, slower motion).

{8{

R Leonis

N

E

5 arcsec

α Orionis

N

E

Fig. 3.| Averages of nulled images for  Orionis and R Leonis (also observed on January 17-18, 1998) showing the 10 m dust nebulae with improved signal-to-noise ratio. The contours are at the level 1%, 10%, and 20% of the non-interfered stellar peak intensity. The small + marks the centroid of the stellar emission. For  Orionis this is clearly oset from the nebula.

{9{

R Leonis

a

5 arcsec

α Orionis

b

Fig. 4.| Stellar diraction patterns obtained by subtracting the nulled images of Figure 3 from the corresponding constructive images. For  Orionis a misalignment of the centroids by 0.3 arcsec was applied to obtain the symmetric diraction pattern shown. The non-interfered images of unresolved stars, when similarly aligned and stacked, are not so perfect because they include the halo components from nebular emission or scattering due to higher order wavefront aberrations. These subtract out in the dierence image.