next previous
Up: Present performance of the coronagraph


Subsections

2 Instrument layout

 

2.1 General view

The dark-speckle coronagraph (DSC) combines features of adaptive optics, Lyot coronagraphy and speckle interferometry. Figure 1 shows the arrangement utilized at the Coudé focus of the 152 cm telescope at Observatoire de Haute-Provence. The stellar beam from the telescope enters the adaptive optics bench BOA, of the ONERA (Office National d'Études et de Recherches Aérospatiales). It has 88 actuators and responds fast enough for seeing correction in the visible range (Conan et al. 1998). Downstream from BOA, the afocal beam reaches a dichroic beamsplitter which transmits the IR (K band) light towards another instrument used simultaneously, the achromatic interfero-coronagraph (Gay & Rabbia 1996), while visible light ($650\,{\rm nm}<\lambda<850\,{\rm nm}$) is reflected to the DSC.
  
Figure 1: Dark speckle camera used at the Haute-Provence 1.52 m telescope (not a scale drawing). The dichroic beam-splitter selects the wavelength range $650-850\,{\rm nm}$ in reflection. The collimated beam received from the telescope with BOA adaptive optics crosses a pair of Risley prisms for atmospheric dispersion compensation and is focussed by concave mirror CM onto the Lyot occulting mask OM. The masked field is re-imaged onto photon-counting camera CP through the pair of Wynne corrector triplets WT which compensate the first-order lateral chromatism of diffraction patterns. A Lyot stop LS removes stellar diffracted light at the edge of the relayed pupil, according to the classical Lyot coronagraph principle

The Strehl ratio reached 10% to 30% depending on seeing conditions. Although the adaptive optics gain is higher in the IR range, the advantage of working with visible light results from the use of a photon-counting camera. Its low dark count allows short exposures with negligible added noise, as required for speckle observations.

The centering of the star's image on the Lyot occulting mask is critical, and should be part of the adaptive optics loop. This is achievable by collecting light reflected from the mask, but the drift had to be corrected manually at this stage.

2.2 Optical bench

2.2.1 Coronagraph

The diffraction rings are removed by a Lyot coronagraph (Lyot 1930, Fig. 1). The occulting masks in the f/122 beam are inter-changeable, with sizes correponding to $0.22\hbox{$^{\prime\prime}$}$, $0.45\hbox{$^{\prime\prime}$}$ and $0.67\hbox{$^{\prime\prime}$}$. They have been vaccum evaporated through pinholes onto a 1.2 mm thick plate of anti-reflection coated glass. The apodizing masks located at the relayed pupil are pinholes carried by a x-y translation stage. No attempt was made to mask the aperture spider and central obscuration although this should decrease the scattered light significantly.

2.2.2 Atmospheric dispersion

The atmospheric dispersion corrector, located near the first pupil image, uses a pair of normal-field Risley prisms, which are rotated. The glasses, Schott PK51A and Corning B29-52, correct a wide spectral bandpass, from $380 \,{\rm nm}$ up to $950 \,{\rm nm}$. The residual chromatism in the $600\,{-}\,850\,{\rm nm}$ band is on average 1/4 of the Airy peak diameter, corresponding to $0.03\hbox{$^{\prime\prime}$}$ at $750\,{\rm nm}$. With the 1.5 m telescope and the 6mm pupil, the prisms correct up to $60\hbox{$^\circ$}$ from zenith. We adjusted visually the prism settings although pre-calculated settings would be preferable, especially for faint stars.

2.2.3 Wynne correctors

For conventional speckle interferometry the detected bandwidth can be increased by transforming the partially proportional wavelength dependence of the speckle scale. Here, the proportionality is more accurate since the residual phase on the wave is made small by the adaptive optics. The photon rate being critical in ground-based dark-speckle imaging, there is much to gain in increasing the usable spectral bandwidth.

   
Figure 2: Wynne corrector for making the speckle pattern wavelength-invariant. A pair of triplets, afocal in yellow light, shrinks the blue pupil while enlarging the red pupil. The exit angle, u', becomes approximately proportional to wavelength, thus ensuring a nearly wavelength-invariant diffraction pattern
One of us (DK) has re-calculated the corrector solution obtained by Wynne (1979) (Fig. 2). The pair of null triplets shrinks the blue pupil, while enlarging the red pupil, so that both Airy patterns, or speckle patterns, are of nearly identical size. With respect to the original Wynne design, we have suppressed the power in the second triplet, in order to keep more flexibility in the final magnification and we have replaced the SF8 glass by SFL5 for optimal correction in the red range ($600-850\,{\rm nm}$). The analytical derivation is given in Appendix.

2.2.4 The photon-counting camera

  Finally L2 focuses the beam onto the camera with an $8\times$ magnification, bringing the aperture to f/976.

The detector is a cooled CP20 photon-counting camera (Vakili 1990) having a first-generation electrostatic intensifier coupled to a second-generation microchannel intensifier and a fiber optics taper, feeding a $384\times 288$ CCD. The resulting amplification is about 105. Such a device allows single photon detection and a very low dark count, less than 10 photons per $20\,{\rm ms}$ exposure for $50\,\mu {\rm m}$ pixels at $-20\,\hbox{$^\circ$}$C (about 0.0045 photons/s/pixel). A low dark noise is necessary to detect accurately any "filling'' of the dark speckles, which could reveal the presence of a faint companion. The drawbacks of such photon-counting cameras are their low quantum efficiency ($<10\%$ at $700\,{\rm nm}$) and their low saturation level of $50000\,{\rm ph/s}$, limited by the acquisition system.

As mentioned above, the coronagraph forms a f/976 focus on the detector in order to achieve the dense sampling required by the dark-speckle technique. For a central wavelength of $635\,{\rm nm}$, it represents 150 pixels/speckle area or 144 pixels/arcsecond. The field of view is limited to a diameter of 250 pixels (about $1.8\hbox{$^{\prime\prime}$}$) by the Wynne corrector. In addition to the Wynne corrector, spectral filters can be inserted in front of the photocathode to select different narrow bandwidths.

2.3 Efficiency of the Lyot stop

 
  
\begin{figure}
\epsfxsize=7cm
\epsfbox {d7614_3.ps}\end{figure} Figure 3: a) Reference image obtained with a white light source inside the BOA adaptive optics system, using the $0.45\hbox{$^{\prime\prime}$}$ mask and without the Lyot stop. The Airy rings remain visible. Diffraction spikes from the spider arms can be seen as dotted lines, influenced by the ring structure. In such conditions a planetary companion would be invisible. In spite of the broad spectral bandwidth ($\Delta\lambda=650-850\,{\rm nm}$) at $\lambda_0 = 635 \,{\rm nm}$, the outer rings retain good contrast, owing to the Wynne corrector
 
\begin{figure}
\epsfxsize=7cm
\epsfbox {d7614_4.ps}

\setcounter{figure}{2}\end{figure} Figure 3: b) The same reference frame obtained with the mask and the $400\,\mu {\rm m}$ Lyot stop. Here, the Airy pattern is markedly attenuated and the gain in sensitivity is about 1.7 magnitudes. Nevertheless, due to the inadapted Lyot stop, some bright features remain, like the rings around the mask and the four symmetrical speckles which could bring out wrong detections. These static defects can be partially removed with frame subtraction
To calibrate the efficiency of the Lyot stop, we have acquired an internally-generated reference image using a single-mode fiber included in BOA. The Strehl ratio (SR) of this reference source is about $80\%$ and does not take into account the atmospheric turbulence or the static aberrations of the adaptive mirror. When the core of the Airy pattern is occulted by the mask, the edges of the pupil image become decorated with two bright fringes (Fig. 1). The complementary spatial filter in the pupil plane should suppress much of the diffracted light, except that caused by the wave bumpiness. The Lyot stop being a simple pinhole of $400\,\mu {\rm m}$ instead of a telescope pupil image, some Airy-like rings remain visible in the final image (Fig. 3b). Moreover, the spider spikes combined with the residual rings produce symmetrical side-lobes, especially bright within the 2 first rings. These artifacts remain on the compensated images despite the smoothing introduced by the atmospheric turbulence.

One can compute the rejection rate of the coronagraph as defined in Malbet 1996:
\begin{displaymath}
R={I_{\rm w/o}\over I_{\rm w}}\end{displaymath} (1)
where $I_{\rm w/o}$ is the total intensity of the reference beam without coronagraph and $I_{\rm w}$ is its intensity with the coronagraph. To characterize the efficiency of the Lyot stop, one defines $R_{\rm w}$ the rejection rate with the Lyot stop and $R_{\rm w/o}$ without it. We can then estimate the gain in magnitude introduced by the Lyot stop with the following relation (Malbet 1996):
\begin{displaymath}
\Delta m=2.52 \,{\rm Log}(2R_{\rm w}/R_{\rm w/o}).\end{displaymath} (2)
A gain of 1.7 magnitude has been measured for the $0.45\hbox{$^{\prime\prime}$}$ mask and the $400\,\mu {\rm m}$ Lyot stop. As this value is averaged over the entire field, it is therefore underestimated far from the axis and overestimated near the mask. An optimized Lyot stop, including secondary mirror and spider arms, should improve the gain by another 1.3 magnitude (Malbet 1996).


next previous
Up: Present performance of the coronagraph

Copyright The European Southern Observatory (ESO)