A Novel Simultaneous Differential Imager for the direct imaging of giant extra-solar planets

Young (100 Myr old) extra-solar planets are 100000 times more self-luminous than old (5 Gyr) extra-solar planets, whereas their primary stars are only slightly (2-5 times) brighter when this young. Currently the majority of such young stars that are nearby (<50pc) are located in the southern star forming regions and associations (d < -20 deg.). To detect a faint point source (a planet) near a bright source (its star) requires high spatial resolution, that is high Strehl ratios at large telescopes. Therefore, NACO is the instrument of choice to detect extra-solar planets directly by imaging. However, NACO (like all AO systems) suffers from a limiting “Speckle-noise” floor which prevents the detection of planets within 1“ of the primary star. Hence NACO requires some method to suppress this limiting “Speckle-noise” floor.

Together with the
Steward Observatory we have implemented and commissioned in NACO an observation mode to calibrate and remove the “Speckle noise” in the AO-images: Exploiting the fact that all extra-solar giant planets have strong CH4 (methane) absorption beyond 1.62 µm in the H band NIR atmospheric window, a novel optical device has been implemented that allows one to obtain simultaneously 4 images of a star through 3 slightly different filters, sampling both inside and outside of the CH4 feature (see Fig. 1 and 4).

Fig. 1:
Schematic view of the new NACO-SDI-mode


Technical realization:

1. A new Camera system has been implemented into CONICA consisting of a f/40 camera and a special quadrant filter just in front of the focal plane (see Fig. 1). To make it fit into the given space, a three lens solution has been chosen with highly effective AR-coatings below 0.1% reflection losses per lens. Special care has been taken to make sure that imaging errors are below the diffraction limit, to minimize the differential static aberrations between the four PSFs.
The complete module is shown in Fig. 2.

Fig. 2: The SDI-Camera modul

To avoid residual speckle pattern produced by variation of the PSF over the detector array, high imaging quality all over the chip is required. The theoretical Strehl ratio is better than 95% all over the FOV. Lens radii have been tested by interferometry.

2. To split the light into four imaging beams of full aperture, a double Calcite Wollaston has been inserted into CONICA“s Grism/Polarizer wheel. The second Wollaston is rotated by 45 deg relative to the first one resulting in a romboid distribution of the four sub-images.

3. To avoid overlapping FOVs, a small 6x6 arcsec mask has been introduced into the focal plane wheel.

In summary, this combination provides four simultaneous narrow band images of the same object with slightly different central wavelengths distributed around the CH4 absorption at 1.62µm. PSFs and residual speckle noise distributions are very close to be identical, thus, the speckle noise is expected to be drastically reduced by this simultaneous differential imaging method (see Marois et al, 2000).

CONICA-SDI has been successfully commissioned during two observational periodes in August 2003 and Fevruary 2004:
The image scale has been confirmed within better than 1%. The measured pixel scale is 17.25 ±0.06 mas/pixel (nominal 17.4mas/pixel).
Distortion is shown to be less than 0.5% between and inside the given FOVs.
Focus position and higher order static aberrations have been measured by the method of phase diversity. The measured differential wave front deviations are less than 10 nm RMS error across all orders measured.
For a number of near-by young stars deep 40 min SDI-images have been taken, to compare the gain in speckle noise reduction to the theoretically expected one. Fig. 5 and 6 shows the result for one star:
The differential SDI-method gains particularly inside the residual seeing halo compared to non-differential methods.

In February 2007, this SDI-mode has been improved significantly by inserting a special YVO4-double prism: The optical axis of the second Wollaston prism is rotated by 45 deg relative to the wedge edge, thus, the four resulting sub-images are using now the full detector quadrants instead of the rhomboid pattern of the first splitting device. The resulting advantages are: Significantly increased FOV (now 8”x8”) and smaller chromatic aberration, which improves the differential accuracy.  

Fig. 4:
IR-Spectrum of Gl229B, a T6 brown dwarf, around the 1.6µm CH4 –Absorption feature. The pass-bands of the three different narrow band filters are indicated.


Fig. 3:
The double Wollaston component

Fig. 4:
The new YVO4-double Wollaston  (SDI+)


Fig. 4: The
Data reduction:

A rather complex FDL pipeline has been developed to get well reduced data just after finishing the exposure in order to be able to identify potential candidates for extrasolar planets: After flat-fielding and sky-subtraction the individual images are extracted and centered relative to each other within 0.01 pixels. Even though the narrow band filters are very near to each other (1.575, 1.600, 1.625 µm), we have corrected for deviation in wavelength by re-scaling the diffraction pattern before subtracting them. The same procedure has been done for a data set obtained after having rotated by 33 deg. The two data sets have been subtracted again, thus, a potential planet should appear as a pair of negative and positive spots, separated by a rotation angle of 33 deg. relative to the star.

Fig. 5: To demonstrate the similarity of speckle pattern, to the right the four individual images are shown as close-ups. To the right we show schematically the way from raw data to fully SDI reduced data: The raw data reflect the romboidal splitting by the double Wollaston device. A data reduction pipeline has been used to flat-field, extract and subtract the individual images. The comparison of standard AO procedure (including unsharp masking) to the SDI method clearly demonstrate the advantage of this differential simultaneous method. It should be noted that the seeing was a rather poor 1.1” for these observations.

Fig. 6: Comparison of AO-Point spread function to residual un-sharp masked and SDI-imaging sensitivities at the 1 sigma level: To the left the image contrast is shown in H-magnitudes as a function of radial distance from the star. While the „optimized conventional AO method“ remains speckle noise limited within the seeing halo regime and farer away, the sensitivity of the SDI method is by about 2.5 to 3 mag deeper inside 0.5arcsec radius and reaches the photon noise limit outside this radius.


  • Lagrange, A.-M., Chauvin, G., Fusco, Th., Gendron, E., Roman, D., Hartung, M., Lacombe, F., Mouillet, D., Rousset, G.,     Drossart, P., Lenzen, R., Moutou, C., Brandner, W., Hubin, N., Clenet,Y., Stolte, A., Schoedel, R., Zins, G., Spyromilio, J.,     2003, First diffraction limited images at VLT with NAOS and CONICA, Proc. SPIE 4841, page 860-868
  • Lenzen, R., Hartung, M., Brandner, W., Finger, G., Hubin, N., Lacombe, F., Lagrange, A.-M., Lehnert, M., Moorwood, A.,     Mouillet, D., 2003, NAOS-CONICA first on sky results in a variety of observing modes, Proc.
SPIE Vol. 4841, page 944-
  • Marois, Ch., Doyon, R., Racine, R., Nadeau, D., 2000, PASP Vol. 112, page 91-96

Responsible: Rainer Lenzen

Last updated: July 11, 2007