Detection of faint stellar companions with Hokupa'a on the Gemini North telescope
Figure 1 illustrates early results obtained with the Hokupa'a adaptive optics system on the Gemini North telescope, and presented at the
telescope dedication. These are images of a young massive star called V1318 Cyg taken at Brackett Gamma. The uncompensated image on the
left was taken with Hokupa'a turned off. The compensated image in the center was taken with Hokupa'a turned on. The compensated and
post-processed image on the right was obtained by processing a set of 13 compensated images as described below. Field rotation was used to
further improve image quality, allowing us to detect the faint companion seen below and slightly left of the main star. This previously
unknown companion is 0.68 arcsec away from the main star and is 5.8 magnitude fainter (the scale is given by the one arcsec bar in the
upper right corner of the right image).
The first two images in Figure 2 are from left to right the first and last images of our sequence of 13 compensated images, the total
observation time being about 10 minutes. A number of speckles (due to defects in the telescope secondary mirror) have been encircled. Note
that they are exactly at the same location on both images except one, in the lower part of the image, which has moved due to field
rotation. This is the companion we have discovered. The last image on the right side of Fig. 2 is the median of the 13 images. Because of
its motion, the companion disappears in the median image, whereas all other speckles remain practically the same. Therefore, the median
image provides an excellent estimation of the telescope point spread function (PSF) from the data themselves. No other reference star was
observed.
The first two images on the left of Fig. 3 are the same as in Fig. 2, but for a deconvolution with the Lucy-Richardson algorithm, using the
median image as a PSF. In addition, the deconvolved images have been rotated to compensate for field rotation, so that all 13 images now
have the same orientation, which is that of the image in the middle of the sequence. The last image on the right side of Fig. 3 is the
median of the 13 deconvolved and rotated images. The companion is clearly seen 0.68 arcsec away of the central star. For accurate
photometry, data were reprocessed using the MCS deconvolution algorithm. The magnitude difference was estimated to be delta-m = 5.8 with a
dispersion of 0.1.
To investigate the detectivity limit of this method, simulated companions were added to the real images with appropriate photon noise. The
location of these artificial companions are indicated by the gaps in the circles drawn in Fig. 4. The radius of the circles are 0.6 arcsec
for the left image and 0.25 arcsec for the right image. For the left image, the same field rotation was simulated as for the real
companion. The magnitude differences are alternatively 6.5 and 7 starting at 6.5 for the artificial companion closest to the real one and
on the left of it. For the right image, the amount of field rotation was increased by a factor 3 to suppress any overlap of the companion
images in the first and last exposure. In this case, the magnitude differences are respectively 4, 4.5, 5, 5.5, 6, and 6 starting clockwise
from the most right one. Figure 4 shows the final processed images. The positions are well recovered with the Lucy algorithm. When delta-m
is larger than 5, the magnitudes recovered with the MCS algorithm. are about 0.5 mag higher than the real ones. This gives us a detection
limit of delta-m =6 +/- 0.5 at a distance of 0.25 arcsec.