What can we do to enter this competition? We need a large telescope (now that Keck has entered the field), and a stable spectrograph with both high resolution and a wide spectral range to allow many lines for cross-correlation. How can we do this at the CFHT?
The CFHT is now equipped with an adaptive optics focus that supplies a 0"15 image of a bright star in the red. In our case, the star being recorded is also the guide star for the Adaptive Optics Bonnette. At the output of the f/20 focus of the AOB, we shall have a small focal reducer that changes the beam to f/8 at a scale of 140 mu/arcsec. We inject this beam into a 50mu core diameter optical fibre, which subtends 0"35 on the sky, and is quite sufficient to collect all the light from the star. At the fibre output, we collect the beam at f/5, to compensate for fibre Focal Ratio Degradation and thereby recover over 90% of the total flux. In the red, the fibre transmission over a 40m length is also more than 90%. We now have our light coming out from a 50mu fibre at f/5. We change this to f/20 giving a 200mu image on the Gecko slit. In this case, the slit projected onto the detector is 34mu wide, covering 2 or 2.5 pixels depending on the CCD used, and providing a spectral resolution of 100,000. It is obvious that there will be no further need for an image slicer.
Now, what about Gecko? I know that it works well with a 4096 pixel long CCD in the dispersion direction because I used the UBC CCD with it. The question is what is the image quality in the perpendicular direction? For this investigation, we have access to 200 pixels image height with ORBIT 1 or UBC CCD or Loral 5. We have to check the image quality inside the known window and verify the chromatism generated by various optical components.
I have a Gecko Th-Ar spectrum centered at H alpha, 656.2nm, taken with a long slit covering 200 pixels. The slit was said to be 200mu but it looks closer to 264mu. (Perhaps the focus was not perfect). The line width varies between 2.6 and 2.8 15mu pixels along the 200 pixel window. The net result is that we could have R=100,000 over the whole 4096 x 200 pixel window in monochromatic light with a filter rather than a grism.
Second step:
I have a Gecko Th-Ar spectrum taken in standard mode, with an image slicer
and centered in the 9th order at 620nm. This image shows three orders:
| Order | centered at | FWHM of line |
| 10 | 558nm | 4 pixels |
| 9 | 620nm | 3 pixels |
| 8 | 697.5nm | 4 pixels |
The net result is that we have a resolution ranging from 75,000 to 100,000 inside a 4096 x 200 pixel window with the present optical layout. This is a conservative figure, and by using a narrower slit and a better focus, we could improve these values. We shall have to await optical ray tracing to see how we can upgrade the design, at low cost, by minor changes to some optical components. The wavelength coverage being explored is 140nm, from one side of order 10 to the opposite side of order 8. This is larger than the 500 to 600nm wavelength interval needed for an iodine cell. To exploit this window, we shall have to change the grating mosaic to a monolithic échelle grating with low line density. It could be supplied, at a reasonable price, by the Vavilov Institut in Russia on a 300 x 500 mm blank, which gives an acceptable level of vignetting. The cross dispersing grisms are small optical components of low cost.
Harvey Richardson has agreed to cast his expert eye over this proposal and to see whether it is possible to increase the usable window to a 4k x 2k EEV CCD with 13.5mu pixels.
What is the present situation at Keck? The cross-dispersed spectrograph is HIRES, with a maximum resolution of 60,000 and a slit width-resolution product of 39,000 arcsec. To achieve this maximum resolution, one needs a slit width of 0"65 which is about the average seeing at Mauna Kea. Therefore only half of the light in the image enters the slit. So Keck is equivalent to about a 7m telescope, gathering four times as much light as full use of the CFHT. We recover partially from this handicap because our spectral resolution is double that of Keck.
Going back to the extrasolar planets search, we need high stability and very good wavelength calibration. The stability will be increased by AOB guiding and optical fibres, with double image scrambling. At Meudon, we have been in this field for a long time and one of our PhD students has developed a solid double fibre scrambler that could be used in this study. Wavelength calibration and spectrograph stability would be monitored with an iodine cell in the classical way. If we want to increase the spectral range, we may have to go to a Fabry-Perot channel for calibration.
