To check that we indeed reach the 1m/s RV accuracy, monitor the short- and long-term RV stability and potentially compensate for small RV drifts of the recorded spectrum, we need to have a specific calibration module providing a very stable RV reference. While evacuated instruments with temperature/pressure control are intrinsically very stable (eg HARPS at a 1 m/s level), having such a wavelength reference module is nevertheless necessary, for regular stability checks at the very least.
Experiments carried out with the PathFinder spectrograph @ Penn State University (Ramsey et al 2008) suggest that a standard ThAr hollow cathode lamp is potentially usable in this purpose, and that a UAr hollow cathode lamp would provide a significantly higher line density in the nIR. However, prior experiments with such lamps (eg Sophie spectrograph) demonstrates that stability problems may arise when lamps are ageing. We therefore aim at another more accurate option (keeping this one only as a backup).
Another possibility is to use an athermal Fabry-Perot etalon (with both pressure and temperature control) in conjunction with a halogen lamp. The advantage of this option would of course be to provide a very high and regular density of lines throughout the whole spectrum (>50 of lines per order). The Geneva group is also exploring this option in the framework of the CODEX/EXPRESSO ESO project at getting un ultra-stable calibration cell (1 cm/s), has started designing and building a stable Fabry-Perot unit and should start testing it on HARPS by early 2009. We are in close interaction with F Pepe - in charge of this experiment and our contact in Geneva - on this issue. Provided that their unit fullfills our less stringent specification of 1 m/s (highly probable), we will simply adapt their design to nIR wavelengths (in collaboration with them) and build a specific Fabry-Perot unit for SPIRou.
A third solution is to implement a laser frequency comb with a Fabry-Perot filtering cavity; this promising solution is also very accurate and expected to provide accuracies of 1cm/s. However, it is not completely clear how adaptable it is to our case (broad spectral domain, nIR) in a reasonable time-scale and budget. We will keep contact with the developers to ensure that we can switch to this third solution if needed.
As the light from the wavelength reference module is brought to the instrument through an optical fiber, we can easily make provisions for future developments (such as the laser comp solution) if it turns out necessary for matching the specifications, or to improve further the accuracy of the reference cell.