ESPaDOnS
Thermal response and spectral stability


General concept

To ensure that ESPaDOnS would be as stable as possible, we decided to follow the advice of the Geneva experts for improving the stability of echelle spectrographs. Without going to the extremes of enclosing the entire spectrograph within a depressurised and thermally regulated container (eg as was done for harps, the eso spectrograph dedicated to ultra high precision measurements of stellar radial velocities and mounted on the la silla 3.6m telescope), we converged towards an intermediate solution involving a double-layer thermal insulation. The concept (recommended by the Geneva experts) features: This ensures in particular that the temperature within the spectrograph is stable to a rms level of a few 0.01deg, provided that operations within the inner enclosure are kept to an absolute minimum. For both scheduling and practical reasons, it was decided that only the inner enclosure is built while ESPaDOnS is at OMP, while the outer enclosure and thermal regulation is implemented in a second step, once ESPaDOnS is installed in the coude room of cfht.

Accurate temperature sensors (with a precision of 0.01deg) are implemented at different points within the inner enclosure to check the stability and estimate potential temperature drifts and gradients. A digital barometer is also implemented within the inner enclosure to monitor pressure fluctuations at the 0.01mbar level. To minimise operations within the inner enclosure, the ccd filling and exhaust pipes are installed permanently and are thermally insulated (within an evacuated tube) from the inner spectrograph environment.

Performance of inner enclosure

The graph on the right shows the room temperature (full line) along with the temperature within ESPaDOnS inner enclosure (dashed line), as recorded in a long test run of several weeks during which the enclosure was kept closed as much as possible. This graph shows the temperature variations during about four consecutive days, where daily fluctuations in outside temperature (with a peak-to-peak amplitude of about 1deg) are clearly visible.

This demonstrates that the inner enclosure smoothes out all short term temperature variations by at least an order of magnitude. In particular, daily changes are no longer detectable within the spectrograph. However, longer term variations (on a timescale of several days) are still present and essentially mimic (as expected) the long term fluctuations in outside temperature.

Spectrograph temperature variations of as much as 0.7deg/d are observed in the present context; they should be reduced by at least a factor of 2 once ESPaDOnS is installed at CFHT, where temperature drifts in the coude room are typically of order 0.1degr/d and rarely exceed 0.3deg/d. Once the outside enclosure and thermal regulation is setup, such drifts should be further reduced by typically an order of magnitude on timescales of days.

Spectral stability

By taking calibration frames during a complete night (at a rate of one every 10min) and by correlating all images with respect to the first one in the series, it is possible to see how the position of the spectrum with respect to the ccd varies with time; this experiment is very useful to estimate how much spurious spectral radial velocity changes are induced by thermal and mechanical relaxation within the spectrograph. The graph on the right show the changes in the radial velocity of the thorium spectrum (in km/s, full line) with respect to the first spectrum of the series, while the 2 other curves depict the corresponding temperature and pressure changes (in deg and mbar, dashed and dash-dot line respectively) throughout one night.

We find that the position of the thorium spectrum with respect to the ccd varies by typically: Once the temperature and pressure effects are subtracted off, the residual changes in radial velocities, equal to about 20m/s rms, indicate what the true absolute stability of the spectrograph is. Note that this experiment demonstrates clearly the need for an outer enclosure with thermal regulation to reduce the shifts with temperature as much as possible and make them depend mostly on pressure.

From such a series of thorium frames, we can also estimate the relative stability of the instrument (with respect to a given spectral reference). Using the even thorium frames as the reference and the odd thorium frames as the test spectrum whose stability is to be checked, we obtain that the relative stability is better than 10m/s rms, for a time lag of less than 10min between the object and reference measurements.



© Jean-François Donati, last update May 10 2004