Spectral domain and resolution

Echelle orders

The image on the right represents an example flat field frame taken with ESPaDOnS in polarimetric mode (using light from a combination of tungsten lamps and filters so that all orders get a reasonable illumination level). Orders are clearly visible on this image, where they show up as bright slightly curved strips running vertically, successive orders being stacked next to each other from the left to the right of the ccd. As obvious from this image, the order separation varies with wavelength, being largest in the blue (right side of image) and smallest in the red (left side of image) as expected from a prism crossdisperser. A close up view of the small scale structure of the orders is displayed in the insert (bottom right of image), where the two spectra associated to each order in polarimetric mode (one spectrum per orthogonal state of the selected polarisation to be measured) are clearly visible.

Up to 40 orders are visible on the image, the first one being order #22 (centred at 1029nm) on the left side of the chip and the last one being order #61 (centred at 372nm) on the right side of the chip. Apart from very small gaps on the edges of the 3 reddest orders (between 922.4 and 923.4, 960.8 and 963.6nm, 1002.6 and 1007.4nm), the wavelength coverage is complete from 369 to 1048nm and can be obtained in a single exposure.

When reducing the data, the first operation consists at tracking the location and shape of all orders across the whole chip to a rms accuracy of better than 0.1pxl.

Wavelength calibration

The image on the right represents an example calibration frame taken with ESPaDOnS in polarimetric mode (using light from a combination of a thorium/argon and a thorium/neon lamp with filters to minimise the amount of strong red lines blooming the chip). As obvious from this image, a very large number of lines are present in each order, from which the accurate relation between pixel number along and across each order can be derived. The spectral resolution achieved is derived from the width of these lines. A close up view of the individual thorium lines is shown in the insert (bottom right of image) where one can see again the dual structure of each order (the gap between the two spectra as well as the instrumental width of the lines (slightly lower than 2pxl).

An average number of about 50 lines per order (about 2,000 lines in total) are automatically searched for by the reduction routine and identified using reference lists of thorium line wavelengths; from these, wavelength calibration polynomials are produced over the full spectral range. The typical accuracy of this calibration is found to be of order of 0.06pxl or 150m/s at each given wavelength.

The few remaining neon lines blooming the ccd in the red part of the domain do not really affect the precision of the calibration procedure.

Spectral resolution

By measuring the full width at half maximum of the individual thorium lines (reflecting mostly the instrumental broadening), one can determine the spectral resolution of ESPaDOnS in the selected instrument configuration (the reduction code does it automatically).

The graph on the right shows an example of such thorium lines (the strongest of the 3 being the ThI line at 550.75385nm). The full line indicates the wavelength calibrated spectrum around this line derived with ESPaDOnS being set in polarimetric mode, while the dashed line depicts that obtained in the 'object only' spectroscopic mode. The respective line widths (at half maximum) are respectively equal to 8.3 and 6.9pm, in agreement with the spectral resolutions of 68,000 and 81,000 associated to these modes.

These resolutions correpond to velocity elements of 4.4 and 3.7km/s respectively, to be compared to the 2.6km/s ccd pixel size and the 1.8km/s bin size on which the spectra are recovered.

© Jean-François Donati, last update May 5 2004