Subtleties-this bit is worth reading!

There are some important alternative ways of specifying the wavelength of the line.

1. If you know the wavelength exactly, but it is not in the line list, you can specify the type as `exact'. So if your response to the wavelength prompt is

      3456.789 e
   

   `e' being short for `exact'-`arc' will use that value as the wavelength. If you have specified 'NONE' as the arc type, `arc' will assume that all wavelengths given are exact values.

2. Once two lines have been identified, `arc' is able to estimate the wavelength of a new line by linear interpolation. It will then tell you this interpolated wavelength as well as the channel number. If more than two lines have been identified, the interpolation is based on the two nearest to the new line. You can now use `i' in your response instead of specifying a wavelength-the result is the same as if you had typed in the interpolated wavelength. `arc' will then look for the listed wavelength nearest to the interpolated value.

3. Similarly, once more than two lines have been identified, `arc' is able to start performing a running fit to the identified lines. The fit is recalculated each time a new line is identified and the RMS of the fit is displayed. So a bad identification will usually show up immediately as a large rise in the RMS figure. `arc' can now display the fitted wavelength as well as the interpolated wavelength when a new line is selected, and you can use `f' for the fitted wavelength in the same way as you can use `i'.

What this means in practice, is that with a good line list, you can identify enough lines to tie down the fit fairly well (it helps to do some lines at one end, then at the other, then work in), and then just respond `f' to each wavelength prompt. `arc' will then use the line in the line list closest to the fitted wavelength. This simplifies the process considerably, but it is still under your control, and you can intervene if a new fit shows a considerable increase in RMS.