Paper detail

Quantifying resolving power in astronomical spectra

The spectral resolving power R = lambda / delta lambda is a key property of any spectrograph, but its definition is vague because the `smallest resolvable wavelength difference' delta lambda does not have a consistent definition. Often the FWHM is used, but this is not consistent when comparing the resolution of instruments with different forms of spectral line spread function. Here two methods for calculating resolving power on a consistent scale are given. The first is based on the principle that two spectral lines are just resolved when the mutual disturbance in fitting the fluxes of the lines reaches a threshold (here equal to that of sinc^2 profiles at the Rayleigh criterion). The second criterion assumes that two spectrographs have equal resolving powers if the wavelength error in fitting a narrow spectral line is the same in each case (given equal signal flux and noise power). The two criteria give similar results, and give rise to scaling factors which can be applied to bring resolving power calculated using the FWHM on to a consistent scale. The differences among commonly encountered Line Spread Functions are substantial, with a Lorentzian profile (as produced by an imaging Fabry-Perot interferometer) being a factor of two worse than the boxy profile from a projected circle (as produced by integration across the spatial dimension of a multi-mode fibre) when both have the same FWHM. The projected circle has a larger FWHM in comparison with its true resolution, so using FWHM to characterise the resolution of a spectrograph which is fed by multi-mode fibres significantly underestimates its true resolving power if it has small aberrations and a well-sampled profile.