Electromagnetic radiation's spectral energy distribution (SED) is a function, i.e., plot of brightness or flux density versus the frequency or wavelength (the choice of which produce different results: see below). When the SED of an astronomical object is measured from Earth, the effects of Earth atmosphere are relevant. Capturing a wide range of wavelengths requires multiple instruments, which causes data-calibration challenges. But some limited bandwidths can provide signatures for specific types sources.
The alternate term, spectral power distribution (SPD) is used in some other (non-astronomy) fields.
Plotting by wavelength versus plotting by frequency has consequences: the peak wavelength differs in the two cases if each is plotted linearly, a consequence of plotting density functions of values that have a reciprocal relationship. Taking black-body spectrum as an example, plotting wavelengths by millimeter (0mm, 1mm, 2mm, 3mm, ...) crowds all the infrared, visible light, ultraviolet, etc., between 0 and 1 and spreads radio over a wide range. Plotting frequency by 100GHz (0GHz, 100GHz, 200GHz, 300GHz, ..) crowds the radio into 0-300GHz and spreads the others over the rest. The CMB SED's peak is cited at various wavelengths and frequencies, but all are valid for the spectrum representing a temperature of 2.72548 K, depending upon how it is plotted:
|plotted by||peak frequency||peak wavelength||peak photon energy|
|linear wavelength||282 GHz||1.063 mm||1.168 × 10-3 eV|
|linear frequency||160.23 GHz||1.871 mm||6.26 × 10-4 eV|
|log of either||222.6 GHz||1.347 mm||9.2 × 10-4 eV|
Plots against the log of either produce an equivalent plot: 1mm≡300GHz, 10mm≡30GHz, 100mm≡3GHz, etc.
SED fitting (aka spectral energy distribution fitting) is a method of classifying distant galaxies, to help determine their age and degree of star formation. A challenge is that a young galaxy with dust has a SED across ultraviolet, visible light, and near infrared as an old galaxy. A fitting method is Markov chain Monte Carlo (MCMC), using random walks around a parameter-space to determine the degree of unique fit between an SED and a known type of object. GalMC is an example of software that takes this approach.