The multi-wavelength shapes of galaxies - Spectral Energy Distributions

In this post I’d like to tell you a bit more about the spectral energy distributions of galaxies, what they tell us and how we can decode the properties of galaxies, such as a galaxy’s mass, using a galaxy's spectral energy distribution.

A spectral energy distribution (SED) shows how the energy output of a galaxy is distributed across the electromagnetic spectrum, meaning how bright the galaxy is at each particular wavelength. As was described in previous posts, galaxies consist of many different components, such as stars, gas and dust, and possibly active galactic nuclei (AGN). Each of these leaves a different and specific imprint on a galaxy's SED. For example, young stars are very bright and emit most of their light in the ultraviolet and blue optical (visible) part of the spectrum while old stars emit most of their light in the red optical (visible) and near-infrared. This difference is due to the different temperature of the stars: young stars are very hot, old stars are cooler. You can see this difference by comparing the different shapes of the SEDs for the elliptical (red curve) and spiral (Sd, blue curve) galaxies in the picture below. For a starbursting galaxy that is undergoing a merging event, such as Arp 220 (purple line), the SED shows that the galaxy is very bright in the infrared compared to its optical emission and compared to normal star-forming galaxies like the two spiral galaxies. That is why Arp 220 belongs to the class of ultra-luminous infrared galaxies which were described in a previous post.

Spectral energy distributions for typical galaxies - an old elliptical galaxy (red), two types of spiral galaxies (Sb in green and Sd in blue), an AGN (Markarian 231, solid black), a QSO (dotted black), and a merging and star-bursting galaxy Arp 220. Template spectra are taken from Polletta et al. 2007.

AGN on the other hand show themselves in the ultraviolet, optical, X-ray, and sometimes also at radio wavelengths. The overall shape of an AGN SED (shown here is that of Mrk 231) is similar to that of a power law, meaning in the figure above, the black line is very flat at optical and IR wavelengths. The SED of a QSO, a quasi stellar object, an object for which the AGN outshines the host galaxy in which it resides, is very steep and shows emission lines in the UV and optical.

If dust is present in a galaxy then some of the ultraviolet and blue optical light is absorbed and re-emitted in the infrared which can be seen as bumps in the purple, blue and green curves (around a wavelength of 5 to 110 micron). You might have also noticed the "spikes" of emission in the SEDs around about 5 to 12 micron, these are caused by so-called polycyclic aromatic hydrocarbons (PAHs), which are a class of organic molecules. They give important clues towards the structures of dust in galaxies, star formation, and the merger histories of galaxies.

Thus, from looking at the spectral energy distribution of a galaxy one can establish what kind of stars live in the galaxy (predominantly old ones or young ones), how much dust is present in the galaxy, with what intensity it formed stars over its lifetime, and if the galaxy has an active galactic nucleus or not and how active it might be.

Spectral energy distribution fitting

In order to extract the information encoded in the SED one can fit the SEDs of model galaxies to the data. This is called spectral energy distribution fitting. Let’s first look at the data needed for this. Instead of taking a spectrum - very detailed data showing how the light of the galaxy is distributed over a narrow wavelength range - one takes images of the galaxy in a number of filter bands (which only let the light in a narrow wavelength range through, see figure below). From each filter band image one obtains a number for the brightness of the galaxy at that particular wavelength (green dots with error bars in the last figure of this post). Combined, these numbers trace the overall shape of the SED of the galaxy. Since the different components of a galaxy (stars, dust, AGN) leave their traces at different wavelengths one aims at imaging galaxies over a wide wavelength range. The CANDELS fields with their vast data coverage are an ideal data set for this sort of analysis.

Shown are some commonly used filter bandpasses from the optical and near-IR wavelength ranges (wavelength on the x-axis). For each filter only the light of an object that is within the curve of the filter is detected, everything outside is excluded. Credit: Janine Pforr

On the other hand model galaxies are needed for comparison. So-called model SEDs or templates, can be created purely based on the theory of stellar evolution or by using real observed SEDs of typical galaxies. The model spectra for different types of galaxies are treated in the same way as the observed, to-be-analyzed data. The brightness of the model galaxies in each filter band is obtained and compared to the brightness of the real galaxy in the same filter band. The properties from the model galaxy – such as age, stellar mass, star formation rate and history, amount of dust - that best compares overall to the real galaxy are then assigned to the real galaxy.

Example of an SED-fit: the blue and red curves are two different model SEDs that are fit
to the green data points (brightness in the listed filter bands). The blue model represents
an old galaxy, nearly 4 billion years old, which does not contain any dust. The red model
represents a very young galaxy that is also very dusty. Both models fit the data equally well,
demonstrating the difficulties astronomers face when analyzing SEDs. Credit: Janine Pforr

Although this might sound simple, there are many galaxy properties that make the SEDs of very different galaxies look very similar over some wavelength range. One of the most prominent difficulties one faces is the ambiguity between the age of the galaxy and its dust content. For example, a young but dusty galaxy looks red (red curve in the picture to the left) but an old dust-free galaxy looks red too (see blue curve in the picture to the left).  

Since the only data required for SED-fitting is obtained from taking many images of a galaxy and with modern instruments it is possible to take images of tens of thousands of galaxies at the same time, SED-fitting is a very popular method to analyze large numbers of galaxies efficiently. The results can then be used to try to understand how galaxies formed and evolved with time. Many CANDELS team members use SED-fitting in their research and you will soon hear more about their exciting results.