How SDSS Uses Light to Measure the Mass of Stars in Galaxies

Screen Shot 2014-11-20 at 12.24.55

Galaxy NGC 3338 imaged by SDSS (the red stars to the right is in our own galaxy). Credit: SDSS

It might sound relatively simple – astronomers look at a galaxy, count the stars in it, and work out how much mass they contain, but in reality interpreting the total light from a galaxy as a mass of stars is fairly complex.

If all stars were the same mass and brightness, it would be easy, but stars come in all different brightnesses, colours and masses, with the lowest mass stars over 600 times smaller than the most massive.

Hertzsprung-Russel Diagram identifying many well known stars in the Milky Way galaxy. Credit: ESO

Hertzsprung-Russell (HR) Diagram, which shows the mass, colour, brightness and lifetimes of different types of stars. This version identifies many well known stars in the Milky Way galaxy. Credit: ESO

And it turns out that most of the light from a galaxy will come from just a small fraction of these stars (those in the upper left of the HR diagram). The most massive stars are so much brighter ounce for ounce than dimmer stars this makes estimating the total mass much more of a guessing game than astronomers would like (while they are 600 times more massive, they are over a million times brighter). So astronomers have to make assumptions about how many stars of low mass are hiding behind the light of their brighter siblings to make the total count.

One of the first astronomers to suggest trying to decode the light from galaxies in this way was Beatrice Tinsley. British born, raised in New Zealand, and working at Yale University in the USA, Dr. Tinsley had a much larger impact on extragalactic astronomy than her sadly shortened career would suggest (she died of cancer in 1981 aged just 40).

Stars of different masses have distinctive spectra (and colours), as first famously classified by Astronomer Annie Jump Cannon in the late 1890s into the OBAFGKM stellar sequence. O stars (at the top left of the HR diagram) are massive, hot, blue and with very strong emission lines, while M stars (at the lower right) are low mass, red and show absorption features from metallic lines in their atmospheres. With a best guess as to the relative abundance of different stars (something we call the “initial mass function“) a stellar population model can be constructed from individual stellar spectra or colours and fit to the total light from the galaxy. Example optical spectra of different types of stars are shown below (or see the APOGEE View of the IR Stellar Sequence)

Example optical spectra of different stellar types. Credit: NOAO/AURA/NSF

Example optical spectra of different stellar types. Credit: NOAO/AURA/NSF.

Using data from SDSS (and other surveys) astronomers use this methods to decode the galaxy light – in fact we can use either the total light observed through different filters in the SDSS imaging, to match the colours of the stars, or if we measure the spectrum of the galaxy we can fit a population of stars to this instead. While in principle the spectrum should give more information, in SDSS (at least before the MaNGA survey) we take spectra through a small fibre aperture (just 2-3″ across), so for nearby galaxies this misses most of the light (e.g. see below), and most galaxies have colour gradients (being redder in the middle than the outskirts), so the extrapolation can add quite a lot of error to the inferred mass.

NGC 3338 with the approximate SDSS fibre size overlaid (note this is an example of a very large galaxy imaged by SDSS). Credit: SDSS, KLM

NGC 3338 with the approximate SDSS fibre size (ie. the part of the galaxy for which we measured spectra) overlaid (note this is an example of a very large galaxy imaged by SDSS, and not representative of most galaxies). Credit: KLM, SDSS


Many astronomers prefer to use models based on the total light through different filters (at least for nearby galaxies). The five filters of the SDSS imaging are an excellent start for this, but extending into the UV with the GALEX survey, and IR with a survey like 2MASS or WISE adds even more information to make sure no stars are being missed. However, these fits are still a “best guess” and will still have error –  there is often more than one way to fit the galaxy light (e.g. model galaxies with certain combinations of ages and metallicities can have the same integrated colours), so there’s still typically up to 50% error in the inferred mass.

The SDSS camera filter throughput curves (from left to right ugriz). Credit: SDSS

The SDSS camera filter throughput curves (from left to right ugriz). Credit: SDSS


But with galaxies spanning more than 3 orders of magnitude in total mass (ie. the biggest galaxies have more than a 1000 times the stellar mass of the smallest) this is still good enough for many purposes. It gives us an idea of the total mass in stars in a galaxy, which (as you know from earlier post for IYL2015) is almost always way less than the total mass we estimate from looking at the dynamics (ie. the “gravitating mass”). And the properties of galaxies correlate extremely well with their stellar masses, so it’s a really useful thing to have even an estimate of.

This post by Karen Masters is part of the SDSS Celebration of the International Year of Light 2015, in which we aim to post an article a month in support of the celebration of light.