Galactic Cartography For Beginners (part 2)

Poking about in the structure of our galaxy with Gaia EDR3 data.
Chris Street, 29 Jan 2022

[Experiment 3] Let's look at the far Milky Way. That sounds like a normal weekend activity.

As we saw in the first experiment, to see the nearby spiral structure of the galaxy, it was sufficient to use the Gaia EDR3 parallaxes directly to calculate distance. For this experiment, however, we'd like to penetrate a little deeper, and that requires a slightly different method.

We want to probe the far portion of the Milky Way in this experiment. We want to see the far spiral arms, distant satellite galaxies, deep halo stars, and whatever other things we can find. If only we had comic book style X-ray vision, and could just see through things like Superman. (Real X-ray vision is of course a thing in astronomy, but here we're working just with visible light so we're going to want the comic book kind that seems to work on everything.) Well, we sort of can have it. We can do it with science, because (in the words of the immortal Mr. Nye) science rules.

Bailer-Jones et al. (2021) have done a few things that will be helpful to us here: first, they apply a correction for a zero-point bias in the Gaia EDR3 parallax, and then, using photometric and geometric models, produce distance estimates that are (on average) more accurate than the estimates you get by simply inverting the measured Gaia parallax. These models take additional data, such as the star's color and extinction maps ("dust maps") of the Milky Way, into account.

(If you're wondering how star color factors into a distance calculation, it's because a star's color gives information about its absolute magnitude. For example, a red giant makes most of its energy by fusing its little remaining hydrogen via the CNO sequence. This information limits the luminosity/absolute magnitude of the star to a rather small interval: thanks to all the hard-working particle physicists, we know exactly how much energy these fusion reactions create and thus how bright such a star must be. If you know the absolute magnitude and apparent magnitude of a star, it's easy to determine its distance; this measurement can be even more accurate than a parallax determination for distant stars, if you can estimate the vabs sufficiently well.)

Without the Bailer-Jones et al. corrections, only about 15% of all Gaia sources have a parallax measurement with a signal-to-noise (PX/PXerr) greater than 5. With the Bailer-Jones et al. corrections, just under half of Gaia sources have distance measurements satisfying that inequality, at least according to their expected standard errors. (The underlying model is partly probabalistic and designed to decrease average distance error across large samples of Gaia stars, so whether it actually refines a specific individual source measurement is anybody's guess. Fortunately we're not using specific sources; we're using the entire Gaia dataset of nearly 2 billion sources in this experiment, so we'll count on non-systemic errors averaging out.) The Gaia sources that still have poor signal-to-noise are overwhelmingly distant objects.

In other words, after applying the corrections, we still don't have very accurate distances for objects on the other side of the galaxy (over 20,000 parsecs away), but we do have sufficiently accurate measurements for closer objects. For the experiment we're about to do, errors as much as 20%-30% either way aren't going to matter a lot, and the great majority of nearby sources will be well within that range.

When you look at the Milky Way, you're not just seeing the galaxy's nearby spiral arms, you're also seeing the far spiral arms shining behind them. How can you distinguish one from the other? If you want to see the stuff in the back -- but you don't know exactly where the back is -- the simplest answer is to subtract the stuff in the front. We know reasonably well which stars are in the galactic foreground, for foregrounds up to 10 Kpc or a little more -- our distance estimates are in aggregate accurate enough to get pretty clean results up to that distance. We can use that information to reveal the Milky Way behind the nearby objects.

So let's check out our new X-ray specs and take a peek behind the foreground stars. (Sorry, but this X-ray vision doesn't work on the galactic dust clouds. Even Superman has his limits, or so I'm told.)

We begin with the visible sky; here it is, our plaything:

Rectangular (Mercator) projection of the full Gaia EDR3 dataset. 0h right ascension is at the left and right edges, and increases right to left; 12h in the middle. -90° declination at bottom (south), 90° declination at top (north). Click the plot for full resolution.

With almost 2 billion stars to plot, it looks a lot like a long exposure photographic plate. (This is a 4,000 × 2,000 pixel image, as are all the other full-sky plots on this page; you are encouraged to click and examine them in full size.)

The foreground dust lanes that follow along the galactic plane, like the Great Rift, are easy to spot in any picture of the galaxy, but the foreground dust lanes away from the galactic plane are much clearer in a plot like this. Those dust lanes are mostly relatively close to us (2K-3K parsecs).

This is the only picture in this series that has the familiar stars and constellations visible (we're about to look straight through them with our unearthly powers), although all but the brightest are difficult to pick out. Orion's sword and the various dark nebulae associated with Orion Molecular Cloud Complex are quite prominent: look for a dark cloud shape in the Milky Way about 1/4 from right, nearly centered vertically. A bright line of vertical stars in front of the lower portion of the dark cloud is the sword; once that is identified the rest of Orion is easy to see. (Note that far northern or southern constellations, like Ursa Major, will be spread out on this map; it's a Mercator projection after all.)

All very good, but nothing we haven't already seen. Let's put these X-ray specs on.

Here is the visible sky, after subtracting 3,000 parsecs of foreground stars:

In this view, almost all of the Orion spur is gone, including nearly every naked eye-visible star, but the Perseus and Scutum-Centaurus spiral arms have barely even been touched. The galaxy is big.

Here is the visible sky, after subtracting 5,000 parsecs of foreground stars:

At this distance, we've eaten up about a third of the Perseus spiral arm rimward (away from the galactic center) from us. This is seen as the Milky Way becoming much dimmer toward the right of the plot.

The visible sky, after subtracting 7,000 parsecs of foreground stars:

7 Kpc looks like a continuation of 5 Kpc: the rimward Perseus arm is even dimmer, Sgr dSph is even more conspicuous, and the disappearance of the Near 3kpc arm and some of the closest bulge stars makes the central region less bright.

The visible sky, after subtracting 10,000 parsecs of foreground stars:

The center of the galaxy itself, and about half the Milky Way's central bar(s), has disappeared at this point. The rimward portions of the Milky Way are down to only a thin haze, not much brighter than the halo. The far spiral arms spread before us as the still-bright parts of the Milky Way in the left half of the plot -- appearing much shorter than the near arms did, thanks to perspective -- and even with the near dust lanes obscuring our view, it's quite clear the Milky Way, despite some messy patches like our own Orion spur, exhibits the symmetrical arm structure seen in most other barred spiral galaxies. Not only is Sgr dSph clearly visible, but its associated stellar streams (and other stellar streams) appear dimly in the galactic halo.

At 10 kpc, we're now getting close to the end of the usefulness of our data, but we might as well go one step farther and see if we find anything new to see in

The visible sky, after subtracting 15,000 parsecs of foreground stars:

15 kpc is just under 50,000 light years, and after taking that big a bite out of the galaxy, essentially only (parts of) the distant spiral arms are left along the galactic plane. Many of the stellar streams in the halo region are more prominent in this plot than the 10 kpc plot. Some non-halo stars are still visible rimward, but these are mostly sources with unreliable parallax, and are likely closer to us than 15 kpc.

At this distance, the errors in the distance estimates get troublesome. If we had perfectly accurate distances, the only sources rimward from us here would be scattered halo or extragalactic stars; what's left in the rimward part of the galactic plane in this plot are mostly the small percentage of sources that have silly high parallax errors, or that fell afoul of one of the model assumptions, and so ended with an overestimated distance.

Removing more halo stars did make the stellar streams more clearly visible looping through the Milky Way's galactic halo here, however, and this plot in its full resolution is an absolutely fantastic view of these extremely dim and distant objects, well worth pushing this experiment out so far.

Visible noise is getting more prominent, and as explained at the start, this is as far as our subtractive view of the galaxy can go while still being rooted in data and not measurement errors or model priors. But stop for a moment and think about some of what we saw with such a simple experiment:

Page content copyright © 2022 Chris Street.
ACKNOWLEDGEMENTS: This work has made use of data from the European Space Agency (ESA) mission Gaia, processed by the Gaia Data Processing and Analysis Consortium (DPAC). Funding for the DPAC has been provided by national institutions, in particular the institutions participating in the Gaia Multilateral Agreement.
Distance corrections taken from Estimating distances from parallaxes. V. Geometric and photogeometric distances to 1.47 billion stars in Gaia Early Data Release 3, C.A.L. Bailer-Jones, J. Rybizki, M. Fouesneau, M. Demleitner, R. Andrae, Astronomical Journal 161 147 (2021).