Follow the reluctant adventures in the life of a Welsh astrophysicist sent around the world for some reason, wherein I photograph potatoes and destroy galaxies in the name of science. And don't forget about my website,

Wednesday 13 March 2013

Galaxies Are Pretty

EDIT : This is an old post from 2013, updated as of March 2016 to include new data.

I've always wanted to understand how the Universe works. This is proving rather difficult, so I'm prepared to settle for knowing what it looks like.

Since I already describe the technical aspects of how the video was produced on my website and in an earlier blog post, I'm not going to do it again because that would be silly. EDIT : I will just add a brief note - you might notice in the video that there's a difference in the number of galaxies detected by ALFALFA and actually shown here. That's because some of them aren't galaxies at all, and a few aren't properly identified by the SDSS automatic algorithms.

 Maybe a more interesting question to ask is : how accurate is the video ? Is this really what the Universe looks like ?

Well, no, it isn't. It's an approximation, but it's grounded in reality. Allow me to try and explain, because it's my blog and you can't stop me.

The complete, full HD video. Partially in 3D if you've got those 
old red-blue glasses. The original can be seen here.

How do we know where all the galaxies are ?

We don't. Just looking at an image isn't always enough to tell what's a galaxy and what isn't. For example... THIS

isn't a galaxy. It's just a crummy old star cluster, and therefore useless. Whereas THIS

is a genuine galaxy and no mistake.

Just like with James Spader and Michael Shanks, there's no way to tell the two apart by just ogling them. But if we measure how fast their stars are moving, we can see that all is well with the star cluster, but the stars in the galaxy are moving way too fast. Without some invisible matter to hold them all together, they should all fly apart (unless we've got something very wrong, which is possible, but in the interests of time let's not go there). So galaxies have dark matter, and star clusters don't.

Whereas distinguishing Spader / Shanks is simply hopeless.
Unfortunately we can't always get accurate measurements of how fast the galaxies are rotating. Sometimes, though, we can measure how far away they are, and that tells us if we're looking at a small bunch of stars very nearby (which would be a star cluster) or a big bunch of stars very far away (which would be a galaxy). More on that in a minute.

Still, the upshot is that identifying galaxies is tricky, and we can't be sure we've found them all. In fact, we know we haven't, because the Milky Way is big, bright, and in the way.

And very pretty.
More important is that the galaxies in the video are selected from the ALFALFA survey. This is a hydrogen gas survey, and galaxies which have hydrogen tend to be forming lots of stars and very blue. So the video is missing a lot of the reddest galaxies, even the really bright ones, because these often lack hydrogen gas. This was a deliberate choice, because I made the video to demonstrate the capabilities of Arecibo in a photogenic way, and neither Michael Shanks nor James Spader seemed to fit the bill.

How do we know how far away the galaxies are ?

We may not know exactly where all the galaxies are, but for those we do know about we generally have pretty good estimates of how far away they are - even though we can't directly measure the distances to other galaxies.

The only way we can directly measure the distances to astronomical objects outside of our own solar system is through parallax, where we see how much a star moves relative to the stars behind it. That's good for nearby stars, but for more distant objects we have to use other methods (calibrated on the ones we can get direct measurements for) to create a 'distance ladder'. And that lets us get pretty good estimates of how far away galaxies are.

Full size image here.
The most important result of this is Hubble's Law. This allows us to use measurements of how fast the galaxies are moving away from us (a relatively easy measurement) and convert it into a distance. And we think this works pretty well... most of the time. The problems start when you get two or more galaxies close together. Like two sumo wrestlers on a trampoline, the force of attraction is irresistible*  and they can't help but fall towards each other. And that screws up the conversion of velocity to distance.

* It's not really irresistible - it's relatively rare for galaxies to collide and merge**, but they do affect each other's velocity.
** Merging is even rarer - and considerably more unpleasant - for sumo wrestlers.

This is particularly bad in galaxy clusters, which are the equivalent not of sumo wrestlers, but a bunch of obese, angry rabid cats on a trampoline, possibly on fire - the Platonic ideal of chaos. Here, the velocity measurements tell us practically nothing about distance. Some galaxies are even coming toward us, which would give a meaningless, negative distance if we were silly enough to take Hubble's Law at face value. So the galaxy distances in the GIF are pretty good, but far from perfect. There's just no way around that, because better data doesn't exist.

Do the galaxies really look like that ?

No they don't. The images used are optical images from the Sloan Digital Sky Survey, but just because they were observed at similar wavelengths that we use to see with doesn't mean they look like what you'd see with your naked eye.

For starters, they're all much too bright. Think about what the Milky Way looks like on a clear, dark night. Does it really look like this ?

Hell no. The real thing looks beautiful, but nowhere near as bright or as colourful - not even in the darkest spot on the darkest night. Images like this are only possible with long exposures, or large telescopes. So out of necessity - because the video would show bugger all otherwise - the galaxies are shown much brighter than you'd really see them.

The colours aren't 'true' either. but that shouldn't bother you. Everyone perceives colour in a slightly different way - it's practically a subjective choice, rather than a measurement. And then, of course, some people are colourblind. The important point is that the colours are relatively correct - things which look bluer really are bluer than things which are redder.

EDIT : Since making the original, the Sloan Digital Sky Survey (which is where the galaxy images come from) updated their data processing techniques. The results make the galaxies look rather redder and blue features almost disappear - especially given the particular technique I'm using to display them in Blender. Fortunately, a bit of simple colour correction restored the galaxies to their former glory.

Left to right : 1) Raw image from the SDSS; 2) Galaxy as displayed in Blender without colour correction; 3) Galaxy displayed in Blender with colour correction.
This makes a very dramatic difference indeed for the whole set of galaxies. Here's what they would look like if I hadn't done the correction :

Flat, dull yellow. Drab, even. In the words of Bart Simpson, "Wow, the Universe is so boring !". Fortunately, that bit of colour adjustment brings out all the shiny details and stops it from looking a particularly unpleasant shade of beige.

A Universe rich in stars, galaxies, flying spaghetti monsters, etc.

What about the sizes of the galaxies ? Like brightness, these are exaggerated out of necessity, by about a factor of 40. That means that you can see the structures in many galaxies at once, without hoping that the viewpoint passes very close to them (otherwise you'd just see a bunch of small, uninteresting fuzzy blobs). However, as with colour, the relative differences are correct : a galaxy that looks twice as big as another really is twice its size.

One thing there was no need to compromise was the structures visible in the galaxies. Each of the 23,935 galaxies in the video is represented by an optical image of the actual galaxy seen at its location. This is pretty unusual - many of these large-scale videos often just use a few hundred template images. Here, there are unique images of every single blasted one. Don't believe me ? Here's the complete sample.

EDIT : This is from the original version of the post, which used a mere 11,710 galaxies. I can't be bothered to remake this.

View this in full, terrifying detail here.
Astute viewers may note that some of the images look strange. Unfortunately, there aren't any good estimates available for how wide the galaxies are. A large survey like this has most of the data processed automatically, and inevitably sometimes something goes wrong - especially on the large galaxies, which are often (incorrectly) broken up in the catalogue into several pieces.

I had a choice here. The first option was to carefully check all 11,710 23,935 images and make sure the galaxies were the correct size. This option was stupid, so I chose the second option, which was to ignore the problem completely.

Finally, we'll never know exactly what some galaxies really look like, because we're seeing them from a jaunty angle. We can correct for this to some extent, but only if we know the minimum and maximum dimensions of the galaxy. Which, for this sample, we don't. And if a galaxy is edge-on to us, there's nothing at all we can do except go home and cry.

In summary then, the video shows what you'd see if you had super-sensitive eyes and all the galaxies were forty times bigger than they actually are, with lots of errors. The large-scale structures of filaments and voids you see are correct, but not complete. That's a correctable problem, and one I might tackle at some point. But since downloading 11,000 images took about 8 hours, it can wait.


  1. Just absolutely wonderful! Keep up the great work. And keep posting to G+, Space, etc -- so I can see it.

  2. This comment has been removed by a blog administrator.


Due to a small but consistent influx of spam, comments will now be checked before publishing. Only egregious spam/illegal/racist crap will be disapproved, everything else will be published.