# Why do galaxies collide?

If the universe is expanding outward, what is the processes for one galaxy to get off track enough to collide with another?

Say, the Andromeda Galaxy and the Milky Way.

• This is a very little bit like asking why do molecules in an expanding cloud of gas collide.
– uhoh
May 14, 2019 at 4:50
• Also 'expansion' in space-time is weird. May 14, 2019 at 15:02
• @uhoh: except for all the differences, this is "same question, different scale". :-) May 14, 2019 at 16:41
• Not to be rude, but I just want to clarify that galaxies can't really get "off track". There's no set track for them to begin with (unless your question is about a universe simulator or something). They just are where they are and the laws of physics determine their motions. Gravity is the primary force that attracts galaxies to each other. May 14, 2019 at 17:50
• @uhoh: Or like "Why do people run into each other on the street if the universe is expanding. ^^ May 15, 2019 at 5:08

The universe is expanding on a large scale. But locally things are always messy.

Locally, galaxies are not set in stone, they move relative to each other, and the directions are random. If they're moving towards each other fast enough, then they will collide.

Also, there's gravity. Some galaxies are bound to each other by gravity, and that will tend to pull them together.

As to why galaxies move at all, relative to each other - well, things in this universe have kinetic energy, and it's distributed randomly. Being distributed randomly, all kinds of scenarios are possible - things running away from each other, zooming past each other, bumping into each other, etc.

It's a messy and random universe, and the order of expansion becomes apparent only on the largest scale.

• Is there proof that the energy distribution is indeed random? If this entails a separate Stack question, just let me know. May 14, 2019 at 13:23
• @Mindwin: CMB is random. Mass is random. What makes you think kinetic energy could be non-random? May 14, 2019 at 14:01
• @aleppke - I think you're confusing two different meanings of "random" - that of something behaving in a way that cannot be modelled (i.e. it is truly unpredictable), and that of something that can be modelled, but has multiple states to choose from (e.g. the roll of a dice). In this case, DevSolar is stating that just like CMB and mass, kinetic energy is not evenly distributed. If you imagine kinetic energy as arrows on paper pointing in random directions - the spaces between them may be moving outwards gradually, but some of the arrows are still pointing at each other. May 14, 2019 at 16:12
• @aleppke That's actually a very big discussion. True randomness does exist, it's the outcome of certain quantum phenomena. But it is true that trajectories of bodies are quite deterministic. However, many bodies moving under gravity make complex, chaotic systems. The behavior after a long time is essentially impossible to predict except on a statistical level. Don't get stuck on the word "random" in the reply as it was not used in a strict, naive sense. What I meant is - it's a complex distribution that simulates some aspects of true randomness well enough for this discussion. May 14, 2019 at 18:23
• @Mindwin: Simply put, it is the most plausible: Considering lots of masses moving independently (i.e., any gravitational influence is secondary), they could be all at rest relatively to each other - but that would give preference to one inertial reference frame (at least regionally); hence they move. Any preferred direction or common rotation would be against the anisotropy we do strongly expect May 14, 2019 at 19:19

The galaxies don't really get "off track" - it's not impossible, but that kind of thing probably doesn't happen anymore (as space continues to expand). What actually happens is that galaxies form gravitationally bound clusters - within the cluster, the acceleration due to gravity is larger than the equivalent expansion of space between the galaxies, so rather than growing more distant, the galaxies in question actually get closer together over time. Eventually, this results in a collision and a merger.

If the expansion remains roughly constant, there will come a point where we'll no longer be able to see any galaxies outside of our own cluster. But for those close enough, this has little effect - just like the expansion of space doesn't cause atoms, planets, solar systems or galaxies to get bigger.

• " the galaxies in question actually get closer together over time. Eventually, this results in a collision and a merger". Really? How do they shed their kinetic energy? Are gravitational waves sufficient? What's the supposed time scale? May 14, 2019 at 13:20
• @PeterA.Schneider gas/dust clouds collide directly, tidal interactions throw streamers of stars away from the combined mass of the 2 galaxies carrying momentum away. AFAIK gravitational waves a total nonfactor in the galaxies merger (but would be involved in an eventual merger of the central black holes). The timescale is hundreds of millions to a billion or two years. May 14, 2019 at 14:42
• @DanNeely I see. The reason being that while intergalactic space is vast, it's not as relatively vast as interstellar space... May 14, 2019 at 14:55
• "Eventually, this results in...a merger" - just wait. "Dear Sol: as you're aware, the recent merger of the Milky Way and Andromeda galaxies has formed the new MilkyMeda super-galaxy. Unfortunately, due to consolidation and restructuring efforts a limited number of positions have been made redundant including yours. We sincerely regret this, and are further saddened to note that no positions requiring your unique set of skills are available. Please gather your personal effects, including all moons and planets, and report to the nearest black hole for repurposing. Wishing you all the best..." May 14, 2019 at 16:57
• @Acccumulation Yes of course; the keyword was "relative", to the sizes of planetary systems and galaxies, respectively. Galaxies in clusters are much closer to each other relative to their diameters than planetary systems to each other, relative to their diameter. That's why we don't have many tidal perturbations between planetary systems, but more frequently between galaxies. May 14, 2019 at 21:51

I'm not sure that anyone has answered the question asked. The root cause is indeed that gravitationally bound structures with freefall timescales that are much shorter than the age of the universe are not greatly affected by the general expansion of the universe (NB: Structures with freefall timescales longer than this are not going to be the source of many galaxy collisions). That is, locally, the expansion within such structures is negligible. However, that does not necessarily lead to collisions on a a timescale shorter than the age of the universe.

The first reason for galaxy collisions is that galaxy clusters have a very large number density - that is, the spacing between galaxies is not vastly larger than the "size" of a galaxy, where here, "size" means the effective interaction cross-sectional radius. As a result of these high densities, the freefall dynamical timescales in rich clusters (and even smaller groups of galaxies) are of order billions of years and so there is plenty of time for the galaxies to interact. As a contrast, think about how you might construct a scale model of stars in the local neighbourhood and compare the sizes of stars with their separations. It would in fact be difficult to make such a scale model with any meaningfully sized stars. On the other hand, you can make a scale model of say the local group of galaxies because their separations are only $$\sim 10$$ times their sizes.

The second reason is that many galaxies contain gas and that gas can easily dissipate kinetic energy and also transfer angular momentum. Another factor is that massive clusters of galaxies contain intracluster gas that can also serve to dissipate kinetic energy. In a gravitationally bound system, then objects that are in orbit around each other or around a common centre of mass need ways in which kinetic energy and angular momentum can be lost in order for a collision to take place. Even without gas, the fact that galaxies exist in groups and clusters means that n-body interactions can serve to dissipate energy and angular momentum to make a collision happen.

• My very vague understanding suggestion that "gravitationally bound structures are not affected by the general exapansion of the universe" is limited to those whose orbital periods are short compared to the inverse Hubble constant. Is that reasonable or am I off track? May 14, 2019 at 17:46
• Also, is it worth being explicit that clusters are multi-body situation meaning that the simple conic sections of two-body orbital mechanics are not the whole story? May 14, 2019 at 17:47
• +1 but I've always been uncomfortable with the oft-repeated phrasing (in various forms) of "gravitationally bound structures are not affected by metric expansion of space" Doesn't metric expansion happen everywhere, but is less observable in gravitationally bound systems because its effect is dominated by local motion due to the system being gravitationally bound? It doesn't actually "repel" or block or shut down metric expansion, as much as it simply dominates it observationally? I can ask this as a separate question if that provides a better format.
– uhoh
May 15, 2019 at 5:26
• Is "facto" a typo or some sort of specialized usage I've never seen before? May 15, 2019 at 14:57
• @phoog I propose that we make "facto" a word. Defined as a a piece of information often presented as fact which is defensible under specific circumstances but usually incomplete or misleading to a level that it should not be presented as fact. This definition makes popular science writing the art of lining up factos in a pleasing manner. May 16, 2019 at 18:35

Galaxies don't get "off track" - to see how collisions happen, we need to go right back to galaxy formation early on.

So, Big Bang happens. Space starts to expand - dramatically and to a huge extent. That's space itself expanding, not galaxies moving within space, by the way - distances themselves change. (Which is why it's called a "metric" expansion, metric being a term for distance-measures, and also why cosmologists say the Big Bang happened "everywhere").

At some tiny fraction of a second, the massive expansion winds down. Space continues to expand, but at a much slower rate. The last of the fundamental forces breaks away, and the cosmos is left as an insanely hot dense mix, so hot that even basic particles like protons, neutrons and electrons can't yet exist - although quarks can.

But there are some very subtle things going on. Even though expansion left us with an incredibly uniform, homogenous universe, the density does sliiightly vary between places. As things cool down, and particles start to condense (and annihilate, and other things), the universe is left with what cosmologists call acoustic waves - basically standing waves. And if you've ever seen videos of a tray of sand being vibrated, you'll know that one effect is that it leaves some places with more sand, some with less, due to interference patterns. So our universe ends up, as it expands, with some areas denser, some less dense.

A second effect comes into play. You'll know (or have heard of) dark matter. We don't know what it is made of, but we know it exists (galaxies couldn't form without it, they'd fly apart or take longer than the age of the universe to form), and we know a lot about how it behaves - what forces it responds to, and what forces it doesn't. Interact via gravity - yes, very weakly. Interact via electromagnetic force - no, not at all. That latter bit is crucial.

When "ordinary" matter collapses, it heats up. That's how we get stars, for example. The radiation released during collapse also acts as a kind of pressure, opposing collapse, slowing it down. That's why stars like our sun are stable for so long. Dark matter doesn't interact electromagnetically (as far as we know) so it can't experience or create electromagnetic radiation. So when it collapses, it doesn't get hot, it doesn't release radiation... I think you can see where this is going. There's no radiation released during collapse to resist further collapse, so it can collapse much faster than ordinary matter. As an aside, because it can't release radiation, it also can't jettison the energy that must be got rid of to allow dense objects to form. So it ends up quickly collapsing to a hazy diffuse "halo", but then can't collapse much more. And no surprise, it collapses in those places where the universe was fractionally denser. So you get what cosmologists call "filaments" and "halos" of dark matter, a bit like a sponge or a swiss cheese, with comparative "voids" separating them. Ordinary matter is more strongly attracted to these already-existing dark matter filaments and halos. It collapses towards them. The self-gravity of the ordinary matter is enhanced by the gravity due to the concentrations of dark matter there - and ordinary matter can lose energy by radiation, so it collapses more than dark matter does, to form the galaxies and their contents which we can see today.

Gravity can do this, because the expansion of the universe has by now slowed down so much from its "heyday", that gravity can pull some of the matter together within space faster than the expansion can add space between them. Over cosmic distances, gravity is much much weaker, and expansion dominates, so clusters and superclusters still move apart, but within clusters, the galaxies and groups of galaxies are accelerated by gravity enough that they mostly stay in their groups and clusters, and move around or orbit within them.

So we end up with a universe that, on a cosmic scale, we see expansion "winning" as gravity is weak, so we see superclusters moving apart. But within clusters and galaxy groups, we see gravity "winning" because it's stronger over smaller distances, so clusters and gravitationally bound entities like galaxies stay together.

What this in turn means, is that galaxies and galaxy groups are bound by gravity more than they are separated by expansion. So they remain moving within their clusters and groups, despite universal expansion. And, occasionally, because motion of 3 or more separate bodies under gravity is chaotic (and clusters can contain billions or trillions of galaxies), entire galaxies will get ejected, or collide, or do whatever galaxies do. And that's how it happens.

(Although you didn't ask, it's a natural question to wonder what happens next. We believe that the rate of expansion has slowly sped up. That means that in the far future (tens and hundreds of billions of years), that galaxies will have to be even closer together, for gravity to dominate expansion. So clusters that are stable now, might break apart in the far future. If expansion accelerates enough then even smaller bodies could ultimately break up, perhaps galaxies themselves, or even stars and atoms. But that's something nobody knows.)

I think it's also worth noting that cosmological expansion is measurable only on the very largest of scales. Hubble's law tells us that the farther away an object is, the faster it's being pulled away by the expansion. That rate is approximately $$70 km.s^{-1}Mpc^{-1}$$ -- in other words, for every megaparsec (Mpc -- about 3 million light years) that you look farther out in space, an additional $$70 km.s^{-1}$$ is tacked onto its recessional speed. For a relatively close galaxy, its actual velocity through space can be much larger than this, and in the specific case of the Andromeda Galaxy, which is only about 2.5 million light years (0.77 MPc) away, it is actually approaching us at around $$110 km.s^{-1}$$. For much, much farther galaxies, billions of light years away, all of those $$70 km.s^{-1}$$ add-ons stack up to a much, much higher recessional speed.

Although the Universe is expanding and, in general, the further away a galaxy is from us the faster it appears to be moving away from us. This does not apply to the galaxies in the Local Group. which is a gravitationally bound structure. The Andromeda galaxy is moving towards the Milky Way at about 400,000 km/h and the Milky Way and Andromeda are expected to collide in about 4 billion years time. When this happens, a large new single galaxy will be formed. The new galaxy which will be formed by the merger is sometimes given the name Milkomeda . For more details see my recent blog post on this topic.

Over billions of years, Milkomeda will gradually absorb the other Local Group members.

In general any gravitationally bound structure such as: stellar systems (e.g. solar system ) our galaxy and groups and clusters of galaxies will not get larger as the Universe expands)