How the Milky Way-Andromeda merger became so uncertain

If you had asked any informed astronomer about the fate of our galaxy here in the 21st century, you would have gotten a near-universal story from practically all of them: in about four billion years, the Milky Way would merge with the only galaxy larger than itself in the entire Local Group, Andromeda. Andromeda holds the distinction of being the very first object correctly identified as being outside the Milky Way itself: when Edwin Hubble identified and measured a special class of star (Cepheid variable stars) in Andromeda back in 1923. These stars allowed Hubble, building on the earlier work of Henrietta Leavitt, to measure the distance to Andromeda, determining that it was far, far outside the full extent of the Milky Way.

While nearly all subsequently-discovered galaxies in the Universe are measured to be receding from us — caught up in the expansion of the Universe — Andromeda is different. Instead of moving away, Andromeda is headed towards us at an impressive 109 km/s: about twice as fast as Halley’s comet moves at its maximum speed. Because of this speed, its distance of about 2.5 million light-years from us, and the mass-based gravitational attraction of the Milky Way and Andromeda, it was simple and straightforward to conclude that in about four billion years, the Milky Way and Andromeda would collide and merge.

Only now, with new data from Hubble and Gaia about galaxies from all across the Local Group, it looks like that picture is too simple and naive, and may not be right at all. Here’s what the newest data teaches us.

A series of stills showing a visualization of the Milky Way-Andromeda merger and how the sky will appear different from Earth as it happens. This merger has long been expected to begin roughly 4 billion years in the future, with a huge burst of star formation leading to a depleted, gas-poor, more evolved galaxy ~7 billion years from now. However, new research challenges the likelihood and timescale of this event, throwing this classic picture into doubt.

From a cosmic perspective, it makes sense that Andromeda and the Milky Way would inevitably collide and merge. After all, the way that cosmic structure forms is as follows.

• The Universe was seeded, during cosmic inflation, with density imperfections (i.e., underdense and overdense regions) on all cosmic scales.
• The small cosmic scales, only hundreds or thousands of light-years across, collapse first, creating star clusters and small galaxies.
• Larger cosmic scales, comparable to the size of the Milky Way or Andromeda, collapse later, leading to large galaxies and the merger and accretion of many smaller ones.
• Even larger cosmic scales, like those of galaxy groups (including the Local Group) and galaxy clusters, are the last to form, requiring billions of years and having not yet completed forming even today, 13.8 billion years after the Big Bang.
• And on the largest of cosmic scales, or scales greater than ~1.5 billion light-years, all apparent structures are nothing more than phantasms, as they will never be gravitationally bound and dark energy will drive them apart.

It would only make sense, then, that we would find ourselves living in a galaxy group or cluster that consisted of many galaxies of various sizes at the present time, with many mergers still to come in the future.

This view of the Perseus cluster of galaxies, from ESA’s Euclid mission, shows over 1000 galaxies all clustered together some 240 million light-years away, with many tens of thousands more identifiable in the background portion of the image. While optically, the image is dominated by the most massive, star-rich galaxies, they are vastly outnumbered by smaller, fainter, low-mass galaxies that are exceedingly difficult to detect, even nearby. Our Local Group is a much smaller, lower-mass group of galaxies than this.

After all, with both dark matter and dark energy present in our Universe, there ought to be a tension between the gravitating effects of dark (and normal) matter, working to pull nearby objects in the Universe together, and the expansive effects of dark energy, working to relentlessly drive distant objects apart. In all cases, we would expect there to be a maximum “size” to any bound structure that was made of matter — whether stars, star clusters, small galaxies, large galaxies, galaxy groups, galaxy clusters, or even collections of galaxy clusters — and that every bound structure would contain many smaller bound structures within it.

Only after a sufficiently long time has passed will every object within a bound structure merge together, giving rise to a single behemoth galaxy that contains all of the normal matter within that structure in a lone object. We find galaxy groups and galaxy clusters all across the Universe, containing many large galaxies within them that are often in the process of interacting, colliding, and merging together. It makes total sense that, eventually, not only will the Milky Way and Andromeda merge together within our own Local Group, but that all of the smaller galaxies that comprise our Local Group as well will eventually merge as well, becoming part of one giant galaxy far (or maybe even not all that far, cosmically speaking) into the future.

Known colloquially as a “perfect 10,” Arp 147 features two interacting galaxies where each one features a ring, almost certainly as the aftermath of a center-on-center collision between the two precursors. The dusty reddish knot at the lower left of the blue ring probably marks the location of the original nucleus of the galaxy that was hit.

You might think — as many astronomers have long thought — that if you can identify any two galaxies within a larger bound structure, like a galaxy group or galaxy cluster, and can find those galaxies moving toward one another, that’s simply a prelude to an inevitable collision and merger. Indeed, we see evidence for this in many cases, with the impressive Arp catalogue of interacting galaxies, also known as the Atlas of Peculiar Galaxies, showcasing hundreds of such galaxies in the process of colliding and merging. Since we only can observe the Universe over very brief periods of cosmic time, as all of recorded human history goes back mere thousands of years, while the timescale of galaxy mergers is tens-to-hundreds of millions of years, it makes sense to measure the snapshots we can measure and infer the rest.

But there’s a limitation to this line of reasoning: we can only easily observe how objects are in motion in the Universe along our line-of-sight to them, or how quickly they’re either moving towards us or away from us. This is important, because space is a three-dimensional place! Sure, galaxies are in motion along our line-of-sight, but they’re also allowed to have transverse motions: motions that are perpendicular (i.e., up-down or left-right) to our line-of-sight. We can’t measure these motions with just a single snapshot; it would take decades or even centuries to detect even slight 3D motions for objects that are millions of light-years away.

The European Space Agency’s space-based Gaia mission has mapped out the three-dimensional positions and locations of more than one billion stars in our Milky Way galaxy: the most of all-time, as well as many stars in sufficiently nearby galaxies as well. Looking toward the center of the Milky Way, Gaia reveals both light-blocking and luminous features that are scientifically and visually fascinating.

However, the last few years have seen two major revolutions in the science of mapping the Universe not just as we see it, projected on the sky and with the ability to measure motions towards or away from us, but in all three dimensions more fully. The first and most major advance comes from the ESA’s Gaia mission, whose goal was to measure a wide variety of objects — mostly nearby, like stars in the Milky Way, but also stars and galaxies that lie outside of our galaxy as well — with exquisite precision over time. For stars, this allowed us to track their variability, measure their parallax, and determine their proper (i.e., three-dimensional) motions through space.

But for galaxies, or large collections of individually identifiable stars within such a galaxy, we could begin to detect their three-dimensional motions as well, especially as those motions are influenced by other galaxies in their vicinities. For example, the Milky Way isn’t an isolated galaxy, but rather has two substantial satellite galaxies located within 200,000 light-years of ourselves: the Large Magellanic Cloud and the Small Magellanic Cloud, where the former is actually quite massive, weighing in as the fourth most massive galaxy in the Local Group, behind only Andromeda, the Milky Way, and Triangulum.

Our Local Group of galaxies is dominated by Andromeda and the Milky Way, but there’s no denying that Andromeda is the biggest, the Milky Way is #2, Triangulum is #3, and the LMC is #4. At just 165,000 light-years away, the LMC is by far the closest among the top 10+ galaxies to our own, and as such it takes up the largest angular span on the sky of all galaxies outside the Milky Way. There are over 100 galaxies within the Local Group, but Andromeda and the Milky Way contain most of the stars, as well as most of the mass.

Meanwhile, Andromeda is much farther away, at 2.5 million light-years, but possesses satellite galaxies of its own. Even though it isn’t technically a satellite of Andromeda, the also-massive Triangulum galaxy is located much closer to Andromeda than it is to the Milky Way, implying that there’s a substantial gravitational interaction occurring between Triangulum and Andromeda, just as there’s one occurring between the Milky Way and the Large Magellanic Cloud. While Gaia might be well-equipped to determine the proper motions of stars in the Large Magellanic Cloud relative to the Milky Way, Triangulum and Andromeda are too far away for Gaia’s specialized capabilities.

But that’s where the second major advance comes in: sustained observations over long (a decade or more) timescales of Andromeda and Triangulum with the Hubble Space Telescope! By conducting long-period observations with Hubble, it becomes possible to measure the three-dimensional motions of stars inside these galaxies, and hence to infer the overall motions of the galaxies themselves. By combining these measured motions (along with the uncertainties in those motions) with estimates of the masses of the relevant galaxies in question (Andromeda, the Milky Way, Triangulum, and the Large Magellanic Cloud), it becomes possible to simulate the most likely trajectories of these galaxies into the far future.

These panels show the simulated trajectories of the Milky Way (MW, green) and Andromeda (M31, pink) in both the face-on (top) and edge-on (bottom) directions. The left columns include simulations of the Milky Way and Andromeda alone, while the right-hand panels also include the Triangulum (M33) and Large Magellanic Cloud (LMC) galaxies.

After all, astrophysics is just the science of physics as applied to astronomical objects, and perhaps the best-understood physical law is the law of gravity. If you can measure:

• the three-dimensional positions of all of your relevant objects,
• the masses of those objects,
• the initial three-dimensional velocities of those objects,
• and you know how gravity works (allowing you to determine the accelerations of those objects),

then you should be able to use physics to calculate and evolve the trajectories and positions of the relevant objects as far into the future as you like.

Of course, there are associated measurement uncertainties with all of these quantities: position, mass, and velocity. Only the law of gravity is exquisitely known. What this enables, however, is the ability to conduct simulations that model those uncertainties and translate the various possible sets of parameters into a likelihood estimate. When we take the best position and velocity data that we have from Hubble and Gaia of these objects, and when we include what we know about their masses from a variety of lines of evidence, we can then look at the probability and timescale of a collision-and-merger occurring between the two most massive galaxies in the entire Local Group: Andromeda and the Milky Way.

This four-panel illustration shows the results of 100 simulations each for a model of the Milky Way-Andromeda system, along with the likelihood and timescale of a merger occurring. The top-left shows only the Milky Way and Andromeda, the lower-left also includes Triangulum (M33), the bottom-right excludes Triangulum but includes the LMC, and the top-right includes all four.

This is precisely the topic of a new article in Nature Astronomy, published by a team led by Dr. Till Sawala, which turns the naive conclusion that these two galaxies will merge in ~4 billion years on its head. When you look at the two galaxies on their own, ignoring the effects of Triangulum (M33) and the Large Magellanic Cloud (LMC), it does indeed appear very likely that a merger will occur between the Milky Way and Andromeda. In fact, the most likely outcome appears to be that:

• the Milky Way and Andromeda will get close to one another (within about 500,000 light-years) about four billion years from now,
• then move away to about 700,000 light-years after about another 2 billion years,
• and then come together to merge between 7-and-8 billion years from now.

If you add in the Triangulum (M33) galaxy, the probabilities of a merger occurring go up, and the timescale for these interactions decreases. For example, including Triangulum means that:

• the Milky Way and Andromeda get even closer (within just ~300,000 light-years) about 4 billion years from now,
• then move away to about 500,000 light-years after another 1 billion years,
• and then come together to merge about 6-7 billion years from now.

Whereas the models of just the Milky Way and Andromeda resulted in 44% of simulations leading to a merger in less than 10 billion years, including the Triangulum galaxy resulted in 63% of simulations leading to a full-on merger in the same amount of time.

This image shows the trajectories of the Milky Way in the Milky Way-Andromeda orbital plane (top) and perpendicular to the plane (bottom) if we consider the effects of Triangulum (left), the LMC prior to its merger with the Milky Way (middle), and the LMC after its perger with the Milky Way. The inclusion of Triangulum makes a near-term merger more likely; the inclusion of the LMC makes a near-term merger less likely, as it pulls the Milky Way away from the direction of Andromeda.

But the Large Magellanic Cloud, often ignored in this type of analysis, turns out to play a major role. The orbit of the Large Magellanic Cloud with respect to the Milky Way can now be inferred, and it turns out it’s in motion in the perpendicular direction with respect to the Milky Way-Andromeda orbit. When we include the LMC in these simulations, the researchers find three major things out about the fate of the massive galaxies in our Local Group.

1. The presence of the LMC serves to pull the Milky Way away from the most likely merger trajectory with Andromeda, making it less likely that such a merger will occur on short timescales.
2. The LMC will merge with the Milky Way in less than 10 billion years; this is inevitable, however the most-merger trajectory tends to drag the Milky Way farther from Andromeda, not closer.
3. And that when you include the effects of all four galaxies — Andromeda, the Milky Way, Triangulum, and the LMC — the probability of having a merger between Andromeda and the Milky Way in the next ~10 billion years becomes a mere 54%.

We can break that down further, if we like. In only 2% of simulations does a merger between the Milky Way and Andromeda occur on timescales of ~4 billion years. The most likely timescale for a merger is around 8 billion years from now, with uncertainties of around 1-2 billion years on that figure. And that in nearly half of simulations, the Milky Way and Andromeda won’t merge until the Universe has gotten much older: to more than double its present age.

This selection of images of external galaxies illustrates three encounter scenarios between our Milky Way and the neighboring Andromeda galaxy. In the top left panel, a wide-field DSS image showing galaxies M81 and M82 serves as an example of the Milky Way and Andromeda passing each other at large distances. The top right panel shows NGC 6786, a pair of interacting galaxies displaying the telltale signs of tidal disturbances after a close encounter. The bottom panel shows NGC 520, a cosmic train wreck as two galaxies are actively merging together. The last is likely the ultimate fate of the Milky Way-Andromeda system, but the question as to when this will occur is greatly uncertain.

The authors correctly conclude that this implies that the conventional story — of an inevitable merger between the Milky Way and Andromeda in ~4 billion years — is too naive to be true, and in fact their analysis supports a low likelihood for this scenario. (Something that can only be known when we include the gravitational effects of the LMC, which was not known until well into the Gaia era.) By adding in new data, as well as using both simulation and semi-analytical techniques, they confirm that while including the Triangulum galaxy makes a near-term merger more likely, adding in the LMC both pushes out the merger timescale and makes a collision-and-merger less likely overall within the next ~10 billion years.

However, the authors themselves end with too strong of a conclusion, claiming:

“Based on the best available data, the fate of our Galaxy is still completely open.”

This, unfortunately, is false. The timetable for the ultimate fate of our galaxy is what’s still completely open. In practically all of the models and simulations — even the most dramatic ones that include the LMC and that experience high-speed near-misses between the Milky Way and Andromeda — the Milky Way and Andromeda remain gravitationally bound to one another and to the greater Local Group, implying that a collision-and-merger will still occur.

Our entire Local Group eventually should turn into a single galaxy, Milkdromeda, but the timescale for this fate is still largely unknown. It may still be ~4 billion years from now; it’s more likely that it will be ~8 billion years from now; it’s plausible that it will be ~10+ billion years from now, and maybe even tens of billions of years into the future. However, it’s extremely unlikely that either of these galaxies will be ejected from the Local Group, and thus an eventual merger really is inevitable. The question of when this will occur, however, just got a whole lot more interesting.