pbsspacetime
patreon
Is Our Model of Dark Energy Wrong? (SCRIPT)
🕑 Added 2025-05-30 18:00:07 +0000 UTCThe biggest news in cosmology in recent years is that the mysterious universe-accelerating entity that we call dark energy may be fading away. The evidence for this is now strong enough that enormous effort is going into confirming the result. So what’s it going to take, and when are we going to know?
___
In the 90s we thought cosmology was solved: the universe is expanding and slowing down under the influence of gravity. In an effort to nail down this result, astronomers tracked the expansion history of the universe by watching the flashes of distant supernovae. They found that expansion wasn’t slowing down at all—it was speeding up. We don’t know what mysterious influence causes this acceleration, but we’ve named it dark energy. The most mainstream interpretation is that the vacuum of space has a native energy to it. Quantum mechanics predicts this to be the case, even if it can’t predict the quantity. This idea also gives us the simplest mathematical description of dark energy—the addition of a single number, the cosmological constant, lambda, to the Einstein equations. As we saw in another video, that this result is a constant energy density and resulting exponential expansion.
This simplest version of dark energy is part of the Lambda-CDM cosmological model, and historically it's done a pretty good job of describing the expansion history and the growth of structure through cosmic time. With Lambda-CDM we thought that we were back to cosmology being mostly solved. Time to tie up the loose ends once again by refining that measurement of the expansion history to verify that dark energy is indeed constant. Many thought that the Dark Energy Science Instrument would do just that—confirm Lambda-CDM. It didn’t. DESI seems to be showing that dark energy isn’t constant at all—rather, it may be decreasing in strength, which puts in doubt the whole vacuum energy thing.
We first reported this result last year, after DESI’s first data release, when the hint of a changing dark energy was just that—a hint. DESI’s second data release has now promoted this tenuous result to a confidence level just shy of formal detection. That was a few months ago now, and so this is not a news piece about the new result. Instead, this episode is really about what comes next. What measurements can and will we make next to confirm that dark energy is changing, to better measure how it’s changing, and what this can tell us about what dark energy really is?
That said, a quick overview of DESI is in order. The Dark Energy Science Instrument is an instrument, the survey conducted by that instrument, and the near-500-person collaboration behind it all. The instrument on the Mayall telescope at Kitt Peak in Arizona. It robotically places 5000 optical fibres to simultaneously capture the light of as many galaxies. That light has traveled to us for billions of years through an expanding universe, and the signature of that expansion is imprinted in the stretching or redshift of each galaxy’s light. By splitting the light of each galaxy with a spectrograph, the DESI team can measure that redshift, which encodes both the distance to each galaxy and the expansion history during that light’s journey.
These two can’t be disentangled without more information and so DESI does something else incredibly cool. Remember, we want to measure the expansion history. Perhaps the simplest way to think about it is that we want to measure how big the universe was at different times in the past. Light takes time to reach us, so light from distant galaxies carries a snapshot from an earlier time. Next we just need some sort of structure that has expanded with the universe, so we can measure its size of the universe at different times.
In the case of DESI, that something is the baryon acoustic oscillations—BAOs—which we discussed in previous episodes. These are gigantic circular patterns in the distribution of galaxies on the sky. They are the remnants of sound waves in the hot, dense plasma that filled the very early universe before the first galaxies formed. These density waves froze at a certain size when the first atoms formed, leaving a ripple pattern in the gas and dark matter that filled the universe. The galaxies that would later form from that matter still reflect the patterning, but the ripples have grown with the universe. They now give us a powerful “standard ruler” to track that expansion.
So, simplistically, we have redshift that encodes the subsequent expansion history from a given time in the past, and then we have a measure of the scale of the universe at those times. This doesn’t quite give an unambiguous expansion history by itself, but the combination does allow DESI to test the consistency of different models of that history—different cosmological models.
In the case of the DESI result, the cosmological models tested included Lambda-CDM with its constant dark energy, as well as a particular model for changing dark energy. In that model, dark energy changed over time according to an equation given in the form of a varying equation of state parameter—something we talked about in the previous video.
By itself, DESI finds that the model with changing dark energy has only a marginal improvement over LCDM. But things get much more interesting when the DESI data is combined with other studies. The first is the constraints from the cosmic background radiation, which provides the starting point—the initial conditions from which cosmic expansion and structure formation measured by DESI will evolve. The other is an independent measure of distance. DESI itself only has the galaxy redshift as a distance indicator, which makes it difficult to pin down the absolute scale of these BAO rings. So they add an independent distance measure—the same type as was first used to discover dark energy—type 1a supernovae.
These exploding white dwarf stars are called standard candles, because they explode with predictable brightnesses. That means we can figure out how far away they are, based on how faint they appear to us due to that distance. One way to think about it is that the supernovae give distances from us to the galaxies, while the baryon acoustic oscillations give distances between those galaxies. Standard candles plus standard rulers. Combined with the initial conditions from the CMB, these measurements give a much tighter constraint on their cosmological parameters. DESI claims that the combination favors a cosmology in which dark energy is actually weakening, and that this differs from the cosmological constant by model by up to 4.8 sigma for the supernovae in DESI’s own survey, but a bit less than 4.2 using other supernova surveys. This doesn’t quite reach the 5-sigma level usually required to claim a discovery, but it’s tantalizingly close.
So did DESI probably discover that dark energy isn’t constant? Not really. Besides not quite hitting 5-sigma, they also test this one version of varying dark energy, which you could argue is a little ad-hoc. When you’re fitting a curve to a bunch of points, adding extra degrees of freedom always improves the apparent fit. That could be what’s happening here. Still, it’s intriguing enough and warrants deeper exploration.
What we really want is a precision measurement of the way expansion and dark energy have changed over cosmic time, because that would allow us a) to be sure that the change is really happening, and b) to distinguish between different models of dark energy, which in turn could tell us what dark energy really is. We’ve talked about those options before, so I won’t go into it now. I want to focus on what’s needed and what’s being done to improve this measurement.
How do we improve on the DESI result? The answer is a little prosaic I'm afraid—we just need to do everything better. Let me remind you of the three things we talked about. Now this is an oversimplification, but it serves to organise the ideas. We need to:
1. Measure the starting state
2. To measure the size of the universe at different times. And 3.
To measure distances to those different times.
We’ll dispense with number 1 quickly. Measuring the starting state means measuring the initial density distribution and various energetic components of the universe from the cosmic microwave background. There are plans to do better than the Planck satellite already has, but honestly, our current CMB map is the most precise part of this whole calculation, so it’s not the limiting factor.
So straight on to number 2: how can we improve our measurement of the size of the universe? Baryon acoustic oscillations are really the gold standard, and so just doing DESI better is probably the way to go. That means we need bigger, better galaxy redshift surveys which beat down the uncertainty in various ways, including the sheer number of galaxies. We also want to push this measurement back further in time, which we can do with more sensitive instruments on bigger telescopes, or just staring at the sky for longer to collect photons from fainter galaxies.
And 3: we also want to improve our measurement of distances to those galaxies. The most obvious thing there is to improve our supernova measurements with more type-1a’s, but also by refining our calibration of these things as standard candles. This is one of the most contentious points in this whole effort because the supernova distances piggyback off a chain of distance measurements and so errors could have crept in unnoticed in various places. Major efforts are underway to weed out potential issues.
It’s also important to find independent ways
to get those distances. One of the most exciting, and a favorite of mine, is to use gravitational lensing. Gravity bend space, and so bends the path of light. Very strong gravitational fields, like those of entire galaxies, bends space so much that light from a distant source can travel to us via multiple separate paths. For example, distant quasars—feeding supermassive black holes—sometimes appear duplicated or even quadrupled around a galaxy that happens to lie directly between us and it. Now when that quasar fluctuates in brightness, all of these lensed images also fluctuate—they’re all the same quasar. But the pattern of fluctuations has an offset in time that reflects the difference in the lengths of these paths. Measuring this time delay actually gives us a measure of cosmic distance that’s completely independent of the supernova measurements. Astronomers have used lensed quasars to measure the expansion rate of the universe. But by doing this for many more such objects, we can also get the expansion history.
Now I was being a little simplistic when I separated the ideas of measuring the size of the universe at different times to measuring the distance to those times. Really, any given cosmological measurement does some of both. For example, we track the size of the universe just by looking at how far apart galaxies are in general, not just the size of the BAO rings. So this size information is also there in, for example, the supernova catalogs even without the BAO analysis. And the expansion history is coded in the redshifts, so we can calculate expansion rates from even just the lensing time delays.
But really, combining these datasets is the key. What we’re trying to do is to fit a cosmological model with many interacting or “covariant” parameters, so we need measurements that will tease apart and constrain all parameters. And on that note, there is one parameter I haven’t mentioned—the clustering of the galaxies. Over cosmic time, galaxies are thrown apart by cosmic expansion, but they also fall together due to their mutual gravitational attraction. The amount of this clustering tells you stuff about the amount of dark matter in the universe and the size of the initial density fluctuations—both important parameters in your cosmological model. So measuring this clustering is a big help in constraining that model. Galaxy redshift surveys like DESI can do this directly to some extent based on the clustering of the galaxies that they can see. But there are also more cunning approaches that can take account of all of the mass, seen or unseen.
For example, through weak lensing. The lensing I already talked about where you see multiple images of one object is called strong lensing. In weak lensing, the gravitational field of galaxies and clusters slightly warp the shapes of background objects, without producing multiple images. In the case of a weakly lensed galaxy, it will be slightly stretched in one direction. You can’t identify this effect with a single galaxy because you don’t know what shape it was originally.
Now imagine many galaxies in a particular direction that normally have random orientations relative to each other. As their light travels to u s, it passes around a particularly dense patch of the universe. The orientations of all these galaxies will be skewed at right angles to that patch and become correlated with each other. By looking at the correlations in the orientations of galaxies across the entire sky, you get a crude density map of the universe, which in turn allows us to constrain galaxy clustering.
The DESI result that made all this news is from their 2nd data release, 3 years into the 5-year survey, and with 30 out of the planned 50 million galaxies and quasars actually. If the deviation from constant dark energy is real, it seems likely that this extra signal will get them over the 5-sigma level.
But DESI isn't the only game in town nor the only thing to reduce that uncertainty. We should not confuse it with DES for example
- that's the Dark Energy Survey. DES observes a factor of 10 more galaxies than DESI, but where DESI takes spectra of known galaxies, DES takes images at different wavelengths and actually discovers galaxies. That makes DES an imaging or photometric survey, where DESI is a spectroscopic survey. DES measures photometric redshifts, which are less accurate that DESI's spectroscopic redshifts but take much less time to observe. DES also doesn't see as far as DESI, so isn't sensitive to the expansion history. Instead, DES is good at measuring the late-time expansion rate as well as the growth of structure via clustering and weak lensing measurements. It's also important for nailing down them pesky supernovae. Really, DES is complementary to DESI, giving the late-time constraints, just as the CMB gives the initial conditions.
But perhaps the greatest improvement will be from very new and upcoming surveys. We have the European Space Agency's Euclid satellite which is studying over a billion galaxies with both imaging and spectroscopy, and its most important cosmological contribution is by measuring galaxy clustering and weak lensing. This should allow us to beat down the uncertainty in the equation of state paremeters to the one-percent level, which moves us towards a precision measurement of dark energy. Euclid is in the first of its 6-year mission, and the first cosmological results are expected next year.
And the last and perhaps most exciting endevour is the Legacy Survey of Space and Time--LSST--on the Rubin Observatory. This is a photometric survey that does something very new--it will map much of the southern sky many, many times for 10 years. Besides complementing or improving on DESI, DES, and Euclid. LSST will add the dimension of time to the analysis. Perhaps the most important benefits to cosmology are measurement of many hundreds of thousands of type 1a supernovae, providing an incredible late-time anchor for other surveys. LSST will enable measurement of those time-delays in gravitationally lensed quasars, which if you remember give an expansion history independent of supernovae--but LSST will only do this
after discovering something like 20 times more lensed quasars than we currently know of, so this method should really come into its own with LSST.
We're entering the era of precision cosmology. It's really a golden era for the field, and in the next decade you can expect the cosmic expansion history and the general cosmological model to become known to a precision that seemed impossible not that long ago. With the precision comes the ability to ask what kind of physics can produce such a universe--what kind of dark energy? dark matter? starting conditions? And what does that mean for how the universe will end? We'll know it all just as soon as we've finished surveying all of space time.
Comments
Phil Drinkwater
What if we’re experiencing a huge gravitational wave from the “big bang” - and it’s reducing, because waves do?
