CAMELS blog

What can we learn from our Universe’s climate?

Some of the most interesting phenomena in Nature involve interactions between vastly different scales. We can illustrate this with an example from our own experience: volcanic eruptions. They are comparatively small-scale, highly energetic events which have the power to influence the entire globe. The recent eruption of the Hunga Tonga-Hunga Ha’apai volcano caused human tragedy in the state of Tonga, but also caused tsunami waves in places as far away as Peru and even on the Atlantic ocean. Looking further back, there have been even more energetic volcano eruptions, for example the Mount Tambora event in 1815 which caused a cool-down in the world’s climate dubbed the “year without a summer”.


Due to the intricate mechanisms affecting an eruption, single events are impossible to accurately predict. However, by examining the geological record (for example ice cores), geologists are able to draw statistical conclusions regarding the frequency and characteristics. This is made possible by the aforementioned climatic changes that volcanic eruptions are able to cause.


In an entirely different field, Cosmology, the situation is very similar. Here, staggering energy input can be caused by active galactic nuclei (AGN) and supernovae. AGN are supermassive black holes residing at the centers of galaxies. Gas, dust and stars are drawn into the AGN and in the process release vast amounts of energy. Similarly, at the end of their lives massive stars explode in supernovae, dispersing highly energetic matter over large scales. These small-scale events are able to affect cosmological scales thanks to the incredible amount of energy released. Unfortunately, they are also extremely difficult to understand if one starts from the basic laws of physics due to the interplay of complex multi-scale processes. In numerical simulations of the cosmos, this problem is circumvented by describing AGN and supernovae with an effective model that is calibrated on astronomical observations.


In the study summarized here, we consider observations of the Universe’s “climate” – very similar to what is done for volcanoes! The two quantities that we can measure are the mean pressure and mean temperature of electrons in the observable universe. While geologists use ice cores to form statistical averages across time, cosmologists can statistically average over the Universe’s volume (which implicitly involves an average over time due to the finite speed of light). By comparing the measured pressure and temperature with the output of numerical simulations, we can pin down which simulations are best able to describe our actual Universe.


The fact that we can even measure our Universe’s climate is thanks to the cosmic microwave background (CMB). The CMB was generated at a time when the Universe was still a hot and almost homogeneous soup of plasma. Photons reaching us from this early stage contain a tremendous amount of information about the cosmos when it was still in its infancy. However, they also act as a back-light enabling us to discern cosmic structures formed much later after the CMB was released. By interacting with free electrons, the CMB photons acquire a distinct signature in their energy distribution. This allows us to infer the thermodynamic properties of the electrons the CMB photons encountered en route to our telescopes. The resulting signal is extremely weak, but by observing it long enough over large sky areas we can measure its sky average. This allows us to pin down the pressure and temperature the electrons have on average.


Such a measurement of the Universe’s mean climate has not been performed yet. Both the US space agency NASA and the European counterpart ESA have developed proposals to carry out the required observations, for which a space mission is necessary. Such a mission would be feasible within the next decade (and it would do some amazing science besides the Universe’s climate). In our study, we ask the simple question: if one (or both) of these missions were to be carried out, what would their measurement of the pressure and temperature imply for our understanding of AGN and supernovae?



In order to answer this question, the CAMELS suite of numerical simulations is the ideal tool. CAMELS simulates thousands of mini-universes, each with a different model for AGN and supernovae. We extract the mean electron pressure and temperature from each of these simulations and infer the relationship between the parameters describing AGN and supernovae and these “climatic” observables. This task is not as simple as it appears, due to the small volume each of the CAMELS simulations is run on. One could compare this to considering only small patches of the Earth’s surface: the conclusions we would draw on volcano eruptions would depend a lot on whether we happen to look at a tectonic fault line or some quiet spot. For this reason, there is substantial “noise” in the CAMELS simulations, depending on how many galaxies of what masses were formed in them. We solve this problem with a combination of two methods. First, we can use our existing understanding of the connection between the number of galaxies and the electron thermodynamics to compute rescaling factors – this is as if one would know the expected number of volcanoes in a given patch of the earth and compare it to the randomly realized one. Second, machine learning tools called neural networks allow us to interpolate smoothly through the noisy data.


Our results are striking. At the present moment, we are uncertain about the AGN and supernovae models in the numerical simulations by perhaps a factor of two. If we were to obtain the described measurements of the Universe’s climate, this uncertainty would go down to mere percents. This vastly increased knowledge would enable us to perform simulations that are much closer to reality than what is currently possible. More accurate simulations would be pivotal to many efforts in Cosmology, since uncertainty on small-scale processes such as AGN and supernovae is beginning to limit our ability to better understand the fundamental laws governing the evolution and composition of our Universe.


Further reading:

Paper: https://arxiv.org/abs/2201.01663

Video: https://www.youtube.com/watch?v=u2tEG1nLwV8&t=66s


Post author:

Leander Thiele

Graduate Student, Princeton University

Princeton, NJ, 08544, USA