A study in Nature with the participation of an Israeli researcher reveals the first detailed simulation of the universe* required three months of running on two supercomputers

Dr. Shay Ganel, a post-doctoral student at the Institute for Theory and Computation at Harvard University and the Harvard-Smithsonian Institution, says in an interview with the science website that the simulation contains initial conditions and the basic laws of nature, including the star formation process, dark matter and dark energy

Scientists have created the first realistic virtual model of the universe using a computer simulation known as Illustris. Illustris can reconstruct 13 billion years of cosmic evolution in a cube measuring 350 million light years in each direction with unprecedented resolution.

"Until now, no single simulation has been able to reproduce the universe on both a large scale and a small scale at the same time," said lead researcher Mark Vogelsberger of MIT and the Harvard-Smithsonian Institute for Astrophysics, who carried out the work with the help of researchers from several institutes including the Heidelberg Institute for Theoretical Studies in Germany. The study is published today, May 8 in the journal Nature.

In an interview with the Hidaen site, Dr. Shay Ganel, a co-author of the article, and a postdoctoral student at the Institute for Theory and Computation at Harvard University and the Harvard Smithsonian Institution, explains that previous attempts to simulate the universe encountered a lack of computing power and the complexity of the physics. As a result, these programs were limited in resolution or had to focus on a small sector of Previous simulations have also struggled to model the complex feedback between star formation, Supernova explosions and massive black holes at the centers of galaxies.

Genel, who was born and raised in Herzliya, served in Unit 8200. In his first degree he studied a combined physics and electrical engineering course at Tel Aviv University: "As a child I was interested in physics and astronomy and in the army I became interested in computer science, it became a very natural and enjoyable connection for me," he explains. Afterwards Ganel continued on a direct path to becoming a doctoral student at the Max Planck Institute near Munich in Germany, after completing his doctorate he moved to a postdoctoral position at the Institute for Theory and Calculations at Harvard University headed by Prof. Avi Leib.

"The Illustris simulation uses sophisticated programming of approximately 100 lines of code to create the evolution of the universe in high quality. The model includes reference to both normal matter and dark matter using 12 billion three-dimensional 'pixels,' or resolution elements."

The team spent five years developing the Illustris software. The actual calculations required three months of running time using 8,000 concurrent processors on two large supercomputers in Europe. If these were normal personal computer processors, it would take two thousand years to complete the calculation.

The computer simulation began about 12 million years after the Big Bang, and it continues to this day. The astronomers counted over 41 thousand galaxies in the cube of the recommended space. More importantly, Illustris produced a very close to reality mix of spiral galaxies such as the Milky Way and elliptical galaxies such as an American football. It also created large-scale structures such as galaxy clusters as well as bubbles and voids of the cosmic web. On a small scale it also reconstructed the chemistry of individual galaxies.

Intermediate: The galaxies formed by themselves
Dr. Ganel explains: "We included in the model the initial conditions that prevailed in the early universe, and we chose an age of 12 million years because this is an area that we can observe using satellites such as the Planck satellite. The radiation from that period is uniform radiation (with the exception of small disturbances created by the gravitational waves - see an article we published on the science website following the discoveries of the gravitational waves, but their effect was in the inflation of the universe a fraction of a second after the big bang, but at the age of 12 million the already inflated universe was indeed uniform AB. ). The universe was then homogeneous and it consisted only of hydrogen and helium atoms. At that time there were no stars yet. Our computer code represents the laws of physics, mainly the laws of gravity, the fact of the expanding universe (dark energy, and on that later, AB) and hydrodynamics - the flow of the gas that fills the universe."

"In addition, we modeled the process of the formation of stars and black holes from the gas, and also took into account the mutual effect of those stars and black holes back on this gas. Shortly after a star or black hole is formed, it begins to release energy and radiation into the environment of and this energy affects the formation of subsequent generations of stars And black holes, it's a self-regulating process. However, it's a very complicated process and therefore we can't solve the equations related to it using a paper and a pencil and therefore We have to use very powerful computers that do a lot of calculations and follow these processes in time."

Because light travels at a constant speed, the farther astronomers look, the farther back in time they see. A galaxy a billion light years away from us looks like it did a billion years ago. Telescopes such as Hubble can observe the early universe by looking at greater distances. However, astronomers cannot use Hubble or any other telescope, however powerful, to follow the evolution of a single galaxy over time. "Illustris is like a time machine. We can go forward and backward in time. We can stop the simulation and focus on one galaxy or a cluster of galaxies and see what is happening there" says Ganel.

"After setting the initial conditions, we ran cosmic time forward and followed the beginnings of the evolution of the gas, stars and black holes and saw it as a miracle, galaxies naturally formed in this random universe without you programming them. They were created by themselves only thanks to the initial conditions and the laws of physics." Genel explains.

In response to the question of the science site, what is special about this simulation compared to previous simulations?

The simulation we are publishing today in Nature has two main innovations: the first innovation is the inclusion of many processes related to the formation of stars and black holes that previous simulations did not include. This allowed the simulation to develop a large population of tens of thousands of galaxies. We have shown that these galaxies are similar to real galaxies that other scientists see in a telescope and this is due to the accuracy in modeling the laws of physics.
The second thing that is special about the simulation is the scope. This is the largest simulation of the universe ever understood. The size advantage allows us to continuously monitor a large volume of the universe, and examine the internal structure of each galaxy individually. The large number of galaxies allows us to statistically compare the population of galaxies we see with the one that exists in reality and compare whether the simulation actually simulates reality. To understand if our models are correct, we need a large number of identified galaxies.

The scientist: How do we know what happened to the stars we see today? After all, in fact, we cannot see the entire universe in its current state, but only the environment close to us, and as we move further away, we also see the universe in a different time?

Ganel: "In the simulation we follow the entire history of the universe. We compare each period in our simulation to the corresponding period as seen from the observations. In telescopes we see objects in different stages of development in different periods, for example galaxies at different ages of the universe, but we cannot follow the development of the same object over time. The work of the observers is a kind of detective work during which they are required to decipher how one type of galaxy evolved into a type Another galaxy but we can't see how it happens. A simulation like ours makes life easier because we can track a specific galaxy and see how it changes over time. This can help observers make the same inferences for the original universe."

How did you treat dark matter and dark energy?

We assume in our simulation the simplest model for dark energy and dark matter. Regarding dark energy, we assume that it is uniformly distributed throughout the universe and its only effect is in determining the rate of expansion of the universe. Dark matter behaves differently, so we implemented it as a dynamic component within the simulation. Since the dark matter reacts to the force of gravity, its distribution in space is different - there are areas that have a lot of dark matter, and other areas are almost completely empty. We follow the dark matter in the simulation under the assumption that it responds only to gravity and no other force. The fact that when we use these assumptions we get a good fit to the observations strengthens the likelihood that the model correctly describes nature. This is not proof because another model can always come along that will also be able to explain the same observations, but at the moment there is no such model and we showed for the first time that under the assumptions I described it is possible to obtain not only the distribution of galaxies on very large scales - (the Cosmic Web) - a scale of hundreds of millions of light years, but also show that the model can reproduce the properties of individual galaxies on a scale of thousands of light years - a much smaller and detailed scale."

Did the galaxies merge with you, as in many astronomical observations?

Indeed we see mergers of galaxies. In one of the films we published, you see an example of the creation of a very massive elliptical galaxy as a result of the merger of six separate galaxies. We have many cases in this simulation and one of the students is trying to quantify the rate of galaxy mergers as a function of cosmological time.

Ganel is currently in Israel on a working visit, because he is conducting joint research with two Israeli researchers - Prof. Amiel Sternberg from Tel Aviv University (whose Ganel was a research assistant during his bachelor's degree), and Prof. Avishai Dekel from the Hebrew University.

In the video you can see some examples of the results in different layers of the simulation (such as the density of the dark matter, the temperature of the gas or the chemistry). Additional short videos (including the video describing the merging of galaxies) and images can be found on the simulation site.