Progress in Computational Cosmology

We’ve had a visitor in Maynooth for the last couple of days in the form of Mathieu Schaller, who works at Leiden University in the Netherlands. Mathieu was here to work with John Regan’s group on cosmological simulations, but also gave a Theoretical Physics seminar yesterday to a general audience including some of our undergraduate students.

Mathieu’s talk was about a project called FLAMINGO – what is it with cosmologists and acronyms? – which is a suite of simulations designed to be virtual “twins” of the next generation of surveys. This suite includes the largest cosmological simulation ever run to the present time so it can simulate redshift surveys encompassing local volumes near redshift z=0 out to very distant sources at high redshift.

It was a very interesting talk which I thought I would mention here because of one thought that struck me, which is how much the field of computational cosmology has moved on since I started in the field in 1985, almost forty years ago. Not for the first time, it was a seminar that made me feel very old. I’ve been a spectator as far as this is concerned, of course, because I don’t do massive simulation work. Nevertheless these calculations have had a huge impact in the field, and play an important role in, for example, the Euclid mission. They are used both for planning survey strategies and for analyzing the result data.

Take a look at these two pictures, which I’ve chosen to illustrate the progress there has been in the field.

The simulation on the left shows the state-of-the-art when I started my PhD DPhil in 1985 from the classic “DEFW” paper by Davis Efstathiou, Frenk & White; the one on the right I took from Mathieu’s Twitter account. These do no simulate the same volume so the scale looks different, but the morphology of the cosmic web looks similar.

The most obvious change over the years has been the ability to generate colour graphics. The standard cosmological model has also evolved: the one on the right shows a model universe dominated by Cold Dark Matter with no dark energy, while the one on the right is the modern variant known as ΛCDM. The one on the left also is gravitational-only, i.e. no hydrodynamic effects arising from baryonic material., just the effect of the cold dark matter. The simulation on the right includes extensive modelling of baryonic physics. The largest gravity-only simulations that I’m aware of is the Euclid flagship simulation which produces mock galaxy catalogues like this:

The thing that struck me as an oldie, however, is the sheer scale of modern simulations. The DEFW simulations were done by moving N=323 particles around in a box in response to their mutual gravitational interactions. That’s just 32768 particles. The simulations Mathieu talked about involve N=50403 = 125,300,240,064 particles. That’s a factor of almost 4 million bigger. The Flagship simulations are about 16 times bigger than that, with about 2 trillion particles. Impressive! Moore’s Law is a wonderful thing…

2 Responses to “Progress in Computational Cosmology”

  1. […] microprocessors, personal computers and supercomputers. I had reason to refer to Moore’s Law on this blog just a couple of days […]

  2. […] The second paper is the 11th paper in Volume 6 (2023) as well as the 76th in all; this one was published last Thursday (23rd March). This is another for the folder marked Cosmology and Nongalactic Astrophysics. The title is “GLASS: Generator for Large Scale Structure” and the paper is about a new code for the simulation of cosmological observables obtainable from galaxy surveys in a realistic yet computationally inexpensive manner. The code can be downloaded here. This is an interesting approach that contrasts with the “brute force” of full numerical simulations like those I discussed a few days ago. […]

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: