An Artificial Intelligence Just Found 56 New Gravitational Lenses

Gravitational lenses are an important tool for astronomers seeking to study the most distant objects in the Universe. This technique involves using a massive cluster of matter (usually a galaxy or cluster) between a distant light source and an observer to better see light coming from that source. In an effect that was predicted by Einstein’s Theory of General Relativity, this allows astronomers to see objects that might otherwise be obscured.

Recently, a group of European astronomers developed a method for finding gravitational lenses in enormous piles of data. Using the same artificial intelligence algorithms that Google, Facebook and Tesla have used for their purposes, they were able to find 56 new gravitational lensing candidates from a massive astronomical survey. This method could eliminate the need for astronomers to conduct visual inspections of astronomical images.

The study which describes their research, titled “Finding strong gravitational lenses in the Kilo Degree Survey with Convolutional Neural Networks“, recently appeared in the Monthly Notices of the Royal Astronomical Society. Led by Carlo Enrico Petrillo of the Kapteyn Astronomical Institute, the team also included members of the National Institute for Astrophysics (INAF), the Argelander-Institute for Astronomy (AIfA) and the University of Naples.

The notable gravitational lens known as the Cosmic Horseshoe is found in Leo. Credit: NASA/ESA/Hubble

While useful to astronomers, gravitational lenses are a pain to find. Ordinarily, this would consist of astronomers sorting through thousands of images snapped by telescopes and observatories. While academic institutions are able to rely on amateur astronomers and citizen astronomers like never before, there is imply no way to keep up with millions of images that are being regularly captured by instruments around the world.

To address this, Dr. Petrillo and his colleagues turned to what are known as “Convulutional Neural Networks” (CNN), a type of machine-learning algorithm that mines data for specific patterns. While Google used these same neural networks to win a match of Go against the world champion, Facebook uses them to recognize things in images posted on its site, and Tesla has been using them to develop self-driving cars.

As Petrillo explained in a recent press article from the Netherlands Research School for Astronomy:

“This is the first time a convolutional neural network has been used to find peculiar objects in an astronomical survey. I think it will become the norm since future astronomical surveys will produce an enormous quantity of data which will be necessary to inspect. We don’t have enough astronomers to cope with this.”

The team then applied these neural networks to data derived from the Kilo-Degree Survey (KiDS). This project relies on the VLT Survey Telescope (VST) at the ESO’s Paranal Observatory in Chile to map 1500 square degrees of the southern night sky. This data set consisted of 21,789 color images collected by the VST’s OmegaCAM, a multiband instrument developed by a consortium of European scientist in conjunction with the ESO.

A sample of the handmade photos of gravitational lenses that the astronomers used to train their neural network. Credit: Enrico Petrillo/Rijksuniversiteit Groningen

These images all contained examples of Luminous Red Galaxies (LRGs), three of which wee known to be gravitational lenses. Initially, the neural network found 761 gravitational lens candidates within this sample. After inspecting these candidates visually, the team was able to narrow the list down to 56 lenses. These still need to be confirmed by space telescopes in the future, but the results were quite positive.

As they indicate in their study, such a neural network, when applied to larger data sets, could reveal hundreds or even thousands of new lenses:

“A conservative estimate based on our results shows that with our proposed method it should be possible to find ?100 massive LRG-galaxy lenses at z ~> 0.4 in KiDS when completed. In the most optimistic scenario this number can grow considerably (to maximally ? 2400 lenses), when widening the colour-magnitude selection and training the CNN to recognize smaller image-separation lens systems.”

In addition, the neural network rediscovered two of the known lenses in the data set, but missed the third one. However, this was due to the fact that this lens was particularly small and the neural network was not trained to detect lenses of this size. In the future, the researchers hope to correct for this by training their neural network to notice smaller lenses and rejects false positives.

But of course, the ultimate goal here is to remove the need for visual inspection entirely. In so doing, astronomers would be freed up from having to do grunt work, and could dedicate more time towards the process of discovery. In much the same way, machine learning algorithms could be used to search through astronomical data for signals of gravitational waves and exoplanets.

Much like how other industries are seeking to make sense out of terabytes of consumer or other types of “big data”, the field astrophysics and cosmology could come to rely on artificial intelligence to find the patterns in a Universe of raw data. And the payoff is likely to be nothing less than an accelerated process of discovery.

Further Reading: Netherlands Research School for Astronomy , MNRAS

 

ESO Survey Shows Dark Matter to be Pretty “Smooth”

Dark Matter has been something of a mystery ever since it was first proposed. In addition to trying to find some direct evidence of its existence, scientists have also spent the past few decades developing theoretical models to explain how it works. In recent years, the popular conception has been that Dark Matter is “cold”, and distributed in clumps throughout the Universe, an observation supported by the Planck mission data.

However, a new study produced by an international team of researchers paints a different picture. Using data from the Kilo Degree Survey (KiDS), these researchers studied how the light coming from millions of distant galaxies was affected by the gravitational influence of matter on the largest of scales. What they found was that Dark Matter appears to more smoothly distributed throughout space than previously thought.

For the past five years, the KiDS survey has been using the VLT Survey Telescope (VST) – the largest telescope at the ESO’s La Silla Paranal Observatory in Chile – to survey 1500 square degrees of the southern night sky. This volume of space has been monitored in four bands (UV, IR, green and red) using weak gravitational lensing and photometric redshift measurements.

All-sky survey data from ESA's Planck space telescope. Credit: ESA
All-sky survey data from ESA’s Planck space telescope. Credit: ESA

Consistent with Einstein’s Theory of General Relativity, gravitational lensing involves studying how the gravitational field of a massive object will bend light. Meanwhile, redshift attempts to gauge the speed at which other galaxies are moving away from ours by measuring the extent to which their light is shifted towards the red end of the spectrum (i.e. its wavelength becomes longer the faster the source is moving away).

Gravitational lensing is especially useful when it comes to determining how the Universe came to be. Our current cosmological model, known as the Lambda Cold Dark Matter (Lambda CDM) model, states that Dark Energy is responsible for the late-time acceleration in the expansion of the Universe, and that Dark Matter is made up of massive particles that are responsible for cosmological structure formation.

Using a slight variation on this technique known as cosmic sheer, the research team studied light from distant galaxies to determine how it is warped by the presence of the largest structures in the Universe (such as superclusters and filaments). As Dr. Hendrik Hildebrandt – an astronomer from the Argelander Institute for Astronomy (AIfA) and the lead author of the paper – told Universe Today via email:

“Usually one thinks of one big mass like a galaxy cluster that causes this light deflection. But there is also matter all throughout the Universe. The light from distant galaxies gets continuously deflected by this so-called large-scale structure. This results in galaxies that are close on the sky to be “pointing” in the same direction. It’s a tiny effect but it can be measured with statistical methods from large samples of galaxies.When we have measured how strongly galaxies are “pointing” in the same direction we can infer from this the statistical properties of the large-scale structure, e.g. the mean matter density and how strongly the matter is clumped/clustered.”

Beautiful image of sprites at La Silla Observatory, captured by ESO Photo Ambassador Petr Horálek. Sprites are extremely rare atmosphere phemomena caused by irregularities in the ionosphere, high above storm clouds, at altitudes of about 80 kilometres. Typically seen as groups of red-orange flashes, they are triggered by positive cloud-to-ground lightning, which is rarer and more powerful than its negative counterpart, as the lightning discharge originates from the upper regions of the cloud, further from the ground. In a short burst, the sprite extends rapidly downwards, creating dangling red tendrils before disappearing. The sprite pictured here was most likely over 500 kilometres away (compare with a satellite image showing the storm over Argentina), spanned a height of up to 80 kilometres and lasted only a fraction of a second. Links: Midsummer Night Brings Sprites — Rare phenomenon caught on camera at La Silla Red Sprites at La Silla Observatory Sprites at Paranal Observatory
A rare phenomena known as “sprites” being seen above the La Silla Observatory in Chile,  Credit: ESO/Petr Horálek

Using this technique, the research team conducted an analysis of 450 square degrees of KiDS data, which corresponds to about 1% of the entire sky. Within this volume of space, the observed how the light coming from about 15 million galaxies interacted with all the matter that lies between them and Earth.

Combining the extremely sharp images obtained by VST with advanced computer software, the team was able to carry out one of the most precise measurements ever made of cosmic shear. Interestingly enough, the results were not consistent with those produced by the ESA’s Planck mission, which has been the most comprehensive mapper of the Universe to date.

The Planck mission has provided some wonderfully detailed and accurate information about the Cosmic Microwave Background (CMB). This has helped astronomers to map the early Universe, as well as develop theories of how matter was distributed during this period. As Hildebrandt explained:

“Planck measures many cosmological parameters with exquisite precision from the temperature fluctuations of the cosmic microwave background, i.e. physical processes that happened 400,000 years after the Big Bang. Two of those parameters are the mean matter density of the Universe and a measure of how strongly this matter is clumped. With cosmic shear, we also measure these two parameters but a much later cosmic times (a few billion years ago or ~10 billion years after the Big Bang), i.e. in our more recent past.”

This model assumes the cosmological principle. The LCDM universe is homogeneous and isotropic. Time dilation and redshift z are attributed to a Doppler-like shift in electromagnetic radiation as it travels across expanding space. This model assumes a nearly "flat" spatial geometry. Light traveling in this expanding model moves along null geodesics. Light waves are 'stretched' by the expansion of space as a function of time. The expansion is accelerating due to a vacuum energy or dark energy inherent in empty space. Approximately 73% of the energy density of the present universe is estimated to be dark energy. In addition, a dark matter component is currently estimated to constitute about 23% of the mass-energy density of the universe. The 5% remainder comprises all the matter and energy observed as subatomic particles, chemical elements and electromagnetic radiation; the material of which gas, dust, rocks, planets, stars, galaxies, etc., are made. This model includes a single originating big bang event, or initial singularity, which constitutes an abrupt appearance of expanding space containing radiation. This event was immediately followed by an exponential expansion of space (inflation).
The LCDM cosmological model assumes the existence of Dark Matter and Dark Energy, and that both played an active role in the formation of the Universe. Credit: Wikipedia Commons/Alex Mittelmann

However, Hildebrandt and his team found values for these parameters that were significantly lower than those found by Planck. Basically, their cosmic shear results suggest that there is less matter in the Universe and that it is less clustered than what the Planck results predicted. These results are likely to have an impact on cosmological studies and theoretical physics in the coming years.

As it stands, Dark Matter remains undetectable using standard methods. Like black holes, its existence can only be inferred from the observable gravitational effects it has on visible matter. In this case, its presence and fundamental nature are measured by how it has affected the evolution of the Universe over the past 13.8 billion years. But since the results appear to be conflicting, astronomers may now have to reconsider some of their previously held notions.

“There are several options: because we do not understand the dominant ingredients of the Universe (dark matter and dark energy) we can play with the properties of both,” said Hildebrandt. “For example, different forms of dark energy (more complex than the simplest possibility, which is Einstein’s “cosmological constant”) could explain our measurements. Another exciting possibility is that this is a sign that the laws of gravity on the scale of the Universe are different from General Relativity. All we can say for now is that something appears to be not quite right!”

Further Reading: ESO, arXiv

New VLT Survey Telescope Opens Wide Eyes to the Universe

[/caption]

There’s a new telescope at the Paranal Observatory in Chile and what big eyes it has! The VLT Survey Telescope (VST) is a wide-field survey telescope with a field of view twice as broad as the full Moon, enabling new, spectacular views of the cosmos. It is the largest telescope in the world designed to exclusively survey the sky in visible light. Over the next few years the VST and its camera OmegaCAM will make several very detailed surveys of the southern sky.

The first image released from these new eyes on the Universe is a spectacular view star-forming region Messier 17, also known as the Omega Nebula or the Swan Nebula, shown above. The VST field of view is so large that the entire nebula, including its fainter outer parts, is captured — and retains its superb sharpness across the entire image.

This new image may be the best portrait of the globular star cluster Omega Centauri ever made. Omega Centauri, in the constellation of Centaurus (The Centaur), is the largest globular cluster in the sky, but the very wide field of view of VST and its powerful camera OmegaCAM can encompass even the faint outer regions of this spectacular object. Credit: ESO/INAF-VST/OmegaCAM. Acknowledgement: A. Grado/INAF-Capodimonte Observatory

The second image is the globular star cluster Omega Centauri. This is the largest globular cluster in the sky, but the very wide field of view of VST and OmegaCAM allows even the faint outer regions to be seen clearly. This view includes about 300,000 stars.

Here’s a look at the new telescope:

The VLT Survey Telescope (VST) is the latest telescope to be added to ESO’s Paranal Observatory in the Atacama Desert of northern Chile. Credit: ESO/G. Lombardi

Below is a timelapse sequences of the VST enclosure at night:

Loading player…

For more info and images see this ESO webpage.