Tyler O'Neal, Staff Editor ASTRONOMY

Penn State astrophysicist builds dark matter map that reveals hidden bridges between galaxies

A new map of dark matter in the local universe reveals several previously undiscovered filamentary structures connecting galaxies. The map, developed using machine learning by an international team including a Penn State astrophysicist, could enable studies about the nature of dark matter as well as about the history and future of our local universe.

Dark matter is an elusive substance that makes up 80% of the universe. It also provides the skeleton for what cosmologists call the cosmic web, the large-scale structure of the universe that, due to its gravitational influence, dictates the motion of galaxies and other cosmic material. However, the distribution of local dark matter is currently unknown because it cannot be measured directly. Researchers must instead infer its distribution based on its gravitational influence on other objects in the universe, like galaxies. An international team of researchers has produced a map of the dark matter within the local universe, using a model to infer its location due to its gravitational influence on galaxies (black dots). These density maps--each a cross section in different dimensions--reproduce known, prominent features of the universe (red) and also reveal smaller filamentary features (yellow) that act as hidden bridges between galaxies. The X denotes the Milky Way galaxy and arrows denote the motion of the local universe due to gravity.

"Ironically, it's easier to study the distribution of dark matter much further away because it reflects the very distant past, which is much less complex," said Donghui Jeong, associate professor of astronomy and astrophysics at Penn State and a corresponding author of the study. "Over time, as the large-scale structure of the universe has grown, the complexity of the universe has increased, so it is inherently harder to make measurements about dark matter locally."

Previous attempts to map the cosmic web started with a model of the early universe and then simulated the evolution of the model over billions of years. However, this method is computationally intensive and so far has not been able to produce results detailed enough to see the local universe. In the new study, the researchers took a completely different approach, using machine learning to build a model that uses information about the distribution and motion of galaxies to predict the distribution of dark matter.

The researchers built and trained their model using a large set of galaxy simulations, called Illustris-TNG, which includes galaxies, gasses, other visible matter, as well as dark matter. The team specifically selected simulated galaxies comparable to those in the Milky Way and ultimately identified which properties of galaxies are needed to predict the dark matter distribution.

"When given certain information, the model can essentially fill in the gaps based on what it has looked at before," said Jeong. "The map from our models doesn't perfectly fit the simulation data, but we can still reconstruct very detailed structures. We found that including the motion of galaxies--their radial peculiar velocities--in addition to their distribution drastically enhanced the quality of the map and allowed us to see these details."

The research team then applied their model to real data from the local universe from the Cosmicflow-3 galaxy catalog. The catalog contains comprehensive data about the distribution and movement of more than 17 thousand galaxies in the vicinity of the Milky Way--within 200 megaparsecs. The resulting map of the local cosmic web is published in a paper appearing online on May 26 in The Astrophysical Journal.

The map successively reproduced known prominent structures in the local universe, including the "local sheet"--a region of space containing the Milky Way, nearby galaxies in the "local group," and galaxies in the Virgo cluster--and the "local void"--a relatively empty region of space next to the local group. Additionally, it identified several new structures that require further investigation, including smaller filamentary structures that connect galaxies.

"Having a local map of the cosmic web opens up a new chapter of the cosmological study," said Jeong. "We can study how the distribution of dark matter relates to other emission data, which will help us understand the nature of dark matter. And we can study these filamentary structures directly, these hidden bridges between galaxies."

For example, it has been suggested that the Milky Way and Andromeda galaxies may be slowly moving toward each other, but whether they may collide in many billions of years remains unclear. Studying the dark matter filaments connecting the two galaxies could provide important insights into their future.

"Because dark matter dominates the dynamics of the universe, it basically determines our fate," said Jeong. "So we can ask a [super]computer to evolve the map for billions of years to see what will happen in the local universe. And we can evolve the model back in time to understand the history of our cosmic neighborhood."

The researchers believe they can improve the accuracy of their map by adding more galaxies. Planned astronomical surveys, for example using the James Web Space Telescope, could allow them to add faint or small galaxies that have yet to be observed and galaxies that are further away.

KU wins grant to use big data from IceCube for studying neutrinos

Tyler O'Neal, Staff Editor ASTRONOMY

Neutrinos are almost intangible subatomic particles that hardly interact with matter in the universe. Billions of neutrinos are shooting through your body as you read this sentence, and you don't even notice. Explosive deep-space events like gamma-ray bursts, merging black holes, neutron stars, and even the Big Bang all emitted high-energy neutrinos that shower the Earth, holding secrets to the nature and history of the cosmos.

Now, a $1.2 million award from the National Science Foundation's EPSCoR program will create a new faculty position at the University of Kansas within the next year, support postdoctoral researchers and graduate students, and fund work to better detect and analyze neutrinos at the IceCube Neutrino Observatory (IceCube) in Antarctica and a new observatory, dubbed RNO-G, in Greenland. A recent grant to scientists the University of Kansas will enable work at the IceCube Neutrino Observatory at the South Pole and support new faculty. CREDIT Benjamin Eberhardt, IceCubeNSF.

The work at KU is part of a larger $6 million collaboration among multiple universities to develop large-scale neutrino detection instruments and harness big data.

"We're going to be instituting a project this summer in Greenland that will pilot a lot of the hardware we want to use for the next phase of the IceCube experiment, so-called IceCube-Gen2, at the South Pole," said co-principal investigator David Besson, professor of physics & astronomy at KU. "I'll be going to Greenland for six weeks over the summer as part of that effort."

Besson said the study of neutrinos will push forward understanding of the cosmos beyond the capability of the telescope, the standard instrument of astronomy for 400 years.

"The universe isn't entirely transparent to light, especially very high-energy electromagnetic radiation, like gamma rays," he said. "The universe is filled with this sort of fog of faint light leftover from a time close to the Big Bang called the cosmic microwave background. Because that faint fog permeates the entire universe, very-high-energy gamma rays produced at the edge of the universe can't make it to our terrestrial observatories -- they're absorbed by that fog. If you wanted to trace some interesting astrophysical object, not by the gamma rays that are produced, but by the protons that it might produce, the protons will also be absorbed on that fog. So, how do you observe the far reaches of the universe at these energy levels? The only particle that's capable of reaching us is the neutrino because it's a very penetrating particle."

Because neutrinos are so elusive, scientists at IceCube in Antarctica and IceCube2 in Greenland transform miles and miles of ice into huge neutrino detectors.

"Because neutrinos are so penetrating you have to have a huge target to stop a neutrino," he said. "It very rarely interacts with matter, and that's how it actually makes it through this fog. So, we're using the Antarctic ice sheet as our very thick target for neutrino interactions. The ultimate goal is that you want to do all the science that you can currently do with a standard telescope. But you want to do it with a 'neutrino telescope.'"

Besson's work will involve improving the calibration of these massive instruments to account for deformations in the ice, radio-frequency interference, and other noise that make it harder to detect neutrino signals. 

"We're working to develop a technique that's going to learn how to classify the different types of background and then specifically target those backgrounds for suppression," Besson said. "We will develop background 'templates' and compare each event against that template -- everything else that isn't background will pass through as something potentially interesting. We already have examples of what this sort of noise would be like. It can be fairly mundane. The South Pole is an area where there's not just IceCube but there are also other experiments -- so there are electronics and those can produce noise. People driving on snowmobiles will produce noise. With very high wind velocities it's possible for the wind to ionize the surface of the snow and then you get static electric effects and those will produce radio frequency."

With massive amounts of data being collected by IceCube and IceCube-Gen2, Besson and his collaborators will work on their "ability to manage big data through advanced data science techniques in data throughput, calibration, simulation, analysis, modeling, and hypothesis testing."

"You want to be able to push the detection threshold as absolutely low as you can, and that means that you're going to be collecting huge reams of data," Besson said. "A lot of that data is, of course, just noise. It's basically just junk. But you want to have algorithms capable of using sophisticated machine learning techniques that winnow and throw away the gazillions of background events and pick up the one interesting event on the fly. You want to feed that back into your data collection. So, your data collection is at the same time getting more intelligent about which events it's writing to disk."

The new EPSCoR grant is led by researchers at the South Dakota School of Mines and Technology. Other institutions from Alabama, Alaska, Delaware, and Nebraska are participating in the project.

Saarland's Jan Reineke wins 2.5 million euros to help solve fundamental problems in the interaction of hardware, software

Tyler O'Neal, Staff Editor ASTRONOMY

Many safety-critical areas of our lives are being controlled by computer systems: from airbag controls in cars and landing gear on airplanes to essential infrastructure such as energy supply and telecommunications. But are these systems reliable? Computer science professor Jan Reineke of Saarland University thinks not - because a crucial component of today's computer systems renders the development of safe and secure IT applications impossible at a fundamental level.

To change this and to improve the interaction between hardware and software, the computer scientist is now being funded through an ERC Advanced Grant with around 2.5 million euros over five years.

Jan Reineke's current project addresses the interaction between hardware and software. For software to be executed, it must first be translated from the higher-level programming language in which it was written by a programmer into a language that the hardware understands. It is this machine language, the so-called instruction set architecture, that Jan Reineke is focusing on in his research. Computer Science Professor of Saarland University CREDIT Oliver Dietze

The instruction set architecture is the interface between hardware and software. It can be thought of as a contract that regulates the interaction of the two components. Accordingly, another name for it is Hardware-Software Contracts. "These 'contracts' provide guarantees about how the machine code is to be executed by the underlying hardware, the microarchitecture. But they also specify what the machine code must look like for the hardware to be able to understand it," the computer scientist adds. According to Reineke, current instruction set architectures have blind spots in two crucial areas: "First, they provide no guarantees whatsoever about how long software takes to execute. Second, they lack basic security guarantees against malicious attacks," says Reineke.

The resulting problems have far-reaching implications: The lack of time guarantees is the reason why time-critical systems such as an airbag control system depend on complex but fundamentally inadequate computing models. Lack of security guarantees on the hardware side led, among other things, to the well-known security vulnerabilities 'Spectre' and 'Meltdown' at the beginning of 2018, which to this day affect almost all modern processors in systems.

"The goal now is to rethink from the ground up how this interface between software and hardware should be defined so that it is both efficient and secure," says computer scientist Reineke. Doing this, many other aspects have to be considered. On the one hand, the instruction set architecture must leave enough creative freedom for hardware developers; on the other hand, it should make the development of safe and secure software as simple as possible.

The project, entitled "Abstractions for Safe and Secure Hardware-Software Systems", is funded by an "Advanced Grant" from the European Research Council (ERC) with around 2.5 million euros over five years. ERC Advanced Grants are among the most prestigious research awards worldwide. A total of 2678 projects were submitted for the current funding period, of which 209 were approved (about 8%).

The research project described is the sixth ERC Advanced Grant and the 24th grant overall from the European Research Council, which has been granted for a project at the Saarland Informatics Campus. There, Jan Reineke is based in the Faculty of Mathematics and Computer Science at Saarland University.