Applying Machine Learning To The Universe’s Mysteries

Computers can beat chess champions, simulate star explosions, and forecast global climate. We are even teaching them to be infallible problem-solvers and fast learners.

And now, physicists at the Department of Energy’s Lawrence Berkeley National Laboratory (Berkeley Lab) and their collaborators have demonstrated that computers are ready to tackle the universe’s greatest mysteries. The team fed thousands of images from simulated high-energy particle collisions to train computer networks to identify important features.

The researchers programmed powerful arrays known as neural networks to serve as a sort of hivelike digital brain in analyzing and interpreting the images of the simulated particle debris left over from the collisions. During this test run the researchers found that the neural networks had up to a 95 percent success rate in recognizing important features in a sampling of about 18,000 images.

The study was published Jan. 15 in the journal Nature Communications.

The next step will be to apply the same machine learning process to actual experimental data.

Powerful machine learning algorithms allow these networks to improve in their analysis as they process more images. The underlying technology is used in facial recognition and other types of image-based object recognition applications.

The images used in this study — relevant to particle-collider nuclear physics experiments at Brookhaven National Laboratory’s Relativistic Heavy Ion Collider and CERN’s Large Hadron Collider — recreate the conditions of a subatomic particle “soup,” which is a superhot fluid state known as the quark-gluon plasma believed to exist just millionths of a second after the birth of the universe. Berkeley Lab physicists participate in experiments at both of these sites.

“We are trying to learn about the most important properties of the quark-gluon plasma,” said Xin-Nian Wang, a nuclear physicist in the Nuclear Science Division at Berkeley Lab who is a member of the team. Some of these properties are so short-lived and occur at such tiny scales that they remain shrouded in mystery.

In experiments, nuclear physicists use particle colliders to smash together heavy nuclei, like gold or lead atoms that are stripped of electrons. These collisions are believed to liberate particles inside the atoms’ nuclei, forming a fleeting, subatomic-scale fireball that breaks down even protons and neutrons into a free-floating form of their typically bound-up building blocks: quarks and gluons.

Researchers hope that by learning the precise conditions under which this quark-gluon plasma forms, such as how much energy is packed in, and its temperature and pressure as it transitions into a fluid state, they will gain new insights about its component particles of matter and their properties, and about the universe’s formative stages.

But exacting measurements of these properties — the so-called “equation of state” involved as matter changes from one phase to another in these collisions — have proven challenging. The initial conditions in the experiments can influence the outcome, so it’s challenging to extract equation-of-state measurements that are independent of these conditions.

“In the nuclear physics community, the holy grail is to see phase transitions in these high-energy interactions, and then determine the equation of state from the experimental data,” Wang said. “This is the most important property of the quark-gluon plasma we have yet to learn from experiments.”

Researchers also seek insight about the fundamental forces that govern the interactions between quarks and gluons, what physicists refer to as quantum chromodynamics.

Long-Gang Pang, the lead author of the latest study and a Berkeley Lab-affiliated postdoctoral researcher at UC Berkeley, said that in 2016, while he was a postdoctoral fellow at the Frankfurt Institute for Advanced Studies, he became interested in the potential for artificial intelligence (AI) to help solve challenging science problems.

He saw that one form of AI, known as a deep convolutional neural network — with architecture inspired by the image-handling processes in animal brains — appeared to be a good fit for analyzing science-related images.

“These networks can recognize patterns and evaluate board positions and selected movements in the game of Go,” Pang said. “We thought, ‘If we have some visual scientific data, maybe we can get an abstract concept or valuable physical information from this.'”

Wang added, “With this type of machine learning, we are trying to identify a certain pattern or correlation of patterns that is a unique signature of the equation of state.” So after training, the network can pinpoint on its own the portions of and correlations in an image, if any exist, that are most relevant to the problem scientists are trying to solve.

Accumulation of data needed for the analysis can be very computationally intensive, Pang said, and in some cases it took about a full day of computing time to create just one image. When researchers employed an array of GPUs that work in parallel — GPUs are graphics processing units that were first created to enhance video game effects and have since exploded into a variety of uses — they cut that time down to about 20 minutes per image.

They used computing resources at Berkeley Lab’s National Energy Research Scientific Computing Center (NERSC) in their study, with most of the computing work focused at GPU clusters at GSI in Germany and Central China Normal University in China.

A benefit of using sophisticated neural networks, the researchers noted, is that they can identify features that weren’t even sought in the initial experiment, like finding a needle in a haystack when you weren’t even looking for it. And they can extract useful details even from fuzzy images.

“Even if you have low resolution, you can still get some important information,” Pang said.

Discussions are already underway to apply the machine learning tools to data from actual heavy-ion collision experiments, and the simulated results should be helpful in training neural networks to interpret the real data.

“There will be many applications for this in high-energy particle physics,” Wang said, beyond particle-collider experiments.

Team Sheds Light On The Mysteries Of Saturn’s Rings

A Skoltech-led team of international scientists has developed a mathematical model that makes sense of one of the great mysteries of Saturn’s rings.

The rings of Saturn are formations of dust and ice that swirl along Saturn’s equatorial plane. They were first spotted in 1610 by Galileo Galilei. At the time, he figured they were part of the planet itself. In-depth study of these unique formations only became possible in the 20th century in the era of space exploration. And many mysteries related to their formation and processes remain unanswered.

Most of the planets in the solar system have remain unchanged for millions of years, but Saturn’s rings exhibit extreme instability. This is particularly true of the F-ring, one of its outermost rings. The F-ring has been known to change over the course of days or even hours. For example, spacecraft have detected clumps of matter in the F-ring that suddenly disappear.

While it is not yet known what causes this erratic behavior, it is known that the processes of particle aggregation (fusion) and fragmentation (decay) play a role. In planetary rings, such as those of Saturn, these processes occur on a particularly large scale, necessitating a consistent balance between fragmentation and aggregation.

Scientists from universities in Russia (Skoltech, Moscow State University, the Institute of Numerical Mathematics of the Russian Academy of Sciences), the United Kingdom (the University of Leicester), and the United States (Boston University) teamed up to shed light on the mysterious F-ring using a mathematical model that had previously been used to study Saturn’s more stable elements.

Using the model, the Skoltech-led team demonstrated the possibility of ceaseless periodic oscillatory regimes for particle fusion and fragmentation within Saturn’s rings. In other words, they have identified a mechanism that could cause the formation of clots in a planetary ring such as the F-ring. Their findings satisfy the law of mass conservation, which holds that mass cannot be created or destroyed in chemical reactions.

The results of the study have been published is the scientific journal Physical Review Letters.

“We spent nearly three years working on this before presenting our results for publication. This is because it took us a great deal of time to work through our doubts and the validation stages. In the end, we found an example of a never-ending periodic oscillatory solution from the mathematical model of the aggregation and fragmentation processes of Saturn’s rings. Furthermore, we determined that these periodic regimes lead to a stable limiting cycle, which is very surprising for a model under the mass-conservation law, where stable equilibrium solutions are expected to be found. As our main results come from numerical simulations, we hope to attract interest from the mathematical community for the elaboration of rigorous analytical proofs of the presented phenomena. Anyway, we infer that those results may shed some new light on the phenomena of periodic clumps arising in the F-ring of Saturn,” said lead author and Skoltech research scientist Sergey Matvee

Astronomers Produce First Detailed Images Of Surface Of Giant Star

An international team of astronomers has produced the first detailed images of the surface of a giant star outside our solar system, revealing a nearly circular, dust-free atmosphere with complex areas of moving material, known as convection cells or granules, according to a recent study.

The giant star, named π1Gruis, is one of the stars in the constellation Grus (Latin for the crane, a type of bird), which can be observed in the southern hemisphere. An evolved star in the last major phase of life, π1Gruis is 350 times larger than the Sun and resembles what our Sun will become at the end of its life in five billion years. Studying this star gives scientists insight about the future activity, characteristics and appearance of the Sun.

Convection, the transfer of heat due to the bulk movement of molecules within gases and liquids, plays a major role in astrophysical processes, such as energy transport, pulsation and winds. The Sun has about two million convective cells that are typically 2,000 kilometers across, but theorists believe giant and supergiant stars should only have a few large convective cells because of their low surface gravity. Determining the convection properties of most evolved and supergiant stars, such as the size of granules, has been challenging because their surfaces are frequently obscured by dust.

In this study, the researchers discovered the surface of the giant star π1Gruis had a complex convective pattern and the typical granule measured 1.2 x 10^11 meters horizontally or 27 percent of the diameter of the star. The findings are published in the journal Nature.

“This is the first time that we have such a giant star that is unambiguously imaged with that level of details,” said Dr. Fabien Baron, assistant professor in the Department of Physics and Astronomy at Georgia State University. “The reason is there’s a limit to the details we can see based on the size of the telescope used for the observations. For this paper, we used an interferometer. The light from several telescopes is combined to overcome the limit of each telescope, thus achieving a resolution equivalent to that of a much larger telescope.”

The star π1Gruis was observed with the PIONIER instrument, which has four combined telescopes, in Chile in September 2014. Baron, who specializes in making images, used interferometric data, image reconstruction software and algorithms to compose images of the star’s surface. Interferometry is relatively new to astronomy, and Georgia State’s Center for High Angular Resolution Astronomy array was the first facility to use interferometry to image a star similar to the Sun in 2007.

This study was also the first to confirm theories about the characteristics of granules on giant stars.

“These images are important because the size and number of granules on the surface actually fit very well with models that predict what we should be seeing,” Baron said. “That tells us that our models of stars are not far from reality. We’re probably on the right track to understand these kinds of stars.”

The detailed images also showed different colors on the star’s surface, which correspond to varying temperatures. A star doesn’t have the same surface temperature throughout, and its surface provides our only clues to understand its internals. As temperatures rise and fall, the hotter, more fluid areas become brighter colors (such as white) and the cooler, more dense areas become darker colors (such as red).

In the future, the researchers would like to make even more detailed images of the surface of giant stars and follow the evolution of these granules continuously, instead of only getting snapshot images.

Puzzling Finding Raises New Questions About Atmospheric Physics Of Giant Planets

The hottest point on a gaseous planet near a distant star isn’t where astrophysicists expected it to be – a discovery that challenges scientists’ understanding of the many planets of this type found in solar systems outside our own.

Unlike our familiar planet Jupiter, so-called hot Jupiters circle astonishingly close to their host star—so close that it typically takes fewer than three days to complete an orbit. And one hemisphere of these planets always faces its host star, while the other faces permanently out into the dark.

Not surprisingly, the “day” side of the planets gets vastly hotter than the night side, and the hottest point of all tends to be the spot closest to the star. Astrophysicists theorize and observe that these planets also experience strong winds blowing eastward near their equators, which can sometimes displace the hot spot toward the east.

In the mysterious case of exoplanet CoRoT-2b, however, the hot spot turns out to lie in the opposite direction: west of center. A research team led by astronomers at McGill University’s McGill Space Institute (MSI) and the Institute for research on exoplanets (iREx) in Montreal made the discovery using NASA’s Spitzer Space Telescope. Their findings are reported Jan. 22 in the journal Nature Astronomy.

Wrong-way wind

“We’ve previously studied nine other hot Jupiter, giant planets orbiting super close to their star. In every case, they have had winds blowing to the east, as theory would predict,” says McGill astronomer Nicolas Cowan, a co-author on the study and researcher at MSI and iREx. “But now, nature has thrown us a curveball. On this planet, the wind blows the wrong way. Since it’s often the exceptions that prove the rule, we are hoping that studying this planet will help us understand what makes hot Jupiters tick.”

CoRoT-2b, discovered a decade ago by a French-led space observatory mission, is 930 light years from Earth. While many other hot Jupiters have been detected in recent years, CoRoT-2b has continued to intrigue astronomers because of two factors: its inflated size and the puzzling spectrum of light emissions from its surface.

“Both of these factors suggest there is something unusual happening in the atmosphere of this hot Jupiter,” says Lisa Dang, a McGill PhD student and lead author of the new study. By using Spitzer’s Infrared Array Camera to observe the planet while it completed an orbit around its host star, the researchers were able to map the planet’s surface brightness for the first time, revealing the westward hot spot.

New questions

The researchers offer three possible explanations for the unexpected discovery – each of which raises new questions:

The planet could be spinning so slowly that one rotation takes longer than a full orbit of its star; this could create winds blowing toward the west rather than the east – but it would also undercut theories about planet-star gravitational interaction in such tight orbits.

The planet’s atmosphere could be interacting with the planet’s magnetic field to modify its wind pattern; this could provide a rare opportunity to study an exoplanet’s magnetic field.

Large clouds covering the eastern side of the planet could make it appear darker than it would otherwise – but this would undercut current models of atmospheric circulation on such planets.

“We’ll need better data to shed light on the questions raised by our finding,” Dang says. “Fortunately, the James Webb Space Telescope, scheduled to launch next year, should be capable of tackling this problem. Armed with a mirror that has 100 times the collecting power of Spitzer’s, it should provide us with exquisite data like never before.”

Middle-Aged Sun Observed By Tracking Motion Of Mercury

Like the waistband of a couch potato in midlife, the orbits of planets in our solar system are expanding. It happens because the Sun’s gravitational grip gradually weakens as our star ages and loses mass. Now, a team of NASA and MIT scientists has indirectly measured this mass loss and other solar parameters by looking at changes in Mercury’s orbit.

The new values improve upon earlier predictions by reducing the amount of uncertainty. That’s especially important for the rate of solar mass loss, because it’s related to the stability of G, the gravitational constant. Although G is considered a fixed number, whether it’s really constant is still a fundamental question in physics.

“Mercury is the perfect test object for these experiments because it is so sensitive to the gravitational effect and activity of the Sun,” said Antonio Genova, the lead author of the study published in Nature Communications and a Massachusetts Institute of Technology researcher working at NASA’s Goddard Space Flight Center in Greenbelt, Maryland.

The study began by improving Mercury’s charted ephemeris — the road map of the planet’s position in our sky over time. For that, the team drew on radio tracking data that monitored the location of NASA’s MESSENGER spacecraft while the mission was active. Short for Mercury Surface, Space Environment, Geochemistry, and Ranging, the robotic spacecraft made three flybys of Mercury in 2008 and 2009 and orbited the planet from March 2011 through April 2015. The scientists worked backward, analyzing subtle changes in Mercury’s motion as a way of learning about the Sun and how its physical parameters influence the planet’s orbit.

For centuries, scientists have studied Mercury’s motion, paying particular attention to its perihelion, or the closest point to the Sun during its orbit. Observations long ago revealed that the perihelion shifts over time, called precession. Although the gravitational tugs of other planets account for most of Mercury’s precession, they don’t account for all of it.

The second-largest contribution comes from the warping of space-time around the Sun because of the star’s own gravity, which is covered by Einstein’s theory of general relativity. The success of general relativity in explaining most of Mercury’s remaining precession helped persuade scientists that Einstein’s theory was right.

Other, much smaller contributions to Mercury’s precession, are attributed to the Sun’s interior structure and dynamics. One of those is the Sun’s oblateness, a measure of how much it bulges at the middle — its own version of a “spare tire” around the waist — rather than being a perfect sphere. The researchers obtained an improved estimate of oblateness that is consistent with other types of studies.

The researchers were able to separate some of the solar parameters from the relativistic effects, something not accomplished by earlier studies that relied on ephemeris data. The team developed a novel technique that simultaneously estimated and integrated the orbits of both MESSENGER and Mercury, leading to a comprehensive solution that includes quantities related to the evolution of Sun’s interior and to relativistic effects.

“We’re addressing long-standing and very important questions both in fundamental physics and solar science by using a planetary-science approach,” said Goddard geophysicist Erwan Mazarico. “By coming at these problems from a different perspective, we can gain more confidence in the numbers, and we can learn more about the interplay between the Sun and the planets.”

The team’s new estimate of the rate of solar mass loss represents one of the first times this value has been constrained based on observations rather than theoretical calculations. From the theoretical work, scientists previously predicted a loss of one-tenth of a percent of the Sun’s mass over 10 billion years; that’s enough to reduce the star’s gravitational pull and allow the orbits of the planets to spread by about half an inch, or 1.5 centimeters, per year per AU (an AU, or astronomical unit, is the distance between Earth and the Sun: about 93 million miles).

The new value is slightly lower than earlier predictions but has less uncertainty. That made it possible for the team to improve the stability of G by a factor of 10, compared to values derived from studies of the motion of the Moon.

“The study demonstrates how making measurements of planetary orbit changes throughout the solar system opens the possibility of future discoveries about the nature of the Sun and planets, and indeed, about the basic workings of the universe,” said co-author Maria Zuber, vice president for research at MIT.

Hubble And Spitzer Team Up To Find Magnified And Stretched Out Image Of Distant Galaxy

An intensive survey deep into the universe by NASA’s Hubble and Spitzer space telescopes has yielded the proverbial needle-in-a-haystack: the farthest galaxy yet seen in an image that has been stretched and amplified by a phenomenon called gravitational lensing.

The embryonic galaxy named SPT0615-JD existed when the universe was just 500 million years old. Though a few other primitive galaxies have been seen at this early epoch, they have essentially all looked like red dots given their small size and tremendous distances. However, in this case, the gravitational field of a massive foreground galaxy cluster not only amplified the light from the background galaxy but also smeared the image of it into an arc (about 2 arcseconds long).

“No other candidate galaxy has been found at such a great distance that also gives you the spatial information that this arc image does. By analyzing the effects of gravitational lensing on the image of this galaxy, we can determine its actual size and shape,” said the study’s lead author Brett Salmon of the Space Telescope Science Institute in Baltimore, Maryland. He is presenting his research at the 231st meeting of the American Astronomical Society in Washington, D.C.

First predicted by Albert Einstein a century ago, the warping of space by the gravity of a massive foreground object can brighten and distort the images of far more distant background objects. Astronomers use this “zoom lens” effect to go hunting for amplified images of distant galaxies that otherwise would not be visible with today’s telescopes.

SPT0615-JD was identified in Hubble’s Reionization Lensing Cluster Survey (RELICS) and companion S-RELICS Spitzer program. “RELICS was designed to discover distant galaxies like these that are magnified brightly enough for detailed study,” said Dan Coe, Principal Investigator of RELICS. RELICS observed 41 massive galaxy clusters for the first time in the infrared with Hubble to search for such distant lensed galaxies. One of these clusters was SPT-CL J0615-5746, which Salmon analyzed to make this discovery. Upon finding the lens-arc, Salmon thought, “Oh, wow! I think we’re on to something!”

By combining the Hubble and Spitzer data, Salmon calculated the lookback time to the galaxy of 13.3 billion years. Preliminary analysis suggests the diminutive galaxy weighs in at no more than 3 billion solar masses (roughly 1/100th the mass of our fully grown Milky Way galaxy). It is less than 2,500 light-years across, half the size of the Small Magellanic Cloud, a satellite galaxy of our Milky Way. The object is considered prototypical of young galaxies that emerged during the epoch shortly after the big bang.

The galaxy is right at the limits of Hubble’s detection capabilities, but just the beginning for the upcoming NASA James Webb Space Telescope’s powerful capabilities, said Salmon. “This galaxy is an exciting target for science with the Webb telescope as it offers the unique opportunity for resolving stellar populations in the very early universe.” Spectroscopy with Webb will allow for astronomers to study in detail the firestorm of starbirth activity taking place at this early epoch, and resolve its substructure.

Archeology Of Our Milky Way’s Ancient Hub

For many years, astronomers had a simple view of our Milky Way’s central hub, or bulge, as a quiescent place composed of old stars, the earliest homesteaders of our galaxy.

However, because the inner Milky Way is such a crowded environment, it has always been a challenge to disentangle stellar motions to probe the bulge in detail.

Now, a new analysis of about 10,000 normal Sun-like stars in the bulge reveals that our galaxy’s hub is a dynamic environment of stars of various ages zipping around at different speeds, like travelers bustling about a busy airport. This conclusion is based on nine years’ worth of archival data from NASA’s Hubble Space Telescope.

The Hubble study of this complicated, chaotic heart of our Milky Way may provide new clues to the evolution of our galaxy, said researchers.

The research team, led by Will Clarkson of the University of Michigan-Dearborn, found that the motions of bulge stars are different, depending on a star’s chemical composition. Stars richer in elements heavier than hydrogen and helium have less disordered motions, but are orbiting around the galactic center faster than older stars that are deficient in heavier elements.

“There are many theories describing the formation of our galaxy and central bulge,” said Annalisa Calamida of the Space Telescope Science Institute, Baltimore, Maryland, a member of the Hubble research team. “Some say the bulge formed when the galaxy first formed about 13 billion years ago. In this case, all bulge stars should be old and share a similar motion. But others think the bulge formed later in the galaxy’s lifetime, slowly evolving after the first generations of stars were born. In this scenario, some of the stars in the bulge might be younger, with their chemical composition enriched in heavier elements expelled from the death of previous generations of stars, and they should show a different motion compared to the older stars. The stars in our study are showing characteristics of both models. Therefore, this analysis can help us in understanding the bulge’s origin.”

The astronomers divided the stars by their chemical compositions and then compared the motions of each group. They determined the stars’ chemical content by studying their colors and divided them in two main groups according to their heavy-element (iron) abundance. The chemically enriched stars are moving twice as fast as the other population.

“By analyzing nine years’ worth of data in the archive and improving our analysis techniques, we have made a clear, robust detection of the differences in the motion for chemically deficient and chemically enriched Sun-like stars,” Clarkson said. “We hope to continue our analysis, which will allow us to make a three-dimensional chart of the rich chemical and dynamical complexity of the populations in the bulge.”

The astronomers based their analysis on Advanced Camera for Surveys and Wide Field Camera 3 data from two Hubble surveys: the Wide Field Camera 3 Galactic Bulge Treasury Program and the Sagittarius Window Eclipsing Extrasolar Planet Search. Sets of spectra from the European Southern Observatory’s Very Large Telescope in Chile were used to help estimate the chemical compositions of stars.

Currently, only Hubble has sharp enough resolution to simultaneously measure the motions of thousands of Sun-like stars at the the galaxy bulge’s distance from Earth. The center of our galaxy is about 26,000 light-years away. “Before this analysis, the motions of these stars was not known,” said team member Kailash Sahu of the Space Telescope Science Institute. “You need a long time baseline to accurately measure the positions and the motions of these faint stars.”

The team studied Sun-like stars because they are so abundant and easily within Hubble’s reach. Previous observations looked at brighter, aging red giant stars, which are not as plentiful because they represent a brief episode in a star’s lifetime. “Hubble gave us a narrow, pencil-beam view of the galaxy’s core, but we are seeing thousands more stars than those spotted in earlier studies,” Calamida said. The Milky Way’s bulge is roughly one-tenth the diameter of our pancake-shaped galaxy. “We next plan to extend our analysis to do additional observations along different sight-lines, which will allow us to make a three-dimensional probe of the rich complexity of the populations in the bulge,” Clarkson added.

The researchers said that this work is also an important pathfinder for NASA’s James Webb Space Telescope to probe the archaeology of the Milky Way. Scheduled for launch in 2019, Webb is expected to more deeply probe stellar populations in the Milky Way bulge.