System Bits: July 23

Superconductivity seen in trilayer graphene
Stanford University and University of California at Berkeley researchers discovered signs of superconductivity in stacking three-layer sheets of graphene, they report.

“It’s definitely an exciting development,” says Cory Dean, a physicist at Columbia University. Dean notes that bilayer graphene superconducts only when the atomic lattices of the two graphene layers are twisted with respect to one another by a “magic” angle of 1.1°—a difficult maneuver to perform on the thinnest material known. “If you’re off by a little bit, it doesn’t work,” Dean says. The trilayer graphene, by contrast, doesn’t have to be twisted. Rather, the atomic lattice of each layer aligns with those above and below, which happens naturally when multilayer graphene is produced.

Stanford’s David Goldhaber-Gordon and Cal’s Feng Weng led the teams involved in the trilayer graphene research. The researchers began by sticking a piece of Scotch tape on a bulk piece of graphite—the “lead” in most pencils—and peeling it off. Repeating the process leaves flakes of graphene sticking to the tape, some just a single sheet thick, but others with two and three layers. Wang’s team previously pioneered a technique to spot unique optical signatures in trilayer graphene.

The team then used these trilayer flakes as a starting material to make electrical devices. They sandwiched trilayer flakes between layers of boron nitride, which isolate the graphene from contaminants and prevent it from buckling. In some places, the atoms in the boron nitride layers line up precisely with the carbon atoms in the graphene layers, but a few nanometers away they are offset. After about 10 nanometers the atoms in the layers align once again, creating a “moiré” repeating pattern that is also apparent in the twisted bilayer graphene. Each repeated moiré cell can hold up to four extra electrons, in addition to those in the material, altering the material’s conductivity.

Next, the researchers patterned metals on top of the flakes, building transistors with “gates” that control the addition of electrons to the material. By manipulating the electric field on their gates, the researchers were able to control exactly how many electrons were present in each repeated moiré cell. When they added three electrons to each cell and dropped the temperature below 2 kelvins, they noticed a sharp drop in electrical resistance, a sign of superconductivity, which they report in the journal Nature. They also noticed that when they applied an external magnetic field to their sample, the near-zero electrical resistance vanished, another sign of superconductivity. “All these things check the boxes [of superconductivity],” Goldhaber-Gordon says. But he adds the signals are not yet definitive. For one, the electrical resistance does not drop completely to zero, which is required for a superconductor. However, he points out, this could be due to impurities in the graphene flakes. “It may not be superconducting everywhere within the device,” he says.

Still, Goldhaber-Gordon notes that the apparent superconductivity from the three extra electrons is similar to what is seen in conventional high-temperature superconductors, the copper-based materials that were discovered in 1986. For Dean, that raises hopes that trilayer graphene will be a good model system for solving that long-standing mystery. Trilayer graphene, he says, “is such a clean system it provides a simple way to explore complex physics.”

Deep neural network detects deepfakes
University of California at Riverside researchers came up with a deep neural network architecture that can identify manipulated images at the pixel level with great precision. The DNN could prove useful in identifying “deepfakes,” images that have been altered, often for disinformation or propaganda purposes. Such altered images are also being employed in business frauds, such as impersonating CEOs in order to steal corporate frauds.

Amit Roy-Chowdhury’s Video Computing Group at UCR is behind the groundbreaking research. Roy-Chowdhury is a professor of electrical and computer engineering and the Bourns Family Faculty Fellow in the Marlan and Rosemary Bourns College of Engineering.

Objects in images have boundaries and whenever an object is inserted or removed from an image, its boundary will have different qualities than the boundaries of objects in the image naturally. Someone with good Photoshop skills will do their best to make the inserted object looks as natural as possible by smoothing these boundaries.

While this might fool the naked eye, when examined pixel by pixel, the boundaries of the inserted object are different. For example, inserted boundaries are often smoother than the natural objects. By detecting boundaries of inserted and removed objects, a computer should be able to identify altered images.

The researchers labeled nonmanipulated images and the relevant pixels in boundary regions of manipulated images in a large dataset of photos. The aim was to teach the neural network general knowledge about the manipulated and natural regions of photos. They tested the neural network with a set of images it had never seen before, and it detected the altered ones most of the time. It even spotted the manipulated region.

“We trained the system to distinguish between manipulated and nonmanipulated images, and now if you give it a new image it is able to provide a probability that that image is manipulated or not, and to localize the region of the image where the manipulation occurred,” Roy-Chowdhury said.

The researchers are working on still images for now, but they point out that this can also help them detect deepfake videos.

“If you can understand the characteristics in a still image, in a video it’s basically just putting still images together one after another,” Roy-Chowdhury said. “The more fundamental challenge is probably figuring out whether a frame in a video is manipulated or not.”

Even a single manipulated frame would raise a red flag. But Roy-Chowdhury thinks we still have a long way to go before automated tools can detect deepfake videos in the wild.

Molecular motions captured in a “movie”
Brown University researchers report using ultra-high-speed X-ray pulses to produce a high-resolution “movie” of a molecule undergoing structural motions. The research was published in the Nature Chemistry journal.

The ability to see molecular motions in real time offers insights into chemical dynamics processes that were unthinkable just a few decades ago, the researchers say, and may ultimately help in optimizing reactions and designing new types of chemistry.

“For many years, chemists have learned about chemical reactions by essentially studying the molecules present before and after a reaction has occurred,” said Brian Stankus, a recent Ph.D. graduate from Brown University and co-lead author on the paper. “It was impossible to actually watch chemistry as it happens because most molecular transformations happen very quickly. But ultrafast light sources like the one we used in this experiment have enabled us to measure molecular motions in real time, and this is the first time these sorts of subtle effects have been seen with such clarity in an organic molecule of this size.”

The work is a collaboration between chemists from Brown, scientists at SLAC National Accelerator Laboratory, and theoretical chemists from the University of Edinburgh in the U.K. The team was led by Peter Weber, professor of chemistry at Brown.

For the study, the researchers looked at the molecular motions that occur when the organic molecule N-methyl morpholine is excited by pulses of ultraviolet light. X-ray pulses from SLAC’s Linac Coherent Light Source were used to take snapshots at different stages of the molecule’s dynamic response.

“We basically hit the molecules with UV light, which initiates the response, and then fractions of a second later we take a ‘picture’ — actually we capture a scattering pattern — with an X-ray pulse,” Stankus said. “We repeat this over and over, with different intervals between the UV pulse and X-ray pulse to create a time-series.”

The X-rays scatter in particular patterns depending on the structure of molecules. Those patterns are analyzed and used to reconstruct a shape of the molecule as the molecular motions unfold. That pattern analysis was led by Haiwang Yong, a graduate student at Brown and the study’s co-lead author.

The experiment revealed an extremely subtle reaction in which only a single electron becomes excited, causing a distinct pattern of molecular vibrations. The researchers were able to image both the electron excitation and the atomic vibration in fine detail.

“This paper is a true milestone because for the first time, we were able to measure in great clarity the structure of a molecule in an excited state and with time resolution,” said Weber, the study’s corresponding author.

“Making these types of nearly noise-free measurements in both energy and time is no small feat,” said Mike Minitti, a senior staff scientist at SLAC and study co-author. “Over the past seven years, our collaboration has learned a great deal on how best to use the various LCLS diagnostics to precisely measure the small fluctuations in X-ray intensities, and to an even greater extent, track the femtosecond timescale changes the molecules evolve on. All of this has informed the development of custom data analysis routines that virtually eliminate pesky, unwanted signals to our data. These results demonstrate the fidelity we can achieve.”

A particularly interesting aspect of the reaction, the researchers say, is that it’s coherent — meaning when groups of these molecules interact with light, their atoms vibrate in concert with each other.

“If we can use experiments like this one to study how exactly light can be used to direct the collective motion of billions of molecules, we can design systems that can be coherently controlled,” Stankus said. “Put simply: If we understand exactly how light directs molecular motions, we can design new systems and control them to do useful chemistry.”