A process developed by Rice University chemical and biomolecular engineer Aditya Mohite and collaborators at Northwestern University, the University of Pennsylvania and the University of Rennes yields 2D perovskite-based semiconductor layers of ideal thickness and purity by controlling the temperature and duration of the crystallization process.

Known as kinetically controlled space confinement, the process could help improve the stability and reduce the cost of halide perovskite-based emerging technologies like optoelectronics and photovoltaics.

“Producing 2D perovskite crystals with layer thicknesses ⎯ or quantum well thickness, also known as ‘n value’⎯ greater than two is a major bottleneck,” said Jin Hou, a Ph.D. student in Rice’s George R. Brown School of Engineering who is a lead author on a study about the process published in Nature Synthesis. “An n value higher than four means materials have a narrower band gap and higher electrical conductivity ⎯ a crucial factor for application in electronic devices.”

As they form into crystals, atoms or molecules arrange themselves into highly organized, regular lattices. Ice, for instance, has 18 possible atomic arrangements, or phases. Like the hydrogen and oxygen atoms in ice, the particles that make up halide perovskites can also form multiple lattice arrangements. Because material properties are phase-dependent, scientists aim to synthesize 2D halide perovskite layers that exhibit only a single phase throughout. The problem, however, is that traditional synthesis methods for higher n value 2D perovskites generate uneven crystal growth, which impacts the material’s performance reliability.

“In traditional methods of 2D perovskite synthesis, you get crystals with mixed phases due to the lack of control over crystallization kinetics, which is basically the dynamic interplay between temperature and time,” Hou said. “We designed a way to slow down the crystallization and tune each kinetics parameter gradually to hit the sweet spot for phase-pure synthesis.”

In addition to designing a synthesis method that can achieve a gradual n value increase in 2D halide perovskites, the researchers also created a map ⎯ or phase diagram ⎯ of the process through characterization, optical spectroscopy and machine learning.

“This work pushes the boundaries of higher quantum well 2D perovskites synthesis, making them a viable and stable option for a variety of applications,” Hou said.

“We have developed a new method to improve the purity of the crystals and resolved a long-standing question in the field on how to approach high n value, phase-pure crystal synthesis,” said Mohite, an associate professor of chemical and biomolecular engineering and materials science and nanoengineering whose lab has pioneered various methods of improving halide perovskite semiconductor quality and performance, from calibrating the initial stage of crystallization to fine-tuning solvent design.

“This research breakthrough is critical for the synthesis of 2D perovskites, which hold the key to achieving commercially relevant stability for solar cells and for many other optoelectronic device applications and fundamental light matter interactions,” Mohite added.

Semiconductors — most notably, silicon — underpin the computers, cellphones, and other electronic devices that power our daily lives, including the device on which you are reading this article. As ubiquitous as semiconductors have become, they come with limitations. The atomic structure of any material vibrates, which creates quantum particles called phonons. Phonons in turn cause the particles — either electrons or electron-hole pairs called excitons — that carry energy and information around electronic devices to scatter in a matter of nanometers and femtoseconds. This means that energy is lost in the form of heat, and that information transfer has a speed limit.

The search is on for better options. Writing in Science, a team of chemists at Columbia University led by Jack Tulyag, a PhD student working with chemistry professor Milan Delor, describes the fastest and most efficient semiconductor yet: a superatomic material called Re₆Se₈Cl₂.

Rather than scattering when they come into contact with phonons, excitons in Re₆Se₈Cl₂ actually bind with phonons to create new quasiparticles called acoustic exciton-polarons. Although polarons are found in many materials, those in Re₆Se₈Cl₂ have a special property: they are capable of ballistic, or scatter-free, flow. This ballistic behavior could mean faster and more efficient devices one day.

In experiments run by the team, acoustic exciton-polarons in Re₆Se₈Cl₂ moved fast — twice as fast as electrons in silicon — and crossed several microns of the sample in less than a nanosecond. Given that polarons can last for about 11 nanoseconds, the team thinks the exciton-polarons could cover more than 25 micrometers at a time. And because these quasiparticles are controlled by light rather than an electrical current and gating, processing speeds in theoretical devices have the potential to reach femtoseconds — six orders of magnitude faster than the nanoseconds achievable in current Gigahertz electronics. All at room temperature.

“In terms of energy transport, Re₆Se₈Cl₂ is the best semiconductor that we know of, at least so far,” Delor said.

A Quantum Version of the Tortoise and the Hare

Re₆Se₈Cl₂ is a superatomic semiconductor created in the lab of collaborator Xavier Roy. Superatoms are clusters of atoms bound together that behave like one big atom, but with different properties than the elements used to build them. Synthesizing superatoms is a specialty of the Roy lab, and they are a main focus of Columbia’s NSF-funded Material Research Science and Engineering Center on Precision Assembled Quantum Materials. Delor is interested in controlling and manipulating the transport of energy through superatoms and other unique materials developed at Columbia. To do this, the team builds super-resolution imaging tools that can capture particles moving at ultrasmall, ultrafast scales.

When Tulyag first brought Re₆Se₈Cl₂ into the lab, it wasn’t to search for a new and improved semiconductor — it was to test the resolution of the lab’s microscopes with a material that, in principle, shouldn’t have conducted much of anything. “It was the opposite of what we expected,” said Delor. “Instead of the slow movement we expected, we saw the fastest thing we’ve ever seen.”

Tulyag and his peers in the Delor group spent the next two years working to pinpoint why Re₆Se₈Cl₂ showed such remarkable behavior, including developing an advanced microscope with extreme spatial and temporal resolution that can directly image polarons as they form and move through the material. Theoretical chemist Petra Shih, a PhD student working in Timothy Berkelbach’s group, also developed a quantum mechanical model that provides an explanation for the observations.

The new quasiparticles are fast, but, counterintuitively, they accomplish that speed by pacing themselves — a bit like the story of the tortoise and the hare, Delor explained. What makes silicon a desirable semiconductor is that electrons can move through it very quickly, but like the proverbial hare, they bounce around too much and don’t actually make it very far, very fast in the end. Excitons in Re₆Se₈Cl₂ are, comparatively, very slow, but it’s precisely because they are so slow that they are able to meet and pair up with equally slow-moving acoustic phonons. The resulting quasiparticles are “heavy” and, like the tortoise, advance slowly but steadily along. Unimpeded by other phonons along the way, acoustic exciton-polarons in Re₆Se₈Cl₂ ultimately move faster than electrons in silicon.

The Semiconductor Search Continues

Like many of the emerging quantum materials being explored at Columbia, Re₆Se₈Cl₂ can be peeled into atom-thin sheets, a feature that means they can potentially be combined with other similar materials in the search for additional unique properties. Re₆Se₈Cl_2, however, is unlikely to ever make its way into a commercial product — the first element in the molecule, Rhenium, is one of the rarest on earth and extremely expensive as a result.

But with the new theory from the Berkelbach group in hand along with the advanced imaging technique that Tulyag and the Delor group developed to directly track the formation and movement of polarons in the first place, the team is ready to see if there are other superatomic contenders capable of beating Re₆Se₈Cl_2′‘s speed record.

“This is the only material that anyone has seen sustained room-temperature ballistic exciton transport in. But we can now start to predict what other materials might be capable of this behavior that we just haven’t considered before,” said Delor. “There is a whole family of superatomic and other 2D semiconductor materials out there with properties favorable for acoustic polaron formation.”

But are machines capable of this type of thinking? In the late 1980s, Jerry Fodor and Zenon Pylyshyn, philosophers and cognitive scientists, posited that artificial neural networks — the engines that drive artificial intelligence and machine learning — are not capable of making these connections, known as “compositional generalizations.” However, in the decades since, scientists have been developing ways to instill this capacity in neural networks and related technologies, but with mixed success, thereby keeping alive this decades-old debate.

Researchers at New York University and Spain’s Pompeu Fabra University have now developed a technique — reported in the journal Nature — that advances the ability of these tools, such as ChatGPT, to make compositional generalizations. This technique, Meta-learning for Compositionality (MLC), outperforms existing approaches and is on par with, and in some cases better than, human performance. MLC centers on training neural networks — the engines driving ChatGPT and related technologies for speech recognition and natural language processing — to become better at compositional generalization through practice.

Developers of existing systems, including large language models, have hoped that compositional generalization will emerge from standard training methods, or have developed special-purpose architectures in order to achieve these abilities. MLC, in contrast, shows how explicitly practicing these skills allow these systems to unlock new powers, the authors note.

“For 35 years, researchers in cognitive science, artificial intelligence, linguistics, and philosophy have been debating whether neural networks can achieve human-like systematic generalization,” says Brenden Lake, an assistant professor in NYU’s Center for Data Science and Department of Psychology and one of the authors of the paper. “We have shown, for the first time, that a generic neural network can mimic or exceed human systematic generalization in a head-to-head comparison.”

In exploring the possibility of bolstering compositional learning in neural networks, the researchers created MLC, a novel learning procedure in which a neural network is continuously updated to improve its skills over a series of episodes. In an episode, MLC receives a new word and is asked to use it compositionally — for instance, to take the word “jump” and then create new word combinations, such as “jump twice” or “jump around right twice.” MLC then receives a new episode that features a different word, and so on, each time improving the network’s compositional skills.

To test the effectiveness of MLC, Lake, co-director of NYU’s Minds, Brains, and Machines Initiative, and Marco Baroni, a researcher at the Catalan Institute for Research and Advanced Studies and professor at the Department of Translation and Language Sciences of Pompeu Fabra University, conducted a series of experiments with human participants that were identical to the tasks performed by MLC.

In addition, rather than learn the meaning of actual words — terms humans would already know — they also had to learn the meaning of nonsensical terms (e.g., “zup” and “dax”) as defined by the researchers and know how to apply them in different ways. MLC performed as well as the human participants — and, in some cases, better than its human counterparts. MLC and people also outperformed ChatGPT and GPT-4, which despite its striking general abilities, showed difficulties with this learning task.

“Large language models such as ChatGPT still struggle with compositional generalization, though they have gotten better in recent years,” observes Baroni, a member of Pompeu Fabra University’s Computational Linguistics and Linguistic Theory research group. “But we think that MLC can further improve the compositional skills of large language models.”

Researchersexamined the exceptionally bright gamma-ray burst GRB 230307A, which was caused by a neutron star merger. The explosion was observed using an array of ground and space-based telescopes, including NASA’s James Webb Space Telescope, Fermi Gamma-ray Space Telescope, and Neil Gehrels Swift Observatory.

Publishing their findings today in Nature(25 Oct), the international research team which included experts from the University of Birmingham, reveal that they found the heavy chemical element tellurium, in the aftermath of the explosion.

Other elements such as iodine and thorium, which are needed to sustain life on earth, are also likely to be amongst the material ejected by the explosion, also known as a kilonova.

Dr Ben Gompertz, Assistant Professor of Astronomy at the University of Birmingham, and co-author of the study explains: “Gamma-ray bursts come from powerful jets travelling at almost the speed of light — in this case driven by a collision between two neutron stars. These stars spent several billion years spiralling towards one another before colliding to produce the gamma-ray burst we observed in March this year. The merger site is the approximate length of the Milky Way (about 120,000 light-years) outside of their home galaxy, meaning they must have been launched out together.

“Colliding neutron stars provide the conditions needed to synthesise very heavy elements, and the radioactive glow of these new elements powered the kilonova we detected as the blast faded. Kilonovae are extremely rare and very difficult to observe and study, which is why this discovery is so exciting.”

GRB 230307A was one of the brightest gamma-ray bursts ever observed — over a million times brighter than the entire Milky Way Galaxy combined. This is the second time individual heavy elements have been detected using spectroscopic observations after a neutron star merger, providing invaluable insight into how these vital building blocks needed for life are formed.

Lead author of the study Andrew Levan, Professor of Astrophysics at Radboud University in the Netherlands, said: “Just over 150 years since Dmitri Mendeleev wrote down the periodic table of elements, we are now finally in the position to start filling in those last blanks of understanding where everything was made, thanks to the James Webb Telescope.”

GRB 230307A lasted for 200 seconds, meaning it is categorised as a long-duration gamma-ray burst. This is unusual as short gamma-ray bursts, which last less than two seconds, are more commonly caused by neutron star mergers. Long gamma-ray bursts like this one are usually caused by the explosive death of a massive star.

The researchers are now seeking to learn more about how these neutron star mergers work and how they power these huge element-generating explosions.

Dr Samantha Oates, a co-author of the study while a postdoctoral research fellow at the University of Birmingham (now a lecturer at Lancaster University) said: “Just a few short years ago discoveries like this one would not have been possible, but thanks to the James Webb Space Telescope we can observe these mergers in exquisite detail.”

Dr Gompertz concludes: “Until recently, we didn’t think mergers could power gamma-ray bursts for more than two seconds. Our next job is to find more of these long-lived mergers and develop a better understanding of what drives them — and whether even heavier elements are being created. This discovery has opened the door to a transformative understanding of our universe and how it works.”

But findings from a new paper published in the journal Nature could lead to reanalysis of that data. An international team of researchers discovered the presence of a molten silicate layer overlying Mars’ metallic core — providing new insights into how Mars formed, evolved and became the barren planet it is today.

Published on October 25, 2023, the team’s paper details the use of seismic data to locate and identify a thin layer of molten silicates (rock-forming minerals that make up the crust and mantle of Mars and Earth) lying between the Martian mantle and core. With the discovery of this molten layer, the researchers determined that Mars’ core is both denser and smaller than previous estimates, a conclusion that better aligns with other geophysical data and analysis of Martian meteorites.

Vedran Lekic, a professor of geology at the University of Maryland and co-author of the paper, compared the molten layer to a ‘heating blanket’ covering the Martian core.

“The blanket not only insulates the heat coming from the core and prevents the core from cooling, but also concentrates radioactive elements whose decay generates heat” Lekic said. “And when that happens, the core is likely to be unable to produce the convective motions that would create a magnetic field — which can explain why Mars currently doesn’t have an active magnetic field around it.”

Without a functional protective magnetic field around itself, a terrestrial planet such as Mars would be extremely vulnerable to harsh solar winds and lose all the water on its surface, making it incapable of sustaining life. Lekic added that this difference between Earth and Mars could be attributed to differences in internal structure and the different planetary evolution paths the two planets took.

“The thermal blanketing of Mars’ metallic core by the liquid layer at the base of the mantle implies that external sources are necessary to generate the magnetic field recorded in the Martian crust during the first 500 to 800 million years of its evolution,” said the paper’s lead author Henri Samuel, a scientist with the French National Center for Scientific Research. “These sources could be energetic impacts or core motion generated by gravitational interactions with ancient satellites which have since then disappeared.”

The team’s conclusions support theories that Mars was at one time a molten ocean of magma that later crystallized to produce a layer of silicate melt enriched in iron and radioactive elements at the base of the Martian mantle. The heat emanating from the radioactive elements would then have dramatically altered the thermal evolution and cooling history of the red planet.

“These layers, if widespread, can have pretty big consequences for the rest of the planet,” Lekic said. “Their existence can help tell us whether magnetic fields can be generated and maintained, how planets cool over time, and also how the dynamics of their interiors change over time.”

NASA’s InSight mission officially ended in December 2022 after more than four years of collecting data on Mars, but the analysis of the observations continues. Samuel, Lekic and their co-authors are among the latest researchers to reexamine prior models of Mars using seismology to confirm the planet’s structure and turbulent history.

“This new discovery of a molten layer is just one example of how we continue to learn new things from the completed InSight mission,” Lekic said. “We hope that the information we’ve gathered on planetary evolution using seismic data is paving the way for future missions to celestial bodies like the moon and other planets like Venus.”