A memory device that can operate at temperatures over 700 °C could enable electronic systems to withstand harsh conditions with less need for cooling. The device, which is a memristor based on graphene, tungsten and a hafnium oxide ceramic, can store data for over 50 hours, has a working voltage of just 1.5 V, and is robust to more than 10⁹ switching cycles. It also has a high switching speed of just tens of nanoseconds, according to its developers at the University of Southern California (USC), US.

“Our work provides one of the most critical electronic components – memory – for a wide range of applications, particularly in extreme environments,” says Joshua Yang, who directs USC’s Center On Neuromorphic Computing undeR ExTreme Environments (CONCRETE). “These include space exploration, deep-Earth drilling (for geothermal energy) and nuclear and fusion energy plants in which intense heat is generated.”

Heat-tolerant electronics could also dramatically reduce the need for energy-intensive cooling systems, cutting both power consumption and fan noise, Yang adds. “Our work also shows that these devices require significantly lower voltage and current to operate at elevated temperatures – meaning higher ambient temperature can actually improve energy efficiency of computing systems.”

A device to remember

Rather than being fixed, the resistance of a memristor (or memory-resistor to give it its full name) changes depending on the current or voltage previously applied to it. This means that specific resistances can be programmed into the devices and subsequently stored. Importantly, the “remembered” value of the resistive state persists even when the power is switched off, making it a non-volatile form of electronic memory.

Memristors are also capable of processing large amounts of data in parallel, making them faster and more energy-efficient than conventional memories for certain calculations such as matrix-vector multiplication. They are therefore useful for in-memory computer technologies, including those that are now routinely employed in artificial intelligence (AI) hardware.

An unexpected discovery

The memristor described in the new CONCRETE Center study consists of a hafnium oxide (HfO₂) layer sandwiched between two electrodes: a tungsten one on top and a graphene one on the bottom. Tungsten has the highest melting point of any metallic element, and the study’s first author, Jian Zhao, notes that graphene (a sheet of carbon just one atom thick) can also withstand high temperatures without degrading. Nevertheless, Yang says they didn’t specifically set out to make a super-high temperature device.

“As often in science, this work originated from an unexpected discovery,” he explains. “We identified a material stack with significantly higher temperature tolerance while investigating something else completely – namely trying to build a different kind of device using graphene.”

Understanding why this stack could withstand such high temperatures and validating their hypotheses took considerable effort, Yang tells Physics World. The team used a combination of advanced electron microscopy, spectroscopy and first-principles calculations to work out the physical mechanisms behind the process, he adds.

The role of graphene

In conventional ceramic-based memristors, like those with a platinum bottom electrode, high temperatures cause the metal atoms from the top electrode to migrate through the ceramic layer until they reach the bottom electrode. When this happens, the two electrodes permanently connect and the devices short-circuit.

In the USC team’s memristor, though, this simply wasn’t happening. “Graphene puts an end to this process,” Yang explains. “Tungsten atoms still drift towards the graphene electrode as expected, but because of its surface chemistry and structure they cannot anchor onto it. These atoms therefore end up migrating away from the electrode, so avoiding short-circuiting and device failure.”

The researchers, who report their work in Science, say that one future research direction might be to search for materials that have a similar surface chemistry to graphene, but are easier to handle. Their next goal, which they acknowledge will be challenging, is to integrate their high-temperature memristors with logic devices (such as those based on SiC substrates) that can also withstand extreme temperatures.

To advance their memristor technology, Yang and his colleagues Glenn Ge, Miao Hu and Qiangfei Xia have founded a start-up company, Tetramem Inc., focused on developing memristor-based machine learning/AI accelerators. Though scaling up their devices will take time – the current examples were made by hand in the lab at the sub-microscale – Yang says that creating high-operating-temperature accelerators could enable intelligent computing in extreme environments, including space applications or datacentres.

The post Memory device breaks high-temperature performance record appeared first on Physics World.

The “bumblebee” bat – a little animal weighing just 2 g – has inspired researchers to make the first palm-sized drone that can efficiently navigate in confined, dark and cluttered environments. The drone, which works using echolocation and operates on a milliwatt of power, could find applications in search and rescue missions in difficult-to-access spaces, say the researchers at the Worcester Polytechnic Institute in the US who developed it.

The bumblebee bat thrives in deep, dark caves and can perceive objects as small as just 0.1 mm thanks to ultrasound-based echolocation. The bat sends short chirps and then listens to the echoes produced as the sound waves bounce off surfaces. This ability is all the more astounding since the animal has only simple biosensory apparatus and just two million neurons.

The new drone, developed by a team led by Nitin Sanket, differs from existing autonomous aerial robots that require sophisticated sensors to work – including light detection and ranging (LIDAR), radio detection and ranging (RADAR), tactile sensors and infrared-based depth cameras, to name just a few. These complicated devices cannot easily be deployed in cluttered environments under difficult environmental conditions, such as fog, dust, smoke, low light and/or snow. This makes them unsuitable for search and rescue missions in disaster zones, where such conditions are often the norm.

Another major problem with existing robots, explains Sanket, is that they generate a lot of propeller noise, making echolocation difficult. “It’s like trying to listen to your friend while a jet engine is taking off next to you,” he says.

The new device, which is detailed in Science Robotics, employs a physical acoustic shield inspired by the ear cartilages of bumblebee bats to overcome this problem. In addition, the team used an artificial-intelligence (AI)-based neural network denoising framework to recover weak echoes from noisy signals.

New device works well in the wild

Ultrasonic sensing is insensitive to most environmental conditions, such as smoke, snow, dust and darkness, that are visually degrading and render light-based sensors like cameras or LIDARs ineffective. As such, they work very well in the wild, says Sanket. “This will allow this new class of robots to be readily deployed for search and rescue in real-world settings where conditions are dynamic, unpredictable and visually degraded, bringing us one step closer to deploying swarms of aerial robots to look for survivors.”

The researchers built their aerial device using standard off-the-shelf parts for motors, and flight- and electronic speed controllers. They custom designed a carbon fibre frame and 3D-printed other structural parts. The on-board computer is a Google Coral Mini development board and the ultrasound sensors are made by TDK Electronics and designed by team member Richard Przybyla. The robot measures around 16 cm across, costs roughly $400 and works using just 1.2 mW of sensing power.

The robot uses echolocation to determine obstacle locations in 3D using trilateration, explains Sanket. “This means that once it has found the obstacles, it plans a path around them to avoid them and go towards a goal direction (like North, for example).”

At the heart of the device is noise reduction using the physical shield and the neural network (dubbed “Saranga” by the team), which reduces noise by looking at echo signatures over time, in the same way as the bat’s neuronal signal processing system does. The researchers trained the network entirely in simulation and say that it can be adapted to the real world without re-training/fine-tuning.

Looking to nature’s experts

The idea for the project actually started out as a joke during Halloween of 2024, remembers Sanket, when he and his students wanted to build a robot that emerged from smoke for a video. “That film was much harder to make than we anticipated, and it turned into an obsession, forcing us to solve a real problem: how to make robots navigate in visually degraded/challenging conditions.”

“To find the answer, we looked to nature’s experts, bats, which not only live but thrive in damp, dark and dusty caves and can pinpoint something as thin as a human hair,” he explained.

In their experiments, Sanket and his colleagues had to study how bats deal with low signal-to-noise ratios. They found that bats change their cartilage stiffness to muffle noise and have peculiar nose-leaves (ridges on their nose) to modulate sound chirps. They based their physical acoustic shield on these structures.

According to the researchers, these highly-functional autonomous tiny aerial robots could be deployed in critical humanitarian applications such as search and rescue, cave exploration and combating poaching – tasks currently infeasible using existing aerial robots. “They could, for example,” says Sanket, “be sent into disaster areas where human or larger helicopter access is limited, thereby alleviating the challenges and pressures associated with saving lives.”

Looking ahead, the Worcester Polytechnic Institute team is now working to increase the robot’s flying speed and reduce its size even further. “We speculate that looking at novel forms of flight mechanisms is the key,” Sanket tells Physics World.

The post Bat-inspired drone can navigate through dense fog and dodge obstacles appeared first on Physics World.

A new highly stable and energy-efficient memristor based on a hafnium oxide material can emulate the behaviour of synapses in the brain. The neuromorphic device could help dramatically cut the energy consumed by artificial intelligence (AI) hardware, say its developers at the University of Cambridge in the UK.

Today’s AI systems rely on conventional digital computers. These have separate processing and storage units and consume huge amounts of energy when performing data-intensive tasks. As global AI use is exploding, this energy consumption has already become unsustainable, says materials scientist Babak Bakhit, who led this new study.

An alternative way to process information

Neuromorphic computers could provide an alternative way to process information. As their name suggests, they are inspired by the architecture of the human brain. The circuits in these computers are made up of highly connected artificial neurons and artificial synapses that simulate the brain’s structure and functions. These machines have combined processing and memory units that allow them to process information at the same time as they store it, in the same way as a multi-tasking human brain. This means they could reduce energy consumption by as much as 70% compared with their digital counterparts.

Memory-resistors, or memristors, have become a fundamental building block of such neuromorphic architectures. This is because they can be engineered to behave very much like neurons in the human brain, which learn by reconfiguring the strengths of the connections (synapses) between neurons. Memristors excel in this respect as they can bring this learning functionality to the connections in electronic circuits.

First described theoretically in 1971, it was not until 2008 that researchers made the first practical version of a memristor. These devices are special in that their resistance can be programmed and subsequently stored. This is because, unlike standard resistors, the resistance of a memristor changes depending on the current previously applied to it – hence the “memory” in its name. What is more, the device “remembers” this resistive state even when the power is switched off.

Randomness in switching behaviour is a problem

All well and good, but most of today’s memristors unfortunately suffer from randomness in their switching behaviour because they rely on the formation of tiny conductive filaments in the materials making them up. These filamentary devices also typically require high forming and operating voltages and extra devices to avoid uncontrolled current changes that lead to permanent device failure. These challenges make such devices difficult to scale up for real-world applications, says Bakhit.

The researchers, who report their work in Science Advances, claim to have overcome the intrinsic stochasticity of memristive switching by exploiting a completely different switching mechanism – based on carefully engineered heterointerface physics rather than random filament switching. They achieved this by adding strontium and titanium to a hafnium-oxide thin film, which results in the formation of a p-n heterointerface. This junction allows the device to change its resistance smoothly by shifting the height of an energy barrier at the bottom interface through the migration of electro-ionic charges, explains Bakhit.

The new interfacial device has an ultralow switching current of less than or equal to 10^-8 A, which is around 10⁶ times lower than those of conventional oxide-based memristors. It also produces hundreds of distinct and stable conductance levels that can be easily modulated, a key prerequisite for analogue “in-memory” computing. And that’s not all: the device can also undergo tens of thousands of switching cycles without losing its programmed states for around a day.

Looking ahead, the researchers say they will now be focusing on translating their material and device breakthrough into a functional computing system. “In particular, we are working on reducing the thin-film growth temperature (which currently stands at around 700 °C) so that it is compatible with standard semiconductor manufacturing (CMOS) tolerances,” says Bakhit. “We will then scale up device arrays to demonstrate large-scale integration.”

Ultimately, the goal is to move from individual devices to fully integrated neuromorphic chips that can compete with, or surpass, conventional AI hardware in both performance and energy efficiency, he tells Physics World.

The post Memristive synapses could reduce AI energy consumption appeared first on Physics World.

Robots tend to move things physically, using arms or other appendages. But what if robots could move objects without physically touching them? Researchers from the Max Planck Institute for Intelligent Systems, the University of Michigan and Cornell University have developed robotic swarms that can manipulate objects using only water, by inducing a fluidic torque.

Strong viscous interactions exist in microscale systems, which can be used to generate fluid flows that actuate passive objects. In their previous research, the researchers found that this manipulation can be influenced by the number of microrobots, the spin rate of microrobots and the position of the microrobots relative to the object. This latest work, published in Science Advances, has gone one step further, demonstrating that a magnetic robot swarm can assemble, transport and reorganize objects that are many times larger than the microrobots themselves.

“This study is the third in a series of papers where our team explores how microscale robot swarms can coordinate using simple global control signals,” says Kirstin Petersen of Cornell University, “Rather than controlling each robot individually, we broadcast the same signal to the entire group and rely on the robots’ interactions with each other and with their environment to produce different collective behaviours. Here, we showed that those interactions could also be used to manipulate external structures through the fluid flows generated by the swarm”.

The robots are microdisks with diameters of about 300 µm and because they are magnetic, they can be rotated using an externally applied magnetic field. When each individual microrobot spins, it drags the fluid around it, which generates a force in the liquid. While this force is small for an individual robot, combining hundreds of robots together that spin in unison (and/or increasing the spin speed of the robots) creates a much larger flow force in the water – generating a high enough torque to move objects.

“The most exciting result is that the robot collective can use the fluidic torque it generates to manipulate structures much larger than the robots themselves, without physical contact. It suggests that you could add actuation to otherwise passive objects simply by introducing microrobots in the surrounding fluid,” Petersen tells Physics World.

To demonstrate this approach, the researchers positioned the microrobots inside and outside of concentric floating ring structures, and used the number of robots, their positions and spin speeds to act as a form of control for moving objects. They found that the robots could spread out and surround the object, rotating it in the process, or they could crawl around the edges of an object, allowing them to reorganize objects. The ability to change these parameters and obtain different torques provided a tuneable and programmable way of using the microrobot swarms.

The researchers extended the principles to mechanical systems, using the microrobot to turn miniature gear trains (after turning the first gear, the other gears moved by conventional mechanical contact). They also rotated 3D floating objects that were 45,000 times the mass of an individual robot. Here, placing the robots on top of the object generated sufficient torque to rotate it, despite the mass difference.

The team also found that the microrobot swarm could dynamically assemble objects using coordinated fluid flows, in which the robots switched between their rotational function and crawling ability to move objects along a surface. This adaptive behaviour not only allowed the manipulation of objects, but also their reorganization – including expelling, dispersing and aggregating objects – based on the environment and task requirements.

The introduction of these small robots into fluids essentially turns the fluid from a passive medium into a small-scale motor. For applications where there is a risk of structural damage from mechanical manipulation, contactless manipulation could be highly beneficial. For example, this type of mechanism could be useful in microscale manufacturing and biomedical engineering, particularly for miniature device assembly, biological matter transport and targeted manipulation within the human body.

When asked about what’s next for this research, Petersen tells Physics World that “the other authors are focusing specifically on innovating microrobots, whereas my lab is studying the broader question of how collectives coordinate through their shared environment while keeping individual agents simple. We are exploring natural and engineered fluid-coupled swarms across a wide range of size scales”.

The post Magnetic microrobot swarm moves objects with water appeared first on Physics World.