The team investigated whether memory behavior in a planar MoS₂ memtransistor could be controlled by modifying its electrical contacts rather than adding a switching material throughout the semiconductor channel. Using a solution-based AuCl₃ treatment, the researchers selectively altered the interface where a metal electrode meets MoS₂. Electrical measurements indicated that charge trapping and detrapping at this interface dynamically changed the effective Schottky barrier controlling carrier injection.

The intrinsic MoS₂ channel remained largely unmodified. Temperature-dependent measurements supported a contact-limited transport mechanism. The extracted effective Schottky barrier height was approximately 40.8 meV in the pristine device and approximately 196 meV after AuCl₃ treatment.

Raman spectroscopy, X-ray photoelectron spectroscopy and atomic force microscopy were used to examine material and chemical changes associated with the modified interface. Under drain-voltage pulse stimulation, the device produced gradual long-term potentiation and depression, spike-amplitude-dependent plasticity and timing-dependent conductance changes. Its paired-pulse response included both faster and slower temporal components, consistent with more than one interfacial charge-relaxation process.

To apply the device’s temporal dynamics, the researchers operated one memtransistor as the physical reservoir in an ECG-processing system. The single device was used with 10 time-multiplexed virtual nodes, rather than a physical array of devices, and its reservoir responses were analyzed for heartbeat classification. Of the 23,818 heartbeat segments, 19,053 were used for training and 4,765 for testing.

The resulting analysis produced an area under the receiver operating characteristic curve of 0.965 and an overall accuracy of 93.9% when discriminating normal and ventricular ectopic heartbeat classes. Because the switching mechanism is localized at the metal–MoS₂ interface, the study demonstrates that a controlled contact can provide the analog synaptic behavior needed for temporal signal processing without modifying the whole MoS₂ channel. The solution-based treatment also provides a route compatible with lithographic device fabrication and toward planar, independently addressable artificial synaptic devices.

The researchers described the ECG result as a proof of concept for processing temporal signals with the dynamic response of a single contact-engineered device.

Subject of Research: Contact-engineered MoS₂ memtransistors for analog synaptic memory and ECG reservoir computing

Article Title: Contact engineering provides MoS₂ memtransistor with brain-inspired memory for cardiac signal processing

Article References: Contact engineering provides MoS₂ memtransistor with brain-inspired memory for cardiac signal processing. Original publication (Bibliographic details unavailable.)

Image Credits: AI Generated

DOI: Not provided

Keywords: MoS₂ memtransistor, contact engineering, artificial synapse, reservoir computing, ECG processing, analog memory, Schottky barrier, charge trapping, neuromorphic computing, ventricular ectopic beats, Contact, engineering

The secret lies in the dielectric layer. Instead of synthetic polymers or metal oxides, the team, led by Shuolei Wei, Linhao Cheng, Jiacheng Huang and Jie Su, used dextran, a glucan polysaccharide whose chains are densely decorated with hydroxyl groups. This chemistry makes the dielectric exceptionally hydrophilic, so it spontaneously harvests water molecules from the ambient air. Under an applied electric field, protons migrate efficiently through the internal hydrogen-bond network of the glucan chains, accumulating at the dielectric/semiconductor interface to form an electric double layer, or EDL. At a modest frequency of 20 Hz, the dextran dielectric exhibits a capacitance of 5.14 microfarads per square centimeter, a value high enough to grant the gate strong control over the transistor channel at very low voltages.

The semiconductor layer is PDBT-co-TT, also known as DPP-DTT, an organic polymer semiconductor well suited to phototransistor operation. The strong interfacial electric field established by the EDL does more than lower the operating power; it dramatically enhances the dissociation of photogenerated excitons in DPP-DTT. Excitons, the bound electron-hole pairs created when light is absorbed, often recombine before they can contribute useful current in organic semiconductors. The dextran-gated field tears them apart, yielding a photoconductive gain of approximately 10^3 under weak illumination. That gain means the device can respond to light signals far too faint for ordinary organic phototransistors to register, addressing one of the field’s most persistent limitations.

With this hardware in place, the device emulates the essential behaviors of a biological synapse. It reproduces paired-pulse facilitation, the phenomenon in which a second neural pulse arriving shortly after the first produces a stronger response, and it demonstrates the transition from short-term to long-term plasticity, the cellular basis of learning and memory. The photoresponse is wavelength-dependent, so the synapse distinguishes between different colors of light much as retinal circuits do. These are not merely demonstrations; they are the computational primitives that neuromorphic engineers need to build brain-inspired systems that process information the way neural tissue does, rather than shuttling data between separate memory and processor units in the von Neumann architecture.

The photoelectric synergy of the device opens unexpected applications. Because light pulses can write, amplify and erase conductance states, the team used the synapse to transmit information optically in Morse code, with selective optical signal amplification and erasure supporting the protocol. A short flash of light becomes a dot, a longer pulse a dash, and the transistor’s plastic conductance states carry the message. This suggests a route toward optical communication interfaces in which the sensing element and the processing element are the same physical device, cutting the energy and complexity that separate photodetectors from neural processing circuits in conventional designs.

Perhaps the most striking application is neuromorphic encryption. The researchers leveraged hardware-derived conductance sequences from the transistor as physical unclonable keys. Because these conductance sequences arise from the stochastic electrochemistry of the dextran EDL and the device’s own photoresponse history, they are effectively impossible to duplicate without possessing the physical hardware itself. Using these sequences, the system encrypted and decrypted images of handwritten digits from the MNIST dataset, a standard benchmark in machine vision. The encryption is neuromorphic in the deepest sense: the same synapse that computes also secures, merging memory, processing and cryptography into a single low-power element.

The work fits into a broader movement toward green organic neuromorphic devices. Earlier efforts in the field have explored dielectrics and active layers derived from wood nanopaper, human hair keratin, chitosan, pectin, chlorophyll and other natural polyelectrolytes, motivated by the promise of flexible, biocompatible and even degradable electronics. The Qingdao team’s dextran dielectric advances this agenda by combining environmental sustainability with the ultra-low power consumption and high optical sensitivity that practical neuromorphic systems demand. Prior natural polysaccharide synapses have shown promise, but few have reached the femtojoule regime while simultaneously delivering photoconductive gains near 10^3 and supporting encryption protocols.

The implications ripple outward. Wearable health monitors, implantable neural interfaces and distributed sensor networks all face the same bottleneck: they must sense, compute and communicate continuously on power budgets measured in microwatts or less. A synapse that senses light, computes with synaptic plasticity and encrypts with physical keys, all while drawing femtojoule-scale pulse energies, could relax those constraints fundamentally. The wavelength-dependent photoresponse adds a further dimension, hinting at neuromorphic vision systems that discriminate color at the sensor level, and the Morse-code demonstration points toward optical inter-synapse communication protocols for future brain-inspired networks.

Challenges remain before dextran-gated synapses reach commercial maturity. The reliance on ambient water molecules for proton transport raises questions about device behavior in extreme humidity or dryness, and the long-term stability of natural polysaccharide dielectrics under continuous electrical stress must be characterized. Yet the conceptual advance is clear and compelling. By gating an organic semiconductor with a sugar, the Qingdao researchers have shown that the path to brain-scale energy efficiency may run through green chemistry, and that the same physical device can sense, learn, communicate and encrypt. As neuromorphic computing races toward real-world deployment, this dextran-gated transistor offers a template for devices that are simultaneously ultra-low power, optically sensitive and environmentally sustainable, a rare and powerful combination in the emerging landscape of brain-inspired hardware.

Subject of Research: Electronic materials

Article Title: Dextran-gated organic optoelectronic synaptic transistors with ultra-low power consumption and neuromorphic encryption computing

Article References: Wei, S., Cheng, L., Huang, J., & Su, J. (2026). Dextran-gated organic optoelectronic synaptic transistors with ultra-low power consumption and neuromorphic encryption computing. Journal of Materials Science. https://doi.org/10.1007/s10853-026-13947-7

Image Credits: AI Generated

DOI: 10.1007/s10853-026-13947-7

Keywords: neuromorphic computing, organic electronics, synaptic transistors, dextran, encryption, low-power devices, Dextran-gated, organic, optoelectronic, synaptic, transistors, ultra-low

The device itself is deceptively simple in architecture. It consists of a vertical stack of silver, hexagonal boron nitride, and indium tin oxide, all deposited on a flexible polyethylene terephthalate substrate. Hexagonal boron nitride, often nicknamed white graphene, is a layered insulating crystal whose atomically flat sheets can be stacked like a deck of cards. In a conventional memristor, electrical pulses program the resistance state, and mechanical deformation is usually treated as a nuisance, a source of noise or failure to be engineered away. The Shiv Nadar team inverted that logic. In their device, bending is the programming signal. When the flexible substrate flexes, the hBN flakes within the stack slide laterally relative to one another, and that sliding perturbs the stacking arrangement and the inter-flake tunneling distance that governs how easily electrons can hop across the layered insulator.

The elegance of the mechanism lies in what does not happen. Because the programming arises from interlayer sliding rather than from stretching the crystal lattice itself, the hBN sheets are never strained at the atomic level. The covalent bonds within each flake remain intact, which means the process is non-destructive and, crucially, reversible. Straighten the device and the flakes can slide back, restoring the previous tunneling gap and the previous conductance state. This stands in sharp contrast to many flexible memristor designs, where repeated mechanical cycling gradually degrades the active material through crack formation, atomic displacement, or electromigration. Here, deformation is not damage; it is information. The researchers describe this as a mechanoelectronic design principle, one in which the physical interaction between device and environment becomes the write mechanism for synaptic plasticity.

What makes the comparison to the neuromuscular junction more than a metaphor is the way the device responds to bending. Biological mechanosensitive neurons transduce deformation into graded changes in synaptic strength, with both the amplitude and the dynamics of the response modulated by the magnitude and history of the mechanical stimulus. The hBN memristor reproduces this co-modulation quantitatively. Different bending conditions produce different families of conductance states, and the temporal evolution of those states mirrors the amplitude and dynamics seen in their biological counterparts. In other words, the device does not simply switch between on and off; it exhibits the continuous, history-dependent plasticity that neuroscientists associate with real synapses, where the same stimulus can produce different outcomes depending on what came before.

To demonstrate that these mechanically programmed states are more than a laboratory curiosity, the team put the device to work as an artificial synapse in a computational brain. The conductance values measured under varying bending conditions were used as analog synaptic weights in a fully connected neural network, the workhorse architecture of modern machine learning. The network was trained on MNIST, the classic benchmark of handwritten digit recognition that has served for decades as a proving ground for neuromorphic hardware. Across all bending conditions tested, the network achieved classification accuracy exceeding 91 percent. That figure matters because it shows that the mechanically induced conductance states are not random or noisy; they are stable, distinguishable, and useful as computational elements. A synapse that changes when you bend it, yet still supports high-accuracy inference, is a synapse that could live inside a prosthetic hand or a soft robot.

The implications for flexible and wearable neuromorphics are considerable. Current approaches to tactile sensing in soft electronics typically separate the sensing element from the computing element: a pressure sensor converts touch into an electrical signal, which is then routed to a processor that decides what it means. That separation costs power, latency, and wiring complexity, all of which are precious commodities in a wearable device. A memristor that programs itself through touch collapses that boundary. The sensing and the memory update are the same physical event, happening at the same place at the same time. For applications such as electronic skin, adaptive prosthetics, or soft robotic grippers that must learn from contact, this kind of co-located mechanosensation and plasticity could dramatically simplify the hardware stack.

The choice of materials is also strategically significant. Hexagonal boron nitride is chemically robust, thermally stable, and widely studied as a protective encapsulant and tunneling barrier in two-dimensional electronics. Its layered structure, with weak van der Waals forces between sheets and strong covalent bonds within them, is precisely what enables the sliding mechanism at the heart of the device. The silver and indium tin oxide electrodes provide the vertical conduction path, and the PET substrate supplies the flexibility that makes the whole concept possible. Because the active mechanism does not rely on electrochemical filament formation or oxygen vacancy migration, the usual culprits behind memristor variability and endurance limits, the device sidesteps several long-standing reliability problems in the field.

There are, of course, hurdles between a promising laboratory demonstration and a deployable technology. The reported work establishes the principle and the benchmark performance, but scaling from individual devices to the dense crossbar arrays needed for practical neuromorphic systems will require careful control of device-to-device uniformity, particularly for a mechanism that depends on the precise stacking of exfoliated or grown two-dimensional flakes. Integration with conventional readout circuitry, long-term cycling stability under millions of bending events, and compatibility with large-area manufacturing are all questions that the field will now press forward. The authors acknowledge support from the Science and Engineering Research Board and the Department of Science and Technology in India, reflecting the growing investment in flexible neuromorphic platforms worldwide.

Still, the conceptual shift is hard to overstate. For half a century, computing has treated memory and processing as separate functions and mechanical deformation as an enemy of reliability. This work suggests a future in which the shape of a device is part of its computational state, in which a bend is a bit, and in which soft machines learn the way our muscles do, through the very act of moving. The neuromuscular junction has been quietly optimizing that trick for hundreds of millions of years. With a few atomic layers of boron nitride and a flexible plastic backing, engineers have finally started to take notes.

Subject of Research: Mechanosensitive hexagonal boron nitride memristors for flexible neuromorphic computing

Article Title: Neuromuscular Junction Inspired Mechanosensitive hBN-based Memristors for Flexible Neuromorphics

Article References: Parida, T., Swain, S. K., Sengupta, D., & Kanjilal, A. (2026). Neuromuscular Junction Inspired Mechanosensitive hBN-based Memristors for Flexible Neuromorphics. npj Flexible Electronics. https://doi.org/10.1038/s41528-026-00647-0

Image Credits: AI Generated

DOI: 10.1038/s41528-026-00647-0

Keywords: memristor, hexagonal boron nitride, neuromorphic computing, flexible electronics, neuromuscular junction, synaptic plasticity, mechanosensitive, interlayer sliding, MNIST, wearable devices, two-dimensional materials, artificial synapse

That search has now produced a striking result. Researchers led by Professor Yang Li at Suzhou University of Science and Technology have built a flexible, ultra-fast artificial synapse printed entirely from room-temperature liquid inks, and the device can dissolve on demand in just six minutes. Writing in the International Journal of Extreme Manufacturing, the team demonstrates that high-performance, brain-inspired hardware does not require extreme vacuum chambers, rare metals, or permanent electronic waste. The finding challenges a long-standing assumption in the electronics industry that cutting-edge neuromorphic devices must be fabricated in energy-hungry cleanroom environments using rigid silicon and heavy metal electrodes, materials that eventually contribute to the millions of tons of non-recyclable electronic waste discarded around the world every year.

The central obstacle that has held back soft, carbon-based computing components is a stubborn material conflict. Organic compounds are naturally poor conductors of electricity, and every past attempt to push their switching speeds into a useful range forced engineers back to rigid metal plates deposited in vacuum chambers, undoing the very advantages that made organic electronics attractive in the first place. Flexible substrates and printable inks promised cheap, lightweight, and biologically compatible devices, but without a way to achieve fast, reliable electrical switching in a purely solution-processed material, the field remained stuck at the laboratory bench.

Professor Li’s team bypassed this trade-off through an elegant piece of molecular engineering. They assembled organic porphyrin molecules into a rigid, two-dimensional lattice known as a hydrogen-bonded organic framework, abbreviated TCPP-HOF. In this material, hydrogen bonds act like microscopic structural beams, locking the molecules into flat sheets stacked just 0.35 nanometers apart, a spacing roughly one hundred thousand times narrower than a single human hair. The result is a material that combines the processability of a liquid ink with the structural order of a crystal, giving charge carriers the precisely aligned pathways they need to move quickly while remaining fully compatible with low-temperature, solution-based manufacturing.

Functionally, this orderly architecture behaves like a microscopic turnstile that remembers every electrical pulse passing through it. When a forward voltage pulse strikes the sheet, electric charges slip into designated molecular resting pockets, paving a low-resistance path and raising the device’s electrical conductance, much as a biological synapse grows stronger each time a memory is reinforced. A reverse voltage pulse draws those charges back out, resetting the connection. Because the molecular lanes are precisely aligned, this switching happens in just 26 nanoseconds, roughly twenty million times faster than a human eyelid can blink, while the device reliably holds ten distinct memory levels. That combination of speed and multi-level storage is exactly what neuromorphic computing demands, since artificial neural networks rely on synapses whose strength can be tuned gradually and retained over time.

To test how well the hardware handles real computational work, the researchers printed a ten-by-ten grid of these synapses on a flexible polyimine plastic base and ran it through standard handwritten-digit recognition trials, a classic benchmark for evaluating neuromorphic hardware. The printed array correctly identified the digits with 97.23 percent accuracy, coming remarkably close to the 99.24 percent theoretical ceiling achieved by standard software running on conventional computers. For a device manufactured entirely from printable liquid inks at room temperature, that performance gap is narrow enough to suggest that solution-processed organic frameworks could eventually compete with far more expensive fabrication routes in practical machine-learning hardware.

Durability, often the weak point of flexible electronics, also held up under stress. The flexible organic layers kept working even after the sheet was bent two hundred consecutive times around a tight five-millimeter curve, a deformation far more severe than the gentle flexing a wearable device would typically experience. Then came the most unusual demonstration: once the device finished its job, dipping it into a warm chemical wash broke down the polymer base and the hydrogen-bonded framework into soluble monomers in six minutes flat, leaving behind only minimal inert residue. In effect, the chip can be programmed to disappear, converting what would normally be permanent electronic waste into simple, recoverable molecular building blocks.

The manufacturing implications are considerable. Because the entire device is built from solution-based inks, brain-inspired chips of this kind could be rolled out using conventional spray-coating and inkjet printers, skipping the multi-million-dollar vacuum systems that silicon fabs depend on. That opens the door to low-cost, distributed production of neuromorphic hardware, potentially in facilities far removed from the specialized foundries that dominate semiconductor manufacturing today. The immediate applications lie in bendable medical patches and short-term wearable monitors that process health data directly on the body and then dissolve when no longer needed, eliminating both the disposal problem and the privacy concerns associated with discarded sensors that still hold personal information.

Challenges remain before such devices can leave the laboratory. The researchers are now working to scale the ten-by-ten grid into larger crossbar arrays, the dense mesh structures needed to run realistic neural network workloads, while simultaneously engineering ultra-thin moisture barriers to keep the water-soluble circuits functioning reliably in humid outdoor air. Protecting a deliberately dissolvable device from everyday moisture without sealing it so thoroughly that it can no longer be broken down is a delicate balancing act, and it will determine whether transient neuromorphic electronics can survive outside controlled environments. The team’s progress on both fronts will be closely watched by a field increasingly interested in sustainable, transient electronics.

Even so, the demonstration marks a meaningful convergence of three goals that have often been pursued separately: brain-inspired computing performance, printable low-temperature manufacturing, and environmentally responsible disposal. A single material system, held together by nothing more exotic than hydrogen bonds and printed from liquid ink, now delivers nanosecond switching, ten stable memory states, near-software-level recognition accuracy, mechanical flexibility, and complete dissolvability in six minutes. If the scaling and moisture-barrier work succeeds, the idea that powerful computing hardware must be permanent, rigid, and wasteful may soon look as outdated as the vacuum tubes it once replaced.

Subject of Research: Solution-processed hydrogen-bonded organic framework artificial synapses for transient neuromorphic computing

Article Title: Printable brain-inspired chips that compute at lightning speed and vanish in minutes

Article References: Printable brain-inspired chips that compute at lightning speed and vanish in minutes. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: neuromorphic computing, artificial synapse, hydrogen-bonded organic framework, porphyrin, transient electronics, printed electronics, flexible electronics, dissolvable devices, wearable sensors, handwritten digit recognition, solution processing, electronic waste