As artificial intelligence systems become more powerful, the movement of data between processors and memory is emerging as one of the largest obstacles to performance. Modern AI workloads require enormous volumes of information to be transferred rapidly, yet conventional memory technologies struggle to keep up without consuming substantial energy. A research team at the Korea Advanced Institute of Science and Technology, or KAIST, has now developed a device structure designed to address one of the most persistent problems in three-dimensional memory: instability in vertically stacked semiconductor channels. By controlling the movement of oxygen at the nanoscale, the researchers have created an “oxygen-tunnel” architecture that could improve the speed, endurance and power efficiency of future AI chips.

The work focuses on oxide vertical channel transistors, or VCTs, a promising technology for increasing memory density beyond the limits of conventional planar designs. In a vertical channel device, electrical current flows through a semiconductor channel oriented vertically rather than horizontally. This architecture allows memory cells and transistors to be stacked in multiple layers, creating a three-dimensional structure that can hold more computing and storage elements within a smaller footprint. Such designs are particularly attractive for compute-in-memory systems, which perform calculations directly where data are stored instead of repeatedly moving information between separate memory and processor units.

The central challenge lies in the behavior of oxygen inside oxide semiconductors. These materials can contain oxygen vacancies, which are sites where oxygen atoms are missing from the crystal or amorphous structure. Although these vacancies may be useful in carefully controlled amounts, excessive or unstable concentrations can create unwanted electronic states, alter the transistor’s threshold voltage and increase leakage or variability. Over time, such defects can cause the electrical properties of the channel to drift, undermining the reliability required for memory and AI hardware. Supplying additional oxygen can compensate for these vacancies, but the solution creates another problem: oxygen may continue migrating beyond the channel and oxidize the nearby metal electrode.

That oxidation at the electrode interface can increase contact resistance, making it more difficult for current to enter and leave the transistor. The result is a fundamental materials trade-off. The semiconductor channel needs a controlled supply of oxygen to remain electrically stable, while the electrode must be shielded from the same oxygen to preserve a low-resistance connection. Conventional interlayer dielectric materials do not provide sufficient directional control, allowing oxygen to diffuse through regions where it can damage the device. KAIST researchers therefore designed a multilayer structure intended to guide oxygen toward the channel and block its movement toward the electrode.

The new structure consists of alternating layers of silicon nitride, silicon dioxide and silicon nitride, abbreviated as SiN/SiO₂/SiN. Rather than acting as a simple insulating barrier, this stack functions as a selective oxygen pathway. The researchers engineered the layers so that oxygen released from the interlayer dielectric could reach the oxide semiconductor, where it helps compensate oxygen vacancies, while its diffusion toward the metal contact was suppressed. The concept resembles a microscopic tunnel with a preferred direction of travel. By regulating the local chemical environment around the channel, the oxygen-tunnel architecture improves semiconductor stability without sacrificing the electrical quality of the electrode interface.

The researchers fabricated indium tin oxide vertical channel transistors incorporating the new structure and examined them using cross-sectional transmission electron microscopy. The imaging confirmed the formation of the intended SiN/SiO₂/SiN stack and showed a well-defined hole profile in the fabricated device. This structural evidence was important because the performance of vertically integrated transistors depends heavily on the precision of nanoscale layers and interfaces. Even small defects, irregularities or unintended diffusion paths can cause large changes in resistance, switching behavior and long-term reliability when thousands or millions of devices are combined in a memory array.

According to KAIST, the oxygen-tunnel transistors achieved high current density while also demonstrating strong data-retention characteristics. The devices maintained their performance during more than ten million cycles of severe electrical stress testing, with the threshold voltage shift remaining below 50 millivolts. Threshold voltage is the gate voltage required to switch a transistor into conduction, and changes in this value can cause a memory cell to be misread or an analog computation to become inaccurate. Limiting the shift to such a small value after repeated operation indicates that the new structure can withstand demanding workloads without the rapid electrical drift that has hindered oxide-based vertical devices.

The potential significance extends beyond memory density. In compute-in-memory hardware, transistors can be used not only to store information but also to perform operations such as multiplication and accumulation, which are fundamental to neural-network processing. These systems can reduce the energy lost when data travel back and forth between memory and a separate processor. The KAIST team evaluated the technology in combination with conventional silicon complementary metal-oxide-semiconductor circuitry and reported results suggesting that the oxide platform could improve the performance of monolithic three-dimensional compute-in-memory systems. Such integration could allow high-density memory layers to be placed directly above logic circuits, shortening interconnects and increasing the number of operations performed per unit of area.

The research was led by Hyeonho Gu, with Yongwoo Lee, Haksoon Jung and Jimin Kwon serving as corresponding authors, in collaboration with scientists from UNIST, Yonsei University, the Korea Research Institute of Chemical Technology, Seoul National University and other Korean institutions. Published in Advanced Functional Materials, the study presents oxygen migration control as a new route to solving reliability problems in vertically stacked oxide electronics. The paper, titled “Oxygen-Tunnel Indium Tin Oxide Vertical Channel Transistors with Enhanced Current Density and Reliability for Monolithic 3D Compute-In-Memory Systems,” was also selected as a Front Cover article. If the approach can be scaled to large arrays and integrated into manufacturing processes, it could help advance the ultra-low-power AI memory systems needed for increasingly data-intensive applications.

Subject of Research: Oxygen-tunnel indium tin oxide vertical channel transistors for reliable, high-density memory and monolithic 3D compute-in-memory systems.

Article Title: Oxygen-Tunnel Indium Tin Oxide Vertical Channel Transistors with Enhanced Current Density and Reliability for Monolithic 3D Compute-In-Memory Systems

Web References: https://doi.org/10.1002/adfm.202531989

References: Advanced Functional Materials, DOI: 10.1002/adfm.202531989

Image Credits: KAIST

Keywords

AI chips, three-dimensional memory, compute-in-memory, oxide semiconductors, vertical channel transistors, indium tin oxide, oxygen vacancies, oxygen migration, silicon nitride, silicon dioxide, semiconductor reliability, low-power computing

The work, led by Professor Kyung Min Kim of the Department of Materials Science and Engineering at the Korea Advanced Institute of Science and Technology, addresses a fundamental difference between conventional computers and biological brains. Digital electronics are generally designed to produce stable, repeatable outputs. Unwanted variations in current or voltage are treated as noise because they can distort data and reduce computational precision. The brain, however, does not operate as a perfectly deterministic machine. Biological neurons may fire at different times, or with different frequencies, even when they receive similar stimuli. These variations arise partly from the stochastic opening and closing of ion channels in neuronal membranes. Rather than simply degrading performance, that probabilistic behavior helps biological neural networks adapt to uncertain and changing environments.

The KAIST team sought to reproduce this useful biological irregularity in a compact semiconductor device. Their platform is based on a memristor, an electronic component whose resistance changes in response to an applied electrical stimulus and can retain that resistance after the stimulus is removed. Because memristors can both store information and participate in computation, they are widely viewed as promising building blocks for neuromorphic systems designed to imitate aspects of brain function. In ordinary memristor circuits, however, fluctuations in current are often regarded as an obstacle, or they are exploited mainly to generate random numbers and support probabilistic computing. The researchers instead asked whether the noise itself could be deliberately shaped and assigned a functional role in signal processing.

Their central discovery was that the noise produced by a memristor is not fixed. Its magnitude and behavior change when the device is placed in a different resistance state. By setting that state in advance, the researchers could alter how likely the device was to produce a spike in response to a given input. A spike is the brief electrical event used by artificial neurons to represent and transmit information. In the new system, the same input can therefore lead to different firing probabilities depending on the memristor’s programmed resistance. This creates a reconfigurable artificial neuron whose response is not only adjustable in strength, but also probabilistic in a way that resembles the variable firing patterns of biological sensory neurons.

The researchers named their device a programmable probabilistic neuron, or PPN. Its operation depends on converting current fluctuations into meaningful spike patterns rather than treating those fluctuations as errors. When an input signal crosses the neuron’s effective response conditions, the noise can help determine whether and when a spike is generated. Repeated presentations of a similar signal do not necessarily produce identical spike trains, but the statistical properties of those trains can be controlled. In technical terms, programming the memristor modifies the neuron’s firing probability and input response range. This gives the circuit a tunable relationship between the temporal characteristics of a signal and the electrical events used to encode it.

That tunability is especially important for time-series data, in which information is carried not only by the size of a signal but also by how quickly it changes. Human activity signals, such as those captured by wearable sensors, generally evolve over relatively slow time scales and occupy frequency ranges around hertz. Speech contains much faster variations, extending into the kilohertz range, where one kilohertz represents 1,000 cycles per second. Conventional hardware may require different filters, processing pathways, or neural circuits to handle such distinct signal regimes. The KAIST approach allows the same neuron architecture to be reconfigured simply by changing the memristor’s resistance state, enabling it to become more responsive to slow or fast inputs without replacing the underlying circuit.

In demonstrations, the team adjusted the artificial neuron for different frequency-selective tasks. With one configuration, the system processed signals associated with human movement and activity. With another, it responded to the more rapidly changing patterns found in speech. The resulting system encoded and classified the two categories of time-series information with accuracies of 94.8 percent for human activity recognition and 95.0 percent for speech recognition. These results indicate that the noise-tuning mechanism can influence not only the behavior of an isolated device, but also the performance of a practical signal-processing pipeline. The technology could be particularly valuable in edge devices, where data must be interpreted locally rather than transmitted continuously to a remote server.

The appeal of this approach extends beyond its recognition accuracy. Neuromorphic hardware is being developed to reduce the energy and latency costs associated with conventional artificial intelligence, especially in sensors, wearable electronics, robotics, and autonomous machines. In many current systems, sensor data are converted into digital form and moved through several layers of memory and computation, a process that consumes energy and introduces delays. Memristor-based neurons could perform aspects of sensing, memory, and computation closer to where data are generated. Because the KAIST neuron can be tuned for different signal speeds through a device-level resistance adjustment, one hardware platform might support multiple applications while reducing the need for separate specialized circuits.

Professor Kim described the significance of the work as a shift in how semiconductor noise is understood. Instead of treating noise solely as an indicator of instability or a source of computational error, the researchers demonstrated that it can be used as a programmable information-processing resource. The concept also reflects an important principle of biological computation: variability does not always have to be eliminated to achieve reliable behavior. When controlled statistically, randomness can help a system distinguish patterns, respond flexibly, and operate across changing conditions. The researchers say that future versions of the technology could contribute to low-power neuromorphic systems capable of processing diverse sensory signals with adaptable hardware.

The study was led by Dr. Do Hoon Kim as first author and was published in Advanced Materials under the title “Noise-Tunable Memristor Enabling Programmable Probabilistic Neurons for Frequency-Selective Time-Series Signal Encoding.” The research was supported by Korea’s Basic Research Program in Science and Engineering and the PIM Artificial Intelligence Semiconductor Core Technology Development Program, funded through the Ministry of Science and ICT and the National Research Foundation of Korea. Although further work will be needed to assess long-term reliability, large-scale integration, and performance under real-world conditions, the findings point to a provocative future for electronic noise: rather than an enemy that every circuit must silence, it may become one of the mechanisms that allows intelligent machines to sense and interpret the world.

Subject of Research:
Noise-tunable memristor-based programmable probabilistic neurons for frequency-selective time-series signal processing and neuromorphic computing.

Article Title:
Noise-Tunable Memristor Enabling Programmable Probabilistic Neurons for Frequency-Selective Time-Series Signal Encoding

News Publication Date:
August 16

Web References:
https://doi.org/10.1002/adma.74529
https://mediasvc.eurekalert.org/Api/v1/Multimedia/833f7795-897f-44eb-969a-97c2c7e4e1c3/Rendition/low-res/Content/Public

References:
Kim, Do Hoon et al., “Noise-Tunable Memristor Enabling Programmable Probabilistic Neurons for Frequency-Selective Time-Series Signal Encoding,” Advanced Materials, DOI: 10.1002/adma.74529.

Image Credits:
KAIST

Keywords

Memristor, probabilistic neuron, neuromorphic computing, semiconductor noise, artificial intelligence, brain-inspired technology, time-series signal processing, frequency-selective encoding, human activity recognition, speech recognition, edge AI, low-power electronics, KAIST.