Topological Field-Effect Memristor Enables Robust, Reconfigurable Computing Beyond von Neumann Architectures

The quest to move beyond the limitations of conventional computing has driven researchers to explore novel devices and computational paradigms, and neuromorphic computing represents a promising avenue. Manuel Meyer, Selena Barragan, Sergey Krishtopenko, and colleagues demonstrate a significant advance in this field with the creation of a topological field-effect memristor. This device uniquely combines the benefits of robust, low-power topological transport with the adaptability of non-volatile memory, overcoming a key challenge in the development of next-generation computing. By integrating electrically tunable coherent and incoherent transport within a single device, the team realises a fundamental topological electronic element, offering a pathway towards hybrid computing architectures that promise enhanced efficiency and performance.

Topological Phase Control in Memristive Devices

This research details the development of a novel memristive device, built from a topological insulator, for potential use in neuromorphic computing. The team successfully integrated topological insulators, materials with robust, dissipationless edge currents, with memristors, devices that change resistance based on past current flow, mimicking synaptic plasticity. They demonstrated the ability to control the topological phase, and thus the edge current conductivity, of a quantum well structure using an applied voltage, allowing precise tuning of the memristive behavior. The device exhibits non-volatile memory characteristics, retaining its resistance state even without power, and can emulate synaptic plasticity, crucial for building artificial neural networks.

Topological insulators are materials that act as insulators internally but conduct electricity along their surfaces or edges, offering robust and efficient current flow. Researchers utilized a specific structure, InAs/GaSb/InAs quantum wells, to create a two-dimensional topological insulator, carefully engineering the band structure to achieve the desired topological properties. A memristor is a two-terminal device whose resistance depends on its history of current flow, making it a promising building block for neuromorphic computing due to its ability to mimic synaptic behavior. Neuromorphic computing is a computing paradigm inspired by the structure and function of the human brain, aiming for energy efficiency and parallel processing.

The InAs/GaSb/InAs quantum well heterostructure is engineered to create a topological phase with conducting edge states. Applying a voltage modulates the band structure of the quantum wells, effectively controlling the conductivity of the edge states. Changes in edge state conductivity alter the device’s resistance, creating a memristive effect. The resistance state is non-volatile, meaning it’s retained even after the voltage is removed. By controlling the voltage and current flow, the device can mimic the behavior of synapses in the brain, including strengthening and weakening. This research presents a promising new approach to building neuromorphic computing systems by leveraging the unique properties of topological insulators to create robust, tunable, and energy-efficient memristive devices.

Topological Memristor Fabricated from Quantum Well Heterostructure

The research team engineered a topological field-effect memristor based on an inverted InAs/GaInSb/InAs trilayer quantum well heterostructure, grown on an AlSb buffer, to explore new computational paradigms beyond conventional architectures. This device leverages the unique properties of the quantum spin Hall insulator state, achieving both coherent transport and non-volatile memory functionality within a single, reconfigurable element. The fabrication process carefully controlled quantum well thicknesses and indium mole fractions to induce a band-inverted regime, essential for establishing the topological state and enabling robust, low-power operation. To realize the memristor, scientists harnessed the intrinsic floating-gate behavior of the gate dielectric, creating a device broadly tunable by electric fields.

The team constructed a six-terminal Hall bar geometry, allowing for precise control and measurement of electrical characteristics. A superlattice of alternating SiO2 and SiN layers served as the gate dielectric, facilitating charge trapping at the semiconductor/oxide interface. This configuration enables memristive switching between a bulk-dominated, incoherent state and a helical edge-dominated, coherent state, leveraging the topological protection inherent in the quantum spin Hall insulator state. The study pioneered a dual-gating approach, employing both a top gate and a back gate to independently manipulate the Fermi energy and perpendicular electric field, providing full electrostatic control over the device.

Transport measurements were conducted at 4. 2K, allowing for detailed characterization of the device’s behavior in both transistor and memristor configurations. By sweeping the top-gate voltage, the team observed a prominent resistance peak when the Fermi energy resided within the topological insulating gap, demonstrating the tunability of the device and the emergence of hysteresis due to charge accumulation at the oxide-semiconductor interface. This intrinsic floating gate functionality is central to the memristive behavior, enabling adaptive electronic behavior.

Topological Memristor Integrates Switching and Robust Transport

Scientists have realized a topological field-effect memristor that uniquely combines quantized helical edge transport with memristive switching between distinct, history-dependent transport regimes. This device, built on a scalable III-V semiconductor platform, exploits the helical edge states of InAs/GaInSb/InAs quantum wells to switch between coherent topological and incoherent bulk conducting states. Experiments demonstrate that the device functions as both a transistor and a memristor, unifying topologically protected transport with adaptive charge-storage functionality within a single material system. The team measured a quantized resistance value of precisely h/2e2 in the microscopic device, confirming the robustness and dissipationless character of the underlying topological edge states even during device-level operation.

This measurement establishes a direct link between topological protection and non-volatile memory functionality, introducing a new class of quantum-coherent memory elements. During sweeps at zero current, the device exhibited significantly lower differential resistance when dominated by incoherent bulk conductivity. The differential resistance in the band-gap region reached approximately 1. 3 h/2e2, exceeding a quantized value due to a reduction in phase coherence length. These results demonstrate that memristive switching between a low-resistance incoherent bulk state and a high-resistance coherent helical edge state is achievable within a single device. The intrinsic dual transistor, memristor operation enables information to be stored and processed locally within the same coherent medium, offering a route to overcoming the von Neumann bottleneck. This breakthrough delivers a prototypical topological electronic device that combines coherence, adaptability, and scalability, opening the prospect of integrated quantum-neuromorphic hardware architectures.

Topological Memristor Unifies Storage and Transport

This work demonstrates a novel topological field-effect memristor, fabricated from an inverted InAs/GaInSb/InAs trilayer well structure, which integrates coherent and incoherent transport with non-volatile memory functionality. The device exploits the helical edge states present in the material to switch between distinct conducting states, achieving a quantized resistance value indicative of dissipationless, topological edge transport. This establishes a direct link between topological protection and the ability to store information, introducing a new type of quantum-coherent memory element. Crucially, the researchers successfully unified topologically protected transport and adaptive charge-storage within a single material system, enabling both information storage and processing to occur locally, offering a potential pathway to overcome limitations inherent in conventional von Neumann architectures. The device’s scalability, built on a III-V semiconductor platform, suggests the possibility of future integrated circuits.