
- Neuromorphic Computing - Home
- Neuromorphic Computing - Introduction
- Neuromorphic Computing - Difference From Traditional Computing
- Neuromorphic Computing - History and Evolution
- Neuromorphic Computing - Types of Technologies
- Neuromorphic Computing - Architecture
- Neuromorphic Computing - Memristors
- Neuromorphic Computing - Synaptic Devices
- Neuromorphic Computing - Hardware Accelerators
- Neuromorphic Computing - Neuromorphic Chips
- Neuromorphic Computing - Analog Circuits
- Neuromorphic Algorithms and Programming
- Neuromorphic Computing - Spiking Neural Networks (SNNs)
- Neuromorphic Computing - Algorithms for SNNs
- Neuromorphic Computing - Programming Paradigms
- Applications of Neuromorphic Computing
- Neuromorphic Computing - Edge Computing
- Neuromorphic Computing - IoT
- Neuromorphic Computing - Robotics
- Neuromorphic Computing - Autonomous Systems
- Neuromorphic Computing - AI and ML
- Neuromorphic Computing - Cognitive Computing
- Neuromorphic Computing Resources
- Neuromorphic Computing - Useful Resources
- Neuromorphic Computing - Discussion
Neuromorphic Computing - Memristors
Memristors are important component in neuromorphic computing technology. These memory devices store information in the form of resistance, by changing the internal distribution of ions such as oxygen anions or metal cations. This way neuromorphic computers are able to store and process data as a single unit. In this section, we will explain detailed overview, working principle and examples of memristors in neuromorphic computers.
How Memristors Work?
A memristor, short for "memory resistor," operates by adjusting its resistance according to the amount and direction of current that passes through it. Memristors can remember the amount of charge that has previously flowed through them, even when the power is turned off. This unique ability allows them to function as both memory and processing elements in neuromorphic systems.

Imagine a neural network that needs to learn a new pattern. To achieve this, multiple memristors are arranged to form the network's synaptic connections. When a new pattern is detected, a voltage is applied across the relevant memristors, causing their resistance to change in according to the incoming pattern. This change allows the network to adjust and "learn" the new information. Here is step-by-step overview of what happens in memristors during this process.
- Initial State: Each memristor begins with a specific resistance value, typically either a high resistance state (HRS) or a low resistance state (LRS).
- Ion Movement: When voltage is applied, it triggers the movement of ions (such as oxygen vacancies or metal cations) within the memristors material. This migration alters the resistance of the memristor, either increasing or decreasing it, depending on the direction of the current.
- Resistance Change: Memristors can store change in resistance values as real numbers. Hence this will facilitate more detailed data storage similar to analog signals. The resistance change can be fine-tuned to reflect complex information patterns, which is not possible in traditional transistors that uses binary on/off states.
- Memory Retention: After the voltage is removed, the memristor retains its adjusted resistance, making it a non-volatile memory element. This feature is critical in neuromorphic computing, where the memristor's behavior closely resembles the synaptic learning process of the human brain, maintaining the learned pattern even without continuous power.
In neuromorphic systems, memristors can act as artificial synapses by adjusting their resistance based on the strength and timing of incoming electrical signals. This enables them to participate in learning processes, similar to how biological synapses adjust their strength through experience (synaptic plasticity).
Key Features of Memristors
Here are key features of memristors:
- Non-Volatile Memory: Memristors retain their resistance state even after the power is turned off. This makes them highly efficient for storing data in neuromorphic systems, similar to how the brain keeps learned information.
- Analog Data Storage: Traditional transistors store data in binaries (0s and 1s), While memristors can store in analog way using decimal values of resistance.
- Energy Efficiency: Memristors require low power to operate, which makes them highly energy-efficient. This is a significant advantage in neuromorphic systems, where thousands or even millions of memristors may be used in parallel.
- Scalability: Memristors are extremely small in size, allowing for high-density integration. This scalability makes them suitable for building large neuromorphic networks that mimic the dense connectivity of the human brain.
- Synaptic Plasticity: Memristors exhibit behavior similar to synaptic plasticity in biological systems. They can strengthen or weaken their connections based on the patterns they encounter, allowing for learning and adaptation over time.
Types of Memristor-Based Neurons
Researchers have come across different types of memristors, such as diffusive memristors (e.g., SiOxNy:Ag) and Mott memristors (e.g., NbO2 or VO2), which have shown their abilities in neuromorphic applications. However, each type comes with its own set of limitations:
- Diffusive Memristors: Known for their fast switching speed and high on/off ratio, but they suffer from poor uniformity, which impacts performance reliability.
- Mott Memristors: Offer better uniformity and speed but require high operating currents (~mA) and have lower on/off ratios, making them less energy efficient.