Tag Archives: neuromorphic devices

It’s not all about simulating the synapse for neuromorphic (brainlike) computing: presenting dendritic integration

Leave a reply

Michael Berger’s May 20, 2026 Nanowerk Spotlight article features a new (to me) aspect (or, if you prefer, challenge) to neuromorphic computing, Note: A link has been removed,

Efforts to design computing systems that operate more like the brain have pushed engineers to rethink how information is processed, transmitted, and stored. Biological neurons are not simple relays. Their ability to process input relies not just on synapses—the connections between neurons—but also on dendrites. These branching structures collect and integrate signals across both time and space, shaping how a neuron responds.

Most neuromorphic devices developed so far have focused on mimicking synaptic functions. Dendritic behavior, which governs how multiple inputs are combined and modulated, remains less explored. This gap limits the capacity of neuromorphic hardware to emulate the full computational complexity of biological neurons.

…

For anyone unfamiliar with dendrites, here’s a description from the Dendrite Wikipedia entry, which follows the image, Note: Links not included in the caption for the image have been removed,

Credity: Curtis Neveu – Own work. Caption: The neuron contains dendrites that receives information, a cell body called the soma, an an axon that sends information. Schwann cells make activity move faster down axon. Synapses allow neurons to activate other neurons. The dendrites receive a signal, the axon hillock funnels the signal to the initial segment and the initial segment triggers the activity (action potential) that is sent along the axon towards the synapse. Please see learnbio.org for interactive version. CC BY-SA 4.0
File:Anatomy of neuron.png
Created: 17 May 2022
Uploaded: 17 May 2022

A dendrite (from Greek δένδρον déndron, “tree”) or dendron is a branched cytoplasmic process that extends from a nerve cell that propagates the electrochemical stimulation received from other neural cells to the cell body, or soma, of the neuron from which the dendrites project. Electrical stimulation is transmitted onto dendrites by upstream neurons (usually via their axons) via synapses which are located at various points throughout the dendritic tree.

Dendrites play a critical role in integrating these synaptic inputs and in determining the extent to which action potentials are produced by the neuron.[1]

…

Berger’s May 20, 2026 article explains how scientists are attempting to create artificial dendrites, Note: Links have been removed,

Artificial dendrites are difficult to construct. Unlike synapses, which can often be replicated with resistive memory elements (memristors), dendrites require spatially distributed signal processing and sensitivity to the timing of input spikes. Biological dendrites perform this by managing ion flow across complex membrane structures, often with localized chemical and electrical variations. Traditional electronic systems, which rely on electrons in solid-state circuits, struggle to reproduce these dynamics.

Ionic devices offer a more faithful analogue. In particular, nanofluidic memristors—devices that transport ions through confined channels—can mimic how neurons regulate ionic currents. Prior work has shown that such systems can simulate synaptic plasticity and memory. Yet most rely on electrical stimulation, which adds complexity to control circuitry.

In contrast, light offers a clean, contactless way to manipulate ion behavior. Optogenetics, a biological technique that uses light to activate ion channels in neurons, has shown how effective this can be. Researchers have started applying similar principles to synthetic systems, but artificial dendrites with full spatiotemporal integration remain rare.

A study published in Advanced Materials (“Optogenetics‐Inspired Nanofluidic Artificial Dendrite with Spatiotemporal Integration Functions”) introduces a nanofluidic device that addresses this challenge. Developed by a team at Northeast Normal University [NENU], the system integrates layered graphene oxide (GO) into a flexible polydimethylsiloxane (PDMS) matrix. It uses light to control sodium ion (Na⁺) transport through nanochannels. This approach simulates how dendrites integrate signals from different spatial locations and over time. It also lays the groundwork for more advanced neuromorphic machines that include artificial sensory-motor reflexes.

…

This work shows how optical modulation of ionic pathways can be used to create functional artificial dendrites. It opens a path toward more realistic neural circuits in hardware, capable not just of memory and learning, but of the nuanced signal processing required for perception and motor control. As components like this are refined, they could play a central role in building autonomous systems that interact more naturally with their environment.

Here’s a link to and a citation for the paper,

Optogenetics-Inspired Nanofluidic Artificial Dendrite with Spatiotemporal Integration Functions by Zhuangzhuang Li, Ya Lin, Xuanyu Shan, Zhongqiang Wang, Xiaoning Zhao, Ye Tao, Haiyang Xu, Yichun Liu. Advanced Materials First published: 16 May 2025 Online Version of Record before inclusion in an issue 2502438 DOI: https://doi.org/10.1002/adma.202502438

This paper is behind a paywall.

If you have the time, Berger’s May 20, 2026 article provides more detail about the device.

Memristors could help AIs overcome ‘catastrophic forgetting’

Leave a reply

A March 20,205 news item on SciencDaily describes a ‘novel’ memristor,

They consume extremely little power and behave similarly to brain cells: so-called memristors. Researchers from Jülich [Forschungszentrum Juelich; Germany], led by Ilia Valov, have now introduced novel memristive components in Nature Communications that offer significant advantages over previous versions: they are more robust, function across a wider voltage range, and can operate in both analog and digital modes. These properties could help address the problem of “catastrophic forgetting,” where artificial neural networks abruptly forget previously learned information.

The problem of “catastrophic forgetting” occurs when deep neural networks are trained for a new task. This is because a new optimization simply overwrites a previous one. The brain does not have this problem because it can apparently adjust the degree of synaptic change; experts are now also talking about a so-called “metaplasticity”. They suspect that it is only through these different degrees of plasticity that our brain can permanently learn new tasks without forgetting old content. The new memristor accomplishes something similar.

“Its unique properties allow the use of different switching modes to control the modulation of the memristor in such a way that stored information is not lost,” says Ilia Valov from the Peter Grünberg Institute (PGI-7) at Forschungszentrum Jülich.

…

A March 20, 2025 Forschungszentrum Juelich press release (also on EurekAlert), which originated the news item, provides context for the work along with more technical details,

Ideal candidates for neuro-inspired devices

Modern computer chips are evolving rapidly. Their development could receive a further boost from memristors—a term derived from memory and resistor. These components are essentially resistors with memory: their electrical resistance changes depending on the applied voltage, and unlike conventional switching elements, their resistance value remains even after the voltage is turned off. This is because memristors can undergo structural changes—for example, due to atoms depositing on the electrodes.

“Memristive elements are considered ideal candidates for learning-capable, neuro-inspired computer components modeled on the brain,” says Ilia Valov.

Despite considerable progress and efforts, the commercialization of the components is progressing slower than expected. This is due in particular to an often high failure rate in production and a short lifespan of the products. In addition, they are sensitive to heat generation or mechanical influences, which can lead to frequent malfunctions during operation. “Basic research is therefore essential to better control nanoscale processes,” says Valov, who has been working in this field of memristors for many years. ”We need new materials and switching mechanisms to reduce the complexity of the systems and increase the range of functionalities.”

It is precisely in this regard that the chemist and materials scientist, together with German and Chinese colleagues, has now been able to report an important success: “We have discovered a fundamentally new electrochemical memristive mechanism that is chemically and electrically more stable,” explains Valov. The development has now been presented in the journal Nature Communications.

A New Mechanism for Memristors

“So far, two main mechanisms have been identified for the functioning of so-called bipolar memristors: ECM and VCM,” explains Valov. ECM stands for ‘Electrochemical Metallization’ and VCM for ‘Valence Change Mechanism’.

  • ECM memristors form a metallic filament between the two electrodes—a tiny “conductive bridge” that alters electrical resistance and dissolves again when the voltage is reversed. The critical parameter here is the energy barrier (resistance) of the electrochemical reaction. This design allows for low switching voltages and fast switching times, but the generated states are variable and relatively short-lived.
     
  • VCM memristors, on the other hand, do not change resistance through the movement of metal ions but rather through the movement of oxygen ions at the interface between the electrode and electrolyte—by modifying the so-called Schottky barrier. This process is comparatively stable but requires high switching voltages.

Each type of memristor has its own advantages and disadvantages. “We therefore considered designing a memristor that combines the benefits of both types,” explains Ilia Valov. Among experts, this was previously thought to be impossible. “Our new memristor is based on a completely different principle: it utilizes a filament made of metal oxides rather than a purely metallic one like ECM,” Valov explains. This filament is formed by the movement of oxygen and tantalum ions and is highly stable—it never fully dissolves. “You can think of it as a filament that always exists to some extent and is only chemically modified,” says Valov.

The novel switching mechanism is therefore very robust. The scientists also refer to it as a filament conductivity modification mechanism (FCM). Components based on this mechanism have several advantages: they are chemically and electrically more stable, more resistant to high temperatures, have a wider voltage window and require lower voltages to produce. As a result, fewer components burn out during the manufacturing process, the reject rate is lower and their lifespan is longer.

Perspective solution for “catastrophic forgetting”

On top of that, the different oxidation states allow the memristor to be operated in a binary and/or analog mode. While binary signals are digital and can only output two states, analog signals are continuous and can take on any intermediate value. This combination of analog and digital behavior is particularly interesting for neuromorphic chips because it can help to overcome the problem of “catastrophic forgetting”: deep neural networks delete what they have learned when they are trained for a new task. This is because a new optimization simply overwrites a previous one.

The brain does not have this problem because it can apparently adjust the degree of synaptic change; experts are now also talking about a so-called “metaplasticity”. They suspect that it is only through these different degrees of plasticity that our brain can permanently learn new tasks without forgetting old content. The new ohmic memristor accomplishes something similar. “Its unique properties allow the use of different switching modes to control the modulation of the memristor in such a way that stored information is not lost,” says Valov.

The researchers have already implemented the new memristive component in a model of an artificial neural network in a simulation. In several image data sets, the system achieved a high level of accuracy in pattern recognition. In the future, the team wants to look for other materials for memristors that might work even better and more stably than the version presented here. “Our results will further advance the development of electronics for ‘computation-in-memory’ applications,” Valov is certain.

Here’s a link to and a citation for the paper,

Electrochemical ohmic memristors for continual learning by Shaochuan Chen, Zhen Yang, Heinrich Hartmann, Astrid Besmehn, Yuchao Yang & Ilia Valov. Nature Communications volume 16, Article number: 2348 (2025) DOI: https://doi.org/10.1038/s41467-025-57543-w Published: 08 March 2025

This paper is open access.