\t\t\t\t\tOpto-electrical excitation of MTJs<\/p>\n
Researchers at the University of Greifswald, International Iberian Nanotechnology Laboratory, Max Planck Institute for the Science of Light, and Aarhus University advanced the use of magnetic tunnel junctions<\/a> (MTJs) for neuromorphic computing.<\/p>\n The team developed a hybrid opto-electrical excitation scheme that combines electrical currents with short laser pulses, making it possible to generate high thermoelectric voltages in MTJs to simulate synapse behavior.<\/p>\n The researchers determined that the generated voltage can be adjusted flexibly depending on the electrical current, similar to the weight of a synapse in the brain, while information exchange between nerve cells was mimicked through spontaneous \u201cspike\u201d signals. In computer simulations, a simple neuromorphic network based on this technology recognized handwritten digits with an accuracy of 93.7%.<\/p>\n \u201cOur results show that MTJs with optical-electrical control represent a compact and energy-saving platform for the next generation of computing,\u201d said Markus M\u00fcnzenberg, a professor at the University of Greifswald, in a press release. \u201cAs the technology is compatible with today\u2019s semiconductor technology, we believe, that in the future, it could be used in everyday devices as well as high-performance computers.\u201d [1]<\/p>\n Nanofluidic memristors<\/p>\n Researchers from the University of Manchester created programmable nanofluidic memristors<\/a> that exhibit all four theoretically predicted types of memristive behavior and mimic the memory functions of the human brain. The memristors use confined liquid electrolytes within thin nanochannels made from 2D materials like molybdenum disulfide (MoS\u2082) and hexagonal boron nitride (hBN).<\/p>\n By tuning parameters such as electrolyte composition, pH, voltage frequency, and channel geometry, the same nanofluidic device can switch between four distinct memory loop styles, two \u201ccrossing\u201d and two \u201cnon-crossing\u201d types, the researchers said. These loop styles correspond to different memory mechanisms, including ion-ion interaction, ion-surface charge adsorption\/desorption, surface charge inversion, and ion concentration polarization. Similar to biological synapses, the devices also exhibit both short-term and long-term memory depending on the applied voltage and electrolyte conditions.<\/p>\n \u201cThis is the first time all four memristor types have been observed in a single device,\u201d said Radha Boya, a professor in the Condensed Matter Physics Group at the University of Manchester. \u201cIt shows the remarkable tunability of nanofluidic systems and their potential to replicate complex brain-like behavior.\u201d [2]<\/p>\n Droplets play tic-tac-toe<\/p>\n Researchers from Lawrence Livermore National Laboratory, University of Southern California, University of California Santa Barbara, and Google Research created a droplet-based platform<\/a> that uses ions to perform simple neuromorphic computations, such as recognizing handwritten digits and playing tic-tac-toe.<\/p>\n The device uses two droplets of salt water coated with lipids that are then suspended in oil, where they touch and form a bilayer that mimics a cell membrane. An electrode inserted into each droplet applies a voltage, and the current response of the droplet pair is measured.<\/p>\n The researchers found that the droplets exhibited a memory effect, producing a slightly different current depending on the voltage that was applied previously. While high voltages typically result in high current outputs, the team conducted an experiment similar to Pavlov\u2019s dog that showed that by giving the droplet system repeated training spikes of low and high voltages, they could induce high current outputs at low voltages.<\/p>\n To recognize handwritten numbers, the droplets were fed a voltage code for each pixel in the image. Due to the memory effect, each code led to a different output current that was mapped to the correct output during the training stage. After training, the droplet was able to identify further handwritten digits.<\/p>\n To play tic-tac-toe, the moves were input as voltage codes into the droplet, and the output was mapped to what move the droplet should make next. After training, the droplet system was able to tie its computer opponent consistently.<\/p>\n The researchers believe the approach is worth further investigation for energy-efficient computing. [3]<\/p>\n References<\/p>\n [1] F. Oberbauer, T.J. Winkel, T. B\u00f6hnert, et al. Magnetic tunnel junctions driven by hybrid optical-electrical signals as a flexible neuromorphic computing platform. Commun Phys 8, 329 (2025). https:\/\/doi.org\/10.1038\/s42005-025-02257-0<\/a><\/p>\n [2] A. Ismail, GH. Nam, A. Lokhandwala, et al. Programmable memristors with two-dimensional nanofluidic channels. Nat Commun 16, 7008 (2025). https:\/\/doi.org\/10.1038\/s41467-025-61649-6<\/a><\/p>\n