image: Illustration of the similarity between a CuMnAs-based antiferromagnetic analog memory device and a biological synapse. a, In a biological synapse, the presynaptic and postsynaptic neurons are separated by a 20 to 40 nm thick gap, called a synaptic cleft. When presynaptic neurons are activated, they release neurotransmitters into the synaptic cleft, which then bind to receptors in a postsynaptic neuron and cause it to become active. Since the release of neurotransmitters is initiated by the intracellular calcium concentration (Ca2+), the complex kinetics of Ca2+ leads to a transient change in the synaptic weight, working as a short-term memory (STM). b, The increase of a number of postsynaptic receptors (shown in green) on longer time scales leads to long-term memory (LTM). c and d, Schematic representation of a CuMnAs-based memory device in which the ultrafast change in reflectivity and heat dissipation from the CuMnAs layer (red arrows) after excitation by a single femtosecond laser pulse (red column) play the role of STM and the device resistance increase due to magnetic switching plays the role of LTM. e, Multistore psychological model of human memory, which illustrates that information from a "sensor" is transferred from (STM) to (LTM) only after frequent rehearsal. f, In the device, laser pulse bursts are used as input stimuli and absorption in CuMnAs serves as the sensor. Note that pulse widths of 150 fs are strongly exaggerated with respect to their time spacing of 700 ps for clarity in f (see the inset). g, Schematic depiction of a transient increase of effective CuMnAs temperature after absorption of the shown number of laser pulses; at room temperature (RT) the required temperature threshold (TTH) for achieving the switching is overpassed only for 8 and 16 pulses. h, Measured resistivity increase DR in a device made from a 20‑nm‑thick CuMnAs epilayer, which shows that the number of rehearsals for information to be transferred from STM do LTM has to be at least 8 within the used 10‑ns‑long memorizing period.
Credit: Prof. Petr Nemec, petr.nemec@matfyz.cuni.cz
As electronic devices generate increasing amounts of waste heat, interest in using heat as a computational medium is growing. In this work, the response of an analog memory device made from thin film of the antiferromagnetic metal CuMnAs to bursts of heat pulses generated by the absorption of femtosecond laser pulses at room temperature was investigated. When the accumulated temperature in the device’s heat-based short-term memory exceeds a threshold, the output of in-memory logic operations is transferred within the same device to long-term memory, mimicking human memory, where repeated rehearsals are needed to transfer information from short-term to long-term memory. The long-term memory relies on magnetoresistive switching from a reference low-resistive uniform magnetic state to high-resistive metastable nanofragmented magnetic states, where information is stored and electrically readable at macroscopic timescales. Remarkably, this entire process – from logic operations to memory transfer – occurs on sub-nanosecond timescales, making it compatible with the GHz frequencies of standard electronics. Rapid resetting of the long-term memory using heat pulses was also demonstrated, completing a full write-read-reset cycle of this bio-inspired memory device entirely within the thermal domain.
Journal
PhotoniX
Method of Research
Experimental study
Subject of Research
Not applicable
Article Title
Sub-nanosecond heat-based logic, writing and reset in an antiferromagnetic magnetoresistive memory
Article Publication Date
4-Nov-2025
COI Statement
The authors declare no competing interests.