Page 70 - Read Online

Page 2 of 22 Na et al. Soft Sci. 2026, 6, 25

scalability, and long-term applicability of zero-power optoelectronic synapses in realistic wearable neuromorphic
platforms. Finally, this review proposes technological strategies for addressing these challenges. We further outline
how these advances could enable practical, scalable, and mechanically compliant synaptic platforms for future
energy-autonomous, body-interfaced neuromorphic systems capable of continuous perception and intelligent
processing.

INTRODUCTION
Wearable electronics are evolving rapidly from passive sensing modules to intelligent systems capable of
perception, learning, and decision-making [1-3] . For body-integrated platforms, minimizing energy
consumption while maintaining mechanical softness is essential to ensure safe, comfortable, and continuous
operation without wired power or frequent recharging . Achieving energy-autonomous operation within
[4-6]
soft, deformable platforms therefore represents a critical milestone in the advancement of wearable
technology.

Conventional von Neumann architectures, where memory and computation are physically separated,
inherently suffer from excessive energy loss during data transfer [7-9] . Neuromorphic systems mitigate this
limitation by emulating biological synapses to enable parallel and adaptive learning with substantially
reduced power demand [10-13] . However, electrically weighted synaptic devices still face persistent challenges,
including Joule heating, static power consumption, and restricted bandwidth, which limit their suitability for
energy-constrained wearable environments [14,15] . These limitations have spurred growing interest in
optoelectronic strategies, where light functions not only as an information carrier but also as an energy
source that directly drives device operation [16-19] . Optoelectronic synapses employ this dual functionality of
light to integrate sensing, transmission, and learning within a single device, achieving ultrafast and
contactless operation [20-22] . When implemented with organic or hybrid materials, they further combine
biocompatibility, deformability, and optical responsiveness, enabling seamless integration with soft and
body-interfaced surfaces [23-26] . Despite these advantages, most existing optoelectronic synapses still rely on
wired electrical power, which restricts genuine autonomy and impedes their use in untethered,
body-interfaced platforms [27,28] .

The emerging concept of zero-power optoelectronic synapses, in which synaptic optoelectronic modulation
occurs without external electrical bias, offers a promising pathway toward energy-autonomous
neuromorphic operation. Recent studies have demonstrated that built-in potentials in Schottky and
heterojunctions, as well as photothermoelectric (PTE) effects, can drive synaptic weight modulation solely
through light energy, enabling sensing and learning without external electrical bias [29-31] . Specifically,
zero-power operation is defined here on the basis of Joule’s law (W = V·I·t, where W is the electrical energy,
V is the electrical bias, I is the peak current, and t is the spike pulse duration) in terms of electrical energy
consumption. By eliminating external electrical bias, zero-power optoelectronic synapses fundamentally
suppress electrical power consumption during synaptic operation. Although fully zero-power operation
across all device functions has not yet been realized, these demonstrations highlight the potential to
substantially reduce energy consumption during the learning stage.

Despite encouraging progress, several fundamental challenges continue to limit the practical deployment of
zero-power optoelectronic synapses in wearable neuromorphic platforms. These include limited light
utilization, insufficient bidirectional weight modulation, instability and variability, mechanical
incompatibility, and lack of system-level integration. These challenges collectively hinder reliable operation,
scalability, and long-term usability in wearable environments. Addressing these challenges requires not only
device-level optimization but also a comprehensive understanding of the trade-offs among different
operating mechanisms.