Page 82 - Read Online

Page 14 of 22 Na et al. Soft Sci. 2026, 6, 25

charge carrier in one semiconductor layer, depending on the incident light wavelength. This
wavelength-dependent carrier selectivity directly enables bidirectional weight modulation by controlling the
transport of photogenerated carriers. Extending this design principle to zero-power optoelectronic synapses
suggests a practical pathway toward stable bidirectional learning without external electrical bias. The
consideration of the optical properties of semiconductor pairs and careful engineering of heterojunction
interfaces, wavelength-selective absorption can be directly translated into controlled carrier dynamics,
thereby improving the precision and repeatability of synaptic weight updates.

Additionally, conventional neuromorphic computing primarily relies on electrically driven conductance
modulation for learning [82-84] . However, in zero-power optoelectronic synapses, the absence of external
electrical bias precludes conductance-based neuromorphic computing. Thus, the development of new
learning paradigms and system architectures for zero-power optoelectronic synapses is essential. Although
these approaches remain in their early stages, exploring such optical pathways will be essential for realizing
fully adaptive and energy-efficient zero-power neuromorphic devices.

Instability and variability
Ensuring reliable and consistent operation remains one of the most critical challenges for zero-power
optoelectronic synapses. Under repeated optical stimulation or prolonged environmental exposure, devices
often exhibit drift in conductance states, inconsistent cycle-to-cycle responses, and noticeable
device-to-device variation [85,86] . In realistic wearable environments, zero-power optoelectronic synapses must
operate reliably under continuous exposure to moisture, sweat, and UV light in daily human life. Ensuring
stable device performance under sweat and high-humidity conditions requires encapsulation to protect the
device from moisture, hydrated complexes inducing irreversible performance degradation. In addition,
photonic safety and feasibility under continuous exposure, especially UV light, must be carefully considered
to ensure long-term operational stability in wearable applications.

This instability not only undermines reproducibility at the individual device level but also becomes
increasingly problematic when multiple synapses are integrated into array networks, where cumulative
variability can severely degrade overall system performance. Unlike conventional electrically driven synapses,
where active biasing can assist stabilization, zero-power synapses rely solely on passive photonic or ionic
mechanisms. This makes them more susceptible to instability caused by trap dynamics, environmental
fluctuations, and mechanical perturbations. Minor differences in material composition, processing history,
or interface quality can therefore translate into large disparities in the synaptic response, even among
nominally identical devices. Such sensitivity to subtle variations poses a serious challenge for large-area
fabrication and scaling, where maintaining precise control over material uniformity and interfacial quality is
inherently difficult. In wearable environments, this challenge is further exacerbated by continuously
changing temperature, humidity, and mechanical strain. As a result, instability and variability represent not
only device-level concerns but also system-level obstacles that fundamentally limit the reliability, scalability,
and long-term stability of zero-power optoelectronic synapses in practical wearable neuromorphic platforms.

Future development should prioritize stability throughout the entire device configuration from active
materials to interfacial contact. Robust material systems with intrinsic chemical and thermal stability, as well
as self-regulated trap-state dynamics under environmental and temporal fluctuations, are required to sustain
consistent performance over time [54,87,88] . In parallel, encapsulation technologies must advance beyond
conventional protective layers to provide not only high environmental resistance but also mechanical
compliance and optical transparency. Moreover, suppressing device-to-device variability is essential for
ensuring reproducible synaptic behavior, particularly when zero-power optoelectronic synapses are
integrated into large-scale arrays. The device variability induced by film thickness, interfacial quality, defect