Page 81 - Read Online

Na et al. Soft Sci. 2026, 6, 25 Page 13 of 22

Future strategies should focus on improving photon management within soft and deformable platforms.
Incorporating microstructured surfaces, photonic crystal layers, or plasmonic nanostructures can enhance
light scattering and confinement without compromising mechanical flexibility [67,68] . In addition, hybridization
with high-absorption nanomaterials such as quantum dots (QDs) or perovskite nanoparticles could broaden
the spectral response, enabling more efficient energy harvesting from broadband illumination [69,70] . The
photoactive components with inherently high absorption coefficients and strong light-matter interactions are
an effective strategy for enhancing photon harvesting, particularly under low-intensity illumination.
Specifically, QDs offer size-tunable bandgaps that allow the absorption spectrum to be precisely engineered,
enabling enhanced light efficiency at user-selective wavelengths. In parallel, device structure engineering
plays a crucial role in enhancing light utilization efficiency by increasing the effective optical interaction
length and facilitating efficient carrier separation. In particular, the shortened channel length of a vertical
architecture can facilitate efficient charge separation, enabling higher light utilization under weak
illumination. Adapting these material and structural designs to zero-power optoelectronic synapses may
enable a stronger photoresponse and more efficient synaptic modulation under low light intensity, thereby
improving device performance under realistic wearable operating conditions. Therefore, beyond
conventional responsivity enhancement, future approaches must focus on maximizing light-matter
interaction under low-intensity and broadband illumination while remaining compatible with soft,
deformable platforms.

Insufficient adaptive learning capability
Most reported zero-power optoelectronic synapses primarily exhibit unidirectional weight modulation,
typically light-induced potentiation followed by passive relaxation. This behavior originates from the limited
capability to reversibly control carrier trapping and detrapping under zero-bias conditions, leading to
gradual saturation of the synaptic state and a loss of learning flexibility [32-39,41-45,66] . For neuromorphic
operation, however, devices must achieve hardware-level adaptive learning through both potentiation and
depression to dynamically adjust synaptic weights in response to varying optical stimuli. In addition,
on-demand resetting, facilitated by bidirectional weight modulation, is essential to restore baseline states and
prevent weight accumulation. This bidirectional modulation is particularly important in zero-power systems,
where no external electrical resetting is available, and synaptic states must be optically reconfigurable without
energy-consuming processes. In artificial neural networks, bidirectional weight modulation is fundamentally
required to enable stable, efficient, and adaptive learning [71-73] . Neuromorphic learning relies on the
coordinated strengthening and weakening of synaptic weight updates across multiple layers. When synaptic
updates occur only through potentiation, synaptic weights tend to increase excessively, leading to overfitting
and imposing fundamental limitations on learning stability, adaptability, and generalization in artificial
neural networks. Once saturation is reached, further learning becomes ineffective because synapses lose the
dynamic range required to encode new information. These limitations severely hinder learning accuracy,
error correction, and feedback. Therefore, bidirectional weight modulation is not merely a desirable device
characteristic but also a prerequisite for neuromorphic networks.

Achieving bidirectional modulation under zero-power operation requires new approaches that enable
controlled depression without external electrical bias. While conventional optoelectronic synapses in
transistor configurations typically rely on electrical gating to induce depression [74,75] , such methods
undermine the zero-power concept. Alternative strategies, such as using different optical wavelengths or
spatially patterned illumination demonstrated in bias-dependent devices [76-78] , may provide optical routes for
bidirectional weight control while maintaining energy autonomy. To achieve bidirectional weight
modulation, heterojunction-based optoelectronic synapses indicate wavelength-selective photoexcitation
pathways through distinct optical absorption spectra, energy bandgaps, and energy band alignments [40,79-81] .
The different optical properties of heterojunction semiconductors enable selective generation of dominant