In traditional optics, photons do not directly interact with one another. For one light beam to modulate, switch, or alter another, it must change the properties of the medium through which they both travel. This light-matter interaction is governed by the material's polarization response:
$$P = \epsilon_0 \left( \chi^{(1)}E + \chi^{(2)}E^2 + \chi^{(3)}E^3 + \dots \right)$$
In standard photonic materials like silicon or silica, the higher-order nonlinear susceptibilities ($\chi^{(2)}$ and $\chi^{(3)}$) are incredibly small. To make the nonlinear terms comparable to the linear response, the electric field ($E$) of the incoming light must be exceptionally large—often approaching the scale of atomic electric fields. Historically, this has meant that achieving optical switching, bistability, or frequency conversion required high-power pulsed lasers.
This high-intensity requirement creates a fundamental bottleneck. If an optical neural network or a logic gate requires a high-power laser simply to execute an activation function, its energy footprint quickly eclipses that of conventional electronic hardware. To realize the promise of low-power, high-speed optical computing and communication, researchers are developing strategies to bypass or drastically lower this power threshold.
Enhancing the Local Field via Structural Resonance
If a material's intrinsic nonlinear coefficient is weak, one way to compensate is to artificially amplify the local electric field ($E$) without increasing the input power. This can be achieved by trapping light within resonant nanostructures, effectively forcing photons to circulate and interact with the material repeatedly over their cavity lifetime.
Felipe Herrera et al. demonstrated this concept theoretically in Quantum Nonlinear Optics with Polar J-Aggregates in Microcavities. By coupling organic dye molecules with permanent dipole moments to a confined microcavity mode, the authors showed that the interaction could be amplified to the single-photon level. In this regime, a single cavity photon alters the absorption profile of the system enough to act as a quantum switch for a secondary probe beam.
Input Light ──> [ Metasurface / Quasi-BIC ] ──> Trapped Field (E-field spikes) ──> Strong Nonlinear Output
Moving from isolated microcavities to scalable surfaces, Jie Fang et al. utilized a quasi-bound-state-in-the-continuum (quasi-BIC) architecture in Free space few-photon nonlinearity in critically coupled polaritonic metasurfaces. By engineering "critical coupling"—where the rate of light entering the metasurface perfectly matches its internal loss rate—they trapped light so efficiently that strong nonlinear modulation was achieved with an average of only six photons per cavity lifetime.
Slowing Light with Epsilon-Near-Zero Metamaterials
A second, highly elegant approach alters the linear parameters of the medium to amplify nonlinear effects. In Epsilon-Near-Zero (ENZ) metamaterials, the real part of the linear permittivity ($\epsilon$) approaches zero at a specific wavelength. This unique state unlocks two distinct physical mechanisms that lower the power threshold:
- Boundary Condition Enhancement: From Maxwell's equations, the normal component of the electric displacement field ($D = \epsilon E$) must be continuous across an interface. When light transitions from a standard medium into an ENZ medium, the drop in $\epsilon$ forces the internal electric field ($E$) to spike dramatically, amplifying nonlinear responses.
- Index Scaling: The nonlinear refractive index coefficient ($n_2$), which dictates the optical Kerr effect, is inversely proportional to the linear refractive index ($n$). Because $n = \sqrt{\epsilon}$, as $\epsilon \to 0$, the linear index vanishes, causing the effective $n_2$ to skyrocket.
This framework was explored by Monika E. Pietrzyk et al. in Epsilon near zero metamaterials for ultra-low power nonlinear applications using alternating layers of silver and silica. Experimentally, Lucia Caspani et al. confirmed this giant enhancement in aluminum-doped zinc oxide (AZO) thin films, reporting ultrafast, large-scale shifts in refractive index at accessible power levels in Enhanced Nonlinear Refractive Index in<mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline"><mml:mi>ε</mml:mi></mml:math>-Near-Zero Materials.
Furthermore, Alessandro Ciattoni et al. demonstrated in Enhanced nonlinear effects in pulse propagation through epsilon‐near‐zero media that because the phase velocity approaches infinity in an ENZ medium, electromagnetic fields vary slowly over large spatial regions. This spatial uniformity maximizes the interaction volume and time, allowing dramatic pulse transformations to occur within ultra-thin material layers.
Exploiting the Extreme Electron Mobility of Graphene
Instead of relying on structural geometry or metamaterial scaling, researchers can also turn to materials with inherently extreme electronic properties. Graphene is a prime candidate due to its linear Dirac cone dispersion relation, which allows its conduction electrons to behave as massless fermions. Because these electrons move with minimal resistance, they respond highly nonlinearly to even modest optical electric fields.
Rujiang Li et al. leveraged this behavior in Graphene induced mode bifurcation at low input power and Bistable scattering in graphene-coated dielectric nanowires. By wrapping standard dielectric nanowires in a single layer of graphene, the authors demonstrated that the extreme nonlinearity of the graphene coating enables complex light-steering effects—such as mode bifurcation and optical bistability—at power thresholds orders of magnitude lower than those required by conventional bulk dielectrics. Their modeling revealed that the ultimate bottleneck for lowering this threshold further is not the material's chemical potential, but rather the electron relaxation time within the graphene layer.
Synthetic Nonlinearity via Linear Optics
A radical alternative to material engineering is to abandon physical nonlinear interactions altogether. In Nonlinear processing with linear optics, Mustafa Yildirim et al. introduced a framework called nPOLO that synthesizes nonlinear mathematical transformations using exclusively linear optical components and low-power, continuous-wave light.
Input Data ──> [ Scattering Layer 1 ] ──┐
▲ │ (Feedback Loop)
└─ Cavity Bounce ┘
Instead of relying on high-intensity fields to change a material's index, the nPOLO system repeatedly routes and mixes input data across multiple linear scattering layers inside a multi-bounce cavity. By strategically feeding the light back onto itself, the system emulates the complex, high-order input-output mappings typically obtained via material nonlinearities. This approach completely bypasses the thermal and energy constraints of high-power pulsed lasers.
The Path to Practical Integrated Photonics
The primary driver behind these low-power breakthroughs is the push to build integrated, chip-scale optical hardware for computing and communications.
In the domain of Optical Neural Networks (ONNs), the necessity of training all-optical layers was established early on. [James Edward Steck et al. developed a continuous backpropagation framework in A Lagrangian Formulation For Optical Backpropagation Training In Kerr-Type Optical Networks, but scaling such systems was historically limited by the high optical power required to sustain the Kerr-type nonlinear layers.
Modern architectures solve this by co-integrating highly nonlinear materials directly onto silicon platforms. For example, Chuyu Zhong et al. designed a dual-functional graphene/silicon heterojunction device inside a microring resonator in Graphene/silicon heterojunction for reconfigurable phase-relevant activation function in coherent optical neural networks. This device acts simultaneously as a modulator and photodetector, providing a reconfigurable, phase-aware activation function that operates at ultra-low power.
As reviewed by Nikolay L. Kazanskiy et al. in Exploring the Frontier of Integrated Photonic Logic Gates: Breakthrough Designs and Promising Applications, overcoming the weak intrinsic nonlinearity of standard silicon platforms is the definitive hurdle for practical photonic logic gates. Similarly, for optical communications, Haipeng Zhang 0002 et al. demonstrated in Silicon Carbide Soliton Microcomb Generation for Narrow-grid Optical Communications that utilizing alternative platforms like silicon carbide can significantly lower the pump-power threshold required for generating soliton microcombs.
By shifting from brute-force laser power to sophisticated light-matter engineering, these diverse strategies are collectively bringing energy-efficient, chip-scale optical processing into reality.
Key Foundational Work
A Lagrangian Formulation For Optical Backpropagation Training In Kerr-Type Optical Networks (1994) — contributed 1 concept elements to this direction.
Low power resonant optical excitation of an optomechanical cavity (2011) — contributed 1 concept elements to this direction.