Electrothermal Optical Effect (ETMOE)

• ETMOE is a thermal-state-mediated modulation mechanism where electrical input induces Joule heating that alters optical properties via refractive index changes.
• It is implemented in diverse platforms such as silicon waveguides, GST-based devices, and PbTe metasurfaces to achieve resonance shifts, nonlinear responses, and programmable optical switching.
• Design strategies integrate heat diffusion models and thermo-optic relations to optimize performance while addressing challenges like thermal inertia and crosstalk.

Searching arXiv for the specified papers and closely related ETMOE literature to ground the article with citations. Searching arXiv for "Thermo-Optically Induced Transparency on a photonic chip (Clementi et al., 2020)" Electrothermally Modulated Optical Effect (ETMOE) denotes a class of optical modulation phenomena in which an electrical drive is converted into an optical response through an intermediate thermal state. In its strict form, an applied bias or current produces Joule heating in a microheater or resistive region; the resulting temperature field changes the complex refractive index, effective index, extinction coefficient, or cavity resonance of a photonic structure; and that thermal perturbation modulates transmission, reflection, phase, resonance position, or switching state. Recent work places ETMOE across several integrated-photonic and nanophotonic settings, including GST-based thermoreflectometry, Pt-heated silicon waveguides and racetrack resonators, and broader thermo-optic analogues in photonic-crystal cavities and dielectric metasurfaces where the same temperature-to-optical transfer physics is isolated even when the heat source is optical rather than electrical (Nobile et al., 2022, Gupta et al., 11 Jul 2025, Clementi et al., 2020, Karaman et al., 2024, Lewi et al., 2017).

1. Definition and conceptual scope

ETMOE is fundamentally distinct from direct electro-optic modulation. In a pure electro-optic modulator, an applied electric field changes refractive index directly via the Pockels or plasma-dispersion effect, generally with much higher speed and without thermal inertia. In ETMOE, the electric input first establishes a temperature excursion, and the optical device responds through the thermo-optic relation $n(T)$, $k(T)$, or the equivalent temperature dependence of cavity or resonator eigenfrequencies. The operative bandwidth is therefore governed by thermal relaxation, thermal confinement, and the optical sensitivity of the resonant structure rather than by an instantaneous field-induced susceptibility change (Clementi et al., 2020, Gupta et al., 11 Jul 2025).

This scope also requires separation from nonvolatile phase-transition behavior. In crystalline GST devices, for example, reversible optical modulation used for thermometry is intentionally thermo-optic and volatile, whereas amorphous-to-crystalline switching produces a permanent reflectance change of a different physical class. ETMOE in this sense refers to the reversible chain from electrical heating to temperature change to optical-property change, not to structural phase transformation itself, even though both may coexist in the same materials platform (Nobile et al., 2022).

A broader, ETMOE-relevant literature studies optothermal analogues. Thermo-optically induced transparency in a silicon photonic-crystal cavity and photo-driven thermo-optical nonlinearities in amorphous-silicon metasurfaces are not electrically actuated in the strict sense, but they isolate the same downstream mechanism: a delayed temperature field reshapes an optical susceptibility or resonance and produces coherent or nonlinear optical modulation. This suggests that ETMOE is best understood not merely as “heating-based tuning,” but as a thermal-state-mediated optical transfer process whose actuator may be electrical, optical, or externally imposed, while the optical consequences remain thermo-optic (Clementi et al., 2020, Karaman et al., 2024).

2. Physical mechanism and governing relations

The canonical ETMOE chain is

$$\text{electrical input} \rightarrow \text{Joule heating} \rightarrow T(\mathbf r,t) \rightarrow \tilde n(T)=n(T)+ik(T) \rightarrow \text{optical observable}.$$

For electrically driven phase-change devices, the heating power may be written as

$P = I^2R = VI = \frac{V^2}{R},$

and the corresponding volumetric heat source as

$$Q = \mathbf{J}\cdot \mathbf{E} = \sigma |\mathbf{E}|^2 = \frac{|\mathbf{J}|^2}{\sigma}.$$

The temperature field then obeys the heat diffusion equation

$$\rho C_p \frac{\partial T}{\partial t} = \nabla \cdot (k \nabla T) + Q,$$

while the optical transduction layer follows, to first order,

$$n(T)=n(T_0)+\beta (T-T_0), \qquad k(T)=k(T_0)+\gamma (T-T_0).$$

In GST thermoreflectometry, the final observable is reflectance, computed as $R(T)$ by transfer-matrix optics for the multilayer stack; at $637$ nm it is approximately linear in temperature, so $$\Delta R \approx \left(\frac{dR}{dT}\right)\Delta T$$ (Nobile et al., 2022).

In silicon waveguides and resonators, the same thermal perturbation is cast as a refractive-index or effective-index update inside a coupled electrical-thermal-optical model. The electrical conduction and Joule-heating relations are written as

$k(T)$0

$k(T)$1

and the heat-transfer problem as

$k(T)$2

The thermo-optic material update is expressed as

$k(T)$3

and the resonator resonance condition as

$k(T)$4

Heating therefore changes $k(T)$5, shifts $k(T)$6, and modulates transmission or switch state (Gupta et al., 11 Jul 2025).

A more elaborate thermal-mediated mechanism appears in thermo-optically induced transparency. There, a strong control field and weak probe beat inside a cavity, generating an oscillating heat source. The thermal mode is described by

$k(T)$7

$k(T)$8

The optical response is thus not driven by a direct electric-field susceptibility, but by a delayed thermal feedback channel that modulates cavity detuning and generates interference between a probe and thermally induced sidebands (Clementi et al., 2020).

3. Representative material platforms and device architectures

ETMOE has been realized, or closely approximated, in several distinct photonic architectures.

Platform	Thermal actuation	Optical manifestation
Crystalline GST on Pt or doped-Si microheaters	Electrical Joule heating	Reflectance thermomodulation and temperature mapping
SOI waveguide-integrated racetrack resonator with Pt heater	Electrical Joule heating	Resonance shift and through/drop switching
PbTe Mie resonators and metasurfaces	External thermal stage	Thermo-refractive resonance tuning
a-Si dielectric metasurfaces	Optical photothermal pumping	Nonlinear transmission modulation
Silicon PhC cavity	Optical absorption-induced heating	Thermo-optically induced transparency or amplification

In GST thermoreflectometry, the optical stack is optimized to maximize $k(T)$9 at $$\text{electrical input} \rightarrow \text{Joule heating} \rightarrow T(\mathbf r,t) \rightarrow \tilde n(T)=n(T)+ik(T) \rightarrow \text{optical observable}.$$0 nm and consists of $$\text{electrical input} \rightarrow \text{Joule heating} \rightarrow T(\mathbf r,t) \rightarrow \tilde n(T)=n(T)+ik(T) \rightarrow \text{optical observable}.$$1 nm SiO$$\text{electrical input} \rightarrow \text{Joule heating} \rightarrow T(\mathbf r,t) \rightarrow \tilde n(T)=n(T)+ik(T) \rightarrow \text{optical observable}.$$2, $$\text{electrical input} \rightarrow \text{Joule heating} \rightarrow T(\mathbf r,t) \rightarrow \tilde n(T)=n(T)+ik(T) \rightarrow \text{optical observable}.$$3 nm GST, and $$\text{electrical input} \rightarrow \text{Joule heating} \rightarrow T(\mathbf r,t) \rightarrow \tilde n(T)=n(T)+ik(T) \rightarrow \text{optical observable}.$$4 nm SiO$$\text{electrical input} \rightarrow \text{Joule heating} \rightarrow T(\mathbf r,t) \rightarrow \tilde n(T)=n(T)+ik(T) \rightarrow \text{optical observable}.$$5. The work compares a Pt microheater with a $$\text{electrical input} \rightarrow \text{Joule heating} \rightarrow T(\mathbf r,t) \rightarrow \tilde n(T)=n(T)+ik(T) \rightarrow \text{optical observable}.$$6 nm Pt layer and an active area of about $$\text{electrical input} \rightarrow \text{Joule heating} \rightarrow T(\mathbf r,t) \rightarrow \tilde n(T)=n(T)+ik(T) \rightarrow \text{optical observable}.$$7, and a doped-silicon microheater with active area about $$\text{electrical input} \rightarrow \text{Joule heating} \rightarrow T(\mathbf r,t) \rightarrow \tilde n(T)=n(T)+ik(T) \rightarrow \text{optical observable}.$$8, $$\text{electrical input} \rightarrow \text{Joule heating} \rightarrow T(\mathbf r,t) \rightarrow \tilde n(T)=n(T)+ik(T) \rightarrow \text{optical observable}.$$9 nm device-layer thickness, and $P = I^2R = VI = \frac{V^2}{R},$0m buried oxide. The thicker oxide and smaller active area make the silicon heater much more thermally efficient and much faster (Nobile et al., 2022).

In silicon ETMOE resonators, the reported architecture is an SOI platform with a $P = I^2R = VI = \frac{V^2}{R},$1 nm Si core, SiO$P = I^2R = VI = \frac{V^2}{R},$2 cladding, a Pt microheater, a Ti adhesion layer, Au pads, and W vias. The selected waveguide width is $P = I^2R = VI = \frac{V^2}{R},$3 nm, below the reported $P = I^2R = VI = \frac{V^2}{R},$4 nm single-TE$P = I^2R = VI = \frac{V^2}{R},$5-mode cutoff width. For the straight-waveguide heater study, the chosen Pt dimensions are $P = I^2R = VI = \frac{V^2}{R},$6 nm thickness, $P = I^2R = VI = \frac{V^2}{R},$7 nm width, $P = I^2R = VI = \frac{V^2}{R},$8 nm length, and $P = I^2R = VI = \frac{V^2}{R},$9 nm heater-to-waveguide separation. The resonator implementation is a double-bus racetrack resonator with a symmetric sectored Pt heater, chosen specifically to mitigate asymmetric heat distribution (Gupta et al., 11 Jul 2025).

PbTe meta-atoms extend the ETMOE concept into the mid-infrared. Although the experiments use a thermal stage rather than integrated resistive heaters, the relevant chain remains temperature-driven refractive-index tuning of resonant elements. PbTe is notable for $$Q = \mathbf{J}\cdot \mathbf{E} = \sigma |\mathbf{E}|^2 = \frac{|\mathbf{J}|^2}{\sigma}.$$0 in the MIR and a large negative thermo-optic coefficient, with room-temperature $$Q = \mathbf{J}\cdot \mathbf{E} = \sigma |\mathbf{E}|^2 = \frac{|\mathbf{J}|^2}{\sigma}.$$1 and stronger low-temperature values such as $$Q = \mathbf{J}\cdot \mathbf{E} = \sigma |\mathbf{E}|^2 = \frac{|\mathbf{J}|^2}{\sigma}.$$2 near $$Q = \mathbf{J}\cdot \mathbf{E} = \sigma |\mathbf{E}|^2 = \frac{|\mathbf{J}|^2}{\sigma}.$$3 K. The demonstrated structures include spheres, cubes on Si or Au, and periodic metasurfaces with $$Q = \mathbf{J}\cdot \mathbf{E} = \sigma |\mathbf{E}|^2 = \frac{|\mathbf{J}|^2}{\sigma}.$$4 values up to $$Q = \mathbf{J}\cdot \mathbf{E} = \sigma |\mathbf{E}|^2 = \frac{|\mathbf{J}|^2}{\sigma}.$$5 in simulated reflectarray configurations (Lewi et al., 2017).

In dielectric metasurfaces of amorphous silicon on fused silica, thermo-optic modulation is realized with a $$Q = \mathbf{J}\cdot \mathbf{E} = \sigma |\mathbf{E}|^2 = \frac{|\mathbf{J}|^2}{\sigma}.$$6 nm PECVD a-Si layer patterned into nanodisks of periodicity $$Q = \mathbf{J}\cdot \mathbf{E} = \sigma |\mathbf{E}|^2 = \frac{|\mathbf{J}|^2}{\sigma}.$$7 nm, diameter $$Q = \mathbf{J}\cdot \mathbf{E} = \sigma |\mathbf{E}|^2 = \frac{|\mathbf{J}|^2}{\sigma}.$$8 nm, and height $$Q = \mathbf{J}\cdot \mathbf{E} = \sigma |\mathbf{E}|^2 = \frac{|\mathbf{J}|^2}{\sigma}.$$9 nm. The measured transmission spectrum exhibits a magnetic-dipole resonance near $$\rho C_p \frac{\partial T}{\partial t} = \nabla \cdot (k \nabla T) + Q,$$0 nm and an electric-dipole resonance near $$\rho C_p \frac{\partial T}{\partial t} = \nabla \cdot (k \nabla T) + Q,$$1 nm with $$\rho C_p \frac{\partial T}{\partial t} = \nabla \cdot (k \nabla T) + Q,$$2. Although the heat source is a $$\rho C_p \frac{\partial T}{\partial t} = \nabla \cdot (k \nabla T) + Q,$$3 nm optical pump, the optical transfer from $$\rho C_p \frac{\partial T}{\partial t} = \nabla \cdot (k \nabla T) + Q,$$4 to $$\rho C_p \frac{\partial T}{\partial t} = \nabla \cdot (k \nabla T) + Q,$$5 to transmission directly parallels ETMOE (Karaman et al., 2024).

4. Dynamical regimes and spectral signatures

The most elementary ETMOE signature is steady-state resonance tuning. In the Pt-heated silicon racetrack resonator, applied voltages of $$\rho C_p \frac{\partial T}{\partial t} = \nabla \cdot (k \nabla T) + Q,$$6, $$\rho C_p \frac{\partial T}{\partial t} = \nabla \cdot (k \nabla T) + Q,$$7, and $$\rho C_p \frac{\partial T}{\partial t} = \nabla \cdot (k \nabla T) + Q,$$8 mV produce resonance shifts of $$\rho C_p \frac{\partial T}{\partial t} = \nabla \cdot (k \nabla T) + Q,$$9, $$n(T)=n(T_0)+\beta (T-T_0), \qquad k(T)=k(T_0)+\gamma (T-T_0).$$0, and $$n(T)=n(T_0)+\beta (T-T_0), \qquad k(T)=k(T_0)+\gamma (T-T_0).$$1 nm, associated with average shadowed-waveguide temperature rises of $$n(T)=n(T_0)+\beta (T-T_0), \qquad k(T)=k(T_0)+\gamma (T-T_0).$$2, $$n(T)=n(T_0)+\beta (T-T_0), \qquad k(T)=k(T_0)+\gamma (T-T_0).$$3, and $$n(T)=n(T_0)+\beta (T-T_0), \qquad k(T)=k(T_0)+\gamma (T-T_0).$$4 K. In the double-bus switch, the state at about $$n(T)=n(T_0)+\beta (T-T_0), \qquad k(T)=k(T_0)+\gamma (T-T_0).$$5 nm changes from through-port transmission $$n(T)=n(T_0)+\beta (T-T_0), \qquad k(T)=k(T_0)+\gamma (T-T_0).$$6 and drop-port transmission $$n(T)=n(T_0)+\beta (T-T_0), \qquad k(T)=k(T_0)+\gamma (T-T_0).$$7 at $$n(T)=n(T_0)+\beta (T-T_0), \qquad k(T)=k(T_0)+\gamma (T-T_0).$$8 mV to through $$n(T)=n(T_0)+\beta (T-T_0), \qquad k(T)=k(T_0)+\gamma (T-T_0).$$9 and drop $R(T)$0 at $R(T)$1 mV; the $R(T)$2 mV state corresponds to an average shadowed-waveguide temperature rise of $R(T)$3 K (Gupta et al., 11 Jul 2025).

Time-domain behavior reveals the thermal bottleneck more explicitly. In GST microheaters, the doped-silicon design yields single-exponential heating and cooling constants of $R(T)$4 and $R(T)$5, whereas the Pt heater is much slower, with $R(T)$6 and $R(T)$7 in a single-exponential fit, and a more complex multi-time-constant interpretation when substrate heating is included. The same work shows that the doped-silicon platform is roughly $R(T)$8 more efficient than the Pt heater at steady state when plotted versus power density (Nobile et al., 2022).

Thermal systems can also generate narrow spectral features that are much sharper than the host optical linewidth. In a silicon photonic-crystal cavity, the optical resonance has linewidth $R(T)$9, while the nominal thermal relaxation rate is $637$0. Pump–probe beating creates an oscillating temperature field within this narrow thermal bandwidth, and the resulting thermal sideband interferes with the probe to produce either induced absorption or induced amplification. The measured maximum group delay is $637$1 with $637$2, while the maximum group advance is $637$3 with loss $637$4 and bandwidth $637$5. Although optothermal rather than electrical, this establishes that a temperature field can act as a coherent intermediate channel for optical susceptibility shaping (Clementi et al., 2020).

A second nontrivial regime is nonlinear transfer-function engineering in resonant metasurfaces. In amorphous-silicon metasurfaces, the $637$6 nm transmission first decreases by about $637$7 and then increases by about $637$8 as pump intensity rises to $637$9, because the electric-dipole resonance redshifts through the probe wavelength. Under $$\Delta R \approx \left(\frac{dR}{dT}\right)\Delta T$$0 kHz pump modulation, the optical characteristic time can collapse to $$\Delta R \approx \left(\frac{dR}{dT}\right)\Delta T$$1 while the thermal characteristic time remains $$\Delta R \approx \left(\frac{dR}{dT}\right)\Delta T$$2, and the optical output can be modulated at $$\Delta R \approx \left(\frac{dR}{dT}\right)\Delta T$$3 kHz, twice the excitation frequency. The validated model further projects optical modulation at MHz speeds with amplitudes up to $$\Delta R \approx \left(\frac{dR}{dT}\right)\Delta T$$4 (Karaman et al., 2024).

PbTe resonators emphasize sensitivity rather than transient speed. A high-$$\Delta R \approx \left(\frac{dR}{dT}\right)\Delta T$$5 PbTe sphere with $$\Delta R \approx \left(\frac{dR}{dT}\right)\Delta T$$6 at $$\Delta R \approx \left(\frac{dR}{dT}\right)\Delta T$$7 shifts by more than one linewidth for $$\Delta R \approx \left(\frac{dR}{dT}\right)\Delta T$$8 K, while simulated metasurfaces show more than one-linewidth tuning for $$\Delta R \approx \left(\frac{dR}{dT}\right)\Delta T$$9 K around $k(T)$00 K or $k(T)$01 K around room temperature, and $k(T)$02 phase shift for $k(T)$03 K around $k(T)$04 K while maintaining reflectivity $k(T)$05. This suggests that ETMOE performance is not set only by thermal actuator speed, but also by how much optical leverage a given $k(T)$06 produces in a high-$k(T)$07 spectral feature (Lewi et al., 2017).

5. Modeling strategies and thermal-response engineering

ETMOE is inherently multiphysics. The most complete formulation among the cited works is a fully $k(T)$08D electronic-photonic co-integrated framework in COMSOL Multiphysics using Electric Currents, Heat Transfer in Solids, and Electromagnetic Waves, Frequency Domain. Its stated novelty is nonlinear numerical coupling of temperature-dependent electrical conductivity, temperature-dependent thermal properties, and temperature- and wavelength-dependent optical material properties. For numerical resolution at $k(T)$09 nm, the recommended mesh sizes are $k(T)$10 for the minimum and $k(T)$11 for the maximum, with finer meshing near the heater and waveguide (Gupta et al., 11 Jul 2025).

A complementary strategy appears in GST thermoreflectometry, where temperature is inferred optically rather than solved only numerically. The calibration chain is: measure GST thermo-optic coefficients $k(T)$12 and $k(T)$13 by ellipsometry, use transfer-matrix modeling to calculate $k(T)$14, and then invert the measured reflected signal through the approximately linear $k(T)$15 relation at $k(T)$16 nm. The ellipsometric calibration uses a single Tauc–Lorentz dispersion model, and the thermal COMSOL model includes thermal contact resistances of $k(T)$17 for Si/SiO$k(T)$18, $k(T)$19 for Si/Al, and $k(T)$20 for GST/SiO$k(T)$21 (Nobile et al., 2022).

Thermal engineering is frequently as important as optical design. In the silicon racetrack ETMOE study, stable confinement requires heater separation of $k(T)$22 nm or greater; best confinement is reported near $k(T)$23 nm, but $k(T)$24 nm is chosen as the practical point because it sacrifices only about $k(T)$25 power confinement relative to the best case while improving fabrication robustness. The same work shows attenuation decreasing from $k(T)$26 at $k(T)$27 nm separation to $k(T)$28 at $k(T)$29 nm, illustrating the classical ETMOE trade-off between optical perturbation and thermal coupling (Gupta et al., 11 Jul 2025).

Distributed thermal networks rather than single thermal time constants are often required. In thermo-optically induced transparency, a refined model with up to three coupled heat capacities is used, with fitted decay rates $k(T)$30, $k(T)$31, and $k(T)$32, bracketing the single-rate estimate $k(T)$33. The same work explicitly argues that heat-flow engineering can be decoupled from optical design, and discusses adding materials such as graphene to increase heat dissipation or structuring the surroundings and support bridges to reduce thermal decay by more than an order of magnitude (Clementi et al., 2020).

6. Applications, advantages, constraints, and unresolved issues

ETMOE is being developed for reconfigurable photonic systems, optical interconnects, programmable photonic networks, neuromorphic photonics, tunable metasurfaces, optical switching, and wavefront-control elements. The silicon racetrack study explicitly frames ETMOE as a route toward energy-efficient, programmable photonic systems. GST thermoreflectometry targets electrically programmable phase-change devices and provides a non-invasive route to verify thermal trajectories, spatial uniformity, and cooling rates. PbTe resonators and metasurfaces indicate a path toward active notch filters, reflective phase shifters, and reconfigurable mid-IR metasurfaces with few-kelvin thermal excursions. Amorphous-silicon metasurfaces show that resonant thermo-optic nonlinearities can support not only amplitude modulation but also frequency multiplication and fast transient shaping (Gupta et al., 11 Jul 2025, Nobile et al., 2022, Lewi et al., 2017, Karaman et al., 2024).

Several recurrent advantages follow from this thermal mediation. First, ETMOE is platform-flexible: it does not require intrinsic Pockels materials, carrier injection, atomic resonances, or sideband-resolved optomechanics. Second, the optical leverage can be large even at moderate $k(T)$34, as shown by TOIT in a cavity with $k(T)$35 and by racetrack resonators with $k(T)$36. Third, heater geometry, oxide thickness, suspended membranes, and surrounding thermal conductance can often be engineered with greater freedom than in mechanisms where the optical and dynamical resonances are inseparable (Clementi et al., 2020, Gupta et al., 11 Jul 2025).

The principal limitations are equally consistent across platforms. Thermal inertia remains the dominant speed bottleneck in directly heated devices, with a $k(T)$37 fall time in the Pt-heated silicon waveguide and microsecond-to-tens-of-microseconds behavior in less confined heaters. Thermal crosstalk, parasitic substrate heating, resonance overlap at larger tuning voltages, and temperature nonuniformity remain practical constraints. Optically driven analogues add absorption dependence and possible onset of unwanted nonlinear absorption at higher powers. Phase-change platforms must also remain below unwanted phase-transition thresholds during volatile ETMOE operation (Gupta et al., 11 Jul 2025, Nobile et al., 2022).

Several open questions remain. PbTe exhibits a very large low-temperature increase in $k(T)$38 that standard thermo-optic models do not fully explain, and the authors explicitly suggest that unknown physical mechanisms may be involved. The most dramatic optical-speed enhancements demonstrated in metasurfaces come from photothermal pumping rather than integrated heaters, so electrical implementations must still establish how electrode design, RC constraints, and thermal spreading affect the same nonlinear transfer-function benefits. Likewise, thermo-optically induced transparency is a strong source paper for ETMOE physics but does not demonstrate electrical drive; its relevance lies in thermal-response engineering and coherent thermal susceptibility shaping rather than in heater hardware (Lewi et al., 2017, Karaman et al., 2024, Clementi et al., 2020).

Taken together, the literature defines ETMOE as a thermal-state-mediated optical control mechanism whose essential variables are heater power, thermal transport, thermo-optic material response, and resonant optical sensitivity. Its central design problem is not merely generating heat, but engineering the full transfer chain from $k(T)$39 or $k(T)$40 to $k(T)$41, and from $k(T)$42 to a spectrally selective optical response. The most consequential recent result is that this transfer chain can be exploited not only for static tuning but also for coherent interference, narrowband delay or advance, nonlinear transmission shaping, and even modulation rates that exceed naive thermal-time-constant expectations when the optical response is engineered around steep or non-monotonic thermo-optic resonances (Clementi et al., 2020, Karaman et al., 2024).