Thermo-Optic Reconfiguration in Effective Media

• Thermo-optic reconfiguration of effective media is a phenomenon where temperature-induced changes in refractive properties collectively tune and modify the resonant behavior of composite photonic systems.
• It employs mechanisms such as modal averaging, collective resonance detuning, and retarded self-interactions to achieve precise control over resonance wavelength, phase, and field profiles.
• The approach underpins advances in temperature-invariant metasurfaces, thermally tunable meta-atoms, and ENZ metamaterials, paving the way for both athermal stabilization and dynamic optical switching.

Searching arXiv for the cited papers to ground the article in current preprints. Thermo-optic reconfiguration of effective media denotes the class of phenomena in which temperature-dependent changes in refractive index, absorption, thermal conductivity, constitutive parameters, or geometry alter the effective optical response of a composite photonic system rather than merely perturbing an otherwise fixed device parameter. In this view, heating can shift, suppress, enhance, or cancel resonant behavior; reweight modal overlap among constituents with opposite thermo-optic dispersions; reshape collective resonances in nonlocal metasurfaces; modify retardation kernels in photon Bose-Einstein condensates; or move the epsilon-near-zero condition of a metamaterial. The topic spans temperature-invariant metasurfaces, thermally tunable Mie meta-atoms, q-BIC/Fano meta-optics, graphene-oxide-integrated microrings, dye-filled microcavities, ENZ multilayers, and thermo-optic metrology in thick semi-transparent media (Cohen et al., 2023, Wu et al., 28 Apr 2026).

1. Conceptual basis and scope

The central thermodynamic variable is the thermo-optic response, commonly expressed through the refractive-index variation with temperature, $dn/dT$. In resonant nanophotonics this parameter is amplified by modal confinement and by the narrow linewidth of high-$Q$ resonances, so small temperature excursions can drift resonance wavelength, amplitude, phase, and field distributions. A key conceptual development in the recent literature is that the relevant quantity is often not the intrinsic $dn/dT$ of a single constituent but the effective response of a hybrid optical mode, an array-supported collective resonance, or a homogenized metamaterial (Cohen et al., 2023, Malek et al., 2020, Wu et al., 28 Apr 2026).

In hybrid subwavelength resonators, the effective-medium target is explicit:

$$\frac{dn_{\mathrm{eff}}}{dT}\approx 0, \qquad \frac{d\lambda}{dT}\approx 0.$$

In integrated microrings, the same logic appears through the resonance condition

$$n_{\mathrm{eff}}\times \frac{2\pi}{\lambda}\times L = m\times 2\pi,$$

so any temperature-driven change in the mode-overlapped effective index shifts the device resonance. In ENZ systems, temperature reconfigures the constitutive parameters that define the effective medium itself, so the ENZ wavelength is displaced as the metal and dielectric layers change permittivity and thickness (Moss, 2024, Wu et al., 28 Apr 2026).

The literature therefore supports a broad definition in which “effective medium” may refer to a homogenized multilayer, a mode-weighted composite resonator, a periodic metasurface with collective band structure, or a photon fluid whose self-induced temperature field mediates an effective interaction kernel. The common element is that thermal variation acts on a composite optical entity.

The following representative manifestations recur across the field (Cohen et al., 2023, Lewi et al., 2017, Moss, 2024, Wu et al., 28 Apr 2026, Stein et al., 2022):

Platform	Reconfigured quantity	Representative outcome
Hybrid Si/PbTe metasurfaces	Effective thermo-optic coefficient	$\lvert S_{\max}\rvert \le 0.001$, including $\Delta\lambda=0$ over 293–393 K
PbTe Mie resonators and metasurfaces	Resonance wavelength and phase	More than 8 linewidths tuning; full $2\pi$ phase shift with greater than 99% reflectivity
GO-coated SiN microrings	Effective index, loss, $dn/dT$, thermal conductivity	$dn/dT$ changes sign; thermal conductivity rises from about 20.6 to about 998 W/(m·°C)
Ag/SiO$Q$0 ENZ multilayers	Constitutive parameters defining ENZ	$Q$1 and $Q$2
Dye-filled microcavities	Effective photon-photon interaction	Retarded, nonlocal thermo-optic interaction determined by diffusion and history

2. Governing mechanisms of thermal reconfiguration

A first mechanism is modal averaging across materials with opposite thermo-optic dispersions. “Temperature Invariant Metasurfaces” uses silicon with $Q$3 and PbTe with $Q$4 so that geometry and field overlap are tuned until the temperature-induced red shift in silicon is compensated by the opposite shift in PbTe. The hybrid resonator is treated as a composite optical medium whose net thermal coefficient is determined by weighted constituent contributions and resonant field overlap (Cohen et al., 2023).

A second mechanism is collective resonance detuning in nonlocal metasurfaces. In “Active Nonlocal Metasurfaces,” the relevant mode is not an isolated meta-atom resonance but a q-BIC or Fano resonance supported by many adjacent meta-units. Because silicon has a thermo-optic coefficient of about $Q$5 near telecom wavelengths, thermal refractive-index tuning shifts the collective resonance. The shift is amplified by the large optical lifetime of the q-BIC, and the device can thereby modulate transmission, polarization conversion, or wavefront shaping at a fixed illumination wavelength (Malek et al., 2020).

A third mechanism is retarded medium-mediated self-interaction. In dye-filled microcavities, photons heat the dye solution through imperfect absorption and re-emission cycles, and the temperature rise changes refractive index according to

$Q$6

“Hartree-Fock Analogue Theory of Thermo-Optic Interaction” formulates this through a second-quantized Hamiltonian coupled to a temperature field,

$Q$7

with temperature obeying a diffusion equation sourced by photon density,

$Q$8

Substitution yields a nonlocal, retarded effective interaction with memory, rather than an instantaneous contact term. “Photon BEC with Thermo-Optic Interaction at Dimensional Crossover” then shows that this interaction is itself geometry dependent: as trap anisotropy increases from 2D toward quasi-1D, the thermo-optic interaction first increases and later saturates once the condensate becomes narrower than the thermal diffusion length (Stein et al., 2022, Stein et al., 2021).

A fourth mechanism is temperature-driven reconfiguration of homogenized constitutive parameters. In ENZ media, temperature changes both $Q$9 and layer thicknesses $dn/dT$0, so the medium transforms from $dn/dT$1 to $dn/dT$2. This is the core physical picture advanced in “Thermo-optic dynamics of effective epsilon-near-zero media” (Wu et al., 28 Apr 2026).

3. Temperature-invariant resonators and metasurfaces

A prominent branch of the field uses thermo-optic reconfiguration not for tuning but for suppression of thermal drift. “Temperature Invariant Metasurfaces” develops a systematic route to zero effective thermo-optic response by designing hybrid subwavelength resonators composed of Si/PbTe/Si layers in spheres, disks, cubes, and periodic arrays. The operational signature is a nearly flat resonance wavelength versus temperature and preserved amplitude and phase spectra across large thermal excursions (Cohen et al., 2023).

The reported temperature span is broad. Most invariance is demonstrated over $dn/dT$3, corresponding to a 500 K swing. For the 11-layer 1D Bragg mirror built from alternating PbTe and Si, the central resonance at $dn/dT$4 shows essentially no spectral shift across the 500 K span while maintaining very high reflectivity $dn/dT$5. The same design logic is extended to 3D multilayer spherical Mie resonators, in which the first three modes—magnetic dipole, electric dipole, and magnetic quadrupole—remain fixed in wavelength, and the phase and amplitude spectra at 143 K and 643 K nearly overlap (Cohen et al., 2023).

The paper quantifies thermal drift through the normalized figure of merit

$dn/dT$6

For the sphere, $dn/dT$7, with a particularly optimized MQ mode showing $dn/dT$8 over 193 K to 343 K. For the cubic meta-atom, the fundamental mode has $dn/dT$9, and for the disk $$\frac{dn_{\mathrm{eff}}}{dT}\approx 0, \qquad \frac{d\lambda}{dT}\approx 0.$$0. In the metasurface implementation, the disk array maintains the resonance frequency, amplitude, and phase of its fundamental modes across the thermal sweep, with $$\frac{dn_{\mathrm{eff}}}{dT}\approx 0, \qquad \frac{d\lambda}{dT}\approx 0.$$1, including a case of “perfect” invariance where $$\frac{dn_{\mathrm{eff}}}{dT}\approx 0, \qquad \frac{d\lambda}{dT}\approx 0.$$2 and $$\frac{dn_{\mathrm{eff}}}{dT}\approx 0, \qquad \frac{d\lambda}{dT}\approx 0.$$3 over 293–393 K (Cohen et al., 2023).

An important technical point is that the invariance is not limited to resonance position. The field profiles are reported to remain essentially unchanged with temperature, which indicates preservation of the same mode rather than a thermally induced mode conversion. The paper also emphasizes preserved $$\frac{dn_{\mathrm{eff}}}{dT}\approx 0, \qquad \frac{d\lambda}{dT}\approx 0.$$4 phase coverage in the metasurface. This matters because in resonant meta-optics the device function depends jointly on spectral alignment, amplitude response, and phase response.

A common misconception is that thermo-optic design is intrinsically a tuning strategy. These results establish the opposite use-case: by controlling the sign and magnitude of thermo-optic dispersion at the mode level, the effective medium can be made thermally “neutral” within a specified window (Cohen et al., 2023).

4. Thermally tunable meta-atoms and nonlocal meta-optics

A complementary branch exploits thermo-optic reconfiguration for large, deliberate switching. “Thermo-optically Reconfigurable PbTe Mie Resonator Meta-atoms” shows that PbTe combines a very high infrared refractive index with an anomalous negative thermo-optic coefficient, so increasing temperature decreases refractive index and blue-shifts resonances. The standard magnetic-dipole Mie condition is written as

$$\frac{dn_{\mathrm{eff}}}{dT}\approx 0, \qquad \frac{d\lambda}{dT}\approx 0.$$5

For a $$\frac{dn_{\mathrm{eff}}}{dT}\approx 0, \qquad \frac{d\lambda}{dT}\approx 0.$$6 PbTe sphere, the magnetic-dipole resonance shifts by 480 nm from 80 K to 293 K, whereas the analogous Ge resonance shift over the same range is 90 nm. The paper reports normalized tunability 0.09 for Ge MD and $$\frac{dn_{\mathrm{eff}}}{dT}\approx 0, \qquad \frac{d\lambda}{dT}\approx 0.$$7 for PbTe MD. For a sharper PbTe resonance around $$\frac{dn_{\mathrm{eff}}}{dT}\approx 0, \qquad \frac{d\lambda}{dT}\approx 0.$$8 with $$\frac{dn_{\mathrm{eff}}}{dT}\approx 0, \qquad \frac{d\lambda}{dT}\approx 0.$$9, the resonance is tuned by more than 8 linewidths over 80–573 K, and more than one linewidth with only $$n_{\mathrm{eff}}\times \frac{2\pi}{\lambda}\times L = m\times 2\pi,$$0 K (Lewi et al., 2017).

When PbTe resonators are arranged into arrays, the single-particle thermo-optic shift becomes a reconfigurable effective-medium response. A 2D array of PbTe spheres shows resonance narrowing to about $$n_{\mathrm{eff}}\times \frac{2\pi}{\lambda}\times L = m\times 2\pi,$$1, normalized tunability $$n_{\mathrm{eff}}\times \frac{2\pi}{\lambda}\times L = m\times 2\pi,$$2, and tuning by more than one linewidth with induced index shift $$n_{\mathrm{eff}}\times \frac{2\pi}{\lambda}\times L = m\times 2\pi,$$3, corresponding to $$n_{\mathrm{eff}}\times \frac{2\pi}{\lambda}\times L = m\times 2\pi,$$4 K at 173 K or about $$n_{\mathrm{eff}}\times \frac{2\pi}{\lambda}\times L = m\times 2\pi,$$5 K around room temperature. A cube array on a perfect electric conductor substrate reaches $$n_{\mathrm{eff}}\times \frac{2\pi}{\lambda}\times L = m\times 2\pi,$$6 and supports an active notch filter in which $$n_{\mathrm{eff}}\times \frac{2\pi}{\lambda}\times L = m\times 2\pi,$$7 moves the resonance by more than a linewidth. The same platform provides a full $$n_{\mathrm{eff}}\times \frac{2\pi}{\lambda}\times L = m\times 2\pi,$$8 phase shift while maintaining greater than 99% reflectivity with $$n_{\mathrm{eff}}\times \frac{2\pi}{\lambda}\times L = m\times 2\pi,$$9, corresponding to about $\lvert S_{\max}\rvert \le 0.001$0 around 173 K (Lewi et al., 2017).

“Active Nonlocal Metasurfaces” shows that thermo-optic reconfiguration can also operate through collective q-BIC/Fano resonances rather than isolated Mie modes. The quality factor follows the scaling law

$\lvert S_{\max}\rvert \le 0.001$1

where $\lvert S_{\max}\rvert \le 0.001$2 is the symmetry-breaking perturbation strength. In a 1D free-space silicon modulator, a 100 °C temperature range produces a 4.6 nm resonance shift and an extinction ratio of 2.4 at $\lvert S_{\max}\rvert \le 0.001$3 nm. In a 2D polarization-insensitive modulator, a 100 °C range gives a 3.2 nm shift and an extinction ratio of 1.18 at $\lvert S_{\max}\rvert \le 0.001$4 nm. Numerical designs extend this to wavefront shaping: changing silicon index from $\lvert S_{\max}\rvert \le 0.001$5 to $\lvert S_{\max}\rvert \le 0.001$6 shifts the resonance by 14.0 nm and gives an extinction ratio of 37.9 at $\lvert S_{\max}\rvert \le 0.001$7 nm, enabling thermal switching of a cylindrical metalens from an “on” state at 25 °C to an “off” state at 275 °C. A multifunctional metasurface is simulated to switch between 35° deflection at 25 °C and 16.7° deflection at 275 °C at $\lvert S_{\max}\rvert \le 0.001$8 nm (Malek et al., 2020).

The contrast between local and nonlocal metasurfaces is important here. In local metasurfaces, each element largely determines its own phase response; in nonlocal metasurfaces, the operative resonance is a delocalized collective mode. Thermo-optic reconfiguration in the latter case is therefore a band-structure and lifetime effect as much as a local material effect.

5. Integrated, two-dimensional, and ENZ composite media

Integrated photonics extends the same principles to guided-wave effective media. “Graphene oxide two dimensional films for thermo-optic photonic integrated devices” treats GO films on silicon nitride microring resonators as a reconfigurable thermo-optic effective medium whose optical constants and heat transport properties can be tuned by thickness, reduction degree, and polarization. For unreduced GO at 1550 nm, the fitted values are about $\lvert S_{\max}\rvert \le 0.001$9 and $\Delta\lambda=0$0; after heating to 200 °C these become approximately $\Delta\lambda=0$1 and $\Delta\lambda=0$2, with $\Delta\lambda=0$3 and an increase in $\Delta\lambda=0$4 by more than 36-fold. The thermo-optic coefficient changes sign with reduction: for 1 layer of unreduced GO in TE polarization, $\Delta\lambda=0$5 is about $\Delta\lambda=0$6, whereas at 200 °C it becomes about $\Delta\lambda=0$7. Thermal conductivity likewise changes strongly, from about 20.6 W/(m·°C) for 1 layer of unreduced GO in TE polarization to about 998 W/(m·°C) after reduction to 200 °C (Moss, 2024).

The same paper shows that photo-thermal reduction can be reversible or permanent depending on pump power. In a 1-layer GO microring, the extinction ratio remains roughly constant at about 14 dB for $\Delta\lambda=0$8 mW in TE polarization, but decreases above that threshold. For $\Delta\lambda=0$9 between about 40 and 72 mW the change is reversible; above about 72 mW the reduction becomes permanent. These material-state changes reconfigure the hybrid waveguide’s effective index, resonance wavelength, extinction ratio, and optical bistability without changing chip geometry (Moss, 2024).

ENZ metamaterials push the concept to the constitutive level. “Thermo-optic dynamics of effective epsilon-near-zero media” studies a 5-period stack of 74 nm SiO$2\pi$0 and 18 nm Ag on quartz, treated with Maxwell-Garnett effective medium theory. For the layered anisotropic medium,

$2\pi$1

The ENZ condition is $2\pi$2. The theoretical design predicts $2\pi$3 nm at normal incidence, while ellipsometric extraction gives $2\pi$4 nm at an incidence angle of $2\pi$5. Heating from 10°C to 70°C causes a blue shift of $2\pi$6, with an experimentally extracted thermal-spectral modulation rate of $2\pi$7 and an inferred temperature resolution of $2\pi$8. The effective thermo-optic coefficient reaches values of order $2\pi$9 in the ENZ enhancement region (Wu et al., 28 Apr 2026).

The same work also reframes ultrafast thermo-optics as time-dependent effective-medium reconfiguration. Under visible-range pump-probe excitation, the response evolves from rapid electron heating to a thermo-optic regime dominated by electron-phonon coupling and lattice heating. The two-temperature model yields electron cooling to half maximum at 3.2 ps and a lattice-temperature peak at 8.1 ps, interpreted as a picosecond-scale transient evolution of the effective ENZ medium (Wu et al., 28 Apr 2026).

A recurring misconception is that thermo-optic phenomena are necessarily slow and secondary. The ENZ results show that this is not generally valid once the temperature dependence of the effective constitutive parameters is strongly enhanced.

6. Retarded effective media, dimensional crossover, and characterization

In photon Bose-Einstein condensates, thermo-optic reconfiguration takes the form of a self-consistently generated medium response. “Photon BEC with Thermo-Optic Interaction at Dimensional Crossover” separates the effective photon-photon interaction into a local Kerr contribution and a delayed, spatially nonlocal thermo-optic term mediated by diffusion. The Green’s function in Fourier space is

$dn/dT$0

with diffusion length $dn/dT$1. As the trap aspect ratio increases, the thermo-optic interaction strength initially grows approximately linearly and then saturates at $dn/dT$2. In the strong quasi-1D limit, the thermo-optic interaction remains saturated while the Kerr contribution continues to scale, so the dominant mechanism reverses from thermo-optic to Kerr. The paper identifies the crossover scale as $dn/dT$3 (Stein et al., 2021).

This dimensional-crossover result is conceptually significant because it makes the “effective medium” neither purely material-defined nor purely structure-defined. The effective nonlinear optical response depends on trap geometry, diffusion length, duty cycle, and experimental timescale. “Hartree-Fock Analogue Theory of Thermo-Optic Interaction” reaches a related conclusion from a quantum many-body direction: during a single pulse, the condensate lifetime is very short compared with the thermal relaxation time, so the temperature profile is effectively determined by the initial photon density rather than the instantaneous density, and the thermo-optic potential behaves like a time-growing external mean field proportional to that initial state (Stein et al., 2022).

The metrology literature reaches a parallel conclusion for bulk media. “Infrared photothermal heterodyne imaging in thermally thick medium for thermo-optic property characterization” generalizes IR-PHI from thin films to thermally thick semi-transparent media, where the probe samples front-surface temperature, rear-surface temperature, and volume-averaged temperature rather than a single uniform temperature. The thick-medium transmittance model is written as

$dn/dT$4

For PDMS with thickness $dn/dT$5 mm at probe wavelength $dn/dT$6, nonlinear least-squares fitting gives $dn/dT$7 and $dn/dT$8. For Borofloat glass, the fitted coefficients are of order $dn/dT$9. The method is reported to detect temperature changes down to about 30 mK in PDMS near the SNR cutoff (Letessier et al., 1 May 2025).

These measurement results underscore an important interpretive point: in thick media the retrieved thermo-optic parameter is already an effective quantity, because the optical probe integrates over a nonuniform thermal field. That perspective aligns naturally with the broader literature on metasurfaces, microrings, photon fluids, and ENZ metamaterials.

7. General implications and boundaries of the field

Across these platforms, thermo-optic reconfiguration of effective media is not a single device concept but a common physical framework. In one limit, the objective is athermalization through cancellation of opposite-sign dispersions, as in Si/PbTe hybrid resonators (Cohen et al., 2023). In another, the objective is strong tuning or switching, as in PbTe Mie meta-atoms and q-BIC/Fano nonlocal metasurfaces (Lewi et al., 2017, Malek et al., 2020). In integrated photonics, chemically and optically reconfigurable films such as GO make the effective medium state variable and polarization dependent (Moss, 2024). In photon condensates, the medium is reconfigured by the optical field itself through a delayed diffusion process, and geometry determines which nonlinear mechanism dominates (Stein et al., 2022, Stein et al., 2021). In ENZ metamaterials, temperature redefines the constitutive parameters that generate the singular optical response (Wu et al., 28 Apr 2026). In thermo-optic metrology, the measured coefficient becomes a depth-weighted effective parameter tied to the spatial thermal profile (Letessier et al., 1 May 2025).

A plausible implication is that the field is converging on a unified vocabulary in which thermal tuning, thermal stabilization, and thermal nonlinearity are all treated as manipulations of an effective optical medium. This suggests a design methodology centered not only on material $dn/dT$0, but also on modal overlap, diffusion length, relaxation time, anisotropy, reduction chemistry, quality factor, and homogenized constitutive parameters. The literature does not reduce these regimes to a single universal model, but it consistently supports the claim that temperature can act as a reconfiguration knob for composite photonic media rather than as a mere source of parasitic drift.