Exact microscopic theory of electromagnetic heat transfer between a dielectric sphere and plate

Near-field electromagnetic heat transfer holds great potential for the advancement of nanotechnology. Whereas far-field electromagnetic heat transfer is constrained by Planck’s blackbody limit, the increased density of states in the near-field enhances heat transfer rates by orders of magnitude relative to the conventional limit. Such enhancement opens new possibilities in numerous applications, including thermal-photo-voltaics, nano-patterning, and imaging. The advancement in this area, however, has been hampered by the lack of rigorous theoretical treatment, especially for geometries that are of direct experimental relevance. Here we introduce an efficient computational strategy, and present the first rigorous calculation of electromagnetic heat transfer in a sphere-plate geometry, the only geometry where transfer rate beyond blackbody limit has been quantitatively probed at room temperature. Our approach results in a definitive picture unifying various approximations previously used to treat this problem, and provides new physical insights for designing experiments aiming to explore enhanced thermal transfer.

💡 Research Summary

This paper presents a rigorously exact computational framework for evaluating near‑field electromagnetic heat transfer between a dielectric sphere and a planar substrate—a geometry that is directly relevant to recent experiments. Starting from classical macroscopic fluctuational electrodynamics (CMFED), the authors introduce stochastic current sources inside the sphere that satisfy the fluctuation‑dissipation theorem. The resulting electric field is expressed through the dyadic Green’s function, which is then expanded in a mixed basis of vector spherical harmonics (centered on the sphere) and vector cylindrical harmonics (associated with the plate). By defining transformation operators that map spherical to cylindrical modes and vice‑versa, and by incorporating the Fresnel reflection coefficients for each body (R_A for the sphere, R_B for the plate), the authors derive a compact linear system (Eq. 8) of the form (I‑R_A R_B) A = A⁰. Solving this system yields the full set of mode amplitudes, from which the Poynting flux through the plate surface is obtained by summing over all source configurations (ℓ, m, p).

The numerical implementation truncates the angular momentum index ℓ at a value ℓ_max that scales with both the sphere radius a and the gap d (ℓ_max ≈ ℓ₀ + ℓ₁(a/λ + a/d)). For experimentally realistic parameters (a = 20 µm, d = 100 nm, giving a/d ≈ 200) convergence to within 1 % is achieved with ℓ_max ≈ 700, leading to a modestly sized linear system that can be solved efficiently. Frequency integration is performed over 47 points chosen to resolve the dominant surface‑phonon‑polariton resonances of silica.

The authors benchmark their exact results against three widely used approximations: (i) the far‑field limit Q_ff, (ii) the dipole approximation (treating the sphere as a point dipole), and (iii) the proximity (Derjaguin) approximation. Their analysis shows that:

• The far‑field expression is valid only when the gap exceeds roughly 1 µm; for smaller gaps the near‑field contribution scales as a/d while the far‑field part scales as a/λ, so the latter dominates for large spheres even at nanometric separations.
• The dipole model accurately reproduces the exact heat flux for a ≲ d, with errors below 10 % for d = 1 µm and sphere radii up to a ≈ d. When a ≫ d the dipole prediction underestimates the exact result by a factor ∼(a/d)², reflecting the neglect of higher‑order multipoles.
• The proximity approximation systematically overestimates the flux for small spheres (a ≲ d) because it ignores curvature‑induced scattering and does not separate propagating from evanescent contributions. Conversely, for very large spheres (a ≫ d) it underestimates the far‑field contribution, since it assumes all radiation originates from the closest points.

A “phase diagram” in the (a, d) plane is constructed, indicating the regions where each approximation is within 30 % of the exact solution. The diagram reveals a crossover region around d ≈ 0.3 µm for a = 1 µm where neither dipole nor proximity approximations are reliable, emphasizing the need for the full calculation.

Comparison with experimental data from Rousseau et al. (Nature Photonics 2009) on a silica sphere (a = 20 µm) shows that both the exact theory and the proximity approximation can fit the measured cantilever deflection within experimental uncertainties, primarily because the calibration parameters (effective gap offset d₀ and proportionality constant H) are not precisely known. Nevertheless, the exact theory predicts a markedly different distance dependence, especially at sub‑100 nm gaps, suggesting that future experiments with better gap control could discriminate between models.

Finally, the authors stress that their scattering‑matrix based approach is not limited to the sphere‑plate configuration; any geometry for which individual scattering matrices are known can be treated similarly. This opens the door to quantitative studies of multi‑body near‑field radiative heat transfer, Casimir forces, optical trapping forces, and vacuum friction in complex nanostructured systems.