Ultrastrong magnetic light-matter interaction with cavity mode engineering

The work addresses the challenge that magnetic light-matter interactions are intrinsically weaker than electric interactions, limiting applications in quantum computing, sensing, NMR/ESR, and masers. The research question is how to engineer electromagnetic cavities to maximize magnetic coupling—i.e., to produce strong single-photon magnetic fields while maintaining long photon lifetimes—by minimizing the magnetic mode volume (VB) and maximizing the quality factor (Q). The authors present a framework for mode engineering in hollow metallic (3D) cavities that supports non-TEM modes, aiming for ultrastrong magnetic interaction suitable for single-photon coupling to spins or superconducting qubits. They introduce three strategies—longitudinal mode squeezing, current engineering, and magnetic field expulsion—and analyze their performance and trade-offs, corroborated by finite-element simulations and a proof-of-principle experiment with NV centers in diamond.

Prior approaches to enhance magnetic coupling largely employed coplanar waveguide (CPW) resonators coupled to superconducting qubits or solid-state spin ensembles, which allow small effective mode volumes but suffer from dielectric losses in substrates, typically limiting Q to below ~10^6. Conversely, 3D hollow metallic cavities (e.g., 3D transmon architectures) reach Q > 10^6, but their supported TE/TM modes generally have large magnetic mode volumes, hindering strong magnetic coupling. Existing cavity concepts such as split-mode, loop-gap, and reentrant cavities have explored localization of fields via geometric features. Dielectric-based light-confinement strategies in nanophotonics can achieve small electric-mode volumes but differ fundamentally from the present approach, which exploits electromagnetic boundary conditions in metallic cavities without relying on high-index dielectrics. The authors unify and extend these cavity designs under a common mode-engineering framework focused on magnetic field concentration and Q preservation.

Theory and definitions: The magnetic interaction Hamiltonian is Hint = m·B. For a cavity mode, the single-photon magnetic field at position r is B_s(r) = hf / sqrt(2 VB(r)/μ), where VB(r) = [∫ dV |B(r)|^2/μ(r)] / [|B(r)|^2/μ(r)] is the effective magnetic mode volume at the dipole position, chosen at the magnetic antinode. Photon lifetime is τ = Q/(2πf). Figures of merit include VB, Q, and Q/VB. The geometric factor G quantifies material-independent cavity losses, with Q = G/Rs, Rs being the surface resistance. Finite-element simulations (COMSOL FEM) compute eigenmodes, fields, VB, and G under realistic geometries and constraints.

Mode-engineering strategies:

1. Longitudinal squeezing of TM010: Cylindrical cavities supporting TMmnp modes are analyzed; for TM010, fields are azimuthally symmetric and independent of z, allowing cavity height h to be reduced without changing resonance frequency (ω010 = c j01/a). Analytical mode volume scales linearly with h and as λ^2: VB ≈ 0.135 (2π J0,max)^2 h λ^2 ≈ 0.366 λ^3 for h ≈ λ; for h ≪ λ, VB can be far below λ^3 and, as h → 2δ (metal penetration depth), VB → 0.73 δ λ^2. Superconductors (e.g., Nb) with small penetration depth and low Rs enable extreme squeezing with minimal ohmic loss, provided operation is below the superconducting gap frequency.

2. Current engineering: Viewing the TM010 cavity as an LC oscillator, magnetic fields correlate with surface currents. Designs that increase local current density boost the peak magnetic field and reduce VB. Implementations include split-mode cavities, reentrant cavities (single post), and doubly reentrant cavities (two posts). The reentrant geometry increases capacitance and funnels current through smaller circumferences, enhancing local B. Scaling laws: for a reentrant post with radius R, the surface magnetic field scales ∝ [R ln(a/R)]^-1 as R → 0; for a doubly reentrant geometry with gap g, the maximum B scales ∝ [R ln(g/R)]^-1 for g ≫ R and ∝ 1/g for R ≪ g. Further enhancement is achieved by inverse-tapering the posts to crowd current—keeping a large top capacitance while narrowing the current path at the base—leading to B ∝ 1/R' (bottom radius) and stronger proportionality than non-tapered designs.

3. Magnetic field expulsion: Inserting a thin conductive plate into the cavity expels magnetic flux (Meissner-like effect), concentrating B near edges. In the quasistatic limit, ∇×E ≈ 0 and ∇^2 B = δ^-2 B (δ is penetration depth). The surface field scales with the demagnetization factor N as Bsurf = B0/(1−N); for very thin, elongated plates N → 1 and surface B is strongly enhanced (finite in practice due to finite δ). A floating thin metal plate (e.g., foil on a low-loss substrate) inserted with a small gap from the sidewall creates a magnetic hotspot and reduces VB approximately linearly with the plate thickness t, while leaving Q nearly unchanged when only t is varied.

Trade-off analysis: The authors map VB versus G (hence Q via Rs) for multiple cavity families under fabrication constraints (e.g., minimum reentrance radius ~100 μm, plate thickness ~10 μm). They show how each method trades VB against Q and how combined strategies can maximize Q/VB.

Numerical methods: FEM with tetrahedral meshes (max size 1–2 mm), minimum mesh < half the smallest feature (e.g., plate thickness), growth rate 1.35–1.45, curvature factor −0.4, narrow-region resolution −0.7. Material losses incorporated via Rs; Nb (superconductor) Rs ≈ 1–10 nΩ (low T), Cu ≈ 1 mΩ (room T).

Experimental validation: A copper doubly reentrant cavity (h ≈ 1.9 cm) resonant at ~2.87 GHz was fabricated (milled C101 Cu, SMA input, optical access hole). An electronic-grade diamond (3×3×0.5 mm^3, ~300 ppb NV density) was placed between posts. NV ensemble optically detected magnetic resonance and Rabi oscillations under strong drive (25 dBm) yielded oscillation frequency to infer B1 and VB. Vector network analysis measured f and Q. Results agreed with simulations.

Coupling estimates: For electronic spins (γ = 28 GHz/T), at 10 GHz a representative cavity with VB = 3.62×10^-8 m^3 gives B1 ≈ 2.06 nT and single-spin coupling g ≈ 57.7 Hz; ensembles enhance cooperativity by √N (or N in cooperativity). For flux qubits (e.g., 6 μm × 5 μm loop with ⟨1|I|0⟩ ≈ 100 nA), m0 ≈ 3 μA·μm^2, yielding g ≈ 2.33 MHz at 5 GHz for the high-Q, small-VB cavity.

• All three strategies—longitudinal squeezing, current engineering, and field expulsion—can, in principle, achieve arbitrarily small magnetic mode volumes limited by electromagnetic penetration depth or fabrication constraints, while maintaining high Q.
• TM010 longitudinal squeezing: Reducing height from h = 2 cm to h = 5 mm in a cylindrical cavity cuts VB from ~0.140 λ^3 to ~0.0349 λ^3 at constant frequency (f ≈ 5.74 GHz), doubling the single-photon magnetic field. Analytical scaling predicts VB ∝ h λ^2 and no diffraction limit for metallic cavities.
• Current engineering: Introducing reentrances drastically reduces VB by concentrating current and increasing capacitance. Representative values (normalized by λ^3): split-mode ≈ 6.76×10^-3; reentrant ≈ 1.44×10^-3 at f ≈ 2.23 GHz; doubly reentrant ≈ 4.03×10^-4 at f ≈ 3.21 GHz. Inverse-tapered posts further reduce VB by ~2 orders of magnitude: tapered reentrant ≈ 1.76×10^-5; tapered doubly reentrant ≈ 5.06×10^-6. Scaling laws show B enhancement ∝ [R ln(…)]^-1 for small post radii and ∝ 1/g for small gaps; with tapering, B ∝ 1/R' (bottom radius), enabling continued reduction.
• Field-expulsion cavities: Inserting a thin, floating metallic plate (e.g., d = 1.5 cm, gap g = 2 mm, thickness t = 10 μm) into a cylindrical cavity yields VB ≈ 3.18×10^-5 λ^3. VB scales approximately linearly with t; thinning the plate reduces VB with negligible change in Q when only t is varied. Increasing plate size d or reducing gap g further reduces VB, with some Q penalty from altered current paths.
• Trade-off and ultimate performance: Using the geometric factor G and Rs for Nb (10 nΩ) or Cu (1 mΩ), the best combined design—double reentrant, inverse-tapered with a thin inserted plate—achieves Q/VB ≈ 3×10^16 λ^-3 (≈ five orders of magnitude improvement over a plain cylindrical cavity). With a 1 μm plate, simulations indicate VB ≈ 3.6×10^-9 λ^3 is feasible. Operating at lower frequencies (e.g., ~10 MHz) with ~100 μm features and ~1 μm plates could reach VB < 10^-11 λ^3 and Q/VB > 10^20 λ^-3.
• Experimental validation: A copper doubly reentrant cavity measured at f = 2.871 GHz had Q = 2,421 (simulated 3,169). NV ensemble Rabi oscillations gave VB = 1.75×10^-3 λ^3 (simulation 1.95×10^-3 λ^3), confirming predictive accuracy for VB, Q, and f.
• Application-impact metrics: For single NV spins at 10 GHz with representative VB, g ≈ 57.7 Hz; ensembles or magnons can achieve high cooperativity. For a flux qubit (6 μm × 5 μm loop, ⟨1|I|0⟩ ~ 100 nA) placed at the magnetic antinode, g ≈ 2.33 MHz at 5 GHz, entering the strong-coupling regime. Overall magnetic interaction strength can exceed free-space coupling by >10^16.

The engineered cavities address the core challenge of weak magnetic dipole coupling by concentrating the single-photon magnetic field while maintaining low loss. Longitudinal squeezing exploits TM010’s z-invariant fields to shrink VB without raising frequency; current engineering increases local current density to create magnetic hotspots at reentrances; field expulsion leverages demagnetization to focus fields near thin metal edges. The resulting gains in Q/VB enable ultrastrong magnetic interactions suitable for high-cooperativity spin–photon coupling, magnetic circuit QED with superconducting qubits, low-noise masers, precision metrology, and searches for dark matter through enhanced magnetic sensitivity. Simulations and experiments agree on VB, Q, and resonance frequency, validating the design principles. The combined approaches extend the achievable trade-off frontier, offering material- and geometry-based routes to tailor coupling for specific applications, including placing qubits at magnetic antinodes to minimize dielectric electric-field loss.

The paper introduces a unified framework for magnetic mode-volume engineering in hollow metallic cavities via three methods—TM010 longitudinal squeezing, current engineering (including inverse-tapered reentrances for current crowding), and magnetic field expulsion using thin plates. All methods can drive VB to arbitrarily small values limited by penetration depth or fabrication, with minimal Q degradation in favorable regimes. Trade-off analyses using the geometric factor show that combined designs can raise Q/VB by about five orders of magnitude relative to plain cavities, and simulations indicate VB down to ~3.6×10^-9 λ^3 with thin plates. A proof-of-principle experiment with NV centers confirms theoretical predictions for VB, Q, and f. Future work includes tailoring designs for fabrication practicality, target frequencies, multimode operation, and application-specific figures of merit (e.g., maximizing Q/VB, 1/VB, or 1/V depending on regime), as well as exploring materials (e.g., high-Tc superconductors) and cascaded cavity architectures to further enhance functionality.

• Practical limits arise from material properties (electromagnetic penetration depth) and fabrication resolution (minimum post radius, plate thickness, and gaps), which bound how small VB can be made.
• Reducing VB often increases surface currents and ohmic losses, imposing Q–VB trade-offs; although superconductors mitigate Rs, operation must remain below the superconducting gap frequency to avoid quasiparticle losses.
• Field-expulsion designs benefit from very thin plates; while thinning reduces VB with little Q penalty, other geometry changes (e.g., larger plates or smaller gaps) can reduce Q.
• Dielectric supports introduce additional loss; although simulations with sapphire suggest small impact at single-photon levels, material interfaces and two-level systems can affect Q in some regimes.
• The study’s experimental validation used copper at room temperature; extrapolation to superconducting operation and extreme miniaturization requires careful materials processing and magnetic-flux management to achieve the predicted Q values.