Asymmetric Purcell Enhancement

Updated 5 July 2026

Asymmetric Purcell enhancement is the selective boost of an emitter’s spontaneous decay rate via engineered geometrical, modal, and non-Hermitian asymmetries.
It utilizes structural and microscopic design to tailor the local density of optical states, enabling high Purcell factors and controlled radiative out-coupling.
Practical implementations demonstrate directional emission control, interference between resonant modes, and enhanced routing in photonic and plasmonic systems.

Asymmetric Purcell enhancement denotes a class of spontaneous-emission phenomena in which the Purcell effect is not distributed equivalently across the available electromagnetic channels. In the literature, the asymmetry may be directional, producing unequal emission into opposite waveguides; structural, arising from deliberately non-mirror-symmetric resonators or waveguides; microscopic, through strong dependence on the emitter position inside a unit cell; or non-Hermitian, through spatially patterned loss or gain that privileges dark and bright channels differently. The common framework remains the modification of the local density of optical states, cavity lifetime, mode volume, and radiative out-coupling. At the same time, several closely related platforms deliver very large Purcell factors without constituting asymmetric Purcell enhancement in the strict left-right sense, which makes the scope of the term intrinsically context-dependent (Krasnok et al., 2016).

1. Conceptual scope and terminological boundaries

The Purcell effect is the modification of the spontaneous decay rate of a quantum emitter by its electromagnetic environment. In asymmetric Purcell enhancement, the central issue is not merely that the decay rate is increased, but that the increase is selective with respect to propagation direction, spatial position, modal branch, radiative port, or loss channel. Structural asymmetry can remove symmetry constraints and enable low-loss long-range supermodes; modal asymmetry can create interference between orthogonal resonances; microscopic asymmetry can make the enhancement depend strongly on the emitter coordinate inside a unit cell; and non-Hermitian asymmetry can be encoded directly in the loss or gain distribution.

A recurrent misconception is that any large Purcell factor in an asymmetric structure is itself an asymmetric Purcell effect. This is not generally correct. A finite chain of silicon nanoparticles exhibits a large Purcell factor through a dark band-edge mode and a Van Hove singularity, but the result is explicitly not presented as a left-right directional emission imbalance (Krasnok et al., 2016). Similarly, a dielectric hemisphere above a ground plane produces strongly enhanced, mode-selective, forward conical radiation, yet the radiation pattern is axisymmetric about the rod axis rather than one-sided (2209.13670). The literature therefore separates asymmetry of geometry from asymmetry of emission channels.

2. Structural asymmetry as a route to large $Q/V$ and controlled out-coupling

A major strand of the subject uses geometric or material asymmetry to relax symmetry constraints that ordinarily limit confinement, propagation loss, or spectral selectivity. In these systems, asymmetry is not incidental; it is the design variable that permits a more favorable balance between mode area, lifetime, and extraction.

Platform	Asymmetry mechanism	Representative reported result
Composite hybrid plasmonic waveguide (Su et al., 2019)	Dissimilar HPW and SPP modes in an asymmetric 4-layer stack	$0.03~\mathrm{dB/\mu m}$ loss, $0.002~\mu\mathrm{m}^2$ mode area, $71\%$ coupling, $29~\mathrm{dB}$ extinction ratio, $1.5\times10^4$ normalized Purcell factor
Asymmetric Tamm structure (Singh et al., 2022)	Two unequal DBR sections and unequal spacer layers around a thin Ag layer	$Q=140$ at $640~\mathrm{nm}$ , Purcell factor $4$, collection efficiency enhancement $5$
Antenna-cavity hybrid (Doeleman et al., 2016)	Detuning asymmetry and interference between antenna and cavity scattering paths	Hybrid total enhancement around $0.03~\mathrm{dB/\mu m}$ 0, bandwidth roughly $0.03~\mathrm{dB/\mu m}$ 1
Quasi-BIC metasurface (Tse et al., 27 Aug 2025)	Asymmetric or lossy port coupling near a BIC	Enhancement maximized at finite detuning, not at the exact BIC

In composite hybrid plasmonic waveguides, asymmetry is encoded both geometrically and modally. The structure couples a bottom hybrid plasmonic waveguide and a top surface plasmon polariton interface across a thin metal layer, so it does not require structural, material, or modal symmetry. When the metal thickness is below the skin depth, the two constituent modes hybridize into short-range and long-range supermodes. The long-range mode is obtained when the longitudinal electric field in the metal is suppressed; the key design quantity is the normalized electric flux through the metal,

$0.03~\mathrm{dB/\mu m}$ 2

which is tuned toward zero to minimize absorption. Experimentally, the optimized devices exhibit $0.03~\mathrm{dB/\mu m}$ 3 propagation loss and $0.03~\mathrm{dB/\mu m}$ 4 mode area, while CHPW microrings reach $0.03~\mathrm{dB/\mu m}$ 5 in/out coupling efficiency, $0.03~\mathrm{dB/\mu m}$ 6 extinction ratio, and a normalized Purcell factor of $0.03~\mathrm{dB/\mu m}$ 7 (Su et al., 2019).

In asymmetric Tamm structures, asymmetry yields dual resonant modes rather than a single dominant Tamm resonance. The specific structure comprises DBR1, DBR2, a thin Ag layer, and two spacers $0.03~\mathrm{dB/\mu m}$ 8 and $0.03~\mathrm{dB/\mu m}$ 9, producing TPR1 near $0.002~\mu\mathrm{m}^2$ 0 and TPR2 near $0.002~\mu\mathrm{m}^2$ 1. The resonance aligned with the NV $0.002~\mu\mathrm{m}^2$ 2 zero-phonon line has $0.002~\mu\mathrm{m}^2$ 3, corresponding to $0.002~\mu\mathrm{m}^2$ 4, compared with $0.002~\mu\mathrm{m}^2$ 5 for the conventional Tamm structure at the same wavelength. With the emitter positioned in spacer-1 about $0.002~\mu\mathrm{m}^2$ 6 from the Ag layer, the field enhancement is about $0.002~\mu\mathrm{m}^2$ 7 at $0.002~\mu\mathrm{m}^2$ 8, yielding a Purcell factor of $0.002~\mu\mathrm{m}^2$ 9 and a collection-efficiency enhancement of $71\%$ 0 (Singh et al., 2022).

A related but distinct asymmetry appears in antenna-cavity hybrids, where the strongest boost is not at zero detuning but typically for red detuning of the cavity relative to the antenna. The asymmetry is spectral and interference-based: constructive interference between antenna-mediated and cavity-mediated multiple-scattering paths can push the hybrid enhancement above that of either bare subsystem, while preserving a near-unity outcoupling efficiency into a single cavity decay channel. In the realistic silicon-nitride whispering-gallery-cavity plus gold-antenna example, the reported hybrid total enhancement is around $71\%$ 1, with the enhancement bandwidth increased by about a factor of $71\%$ 2 to roughly $71\%$ 3 (Doeleman et al., 2016).

3. Directional asymmetry and spin-dependent routing

The most direct realization of asymmetric Purcell enhancement is a system in which the Purcell-boosted emission is routed preferentially into one output direction. In this setting, the asymmetry is a property of the emitter-cavity-waveguide interface rather than of the cavity $71\%$ 4 factor alone.

A waveguide-coupled H1 nanocavity containing InAs quantum dots in a GaAs p-i-n photonic-crystal membrane exemplifies this regime. The cavity supports two orthogonal modes, $71\%$ 5 and $71\%$ 6, which couple both to the emitter and to two opposite W1 waveguides. Directionality arises because the two cavity modes couple to the waveguides with opposite phase structure, while a spin-polarized emitter couples to $71\%$ 7 and $71\%$ 8 with a relative phase of approximately $71\%$ 9. For Device 1, the cavity resonances are $29~\mathrm{dB}$ 0 with $29~\mathrm{dB}$ 1 for the $29~\mathrm{dB}$ 2 mode and $29~\mathrm{dB}$ 3 with $29~\mathrm{dB}$ 4 for the $29~\mathrm{dB}$ 5 mode. The measured lifetime is $29~\mathrm{dB}$ 6, compared with a bulk lifetime of $29~\mathrm{dB}$ 7, giving a Purcell factor of $29~\mathrm{dB}$ 8. The same device reaches a peak directional contrast of $29~\mathrm{dB}$ 9, while a second device shows electrical tuning of the directional contrast from $1.5\times10^4$ 0 to $1.5\times10^4$ 1 (Martin et al., 17 Jan 2025).

In this class of device, asymmetric Purcell enhancement is therefore a compound effect: the spontaneous-emission rate is increased by the low-mode-volume nanocavity, and the emission amplitudes into the left and right waveguides are made unequal by interference between the two cavity modes. The observed asymmetry is not captured by a purely scalar LDOS picture; the relative phase of the mode couplings and the fine-structure-dependent emitter polarization are essential. The three-level quantum-trajectory model introduced for Device 2 shows that asymmetric directional contrast between Zeeman lines is reproduced only when coherent population transfer and imperfect circular polarization are included (Martin et al., 17 Jan 2025).

By contrast, the microwave Purcell-efficiency demonstration based on a structured dielectric hemisphere and ground plane illustrates a different form of channel selectivity. The system supports two essentially omnidirectional radiation modes at $1.5\times10^4$ 2 and $1.5\times10^4$ 3, with Purcell enhancement factors of $1.5\times10^4$ 4 and $1.5\times10^4$ 5, and radiation efficiencies of $1.5\times10^4$ 6 and $1.5\times10^4$ 7, respectively. The far field forms diverging conical beams in the forward direction, with main lobes at approximately $1.5\times10^4$ 8 for the $1.5\times10^4$ 9 mode and $Q=140$ 0 for the $Q=140$ 1 mode, but the system remains axisymmetric about the rod axis (2209.13670). This demonstrates that strong directional structure and strong Purcell enhancement do not automatically imply left-right asymmetry.

4. Microscopic and emitter-selective asymmetry

In many systems, the asymmetry resides neither in macroscopic left-right routing nor in deliberately broken geometric parity, but in the emitter’s microscopic placement and orientation relative to the local field. In these cases, the Purcell factor can vary by orders of magnitude inside a single unit cell.

A microscopic model of hyperbolic metamaterials makes this point explicit. The medium is represented as an infinite cubic lattice of resonant point dipoles with lattice constant $Q=140$ 2, where only the $Q=140$ 3-component of the polarizability is nonzero. The emitter is a radiating dipole at $Q=140$ 4, and the Purcell factor $Q=140$ 5 and Lamb shift $Q=140$ 6 are defined by

$Q=140$ 7

The resulting enhancement depends jointly on the Bloch-mode pole condition, the local field amplitude, and the emitter position $Q=140$ 8. In the hyperbolic regime, the Green function has a characteristic cross-like spatial shape and is modulated by ripples associated with Brillouin-zone-edge states. The maximum Purcell factor is expected on the vertical edge of the unit cell, $Q=140$ 9, $640~\mathrm{nm}$ 0, because the local field from nearby lattice dipoles is strongest there. Both the Purcell factor and the Lamb shift increase strongly as the emitter approaches lattice nodes, making the enhancement intrinsically position-dependent and asymmetric (Poddubny et al., 2012).

A different emitter-selective asymmetry arises in gain-compensated plasmonic dimers. Here the system is a gold nanorod dimer surrounded by a finite cylindrical gain region, and the relevant resonance is a plasmonic quasinormal mode with complex eigenfrequency $640~\mathrm{nm}$ 1. The bare dimer has $640~\mathrm{nm}$ 2, but with strong gain the quality factor increases to $640~\mathrm{nm}$ 3, while the effective mode volume remains nearly unchanged. In the linear-gain regime, the spontaneous-emission rate is not described by the projected LDOS term alone; it becomes

$640~\mathrm{nm}$ 4

The asymmetry is tied to emitter position $640~\mathrm{nm}$ 5 and dipole orientation through the overlap with the cavity mode, not to a left-right geometric bias. Numerically, the peak LDOS Purcell factor increases from $640~\mathrm{nm}$ 6 to $640~\mathrm{nm}$ 7, and the total Purcell factor reaches $640~\mathrm{nm}$ 8. At the same time, the radiative beta factor changes only moderately, from $640~\mathrm{nm}$ 9 to $4$0, confirming that the dominant change is the resonant amplitude and lifetime rather than a radical redistribution of radiation versus loss (VanDrunen et al., 2023).

5. Non-Hermitian asymmetry: exceptional points, dark-bright partitioning, and engineered loss

Non-Hermitian systems introduce a further generalization of asymmetric Purcell enhancement. In this literature, asymmetry is encoded in the coupling topology between dark and bright resonators and, crucially, in the spatial distribution of loss or gain. The resulting local density of states is no longer constrained to a single Lorentzian resonance.

RF cavities tuned to higher-order exceptional point degeneracies provide an experimentally explicit example. In these systems, the emitter is coupled to a dark resonator, while emission is extracted through bright resonators with engineered dissipation. At an order-$4$1 EPD, $4$2 eigenfrequencies and their eigenmodes coalesce, and the local density of states becomes an $4$3-th-power Lorentzian rather than a conventional Lorentzian. For EPD-2, the LDoS is a Lorentzian squared; for EPD-3, cubic and quadratic terms appear, with the cubic term dominating only under judiciously designed dissipation. The reported enhancement relative to an isolated-resonance regime is about $4$4 for EPD-2 and about $4$5 in theory for optimally tuned EPD-3. In the experimental trimer, the observed enhancement is around $4$6 to $4$7, while the dimer shows a smaller enhancement in direct transmission-line measurements. The asymmetry is therefore primarily spatial and dissipative: the emitter is attached to a specific dark site, and the loss rates $4$8 are deliberately made nonuniform to promote the highest-order singular term (Wyszkowski et al., 6 Dec 2025).

Exceptional points also produce nonlinear Purcell asymmetry in triply resonant $4$9 cavities composed of a dark mode, a leaky mode, and a second-harmonic mode. At the EP, the linear LDOS peak is enhanced by a factor of $5$0 for a monochromatic source on resonance, but the integrated linear LDOS remains conserved. The nonlinear situation is different. When the emitter couples only to the dark fundamental mode and the dominant nonlinear coefficient is the corresponding $5$1, the ideal monochromatic second-harmonic emission enhancement reaches $5$2 at zero detuning, and the broadband frequency-integrated enhancement can reach $5$3. In the inverse-designed 2D photonic-crystal slab proof of concept, the reported monochromatic enhancement is about $5$4, the frequency-integrated enhancement is about $5$5, and the spatially averaged integrated SDOS enhancement is about $5$6 (Pick et al., 2017). This is asymmetry in modal role rather than in physical propagation direction: the source excites the dark state, while radiation escapes through the leaky partner.

A broader view of asymmetric Purcell enhancement requires attention to the tradeoff between resonant lifetime and radiative escape. Quasi-BIC metasurfaces make this explicit. Their measurable photoluminescence enhancement is expressed as

$5$7

The crucial result is that enhancement is absent at the exact BIC because the out-coupling channel closes, even though the radiative $5$8 diverges formally. Strong enhancement is observed only at a quasi-BIC, and in realistic lossy or asymmetric metasurfaces the optimum occurs at finite detuning from the BIC. For asymmetric accidental quasi-BICs, the port-resolved radiative rates for the top and bottom ports differ, so the optimal detuning depends on which emission side is targeted (Tse et al., 27 Aug 2025).

ENZ Bragg-reflection microcavities show a related but distinct constraint. The analytical theory assumes a dipole emitter at the center of the cavity and a $5$9-polarized ENZ mode propagating along the $0.03~\mathrm{dB/\mu m}$ 00-axis, with output coupling along $0.03~\mathrm{dB/\mu m}$ 01 and optical leakage along $0.03~\mathrm{dB/\mu m}$ 02 and $0.03~\mathrm{dB/\mu m}$ 03. The paper does not explicitly study asymmetric emitter placement, but the cavity architecture is direction-dependent by construction. In all-dielectric ENZ Bragg-reflection microcavities, the Purcell and quality factors scale as $0.03~\mathrm{dB/\mu m}$ 04 and $0.03~\mathrm{dB/\mu m}$ 05, respectively, and the BRW estimate gives $0.03~\mathrm{dB/\mu m}$ 06 in the low-loss limit (Panahpour et al., 2024). This suggests that directional loss channels and modal anisotropy can be decisive even when the emitter itself is treated symmetrically.

Finally, all-dielectric nanoparticle chains delineate the boundary between asymmetric Purcell enhancement and adjacent Purcell-engineering mechanisms. In these structures, a dipole emitter couples to a dark collective band-edge mode with nearly zero group velocity, producing a Van Hove singularity in the density of states. The theoretical Purcell factor is about $0.03~\mathrm{dB/\mu m}$ 07 for a dimer and about $0.03~\mathrm{dB/\mu m}$ 08 for a $0.03~\mathrm{dB/\mu m}$ 09-particle chain, while the microwave proof-of-concept shows a $0.03~\mathrm{dB/\mu m}$ 10-fold enhancement for a $0.03~\mathrm{dB/\mu m}$ 11-particle chain. The emitted light is localized to the chain and the radiation pattern has two narrow lobes directed along the chain, but the work does not emphasize nonreciprocal emission, chiral coupling, or left-right asymmetry (Krasnok et al., 2016). The broader implication is that the term asymmetric Purcell enhancement is most precise when the privileged channel—direction, port, position, mode, or loss pathway—is explicitly identified.