Seeded Axion-Photon Conversion Scheme

• The scheme uses an external seed, such as a coherent electromagnetic field or structured background, to trigger resonant axion-photon conversion.
• Analytical models employ two-level systems and 3D geometric-optics to quantify mixing, phase matching, and conversion probabilities across diverse environments.
• Applications span neutron-star magnetospheres, gamma-ray propagation, and advanced laboratory experiments like light-shining-through-wall setups.

Seeded axion-photon conversion scheme denotes a class of axion-electrodynamic constructions in which conversion is initiated, amplified, maintained, or diagnostically tagged by a seed. The seed may be an injected coherent electromagnetic field, a pre-existing polarized photon population, a structured magnetic or plasma background, or a trajectory choice that preserves resonance. Across these realizations the common interaction is

$$\mathcal{L}_{a\gamma}=-\frac{1}{4}g_{a\gamma}\,a\,F_{\mu\nu}\tilde F^{\mu\nu} = g_{a\gamma}\,a\,\mathbf{E}\cdot\mathbf{B},$$

but the operational meaning of “seeded” depends strongly on context: in neutron-star magnetospheres it can mean LO-mode injection and trajectory control; in gamma-ray propagation it can mean source-side photon-to-axion conversion before intergalactic transport; in laboratory searches it can mean resonant regeneration, interferometric local oscillators, or explicit seed fields in the regeneration region (Millar et al., 2021, Wang et al., 2023, An et al., 25 Sep 2025, Batllori et al., 20 Sep 2025).

1. Scope and meanings of “seeded”

The term is used in at least three distinct but related senses. First, it may denote stimulated or interferometric seeding, where a coherent electromagnetic field is deliberately injected so that the axion-induced field interferes with it or is regenerated into a monitored receiver mode. Second, it may denote resonance seeding, where magnetic-field modulation, plasma shaping, or trajectory selection is used to keep the system near the axion-photon level crossing. Third, it may denote environmental seeding, where an ambient photon population or magnetic structure provides the initial conversion stage or the only mixing polarization available in the medium (Millar et al., 2021, Perna et al., 2012, Seong et al., 2023, Chen et al., 5 Jul 2026).

Seed meaning	Physical role	Representative setting
Injected coherent field	Interference, regeneration, readout	Penning-trap LSW, short-pulse LSW, WINTER
Ambient photon population	Supplies the mode that mixes	Neutron-star thermal O-mode, GRB source photons
Structured field or plasma	Maintains phase matching or selects eigenmode	AMR magnets, magnetospheres, primordial magnetic fields

This semantic spread is not incidental. In strongly magnetized anisotropic plasmas the seed must match the propagating plasma eigenmode rather than a vacuum transverse photon. In resonant-regeneration searches the seed is often a local oscillator or stored cavity field. In cosmological and astrophysical transport problems the “seed” is frequently the first conversion stage that moves energy into the axion channel, after which the axion propagates through otherwise opaque regions (Millar et al., 2021, Wang et al., 2023, Cuesta et al., 2023).

2. Common mixing formalism and resonance engineering

A standard baseline description uses a two-level system for one photon mode $A$ and the axion $a$,

$$i\,\partial_z \begin{pmatrix} A\ a \end{pmatrix} = \begin{pmatrix} \Delta_\gamma & \Delta_{a\gamma}\ \Delta_{a\gamma} & \Delta_a \end{pmatrix} \begin{pmatrix} A\ a \end{pmatrix},$$

with

$$\Delta_{a\gamma}=\frac{1}{2}g_{a\gamma}B_\perp,\qquad \Delta_a=-\frac{m_a^2}{2\omega},\qquad \Delta_\gamma=-\frac{\omega_p^2}{2\omega}+\Delta_{\rm QED},$$

and

$$\tan(2\theta_{\rm mix})=\frac{2\Delta_{a\gamma}}{\Delta_\gamma-\Delta_a}.$$

In this baseline picture resonance occurs when $\Delta_\gamma=\Delta_a$, i.e.

$\omega_p^2 \approx m_a^2,$

and adiabaticity is quantified by

$$\gamma_{\rm ad}=\frac{2\Delta_{a\gamma}^2}{\left|\,d(\Delta_\gamma-\Delta_a)/ds\,\right|}, \qquad P^{\rm LZ}_{a\to\gamma}=1-\exp\!\left[-\frac{\pi}{2}\gamma_{\rm ad}\right].$$

These relations underlie both astrophysical conversion estimates and laboratory designs (Millar et al., 2021, Perna et al., 2012).

The 3D geometric-optics treatment generalizes the baseline by integrating along the actual photon worldline rather than along a fixed Cartesian axis. For a photon eigenmode with polarization $\hat{\boldsymbol{\epsilon}}$, the WKB transport equation yields an amplitude integral along the curved ray, and stationary phase gives the resonant conversion probability

$A$0

This formulation incorporates refractive ray bending, polarization selection, and inhomogeneous dispersion without collapsing the problem to one dimension (McDonald et al., 2024).

Seeded schemes often attempt to cancel the phase mismatch explicitly. Axion Magnetic Resonance achieves this by modulating the external magnetic field spatially or temporally. For a periodically modulated transverse field,

$A$1

the resonant condition is

$A$2

and the conversion probability near phase matching becomes

$A$3

A helical magnetic profile gives the same basic effect through a geometric shift $A$4 in the effective detuning, and the paper argues that this can extend the projected ALPS II reach in $A$5 by two orders of magnitude at $A$6 (Seong et al., 2023).

3. Strongly magnetized anisotropic plasmas

In strongly magnetized plasma, the mode that mixes with the axion is generally not a purely transverse vacuum photon. Millar, Baum, Lawson, and Marsh showed that the relevant propagating mode is the Langmuir–O mode, whose polarization has both transverse and longitudinal electric components. In that setting the axion-driven envelope obeys

$A$7

with the evolution coordinate $A$8 itself depending on the anisotropic plasma tensor. The crucial point is that the LO-mode amplitude evolves along $A$9, not along the axion momentum direction $a$0, and that $a$1 is locally perpendicular to the LO polarization. One-dimensional intuition therefore fails both kinematically and polarimetrically (Millar et al., 2021).

The resonance condition is correspondingly modified to

$a$2

which reduces to $a$3 in the nonrelativistic limit but interpolates between longitudinal and transverse resonances as $a$4 varies. The stationary-phase conversion length is

$a$5

and in the nonrelativistic, slowly varying limit simplifies to

$a$6

Large conversion is therefore associated not only with strong $a$7 but also with weak gradients along the anisotropic evolution direction (Millar et al., 2021).

For seeded operation this changes the design rules. A coherent injected field must be LO-like, must maximize $a$8, and must be aligned with the $a$9-evolution direction rather than with the axion momentum. In the nonrelativistic limit the near-resonance flux-transfer ratio becomes

$$i\,\partial_z \begin{pmatrix} A\ a \end{pmatrix} = \begin{pmatrix} \Delta_\gamma & \Delta_{a\gamma}\ \Delta_{a\gamma} & \Delta_a \end{pmatrix} \begin{pmatrix} A\ a \end{pmatrix},$$0

whereas the 1D estimate is

$$i\,\partial_z \begin{pmatrix} A\ a \end{pmatrix} = \begin{pmatrix} \Delta_\gamma & \Delta_{a\gamma}\ \Delta_{a\gamma} & \Delta_a \end{pmatrix} \begin{pmatrix} A\ a \end{pmatrix},$$1

For neutron-star parameters such as $$i\,\partial_z \begin{pmatrix} A\ a \end{pmatrix} = \begin{pmatrix} \Delta_\gamma & \Delta_{a\gamma}\ \Delta_{a\gamma} & \Delta_a \end{pmatrix} \begin{pmatrix} A\ a \end{pmatrix},$$2, $$i\,\partial_z \begin{pmatrix} A\ a \end{pmatrix} = \begin{pmatrix} \Delta_\gamma & \Delta_{a\gamma}\ \Delta_{a\gamma} & \Delta_a \end{pmatrix} \begin{pmatrix} A\ a \end{pmatrix},$$3, $$i\,\partial_z \begin{pmatrix} A\ a \end{pmatrix} = \begin{pmatrix} \Delta_\gamma & \Delta_{a\gamma}\ \Delta_{a\gamma} & \Delta_a \end{pmatrix} \begin{pmatrix} A\ a \end{pmatrix},$$4, and $$i\,\partial_z \begin{pmatrix} A\ a \end{pmatrix} = \begin{pmatrix} \Delta_\gamma & \Delta_{a\gamma}\ \Delta_{a\gamma} & \Delta_a \end{pmatrix} \begin{pmatrix} A\ a \end{pmatrix},$$5, the full 3D flux transfer can differ by up to three orders of magnitude from the 1D estimate, and regions disfavored in 1D can become highly efficient once anisotropy is included (Millar et al., 2021).

4. Astrophysical and cosmological realizations

In magnetized neutron stars, seeded conversion appears in two distinct ways. For thermal emission, the seed is the polarized atmosphere radiation itself. Only the ordinary mode, with electric field in the $$i\,\partial_z \begin{pmatrix} A\ a \end{pmatrix} = \begin{pmatrix} \Delta_\gamma & \Delta_{a\gamma}\ \Delta_{a\gamma} & \Delta_a \end{pmatrix} \begin{pmatrix} A\ a \end{pmatrix},$$6 plane, mixes with the axion, whereas the extraordinary mode decouples. Magnetized atmosphere calculations including vacuum polarization show that photon-axion conversion can alter spectra, light curves, and polarization; among the identified signatures are an increase of the effective area of a hot spot as it rotates away from the line of sight, apparent radii that can be either larger or smaller than neutron-star equation-of-state limits, and inversion of the plane of polarization for phase-on views (Perna et al., 2012). In axion-dark-matter conversion near the magnetosphere, the seed can instead be an injected LO-like field or a trajectory choice that keeps $$i\,\partial_z \begin{pmatrix} A\ a \end{pmatrix} = \begin{pmatrix} \Delta_\gamma & \Delta_{a\gamma}\ \Delta_{a\gamma} & \Delta_a \end{pmatrix} \begin{pmatrix} A\ a \end{pmatrix},$$7 and $$i\,\partial_z \begin{pmatrix} A\ a \end{pmatrix} = \begin{pmatrix} \Delta_\gamma & \Delta_{a\gamma}\ \Delta_{a\gamma} & \Delta_a \end{pmatrix} \begin{pmatrix} A\ a \end{pmatrix},$$8 favorable along $$i\,\partial_z \begin{pmatrix} A\ a \end{pmatrix} = \begin{pmatrix} \Delta_\gamma & \Delta_{a\gamma}\ \Delta_{a\gamma} & \Delta_a \end{pmatrix} \begin{pmatrix} A\ a \end{pmatrix},$$9, with strong beaming and spin-modulated polarization at the axion line (Millar et al., 2021).

For extragalactic gamma rays the seeded logic is three-stage. In GRB221009A, photons in the host galaxy convert to axionlike particles in a cellular $$\Delta_{a\gamma}=\frac{1}{2}g_{a\gamma}B_\perp,\qquad \Delta_a=-\frac{m_a^2}{2\omega},\qquad \Delta_\gamma=-\frac{\omega_p^2}{2\omega}+\Delta_{\rm QED},$$0 magnetic field, the ALPs free-stream through the EBL-opaque intergalactic medium, and reconvert in the Milky Way field. In the source galaxy,

$$\Delta_{a\gamma}=\frac{1}{2}g_{a\gamma}B_\perp,\qquad \Delta_a=-\frac{m_a^2}{2\omega},\qquad \Delta_\gamma=-\frac{\omega_p^2}{2\omega}+\Delta_{\rm QED},$$1

and the overall survival fraction at Earth is

$$\Delta_{a\gamma}=\frac{1}{2}g_{a\gamma}B_\perp,\qquad \Delta_a=-\frac{m_a^2}{2\omega},\qquad \Delta_\gamma=-\frac{\omega_p^2}{2\omega}+\Delta_{\rm QED},$$2

For GRB221009A the paper finds penetration probabilities $$\Delta_{a\gamma}=\frac{1}{2}g_{a\gamma}B_\perp,\qquad \Delta_a=-\frac{m_a^2}{2\omega},\qquad \Delta_\gamma=-\frac{\omega_p^2}{2\omega}+\Delta_{\rm QED},$$3, with $$\Delta_{a\gamma}=\frac{1}{2}g_{a\gamma}B_\perp,\qquad \Delta_a=-\frac{m_a^2}{2\omega},\qquad \Delta_\gamma=-\frac{\omega_p^2}{2\omega}+\Delta_{\rm QED},$$4 saturating at $$\Delta_{a\gamma}=\frac{1}{2}g_{a\gamma}B_\perp,\qquad \Delta_a=-\frac{m_a^2}{2\omega},\qquad \Delta_\gamma=-\frac{\omega_p^2}{2\omega}+\Delta_{\rm QED},$$5 for $$\Delta_{a\gamma}=\frac{1}{2}g_{a\gamma}B_\perp,\qquad \Delta_a=-\frac{m_a^2}{2\omega},\qquad \Delta_\gamma=-\frac{\omega_p^2}{2\omega}+\Delta_{\rm QED},$$6, and viable parameters

$$\Delta_{a\gamma}=\frac{1}{2}g_{a\gamma}B_\perp,\qquad \Delta_a=-\frac{m_a^2}{2\omega},\qquad \Delta_\gamma=-\frac{\omega_p^2}{2\omega}+\Delta_{\rm QED},$$7

in the source-galaxy domain model (Wang et al., 2023). The same transport logic was proposed as an explanation for LHAASO multi-TeV and PeV events, with source-side conversion, negligible intergalactic reconversion for sufficiently weak $$\Delta_{a\gamma}=\frac{1}{2}g_{a\gamma}B_\perp,\qquad \Delta_a=-\frac{m_a^2}{2\omega},\qquad \Delta_\gamma=-\frac{\omega_p^2}{2\omega}+\Delta_{\rm QED},$$8, and Milky-Way reconversion before detection (Zhang et al., 2022).

Around black holes with superradiant axion clouds, the seed is the background magnetic field. For a uniform field the cloud decay rate scales as

$$\Delta_{a\gamma}=\frac{1}{2}g_{a\gamma}B_\perp,\qquad \Delta_a=-\frac{m_a^2}{2\omega},\qquad \Delta_\gamma=-\frac{\omega_p^2}{2\omega}+\Delta_{\rm QED},$$9

while for a monopole background it scales as

$$\tan(2\theta_{\rm mix})=\frac{2\Delta_{a\gamma}}{\Delta_\gamma-\Delta_a}.$$0

For the Galactic-center black hole, the uniform-field decay rate is reported to be comparable to the superradiant growth rate for $$\tan(2\theta_{\rm mix})=\frac{2\Delta_{a\gamma}}{\Delta_\gamma-\Delta_a}.$$1, $$\tan(2\theta_{\rm mix})=\frac{2\Delta_{a\gamma}}{\Delta_\gamma-\Delta_a}.$$2, and $$\tan(2\theta_{\rm mix})=\frac{2\Delta_{a\gamma}}{\Delta_\gamma-\Delta_a}.$$3, whereas the monopole case is larger by $$\tan(2\theta_{\rm mix})=\frac{2\Delta_{a\gamma}}{\Delta_\gamma-\Delta_a}.$$4 at the same parameters (Yoo et al., 2021). With multipole magnetic backgrounds the angular structure itself becomes the seed: the conversion rate

$$\tan(2\theta_{\rm mix})=\frac{2\Delta_{a\gamma}}{\Delta_\gamma-\Delta_a}.$$5

depends sharply on the multipole order $$\tan(2\theta_{\rm mix})=\frac{2\Delta_{a\gamma}}{\Delta_\gamma-\Delta_a}.$$6, and the coefficient $$\tan(2\theta_{\rm mix})=\frac{2\Delta_{a\gamma}}{\Delta_\gamma-\Delta_a}.$$7 in $$\tan(2\theta_{\rm mix})=\frac{2\Delta_{a\gamma}}{\Delta_\gamma-\Delta_a}.$$8 peaks near $$\tan(2\theta_{\rm mix})=\frac{2\Delta_{a\gamma}}{\Delta_\gamma-\Delta_a}.$$9, so that conversion can compete with or exceed superradiant growth (Sakurai et al., 2023).

Cosmological implementations use the background magnetic field itself as the seed. During Big Bang Nucleosynthesis, resonant conversion of CMB photons into a majoron-like ALP in a primordial field $\Delta_\gamma=\Delta_a$0 occurs when $\Delta_\gamma=\Delta_a$1. In the baseline model with $\Delta_\gamma=\Delta_a$2, $\Delta_\gamma=\Delta_a$3, and $\Delta_\gamma=\Delta_a$4, the resonance at $\Delta_\gamma=\Delta_a$5 converts $\Delta_\gamma=\Delta_a$6 of photons and $\Delta_\gamma=\Delta_a$7 of the CMB energy, lowering the photon temperature by $\Delta_\gamma=\Delta_a$8, increasing $\Delta_\gamma=\Delta_a$9 from $\omega_p^2 \approx m_a^2,$0 to $\omega_p^2 \approx m_a^2,$1, and yielding $\omega_p^2 \approx m_a^2,$2 with $\omega_p^2 \approx m_a^2,$3 in the fitted ALM cosmology (Cuesta et al., 2023). A different cosmological realization uses resonant ALP-to-photon conversion to explain the ARCADE2 radio excess and the EDGES 21 cm absorption depth. There the brightness temperature scales as

$\omega_p^2 \approx m_a^2,$4

and the resulting excess obeys $\omega_p^2 \approx m_a^2,$5 over $\omega_p^2 \approx m_a^2,$6–$\omega_p^2 \approx m_a^2,$7, with an additional absorption trough predicted below $\omega_p^2 \approx m_a^2,$8 for higher-redshift resonances (Addazi et al., 2024).

5. Laboratory implementations

Microwave light-shining-through-wall proposals supplied some of the earliest explicitly seeded laboratory architectures. STAX drives a high-$\omega_p^2 \approx m_a^2,$9 Fabry–Perot cavity with a $$\gamma_{\rm ad}=\frac{2\Delta_{a\gamma}^2}{\left|\,d(\Delta_\gamma-\Delta_a)/ds\,\right|}, \qquad P^{\rm LZ}_{a\to\gamma}=1-\exp\!\left[-\frac{\pi}{2}\gamma_{\rm ad}\right].$$0 gyrotron or klystron, uses a second high-$$\gamma_{\rm ad}=\frac{2\Delta_{a\gamma}^2}{\left|\,d(\Delta_\gamma-\Delta_a)/ds\,\right|}, \qquad P^{\rm LZ}_{a\to\gamma}=1-\exp\!\left[-\frac{\pi}{2}\gamma_{\rm ad}\right].$$1 cavity for resonant regeneration, and detects regenerated single photons with a TES at $$\gamma_{\rm ad}=\frac{2\Delta_{a\gamma}^2}{\left|\,d(\Delta_\gamma-\Delta_a)/ds\,\right|}, \qquad P^{\rm LZ}_{a\to\gamma}=1-\exp\!\left[-\frac{\pi}{2}\gamma_{\rm ad}\right].$$2. The reference source powers are $$\gamma_{\rm ad}=\frac{2\Delta_{a\gamma}^2}{\left|\,d(\Delta_\gamma-\Delta_a)/ds\,\right|}, \qquad P^{\rm LZ}_{a\to\gamma}=1-\exp\!\left[-\frac{\pi}{2}\gamma_{\rm ad}\right].$$3 and $$\gamma_{\rm ad}=\frac{2\Delta_{a\gamma}^2}{\left|\,d(\Delta_\gamma-\Delta_a)/ds\,\right|}, \qquad P^{\rm LZ}_{a\to\gamma}=1-\exp\!\left[-\frac{\pi}{2}\gamma_{\rm ad}\right].$$4, the cavities have $$\gamma_{\rm ad}=\frac{2\Delta_{a\gamma}^2}{\left|\,d(\Delta_\gamma-\Delta_a)/ds\,\right|}, \qquad P^{\rm LZ}_{a\to\gamma}=1-\exp\!\left[-\frac{\pi}{2}\gamma_{\rm ad}\right].$$5, and the projected reach improves present laboratory exclusion limits by at least four orders of magnitude for $$\gamma_{\rm ad}=\frac{2\Delta_{a\gamma}^2}{\left|\,d(\Delta_\gamma-\Delta_a)/ds\,\right|}, \qquad P^{\rm LZ}_{a\to\gamma}=1-\exp\!\left[-\frac{\pi}{2}\gamma_{\rm ad}\right].$$6 (Ferretti, 2016).

A more recent seeded LSW scheme replaces radiometric readout by Penning-trap field sensing. In that design coherent microwaves are stored in a rectangular $$\gamma_{\rm ad}=\frac{2\Delta_{a\gamma}^2}{\left|\,d(\Delta_\gamma-\Delta_a)/ds\,\right|}, \qquad P^{\rm LZ}_{a\to\gamma}=1-\exp\!\left[-\frac{\pi}{2}\gamma_{\rm ad}\right].$$7 cavity, axions are generated in a production field $$\gamma_{\rm ad}=\frac{2\Delta_{a\gamma}^2}{\left|\,d(\Delta_\gamma-\Delta_a)/ds\,\right|}, \qquad P^{\rm LZ}_{a\to\gamma}=1-\exp\!\left[-\frac{\pi}{2}\gamma_{\rm ad}\right].$$8, and regenerated RF fields are read out by a Penning-trapped ion crystal in a $$\gamma_{\rm ad}=\frac{2\Delta_{a\gamma}^2}{\left|\,d(\Delta_\gamma-\Delta_a)/ds\,\right|}, \qquad P^{\rm LZ}_{a\to\gamma}=1-\exp\!\left[-\frac{\pi}{2}\gamma_{\rm ad}\right].$$9 receiver. In the coherent limit,

$\hat{\boldsymbol{\epsilon}}$0

and the total LSW probability is their product. For $\hat{\boldsymbol{\epsilon}}$1, $\hat{\boldsymbol{\epsilon}}$2, $\hat{\boldsymbol{\epsilon}}$3, $\hat{\boldsymbol{\epsilon}}$4, $\hat{\boldsymbol{\epsilon}}$5, $\hat{\boldsymbol{\epsilon}}$6, $\hat{\boldsymbol{\epsilon}}$7, $\hat{\boldsymbol{\epsilon}}$8, $\hat{\boldsymbol{\epsilon}}$9, and one day of averaging, the projected sensitivity is

$A$00

The method is explicitly seeded because the production cavity provides the coherent pump and the receiver is tuned to the regenerated mode rather than to broadband power (Chen et al., 5 Jul 2026).

WINTER uses seeded interferometry rather than regeneration. It places a $A$01, $A$02 magnetic arm with a Fabry–Perot cavity of finesse $A$03 inside a Mach–Zehnder interferometer operated near a dark fringe with $A$04. The dark-port observable after amplitude modulation at $A$05 is

$A$06

With $A$07, $A$08, $A$09, $A$10, and one year of integration, the projected sensitivity is $A$11 up to $A$12 (Batllori et al., 20 Sep 2025).

Short-pulse LSW motivates a different seeded strategy because high-finesse cavities cannot ring up on fs–ps timescales. Injecting a coherent seed into the regeneration region gives

$A$13

and the detected photon-number variation is

$A$14

so that

$A$15

For constructive interference $A$16 and $A$17,

$A$18

At fixed repetition rate and integration time this changes the coupling scaling from the unseeded $A$19 to

$A$20

which is the central advantage when cavity enhancement is unavailable (An et al., 25 Sep 2025).

Laser-driven wakefields provide yet another seeded architecture. There the drive laser is itself the seed, the wakefield supplies effective fields $A$21, and the produced axions can reconvert into axion-regenerated electromagnetic fields with distinctive polarization, frequency, and transverse-mode content. For a $A$22 guided interaction the paper reports $A$23, compared with $A$24 for ALPS-II, and argues that replacing the generation stage of a conventional LSW setup by the wakefield stage can reach $A$25. The regenerated field carries specific Laguerre–Gaussian signatures, including $A$26 and $A$27 components at $A$28, $A$29, and $A$30, which serve as intrinsic seed tags for filtering (An et al., 16 Apr 2025).

6. Limitations, misconceptions, and disputed points

A central misconception is that any seed photon field must increase the average conversion rate. The quantum-field treatment of dielectric-interface conversion shows otherwise. In that framework the average photon production rate is governed by the overlap integral with the Garibian photon wave function, and if the final photon mode already contains $A$31 quanta then spontaneous plus stimulated emission scales as $A$32 while stimulated absorption scales as $A$33, so the net average rate is proportional to $A$34. A coherent seed is therefore useful as a local oscillator or mode-matching tool, but it does not by itself increase the average conversion power unless nonreciprocal or dissipative elements are added (Ioannisian et al., 2017).

Another persistent misconception is that one-dimensional transverse mixing remains adequate in strongly magnetized plasmas. The 3D anisotropic calculations show that the correct evolution occurs along curved photon worldlines and, in neutron-star magnetospheres, along the LO-mode direction $A$35 rather than the axion direction $A$36. Absorption, scattering, refraction, curvature-induced dephasing, and WKB breakdown near regions such as the $A$37 throat can all invalidate simplified seeded designs that ignore full ray tracing (Millar et al., 2021, McDonald et al., 2024).

The notion of seeding also changes in parametric-instability problems. In axion dark matter with photon-pair production, a tiny seed would classically trigger rapid coherent mixing, but the many-mode quantum treatment finds that commutators supply an intrinsic “no-seed” source term, eliminating the need for external seeding and reducing the logarithmic break-time factor from $A$38 to $A$39 in the multimode case (Sawyer, 2018). In the related stability analysis with a background magnetic field, the coupled axion-photon system becomes Mathieu-like, the usual instability bands are shifted, and new bands appear, including the $A$40 band and the $A$41-shifted structures near $A$42 (Masaki et al., 2019). In such systems, “seeded” and “self-seeded” are not operationally equivalent.

Magnetic-field configuration has also been contentious. For domain and helical field networks, the asymptotic behavior of the polarization variances is insensitive to the detailed configuration, but the early transient dynamics are not. The helical-domain analysis finds that the “peculiar” behavior previously claimed for continuous helical connections is not generic: when helicity is randomized domain by domain, the results approach those of the conventional random-domain model, and the difference is traced to the domain-connection prescription rather than to a fundamentally new conversion mechanism (Masaki et al., 2017).

Taken together, these results define a technical criterion for the term. A seeded axion-photon conversion scheme is not merely any setup with a background field. It is a scheme in which the seed determines the accessible eigenmode, the phase-matching geometry, or the readout channel, and whose efficacy is controlled by the full mode structure, resonance sweep rate, and loss mechanism of the medium. That criterion is now common to neutron-star LO-mode conversion, source-seeded gamma-ray transport, resonant-regeneration searches, interferometric amplitude experiments, wakefield generation, and cosmological resonance models (Millar et al., 2021, Batllori et al., 20 Sep 2025).