Dielectric Haloscope: Boosted Axion Detection

• Dielectric haloscope is an axion-to-photon converter that employs a mirror and dielectric disk stack to generate constructive interference from weak electromagnetic emissions.
• It uses a multilayer design where tailored disk spacings and material properties create a boost factor of up to ~5×10⁴ over a calibrated bandwidth.
• The technology bridges axion and dark-photon searches, with prototypes demonstrating precise mechanical control and resonant enhancement across microwave to optical regimes.

A dielectric haloscope is a dark-matter haloscope in which a mirror and a stack of dielectric disks are placed in a strong external magnetic field so that axion-induced electromagnetic radiation from many dielectric interfaces adds constructively and can be detected as a boosted microwave or millimeter-wave signal. In the foundational MADMAX formulation, the concept was proposed to cover the high-frequency range of $10\text{--}100$ GHz, corresponding to axion masses of $40\text{--}400~\mu\mathrm{eV}$, where traditional cavity resonators have difficulties reaching the required volume (Group et al., 2016, Millar et al., 2016). In a broader experimental usage, related multilayer dielectric stacks have also been used as dielectric haloscopes for dark-photon dark matter in the millimeter-wave, optical, and near-optical regimes (Wei et al., 30 Mar 2025, Manenti et al., 2021, Li et al., 3 Jul 2026).

1. Definition, scope, and conceptual variants

In the strict MADMAX sense, a dielectric haloscope is an open, broadband axion-to-photon converter built from a mirror plus a stack of dielectric disks in a magnetic field. The purpose of the stack is not to support a single closed cavity eigenmode, but to create constructive interference and resonant build-up from the weak electromagnetic waves emitted at many dielectric boundaries. This is the sense used in the original proposals “MADMAX: A new Dark Matter Axion Search using a Dielectric Haloscope” (Majorovits et al., 2016), “Dielectric Haloscopes: A New Way to Detect Axion Dark Matter” (Group et al., 2016), and “Dielectric Haloscopes to Search for Axion Dark Matter: Theoretical Foundations” (Millar et al., 2016).

The foundational MADMAX design in “MADMAX: A new way of probing QCD Axion Dark Matter with a Dielectric Haloscope -- Foundations” specifies 80 high dielectric disks in a $10$ T magnetic field, with disk thickness $d = 1\ {\rm mm}$, disk area $A = 1\ {\rm m^2}$, high dielectric constant $\epsilon \approx 25$, loss tangent $\tan\delta \sim 10^{-5}$, and an achievable boost factor of about $\beta^2 \sim 5\times 10^4$ over a bandwidth of about $50\ {\rm MHz}$ (Knirck, 2017). The same literature emphasizes that the dielectric haloscope is intended to probe a mass region that is difficult for conventional cavity searches, especially the range around $m_a \sim 100~\mu\mathrm{eV}$ and, more broadly, $40\text{--}400~\mu\mathrm{eV}$0 (Majorovits et al., 2016, Group et al., 2016).

The term has also been extended to dark-photon searches in which the same interface-conversion principle is exploited without an external magnetic field. “Search for dark photons using a multilayer dielectric haloscope equipped with a single-photon avalanche diode” uses a 23-bilayer SiO$40\text{--}400~\mu\mathrm{eV}$1/Si$40\text{--}400~\mu\mathrm{eV}$2N$40\text{--}400~\mu\mathrm{eV}$3 multilayer stack at the eV scale (Manenti et al., 2021). “New Constraints on Dark Photon Dark Matter with a Millimeter-Wave Dielectric Haloscope” uses four LaAlO$40\text{--}400~\mu\mathrm{eV}$4 dielectric disks and a mirror in the W band (Wei et al., 30 Mar 2025). “Searching for Dark Photons with a room-temperature dielectric haloscope” uses a 47-pair TiO$40\text{--}400~\mu\mathrm{eV}$5/SiO$40\text{--}400~\mu\mathrm{eV}$6 stack with a CMOS focal plane near $40\text{--}400~\mu\mathrm{eV}$7 eV (Li et al., 3 Jul 2026).

A persistent terminological boundary separates these devices from dielectric-loaded resonant cavities. “A high-Q microwave dielectric resonator for axion dark matter haloscopes” explicitly states that it is not a dielectric haloscope in the MADMAX sense; it is a microwave haloscope resonator that uses dielectric structures inside a copper cavity (Vora et al., 2022). That distinction is reinforced by “Tunable Super-Mode Dielectric Resonators for Axion Haloscopes,” which treats dielectric-engineered TM resonators, super-modes, and Bragg-like confinement inside cylindrical conducting cavities rather than mirror-plus-disk interface boosters (McAllister et al., 2017).

2. Axion electrodynamics and interface emission

The theoretical starting point is the axion-photon interaction,

$40\text{--}400~\mu\mathrm{eV}$8

together with axion-modified Maxwell equations in an external magnetic field (Millar et al., 2016, Millar et al., 2017). For non-relativistic galactic axion dark matter, the field can be treated as a classical oscillating source, and in a homogeneous medium the axion-induced electric field is

$40\text{--}400~\mu\mathrm{eV}$9

Because this induced field depends on the dielectric constant, it is discontinuous across an interface where $10$0 changes (Knirck et al., 2021).

That discontinuity is the central physical mechanism of the dielectric haloscope. Maxwell boundary conditions require the tangential electric field to remain continuous, so the interface emits ordinary electromagnetic waves to restore the full solution. The original theoretical papers emphasize that every interface between different dielectric media inside the magnetic field radiates weakly, and that the emitted waves emerge in both directions perpendicular to the surface (Millar et al., 2016, Group et al., 2016). A mirror is then used to force the useful power toward the receiver, so the device acts as a one-sided photon booster (Knirck, 2017).

For a single metallic mirror in vacuum, the emitted power per unit area provides the natural normalization. One formulation gives

$10$1

showing that the raw signal is extremely small (Millar et al., 2016). Another formulation states that the power emitted by a single boundary is

$10$2

with $10$3 describing the dielectric contrast (Knirck, 2017). The need for a large enhancement factor follows directly from these scales.

The dark-photon variant uses the same interface logic but a different source term. In the W-band dielectric haloscope, the dark photon is described by

$10$4

and the emitted power from the stack is written as

$10$5

For a perfect half-wave stack at the center frequency,

$10$6

which makes the coherent $10$7 enhancement explicit (Wei et al., 30 Mar 2025).

3. Multilayer booster physics, boost factor, and bandwidth

The defining figure of merit of a dielectric haloscope is the boost factor $10$8, usually defined as the ratio of the emitted power from the multilayer system to the power emitted by a mirror alone (Knirck, 2017, Egge et al., 2020). In practical terms, $10$9 measures how much the dielectric structure amplifies the axion signal. The multilayer stack functions by combining two effects that the foundational papers treat on equal footing: coherent addition of interface-emitted waves and resonant build-up between partially reflecting surfaces (Group et al., 2016, Millar et al., 2016).

The MADMAX literature stresses that disk thicknesses and especially disk spacings determine whether the booster operates in a more transparent, coherently summed regime or in a more resonant regime. A single disk becomes transparent for special optical thicknesses, and a many-disk stack can then be configured so that emissions from all interfaces arrive in phase. In that transparent-mode limit, the boost scales as $d = 1\ {\rm mm}$0 while the bandwidth scales as $d = 1\ {\rm mm}$1, so the integrated response remains constrained (Millar et al., 2016).

This tradeoff is formalized by the Area Law: $d = 1\ {\rm mm}$2 or equivalently that one cannot arbitrarily increase both peak boost and bandwidth at the same time (Knirck, 2017, Group et al., 2016). The practical consequence is that a broad discovery scan uses a moderate boost spread over a wider band, whereas a follow-up rescan can use a narrower but larger peak boost. The MADMAX papers explicitly frame scan strategy in these terms (Knirck, 2017).

The same logic appears in small-scale booster demonstrations. In the proof-of-principle MADMAX booster, the relevant experimental observable is the reflectivity phase and its group delay,

$d = 1\ {\rm mm}$3

because the group delay is correlated with the boost factor even when the axion signal itself cannot be measured directly in the laboratory (Egge et al., 2020). In the DALI program, the multilayer system is described as a Fabry–Pérot-type dielectric haloscope, and the quality factor is defined as

$d = 1\ {\rm mm}$4

In that formulation, $d = 1\ {\rm mm}$5 is the average photon lifetime or, equivalently, the number of effective round trips in the resonator (Hernández-Cabrera et al., 2024).

The power scale targeted by large axion dielectric haloscopes is often quoted as

$d = 1\ {\rm mm}$6

which makes clear why large area, strong field, and large boost are simultaneously required (Egge et al., 2020, Knirck et al., 2021).

4. Modeling, tuning, reciprocity, and 3D tolerances

The first-generation theoretical description of dielectric haloscopes is predominantly one-dimensional. Both the MADMAX foundations and the proof-of-principle booster use a transfer-matrix formalism in which each dielectric layer and vacuum gap is represented by propagation and boundary matrices, allowing the boost factor and reflectivity to be computed as functions of spacing and frequency (Knirck, 2017, Egge et al., 2020). In the proof-of-principle system, the 1D model was implemented in circuit-simulator language (ADS), and the measured group delays agreed well with 1D calculations (Egge et al., 2020).

A major refinement is the transition from idealized 1D stacks to real 3D finite-aperture devices. “Simulating MADMAX in 3D: Requirements for Dielectric Axion Haloscopes” finds that finite disk diameter reduces the emitted power by up to $d = 1\ {\rm mm}$7 compared to earlier 1D calculations; this reduction is interpreted as a geometrical form factor rather than diffraction loss (Knirck et al., 2021). The same study derives the emitted beam shape and finds that the fundamental mode is very close to a Gaussian beam with waist $d = 1\ {\rm mm}$8, while a Gaussian antenna matched to that beam has a frequency-independent coupling of about $d = 1\ {\rm mm}$9 to the fundamental mode (Knirck et al., 2021).

That 3D study also establishes the mechanical tolerances required for a benchmark boost factor with a bandwidth of $A = 1\ {\rm m^2}$0 at $A = 1\ {\rm m^2}$1. The maximum allowed disk tilt is $A = 1\ {\rm m^2}$2 divided by the disk diameter, the required disk planarity is $A = 1\ {\rm m^2}$3 (min-to-max) or better, and the maximum allowed surface roughness is $A = 1\ {\rm m^2}$4 (min-to-max) (Knirck et al., 2021). Realistic dark matter axion velocities of $A = 1\ {\rm m^2}$5 and magnetic-field inhomogeneities at the scale of $A = 1\ {\rm m^2}$6 have negligible impact on the sensitivity of MADMAX in that benchmark configuration (Knirck et al., 2021).

The dedicated velocity analysis sharpens that conclusion. “Dielectric haloscopes: sensitivity to the axion dark matter velocity” finds that velocity effects become important only when the haloscope size is larger than about $A = 1\ {\rm m^2}$7 of the axion de Broglie wavelength, and concludes that for the planned MADMAX experiment with 80 dielectric disks the velocity dependence can safely be neglected (Millar et al., 2017). The same work argues that an augmented MADMAX or a second generation experiment would be directionally sensitive to the axion velocity and therefore a sensitive measure of axion astrophysics (Millar et al., 2017). This suggests a future transition from discovery-mode haloscopes to directional axion astronomy if the physical size of the booster is increased sufficiently.

Calibration of open dielectric haloscopes introduces additional complications not captured by idealized multilayer models. “Experimental determination of axion signal power of dish antennas and dielectric haloscopes using the reciprocity approach” uses a bead-pull measurement and a reciprocity relation to infer the expected axion signal power directly from a reflection-induced test field (Egge et al., 2023). That study quantifies, for the first time in a dielectric haloscope, the effect of antenna standing waves and higher order mode perturbations. In both the dish antenna and the minimal dielectric haloscope, antenna-booster resonances can alter the signal power by up to about $A = 1\ {\rm m^2}$8 relative to the baseline (Egge et al., 2023). This is a central practical result because large open systems do not admit a simple discrete-eigenmode calibration.

Closed dielectric haloscopes permit a different modeling strategy. “Sensitivity of a closed dielectric haloscope to axion dark matter” develops a simple transmission-line-based model for the CB200 prototype and uses it to infer the boost factor and coupling reach from reflectivity and noise spectra rather than from expensive full 3D simulations (Ivanov et al., 5 Mar 2026). In that formulation, the axion coupling is mediated by a TE$A = 1\ {\rm m^2}$9-like booster mode and a transverse overlap factor

$\epsilon \approx 25$0

which functions as the transverse analog of a cavity form factor (Ivanov et al., 5 Mar 2026).

5. Axion-oriented realizations and measured performance

Experimental work on dielectric haloscopes has proceeded through staged realizations. The first MADMAX proof-of-principle booster consisted of a plane copper mirror and up to five sapphire disks, each $\epsilon \approx 25$1 in diameter and $\epsilon \approx 25$2 thick, probed in the range $\epsilon \approx 25$3 GHz (Egge et al., 2020). Its measured group delays agreed well with 1D calculations, the repeatability of the tuning was at the percent level, the boost-factor frequency uncertainty was less than $\epsilon \approx 25$4 MHz, and the boost-factor amplitude uncertainty was less than $\epsilon \approx 25$5 (Egge et al., 2020). The same study showed that tuning four equally spaced disks from about $\epsilon \approx 25$6 to $\epsilon \approx 25$7 moves the boost factor across roughly $\epsilon \approx 25$8 to $\epsilon \approx 25$9 (Egge et al., 2020).

The mechanical feasibility of a cryogenic, field-compatible booster was demonstrated by “First mechanical realization of a tunable dielectric haloscope for the MADMAX axion search experiment” (Collaboration et al., 2024). The OB200 prototype uses one movable sapphire disk of $\tan\delta \sim 10^{-5}$0 diameter plus a copper mirror of the same diameter, actuated by three piezoelectric motors and monitored by an interferometer. It was tested in a magnetic field up to $\tan\delta \sim 10^{-5}$1 T and at cryogenic temperatures down to $\tan\delta \sim 10^{-5}$2 K (Collaboration et al., 2024). At room temperature and no magnetic field, the motors reached their target positions after each $\tan\delta \sim 10^{-5}$3 jump with better than $\tan\delta \sim 10^{-5}$4 precision and final position stability better than $\tan\delta \sim 10^{-5}$5; under the $\tan\delta \sim 10^{-5}$6 T field the behavior was essentially unchanged (Collaboration et al., 2024). In cryogenic operation, the target-position precision improved to better than $\tan\delta \sim 10^{-5}$7, the relative position of motor 1 with respect to the other motors was always within $\tan\delta \sim 10^{-5}$8, and the inter-motor spacing remained within $\tan\delta \sim 10^{-5}$9 (Collaboration et al., 2024).

The DALI program represents another branch of dielectric-haloscope development. “Echo-free quality factor of a multilayer axion haloscope” uses fixed-plate Fabry–Pérot resonators with $\beta^2 \sim 5\times 10^4$0 yttria-stabilized zirconia layers, each $\beta^2 \sim 5\times 10^4$1 and $\beta^2 \sim 5\times 10^4$2 thick, with plate spacings of 6.04 mm and 6.21 mm (Hernández-Cabrera et al., 2024). An anechoic chamber permits echo-free determination of $\beta^2 \sim 5\times 10^4$3, and for $\beta^2 \sim 5\times 10^4$4 the experiment observes $\beta^2 \sim 5\times 10^4$5 around 7 GHz (Hernández-Cabrera et al., 2024). The frequency-normalized enhancement scales approximately linearly with the number of layers, with fitted slopes $\beta^2 \sim 5\times 10^4$6 and $\beta^2 \sim 5\times 10^4$7 for two devices (Hernández-Cabrera et al., 2024). The paper projects that the full-scale DALI could reach $\beta^2 \sim 5\times 10^4$8 and sensitivity at $\beta^2 \sim 5\times 10^4$9 over the entire $50\ {\rm MHz}$0 range (Hernández-Cabrera et al., 2024).

The first axion-search interpretation framework for a closed dielectric haloscope is developed in the CB200 analysis (Ivanov et al., 5 Mar 2026). The prototype consists of three sapphire disks inside an aluminum cylinder, with a taper and dielectric lens to couple the booster mode to the receiver chain (Ivanov et al., 5 Mar 2026). Applied to data taken in the $50\ {\rm MHz}$1 T Morpurgo dipole magnet at CERN, the model yields boost factors reaching about

$50\ {\rm MHz}$2

with combined relative uncertainty of roughly $50\ {\rm MHz}$3, for two configurations centered near $50\ {\rm MHz}$4 GHz and $50\ {\rm MHz}$5 GHz (Ivanov et al., 5 Mar 2026). The authors state that this work underpins the first axion dark matter search using a dielectric haloscope (Ivanov et al., 5 Mar 2026).

6. Dark-photon implementations and broadening of the concept

Dark-photon searches have extended dielectric-haloscope methods beyond the original axion setting because they do not require an external magnetic field. In the optical regime, MuDHI uses a 23-bilayer stack of alternating SiO$50\ {\rm MHz}$6 and Si$50\ {\rm MHz}$7N$50\ {\rm MHz}$8 thin films with linearly increasing thicknesses, described as a mirrored chirped stack optimized around the peak sensitivity of the SPAD detector near $50\ {\rm MHz}$9 nm (Manenti et al., 2021). The search reports no significant signal excess, and the 90% CL upper bound reaches

$m_a \sim 100~\mu\mathrm{eV}$0

at

$m_a \sim 100~\mu\mathrm{eV}$1

with sensitivity over the region around $m_a \sim 100~\mu\mathrm{eV}$2 (Manenti et al., 2021).

In the millimeter-wave regime, the W-band dielectric haloscope of “New Constraints on Dark Photon Dark Matter with a Millimeter-Wave Dielectric Haloscope” uses four LaAlO$m_a \sim 100~\mu\mathrm{eV}$3 dielectric disks and one gold-coated aluminum mirror (Wei et al., 30 Mar 2025). The nominal geometry is disk thickness $m_a \sim 100~\mu\mathrm{eV}$4, spacing $m_a \sim 100~\mu\mathrm{eV}$5, holder inner diameter $m_a \sim 100~\mu\mathrm{eV}$6, and antenna aperture $m_a \sim 100~\mu\mathrm{eV}$7, with refractive index $m_a \sim 100~\mu\mathrm{eV}$8 (Wei et al., 30 Mar 2025). The search covers $m_a \sim 100~\mu\mathrm{eV}$9 to $40\text{--}400~\mu\mathrm{eV}$00, corresponding to $40\text{--}400~\mu\mathrm{eV}$01 to $40\text{--}400~\mu\mathrm{eV}$02, finds no evidence for dark photon dark matter, and sets 90% confidence limits with strongest bound

$40\text{--}400~\mu\mathrm{eV}$03

at $40\text{--}400~\mu\mathrm{eV}$04, together with $40\text{--}400~\mu\mathrm{eV}$05 across the full studied mass interval (Wei et al., 30 Mar 2025). The paper states that these constraints improve existing limits by more than two orders of magnitude in the reported range (Wei et al., 30 Mar 2025).

A room-temperature optical implementation is reported in “Searching for Dark Photons with a room-temperature dielectric haloscope” (Li et al., 3 Jul 2026). The SPECTRA experiment uses a 47-pair TiO$40\text{--}400~\mu\mathrm{eV}$06/SiO$40\text{--}400~\mu\mathrm{eV}$07 dielectric stack on a 1-inch sapphire substrate with a silver mirror underlayer and a Ta$40\text{--}400~\mu\mathrm{eV}$08O$40\text{--}400~\mu\mathrm{eV}$09 protective cap, read out by a cooled Sony IMX533CLK-D CMOS sensor (Li et al., 3 Jul 2026). The search uses 904 h of stack-on data and 404 h of stack-off background-control data, observes no statistically significant excess, and sets a 90% CL upper limit

$40\text{--}400~\mu\mathrm{eV}$10

at the most sensitive point, corresponding to a dark-photon mass of $40\text{--}400~\mu\mathrm{eV}$11 (Li et al., 3 Jul 2026). The same paper shows that the template-based spatial analysis is stronger than a pure counting analysis, which would have given $40\text{--}400~\mu\mathrm{eV}$12 (Li et al., 3 Jul 2026).

A further conceptual broadening appears in “Dark Matter Haloscope with a Disordered Dielectric Absorber,” which replaces a carefully engineered 1D periodic stack with a disordered dielectric powder (Koppell et al., 30 May 2025). The central claim is that, in the broadband limit, the relevant resource is the total surface area of dielectric interfaces. The paper summarizes the scanning-speed scaling as

$40\text{--}400~\mu\mathrm{eV}$13

and proposes DPHaSE, the Dielectric Powder Haloscope SNSPD Experiment (Koppell et al., 30 May 2025). The projected reach is in the $40\text{--}400~\mu\mathrm{eV}$14 mass range, sensitive to QCD axion-photon couplings and exceeding current constraints on dark photon dark matter by up to 5 orders of magnitude (Koppell et al., 30 May 2025). This suggests that the dielectric-haloscope principle can be generalized from ordered stacks to broadband, interface-dominated disordered media when the readout is based on single-photon detection.

7. Relation to dielectric resonator haloscopes and recurrent misconceptions

A common misconception is that any axion haloscope containing dielectric material is a dielectric haloscope. The literature draws a sharper distinction. MADMAX-style dielectric haloscopes rely on interface emission from a mirror-plus-disk stack and are usually analyzed through boost factors, transfer matrices, group delay, and open or weakly confined propagation (Millar et al., 2016, Egge et al., 2020). By contrast, dielectric-loaded resonator haloscopes remain cavity devices whose signal power is organized by the usual $40\text{--}400~\mu\mathrm{eV}$15 logic.

This distinction is explicit in “A high-Q microwave dielectric resonator for axion dark matter haloscopes,” which describes a right circular copper cavity containing two concentric hollow sapphire cylinders and targeting the TM$40\text{--}400~\mu\mathrm{eV}$16 mode at about $40\text{--}400~\mu\mathrm{eV}$17 GHz (Vora et al., 2022). Its total cavity volume is $40\text{--}400~\mu\mathrm{eV}$18, the relevant axion form factor is $40\text{--}400~\mu\mathrm{eV}$19, and the figure $40\text{--}400~\mu\mathrm{eV}$20 is $40\text{--}400~\mu\mathrm{eV}$21 (Vora et al., 2022). At $40\text{--}400~\mu\mathrm{eV}$22 K the measured $40\text{--}400~\mu\mathrm{eV}$23 is $40\text{--}400~\mu\mathrm{eV}$24, and at $40\text{--}400~\mu\mathrm{eV}$25 T the measured quality factor exceeds $40\text{--}400~\mu\mathrm{eV}$26, with the quality factor increasing by about $40\text{--}400~\mu\mathrm{eV}$27 relative to zero field (Vora et al., 2022). Under assumptions including $40\text{--}400~\mu\mathrm{eV}$28, $40\text{--}400~\mu\mathrm{eV}$29 T, quantum-limited readout, KSVZ sensitivity, and $40\text{--}400~\mu\mathrm{eV}$30, the estimated scan rate is about $40\text{--}400~\mu\mathrm{eV}$31, compared with $40\text{--}400~\mu\mathrm{eV}$32 for a conventional copper TM$40\text{--}400~\mu\mathrm{eV}$33 cavity at the same frequency (Vora et al., 2022). The paper is directly relevant to axion searches, but it belongs to resonant cavity haloscopes rather than to the non-cavity dielectric-haloscope class.

The same boundary appears in “Tunable Super-Mode Dielectric Resonators for Axion Haloscopes,” which introduces tunable dielectric disks and Bragg-like hollow dielectric cylinders inside cylindrical cavities (McAllister et al., 2017). The paper reports a $40\text{--}400~\mu\mathrm{eV}$34 MHz tuning range from a starting frequency of $40\text{--}400~\mu\mathrm{eV}$35 GHz for the disk super-mode design, and a tuning span of about $40\text{--}400~\mu\mathrm{eV}$36 GHz from $40\text{--}400~\mu\mathrm{eV}$37 GHz for the split-ring super-mode design (McAllister et al., 2017). For a sapphire ring configuration, measured quality factors include $40\text{--}400~\mu\mathrm{eV}$38 for TM$40\text{--}400~\mu\mathrm{eV}$39 at $40\text{--}400~\mu\mathrm{eV}$40 GHz (McAllister et al., 2017). The optimized DBAS TM$40\text{--}400~\mu\mathrm{eV}$41-like resonator reaches $40\text{--}400~\mu\mathrm{eV}$42, compared with $40\text{--}400~\mu\mathrm{eV}$43 for an empty TM$40\text{--}400~\mu\mathrm{eV}$44 cavity, and the paper states that the DBAS resonator outperforms a traditional conducting-rod-tuned TM$40\text{--}400~\mu\mathrm{eV}$45 haloscope by 1–2 orders of magnitude in $40\text{--}400~\mu\mathrm{eV}$46 at around 5 GHz (McAllister et al., 2017). These are dielectric-engineered haloscope resonators, not dielectric haloscopes in the MADMAX sense.

The conceptual boundary is therefore operational rather than semantic. A dielectric haloscope in the narrow sense uses dielectric interfaces as the primary conversion and boosting resource; a dielectric-loaded resonator uses dielectric material to improve a cavity mode. Both are part of the broader axion-haloscope landscape, and the later literature increasingly treats them as complementary responses to the same high-frequency problem: maintaining sensitivity when the target mass pushes the experiment into the GHz, W-band, or optical regimes (Vora et al., 2022, McAllister et al., 2017).