Dark Photon Portal: Mechanisms & Experiments

• Dark photon portal is a framework where a hidden U(1) gauge boson mediates suppressed interactions with the SM via kinetic mixing, Higgs, or axion channels.
• It impacts dark matter thermal history, relic abundance, and provides experimental targets through direct detection, colliders, and astrophysical probes.
• The mechanism encompasses diverse realizations—with distinct coupling structures and UV completions—that tailor its phenomenological and cosmological implications.

Dark photon portal denotes a class of interactions in which a hidden Abelian gauge boson—usually written $A'_\mu$, $V_\mu$, or $\gamma'$—provides the leading non-gravitational connection between a dark sector and the Standard Model (SM). In the minimal realization, the portal is a kinetic mixing between the dark $U(1)_D$ field strength and the electromagnetic field strength, so that after diagonalization the dark photon couples to the SM electromagnetic current with strength $e\epsilon$; related constructions replace or supplement kinetic mixing by Higgs-portal scalar mixing or by mixed axion–photon–dark photon operators (Jia, 2018, Dutra et al., 2018, Hadjimichef, 2016, Kaneta et al., 2016). Across these realizations, the portal controls thermal contact, relic abundance, late-time decays, direct detection, collider signatures, and in some models the transfer of entropy or asymmetry between visible and dark sectors (Ibe et al., 2018, Ibe et al., 2018).

1. Canonical effective-field-theory structure

In the minimal kinetic-mixing framework, the dark sector contains a new gauge symmetry $U(1)_D$ with gauge boson $A'_\mu$, and dark matter or other hidden states carry the corresponding dark charge. A standard low-energy description introduces the mixing term

$$\mathcal{L} \supset -\frac{1}{4}F_{\mu\nu}F^{\mu\nu} -\frac{1}{4}F'_{\mu\nu}F'^{\mu\nu} -\frac{\varepsilon}{2}F_{\mu\nu}F'^{\mu\nu} +\frac{1}{2}M_{A'}^2 A'_\mu A'^\mu,$$

or, in the electroweak-normalized form used in one MeV-scale model,

$$\mathcal{L} \supset \frac{1}{2}\,\frac{\epsilon}{\cos\theta_W}\,F_{\mu\nu}A'^{\mu\nu}.$$

After diagonalization, the dark photon inherits a suppressed coupling to the SM electromagnetic current,

$$\mathcal{L}_{\rm int}^{\rm SM}= e\,\epsilon\,A'_\mu J_{\rm em}^\mu,$$

with $V_\mu$0 (Jia, 2018, Dutra et al., 2018).

A representative minimal dark-matter implementation contains a Dirac fermion $V_\mu$1 charged under $V_\mu$2 with coupling $V_\mu$3, so that in the physical basis the interaction terms take the form

$V_\mu$4

The portal is therefore factorized: hidden-sector states couple to $V_\mu$5 through $V_\mu$6, while SM charged particles couple through $V_\mu$7. In this sense, the dark photon behaves as a photon copy with suppressed electric couplings (Dutra et al., 2018).

The same logic appears in direct-detection analyses of “photon portal” dark matter, where a hidden $V_\mu$8 gauge boson $V_\mu$9 kinetically mixes with the ordinary photon, and a fermionic dark matter field $\gamma'$0 couples to $\gamma'$1 through a covariant derivative $\gamma'$2. After diagonalization, SM charged particles acquire effective $\gamma'$3 couplings proportional to $\gamma'$4, while dark-sector matter retains its $\gamma'$5 coupling (Ge et al., 2017).

2. Principal realizations of the portal

The term “dark photon portal” is often used for kinetic mixing, but the literature also contains Higgs-mediated and axion-mediated realizations in which the dark photon is still the operative connector between sectors.

Realization	Characteristic interaction	Distinctive feature
Kinetic mixing	$\gamma'$6 or $\gamma'$7	Dark photon inherits suppressed SM electric couplings
Higgs-mediated	$\gamma'$8 with $\gamma'$9	Dark photon mass from singlet-scalar vev; no kinetic mixing
Dark axion portal	$U(1)_D$0, $U(1)_D$1	Mixed ALP–photon–dark-photon phenomenology

In the Higgs-portal realization, the hidden sector contains a real singlet scalar $U(1)_D$2, a Dirac fermion $U(1)_D$3, and a vector boson $U(1)_D$4. The singlet couples to the SM only through

$U(1)_D$5

while the dark photon mass is generated by

$U(1)_D$6

After $U(1)_D$7, one obtains $U(1)_D$8, and after scalar mixing the SM-like Higgs $U(1)_D$9 couples to $e\epsilon$0. This model explicitly does not introduce kinetic mixing; all dark-photon interactions with the SM are scalar-mediated, and processes such as $e\epsilon$1 are correspondingly tiny (Hadjimichef, 2016).

The dark axion portal introduces a different structure. Its defining interactions are

$e\epsilon$2

or, in alternate conventions,

$e\epsilon$3

These couplings are “genuinely new couplings, not just from a product of the vector portal and the axion portal,” because their leading pieces arise directly from anomaly diagrams with heavy fermions charged under PQ symmetry, electromagnetism, and the dark $e\epsilon$4 (Kaneta et al., 2016, Kaneta et al., 2017).

A further variant appears in mirror-world models. There the visible and mirror sectors each contain their own photon and dark photon, and the portal is realized through kinetic and mass mixing between the visible dark photon $e\epsilon$5 and the mirror dark photon $e\epsilon$6. After diagonalization, the physical dark-photon states couple to both ordinary and mirror currents, inducing effective interactions of the form $e\epsilon$7 (Alizzi et al., 2021).

3. Dark matter, thermal history, and cosmological roles

Dark photon portals are used in several distinct cosmological roles. In minimal MeV dark matter, a Dirac fermion $e\epsilon$8 annihilates through an off-shell dark photon into SM fermions, with the non-resonant thermally averaged cross section scaling as

$e\epsilon$9

For $U(1)_D$0 MeV and $U(1)_D$1 MeV, the thermal relic solution is strongly constrained by CMB energy injection, and the viable thermal parameter space is a narrow region near the resonance $U(1)_D$2; the same studies also show that freeze-in becomes viable for $U(1)_D$3 (Dutra et al., 2018).

In the EDGES-motivated millicharged scenario, the portal does not provide baryon cooling directly. Instead, baryon cooling is controlled by photon exchange through a millicharge $U(1)_D$4, while the dark photon portal determines the small relic fraction of millicharged dark matter. The construction assumes $U(1)_D$5 slightly above threshold,

$U(1)_D$6

so that annihilation $U(1)_D$7 is p-wave and resonantly enhanced during freeze-out. The viable ranges quoted are $U(1)_D$8–$U(1)_D$9, millicharged fraction $A'_\mu$0–$A'_\mu$1, millicharge $A'_\mu$2–$A'_\mu$3, and kinetic mixing $A'_\mu$4–$A'_\mu$5 (Jia, 2018).

Composite asymmetric dark matter uses the portal differently. Here dark matter is a dark baryon of a confining $A'_\mu$6, while a sub-GeV dark photon with kinetic mixing transfers the large dark-sector entropy to the SM. The required ordering is

$A'_\mu$7

so that dark mesons annihilate or decay into dark photons and the dark photons decay into SM leptons. In this framework, the portal is needed to avoid overclosure or excessive dark radiation, and the paper concludes that the viable parameter space is “largely tested by direct detection experiments” (Ibe et al., 2018). A UV-complete version ties this to a product GUT and uses a favored range

$A'_\mu$8

so that dark photons can dump entropy into the visible sector while remaining compatible with cosmology (Ibe et al., 2018).

The portal can also contribute to relativistic energy density. In a gauged Higgs-portal model with a very light dark photon, the dark photons can constitute about $A'_\mu$9 or $$\mathcal{L} \supset -\frac{1}{4}F_{\mu\nu}F^{\mu\nu} -\frac{1}{4}F'_{\mu\nu}F'^{\mu\nu} -\frac{\varepsilon}{2}F_{\mu\nu}F'^{\mu\nu} +\frac{1}{2}M_{A'}^2 A'_\mu A'^\mu,$$0 of the effective number of light neutrino species, depending on whether they decouple after or before the QCD transition. Combining the freeze-out condition with collider constraints on the Higgs invisible width requires the dark Higgs mass to be less than a few GeV (Ng et al., 2014).

4. Ultraviolet completions and non-minimal theoretical embeddings

Several papers embed the portal in explicit high-scale constructions. One composite asymmetric-dark-matter completion unifies the visible and dark sectors into

$$\mathcal{L} \supset -\frac{1}{4}F_{\mu\nu}F^{\mu\nu} -\frac{1}{4}F'_{\mu\nu}F'^{\mu\nu} -\frac{\varepsilon}{2}F_{\mu\nu}F'^{\mu\nu} +\frac{1}{2}M_{A'}^2 A'_\mu A'^\mu,$$1

with the dark photon emerging from

$$\mathcal{L} \supset -\frac{1}{4}F_{\mu\nu}F^{\mu\nu} -\frac{1}{4}F'_{\mu\nu}F'^{\mu\nu} -\frac{\varepsilon}{2}F_{\mu\nu}F'^{\mu\nu} +\frac{1}{2}M_{A'}^2 A'_\mu A'^\mu,$$2

Because $$\mathcal{L} \supset -\frac{1}{4}F_{\mu\nu}F^{\mu\nu} -\frac{1}{4}F'_{\mu\nu}F'^{\mu\nu} -\frac{\varepsilon}{2}F_{\mu\nu}F'^{\mu\nu} +\frac{1}{2}M_{A'}^2 A'_\mu A'^\mu,$$3 and $$\mathcal{L} \supset -\frac{1}{4}F_{\mu\nu}F^{\mu\nu} -\frac{1}{4}F'_{\mu\nu}F'^{\mu\nu} -\frac{\varepsilon}{2}F_{\mu\nu}F'^{\mu\nu} +\frac{1}{2}M_{A'}^2 A'_\mu A'^\mu,$$4 are embedded in non-Abelian groups above the breaking scales, renormalizable kinetic mixing is absent in the ultraviolet. The mixing instead arises from the Planck-suppressed operator

$$\mathcal{L} \supset -\frac{1}{4}F_{\mu\nu}F^{\mu\nu} -\frac{1}{4}F'_{\mu\nu}F'^{\mu\nu} -\frac{\varepsilon}{2}F_{\mu\nu}F'^{\mu\nu} +\frac{1}{2}M_{A'}^2 A'_\mu A'^\mu,$$5

which induces

$$\mathcal{L} \supset -\frac{1}{4}F_{\mu\nu}F^{\mu\nu} -\frac{1}{4}F'_{\mu\nu}F'^{\mu\nu} -\frac{\varepsilon}{2}F_{\mu\nu}F'^{\mu\nu} +\frac{1}{2}M_{A'}^2 A'_\mu A'^\mu,$$6

This same construction also generates the $$\mathcal{L} \supset -\frac{1}{4}F_{\mu\nu}F^{\mu\nu} -\frac{1}{4}F'_{\mu\nu}F'^{\mu\nu} -\frac{\varepsilon}{2}F_{\mu\nu}F'^{\mu\nu} +\frac{1}{2}M_{A'}^2 A'_\mu A'^\mu,$$7 portal needed for asymmetry transfer while suppressing unwanted washout operators (Ibe et al., 2018).

Mirror-world models motivate a different ultraviolet picture. Starting from heterotic $$\mathcal{L} \supset -\frac{1}{4}F_{\mu\nu}F^{\mu\nu} -\frac{1}{4}F'_{\mu\nu}F'^{\mu\nu} -\frac{\varepsilon}{2}F_{\mu\nu}F'^{\mu\nu} +\frac{1}{2}M_{A'}^2 A'_\mu A'^\mu,$$8 or an effective $$\mathcal{L} \supset -\frac{1}{4}F_{\mu\nu}F^{\mu\nu} -\frac{1}{4}F'_{\mu\nu}F'^{\mu\nu} -\frac{\varepsilon}{2}F_{\mu\nu}F'^{\mu\nu} +\frac{1}{2}M_{A'}^2 A'_\mu A'^\mu,$$9 GUT, one obtains a visible sector and a mirror sector, each with its own extra $$\mathcal{L} \supset \frac{1}{2}\,\frac{\epsilon}{\cos\theta_W}\,F_{\mu\nu}A'^{\mu\nu}.$$0. The portal is then a mixing of visible and mirror dark photons. The effective low-energy interaction contains a mass-mixing term $$\mathcal{L} \supset \frac{1}{2}\,\frac{\epsilon}{\cos\theta_W}\,F_{\mu\nu}A'^{\mu\nu}.$$1, and the induced visible–mirror interactions scale with products of small mixings. The paper emphasizes that in the symmetric mirror scenario $$\mathcal{L} \supset \frac{1}{2}\,\frac{\epsilon}{\cos\theta_W}\,F_{\mu\nu}A'^{\mu\nu}.$$2, the positronium–mirror-positronium amplitude behaves as $$\mathcal{L} \supset \frac{1}{2}\,\frac{\epsilon}{\cos\theta_W}\,F_{\mu\nu}A'^{\mu\nu}.$$3, and the induced photon–mirror-photon mixing is of order $$\mathcal{L} \supset \frac{1}{2}\,\frac{\epsilon}{\cos\theta_W}\,F_{\mu\nu}A'^{\mu\nu}.$$4, rendering the portal effectively too weak for phenomenology (Alizzi et al., 2021).

Dark-KSVZ constructions generate dark axion portal couplings directly. In that setup, anomaly coefficients yield $$\mathcal{L} \supset \frac{1}{2}\,\frac{\epsilon}{\cos\theta_W}\,F_{\mu\nu}A'^{\mu\nu}.$$5 and $$\mathcal{L} \supset \frac{1}{2}\,\frac{\epsilon}{\cos\theta_W}\,F_{\mu\nu}A'^{\mu\nu}.$$6 terms that survive even when ordinary kinetic mixing is tuned small. One cosmological application uses the “dark Primakoff” process

$$\mathcal{L} \supset \frac{1}{2}\,\frac{\epsilon}{\cos\theta_W}\,F_{\mu\nu}A'^{\mu\nu}.$$7

to freeze in dark photons from a thermal axion bath, producing a two-component dark matter sector. The same framework allows $$\mathcal{L} \supset \frac{1}{2}\,\frac{\epsilon}{\cos\theta_W}\,F_{\mu\nu}A'^{\mu\nu}.$$8 and has been proposed as a way to address the reported 3.5 keV $$\mathcal{L} \supset \frac{1}{2}\,\frac{\epsilon}{\cos\theta_W}\,F_{\mu\nu}A'^{\mu\nu}.$$9-ray excess through a 7 keV dark photon (Kaneta et al., 2017, Kaneta et al., 2016).

Gauge invariance can further enlarge the portal structure. In dark-axion-portal models written in the electroweak-symmetric basis, the mixed hypercharge–dark-photon operator implies after electroweak symmetry breaking that

$$\mathcal{L}_{\rm int}^{\rm SM}= e\,\epsilon\,A'_\mu J_{\rm em}^\mu,$$0

This means that a photon–dark-photon–ALP coupling automatically induces a $$\mathcal{L}_{\rm int}^{\rm SM}= e\,\epsilon\,A'_\mu J_{\rm em}^\mu,$$1-boson–dark-photon–ALP coupling with fixed strength, which becomes directly relevant at $$\mathcal{L}_{\rm int}^{\rm SM}= e\,\epsilon\,A'_\mu J_{\rm em}^\mu,$$2-boson factories (Jodłowski, 2024).

5. Experimental probes and parameter-space coverage

The experimental program is correspondingly diverse. In MeV-scale invisible-dark-photon searches relevant for millicharged dark matter, NA64 and BaBar set $$\mathcal{L}_{\rm int}^{\rm SM}= e\,\epsilon\,A'_\mu J_{\rm em}^\mu,$$3–$$\mathcal{L}_{\rm int}^{\rm SM}= e\,\epsilon\,A'_\mu J_{\rm em}^\mu,$$4 for $$\mathcal{L}_{\rm int}^{\rm SM}= e\,\epsilon\,A'_\mu J_{\rm em}^\mu,$$5–$$\mathcal{L}_{\rm int}^{\rm SM}= e\,\epsilon\,A'_\mu J_{\rm em}^\mu,$$6, and future Belle II and LDMX are explicitly identified as probes of much of the remaining parameter space (Jia, 2018). In the MeV Dirac-fermion portal, direct detection is controlled by dark-matter–electron scattering,

$$\mathcal{L}_{\rm int}^{\rm SM}= e\,\epsilon\,A'_\mu J_{\rm em}^\mu,$$7

with XENON10/100 and projected SuperCDMS sensitivities translated into exclusions in the $$\mathcal{L}_{\rm int}^{\rm SM}= e\,\epsilon\,A'_\mu J_{\rm em}^\mu,$$8–$$\mathcal{L}_{\rm int}^{\rm SM}= e\,\epsilon\,A'_\mu J_{\rm em}^\mu,$$9 plane (Dutra et al., 2018).

Coherent-scattering and reactor data provide another direct route. In a hidden-$V_\mu$00 model with kinetic mixing, the recent COHERENT data rule out previously allowed regions favored by the thermal relic hypothesis, and when mapped onto the DM–electron cross section COHERENT gives the leading direct constraints for DM masses $V_\mu$01 MeV (Ge et al., 2017).

Collider and fixed-target searches increasingly explore nonstandard production modes. One recent proposal studies photon-induced exclusive diffractive processes and identifies “a currently unexplored region around dark photon masses of 1 GeV/$V_\mu$02 and coupling suppression $V_\mu$03 values around $V_\mu$04” where dark photons could be potentially found at the LHC. For $V_\mu$05 GeV and $V_\mu$06, the quoted HL-LHC event yield is $V_\mu$07 in $V_\mu$08 collisions, although the paper stresses that the measurement would be extremely difficult (Cepila et al., 2024).

In Higgs-portal dark-photon models, the dominant portal observables are exotic Higgs decays

$V_\mu$09

and scalar-mediated pair production at future $V_\mu$10 colliders. The partial width

$V_\mu$11

can contribute an exotic Higgs branching ratio, while the paper finds $V_\mu$12 cross sections too small to be easily observed without very high luminosity (Hadjimichef, 2016).

Dark-axion-portal searches broaden the landscape further. At $V_\mu$13 factories, the gauge-invariance relation $V_\mu$14 implies on-shell

$V_\mu$15

production, followed by semi-visible displaced decays of the heavier dark-sector state to $V_\mu$16 invisible or $V_\mu$17 invisible. LEP already gives strong terrestrial limits for masses above $V_\mu$18, while FCC-ee, FASER, and MATHUSLA offer complementary sensitivity to short- and long-lived regimes (Jodłowski, 2024). A related LLP study shows that including vector meson decays and secondary Primakoff-like production on tungsten layers in FASER$V_\mu$192 lets FASER2 cover a significant portion of the $V_\mu$20 m region that is otherwise difficult for typical beam-dump geometries (Jodłowski, 2023).

In a more elaborate dark axion portal setting, LUXE-NPOD can probe kinetic mixing in regions inaccessible to ordinary dark-photon searches. The projected reach includes “novel constraints on DP kinetic mixing parameters smaller than $V_\mu$21” and the statement that “restrictions on $V_\mu$22 kinetic mixing can be extracted for arbitrarily small DP masses” when ALP-assisted processes are included (Ness et al., 12 Dec 2025).

6. Conceptual lessons, recurring misconceptions, and model dependence

A first recurrent misconception is that “dark photon portal” always means kinetic mixing and nothing else. The literature in fact contains kinetic-mixing models, Higgs-mediated models with no $V_\mu$23 term, and axion-mediated models in which the dominant couplings are $V_\mu$24 and $V_\mu$25 (Hadjimichef, 2016, Kaneta et al., 2016). This suggests that the term is best understood functionally—as the mechanism by which a dark photon communicates with the SM—rather than as a single operator identity.

A second misconception is that the portal necessarily mediates every relevant dark-sector process. In the EDGES-inspired millicharged model, the paper states explicitly that the dark photon portal “does not primarily mediate the cooling that explains EDGES”; the cooling arises from photon exchange and the millicharge $V_\mu$26, whereas the dark photon determines the thermal history and the small relic fraction through resonant p-wave annihilation (Jia, 2018).

A third misconception is that the mere existence of a portal implies observable rates. Mirror-world constructions provide a counterexample: the portal exists in principle, but once $V_\mu$27 is imposed, induced visible–mirror interactions become so small that the model is pessimistic about direct detection (Alizzi et al., 2021). A plausible implication is that viability and detectability are often anti-correlated in ultraviolet-complete embeddings.

Finally, thermal relic intuition is strongly model dependent. For MeV dark matter annihilating through a dark photon into $V_\mu$28, CMB bounds push standard s-wave thermal freeze-out into tension and leave only small windows near resonance, while freeze-in or modified cosmology readily reopen parameter space (Dutra et al., 2018). In composite asymmetric dark matter, by contrast, the portal is not primarily a relic-annihilation channel but an entropy-transfer mechanism and, in some UV completions, part of a broader $V_\mu$29-sharing structure (Ibe et al., 2018, Ibe et al., 2018).

Taken together, these results define the dark photon portal as a family of portal mechanisms with a common mediator but highly non-universal dynamics. In some models it is a thermal relic setter, in others an entropy dump, an asymmetry-transfer relay, a dark-radiation source, an LLP trigger, or an axion-assisted connector. The modern literature therefore treats the portal less as a single benchmark and more as a flexible organizing principle for hidden-sector phenomenology (Jia, 2018, Cepila et al., 2024, Jodłowski, 2024).