Photon-Recycling Technique

• Photon-recycling technique is an optical design principle that recovers unused photon energy by redirecting it into beneficial signal and energy channels.
• It is applied across semiconductors, cavity systems, interferometers, and detectors to enhance efficiency, voltage, and sensing sensitivity.
• By reusing internally emitted photons and reflected fields, the technique transforms loss mechanisms into performance gains with practical tradeoffs.

Photon-recycling technique denotes a family of optical and optoelectronic strategies that recover a channel ordinarily treated as loss and convert it into useful signal, excitation, or energy transfer. In different subfields, the recycled quantity may be an internally emitted photon reabsorbed in a semiconductor, a cavity reflection returned after a controlled delay, a normally discarded interferometer output coherently reinjected, or detector-reflected light sent back for a second absorption opportunity. This suggests a unifying abstraction: photon recycling is not a single device architecture, but a recurrent design principle in which optical energy that has left the primary interaction path is redirected into a useful one (Wong et al., 1 Jun 2025, Markvart, 2016, Li et al., 2023, Nagano et al., 2017).

1. Terminological scope and physical principle

In semiconductor radiative-balance theory, photon recycling usually means the reabsorption of internally emitted photons. Markvart formulates this through a surface-volume balance, in which the escaping surface flux is the non-reabsorbed fraction of the internally generated volume flux, $\Phi_S=(1-r)\Phi_V$, with $r$ the average recycling probability. In this usage, recycling is tied to absorptivity, absorption coefficient, refractive index, and geometry, and $1-r$ is the escape probability (Markvart, 2016).

In coherent optical control and cavity-waveguide problems, the same phrase denotes the coherent reuse of a field that would otherwise be reflected away. The two-pass cavity-capture protocol routes the first-pass reflection into a delay line and sends it back for a second interaction, so that the returned field interferes destructively with the would-be output and constructively with the intracavity excitation (Wong et al., 1 Jun 2025).

In interferometric metrology, recycling refers to coherent feedback of a normally discarded output mode. In a Mach-Zehnder interferometer, the final output mode $b$ can be phase shifted, attenuated by loop loss, and re-injected into input port $b$, thereby increasing the number of photons circulating inside the interferometer and strengthening the phase dependence of the detected quadrature (Li et al., 2023). In detector engineering, it can mean sending the photodiode’s specular reflection back to the same detector for a second absorption attempt, provided the recycled beam is misaligned enough to avoid excess backscatter (Nagano et al., 2017).

The common misconception is that photon recycling always means luminescence reabsorption in a semiconductor. The literature is broader. In some papers the recycled object is a photon emitted inside matter; in others it is a reflected cavity field, an interferometer output, or even the optical field that remains after electron acceleration in a photonic loop (Li et al., 9 Jan 2025).

2. Radiative reabsorption in semiconductors, photovoltaics, and thin films

The radiative-balance formulation provides the most compact classical statement of photon recycling. Markvart derives $r$ from the equality of Kirchhoff surface emission and Planck volume emission, with the central balance equation $\Phi_S=(1-r)\Phi_V$. For a planar slab, the optical length becomes $\ell_{\mathrm{opt}}=4n^2 d$, so recycling is controlled by $a$, $\alpha$, $r$0, and $r$1; in LED language, $r$2 is the photon escape probability $r$3 (Markvart, 2016).

In bulk $r$4-InP, photon recycling appears as repeated interband absorption, minority-carrier generation, radiative recombination, and re-emission. The reported $r$5 wafers showed a transmitted-to-reflected luminescence ratio rising from about $r$6 at $r$7 to $r$8 at $r$9 at room temperature, and to about $1-r$0 and $1-r$1 for those same doping levels at $1-r$2. At lower doping, around $1-r$3, the ratio rose again to about $1-r$4 while the total luminescence intensity became more than ten times larger than in heavily doped samples, which the authors interpret as stronger recycling enabled by lower free-carrier absorption and higher radiative efficiency (Luryi et al., 2010).

In perovskite photovoltaics, photon recycling is treated as a voltage- and efficiency-enhancing process that raises steady-state carrier density and lowers the effective external radiative saturation current. For $1-r$5, the radiative-limit comparison gives $1-r$6 in both cases, but $1-r$7 increases from $1-r$8 to $1-r$9, $b$0 from $b$1 to $b$2, and PCE from $b$3 to $b$4, corresponding to $b$5, $b$6, and $b$7 absolute. With finite non-radiative recombination, the reported thresholds are $b$8 for a $b$9 benefit and $b$0 for $b$1 and practical power gain; the abstract states benefits when $b$2 and $b$3 (Brenes et al., 2019).

A distinct perovskite result concerns the extraction of internal radiative quality from external photoluminescence. For $b$4 films, a spectral-shape analysis that includes scattering-induced outcoupling gives $b$5 for $b$6, $b$7 for $b$8, and $b$9 for $r$0, with a champion $r$1 yielding $r$2 rather than the $r$3 inferred under smaller escape-probability assumptions. The same work reports that scattering-induced outcoupling makes $r$4 more than $r$5 higher in absolute terms than earlier assumptions (Fassl et al., 2020).

Rigorous thin-film and near-field analyses show that a single recycling factor can be inadequate when optical transport is strongly architecture dependent. In near-field thermophotovoltaics, the nearby emitter opens frustrated-mode leakage channels and can also modify the spontaneous-emission rate, so “using a radiative recombination model with a spatially uniform radiative lifetime, even corrected by a photon recycling factor, is inappropriate” (DeSutter et al., 2016). In planar thin films, the interference radiative transfer–drift-diffusion framework computes radiative generation and recombination self-consistently from $r$6, mode-resolved photon transport, and the local divergence of spectral radiance, thereby treating recycling as a nonlocal redistribution of excitation rather than as a heuristic correction (Kivisaari et al., 2021).

3. Coherent recapture in cavities and guided-wave systems

In cavity capture, the basic limitation of the single-pass protocol is already visible in the one-port Langevin equations,

$r$7

Perfect absorption requires $r$8, equivalently $r$9, but for arbitrary pulses this impedance-matching condition fails during some interval when the capture window and tunable coupling are finite. The two-pass protocol replaces discarded reflection by coherent reuse: the reflected field is delayed and returned as a second input,

$\Phi_S=(1-r)\Phi_V$0

with $\Phi_S=(1-r)\Phi_V$1. For arbitrary positive bounded packets, the control law is

$\Phi_S=(1-r)\Phi_V$2

For the square-pulse worst case with $\Phi_S=(1-r)\Phi_V$3, the exact sufficient condition is $\Phi_S=(1-r)\Phi_V$4 for perfect two-pass capture of any bounded positive packet. For the exponentially decaying packet $\Phi_S=(1-r)\Phi_V$5, perfect capture becomes possible once $\Phi_S=(1-r)\Phi_V$6, where $\Phi_S=(1-r)\Phi_V$7. The same formalism extends by time reversal to photon emission and arbitrary pulse shaping (Wong et al., 1 Jun 2025).

The exactness of “perfect capture” in that model is conditional. The stated assumptions are no intrinsic loss in cavity or delay line, an ideal coherent delay line, perfect phase stability, tunable couplings with sufficient bandwidth and peak strength, the Markov/input-output approximation, and operation in the linear or single-excitation regime. When intrinsic losses $\Phi_S=(1-r)\Phi_V$8 are introduced, the benefit remains but the scaling changes: recycling removes the coupling-limited error and leaves performance limited mainly by intrinsic dissipation, with the paper summarizing the two-pass scaling as $\Phi_S=(1-r)\Phi_V$9 (Wong et al., 1 Jun 2025).

An integrated-photonics analogue appears in the photon-recycling dielectric laser accelerator. There, the post-interaction guided field is not dumped after accelerating electrons; it is routed around a silicon photonic racecourse loop, combined coherently with newly injected light, and sent through the accelerator again. The loop model uses $\ell_{\mathrm{opt}}=4n^2 d$0, with the combiner transmission written as $\ell_{\mathrm{opt}}=4n^2 d$1, $\ell_{\mathrm{opt}}=4n^2 d$2. The reported optimum requires only $\ell_{\mathrm{opt}}=4n^2 d$3 injected power to sustain $\ell_{\mathrm{opt}}=4n^2 d$4 circulating power, reaches steady state in fewer than $\ell_{\mathrm{opt}}=4n^2 d$5 cycles or about $\ell_{\mathrm{opt}}=4n^2 d$6, and at an optimal beam current of $\ell_{\mathrm{opt}}=4n^2 d$7 gives $\ell_{\mathrm{opt}}=4n^2 d$8 with photon efficiency $\ell_{\mathrm{opt}}=4n^2 d$9 (Li et al., 9 Jan 2025).

4. Interferometric metrology, weak-value schemes, and magnetic sensing

In a photon-recycling Mach-Zehnder interferometer, the output mode $a$0 that would normally be ignored is phase shifted by $a$1, subjected to loop loss $a$2, and re-injected into input $a$3. The output mode $a$4 remains a displaced Gaussian state with $a$5, so the improvement does not come from noise squeezing; it comes from a steeper phase dependence of the mean quadrature. For coherent-state input and homodyne detection, the sensitivity is written as $a$6. With $a$7 loss in the recycled arm, the reported optimum is $a$8 near $a$9 and $\alpha$0. The same work shows a lower QCRB than the conventional MZI and emphasizes that the improvement region is horn-shaped in the $\alpha$1 plane rather than global (Li et al., 2023).

In recycled weak-value amplification, the rejected bright-port photons of a nearly dark Sagnac interferometer are actively routed back through the same weak measurement. The shot-noise-limited single-photon experiment used a pulsed $\alpha$2 laser, a Pockels-cell recycling loop, and up to $\alpha$3 passes. The measured signal enhancement was $\alpha$4, and the SNR enhancement was $\alpha$5 relative to single-pass weak-value amplification and also relative to a conventional direct measurement under the same photon-number conditions. The paper adds that with lower-loss optics the SNR improvement could reach around $\alpha$6 for the chosen postselection probability $\alpha$7 (Krafczyk et al., 2021).

Magnetic-field sensing via Faraday rotation provides a third metrological realization. There, $\alpha$8 is encoded in one arm of a Mach-Zehnder interferometer, while a phase $\alpha$9 in the other arm creates the intentional near-dark-port postselection. Two recycling schemes are analyzed. In scheme (I), the bright-port light is sent through an external recycling loop; if $r$00 is the one-pass bright-port probability and $r$01 the loop loss, the total dark-port probability becomes $r$02, while the dark-port polarization signal retains the single-pass amplified form $r$03. The corresponding SNR enhancement over conventional measurement is $r$04. In scheme (II), the bright-port light is returned internally through the interferometer; the enhancement is written as $r$05, with $r$06 determined by the recycled-path amplitudes. The paper reports that the combined method is especially advantageous for smaller $r$07 and smaller $r$08, because signal amplification is strong while polarization-filter loss remains small (Niu et al., 8 Oct 2025).

These metrological uses correct a recurrent misunderstanding. Postselection alone does not improve the ideal shot-noise scaling, because the amplified signal is offset by the reduced count rate. Recycling changes that bookkeeping by increasing the total number of photons that eventually contribute to the amplified output, while leaving the per-photon amplification mechanism intact (Krafczyk et al., 2021).

5. Detector reflection recovery and external quantum efficiency

In photodiode engineering, photon recycling is implemented by giving the surface-reflected beam a second absorption opportunity. The theoretical model for $r$09-fold recycling is

$r$10

which for single-fold recycling becomes $r$11. In the ideal limit $r$12 and negligible scattering, repeated recycling can recover all external reflection loss so that the asymptotic EQE approaches the internal QE (Nagano et al., 2017).

The experiment used a $r$13 InGaAs photodiode at $r$14, with a recycling mirror of reflectivity $r$15, radius of curvature $r$16, and main distance $r$17 from the diode. The EQE increased from $r$18 to $r$19, i.e. by $r$20, over incidence angles $r$21 to $r$22. The technical novelty is backscatter evasion: the recycled beam was intentionally misaligned so that the secondary reflection had poor Gaussian-mode overlap with the original input. With $r$23, the measured backscatter reflectivity remained $r$24, statistically indistinguishable from the no-recycling value $r$25; with $r$26, a minor increase was observed (Nagano et al., 2017).

The practical tradeoff is explicit. To maximize EQE gain, the recycled beam should couple efficiently back to the diode; to suppress backscatter, the secondary reflection should have poor overlap with the incident mode and preferably scatter at larger angles. The paper resolves this by a deliberate but limited misalignment, together with a beam dump for the final exiting reflection (Nagano et al., 2017).

6. Conceptual extensions, boundary conditions, and broader variants

The term also appears in contexts where what is “recycled” is not a propagating photon in the usual radiative-balance sense, but an output state, an internal thermal photon, or the energy of a failed photoelectron. In large-Fock-state preparation, recycling means storing and reusing the leftover output Fock state from a beamsplitter fusion event rather than discarding every non-ideal outcome. If $r$27 and $r$28 are fused and $r$29 photons are detected in one output, the surviving mode is $r$30; with recycling it is stored for future use. The strongest simulated strategies change the rate scaling from exponential to polynomial, with a recycled balanced strategy $r$31, a recycled frugal strategy $r$32, and an improvement of about $r$33 over single-shot preparation at $r$34 photons (Motes et al., 2016).

In photoemission, the proposed Auger-assisted tertiary photoemission mechanism recycles the energy of failed primary or secondary photoelectrons that lie below $r$35. Minority electrons stored in a $r$36-band and minority holes stored in an $r$37-band recombine through a novel Auger process that promotes another electron from an $r$38-band into an above-vacuum $r$39-band, satisfying $r$40. The paper predicts linewidth narrowing, population inversion, and photoemission even for $r$41 when the band placement is favorable (Matzelle et al., 2024).

A still broader thermodynamic usage appears in thermal radiation sourced within matter. There, photon recycling means the internal reflection and later annihilation of thermally created photons, which returns energy as heat. The supplement derives

$r$42

so high internal hemispherical reflectance $r$43 amplifies the internal heating rate relative to the externally supplied heat flow (Smith et al., 2022). This usage is conceptually distant from semiconductor luminescence recycling, but it preserves the same structural motif: a photon that does not escape on first encounter continues to participate in the system’s internal energetics.

The boundary conditions for useful photon recycling are consequently platform specific. In cavity capture, exact unit efficiency requires ideal coherence, tunable couplings, negligible intrinsic loss, and phase-stable recycling (Wong et al., 1 Jun 2025). In photovoltaics, the open-circuit-voltage enhancement obeys

$r$44

so escape probability $r$45, parasitic reabsorption $r$46, and internal luminescence efficiency $r$47 are explicit control parameters. In the perovskite thin-film analysis, a parasitic reabsorption probability of only $r$48 reduces the strongly angular-restricted gain from about $r$49 to about $r$50 (Abebe et al., 2018). In detector and interferometer implementations, low loop loss and stable phase are equally central (Nagano et al., 2017, Li et al., 2023).

Taken together, these studies establish photon recycling as a general optical strategy for converting nominal loss channels into resources. What varies across the literature is the ontology of the recycled object, the governing balance equations, and the meaning of “efficiency”: carrier persistence in semiconductors, unit-fidelity capture in cavities, phase sensitivity in interferometers, EQE in detectors, preparation rate in linear optics, or internal-energy amplification in thermal radiation. The technical content of the term is therefore inseparable from the physical system in which it is used.