Distributed Coupled-Cavity Laser (DCCL)

• Distributed coupled-cavity lasers (DCCL) are laser architectures where multiple optically linked cavities share feedback, phase accumulation, and loss, replacing the traditional single resonator.
• These systems employ nonlinear and linear coupling mechanisms—such as saturable absorbers and cross-gain coupling—to enhance phase locking, pulse formation, and operational coherence.
• Realizations include degenerate-cavity arrays, THz quantum-cascade lasers with graphene reflectors, microcomb lasers, and resonant-beam systems for wireless power transfer with intrinsic safety.

Distributed coupled-cavity laser (DCCL) denotes a class of laser architectures in which lasing dynamics are governed by at least two optically linked cavities or cavity sections rather than by a single monolithic resonator. In the recent literature, the term covers several related realizations: degenerate-cavity laser arrays with intracavity nonlinear feedback, tightly coupled on-chip quantum-cascade-laser and reflector structures, spatially distributed resonant-beam systems for wireless power transfer and simultaneous light information and power transfer, and nested microresonator–amplifier lasers for self-starting pulsed operation (Mahler et al., 2019, Mezzapesa et al., 2020, Liu et al., 29 Jul 2025, Xiong et al., 22 Feb 2026, Cutrona et al., 2021). Across these embodiments, the common principle is that the effective cavity is distributed over multiple optical subsystems, so that phase selection, loss minimization, comb coherence, safety response, or pulse formation emerge from collective cavity dynamics rather than from a single local element.

1. Definition, terminology, and architectural scope

A DCCL is a coupled resonator laser in which the feedback, phase accumulation, and loss are distributed across multiple linked cavity sections. In the wireless-power and resonant-beam literature, this structure is described as a coupled resonator laser with at least two optically linked cavities, and in closely related work it is also termed a spatially distributed cavity (SDC) laser or resonant beam system (RBS) (Liu et al., 29 Jul 2025, Xiong et al., 22 Feb 2026). In array and comb contexts, the same organizing idea appears in systems where many emitters or subcavities are embedded in a common resonator and are coupled by shared intracavity feedback rather than by isolated pairwise couplers (Mahler et al., 2019, Mezzapesa et al., 2020).

The architectural range is unusually broad. One realization uses a degenerate cavity laser (DCL) formed by two flat mirrors, a near-field mask defining many laser sites, and a 4f telescope that images each hole onto itself after one round trip; this creates many spatially distinct emitters within a common cavity (Mahler et al., 2019). Another realization uses a heterogeneous GaAs/AlGaAs metal–metal THz QCL monolithically coupled to a solution-processed graphene saturable-absorber reflector across a 15 μm air gap, producing a tightly coupled two-part photonic structure described as Gires–Tournois interferometer-like (GTI-like) (Mezzapesa et al., 2020). In DCCL-based wireless power transfer (WPT), the cavity is split between a main cavity containing the gain medium and a free-space cavity relevant to over-the-air delivery, implemented with four cat’s-eye retroreflectors (CRRs) (Liu et al., 29 Jul 2025). In the SLIPT literature, the cavity is physically distributed between transmitter and receiver retroreflectors and may include an internal telescope that creates two coupled sub-cavities while preserving self-alignment (Xiong et al., 22 Feb 2026). A further realization nests a nonlinear Kerr micro-resonator inside and linearly couples it to an amplifying laser cavity, creating a two-cavity pulsed-laser system without saturable absorption (Cutrona et al., 2021).

This diversity has sometimes obscured the concept. The literature does not present DCCL as a single canonical hardware platform; rather, it uses the term for a family of systems in which distributed cavity physics is exploited for different ends. A plausible implication is that DCCL is best understood at the level of cavity organization and coupling mechanism, not at the level of any one gain medium, wavelength, or application.

2. Coupling mechanisms and cavity physics

The defining physical distinction within DCCL research is how one cavity section influences another. Some realizations are dominated by nonlinear loss-mediated coupling, others by linear cross-coupling, and others by distributed resonator boundary conditions.

In the degenerate-cavity array reported in "Improved phase locking of laser arrays with nonlinear coupling" (Mahler et al., 2019), nonlinear coupling is produced by placing a Cr:YAG saturable absorber in the far-field plane of the cavity. The far-field intensity depends on the relative phases of the array elements: phase-locked states produce sharp bright peaks, whereas uncorrelated states produce a broad diffuse pattern. Because the saturable absorber transmission increases with intensity, phase-locked states saturate the absorber more strongly and therefore incur lower effective loss. The coupling is therefore nonlinear in the precise sense that the coherent sum of laser fields sets the intensity at the absorber, the absorber transmission depends on that intensity, and the resulting loss feeds back onto the lasers (Mahler et al., 2019).

The THz QCL comb architecture introduces a different but related mechanism. The QCL back facet and graphene reflector form a coupled cavity with a 15 μm air gap, and the reflector is not merely passive: graphene provides field-dependent absorption bleaching, the reflected field re-enters the cavity, and the boundary conditions seen by the QCL modes are modified by both phase delay and intensity-dependent loss (Mezzapesa et al., 2020). The paper explicitly states that, at the relevant geometry, the GTI GDD contribution is negligible compared with the intrinsic QCL gain dispersion; stabilization is instead attributed to fast saturable absorption, spatial separation of gain and loss, reinjected feedback, and a contribution from Fresnel reflection at the graphene surface (Mezzapesa et al., 2020).

In the nested microcomb laser, the coupling is not absorptive but real linear cross-coupling between a Kerr microcavity field $a$ and an amplifying cavity field $b$. The reduced coupled model is written as $b$14 $b$15 with $\kappa$ serving as the linear coupling strength and also contributing to effective loss in the microcavity equation (Cutrona et al., 2021). The authors emphasize that these coupling terms have the same sign and represent cross-gain coupling, unlike conservative waveguide coupling.

In WPT and SLIPT systems, the coupling mechanism is often formulated through self-consistent propagation between cavity segments. In the DCCL-WPT model, the intracavity field is solved by Fox–Li iteration using operators $\mathcal{A}$, $\mathcal{B}$, $\mathcal{C}$, and $\mathcal{D}$ for the main cavity, free-space cavity, and cross-coupling terms, with obstruction introduced through $\mathcal{T}_{\rm inv}$ (Liu et al., 29 Jul 2025). In the SDC/SLIPT work, cavity existence is determined by the round-trip ABCD matrix and the stability condition

$-1<g<1, \qquad g=\frac{A_0+D_0}{2}$

which is the distributed-cavity analogue of the standard stable-resonator criterion (Xiong et al., 22 Feb 2026).

3. Phase locking, mode competition, and coherence selection

One of the principal motivations for DCCL architectures is their ability to select or preserve a global coherent state when multiple competing states are close in loss. This issue appears in both spatial arrays and frequency-comb devices, albeit with different observables.

In the degenerate-cavity array experiment, the relevant competition is among phase-locked spatial states such as in-phase, out-of-phase, and coexistence states (Mahler et al., 2019). For linear Talbot coupling, the output coupler is shifted so that the round-trip distance to the mask equals the Talbot length

$Z_T = \frac{2a^2}{\lambda},$

where $b$0 is the lattice period and $b$1 is the wavelength (Mahler et al., 2019). At the Talbot condition, the in-phase and out-of-phase patterns can be exactly degenerate in loss. This is problematic near threshold because different longitudinal modes can independently select different spatial patterns, yielding coexistence states. With the far-field saturable absorber, by contrast, the coexistence state produces many separate far-field peaks, each of lower intensity, so the absorber saturates less efficiently and imposes higher loss. The nonlinear element also couples longitudinal modes to one another, forcing them toward the same phase pattern and suppressing mixed-state coexistence (Mahler et al., 2019).

The reported quantitative changes are substantial. The array phase locked more than 30 lasers and was temporally Q-switched (Mahler et al., 2019). Eigenvalue modal analysis over the range $b$2 to $b$3 showed that linear coupling gives in-phase and out-of-phase degeneracy at $b$4, whereas nonlinear coupling shifts the degeneracy point to about $b$5 and changes the relative loss by about 0.5% in favor of one state over the studied range (Mahler et al., 2019). Reliable selection of the lowest-loss state was possible with a loss difference of only about 0.2% for nonlinear coupling, compared with about 5% for linear coupling, corresponding to approximately 25 times higher sensitivity to loss differences. The transition between states was about 5 times sharper, interpreted as 5 times faster convergence to the lowest-loss phase-locked state. State statistics over 50 pump realizations and convergence measurements over 100 realizations showed that linear coupling always exhibited coexistence near degeneracy, whereas nonlinear coupling largely suppressed coexistence near threshold (Mahler et al., 2019).

In the THz QCL case, the coherence problem is not spatial-state coexistence but preservation of comb phase locking as bias-dependent dispersion increases. Conventional THz QCL combs rely on four-wave mixing (FWM) in the intersubband gain medium and usually sustain stable comb operation over only about 16–20% of the laser operating range (Mezzapesa et al., 2020). The graphene-coupled design preserves phase coherence between lasing modes even when FWM alone no longer provides sufficient dispersion compensation. The device produced 8 mW CW output power from the front facet, 40 mW peak power in pulsed mode, over 90 equally spaced optical modes, a continuous bandwidth of 0.94 THz from 2.55 to 3.49 THz, a discontinuous total bandwidth of 1.25 THz from 2.30 to 3.55 THz, and stable comb behavior over >55% of the laser operational range (Mezzapesa et al., 2020). The intermode beatnote linewidth was reported as narrow as 780 Hz and as low as 600 Hz, about 5× narrower than in the reference laser at some biases, whereas the broad beatnote regime of the bare device exceeded 100 MHz and could reach 300 MHz (Mezzapesa et al., 2020). Stable all-electrical injection locking was demonstrated near 11 GHz at both 425 mA and 980 mA, with locking range following the square-root dependence expected from Adler-like behavior (Mezzapesa et al., 2020).

These examples show a common DCCL function: distributed cavity interactions reshape the effective loss or phase landscape so that a collective coherent state is favored over fragmented or incoherent alternatives.

4. Temporal dynamics, Q-switching, and pulse formation

DCCL architectures support more than static coherence selection; they also govern temporal pulse formation through either saturable-absorber dynamics or coupled-cavity modulational instability.

In the degenerate-cavity array, the Cr:YAG saturable absorber simultaneously provides nonlinear spatial coupling and induces passive Q-switching (Mahler et al., 2019). Without the absorber, the output pulse lasted about 200 μs and exhibited complicated oscillations; with the absorber, the pulse duration decreased to about 100 ns (Mahler et al., 2019). The absorber thus converts the same intracavity element into both a temporal and spatial organizing mechanism: it blocks lasing until the intracavity intensity rises enough to saturate it, then sharply lowers cavity loss while also favoring high-coherence far-field states.

The THz graphene architecture also uses saturable absorption, but in a fast reflector geometry rather than as a bulk intracavity Q-switch. The graphene reflector behaves as an ultrafast intraband saturable absorber with ~80% transparency modulation, saturation intensity $b$6, and measured parameters

$b$7

with $b$8, $b$9, and $\kappa$0 (Mezzapesa et al., 2020). In this system, the saturable loss is presented not as a Q-switching mechanism but as a means to preserve comb phase coherence under high-bias operation where FWM alone is insufficient.

A conceptually distinct route is developed in the nested microcomb laser, where self-starting pulses arise without saturable absorption (Cutrona et al., 2021). The system is designed so that its trivial zero state $\kappa$1 becomes unstable to selected perturbations. For $\kappa$2-independent perturbations, the instability eigenvalue obeys $b$16 and the onset of instability occurs at a threshold gain $\kappa$3 satisfying the paper’s Eq. (4) (Cutrona et al., 2021). More generally, the modulational-instability boundary for perturbations with spatial modulation frequency $\kappa$4 is controlled by detuning $\kappa$5, walk-off $\kappa$6, gain dispersion $\kappa$7, and dispersion mismatch $\kappa$8 (Cutrona et al., 2021). The governing principle is that self-starting from noise requires the zero state to be unstable to the perturbations that seed the desired pulse pattern, while remaining stable to irrelevant perturbations.

The paper gives explicit startup examples selected by engineered MI spectra: single soliton at $\kappa$9; two solitons at $\mathcal{A}$0; three solitons at $\mathcal{A}$1; three-soliton noise startup at $\mathcal{A}$2; and five-soliton startup with shaped gain at $\mathcal{A}$3 (Cutrona et al., 2021). Gain shaping is introduced through $b$17 with $\mathcal{A}$4 and $\mathcal{A}$5, producing a two-peaked gain profile that moves the MI maximum away from $\mathcal{A}$6 and enables direct multisolition startup (Cutrona et al., 2021).

This body of work refines a common misconception. DCCL pulse formation is not intrinsically tied to saturable absorbers. The literature includes both absorber-mediated temporal organization and purely coupled-cavity, MI-engineered self-starting soliton formation.

5. Wireless power transfer, SLIPT, and intrinsic safety

A major branch of DCCL research treats the distributed cavity not primarily as a coherence-engineering device but as the optical core of safe, self-aligning free-space power links. Here the salient features are field of view, power density in the free-space cavity, and shutdown behavior under intrusion.

In the DCCL-WPT architecture, CRR1 and CRR2 form the main cavity containing the gain medium and pump source, while CRR3 and CRR4 form the free-space cavity; the output beam emerges through a partially transmitting mirror at the receiver and is converted to electricity by a photovoltaic (PV) panel (Liu et al., 29 Jul 2025). This is explicitly contrasted with the earlier distributed single-cavity laser (DSCL), in which the gain medium is more tightly coupled to the free-space path. By decoupling the gain mechanism from the free-space cavity, DCCL aims to expand mobility tolerance / FoV, lower free-space intracavity power density, and preserve intrinsic safety because intrusion raises cavity loss and collapses lasing (Liu et al., 29 Jul 2025).

The paper formulates a unified safety analysis framework based on diffraction propagation, Fox–Li cavity iteration, gain dynamics, intrusion models, and PV charging estimation (Liu et al., 29 Jul 2025). The irradiance on an intruding object is

$\mathcal{A}$7

where $\mathcal{A}$8 and $\mathcal{A}$9 are forward- and backward-pass irradiance, and the system is safe when $\mathcal{B}$0 remains below the relevant $\mathcal{B}$1 (Liu et al., 29 Jul 2025). The propagation operator is written as

$\mathcal{B}$2

with

$\mathcal{B}$3

The gain update uses a nonlinear amplification relation with small-signal gain coefficient

$\mathcal{B}$4

and charging power is obtained from a diode-equation PV model as

$\mathcal{B}$5

under MPPT (Liu et al., 29 Jul 2025).

The reported case studies cover skin safety, eye safety, and small-object intrusion. For skin safety, using $\mathcal{B}$6 nm, CRR radius $\mathcal{B}$7 mm, focal length $\mathcal{B}$8 mm, reflectivities $\mathcal{B}$9, $\mathcal{C}$0, $\mathcal{C}$1, gain medium radius $\mathcal{C}$2 mm, gain medium length $\mathcal{C}$3 mm, saturation intensity $\mathcal{C}$4, cavity length $\mathcal{C}$5 m, and PV responsivity $\mathcal{C}$6 A/W, the simulation produced $\mathcal{C}$7 W at $\mathcal{C}$8 W (Liu et al., 29 Jul 2025). Relative to DSCL, DCCL produced nearly 50% lower irradiance on large intruding objects and remained below the skin-safety guideline of approximately

$\mathcal{C}$9

Under skin-safe conditions at 5 m, the system achieved over 600 mW charging power, with 100 mW over 16° FoV (Liu et al., 29 Jul 2025).

The eye-safety analysis is more distinctive. Rather than model an isolated eye, the paper uses an STL mesh of the whole human head, computes the convex hull

$\mathcal{D}$0

derives tangent-plane normals

$\mathcal{D}$1

and evaluates eye-to-plane distances and angles through $\mathcal{D}$2 and $\mathcal{D}$3 (Liu et al., 29 Jul 2025). The reported safety thresholds are approximately $\mathcal{D}$4 mW/cm$\mathcal{D}$5 for retina and $\mathcal{D}$6 mW/cm$\mathcal{D}$7 for cornea. Ray tracing in Ansys Zemax OpticStudio showed that, under the relevant geometry, the beam is not focused onto the retina, so corneal exposure becomes the primary concern (Liu et al., 29 Jul 2025). The minimum reported angle between tangent-plane normal and corneal normal was $\mathcal{D}$8, corresponding to an incident beam angle of $\mathcal{D}$9 in the eye model setup. With input power around 85.35 W, output exceeded 650 mW while intracavity irradiation on the intruding surface remained below

$\mathcal{T}_{\rm inv}$0

and as intrusion depth approached about 3 mm the cavity was fully disrupted and output fell to zero (Liu et al., 29 Jul 2025). Under these conditions, the paper reports about 150 mW charging power at 5 m, using 650 mW output at 1064 nm, far above the typical discussion-context eye-safe output of around 10 mW (Liu et al., 29 Jul 2025).

The small-object study models hair as a narrow opaque rectangle with $\mathcal{T}_{\rm inv}$1 mm and $\mathcal{T}_{\rm inv}$2 (Liu et al., 29 Jul 2025). As the hair approaches the beam center, the field eventually collapses abruptly; at about $\mathcal{T}_{\rm inv}$3 mm the output drops sharply. This suggests that DCCL-based WPT is highly sensitive to narrow obstructions, which the paper treats as beneficial for hazard mitigation.

6. Stability theory, tolerances, and practical feasibility boundaries

For meter-scale and longer distributed cavities, stability is not merely a secondary design issue; it determines whether a resonant beam can exist at all. The SDC/DCCL/SLIPT analysis provides the most explicit mathematical treatment of this problem (Xiong et al., 22 Feb 2026).

The system consists of transmitter and receiver retroreflectors, an Nd:YVO$\mathcal{T}_{\rm inv}$4 gain medium pumped at $\mathcal{T}_{\rm inv}$5 nm and lasing at $\mathcal{T}_{\rm inv}$6 nm, and optionally a transmitter-side telescope formed by $\mathcal{T}_{\rm inv}$7 and $\mathcal{T}_{\rm inv}$8 with magnification

$\mathcal{T}_{\rm inv}$9

The cavity is distributed across free space over working distance $-1<g<1, \qquad g=\frac{A_0+D_0}{2}$0, while the telescope creates coupled sub-cavities that can compress the gain-medium spot size or enlarge the field of view (Xiong et al., 22 Feb 2026).

Using the standard ABCD method, the complex beam parameter is

$-1<g<1, \qquad g=\frac{A_0+D_0}{2}$1

transforms as

$-1<g<1, \qquad g=\frac{A_0+D_0}{2}$2

and in a resonator must satisfy

$-1<g<1, \qquad g=\frac{A_0+D_0}{2}$3

The physically admissible solution is

$-1<g<1, \qquad g=\frac{A_0+D_0}{2}$4

yielding

$-1<g<1, \qquad g=\frac{A_0+D_0}{2}$5

The corresponding cavity-stability criterion is

$-1<g<1, \qquad g=\frac{A_0+D_0}{2}$6

equivalently

$-1<g<1, \qquad g=\frac{A_0+D_0}{2}$7

This is the paper’s decisive condition for stable resonant-beam formation (Xiong et al., 22 Feb 2026).

The paper further develops explicit engineering measures. For a focusing retroreflector,

$-1<g<1, \qquad g=\frac{A_0+D_0}{2}$8

so $-1<g<1, \qquad g=\frac{A_0+D_0}{2}$9 gives focusing behavior and $Z_T = \frac{2a^2}{\lambda},$0 gives defocusing behavior (Xiong et al., 22 Feb 2026). The stable-region width with respect to an adjustable parameter such as $Z_T = \frac{2a^2}{\lambda},$1 is defined as

$Z_T = \frac{2a^2}{\lambda},$2

and manufacturing tolerances are modeled by

$Z_T = \frac{2a^2}{\lambda},$3

with robustness requirement

$Z_T = \frac{2a^2}{\lambda},$4

for all admissible perturbations (Xiong et al., 22 Feb 2026). A first-order approximation gives

$Z_T = \frac{2a^2}{\lambda},$5

and hence

$Z_T = \frac{2a^2}{\lambda},$6

with

$Z_T = \frac{2a^2}{\lambda},$7

To estimate tolerances more robustly, the authors introduce a binary-search-based Monte Carlo (BMC) method that samples

$Z_T = \frac{2a^2}{\lambda},$8

computes the worst-case $Z_T = \frac{2a^2}{\lambda},$9, and updates a binary search over $b$00 (Xiong et al., 22 Feb 2026). A fast linear approximation is used as a complementary estimate.

The central numerical conclusion is that the stable region contracts sharply as working distance increases. With fixed manufacturing tolerance

$b$01

the achievable transmission distance is less than 2 m (Xiong et al., 22 Feb 2026). For kilometer-level distances, the stable-region width in the basic design is only about 0.001 mm; scaling focal lengths by $b$02 improves the width to the order of 0.01 mm, which the paper states could support kilometer-scale SLIPT in principle (Xiong et al., 22 Feb 2026). The value of $b$03 giving $b$04 decreases with increasing $b$05, and $b$06 is identified as an optimal design point because the beam radius at the gain medium is minimized there (Xiong et al., 22 Feb 2026). Under a chosen parameter set, a telescope with $b$07 mm and $b$08 mm can support a maximum transmission distance of 76.8 m (Xiong et al., 22 Feb 2026). Experimentally, a setup with cat’s-eye retroreflectors, $b$09 mm, 20 W pump power, and telescope embedding initially supported only about 0.5 m reliable resonance; after tuning the stable region during assembly, stable operation exceeded 2 m, reaching about 2.18 W output and 10.9% transmission efficiency at $b$10 m, about 1.47 W at $b$11 m, and becoming difficult beyond 2.8 m because tuning precision was exhausted (Xiong et al., 22 Feb 2026).

These results establish a practical boundary in DCCL design: fixed-geometry distributed cavities are severely range-limited, whereas assembly-tunable cavities can extend operation but only within tolerance windows set by classical resonator stability.

7. Conceptual significance and recurring themes

Across its diverse implementations, DCCL research converges on several recurring themes. First, the cavity is treated as a distributed dynamical system rather than as a single passive enclosure. The relevant control variables are therefore global quantities such as collective phase configuration, round-trip trace $b$12, reinjected feedback phase, or the MI spectrum of the zero state (Mahler et al., 2019, Xiong et al., 22 Feb 2026, Cutrona et al., 2021). Second, DCCL frequently exploits a separation of functions across cavity sections: one section supplies gain, another imposes nonlinear loss, another provides free-space transport, and another defines boundary conditions or spectral filtering (Mezzapesa et al., 2020, Liu et al., 29 Jul 2025). Third, several papers emphasize that the architecture can make the system behave as a single collective system, whether for coherent beam combining, frequency-comb stabilization, or self-aligned power delivery (Mahler et al., 2019, Liu et al., 29 Jul 2025).

The literature also clarifies two important limits. One is that DCCL does not imply a single mechanism of coherence control: saturable absorbers, linear cross-coupling, and classical resonator-stability tuning all appear as central ingredients in different realizations (Mahler et al., 2019, Cutrona et al., 2021, Xiong et al., 22 Feb 2026). The other is that increased functionality does not remove conventional optical constraints. Near-degenerate phase states still require loss discrimination, THz combs still confront gain dispersion, and long-range resonant beams still obey the strict criterion $b$13 and can be defeated by sub-0.01-mm tolerance limits (Mezzapesa et al., 2020, Xiong et al., 22 Feb 2026).

Taken together, these works portray DCCL as a unifying framework for lasers whose operative physics is distributed across multiple cavity elements. In spatial arrays, DCCL-like nonlinear coupling enables selection of a unique minimum-loss phase-locked state from several near-degenerate candidates (Mahler et al., 2019). In THz microphotonics, tightly coupled reflector sections preserve comb coherence across much more of the operational range than conventional FWM-only designs (Mezzapesa et al., 2020). In WPT and SLIPT, distributed cavities combine self-alignment, dynamic interruption under obstruction, and broader field of view with quantified safety and assembly-tolerance boundaries (Liu et al., 29 Jul 2025, Xiong et al., 22 Feb 2026). In pulsed microcomb lasers, nested coupled cavities permit self-starting temporal cavity solitons without saturable absorption through engineered modulational instability (Cutrona et al., 2021). The cumulative record suggests that DCCL is less a narrow device label than a general cavity-engineering paradigm for collective optical dynamics.