Directional Lasing Metasurface

• Directional lasing metasurfaces are engineered planar photonic systems that funnel coherent emission into specific momentum channels using integrated phase masks or resonant feedback.
• They employ mechanisms like bound states in the continuum, guided-mode resonances, and non-Hermitian exceptional points to control output angles and enhance quality factors.
• These systems combine static design and dynamic control through programmable phase profiles and pump modulation, enabling tailored beam shapes for integrated optical applications.

Searching arXiv for additional context on directional lasing metasurfaces and related BIC/exceptional-point laser work. Directional lasing metasurfaces are planar photonic systems in which a metasurface is used either as the lasing resonator itself or as an integrated wavefront-shaping layer, so that coherent emission is funneled into specified momentum channels rather than radiating isotropically. Across the recent literature, this category includes monolithically integrated metasurfaces on VCSELs for beam steering and shaping, dielectric and plasmonic lasing metasurfaces exploiting guided-mode resonances, Rayleigh anomalies, and bound states in the continuum, exciton-polariton metasurfaces driven to exceptional points, and pump-programmed or random-lasing structures whose emission angle is selected by lattice symmetry, band topology, or diffraction conditions (Xie et al., 2019, Azzam et al., 2020, Masharin et al., 2022).

1. Architectural classes

Two broad device classes recur. In one class, the metasurface is not the laser cavity but a monolithically integrated output coupler. Xie et al. placed a dielectric metasurface on the planar backside of a standard VCSEL wafer in a non-intrusive “back-emitting” configuration, with no change made to the DBR or gain region. In that architecture, the metasurface acts as a passive phase mask that converts the native VCSEL output into self-collimated, Bessel, vortex, or steered beams while preserving the basic lasing characteristics of the source (Xie et al., 2019).

In the second class, the metasurface itself supplies the decisive feedback, dispersion engineering, or out-coupling. Examples include a MAPbBr$_3$ perovskite grating supporting exciton-polariton branches and exceptional points, a WS$_2$ monolayer on a dual-resonance Si$_3$N$_4$ nanohole metasurface, TiO$_2$/SU8/Rhodamine 6G systems in which guided-mode resonances, surface-lattice resonances, and quasi-bound states in the continuum are co-engineered, and dielectric nanoresonator arrays in which electric and magnetic dipoles collectively couple to the lattice Rayleigh anomaly (Masharin et al., 2022, Barth et al., 2023, Bashiri et al., 15 Jan 2026, Azzam et al., 2020).

A further distinction concerns how directionality is encoded. Some devices emit near normal incidence with very small divergence because the lasing mode is pinned near $\Gamma$ or near a quasi-BIC. Others lase at a deliberately oblique angle set by an off-$\Gamma$ valley, an off-$\Gamma$ Friedrich–Wintgen BIC, or a grating out-coupling condition. Still others provide multiple discrete output angles or momentum channels from a single patterned platform (Nguyen et al., 2018, Mermet-Lyaudoz et al., 2022, Fortman et al., 15 Dec 2025).

2. Wavevector selection and non-Hermitian mechanisms

In passive metasurface-integrated lasers, directionality is imposed directly by a spatially varying phase profile. In the Fraunhofer approximation, a metasurface that adds a local phase $\phi(x,y)$ imparts a transverse momentum shift

$$k_x=\frac{\partial \phi}{\partial x},\qquad k_y=\frac{\partial \phi}{\partial y}.$$

For steering to an angle $_2$0, the prescribed phase can be written as

$_2$1

whereas collimation requires compensation of the spherical wavefront emitted through the substrate (Xie et al., 2019).

In lasing metasurfaces, momentum selection is more commonly tied to resonant band structure. Guided-mode resonances satisfy a grating-coupling condition of the form

$_2$2

with $_2$3 the reciprocal-lattice vector and $_2$4 the in-plane modal propagation constant. Near a Rayleigh anomaly, surface-lattice resonance, or symmetry-protected BIC, the radiative linewidth can collapse and the quality factor can become very large. In the TiO$_2$5 visible multimode platform, the quasi-BIC quality factor is controlled by the symmetry-breaking notch, with $_2$6 and

$_2$7

while near a symmetry-protected BIC one has $_2$8 as $_2$9 (Bashiri et al., 15 Jan 2026).

The perovskite polariton metasurface adds a distinctly non-Hermitian mechanism. Its lower and upper polariton branches are modeled by

$_3$0

with exceptional points occurring when the eigenvalues coalesce, namely when

$_3$1

In that system, the exceptional points appear in the spectral vicinity of a symmetry-protected BIC, and the BIC-enhanced local density of states scales as $_3$2, which is beneficial for polariton condensation and produces one-axis collimation when the EPs form a continuous stripe in $_3$3-space at $_3$4 (Masharin et al., 2022).

Oblique directional emission is frequently realized by engineering the dispersion away from $_3$5. Nguyen et al. obtained a W-shaped lower branch with lasing at a valley extremum located at $_3$6, so that momentum conservation yields

$_3$7

and therefore $_3$8 for $_3$9 nm (Nguyen et al., 2018). Mermet-Lyaudoz et al. used Friedrich–Wintgen interference between leaky resonances, with a BIC condition requiring both frequency degeneracy and destructive interference of radiative losses,

$_4$0

to realize a lasing polarization singularity at an on-demand tilted angle of about $_4$1 (Mermet-Lyaudoz et al., 2022).

3. Materials, geometry, and fabrication

The materials span conventional semiconductor laser wafers, solution-processed perovskites, 2D semiconductors, polymers doped with organic dyes, and plasmonic or all-dielectric resonator arrays. The perovskite exciton-polariton platform uses MAPbBr$_4$2 on fused silica, with refractive index $_4$3–2.35, exciton resonance $_4$4 eV, homogeneous linewidth $_4$5–7 meV, light–matter coupling $_4$6 meV, and exciton binding energy $_4$7–60 meV, which ensures room-temperature stability. The lasing sample is a nanoimprinted grating with period $_4$8 nm, ridge width $_4$9 nm, ridge height $_2$0 nm, and total film thickness $_2$1 nm (Masharin et al., 2022).

The VCSEL-integrated platform instead uses GaAs nanopillars on the backside of a standard VCSEL wafer. The metasurface consists of circular pillars in a square lattice of period $_2$2 nm, with nominal height $_2$3 nm and diameters spanning roughly $_2$4 to $_2$5 nm to cover a full $_2$6–$_2$7 phase range with $_2$8 transmission per pillar. Alignment registers each $_2$9 metasurface disk to a $\Gamma$0 oxide aperture with $\Gamma$1 lateral error (Xie et al., 2019).

In the 2D-material case, the dual-resonance metasurface is a rectangular nanohole array in a $\Gamma$2 nm Si$\Gamma$3N$\Gamma$4 film on Borofloat-33 glass, with periods $\Gamma$5 nm and $\Gamma$6 nm, hole diameter $\Gamma$7 nm, and a $\Gamma$8 nm PMMA superstrate that increases overlap with the WS$\Gamma$9 monolayer. A TE guided-mode resonance with $\Gamma$0 enhances pumping at $\Gamma$1 nm, while TM-polarized resonances near $\Gamma$2 nm include a leaky TM-GMR and a symmetry-protected TM-BIC with $\Gamma$3 (Barth et al., 2023).

Visible multimode lasing has also been demonstrated in symmetry-broken TiO$\Gamma$4 metasurfaces integrated with an SU8 slab waveguide containing Rhodamine 6G. There the resonators are L-shaped TiO$\Gamma$5 blocks of height $\Gamma$6 nm and lateral footprint $\Gamma$7 nm with a $\Gamma$8 nm $\Gamma$9 $\Gamma$0 nm notch, on a square lattice with period tunable from $\Gamma$1 to $\Gamma$2 nm. The notch converts symmetry-protected BICs into leaky quasi-BICs while the SU8 thickness $\Gamma$3 nm sets the slab-mode dispersion (Bashiri et al., 15 Jan 2026).

These fabrication routes are not equivalent. Some are explicitly solution-processable and large-area, as in the nanoimprinted perovskite metasurface; some are fully compatible with wafer-level semiconductor processing, as in the VCSEL platform; and some depend on transferable or wafer-scale gain media such as a CVD-grown WS$\Gamma$4 monolayer (Masharin et al., 2022, Xie et al., 2019, Barth et al., 2023).

4. Representative demonstrations and reported metrics

Experimental realizations cover visible, telecom, polaritonic, and random-lasing regimes. The reported figures differ because the underlying directionality mechanisms differ.

System	Directional signature	Reported metrics
VCSEL with backside GaAs metasurface (Xie et al., 2019)	Divergence reduced from $\Gamma$5 to $\Gamma$6; steering in a $\Gamma$7 array for $\Gamma$8	$\Gamma$9–$\phi(x,y)$0 mA; $\phi(x,y)$1; deflection efficiency $\phi(x,y)$2–$\phi(x,y)$3
MAPbBr$\phi(x,y)$4 polariton metasurface (Masharin et al., 2022)	$\phi(x,y)$5, $\phi(x,y)$6; Fourier-space stripe at $\phi(x,y)$7	$\phi(x,y)$8; $\phi(x,y)$9 nm; $$k_x=\frac{\partial \phi}{\partial x},\qquad k_y=\frac{\partial \phi}{\partial y}.$$0 meV; $$k_x=\frac{\partial \phi}{\partial x},\qquad k_y=\frac{\partial \phi}{\partial y}.$$1; $$k_x=\frac{\partial \phi}{\partial x},\qquad k_y=\frac{\partial \phi}{\partial y}.$$2–$$k_x=\frac{\partial \phi}{\partial x},\qquad k_y=\frac{\partial \phi}{\partial y}.$$3
WS$$k_x=\frac{\partial \phi}{\partial x},\qquad k_y=\frac{\partial \phi}{\partial y}.$$4 dual-resonance metasurface (Barth et al., 2023)	Narrow Gaussian along $$k_x=\frac{\partial \phi}{\partial x},\qquad k_y=\frac{\partial \phi}{\partial y}.$$5 with divergence $$k_x=\frac{\partial \phi}{\partial x},\qquad k_y=\frac{\partial \phi}{\partial y}.$$6 mrad or $$k_x=\frac{\partial \phi}{\partial x},\qquad k_y=\frac{\partial \phi}{\partial y}.$$7 mrad; stripe in $$k_x=\frac{\partial \phi}{\partial x},\qquad k_y=\frac{\partial \phi}{\partial y}.$$8-space	$$k_x=\frac{\partial \phi}{\partial x},\qquad k_y=\frac{\partial \phi}{\partial y}.$$9; output power $_2$00 nW; $_2$01
Dielectric nanoresonator array (Azzam et al., 2020)	Single-mode emission at $_2$02 with $_2$03; dual-mode at $_2$04 and $_2$05	Single-mode threshold $_2$06; dual-mode thresholds $_2$07 and $_2$08
Multivalley telecom metasurface (Nguyen et al., 2018)	Lasing from valley extremum at $_2$09	$_2$10; $_2$11 nm; $_2$12
FW-BIC vortex laser (Mermet-Lyaudoz et al., 2022)	Emission at $_2$13, $_2$14, $_2$15	$_2$16; $_2$17 nm; $_2$18
TiO$_2$19/SU8/Rh6G multimode metasurface (Bashiri et al., 15 Jan 2026)	Normal and oblique outputs; oblique lasing at $_2$20; normal-mode divergence $_2$21	$_2$22–$_2$23 nm span; thresholds $_2$24–$_2$25 nJ per $_2$26 ns pulse $_2$27–$_2$28; up to four modes

The random-lasing case forms a distinct subfield. In a dye-doped nematic liquid crystal infiltrated into a nanostructured silica metasurface, the emission evolves from uniform angular photoluminescence to a strongly directional peak at large angles. The principal emission angle follows

$_2$29

which evaluates to $_2$30 for the reported system, with an experimental lobe width $_2$31 and a random-lasing threshold $_2$32 (Pham et al., 10 Nov 2025).

These demonstrations show that “directional” does not refer to a single beam geometry. Reported outputs include one-axis-collimated stripes in momentum space, nearly normal diffraction-limited spots, high-oblique-angle beams, six-spot and twelve-spot Brillouin-zone emission patterns, and wide-angle random-laser out-coupling (Masharin et al., 2022, Nguyen et al., 2018, Fortman et al., 15 Dec 2025).

5. Programmability and dynamic control

Programmability enters at several levels. In the VCSEL platform, each device in a $_2$33 array carries a different linear phase gradient $_2$34, so that beam angle is lithographically assigned while the laser remains electrically pumped and wafer-level integrated. The same phase-composition framework yields collimated Gaussian, zero-order Bessel, vortex, and steered Gaussian outputs in one fabrication paradigm (Xie et al., 2019).

A more dynamic form of control is achieved by spatially structured pumping. In the hexagonal plasmonic metasurface laser of de Gaay Fortman et al., a digital micromirror device imposes a binary pump mask $_2$35, with local pumping rate

$_2$36

Hexagonal or triangular pump regions select K-point lasing, whereas long thin rectangles select M-point lasing. The corresponding back-focal-plane patterns are six spots at the $_2$37 points with divergence $_2$38 rad or twelve spots along the M directions with divergence $_2$39 rad. K-point lasing exhibits spontaneous symmetry breaking in the relative intensity between degenerate $_2$40 and $_2$41 modes, while M-point lasing allows deterministic control over emission channels via asymmetric pumping (Fortman et al., 15 Dec 2025).

Pump- and material-dependent steering also appears in random lasers. In the liquid-crystal-infiltrated silica metasurface, changing pump energy modifies the effective refractive index $_2$42 of the guided random-laser mode, and pump polarization or an applied voltage alters the liquid-crystal orientation and hence $_2$43. The principal emission angle remains locked near $_2$44 for fixed geometry, but the angular distribution width and peak angle can shift by several degrees in simulation when the Maier–Saupe order parameter is varied from $_2$45 to $_2$46 (Pham et al., 10 Nov 2025).

At the theoretical limit of reconfigurability, subwavelength atomic arrays provide a quantum metasurface analogue. Fernández-Fernández and González-Tudela showed that directional emission can be tuned by the relative dipole orientation between auxiliary atoms and the array, by entangled clusters, or by bilayer geometries, with a one-dimensional directionality metric $_2$47 obtained for $_2$48, in-plane dipoles, and an auxiliary emitter positioned at $_2$49 above a plaquette center (Fernández-Fernández et al., 2021). This suggests that directional lasing metasurfaces need not be static optical elements; the momentum channel itself can be selected by pump geometry, dipole configuration, or collective-state preparation.

6. Trade-offs, validation, and design methodology

The central trade-off is between quality factor, gain overlap, and useful out-coupling. In quasi-BIC devices, smaller asymmetry produces higher $_2$50 but weaker extraction; the TiO$_2$51 multimode metasurface explicitly states that smaller notch depth $_2$52 yields higher $_2$53 but weaker outcoupling, with a practical target $_2$54–$_2$55 (Bashiri et al., 15 Jan 2026). The dielectric nanoresonator platform makes the same point differently: approaching the perfect BIC lowers threshold because $_2$56, but it also slows turn-on and increases sensitivity to absorption and fabrication tolerances (Azzam et al., 2020).

Directional lasing also does not always require modifying the cavity or paying a threshold penalty. The backside metasurface on a VCSEL is outside the cavity, and the reported lasing characteristics show that integrating the metasurface does not appreciably shift threshold current or slope efficiency (Xie et al., 2019). Conversely, when the metasurface itself is the cavity, directionality and threshold are inseparable because the same resonances set both the radiation channel and the modal loss balance. A generic threshold statement appearing across several platforms is that gain must overcome radiative and material losses, for example $_2$57 in the WS$_2$58 metasurface and $_2$59 in dielectric nanoresonator arrays (Barth et al., 2023, Azzam et al., 2020).

A recurrent methodological issue is the distinction between true lasing and “laser-like” phenomena such as amplified spontaneous emission. This is particularly acute for atomically thin gain media. The WS$_2$60 study emphasized output power, directionality, and spatial coherence, measuring interference visibility with a Young’s double slit and extracting a first-order coherence length $_2$61, precisely because limited output power in previous 2D-material devices made it difficult to distinguish true laser operation from other phenomena (Barth et al., 2023). The same caution is relevant in random lasers, where linewidth collapse, S-shaped intensity curves, and strong temporal fluctuations are used to mark the transition from photoluminescence to ASE and random lasing (Pham et al., 10 Nov 2025).

Optimization methods are increasingly symmetry-aware. Parry et al. formulated metasurface design as a multi-objective eigenmode problem with projection operators

$_2$62

and symmetry metric

$_2$63

so that undesired mode mixing can be filtered during surrogate-based optimization. In their C$_2$64-symmetric up-conversion laser design, the optimized lasing mode appeared at $_2$65 nm with $_2$66, while a pump mode near $_2$67 nm had $_2$68, and the resulting emission was nearly normal with beam divergence $_2$69 (Parry et al., 2023).

The application space follows directly from these device-level properties. Reported or proposed uses include on-chip light sources for optical interconnects, lab-on-a-chip spectroscopies, integrated quantum photonics, wide-field beam shaping, directional displays, optical sensing, face recognition, LiDAR, optical communications, and mode-division multiplexing using OAM channels (Masharin et al., 2022, Xie et al., 2019). The accumulated literature indicates that directional lasing metasurfaces are best understood not as a single device type but as a design paradigm: a flat optical platform in which dispersion engineering, symmetry selection, non-Hermitian mode control, and wavefront synthesis are combined so that lasing occurs in a prescribed angular and modal channel.