GaN-based devices are widely found in a diverse range of optoelectronic and electronic applications including lighting, display, optical storage, and power supplies. With growing demand for ultraviolet–visible light laser diodes for current and upcoming applications, such as laser lighting, LiDAR, and visible-light communications (VLC), research on GaN-based lasers has constantly been gaining momentum. Among them, the GaN-based microdisk lasers have drawn significant interest due to their wide wavelength coverage, high optical gain, small modal volume, uncomplicated structure, and simple fabrication process. (1) The low-loss characteristics of an ideal microdisk laser at the cavity boundary, where total internal reflection (TIR) occurs, result in higher optical finesse and lower lasing thresholds.

In the design of microdisk structures, the primary focus lies on their optical confinement ability and the overlap factor between the whispering gallery modes (WGMs) and the gain region. The lateral optical confinement in the cavity is achieved by confining light in the circular high refractive index GaN microdisk achieve TIR at the cavity boundary. (2−4) As for vertical confinement, the most common approach is to expose the upper and lower surfaces of the GaN microdisk to air to confine light via the large refractive index contrast. (5−14) However, to expose the bottom surface of the GaN microdisk to form the undercut structure, the material underneath the GaN layer needs to be removed by selective etching. This implies that GaN microdisks with an undercut structure are typically thick (as thick as the entire epitaxial structure which are often in excess of a few microns) resulting in low overlap factors, (15,16) and suffer from a rough bottom surface resulting from selective wet etching, (5,6,9) which could significantly reduce optical confinement. Furthermore, the mechanical stability of such a suspended structure is compromised, and it faces challenges with heat sinking, leading to performance instabilities.

To address these issues, thin-film GaN microdisks have been developed utilizing wafer bonding, substrate removal, and thinning techniques, (10,17−21) enabling optimization of the microdisk thickness without altering the growth process to maintain the high internal quantum efficiency (IQE) of InGaN/GaN multiquantum wells (MQWs). In this approach, the metallic bonding layer serves as the bottom mirror to provide optical confinement for the microdisks. However, compared to the undercut structure, the lasing threshold and Q factor of these thin-film microdisks with metallic mirrors did not improve significantly due to strong optical absorption of the metal. (18) A previous study explored the utilization of epitaxial 25-pair AlN/AlGaN distributed Bragg reflector (DBR) instead of the undercut structure to achieve optical isolation from the substrate, realizing an optical-pumped ultraviolet GaN microdisk laser with a Q factor of approximately 400. (22) Furthermore, due to the angular dependency of DBR reflection, especially when applied to cavities of high refractive index materials, the reflectance drops sharply as the angle of incidence increases, making it challenging to achieve omnidirectional high reflectivity which is needed for microdisks. While the reflectance and angular range may increase with an increasing number of pairs, thicker DBRs would directly lead to failures in the wafer bonding process. To realize high-quality thin-film GaN microdisk lasers with superior optical isolation from the substrate, an incidental solution presents itself through the incorporation of several pairs of DBR between the high refractive index material cavity and the metallic layer which is needed for the wafer bonding process. This configuration could provide high reflectivity with low absorption losses across a large angular range of the cavity. Here, we report on a GaN microdisk laser with a hybrid omnidirectional reflector (ODR) structure combining an 8-pair SiO₂/TiO₂ DBR with an Al mirror to achieve bottom optical confinement, instead of the undercut structure or the thin-film structure with metallic mirror. This hybrid ODR not only provides a higher reflectivity across a wide range of angles, this nonsuspended structure also opens up possibilities for the design of subsequent electric injection structures. For instance, a transparent conducting oxide (TCO) layer, such as indium tin oxide (ITO), could be inserted between the p-GaN and dielectric layers to realize electroluminescence (EL).

The microdisks are fabricated on blue-light-emitting diode (LED) wafers as depicted in Figure 3a. The wafers are grown by metal–organic chemical vapor deposition (MOCVD) on silicon substrates, with epitaxial structures comprising a 1500 nm-thick AlGaN buffer layer, a 1500 nm-thick n-GaN layer, 150 nm-thick InGaN/GaN MQWs, and a 150 nm-thick p-GaN layer. The device fabrication begins with the hybrid ODR structure. The 8 pairs of SiO₂/TiO₂ (50 nm/80 nm) dielectric DBR (24,25) are deposited on the p-GaN surface of a GaN LED wafer on the Si substrate, followed by annealing at 800 °C for 1 h in an air environment, as indicated in Figure 3b. The Ti/Al/Ti/Au (5 nm/100 nm/25 nm/50 nm) layers are then deposited onto the DBR by e-beam evaporation to form the hybrid ODR as shown in Figure 3c, which also serves as a seeding layer for the electroplating of Cu–Sn for wafer bonding. The sample is then wafer-bonded to a Si substrate, corresponding to Figure 3d. Wet etch process is then carried out to remove the Si substrate from the GaN LED wafer, as depicted in Figure 3e, after protecting the Si carrier substrate by wax. The exposed GaN surface is further thinned down to about 500 nm by BCl₃-based inductively coupled plasma (ICP) etching as presented in Figure 3f; fabrication of microdisks on such thin-films ensures optimal overlap between optical modes and gain regions. The microdisk patterns are defined by the microsphere lithography (see Figure S3): silica spheres suspended in DI water with diameters of about 8 μm are spin-coated onto the sample (see Figure 3g), followed by ICP etching for the pattern transfer (see Figure 3(h)). After removal of the residual microspheres by sonification in DI water, the fabrication of microdisks is completed, as illustrated in Figure 3i. The configuration of a microdisk is shown in Figure 4a, while top and side view SEM images of the fabricated structure are shown in Figure 4b,c, respectively.

To investigate the lasing characteristics of these thin-film GaN microdisks with hybrid ODR, room temperature microphotoluminescence (μ-PL) spectroscopy is conducted using a TUN-TiA-393–408 CNI pulsed laser at a frequency of 393 nm, a repetition frequency of 50 kHz, and a pulse width of 29.25 ns to pump the microdisk via a 10× microscope objective with a numerical aperture of 0.3 in the normal direction. The μ-PL signals are collected by an optical fiber bundle placed 1 cm away from the sample at an inclination angle of approximately 20°. These optical signals are then coupled to a Horiba iHR-550 spectrograph, dispersed by a 2400 g/mm grating, and detected by a Syncerity BI UV–vis deep-cooled charge-coupled device (CCD). Details of the μ-PL measurement setup are presented in Figure S4. The measured μ-PL spectra at increasing excitation power densities from 44.07 to 60.6 W/cm² are illustrated in Figure 5. It is observed that as the excitation power increases, a single spectral peak appears at the center wavelength of 449.96 nm, whose intensity grows rapidly with increasing excitation powers.

Figure 6 plots the integrated intensity and the full width at half-maximum (fwhm) spectral line width of this optical mode as a function of the excitation power density. The PL intensity rises nonlinearly beyond the lasing threshold (P_th) of 46.5 W/cm² (corresponding to a peak threshold power 31.8 kW/cm²) as determined from the distinctive S-shaped input–output curve. For a clearer view of the distinctive features of individual spectra, Figure 7 plots the Y-offset-normalized μ-PL spectra at several excitation powers below and above the threshold. When the excitation power is just below the threshold (∼46.48 W/cm²), the spectral line width is evaluated to be 0.0248 (see Figure S5). This corresponds to a Q factor of 18200, determined from λ / Δλ where λ is the cavity mode wavelength and Δλ the line width of the mode. The broader spectrum which covers the entire spontaneous emission peak is shown in Figure S6, measured with a low-groove density grating of 600 g/mm. As the excitation power increases beyond the lasing threshold, the line width of the mode narrows further, reaching its minimum line width of approximately 0.02 nm, indicating an increase in time coherence of the emitted light in the lasing state. Limited by the optical resolution of the spectrometer of about 0.02 nm, the actual line width might be even narrower. With further increase of the excitation power, the lasing intensity saturates, while the line width increases to ∼0.022 nm, indicating the saturation of gain. When the excitation power exceeds 50 W/cm², a second mode begins to appear at the center wavelength of 446.1 nm with a line width of 0.024 nm, although the initial mode remains dominant. With an observed spacing of 3.7 nm between the first and side spectral peak, vertical mode Fabry–Pérot (FP) lasing can be ruled out which would otherwise have a spectral spacing in the range of micrometers for microdisk thickness of ∼500 nm. Additionally, for resonant cavities with diameters much larger than their thickness, the Q factor of an FP mode is much lower than that of a WGM as the FP-type Q factor converges to values predicted by the Fresnel equation while the WGM-type Q factors increase without bound. (26) The angle-dependent μ-PL spectra (see Figure S7) indicate that the lasing signal is emitted at a small angle to the in-plane of the disk.

For WGMs of the same order (with the same number of radial and vertical nodes), the free spectral range (FSR) is given by the equation, $\mathrm{\Delta }\lambda ～\frac{{\lambda }^2}{2\pi a_{\mathrm{eff}}{n}_{\mathrm{eff}}}$, where λ is the mode wavelength, a_eff is the effective radius of the optical mode, and n_eff is the group effective index of refraction. With an observed spacing between two peaks of ∼3.7 nm and a calculated FSR of ∼2.7 nm (n_eff is estimated to be 2.9 at blue regime), (27) it can be deduced that the two observed modes are of different orders (see Figure S8). Figure 8 shows the mode profiles of three types of low-order WGMs in a thin-film microdisk cavity. A first-order WGM without nodes in both the radial and vertical directions is depicted in Figure 8a, while a higher-order (in the vertical direction) WGM is illustrated in Figure 8b, which has no nodes in the radial direction but with two nodes in the vertical direction (WGMs with one node in the vertical direction are not considered due to low overlap with the MQWs in the thin-film microdisk). The mode in Figure 8c is typically termed a higher-order (in the radial direction) WGM, characterized by one node in the radial direction and no nodes in the vertical direction. From the 3D-FDTD simulation results, the mode centered at 449.96 nm is assigned as a first-order mode with m = 124, which has no nodes in both the radial and vertical directions. On the other hand, the mode centered at 446.1 nm is assigned as a WGM with m = 110, with no nodes in the radial direction but two nodes in the vertical direction. More details of the mode analysis are provided in Supporting Information section 8. Obviously, a_eff of the m¹¹⁰ WGM is smaller than that of the m¹²⁴ WGM, which could also be seen from the simulated planar electric field distribution maps in Figure 8a,b for the respective modes, where the distance of the electric field antines to the centers of the microdisks for the WGM at 446.1 nm is slightly smaller than that of the mode at 449.96 nm. Due to the presence of two nodes in the vertical direction for the mode at 446.1 nm, the overlap factor of this mode is ∼30%, a fraction of that for the 449.96 nm mode.

Multimode lasing is expected for an 8-μm diameter GaN microdisk based on both theory (6,28) and our previous experimental observations. (17−21) However, single-mode lasing is observed under a low excitation power, which remains dominant until it reaches saturation. This is attributed to spatial hole burning (SHB), where the mode with the highest gain depletes all the inverted carriers near the mode’s antinodes within the resonant cavity. (29) As a result, although the gain spectrum covers a wide spectral range (see Figure S3), SHB prevents other WGMs (m = 122,123,125, and 126) from developing, enabling low threshold single-mode (m = 124) lasing. This phenomenon also underscores both the low optical loss and the strong coupling between the optical mode and the gain region of this proposed structure. Emergence of the second spectral peak at shorter wavelengths at higher excitation powers is commonly attributed to spectral blue-shifting of the InGaN/GaN MQW gain spectra. However, the emission peak of the gain does not exhibit a significant blue-shift (see Figure S9) as measured from an as-prepared thin-film sample due to significantly reduced quantum-confined Stark effect (QCSE) effects from the strain-relaxed thin films. Compared to the first order WGM near 450 nm (m = 122,123,125, and 126), the field antinodes of this higher-order WGM with m = 110 is located further away from that of the first mode (m = 124), leading to an increase in gain. For the second radial order WGM which has a smaller effective radius, besides having much lower Q factor (see Figure S8), its strongest field is further away from the sidewall, resulting in lower out-coupling efficiency compared to the other modes, so that it is not observed in the measured spectra. However, as the excitation power increases, the intensity of the second spectral peak did not strengthen as rapidly as that of the first peak did. This is attributed to the lower overlap factor of the third vertical order WGM.

Figure 9 presents the lasing threshold as a function of Q factor for 7 microdisk lasers. The lasing threshold ranges from 44 to 56 W/cm² (corresponding to the peak threshold powers from 30 to 38 kW/cm²), while the Q factors vary from 13240 to 18200. The microdisk exhibiting the highest Q factor is picked for detailed analysis in the discussion above. The lasing modes are distributed around the wavelength of 450.5 nm (see Figure S10), aligning well with the center of the MQW emission gain spectrum. This alignment enabled a significant proportion of spontaneous emission to couple into the lasing mode, which also accounts for the low lasing thresholds. From the same figure, a negative correlation between the lasing threshold and the Q factor of the microdisk is observed (through the red dotted line). The fluctuations among devices could be attributed to the slight differences in diameters of the microdisks due to the variations in the dimensions of the silica microspheres, process nonuniformity and defects at the sidewall of GaN microdisks.

A summary of key reported figures, including Q factors and lasing threshold, of room temperature optically pumped GaN microdisk lasers are tabulated in Table 1. It is evident that our approach provides the most significant simultaneous enhancement in the Q factor and reduction in lasing threshold. A high Q factor suggests both efficient energy preservation and the effective interaction between the microcavity and the optical mode. In a microdisk laser cavity, the 1/Q value, obtained at the emission wavelength, is determined by four primary factors $Q^{?1}=Q_r^{?1}+Q_a^{?1}+Q_s^{?1}+Q_{ex}^{?1}$. The factor 1/Q_r is due to radiation loss from the disk, 1/Q_a is attributed to the optical absorption of the medium, 1/Q_s is caused by scattering losses related to surface roughness and other defects, and 1/Q_ex is ascribed to the energy coupled to the probe. Radiation loss refers to the energy leakage of light during propagation at the interface, and the integration of the hybrid ODR structure in our microdisk plays an important role in minimizing such losses. On the other hand, the smoothness of the top surface of the microdisk laser as observed from the SEM image of Figure 4b, is ensured by the wet etch removal of the Si substrate as compared with thin-film microdisk lasers processed by laser liftoff. (19) Together with the smoothness of the sidewalls, as observed from the SEM image of Figure 4c, scattering losses are minimized. Internal absorption loss is caused by electronic transitions between band edges and defect states. The high dislocation and defect density buffer GaN layer is removed from the thin-film structure to produce a substantial increase in overlap coefficient, which significantly reduces internal absorption loss at the same time. With optimization of all parameters enabled by the thin-film microdisk with ODR combination, the achievement of simultaneous high Q and low threshold lasing is realized.