Single-mode semiconductor lasers reduce the dispersion associated with long-distance transmission in fiber-optic communication systems, thereby increasing the transmission distance and rate^[1]. Moreover, the excellent beam quality facilitates efficient coupling to single-mode fibers or other photonic devices, which further improves the integration of optical systems. As a result, single-mode lasers have been widely exploited in optical communications, measurement, holography, and optical integrated circuits^[2–4].

So far, various schemes for single-mode lasers have been proposed. Distributed feedback (DFB) is a common structure with a period of hundreds of nanometers, which is generally fabricated by electron beam lithography (EBL)^[5,6]. As a dot-by-dot process, EBL offers high flexibility and accuracy but is also time- and cost-consuming^[7]. Nanoimprint lithography (NIL) transfers patterns via stamps, reducing the costs by multiple uses and offering a cost-effective option compared to EBL. However, stamps suffer production difficulties and high fragility^[8,9]. The reconstruction equivalent chirp (REC) technology attains precise control of the wavelength and can be achieved by holographic lithography (HL) combined with standard lithography, but it undergoes a complicated combination process^[10,11]. Besides the need for high-precision lithography, these solutions consist of the re-epitaxial growth process, which increases the complexity of manufacturing as well^[12]. As a result, large-scale and low-cost production of single-mode lasers is still challenging.

High-order surface gratings, introducing distributed reflection defects by etching micron-sized slots on the ridge waveguide of the laser, can be fabricated simply by standard lithography^[13–15]. High-precision lithography and regeneration techniques could be avoided, offering a potential candidate for single-mode lasers. We have reported a four-channel laser array with single-mode operation based on a narrow ridge and slots^[16]. To achieve single-lateral mode lasing, the ridge waveguide width was designed to be 3 µm in conjunction with practically achievable process conditions. The narrow ridge corresponds to a small current-injection active region, resulting in a significant limitation of the output power enhancement. Hence, great challenges have emerged to obtain the high power output of the laser while maintaining single-mode characteristics.

In this work, a photonic crystal (PC) structured 1.3 µm high-power single-mode laser is demonstrated. A set of slots is introduced as a longitudinal photonic crystal (LPC) onto the surface of the ridge waveguide to achieve a single longitudinal mode. An 8 µm broad ridge waveguide is employed to increase the output power, and two sets of asymmetric photonic crystals are designed on either side of the ridge to achieve filtering of high-order lateral modes. To be more distinguishable, the PCs introduced laterally are referred to as transverse photonic crystals (TPCs). The PC laser shows a high power up to 120 mW, a narrow horizontal far-field (HFF) distribution with a full width at half-maximum (FWHM) of only 8.8°, and an O-band single-mode output with a side-mode suppression ratio (SMSR) of 41.74 dB at 700 mA. The fabrication process is fully compatible with standard lithography, resulting in low production costs and a simple process.

Figure 1(a) shows the schematic of the PC laser. Figure 1(b) shows the cross-section of the LPC, which consists of a set of periodic high-order slotted gratings on the surface of the main ridge waveguide, distributed along the resonance direction of the optical cavity. The loss of high-order gratings is non-negligible, so the scattering matrix method (SMM) is used to simulate the slots^[17]. The effect of the slot parameters on the power reflection R and power transmission T is calculated, and further, the power loss L is defined as L=1−R−T. The parameters of the slots include the slot depth (hs), number of slots (Ns), slot width (ds), and slot spacing (dw). Figures 2(a) and 2(b) show the respective relation between hs and Ns in terms of R and L. Both R and L show an overall increasing trend as the hs and Ns increase. Large R implies stable single-mode characteristics, while large L implies significant power loss, since a trade-off between the R and L is required. Based on previous research experience^[16], we choose hs=1.25μm and Ns=10.

The design of the slot satisfies Bragg’s law: 4nsds=(2ms+1)λB and 4nwdw=(2mw+1)λB, where ms and mw are the integer numbers, which are the orders corresponding to the slot width and spacing, respectively; ns and nw are the effective refractive indices corresponding to slots and waveguide regions, respectively; and λB is the Bragg wavelength, which is set to 1330 nm. The total order is defined as m=ms+mw+1. To satisfy the processing conditions of standard lithography, ms=5 is chosen. The larger m leads to a denser reflection peak distribution, along with a smaller free spectral range (FSR), and as a result, the single wavelength becomes more unstable. Therefore, a smaller m is preferred under the premise of satisfying the process feasibility. Figures 3(a) and 3(b) show the respective dependence on λ and m of both R and L. It is noted that there is a reflection maximum near m=30, and considering the loss, m=30 is chosen. The parameters are designed to realize steady single-mode output under controllable manufacturing.

Figure 1(c) shows a schematic cross-section of the TPC. To enhance the output power, the width of the main ridge is set to 8 µm, which theoretically supports three guided lateral modes: TE0, TE1, and TE2. To selectively filter out high-order modes, two asymmetric groups of TPC, which are distributed on both sides of the main ridge waveguide, are designed to match with TE1 and TE2, respectively. The etching depth of the waveguide (Hr) is determined as 1.45 µm. To simplify the calculation, the effective index method (EIM) is employed to convert the cross-section of the device into an equivalent one-dimensional (1D) multilayer waveguide, as shown in Fig. 4.

The transmission matrix method (TMM) is used to calculate the propagation constants (β) of the guided modes within the main ridge waveguide and the photon energy bands of the TPC (the number of TPC waveguides is assumed to be infinite), as shown in Fig. 5. The blue region in Fig. 5(a) represents the photon energy band of TPC1 with the waveguide width (w1) of 1.7 µm, and the pink region in Fig. 5(b) represents the photon energy band of TPC2 with the waveguide width (w2) of 3.4 µm. The yellow, blue, and green dashed lines in the two figures indicate the β of the guided modes TE0, TE1, and TE2 in the main ridge, respectively, while the red dashed line corresponds to the selected TPC waveguide spacing (d1 and d2) of 1 µm. It is observed that the propagation constants of TE2 and TE1 modes are situated within the photon-allowed bands of TPC1 and TPC2, respectively, but the TE0 mode is in the photon forbidden bands of the two sets of TPC. Therefore, the high-order lateral modes will be efficiently coupled into the TPC, while the fundamental lateral modes will not.

Figure 6 shows the simulated light field distributions of different lateral modes, and the optical confinement factor of the modes in quantum wells within the width of the main ridge region is labeled. Figure 6(a) corresponds to a single broad ridge waveguide of 8 µm width, and all three modes are mostly distributed within the ridge region. Figure 6(b) corresponds to an 8 µm broad ridge waveguide with TPC1 and TPC2 on both sides, and the TE1 and TE2 modes are obviously coupled into the TPC waveguides and split into new supermode pairs, which significantly reduces their confinement factors (Γ1 and Γ2) in the main ridge region. TE0 is not coupled and is able to be localized in the main ridge, so its confinement factor (Γ0) remained almost unchanged. Γ0 is more than twice as large as Γ1 and Γ2, proving that the TE0 mode is more competitive for lasing. It should be noted that, in order to reduce the duration of the simulation, the number of TPC waveguides in each group is only set to 5 here, while the actual fabrication is 30. This means that the actual coupling of the high-order modes will be stronger and the confinement factor will be smaller. Since the current is injected only above the surface of the main ridge, the TE0 mode is bound to get a larger gain and easier lasing, while the TE1 and TE2 modes will experience a larger loss and be filtered out; thus, TPC can achieve high power and a single lateral mode of the laser.

In the fabrication, SiO2 is first deposited by plasma-enhanced chemical vapor deposition (PECVD). The pattern is formed on the photoresist by standard lithography and then transferred to SiO2 by reactive ion etching (RIE), and further transferred to the wafer by inductively coupled plasma (ICP). This two-step etching method ensures that the distortion of deeply etched LPC and TPC is avoided during pattern transfer. Next, SiO2 is deposited as an insulating layer and formed into an electrically injected window pattern by RIE. Then, TiAu is sputtered as the p-electrode, and the metal within the slots is removed by wet etching to avoid significant absorption of light. The substrate is thinned to 130 µm for better heat dissipation. Sputtered Au/GeNiAu is the n-electrode and is annealed to complete the alloy. An anti-reflective coating with 1% reflectivity is on the front facet, and a highly reflective coating with 99% reflectivity is on the rear facet for higher output power. Figures 7(a) and 7(b) show the scanning electron microscope (SEM) images of the fabricated LPC and TPC, respectively.

For comparative analysis, a single-ridge Fabry–Perot (FP) laser with a ridge width of 8 µm is also fabricated on the same wafer. Both devices are encapsulated on heat sinks, which are mounted on a thermoelectric cooler to control the measurement temperature to 25°C. The cavity lengths of the lasers are both 1 mm. Figures 8(a) and 8(b) show the light–current–voltage (L–I–V) curves of the FP laser and PC laser, respectively. It is observed that the FP laser has a threshold current of 61 mA and a slope efficiency of about 0.30 W/A, and an output power of 181 mW at 700 mA. For the PC laser, the threshold current is 92 mA, the slope efficiency is 0.27 W/A, and the maximum output power reaches 120 mW at 700 mA. In addition, the slope of the I–L curve gradually decreases with the increasing current, indicating that the power of the PC laser saturates gradually. The lower threshold and higher power of the FP laser are due to the lasing of multiple modes, whereas the PC laser is a single-mode output. Compared to the relatively low output power (typically<50mW) of previously reported single-mode lasers with narrow ridges (typically≤3μm) of slotted lasers^[13–16], the PC laser achieves a significant increase in output power.

In order to further confirm the single lateral mode output characteristics of the PC laser, the HFF distributions of the two lasers are tested at 700 mA. It is clearly seen in Fig. 9(a) that the HFF of the FP laser has a multi-flap distribution with an FWHM as high as 34.6°, and the morphology is consistent with the superposition effect of the three guided modes (TE0, TE1, and TE2), which verifies the existence of lasing high-order modes in the broad ridge under the high injection current. As shown in Fig. 9(b), the HFF of the PC laser presents a single flap, and the morphology is consistent with the Gaussian distribution of the single fundamental mode. In addition, the TE0-mode near-field optical field distribution of the PC laser is simulated, and the HFF distribution is calculated by Fourier transform, as shown in Fig. 9(c), which is almost in accordance with that of the practical measurement. It proves that the output mode of the PC laser is exactly the TE0 mode, so the TPC structure successfully filters out the high-order lateral modes and achieves single-mode output.

The FWHM of the HFF distribution of the PC laser is only 8.8°, which is close to the calculated result of 8.1° and is significantly reduced compared with the FP laser. A smaller divergence angle indicates a better spot quality, which greatly reduces the difficulty of coupling the laser to the fiber and the loss caused by the optical lens. Furthermore, a wider ridge corresponds to a larger near-field dimension and smaller far-field divergence angle, so the PC laser not only increases the single-mode output power but also reduces the divergence angle compared with the previous narrow ridge structure.

To characterize the different longitudinal mode outputs of the two lasers, the spectra are measured. As presented in Fig. 10(a), the spectrum of the FP laser has multi-longitudinal modes with no obvious peak phenomenon. In Fig. 10(b), the spectrum of the PC laser is shown as a single-peak state, indicating that the LPC achieves single longitudinal mode operation due to the introduction of distributed reflection defects. The peak wavelength of the PC laser is 1329.1 nm at 700 mA, which is close to the designed 1330 nm, proving that the theoretical design of the LPC coincides well with the experimental results. With an SMSR of up to 41.74 dB, the LPC achieves good single longitudinal mode performance. In addition, it is expected that the SMSR can be further improved by parameter optimization such as increasing the number of slots.

Table 1 shows the output power and SMSR characteristics of different reported slotted single-mode lasers. Our PC laser demonstrates an ultrahigh output power compared to other works.

In summary, a high-power single-mode laser with a longitudinal and TPC structure is demonstrated. The output power is up to 120 mW, and the FWHM of the HFF distribution is only 8.8°. The lasing wavelength is 1329.1 nm with an SMSR of 41.74 dB. The fabrication process is completely based on standard lithography, which makes it a low-cost, simple-process solution with the potential to be applied for O-band fiber-optic communication systems.