In recent years, the applications of multi-mode vertical-cavity surface-emitting lasers have developed rapidly. They have been widely used in 3D face recognition, for automatic driving, large bit rate data communication, and other fields [1–7]. Significant enhancements have been achieved in both the power and efficiency of multi-mode vertical-cavity surface-emitting lasers (VCSELs). However, the performance improvement in power and efficiency of single-mode VCSELs has always been progressing slowly. So, single-mode VCSELs, limited by their low power, have always been used only in data centers with short-distance requirements and quantum sensing fields with lower precision demands [8–10]. Especially in the high-speed communication field, power has become crucial, with the adoption of PAM4+ modulation schemes [11]. Therefore, the development of high-power single-mode VCSELs is of utmost importance. Since 2005, the output power of single-mode VCSELs has remained stable in the range of 6 to 8 mW [12,13]. Until 2021, Su et al. achieved a laser output of 10 mW through the improved Zn diffusion process [14], but its electro-optical conversion efficiency was also very low, only around 12%. We also emphasize achieving high efficiency in single-mode VCSELs, which is currently a focus in VCSEL research in the field of communication, reducing the total power budget of data centers [15–19]. Especially in recent years, the swift advancement of AI technology has led to a massive increase in data throughput, making energy consumption a critical concern. It is projected to grow by an order of magnitude by 2030 [20]. As Professor Jim Tatum mentions in his book, reducing power consumption in various parts of transceivers remains very important [21]. In addition, emerging applications also value the potential for energy efficiency of the single-mode VCSEL platform, such as in photonic brain-inspired computing [22] and VCSEL-based deep learning [23,24]. However, existing methods find it challenging to significantly enhance the gain volume and discrimination capability of surface microstructures, making it difficult for single-mode VCSELs to achieve significant breakthroughs in power and electro-optic conversion efficiency. There has been a lack of theoretical analysis and experimental solutions that can significantly improve the power and efficiency of single-mode VCSELs. Meanwhile, a few articles focus on how to improve the efficiency of single-mode VCSELs. Therefore, exploring the realization of high-power, high-efficiency VCSELs holds significant theoretical and practical value; achieving high electro-optical conversion efficiency devices plays a crucial role in the development of the green photonic intelligence era.

Single-mode VCSELs’ limited power and efficiency arise from several key factors. First, as a microcavity laser, VCSELs have a very short longitudinal gain volume, leading to a significant reduction in round-trip gain. Second, to guarantee single-mode output, VCSELs have restricted lateral dimensions, often with an oxidation aperture under 3 μm. This small aperture curtails emission power and raises series resistance, amplifying the device’s self-heating effect. This self-heating effect further limits the upper threshold of its driving current, hindering the realization of high-power single-mode laser output. To attain high-power single-mode output, diverse strategies have been researched. A primary approach is enlarging the light-emitting aperture to boost lateral gain and reduce series resistance. This involves adjusting the threshold gain of higher-order modes while expanding the aperture. Researchers have employed various techniques, such as Zn diffusion [25], high-contrast gratings [26], surface relief [27], triangular holey structures [28], photonic crystals [29], carrier injection design [30,31], and anti-waveguide cavities [12], to modulate the threshold gain difference between the fundamental and higher-order modes. Yet, in single-junction VCSELs, these methods face challenges in effectively controlling this threshold gain difference, leading to a constrained single-mode operational range. Therefore, the limited increase in lateral gain size and insufficient suppression ratio between the fundamental mode and higher-order modes make it challenging to achieve significant power enhancement. Commonly, VCSEL designs utilize top distributed Bragg reflectors (DBRs) with reflectivity often surpassing 98%. However, such a design significantly increases the series resistance, leading to high Joule heating. The doped DBR also introduces free carrier absorption, and the introduction of microstructures results in increased internal losses, making it difficult for single-mode VCSELs to achieve high efficiency. Simultaneously, surface microstructures, including meta-surfaces [32], BIC photonic crystals [33], topological photonics [34,35], and high-contrast gratings [36,37], are employed for VCSEL beam modulation. Yet, the pronounced surface reflectivity in these designs results in feeble coupling between the surface and cavity, limiting effective mode discrimination.

However, it is worth noting that as early as 1984 [38], Iga et al. proposed an innovative method, which involved using a reverse tunneling junction to achieve cascading in the active region, thereby increasing the longitudinal gain volume. This design strategy allows carriers to undergo multiple stimulated emission processes, enabling the device to not only exhibit high differential quantum efficiency but also maintain a low threshold current. The high differential quantum efficiency allows for an increased slope, thereby enhancing power. At the same time, it can maintain a low threshold current with low reflectivity output, which can enhance the mode discrimination capability of the surface structure to the cavity. Most importantly, through the cascaded active region, multi-junction VCSELs can achieve a substantial increase in gain, but the series resistance and internal losses will not multiply, leading to an improvement in electro-optical conversion efficiency. In 2021, we reported that a 940 nm three-junction multi-mode VCSEL achieved a power conversion efficiency (PCE) of 61.3% under continuous operation at room temperature [39]. However, current multi-junction VCSEL research focuses on highlighting the advantage of high slope efficiency to achieve higher power [40–44]. There were few reports on the analysis and experimentation of multi-junction VCSELs in achieving high-power and high-efficiency single-mode VCSELs. Therefore, we analyzed the potential of multi-junction VCSELs to achieve enhanced mode discrimination capability of surface microstructures and high-power, high-efficiency single-mode laser output while maintaining low threshold operation.

This paper demonstrates the advantages of surface microstructure-based multi-junction cascaded vertical-cavity surface-emitting lasers (multi-junction VCSELs) in delivering ultra-high-power and high-efficiency single-mode laser output. We conducted simulations and experimental research on the single-mode multi-junction VCSEL based on surface microstructures. Theoretical analysis revealed that multi-junction VCSELs, while maintaining a low threshold operation, can achieve enhanced mode discrimination capabilities for higher-order modes of surface microstructures, ensuring the potential for high power and high efficiency output. The 6 μm oxide aperture VCSELs with surface relief of different diameters were fabricated. The 940 nm single-mode VCSELs with the output power of 20.2 mW and side-mode suppression ratios greater than 35 dB under continuous-wave operation were demonstrated; the corresponding electro-optical conversion efficiency was 42%, and the divergence angle was 9.8°. Near-field images confirm single fundamental mode laser operation. To the best of the authors’ knowledge, this is the highest single-mode power recorded for a single VCSEL to date, almost twice the currently known record, and achieving a very high electro-optical conversion efficiency. Our approach opens the door for designing high-power, high-efficiency single-mode VCSELs at different wavelengths. Our results will guide further development and applications of high-power, high-efficiency single-mode semiconductor lasers, and will play an important role in promoting the development of green (meaning energy-efficient) photonics.

Figure 1(a) depicts the schematic of the six-junction VCSEL structure. The structure comprises the top and bottom DBRs. The central section is made up of cascaded active areas. Each active region includes multiple quantum well structures, an oxidation layer, and a reverse-biased tunnel junction. The oxidation layer and tunnel junction are located at the node of the standing-wave light field. It can reduce the restriction on light and the absorption of free carriers, respectively. A layer of Si3N4 is situated above the top DBR.

It serves to isolate moisture. This ensures that the VCSEL is not corroded by the external environment. Our surface microstructure design is on the Si3N4 layer. Figures 1(b) and 1(c) display two types of p-DBR designs. We will adjust the thickness of the Si3N4 layer on the DBR surface to simulate changes in its reflectivity. As shown in Fig. 1(d), the reflectance of the top mirror varies with the thickness of the surface Si3N4 under different numbers of top DBR pairs. It can be observed that the reflectance exhibits periodic changes with the optical thickness of the surface Si3N4. The reflectivity is at its minimum when the optical thickness of Si3N4 equals an integer multiple of 0.5λ added to λ/4. Thus, through microstructure design, the central region can have high reflectance, while the peripheral higher-order mode area has low reflectance. As a result, a lower reflectivity causes an increase in the threshold gain. This allows for the suppression of higher-order mode lasing within a certain current range, thereby achieving single-mode operation. As can be seen from Fig. 1(b), the reflectance modulation of the surface microstructure for 19 pairs of p-DBR is very small, whereas the reflectance modulation range for nine pairs of p-DBR is 20%. This will result in a significant difference in threshold gain between the higher-order modes and the fundamental mode. This enhances the mode discrimination capability of the surface microstructure for higher-order modes. Typically, single-junction VCSELs have a very small gain volume. To ensure low-threshold operation, their reflectance usually needs to be greater than 98%. Therefore, the current single-junction VCSELs have limited surface microstructure mode discrimination capability.

To further investigate the impact of surface microstructures on single-junction and multi-junction VCSELs, we simulated the scaling characteristics of threshold gain and efficiency for single-junction and multi-junction structures based on surface microstructures. The threshold gains and electro-optical conversion efficiency for single-junction VCSEL and multi-junction VCSEL are as follows: Γg1=α+1lc1ln1RbRt,(1)gn=g1N+αiΔlcΓla1N−lnζΓla1N,(2)ηPCEN=ηPCE1lnζRbRtlnRbRt+ηd1/ηilnζ·Na(V1+IR)NaV1+IRN|g1=gN.(3)

In the equation, RbRt′=ζRbRt, Γ is the confinement factor, N is the number of cascaded active regions, αi is the internal loss of a single active region, and ηi is the internal quantum efficiency.

The 6 μm oxide aperture VCSELs with surface relief of different diameters were fabricated. The structural schematic is shown in Fig. 1(a), where surface microstructures are etched on the surface of Si3N4. Surface relief structures with different aperture sizes are etched. The top-emitting six-junction VCSEL is grown on an N-type GaAs substrate using metal-organic chemical vapor deposition (MOCVD) technology. Each cascaded active region consists of a multi-quantum well structure made up of three pairs of In0.14Ga0.86As and GaAs0.88P0.12 quantum wells and barriers. The bottom and top DBRs consist of 41 pairs and nine pairs of N-doped Al0.07Ga0.93As/Al0.9Ga0.1As, respectively. The optical thickness of the cavity length for each cascaded active region is 2λ. The oxidation layer is composed of 20 nm Al0.98Ga0.02As. The tunnel junction is made up of 15 nm thick P++ and N++ GaAs, where the doping concentration for P-type is 1020cm−2 and for N-type is 1019cm−2. In 2020, we reported the fabrication process and performance of the tunnel junction in this device [48]. The manufacturing process is as follows. First, P-type contact metal is deposited on the top of the P-side P+ contact layer, forming a ring-shaped electrode structure. Then, dry etching is performed using chlorine-based gas, exposing the high-Al oxidation layer. Then, using a high-temperature and high-humidity oxidation furnace process, an oxidation aperture with a diameter of 6 μm is formed. To isolate the device, we deposit Si3N4 as a passivation layer using PECVD technology. The optical thickness of the deposited Si3N4 surface is 1.25λ. Subsequently, photolithography and wet etching are performed, with an etching optical depth of 0.25λ, resulting in microstructures with Si3N4 optical thicknesses of 1λ at 1, 2, 3, and 4 μm. The P-type electrode consists of Ti/Pt/Au, while the N-type electrode consists of Pd/Ge/Ti/Au. Both N-type and P-type electrodes are annealed at a temperature of 290°C.

Subsequently, through grinding and polishing processes, the substrate thickness is reduced to 80 μm. The final step is to deposit gold on the substrate to form an N-type electrode. Figure 3 shows the SEM image of a six-junction VCSEL with SR=3μm. It can be observed that the surface SR morphology has a gradual step, which is due to the imprecise control of wet etching. To ensure high-order mode suppression characteristics, the steepness of the step can be enhanced through dry etching.

To evaluate the electrical and optical performance of a single-emitting unit six-junction VCSEL with different SR sizes, we conducted tests directly on an unpackaged 6-inch wafer at room temperature using a probe, in conjunction with a DC source for light-current-voltage (L-I-V) testing. The I-V characteristics were measured using Keithley 2425, while the optical output power (L) was tested using a calibrated power meter (Coherent, PM150). Figure 4 shows the L-I-V characteristic curves of the six-junction VCSEL with different SR sizes. The spectral plots in the Fig. 4 insert are in linear coordinates.

From the L-I-V curve of the large-sized SR device, we observed a kink phenomenon as the current increased. The reason is that at this current, higher-order modes begin lasing, resulting in multi-mode lasing, thereby increasing the slope of the curve. Additionally, as the SR size increases, the mode modulation effect of the higher-order modes gradually weakens, causing the current value at which the kink appears to gradually decrease. This also implies that as the SR size increases, the threshold gain difference between the higher-order modes and the fundamental mode gradually narrows. Notably, when SR=1μm, the L-I-V curve did not show a kink but experienced thermal rollover at 5.8 mA. When SR=2μm, due to the increased area of the lateral high-reflectivity region, the threshold gain of the fundamental mode decreases, leading to a reduced threshold. At 5.5 mA, a kink appears in the curve, with the output power reaching 21.5 mW at this point. Figure 4 displays the spectrum (Yokogawa AQ6370D) at a current of 5.9 mA. The spectra shown in Fig. 5 use linear coordinates. It is evident that after the kink manifests, they transition to multi-mode operations.

To gain a deeper understanding of its single-mode characteristics, we conducted spectral tests on devices with SR=1μm and 2 μm under different currents. Spectral measurements are conducted using a free-space coupling system. We utilized the spectrometer (Yokogawa AQ6370D) for the tests, with the outcomes presented in Fig. 5. Figure 5(a) illustrates the spectral characteristics of the six-junction VCSEL with SR=1μm at varying currents. The figure reveals that the device consistently operates in a single mode across all current settings. Even at a current of 6 mA, where thermal rollover occurs, it still maintains its single-mode characteristics, with a maximum single-mode output power of 20.2 mW and an electro-optic conversion efficiency of 36%. Figure 5(b) shows the spectrum of the six-junction VCSEL with SR=2μm under different currents. When the current is 5.6 mA, two modes appear in the spectrum, corresponding to the kink phenomenon on the L-I-V curve. At a current of 5 mA, although it remains single mode, a secondary peak appears in the spectrum. The output power at this time is 20.2 mW, with an electro-optic conversion efficiency of 42%, and a side-mode suppression ratio exceeding 35 dB. To the best of the authors’ knowledge, this is the highest single-mode output power for a single VCSEL to date, and this power value is nearly twice the current known record. This further proves that multi-junction VCSELs based on surface microstructures can improve their slope efficiency, thereby increasing output power. Additionally, as seen from Fig. 5, as the current increases, the central wavelength of the spectrum also increases. This phenomenon is attributed to the increase in thermal power produced by the device as the current rises, resulting in a redshift of the cavity mode.

For a comprehensive insight into the beam properties of single-mode lasers, we undertook optical tests in both near-field and far-field configurations, as depicted in Fig. 6. Near-field and far-field measurement setups are shown in Fig. 6. We utilize the testing method in Ref. [49]. Figure 6(a) shows the near-field measurement setup. The VCSEL wafer is placed on a cooling system and maintained at 25°C. The current is controlled by a source meter (Keithley 2425). Imaging on the CCD (Mightex MCE-B013-U) is achieved through a microscope system with a 50× objective lens. Figure 6(b) shows the far-field measurement setup. The distance from the surface of the VCSEL chip to the CCD (Visiondatum, Mars25MP-43Tgm) camera is 31.6 mm. The near-field image presented in Fig. 6(c) exhibits a quasi-Gaussian beam intensity distribution, further confirming that our high-power, high-efficiency single-mode laser operates in the fundamental mode rather than single higher-order modes. Figure 6(d) shows the far-field image, further elucidating its modal attributes. As inferred from Fig. 6(e), the laser exhibits a far-field divergence angle of 9.8° (at 1/e2). This indicates that we have successfully achieved a single-mode laser with both high power and small divergence angle.

Table 2 summarizes the developmental history of single-mode VCSELs. It is evident that since 2006, the power of single-mode VCSELs has been developing slowly, maintaining around 10 mW, and simultaneously, their electro-optical conversion efficiency has also been relatively low. The experimental results of this paper indicate that the highest single-mode power has nearly doubled, while the efficiency has also greatly improved. This also provides a beneficial reference for single-mode VCSELs in simultaneously achieving enhanced power, efficiency, and surface microstructure discrimination capability. This also represents a significant advancement in VCSEL laser sources.Table 2.

Summary of Single-Mode VCSELs

In this work, we demonstrate the advantages of surface microstructure-based multi-junction cascaded vertical-cavity surface-emitting lasers (multi-junction VCSELs) in delivering high-power and ultra-high-efficiency single-mode laser output. We conducted simulations and experimental research on the single-mode multi-junction VCSEL based on surface microstructures. Theoretical analysis revealed that multi-junction VCSELs, while maintaining a low threshold operation, can achieve enhanced mode discrimination capabilities for higher-order modes of surface microstructures, ensuring the potential for high-power and high-efficiency output. It makes it possible for the efficiency of high-power single-mode VCSELs to exceed 60%. Such high efficiency is very important for solving the energy consumption problem in the future intelligent era and also provides very important guidance for the development of green photonics. With the increase in gain volume, we can not only maintain a lower threshold current but also reduce the reflectivity at the top DBR. This design helps to shorten the lifetime of photons and enhance the modulation bandwidth. Through the series PN structure formed by cascading active regions, we can effectively reduce the capacitance of the device, thereby creating conditions for achieving a wider modulation bandwidth. It is worth noting that compared to single-junction VCSEL, the slope of multi-junction VCSEL is higher. This means that while achieving high power output, it can maintain a lower operating current and a higher operating voltage, which is very advantageous for driver sources requiring high power and high modulation rates. The 940 nm single-mode VCSELs with the output power of 20.2 mW and side-mode suppression ratios greater than 35 dB under continuous-wave operation were demonstrated, the corresponding electro-optical conversion efficiency was 42%, and the divergence angle was 9.8°. Near-field images confirm single fundamental mode laser operation. To the authors’ knowledge, this is the highest single-mode power for single-unit VCSEL so far, and this power value is nearly twice the current known record.

This method can be extended to any research that modulates the output characteristics of VCSEL based on surface microstructures, making it possible to achieve ultra-high efficiency and power special VCSEL light sources. In the future, we will also explore the possibility of achieving single-mode VCSEL with a larger number of junctions. This research will provide valuable references for the further development and application of high-power, high-efficiency single-mode semiconductor lasers, and play an important role in promoting the development of green photonics.

Acknowledgment. The authors thank Everbright Photonics Co., Ltd. for providing the R&D platform, and they also appreciate all members of the VCSEL project team, including engineers and technicians.