1. Introduction

The vertical-cavity surface-emitting laser (VCSEL) array has emerged as a pivotal component in applications such as face ID, 3D sensing, lidar and free-space optical communication [1–7] due to its advantages of small size, circular output beam and easy integration. However, challenges such as thermal crosstalk [8] and current crowding effects [9] hinder its performance and broader applications. The isotropic diffusion of heat leads to an inhomogeneous temperature distribution, known as the thermal crosstalk effect. It will lead to an increase in spectral width of VCSEL. Meanwhile, localized overheating in VCSEL array will decrease its life-time. Additionally, the electrode's inner diameter exceeds the aperture diameter in traditional VCSEL, resulting in higher current density at the aperture's edge, known as the current crowding effect. According to the linear polarization mode theory [10], the fundamental mode and the higher-order modes are located at center and edge of waveguide, respectively. The current crowding effect can lead to an increased gain of higher-order modes. Higher-order modes exhibit a shorter wavelength compared to fundamental mode [9]. Therefore, the appearance of higher-order modes leads to expansion of spectrum, blue-shift of peak position and larger divergence angle in VCSEL array. These thermal accumulative effects significantly restrict the utilization of VCSEL array in the next-generation wireless communication systems [5–7], silicon photonic integrated circuits [11,12] and time-of-flight (ToF) radar [3,4]. A way to address the issue of higher-order modes is to suppress the current crowding effect, but methods like double oxide apertures [13] have complex manufacturing processes, distributed-ring contacts [14] are challenging to apply to small-aperture VCSEL. Another strategy involves increasing the loss of higher-order modes, such as Zn-diffusion [15,16] and surface relief [17,18]. However, the energy which is wasted in these methods will be converted into self-heating. It will worsen the problem of thermal crosstalk and result in reduced thermal stability.

Several high-beam-quality VCSEL arrays have been reported. Ning et al. reported a 4 × 4 VCSEL array integrated with micro-lens array [19], exhibits an output power of 1 W, a divergence angle of 6.6 °, and a spectral full width at half maximum (FWHM) of 0.7 nm. Shi et al. developed a 26 × 26 Zn-diffusion VCSEL array [16], which achieves a quasi-single mode emission of 200 mW under continuous-wave (CW) operation, with series resistance of 4 Ω and a divergence angle of 14 °. Although these methods can enhance the beam quality, they are unable to mitigate the thermal crosstalk effect within the array. The bottom emitting structure reduces self-heating by avoiding p-type distributed Bragg reflector (p-DBR), and heat source is closer to heat sink [20]. However, this method is unsuitable for VCSEL with wavelengths less than 850 nm due to the high absorption of GaAs substrates. Substrate removal process [21] is complex, and the active cooling module [22] hampers VCSEL integration. The above research can illustrate that reducing series resistance and improving current distribution have become key factors in reducing thermal crosstalk and suppressing higher-order modes for high-beam-quality VCSEL array. Graphene, as a novel transparent conductive material, has high transparency, ultra-fast mobility, high thermal conductivity and high mechanical strength. Besides graphene as an electrode has found extensive applications in light emitting diode (LED) technology [23–25].

In this paper, a novel design of VCSEL array with graphene electrode (Gr-VCSEL array) was presented, enabling vertical injection of current through the oxide aperture. By comparison to traditional VCSEL array, this novel design effectively reduces transverse current transport within the VCSEL, and minimizes series resistance and self-heating of array. The output power and spectrum results show that it successfully suppresses the thermal crosstalk effect. Additionally, the optimization of current density distribution in the active region helps to suppress higher-order modes. As a result, the proposed Gr-VCSEL array achieves high-power quasi-single mode emission.

2. Devices fabrication and simulation

A 10 × 10 traditional VCSEL array is prepare by lithography and wet oxidation techniques. The epitaxial layer structure of VCSEL from bottom to top includes a n-type GaAs substrate, 36 pairs of Al_0.9Ga_0.1As/Al_0.12Ga_0.88As n-DBR, 3 pairs of In_0.1Ga_0.9As/Al_0.4Ga_0.6As multilayer quantum wells with thickness of 4.2 nm/7 nm, an Al_0.98Ga_0.02As oxidation layer, 24 pairs of Al_0.12Ga_0.88As/Al_0.9Ga_0.1As p-DBR and a GaAs top contact layer. The mesa of VCSEL array is obtained by wet etching, and the height of mesa is 3.1 µm. The mesa profile is trapezoidal after wet etching, and the angle between the side and bottom is approximately 45 °. In wet oxidation process, water vapor is transported by nitrogen into the tube furnace, and partially oxidizes Al_0.98Ga_0.02As to Al₂O₃ at 400 ℃. By controlling the oxidation time, the center is a circular unoxidized region with a diameter of 6 µm. To protect active region and achieve insulation, a 200 nm thick SiO₂ is grown on the surface of array by plasma enhanced chemical vapor deposition (PECVD) as a passivation layer. A 40 µm diameter circular window on mesa surface is opened to enable the electrical contact between top electrode and VCSEL surface. The top metal electrode material, Ti/Au, is deposited by magnetron sputtering with a thickness of 15 nm/150 nm. Units in array are parallel connected through top metal electrode. The surface is then opened a 20 µm diameter window to output laser. Finally, AuGeNi/Au is deposited on the bottom of array as the bottom electrode with a thickness of 50 nm/300 nm. Figure 1(d) shows the array in detail. It consists of 100 units in a square arrangement, the diameter of each unit’s mesa is 70 µm, and the spacing between the center of unit is 100 µm.

 

Fig. 1. (a) Preparation process of Gr-VCSEL array. (b) Scanning electron microscope images of graphene electrode. (c) Raman spectra and transmittance of graphene electrode. (d) Microscope image of Gr-VCSEL array. (e) Current transport path of two structures.

Download Full Size | PDF

Figure 1(a) shows the preparation process of Gr-VCSEL array. After the preparation of traditional VCSEL array, graphene is transferred to the surface of array by wet transfer method. The sheet resistance of graphene electrode is 500 Ω/sq. The photoresist is used as a mask for oxygen plasma etching. After oxygen plasma etching process, graphene electrode only exists on mesa region, which contact with both the metal and the GaAs at the top of VCSEL simultaneously, as shown in Fig. 1(a). The array is annealed at 430 °C to form an ohmic contact between metal and VCSEL.

Figure 1(b) shows the scanning electron microscope images of the graphene electrode. Graphene is covered on the surface of VCSEL mesa. Due to the flexibility of graphene, the graphene remains intact even in areas with height variations and forming an inclined plane. In Fig. 1(c), the Raman spectrum and transmissivity of the graphene electrode are presented. The three peaks observed in Raman spectrum are commonly referred to as D peak, G peak, and 2D peak. D peak gives information about the amount of defects and impurities in graphene, G peak is a signature of graphitic material, and 2D peak is the trademark for graphene. The ratio of 2D peak to G peak (2D/G) decreases with the increase in the amount of graphene layers [26,27]. When 2D/G is less than 1, it can be regarded as multilayer graphene. Thus, the 2D/G ratio of 2 - 3 layers graphene we utilized is 0.68 and the thickness is about 1 nm [28]. The transmissivity of the graphene electrode exceeds 94% within the 750 nm - 950 nm range. The Gr-VCSEL array achieved higher optical power despite a 6% increase in loss, attributed to the graphene electrode's ability to effectively reduce self-heating. The current transport paths of Gr-VCSEL are depicted on the right side of Fig. 1(e). The current in Gr-VCSEL can be injected vertically into the oxide aperture through the graphene electrode on the surface, which mitigating current crowding at the edge of aperture in traditional VCSEL, as shown in the left side of Fig. 1(e).

To investigate the different of joule heat and current transport path between Gr-VCSEL and traditional VCSEL, the PICS3D software is utilized for the simulation of two structures. The simulation parameters are shown in Table 1, σ_L and σ_V are the transverse electrical conductivity and vertical electrical conductivity, respectively. Figure 2(a) and Fig. 2(b) illustrate the simulated joule heat distribution and current flow in both the Gr-VCSEL and traditional VCSEL with 6 µm aperture. In these figures, colors represent joule heat, and arrows indicate the direction of current transport. In traditional VCSEL with annular electrode, the current is injected outside the aperture, and transported along the shortest path, ultimately crowding at the edge of aperture, as shown by the arrow in Fig. 2(b). Conversely, the current in Gr-VCSEL is vertically injected from above the oxidation aperture through the surface graphene electrode, as depicted by the arrow in Fig. 2(a). This change of transport path mitigating the current crowding effect and increasing the current density at waveguide’s center in Gr-VCSEL. We denote the current density at waveguide’s center as D_c and the current density at waveguide’s edge as D_e. The ratio of D_c to D_e is calculated to be 0.58 in traditional VCSEL and 0.68 in Gr-VCSEL. According to the linear polarization theory, the fundamental mode is located at waveguide’s center and the higher-order mode is located at waveguide’s edge. Consequently, a higher current density at waveguide’s center results in an increased gain for fundamental mode in Gr-VCSEL.

 

Fig. 2. (a) Simulated joule heat distribution and current transport path of Gr-VCSEL at 10 mA. (b) Simulated joule heat distribution and current transport path of traditional VCSEL at 10 mA. (c) Simulated resistance ratio of two structures at different current transverse transport distance. (d) Simulated temperature distribution on the surface of two arrays at 10 mA.

Download Full Size | PDF

Table 1. Material properties for each layer of 850 nm VCSEL

View Table

The graphene electrode also reduces the transverse resistance caused by the transverse transport of current, thereby reducing the series resistance of Gr-VCSEL. The calculated series resistance is 142.58 Ω for traditional VCSEL and 126.82 Ω for Gr-VCSEL. The transverse resistance in traditional VCSEL is determined by the transverse transport distance of the current, which can be expressed as [29]:

3. Results and discussion

Figure 3 presents the light-current-voltage (LIV) characteristics and far-field pattern of both VCSEL arrays. The tests are conducted under CW conditions and without temperature control. The threshold current of both arrays is 0.05 A. The array is composed of multiple devices connected in parallel, resulting in a lower resistance compared to individual device. The series resistance of Gr-VCSEL array and traditional VCSEL array are 3.83 Ω and 6.35 Ω, respectively. By reducing the current transverse transport, the series resistance of Gr-VCSEL array is reduced by 41%. Due to the effect of series resistance, the self-heating of traditional VCSEL array is higher compared with Gr-VCSEL array. The thermal rollover current of traditional VCSEL array is smaller (0.3 A). Despite a 6% power loss, the graphene electrode resulted in 57% enhancement in Gr-VCSEL array's power (from 0.179 W to 0.282 W), attributed to its effective mitigation of self-heating. The improvement in resistance and reduction in thermal crosstalk due to the graphene electrodes result in a higher efficiency for the Gr-VCSEL array. In the condition of without any temperature control or thermal sink, the maximum Power Conversion Efficiency (PCE) of Gr-VCSEL array is 33.33%, which significantly surpassing the 23.93% of the traditional VCSEL array. Affected by the current crowding effect, the higher-order modes are more pronounced in traditional VCSEL array, and the far-field pattern shape of traditional VCSEL array is converted to flat-top shape at 0.25A. However, the Gr-VCSEL array maintains a Gaussian shape at the same current.

 

Fig. 3. The Light-Current-Voltage characteristics and far-field pattern of two arrays.

Download Full Size | PDF

Figure 4(a) and Fig. 4(b) display the spectra of both VCSEL arrays, the tests are under CW conditions and without temperature control. By optimizing the current density distribution in active region, the current crowding effect in Gr-VCSEL array is obviously mitigated. The intensity of LP₀₁ mode of Gr-VCSEL array is measured at -11.03 dBm, when the current is set to 0.25 A, and the intensity of LP₁₁ mode is -44.01 dBm. The side mode suppression ratio (SMSR) of Gr-VCSEL array is 32 dB, which indicating the quasi-single mode emission. The traditional VCSEL array exhibits stronger LP₁₁ mode influenced by the current crowding effect, resulting in a maximum SMSR of only 11 dB, as shown in Fig. 4(c). The high temperature region in traditional VCSEL is located at the edge of oxidation aperture, which leads to a slightly red shift in the wavelength of LP₁₁ mode. The reduced self-heating effectively mitigates the thermal crosstalk, resulting in enhanced spectral characteristics of Gr-VCSEL array. The spectral FWHM of Gr-VCSEL array is approximately 0.22 nm, and the normalized spectral area is increased from 0.62 to 0.74, as depicted in Fig. 4(d). In traditional VCSEL array, the severe thermal crosstalk effect leads to an uneven temperature and wavelength distribution in the array, so the normalized spectral area is increased from 1.09 to 1.99. The lower self-heating of Gr-VCSEL array contributes to a reduced wavelength red-shift rate. As shown in Fig. 4(e), the wavelength red-shift rate of the Gr-VCSEL array is 15.4 nm/A. Whereas the traditional VCSEL array exhibits a higher rate of 22.1 nm/A. This reduction in the wavelength red-shift rate further prominent the superiority of the Gr-VCSEL array in terms of its thermal performance.

 

Fig. 4. (a) Spectrum of Gr-VCSEL array under different operation current. (b) Spectrum of traditional VCSEL array under different operation current. (c) SMSR of two arrays. (d) Normalized spectral area of two arrays. (e) Wavelength red-shift rate of two arrays.

Download Full Size | PDF

In accordance with linear polarization mode theory, the intensity distribution of the LP₀₁ mode takes on a Gaussian shape, while the LP₁₁ mode exhibits a ring shape. When multiple modes exist, these modes will be superimposed to form a far-field pattern, and the intensity of each mode contributes to this pattern depending on its power. As the increase of LP₁₁ mode intensity, the far-filed pattern of VCSEL array will change from Gaussian shape to flat-top shape, and ultimately to ring shape. Overall, the far-field beam profile serves as a critical tool for characterizing and optimizing laser beams, ensuring their suitability for specific applications and enhancing the overall performance of laser systems. The far-field profile and beam combination process of both arrays are illustrated in Fig. 5(a) and Fig. 5(b), and the far-field pattern are shown in Fig. 5(d) and Fig. 5(e). The tests are conducted under CW conditions and without temperature control. Each VCSEL in an array emits coherent light, but since the VCSELs are optically uncoupled, the combined far field pattern consists of a superposition of individual coherent beams. Therefore, the far field pattern of array is influenced by each individual unit that comprises the array. In the case of Gr-VCSEL array, the optimization of current density distribution through the graphene electrode make the array quasi-single mode emission, and the pattern is Gaussian shape, as shown in Fig. 5(a). However, in the traditional VCSEL array, the LP₁₁ mode exhibits higher intensity, resulting in a flat-topped far-field pattern as shown in Fig. 5(b). The fundamental mode has a smaller divergence angle, the divergence angle of the Gr-VCSEL array measuring is approximately 12 °, and the traditional VCSEL array is about 17°, as illustrated in Fig. 5(c). This demonstrates the ability of Gr-VCSEL array to generate a high-quality beam with a narrower divergence angle compared with traditional VCSEL array.

 

Fig. 5. Schematic diagram of incoherent beam combining and far-field profile (a) Gr-VCSEL array. (b) Traditional VCSEL array. (c) The relationship between divergence angle and current. (d) Far-field pattern of Gr-VCSEL array. (e) Far-field pattern of traditional VCSEL array.

Download Full Size | PDF

4. Conclusion

We suggested and demonstrated a vertical current injection VCSEL array based on graphene electrode. By optimizing the current transport paths, graphene electrode could reduce the adverse effects of self-heating and thermal crosstalk on VCSEL array. Compared with traditional VCSEL array, the Gr-VCSEL array exhibits 41% reduction in series resistance, 9.4% increase in maximum PCE, and 30% reduction in wavelength red-shift rate. Moreover, the Gr-VCSEL array structure could suppresses the higher-order modes through optimizing the current density distribution in active region of VCSEL. Without any temperature control or thermal sink, the Gr-VCSEL array achieves quasi-single mode emission. In spectrum test, the SMSR, FWHM and divergence angle of Gr-VCSEL array could achieve 31 dB, 0.22 nm, and 12 °, respectively, which exhibits high spectral quality. This novel structure could improve the beam quality and reduce the thermal crosstalk effect in VCSEL array. Meanwhile, the demonstrated graphene electrode could be flexibly applied to VCSEL with different wavelengths. It provides a new method for the development of high-beam quality VCSEL array, offering new possibilities for various applications.