Fabrication of Circular Defects in 2-Dimensional Photonic Crystal Lasers with Convex Edge Structure

Abstract

We have developed circular defects in 2-dimensional photonic crystal lasers that allow current injection for interconnected optical communications. However, when cleaving the sample to measure the output light, the output light intensity changes due to the cleaving position. In a previous study, we proposed a new end face structure called a convex edge structure. In this paper, we design the electron beam lithography patterns to fabricate this structure. With this design, it is possible to eliminate the effect of different cleaving positions and ensure that the cleavage tolerance is larger than the cleavage position error. We also develop the fabrication technology for this structure, fabricate samples, and measure the output light experimentally. The optical properties of the fabricated sample are similar to well-fabricated samples with normal cleavage edge faces. We are assured that these results contribute to future work such as accurate manufacturing and improving the end face configuration to obtain higher outputs.

1. Introduction

In recent years, with the rapid development of generative artificial intelligence led by ChatGPT, as well as electric vehicles, the constant demand for chips with higher communication capacity has significantly triggered a semiconductor shortage [1,2]. Meanwhile, there has also been an explosion in the volume of global data. The International Data Corporation predicts that the global datasphere will grow from 33 ZB in 2018 to 175 ZB by 2025 [3]. Due to bandwidth, heat, and latency limitations, traditional electrical communication systems face significant challenges with expanding data volumes [4,5,6]. Using optical signals instead of electrical signals to transmit information, ultra-short-range optical communications have gained attention due to their high bandwidth capacity, low latency, and high security [7]. Some designs or concepts have been proposed regarding intra-chip optical communication [8,9,10,11]. It has been calculated that a communication capacity density of 10 Pbps cm⁻² is required to achieve intra-chip optical communications [12]. The wavelength division multiplexing (WDM) is mandatory to attain such high communication capacity densities. There are high expectations for photonic crystals (PhCs) to achieve intra-chip optical communications due to their strong light confinement and low propagation loss. A PhC is a nanostructure consisting of a periodic arrangement of materials with different refractive indexes [13,14,15,16,17,18]. This configuration creates a photonic bandgap, and no photon whose energy lies within the bandgap can propagate through the PhC. By setting defects in PhC structures, photons located in the bandgap can be confined within the defects. Changing the position or shape of the defects allows the defects to be used as cavities or waveguides, enabling fine manipulation of light at the nanometer level. Due to this feature, various PhC lasers have been researched and developed [19,20,21,22,23,24,25,26]. However, current-injected lasers are essential for optical interconnects, but there are few reports on room-temperature continuous-wave (RT-CW) lasing in current-injected PhC defect lasers. G. Crosnier et al. reported laser diodes based on 1-dimensional PhC nanocavities made in InP nanoribs. RT-CW single-mode operation was obtained, but the threshold current of 100 μA was too high for optical interconnects [27]. The lambda-scale embedded active region photonic-crystal (LEAP) laser developed by Nippon Telegraph and Telephone Corporation (NTT) is a PhC defect laser that allows lateral current injection. It achieved RT-CW lasing and has a low threshold current of 4.8 μA and an operating energy of 4.4 fJ bit⁻¹ [28,29,30,31]. E. Dimopoulos et al. reported a nano-buried-heterostructure PhC laser with a structure similar to the LEAP laser in 2023, featuring an ultra-low threshold current of 730 nA. The laser can be modulated at 3 GHz at an energy cost of 1 fJ bit⁻¹ [32]. However, neither of these lasers is available for WDM, and the electrodes for current injection occupy a large area, so they are still far from practical application.

We proposed a circular defect in a 2D photonic crystal (CirD) laser that has the potential to be used as an intra-chip optical communication device [12,33] and allows vertical current injections. We need to develop fabrication technology due to the large thickness of the device for vertical current injection. It is fabricated on a GaAs/AlGaAs heteroepitaxial wafer. The structure consists of a GaAs core layer containing InAs quantum dots (QDs), sandwiched between two AlGaAs cladding layers, and a GaAs contact layer covering the upper cladding layer. The AlGaAs cladding layers are oxidized to AlGaO_x, which confines the light to the core layer [34]. The center of the CirD cavity remains unoxidized to form an AlGaAs funnel that allows a current injection into the cavity. The whispering gallery mode (WGM) is generated by injecting power into the cavity. If the frequencies of the WGM match those of the waveguide mode, the light enters the waveguide and propagates through it, eventually exiting its end face. The lasing wavelength can be changed by adjusting the cavity radius. In the future, WDM can be achieved easily by coupling multiple CirD cavities, each with a different lasing wavelength, to a single output waveguide. We previously proposed an orthogonal lattice waveguide (OLW) [35]. We demonstrated experimentally that it has a bandwidth of 20 nm and can support 20-channel WDM with a bandwidth interval of 1 nm [36]. The size of a CirD laser array containing 20 cavities can be limited to 10⁻⁴ cm² [12], so it is possible to achieve optical communications within the chip due to its ultra-small size.

As mentioned above, in the future conception, 20 CirD cavities would combine with a single waveguide and output light (see Figures 3 and 4 in [12]). The CirD cavity furthest from the end face is approximately 100 μm from the end face. We found that even though the distance between the cavity and the end face is 10 μm, the light emitted from the waveguide is very weak, due to the absorption of light by the QDs in the waveguide as the light is transmitted through the waveguide. To solve this problem, we are trying to remove the QDs outside the CirD cavities in another work [37,38,39]. Until this work is completed, to accurately evaluate the light coming out from the waveguide, it is a reasonable and feasible plan to investigate a single CirD cavity combined with a waveguide (see Figure 1 in [12]), which is exactly what we are working on. When fabricating a sample with a single cavity, the output end face is usually formed by cleaving the sample along the direction perpendicular to the waveguide. However, since the cleavage position has an error of at least 20 μm, the actual cleavage position affects the output optical power for a nanoscale sample. We proposed a convex edge structure in a previous study and performed simulations. The optical output can be improved by setting semicircular ports on the output end face and changing the waveguide near the port to a line defect (W1) waveguide (see Figure 4 in [40]). In this paper, we develop fabrication techniques for a CirD laser with a convex edge structure that can eliminate the effects caused by the cleavage position. We also fabricate samples and measure their optical characteristics by photoexcitation.

2. Structure Design and Fabrication

The schematics of the edge structure produced by conventional cleavage and the convex edge structure are shown in Figure 1a and Figure 1b, respectively. In the convex edge structure, only the GaAs core layer is retained near the output end face and floats in the air, which we refer to as the membrane structure. As can be seen from the top view, the membrane structure is connected to the wafer only on the left side. As long as the cleaving position is within the membrane structure, the membrane structure will not be damaged after cleavage. In addition, the cavity retains upper and lower AlGaAs/AlGaO_x cladding layers that still allow current injection.

We initially designed an electron beam (EB) lithography pattern to fabricate the structure shown in Figure 2a. By setting a deep trench shaped like the alphabet C around the end face, this structure can be formed by cleaving within the cleavage range shown in this figure. However, when cleaved within the 3 μm range, the GaAs substrate will protrude more than the membrane structure, obstructing the optical fiber used to collect the output light close to the waveguide during optical measurements. Therefore, we improved the design to a symmetric construction shown in Figure 2b. Due to the symmetrical construction, one side of the GaAs substrate will not obstruct the fiber after cleaving the sample. However, this structure has a cleavage range of only 10 μm, which is still less than the cleavage error. As shown in Figure 2c, by aligning multiple patterns in the y direction and shifting in the x direction, the cleavage ranges of two neighboring patterns overlap to some extent, and the cleavage tolerance can be increased. In this work, we aligned three patterns in the y direction and shifted them by 7 μm in the x direction, resulting in a cleavage tolerance larger than 20 μm. Although only one of the six samples can be evaluated under this design, the current goal of our study is to accurately assess a single CirD laser, which will pave the way for future array integration.

The process flow diagram is shown in Figure 3. The thicknesses of the GaAs contact layer, upper AlGaAs cladding layer, GaAs core layer containing three InAs QD layers, and lower AlGaAs cladding layer are 180, 550, 220, and 550 nm, respectively. The photoluminescence peak of the QDs is 1289 nm. Steps 1 and 2: We drew PhC and edge structure patterns on the wafer by EB lithography and etched the patterns by dry etching. Here, the lattice constant of triangular lattice a and the airhole radius r are 365 and 115 nm, respectively. The edge face position d is 165 nm, which is the distance from the edge face to the center of the airhole row closest to the edge face. The radius of the semicircular port R_p is 420 nm (see Figure 1). The radius of CirD laser cavity R is 2.78a. The values of these parameters were adopted from the optimal values previously obtained by calculation [40]. Step 3: We oxidized AlGaAs to AlGaO_x in H₂O/N₂ steam at 395 °C for 45 min. At this point, the AlGaAs at the cavity would be completely oxidized to AlGaO_x, and current injection would be impossible. This is because ultimately only optical measurements would be performed, and complete oxidation can enhance light confinement. Steps 4 and 5: We performed an overlaid EB lithography and another dry etching to remove the contact layer of the membrane structure. Step 6: We removed the AlGaO_x layers of the membrane structure by buffered hydrogen fluoride wet etching. Steps 4 and 5 were not taken in the initial fabrication process. But even without these two steps, the fabricated sample’s membrane structure would float in the air, and only the contact layer that was not removed by wet etching would remain above it. However, after fabricating some samples, we noticed that the contact layers collapsed and bent the core layer. Figure 4 shows the cross-sectional SEM image of a sample with a collapsing contact layer. Therefore, we added steps 4 and 5 to remove the contact layer of the membrane structure. Step 7: We cleaved the sample. The top-view SEM image of the sample after cleavage is shown in Figure 5. It shows that the sample was successfully fabricated as designed. On the surfaces of the membrane structure, unknown particles can be observed in some of the holes. However, since they are far away from the waveguide and there are no particles on the waveguide, we do not consider that they will affect the optical confinement.

3. Optical Measurement and Results

We used a semiconductor laser with a wavelength of 785 nm as the pump light source. The laser beam was focused on the surface of the CirD cavity using a microscope’s objective lens, resulting in a spot size of 2 μm. The convex edge emitted light collected with a lensed fiber that was split into two paths. One path directed the light to a photodiode (PD) array (ANDOR, iDus InGaAs 1.7 µm) via a monochromator (LUCIR, Z-3008). The other path directed the light to an optical spectrum analyzer (OSA: Yokogawa, AQ6370D).

The PD array was used for quick measurements of the spectra to determine the threshold powers. The OSA was employed to measure the spectra with higher precision and determine the full width at half maximum (FWHM). The monochromator and the OSA had resolutions of approximately 0.4 and 0.02 nm, respectively. These instruments allowed for accurate spectral analysis in the experiments.

Figure 6 shows the relationship between output power and input power for the sample measured by the PD array and monochromator. The lasing threshold power (P_th) can be obtained as P_th = 50 μW. This value is close to the threshold power reported in our previous study for the well-fabricated sample cleaved by the conventional method [41].

Figure 7a shows the spectrum of this sample measured by the OSA, and the input power was 210 μW. A spectrum with a wavelength span = 60 nm at the same input power was also measured, and single mode lasing due to WGM was observed in the 1.3 μm range. Figure 7b shows the spectrum measured at the input power of 48 μW, which is close to the threshold power. Its FWHM or Δλ was estimated to be 0.29 nm by fitting it to a Gaussian function. The quality factor (Q) of a CirD cavity can be calculated as 4433 using the formula Q = λ/Δλ. Since the convex edge structure hardly changes the Q of the CirD laser, it is essentially the same value of Q as the well-fabricated samples obtained by the conventional cleaving method, which is consistent with the simulation results for the conventional structure [42]. The optimum Q of PhC defect lasers for optical interconnects is 3000–4000. A previous study has shown that too high Q negatively impacts high-speed operation [43]. These results indicate that CirD lasers with convex edge structures have similar optical properties to the well-fabricated CirD lasers cleaved by conventional methods in previous studies.

4. Conclusions

Since the output light intensity of the CirD laser changes due to the different cleaving positions, we designed the EB lithography patterns of convex edge structures to eliminate the effect of different cleaving positions and ensure that the cleavage tolerance is larger than the cleavage position error. We also developed fabrication techniques for CirD lasers with convex edge structures. By performing an overlaid EB lithography and dry etching to remove the contact layer of the membrane structure, we prevented the collapse of the contact layer and thus fabricated samples successfully. We measured the optical characteristics of samples by photoexcitation. The RT-CW lasing of the WGM was observed, and it has almost the same optical characteristics as the well-fabricated samples reported in previous studies. We expect these results to contribute to future work such as accurate manufacturing and improving the end face configuration to obtain higher outputs.