Improvement of output efficiency of p-face up photonic-crystal surface-emitting lasers

: We optimized the p-side emission device configuration of photonic-crystal surface-emitting laser (PCSEL) to facilitate the easier chip process and wafer level testing as well as the feasibility of lasing at shorter wavelength. Typically, in order to obtain uniformly distributed current for larger emission area of PCSELs, laser output is designed through the n-side window due to the low hole mobility and thin p-side cladding layer. However, the substrate as well as the epi-layers have to be isolated before the test of each single die on the wafer, which compromised the advantage of wafer-level test of surface emitters. On the other hand, for lasers with emission photon energy higher than the bandgap energy of GaAs substrate, the power will be entirely attenuated. In this study, the optimized p-side emission by applying the transparent conduction layer on top of the p side contact layer to enhance the current distribution and breaking the symmetry of conventional circle pattern in a unit cell to boost the output efficiency is investigated. Through this approach, a high efficiency p-side up PCSEL platform with lower fabrication cost is developed, which is also applicable for short wavelength PCSELs.


Introduction
Surface emitting lasers exhibit several advantages over edge emitting lasers such as facilitation of two-dimensional (2D) array integration, high yield, wafer level testing, and circular output beam for better coupling efficiency to the optical system, leading to the rapidly soaring demand in many applications. The vertical-cavity surface-emitting lasers (VCSELs) have been vastly applied as the light sources for data communication and is promising for the wavelength division multiplexing (WDM) platform for increasing demand of communication bandwidth [1][2][3][4][5]. The blue VCSEL array demonstrating to deliver optical power up to watt class is heading for the automobile applications [6]. The near-IR VCSELs arrays are also adopted as the light sources of the highly increasing demand of three dimensional (3D) sensing and light detection and ranging (LIDAR) technologies for consumer electronics [7][8][9][10]. With the rapid development of surface emitting lasers, the prevailing of various applications as mentioned above are accelerated.
For applications required watt-class optical power, arrays composed of hundreds of individual VCSEL device are typically applied [11,12]. While for the photonic-crystal surface-emitting laser (PCSEL), benefited from the distributed feedback oscillation of the entire surface area, enables the delivering of high power with single mode and small divergence angle laser beams [10,[13][14][15][16][17][18][19]. In comparison to VCSELs, PCSELs are capable of manipulating the output beam angle without external optical components, which is advantageous for applications required multiple output beam directions [10,[20][21][22]. To achieve large area emission, uniformly distributed current and hence the evenly distributed gain is required. Hence, the laser emission window is typically designed at the n-side, utilizing the advantage of the thick and highly conducted substrate to achieve uniformly distributed current. While the p-side is covered with the metal layer down to the submount to assist uniform current distribution [9]. Under this device structure, the testing of device properties requires the isolation of individual die to prevent the parallel connecting through the n-type substrate and cladding layer. On the other hand, the band-to-band absorption of the GaAs hinders the shorter wavelength emitting from n-side. In contrast, the isolation of p-side up device structure can be achieved by a shallow mesa etching or the photonic crystal (PC) surrounding proposed in this study, which facilitates the easier wafer-level testing. Otherwise, the light propagating toward the substrate can be reflected by distributed Bragg reflector (DBR) to avoid the absorption of the GaAs [23,24], hence the shorter wavelength can be achieved.
To surmount the current distribution issue, thin indium-tin-oxide (ITO) layer is applied on top of the p-type contact layer to enhance the lateral carrier transportation [16,[25][26][27]. The p-side up PCSEL with and without ITO layer on top are compared including the light-current-voltage (L-I-V) characteristics and the carrier distribution represented by the near field distribution. In this study we also investigate the effect of destructive interference of the electric field resulted from symmetric circular airhole structure, which leads to low optical output power [14]. In order to improve the output efficiency, different air hole shape including the equilateral triangular (ET) and isosceles right triangle (IRT) are applied for PC layer to compare the output efficiency on the p-side up configuration with the circular (CC) PC pattern.

Figures 1(a) and (b)
show the respective schematic of p-side up PCSEL without (sample A) and with (sample B) ITO current conduction layer. The epitaxial layers for PCSEL grown by metal organic chemical vapor deposition (MOCVD) on n-type GaAs substrate was primarily consisted of an n-Al 0.45 Ga 0.55 As cladding layer, an active region, a PC layer, a p-Al 0.45 Ga 0.55 As cladding layer, and a p+-GaAs contact layer. The active region sandwiched between n-type and p-type cladding layers was consisted of three In 0.25 Ga 0.75 As quantum wells (QWs) separated by GaAs barrier layers aiming the lasing wavelength around 945 nm. The PC consisted of square lattice with periodic ET air holes was fabricated through e-beam lithography and inductively coupled plasma -reactive ion etching (ICP-RIE) at the p-GaAs PC layer, and the PC area is 125 µm × 125 µm square. The p-Al 0.45 Ga 0.55 As cladding layer and succeeding layers was grown on the processed wafer through MOCVD and the growth parameters have been adjusted to retain the air holes as much as possible. The filling factor (FF) of the PC structure was designed to be 18% after regrowth of the p type cladding layer, and the lattice constant (the period of the pattern) was 281 nm for the lasing wavelength near 945 nm. The PCSEL device process on the epi-wafer was began with the mesa etching through ICP-RIE followed by the SiN x passivation layer deposition through plasma enhanced chemical vapor deposition (PECVD) to constrain the current within the emission area. A current aperture revealing the p+-GaAs for the formation of ohmic contact was defined by photo lithography and etching off the SiN x layer by buffered oxide (BOE) solution. To investigate the effect of current distribution, an additional 200 nm-thick ITO layer was deposited on top of the p+-GaAs contact layer of sample B through e-gun evaporation in prior to the deposition of p-metal. The p-metal was composed of Ti-Au to form ohmic contacts both on p+-GaAs (sample A) and ITO (sample B).
Figures 1(c) and (d) show the plane view images of sample A and sample B observed through optical microscopy (OM), respectively. The current injection is through the metal ring with inner and outer diameter of 60 µm and 80 µm for sample A, which is the only ohmic contact region with p+-GaAs as shown in Fig. 1(c). While for sample B, the current may transport through the ITO layer and injected into the entire aperture with a 100 µm diameter which is shown in the purple color region inside the metal ring of Fig. 1(d). The PCSELs with and without ITO were process on the same epi-wafer to control the variables. The light-current-voltage (L-I-V) characteristics, near and far field distribution of the two PCSEL configurations were investigated under driving current of pulse width, 1µs; duty cycle, 0.1%; and temperature, 300 K.

Results and discussion
Figure 2(a) shows the L-I-V curve of sample A. The respective threshold current density (J th ) and slop efficiency of sample A are 1.28 kA/cm 2 and 0.016 W/A, respectively. The J th of sample B is 0.4 kA/cm 2 , which is nearly one-third of that of the sample A as shown in Fig. 2(b). The slop efficiency of sample B also shown in Fig. 2(b) is 0.055 W/A, which is approximately 3∼4 times of that of the sample A. The dynamic resistance of sample B appeared to be higher than sample A. Although the carrier concentration and mobility of ITO are 2.2 × 10 20 cm −3 and 27 cm 2 /V-s, respectively, which leads to a highly conductive layer, the larger resistance was most likely resulted from the contact resistance between ITO and P+-GaAs. The annealing process performed for Ohmic contact formation should be further refined. The spectra for the two samples are nearly identical as shown respectively in Figs. 2(c) and 2(d), indicating the lasing wavelength is mainly dependent on the PC lattice. The broader lasing spectrum of sample A should possibly be resulted from the unevenly distributed current around the p type electrode as discussed below, which caused slightly variation of the refractive index of each region.
The near-field pattern of sample A measured at 0.8×Ith shows that the light emitted only in the vicinity of p type metal contact as shown in Fig. 3(a). While for sample B, the radiation pattern uniformly distributed in the aperture as displayed in Fig. 3(d). The x-y axis indicated in emission pattern is corresponding to the OM pictures in Fig. 1 that the relevant section can be identified. As measurement current increased to 1.2×I th , the current crowding effect was even worse for sample A as shown in Fig. 3(b). From this picture, the square boundary of PC region is clearly observed and the emission concentrated between the p-metal ring and the four edges of the PC boundary. This could be resulted from the current that is apt to transport This could be resulted from the current that is apt to transport toward the region without the PC in which lower resistance is experienced. The distinct boundary of emission pattern is due to the propagated light being scattered vertically by the PC, while outside the PC region, the in-plane propagation light is below the light line and evanescent eventually. In contrast, the near-field image of sample B above threshold shows concentrated emission in the aperture as displayed in Fig. 3(e). This emission pattern is due to the carrier restricted to transport through the aperture inside the metal ring defined by the SiN x passivation layer. The highly conducting ITO layer above the p+-GaAs assisted the lateral transport of injected carrier and hence the uniform emission in the aperture is preserved. Figure 3(f) shows the far-field pattern of sample B from which a singled peak with low divergence is observed as predicted. While for sample A, a main peak with side lobes is shown in Fig. 3(c), which is caused by the interference of the peaks shown in the near-filed pattern. The measured beam quality factor (M 2 ) for sample A and B are 1.4 and 3.1, respectively. The beam quality factor of sample B enables the delivery of small focused spot size and expected to be smaller as previous report as further increase the emission area [9]. According to the emission pattern analyzed above, the higher J th of sample A can be attributed from the higher mirror loss owing to its carrier concentrated near the boundary of PC layer where the optical feedback is much lower than that of central region of the PC. The lower slop efficiency can be explained that the generated light propagating toward the region without PC is under the light line and evanescent eventually. As observed on sample B, combining the ITO conducting layer and the SiN x insulator, the current flow can be guided into the desired region for maximized efficiency. Based on the optimized p-side emission configuration, we investigated the correlation of the air hole geometries of the PC layer to the output efficiency of the PCSELs theoretically and experimentally.
The output efficiency was evaluated through the calculation of radiation constant of PCSELs with different FF of respective PCs geometry composed of CC, ET, and IRT air holes which represented different symmetries. The electric field distributed evenly in symmetric PC pattern, and resulted in destructive interference entirely [28]. The introduction of asymmetry breaks the even distribution of the field and hence alleviated destructive interference accompanied with higher radiation. The three fundamental geometries with different degree of symmetry were chosen for facilitation of comparison. The FF in the simulation assumes the air holes are invariant along the z direction and the FF was calculated as the surface area of one air hole divided by the area of one unite lattice. The radiation constant correlated to the light out-coupled by the PC can be calculated according to the following equation, where the quality factor Q is obtained by solving electromagnetic wave equation in the unit cell of the PCSEL with infinite periodicity using the finite element mode solver [28,29]. The in-plane propagating loss can be ignored with this infinite periodic structure Where β 0 and a represent the propagation constant and lattice constant of the PC, respectively. The results depicted in Fig. 4 reveal the radiation constants are increased with the asymmetry of the PC pattern that the IRT geometry delivers the highest optical power. The radiation constant of CC pattern is almost irrelevant to the filling factor from the calculation, namely protected by the symmetry. The radiation constant of ET and IRT peaked around the FF of 25% as shown in the figure. At lower FF, the scattering of radiation increased with the size of air hole. While for the larger FF, the transverse mode profile is shifted to the active region due to the average index of PC layer is getting smaller, and hence less field extended to the PC region to be scattered. Figures 5(a), (b) and (c) represent the SEM images of the CC, ET, and IRT air holes before the regrowth of p cladding layers, respectively. Figures 6(a), (b) and (c) show the experimental results of the PCSELs with FF, 18%; a, 281 nm; and PCs air hole shape of CC, ET, and IRT, respectively. The respective threshold current densities of PCSELs with CC, ET, and IRT air hole shapes are 0.47 kA/cm 2 , 0.4 kA/cm 2 , and 0.62 kA/cm 2 as shown in Fig. 6. The slop efficiency of the PCSELs with CC, ET, and RIT air hole shapes are 0.014 W/A, 0.055 W/A, and 0.114 W/A, respectively.
The slop efficiency increases with the degree of asymmetry, which agrees well with the trend of simulation. The effect of FF of the PC was also studied, the comparison IRT air hole with FF of 0.16, 0.18, and 0.2 is shown in Fig. 7. The J th of the sample with FF of 0.16, 0.18, and 0.2 are 0.66 kA/cm 2 , 0.71 kA/cm 2 , and 0.73 kA/cm 2 , respectively. The corresponding slop efficiency of the three samples are 0.12 W/A, 0.21 W/A, and 0.23 W/A. The radiational loss increases with FF results in higher J th as well as higher slop efficiency, which again agrees well with the calculation.

Conclusion
In this study, we optimized the p-side emission structure of PCSEL which will be beneficial for easier chip processing, wafer level testing and further development of short-wavelength PCSELs. With the aid of ITO transparent conducting layer, the injected carrier distributed uniformly in the emission window, and hence resulting in the improvement of device efficiency. This device configuration facilitates the wafer level testing and shorter wavelength applications that suffer from the absorption of GaAs substrate. The influence of PC geometries and FF were studied on the optimized platform. The slop efficiency of the laser device increases with the degree of asymmetry of the air hole shape. The trend of the experimental results agrees well with the calculation of radiation constant based on the Q factor of each simulated PCSEL devices.