We demonstrate a novel high yield fabrication process for single-mode ridge-waveguide GaAs/AlGaAs ring lasers with significantly lower threshold currents than previously reported for similar devices. In this fabrication process, the ridge waveguide structure is patterned using a metallic etch mask, which survives ensuing fabrication steps to form a continuous metallic cover over the entire resonator structure. This metallic cover improves the uniformity of electrical contact between the resonator structure and the metallic biasing layer deposited at the conclusion of the fabrication process. This leads to optimum electrical pumping of the fabricated devices. This fabrication process also allows for the passivation of the ridge-waveguide device sidewalls and separation of the metallic biasing layer from the optical mode.
©2010 Optical Society of America
Semiconductor ring lasers (SRLs) can be integrated in a myriad of opto-electric applications. SRLs with ring resonators of various geometries, including circular [1–4], racetrack [5,6], square [7,8], and triangular , have been proposed, fabricated and characterized. Typically, these devices feature micron-scale ridge-waveguide mesa structures, i.e., resonator cavity and evanescently coupled waveguide, etched into substrates with quantum wells. The ridge-waveguide structure serves to confine and guide the optical mode generated at its base when the device is biased.
Precise control over the etch depth of these ridge-waveguide structures is the most critical aspect of realizing a successful SRL device. Whereas etching less than the optimal depth leads to bending loss, etching too deep leads to scattering losses . Sidewall roughness of the ridge-waveguide structure can also be a major impediment towards achieving viable devices since rough sidewalls lead to scattering losses in the generated optical mode, excitation of higher-order propagating modes and reduced quality factor of the resonator [10,11]. The output coupler design can also significantly influence device performance. Various output coupling designs, such as evanescent coupling, Y-junction, and multimode interference (MMI) couplers, have been studied extensively [12,13].
Another critical factor in SRL fabrication involves avoiding metallization in the vicinity of the optical mode during electrode layer deposition in order to prevent absorption losses. At the same time, it is also important that the electrode layer maintain optimum electrical contact with the top surface of the resonator structure for adequate electrical pumping of the entire ring resonator. There are established methods that can achieve the above, which involve the application and selective removal of dielectric materials from etched substrate. These methods are also effective in passivating the ridge-waveguide structures. However, these methods can frequently lead to either remnant dielectric material on top of the ridge-waveguides, resulting in inadequate electrical contact, or metallization near the optical mode, which introduces absorption loss.
In this study we present a fabrication method that is more effective at achieving separation between metallic electrode layer and the optical mode and, at the same time, ensures optimum electrical pumping of the SRL. This fabrication process involves the use of a metallic mask for etching the device structure on to the substrate. As such, this etch mask remains intact through the subsequent processing steps, which includes a polymer planarization process, and forms a continuous metallic cover over the entire device structure. Consequently, this metallic cover ensures optimal electrical pumping of the entire ring resonator even when only a part of it is in proper electrical contact with the metallic biasing layer due to planarization non-uniformity. The fabrication method presented in this study has led to devices with remarkably low threshold currents and single mode behavior with a very high yield.
1.1 Separation of electrode layer from optical mode and passivating devices sidewall
There are several methods that achieve sidewall passivation and electrode layer separation from the optical mode. One of the more popular among these involves an oxide layer deposition over etched substrates followed by photo lithographically opening etch-windows over the device structure where the oxide layer is removed to expose the top of the resonator structure . A disk-shaped electrode layer is then patterned on top of the resonator by standard photolithography, deposition and lift-off. However, for fabrication of viable devices with this method, it is important that the width of the oxide etch window be equal to or less than that of the resonator in order to avoid metal deposition at the base of the ridge-waveguides where they can interact with the optical mode and result in absorption losses . Also, exposure of the ridge-waveguide sidewalls can give rise to chemically active dangling bonds and surface states which also results in optical losses . Hence, complicated and precise alignment procedures are required to ensure that the etch windows are smaller than the width of the ridge-waveguide, which for single-mode devices, are on the order of 2μm.
Polyimide planarization is another method of separating the electrode layer from the base of the ridge-waveguide structures . This involves spin-coating and curing of a thick layer of a planarizing polyimide directly above the etched substrate followed by a carefully calibrated etch process to remove the planarizer layer until the top of the device structures have been exposed. Metallization is then done for biasing. However, spin-coating the planarizing polyimide layer on top of the non-uniform etched sample and polyimide reflow during curing can bring about non-uniformity in the planarizer layer coverage across the sample . Therefore, after planarizer etching, planarizer material may persist on top of some portion of the resonator structure while the rest of it is already exposed. Proceeding to metallization from here will result in inadequate electrical pumping of the resonator structure when the device is biased . On the other hand, prolonging the etch process to remove remnant planarizer material from the top of the resonator structure might expose the ridge-waveguide sidewalls whereby dangling bonds can lead to optical losses when the device is pumped [16–18]. Over-etching can also reduce or even remove the planarizer coverage at the base of the ridge-waveguide devices that are already exposed. This will result in metal deposition at the bottom of the resonator structure during metallization, and eventually absorption losses when the device is biased [4,15].
A GaAs/AlGaAs substrate with three 70Ǻ GaAs quantum wells separated by AlGaAs barrier layers has been used for fabricating the devices in this study. The top (p-type) cladding for these substrates comprises a 1.4μm thick layer of p-doped AlGaAs overlying the active layers. The n-type cladding consists of a 1.4μm thick n-doped AlGaAs layer underlying the active layers. A 200nm thick highly p-doped GaAs layer is used as the contact layer (top-surface of the wafer). The substrate handle is highly n-doped GaAs.
Figure 1 illustrates the fabrication process presented in this study. After backside metallization, negative-tone UV Lithography is performed on the substrate using a bright-field mask and AZ5214E photoresist in the image reversal mode. This yields a photoresist mask comprising grooves corresponding to the device structures onto the substrate as shown in step 1 of Figure. 1. The height of the resist structures is around 1µm.
The samples are then transferred to an e-beam evaporator where a 20nm Ti layer is deposited for adhesion promotion of the subsequently deposited 300nm Au layer, as shown in step 2 of Fig. 1. This is followed by a lift-off process in which the samples are immersed in acetone and rinsed in isopropanol to yield the 300nm Au etch mask as illustrated in step 3.
In step 4, the samples are transferred in to an Inductively Coupled Plasma (ICP) tool for anisotropic etching of the metal masked substrate, using Cl2 (3sccm), SiCl4 (3sccm) and BCl3 (1sccm) gasses at source and bias power of 100W and 90W respectively under a vacuum of 1Pa, to yield the ridge-waveguide device structures.
The planarizing polymer Dow Cyclotene 3022 is then spin-coated on top of the etched samples, prebaked at 95°C for 5 minutes, and cured in a N2 environment at 300°C for 30 minutes, as shown in step 5. The Cyclotene layer is approximately 4μm thick. After the planarizer layer has been cured, the samples are transferred into an ICP tool where the planarizer is etched using fluorine C4F8 and SF6 plasma in a carefully calibrated process until the top of the Au deposited ridge-waveguide devices are exposed, as depicted in step 6.
After the conclusion of the planarizer etch-back, disc-shaped deposition windows are opened on top of the ring resonator structures using UV lithography with the AZ5214E resist in image reversal mode. The samples are then transferred into an e-beam evaporator where a 20nm of Ti is deposited for adhesion promotion prior to the deposition of a 300nm layer of Au. After metal deposition, the samples are immersed into acetone and rinsed with isopropanol to remove the resist patterns and the Au deposited over them to yield the disc electrodes on top of the ring resonators as shown in step 7.
2.1 Advantages of the fabrication process presented in this study
The Cyclotene planarization process used for the fabrication process presented in this study, and indeed the other methods to passivate the ridge-waveguide structures and to separate the optical mode from the electrode layers, can bring about inadequate electrical contact between the electrode layer and the devices as discussed earlier. These difficulties are vividly illustrated in Fig. 2 , which depicts a sample after planarizer etch-back. One can clearly observe that whereas some parts of the ring resonator are exposed, there are portions of the ring that the planarizer still covers. Proceeding to metallization from here would result in poor electrical contact uniformity between the electrode layer and parts of the ring resonator which are still planarizer covered, which would lead to poor device performance. On the other hand, opting to prolong the planarizer etch-back to remove the planarizer cover persisting on portions of the ring will result in thinning or even removal of the planarizer coverage near the base of the resonator structure. Metallization at this point will result in metal deposition at or very near the base of the resonator structures, which would lead to absorption losses when the device is pumped. Prolonging the planarizer etch-back can also lead to exposed ridge waveguide sidewalls, which can lead to optical losses as described earlier.
For the devices fabricated in this study, after the planarizer etch-back (step 6 in Fig. 1), about 200nm thickness of the original Au etch mask survives. This can be observed clearly in Fig. 3 , which provides a cross-section view of a device after the planarizer etch-back. Thus, even when planarization is non-ideal, as seen in Fig. 2, the entirety of the device is adequately pumped if any one part of the metal-capped ring resonator is in proper contact with the metal electrode deposited at the conclusion of the fabrication process. The surviving metal etch mask also provides some leeway for the planarizer etch-back step, i.e. the etch-back process can be extended somewhat without exposing sidewalls. These advantages have resulted in significant improvement in reproducibility of successful devices. Better electrical contact uniformity also leads to the reduction of threshold current, which bring about improved device performance.
The devices fabricated in this study had ring resonators 2μm wide and 500μm in radius, and a 2μm wide straight output waveguide directionally coupled to the resonators, as shown in the optical image of a fabricated device in Fig. 4 . The etch depth for all of the fabricated devices has been carefully calibrated to be approximately 1.3µm, which our simulations have indicated to be optimum for confinement of the optical mode within the ridge-waveguides. Bending loss at 1.3μm etch depth has been calculated to be 0.67dB/cm. However, bending loss can change dramatically depending on etch depth. Our simulations show that if the etch depth is changed to 1.25μm, the bending loss goes as high as 6dB/cm. On the other hand, increasing etch depth to 1.35μm reduced bending loss to 0.002dB/cm. However, at this etch depth the optical mode has increased interaction with the ridge-waveguide sidewalls where sidewall-roughness can lead to significant scattering losses, whereby threshold current values will increase dramatically. The device yield with the proposed fabrication process has been more than 95%.
The optical output power with respect to pump current for one of the devices fabricated with the process presented in this study is shown in the graph depicted in Fig. 5 . For this device, the threshold current, which corresponds to the current value where the curve spikes upwards (indicating onset of lasing), is approximately 62mA. The erratic nature of the plot in Fig. 6 is partly due to mode competition of the counter propagating modes within the resonator structure. It is also due to the fact that the device substrates were not cooled during the measurement. This results in expansion of the resonator as the substrate heats up with increasing pump current which leads to slight shifts in the lasing wavelength. This can result in erratic optical output power versus pump current plots.
The threshold currents for the devices fabricated in this study have varied between 60mA (0.95kA/ cm2) and 75mA (1.20kA/cm2). The variability of the threshold currents is most likely due to day to day fluctuations of the ICP equipment whereby etch depth can vary slightly from sample to sample resulting in a variability of bending loss between samples. Etch depths for the devices fabricated in this study have varied from 1.28μm to 1.32μm.
The significance of the above threshold current values becomes clear when they are compared to those for devices that we fabricated using a fabrication process reported in reference . These devices do not feature a metallic etch mask. For these devices (without metallic masks), a polymer etch mask was lithographically defined on the substrate (identical to the ones for the devices with metal etch masks) using a dark-field mask with exactly the same device patterns and lithography parameters, (e.g., exposure dose, bake time, development time) used for the samples with metallic etch masks. The samples were then etched to the optimal depth of 1.3µm using the same dry-etch recipe described earlier. This was followed by planarization and contact-layer deposition steps identical to those performed for the devices with metallic etch masks. For contact-layer deposition, the same evaporation crucibles and e-beam evaporator were used for both processes, whereby the quality of the metal layers deposited for both cases was the same. Thus, the devices fabricated with the polymer etch masks are essentially the same as the devices with the metallic etch mask except for the metallic top-surface of the latter. Therefore, many of the parameters, (e.g., sidewall roughness, etch depth, output coupler design, coupling loss, bending loss,) which can significantly affect threshold current levels are on the average the same for polymer masked and metal masked devices.
Notwithstanding the above, the threshold currents for the devices with the polymer etch masks were significantly higher than the devices with metallic etch masks, ranging between 80mA (1.27kA/cm2) and 120mA (1.97kA/cm2). The lower threshold current densities of the devices with metallic etch masks clearly shows an improvement of device performance attributable to the enhanced electrical contact provided by the new fabrication process.
The improved electrical contact for the metal masked devices is further confirmed in Fig. 6, which plots forward resistance with respect to pump current for a device fabricated using each of the processes. The polymer masked device used for this plot had a threshold current of 92mA as compared to 67mA for the metal masked device. All of the samples fabricated for this study were cleaved after testing to measure the exact height of the ridge-waveguides. Both of these samples had a ridge-waveguide etch depth of approximately 1.28μm. Thus, it can be safely assumed that the bending and scattering losses for both of these devices were similar. The higher resistance of the polymer masked samples indicates that parts of the ring resonator were likely still planarizer covered. The planarizer obscured portions of the ring fail to provide any gain and hence act as passive waveguides contributing only towards the optical losses in the resonator. Hence a larger pump power is necessary for lasing. On the other hand, for the metal masked devices the metallic cover overlying the device structure ensures that the entire resonator is pumped even when parts of the resonator structure are planarizer covered. This results in lower forward resistance and lower threshold currents.
The threshold current densities for the metal masked devices fabricated in this study are about 40% lower as compared to similar devices reported in references  and . The coupling losses for these studies are 1-5%  and 10%  as compared to approximately 2% for our devices at the optimal etch depth of 1.3µm, as per simulation results.
Figure 7 depicts the output spectrum of one of the fabricated ring lasers acquired using an optical spectrum analyzer across a 1-nm wavelength span at a 65mA pump current. One can observe the dominance of a single mode around 873.35 nm with a side mode rejection of at least 10dB around the dominant mode.
A novel method for high yield fabrication of single-mode large-diameter GaAs/AlGaAs ring lasers has been presented which has yielded devices with significantly lower threshold current levels as compared to similar devices reported earlier. This fabrication method also achieves sidewall passivation of the device structure and the separation of the optical mode from the electrode layers that improves performance of the realized devices. This method involves the use of a metallic etch mask, which remains intact through the entire fabrication process to form a continuous metallic cover on top of the entire ridge-waveguide device structure which ensures optimal pumping of the entire resonator structure.
References and links
1. T. F. Krauss, P. J. R. Laybourn, and J. S. Roberts, “CW operation of semiconductor ring lasers,” Electron. Lett. 26(25), 2095–2097 (1990). [CrossRef]
2. M. Sorel, G. Giuliani, A. Scir`e, R. Miglierina, S. Donati, and P. J. R. Laybourn, “Operating regimes of GaAs-AlGaAs semiconductor ring lasers: Experiment and model,” IEEE J. Quantum Electron. 39(10), 1187–1195 (2003). [CrossRef]
3. M. Sorel, P. J. R. Laybourn, G. Giuliani, and S. Donati, “Unidirectional bistability in semiconductor waveguide ring lasers,” Appl. Phys. Lett. 80(17), 3051–3053 (2002). [CrossRef]
4. J. P. Hohimer, D. C. Craft, G. R. Hadley, and G. A. Vawter, “CW room temperature operation of y-junction semiconductor ring lasers,” Electron. Lett. 28(4), 374–375 (1992). [CrossRef]
5. G. Griffel, J. H. Abeles, R. J. Menna, A. M. Braun, J. C. Connolly, and M. King, “Low threshold InGaAsP ring lasers fabricated using bi-level dry etching,” IEEE Photon. Technol. Lett. 12(2), 146–148 (2000). [CrossRef]
6. J. P. Hohimer and G. A. Vawter, “Unidirectional semiconductor ring laser with racetrack cavities,” Appl. Phys. Lett. 63(18), 2457–2459 (1993). [CrossRef]
7. S. Oku, M. Okayasu, and M. Ikeda, “Low-threshold operation of square-shaped semiconductor ring lasers (orbiter lasers),” IEEE Photon. Technol. Lett. 3(7), 588–590 (1991). [CrossRef]
8. H. Han, D. V. Forbes, and J. J. Coleman, “InGaAs-AlGaAs-GaAs strained-layer quantum-well heterostructure square ring lasers,” IEEE J. Quantum Electron. 31(11), 1994–1997 (1995). [CrossRef]
9. C. Ji, M. H. Leary, and J. M. Ballantyne, “Long-wavelength triangular ring laser,” IEEE Photon. Technol. Lett. 9(11), 1469–1471 (1997). [CrossRef]
10. K. K. Lee, D. R. Lim, L. C. Kimerling, J. Shin, and F. Cerrina, “Fabrication of ultralow-loss Si/SiO(2) waveguides by roughness reduction,” Opt. Lett. 26(23), 1888–1890 (2001). [CrossRef]
12. T. F. Krauss, R. M. De La Rue, and P. J. R. Laybourn, “Impact of output coupler configuration on operating characteristics of semiconductor ring lasers,” J. Lightwave Technol. 13(7), 1500–1507 (1995). [CrossRef]
13. R. Roijen, E. C. M. Pennings, M. J. N. van Stalen, T. van Dongen, B. H. Verbeek, and J. M. M. van der Heijden, “Compact InP-based ring lasers employing multi-mode interference couplers and combiners,” Appl. Phys. Lett. 64(14), 1753–1755 (1994). [CrossRef]
14. C. P. Chao, S. Y. Hu, K.-K. Law, B. Young, J. L. Merz, and A. C. Gossard, “Low-threshold InGaAWGaAs strained layer single quantum well lasers with simple ridge waveguide structure,” J. Appl. Phys. 69(11), 7892–7894 (1991). [CrossRef]
15. T. F. Krauss, R. M. De La Rue, P. J. R. Laybourn, B. Vogele, and C. R. Stanley, “Efficient semiconductor ring lasers made by a simple self aligned fabrication process,” IEEE J. Sel. Top. Quantum Electron. 1(2), 757–761 (1995). [CrossRef]
16. H. V. Demir, J.-F. Zheng, V. A. Sabnis, O. Fidaner, J. Hanberg, J. S. Harris, and D. A. B. Miller, “Self-aligning planarization and passivation for integration applications in III-V semiconductor devices,” IEEE Trans. Semicond. Manuf. 18(1), 182–189 (2005). [CrossRef]
17. M. Haverlag, D. Vender, and G. S. Oehrlein, “Ellipsometric study of silicon surface damage in electron cyclotron resonance plasma etching using CF4 and SF6,” Appl. Phys. Lett. 61(24), 2875–2877 (1992). [CrossRef]
19. M. Sorel, P. J. R. Laybourn, A. Scirè, S. Balle, G. Giuliani, R. Miglierina, and S. Donati, “Alternate oscillations in semiconductor ring lasers,” Opt. Lett. 27(22), 1992–1994 (2002). [CrossRef]