Open Access
Add to BibSonomyAbstract
We designed and fabricated a vertical-cavity surface-emitting laser (VCSEL) incorporating a polarization-independent high-index-contrast subwavelength grating (HCG) mirror on silicon-on-insulator (SOI) for a novel polarization-bistable device on a silicon substrate. The VCSEL consists of the HCG mirror, an active layer with InGaAsP quantum wells having optical gain around 1.55 μm, and an Al0.9Ga0.1As/Al0.16Ga0.84As DBR. We used direct wafer bonding for the bonding between the active layer and the AlGaAs DBR, and benzocyclobutene (BCB) bonding for the bonding between the active layer and the polarization-independent HCG mirror. The reflectivity of the HCG embedded with BCB was measured, resulting in a 200-nm-high reflectivity band with reflectivity higher than 99% and a small polarization dependence of ± 1%. We achieved lasing of the fabricated HCG-VCSEL at 1527 nm under an optical short pulse excitation with an average power of 50 mW (~0.2 mJ/cm2) at 240 K.
© 2013 Optical Society of America
1. Introduction
High-index-contrast subwavelength gratings (HCGs) with a thickness of only a few hundred nanometers exhibit a broadband and high reflectivity comparable or even superior to that of distributed Bragg reflectors (DBRs), which are based on in-plane guided mode resonance [1]. Recently, lasing in vertical-cavity surface-emitting lasers (VCSELs) incorporating HCGs as mirrors instead of DBRs has been demonstrated [2, 3]. An optical waveguide-coupled HCG-VCSEL, in which the VCSEL output is coupled to an in-plane waveguide, has also been proposed [4]. These HCGs having one-dimensional periodicity show a broadband and high reflectivity only for a linear polarization that is orthogonal or parallel to the grating [5]. To apply HCGs to polarization-bistable VCSELs [6], which exhibit bistability between two orthogonal linear polarization modes, the reflectivity has to be the same for these two modes. We thus previously proposed a polarization-independent HCG mirror that consists of a single-layer cross-stripe structure having 90º rotational symmetry [7]. We then proposed a waveguide-coupled HCG-VCSEL incorporating the polarization-independent HCG coupled with two in-plane waveguides in orthogonal directions [8], as shown in Fig. 1. The VCSEL has a square mesa structure to define the transverse mode and polarization direction of the cavity. The in-plane waveguides are aligned to the sides of the square mesa parallel to the grating direction. It is assumed that the VCSEL oscillates with one of the two orthogonal linear polarizations. Under these conditions, we expected that the output waveguide can be switched by changing the lasing polarization of the VCSEL and that the lasing polarization of the VCSEL can be controlled by the optical input from the adjacent VCSEL propagating in the waveguide. If these functions are possible, polarization-bistable VCSEL arrays used for optical buffer memories [6] will be drastically miniaturized because each VCSEL can be connected by waveguides instead of free-space optics. Recently, we numerically investigated a waveguide-coupled HCG-VCSEL and found that the output waveguide can be switched with an extinction ratio of 11.9 by changing the lasing polarization [9].
Fig. 1 Conceptual diagram of waveguide-coupled HCG-VCSEL incorporating polarization-independent HCG coupled with two in-plane waveguides in orthogonal directions
For the present study, we designed and fabricated a 1.55-μm HCG-VCSEL with a polarization-independent HCG, but with no output waveguide, and achieved optically pumped lasing as the first step in the fabrication of a waveguide-switchable HCG-VCSEL.
2. Device structure
Figure 2 shows the structure of the fabricated HCG-VCSEL. The VCSEL cavity consists of a 35-pair Al0.9Ga0.1As/Al0.16Ga0.84As DBR with compositional transition layers grown on a GaAs substrate, an active layer, a 2-pair InP/In0.66Ga0.34As0.74P0.26 DBR, a benzocyclobutene (BCB) bonding layer, and a polarization-independent HCG on a silicon-on-insulator (SOI) substrate. We used the AlGaAs DBR since the number of DBR pairs can be significantly reduced compared to InGaAsP DBRs. This DBR has a high reflectivity band ranging from 1480 to 1580 nm. The active layer contains nine 5.4-nm-thick In0.76Ga0.24As0.82P0.18 quantum wells (QWs) separated by 8.1-nm-thick In0.78Ga0.22As0.48P0.52 barriers and spacers. The photoluminescence (PL) spectrum of QWs ranges from 1300 to 1600 nm having the peak at 1550 nm at 300 K. The 2-pair InP/InGaAsP DBR was included to protect the active layer from the damage which may be caused by wet etching in our fabrication process described in section 4. We used direct wafer bonding [10, 11] for the bonding between the active layer and AlGaAs DBR, and BCB bonding [12–14] for the bonding between the active layer and the HCG. The BCB bonding was used for fabricating a light-emitting diode and an edge-emitting laser [13]. In the BCB bonding process, no precise surface treatment is required, and the curing temperature is relatively low (~250ºC) compared to polyimide (~300ºC) [14]. High controllability of BCB thickness is required in our device since it determines the cavity length of the VCSEL. The BCB thickness is uniform and reproducible because it can be controlled by the dilution ratio, rotation speed in spin-coating, and bonding pressure [12–14].
3. Polarization-independent HCG embedded with BCB
The gaps of the HCG are embedded with BCB in our structure since BCB is spin-coated on the HCG for the bonding. Although we previously obtained the optimum structure of the polarization-independent HCG filled with air (Air-BCB) [7], the optimum structure embedded with BCB (BCB-HCG) changes because of the change in the refractive index difference in the gratings. Thus, we first designed a polarization-independent BCB-HCG embedded with BCB using a three-dimensional finite-difference time-domain (FDTD) method (FullWAVE, Synopsys Inc.). Figure 3(a) shows the structure of the HCG used for our simulation. The SiO2 thickness of the SOI substrate and refractive indexes of Si and SiO2 are 3000 nm, 3.48, and 1.48, respectively. We investigated the optimum HCG structure exhibiting the widest bandwidth and highest reflectivity by changing the Si layer thickness TSi, grating period Λ, grating width W, and refractive index of the embedding material nem (1 for air and 1.55 for BCB [12] at a wavelength of 1.55 μm). We used the perfect matched layers for the z direction and the periodic boundary condition for the x and y directions to calculate the unit cell exploiting the periodicity of the HCG, which are denoted by the white and red lines in Fig. 3(a), respectively.
Fig. 3 (a) Structure for simulation of polarization-independent HCG. (b) Optimized reflectivity and transmittance spectra. Red and black lines show Air- and BCB-HCG, respectively. Solid and dashed lines show reflectivity and transmittance spectra, respectively.
Figure 3(b) shows the optimized reflectivity and transmittance spectra with TSi = 415 nm, Λ = 906 nm, and W = 278 nm for the Air-HCG and TSi = 466 nm, Λ = 853 nm, and W = 178 nm for the BCB-HCG. We note that the HCGs are theoretically polarization independent, which was numerically confirmed by our simulations with 0º-, 90º- and 45º-polarized incident light resulting in the same reflectivity spectrum as Fig. 3(b). Furthermore, to remove an ambiguity for the definition of 45º linear polarization in the FDTD software used, we performed another calculation with 0º-polarized incident light and a 45º-rotated HCG structure, and obtained the same reflectivity spectrum. The high reflectivity band with reflectivity higher than 99% (99.7%) is 95 nm (67 nm) and 260 nm (170 nm) for the Air- and BCB-HCG, respectively. The bandwidth of the BCB-HCG is about 2.5 times wider than the Air-HCG. Generally, the high reflectivity in HCGs is obtained at a wavelength under the guided-mode resonance, where the incident wave couples to a leaky waveguide mode. When the diffracted incident wave excites several leaky modes and their resonant wavelengths are close to each other, the reflectivity band broadens [15]. The transmittance spectra in Fig. 3(b) indicate sharp dips at the resonant wavelengths. The BCB-HCG has three dips separated by ~100 nm while the Air-BCB has two dips separated by 30 nm, resulting in a wider reflectivity band in the BCB-HCG. This result indicates that the BCB-HCG is suitable for the HCG-VCSEL.
We fabricated a polarization-independent BCB-HCG and measured the reflectivity spectra. Typical scanning electron microscope (SEM) images of the HCG are shown in Fig. 4. Figure 4(a) shows the angled view of the HCG after reactive ion etching (RIE), and Fig. 4(b) shows the cleaved facet of the HCG after BCB curing. As shown in Fig. 4(b), the gaps between the gratings were completely embedded with BCB. We extracted the sizes of the fabricated HCG from Fig. 4(b), which are schematically represented in Fig. 4(c). Note that the BCB thickness measured from the top of the HCG to the BCB surface was 550 nm.
Fig. 4 SEM images of fabricated polarization-independent HCG. (a) Angled view of HCG, (b) cleaved facet of HCG after BCB was coated and cured, and (c) schematic view of (b).
We measured reflectivity spectra for different linear polarizations to demonstrate polarization independence. Light from a broadband light source was linearly polarized using a polarizer and focused on the HCG with a spot diameter of about 180 μm and a convergent angle of about 0.6º. The reflected light was measured using an InGaAs array spectrometer with a resolution of 5 nm at every 15º rotation of the HCG in relation to the direction of linear polarization. The reflectivity was defined as the ratio of the reflected power from the HCG to that from a dielectric multilayer mirror having a broadband (1300–1800 nm) and high (>99.5%) reflectivity. Figure 5(a) shows the measured reflectivity spectra for different linear polarizations and the simulated spectrum using the sizes in Fig. 4(c), which are in good agreement. Ripples were observed in the measured spectra probably due to the interference in the measurement system. The bandwidth with a spectrally averaged reflectivity between 99 and 100% was 200 nm (1490–1690 nm) for all polarization directions. The HCG exhibited a high reflectivity band around 1.55 μm, which shifted to a longer wavelength compared to that in Fig. 3(b) due to the deviation from the designed sizes (TSi = 466 nm, Λ = 853 nm and W = 178 nm) and the tapered shape as shown in Fig. 4(b), 4(c). Figure 5(b) shows the polarization dependencies of reflectivity at 1500, 1550, and 1600 nm, which indicate variations of less than ± 1% for each wavelength. We may attribute the variations to the change in the optical alignment and interference in the measurement system.
Fig. 5 (a) Measured reflectivity spectra of fabricated BCB-HCG for different incident polarizations, (b) polarization dependence of reflectivity at 1500, 1550, and 1600 nm.
4. Fabrication of HCG-VCSEL
Figure 6 shows the fabrication process flow of the HCG-VCSEL. An MOCVD-grown InP wafer containing a 2-pair InP/InGaAsP DBR and nine InGaAsP QWs was bonded with an MOCVD-grown GaAs wafer containing a 35-pair Al0.9Ga0.1As/Al0.16Ga0.84As DBR by direct wafer bonding. Before the bonding, 150-μm square mesas with a height of 1.2-μm were fabricated on the InP wafer by photolithography and wet-etching with a solution of H2SO4:H2O2:H2O (3:1:1) for InGaAsP and HCl:H3PO4 (5:1) for InP. After the photo-resist mask was removed, ultraviolet-ozone cleaning was done to remove organic contamination for 10 minutes, followed by cleaning with an alkaline semiconductor cleaning solution for 5 minutes for both wafers. Then, the wafers were rinsed using deionized water and treated with a buffered HF solution for 5 minutes (Fig. 6(a)).
Fig. 6 Fabrication process flow of HCG-VCSEL. (a) Fabrication of 150-μm square mesas with height of 1.2-μm on InP wafer, (b) direct wafer bonding for bonding between active layer and Al0.9Ga0.1As/Al0.16Ga0.84As DBR, and complete removal of InP substrate, (c) HCG fabrication and spin-coating of diluted BCB onto HCG, and (d) BCB bonding of HCG and InP/InGaAsP DBR.
The active layer with the InP substrate was placed on the AlGaAs DBR wafer in an ammonium solution after treatment with the same ammonium solution for 5 minutes. Right after removing from the solution, a weight of about 100 kg/cm2 was loaded on the wafers for 5 minutes. The wafers were then transferred to an annealing furnace and heated at 600ºC for 30 minutes in an ambient gas of H2 diluted with N2 with a bonding pressure of about 260 kg/cm2. The InP substrate of the bonded wafer was mechanically polished down to a thickness of about 150 μm then InP was completely etched out with a solution of HCl:H3PO4 (5:1) until the InGaAs etch-stop layer was exposed. The etch-stop layer was also etched out with a solution of H3PO4:H2O2:H2O (6:1:100) (Fig. 6(b)).
We then proceeded to fabricate an HCG on SOI. First, we adjusted the Si layer thickness of SOI to 466 nm, which is the optimum thickness of the BCB-HCG. For this, the Si layer was slightly etched by RIE with a gas mixture of CHF3 (80 sccm) and SF6 (5 sccm). Electron beam lithography was conducted on the electron beam resist spin-coated on the thinned SOI to define the HCG pattern, followed by the same RIE to fabricate the HCG. After the adhesion promoter (AP3000, Dow Chemicals) was spin-coated onto the HCG at 3000 rpm for 20 seconds, BCB (CYCLOTENE 3022-35, Dow Chemicals) diluted with mesitylene was spin-coated onto it at 5000 rpm for 40 seconds (Fig. 6(c)), which led to a layer thickness of about 720 nm. The BCB thickness can be controlled from 200 nm to 1 μm by changing the dilution ratio [12]. The SOI substrate was then baked on a hotplate at 150ºC for 5 minutes.
Finally, the HCG coated with BCB was attached to the InP/InGaAsP DBR having an active layer and AlGaAs DBR behind, and cured in N2 at 290ºC with a bonding pressure of about 140 kg/cm2 for 1 hour (Fig. 6(d)). We estimated that the BCB thickness between the top of the HCG and top of the 2-pair InP/InGaAsP DBR would become about 650 nm after our BCB bonding process. We extracted the fabricated HCG sizes from the SEM images as TSi = 460 nm, Λ = 870 nm, and W = 170 nm, which were close to our design (TSi = 466 nm, Λ = 853 nm, and W = 178 nm).
5. Measurements
We used a mode-locked Ti:sapphire laser as a pump source with a wavelength of 920 nm, a pulse width of 80 fs, and a repetition rate of 80 MHz. The fabricated HCG-VCSEL was pumped with a spot size of 20 μm from the DBR side (Fig. 2). The pump light is transparent for the Al0.9Ga0.1As/Al0.16Ga0.84As DBR and GaAs substrate and absorbed in the QWs, barriers, and InGaAsP layers of the DBR. The laser output was measured using an InGaAs array spectrometer with a resolution of 5 nm. Figure 7(a) shows VCSEL output intensity as a function of average pump power at 240 K. The threshold pump power was about 50 mW (0.2 mJ/cm2). Figure 7(b) (i) and (ii) show the spectra for the VCSEL output with average pump powers below and above the lasing threshold at 240 K, respectively. The broad component around 1400 nm in Fig. 7(b) (i) is the PL light transmitting through the top DBR which has a low reflectivity around 1400 nm. The lasing wavelength was found to be about 1527 nm. From our simulation, the resonant wavelength for lasing is 1550 nm when the BCB thickness is 650 nm, and a change in the BCB thickness of ± 100 nm results in a variation in the resonance wavelength of about ± 25 nm. Therefore, the BCB thickness was estimated to about 550 nm since the lasing wavelength was about 1527 nm.
Fig. 7 (a) VCSEL output intensity as function of average pump power at 240 K and (b) (i) and (ii) show spectra of VCSEL output with average pump powers below and above lasing threshold at 240 K, respectively.
Figure 8(a) shows the lasing threshold average pump power and lasing wavelength plotted as a function of the measured temperatures. The HCG-VCSEL showed lasing at temperatures between 180 and 240 K, while it did not show lasing at 160 and 260 K with the average pump power of 400 mW. The lasing wavelength was almost constant within the 5-nm resolution and about 1527 nm for each measured temperature. From the PL intensity at 1527 nm, shown in Fig. 8(b), the maximum intensity was obtained at 230 K, which was a slightly higher temperature than that for the lowest threshold. Note that the PL spectrum of the QWs on an InP substrate was measured before the fabrication process. This is consistent with the fact that the wavelength of the maximum gain of a semiconductor laser exists at a longer wavelength than that of the PL due to absorption in the semiconductor. Heating of the QWs due to the strong optical excitation might also result in a part of the difference between the lowest threshold temperature and the maximum PL temperature.
Fig. 8 (a) Lasing threshold average pump power and lasing wavelength plotted as function of measured temperatures and (b) PL intensity at 1527 nm as function of temperature.
6. Conclusion
We designed and fabricated an HCG-VCSEL with a polarization-independent HCG as the first step in fabricating a waveguide-switchable HCG-VCSEL. The gaps in the HCG were embedded with BCB for our structure because of the BCB bonding process. The BCB-HCG was numerically optimized and exhibited a 260-nm-high reflectivity band with reflectivity higher than 99%. The reflectivity spectra for different linear polarizations were measured, resulting in a 200-nm-high reflectivity band with reflectivity higher than 99% and a small polarization dependence of ± 1%. The fabricated HCG-VCSEL with the polarization-independent HCG exhibited lasing at 1527 nm under an optical short pulse excitation of 50 mW (0.2 mJ/cm2) at 240 K. We achieved lasing at room temperature by optimizing the HCG structure parameters after this paper was submitted. The results will be presented elsewhere. The next step for the proposed waveguide-switchable HCG-VCSEL will be to demonstrate electrically pumped lasing.
Acknowledgment
This work was supported in part by JSPS KAKENHI Grant Number 24226011. The authors thank Dr. Yuuki Sato and Dr. Satoshi Hattori for their early work on the direct wafer bonding process.
References and links
1. C. F. R. Mateus, C. Y. Huang, L. Chen, C. J. Chang–Hasnain, and Y. Suzuki, “Broad–band mirror (1.12–1.62 μm) using a subwavelength grating,” IEEE Photon. Technol. Lett. 16(7), 1676–1678 (2004). [CrossRef]
2. M. C. Y. Huang, Y. Zhou, and C. J. Chang–Hasnain, “A surface–emitting laser incorporating a high–index–contrast subwavelength grating,” Nat. Photonics 1(2), 119–122 (2007). [CrossRef]
3. C. Sciancalepore, B. B. Bakir, X. Letartre, J. Harduin, N. Olivier, C. Seassal, J. M. Fedeli, and P. Viktorovitch, “CMOS–compatible ultra–compact 1.55–μm emitting VCSELs using double photonic crystal mirrors,” IEEE Photon. Technol. Lett. 24(6), 455–457 (2012). [CrossRef]
4. I.-S. Chung and J. Mørk, “Silicon–photonics light source realized by III–V/Si–grating–mirror laser,” Appl. Phys. Lett. 97(15), 151113 (2010). [CrossRef]
5. Y. Zhou, M. C. Y. Huang, and C. J. Chang-Hasnain, “Tunable VCSEL with ultra-thin high contrast grating for high-speed tuning,” Opt. Express 16(18), 14221–14226 (2008). [CrossRef] [PubMed]
6. H. Kawaguchi, T. Mori, Y. Sato, and Y. Yamayoshi, “Optical buffer memory using polarization–bistable vertical–cavity surface–emitting lasers,” Jpn. J. Appl. Phys. Lett. 45(34), L894–L897 (2006). [CrossRef]
7. K. Ikeda, K. Takeuchi, K. Takayose, I.-S. Chung, J. Mørk, and H. Kawaguchi, “Polarization-independent high-index contrast grating and its fabrication tolerances,” Appl. Opt. 52(5), 1049–1053 (2013). [CrossRef] [PubMed]
8. H. Kawaguchi, T. Katayama, K. Ikeda, and Japan Patent Application 2012–183539, (2012).
9. Y. Tsunemi, K. Ikeda, and H. Kawaguchi, “Lasing–polarization–dependent output from orthogonal waveguides in high–index–contrast subwavelength grating vertical–cavity surface–emitting laser,” Appl. Phys. Express 6(9), 092106 (2013). [CrossRef]
10. H. Wada, Y. Ogawa, and T. Kamijoh, “Electrical characteristics of directly–bonded GaAs and InP,” Appl. Phys. Lett. 62(7), 738–740 (1993). [CrossRef]
11. D. I. Babić, J. J. Dudley, K. Streubel, R. P. Mirin, J. E. Bowers, and E. L. Hu, “Double–fused 1.52 μm vertical–cavity lasers,” Appl. Phys. Lett. 66(9), 1030–1032 (1995).
12. G. Roelkens, D. Van Thourhout, and R. Beats, “Ultra–thin benzocyclobutene (BCB) bonding of III–V dies onto an SOI substrate,” Electron. Lett. 41(9), 561–562 (2005). [CrossRef]
13. I. Christiaeans, G. Roelkens, K. D. Mesel, D. V. Thourhout, and R. Beats, “Thin–film devices fabricated with benzocyclobutene adhesive wafer bonding,” J. Lightwave Technol. 23(2), 517–523 (2005). [CrossRef]
14. F. Niklaus, P. Enoksson, E. Kälvesten, and G. Stemme, “Low–temperature full wafer adhesive bonding,” J. Micromech. Microeng. 11(2), 100–107 (2001). [CrossRef]
15. R. Magnusson and M. Shokooh-Saremi, “Physical basis for wideband resonant reflectors,” Opt. Express 16(5), 3456–3462 (2008). [CrossRef] [PubMed]
