Open Access
Add to BibSonomy
Abstract
This paper presents our research on quantum well intermixing (QWI) of InP-based AlGaInAs/AlGaInAs multi-quantum wells using impurity-free vacancy-disordering (IFVD) and the QWI mask proximity effect and its application in the design and fabrication of a teardrop laser. Using a Si3N4 film deposited by plasma-enhanced chemical vapor deposition (PECVD) as a QWI promoter mask and annealing under 700°C for 2 minutes, a 70 nm wavelength blue shift of a FP laser is achieved using InP-based AlGaInAs quantum well laser material. It is found that a 5 µm separation is needed between the QWI mask edges and the non-QWI area during the QWI process. Based on the QWI technique and proximity effect, the designed and fabricated teardrop laser demonstrated continuous wave (CW) lasing above 40 mA and single frequency operation with a side mode suppression ratio of 32.6 dB at 77.3 mA.
© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Photonic integration is becoming increasingly important in lowering the cost and footprint of data center photonics [1]. Quantum well intermixing (QWI) is a post-growth technique that allows the energy band gap of a quantum well (QW) to be modified without any regrowth and thus can be used in the fabrication of photonic integration circuits (PICs) to reduce or eliminate epitaxial regrowth [2]. During the QWI process, the diffusion of different atoms into the QW changes the material composition of the well. The bandgap of the QW typically increases (thus blue-shifts) as a result of the QWI. Several QWI techniques including impurity induced disordering (IID) [3], photo-absorption-induced disordering (PAID) [4] and impurity-free vacancy-disordering (IFVD) have been reported [5–7]. Among them IFVD is the simplest method to create different amounts of inter-diffusion in different regions of a sample by using dielectric coatings like Si3N4 as a promoter, which is compatible with common semiconductor processing. IFVD results in little or no damage to the surface of the epitaxial wafer and requires no extra treatment to the epitaxial wafer, such as a sacrificial layer. Si3N4 has been reported as a QWI promoter and SiO2 as a prohibitor in InP based QWs material systems, which attributes to more group III and V vacancies created in the Si3N4 and the semiconductor cap layer of the InP interface than that of the SiO2-InP interface during QWI process [8]. However, the degree of QWI is dependent on the film quality and/or the film growth process, as well as the semiconductor cap layer, the quantum wells, and the distance between cap layer and the QWs [9,10]. Therefore, for specific epitaxial wafers, the QWI effect needs to be characterized before being applied in the fabrication of the components and PICs.
The main original contribution of this work is our study on the effect of the QWI mask’s proximity in the InP based AlGaInAs-MQW material, i.e. the required margin between the QWI and non-QWI region. This provides additional flexibility in designing PICs requiring multiple bandgaps, such as PICs made up of lasers, EAMs and passive regions. Here, we present our research on the QWI of an InP-based AlGaInAs-MQW material and demonstrate the effect of the QWI's mask proximity on the laser wavelength shift by making and measuring Fabry-Perot (FP) ridge waveguide lasers using our QWI technique.
Based on the characterized QWI effect and the proximity results, we optimized the design of the single-facet teardrop laser we reported earlier, which demonstrates promising tunable single frequency and low linewidth performance [11–13], to make it less dependent on the injected current through multiple electrodes and to lower power consumption through integrating a QWI area that is used as a passive waveguide. The QWI teardrop laser will not only inherit the advantages of the single-facet teardrop laser but also has a much lower power consumption due to the passive region.
2. Design and experiment of QWI by IFVD
Firstly, the quantum well intermixing with the IFVD method is investigated using InP-based AlGaInAs/AlGaInAs multiple quantum well epitaxial material as show in Fig. 1 which was grown by Metal-organic Chemical Vapor Deposition (MOCVD) including 5 pairs of compressive strained quantum wells (QW, +1.2% strain, 6 nm thick quantum well and 10 nm thick barrier, λPL=1.55 µm, by photoluminescence) on an n-doped InP substrate. In the experiment, the effect of both plasma-enhanced chemical vapor deposition (PECVD) and sputtered Si3N4 and SiO2 films were investigated. All four variants of single layer Si3N4 and SiO2 films were deposited on the top surface of four pieces of as-grown epitaxial wafers, where 50 nm thick Si3N4 (PECVD and sputtered) and 200 nm thick SiO2 (PECVD and sputtered) films were used. These were compared with a non-capped as-grown wafer for the QWI process. Then the samples with different dielectric coatings and the as-grown wafer were annealed under the same conditions by rapid thermal annealing (RTA) at 675°C, 700°C and 725°C, and the photoluminescence (PL) spectra of the samples after annealing were measured at room temperature (20 °C) with the top p-doped GaInAs, GaInAsP and InP layers removed and plotted in Fig. 2(a) and Fig. 2(b). From Fig. 2(a), it can be seen that the PL wavelength of the as-grown non-QWI material is at 1530 nm while the PL wavelength of the sample with Si3N4 by PECVD blue shifts after annealing and further as the temperature increases. It shifts to 1521 nm at 675°C, to 1457 nm at 700°C and to 1410 nm annealing at 725°C. From Fig. 2(b), the PL wavelength of the samples with a surface coating of SiO2 by PECVD is shown to blue shift by 5 to 15 nm after annealing at 675°C, 700°C and 725°C. In both cases, a noisier and degraded PL signal has been observed on the samples after QWI process compared to that of non-QWI as-grown wafer due to the defects generated in the material. The samples with SiO2 and Si3N4 coatings by sputtering showed non-repeatable results with some samples cracking after annealing, thus they were not suitable for the QWI process. Samples without any dielectric coating had surface damage as a result of the annealing process. Based on the experiment, Si3N4 by PECVD results in a PL wavelength shift of more than 100 nm during the annealing at 725°C, while a SiO2 film using PECVD prevents the wavelength shift during the annealing and protects the surface. Thus, the Si3N4 by PECVD works as the promotor while the SiO2 by PECVD acts as inhibitor in our QWI technique.
Fig. 2. a) PL spectra of the samples capped with Si3N4 (left) by PECVD and Figure. b) PL spectra of the samples capped with SiO2 films by PECVD (right) annealed under different temperatures. The PL was measured at room temperature (20 °C).
3. Study on the proximity effect on the QWI process
3.1 Design
Based on the IFVD QWI technique described in section 2, FP lasers with the same structure but with different Si3N4 mask boundaries were designed, as shown in Fig. 3(a) where 12 FP lasers in two arrays with metal coated etched facets back to back and a cleaved facet on the other end were designed. One of them has no Si3N4 mask marked as “NON” while the other 11 lasers have a 40 µm wide Si3N4 mask with different distances between the centre of the ridge waveguide of the FP laser to the edge of the mask used to define the proximity of the QWI, from: -20 µm (negative numbers mean that the ridge is fully covered with the mask as marked with “ON”) to 21.25 µm (as marked with “20”, plus half of the ridge width). All the lasers were 500 µm long with 2.5 µm wide ridges.
Fig. 3. a) 12 lasers with different distances (the numbers on the chip design represent the distance from the edge of the mask to the edge of the ridge) on the QWI boundary (left) and Figure. b) Microscopic picture of the fabricated FP lasers array (right).
3.2 Fabrication
The same epitaxial material as used for the QWI experiments in section 2 was used in fabricating the FP lasers. The fabrication process is a typical Fabry-Perot semiconductor laser process using conventional photolithography and etching techniques, except that QWI areas were first defined by using PECVD SiO2 as protector and PECVD Si3N4 as promoter. First, 50 nm thick Si3N4 was deposited on top of the wafer by PECVD, followed by standard photolithography and a wet etch, which was performed to define the QWI zone (promoted zone). Then 200 nm thick SiO2 was deposited on top of the wafer by PECVD (the III-V- SiO2 contacted surface defined the inhibited zone). The wafer was then annealed in a nitrogen ambient at 700°C for 2 minutes and then the SiO2 and Si3N4 masks were removed from the wafer. Next, 650 nm thick SiO2 was deposited on top of the wafer by PECVD, followed by standard photolithography, which was performed to define the waveguide. Inductively coupled plasma (ICP) dry etching of SiO2 with CF4/CHF3 was then performed to transfer the pattern into the SiO2 mask. 1.92 µm deep ridge waveguides were then formed using a room temperature ICP dry etch with Cl2/CH4/H2. After wafer passivation using 300 nm of SiO2, a window was formed in the SiO2 by ICP dry etching, followed by a p metal (Ti/Au: 20/300 nm) deposition. Finally, the sample was thinned to 120 µm, and an n metal (Au/Ge/Au/Ni/Au: 14/14/14/11/250 nm) stack was deposited on the back side of the wafer, which was then annealed at 400°C for 5 minutes in a nitrogen oven.
3.3 Characterization
The fabricated FP lasers were cleaved into bars as shown in Fig. 3(b) with two arrays of 500 µm long single facet (the other is a metal-coated etched facet) lasers back to back. The light - current - voltage (LIV) characteristics were tested at room temperature (20°C) with electric probing under continuous wave (CW) current injection and using an integrating sphere to collect the light output from the cleaved facet.
In the test, all the FP lasers after the QWI process lase continuously with threshold currents from 23 to 39 mA. This proves that the fabrication of the lasers was successful. The non-QWI laser (the laser area is fully covered by SiO2 during the QWI process) lases continuously with a threshold current of 23 mA and the output power linearly increases with current up to around 31.6 mW at a 300 mA bias. The full-QWI FP laser (with the ridge waveguide in the center of the Si3N4 mask during the QWI process) lases CW at a 39 mA threshold current and the output power reaches 18.5 mW at 300 mA. The turn-on voltage of the full QWI laser increases to 1.1 V, compared to 0.9 V for the non-QWI FP laser, which indicates a bandgap increase of the material during the QWI process. The LIV results are shown in Fig. 4. The threshold current, output power and the turn-on voltage of the half-QWI FP laser (with the ridge half covered by the Si3N4 mask) are in between.
Fig. 4. Characteristics of current–voltage and output power–current at 20°C of non-QWI FP laser, ‘Half’-QWI FP laser and Full-QWI FP laser.
The lasing spectra of the FP lasers were then measured by coupling light into a lensed fiber and using an optical spectrum analyzer (OSA). Figure 5(a) shows the spectra at 45 mA current injection for the FP lasers, which were annealed at 700°C with different QWI mask conditions (with different distances between the ridge center and the edge of the Si3N4 mask). The peak lasing wavelength of the non-QWI laser is 1541 nm due to the SiO2 mask protection during the QWI, while the peak wavelength of the full-QWI laser blue-shifts to 1471 nm, which proves the success of the QWI process. However, the FP lasers with the ridge waveguide non-fully covered or several microns (0 to around 5 µm) away from the edge of the Si3N4 mask also show a blue-shift of the lasing peak wavelength, but show less blue shift as the distance increases. When the distance is more than 5 µm, the QWI effect is no longer observed to affect the lasing performance. This can be clearly seen from Fig. 5(b), which plots the peak lasing wavelengths of the FP lasers annealed at 700°C and 725°C with the different distances between the ridge waveguide center and the edge of the QWI mask.
Fig. 5. a) The lasing spectra of the FP laser annealed at 700°C with different distance between the ridge centre to the edge of the QWI mask (left). Figure. b) The lasing peak wavelength vs the distance between the ridge centre to the edge of the QWI area mask (all the FP lasers are tested at 45 mA and the same other conditions). Here, “-20” and “0” represent the ridge is in the centre of the mask and half covered by the mask (right).
3.4 Summary
We have successfully demonstrated a QWI technique using the IFVD method with Si3N4 by PECVD as promoter and SiO2 by PECVD as inhibitor and protector. A 70 nm blue shift was obtained using InP-based AlGaInAs/AlGaInAs multiple quantum wells, further proved by Fabry-Perot ridge waveguide lasers fabricated with this technique. The QWI mask proximity effect was investigated, which shows the QWI mask edge needs to be more than 5 µm away from the area where the intermixing is not desired. This QWI technique and the mask proximity effect result will be beneficial for the monolithic integration of photonic devices using the AlGaInAs quantum well material.
4. Study on a teardrop laser using the QWI technique
4.1 Design
Previously we have reported a single-facet 1 × 2 MMI teardrop laser which demonstrated excellent tunable single frequency and narrow linewidth emission [11–13]. However, in the previous design, both the MMI and the ring needed to be biased with current injection to be made transparent, which consumes power thus leads to lower overall power efficiency.
Using the IFVD QWI technique described in section 3, teardrop lasers with different QWI zones (blue shadow) using the Si3N4 mask were designed as shown in Fig. 6. From top to bottom, the first teardrop laser is a reference, as all the sections are non-QWI. The second teardrop laser has a QWI zone for the MMI, while the gain waveguide and ring are non-QWI. The third teardrop laser has a QWI zone for the ring, while the gain and MMI sections are non-QWI. The fourth teardrop laser has a QWI zone covering both the ring and the MMI, while the gain waveguide is non-QWI. 7°C angled slots etched down to the etch stop layer above the QWs were designed between the ring and the MMI sections as well as between the gain and the MMI sections to achieve electrical isolation while minimizing the optical reflection at the slots. Figure 6 shows the designed angled slots which are $1\mathrm{\mu }\textrm{m}$ wide. In the bottom design, the slot and $5\mathrm{\mu }\textrm{m}$ waveguide on each side were left as a non-QWI region for comparison and consistency between the designs. The proximity effect will ensure part of the region will be experience QWI to transparency, while in future laser designs, this part should be all included in the QWI area to minimize absorption. All the dimensions of the teardrop laser structures are kept the same as the reference laser.
4.2 Fabrication
The designed teardrop lasers were fabricated using the same material as used for the QWI experiment in section 3 and the reported single-frequency single-facet 1 × 2 MMI teardrop laser [11–13]. The processing is the same as our typical processing for a Fabry-Perot semiconductor laser with common photolithography and etching techniques, except a QWI processing in the beginning and a self-aligned technique is used to achieve the designed two etch depths [11–15].
QWI areas were first defined as previously described. After the removal of the SiO2 and Si3N4 masks, 600 nm thick SiO2 is deposited on top of the wafer by PECVD, followed by standard photolithography, which was performed to define the whole structure including MMI, waveguides, slots and the loop. ICP dry etching of the SiO2 with a CF4/CHF3 was then done to transfer the pattern into the SiO2 mask. After removing the photoresist, 450 nm Si3N4 was deposited on top of SiO2, followed by photolithography and selective ICP dry etching using SF6 to remove the Si3N4 over the SiO2 in the deep-etch region. Then room temperature ICP dry etching with Cl2/CH4/H2 was used to first etch around 2.25µm of the deep- etch region. The Si3N4 was then fully removed by selective dry etching and both the shallow and deep etched regions were dry etched another 1.85µm so that the deep-etch waveguides are 4.1µm deep in total (through the QWs) while the shallow-etch ridges are 1.85µm deep (above the QWs). Following a wafer passivation using 300 nm of SiO2, a window opening was made in the SiO2 by ICP dry etching which was followed by a p metal (Ti/Au: 20/300 nm) deposition. Finally, the samples were thinned to 100µm and an n metal (Au/Ge/Au/Ni/Au: 14/14/14/11/250 nm) stack was deposited on the back side of the wafer, followed by annealing at 400°C for 5 minutes in a nitrogen furnace. Figure 7(a) shows the scanning electron microscopic (SEM) pictures of the fabricated teardrop lasers with 75µm radius loops. Figure 7(b) shows the inset in the microscopic picture of the fabricated teardrop laser.
Fig. 7. a). SEM pictures of the fabricated teardrop lasers with 75 µm radius loops array (left) and Figure. b). Microscopic picture of the fabricated teardrop lasers array (right).
4.3 Characterization
The fabricated teardrop lasers were cleaved into single facet devices for characterization (the cleaved length of the gain waveguide was 400 µm). Firstly, the LIV characteristics of the bare die were tested at room temperature (20°C) with electric probing under CW current injection and using an integrating sphere to collect the light output from the cleaved facet and the results are shown in Fig. 8(a). The thermal dissipation between the laser die and the test system were non-ideal, resulting the rollover of the LI curves, a characteristic that was repeatable between different die. The teardrop laser with a QWI region covering both the MMI and ring lases CW with a bias current above 40 mA only applied to the gain section and the output power is around 12.6 mW at a bias of 108 mA. However, the reference non-QWI teardrop laser does not show lasing if only the gain section is biased between 60 and 160 mA. This is due to light absorption in the non-QWI MMI and loop reflection regions which are non-transparent to the light without biasing.
Fig. 8. a) Characteristics of current - voltage and output power - current at 20°C of non-QWI teardrop laser, QWI MMI & Ring of teardrop laser (left). Figure. b) The lasing spectra of the teardrop laser annealed at 700°C with difference QWI Zone and Comparing them at the same output power, the required current is different (right).
The lasing spectra of the teardrop lasers are shown in Fig. 8(b), which are measured using an OSA with light collected from a lensed-fiber coupled to the gain waveguide facet. It shows the lasing spectra with similar lasing peak power of the four teardrop lasers with different QWI regions for comparison. All the lasers can lase when current is injected into the non-QWI regions to a certain level. To obtain a similar level of fiber-coupled lasing peak power around -7dBm shown in Fig. 8(b), the reference non-QWI teardrop laser is biased with the current on the gain biased at 42 mA, MMI at 36 mA and ring at 33 mA respectively, the teardrop laser with a QWI zone only covering the MMI section is biased with current on the gain at 45 mA and the ring at 53 mA, the teardrop laser with a QWI zone covering only the ring section is biased with a current on the gain of 42 mA and the MMI of 46 mA, while the teardrop laser with a QWI Zone in the MMI and ring section only needs biasing on the gain section with 64 mA current.
From the comparison of the teardrop lasers using the QWI process in the different sections, it can be seen that the QWI process successfully makes the section transparent and therefore removes the need for electrical biasing thus reducing power consumption. The teardrop laser with both transparent MMI and ring sections through QWI works efficiently with only the gain section requiring current injection to achieve single frequency.
The lasing spectrum of the teardrop laser with QWI in both MMI and ring sections (annealed at 700°C) was measured with an OSA. Figure 9(a) plots the spectra under 77.3 mA current injection demonstrating that the laser operates in a single longitudinal mode at 1552.53 nm with a SMSR of 32.6 dB. The Fourier Transform [16] of the wavelength data was taken near threshold, to examine the resonant cavities present in this lasing spectrum and the result is shown in Fig. 9(b). The main peak is at a cavity length of 1043µm, which corresponds to a length from the cleaved facet to the centre of the ring resonator. This indicates that the main cavity of the laser is through the ring, and the other small peaks show the intra-cavities due to the light reflection at the interface between the waveguide and MMI. The coupled cavity effect of these extra reflections at the MMI leads to the single frequency output, which agrees with the design.
Fig. 9. a) The measured lasing spectrum of the QWI MMI & RING teardrop laser under 77.3 mA (left) and Figure. b) The Fourier analysis of the spectrum (right).
5. Conclusions
In summary, QWI with the IFVD method was successfully developed on InP based AlGaInAs MQWs and the proximity effect was studied. A teardrop laser was successfully designed and fabricated with the MMI and ring fully QWI to transparency and thus reduces the power consumption. The laser operated at a single frequency using a single bias on the gain region.
Funding
Science Foundation Ireland (12/RC/2276, 13/IA/1960).
Disclosures
The authors declare no conflicts of interest.
References
1. M. Smit, K. Williams, and J. van der Tol, “Past, present, and future of InP-based photonic integration,” APL Photonics 4(5), 050901 (2019). [CrossRef]
2. L. Hou and J. H. Marsh, “Photonic Integrated Circuits Based on Quantum well Intermixing Techniques,” Procedia Eng. 140, 107–114 (2016). [CrossRef]
3. W. D. Laidig, N. Holonyak Jr., M. D. Camras, K. Hess, J. J. Coleman, P. D. Dapkus, and J. Bardeen, “Disorder of an AlAs -GaAs superlattice by impurity diffusion,” Appl. Phys. Lett. 38(10), 776–778 (1981). [CrossRef]
4. A. McKee, C. J. McLean, G. Lullo, A. C. Bryce, R. M. Rue, and J. H. Marsh, “Monolithic integration in InGaAs-InGaAsP multiple quantumwell structures using laser intermixing,” IEEE J. Quantum Electron. 33(1), 45–55 (1997). [CrossRef]
5. J. H. Marsh, P. Cusumano, A. C. Bryce, B. S. Ooi, and S. G. Ayling, “GaAs/AlGaAs photonic integrated circuits fabricated using impurity-free vacancy disordering,” Proc. SPIE , 2401, 74–85 (1995). [CrossRef]
6. B. S. Ooi, K. McIlvaney, M. W. Street, A. S. Helmy, S. G. Ayling, A. C. Bryce, J. H. Marsh, and J. S. Roberts, “Selective quantum-well intermixing in GaAs-AlGaAs structures using impurity-free vacancy diffusion,” IEEE J. Quantum Electron. 33(10), 1784–1793 (1997). [CrossRef]
7. Y. Alahmadi and P. LiKamWa, “Effects of selective area intermixing on InAlGaAs multiple quantum well laser diode,” Semicond. Sci. Technol. 34(2), 025010 (2019). [CrossRef]
8. Y. An, H. Yang, T. Mei, Y. Wang, J. Teng, and C. Xu, “Cap Layer Influence on Impurity-Free Vacancy Disordering of InGaAs/InP Quantum Well Structure,” Chin. Phys. Lett. 27(1), 017302 (2010). [CrossRef]
9. W. J. Choi, H. T. Yi, J. I. Lee, and D. H. Woo, “Dependence of the Intermixing in InGaAs/InGaAsP Quantum Well on Capping Layers,” J. Korean Phys. Soc. 45(3), 773–776 (2004).
10. J. S. Yu, J. Song, Y. Lee, and H. Lim, “Effects of the thickness of dielectric capping layer and the distance of quantum wells from the sample surface on the intermixing of In0.2Ga0.8As/GaAs multiple quantum well structures by impurity-free vacancy disordering,” J. Korean Phys. Soc. 42, S458–S461 (2003).
11. H. Yang, M. Yang, Y. Zhao, L. Zhang, Z. Jia, J. Alexander, L. Zhao, D. C. Hall, and F. H. Peters, “Butterfly Packaged Ultra-Narrow Linewidth Single Frequency Teardrop Laser Diode,” IEEE Photonics Technol. Lett. 29(18), 1537–1539 (2017). [CrossRef]
12. H. Yang, M. Yang, Z. Jia, D. C. Hall, and F. H. Peters, “InP-based Single-frequency Single-facet 1 × 2 MMI Teardrop Laser Diodes,” 2017 Conference on Lasers and Electro-Optics Pacific Rim, (Optical Society of America), paper s1777 (2017).
13. H. Yang, M. Yang, P. E. Morrissey, D. Lu, B. Pan, L. Zhao, B. Corbett, and F. H. Peters, “Three-coherent-output narrow-linewidth and tunable single frequency 1 ( 2 multi-mode-interferometer laser diode,” Opt. Express 24(6), 5846–5854 (2016). [CrossRef]
14. J. K. Alexander, P. E. Morrissey, H. Yang, M. Yang, P. J. Marraccini, B. Corbett, and F. H. Peters, “Monolithically integrated low linewidth comb source using gain switched slotted Fabry–Perot lasers,” Opt. Express 24(8), 7960–7965 (2016). [CrossRef]
15. M. Dernaika, L. Caro, N. P. Kelly, J. K. Alexander, F. Dubois, P. E. Morrissey, and F. H. Peters, “Deeply Etched Inner-Cavity Pit Reflector,” IEEE Photonics J. 9(1), 1–8 (2017). [CrossRef]
16. P. Zhang, Y. Zhu, and W. Chen, “A Study on Fourier Transformation Demodulating Theory of the Gap of Optical Fiber Fabry-Perot Sensor,” Acta Photonica Sinica 33(12), 1449–1453 (2004).
