Open Access
Add to BibSonomy
Abstract
Lasing is reported for ridge-waveguide devices processed from a 40-stage InP-based quantum cascade laser structure grown on a 6-inch silicon substrate with a metamorphic buffer. The structure used in the proof-of-concept experiment had a typical design, including an Al0.78In0.22As/In0.73Ga0.27As strain-balanced composition, with high strain both in quantum wells and barriers relative to InP, and an all-InP waveguide with a total thickness of 8 µm. Devices of size 3 mm x 40 µm, with a high-reflection back facet coating, emitted at 4.35 µm and had a threshold current of approximately 2.2 A at 78 K. Lasing was observed up to 170 K. Compared to earlier demonstrated InP-based quantum cascade lasers monolithically integrated onto GaAs, the same laser structure integrated on silicon had a lower yield and reliability. Surface morphology analysis suggests that both can be significantly improved by reducing strain for the active region layers relative to InP bulk waveguide layers surrounding the laser core.
© 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
The monolithic integration (i.e., direct growth) of Quantum Cascade Lasers, or QCLs, onto silicon substrates [1] paves the way for the development of ultra-compact and potential low cost infrared platforms with QCL elements used for chip-level wireless communications, sensing, and illumination. Compared to heterogeneous integration of QCL structures onto silicon employing the wafer bonding approach [2,3], monolithic integration offers the advantage of scalability to large-diameter wafers. The authors in [1] reported the first III-V QCLs grown directly on Si substrates, using an InAs-based alloy system for long wavelength infrared (LWIR) around 11 µm. Our work reports the first InP-based QCLs grown directly on a Si substrate for mid-wavelengths (MWIR).
InP-based QCLs have demonstrated the highest performance essentially throughout the entire MWIR and LWIR spectral regions [4] with pulsed efficiency approaching 30% in the 5 µm to 6 µm range [5,6]. Monolithic integration of InP-based QCLs onto silicon without modifying the optimal QCL configuration is the main goal for this work. The integration is challenging due to the large lattice-mismatch between InP and Si: The resultant mechanical stress is released via formation of threading dislocations whose presence in the active region leads to a reduced laser performance.
Dislocation density can be dramatically reduced by employing metamorphic buffers (M-buffers). Using this approach, we have recently demonstrated the integration of an InP-based QCL structure onto a GaAs substrate [7], a material that has approximately half lattice mismatch with InP compared to that for Si. Specifically, room temperature operation and a good short-term reliability were demonstrated for ridge-waveguide devices processed from the InP-based QCL-on-GaAs material. In this current work, we report on our first attempt to extend the monolithic integration of InP-based QCLs to silicon substrates.
2. Device fabrication
This work employed the same active region and waveguide QCL designs reported in [7], which allowed for direct comparison between experimental data for the two structures. The active region design had Al0.78In0.22As barriers and In0.73Ga0.27As quantum wells (total thickness 173 Å and 262 Å, respectively), both highly strained relative to InP bulk waveguide layers surrounding the active region (−2.04% and 1.34%, respectively.) The employment of the high strain active region layers increases the band offset, which improves carrier confinement for MWIR QCLs and, as a consequence, increases material gain. However, while the active region was designed to have a net zero strain relative to InP, the strain of the individual layers plus any residual net strain in the active region can make the InP-based QCL integration onto a lattice-mismatched substrate more challenging.
The QCL epiwafer was produced by IQE. The 150 mm diameter Ge-on-Si substrate was created by chemical vapor deposition of a 500 nm thick Ge epilayer on (100) Si substrate with a 6° miscut towards the [111]. Then, the III-V M-buffer layers and QCL device layers were grown on the Ge-Si substrate using molecular beam epitaxy (MBE). Off-axis substrates were used to help suppress anti-phase domain formation originating from polar (III-V) semiconductor on non-polar (Si) semiconductor growths by promoting bi-atomic steps on the Ge/Si surface [8].
For direct integration of InP-based QCL device structures on Si, we chose a composite metamorphic buffer consisting of GaAs (approximately 4% lattice mismatch with Si) and graded InAlAs (compensating for the additional approximately 4% lattice mismatch between InP and GaAs) [9]. A schematic of a composite buffer is shown in Fig. 1. The QCL-on-GaAs in [7] used the exact same graded InAlAs M-buffer and QCL design. Thus, the added complexity in this current work on the Ge-Si substrate is the GaAs-on-Ge nucleation and the GaAs M-buffer growth to accommodate the first half of the mismatch strain. The GaAs M-buffer unavoidably will have some residual strain, so growing the graded M-buffer plus QCL layers on it is not exactly like growing directly on a GaAs substrate. While we have not performed TEM on the current QCL structure growths, our prior experience with growing InP-based structures using this type of M-Buffer on Ge and Ge-terminated engineering substrates has shown threading dislocation densities (TDD) < 5·107/cm2 [10]. The TDD for the starting Ge/Si template used in this work is in the mid 107/cm2 range so our best estimate for TDD in the InP layers is mid 108/cm2. A full epi-layer sequence grown on the silicon substrate is shown in Fig. 2.
Fig. 1 Schematic of composite M-buffer design that utilizes an inverse step grade for complete compensation of residual strain [8]. InAlAs composition at the end of the M buffer is lattice matched to InP.
3. Results
Surface morphology of the QCL on Si epiwafer was evaluated at various stages of the epi-stack growth using Nomarski optical contrast microscopy and atomic force microscopy (AFM), as shown in Fig. 3. A good surface morphology with a relatively low roughness was observed until QCL active region growth was carried out. The ultimate quality of the QCL-on-Si epiwafer is certainly degraded compared to the QCL-on-GaAs wafer of [7], which is not unexpected due to the 2 × higher mismatch strain. The highly strain-compensated QCL design also affects the surface morphology. Surface areas with increased defects and/or roughness hurt device performance and yield as described below. However, there are many clear areas across the wafer where good devices were fabricated.
Fig. 3 Nomarski (top) and 5 µm × 5 µm AFM images (bottom) showing cross-hatch surface morphology for (a) composite M-buffer (graded InAlAs + GaAs) on Ge-coated Si substrate, (b) subsequent growth of a 2 µm InP layer, and (c) region of full InP QCL device structure.
Upon epi-growth, the wafer was processed into ridge-waveguide devices with lateral current injection following the processing sequence described in [7]. The processed wafer was cleaved into 3 mm chips. Finally, the chips were inspected under a high resolution microscope.
Chips with good facet quality were mounted on submounts, wirebonded, high reflection (HR)-coated, soldered to a heat spreader, and placed in a cryostat for pulsed testing (350 ns; 2 kHz) across a broad temperature range. While lasing was observed for the very first chip tested at 78 K, overall yield of lasing devices was relatively low (approximately 10%). In addition, in contrast to the QCL-on-GaAs material, performance for the first two chips that lased showed signs of degradation after approximately thirty minutes of operation at 78 K. To minimize testing time and to reduce the additional stress for the device due to applied bias, power vs current characterization was done only up to current exceeding threshold by approximately 0.5 A.
Figure 4 shows that an HR-coated 3 mm x 40 µm device had a threshold current of 2.22 A at 78 K (the devices did not lase in continuous wave mode). Corresponding threshold current density of 1.85 kA/cm2 is comparable to that measured for QCL-on-GaAs chips in the same configuration. At the same time, slope efficiency was approximately three times lower. The relatively low quality of the QCL-on-Si material and the relatively low yield of processed devices does not allow for a rigorous analysis of the change in laser characteristics. Lasing for the tested chip was observed up to approximately 170 K. It is important to mention here that only a small portion of the overall 6-inch wafer (2 cm2) was processed into functional devices and a higher laser performance can likely be achieved by additional material screening.
Fig. 4 Pulsed optical power vs. current characteristics for a 3 mm x 40 µm device with an HR-coated back facet measured at in temperature range from 78 K to 170 K. Inset: spectrum measured at 2.40 A and 78 K.
The inset shows that emission spectrum was centered essentially at the same wavelength of 4.35 µm as that for the QCL-on-GaAs material. The latter result is especially important as it demonstrates that despite the complexity of state-of-the-art QCL structures, an emission wavelength for a specific active region design can be reproduced for epi-growth on substrates having a wide range of lattice-mismatch, provided that a proper M-buffer is used for threading dislocation density control.
The observed correlation between surface morphology and laser yield/reliability for the two structures suggests that both yield and reliability can be improved by lowering the strain of the active region layers relative to the bulk InP waveguide layers. As mentioned above, a high strain composition was used in both cases to achieve a higher material gain. However, a high MWIR QCL performance can still be achieved employing a much lower strain composition [11]. In addition, the AlInAs/InGaAs composition lattice-matched or nearly lattice-matched to InP is traditionally used in the LWIR QCL design [12]. Therefore, future work on improvement in QCL-on-Si performance should include a systematic study on finding an optimal combination of strain for the active region layers and composition of the M-buffer. Results of such a study will likely be dependent on QCL emission wavelength. We are also working on improving the quality of the starting Ge/Si template, with the goal of reducing TDD down to the low 106 /cm2 range.
4. Conclusion
In conclusion, lasing was demonstrated for an InP-based QCL structure grown on a Ge-coated silicon wafer via a III-V M-buffer. Measured emission wavelength was close to its design value. The presented data suggest that the observed low yield and reliability can be both improved by reducing strain of the active region layer relative to the bulk InP waveguide layers surrounding the active region.
Funding
Nanoscience Technology Center; University of Central Florida.
References
1. H. H. Nguyen-Van, A. N. Baranov, Z. Loghmari, L. Cerutti, J.-B. Rodriguez, J. Tournet, G. Narcy, G. Boissier, G. Patriarche, M. Bahriz, E. Tournié, and R. Teissier, “Quantum cascade lasers grown on silicon,” Sci. Rep. 8, 7206 (2018).
2. A. Spott, J. Peters, M. Davenport, E. Stanton, C. Merritt, W. Bewley, I. Vurgaftman, C. Kim, J. Meyer, J. Kirch, L. Mawst, D. Botez, and J. Bowers, “Quantum cascade laser on silicon,” Optica 3, 545–551 (2016).
3. A. Spott, J. Peters, M. Davenport, E. Stanton, C. Zhang, C. Merritt, W. Bewley, I. Vurgaftman, C. Kim, J. Meyer, J. Kirch, L. Mawst, D. Botez, and J. Bowers, “Heterogeneously integrated distributed feedback quantum cascade lasers on silicon,” Photonics 3, 35 (2016).
4. P. Figueiredo, M. Suttinger, R. Go, A. Todi, H. Shu, E. Tsvid, C. K. N. Patel, and A. Lyakh, “Continuous wave quantum cascade lasers with reduced number of stages,” IEEE Photon. Technol. Lett. 29, 1328–1331 (2017).
5. Y. Bai, N. Bandyopadhyay, S. Tsao, S. Slivken, and M. Razeghi, “Room temperature quantum cascade lasers with 27% wall plug efficiency,” Appl. Phys. Lett. 98, 181102 (2011).
6. A. Lyakh, M. Suttinger, R. Go, P. Figueiredo, and A. Todi, “5.6 μm quantum cascade lasers based on a two-material active region composition with a room temperature wall-plug efficiency exceeding 28%,” Appl. Phys. Lett. 109, 121109 (2016).
7. R. Go, H. Krysiak, M. Fetters, P. Figueiredo, M. Suttinger, J. Leshin, X.-M. Fang, J. M. Fastenau, D. Lubyshev, A. W. K. Liu, A. Eisenbach, M. J. Furlong, and A. Lyakh, “Room temperature operation of quantum cascade lasers monolithically integrated onto a lattice-mismatched substrate,” Appl. Phys. Lett. 112, 031103 (2018).
8. S. M. Ting and E. A. Fitzgerald, “Metal-organic chemical vapor deposition of single domain GaAs on Ge/GexSi1-x/Si and Ge substrates,” J. Appl. Phys. 87, 2618–2628 (2000).
9. Y. Cordier and D. Ferre, “InAlAs buffer layers grown lattice mismatched on GaAs with inverse steps,” J. Cryst. Growth 201-202, 263–266 (1999).
10. W. K. Liu, D. Lubyshev, J. M. Fastenau, Y. Wu, M. T. Bulsara, E. A. Fitzgerald, M. Urteaga, W. Ha, J. Bergman, B. Brar, W. E. Hoke, R. R. LaRoche, K. J. Herrick, T. E. Kazior, D. Clark, D. Smith, R. F. Thompson, C. Drazek, and N. Daval, “Monolithic integration of III-V As-and P-based devices on Si through direct MBE growth and using lattice engineered substrates,” J. Cryst. Growth 311, 1979–1983 (2009).
11. A. Lyakh, C. Pflügl, L. Diehl, Q. Wang, F. Capasso, X. Wang, J. Fan, T. Tanbun-Ek, R. Maulini, A. Tsekoun, R. Go, and C. K. N. Patel, “1.6 W high wall plug efficiency, continuous-wave room temperature quantum cascade laser emitting at 4.6 μm,” Appl. Phys. Lett. 92, 111110 (2008).
12. F. Xie, C. Caneau, H. Leblanc, D. Caffey, L. Hughes, T. Day, and C. Zah, “Watt-level room temperature continuous-wave operation of quantum cascade lasers with λ > 10 μm,” IEEE J. Sel. Top. Quantum Electron. 19, 1200508 (2013).
