Open Access
Add to BibSonomy
Abstract
This research successfully developed an independent Ge-based VCSEL epitaxy and fabrication technology route, which set the stage for integrating AlGaAs-based semiconductor devices on bulk Ge substrates. This is the second successful Ge-based VCSEL technology reported worldwide and the first Ge-based VCSEL technology with key details disclosed, including Ge substrate specification, transition layer structure and composition, and fabrication process. Compared with the GaAs counterparts, after epitaxy optimization, the Ge-based VCSEL wafer has a 40% lower surface root-mean-square roughness and 72% lower average bow-warp. After device fabrication, the Ge-based VCSEL has a 10% lower threshold current density and 19% higher maximum optical differential efficiency than the GaAs-based VCSEL.
© 2024 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Vertical cavity surface emitting lasers (VCSELs) have seen a remarkable rise in demand as near-infrared illumination sources, driven by their integral roles in optical communications and 3D sensing with applications in data center optical links, virtual reality (VR), augmented reality (AR) technologies, autonomous driving and smartphones [1], [2]. Because of their scalable output power, high-quality optical beam, high wall-plug efficiency, stable wavelength over temperature, low spectral width, and easy testing and packaging, VCSELs are positioned to revolutionize light detection and ranging (LiDAR) systems in autonomous driving and industry robots [3]. Yet, as the VCSEL market is expected to increase from $\$$1.6 billion in 2022 to nearly $\$$3.9 billion by 2027, the prevailing VCSEL substrates used are still 4- to 6-inch GaAs wafers, which hinders production scalability [4–7].
A typical AlGaAs VCSEL structure, ranging from 5-15 microns in thickness, has the thickness mostly from the AlxGa1-xAs/AlyGa1-yAs superlattice (y > x) distributed Bragg reflectors (DBRs) [8]. For epitaxy on GaAs substrates, the inherent growth strain, caused by the lattice mismatch between GaAs (5.653 Å, the lower limit of AlxGa1-xAs) and AlAs (5.660 Å, the upper limit of AlyGa1-yAs), presents a significant impediment. On larger GaAs wafers, this mismatch strain-induced wafer bow and warp intensify, adversely impacting chip yield and reliability. Bulk Ge substrates emerge as a promising solution, offering a suitable lattice constant (5.658 Å, between GaAs and AlAs), resulting in reduced strain (up to 40% less lattice mismatch than GaAs) and enhanced post-growth stability [9]. Beyond this, Ge wafers offer enhanced mechanical resilience, reduced threading dislocations, and a competitive cost advantage, especially considering their availability in sizes up to a substantial 12 inches [10].
In 2020, IQE showcased 940 nm VCSELs on a 6-inch Ge wafer, achieving performance similar to co-processed GaAs-based VCSELs, which is the first report on Ge-based VCSELs [11]. However, details on their Ge substrate, transition layer structure and composition, and growth conditions were not disclosed. Our recent independent investigations successfully integrated AlxGa1-xAs n-DBRs and multi-quantum wells (MQWs) on bulk Ge with an InGaP nucleation layer and InGaAs and GaAs transition layers on bulk Ge [12,13]. Our bulk Ge-based n-DBRs and half VCSELs structure (n-DBRs and MQWs) have revealed comparable stopbands, photoluminescence (PL) spectra, and surface morphology with less bow-warp, compared to the same structures grown on co-processed bulk GaAs substrates. These studies paved the way for Ge-VCSELs with our nucleation and transition layer design and growth techniques.
In this study, we successfully demonstrated the monolithic integration of full VCSELs (including n-DBRs, MQWs, and p-DBRs) on 4-inch Ge wafers with an InGaP nucleation layer and InGaAs and GaAs transition layers. Our results show that the bulk Ge-based full VCSELs produced comparable epitaxy quality with less wafer bow and warp. Successful lasing was achieved. Details in the epitaxy transition layer materials, processing, materials characterizations, and electrical testing are discussed below.
2. Epitaxy structure and growth
A full VCSEL structure, targeting at 940 nm wavelength under operation temperatures, was designed and grown on bulk Ge and GaAs substrates, which consist of 40 pairs of AlGaAs-based n-DBRs in the bottom, three InxGa1-xAs/GaAs1-xPx MQWs in the middle, and 20 pairs of AlGaAs based p-DBRs in the top, as shown in Fig. 1. The top and bottom DBR layers are composed of Al0.9Ga0.1As/Al0.12Ga0.88As, incorporating highly doped graded interfaces for energy band continuity. Modulation doping is employed at the interfaces, adequately aligned with the node and anti-node of the DBRs’ standing wave distribution.
Fig. 1. Schematic of the design 940 nm full VCSEL structures grown on 4-inch (a) bulk Ge (with GaAs/InGaAs/InGaP transition layers and SiN backside coating) and (b) bulk GaAs substrates (not to scale).
The bulk Ge substrates used in this work were n-type, 4-inch, 425 µm thick (100) Ge wafers from Umicore, which has a 6-degree off-cut towards <111 > to avoid antiphase-domains (APDs) and enhances step-flow growth. The thickness of the Ge wafers is much thinner than the GaAs counterparts due to the more robust mechanical properties of Ge. A 100 nm backside SiN was deposed to prevent Ge sublimation. For a better transition from Ge, an InGaP nucleation layer, a 950 nm lattice-matched n-Ga0.985In0.015As layer, and a 50 nm n-GaAs layer were grown successively on the front side of Ge wafers. Then, the Ge wafers were transferred to another epitaxy reactor for the full VCSEL growth with surface cleaning conducted. A few 4-inch 625 µm thick bulk GaAs wafers with a 2-degree off-cut towards <111 > were used as the control samples for comparison. Another 500 nm GaAs was grown to obtain a clean surface.
The full VCSEL structures were grown on the 4-inch Ge and GaAs wafers using a production-scale metal-organic chemical vapor deposition (MOCVD) process. The vapor pressure during growth was 50 mbar, and the growth temperature was 760 °C. The sources used for GaAs growth were H2, SiH4, AsH3, and Ga(CH3)3 (TMGa). During the AlxGa1-xAs growth for DBR layers, Al2(CH3)6 (TMAl) was added to provide the Al source. For the MQW growth process, the sources to add In and P were In(CH3)3 (TMIn) and PH3. Notably, in the full VCSEL growth, the Ge wafers and GaAs wafers were grown in separate runs that were calibrated with the corresponding test wafers. This approach was chosen such that both Ge-VCSELs and GaAs-VCSELs were grown with proper calibration.
3. Materials characterization
To assess the epitaxy growth quality of the full VCSEL structures on the 4-inch Ge and GaAs wafers, we employed atomic force microscopy (AFM), scanning electron microscopy (SEM), bow-warp mapping, reflectance spectrum, and stopband measurement and mapping.
3.1 Cross-section imaging
The cross-sections of full VCSEL structures were obtained on a freshly cleaved surface made by a diamond scriber using an FEI Nova NanoSEM in backscattered electrons (BSE) mode. Both the GaAs/InGaAs/InGaP transition layers and the full VCSEL structures (N-DBR, MQW, P-DBR layers) are in good agreement with the design values, as shown in Fig. 2. In addition, good layer periodicity and cross-wafer uniformity of full VCSEL structures on both Ge and GaAs substrates can be confirmed in Fig. 3.
Fig. 2. Comparison of (a) the design Ge-VCSEL structure (the Ge substrate thicknesses are not to scale) and (b) the cross-section SEM of Ge-VCSEL after epitaxy (captured by a BSE detector at 25KX magnifications). Only a small portion of the Ge substrate is shown.
Fig. 3. Comparison of cross-section SEM images of full VCSEL structures on (a) bulk Ge substrates and (b) bulk GaAs substrates at 30KX magnifications, captured by a BSE detector. Only small portions of the Ge and GaAs substrate are shown.
3.2 Surface morphology
The surface morphology of Ge and GaAs wafers after full VCSEL growth was analyzed by atomic force microscopy (AFM) images in Fig. 4. Compared to the GaAs wafer, a much smoother surface with no cracks or antiphase-domains was observed on the Ge wafer. In addition, the Ge wafer exhibits a 40% reduction in both average roughness (Ra) and root mean square roughness (Rq). These results indicate that the Ge wafer has much better surface quality than GaAs wafer after the full VCSEL epitaxy growth, which is primarily due to the smaller lattice mismatch between Ge substrate and AlxGa1-xAs DBRs.
Fig. 4. AFM images and roughness measurement on $10 \times 10$ µm surfaces of full VCSELs on (a) Ge substrates and (b) GaAs substrates. Ra and Rq represent average roughness and root mean square roughness respectively.
Fig. 5. Bow-warp mapping on (a) the 4-inch Ge wafer and (b) the 4-inch GaAs wafer after the full VCSEL epitaxy growth. All the units are in micron meter (µm).
Fig. 6. Room temperature reflectance spectra at center points on (a) the 4-inch Ge wafer and (b) the 4-inch GaAs wafer after the full VCSEL epitaxy growth by Nanometrics RPM Blue Wafer Laser Measurement Tool.
Fig. 7. Stopband center mapping on (a) the 4-inch Ge wafer and (b) the 4-inch GaAs wafer after the full VCSEL epitaxy growth by Nanometrics RPM Blue Wafer Laser Measurement Tool.
As the predominant thickness component of a typical AlGaAs VCSEL structure is derived from the AlxGa1-xAs/AlyGa1-yAs superlattice (y > x) DBRs, the large lattice mismatch (0.13%) between the GaAs substrate (GaAs, 5.653 Å) and the high index layer of the DBRs (Al0.9Ga0.1As, 5.660 Å) will induce considerable lattice strain during growth. However, Ge substrate, with intermediate lattice constant (Ge, 5.658 Å) between GaAs and AlAs, has only 1/3 of the lattice mismatch (0.04%) between itself and the high index layer of the DBR layers (Al0.9Ga0.1As). The suitable lattice constant of Ge and less lattice mismatch leads to a much smaller bow-warp of Ge-based VCSELs compared to GaAs-based VCSELs.
3.3 Wafer bow-warp maps
The bow-warp maps of Ge- and GaAs-VCSEL wafers were measured by FLX 2320 made by Toho Technology. As shown in Fig. 5, the Ge wafer has a 72% less average bow-warp value (10.28 µm) than that of the GaAs wafer (36.59 µm), which resulted from the more minor lattice mismatches between the Ge substrate and AlxGa1-xAs DBRs as discussed above.
3.4 Reflectance spectrum and stopband center maps
Reflectance spectrum is an important crucial characterization to evaluate the design and epitaxy quality of VCSEL structure, which measures how much light at different wavelengths is reflected off the surface. The reflected light is normally captured by a photodiode that converts the light intensity into voltage signals, proportional to the intensity of the reflected light. All the reflectance spectra in this paper were obtained by Nanometrics RPM Blue Wafer Laser Measurement Tool. Before the measurement, the voltage readings of the equipment have been calibrated using a standard silver mirror.
The reflectance spectra of our Ge- and GaAs-VCSEL wafers were measured at the center point of each wafer and under room temperature. As shown in Fig. 6, the overall reflectance spectra, stopband center, and the Fabry–Perot dip of both Ge- and GaAs-VCSEL wafers are very similar and in good line with the design.
The uniformity in the stopband is crucial for consistent VCSEL performance across the wafer, influencing aspects like emission wavelength and efficiency. Room-temperature stopband center mapping by the same equipment was conducted to check the stopband uniformity throughout the 4-inch wafer. The measurements were taken at 1 mm intervals across the wafers in both X and Y directions. The constructed stopband center maps, as shown in Fig. 7, reveal that Ge-VCSEL wafer has an 8.4 nm peak-valley value, while GaAs-VCSEL wafers have a 12.7 nm peak-valley value, which suggests that Ge-VCSEL has a better stopband uniformity across the wafer. These results from the smaller average lattice mismatch between Ge substrate and AlxGa1-xAs DBRs discussed above.
3.5 Summary of materials analysis
The critical material characterization results are summarized in Table 1. From the material analysis perspective, the full VCSEL epitaxy on 4-inch Ge wafers is generally better than that grown on 4-inch GaAs wafers, including better surface conditions, less bow-warp, and more uniform stopband. These results confirm that Ge is more suitable for mass production of VCSEL on larger substrates.
Table 1. Summary of materials characteristics of GaAs- and Ge-VCSEL epitaxy.
4. Device fabrication
The VCSEL device fabrication began with the formation of p-type ohmic contacts using physical vapor deposition (PVD) of Ti/Pt/Au. Subsequently, a nitride layer was deposited as a hard mask for precise dry etching. Circular mesa patterns were formed using inductively coupled plasma reactive ion etch (ICP-RIE) to define the desired device structure. Then, the device underwent a furnace process with H2O and N2 gas, creating a 9.7 µm diameter oxide aperture for GaAs-based VCSEL and an 11.7 µm diameter oxide aperture for Ge-based VCSEL.
Next, an Au/Ge/Ni/Au film was evaporated and annealed to form the n-type ohmic contacts. Planarization was achieved using Benzo-cyclobutene (BCB) material to reduce external parasitic capacitance. Finally, the device fabrication was completed by via-hole etching and deposition of Ti/Au contact pads to connect the VCSEL ring contact. The SEM images of the fabricated VCSEL devices after via-hole process and after the final lithography process are shown in Fig. 8(a) and (b) respectively.
Fig. 8. (a) The SEM image of the VCSEL after via-hole process, showing both the P-type and N-type metals. (b) The SEM image of the VCSEL after the final lithography process, displaying the P-type and N-type bonding pads, which were designed in cooperation with coplanar top contacts in a ground-signal (GS) configuration.
5. Device characterization
To verify the lasing under electrical pumping and characterize the static performance of the fabricated VCSEL devices on Ge and GaAs substrates, a typical setup is configured for device characterization. The devices were placed on a probe station with a temperature controller as shown in Fig. 9(a) and (b). The optical emission spectra were measured by an optical spectrum analyzer using OM4 multimode fiber to couple the light output from the VCSELs directly. A Precision IV Analyzer was used to provide the bias current, and the optical output power is measured by a large-area photodetector. Figure 9(c) presents the optical spectra of 940 nm VCSEL devices grown on bulk GaAs (upper) and Ge substrates (bottom), measured at room temperature with a bias current of 5 mA. The peak wavelength locates at 940.14 nm for GaAs-based VCSELs and 940.57 nm for Ge-based VCSELs, both of which align closely with the target wavelength. The data indicates successful lasing for 940 nm VCSELs on both substrates.
Fig. 9. (a) Setup for VCSEL static performance measurement. (b) Microscopic image of our fabricated full VCSEL devices on Ge substrates being tested on a probe station. (c) Room temperature optical emission spectra for 940 nm VCSELs grown on bulk GaAs (upper red line) and Ge (bottom blue line) substrates at a bias current of 5 mA.
The room temperature light-current density-voltage (L-J-V) characteristics of the overall best VCSEL devices fabricated on GaAs and Ge substrates, along with optical power and optical differential efficiency as a function of bias current, are shown in Fig. 10. Compared to GaAs-based VCSEL devices, the Ge-based VCSEL devices demonstrate a lower threshold current density of 2.33 kA/cm2, a higher optical output power of approximately 3.98 mW at a corresponding current density of 15 kA/cm2, and superior maximum optical differential efficiency of 0.37 mW-kA−1-cm2. The GaAs-based VCSEL has a threshold current density of 2.58 kA/cm2, an optical output power of approximately 3.35 mW at a corresponding current density of 15 kA/cm2, and a maximum optical differential efficiency of 0.31 mW-kA−1-cm2. A full comparison of the static performance between Ge-VCSEL and GaAs is shown in Table 2.
Fig. 10. (a) Room temperature L-J-V curves for 940 nm VCSELs grown on GaAs (red line) and Ge (blue line) substrates. (b) Optical power and optical differential efficiency as a function of bias currents.
Table 2. Summary of static characteristics of 940 nm VCSELs on GaAs and Ge substrates.
Due to the unoptimized device fabrication process and the limited sample numbers, our selection of devices with similar aperture sizes is restricted. We are actively optimizing the device fabrication process and conducting more detailed studies to understand the performance impact of various aperture sizes in both GaAs-based and Ge-based VCSELs. Despite the current size differences, our results demonstrate that Ge-based VCSELs show promising lasing performance comparable to GaAs-based VCSELs, highlighting their potential for mass production at lower costs and higher yields.
Nevertheless, the Ge-VCSELs have lower roll-over currents (approximately 29.49 kA/cm2) owing to an elevated density of defects stemming from crystal orientation challenges within the substrate. The lattice mismatch between the Ge substrate and the epitaxial layers grown atop it can induce dislocations and strain in the epitaxial layers. This phenomenon gives rise to the formation of defects and non-radiative recombination centers that release energy in the form of heat, potentially imposing limitations on the maximum operational currents before reaching thermal roll-over currents. In addition, the rollover currents for VCSELs are highly related to the thermal impedance and the junction temperature. We are conducting temperature-dependent studies to further explore and validate these mechanisms, which will deepen our understanding and facilitate the enhancement of Ge-based VCSELs.
6. Conclusions
We independently developed a 940 nm AlGaAs full VCSEL technology on bulk Ge substrates. This is the second successful Ge-VCSEL technology reported worldwide and the first Ge-VCSEL technology with technology details disclosed. Before the VCSEL epitaxy growth, the process was optimized for Ge and GaAs substrates respectively to ensure good alignment between Fabry–Perot dip and peak emission at operating temperature. Thanks to Ge’s much lower threading dislocation density, Ge’s higher fracture toughness, and the high quality of GaAs/InGaAs/InGaP transition layers, our Ge-VCSEL has 40% less surface roughness and 72% less bow-warp, compared to the GaAs counterpart. After the device fabrication, our Ge VCSEL device has a 10% lower threshold current density and a 19% higher maximum optical differential efficiency. This study not only demonstrated the feasibility and revealed the key details of fabricating VCSELs on Ge substrates for mass production, but also paved the way to integrate AlGaAs semiconductor devices on bulk Ge substrates for a wide range of optical and electrical applications.
Funding
Huawei Technologies Canada; University of British Columbia.
Acknowledgment
Umicore N. V., Belgium, is acknowledged for providing the bulk Ge wafers with GaAs/InGaAs/InGaP epitaxy layers and SiN back coating.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. A. Extance, “Faces light up over VCSEL prospects,” SPIE Professional magazine, Apr. 2018.
2. M. Dummer, K. Johnson, S. Rothwell, et al., “The role of VCSELs in 3D sensing and LiDAR,” Proc. of SPIE 11692(116920C), 9 (2021). [CrossRef]
3. K. Iga, “Surface-emitting laser-its birth and generation of new optoelectronics field,” IEEE J. Sel. Top. Quantum Electron. 6(6), 1201–1215 (2000). [CrossRef]
4. J. F. Seruin, D. Zhou, G. Xu, et al., “High-efficiency VCSEL arrays for illumination and sensing in consumer applications,” Proc. of SPIE 9766(97660D), 97660D (2016). [CrossRef]
5. H. Moench, M. Carpaij, P. Gerlach, et al., “VCSEL-based sensors for distance and velocity,” Proc. of SPIE 9766(97660A), 97660A (2016). [CrossRef]
6. Yole Dévelopment, “VCSEL – Technology and Market Trends 2022 report,” Technical report, 2022.
7. D. Wiedenmann, M. Grabherr, R. Jäger, et al., “High volume production of single-mode VCSELs,” Proc. of SPIE 6132(613202), 613202 (2006). [CrossRef]
8. C. Bocchi, C. Ferrari, P. Franzosi, et al., “Accurate determination of lattice mismatch in the epitaxial AlAs/GaAs system by high-resolution X-ray diffraction,” J. Cryst. Growth 132(3-4), 427–434 (1993). [CrossRef]
9. M. Bosi and G. Attolini, “Germanium: Epitaxy and its applications,” Prog. Cryst. Growth Charact. Mater. 56(3-4), 146–174 (2010). [CrossRef]
10. B. Depuydt, A. Theuwis, and I. Romandic, “Germanium: From the first application of Czochralski crystal growth to large diameter dislocation-free wafers,” Mater. Sci. Semicond. Process. 9(4-5), 437–443 (2006). [CrossRef]
11. A. Johnson, A. Joel, D. Pearce, et al., “High performance 940 nm VCSELs on large area germanium substrates: the ideal substrate for volume manufacture,” Vertical-Cavity Surface-Emitting Lasers XXV. 11704, 1 (2021). [CrossRef]
12. Y. Zhao, J. Guo, M. Feifei, et al., “Monolithic integration of 940 nm AlGaAs distributed Bragg reflectors on bulk Ge substrates,” Opt. Mater. Express 12(3), 1131–1139 (2022). [CrossRef]
13. Y. Zhao, Z. Wan, J. Guo, et al., “Monolithically Integrated 940 nm Half VCSELs on Bulk Ge Substrates” IEEE Photonics Technology Letters, 2023.
