New Generation of Vertical-Cavity Surface-Emitting Lasers for Optical Interconnects

N. Ledentsov Jr, V. A. Shchukin, N. N. Ledentsov, J.-R. Kropp, S. Burger and F. Schmidt

VI Systems GmbH, Hardenbergstraße 7, Berlin, 10623, Germany

Zuse Institute Berlin (ZIB), Takustraße 7, Berlin, 14195, Germany and JCMwave GmbH, Bolivarallee 22, Berlin 14050, Germany

1 Introduction

As silicon downscaling continues, the pitch size is gradually decreasing. The number of transistors per chip, consequently, further increases. Presently, the major IC manufacturers are planning to market 10 nm technology in 2015 with a further upgrade to 7 nm anticipated in 2017.^{1 2}

With the growing number of transistors per chip and upgrades in the architecture, processor productivity continues to approximately double every year, increasing demands on the processor communication bandwidth. Consequently, the speed of the input/output (I/O) ports must also increase. Until recently, Moore's Law for data communications predicted that a fourfold increase in the I/O speed would be needed every 4–5 years. Thus far, this trend has been generally maintained in major communication standards. According to the IEEE Ethernet Roadmap, the single I/O bitrate should approach 100 Gb/s by 2017. Indeed the presently active IEEE 400G Ethernet Task Force for the related standard recently agreed on serial single channel bit data rate of 400 Gb/s in short distance communications.³ The aggregated transmission rate is to reach only 400G by 2017, but the variety of standard combinations and evolution of mid-board transceivers will allow customized solutions to reach 1 Tb/s or higher. However, at I/O speeds of well above 10 Gb/s, copper cables and connectors are becoming too energy consuming, bulky, and susceptible to electromagnetic pollution. These factors make copper links difficult even inside the box. Thus, massive deployment of optical interconnects is already starting in advanced areas, such as high-performance servers with 60,000 optical links per single rack.⁴

Beyond supercomputers and data centers, even consumer applications are beginning to require unprecedented data transmission rates. Recently, 4K and 5K displays became broadly available and the data traffic demands increased up to 80–100 Gb/s per screen for the highest image quality specs (color tones, frames per second). Next we will have 8K and volumetric displays, with IMAX-like bandwidth requirements of up to 175 Gb/s. The Thunderbolt 3 interface from Apple due in 2015 is already at 40 Gb/s. At higher data bitrates, the use of copper interfaces steadily shrinks while the numbers of optical links explode to tens of millions per single supercomputer, as a result of the increased bandwidth demand.

In this chapter, we address requirements for vertical-cavity surface-emitting lasers (VCSELs), one of the major devices for data communication to meet the bandwidth demand. Single-mode (SM) operation makes it possible to overcome effects related to significant spectral dispersion of the multimode fiber (MMF) standardized for the 840–860 nm wavelength range.

2 VCSEL requirements

To match the needs of the new generation of optical links, VCSELs have to meet a number of key requirements. They need to provide very high data transmission speeds at high energy efficiency, allowing close packaging into parallel links run by multichannel driver electronics. They also need to ensure data transmission over the necessary lengths of fiber. As the critical distance-limiting factor in modern fiber chromatic dispersion, VCSELs must provide an ultranarrow emission spectrum and low wavelength chirp under signal modulation.

Several designs have been proposed for high-temperature, energy-efficient VCSEL operation.^5–7 These include antiwaveguiding cavity design with AlAs-rich core, increased optical confinement, engineering of the density of states by quantum well (QW) strain engineering, thick oxide-confined apertures, and superlattice barriers aimed at prevention of the leakage of nonequilibrium carriers.⁶

This progress, for example, enabled serial digital data transmission up to ∼50 Gb/s without using advanced electronics.⁷ Even higher bitrates can be achieved with equalization schemes, at a cost of extra energy consumption and more ancillary electronics.

Today's technology is based on current-modulated VCSELs, which still show a high potential for future technology generations. Electrooptically modulated VCSELs represent a potential alternative. In such devices, an additional “shutter” section based on the electrooptic effect extends the optical modulation bandwidth beyond 35 GHz and the electrical bandwidth beyond 60 GHz, putting data transmission at 10 Gb/s within reach.

Special VCSEL designs may also allow uncooled wavelength division multiplexing (WDM) within the narrow 840–860 nm spectral range of low modal dispersion in the standard MMFs. Complete temperature stability of the VCSEL can be achieved by using the passive cavity concept. The gain medium is placed in the bottom part of the bottom semiconductor distributed Bragg reflector (DBR), while the upper part of the bottom DBR, the cavity region and the top DBR are all made of dielectric materials. Due to the fairly weak or zero dependence of the refractive index on temperature for certain combinations of dielectric materials, temperature stabilized VCSEL operation without cooling becomes possible. Furthermore, the dielectric DBRs and cavity extend a simple VCSEL technology with a high optical confinement factor and good heat conductivity even to materials normally less suitable for VCSEL fabrication, for example, for InP-based 1300–1550 nm VCSELs. This extends the range of VCSEL applications.

All VCSEL applications in the datacom space require data transmission over long sections of MMF, because the size of modern data centers is growing continuously. On the other hand, even at 25 Gb/s, the transmission distance for multimode VCSELs shrinks to 20–100 m only, depending on the particular standard and application. Single-mode low wavelength chirp VCSELs can be used to overcome this problem. Such devices allow 25G × 12 parallel transmission over 1 km of standard MMF, whereas arrays of conventional high-speed multimode VCSELs can hardly exceed 150 m distances.⁸

WDM based on the diffraction-induced angle separation of the wavelengths also requires single-mode VCSELs to ensure a stable, reproducible far-field pattern under high-speed modulation. The same requirement applies to VCSELs mated to multicore fibers with reduced diameters and close spacing of the core regions.

Single-mode VCSEL operation can be achieved in devices with very small aperture sizes (∼2–3 µm). However, such devices show fairly low power and high series resistance. Industrial applications generally require VCSELs with moderate oxide apertures (5–6 µm) capable of producing 2–3 mW of output power at moderate current densities <20 kA/cm² over a broad temperature range. The differential resistance should be kept below 100 Ω to match standard driver electronics.

Progress in the field of high-speed VCSEL devices has been quite rapid. Very recently, single-mode (SM) VCSELs demonstrated 54 Gb/s transmission over 1 km of OM4 MMF under nonreturn to zero (NRZ) modulation⁹ and 100 Gb/s over 300 m of MMF using multitone transmission.¹⁰

Several approaches are possible for producing SM VCSELs. For example, surface patterning of the device adjacent to the aperture boundary can create higher scattering loss for high-order modes, favoring the fundamental mode in the lasing process.¹¹ Alternatives include increasing the thickness of the optical cavity to reduce the impact of the oxide aperture on mode confinement and placing thin oxide aperture layers in the mode position of the longitudinal optical mode. Unfortunately, these approaches either require high process precision to align the sizes of the related features or suffer from the unstable far field, fairly low optical confinement factor in the gain medium and thus a low speed and high capacitance. The leaky VCSEL concept can allow SM operation without any adjustments in the oxide-confined VCSEL technology and does not require sacrificing the basic VCSEL parameters.¹² Such a device is the main topic of the present chapter.

3 Optical leakage

The optical leakage concept allows the realization of high-power SM VCSELs, which can be used in multiple fields such as illumination, gas sensing, gesture recognition, data transmission, and other fields.^{6 13} There is a growing interest in the development of leaky-mode VCSELs.^{14 15}

The leaky VCSEL concept, as compared to other SM VCSEL approaches, provides a possibility to couple several devices through the leaky emission into a coherent VCSEL array to achieve increased brightness, realize two-dimensional beam steering,¹⁴ or on-chip lateral integration with other types of devices such as monitor photodetectors, slow light waveguides, and 2D optical logic gates for optical computing.¹⁶ Electrooptically controlled leakage can be applied for VCSEL modulation at ultrahigh speed.^{17 18}

Leaky-mode VCSELs can be fabricated using different techniques. In one approach, VCSEL mesas are formed by etching with subsequent overgrowth using a material with a higher refractive index.^{19 20} The VCSEL cavity modes can leak out into the high refractive index material. As the high-order transverse modes have a higher intensity at the boundary of the aperture, they exhibit a much higher leakage loss than the fundamental mode. Another approach is based on etching the top DBR over the aperture region, resulting in a locally shorter effective cavity length that allows the leakage of cavity modes into the surrounding parts of the structure. A similar concept can be realized by thinning the cavity layer in the aperture region prior to the DBR deposition or using photonic crystal patterning of the VCSEL surface, filtering out undesirable high-order modes from the aperture region.²¹

Recently, it was shown that leaky VCSELs can be realized by a proper epitaxial design of an oxide-confined device without any need for surface patterning or etching and overgrowth.¹⁵ As a result, leaky VCSELs can be fabricated in a process that is fully compatible with the standard oxide-confined VCSEL technology without additional processing steps.⁶

In spite of active research in the field of leaky VCSELs, no studies of the impact of the leakage effect on the far-field pattern of the device have been reported so far. This is in a strict contrast to leaky-design edge-emitting lasers, where the leakage effect is clearly manifested by a specific emission tilted away from the normal to the facet of the device.²² Such emission should exist in leaky-mode VCSELs and may serve as a fingerprint of this design, as was recently proposed.²³

In the present chapter, we investigate oxide-confined, leaky-design VCSELs and observe the characteristic leakage-effect-induced features in the vertical far-field pattern of the device. No such emission is detected in conventional oxide-confined VCSELs. A perfect match is found between the near and far fields, in agreement with cold-cavity, three-dimensional (3D) modeling. Clearly resolved narrow emission lobes at high tilt angles are shown to be a unique property of the leaky-mode device. The leaky emission provides a fingerprint and a quantitative measure of the optical leakage in the lateral direction and represents a powerful tool for proper design of the related vertical microcavity devices, their coherent assemblies, and on-chip combinations with other devices.

4 Experiment

Epitaxial wafers for oxide-confined 850 nm VCSELs were grown by MOCVD in an industrial multiwafer reactor. The epitaxial design, optimized for ultrahigh-speed operation, included conventional AlGaAs DBRs with graded compositional profiles, an AlGaAs-based microcavity, and an active region composed of five compressively strained InGaAs quantum wells (QWs).

The VCSEL wafer was fabricated in a high-frequency contact pad configuration with BCB planarization and was tested without soldering to a heat sink. The design and the basic properties of the oxide-confined leaky VCSELs are presented in Refs 12 15.

A CCD camera was used to evaluate the near-field pattern. Far-field measurements were performed over the full 180° range with 2° resolution by using a Si detector.

• Benchmarking to conventional VCSELs

For benchmarking, we first studied conventional double aperture VCSEL device similar to the one described in Ref. 6.

Double oxide aperture and a 3λ/2 cavity were applied both to the reference and the leaky VCSELs. A characteristic feature of the leaky VCSEL is either single-mode operation at moderate aperture diameters (∼5 µm) or a multimode spectrum at larger ∼8–10 µm apertures but at a strongly reduced overall spectral width as compared to the nonleaky VCSEL design at similar aperture values.^{15 24} Modeling of the near and far fields of conventional VCSELs has a long history and is well understood.^{6 12, 15 23, 24} Oxide-confined leaky VCSELs were predicted to have a specific emission at large tilt angles with a narrow angular width.²³ This feature originates from the leaky emission in the direction outside of the aperture. As no confinement for this emission is provided by the aperture diameter, the diffraction-related broadening is suppressed and the feature in the far field can be narrow.

In Fig. 1, we show electroluminescence spectra of a conventional double oxide aperture 3λ/2 VCSEL at room temperature as a function of current. It follows from Fig. 1 that the conventional VCSEL is heavily multimode even at low currents in spite of the narrow ∼4 µm diameter aperture, as revealed by the wavelength splitting between the fundamental and the excited modes of 1.2 nm.

Figure 2(a) shows the far-field diagrams of the same VCSEL as a function of current. At low current, a signal background having a cosine intensity distribution is observed due to the spontaneous emission, which saturates when the lasing evolves. Lasing is revealed in the far field by appearance of the tilted overlapping features (at ∼7° in the angular space) representing high-order transverse mode(s). Measurements made with a CCD camera (sometimes called “near-field” images) clearly reveal multimode behavior as well (see Fig. 2(b) and (c)).

• Leaky VCSELs

A very different behavior is observed for the leaky VCSEL. Even the device with a mode spacing of ∼0.85 nm and an aperture diameter of ∼5 µm shows predominantly single-mode lasing up to ∼4 mA (see Fig. 3).

The far-field pattern emitted by the device above threshold confirms the single-mode behavior. At low intensity, spontaneous emission can be revealed. At currents below 4 mA, the excited mode is not manifested in the spectra. Figure 4 presents the far field of the leaky VCSEL with double oxide aperture as a function of current. At low current, we find the spontaneous emission background with a cosine intensity distribution. Lasing is revealed by appearance of a single-lobe structure in the angular space at currents above threshold.

As follows from Fig. 4, the leaky VCSEL device of the leaky design is single mode up to 4 mA. Electroluminescence spectra of the leaky VCSEL at larger aperture size of ∼5.5 µm and smaller mode spacing of ∼0.75 nm are shown in Fig. 5 for completeness.

The CCD camera images and the far-field diagrams of the device are shown in Fig. 6. The transformation in the near-field pattern agrees with the appearance of the tilted lobes in the far-field pattern at tilt angles ∼7°. The spectra are still narrower both in wavelength and angular space as compared to those of the conventional nonleaky VCSEL design of Figs 1 and 2 used for benchmarking. The most interesting phenomenon is the emergence of an angularly narrow emission at large tilt angles ∼35°. The full width at half maximum of the emission is only ∼2°, much narrower than the angular width of the fundamental and excited lobes that are broadened by diffraction due to the small oxide aperture.

5 Simulation

To understand the experimental results, we now turn to numerical modeling of the optical modes in the VCSEL structure in the cold cavity approach.

Our 3D simulations of electromagnetic fields were performed with the JCM Wave finite element software package based on full vector Maxwell's equations. Cylindrical symmetry was applied,²⁵ allowing a 2D solution to represent the entire 3D picture. Refractive indices of materials for the simulation were taken from Ref. 26.

• Leaky design

The electromagnetic field was modeled for in-plane and cross-section intensity distributions, as shown in Fig. 7. For clarity in transverse images, we simulated the optical field transformation up to 10 µm above the surface of the VCSEL.

In Fig. 7, the bright areas represent the maximum of intensity, and dark areas represent the minimum of intensity (see the logarithmic intensity scale). In the fundamental (HE11) mode, the maximum intensity is located in the center of the VCSEL and lasing emission is visible above the surface as a homogeneous column in the center. Simulation also picks up a small in-plane leakage parallel to the surface and even weaker leakage of the light into the air at some tilt angle.

The first excited mode (HE21), on the other hand, has the maximum of intensity not in the center but closer to the boundary of the oxide aperture, at a radius r = 1.5 µm.

As expected, since the intensity maximum of the higher order lateral mode is located closer to the boundary of the oxide layers, leading to far greater in-plane leakage. Consequently, the strongly tilted emission in the air that is a manifestation of the leakage process is much stronger for the high-order transverse mode.

Based on this simulation, the leakage emission has an angle of 35°. This result corresponds well to the 1D simulation of the optical reflectance spectra of the structure in oxidized and nonoxidized regions and with the experimental results of Fig. 7.¹⁵

In Refs 15 24, the optical power reflectance (OPR) spectra of the same leaky VCSEL structure was simulated by the transfer matrix method applied separately to a multilayer all-semiconductor structure in the nonoxidized core region and in the oxidized periphery region of the device. The OR spectrum of the nonoxidized region at normal incidence contains a reflectivity dip at the lasing wavelength 850 nm, which matches a dip in the OR spectrum in the oxidized region calculated for propagation of light at a tilt angle of ∼10° in the semiconductor. Thus, the VCSEL mode of the nonoxidized cavity is in resonance with a tilted optical mode of the oxidized region. The angle of ∼10° in the semiconductor transfers, upon refraction, into an angle of ∼33° in air, which agrees well both with the 3D simulation and experiment. The discrepancy in the angle between 1D and 3D simulation can be due to the additional lateral confinement taken into account in the 3D simulation.

Figure 8 shows the simulated far-field profiles of the fundamental and the first excited modes, whereas Fig. 9 compares the experimentally measured and the simulated far-field profiles.

Since the experimental setup has an angular resolution of 2°, the observed leakage features are broadened compared to simulation. Furthermore, the leakage lobes in the simulation are narrower compared to the experiment because the simulation assumes an effectively infinite surface area for the leakage, since optical field propagation is limited only by the gradual decay of the optical field intensity due to attenuation. In the real device, the area is also limited by the contact metal ring opening of only 11 µm in diameter that diffracts the leaky emission.

The simulated integral intensity of the leakage-related lobes is comparable to the one found in the experiment within the experimental resolution. This is an important observation, proving that in spite of all the uncertainties – for example, strain fields in and around the oxide layers, tapering of the oxide aperture caused by the side oxidation of the graded Al_xGa_1−xAs layers, free carrier absorption, and other effects – it is the leakage that mainly determines the emission pattern.

• Leaky design with oxide relief

The leakage effect is easier to control when the difference in the refractive index between the semiconductor inner region and the oxide outer region of the aperture-forming layer is maximized. Selective etching of the oxide layer provides an extra tool to control and enhance the leakage effect. Thus, we model VCSELs based on the oxide relief concept^{27 28} and evaluate the impact on the leakage effect. In the simple case, we just replace the oxide apertures with air gaps.

Influence of oxide relief on conventional VCSELs

First we look at a conventional VCSEL design, similar to the one described in Figs 1 and 2. As mentioned earlier, no directed leakage is observed for this type of device, which is confirmed by far-field modeling. When oxide relief is applied to the VCSEL, narrow tilted lobes appear in the modeling of the far-field profile at ∼60°. Far-field profiles before and after oxide relief are shown in Fig. 10.

The modal lifetimes are compared in Fig. 11. The lifetimes of the first excited modes are relatively shorter than the fundamental mode in structures with oxide relief. The difference is particularly strong for 6–9 µm aperture sizes. This promises increased mode selectivity and improved range of single-mode operation.

Influence of oxide relief on leaky VCSELs

Now we turn to applying oxide relief to the leaky VCSELs. Simulated electric field intensity distribution in a device with 5 µm aperture shows that after applying the oxide relief, strong field intensity inside the air gaps can be seen. In the simulation of the far-field profiles, we observe an increase in the intensity and a slight decrease in the angle of leakage emission – see Fig. 12.

Comparison of the lifetimes of the fundamental and first excited modes presented in Fig. 13 shows that the structure with oxide relief has decreased lifetimes, but there seems to be no significant improvement in the mode selectivity. Oscillatory behavior of the mode lifetimes in the leaky structures is evident in both cases, with the reasons of the effect discussed in Ref. 16.

Our modeling shows that the leakage effect can be further improved in leaky VCSELs by employing the oxide relief technique, but the impact is weaker than in otherwise nonleaky-design, oxide-confined VCSELs.

Our modeling approach can be further extended to oxide-relief VCSELs with locally Zn-intermixed DBRs.²⁹ It can also be extended to VCSELs with different types of surface relief,³⁰ photonic crystals, and coherent arrays of leaky VCSELs.^{31 32}

6 Conclusion

To conclude, we have considered requirements for modern VCSELs used in data communications. Single-mode VCSEL technologies at high data rates of up to 100 Gb/s are needed for future communication networks. As any deployed VCSEL technology should meet reliability of standard oxide-confined VCSELs, the leaky design approach appears particularly promising. We studied both theoretically and experimentally the leakage-related effects in oxide-confined VCSELs applying 3D vector modeling of the optical field and evaluated spectral, near- and far-field properties of the device. In line with the theory, we experimentally observed the leakage-induced emission revealed as narrow tilted lobes in the far field of the VCSEL. This observation confirms the validity of the leaky VCSEL concept, allows for better understanding of the device properties such as possible strain and heat gradients, oxide layer tapering, in-plane light scattering, and absorption. Our work enables the engineering of advanced devices and photonic-integrated circuits with targeted design of oxide apertures or air gaps through quantitative evaluation of the leaky emission.

Acknowledgments

The authors acknowledge support by ADDAPT project of the FP7 Program of the European Union under Grant Agreement No. 619197.

References

1. 1. W. M. Holt, “Moore's law: A path forward,” Proc. Intern. Solid-State Circ. Conf. (ISSCC) (2016), p. 1.
2. 2. Taiwan Semiconductor Manufacturing Company, “Future R&D plans,” see www.tsmc.com/ (accessed on Apr. 21, 2016).
3. 3. Documents of IEEE P802.3bs 400 Gb/s Ethernet Task Force, (2015), see www.ieee802.org/3/bs/index.html (accessed on Apr. 21, 2016).
4. 4. C. Nitta, R. Proietti, Y. Yin, et al., “Leveraging AWGR-based optical packet switches to reduce latency in petascale computing systems,” technical report CSE-2013-68, University of California–Davis, 2012.
5. 5. N. N. Ledentsov and J. Lott, “New-generation vertically emitting lasers as a key factor in the computer communication era,” Phys.-Usp. 54, 853–858 (2011).
6. 6. N. N. Ledentsov, “Ultrafast nanophotonic devices for optical inter-connects,” chapter in: S. Luryi, J. M. Xu, and A. Zaslavsky, eds., Future Trends in Microelectronics: Into the Cross Currents, New York: Wiley, 2013, pp. 43–48.
7. 7. N. N. Ledentsov, J.-R. Kropp, V. A. Shchukin, et al., “High-speed modulation, wavelength, and mode control in vertical-cavity surface-emitting lasers,” Proc. SPIE 9381, 93810F (2015).
8. 8. J.-R. Kropp, V. A. Shchukin, N. N. Ledentsov, et al., “850 nm single mode VCSEL-based 25G×16 transmitter/receiver boards for parallel signal transmission over 600 m of mm fiber,” Proc. SPIE 9390, 93900C (2015).
9. 9. G. Stepniak, A. Lewandowski, J.-R. Kropp, et al., “54 Gbit/s OOK transmission using single-mode VCSEL up to 2.2 km MMF,” Electron. Lett. 52, 633–635 (2016).
10. 10. B. Wu, X. Zhou, N. Ledentsov, and J. Luo, “Towards 100 Gb/s serial optical links over 300 m of multimode fibre using single transverse mode 850 nm VCSEL,” Asia Commun. Photonics Conf. (ACP) (2015), paper Asu4C.3.
11. 11. E. Haglund, Å. Haglund, J. S. Gustavsson, et al., “Reducing the spectral width of high speed oxide confined VCSELs using an integrated mode filter,” Proc. SPIE 8276, 82760L (2012).
12. 12. V. Kalosha, N. N. Ledentsov, D. Bimberg, et al., “Design considerations for large-aperture single-mode oxide-confined vertical-cavity surface-emitting lasers,” Appl. Phys. Lett. 101, 071117 (2012).
13. 13. M. Grabherr, “New applications boost VCSEL quantities: Recent developments at Philips,” Proc. SPIE 9381, 938102 (2015).
14. 14. C. Lei and K. D. Choquette, eds., “Vertical-cavity surface-emitting lasers X,” Proc. SPIE 6132, 61320J (2006).
15. 15. V. A. Shchukin, N. N. Ledentsov, J.-R. Kropp, et al., “Single-mode vertical cavity surface emitting laser via oxide-aperture-engineering of leakage of high-order transverse modes,” IEEE J. Quantum Electron. 50, 990–995 (2014).
16. 16. D. K. Serkland, K. M. Geib, G. M. Peake, et al., “Final report on LDRD project: Leaky-mode VCSELs for photonic logic circuits,” U.S. Department of Energy, SAND 7118 (2005).
17. 17. V. A. Shchukin, N. N. Ledentsov, J. A. Lott, et al., “Ultra high-speed electro-optically modulated VCSELs: Modeling and experimental results,” Proc. SPIE 6889, 68890H (2008).
18. 18. A. Paraskevopoulos, H.-J. Hensel, W.-D. Molzow, et al., “Ultra-high-bandwidth (>35 GHz) electrooptically-modulated VCSEL,” Proc. Optical Fiber Commun. Conf. (OFC/NFOEC) (2006), pp. 1–3.
19. 19. Y. A. Wu, G. S. Li, R. F. Nabiev, et al., “Single-mode, passive antiguide vertical cavity surface emitting laser,” IEEE J. Sel. Top. Quantum Electron. 1, 629–637 (1995).
20. 20. L. Bao, N. H. Kim, L. J. Mawst, et al., “Near-diffraction-limited coherent emission from large aperture antiguided vertical-cavity surface-emitting laser arrays,” Appl. Phys. Lett. 84, 320–322 (2004).
21. 21. L. Bao, N. H. Kim, L. J. Mawst, et al., “Single-mode emission from vertical-cavity surface-emitting lasers with low-index defects,” IEEE Photonics Technol. Lett. 19, 239–241 (2007).
22. 22. D. R. Scifres, W. Streifer, and R. D. Burnham, “Leaky wave room temperature double heterostructure GaAs/GaAlAs diode laser,” Appl. Phys. Lett. 29, 23–25 (1976).
23. 23. V. P. Kalosha, D. Bimberg, and N. N. Ledentsov, “Leakage-assisted transverse mode selection in vertical-cavity surface-emitting lasers with thick large-diameter oxide apertures,” IEEE J. Quantum Electron. 49, 1034–1039 (2013).
24. 24. V. A. Shchukin, N. N. Ledentsov, J.-R. Kropp, et al., “Engineering of optical modes in vertical-cavity microresonators by aperture placement: Applications to single-mode and near-field lasers,” Proc. SPIE 9381, 93810V (2015).
25. 25. M. Rozova, J. Pomplun, L. Zschiedrich, et al., “3D finite element simulation of optical modes in VCSELs,” Proc. SPIE 8255, 82550K (2012).
26. 26. S. Adachi, “GaAs, AlAs, and Al_xGa_1−xAs: Material parameters for use in research and device applications,” J. Appl. Phys. 58, R1–R29 (1985).
27. 27. J.-W. Shi, W. C. Weng, F. M. Kuo, et al., “Oxide-relief vertical-cavity surface-emitting lasers with extremely high data-rate/power-dissipation ratios,” Proc. Optical Fiber Commun. Conf. (OFC/NFOEC) (2011), pp. 1–3.
28. 28. J.-W. Shi, C. C. Wei, J. J. Chen, et al., “850-nm Zn-diffusion vertical-cavity surface-emitting lasers with oxide-relief structure for high-speed and energy-efficient optical interconnects from very-short to medium (2 km) reaches,” Proc. SPIE 9381, 93810E (2015).
29. 29. J.-W. Shi, C. C. Wei, J. J. Chen, et al., “High-performance Zn-diffusion 850-nm vertical-cavity surface emitting lasers with strained InAlGaAs multiple quantum wells,” IEEE Photonics J. 2, 960–966 (2010).
30. 30. J. S. Gustavsson, Å. Haglund, J. Bengtsson, et al., “Dynamic behavior of fundamental-mode stabilized VCSELs using shallow surface relief,” IEEE J. Quantum Electron. 40, 607–619 (2004).
31. 31. D. Zhou and L. J. Mawst, “Two-dimensional phase-locked antiguided vertical-cavity surface-emitting laser arrays,” Appl. Phys. Lett. 77, 2307–2309 (2000).
32. 32. D. F. Siriani, S. P. Carney, and K. D. Choquette, “Coherence of leaky-mode vertical-cavity surface-emitting laser arrays,” IEEE J. Quantum Electron. 47, 672–675 (2011).