1. INTRODUCTION

Over the past 30 years, there have been remarkable advances in the technology of vertical cavity surface-emitting lasers (VCSELs) emitting in the near-infrared (NIR, $\lambda = {{800}} {-} {{1000}}\;{\rm{nm}}$) range using AlGaAs structures. Many applications have been enabled by the low-cost coherent sources, including short-haul fiber links in data centers and local networks, mobile face recognition, commercial and industrial machine vision, and light detection and ranging (lidar) [1,2]. Compared to the more mature edge-emitting lasers (EELs), VCSELs offer appealing features, including wafer-level manufacturing and testing for high-yield mass production, flexible configuration into 1D or 2D patterns, a circular beam profile, and scalability to address both low- and high-power applications. Combined with recent breakthroughs in artificial intelligence (AI), VCSEL-based 3D sensing is destined to play a pivotal role in autonomous vehicles, robotics, augmented/virtual reality, manufacture quality control, surveillance, and security. At the same time, there is also a growing recognition that the NIR wavelength (800–1000 nm) may not be the most suitable wavelength for laser-based 3D sensing, in terms of the achievable signal-to-noise ratio (SNR). Specifically, the laser emission in the NIR range can cause significant ocular damage and is regulated with a very low permissible exposure threshold (Fig. 1, light blue solid line) [3], thus limiting the signal strength in many use cases. At the same time, solar blackbody irradiation contributes inevitable background noise to NIR emissions, even at the 940 nm ${{\rm{H}}_2}{\rm{O}}$ absorption window (Fig. 1, orange solid curve).

 

Fig. 1. Solar irradiance spectrum at sea level (orange), and maximum permissible exposure (MPE) of a single pulse with a pulse width of 1 ns–1 µs (light blue).

Download Full Size | PDF

Both limitations in signal and noise can be mitigated if the operating wavelength is moved to the longer, short-wavelength infrared (SWIR) wavelength (1300–2000 nm) [4], for example to 1380 nm, where the solar background becomes negligible and, simultaneously, the eye safety factor is relieved by more than 1 order of magnitude (Fig. 1). Additionally, VCSELs operating in SWIR wavelengths can pave the way for transformative applications in noninvasive health monitoring [5,6], chemical/biosensing, silicon photonics [7–10], and data communication [11,12]. It is also important to note that InP-based edge-emitting and DFB lasers emitting in SWIR wavelengths have been a key enabling factor for fiber-optic communication and the Internet [13]. However, SWIR VCSELs have yet to reach technological viability, despite 30 years of sustained research and development efforts. Before describing our solution of a viable InP SWIR VCSEL, we will briefly summarize the 30-year history and prior arts of SWIR VCSELs so their challenges and significance can be placed into perspective.

Crucial building blocks for VCSEL devices include (i) an active region that can provide high optical gain; and (ii) high-reflectivity distributed Bragg reflector (DBR) mirrors at the bottom and top to form a vertical cavity. The success in NIR VCSEL technology can be attributed to the unique material properties in the GaAs–AlAs alloy system. GaAs and AlAs are almost lattice-matched to each other, yet with a remarkable contrast in the index of refraction ($\Delta n = {0.5}$). Highly reflective GaAs/AlAs DBR can be prepared in a homoepitaxial manner on GaAs substrates. Meanwhile, active regions with high optical gains in the NIR wavelength range can be epitaxially grown on the GaAs/AlAs DBRs. When it comes to SWIR VCSELs [14], the most developed active regions (InGaAlAs or InGaAsP) have to be prepared on InP substrates. Two common epitaxial DBRs grown on the InP substrate, InGaAsP/InP and InAlGaAs/InP [15,16], do not support a high index contrast ($\Delta n \sim {0.2}$); more than 60 pairs of the ${{1/4}}\lambda$ layers are required for these DBRs to reach a reflectance of 99.9%. The Sb-based quaternary AlGaAsSb DBR on InP supports a greater contrast in optical index [17], but Sb-based epitaxy has not reached the state of maturity compared with As- and P-based compound semiconductors [16]. To circumvent the challenges in heteroepitaxial DBRs on InP, other approaches have also been pursued, including (1) growing long-wavelength active regions on GaAs substrates (InAs QDs, and dilute nitrides) [18–21]; (2) employing InP/airgap DBR [16]; (3) double-dielectric DBR [22,23]; and (4) wafer fusion of InP active region with AlGaAs DBRs [24–26]. It is remains unclear, however, whether any of the aforementioned approaches provides a pathway to scalable and mass-producible SWIR VCSELs.

This present study provides an alternative pathway of creating highly uniform, monolithic, and homoepitaxial DBRs on InP substrates. By manipulating the index contrast through photonic nanostructures, we can achieve near-unity reflection with around 10 pairs. This DBR design enables continuous-wave (CW) operation of SWIR VCSELs at milliwatt power levels and offers a viable route toward manufacturable and reliable SWIR VCSELs.

2. RESULTS AND DISCUSSION

A. Engineering the Refractive Index of InP through Sub-Wavelength Porosification

We address fundamental challenge of forming a manufacturable DBR structure on InP with a sufficient index contrast to support SWIR VCSELs by transforming InP into a nanoporous form in a controlled manner that allows us to precisely tune the optical index. Porous nanostructures have been demonstrated in a wide range of electronic materials, including silicon [27], anodic aluminum oxide (AAO) [28,29], GaN [30–32], GaAs [33], and InP [34–36]. When the size of the pores is much smaller than the wavelength of interest, the porous medium can be considered as an optically homogeneous medium, with an index of refraction that is tunable as a function of porosity (effective medium approximation) [37].

Electrochemical (EC) porosification is a field-driven process that happens selectively and locally around the pore tips, where the electric field is highest due to the local curvature. The etching proceeds by the injection of holes across the electrolyte/semiconductor interface (to initiate the chemical etching), a method that allows more control compared to other techniques to create nanoporous semiconductors. Similar to current transport in a reverse-biased tunnel diode (TD), the rate and morphology of EC etching depends sensitively on two parameters: doping concentration and bias voltage [30]. By varying the doping concentrations, we can design InP layered structures with precisely tailored porosities and index profiles.

To examine the idea of conductivity-selective porosification of InP, we prepared an InP structure by metal–organic chemical vapor deposition (MOCVD), which is schematically shown in the top panel of Fig. 2(a). The structure consists of eight pairs of alternating undoped (${\rm{n}} \lt {{1}} \times {{1}}{{{0}}^{17}}\;{\rm{c}}{{\rm{m}}^{- 3}}$, light gray) and ${{\rm{n}}^ +}$-InP (${\rm{n}} = {{5}} \times {{1}}{{{0}}^{18}}\;{\rm{c}}{{\rm{m}}^{- 3}}$, dark gray) layers grown on an n-type InP substrate. The design is such that the EC porosification takes place selectively in the doped n-type conductive layers only. Next, the InP wafer was patterned with inductively coupled plasma (ICP) etching to create trenches to expose the sidewalls of the ${{\rm{n}}^ +}$-InP layers [middle panel in Fig. 2(a)]. Subsequently, we immersed the patterned InP wafer with a metal contact formed on the substrate (anode) and a platinum (Pt) counter electrode (cathode) together in a dilute hydrochloric acid (HCl, 5%–8%), and performed EC porosification of InP by applying an anodization bias (1.5–2.5 V) between two electrodes. EC porosification of the InP wafer is illustrated in the bottom panel of Fig. 2(a). Ideally, EC etching initiates from both edges of a stripe, and the etching front [blue dashed line in the bottom panel of Fig. 2(a)] from each end moves gradually to the center of the stripe. Cleaved perpendicular to the EC etching direction, cross-sectional SEM images of this structure before [Fig. 2(b)] and after [Fig. 2(c)] the porosification reveal a well-controlled and selective porosification process for ${{\rm{n}}^ +}$-InP layers [dark regions in Fig. 2(b), marked by the yellow dashed lines], while the undoped InP layers [bright regions in Fig. 2(b), marked by the incarnadine dashed lines] remain intact [Fig. 2(c)]. From a comparison of the periodic layer thickness before [Fig. 2(b)] and after [Fig. 2(c)] porosification, no noticeable change in thickness of the NP layer was observed, with an accuracy to within 1%. Since the InP DBR layers were prepared on native InP substrates, we do not anticipate any stress generation or relaxation as a result of porosification. It is also worth mentioning that an InP layer with doping concentration of ${{5}} \times {{1}}{{{0}}^{17}}\;{\rm{c}}{{\rm{m}}^{- 3}}$ (not shown) between the substrate and the periodic structure was also not attacked under the current EC etching condition. Figure 2(c) is remarkable in rendering the first evidence of conductivity-selective porosification for InP. Photonic design and index engineering can therefore be implemented through EC porosification, which is in turn controlled effectively and precisely down to nanometer-scale based on the semiconductor doping.

 

Fig. 2. (a) Schematics of the process flow in making NP-InP DBR; upper panel, the process starts with an InP epitaxial structure with eight pairs of alternating undoped and ${{\rm{n}}^ +}$-InP layers; middle panel, trenches are formed through a patterned dry-etching process to expose the sidewall of ${{\rm{n}}^ +}$-InP layers; bottom panel, subsequently, EC etching is performed where the porosification process starts from the both edges of stripes and gradually moves to the center (incarnadine arrow). The blue dashed lines represent the etching front. Cross-sectional SEM images of the same sample before and after the EC porosification process are shown in (b) and (c), respectively. The bright and dark bands in (b) correspond to undoped and ${{\rm{n}}^ +}$-doped InP regions, respectively. These two images reveal a well-controlled and selective porosification process for ${{\rm{n}}^ +}$-InP layers, while the undoped InP layers remain intact. (d) A plan view optical image of the porosified sample; (e) microreflectance is measured (black) within the region circled with a white dashed line in (d). The orange dashed line is a simulated reflectance spectrum. The high correlation (overlap) between the measured and simulated spectra confirms a nearly ideal DBR reflectance centered around 1150 nm.

Download Full Size | PDF

Figure 2(d) shows a plan-view optical image of the porosified sample, where the yellow-gold vertical stripes correspond to ICP-etched trenches. The electrochemistry proceeded in-plane along the directions labeled by the incarnadine arrows. The two EC etching fronts, one moving left from the right trench and the other moving right from the left trench, did not merge, leaving a region (tan-colored center stripe) that is non-EC-etched. The light-green regions, on the other hand, correspond to the regions where the ${{\rm{n}}^ +}$-InP layers have been selectively porosified into multilayered porous structures. Using microreflectance taken approximately around the white dashed circle [Fig. 2(d)], we measured a reflectance spectrum centered around 1150 nm (black solid line in Fig. 2(e)]. We note that the structure in Fig. 2(a) was procured only for initial EC etching and was not designed accurately to meet a specific Bragg condition as a reflector. Nevertheless, given the physical dimensions of the porous and nonporous InP layers, we were able to achieve good fitting [orange dashed line in Fig 2(e)]. Here, a known chromatic dispersion of refractive index was employed for the InP layer [38], and the index of NP–InP layer was assumed to scale proportionally with InP. The refractive indices of the InP and NP–InP layer at 1150 nm are around 3.2 and 2.2, respectively. Furthermore, the use of NP–InP introduces an unprecedented tunability in the optical index ($\Delta {\rm{n}}\sim{{1}}$) that has never been possible from heteroepitaxy. The porosification process is very fast, with a lateral porosification rate of 10 µm/min. The homogeneity in color over ${\gt}{{100}}\;\unicode{x00B5}{\rm m}$ in Fig. 2(d) indicates the overall large-area uniformity across the wafer surface. A careful investigation using spatially resolved microreflectance measurements (not shown) indicates a slight spatial gradient in the refractive index of the NP–InP layer along the pore propagation direction. The gradient or variation in the index of refraction (${\rm{d}}{{\rm{n}}_{\textit{NP}}}/{{\rm{n}}_{\textit{NP}}}$) was determined to be ${\sim}{0.3}\%$ over 10 µm, which is expected to be reduced with a further optimization in the experimental procedure. By using post-growth, conductivity-selective porosification to transform ${{\rm{n}}^ +}$-InP layers to NP-InP layers, we provided a promising solution to overcome the impasse of making highly-reflective DBRs on InP substrates, and reduced this decades-old challenge into a homoepitaxial exercise followed by a uniform, wafer-level EC process that is in principle manufacturable.

Figure 3 provides a comparison between the homoepitaxial NP–InP DBR, and previously reported DBRs by heteroepitaxy on InP substrates at 1300 nm [15–17]. The great index contrast ($\Delta {\rm{n}}\sim{{1}}$) in this work allows us to achieve DBRs with a near-unity reflectivity using ${\sim}{{10}}$ pairs. Compared with heteroepitaxial DBRs (excluding InP/Air), the required number of pairs for NP–InP DBRs is reduced by a factor of 3–5, which improves the manufacturability and performance of SWIR VCSELs. The InP/Air DBR is also included in Fig. 3 for completeness, but this mirror suffers from structural instability [14].

 

Fig. 3. Plot of index contrast and the corresponding required number of pairs to achieve near-unity reflection DBR, centered at 1300 nm. These DBRs were assumed to be sandwiched between InP-based cavity layers (above DBR) and InP substrates. Details of the calculation are given in Supplement 1, Section 1. Green dots are epitaxial DBRs on an InP substrate, and the red star is from this work. We note the AlAs/GaAs DBR grown on the GaAs substrate (not shown) locates near the point of AlGaAsSb/AlAsSb. The dashed curve is the theoretical relation between the index contrast and required number of pairs to reach 99.9% reflectance. Here, we set the refractive indices of the high-index and low-index layer to be 3.2 and 3.2-${\Delta}{\rm n}$, respectively.

Download Full Size | PDF

B. Design and Fabrication of SWIR NP–InP VCSEL Structure

We conducted a proof-of-concept demonstration wherein we developed and produced two NP–InP SWIR VCSELs operating at 1380 and 1550 nm, respectively. Both VCSELs featured a similar device structure, as illustrated in Fig. 4(a). The bottom mirror consisted of NP–InP DBR, while the top mirror was made up of dielectric DBR, centered at respective wavelengths with a calculated optical power reflectance of 99.35%. Before the deposition of quarter-wavelength top dielectric DBR, a 5 nm amorphous silicon (a-Si) buffer layer and a hydrogenated a-Si (a-Si:H) spacer layer were deposited on devices as part of the 3-$\lambda$ cavity [shown in top-right panel of Fig. 4(a)].

The active region of the VCSELs included five compressively strained AlGaInAs quantum wells (QWs), separated by tensile-strained AlGaInAs quantum barriers. Material compositions and thicknesses of QWs and quantum barriers are given in Supplement 1, Table S1. To confine the current, we utilized proton implantation, which selectively destroyed the conductivity of p-InAlAs outside the designated current aperture region [shown in the top-right panel of Fig. 4(a)]. Devices with four different aperture sizes were fabricated (7, 10, 13, and 16 µm in diameter). A tunnel junction (TJ) was incorporated, consisting of 15 nm of ${{\rm{p}}^{+ +}}$-InAlAs ($[{\rm{C}}] \gt {{2}} \times {{1}}{{{0}}^{19}}\;{\rm{c}}{{\rm{m}}^{- 3}}$) and 15 nm of ${{\rm{n}}^{+ +}}$-InP ($[{\rm{Si}}] \gt {{2}} \times {{1}}{{{0}}^{19}}\;{\rm{c}}{{\rm{m}}^{- 3}}$), positioned above p-InAlAs. The active region, p-InAlAs, and TJ were sandwiched between two n-InP layers. The detailed design of the VCSEL cavity and the fabrication process are given in Supplement 1, Sections 2 and 3, respectively.

To ensure optical confinement, a shallow surface relief was created just above the current aperture on the surface of the top n-InP, demonstrated in the top-right panel of Fig. 4(a). This relief was realized through reactive-ion etching, providing in-plane index guiding [39]. Although the n-InP layer above the TJ also underwent implantation outside the current aperture region, its conductivity was successfully restored by employing appropriate post-implantation annealing.

C. Results and Analysis of SWIR NP–InP VCSELs

Figure 4(b) displays a top-view microscope image with 20 fabricated VCSELs in the field of view, and a zoom-in image [Fig. 4(c)] of a VCSEL with 7 µm aperture. Fully fabricated VCSEL devices were tested at wafer level on a probe station. All the testing was performed under continuous-wave (CW) operation. The current source used for the characterization is Keithley 2400, and the optical power was measured with a photodiode (Thorlab S122C) positioned directly above the device. The photodiode with factory calibration was further calibrated with a commercial 1550 nm EEL to ensure accuracy. A lasing spectrum was collected with an optical spectrum analyzer (Thorlab OSA 203). Figures 4(d)–4(g) present the light-current-voltage (L-I-V) characteristics and optical spectra of 1380 and 1550 nm VCSELs with a current aperture of 7 µm in diameter.

Figure 4(d) displays the L-I plots (solid lines) from a 1380 nm VCSEL under CW operation for stage temperatures of 25°C–85°C, and I-V relation of the same device at 25°C and 85°C (dashed lines). At room temperature (RT, 25°C), a clear nonlinear increase of output power was observed at around 0.7 mA, corresponding to a threshold current density of ${1.8}\;{\rm{kA/c}}{{\rm{m}}^2}$. The extracted slope efficiency (SE) is approximately 0.42 W/A, equivalent to a differential quantum efficiency (DQE) of ${\sim}{{47}}\%$. The power conversion efficiency (PCE) of the present device reaches a maximum of 17% at an injection current of 2.5 mA. Figure 4(e) presents the optical spectra around the lasing threshold under RT. At or below 0.5 mA, only the background noise was collected. Increasing the input current at a step of 0.1 mA, a peak appeared at around 1381 nm, and increased dramatically afterwards. Such nonlinear behavior from spectra corroborates with the one demonstrated in the L-I curve [Fig. 4(d)]. Laser operation was possible up to 85°C for this VCSEL, as shown in Fig. 4(d).

 

Fig. 4. (a) Left panel, Schematic of the NP-InP VCSEL structure with the current aperture formed by proton implantation; upper-right panel, magnified illustration of the upper part of the VCSEL cavity; bottom-right panel, enlarged-view of the NP-DBR region; (b) plan view optical image of fabricated VCSELs; each device has a dimension of ${{150}}\;{\rm{\unicode{x00B5}{\rm m}}} \times {{150}}\;{\rm{\unicode{x00B5}{\rm m}}}$. (c) Zoom-in image of a VCSEL with 7 µm aperture; (d) solid lines, L-I relation of the 1380 nm VCSEL with 7 µm aperture for temperature of 25°C–85°C; dashed lines, I-V characteristics of the same VCSEL at 25°C and 85°C; (e) optical spectra of the 1380 nm VCSEL with 7 µm aperture near lasing threshold at 25°C; (f) solid lines, L-I relation of the 1550 nm VCSEL with 7 µm aperture for temperature of 25°C–85°C; dashed lines, I-V characteristics of the same VCSEL at 25°C and 85°C; (g) optical spectra of the 1550 nm VCSEL with 7 µm aperture near lasing threshold at 25°C.

Download Full Size | PDF

CW operation of a 1550 nm VCSEL up to 85°C was also achieved from a separate InP VCSEL structure. The temperature-dependent L-I-V curve and optical spectra at 25°C of a 7 µm aperture device are presented in Figs. 4(f) and 4(g). A similar nonlinear increase in the output power was identified starting at 0.7 mA in the L-I plot of Fig. 4(f) at 25°C, equivalent to a threshold current density of ${1.8}\;{\rm{kA/c}}{{\rm{m}}^2}$. The nonlinear behavior was also observed in the spectra close to threshold, shown in Fig. 4(g). From L-I plot in Fig. 4(f), we extracted an SE of approximately 0.4 W/A at 25°C, and the DQE reached 50%. Below 65°C, we observed similar behavior and performance of the device L-I-V curves between both wavelengths (1380 and 1550 nm), which provides strong evidence of the high reproducibility of NP–InP DBR VCSELs. When the stage temperature is 75°C and higher, 1380 nm VCSEL exhibited thermal rollover under a lower injection current. This can be explained by the fact that the same quantum barriers (QBs) were employed in both VCSELs, but the 1380 nm VCSEL was expected to experience more serious carrier leakage from QWs under a high operating temperature due to shallower barriers.

Figures 5(a) and 5(b) represent lasing spectra well above the lasing threshold for 1380 and 1550 nm VCSELs at 25°C, respectively. Single-mode operations with a side-mode-suppression ratio (SMSR) close to 30 dB were maintained, and the laser linewidth is at the spectral resolution of OSA 203 (60 pm) for both VCSELs. In addition, a coherence length of ${\gt}{{80}}\;{\rm{mm}}$ (detection limit of OSA 203) was measured for these two VCSELs. It is worth mentioning that for VCSELs with an aperture size of 10 µm and larger, multimode lasing was observed at both wavelengths (not shown). The systematic redshift in Figs. 5(a) and 5(b) with an increasing current is due to device heating (at a given operation temperature). Heat transfer or thermal resistance (${{\rm{R}}_{\rm{th}}}$) in a nanoporous medium is a point of potential concern, especially when the wall thickness approaches the phonon mean-free path [40,41]. For the initial assessment in this report, we chose to determine the baseline parameters of heat transport in operando. The dependencies of emission wavelength on the stage temperature at a constant dissipated power, and of emission wavelength on dissipated power at a constant stage temperature are presented in Fig. S3 for 1380 and 1550 nm VCSELs with 7 µm aperture. Through a linear fitting, a thermal tuning coefficient of the emission wavelength ($\Delta \lambda /\Delta {\rm{T}}$) of ${\sim}{0.1}\;{\rm{nm/K}}$ at a constant dissipated power, and a power tuning coefficient ($\Delta \lambda /\Delta {{\rm{P}}_{\rm{diss}}}$) of 0.26 nm/mW at a constant stage temperature (room temperature) were experimentally determined. Based on the expression of ${{\rm{R}}_{\rm{th}}}$,

 

Fig. 5. (a) Single-mode lasing spectra of 1380 nm VCSEL with a 7 µm aperture under an injection current ranging from 1 to 6 mA; (b) single-mode lasing spectra of 1550 nm VCSEL with a 7 µm aperture under an injection current ranging from 1 to 6 mA; the side-mode-suppression ratio for both VCSELs is above 30 dB.

Download Full Size | PDF

Finally, we wish to comment on the reliability of VCSELs with nanoporous InP DBRs. In Fig. S4, we showed a preliminary reliability test of a 1550 nm NP–InP VCSEL device with a 7 µm aperture. The operating current was fixed at 3 mA. Over a span of 100 h, we did not observe any noticeable change in optical power.

3. CONCLUSIONS

VCSELs emitting in the 1300–1600 nm range have been pursued actively for almost 30 years, yet a technology-viable solution remains elusive. In this work, we proposed and demonstrated the use of EC etching as a new solution to realize SWIR VCSELs. The results reported here, including RT CW operation with a threshold current density of below ${{2}}\;{\rm{kA/c}}{{\rm{m}}^2}$, milliwatt output power, a power-conversion efficiency of 17%, and an initial life test of greater than 100-h continuous operation with no degradation, are compelling evidence that the nanoporous InP DBR is a promising pathway to overcome the 30-year impasse. The demonstration of SWIR VCSEL operations in two distinct and important wavelengths (1380 and 1550 nm) demonstrate the versatility of the InP NP DBR designs. SWIR wavelengths have been identified in recent years to possess unique properties for 3D sensing, ranging, data communication, biophotonics, and silicon photonics. The availability of mass-producible VCSELs in this wavelength range is expected to enable many technologies and applications.