1. Introduction

Vertical-cavity surface-emitting lasers (VCSELs) have broad applications in data communication, sensing, and consumer electronics. VCSELs have many advantages, such as circular beam, low power consumption, high modulation speed, low cost, high reliability, and easy fabrication of two-dimensional array configurations [1–4]. High-speed, energy-efficient, and temperature-stable VCSELs for data centers and high-performance computer systems, have made enormous progresses in recent years [3,4]. Recently high-efficiency VCSELs have also been developed for face recognition, flood illumination, and LiDAR [5,6]. Especially, VCSELs with a narrow linewidth are very promising laser sources for atomic magnetometer [7], coherent communication [8], and frequency standards [9], and have attracted much attention.

Typically, a VCSEL structure is composed of two distributed Bragg reflectors (DBRs), multiple quantum wells, and an oxide aperture for current and optical mode confinements. DBRs have a high reflectivity over a broad wavelength range because of multiple reflections with constructive interference among the waves reflected from different interfaces between high-index and low-index layers. Dozens of DBR layers are needed to achieve the required reflectivity being larger than 99.5% for the VCSEL application, making the VCSEL bulky. For DBR-based optoelectronic devices at blue-green or infrared wavelengths [10,11], it is not easy to grow DBRs.

Metastructures like high-index-contrast gratings (HCGs) have attracted much attention in the past few years [12–20]. A typical HCG has a thickness of a few hundreds of nanometers and has a nearly 100% reflectivity across a broad wavelength range [12], and can be used as a mirror in a VCSEL. HCG-VCSELs are now working at wavelengths of 850 nm [18], 980 nm [19,21], 1060 nm [22,23], and 1550 nm [24–26].

The previously reported HCG-VCSELs always employ suspended subwavelength gratings surrounded by air to realize a high index contrast for mirrors. The fabrication of such suspended HCGs is complex because selective etching and grating release process are required. Suspended HCGs in addition suffer the problem of mechanical stability, and device lifetime might be shortened, because the low thermal conductivity of the air can negatively affect the overall performance of devices.

Some attempts have been made to realize HCG bars sitting on a low-index medium, without selective etching and a grating release process [27–29]. GaAs is a well researched material for VCSELs covering a wide wavelength range. GaAs-based HCGs with an AlO_x layer as the low-index medium are highly desirable for a mirror used in VCSELs. Almuneau et al. [27] used 150-nm AlAs to create AlO_x by wet oxidation as the low-index medium below a GaAs-based grating. This thickness of 150 nm was carefully chosen after considering the mechanical strain by the volume shrinkage of the AlAs layer.

The thickness of the low-index medium is an important parameter in HCG-VCSELs. A small thickness of the low-index medium can cause evanescent coupling between the HCG layer and the layer under the low-index medium, resulting in a low reflectivity of the part above the active region in the HCG-VCSEL. Thus, 5/4*λ and 1/2*λ air gaps were used for the HCG-VCSELs demonstrated in [18,19]. A thick (1/2*λ and beyond) low-index medium like AlO_x is preferred to realize a long cavity in the HCG-VCSEL with a narrow linewidth [30]. Thus, when GaAs-based HCG bars are sitting on a thick AlO_x layer as the low-index medium, realization of such thick AlO_x layers is challenging.

In this work, we present a HCG based on GaAs/AlO_x layers. The HCG doesn’t require selective etching and grating release process which are used in [18,19]. The subwavelength grating is defined on the GaAs layer on a 640-nm Al_0.9Ga_0.1As layer oxidized to AlO_x (n = 1.6) with a layer contraction of 4.9%. The thickness of the low-index AlO_x layer is larger than the one in [27], and the HCG here has a better mechanical stability. The HCG shows a high reflectivity, consistent with the simulation results, and can be used as the mirror in narrow-linewidth VCSELs.

2. Design and fabrication

The Aluminum content and thickness of an AlGaAs layer affect the oxidation rate for given oxidation conditions. A high Aluminum content and a thick AlGaAs lead to a large oxidation rate [31]. There is a layer contraction of more than 6.7% for AlGaAs for an Aluminum content of 98% after wet oxidation, introducing considerable strain [32]. This strain may lead to layer delamination when the thickness of an Al_0.98Ga_0.02As layer is too large. We tried a 600-nm Al_0.98Ga_0.02As layer underneath the GaAs layer and layer delamination was observed. Thus, Al_0.9Ga_0.1As layers for oxidation were adopted for less strain after wet oxidation.

The schematic and a scanning electron microscope (SEM) image of one of our HCGs are shown in Fig. 1. The grating period is Λ, and the width of the grating bar is a. The duty cycle (DC) is defined as a/Λ, and t_g is the thickness of the grating. The reflectivity contour of the HCG is calculated by the rigorous coupled wave analysis method. Results are shown in Fig. 2(a) for the normalized wavelength (λ/Λ) and the normalized thickness (t_g/Λ) of the grating for transverse magnetic polarization (TM, the electric field is perpendicular to the grating bar). The thickness and period of the HCGs are 362 nm and 560 nm, respectively.

 

Fig. 1. (a) Schematic of the HCG. The grating period is Λ; a is the width of the grating bar; the duty cycle (DC) is defined as a/Λ; and t_g is the thickness of the grating. (b) SEM image of the HCG.

Download Full Size | PDF

 

Fig. 2. (a) Reflectivity contour of the HCG as a function of normalized thickness (t_g/Λ) and normalized wavelength (λ/Λ) under normal incidence at DC = 0.55. (b) Reflectivity contour of the HCG as a function of wavelength and thickness of the AlO_x layer.

Download Full Size | PDF

Figure 2(b) shows the reflectivity contour of a HCG for a thickness of the AlO_x layer less than 1 µm. The reflectivity spectrum is periodically varying when the thickness of the AlO_x layer varies. When the thickness of the AlO_x layer is around 605 nm (∼3/4*λ), the reflectivity spectrum has its maximum bandwidth for a high reflectivity (more than 99.5%) for TM polarization. Considering the estimated layer contraction for the Al_0.9Ga_0.1As layer after wet oxidation, the thickness of the Al_0.9Ga_0.1As layer before oxidation is chosen to 640 nm. The HCG has a high reflectivity across the whole 1300-nm range.

The fabrication of the HCG is similar to that in [33] and shown in Fig. 3. A Si₃N₄ layer is first deposited on the wafer by plasma-enhanced chemical vapor deposition (PECVD). PMMA electron resist is put on the top and then electron beam lithography (EBL) is used for defining the grating pattern. After development, the grating pattern is transferred to the Si₃N₄ layer by reactive ion etching (RIE) with a gas mixture of CHF₃/O₂. The grating structure is then transferred to the GaAs layer with the patterned Si₃N₄ layer as the mask by inductively coupled plasma (ICP) etching with a gas mixture of Cl₂/BCl₃/Ar. Then the wafer is loaded to the wet oxidation furnace. The Al_0.9Ga_0.1As layer beneath GaAs layer is oxidized at 420 degrees by the water vapor flowing through the air gaps of the HCG. The oxidation rate is isotropic for Al_0.9Ga_0.1As [34] and about 3 nm/min for the test structure. The SEM images of the cross section of the test structure and of the HCG are shown in Fig. 4 and Fig. 1(b), respectively. In Fig. 4, we can find that compared with the unpatterned GaAs region the GaAs-based bars in the patterned region shifts down by 21 nm after oxidation, which is caused by the layer contraction. Thus, we conclude that there is a 4.9% layer contraction for Al_0.9Ga_0.1As layer after wet oxidation. Thus, the layer contraction of Al_0.9Ga_0.1As is much smaller than that of Al_0.98Ga_0.02As after wet oxidation [31], resulting in a more stable AlO_x layer. The HCG has a high fabrication tolerance. The simulation results show that the reflectivity of the HCG at 1300-nm range is enhanced slightly when the Al_0.9Ga_0.1As layer is partly etched as shown in Fig. 4.

 

Fig. 3. Schematic of fabrication process of the HCG. PMMA, poly methyl methacrylate; EBL, electron beam lithography; RIE, reactive ion etching; ICP, inductively coupled plasma.

Download Full Size | PDF

 

Fig. 4. SEM image of the cross section of a test structure for measuring the layer contraction and oxidation rate.

Download Full Size | PDF

3. Results and discussions

The reflectivity spectrum of the HCG is measured with the setup reported in [33]. The broadband amplified spontaneous emission (ASE) (1270-1330 nm) from a praseodymium-doped fiber amplifier (PDFA) is collected by a fiber and then collimated by a lens. A Glan-Thomson polarizer polarizes the light. Finally, the light is focused on the HCG by an objective lens with 50× magnification and numerical aperture (NA) of 0.45. The reflected light is collected by another fiber via a planar beam splitter and then measured by an optical spectrum analyzer (OSA). The reflected light is partly led by a cube beam splitter into a charge-coupled device (CCD) camera to observe the position of the HCG. The reflectivity of a gold-coated mirror is measured for a reference. The reflectivity of the HCG is normalized to the reflectivity of the gold-coated mirror. The measured reflectivity spectrum of the HCG is shown in Fig. 5. The measured reflectivity of the processed HCG agrees well with the simulated reflectivity spectrum calculated by two-dimensional finite difference time domain (2D-FDTD) method for the finite-size HCG. The discrepancy between the simulated reflectivity spectrum and the measured result arises from the angular components of the focused incident beam with a finite size [35].

 

Fig. 5. Simulated and experimental results of the HCG (the measured width of the HCG bars is a = 290 nm). The simulated reflectivity spectrum is calculated by the 2D-FDTD method for a finite-size HCG.

Download Full Size | PDF

In the measurement, the incident beam at normal incidence is a finite-size and focused beam and consists of a lot of angular components (i.e. a spread of wave-vector k points) after the objective lens. The angular components of the incident beam shape the reflectivity spectrum of the HCG, because the reflectivity of the HCG depends on the incident angle [33,35]. As an approximation, the average parallel wave-vector of the mode excited by a finite-size incident beam is inversely proportional to the lateral size of the incident beam, ${k_{{\parallel} \textrm{mod}}} \propto \frac{1}{L}$, where L is the lateral size of the incident wave. The focused incident beam after the objective lens (NA: 0.45) has a divergence angle, also resulting in a spread of wave-vector k points. The average parallel wave-vector of the incident beam is proportional to the divergence angle of the incident beam, ${k_{{\parallel} \textrm{source}}} \propto \theta$, where θ is the divergence angle of the incident beam [36,37]. Figure 6 shows the reflectivity spectra of the HCG under different incident angles (φ) and angular dependence of the HCG reflectivity. As the incident angle increases, the maximum reflectivity shifts to the longer wavelength side. The reflectivity at the shorter wavelength side is reduced, and the reflectivity at the longer wavelength side is increased. The reflectivity is sensitive to the incident angle [38]. The reflectivity is increased and then reduced when the incident angle varies from 0 degree to 30 degrees. In our experiments, the finite-size and focused incident beam can be considered as a sum of plane incident waves with different incident angles (i.e. higher-order angular components). The measured reflectivity of the HCG can be approximately expressed by $R = \int {{F^2}(\varphi ,\lambda )} R(\varphi )d\varphi$, where $F(\varphi ,\lambda )$, $\varphi$, and $R(\varphi )$ are the angular spectrum, the incident angle, and the angle-dependent reflectivity for a plane wave, respectively [39]. The higher-order angular components in the incident beam in the experiment and the imperfect sidewalls of the HCG may contribute to the discrepancy between the simulated reflectivity spectrum and the measured result, shown in Fig. 5.

 

Fig. 6. (a) Reflectivity spectra of the HCG with plane incident waves with different incident angles ($\varphi$); (b) angular dependence of HCG reflectivity for plane incident waves.

Download Full Size | PDF

4. Conclusions

GaAs-based HCG reflectors with AlO_x layer as the low-index medium oxidized from a 640-nm Al_0.9Ga_0.1As layer were fabricated and measured. The HCG exhibits a reflectivity around 90% across the 1300-nm wavelength range, consistent well the simulation result. The layer contraction of the Al_0.9Ga_0.1As layer is 4.9%, much smaller than that of the Al_0.98Ga_0.02As layer after wet oxidation, resulting in a more stable AlO_x layer for real devices. In the fabrication of the HCG-VCSEL, the realization of the AlO_x layer oxidized from the Al_0.9Ga_0.1As layer beneath the top patterned GaAs layer can be made simultaneously when the oxide aperture is formed. This HCG can be used in novel vertical cavities for narrow-linewidth VCSELs and for studies of fundamental physics like cavity quantum electrodynamics and cavity polaritons [40].