1. Introduction

In 1966, shortly after the advent of semiconductors, COOPER et al. ^[1] discovered that increasing the output power of GaAs homojunction semiconductor lasers to a certain level would result in Catastrophic Optical Damage (COD) and failure. In 1977, CHINONE et al. ^[2] discovered that an AlGaAs/GaAs double heterojunction semiconductor laser operated continuously for a certain period resulted in Catastrophic Optical Mirror Damage (COMD) on its cavity surface. Using Scanning Electron Microscope (SEM) observation, it was found that high power density light output and cavity surface oxidation were important factors leading to its COMD^[3].

For InGaAs/AlGaAs high-power Quantum Well (QW) semiconductor lasers, COMD suppression should start from its induced mechanism ^[4]. According to test results, methods such as reducing non-radiative recombination at the cavity surface, suppressing light absorption of the cavity surface material, lowering the carrier concentration at the cavity surface, and improving the heat dissipation capacity at the cavity surface ^[5] can significantly suppress COMD. The preparation of non-absorbing windows based on Quantum Well Intermixing (QWI) technology is a low-cost and effective method to suppress the light absorption of cavity materials ^[6-7]. Commonly used QWI methods include Impurity Induced Disordering (IID), Impurity Free Vacancy Induced Disordering (IFVD), Laser Induced Disordering (LID), etc. ^[8-11]. Among them, in IID technique, a large number of point defects are induced by introducing impurities, and in combination with thermal annealing and other methods, the impurities and point defects are activated to obtain diffusion kinetic energy, ultimately causing changes in the composition and structure of quantum wells. In the 1980 s, LAIDIG ^[12] first found that QWI phenomenon occurred in AlAs/GaAs superlattice structures with the introduction of Zn impurities and heat treatment, and the heat treatment temperature in this method was only 575 °C, far below the temperature required for impurity free induced disordering. Until 1985, KALISKI ^[13] found that the effect of Si impurity inducing AlGaAs/GaAs superlattice QWI was better than that of other impurities. In 1987, MEI et al. ^[14] used Secondary Ion Mass Spectroscopy (SIMS) to test and found that the diffusion coefficient of Al atoms in AlGaAs materials increased significantly with the diffusion of Si impurities. Comprehensive research results show that Si impurities can form defect pairs with larger diffusion coefficients with Al atoms, and Si impurities can also increase the density of point defects in the QW system, thus effectively promoting the QWI of the AlAs/GaAs superlattice structure ^{[6, 15]}.

This paper presents a Non-Absorbing Window (NAW) preparation scheme for InGaAs/AlGaAs high-power QW semiconductor lasers using the method of Si impurity induced QWI. This method is based on the principle that the Si impurity is used as an induction source, which can efficiently induce the atomic interdiffusion between the materials in the QW and the materials in the barrier of the InGaAs/AlGaAs semiconductor QW laser, eventually broadening the band gap of the active region material and suppressing its absorption of the self-generated laser. The preparation of NAW using the Si IID method not only reduces the optical absorption at the cavity surface of the laser, but also serves as an N-type doping element to form a non-carrier injection region at the cavity surface of the device, thus reducing the non-radiative composite here. This design does not require expensive equipment or complex processing, and can effectively increase the COMD threshold triggering power of the laser without changing its characteristic parameters.

2. Simulation analysis of QWI

2.1 Chip development and performance analysis

The primary epitaxial wafers of the InGaAs/AlGaAs QW laser used in this paper were grown by Metal Oxide Chemical Vapor Deposition (MOCVD), with a reaction chamber growth temperature of 550−700 °C and a pressure of 5 kPa ^[16]. The substrate is n-GaAs with a (100) plane offset [111] A-crystal-orientation of 15°. The schematic diagram of the ridge laser structure formed based on this primary epitaxial wafer is shown in Figure 1(color online).

Figure  1.  Epitaxial structure of InGaAs/AlGaAs QW laser diode

DownLoad: Full-Size Img PowerPoint

For In_(1-x-y)Ga_xAl_yAs quaternary compound semiconductor material, its band gap is shown in formula (1), so the increase of Al component will lead to the increase of E_g. Therefore, we determine whether QWI has occurred in the material by the central wavelength position. If a QWI occurs, it is proved that the Al component has entered the QWI material, and the band gap becomes wider, which is shown by the change of the luminescence wavelength toward the short wavelength, that is, the blue shift occurs.

Photoluminescence (PL) spectroscopy test is a commonly used method to obtain the central wavelength of lasers. The original PL test results of the primary epitaxial wafer of InGaAs/AlGaAs QW lasers in this paper are shown in Figure 2. According to the mapping scan results, the luminescence intensity is uniform, indicating that the composition of each layer of the epitaxial wafer is uniform. From the single-point PL signal peak, it can be seen that the peak center wavelength is 1002.2 nm, and the Full Width at Half Maximum (FWHM) is about 23 nm.

Figure  2.  The PL spectrum of InGaAs/AlGaAs QW primary epitaxial wafer

DownLoad: Full-Size Img PowerPoint

2.2 The effect of temperature on QWI

The existence of point defects in crystals leads to the breaking of the perfect arrangement rules of lattice atoms, changes the vibration frequency of atoms around the defects, increases entropy, and deteriorates the thermodynamic stability^[4]. By combining the diffusion coefficient equation of group III atomic point defects, it can be concluded that:

where ${A}$ is a function related to the vibration entropy ${{S_{{f}}}}$ and vacancy, B is a function related to the vibration entropy ${{S_{{f}}}}$ and interstitial atoms, ${{E_{\rm{I}}}}$ is the energy required to form a interstitial atom, ${{f_1}}$ and ${{f_2}}$ are constants, ${{D_{{\rm{{V_{III}}}}}}}$ is the diffusion coefficient of Group III vacancies, ${{D_{{\rm{{I_{III}}}}}}}$ is the diffusion coefficient of Group III interstitial atoms, ${{K_{\rm{B}}}}$ is the Boltzmann constant, and its value is$1.38 \times {10^{ - 23}} {{\rm{J/K}}}$. Under the thermal equilibrium state, approximation can be considered as: ${ A{f_1}} = {B{f_2}}$, $100{{D_{{\rm{{V_{III}}}}}}} = {{D_{{\rm{{I_{III}}}}}}}$ and $2{{E_{\rm{V}}}} = {{E_1}}$, and the relationship curve between the relative interdiffusion coefficient of group III atoms and temperature can be fitted qualitatively according to formula (2), as shown in Figure 3. It can be seen that the diffusion coefficient of point defects in the group III-V material system is exponentially positively correlated with temperature, indicating that increasing the temperature is very beneficial for promoting the diffusion of point defects and enhancing the effect of QWI.

Figure  3.  The relationship between relative interdiffusion coefficient and temperature

DownLoad: Full-Size Img PowerPoint

2.3 The effect of stress on QWI

At the interface of two materials with high lattice mismatch, there will be a certain amount of stress, which will cause compressive or tensile stress on the surface of the material. The surface compressive stress will cause the GaAs lattice atoms to be squeezed, and some atoms, especially Ga atoms, will be squeezed out of the interface, leaving a certain number of vacancy defects on the GaAs surface ^[17]. To study the interface deformation during annealing process, the COMSOL multi-physical field modeling software was used to simulate the stress-strain behavior of GaAs with Si dielectric layers after annealing.

It is assumed that the epitaxial wafers are annealed at 850 °C, and stress is released when the annealing temperature drops to 200 °C, and finally stable deformation occurs at room temperature. The relevant parameters used in the calculation are shown in Table 1. The substrate material of the primary epitaxial wafer is 450 μm n-GaAs, the total thickness of the epitaxial wafer is approximately 4.5 μm, and both contain a large proportion of Ga and As components. To avoid calculation errors caused by excessive relative tolerance, the simulated substrate and epitaxial wafer are both 25-μm GaAs, with a dielectric layer of 200-nm Si. The simulation results based on COMSOL and magnified by 100 times are shown in Figure 4 (color online). It can be seen that the surface of GaAs undergoes compression caused by compressive stress after annealing, indicating that the Si dielectric layer will provide compressive stress to the GaAs surface and induce more Ga vacancies in GaAs, which is conducive to the QWI process.

Table  1.  Young's modulus, Poisson’s ratio, density and coefficient of thermal expansion of related materials

Sample	GaAs	Si	SiO2
Young's modulus(Pa)	8.50×1010	1.77×1011	7.31×1010
Poisson's ratio	0.31	0.2891	0.17
Density(kg/m3)	5500	2328	2203
Coefficient of thermal expansion(1/K)	6.40×10−6	2.60×10−6	5.50×10−7

 | Show Table

DownLoad: CSV

Figure  4.  Deformation results of primary epitaxial wafer simulated by COMSOL after annealing

DownLoad: Full-Size Img PowerPoint

3. Experiment and result analysis

3.1 Research on the effect factors of QWI

3.1.1   Effect of cover layer

During the annealing process, covering with GaAs cover plates not only reduces surface contamination, but also provides a certain pressure for As concentration, which can inhibit the decomposition and volatilization of As on the surface of the epitaxial wafer to some extent. The surface morphology of the primary epitaxial wafer at 875 °C/90 s RTA is shown in Figure 5 (color online). Figures 5 (a) and 5 (b) show the surface morphology of primary epitaxial wafers with and without GaAs cover plates, respectively. Similar to the predicted results, the surface of epitaxial wafers with GaAs covers is smoother, and there are fewer ablative holes generated during annealing, indicating that the GaAs covers have a certain protective effect on the surface of the Si dielectric layer. Therefore, subsequent RTAs were conducted in the environment with GaAs covers.

Figure  5.  Surface morphology (a) with and (b) without epitaxial wafers after RTA

DownLoad: Full-Size Img PowerPoint

3.1.2   The effect of temperature

The calculation results show that temperature has a significant effect on the diffusion coefficients of impurities and point defects. Therefore, the effect of temperature on QWI is investigated first. By using MOCVD, a 20-nm single crystal Si was grown on the surface of GaAs primary epitaxial wafers at the growth temperature of 800 °C. Then, a 90 s Rapid Thermal Annealing (RTA) was performed in the interval of 775 to 900 °C, and the PL results after annealing are shown in Figure 6 (color online). It can be seen that the effect of wavelength blue shift increases with the increase of heat treatment temperature. Compared to the original primary epitaxial wafers, a maximum wavelength blue shift of about 90 nm was obtained at 900 °C, but at this point, the FWHM was significantly widened and the waveform was severely deteriorated, indicating significant material damage. At 875 °C, the wavelength blue shift is about 57 nm, and the FWHM is well maintained. Therefore, it is believed that heat treatment at 875 °C can achieve a good QWI effect while also ensuring the lattice quality of the material.

Figure  6.  Effect of RTA temperature on wavelength blue shift of primary epitaxial wafers

DownLoad: Full-Size Img PowerPoint

3.1.3   The effect of heat treatment time

The effect of heat treatment time on QWI is further investigated. The annealing temperature is always 875 °C, and the annealing time is set to 60 s, 90 s, and 120 s respectively. The PL results of the primary epitaxial wafers after annealing are shown in Figure 7 (color online). As the annealing time increases, the wavelength blue shift of the primary epitaxial wafer introducing Si impurities also gradually increases. However, when the annealing time reaches 120 s, the peak of the PL spectrum is already deformed. It indicates that after 90 s RTA treatment, a good blue shift effect can be achieved, and the peak intensity of the PL spectrum and the FWHM remain good.

Figure  7.  Effect of RTA time on wavelength blue shift of primary epitaxial wafers

DownLoad: Full-Size Img PowerPoint

3.1.4   The effect of the properties of the dielectric layer

If the Si grown on the epitaxial wafer surface is too thick, the lattice mismatch and the difference in coefficient of thermal expansion will be amplified, which will trigger the stress release during thermal annealing. The ability of a thinner Si layer to suppress the decomposition and outward volatilization of Ga and As atoms in the GaAs ohmic contact layer will also be weakened, so it is necessary to consider the effect of Si characteristics. The Si dielectric layer grown by MOCVD equipment is single crystal, and its lattice quality and density are affected by the reaction source, growth temperature and other conditions, so the Si dielectric layer grown under different conditions will also affect the QWI effect. Therefore, three types of Si epitaxial layers were prepared: 20 nm high-temperature Si grown at 800 °C, 20 nm low-temperature Si grown at 650 °C, and 50 nm low-temperature Si, set as # 1, # 2, and # 3, respectively, to investigate the optimal growth conditions for Si dielectric layers that induce best QWI effect.

Similarly, a single RTA treatment at 875 °C/90 s was applied to the group of the primary epitaxial wafers, and the PL spectra of the primary epitaxial wafers were tested after heat treatment, as shown in Figure 8 (color online). It can be seen that the difference of QWI effect caused by the three types of Si layers is relatively small. For Si layers with the same thickness, the effect of Si layer growth temperature on wavelength blue shift is relatively small, but the FWHM is narrower for the high-temperature Si layers. For Si layers with the same growth conditions, thicker Si layers cause more wavelength blue shifts, reaching about 57 nm, but their FWHM is also larger, indicating that the material quality is greatly affected.

Figure  8.  Effect of different silicon layers on wavelength blue shift of primary epitaxial wafers

DownLoad: Full-Size Img PowerPoint

3.2 Microscopic characterization of QWI effect

In order to accurately understand the diffusion depth of Si atoms, EDS was used to test the element distribution at different depths on the epitaxial wafer. The Si IID primary epitaxial wafers treated with 875 °C/90 s RTA were carried out Si layer removal treatment, and then corroded for 0 s, 15 s, 30 s, and 45 s using a special solution. The test results are shown in Figure 9 (color online). Experience shows that the corrosion rate of the corrosive solution is about 25−35 nm/s, so the surface of the etched epitaxial wafer corresponds to different depths. From the EDS results, it can be seen that the p-type doping element of the primary epitaxial wafer is C, so the element C content is ligher when the surface layer of GaAs is not corroded moreover, the element Si content is also higher, and The content of both in the same order of magnitude; with the corrosion time increases to 15 s, the element C content gradually decreases, and the element Si content decreases significantly; when the corrosion time reaches 30 s, i.e., when the corrosion depth reaches approximately the upper limiting layer, the Si content has decreased to 22.2% of the original Si content in the surface layer; when the corrosion time further increases to 45 s, i.e., when the corrosion depth reaches approximately the upper waveguide layer or near the QW region, the Si content basically decreases to 0. This result shows that the Si impurities can diffuse into the upper waveguide layer of the primary epitaxial wafer after 875 °C/90 s RTA treatment, and then produce an effective QWI induction effect.

Figure  9.  Surface EDS results of element composition at different corrosion times of primary epitaxial wafers after 875 °C/90 s RTA. (a) Untreated sammple; (b) corrosion for 15 s; (c) corrosion for 30 s; (d) corrosion for 45 s

DownLoad: Full-Size Img PowerPoint

4. Conclusion

In order to comprehensively improve the performance index of InGaAs/AlGaAs semiconductor QW lasers, a feasible scheme for Si impurity induced QWI is investigated in this paper. The relationship between the effect of Si impurity-induced QWI and the nature of dielectric layer and heat treatment conditions was investigated by using the variable-controlling method with multiple sets of control conditions. The PL test results show that growing a 50 nm Si epitaxial dielectric layer at 650 °C in combination with 875 °C/90 s RTA heat treatment results in a wavelength blue shift of about 57 nm. Combined with EDS test, it is found that Si impurity atoms can diffuse into the upper waveguide layer or QW of the InGaAs/AlGaAs semiconductor QW laser primary epitaxial layer after 875 °C/90 s RTA, resulting in a significant QWI effect. In the future, Si impurity induced QWI NAW can be prepared by combining epitaxial growth technology and RTA technology to suppress CODs and continuously improve the output power of InGaAs/AlGaAs semiconductor QW lasers.