Effect of Annealing Time of GaN Buffer Layer on Curvature and Wavelength Uniformity of Epitaxial Wafer

Abstract

In this study, the curvature changes of an unintentionally doped GaN end and third quantum well were observed in situ when the annealing times of a GaN buffer layer were 40 s, 50 s and 55 s, respectively. When the annealing time was increased from 40 s to 50 s, the concave curvature of the unintentionally doped GaN end and the third quantum well became smaller. When the annealing time was increased to 55 s, there was no significant change in curvature. These curvature changes are related to the relaxation of the stress in the epitaxial wafer with different annealing times. With the increase in buffer annealing time, the compressive stress and warpage decreased gradually, and the photoluminescence wavelength of the sample became longer. Meanwhile, the standard deviation yield of the dominant wavelength was increased by 5.46%, and the wavelength yield was increased by 19.45% when the annealing time was changed from 40 s to 50 s.

1. Introduction

Due to the lack of natural GaN single-crystal materials, GaN-based blue light LEDs are usually grown on sapphire (Al₂O₃) by heteroepitaxy. However, there is a significant difference in the thermal expansion coefficient and lattice constant between the Al₂O₃ substrate and GaN material, which inevitably leads to significant stress during the growth process [1,2,3]. In addition, the epitaxial film is composed of a variety of functional layers, and the composition, growth temperature, growth pressure and other conditions between different functional layers are different, which will inevitably lead to internal stress in the epitaxial layer [4,5]. Due to the existence of these microscopic stresses, the epitaxial layer warps and deforms. With the constant change in stress state, the warping state of the epitaxial layer changes constantly, and cracks appear. Meanwhile, warping deformation is also one of the main factors restricting the production of large-sized epitaxial wafers. Therefore, the study of epitaxial wafer warping is crucial.

Scholars have conducted extensive and in-depth research on the warping problem of epitaxial films, such as using substrates with different warping degrees [6,7], AlN insertion layers [8], AlGaN buffer layers with gradually decreasing Al components [9,10], step-graded AlGaN buffer layers [11,12], patterned sapphire substrates [13], and AlN/GaN superlattice structures [14], to reduce stress in epitaxial films and control the warping degree of epitaxial films. For all these techniques, it has been noted that the growth parameters of the buffer layer or the intermediate layer have a large influence on the quality of GaN, indicating that it is crucial to precisely optimize the growth parameters and study the stress relaxation mechanism in detail [15]. This is especially important for the growth of GaN-based LED structures with a homogenous emission wavelength, which involves the additional difficulty when using 4 in or larger substrates.

Previous work revealed the different sources of growth-induced stress due to thermal mismatch and lattice mismatch between the epilayer and substrate [16,17]. However, quasidirect measurement of stress during epitaxy by recording the wafer curvature has become more and more popular [18,19,20]. Therefore, in situ techniques play an important role in the growth process to observe the influence of growth parameters on stress evolution. In this paper, by changing the annealing time of the GaN buffer layer, the warp change of the wafer during the growth process is monitored in real time by the in situ warpage monitoring equipment of metal–organic chemical vapor deposition (MOCVD), and the influence mechanism of annealing time of the GaN buffer layer on wafer warping was studied. In addition, the relationship between the wavelength of photoluminescence spectrum and the warping of epitaxial wafers was studied.

2. Experiment

This experiment used the low-voltage K465i MOCVD equipment produced by Veeco company. The graphite disk of the equipment was divided into two rings, inner and outer, and each ring was placed between 9 and 15 chips, respectively. Each piece had a diameter of 4 inches. During the growth process, trimethylgallium (TMGa) and triethylgallium (TEGa) were used as gallium sources. Trimethylindium (TMIn) and ammonia (NH₃) were used as indium and nitrogen sources, respectively. Silane (SiH₄) and ferrocene (CP₂Mg) were used as doping sources for n-type gallium nitride (n-GaN) and p-type gallium nitride (p-GaN), respectively. Nitrogen (N₂) and hydrogen (H₂) were used as carrier gases for different functional layers. The preparation process was as follows.

First, the surface of the sapphire substrate was cleaned in a H₂ atmosphere of 1100 °C to remove surface impurities and particles. Then, the GaN buffer layer was grown at 550 °C. The temperature was raised to 1020 °C for high-temperature annealing treatment. The unintentionally doped GaN layer (u-GaN) was then grown at 1060 °C. The n-GaN was grown on u-GaN at 1080 °C. The InGaN/GaN multi-quantum wells were grown at 750 °C to 860 °C for 13 cycles. The AlGaN electron barrier layer was grown at about 850 °C, and p-GaN was finally grown at 960 °C. Figure 1 shows the schematic diagram of the epitaxial structure of the InGaN-based LED. This article mainly studies the effect of buffer layer annealing time on warping, and focuses on samples with three buffer layer annealing times of 40 s, 50 s, and 55 s, respectively. They were labeled as furnace A, furnace B, and furnace C, respectively.

The experiment used DTR-210 in situ measurement equipment to obtain the warping information of epitaxial wafers. Room-temperature photoluminescence (PL) mapping was examined using the 325 nm line of a He–Cd laser to investigate the dominant wavelengths of all samples. The target dominant wavelength of samples for this time is 458 nm.

To ensure the accuracy of the experimental data, only the annealing time of the buffer layer was changed this time, and no other parameters were changed. The curvature values of the u-GaN end and the third quantum well were collected. The locations of the collection point are marked with U-2 and QW-3 in Figure 2.

3. Results and Discussion

Figure 2 shows the variation in curvature over time during sample growth obtained from the DRT-210 in situ curvature monitoring system. Figure 2 shows the curvature curve of a single epitaxial wafer, while the curvature curves of other epitaxial wafers have roughly the same trend. The difference lies in the different curvature values of each segment.

It can be seen from Figure 2 that the warping state of the sample constantly changes during the growth of the epitaxial layer. When the u-GaN layer and n-GaN layer were deposited (left of the yellow dotted line in Figure 2), the epitaxial layer showed a large “concave” deformation state, which was due to the large lattice mismatch and thermal mismatch between the Al₂O₃ substrate and the GaN epitaxial layer. Although a large number of dislocations were formed in the process of nucleation and island merger, thereby releasing a large amount of misfitting stress, the stress could be completely released in this process [21], so there were still some residual stresses during the growth of u-GaN, resulting in the warping of the epitaxial film. In addition, in the n-GaN growth stage (the yellow arrow points to the growing end of n-GaN in Figure 2), the curvature decline rate of the growing end of n-GaN was significantly faster than that of the U-2, which can be explained by silicon doping introducing external stress [22]. The original Ga atoms are replaced by Si atoms in a substitutional way. The dislocation morphology changes [23]. In addition, when the thick u-GaN layer is deposited, a large stress will accumulate, resulting in a large relaxation degree of the epitaxial film, and part of the stress will be released, resulting in a small stress value introduced during the growth of n-GaN. In the subsequent deposition of low-temperature MQWs, the thermal stress decreases the “concave” deformation degree of the epitaxial film until the surface is flat or “convex” deformation occurs. The curvature value of the epitaxial film also changes with the change in the growth temperature of the well layer and the barrier layer. In the growth stage of the quantum well, the curvature of the epitaxial film changes periodically, which is caused by the different thermal expansion coefficients between the well layer and the barrier layer, and the curvature changes continuously in the process of alternating growth temperature.

Figure 3 shows the curvature scatter diagram and linear fitting curves of U-2. It can be seen from Figure 3 that the curvature is basically above 100 km⁻¹, and the warping is concave. During the epitaxial growth of the buffer layer, GaN islands blend and approach each other, and adjacent atoms interact to generate internal tensile stress. Although a large number of dislocations are formed during the nucleation process and island merging, releasing a large amount of mismatched stress, the stress cannot be completely released during this process, so there are still some residual stresses during the growth of u-GaN, which cause the epitaxial film to warp. The curvature value of furnace A is greater than that of furnace B, indicating that the curvature of epitaxial film decreases and the concave deformation of epitaxial film decreases with the increase in annealing time of the buffer layer. The curvature of furnace B and furnace C does not differ significantly with increasing annealing time. With the increase in annealing time of the buffer layer, the curvature of furnace B decreases compared with furnace A, which may be related to the following factors. On the one hand, the annealing time after the buffer is under the condition of high temperature and pressure, and the parts of the buffer with poor crystal quality are baked off, leaving the crystal nuclei with higher crystal quality, which is conducive to the improvement of the crystallization quality. This process is equivalent to the secondary crystallization of the epitaxial layer, which is conducive to reducing the stress of the epitaxial layer during the subsequent epitaxial growth. On the other hand, with the increase in annealing time, the low-quality crystal nuclei in the buffer will be roasted and adjusted, and the overall thickness of the buffer will become thinner. According to the quantitative relationship that the curvature of the epitaxial film changes with the structure of the film, the following equation can be obtained [24]:

$${\displaystyle \frac{1}{R_c}}={\displaystyle \frac{6{\sigma }_f{h}_f}{M_s{h}_s^2}}$$

(1)

where ${\sigma }_f$ is the stress of the epitaxial film. The $h_f$ and $h_s$ are the thickness of the epitaxial film and the substrate, respectively, and $M_s$ is the biaxial elastic modulus of the substrate. As $h_f$ and ${\sigma }_f$ decreases, the curvature radius $R_c$ increases and the curvature of the epitaxial film decreases.

In order to further analyze the relationship between the curvature and the stress of the sample, the Stoney equation [24] was used to calculate the stress on the film. Because the buffer layer thickness is relatively thin and has little influence on the calculation results, the entire epitaxial structure can be assumed as a single GaN layer with uniform warping. The Stoney formula after deformation is as follows:

$${\sigma }_f={\displaystyle \frac{h_s^2{M}_s\kappa }{6h_f}}$$

(2)

where κ is the curvature of the epitaxial film. The biaxial modulus of Al₂O₃ used in this paper is equal to 603 GPa [25]. $h_s$ is 420 μm. By calculation, the stress values of the tenth epitaxial film of furnace A, B and C are 0.83 GPa, 0.65 GPa and 0.62 GPa, respectively.

In addition, due to the influence of the thermal conductivity of substrate and epitaxial layer itself, the actual temperature of the upper and lower surfaces of the substrate film will be different to some extent, resulting in the warping deformation of the epitaxial film. Because the thickness of the epitaxial layer of furnace B is smaller than that of the epitaxial layer of furnace A, the following formula can be used [20]:

$${\displaystyle \frac{1}{R_c}}={\displaystyle \frac{{∆T(\alpha }_{film}-{\alpha }_{substrate})}{h}}$$

(3)

where ${\alpha }_{film}$ and ${\alpha }_{substrate}$ are the thermal expansion coefficients of the epitaxial film and substrate, and h is the thickness of the substrate. $∆T$ is the temperature difference between the upper surface and undersurface of the substrate. Because of the difference in thickness of the epitaxial film, $∆T$ of furnace B is slightly smaller than $∆T$ in furnace A, making the curvature of the epitaxial film of furnace B slightly smaller than that of the epitaxial film of furnace A, but the effect is very small.

Although the annealing time of the buffer layer in furnace C is 5 s longer than that of the buffer layer in furnace B, the curvature of U-2 is almost the same for both. This indicates that the warpage may have a critical value. When the critical value is reached, the warpage no longer changes, or even develops in the opposite direction, resulting in convex deformation. It may also be that 5 s is too little time to cause a noticeable change in curvature. These outcomes need to be further studied.

Figure 4 shows the curvature of the QW-3 of furnace A, B and C, which is much smaller than the previous U-2 curvature. From Figure 4, it can be seen that as the annealing time of the buffer layer increases, the curvature gradually decreases. The QW-3 mean curvature of furnace A, B, and C are 30.24, 19.67 and 17.09 km⁻¹ respectively. With the increase in annealing time of the buffer layer and the subsequent growth of u-GaN and n-GaN, the stress in the epitaxial layer is greatly relaxed before the growth of the quantum well, resulting in the flat surface and small warpage of the epitaxial film when the multi-quantum well with a low growth temperature is deposited. In the growth stage of multi-quantum wells, due to the influence of thermal mismatch, when the multi-quantum wells with a low growth temperature are deposited, the absolute curvature decreases.

In order to analyze the effect of buffer annealing time on the optical properties of the sample, a PL test was performed. Figure 5 shows the PL-mapping chart of dominant wavelengths of all samples in furnace A, B, and C. The uniformity of wavelength distribution can be observed through the standard deviation of dominant wavelength. The smaller the standard deviation, the better the wavelength uniformity. The larger the standard deviation, the worse the wavelength uniformity. The blue color indicates that the actual wavelength is shorter than the target wavelength, and the red color indicates that the actual wavelength is longer than the target wavelength. From Figure 5, we can see that the overall color is blue, mostly presenting a distribution of concentric circles, indicating a trend of overall shorter wavelength in furnace A. The trend of shorter wavelengths in furnace B and furnace C is not as obvious as in furnace A. This is consistent with the previous results; with the increase in annealing time, the compressive stress decreases and the curvature gradually decreases. Because the band gap energy of the GaN epitaxial layer is related to the stress in the film, the band gap decreases due to the tensile stress, while the band gap increases due to the compressive stress [26], which gradually increases the PL spectrum wavelength of the epitaxial film in furnace A, B, and C. When the annealing time of buffer layer of furnace B is increased by 10 s, it can be seen from the mapping diagram that the wavelength change from short to long in the inner circle is more obvious than that in outer circle. On the whole, the trend of the wavelength of furnace C from short to long has been significantly improved.

The wavelength yield and wavelength standard deviation yield are important indicators of the uniformity and consistency of the photoluminescence spectrum wavelength of epitaxial wafers.

Table 1. The results of dominant wavelength uniformity analysis. SD: standard deviation.

	Inner SD	Outer SD	All SD	Wavelength Yield	SD Yield
Furnace A	1.620	3.324	2.482	72.02%	36.01%
Furnace B	1.569	2.815	2.152	91.47%	41.47%
Furnace C	1.543	2.777	2.137	83.13%	41.47%

shows the standard deviation of wavelengths, as well as the wavelength yield and standard deviation yield for each furnace. It can be seen from the data in the table above that after the annealing time of the buffer layer is increased, the issue of the wavelength being short is effectively improved, which is in line with the conclusion of the previous curvature monitoring curve. Furthermore, compared to furnace A and furnace B, the standard deviation yield increased by 5.46%, and the wavelength yield also increased by 19.45%, indicating that the improvement in warping has a positive effect on the standard deviation yield and wavelength yield. The difference between furnace B and furnace C is not significant, but overall, the standard deviation of the dominant wavelength in furnace C is smaller than that in furnace B. Despite the addition of only 5 s, it is still effective.

4. Conclusions

In this paper, by changing the annealing time of the buffer layer at the bottom of the epitaxial layer, the relationship of curvature with annealing time during sample growth is obtained based on an in situ curvature monitoring system. The curvature variation of the unintentionally doped GaN end and third quantum well is studied, and it is proved that increasing the buffer annealing time can impact the concave deformation. When the annealing time is increased from 40 s to 50 s, the concave curvature of the unintentionally doped GaN end and the third quantum well became smaller. In addition, increasing the annealing time of the buffer layer can also improve the uniformity of wavelength distribution, resulting in an overall trend of wavelength increasing from short to long. Meanwhile, the standard deviation yield and wavelength yield of the dominant wavelength are increased. Thus, it can be concluded that the improvement of warping has a positive effect on wavelength uniformity of the epitaxial wafer.