Crack free ZnO thick film on sapphire substrate without GaN template grown by MVPE

1. Introduction

As a wide-bandgap semiconductor, ZnO has unique potentialities in the fields of optoelectronics, spintronics or nanostructures. In addition, the small lattice-mismatch between ZnO and GaN motivates workers to study growth of GaN on ZnO bulk or film substrate. Nowadays, several vendors can offer ZnO substrates of polar face, non-polar face as well as semi-polar face, grown by hydrothermal or melt growth methods. However, the price of ZnO bulk substrate is still high. By comparison, heteroepitaxy growth is still an alternative way to fabricate wide bandgap material-based devices. Earlier, we described a method to grow thick ZnO film of ∼10 μm by metal vapor phase epitaxy (MVPE) . The advantage is that MVPE is a kind of epitaxy growth method with growth rate as high as 120 μm/h. Therefore, it is easy to make large size of ZnO bulk substrate with low cost and high efficiency, considering the ready availability of sapphire substrate with large diameter (in the range of 50–200 mm). However, crack is a crucial problem for thick film growth. Cracks appear when ZnO film is a little thicker than 10 μm, not to mention the fact that the thickness of a free-standing substrate is required to be larger than 100 μm.

Although research has been extended to investigate the stress in fine structures such as quantum wells, there is still meaning to inquire into the problems caused by stress in large lattice-mismatch system, such as ZnO films grown on sapphire substrates. Unlike thin films (with thickness less than 10 μm), thick films form cracks motivated by the large residual tensile stress, composed by tensile growth stress, thermal mismatch stress and lattice mismatch stress. We found that ZnO films, grown on sapphire substrates with GaN templates, can be either tensile or compressive. A simple way to investigate the residual stress in film is to divide it into two groups, one thermal stress and another non-thermal stress, including lattice mismatch stress and growth stress. In this letter, the value of non-thermal stress was deduced from the crack density in ZnO film, while the data of the thermal stress was evaluated. Annealing was used to decrease the non-thermal stress. And annealed ZnO films were used as pseudo-substrates to grow crack-free ZnO films.

2. Experiment

In our experiment, ZnO samples were grown on c-plane Rubicon^® sapphire substrates by a home-made MVPE. Zinc vapor and water vapor were used as Zn and O sources, respectively. Nitrogen was used as the carrier gas at a flow rate of 6 L/min. Other growth parameters like total reactor flow (9.4 L/min), Zn carrier gas flow (0.4 L/min) and H₂O carrier gas flow (3 L/min) were kept identical to allow a direct comparison. The growth temperature is 1073 K. All the as-grown ZnO samples studied were grown using a two-step growth method. In the first step, a layer of ZnO film of about 2 μm is deposited on sapphire directly, and then etched off by HNO₃ dilution. It is found that there are residual Al₂ZnO₄ crystal nuclei or Al₂ZnO₄ buffer layer left on the etched sapphire substrate. The second step is to grow another layer of ZnO film at almost the same conditions. The goal of the two-step growth method is to obtain a stable high quality Zn-polar ZnO film on sapphire substrate. Normally polycrystalline or mixed polar ZnO films were obtained after the first growth, while only Zn-polar ZnO films after the second growth. Two ZnO films were grown by MVPE. One film with cracks has a thickness of 15 μm, labeled as sample A. The other film without cracks has a thickness of 12 μm, labeled as sample B. Sample A was used to evaluate the values of the non-thermal stress and the thermal stress. Sample B was used as a pseudo-substrate to re-grow ZnO films. Fig. 1 is the schematic drawing showing the growth of sample A and B, the annealing of sample B, and the re-growth of ZnO films on annealed sample B. The morphologies of the ZnO samples were observed by Hitachi S-4800 scanning electron microscope (SEM) and Digital optical microscopy. The residual stress was evaluated through the curvatures of the ZnO samples using the Dektak-8 surface profiler of Veeco.

Fig. 1. A schematic drawing of the growth of sample A and B, the annealing of sample B, and the re-growth of ZnO films on annealed sample B.

3. Results and discussions

Fig. 2(a) is the surface morphologies of as-grown sample A. Few cracks along 〈1 1 0〉 directions are observed on the surface. The regular hexagonal mesas on the surface have a diameter of about 25–100 μm, indicating the columnar growth mechanism. This means that cracks may form and then be buried or healed during the subsequent film growth. A wet-etching method was employed to dredge out the buried cracks. Sample A was dipped in diluted HCl (1:100) at room temperature and then picked out every 30 s to observe the changes of the morphologies by optical microscopy. To clarify the etching process, the surface morphologies are recorded at almost the same locations at different times (Fig. 2(a)–(f)). The etching rate is ∼0.3 μm/min. After 30 s etching, new cracks were observed and etching pits appeared. With time increasing, more and more cracks emerge and the widths of the cracks were widened. This phenomenon affirms that many of the cracks had been buried at different depth during the growth, and were dredged up again during the etching. The more there are cracks, the larger the stress is. It is found that, however, the crack density becomes saturated after 240 s etching. The corresponding average crack space at saturation, determined to be ∼20 μm by dividing the observation area by the whole length of the cracks, can be used to evaluate the maximum stress during film growth. The tensile residual strain in film with cracks can be deduced by the following expression.

   (1)

where Γ is the fracture resistance, γ is Poisson's ratio, Y is the Young's module, L_sat is the crack space at saturation for an array of cracks at 120°. Set Γ as 2.32 J/m². The Young's modulus and Poisson's ratio are 154 GPa and 0.30 for ZnO. Therefore, the maximum residual tensile strain is about 2.88×10⁻³, corresponding to a stress of about 630 MPa.

Fig. 2. Surface morphologies of sample A, etched in diluted HCl (1:100) at different times at room temperature; (a) 0 s; (b) 30 s; (c) 90 s; (d) 120 s; (e)180 s; (f) 240 s.

The value of thermal stress can be computed according to Reeber's data in the growth temperature range. The thermal stress is tensile for ZnO/sapphire grown at temperatures above 700 K rather than compressive below 700 K. In our experiment, ZnO samples were grown at 1073 K. Thus, the corresponding maximum thermal strain is 1.1×10⁻⁴ emerging at 500 K. It should be noted that strain instead of stress is used for comparison because strain is a non-unit quantity irrelative with film properties.

We estimated the non-thermal strain by subtracting the thermal strain from the residual strain. The maximum non-thermal strain should be about 2.77×10⁻³, indicating that non-thermal strain is the predominant component causing cracks during film growth. Therefore, it is essential to reduce the non-thermal strain in the as-grown ZnO films.

Annealing is one effective method to decrease non-thermal stress. To evaluate the effect of annealing, sample B without cracks was grown. The first annealing temperature was 900 °C for 10 h. The residual stress in ZnO was determined directly using wafer curvature method, a direct way to measure the change of radian. The strain in ZnO epilayer can be calculated by Stoney formula

   (2)

where Y is the Young's module, t is the thickness, Q is the curvature, s and f are subscripts denoting substrate and film, respectively. The Young's modulus for sapphire is 400 GPa. The thickness is 12 μm for the ZnO film, and 430 μm for the sapphire substrate. The wafer curvatures are evaluated by subtracting the reciprocal of the radius of the ZnO film from that of the sapphire substrate. The substrate curvature is 1.78×10⁻² m⁻¹. The modified curvatures (Qf–Qs) of the as-grown sample B and the annealed films are listed in Table 1. It is shown that there is little change after the first time annealing. Hence, another annealing was operated at 1000 °C for 10 h. It is shown in Table 1 that the strain was reduced to 56.2% after the second time annealing. Almost half of the non-thermal stress was relaxed after annealing.

Table 1.

Curvatures and the corresponding strain and stress in as-grown sample B, and the 1st and 2nd annealed sample B, as well as the deduced critical thickness h_c.

Samples	Modified curvature (Qf–Qs) (m−1)	Strain σ	Stress (MPa)	hc(μm)
As-grown sample B	5.55×10−2	3.55×10−4	77.59	16.51
Sample B after the 1st annealing	4.30×10−2	2.75×10−4	60.11	27.50
Sample B after the 2nd annealing	2.77×10−2	1.77×10−4	38.69	66.39

Then the 2nd annealed ZnO film was used as a pseudo-substrate on which to re-grow ZnO film. Fig. 3 shows the microscopy images of samples grown on the 2nd annealed sample B and as-grown sample B. First, another layer of ZnO was grown on as-grown and annealed sample B in the same batch. The growth time is 10 min., and the accumulated thickness is about 28 μm. It is shown in Fig. 3(a) that serious cracks appear on B₀, grown on as-grown sample B, with part of the film flaking off the substrate. Instead of cracks, however, isolated islands with flat tops are found on sample B₁ (Fig. 3(b)), grown on the 2nd annealed sample B. This can be explained by the selective growth model, caused by the redistribution of the stresses and the defects after annealing. Then, sample B₁ was reloaded into the growth chamber to re-grow 10 min. This time the total thickness is 45 μm for the sample, labeled as sample B₂. There is obvious coalescence of the isolated islands in sample B₂ (Fig. 3(c)). After that, another 10 min. growth of ZnO was grown on sample B₂ and the sample was called sample B₃ in order. The accumulated thickness reaches ∼65 μm. It is shown in Fig. 3(d) that a flat surface occurs, indicating the typical Zn-polar morphology. The cross-section image is shown inFig. 3(e), exhibiting an even interface without cavies. Finally, another 3 min. growth of ZnO was performed on sample B₃. The sample is called sample B₄. Few cracks in an interval of 3–4 mm were observed on sample B₄ (Fig. 3(f)). The total thickness of sample B₄ is about 70 μm.

Fig. 3. Surface morphologies of samples: (a) B₀; (b) B₁; (c) B₂; (d) B₃; (e) the cross section image of Sample B₃; (f) B₄. Sample B₀ is grown on as-grown sample B, regarded as the comparative counterpart of sample B₁. Sample B₁ is grown on the 2nd annealed sample B, and samples B₂, B₃ and B₄ are the consecutive growth based on the predecessor with the beginner of sample B₁. That is, sample B₂ is grown on B₁, B₃ on B₂, and B₄ on B₃, respectively.

Cracks appear when the film thickness is above its critical thickness. And the beginning formation of cracks can be evaluated as the experimental critical thickness. Therefore, the critical thickness is about 70 μm (sample B₄) for ZnO film grown on annealed ZnO pseudo-substrates, much higher than the value of about 15 μm (sample A) for as-grown ZnO samples.

The theoretical critical thicknesses for ZnO pseudo-substrate can be evaluated using the following expression:

   (3)

where εf is the co-axial strain in ZnO pseudo-substrate obtained from Eq. (2), and Z is a dimensionless parameter set at 3.951 for small, penny-shaped cracks at the film surface. The theoretical critical thickness for the as-grown sample B and the 1st and 2nd annealed sample B used for pseudo-substrates are listed in Table 1. The results show that the computed data are close to the experimental values, and that the critical thickness increases much after annealing, which can be explained by re-crystallization during annealing.

Table 2 lists the full-width at half-maximum (FWHM) of DCXRD spectra for (0 0 2) and (1 0 2) planes in as-grown sample B and the re-growth samples including sample B₁, B₂ and B₃. The FWHMs of (0 0 2) and (1 0 2) planes decrease when the ZnO films grow thicker, indicating a great improvement in the crystal quality. The FWHM of (0 0 2) and (1 0 2) in sample B₃ is almost half of those in the as-grown sample B. The decreases of the FWHMs support that there exists a process of re-crystallization during annealing.

Table 2.

FWHMs of DCXRD curves for (0 0 2) and (1 0 2) planes in as-grown sample B and the re-growth samples (B₁, B₂ and B₃).

Samples	(0 0 2) (arcsec)	(1 0 2) (arcsec)
As-grown sample B	467	824
Sample B1	367	554
Sample B2	305	484
Sample B3	270	465

Crack-free ZnO films with thickness beyond 65 μm should be obtained if optimized annealing conditions are to be used, such as temperature, pressure and period. To get a crack-free ZnO film of about 130 μm, the tensile strain in ZnO should be reduced to 1.25×10⁻⁴, a little larger than that of the thermal strain. Therefore, though more efforts should be paid to reduce the non-thermal stress, it is feasible to get crack-free ZnO film with thickness enough to serve as free-standing substrate. GaN template, if added, is expected to speed up the process, as there have been many reports about quality improvement in ZnO films after using GaN templates.

4. Conclusions

In summary, we have obtained crack-free ZnO film with thickness of ∼65 μm, grown directly on sapphire substrate without GaN template. The measurements using wafer curvature method show that non-thermal strain is larger than thermal strain in ZnO/sapphire growth system. Annealing was used to reduce the tensile stress in ZnO/sapphire films. Crack-free ZnO film was obtained by re-growing ZnO on the annealed ZnO pseudo-substrate. To further improve the critical thickness of ZnO film on sapphire substrate, a combination of using annealed ZnO pseudo-substrate and exploiting GaN template is provided.

Source:Materials Science in Semiconductor Processing

If you need more information about Crack free ZnO thick film on sapphire substrate without GaN template grown by MVPE, please visit: http://www.powerwaywafer.com or send us email at gan@powerwaywafer.com.