Low-defect-density ZnO homoepitaxial films grown by low-temperature ALD

ZnO is a widely studied semiconductor with a unique set of physical properties and tremendous potential for various applications.^1,2 Its wide bandgap (3.37 eV) and large exciton binding energy (60 meV), high thermal and chemical stability, inexpensive growth technology, and non-toxitity make ZnO a prospective material for UV light emitters,³ gas⁴ and biosensors,^5,6 transparent electronics,⁷ spintronics,⁸ and solar cells.⁹ Nevertheless, many technological problems remain to be solved to realize device-grade ZnO. In particular, many of the device applications require both donor and acceptor doping of ZnO. Whereas ZnO can be easily doped n-type,¹⁰ stable and reproducible p-type doping has been difficult to attain.¹¹ Since ZnO is known to dope itself n-type, it is not clear what dopant could compensate the background shallow donors and ensure adequate abundance of shallow acceptors for p-type ZnO.

In the last two decades, there has been a significant effort to grow ZnO thin films with low concentrations of point and extended defects to reduce the unintentional n-type doping (for a recent review, see Ref. 12). So far, ZnO films of the highest crystalline quality have been obtained by homo- and heteroepitaxy, using methods that require growth temperatures in excess of 550 °C (see, e.g., Refs. 13–16). Here, we use atomic layer deposition (ALD) of ZnO thin films on a single-crystalline ZnO substrate at relatively low temperatures (⁠$≤200 °$C) to further lower the defect density of homoepitaxial ZnO. For this purpose, ALD is a promising technique in many respects.¹⁷ It provides high uniformity of the grown layers, precision, repeatability, simple doping protocols, and low thermal budget. In this Letter, we focus on extended defects and their impact on structural and optoelectronic properties of the ALD-grown homoepitaxial films, which are studied by transmission electron microscopy (TEM), photoluminescence (PL), and surface photovoltage (SPV) spectroscopy.

The pressurized melt-grown, O-terminated (0001) ZnO substrates used for the deposition are supplied by Cermet. The manufacturer reports the x-ray $θ−2θ$ rocking curve FWHM of 49 arc sec.¹⁸ Before being loaded into the ALD chamber, the substrates were degreased in ethanol and rinsed with de-ionized (DI) water. A Cambridge NanoTech Savannah 200 system was used for the deposition. The ZnO films are grown using electronic-grade diethylzinc and DI water as precursors and semiconductor-grade nitrogen with a constant flow of 20 sccm as a carrier and a purge gas. The temperature of the liquid precursors was 35 °C. Each ALD cycle consisted of a water dose of 15 ms, a nitrogen purge of 5 s, a diethylzinc dose of 15 ms, and a nitrogen purge of 5 s. All the films are deposited with 500 ALD cycles. The substrate temperatures were 120, 150, and 200 °C. When choosing the deposition temperature, we took into consideration reports of ZnO ALD on sapphire,¹⁹ glass, and silicon,²⁰ using the same precursors and suggesting an optimum growth window of 130–180 °C, though ZnO homoepitaxy by ALD at 300 °C was reported.²¹ 

Structural properties of the ZnO homoepitaxial films are studied by TEM. The electron transparent samples were prepared using a focused ion beam (FIB) Zeiss Crossbeam 340 with 30–2 kV Ga⁺ to reduce the amorphized ion-damaged layer. TEM images and electron diffraction patterns were obtained using a JEOL 2010F microscope, while high-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) was performed with an aberration-corrected JEOL ARM200F. Both microscopes were operated at 200 keV.

The film thicknesses are determined from the TEM cross sections with the results listed in Table I. We observe a nearly linear increase in the growth rate with temperature, rather than a constant rate expected in the case of the self-limiting reaction. The increase in the film growth rate above the ALD temperature window was previously attributed to formation of atomic layers via Zn–Zn bonding.¹⁹ The above assertion, however, is not supported by our TEM studies.

The films are imaged in the dark-field mode using the combined $(01¯10)$ and $(0002¯)$ reflections, along the $⟨211¯0⟩$ zone axis. The TEM cross sections of all three samples are shown in Fig. 1, where the shorter red arrows indicate the $[0001]$ growth direction. The selected area electron diffraction (SAED) patterns are obtained from the electron beam focalization on the film/interface regions. The SAED patterns, which are shown in the insets, provide interplanar distances of 0.5125 nm and 0.1382 nm for the (0002) and $(01¯10)$ planes, respectively. These values are in accordance with the c and $a3/4$ parameters for ZnO. In addition, a magnified image of the interface region of the 150 °C sample and the fast Fourier transforms for the film, the substrate, and the interface are shown in the supplementary material, Fig. S1. From the TEM images and SAED patterns, we, thus, observe a monocrystalline hexagonal wurtzite crystal structure in all three samples with no noticeable change in the crystal orientation across the film/substrate interface, attesting high-quality homoepitaxial growth over the $[0001]$ direction. Moreover, in the TEM images, there is no noticeable diffraction contrast between the substrate and the films, indicating their identical chemical composition.

The TEM imaging is particularly useful in identifying structural defects, which is important for optimizing the growth and understanding their impact on optoelectronic properties. Figure 1(a) shows the presence of ZnO crystallite (circled with a broken green line), misoriented with respect to the surrounding crystal. The crystallite inclusions have been observed only in the 120 °C ZnO film. The extended defects, such as basal and prismatic stacking faults (BSFs and PSFs), are also noticeable in Fig. 1. The BSFs, which are parallel to the film/substrate interface and highlighted with longer yellow arrows, can be seen both in the films and in the substrate. The PSFs (shown with shorter red arrows) originate at the film/substrate interface and extend to the film surface (see also the supplementary material, Fig. S1). Because the film samples differ in thickness, the amount of the defects in a single image can be misleading. In order to obtain the defect densities, we counted and averaged the stacking faults over several TEM images of the same cross-sectional area for each sample. The area densities of BSFs and PSFs in the substrate and the films are listed in Table II. These defect densities (standard uncertainty, $≤5%$⁠) are less than in post-annealed ZnO films on sapphire grown by ALD²² and are comparable to that in ZnO films on sapphire grown by molecular beam epitaxy.²³ Moreover, the BSF densities in all ZnO films are close to each other and to that of the substrate.

In polar crystals such as the group III nitrides (Al, In, Ga)N, the PSFs are often identified as inversion domain boundaries (IDBs) by multiple dark-field TEM imaging along a non-centrosymmetric axis.^24,25 This method is based on the violation of Friedel's law in wurtzite crystals,²⁶ resulting in a strong contrast and contrast reversal in two adjacent regions of the image acquired under multiple beam conditions with $g=±(0002)$⁠. In Fig. 2, the inversion domains in the 200 °C ZnO film are shown in red rectangles with the IDBs indicated by white arrows. Conversely, no contrast reversal is observed under the multiple beam conditions in the absence of inversion domains. The IDBs are known to have an effect on bulk piezoelectricity in ZnO.²⁷ By altering the strain fields, they create a local piezoelectric polarization.²⁸ As a result, an electric field parallel to the stress in one of the inversion domains is canceled by an anti-parallel field in the other, thereby reducing the bulk response.²⁹ The formation of inversion domains in the homoepitaxial ZnO films may be nucleated by accidental growth on the substrate or by nucleation and growth within a columnar grain due to the interaction with a BSF.

The BSFs are responsible for the striation along the film/substrate interface in Fig. 1 [see also the supplementary material, Fig. S2(a)]. In hexagonal wurtzite structures, the BSFs can be looked at as a local deviation to the cubic zinc blende crystal structure.³⁰ Since they introduce electronic levels just above the valence-band maximum, they can contribute to optical transitions and become noticeable in PL spectra.³¹ In contrast to GaN epilayers,³² long extended BSFs are not observed in the ZnO films, which can be related to the homoepitaxial growth. Otherwise, the observed stacking faults are closely similar to those in GaN epitaxial films.^30,32 Figure 3(a) shows a HAADF-STEM image of the 200 °C ZnO film, along the $⟨211¯0⟩$ zone axis, featuring a BSF of I₂ type. Following Ref. 32, we highlight the two ${101¯0}$ planes of the perfect wurtzite structure in yellow, separated by 15 lattice spacings, so the wrong stacking sequences (underlined by a longer yellow line) can be seen. The I₂ BSF is terminated by Shockley partial dislocations shown by shorter yellow lines. Since the two partial dislocations have opposite Burgers vectors, $b=±13⟨101¯0⟩$⁠, the Burgers circuit drawn around the fault does not exhibit any closure failure. As suggested in Ref. 32, the I₂ fault can be formed by a shear rather than dissociation of a perfect dislocation into two Shockley partials. A different case where an I₁ BSF arrives at an IDB is shown in Fig. 3(b). While the IDB continues toward the film surface, the I₁ BSF is terminated by a Shockley partial dislocation. Figure 3(c) shows a strain map of the interaction between the BSF and IDB, obtained from geometrical phase analysis.³³ Normal strain component $εyy$ is obtained in the direction perpendicular to the scanning direction in STEM to avoid a “fly-back” error.³⁴ In contrast to the case of nearly strain-free I₂ BSF of Fig. 3(a), the interaction between the I₁ BSF and IDB results in a high strain of $±45%$ at the Shockley partial dislocation as shown in the inset in Fig. 3(c). The BSFs causing the lattice strain are also observed in the 120 and 150 °C films (supplementary material, Figs. S2 and S3). According to Ref. 35, the in-plane lattice strain has an effect on excitonic luminescence of ZnO homoepitaxual films.

Optoelectronic properties of the ZnO films are studied with PL and SPV spectroscopies. The PL measurements employed a CW Kimmon IK5452R-E HeCd laser at a wavelength of 325 nm. The samples were mounted inside an evacuated Janis CCS-150 cryostat operating within a controllable 8–325 K temperature range. The PL spectra were probed by a Spex 1401 monochromator with a spectral resolution of 0.18 cm⁻¹ and an RCA C31034 photomultiplier tube detector. As to SPV spectroscopy, it was performed under ultra-high vacuum (⁠$∼10−9$ Torr). The SPV response was excited by monochromatic light of variable frequency (the energy range, 1.13–3.65 eV), sourced from a 250 W QTH lamp, filtered with an Oriel Cornerstone monochromator, and probed with a Besocke Kelvin Probe S and Kelvin Control 07.

Figure 4 shows PL spectra of the ZnO films at room temperature and 8 K. It should be noted that the film thicknesses are of the same order as the absorption depth (⁠$∼70$ nm) of the laser radiation used for excitation. Thus, the PL spectra can be attributed to both the grown films and the underlying substrate (see the supplementary material, Fig. S4). The deep (⁠$∼2.5$ eV) defect signatures in the spectra have relatively weak PL intensities, indicating low concentrations of visible luminescent centers. By the same token, the near-band edge luminescence is several orders of magnitude stronger than the deep level emission [see both Figs. 4(a) and 4(b)]. Furthermore, the SPV spectrum obtained for the 150 °C sample indicates a low content of surface mid-gap traps, while the bandgap transition is sharp and well-defined (supplementary material, Fig. S5).

The low-temperature PL spectra of Fig. 4(a) exhibit several emission features near the band edge, which originate from the radiative recombination of bound excitons (BEx). These exciton recombination lines are plotted in Figs. 4(c) and 4(d) with complementary spectral ranges. Arguably, all three ZnO samples produce rather similar PL spectra. The dominant BEx peaks have the same peak positions and similar intensities, except for the 3.330 eV peak (Y, indicated by a vertical blue arrow), which has a significantly higher relative intensity in the 120 °C sample. The same peak is observed in the substrate as well [supplementary material, Fig. S4(a)]. We, thus, attribute the 3.330 eV peak to excitons bound to the extended structural defects observed with TEM. For comparison, in Ref. 22, a peak at 3.321 eV was attributed specifically to the BSFs in ZnO epitaxial films grown on c-plane sapphire by ALD; in Ref. 36, a 3.333 eV line was ascribed to excitons bound to structural defects in ZnO single crystals. Furthermore, Ref. 37 reported sharp emission lines around 3.333 eV, accompanied by very weak green luminescence, in Cermet substrates such as used in the present study. Since the green luminescence originates from mobile point defects such as Zn or O vacancies, it was argued that these intrinsic defects are collected by dislocations, which result in the reduction of the 2.45 eV band and the intensity increase in the luminescence lines of excitons bound to extended structural defects.³⁷ In Fig. 4(a), the 2.5 eV band is indeed relatively weak at 8 K in all ZnO samples. Interestingly, at room temperature, this luminescent band is stronger and red shifted in the 120 °C sample compared to the other samples (supplementary material, Fig. S6).

As to other emission lines in Fig. 4(c), we attribute a 3.307 eV peak to two electron satellite (TES) transitions of the I₈ BEx with two phonon replicas at 3.235 and 3.163 eV, and a 3.288 eV peak to a phonon replica of the I₈ BEx. In Fig. 4(d), the dominant peaks are related to impurity BEx lines in the ground state. Following Ref. 36, we assign the I₄, I₆, and I₈ lines to excitons bound to hydrogen, aluminum, and gallium neutral donors, respectively. Note that the structure of the bound to impurities' exciton lines in the grown films is different from that of the substrate [see the supplementary material, Fig. S4(a)], except the I₈ peak, which seems to originate from the substrate. This is hardly surprising because hydrogen participates in the film growth reaction, and aluminum is an omnipresent contaminant in ZnO growth.

The low-temperature PL spectra in the range of neutral donor BEx are analyzed with a Lorentzian peak fitting procedure to obtain temperature dependence of the peak intensities, spectral positions, and full widths at half maximum (FWHM). The temperature dependences of the peak positions for all ZnO samples are shown in Fig. 5(a). The observed red shifts of the BEx emission lines I₄, I₆, and I₈ with increasing temperature are the same in all three samples, regardless of the growth temperature. The temperature dependence of the peak intensity of the I₈ BEx emission line in all ZnO samples is shown in Fig. 5(b). It can be described by^38,39

where I₀ is the peak intensity at 0 K, A is a constant, E_a is the activation energy of the thermal process, k is the Boltzmann constant, and T is the temperature. In Fig. 5(b), the black solid lines are fits to the experimental data (magenta circles). The obtained activation energies E_a (shown in the graph) are similar to all ZnO samples and are in agreement with previously reported data.³⁷ The impurity BEx lines are remarkably narrow in all three samples,with a FWHM of $∼2$ meV at 8 K [Fig. 4(d)]. For comparison, FWHM as small as of 1 to 8 meV were reported in ZnO films of the highest crystalline quality.^16,21,40

The PL and SPV measurements, thus, confirm the high quality of the grown ZnO films, which seems to be independent of the growth temperature. The presence of crystallite inclusions in the 120 °C film does not affect the impurity BEx linewidth. It is larger in the films than in the substrate though, due to the presence of PSFs and higher impurity concentrations. On the other hand, the crystallite inclusions can be linked to the intrinsic defects and their spatial distributions, as discussed above. The increased intensities of the 2.5 and 3.330 eV peaks in the 120 °C ZnO sample are the signatures on that link.

In conclusion, we have used ALD to grow ZnO thin films on single-crystalline ZnO substrates at the relatively low temperatures of 120, 150, and 200 °C. The structural and optoelectronic properties of the homoepitaxial films are studied with high-resolution TEM, PL, and SPV spectroscopy. The TEM has revealed very good crystallinity, perfect alignment of the films with the substrate, and similar defect densities in all the films and substrates. The BSFs and PSFs are identified as dominant defect types in the ZnO films. The lack of prismatic defects on the ZnO substrate highlights the differences between our ALD method and the high-pressure melt growth process. Although the 120 °C film has crystallite inclusions, the growth temperature was found to have slight effect on the overall quality of the films. Narrow BEx emission lines and sharp SPV bandgap transition confirm the high quality of the ZnO films. In addition, the PL spectra identify the presence of point defects and unintentional contaminants (Al, H, and Ga) in the grown films.

The high quality of the homoepitaxial ZnO films grown under less than optimum conditions can be explained by the self-limiting nature of the ALD growth process. In the ideal case, the film is grown one atomic monolayer at a time. This controlled slow growth gives the atoms in the new layer enough time to settle into a proper lattice position. With less controlled and faster growth processes, such as chemical vapor deposition and sputtering, the atoms have less time to reach their lowest energy positions, and thus, more thermal energy is required to grow a good crystalline film. We expect that the ALD method can be a good foundation for fabrication of high-quality ZnO nanostructures and production of ZnO p – n homojunctions.

See the supplementary material for Figs. S1–S6 referred in this Letter.

The TEM work has been performed at the Kleberg Advanced Microscopy Center. This research has been supported by the Air Force Office of Sponsored Research under Award No. FA9550-19-1-0359) (A.A.C.) and the U.S. Department of Defense under Award No. W911NF-18-1-0439 (A.P.).