4-1 Results of X-ray measurements
4-1-1 Growth temperature dependent crystallinity of the ZnO films
ZnO films were deposited at a substrate temperature ranging from 200 to 500oC.
Two batches of samples were grown on two composite substrates, which were prepared nominally under the same deposition conditions.
2 3
Figure 4-1 illustrates the radial scan along surface normal of the ZnO film grown at 300oC, where the abscissa qz = 4πsinθ/λ denotes the momentum transfer along the surface normal and λ is the incident x-ray wavelength. According to JCPDS (Joint Committee on Powder Diffraction Standards), the three pronounced peaks are assigned as the (0002) reflection of wurzite ZnO, (222) peak of cubic γ-Al2O3 and (111) peak of
cubic Si, elucidating the cube-on-cube growth of γ-Al2O3 on Si and a crystalline orientation relationship of along surface normal. The
pronounced thickness fringes observed near the γ-Al2O3(222) reflection indicates the sharpness of the γ-Al2O3 interfaces; the fringe period yields a layer thickness of ~15.3 nm. From the FWHM of the γ-Al2O3(222) reflection, we derived the vertical coherence length of the buffer layer, using the Scherrer’s equation, to be ~15 nm, which is close to the layer thickness. This implies that the structural coherence of the buffer
ZnO γ-Al O Si
(0002) || (111) || (111)
layer is maintained over almost the entire layer thickness.
Azimuthal cone scans (φ-scans) across the off-normal ZnO{1010}, γ-Al2O3{440}, and Si{220} reflections, as shown in Fig. 4-2, were measured to examine their in-plane epitaxial relationship. Six ZnO diffracted peaks evenly spaced 60o apart confirmed that ZnO film with a six-fold rotational symmetry against surface normal was grown epitaxially on the γ-Al2O3/Si(111) composite substrate. Furthermore, two sets of peaks with 3-fold symmetry were observed in the γ-Al2O3{440} φ-scan, revealing the cube-on-cube growth of γ-Al2O3 on Si and the coexistence of two in-plane rotated variants. The dominant one has the same angular position as that of the Si{220}, A-type (111) orientated domain, and the minor one has its in-plane orientation rotated 60o from that of Si substrate, B-type domain [32]. These results suggest that the in-plane epitaxial relationship of this hetero-epitaxial system follows {1010}ZnO ||
-Al O2 3
{224}γ or
-Al O2 3
{422}γ || {224}Si. In this geometry, the two-dimensional
hexagonal unit cell of the γ-Al2O3(111) plane is aligned with the ZnO basal plane having its lattice constant equal to √2·a(γ-Al2O3) = 11.186 Å, about 3.45 times larger than that of ZnO.
1.0 1.5 2.0 2.5 3.0
Fig. 4-1 Radial scan along the surface normal of a 300 nm thick ZnO layer grown on γ-Al2O3/Si(111) composite substrate
0 60 120 180 240 300 360
Fig. 4-2 The profile of φ–scans across ZnO{1010}, γ-Al2O3(440) and Si{220} reflections
Within the employed growth temperatures, all the samples exhibit the same structural characteristics with a small variation of less than 0.2% in ZnO lattice parameters. The lattice parameters of ZnO bulk are a = 3.2438 Å and c = 5.2036 Å, derived from the XRD data of a ZnO wafer. As compared with the lattice parameters of the grown ZnO films, we found that all the ZnO films experience a tensile strain (~0.28%) in the lateral direction and correspondingly a compressive strain (~0.19%) along the surface normal. To characterize the structural quality of the film under different growth temperature, the θ-rocking curve of ZnO(0002) reflection was performed. The small values of FWHM, varying between 0.32o and 0.61o with the minimum obtained at 300oC, manifest good crystalline quality of ZnO epi-layer even for such a thin film with thickness of ~0.3 μm. This value is as good as that of the
ZnO epi-film grown on non-oxide buffer layer, such as ZnS [33], Al [34] and 3C-SiC [35]. The FWHMs of φ-scans across the off-normal ZnO(1010) reflection fall between 1.4o and 3.8o. It is worth noting that for the ZnO epi-layers grown on γ-Al2O3/Si(111) under different growth temperatures, those have smaller FWHM of
(0002) θ-rocking curve always have larger width in φ-scans across the off-normal )
0 1 0 1
( reflection, as shown in Fig. 4-3(a) and (b). This is different from ZnO grown on other substrates, such as c-plane sapphire, where both FWHMs always exhibit the same trend of increase/decrease with growth conditions [36]. Furthermore, we grew
ZnO film on two batches of buffer layers and found the structural quality of ZnO is sensitive to the structural perfection of γ-Al2O3 buffer layer. The FWHM of θ-rocking curve of γ-Al2O3(222) of the first and the second batches are 0.023o and 0.024o, respectively, while the FWHM of φ-scan across the off-normal γ-Al2O3(440) reflection are 3.97o and 4.27o for the first and the second batches, respectively. The quality of buffer layer of the first batch is better than that of the second batch. The ZnO films grown on the first batch are found averagely better than that of the second batch, especially for the FWHMs of φ−scan of ZnO(1010). Nevertheless, the opposite trend of variation in Δθ and Δφ observed in both batches of samples. The factors, Δθ and Δφ are associated with different types of structural defect and the opposite variation trend of Δθ and Δφ allows us to identify their influence on the optical properties of ZnO
film, which will be discussed later.
200 250 300 350 400
Fig. 4-3(a) Correlation between growth temperature and ZnO Δθ(0002) / ΔΦ )
0 1 0 1
( distribution for the first batch of substrate
200 250 300 350 400 450 500
0.42
Fig. 4-3(b) Correlation between growth temperature and ZnO Δθ(0002) / ΔΦ
4-1-2 Analysis of threading dislocation density by XRD
The major defect structure in the ZnO film is TDs. In our case, the majority of TDs have their lines lying along the [0001] growth direction that is the same as the ZnO film grown on commonly used c-sapphire [34]. For a (0001)-oriented thin film with wurtzite structure, the TDs are classified into three different types according to the direction of the corresponding Burgers vector (b) relative to the [0001] line direction.
They are edge, screw and mixed TDs characterized with bE = 1/3⋅<1120>, bC = <0001>
and bM = 1/3⋅<1123>, respectively, where bM is the mix of bE and bC. In order to address the defect structures of wurtzite ZnO by XRD, rocking curves of the ( 0 0)h h and (000l) reflections were measured. Note that the broadening of and (000l) rocking curves are associated with the lattice misalignment along the in-plane and growth directions, respectively. In other words, pure edge TDs twist the surrounding ZnO lattice about [0001], leading to the formation of vertical grain boundaries [37, 38, 39] and the crystalline planes are distorted. On the other hand, pure screw TDs tilt the ZnO lattice and generate a pure shear strain field [40], causing the (000l) planes being deformed. Thus, we measured the and (000l) reflections to investigate the influence of edge and screw TDs, respectively. The FWHMs of the θ-rocking curves of (000l) Bragg peaks reflect the lattice twist/tilt and the line
width of their radial scans are related to the lateral/vertical inhomogeneous strain field
/
( 0 0)h h
( 0 0)h h
( 0 0)h h
( 0 0)h h
and domain size.
To obtain a quantitative results, we employed the Williamson-Hall (WH) plot (Δq vs.
q, where q = 4πsin(2θ/2)/λ denoting the scattering vector andΔq is the line width in q along the radial direction) to separate the reflection peak broadening due to finite structural coherent length from strain-induced broadening. According to Eq. 2-6, the inverse of the ordinate intercept yields the coherence length which corresponds to the effective domain size and the slope yields the root-mean-square (rms) inhomogeneous strain. Figures 4-4 (a) and (b) are the typical WH plots of radial scans along the ZnO (h0h0) and (000l), respectively. For ZnO film grown at 300oC, the coherence length along surface normal is ~293.6 nm, indicating its structure maintains coherent almost over the entire film thickness. The in-plane domain size is 117 nm. The average lateral strain is ~3.83×10-3 about three times that along the surface normal (~1.3×10-3), indicating that the dominant cause of ZnO lattice distortion comes mainly from edge dislocation. Figures 4.5 (a) and (b) are plots of Δqt vs. q (where Δqt = q·Δθ denotes the
spread of θ in the transverse direction) for θ−rocking scans across the (h0h0) and (000l), the slopes yield the spread of twist and tilt angles, respectively. The twist angle (αΦ) and tilt angle (αΩ) of ZnO fall within the ranges of 1.38o~3.8o and 0.28o~0.53o, respectively. Note that for samples grown under different temperatures, the ZnO films exhibit the opposite trend of variation for tilt and twist angles, which are similar to the
trend of θ−rocking curve linewidths of ZnO(0002) and (1010) reflections because Δθ(0002) and Δφ(1011) bear the same physical meaning as the tilt and twist angles. The
TD density can be estimated from the corresponding Burgers vectors and tilt or twist angles. The density of screw type dislocations, Ns, is obtained using the equation:
, where αΩ is the tilt angle and bC is the magnitude of the corresponding
Burgers vector bC, which is [0001] with bC = 0.519 nm in our case [41]. The edge type TD density, NE, can be calculated by adapting , where
α
φ , bE and Lare twist angle, the length of corresponding Burgers vector (bE = 0.325 nm) and correlation length along the in-plane direction, respectively [39]. This formula is based on the model of dislocation piling up in small angle boundaries and forming subgrains with an average size L along the in-plane direction. As for our ZnO films grown at different temperatures, it is noticed that the density of edge TDs (1.39~9.97×1010 cm-2) is about 10 times higher than that of screw TDs (2.14~5.75×109 cm-2). Thus, the dominant type of dislocation of ZnO grown on γ-Al2O3/Si(111) is
0 2 4 6 8 10
Fig. 4-4 Williamson-Hall plots of ZnO layers grown at various temperatures.
The symmetric radial scans were measured for (a) (h0h0) surface peaks and (b) (000l) normal reflections. Lines are linear fits of the data.
0 1 2 3 4 5 6 7 8 (a) surface peaks and (b) normal reflection
4-2 Structural characterization by TEM
4-2-1 cross sectional TEM
Illustrated in Fig. 4-6 (a) is a selected area electron diffraction (SAED) pattern along
[112] direction. The diffraction peaks from Si substrate, γ-AlSi 2O3 buffer and ZnO epi-layer can be well identified and confirmed. The crystalline orientation relationship is (0002) || (222)ZnO γ-Al O2 3|| (111)Si at surface normal direction and
ZnO -Al O2 3 Si
0} || {440}γ &{220}
{112 at in-plane direction that are consistent with the
results from XRD observation. Figure 4-6(b) is the high magnification TEM image taken at the interfacial region with [112] projection. It reveals unequivocal Si interfaces of ZnO/γ-Al2O3 and γ-Al2O3/Si; no intermediate layer was observed in either interface. These results demonstrate that the growth of high-quality epi-ZnO film on Si by using a γ-Al2O3 buffer layer is achievable and the buffer serves as a good template for subsequent ZnO growth.
4-2-2 Analysis of threading dislocation density
The types of dislocations in ZnO films were further characterized by cross-sectional TEM under a two-beam contrast condition. The diffraction contrast of dislocations in TEM images arises from the bending of lattice planes caused by the dislocation induced strain field. For pure screw dislocations, planes with their normals perpendicalar to the Burgers vector bC are undistorted. Therefore, all electron beams that are diffracted by planes containing bC show no image contrast. With g being the operating diffraction vector, and bC the Burgers vector of the dislocation, the invisibility criterion for pure screw TDs is g • bC = 0. For pure edge dislocations, only lattice planes perpendicular to the dislocation line direction will be undistorted and edge dislocations will be invisible in images that are formed using electron beams diffracted from these planes.
A pure edge dislocation gives rise to two displacement components of lattice. One is parallel to the Burgers vector bE, and the other perpendicular to the slip plane, those normal can be expressed by bE × u with u denoting the unit vector along the positive direction of the dislocation line. Therefore, to ensure complete invisibility for an edge dislocation, in addition to g • bE = 0, the condition g • (bE × u) = 0 has to be satisfied, too [27]. Based on these invisibility criteria, we can distinguish the types of the TDs by analyzing images taken with different operating diffraction vector g. Figures 4-7(a)-(c) show the images of the ZnO film grown at 300oC taken with the diffraction
vector g equal to ZnO(0002), (1120), and (1122), respectively. According to the invisibility criteria, g • b = 0, screw type dislocations with bC = <0001> are invisible with g =(1120) but visible under g = (0002) and (1122). In contrast, edge type dislocations with bE = 1/3⋅<1120> and (bE × u) // <1100>should be invisible under g
= (0002) but visible under g = (1120) and (1122). As to the TDs of mixed type
character with bM = 1/3⋅<1123>, they are visible under all three g vectors. In all three micrographs, TDs are seen as dark lines stemming from the ZnO/γ-Al2O3 interface
with their dislocation lines primarily along the [0001] direction. The dislocation density, D, was determined by formula, D = n/lh, where n is the number of dislocations, l is foil length and h is foil thickness. The foil thickness was determined by measuring
the number of extinction distance fringes in two-beam images (g = (0002)) and using the extinction distance as a reference to derive specimen thickness. We calculated the densities of edge, screw and mixed TDs to be approximate 1.04×1010 (~ 27%), 3.46×109 (~ 10%), and 2.42×1010 cm-2 (~ 63%), respectively. Thus, about 90% of total TDs contain the edge component consistent with the result calculated from XRD data that edge is the dominant type TDs.
G=(1120)
G=(1122) G=(0002)
‐
‐ (a)
(c) (b)
Fig. 4-7 Two-beam bright field cross sectional micrograph of a ZnO thin film taken
with (a) g = (0002), (b) g = (1120) and (c) g = (1122).
4-3 Optical characterization by PL
4-3-1 Photoluminescence spectra of as-grown films
We carried out low temperature (LT) PL at 15K to characterize the optical properties of the ZnO films. The PL spectra of ZnO films are shown in Fig. 4-8. Regardless of the growth temperatures, the main features on PL spectra are common for all samples.
The spectra can be divided into two parts: sharp near band-edge emission (NBE) and broad deep level emission (DLE), which are centered at ~3.365 and ~2.196 eV, respectively. The dominant peak at NBE region is attributed to the recombination of excitons bound to neutral donor (DoX) [42] and the broad DLE results from point defects such as O vacancies and Zn interstitials [43]. The assignments of NBE peaks are also shown in Fig. 4-9, which is the PL spectrum of the ZnO grown at 300oC. The peak at 3.375 eV was designated as the free A-exciton (FXA) line; the binding energy of the corresponding A-exciton was obtained to be 58.875 meV by fitting the temperature dependent PL data using the Arrhenius expression, as shown in Fig. 4-10. The dominant peak at 3.364 eV in the NBE region was assigned to the recombination of excitons bound to neutral donor (DoX) [44] and its FWHM is 9.4 meV. The DoX emission accompanied with single phonon (DoX–1LO) and dual phonon (DoX–2LO) replica appear at 3.288 and 3.215 eV. The peak at 3.23 eV is attributed to the donor–acceptor pair (DAP) transition. Another strong line at 3.328 eV originates from
the transition involving radiative recombination of an exciton bound to a neutral donor (DoX) and leaving the donor in the excited state. This process is also known as the two-electron satellite (TES). We made such an assignment based on the ratio of donor binding energy to exciton binding energy 0.35, as reported by Teke et al [42].
Similarly to the E2-high mode in the Raman spectra of ZnO thin films [43], the position of the DoX is sensitive to the strain state of the films. The observed red shift of DoX energy relative to that of ZnO bulk wafer (3.365 eV), marked by the dashed lines, is consistent with the biaxial tensile strain determined by XRD measurements.
1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6
400 o C
300 o C
Normaliz ed int. (a.u.)
Photon energy (eV)
200 o C
(c) (b)
16.09 meV 9.43 meV
Near band edge
Deep level
13.44 meV
(a)
Fig. 4-8 Typically PL spectra measured at 15K for ZnO epi-layers deposited on γ-Al2O3/Si (111) at (a) 200oC, (b) 300oC and (c) 400oC, respectively. The spectra can be divided into two major parts: NBE and DLE. The FWHMs of NBE are marked near the peaks.
3.00 3.05 3.10 3.15 3.20 3.25 3.30 3.35 3.40 3.45 3.50
Fig. 4-9 The extended spectra of NBE emission of ZnO grown at 300oC. The dashed line marks the DoX peak position of bulk ZnO.
The binding energy of exciton : 58.875 +- 5.28481 meV
Fig. 4-10 Dependence of FXA PL peak intensity on sample temperature for the ZnO film grown at 300oC.
4-4 Correlation between structural and optical properties of ZnO films
The ZnO films with narrower FWHM of NBE and lower ratio of DLE to NBE intensity (IDLE/INBE) are considered to have better optical properties. For the ZnO films grown at different temperatures and on different batches of γ-Al2O3/Si substrates, we found that the optical performance of films grown at 300oC always exhibits the best characteristics in NBE region but the worst in DLE region among the samples grown on the same substrate. In other words, FWHM of NBE and (IDLE/INBE) ratio exhibited an opposite trend of variation vs. growth condition. From the previous XRD results, we also observed the opposite trend of Δθ(0002) and Δφ(1011) for our specimen. The (IDLE/INBE) ratio versus Δφ(1011), the FWHM of the azimuthal scan across ZnO off-normal (1010) reflection, and the FWHM of NBE versus Δθ(0002), the FWHM of the rocking curve of the (0002) specular reflection, are illustrated in Figs. 4-11(a) and (b), respectively. Here, Δθ(0002)/Δφ(1011) bears the same physical meaning as the
tilt/twist angle. Besides, tilt angle is coupled with screw type TDs while twist angle is coupled with edge type TDs. It suggests that we can individually observe the influence of different types of TDs on the optical property of ZnO film grown on γ-Al2O3/Si(111). Figure 4-12(a) shows the correlation between (IDLE/INBE) ratio and edge dislocation density. It elucidates that the ZnO with larger (IDLE/INBE) ratio
possesses the higher edge dislocation density. Figure 4-12(b) displays the FWHM of NBE as a function of screw dislocation density. Similarly, it exhibits a positive correlation between them. In contrast, both (IDLE/INBE) ratio vs. screw dislocation density and NBE line width vs. edge dislocation density plots do not show any clear correlation, as shown in Figs. 4-11(c) and (d). Thus, it is evident that the NBE is dominantly affected by screw type TDs and DLE is mainly influenced by edge type TDs.
It is speculated that the edge TDs induced the aggregation of point defects due to stress field near the dislocation core and result in the enhancement of DLE intensity.
Consequently, reducing TDs density of both types in the growth of ZnO epitaxial film is still an important issue in the applications to photonic devices.
1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 0.0
0.4 0.8 1.2 1.6
Ratio ( I
DLE/I
NBE)
ΔΦ
(1011)(deg.)
(a)
0.35 0.40 0.45 0.50 0.55 0.60 5
10 15 20 25 30 35 (b)
Δθ
(0002)(deg.)
FWH M of NB E
(meV)Fig. 4-11 The dependence of (IDLE/INBE) ratio on Δφ(1011) (a) and the dependence of NBE width Δθ(0002)(b). The dashed curves were plotted to guide the eyes.
0 2 4 6 8 10 12 14 16 18
edge dislocation density X 1010 cm-2 (a)
screw dislocation density X109 cm-2 (b)
screw dislocation density X109 cm-2 (c)
edge dislocation density X 1010 cm-2 (d)
Fig. 4-12 PL (IDLE/INBE) intensity ratio vs. edge dislocation density plot (a); NBE FWHMs vs. screw dislocation density (b), the dashed curves were fitted to guide the eyes. (c) and (d) are the plots of exchanging abscissa of (a) and (b). There is no obvious correlation after exchanging the abscissas.