1. Introduction
Oxides with spinel structures are scientifically and technologically interesting, with a wide range of applications. Zinc chromite (ZnCr
2O
4) is a spinel with a direct and wide band gap of ~3.8 eV [
1,
2], and has been proposed as a material for gas and humidity sensing [
3,
4,
5], as a depollution catalytic material for reactions like the oxidation of hydrocarbons and the oxidative dehydrogenation of hydrocarbons [
4,
6,
7], as a photocatalyst [
8,
9], and in magnetic applications [
10]. Moreover, combined with another wide band gap oxide such as zinc oxide (ZnO), it has been proposed as a catalyst for the dehydrogenative condensation of glycerol for improving the economics of biodiesel production [
11], as a photocatalytic material for wastewater treatment [
12], and in gas sensing [
13]. In such mixed oxide or heterojunction applications, the ZnO/ZnCr
2O
4 interfaces are of utmost importance. That is, a defect- and strain-free interface between ZnO and ZnCr
2O
4 is a prerequisite. For example, density functional theory calculations have shown that the (100) surface of the ZnCr
2O
4 spinel has a favorable oxygen vacancy formation energy [
14]. Thus, in order to find the optimal heterostructure interface, a large range of orientations must be studied. However, investigating the strain across a range of different thin film heterostructures having different interface orientations can be exhaustive. An alternative approach is to utilize, for example, a ZnCr
2O
4 inclusion embedded in a ZnO matrix where the different interface orientations are readily available. This approach can be realized by, for example, ball-milling, dry pressing, and sintering at high temperatures [
15].
To study the strain across heterojunction interfaces, one can adopt high-angle annular dark-field (HAADF)-scanning transmission electron microscopy (STEM), which makes use of incoherently and elastically scattered electrons to obtain images with high spatial resolution and atomic-number (Z) contrast [
16,
17,
18]. Characterization of strain associated with extended defects has been further boosted by the successful application of geometric phase analysis (GPA) in probe-corrected high-resolution HAADF images [
19,
20,
21].
In addition, as the size of a semiconductor particle reduces down to nanoscale (i.e., below 10 nm), its band gap increases as a result of quantum confinement effects, providing intriguing optical and electronic properties [
22,
23,
24]. Previously, UV-Visible reflectance spectroscopy was employed to analyze the average band gap of ZnCr
2O
4 nanoparticles [
2]. However, conventional techniques such as photoluminescence (PL), cathodoluminescence (CL), optical absorption, and UV-Visible reflectance spectroscopy have difficulty in measuring the band gap of one single nanoparticle because of poor spatial resolution. The rapid advances in the field of monochromatic electron energy-loss spectroscopy (EELS) during the past decade have attracted wide interest in EELS band structure determination [
25,
26,
27]. By means of monochromated EELS in combination with probe-corrected STEM, recent research has made band gap mapping on the nanometer scale possible [
28], opening the possibility of observing quantum confinement as the size of the particles are reduced. However, due to experimental and data process complexities, to date few EELS investigations have been carried out which measured the band gap of single embedded nanoparticles.
In this work, probe-corrected atomic-resolution STEM imaging coupled with GPA was utilized to characterize the structure and strain at ZnO/ZnCr2O4 interfaces. Furthermore, core-loss EELS revealed a chemically pure ZnCr2O4 phase. In addition, making use of monochromated low-loss EELS, the absorption onset originating from the ZnCr2O4 band gap was extracted, and a two-dimensional mapping of single particles with a diameter around 500 nm was demonstrated. These results also provide the necessary foundation to study smaller particles in the future, for the potential direct observation of quantum confinement leading to an increase in the band gap.
2. Experimental Methods
The specimen of ZnO with ZnCr2O4 nano-inclusion was prepared as follows. ZnO powder was mixed with α-Cr2O3 powder (ZnO: Sigma-Aldrich, St. Louis, MI, USA, 99.999% 6.10 g; α-Cr2O3: Sigma-Aldrich, 99.95%, 0.76 g), and the molar ratio ZnO:α-Cr2O3 was 15:1. The mixed powders were ball-milled for 3 h, then uniaxially dry-pressed at a pressure of 2805 MPa, and sintered at 1350 °C for 24 h in air, with a temperature increase and decrease rate of 450 °C/h.
Parts of the sintered specimen were ground into powder for X-ray diffraction (XRD) measurement. Silicon standard ASM 640d (NIST, Gaithersburg, MD, USA) powder was mixed with the specimen as a reference for calibration of the peak positions. The XRD experiment was performed on Bruker ASX D8 Discover (Billerica, MA, USA) with a Cu Kα1 source. Unit cell parameters were determined via Rietveld-software TOPAS Version 4.2.
The specimen for STEM studies was prepared from the sintered material by mechanical cutting, grinding/polishing, and finally ion beam thinning. Immediately before carrying out the STEM experiments, the specimen was cleaned in a Fischione Model 1020 plasma chamber. All the images were acquired by a probe-corrected and monochromated FEI Titan G2 60-300 (Hillsboro, OR, USA) equipped with four FEI super-X Energy-dispersive X-ray spectroscopy (EDX) detectors. The spatial resolution was approximately 0.8 Å for STEM imaging at a high tension of 300 kV. Energy-dispersive X-ray spectroscopy (EDX) was utilized to map the various elements. For simultaneous HAADF (Fischione Model 3000, Export, PA, USA) and annular bright field (ABF) observations, the probe convergence angle was set at 22 mrad, while the collection angles were 98.7–200 mrad and 10.6–21.5 mrad, respectively. To directly address the atomic structure, STEM image simulations were carried out using the QSTEM program [
29] based on the multi-slice method.
All EELS experiments were performed using a Gatan GIF Quantum 965 (Gatan Inc., Pleasanton, CA, USA). For energy-loss near-edge fine structure (ELNES) analysis the energy resolution was approximately 0.8 eV as determined by the full width at half maximum of the zero-loss peak (ZLP). Monochromated EELS were conducted to measure the band gap of ZnCr
2O
4 nanoparticles two-dimensionally. In order to limit the Cherenkov radiation effect, an accelerating voltage of 60 kV was utilized: the spectral resolution was approximately 0.15 eV. The TEM specimen was finally thinned to be about 30 nm, further reducing the Cherenkov retardation losses efficiently. The theoretical spatial resolution of EELS is defined by the inelastic delocalization length (
). According to Equations (4)–(7) reported by R.F. Egerton [
30], for 60 keV incident electrons, EELS spatial resolutions in measuring band gaps of ZnO (~3.3 eV) and ZnCr
2O
4 (~3.8 eV) are about 5.4 nm and 4.9 nm, respectively, which is highly suitable to characterize the band gap of the ZnCr
2O
4 particles in this work. The band gap extraction procedures used here were the same as those described previously [
28,
31,
32]. In brief, a power-law model was employed to remove the background from the ZLP. The fitting range for background subtraction was 2.4–2.9 eV. Then, the direct band gap value was extracted by performing a curve fitting based on the parabolic band approximation.
Strain field across ZnO/ZnCr
2O
4 interfaces were evaluated with a GPA script developed for the DigitalMicrograph package (Gatan Inc., Pleasanton, CA, USA) [
19,
20,
21,
33,
34,
35]. This Fourier-space image processing technique translates the displacements affecting specific periodicities in the image into phase shifts of the average or reference passband, which are then mapped across the image and used to calculate the local strain. Real-space peak detection is not required, and there is no need to assign appropriate sublattices, therefore robust and straightforward two-dimensional strain mappings can be obtained from atom-resolved images of interfaces provided some degree of coherency.
3. Results and Discussion
XRD measurement of the sample described in the experimental section is shown in
Figure 1. Two crystalline phases were identified: a wurtzite ZnO and a spinel ZnCr
2O
4 phase. Except for the Si pattern used as a reference, no other phases were observed. Furthermore, the peaks were sharp and narrow, indicating high crystalline quality. The ZnO pattern was characteristic of the non-centrosymmetric wurtzite structure, while ZnCr
2O
4 crystallized in the antiferromagnetic spinel structure. The refined cell parameters of the ZnO and ZnCr
2O
4 can be found in
Tables S1 and S2 (Supplementary Materials), respectively. The atomically resolved STEM images in
Figure S1 also confirm the existence of these two phases. Experimentally determined unit cell dimensions for the ZnO phase were
a =
b = 3.2505(4) Å and
c = 5.2059(6) Å. There was a slight shift from the original unalloyed ZnO with
a =
b = 3.2555(2) Å and
c = 5.2152(3) Å [
36]. Note that the ionic radius of Cr
3+ (0.064 nm) was smaller than that of Zn
2+ (0.074 nm) in the tetrahedral coordination. The peak shift and decreased lattice constant
c confirmed the successful incorporation of Cr
3+ ions into the ZnO host lattice, substituting for the Zn
2+ sites [
37]. The concentration of Cr was approximately 0.8 at% based on the EDX quantification in this work.
The inclusion of ZnCr
2O
4 particles in a matrix of ZnO can be seen in the annular dark field (ADF) image in
Figure 2a and in the corresponding EDX maps of Zn, Cr, and O in
Figure 2b–d, respectively. The sizes of the two particles in
Figure 2 were 450 nm (upper) and 500 nm (bottom) in diameter, respectively. EDX mapping such as those in
Figure 2b–d confirmed that the diameter of ZnCr
2O
4 nanoparticles was approximately 100–500 nm. Taking advantage of probe-corrected and monochromated STEM-EELS, we performed two-dimensional band gap measurements of ZnCr
2O
4 nanoparticles in ZnO matrix.
Figure 2e displays the band gap mapping outcome, which was extracted from the same region as elemental EDX maps. The size of this band gap map was approximately 592 nm (width) × 1073 nm (height). The yellow and blue colors represent high and low band gaps, respectively. Interestingly, the contrast shift observed in the ADF and EDX images was closely followed by a change in the absorption onset from the increased band gap of ZnCr
2O
4 as compared to ZnO. With a fitting range 2.4–2.9 eVfor background subtraction, the average band gap of ZnO was found to be approximately 3.22 eV, with standard deviation of 0.01 eVas extracted from 80 pixels. This is consistent with previous measurements [
28,
31]. The average band gap of ZnCr
2O
4 particles was measured to be 3.84 ± 0.03 eVs, being in line with existing reports for bulk values [
1,
2]. These results suggest that STEM-EELS produced reliable band gap measurements of single embedded particles. Furthermore, both sharp and non-sharp interfaces were available by considering different edges of the ZnCr
2O
4 nanoparticle—see the labels “
S” and “
N”, respectively (
Figure 2a). In the EELS core loss region across a sharp interface (see Figure 4), the delocalization effect was reduced, and an improved lateral resolution could be obtained as compared to the low-loss EELS, but also as compared to the secondary X-ray emission process probed in EDX.
Figure 3 shows low-loss EELS spectra extracted from the spectrum image used in
Figure 2e. The spectra were taken from one pixel at the interface, and the two immediately adjacent pixels—one on the ZnCr
2O
4 side and one on the ZnO side. The three neighbor spectra positions are indicated by the white arrow in
Figure 2e. As predicted, it can be clearly seen from the onsets that the band gap of ZnCr
2O
4 was higher than that of ZnO. The size of each pixel is about 22 nm × 22 nm, which means that for both materials
was shorter than the size of each pixel. Thus, the ineleastic delocalization was limited to the interface pixel, where the spectrum revealed a mixture of the ZnO and ZnCr
2O
4 due to the inelastic delocalization of the bandgap signal [
30].
The core-loss EELS spectra of O and Cr obtained across a sharp ZnO/ZnCr
2O
4 interface are illustrated in
Figure 4. It can be seen from
Figure 4a that the O-
K edge was detected in both ZnO and ZnCr
2O
4, with similar intensity in ZnCr
2O
4. In
Figure 4b, the Cr-
L2,3 edge was high in intensity at the ZnCr
2O
4 phase, but this signal dropped abruptly across the sharp interfacial region, and disappeared when it entered ZnO. This confirms the non-existence of other phases, as discussed previously. It is well established that the EELS O-
K and Zn-
L2,3 edges probe the unoccupied O 2
p and Zn 3
d states, respectively. We observed significant differences between ZnO and ZnCr
2O
4 as shown in
Figure 4c, indicative of the different bonding schemes in the two materials. For the O-
K edge, there was only one peak observed in the ZnO matrix, in accordance with previous measurements [
38,
39,
40], while an obvious splitting occurred in ZnCr
2O
4, agreeing well with earlier reports [
41,
42,
43], as also illustrated in
Figure 4d. This was because in the ZnO unit cell, there was only one tetrahedron, which was composed of Zn and its four surrounding O, and each Zn-O bond exhibited a length of 1.98 Å. In comparison, both tetrahedral and octahedral configurations existed in the ZnCr
2O
4 unit cell. One Zn and its four surrounding O constitute a tetrahedron, while one Cr and its six surrounding O form an octahedron. The bond lengths of Zn-O and Cr-O were 2.04 Å and 1.95 Å, respectively. Furthermore, the O-
K edge shifted from one peak to two peaks across the interface region. At the same time, the intensity of the Cr-
L2,3 edge reduced progressively from ZnCr
2O
4 to ZnO at the interfacial region and disappeared when the probe was moved into the ZnO matrix. The oxidation states of the transition metal oxides to a large extent determine their physiochemical properties. ELNES analysis of the EELS spectrum, with its special valence sensitivity to 3
d transition metals [
44,
45,
46], is capable of revealing the electronic structure of target atoms (i.e., valence state, atomic coordination, and spin state). For instance, making use of the white line ratio (
L3/
L2) of the Cr-
L2,3 edge, a number of studies have successfully analyzed the oxidation state in chromium oxides [
42,
43,
47]. According to our experiments, the
L3/
L2 ratio remained at approximately 1.6 across the interface, indicative of no valence change for Cr. This was confirmed by the high similarity of the normalized intensities of
L3 and
L2 edges between the ZnCr
2O
4 phase and the interface position (
Figure 4e).
The structure and strain of some sharp ZnO/ZnCr
2O
4 interfaces (similar to the interface “
S” as marked in
Figure 2a) were found and investigated in this work.
Figure 5 presents a typical sharp interface of ZnO [2
3]/ZnCr
2O
4 [1
0]. This interface area is indicated in
Figure S2 (Supplementary Materials). The size of the HAADF image in
Figure 5a is about 9.7 nm × 9.7 nm. The interface abruptness can be seen in
Figure 5a. In order to clearly reveal the structure, the ABF image, as displayed in
Figure S3a (Supplementary Materials), was recorded simultaneously with the HAADF image. There were dislocations at the lattice mismatched interfacial region, as indicated in the Fourier-filtered image produced with the (
011)
ZnO/(222)
ZnCr2O4 lattice fringes (
Figure 5b). The d-spacings of the (
011)
ZnO and (222)
ZnCr2O4 planes were similar, but the nominal mismatch was enhanced to approximately 9% by the 9° angle adopted at the interface. The strain was localized at misfit dislocations with a spacing in line with the expected mismatch. The localization of the strain was quantified by GPA using the (
011)
ZnO/(222)
ZnCr2O4 reflections, as displayed in
Figure 5c. An unstrained region selected from homogeneous ZnCr
2O
4 was used as reference. A close-up of the white square in
Figure 5c is given in
Figure 5d showing the strain field of a positive edge dislocation with projected Burgers vector
b = [222]
ZnCr2O4.
HAADF imaging was also used to investigate ZnO/ZnCr
2O
4 interfaces in projection such as the interface “
N” shown in
Figure 2a. A representative example is given in
Figure 6, which illustrates the atomically resolved HAADF image and its filtered image at the non-sharp interface of ZnO [1
10]/ZnCr
2O
4 [112] in a non-edge on orientation. An overview HAADF image of the particle and the positions of the interface region can be found in
Figure S4 (Supplementary Materials). The HAADF image size in
Figure 6a is approximately 12.5 nm × 12.5 nm. The ABF image, which was acquired simultaneously with the HAADF image, is given in
Figure S5a (Supplementary Materials) as a support for the structure determination. The Fourier-filtered image of (0002)
ZnO/(
20)
ZnCr2O4 lattice fringes in
Figure 6b evidences the existence of parallel dislocations running at the interface. The angle difference between the (0002)
ZnO and (
20)
ZnCr2O4 planes was negligible, but there was a large lattice mismatch, and the d-spacings were 2.6 Å and 2.94 Å, respectively. This resulted in several misfit dislocations, as marked by blue arrows in
Figure 6b that ensured the semi-coherency of the interface but imply chemical bond rearrangement. In general, more continuity at an interface leads to less rearrangement and reduced interfacial energy [
48]. Other interfaces have been studied as well (not shown), and the strain deduced accordingly. However, dislocations remain a key parameter for the introduction of strain in the ZnO/ZnCr
2O
4 heterojunction interface.