Results and Discussion - A novel hybrid energy harvesting cell by combining thermoelectric and

3-1 Thermoelectric devices

Highlighted results that are World-leading or Taiwan-leading:

1. Two types of homologous In2O3(ZnO)n nanowires with vertical and horizontal superlattices first synthesized

Figure 1 shows the FE-SEM image of the as-grown products, consisting of wire- and belt-like structures. Of which, the typical nanowires are characterized by 5-15μm in lengths and 50-120 nm in diameter, which are composed of homologous In2O3(ZnO)n superlattice structures. Figure 2(a) shows the low magnification TEM image of such a single nanowire. Like the other nanowires, this nanowire has a gold nanoparticle at its tip acting as a catalyst via the vapor–liquid–solid growth mechanism. Figure 2(b) is the HRTEM image, showing modulation contrast along the longitudinal axis, confirming to a In2O3(ZnO)n superlattice structure. This superlattice structure consists of In-O and In/Zn-O layers stacked alternately along the c axis with the spacing between the In planes and the nearest Zn planes being 0.32nm, which is larger than the Zn (0002) interplanar spacing of 0.27nm. The increase in the d spacing is due to the larger ionic radius of In than that of Zn. In addition, the HRTEM image reveals periodic superlattice structure with identical distance between two In-O layers, being around 7-8 In/Zn-O atomic layers. Therefore, this nanowire can be described by a precisely defined unit cell as expected for the In2O3(ZnO)n compounds in theory. Moreover, this structure is obviously reflected in Figure 2(c) of the SAED pattern, with a series of satellite spots between two main spots of ZnO.[11]

By referring to the enlarged HR-TEM, shown in Figure 2(d), M-modulated structures in zigzag shape become evident within each ZnO slabs. Elastic strain can build up in the ZnO layer (with the strain energy increasing as increasing the ZnO layer number n) since the ideal in-plane lattice constants of this plane differ significantly from that of ZnO. Hence, the formation of M modulation in the In/Zn-Olayer can reduce the total energy of the structure. From the total energy calculation results[12], the total energy of the modulated structure is lower than the structure with the flat-layer In-O boundary. There are reasons to explain why modulated structure is energetically favorable. First, considerable strain will be induced around the In–O for the flat In–O layers inside the ZnO slab resulting from highly mismatched lattices with ZnO. By forming the modulated structure, this strain can be significantly reduced, as In and Zn atoms arrange alternatively. Besides, O atoms are threefold coordinated in the flat In–O layers and becomes energetically unfavorable whereas in the modulated structure, all O atoms in the zigzag boundary are fourfold coordinated, which is energetically more favorable[12]. In addition, the In-O octahedral layer acts as the inversion boundary so that the polarities of ZnO at its two sides are inverted. Consequently, the polarities must be inverted again inside the Zn/In-O slabs. In other words, the zigzag boundaries provide the required second polarity inversion[13].

To confirm the microstructure of the In2O3(ZnO)n nanowires, Z-contrast STEM imaging is employed as shown in Fig. 2(f), where the intensity is approximately proportional to the square of the atomic number of the constituent atoms. Thus, using Z-contrast imaging, the location of In (Z= 49) can be unambiguously determined, but oxygen cannot be imaged due to its relatively small atomic number[1]. Fig. 2(e) clearly shows the presence of In-enriched layers (bright lines) oriented perpendicular to the [002] direction. As observed in Fig. 2(f), the In atoms segregate on individual planes and are separated by wurtzite In/Zn-O slabs. Only single layers of In atoms are observed, consistent with the layers being composed of octahedrally-coordinated InO2-, which has been shown to be the most stable configuration for In within the superlattice structure.

Besides the nanowires with a longitudinal superlattice structure, nanowires with transverse superlattice structure have also been observed in the product. As shown in Figure 3(a), layers do not stack along the growth direction [10-10] of the nanowire but along the transverse direction of the ZnO [0001] direction, which is confirmed the corresponding SAED pattern, in Figure 3(b). If we enlarge the Z-contrast image, shown in Figure 3(c), the zig-zag modulation is more apparent than the transverse superlattice structure. However, unlike the longitudinal superlattice structure, the In-O stacking sequence along the c axis is not identical. Therefore, this nanowire cannot be described by a precisely defined unit cell. Lastly, from the Z contrast image shown in Figure 3(d), the spacing between the In planes to the nearest Zn planes is 0.31nm whereas the In/ZnO (0002) interplanar spacing is about 0.27nm.

Figure 1: FE-SEM image of the as-synthesized In2O3(ZnO)n nanowires.

Fig.2: (a) Low magnification TEM image of a In2O3(ZnO)n nanowire. (b) High-Resolution (HR)-TEM image showing different stacking sequences. (c) SAED pattern. (d) Enlarged HRTEM

image from (b). (e) STEM Z-contrast image. (f) Enlarged Z-contrast image.

Fig.3: (a) HR-TEM image of a transverse superlattice structure with different stacking sequences.

(b) SAED pattern. (c) Enlarged HRTEM image from (a). (d) Z contrast image.

2. Enhanced thermoelectric properties with the In2O3(ZnO)n superlattice nanowires

Various nanodevices are first devised to take measurements for thermal conductivity and Seeback coefficient from individual nanowires. As for nanodevices measuring thermal conductivity, the nanowire is designed to be suspended as shown in Fig. 4 so that heat is ensured to pass only through the nanowire without any excursion through the underlined layer. The heat is self-generated through resistive heating or Joule heating by a constant DC current through the nanowire. Therefore, the heating power of the nanowire can be described as P = I²R. The both gold electrodes are regarded as heat sinks due to inherently excellent thermal conductivity to establish equilibrium temperature with the ambient. Therefore, during Joule heating, heat flows from the highest temperature at the central part of the nanowire toward both ends of the electrodes until a steady state for establishing temperature gradients is reached where the power input through Joule heating is equal to the power output leaving from both electrodes. Meanwhile, to avoid heat dissipation by thermal convection caused by ambient gas molecules in the theoretical derivation, the measurements are carried out under vacuum. Based on this model, thermal conductivity has been derived and given by: [14]

where m = R/T. R is determined from the change of the resistance between initial state and steady state during Joule heating. The parameter of m has to be determined independently by reading the resistance value against ambient temperature when heating the nanowire as a whole on a hot plate. The rest of the parameters including length, L, and area, A, can be measured from the SEM image in Fig. 4. Ultimately, thermal conductivity is calculated by the above formula.

(a)

(b)

(c)

(d)

Fig. 5 shows an optical microscopy image of a real device measuring Seeback coefficient by using electron beam lithography process. First, the temperature coefficient of resistance, α, of the Ti/Au electrode on the nanowire is measured by four-point probe measurements. Then, a dc current is applied to one of the two micro-heaters to generate heat through Joule heating, establishing a temperature gradient along the nanowire. The temperature of each Ti/Au electrode on both terminal ends can be determined by measuring resistance of each electrode with the pre-determined  value to calculate the temperature difference across the nanowire. The voltage difference across the nanowire is read out directly. Consequently, Seeback coefficient can be determined by sweeping the applied dc current for different heating power. Table 1 compares the measured results between the In2O3(ZnO)n with pristine ZnO nanowires. Compared to ZnO, both thermal conductivity and seekbeck coefficient of the In2O3(ZnO)n nanowire are enhanced by introducing superlattice, at the expense of electrical conductivity. Therefore, ZT value of the In2O3(ZnO)n nanowire is only two times larger than ZnO. We, therefore, come out with another novel microstructure as further improvement in the next section.

Fig. 4: SEM image of a suspended In2O3(ZnO)n nanowire device allowing to measure the change of resistance after Joule heating by applying a constant current to the nanowire

Fig.5: OM image of an In2O3(ZnO)n nanowire device for measuring Seeback coefficient.

Table1: Thermoelectric properties of ZnO and In2O3(ZnO)n nanowires

3. A novel heterojunction belts by coupling homologous In2O3(ZnO)n with In-doped ZnO for further enhanced thermoelectric properties

The coupled structure of homologous In2O3(ZnO)n with In-doped ZnO into heterojunction belts is successfully fabricated by alloy-evaporation deposition. This work demonstrates a new epitaxial layered structure by combining doping in one layer and homologous structures in another.

Figure 6(a) is a low-magnification SEM image, showing belts of lower density formed accompanied by higher density nanodisks (NDs) on a silicon substrate, where the red dotted circles mark the nucleation sites of the belts. The belts exhibit a tapered morphology, shrinking from bottom to top, with typical lengths and widths falling in the range of 10-20 µm and 1-2 µm, respectively. Figures 6(b) and 6(c) show higher magnification SEM images of the NDs and belts, respectively. The NDs are partially hexagons and partially half-hexagons, with a diagonal size of around 0.8~1 m randomly oriented on the substrate. Careful examination of some vertically oriented NDs reveals that their thicknesses are in the range of 60-100 nm, with most having two parallel side edges with a cone on top. Interestingly, the belts with thicknesses in the range of 250~350 nm also show a sword-like shape, with two parallel side edges and a tip of around 60~70° on top. The similar features of these structures, except for the aspect ratio and size, indicate that the belts grow longer, extending from nanodisks as embryos.

The SE images acquired from the lower and upper detectors at an accelerating voltage of 1 kV with a working distance of around 3 mm are shown in Figures 6(d) and 6(e), respectively. The image from the lower detector is obtained by collecting both back-scattered and secondary electrons, and is an unfiltered image for greater topographic contrast. In contrast, the upper detector can only collect SEs up to a certain energy, depending on the bias applied to the side electrodes, forming a low-pass energy filtered image for voltage contrast, which is sensitive to the electronic structure, as dictated by parameters such as doping concentration.[15] Figure 6(d) reveals no clear change in contrast throughout the belt, as also indicated in the inset for the intensity line profile projected from the marked white rectangular area. The near constant

contrast implies that the belt is characterized by a flat surface morphology. In contrast, Figure 6(e) exhibits a distinct sandwich-like contrast for the belt, indicative of being composed of a heterojunction layered structure. As revealed in the projected intensity line profile in the inset, two dark bands are sandwiched by the bright central area and two bright bands at either side, forming a five-layer structure, where the two brightest lines at the edges due to edge effect are excluded. The symmetric five-layer structure can also be seen in a higher magnification SEM image in Figure 6(f), which clearly shows a heterojunction structure rendered by non-uniform In doping. In order to confirm this conjecture, Figures 6(g)-(i) show EDS elemental maps of indium, zinc, and oxygen, respectively, with reference to the SEM image in Figure 6(f). Intriguingly, these three elements demonstrate drastically different spatial distributions. Whereas oxygen is uniformly distributed over the entire structure, In is richer at the bottom and two sides covering from the area corresponding to the side bright bands, and perhaps extending to the two dark bands in Figure 6(f). Meanwhile, zinc seems to decay slightly and symmetrically toward the two sides, and probably also in the two bright bands. These results reveal the complex dependence of indium doping on the five heterojunction layers through the influence of the electronic band structure, responsible for the strong SE contrast.

Figure 7(a) shows a low-magnification TEM bright-field image of a single belt with the width and length of about 800 nm and 1 m, respectively. By comparing this with the SEM results, we suggest that the “region Ⅰ” feature with a larger width may correspond to the bottom part of the heterojunction belt, while the sharp tip feature of “region Ⅲ” may correspond to the top of the belt. The corresponding SAED pattern on the [0001] projection in Figure 7(a) shows that the entire belt is a single-crystalline wurtzite structure growing along the [10-10] direction without any extra diffraction spots, indicative of epitaxial interfaces across the heterojunctions. However, the average indium concentrations with respect to ZnO acquired from EDS reveal large variations in spatial distribution, at about 10.3, 2.3, and 2.5 at. % for the bottom (Ⅰ), middle (Ⅱ) and top (Ⅲ) regions of the belt, respectively. Notably, the nucleation region corresponds to the highest indium concentration.

To gain more insight into the microstructure and composition of the HB, the region marked with a white rectangle in Figure 7(a) is cut by FIB for TEM cross-sectional characterization.

Figure 7(b) shows a bright field image in the upper part and STEM HAADF image in the lower part, which are analogous to the two SE images in Figure 6 with regard to contrast variation.

Whereas the TEM bright field image indicates the smooth top and bottom surfaces as revealed by the SE image in Figure 6(d), the HAADF image demonstrates similar layer contrast to the SE image from the upper detector in Figure 6(e), in that the two side regions are brighter than the central part. The brighter regions are ascribed to a higher indium concentration, consistent with the EDS In map in Figure 6(g). The FIB-cut sample is further studied by TEM, EDS and SAED in Figure 7 (c) for image contrast, indium concentration, and diffraction pattern, respectively, by which three distinct regions (i ii, and iii) are assigned. Region i (central part) is characterized by the single-crystalline wurtzite structure of ZnO with negligible indium concentration, as shown in the SAED pattern in the inset, and evidenced by the darkest contrast in the HAADF image and

the EDS characterization. In region ii, the average indium concentration is around 1.7 at%, which is associated with the brighter contrast in the HAADF image. The DP in the inset also corresponds to the single crystalline wurtzite structure, with the diffraction spots shifted to a slightly lower angle compared with that of region i, suggesting that this region is single crystalline ZnO homogenously doped with In. Upon careful examination of the HAADF image, region iii is found to exhibit an amazingly complex structure composed of many indium-enriched layers, as evidenced by bright lines oriented parallel to the <11-20> direction, with a few lines extending slightly into region ii, causing weak superlattice spots and prominent streaks in the SEAD pattern. The average In concentration in this region is even higher than region ii of around 2.4 at%, as measured by EDS, which in combination with the superlattice results suggest that this region has a modulated homologous In2O3(ZnO)n structure. Notably, a slight increase in indium incorporation switches the homogeneous In doping in region ii to the superlattice structure in region iii.

Similarly, the In2O3(ZnO)n superlattice structure generally consists of atomic layers of InO2- octahedra separated by slabs of wurtzite InZnnO(n+1)+ of varying thicknesses [1, 16], exactly as observed in the high-magnification HADDF image of region iii in Figure 7(d). The formation of a superlattice structure will induce a large strain between the ZnO slabs and InO2 insertion layers.

The ends of the partial inclusions through incomplete indium lattice diffusion are thus usually associated with edge dislocations, as shown in the white dashed circle in Figure 7(d). The accumulative strain is relaxed via two routes. First, indium atoms distribute along the hexagonal c axis in zigzag modulated structures with the inclined angle of 53°, thus reducing total strain, as shown in Figure 7(e), preferably for thicker ZnO slabs, consistent with theoretical predictions [17]. In addition, the octahedral indium inclusion layers act as inversion domain boundaries (IDB) to relax the strain, as shown in Figure 7(f) in a high-resolution HAADF image, where the stacking sequence of the close-packed metal layers switches from ABAB to CACA passing the IDB. Figure 7(g) shows a projected line profile across an IDB, where the d- spacing in the InZnnO(n+1)+ layer remains almost the same as in pure ZnO (0002) planes (0.26 nm), but with a 19

% increase in the d-spacing (0.31 nm) on the either side of the In-O insertion layer.

From the above analyses, we demonstrate that a novel 5-layer heterojunction ZnO belt comprised of In2O3(ZnO)n / In:ZnO / ZnO / In:ZnO / In2O3(ZnO)n is synthesized by a facile growth method. The possible growth mechanisms could be rationalized by four steps, as shown in Figure 7(h). (1) Self-catalyzed vapor-liquid-solid (VLS) combined with vapor-solid (VS) mechanism: In the beginning, alloy droplets are formed on the substrate from the co-condensation of zinc and indium vapors, followed by the nucleation of ZnO belts with supplied oxygen via a self-catalyzed VLS process. At the same time, indium is incorporated as a dopant. According to the calculations [18], the surface energy of ZnO (0001) will become smaller than that of the other major planes once indium doping exceeds 1/4. The fact that nucleation region I at the bottom is rich in indium makes the 1D ZnO grow in the form of a belt along the thermodynamically favored [01-10] direction rather than the more common [0002] one.

Moreover, it is worth noting that the width is non-uniform throughout the belt, indicating that the

VS process is also operating[19], and that the decrease in width may be as a result of the varied growth rate in the [11-20] direction due to the gradual decrease in indium concentration. (2) Surface diffusion of indium: The indium concentration decreases from the bottom to top, implying that indium diffuses upward. Moreover, the large amount of indium available at the bottom establishes a large concentration gradient as the driving force for diffusion. Under this condition, surface diffusion is much more favored than slow lattice diffusion. (3) and (4) Redistribution of indium to form a heterojunction structure: Subsequently, indium diffuses inward to result in a three-zone sandwich-like belt structure, where each layer structure is determined by the average indium concentration and strain involved. Since the average indium concentration may exceed the solid solubility limit in ZnO in the outermost region (region iii), an In2O3(ZnO)m homologous structure is generated to compromise between indium concentration and strain. Region ii certainly results from residual diffusion of indium into ZnO lattices, a process that is limited by indium supply and temperature. The equilibrium structure is finally established through competition between belt growth and indium diffusion inward from the surfaces, causing region i top shrink more than regions ii and iii with regard to the width towards the top of the belt.

Figure 6. (a) Low-magnification SEM top-view image of the as-synthesized products on Si substructure. (b) and (c) high-magnification SEM images of nanodisks and microbelts, respectively. (d) and (e) SE images of a In2O3(ZnO)n/In:ZnO/ZnO belt recorded using ET and TTL E x B detectors, respectively. The insets show the corresponding projected intensity line profiles. (f)-(i) FESEM images and the corresponding EDS element mapping results of In, Zn, and O, respectively.

Figure 7. (a) TEM bright-filed image of a In2O3(ZnO)n/In:ZnO/ZnO belt with indium concentrations in distinct regions (I, II, and III) obtained from EDS. The inset is the corresponding SAED pattern recorded from the entire belt structure. (b) TEM cross-sectional bright-field (up) and HAADF (down) images of an belt. (c) Medium-magnification HAADF image recorded from the rectangular area in (b) with SAED patterns and In EDS data obtained from regions i, ii, and iii. High-magnification HAADF images obtained from region iii in (c) showing (d) alternate stacking of In-O layers and In/Zn-O slabs along the c-axis, and (e) zigzag modulated structures in the In/Zn-O slabs. (f) Atomic-resolution HAADF image showing columns of Zn and In cations. (g) Intensity line profile and d-spacing across an IDB. (h) Schematic diagram of the proposed growth mechanisms of the belts.

4. Even higher enhanced optical and thermoelectric properties of the novel heterojunction structure

This novel heterostructure provides an ideal pathway to enhance electron conduction through the indium doped ZnO layer, and the homologous In2O3(ZnO)n layer contains numerous interfaces to impede phonon transportation. This structure can offer an efficient way of separating the transportation of electrons and phonons for high thermoelectric power factors in

在文檔中 A novel hybrid energy harvesting cell by combining thermoelectric and piezoelectric devices with one dimensional ZnO nanostructures (頁 8-26)