Enhanced Thermoelectric Performance of ZnO-Based Thin Films via Interface Engineering

: Zinc oxide (ZnO) is a potential thermoelectric material with good chemical and thermal stability as well as an excellent Seebeck coefﬁcient. However, the extremely low carrier concentration brings poor electrical transport properties. Although Gallium (Ga) doping could increase the carrier concentration of ZnO ﬁlm, its thermoelectric performance is still limited due to the deteriorated Seebeck coefﬁcient and enhanced thermal conductivity. Interface engineering is an effective strategy to decouple electron-phonon interaction for thermoelectric materials. Thus, in this work, GZO (Ga-doped ZnO)/NAZO (Ni, Al co-doped ZnO) multilayer ﬁlms were designed to further improve the thermoelectric properties of GZO ﬁlms. It was found that GZO/NAZO multilayer ﬁlms possessed better electrical conductivity, which was attributed to the increased carrier concentration and Hall mobility. Meanwhile, beneﬁting from the energy ﬁltering that occurred at GZO/NAZO interfaces, the density of states effective mass increased, resulting in comparable Seebeck coefﬁcient values. Ultimately, an enhanced power factor value of 313 µ W m − 1 K − 2 was achieved in the GZO/NAZO multilayer ﬁlm, which is almost 46% larger than that of GZO ﬁlm. This work provides a paradigm to optimize the thermoelectric performance of oxide ﬁlms and other thermoelectric systems by multilayer structure design with coherent interfaces.


Introduction
Thermoelectric materials can realize direct conversion between heat and electricity via the Seebeck and Peltier effects, which is suitable for power generation and precise cooling [1][2][3][4]. Advances in 5G communications and the integrated circuits industry promoted portable and wearable electronics, and there is an urgent need for self-powered energy and chip cooling technology [5,6]. In order to fulfill the requirement, thermoelectric thin films with lightweight, small size and controllable structures are developed [7]. An ideal thermoelectric material should have excellent electrical properties and low thermal conductivity, as the performance is related to a dimensionless figure of merit (ZT) and can be described as ZT = S 2 σT/κ. Here, S, σ, κ and T stand for the Seebeck coefficient, electrical conductivity, thermal conductivity and absolute temperature, respectively [8]. Compared to bulks, thin films usually have lower κ due to the lower dimension [9,10]. Thus, it is important to improve electrical transport properties for thin films, which is related to the power factor (PF = S 2 σ).
Zinc oxide (ZnO) thin films are a kind of potential thermoelectric materials due to their merits such as non-toxicity, low cost, outstanding chemical and thermal stability as well as high visible-light transparency [11]. However, intrinsic ZnO possesses relatively low σ, which originates from the extremely low carrier concentration (n) [12]. In order to solve this problem, elements such as Al, Ga, In, Sb and Sn were doped in the Zn site. Among the dopants, Al and Ga are much more effective in improving the electrical conductivity for ZnO [13][14][15][16][17][18]. Although Al or Ga doping could largely increase n, the decreased S and enhanced carrier thermal conductivity still limited the further improvement of thermoelectric performance according to the strongly coupled correlations among S, σ, κ and n [19]. Thus, how to simultaneously optimize S, σ and κ should be considered for obtaining better thermoelectric properties of ZnO thin films.
Interface engineering has been regarded as an effective strategy to decouple electronphonon interaction [20]. On the one hand, energy filtering effects may occur when carriers go across the interfaces with certain potential barriers, where low energy carriers cannot transmit through the energy barrier at the formed interface, which is beneficial to improving Seebeck coefficient values [21]. On the other hand, phonons could be scattered by interfaces, further decreasing lattice thermal conductivity [22]. In our previous work, a GZO-ZnO-GZO sandwich structured thin film was designed, and the simultaneous optimization of σ and S was successfully realized [23]. Thus, it is supposed that multilayer structure may have much more significant effects in optimizing thermoelectric performance for ZnO-based thin films.
Compared with pure ZnO, Ni and Al co-doped ZnO (NAZO) has higher electrical conductivity and closer lattice constants to GZO, which is beneficial for electron transport [24,25]. By considering this, GZO/NAZO multilayer thin films were designed and fabricated by using the pulsed laser deposition (PLD) technique in this work. Moreover, to explore the effects of the GZO/NAZO thickness ratio on their thermoelectric performance, two different GZO/NAZO thin films were designed. The introduction of NAZO increased the σ for GZO thin films, which is attributed to the enhanced n and Hall mobility (µ H ). Benefiting from energy filtering effects functioned at the interfaces between GZO and NAZO, the density of states (DOS) effective mass (m d * ) increased and compensated for the detrimental effects of increased n on the Seebeck coefficient. Consequently, an improved PF value of 313 µW m −1 K −2 was achieved in GZO:NAZO = 8:2 thin film, which is almost 46% larger than that of the GZO thin film. This work provides a facile and simple strategy to optimize thermoelectric performance for thin films via interface engineering.

Sample Preparation
The as-deposited GZO/NAZO thin films with different GZO/NAZO ratios were grown on sapphire (Al 2 O 3 , 0001, Hefei Kejing Materials Technology Co., Ltd., Hefei, China) single-crystal substrates with the size of 12 × 4 mm 2 by using pulsed laser deposition (PLD) technique. The KrF excimer pulsed laser (λ = 248 nm) at a repetition rate of 5 Hz and 3 Hz was used to grow GZO layers and NAZO layers, respectively. As a comparison, a GZO thin film with similar thickness was also prepared by the same method. The deposition temperature and deposition oxygen pressure were set as 673 K and 6.7 Pa. After cooling at the rate of 20 K min −1 , the thin films were then put in another furnace and annealed at 673 K for 1 h under 5% H 2 /95% Ar mixed gas atmosphere. The used targets are GZO

Characterization
The phase purity and crystal structure of the thin films were detected by X-ray diffraction (XRD, Rigaku smartlab, Tokyo, Japan) with Cu Kα radiation by using the 2θ-ω scanning method. In order to ensure the thicknesses of the samples, high-resolution X-ray reflectometry (XRR, Bruker D8 discover, Karlsruhe, Germany) was adopted, and the results were then analyzed by LEPTOS fitting software. The cross-section information, such as interfacial structure, was characterized by transmission electron microscopy (TEM, JEM-2100F, JEOL, Tokyo, Japan) as well as high-resolution TEM (HRTEM, JEM-2100F, JEOL, Tokyo, Japan). The TEM specimen was prepared by a focused ion beam system using Xe + ions equipped with a high-resolution and high-sensitivity surface analysis system (TESCAN S9000X, Brno, Czech). The electrical conductivity (σ) and the Seebeck coefficients (S) were measured simultaneously by using a commercial apparatus (ZEM-3, ULVAC-RIKO, Yokohama, Japan) from 300 K to 673 K under a helium atmosphere with a homemade Al2O3 holder for thin film measurements. The related details for measuring the electrical properties of thin films by ZEM-3 were provided in our previous work [11]. In order to obtain a better understating of the underlying electrical transport mechanisms, room temperature Hall coefficient (RH) values were determined on a homemade system equipped with a maximum 5 T superconducting magnet (Cryogenic Limited, London, UK). The carrier concentration (n) and Hall mobility (μH) were then calculated by equations, n = r/e|RH| (assuming the Hall factor, r = 1.0) and μH = σ|RH|. The measured Seebeck coefficient and electrical conductivity were from the in-plane direction with an uncertainty of approximately 3% and 5%, respectively. Thus, the uncertainty of the PF value is estimated to be about 10%.

Results and Discussion
The phase structure of GZO and GZO/NAZO thin films is reflected in X-ray diffraction (XRD) patterns, as shown in Figure 1c. In addition to the peaks of Al2O3, all the thin films showed only (002) peaks around 34.5°, indicating they are highly c-axis oriented [11,12,23]. In addition, it was observed that there is no evident shift in (002) peaks for the GZO and GZO/NAZO thin films, as presented in the enlarged image of the (002) peaks ( Figure 1d). The reason can be attributed to the similar ionic radius of Ga 3+ (r = 0.062 nm), and co-dopant Al 3+ (r = 0.054 nm)-Ni 2+ (r = 0.069 nm). Furthermore, the close lattice

Characterization
The phase purity and crystal structure of the thin films were detected by X-ray diffraction (XRD, Rigaku smartlab, Tokyo, Japan) with Cu Kα radiation by using the 2θ-ω scanning method. In order to ensure the thicknesses of the samples, high-resolution X-ray reflectometry (XRR, Bruker D8 discover, Karlsruhe, Germany) was adopted, and the results were then analyzed by LEPTOS fitting software. The cross-section information, such as interfacial structure, was characterized by transmission electron microscopy (TEM, JEM-2100F, JEOL, Tokyo, Japan) as well as high-resolution TEM (HRTEM, JEM-2100F, JEOL, Tokyo, Japan). The TEM specimen was prepared by a focused ion beam system using Xe + ions equipped with a high-resolution and high-sensitivity surface analysis system (TESCAN S9000X, Brno, Czech). The electrical conductivity (σ) and the Seebeck coefficients (S) were measured simultaneously by using a commercial apparatus (ZEM-3, ULVAC-RIKO, Yokohama, Japan) from 300 K to 673 K under a helium atmosphere with a homemade Al 2 O 3 holder for thin film measurements. The related details for measuring the electrical properties of thin films by ZEM-3 were provided in our previous work [11]. In order to obtain a better understating of the underlying electrical transport mechanisms, room temperature Hall coefficient (R H ) values were determined on a homemade system equipped with a maximum 5 T superconducting magnet (Cryogenic Limited, London, UK). The carrier concentration (n) and Hall mobility (µ H ) were then calculated by equations, n = r/e|R H | (assuming the Hall factor, r = 1.0) and µ H = σ|R H |. The measured Seebeck coefficient and electrical conductivity were from the in-plane direction with an uncertainty of approximately 3% and 5%, respectively. Thus, the uncertainty of the PF value is estimated to be about 10%.

Results and Discussion
The phase structure of GZO and GZO/NAZO thin films is reflected in X-ray diffraction (XRD) patterns, as shown in Figure 1c. In addition to the peaks of Al 2 O 3 , all the thin films showed only (002) peaks around 34.5 • , indicating they are highly c-axis oriented [11,12,23]. In addition, it was observed that there is no evident shift in (002) peaks for the GZO and GZO/NAZO thin films, as presented in the enlarged image of the (002) peaks (Figure 1d). The reason can be attributed to the similar ionic radius of Ga 3+ (r = 0.062 nm), and co-dopant Al 3+ (r = 0.054 nm)-Ni 2+ (r = 0.069 nm). Furthermore, the close lattice parameters and the same crystal structure between GZO and NAZO are beneficial to electron transport.
In order to further explore the microstructure of GZO/NAZO thin film, the crosssection TEM was performed on sample GZO:NAZO = 8:2. As Figure 2a shows, the thickness of GZO:NAZO = 8:2 was around 36 nm, which was consistent with the results from XRR. In order to shed more light on the interfaces, the two areas marked b and c in Figure 2a were investigated. Figure 2b,c obtained from high-resolution TEM (HRTEM), demonstrate the regions of Al 2 O 3 -GZO and GZO-NAZO. Both regions indicated that the GZO/NAZO thin film was highly c-axis oriented, which was also revealed by XRD. Due to the large lattice mismatch (~18%) between Al 2 O 3 and GZO [26], lattice distortions around the interface region existed, which may impede the carrier transport but strengthen the phonon scattering. As for the interface between GZO and NAZO, because of the close lattice parameters and same crystal structure, the interface is coherent with little lattice distortion. Additionally, such a coherent interface can strongly scatter phonons without hindering electron transport, which can simultaneously optimize electrical and thermal performance [27]. parameters and the same crystal structure between GZO and NAZO are beneficial to electron transport.
In order to further explore the microstructure of GZO/NAZO thin film, the crosssection TEM was performed on sample GZO:NAZO = 8:2. As Figure 2a shows, the thickness of GZO:NAZO = 8:2 was around 36 nm, which was consistent with the results from XRR. In order to shed more light on the interfaces, the two areas marked b and c in Figure  2a were investigated. Figure 2b,c obtained from high-resolution TEM (HRTEM), demonstrate the regions of Al2O3-GZO and GZO-NAZO. Both regions indicated that the GZO/NAZO thin film was highly c-axis oriented, which was also revealed by XRD. Due to the large lattice mismatch (~18%) between Al2O3 and GZO [26], lattice distortions around the interface region existed, which may impede the carrier transport but strengthen the phonon scattering. As for the interface between GZO and NAZO, because of the close lattice parameters and same crystal structure, the interface is coherent with little lattice distortion. Additionally, such a coherent interface can strongly scatter phonons without hindering electron transport, which can simultaneously optimize electrical and thermal performance [27].  Figure 3a shows the temperature-dependent electrical conductivity of all the thin films. It can be seen that the electrical conductivity of GZO increased with increasing temperature, indicating that GZO possessed semiconductor-like behavior. Differently, GZO/NAZO thin films behaved like metals. Additionally, GZO:NAZO = 8:2 had better electrical conductivity than those of GZO and GZO:NAZO = 7:3 at the whole temperature range from 300 K to 673 K. As the electrical conductivity is related to carrier concentration (n) and Hall mobility (μH), the underlying reasons for these results can be revealed by Hall measurement. As Figure 3c and Table 1 show, compared to GZO thin films, GZO/NAZO thin films had larger n and μH, where the n increased from 2.15 × 10 20 cm −3 to 3.43 × 10 20 cm −3 and μH improved from 4.70 cm 2 V −1 s −1 to 15.22 cm 2 V −1 s −1 , which is related to the Ni-Al co-doping and coherent interfaces between GZO and NAZO. As for GZO:NAZO = 8:2 and GZO:NAZO = 7:3, GZO:NAZO = 8:2 with a thinner NAZO layer showed relatively lower n but larger μH, which means that a suitable thickness ratio is important to obtain better performance.  Figure 3a shows the temperature-dependent electrical conductivity of all the thin films. It can be seen that the electrical conductivity of GZO increased with increasing temperature, indicating that GZO possessed semiconductor-like behavior. Differently, GZO/NAZO thin films behaved like metals. Additionally, GZO:NAZO = 8:2 had better electrical conductivity than those of GZO and GZO:NAZO = 7:3 at the whole temperature range from 300 K to 673 K. As the electrical conductivity is related to carrier concentration (n) and Hall mobility (µ H ), the underlying reasons for these results can be revealed by Hall measurement. As Figure 3c and Table 1 show, compared to GZO thin films, GZO/NAZO thin films had larger n and µ H , where the n increased from 2.15 × 10 20 cm −3 to 3.43 × 10 20 cm −3 and µ H improved from 4.70 cm 2 V −1 s −1 to 15.22 cm 2 V −1 s −1 , which is related to the Ni-Al co-doping and coherent interfaces between GZO and NAZO. As for GZO:NAZO = 8:2 and GZO:NAZO = 7:3, GZO:NAZO = 8:2 with a thinner NAZO layer showed relatively lower n but larger µ H , which means that a suitable thickness ratio is important to obtain better performance.  The temperature-dependent Seebeck coefficient of all the thin films is presented in Figure 3b. Although GZO/NAZO thin films have higher carrier concentration, the GZO/NAZO thin films still have comparable Seebeck coefficient values, and the Seebeck coefficient values are even higher than that of GZO thin films when the temperature is above 600 K. For degenerate semiconductors, Seebeck coefficient can be expressed as [28] S= where e is the electron charge, kB is the Boltzmann constant, h is the Plank constant and md * is the density of states (DOS) effective mass. Thus, a lower n and a larger md * are good for obtaining a higher Seebeck coefficient.  The temperature-dependent Seebeck coefficient of all the thin films is presented in Figure 3b. Although GZO/NAZO thin films have higher carrier concentration, the GZO/NAZO thin films still have comparable Seebeck coefficient values, and the Seebeck coefficient values are even higher than that of GZO thin films when the temperature is above 600 K. For degenerate semiconductors, Seebeck coefficient can be expressed as [28] where e is the electron charge, k B is the Boltzmann constant, h is the Plank constant and m d * is the density of states (DOS) effective mass. Thus, a lower n and a larger m d * are good for obtaining a higher Seebeck coefficient. In order to have a better understanding of the underlying mechanisms, we calculated the m d * of all the thin films. The calculations of m d * are based on the assumption of the single parabolic band (SPB) mode and acoustic phonon scattering with the scattering factor, r = −1/2. The correlations are expressed as follows [29,30]: where F i (η F ) is the Fermi-Dirac integral and η F is the reduced Fermi level. It can be seen in Table 1 that the m d * increased from 0.78 m 0 for GZO to 0.82 m 0 for GZO/NAZO. In addition, GZO:NAZO = 8:2 and GZO:NAZO = 7:3 had the same m d * value, indicating the thickness ratio is not the key parameter to influence m d * . In order to directly demonstrate the correlations among S, n and m d * , the calculated room-temperature Pisarenko plots were illustrated in Figure 3d. The improvement of m d * can be ascribed to energy filtering effects, which always occur when carriers go across the interfaces with certain potential barriers. It is also noted that the m d * just slightly increased for the GZO/NAZO multilayer thin films, which can be explained by the increased mobility. The m d * is generally related to degeneracy valleys (N v ) and single valley DOS effective mass (m b * ), and their expression is The m b * is inversely proportional to mobility, and the details can be found elsewhere [31]. Thus, increased mobility may impede the m d * to some extent. Consequently, attributed to the improved electrical conductivity and relatively high Seebeck coefficient, the GZO:NAZO = 8:2 thin film showed the best thermoelectric performance with the PF value of 313 µW m −1 K −2 at 673 K, which resulted in an almost 46% improvement when compared to that of GZO film (Figure 4a). The reasons for the significant improvement are as follows. The increased carrier concentration and Hall mobility originated from the Ni-Al co-doping and GZO-NAZO coherent interfaces, contributing to the increased electrical conductivity. The energy filtering effects brought the increase in DOS effective mass, resulting in comparable Seebeck coefficient values. In addition, as Figure 4b shows, when compared with other representative n-type oxygencontaining thermoelectric materials including Bi 2 O 2 Se [32], SrTiO 3 [33], CaMnO 3 [34] and Bi 5.9 Zr 0.1 Cu 2 Se 3.6 Cl 0.4 O 6 [35], the power factor value of GZO/NAZO at similar working temperature is still competitive. This work presents that optimization of thermoelectric performance could be realized by structure design. By considering the lower cost of ZnO-based materials, GZO/NAZO thin films show potential for future applications.

Conclusions
In summary, we successfully prepared GZO/NAZO multilayer thin films and investigated the effects of interface engineering on thermoelectric performance for such thin films. Due to the close lattice parameters and the same crystal structure between GZO and NAZO, the interfaces are coherent and smooth, resulting in large Hall mobilities. In addition, Ni-Al co-dopants improved the carrier concentration. The optimized carrier concentration and Hall mobility contributed to a largely enhanced electrical conductivity from 175 S cm −1 for GZO thin film to 767 S cm −1 for GZO:NAZO = 8:2 thin film at room temperature. Moreover, the DOS effective mass was increased by energy filtering effects functioned at the GZO/NAZO interfaces, making a comparable Seebeck coefficient. Ultimately, an enhanced power factor value reached 313 μW m −1 K −2 at 673 K for GZO/NAZO multilayer thin film, which resulted in an almost 46% improvement when compared to that of GZO film. The significant improvement proves the beneficial effects of interface engineering on optimizing thermoelectric performance. Although the thermal conductivity was not measured, the thermal conductivity of multilayered GZO/NAZO should be low due to the stronger phonon scattering from the interfaces. Our method of multilayer structure design with coherent interfaces for thin films is effective and efficient in improving their thermoelectric properties, which can also be widely used in other materials and promote the application of thin film thermoelectric devices.

Conclusions
In summary, we successfully prepared GZO/NAZO multilayer thin films and investigated the effects of interface engineering on thermoelectric performance for such thin films. Due to the close lattice parameters and the same crystal structure between GZO and NAZO, the interfaces are coherent and smooth, resulting in large Hall mobilities. In addition, Ni-Al co-dopants improved the carrier concentration. The optimized carrier concentration and Hall mobility contributed to a largely enhanced electrical conductivity from 175 S cm −1 for GZO thin film to 767 S cm −1 for GZO:NAZO = 8:2 thin film at room temperature. Moreover, the DOS effective mass was increased by energy filtering effects functioned at the GZO/NAZO interfaces, making a comparable Seebeck coefficient. Ultimately, an enhanced power factor value reached 313 µW m −1 K −2 at 673 K for GZO/NAZO multilayer thin film, which resulted in an almost 46% improvement when compared to that of GZO film. The significant improvement proves the beneficial effects of interface engineering on optimizing thermoelectric performance. Although the thermal conductivity was not measured, the thermal conductivity of multilayered GZO/NAZO should be low due to the stronger phonon scattering from the interfaces. Our method of multilayer structure design with coherent interfaces for thin films is effective and efficient in improving their thermoelectric properties, which can also be widely used in other materials and promote the application of thin film thermoelectric devices.