Influence of thermal treatment on electronic properties of inkjet-printed zinc oxide semiconductor

ABSTRACT Additive manufacturing of electronic devices using inkjet printing provides a potential alternative approach in substitution for conventional electronic fabrication processes. However, the complex nature of inkjet printing involves the liquid deposition and film formation from the vaporization of solvent, which makes it different from film created by conventional deposition methods. Inkjet printing of zinc oxide (ZnO), which is a widely utilized semiconductor, produces polycrystalline film composed of nano-size grains, which could significantly influence the properties of printed film. In this study, low-temperature annealing was employed to treat inkjet-printed ZnO for UV photodetection application, and its influence on electrical properties was studied. Band bending was characterized using the Mott-Schottky plot which examines the charge distribution of the films. It is found that the annealing of inkjet-printed polycrystalline ZnO film has improved its electrical properties, which could be attributed to the reduction of band bending due to the merging of grains. The treatment also helps to reduce impurities of the film, such as zinc hydroxide complexes, which are common for solution-derived films. Hence, the study could pay the way for the improvement of electrical properties of inkjet-printed functional materials. Graphical abstract


Introduction
Printing of functional material could promote the application of electronic devices via additive manufacturing approaches, whereby the material is gradually deposited to form the desired physical structure from a digital design [1,2]. Among different additive manufacturing technologies, inkjet printing is one of the most suitable techniques for electronic printing thanks to its merits such as non-contact deposition, high precision, being capable of printing multiple materials, and being compatible with flexible substrates [3]. Printing of conductive materials, which are the fundamental of electronic printing, have been extensively explored, such as the printing of silver nanoparticles to create conductive patterns [4]. These printings were conducted on flexible substrates such as polymer substrates, which demonstrated the feasibility of flexible electronic manufacturing by additive manufacturing [5,6]. Considerably, printing of transparent conductive film is possible, which widens the capability of additive manufacturing [7]. Beside conventional transparent electrode material such as indium-tin-oxide-based, printing in large-area of transparent conductive film could be implemented by alternative inks such as silver nanoparticles [8], conductive polymer [9,10]. Furthermore, the printing of different materials with specific functionality could construct devices with performance comparable to conventional fabrication approaches [11][12][13].
ZnO is an attractive wide bandgap semiconductor that has gathered a vast interest in various optoelectronic applications, such as transistors [14,15], photodetectors [16,17], and solar cells [18,19]. Recent developments of ZnO-based device witness the employment of this multi-functionals material on a rapid immunoassay test for SARS-CoV-2 virus [20].
Hence, ZnO nanostructures, such as nanowires and nanoparticles [21,22], have been an attractive source for various effort in inkjet printing of semiconductor materials for construction of functional devices.
Noticeably, the characteristics of the printed material determine the performance of the fabricated device. Because the polycrystalline structure is the main feature of the inkjet-printed ZnO, the grain boundaries have a major effect on electron transportation via the band bending at the grain interface [23,24]. It is well known that the electrical characteristics engineering of polycrystalline semiconductors could be achieved by modification of grain boundary as proven by previous studies [25,26]. For example, nano junction was adjusted to control band bending and enhance the efficiency of photosensing [27]. The band bending was also be tuned by rapid thermal treatment to obtain enhanced optoelectrical performance [28]. Notably, low temperature annealing of ZnO has been investigated using laser treatment to promote the coalescence of the ZnO nanoparticle [29,30]. Therefore, engineering the electronic properties of inkjet-printed ZnO is possible via modification of band bending.
In this work, the engineering of inkjet-printed ZnO electrical properties using relatively low annealing temperatures was studied. The photosensitivity of printed UV photodetector exhibits a strong dependence on annealing temperature, such as the higher the temperature, the higher photocurrent. The increment of photocurrent was attributed to the reduction of the flat band potential of printed ZnO thin film as a result of thermal annealing. Surface states were studied using X-ray photoelectron spectroscopy (XPS), which is linked to the response characteristics of photocurrent under UV illumination. This work presents a simple method for the treatment of inkjet-printed ZnO thin film for optoelectronic application by relatively low-temperature annealing. The treatment helps reducing impurities of inkjet-printed ZnO film, such as zinc hydroxide complexes [31]. The significance of band bending to the characteristics of inkjet-printed UV sensor is also demonstrated, and a higher responsivity could be achieved by reducing band bending of the inkjet-printed semiconductor thin film.

Inkjet printing and formation of ZnO from zinc precursor
Zinc acetate dihydrate solution in ethanol with a concentration of 50 mM was employed as the ink for printing of zinc precursor in the process published in our previous report [16]. A commercial inkjet printer (Dimatix 2831, Fujifilm) was employed for the printing. For fabrication of UV photodetector, both the semiconductor, i.e. ZnO, and conductor, i.e. silver, were printed on Si/SiO 2 substrate. Firstly, two silver patterns were printed to create the electrodes for the device. These silver electrodes are annealed at 150°C to sinter the nanoparticles and improve their conductivity. The silver electrodes are about 100 µm in width, and the space between the two electrodes is 110 µm by 600 µm, which are later covered by the ZnO thin film to form the UV photodetector structure. The zinc salt was printed on top of the silver electrodes to subsequently create metal-semiconductor junctions. The printed zinc salt was then undergone a heating step on a hotplate for 10 minutes at 200°C to form ZnO polycrystalline thin film via the thermal decomposing of the zinc precursor. The printed films were then annealed in a quartz tube furnace (Thermo scientific) for haft an hour at various temperatures from 250°C to 350°C to further improve the film quality.

Characterization
Field emission scanning electron microscope (FE-SEM, JEOL 7600) and transmission electron microscope (TEM, JEOL-JEM 2010) were employed to study the printed ZnO thin film surface structure. The crystallite size was quantified using an X-ray diffractometer (XRD, PANalytical Empyrean), which is equipped with a 0.154 nm CuKα radiation. Surface states of the printed film were studied using X-ray photoelectron spectroscopy (XPS, Kratos Analytical), which has source is a K-alpha X-rays of aluminum with photon energy of 1486.6 eV.
The electrical characteristics of the printed device were characterized using a source measure unit (SMU B2902A, Agilent) to provide a bias voltage of 5 V and measure the current running through the sensor. Optoelectrical performance was studied by illuminating the device with 365 nm UV light filtered from a xenon arc lamp source, the illumination power was 1.08 mW/cm 2 . The wide wavelength provided by the xenon arc lamp was filted by a bandpass filter to obtain 365 nm wavelength only. The transition between light and dark conditions was triggered by a homemade mechanical shutter.
Mott-Schottky plots, which characterize the charge distribution of films, were employed to study the band bending property of the film via flat band potential values. The working electrodes were prepared by printing square patterns of ZnO film with an exposing area of 100 mm 2 on ITO-coated glass. A platinum plate was used as the counter electrode, while silver/silver chloride was used as the reference electrode. All electrodes were immersed in Na 2 SO 4 electrolyte solution (0.5 M) for the measurement. A potentiostat (CHI 760D, CH Instrument) was used to measure the impedance of the film versus applied potential with a frequency fixed at 463 Hz.

Morphology and grain structure of the printed film
Printed ZnO thin films exhibit wrinkle surfaces as shown in (Figure 1(a-b)). The crumple structure might be the result of internal stress during the formation of film by thermal decomposition, which is widely observed in solution-derived thin film [32,33]. The wrinkle structures may increase the surface area of the device, hence increasing the interaction with surrounding environment. Because of the involvement of air interaction to the UV photodetector response to UV radiation, the wrinkle surface mean that there is more area to interact with air and it will affect to the device performance. Furthermore, the wrinkle structures may improve more light absorbance, and lead to a higher efficiency of the device. For example, photovoltaic efficiency of solar cells was proven to be enhanced thanks to the winkle surface [34].
TEM image (Figure 1(c-d)) reveals that printed ZnO films have a polycrystalline structure composed of small grain with an estimated size smaller than 10 nm. However, there is no significant difference in the surface structure that could be observed between films with and without thermal annealing from these electron microscopic images.
The cross-sectional image of the ZnO thin films are shown in (Figure 2), which clearly demonstrates the structure of the crumple films, composed of wrinkles due to thermal stress during the film processing. It is worth noting that general thickness of the film, which is about 2 µm (Figure 2(a-b)). However, it is in fact the height of the wrinkle film. The enlarge microscopic images show that the actual film thickness is about 200 nm only ( Figure 2(c-d)).
To study the film structure, the crystallite size was quantified by XRD analysis, as shown in  [35]. These broad peaks can be attributed to the polycrystalline form of the printed ZnO thin films, which have a small crystallite size of a few nanometers. While the peak (002) has a relatively stable full width at half maximum (FWHM) value of about 1.4° which corresponds to a crystallite size that slightly varies around 6.1 nm, those of other peaks show a noticeable decrease, such as the peak (101) has FWHM value changing from 2.3° to 1.0°. This peak width reduction could be attributed to the increase in size along the peak, from 3.8 nm to 8.3 nm according to Scherrer's equation [36]. These results demonstrate that the annealing has promoted the merging of grains, which results in slightly larger grains of the annealed film. Furthermore, residual Zn(OH) 2 may be identified by the peak located at 39° [37]. Under the effect of thermal annealing, this peak intensity tends to reduce, which is evidence for the transformation from the residual to ZnO [38].
Along with annealing temperatures increasing, there was a change of film structure in term of crystal preferred orientation. It could be observed via the ratio of peak intensity of the axis (002) over the axis (100). While that ratio of the film with treatment at 200°C is about 2.2, the film with treatment at 350°C has a (002)/(100) ratio of 1.3. This means the film tends to transform from slightly c-axis preferred orientation to random orientation, which suggests the growth of crystal in other direction rather than (002).

Electrical properties
The electrical properties of the printed ZnO sensor were characterized by applying a bias voltage of 5 V and measuring the current under dark and UV illumination, and the result is plotted in (Figure 4(a-b)). It could be noticed that the electrical properties of printed ZnO are strongly affected by annealing conditions, such as the current under dark and UV illumination increasing with the increase of annealing temperature. By annealing the film at 350°C, the dark current had raised 78 times, from 3.86 × 10 −11 A to 3.03 x10 −9 ; and the photocurrent had notably raised 1386 times, from 1.89 × 10 −8 A to 2.62 × 10 −5 A.
Interaction of semiconductor lattice with photons that have energy equal to or larger than the bandgap of the semiconductor generates electron-hole pairs. These photongenerated charge carriers increase the charge concentration in the semiconductor film and lead to a surge of photocurrent. To evaluate the efficiency of the conversion from light to electrical current, the responsivity of the device under UV illumination is calculated by the ratio of photocurrent, J ph , with light intensity P and illuminating area S [26]: The responsivity of printed sensors with different heat annealing temperatures is given in (Figure 4(c)), which shows that the responsivity has increased from 2.5 × 10 −2 A/W to 33.9 A/W, which is over one thousand times. Hence, it could be concluded that the thermal annealing indeed improved the responsivity of the printed ZnO UV photodetector, which could be attributed to the improvement of electron transportation in the film. The specific detectivity was calculated by D* = R/(2e I dark /S) 1/2 [26], where e is the absolute value of electron charge, and the value for sample 200°C, 250°C, 300°C and 350°C are 1.86x10 11 , 8.67x10 12 , 2.87 × 10 12 and 2.91x10 13 , respectively. The external quantum efficiency (EQE) was calculated by EQE = 124000I ph /(λP) [39], where I ph is the photocurrent density. It is found that the EQE of sample 200°C, 250°C, 300°C and 350°C are 8.33%, 574.4%, 694.5% and 11,522.7%, respectively. The results demonstrate that the thermal annealing indeed improves the performance of the printed UV photodetector.
The I-V characteristic of inkjet-printed ZnO UV photodetectors are shown in ( Figure 5). It can be seen that the device with only pre-heat treatment at 200°C (Figure 5(a)) exhibits a nearly straight curve. On the other hand, the device with annealing at 350°C shows a linear I-V characteristics, which is the Ohmic property of the semiconductor-conductor contact ( Figure 5(b)). This could be the result of thermal annealing which has helped improve the contact between ZnO film and silver electrodes.
Analytically, the conductivity of semiconductor film containing small grains could be expressed as [40]: where q is the electronic charge, N d is the free carrier concentration, d is the crystallite size, m* is the effective electron mass and V d is the potential barrier between the grains. Therefore, crystallite size and charge concentration are directly proportional to the conductivity of the film, while it is an exponential decay with barrier voltage. Further investigation on these parameters will be discussed to understand the mechanism of improved responsivity of annealed ZnO.

Band bending study
Band bending at the grain boundary was characterized by impedance analysis to study the charge distribution at the film surface. By applying a potential between the ZnO thin film and the counter electrode, band bending level could be adjusted, and the applied value where the band bending is suppressed is the flat band potential, V fb [41]. In this work, the flat band potential value is employed to study the band bending of the printed film. The capacitance of the depletion layer could be adjusted by applying a potential between the ZnO film, which is the working electrode, and the counter electrode. The relationship between the capacitance, C, and the potential, V, could be described by the Mott-Schottky equation [42]: where A is the working electrode exposing area, N d is the carrier concentration of the crystal, k is the Boltzmann constant, and the thermal voltage at room temperature, i.e. kT/ e, could be found at 0.0257 V. The Mott-Schottky analysis is shown in (Figure 6), which plots the value of 1/C 2 by the potential, with a conventional resistance-capacitance model is employed [43]. From the plot, the fitted linear section intercepts with the horizontal axis at a value of V. Then, the flat band potential of each sample could be determined using the equation V fb = V-kT/e. It could be seen that the magnitude of flat band potential of ZnO thin film reduces when thermal annealing was applied, such as the sample with only 200°C treatment has the flat band potential of −0.43 V, as compared to −0.11 V of the film with annealing at 350°C. This reduction in the magnitude of the flat band potential is evidence that band bending has been reduced by thermal annealing.
In addition, the slope of linear fitting, m, could be determined from the plot, and carrier concentration could be estimated by N d = 2/(meA 2 εε 0 ) [42]. It is found that the concentration values of the films have slightly increased when annealing was applied, such as from 3.37 × 10 20 cm −3 of the sample with 200°C treatment to 4.23 × 10 20 cm −3 of the sample with 350°C treatment. These values are comparable to those of the nanoflower ZnO as reported by Allagui et al. [43].
Furthermore, the depletion layer width could be determined by D = (Aεε 0 )/C. At 0 V potential, the depletion width witnesses a reduction from 0.82 nm to 0.43 nm when applying thermal annealing. The reduction of the depletion layer means that there is less barrier between the grains, and therefore electrons can be transported through the boundaries easier.

Film conductivity by band bending effect
From the previous discussion, the conductivity of the film is proportional to crystallite size and charge concentration, which are both slightly increase by the thermal annealing. Hence, they would have a minor effect on the conductivity. However, it could be seen that the conductivity increases significantly, which might be mainly due to the influence of band bending reduction. An illustration of this mechanism is shown in (Figure 7), whereby the film is assumed to be a series of grains connected at the grain boundaries. Band bending at the grain boundaries tends to resist electron transportation through the film. Flat band potential analysis shows that film annealed at a higher temperature exhibits a lower band bending magnitude, which might be the result of the merging of grain boundary as suggested by the XRD analysis.
Furthermore, XRD analysis also suggests that there is a transformation from the residual to ZnO under the effect of heat treatment. Those residual defects could play the role of recombination centers, where photo-generated electrons are recombined with holes which reduces the carrier concentration in the printed film [44,45]. Hence, decreasing the residual could prolong the lifetime of photo-generated electrons and improve the photocurrent, as can be observed in the annealed sample.

Dynamic response of the printed sensor to UV illumination
When the sensor was exposed to UV light, the electrical current raised until it reaches the maximum value. When the UV light was removed, the current gradually decays to the original value. These dynamic responses are presented in (Figure 8(a)). The rise time and fall time of the response are defined by the time the signal raises from 10% to 90% and vice versa.   In general, the response time of the printed device to UV illumination is long, which is the result of air desorbing from the surface of the semiconductor due to photo-generated charge carrier. Without UV illumination, oxygen absorbs to ZnO surface by receiving a free electron, causing the lack of charge carrier. Under UV illumination, the generated hole recombines with trapped electron, releasing oxygen and increasing electrical current running through the device [26]. Such air exchanging is the main cause of the long response to UV light of printed device.
The difference in time response characteristics between printed devices with different thermal treatment conditions could be correlated to the surface defects modification as the effect of heating. It is because the transition time of the photodetector reflects the interactions of the semiconductor with ambient air, which are the absorption and desorption of air molecules to the defect sites [46]. Hence, investigating chemical states on the film surface is necessary to understand the photo-response behavior of the printed device.
To evaluate the charge states on the surface of printed films, XPS spectra of the films with a particular focus on the oxygen spectra O1 s are investigated as shown in (Figure 8 (b)). The asymmetric patterns are observed in both samples, which could be deconvoluted to three energy peaks, such as 530.1 eV, 531.0 eV, and 532.4 eV, which are linked to lattice oxygen (O1), oxygen vacancies (O2), and surface oxygen (O3), respectively [47,48].
In general, lattice oxygens of samples processed at 200°C, 250°C, 300°C and 350°C are 33.8%, 36.0%, 24.8% and 41.0%, respectively. Meanwhile, oxygen vacancies slightly change from 26.1% to 31.0%, 24.9% and 25% when higher annealing temperature were applied. Notably, excess oxygen shows a significant variation from 40.1% to 33.1%, 50.3% and 29.0%, according to sample processed at 200°C, 250°C, 300°C and 350°C, respectively. These analyses demonstrate the transformation of excess oxygen to lattice oxygens [48], which is in good agreement with XRD analysis. The reduction of excess oxygen may also lead to less interaction with ambient air, which shows a slower response to UV illumination. In addition, a higher lattice oxygen percentage means that there are more sites are interacting with UV photons and more free charger carriers are generated, which could also promote the photocurrent. Furthermore, at temperature lower than 300°C, the excess oxygen tends to be transformed to oxygen vacancies, while oxygen vacancies tends to be transformed to lattice oxygen due to the filling of free oxygen to the vacancies at a higher temperature. This observation is agreed with previous report which suggested that high processing temperature could widen the gaps of ZnO crystal, hence promoting the filling of oxygen atoms to the gaps [48].
Beside the enlargement of the grain size, the thermal annealing of inkjet printed ZnO could help improve the crystal quality. As can be seen in (Figure 7(b)), there is a generally increasing of lattice oxygen, which could benefit the generation of electron-hole under high energy photon and boost the current transportation. Table 1 presents the comparison of the UV photodetector performance of this work with some representative studies which employ the same polycrystalline thin film of ZnO. It can be seen that in terms of rise time and decay time, this work shows comparable performance with films prepared by flame spray pyrolysis approach [49,50], but much slower than those prepared by sol-gel thin film deposition [51,52]. The root cause of this poorer performance could be due to the nature of wrinkling film prepared in this work, which allows more air interaction with the film. This wrinkling structure may also be responsible for the better responsibility of the inkjet printed film as compared to films prepared by flame spray pyrolysis approach.

Conclusion
This work studied the electrical characteristics of inkjet-printed ZnO for electrical applications under thermal annealing, which shows a significant responsivity enhancement of more than one thousand times. Band bending at the grain boundaries was also found to be altered by the thermal annealing, which suggests that the annealing might affects the electrical characteristics of the printed film via the band bending modification. Whereby, the merging of grain could lower the barrier between nanoparticles and improve electron mobility. The response characteristics of the semiconductor to UV illumination are also affected by thermal treatment due to the change in surface states as found by XPS analysis. These results provide a fundamental understanding of structural and electrical properties of inkjetprinted semiconductor film using precursor as ink, which is crucial in additive manufacturing of electronic devices.

Acknowledgments
We appreciate the equipment support from Singapore Centre for 3D Printing (SC3DP). We thank Assistant Professor Lau Gih Keong (NTU) for allowing us to use the source metering unit (Agilent B2902A). We thank Associate Professor Fei Duan (NTU) for allowing us to use the tube furnace.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
This work was supported by Nanyang Technological University and the Ministry of Education of Singapore through a Ph.D. Scholarship and AcRF Tier 1 research grant (RG 96/18).