Low-temperature facile solution-processed gate dielectric for combustion derived oxide thin film transistors

Abstract

In this work, acetylacetone assisted solution-processed In–Ga–Zn–O (IGZO) thin film transistors (TFTs) using Al₂O₃ as gate dielectrics were investigated. Normally, fully covered Al₂O₃ thin films are difficult to achieve by spin coating with conventional solvent, such as 2-methoxyethanol, due to the poor wettability of highly doped silicon. Here, a conventional aluminum nitrate solution with an additive was designed to spin coat robust continuous Al₂O₃ thin films, resulting from improved solution hydrophilic with a contact angle of 17°. For active layer fabrication, we utilized the previous reported combustion process to lower treatment temperature, which could be confined in the range from 220 °C to 300 °C, without losing the device performance. Results show that all the devices performed well. Especially, after 240 °C annealing of both Al₂O₃ (in thickness of around 45 nm) and IGZO thin films (in thickness of around 30 nm), we have obtained the following device parameters, namely a Al₂O₃ dielectric breakdown electric field at 7.8 MV cm⁻¹, a current density of around 1 × 10⁻⁶ A cm⁻² in the voltage range of −3 V to 3 V, a areal capacitance of 291 nF cm⁻² at 100 Hz, a carrier mobility of 0.74 cm² V⁻¹ s⁻¹, a threshold voltage of −0.4 V, a current on–off ratio of 6 × 10³, a subthreshold swing of 375 mV per decade. Fabrication of combustion-processed active layers and our facile solution processed high-k dielectrics provides a feasible approach for low cost oxide flexible TFTs applications.

---

1. Introduction

Oxide based semiconductor devices, especially thin film transistors (TFTs), have been studied for decades, as they are important building blocks in optoelectronic devices.^1,2 Until now, oxide materials have been widely explored because of their earth abundance, low cost and environmental friendly features as well as easy processing, which are favorable for industry production and sustainable development.^3,4 To replace traditional silicon dioxide dielectrics, considerably smaller working voltage with an augmented TFTs mobility were anticipated from some oxide TFTs gate dielectric candidates such as Al₂O₃, ZrO₂, and TiO₂ due to proper trade-off among dielectric constant, energy gap and conduction band offset on silicon.^5–11 On the other hand, in comparison with crystalline active-layer TFT, ZnO based semiconductors, such as In–Zn–O (IZO), Zn–Sn–O (ZTO), IGZO are superior candidates towards TFTs active channel, featured by amorphous phase but excellent field effect performance under a relative low processing temperature.^12–15 As a result, zinc oxide based TFTs have received an extensive attention on lowering processing temperature, so as to ensure their various potential applications.^16–18

Recently, solution-processed approach turns to be a major fabrication route for TFT active channels due to its advantages such as easy operation, being vacuum free, and having a controllable element stoichiometric ratio.^19–22 However, in comparison with conventional physical deposition processes, namely radio frequency magnetron sputtering, solution-processed techniques still require a high annealing temperature to facilitate the precursor transformation to oxide and to dense the film at the same time. Therefore, to address that problem, a combustion route has been developed by successfully lowering the annealing temperature by Tobin Mark's group.^23–26 The key point is to introduce either urea or acetylacetone to serve as fuel. As soon as a moderate temperature is provided to drive the reaction, coordinating with oxidizer of NO³⁻, the fuel gives rise to a high enthalpy change in the exothermic reaction, soon afterwards forming a self-energy-generation system that continues a thorough chemical reaction.

Gate dielectric also plays a significant role on TFT performance. To ensure a low leakage current density and high breakdown voltage, the frequently used fabrication route is physical deposition. For example, pulsed laser deposition and radio frequency magnetron sputtering.^27,28 In addition, with regard to the solution process, the thorny challenge of the poor wettability of the highly doped silicon and the inferior hydrophilicity of the general solvent also needs to be settled. 2-Methoxyethanol is one of the general solvent in spin-coating process on thermal grown SiO₂ substrates, but its spin window is quite narrow on highly doped silicon substrates, which implies that a high quality film is hardly obtained even after an effective oxygen plasma or piranha solution treatment to obtain a super hydrophilic surface on substrates by attaching hydroxyl groups.^8,10

In this report, a bunch of all solution processed TFTs have been systematically investigated, aiming to fabricate high performance oxide TFTs via a low processing temperature and facile procedures. First, Al₂O₃ thin film, a high-k oxide dielectric, was successfully fabricated by an additive-assisted solution process. By utilizing acetylacetone and aqueous ammonia as an additive in 2-methoxyethanol, Al(NO₃)₃ solution can be spin-coated and fully covered onto highly doped silicon substrates easily. This indicates that the novel solution system broadens the fabrication window tremendously. In combination with combustion processed IGZO active layers, we demonstrated that by utilizing novel chemical processed Al₂O₃ as dielectric, after a mere 240 °C annealing on both gate dielectric and active channel, TFT devices showed a mobility of 0.74 cm² V⁻¹ s⁻¹, an operating voltage range of −3 V to 3 V and threshold voltage of −0.4 V. Beyond that, when processing temperature increased to 260 °C, the mobility was increased to 2.26 cm² V⁻¹ s⁻¹, and the threshold voltage was reduced to 0 V. The trade-off between treatment temperature and field effect performance implies a feasible trail of low cost and easy processing of flexible devices fabrication. Although acetylacetone is utilized in both the dielectric and semiconductor layer fabrication, the role is totally different, which means it serves as an additive in Al³⁺ solution to improve wettability (after spin coating, acetylacetone volatilizes out at 120 °C), while serving as a fuel in IGZO solution to participate in the reaction at no less than 220 °C.

2. Results and discussion

To demonstrate the improved wettability of the novel Al₂O₃ precursor solution, we compared the contact angles of three solutions dropping onto highly doped silicon substrates without oxygen plasma treatment: conventional solution without additive; conventional solution with acetylacetone only and our novel solution. The measurements are shown in Fig. 1. The contact angles were 26°, 24° and 17°, respectively. Although the difference is small and in principle it could be ignored, after the solution is dropped onto oxygen plasma treated highly doped silicon substrates, just aging without any further action, the wettability varies. The contact angles cannot be measured through contact angle microscope due to the super hydrophilicity resulting from oxygen plasma treatment. Then, to demonstrate the wettability of three solutions, we pictured the spreadability of these three solutions, after a drop of solution from a syringe is added on the oxygen plasma treated highly doped silicon substrates and aging for 5 min, as shown in Fig. 2. After the addition of acetylacetone and aqueous ammonia, solution wettability is improved significantly. The reason is that the carbonyl group of acetylacetone and the amido group of aqueous ammonia are both hydrophilic, hence they are beneficial for solution wettability. In addition, the conventional solution of the additive of aqueous ammonia alone cannot be obtained because of the product of unsolvable Al(OH)₃ obtained from the reaction. The media of acetylacetone obstructs the reaction between Al³⁺ and aqueous ammonia, so that aqueous ammonia can be added successfully with a concentration of no more than 60% of acetylacetone. To further support the competence of the novel solution, a comparison of thin film qualities after spin coating and annealing between the conventional and novel solutions was also conducted. Fig. 3(a)–(c) show optical microscope pictures of the samples, in which all of them were of 2 layers, and the final annealed temperatures were 220 °C. Without any additives, spin-coated thin film is partially covered on substrates and seems uneven. The blue parts have2 layers, the brown parts have 1 layer, and the white round shaped part is uncovered substrate. When conducting a combustion process, which means annealing the thin film at 220 °C directly, with the addition of acetylacetone and aqueous ammonia, poor quality films with a big mass of pores form (the white random-shaped dots). Nevertheless, to avoid the combustion process, while conducting a step annealing starting at 120 °C, the spin-coated thin film (the brown part) is fully covered on substrates, and seems very uniform. The white round ones are Al electrodes with a radius of 100 μm. A superior wettability of solution guarantees the good quality of thin films.

Fig. 1 Contact angle measurements of solutions without additive, with acetylacetone only and with acetylacetone plus aqueous ammonia. The substrates are highly doped silicon without oxygen plasma treatment.

Fig. 2 Wettability comparison of a conventional solution, a conventional solution with acetylacetone only and our novel solution. The substrates are oxygen plasma treated highly doped silicon.

Fig. 3 Optical microscope pictures of the samples, all of which are of 2 layers, annealed at 220 °C, and magnified ×50. (a) Using a conventional solution without additive. (b) Using a novel solution to conduct a combustion reaction, annealing at 220 °C directly. (c) Using a novel solution and conducting step annealing, starting at 120 °C (to avoid combustion reaction).

As novel chemically-processed Al₂O₃ gate dielectrics, X-ray diffraction characterization of Al₂O₃ obtained after annealing at 300 °C is shown in Fig. 4. Since no obvious diffraction peaks were observed, the fabricated Al₂O₃ dielectrics were amorphous, thus minimizing leakage current and maximizing breakdown voltage because of the absence of crystal grain boundaries. Surface morphologies of the Al₂O₃ film surfaces obtained after annealing at 220 °C and 300 °C were examined by AFM under tapping mode, as shown in Fig. 5(a) and (b). Scan scale was 1 μm × 1 μm. It revealed that after being spin-coated twice through a 0.3 M Al(NO₃)₃ solution, the surfaces were not as smooth as thermal grown SiO₂ surface (RMS ∼ 0.19 nm), but they reached a RMS roughness of 0.5 nm and 1.1 nm, respectively. The roughness mainly came from the film growth and the densification processes. Because of nucleation and coalescence of chemical species, the film growth resulted in noticeable ups and down on the substrate plane, while the densification process resulted in the formation of mesoscopic holes.

Fig. 4 X-ray diffraction characterization of Al₂O₃ gate dielectric thin film annealed at 300 °C.

Fig. 5 (a) AFM characterization of Al₂O₃ gate dielectric thin film surface annealed at 220 °C. RMS is about 0.5 nm. (b) AFM characterization of Al₂O₃ gate dielectric thin film surface annealed at 300 °C. RMS is about 1.1 nm.

X-ray photoemission spectroscopy analysis of Al₂O₃ films obtained after annealing was performed to interrogate the chemical states of oxygen, shown in Fig. 6(a). After deconvolution of the O 1s peak of Al₂O₃ thin films, three peaks centered at 532.0 eV, 532.8 eV and 534.0 eV can be found separately. The peak centered at 532.0 eV can be assigned to lattice oxygen, which forms fully bonds with aluminum atoms. The peak centered at 532.8 eV can be assigned to the oxygen atoms that are not fully bound within the oxide, which are usually related to the oxygen vacancy in oxygen-deficient regions. Because of the partial bound oxygen atoms within the oxide, the peak of XPS feature towards to a higher binding energy than fully bound oxygen atoms. The peak centered at 534.0 eV can be assigned to the oxygen in hydroxide. Since hydrogen is more electronegative than metals, the M–OH oxygen atoms are less negatively charged than those in oxides, shifting the XPS feature toward a considerably higher binding energy.^12,15,20 The relative fractions of these three species were calculated from the XPS semi-quantitative analyses. The relation between the percentages of oxygen species and the annealing temperatures are shown in Fig. 6(b). As the annealing temperature increases, the change of concentration in oxygen vacancies is unobvious, but the concentration of oxygen in hydroxide decreases and the lattice oxygen one increases. The results suggest that a higher annealing temperature favors the decomposition Al(OH)₃ to Al₂O₃.

Fig. 6 (a) O_1s XPS spectra of Al₂O₃ gate dielectric thin films annealed at 220 °C, 240 °C, 260 °C, 280 °C and 300 °C. (b) Dependency of percentage of oxygen species and annealing temperature.

Dielectric characteristics were measured through the (metal–insulator–metal) MIM structure of Al₂O₃ thin film. Current densities versus breakdown electric field characteristics are shown in Fig. 7(a). The breakdown electric field of all samples were no lower than 6.5 MV cm⁻¹. Due to the increase of oxygen vacancies as the annealing temperature raised, the breakdown electric field was degraded. However, after spin coating twice and film thickness increasing to around 45 nm, a significantly increased breakdown electric field is achieved. Thus, even with an annealing temperature as high as 300 °C, the breakdown electric field was still close to 7.5 MV cm⁻¹, laying a good foundation for TFT applications. The current density versus voltage of Al₂O₃ dielectrics is shown in Fig. 7(b). In order to match TFTs operating voltage range in convenience, here we adopt voltage instead of electric field as abscissa. The results indicated that leakage current densities were approximately 10⁻⁶ A cm⁻² in the operating voltage range of −3 V to 3 V (the gate voltage of fabricated TFT was −3 V to 3 V), regardless of the annealing temperature. The performed leakage current is acceptable in TFTs applications. The relationship between the areal capacitance of Al₂O₃ films and the frequency obtained in the capacitance–voltage measurements in both ambience and vacuum are shown in Fig. 8(a) and (b). The capacitance density increases along with the annealing temperature, because of the increase of Al₂O₃ percentage. The results also demonstrate an obvious frequency dispersion in the low frequency range, which indicates the increase of capacitance density with decreasing frequency.^10,29 Comparing with capacitance in ambient and vacuum environments, it is obvious that the frequency dispersion and capacitance in vacuum are quite smaller than those in ambience, which may suggest an existence of proton conduction in ambient conditions.^30–33 The proton mobile ion from absorbed water in ambience induces a significant capacitance through its migration and formation of an electric double layer (EDL) under a voltage bias. When the gate electrode is positive biased, proton migrate to the interface of the dielectric and semiconductor, composing a Helmholtz layer. However, because of the slow migration rate, the induced capacitance only responses in the low frequency range, bringing an obvious frequency dispersion. Characterization of phase angle and conductivity versus frequency in ambient environment were shown in Fig. 9(a) and (b). It is known that the phase angle of an ideal capacitor is −90° while for an ideal resistor it is 0°.^30,31 The fabricated Al₂O₃ dielectrics perform as a capacitor in all frequency ranges. This result suggests that the polarization within the dielectric media dominates the capacitance and frequency dispersion. The conductivity is also frequency-dependent. A similar phenomenon has also been found in previous work.³⁴ This is due to the ionic migration in ion-conducting disordered solid materials, causing the leakage current.³⁴

Fig. 7 (a) Current density versus electric field characteristics of MIM structure of Al₂O₃ gate dielectric thin films annealed from 220 °C to 300 °C. (b) Current density versus applied voltage characteristics of MIM structure of Al₂O₃ gate dielectric thin films annealed from 220 °C to 300 °C.

Fig. 8 (a) Dielectric capacitance versus frequency of Al₂O₃ gate dielectric thin films annealed from 220 °C to 300 °C, characterized in ambience, voltage bias was 3 V. (b) Dielectric capacitance versus frequency of Al₂O₃ gate dielectric thin films annealed from 220 °C to 300 °C, characterized in vacuum, voltage bias was 3 V.

Fig. 9 (a) Phase angle versus frequency of Al₂O₃ gate dielectric thin films annealed from 220 °C to 300 °C, characterized in ambience, voltage bias was 3 V. (b) Conductivity versus frequency of Al₂O₃ gate dielectric thin films annealed from 220 °C to 300 °C, characterized in ambience, voltage bias was 3 V.

Moreover, the proton conduction, here we also understand the frequency dispersion phenomenon through the fundamental of intrinsic dielectric, as the Al₂O₃ dielectrics perform a capacitor behavior all through the frequency range. Dielectric constants comes from dipole moment, in other words, coming from polarization of dielectrics. There are basically three kinds of polarizations in the system: (1) the electronic polarization that comes from Al₂O₃, (2) the space charge polarization that comes from trapped charges, which is related to oxygen vacancies and mobile charges, which at the same time is related to hopping from hydroxyl sites, (3) and the orientation polarization that also comes from hydroxyl groups. Electronic polarization, which due to elastic displacement of electron clouds requires a very short time to perform, is almost frequency independent. Space charge and orientation polarization both come from the movements of charged particles, such as orientation of dipoles and migration of electrons, hence they require a longer time to perform, so they are largely frequency dependent. Within the low frequency range, electronic polarization has enough time to respond to external electric fields, but space charge and orientation polarization have not enough time, so it induces frequency dispersion. A semi-quantitative analysis is followed to illustrate the contribution of the three kinds of polarizations.

The total polarization can be written as:

P = (n_eα_e + n_sα_s + n_oα_o)E (1)

where n mean the concentrations of lattice oxygen (which are related to the electronic polarization), oxygen vacancies (which are related to trapped charges as part of space charge polarization) and oxygen coming from hydroxide (which are related to orientation polarization and mobile charge hopping as part of space charge polarization). α mean polarizability, and the subscripts e, s and o mean electron, space charge and orientation, respectively. All the polarizabilities are functions of the frequency.

Following our discussions, the electron polarizability α_e should be a constant. The other two polarizabilities have similar formulas to the Debye's model:

where τ is the relaxation time, α₀ is the optical polarizability and α₀ + α₁ is static polarizability. In our model, we assume they change little with the annealing temperature. Considering that the electron polarizability is almost independent of the frequency, after combining the coefficients, we have:where the subscripts 0, 1 and e, s, o have the same meanings as mentioned above. Considering that the capacitance density and the polarization have a linear relationship about the macro shape of the device, we use A and B to replace α₀ and α₁ because the difference between them is just a constant:

At the low frequency range, the capacitance difference between 20 Hz and 5000 Hz of the five samples increases along with temperature, but the percentage of hydroxyl groups decreases, which indicates that the orientation polarization has little responsibility on the frequency dispersion in the low frequency range. Now, the total capacitance can be rewritten as:

The real part of (5) can correspond to our experiment results of the capacitance density. However, because the space charge polarization relates to both oxygen vacancies and hydroxyl sites, here we divide the total space charge polarization into two parts. One part comes from the contribution of trapped charges and another comes from mobile charge hopping. The two parts only differ in the constant of proportionality. Now, the real part of total capacitance density, which relates to the experimental capacitance – frequency results, can be rewritten as:

This equation will be used to fit the experimental results, which were obtained in vacuum to exclude the proton conduction effect.

We can estimate A_e with the permittivity of pure Al₂O₃, assuming our devices are fulfilled with Al₂O₃, we get A_e = 197 nF cm⁻². As we know, the concentrations obtained from the XPS data are not strictly quantitative, so we selected the data of annealing temperatures of 240 °C and 260 °C to fit with formula (6). We let n_e, n_s-vac and n_s-OH to be the same to the concentrations of lattice oxygen (Al–O–Al), oxygen vacancies (O_vac) and oxygen in hydroxide (–OH) separately. A comparison between the fitting and experiment results is shown in Fig. 10. The fitted data are listed in Table 1. It suggests the relaxation time for space charge polarization is approximately in the order of 10⁻⁴ to 10⁻³.

Fig. 10 Least Square fitting of capacitance – frequency dependency.

Table 1 Least Square fitting data of capacitance – frequency dependency characterized in vacuum

Capacitance – frequency dependency fitting results
T (°C)	As-vac (nF cm^−2)	Bs-vac (nF cm^−2)	τs-vac (s)	As-OH (nF cm^−2)	Bs-OH (nF cm^−2)	τs-OH (s)
240	239	41	0.0001	0.7	113	0.0029
260	267	34	0.0001	0.3	80	0.0022

---

To further demonstrate the dielectric polarization mechanism, temperature dependence of dielectric performances of the sample annealed under 240 °C were measured in vacuum. Capacitance-frequency dependency and Cole–Cole plot are shown in Fig. 11(a) and (b), respectively. When the measuring temperature increases, the sample exhibits increased frequency dispersion in the low frequency range. This phenomenon is attributed to an increased hopping probability of space charge, as a result of raised activation energy. By combination of the capacitance performances that are close to the end of the high frequency range (10 KHz to 1 MHz) and Cole–Cole plot, the corresponding Debye relaxation times can be obtained using:

τ = 1/2πf (7)

where f is the frequency with maximum imaginary part on the arc, being 700 KHz at 330 K and 630 KHz at 370 K. The Debye relaxation times are 2.3 × 10⁻⁷ s at 330 K and 2.5 × 10⁻⁷ s at 370 K. At 293 K, the frequency with maximum imaginary part is beyond our measuring range (>1 MHz). Higher measuring temperature leads to an increase of the relaxation time. Therefore, this relaxation in the 10 KHz to 1 MHz range is related to the orientation of polarization, featured by the rotational process, which encounters a resistance due to thermal agitation.

Fig. 11 (a) Capacitance-frequency dependency of samples annealed at 240 °C via variation of measuring temperature from 293 K to 370 K in vacuum. (b) Cole–Cole plot (ε′′ vs. ε′) of samples annealed at 240 °C, measured in vacuum.

Structural properties of IGZO thin films were also characterized by X-ray diffractometer, a shown in Fig. 12. Due to the weak intensity of the diffraction peaks, which were almost in the same order of background noise, the thin films were regarded to have amorphous phase. Measurements of TFT transfer and output characteristics were also carried out. Devices configuration and electrical characteristics are shown in Fig. 13(a)–(c) and S(1).†³⁵ The field effect mobility (μ_sat) of the TFTs is extracted in the saturation region of the transfer curve by the following equation:

where I_D is the drain to source current, V_G is the gate to source voltage, V_T is the threshold voltage, W and L are the channel width and length, respectively, and C_ox is the capacitance per unit area of the gate dielectric. By matching the transfer performance measurement, the capacitance at low frequency is often adopted to calculate mobility. Here, the value at 100 Hz of Al₂O₃ dielectric capacitance-frequency characteristic was selected. Sub-threshold swing (SS) was calculated using the following equation:

SS = dV_G/d(log I_D)(V_G < V_T) (9)

Fig. 12 X-ray diffraction characterization of IGZO active layer annealed at 220 °C and 300 °C.

Fig. 13 Device configuration and electrical characteristics of TFTs composed of combustion-processed IGZO active channel with spin-coated Al₂O₃ gate dielectrics. (a) Device configuration; (b) transfer and (c) output characteristics of IGZO TFTs annealed from 220 °C to 300 °C.

By utilizing SS value, the maximum interface trap states (N^max_trap) can also be calculated from the following equation:

Device parameters were summarized in Table 2. By comparing device parameters comprehensively, samples obtained after annealing at 260 °C showed the best performances, with a mobility of 2.26 cm² V⁻¹ s⁻¹, a zero threshold voltage, an I_on/I_off of around 3 × 10⁴, a sub-threshold swing of 186 mV per decade, and a maximum trap density of 4.2 × 10¹² cm⁻² eV⁻¹. However, samples obtained after annealing at 220 °C are also operable, albeit with an inferior mobility of 0.12 cm² V⁻¹ s⁻¹, a threshold voltage of only −1.27 V, an I_on/I_off of 2 × 10³, a sub-threshold swing of 447 mV per decade, and a maximum trap density of 1.3 × 10¹³ cm⁻² eV⁻¹. In regards to this type of semiconductors, for example, amorphous IGZO, their carrier transport results from direct overlap among the neighboring metal ns orbitals.¹⁵ The magnitude of this overlap is insensitive to the distorted metal–oxygen–metal (M–O–M) chemical bonds that intrinsically exist in amorphous materials.¹⁴ Owing to both oxygen vacancies and traps, the TFT mobility varies with sample annealing temperature. When the temperature increases from 220 °C to 300 °C a continuous mobility increase is observed, which is believed to be related to pre-filling trap sites generated by oxygen vacancies, and/or distortional relaxation reduced trap sites.³⁶ When annealing temperature is higher than 260 °C, the mobility enhances slowly and approaches its saturation. A negative threshold voltage indicates that the TFT operation mode is depletion one, in which the channel is activated as normally on. This implies that electrons are accumulated at the interface between the dielectric and the active layer, requiring a negative gate bias to deplete them. Moreover, the interface charges, the trap density also plays an important role in the threshold voltage, such as hydroxyl groups, which can be viewed as electron trap sites that attract electrons and limit their movement. Therefore, trap density affects free carriers number. A higher number means it is easier to make carriers accumulate in the semiconductor/dielectric interface, hence it contributes to a small threshold voltage. In other words, a reduction of the trap density always gives rise to a shift in the threshold voltage toward a negative bias direction. It also means that a higher trap density requires a larger gate voltage to turn on the device. A relative phenomenon has previously been found.³⁷ In comparison with their electric hysteresis from dual drain voltage sweep, ΔV_T is less than −0.1 V, which can be ignored. The clockwise hysteresis effect, as shown in Fig. 13(b), is due to the IGZO/Al₂O₃ interface. Under a positive gate bias, it is possible for electrons to be injected from IGZO into Al₂O₃ dielectrics, accompanied by a decrease in the carrier density in the vicinity of the IGZO/Al₂O₃ interface, thus lowering the I_D level and enhancing a more positive threshold voltage in the transfer reverse sweep.⁹ A similar phenomenon was also found in sputtered Al₂O₃ dielectrics. I_on/I_off is determined by both off-current and on-current. A small off-current will be obtained as long as the dielectric exhibits a low leakage current density and the semiconducting active channel has a low concentration of oxygen vacancies. A large on-current will be obtained if the device performs a high carrier mobility. Therefore, they contribute to a good I_on/I_off. Devices sub-threshold swing are 447 mV per decade, 375 mV per decade, 186 mV per decade, 265 mV per decade, 372 mV per decade, for samples annealed at 220 °C, 240 °C, 260 °C, 280 °C and 300 °C, respectively. In general, the sub-threshold swing is an indicator of the total trap density, including the bulk trap density of the semiconductor itself and the interface trap density in the vicinity of the interface between the semiconductor and dielectric layers.³⁸ The 260 °C annealed samples exhibit minimum sub-threshold swing, giving a relative fast response or switch speed. From our previous experiments, we know that both the dielectric and the semiconductor layers account for the non-monotonic dependence. The higher roughness of dielectric surface has a significant role, degrading TFTs performance due to the scattering of electrons. Moreover, when the solution processed semiconductor layer is under annealing, the thermal treatment also affects the dielectric layer. The reason might be the interdiffusion among Al³⁺, In³⁺, Ga³⁺ and Zn²⁺, which has also been previously discussed by other authors.^39,40 In our experiments, it is suggested that In³⁺, Ga³⁺ and Zn²⁺ in the combustion solution should be more likely to diffuse into Al₂O₃ layer at higher treatment temperature because of (1) they are more free to move in the solution than bonded Al³⁺ in the Al₂O₃ layer, and (2) on account of the self-energy generated combustion process, those ions are easily activated to diffuse into the Al₂O₃ layer and reduce the device performance.

Table 2 Electrical parameters of combustion processed IGZO TFTs fabricated using Al₂O₃ gate dielectric films annealed from 220 °C to 300 °C

TFTs electrical performances
T (^oC)	Cox (nF cm^−2)	μ (cm^2 V^−1 s^−1)	VT (V)	Ion/Ioff	SS (mV per decade)	N^maxtrap (cm^−2 eV^−1)
220	317	0.12	−1.3	2 × 10^3	447	1.3 × 10^13
240	291	0.74	−0.4	6 × 10^3	375	9.6 × 10^12
260	319	2.26	0	3 × 10^4	186	4.2 × 10^12
280	405	2.28	−0.5	7 × 10^3	265	8.7 × 10^12
300	377	2.82	−1.9	4 × 10^3	372	1.2 × 10^13

---

3. Experimental procedures

Material synthesis

All chemicals were supplied by Sigma-Aldrich. Aluminum nitrate solution was prepared by dissolving 1.149 g aluminum nitrate nonahydrate (Al(NO₃)₃·9H₂O, 99.997%) in 10 ml 2-methoxyethanol, which was vigorously stirred at 50 °C until a transparent solution formed. Then, 0.4 ml acetylacetone and 0.2 ml NH₃·H₂O were dropped in, and the solution was vigorously stirring overnight at 50 °C to achieve a homogeneous and stable mixture. For the preparation of the IGZO active layer, doping of indium was aimed to increase the TFTs mobility due to the overlap of 5s orbitals of indium atoms, while doping of gallium was aimed at enhancing the semiconductor stability owing to a strong bond energy between gallium and oxygen atoms. However, when the concentration of gallium was greater than 20%, a degradation of mobility occurred.²⁶ Hence, we selected the mole ratio of In : Ga : Zn of 65 : 10 : 25. IGZO solution was prepared by dissolving 0.075 g zinc nitrate hexahydrate (Zn(NO₃)₂·6H₂O, 99.999%), 0.196 g indium nitrate hydrate (In(NO₃)₃·xH₂O, 99.9%), 0.026 g gallium nitrate hydrate (Ga(NO₃)₃·xH₂O, 99.9%) in 10 ml 2-methoxyethanol, and it was vigorously stirred at 50 °C until all nitrates were dissolved. Afterwards, 0.2 ml acetylacetone and 0.1 ml NH₃·H₂O were dropped in, and the solution was vigorously stirred overnight at 50 °C to form a homogeneous and stable mixture.

Thin film fabrication

P⁺⁺–Si substrates were sonicated with acetone and ethanol, followed by nitrogen gas purging. Afterwards a 5 min oxygen plasma treatment was conducted to clean the surface. Before spin coating, both solutions were filtered through 0.22 μm PTFE membrane syringe filters. P⁺⁺–Si substrates were heated at 120 °C for 10 min on a hot plate before spin coating. Then, aluminum nitrate solution was spin-coated at 3000 rpm for 10 s. After spin coating, the substrates were placed on a hot plate and annealed at 120 °C for 10 min, then annealed at 160 °C for 10 min, followed by annealing at 220 °C for 30 min, and finally annealed at intended IGZO treatment temperature for 1 h. The reason for step annealing was to the high-reactivity of the aluminum ion. If a combustion process for aluminum oxide layer was categorically conducted, which means the spin-coated films were heated directly at 220 °C, on account of sharp combustion reaction, poor quality films would be formed with a mass of pores, leading to a high leakage current. This relevant phenomenon has also been reported by other authors.⁴¹ However, if the films were heated at a low temperature first, the input energy was insufficient to initiate the combustion process, meanwhile the fuel – acetylacetone, would also volatilized. Thus, compared to the combustion process, the oxidation reaction slowed down to avert gas pore formation. Therefore, in this dielectric film fabrication process, the acetylacetone was conducive to a definitely successful spin-coating process to attain the hydrophilicity of the solution, matching significantly with highly doped silicon, but did not serve as a fuel. Thus, the process is just a conventional oxidizing process instead of a combustion, in which the additive plays a role to improve the solution wettability, albeit the same recipe as IGZO solution. To guarantee a high breakdown voltage and a low leakage current, another spin-coating process was repeated. After the substrate coated with the aluminum oxide layer, spin-coating time could be extended to 20 s. The 2-layer dielectrics had a thickness of ∼45 nm. Before IGZO thin film spin-coated, aluminum oxide was pre-heated to 120 °C for 10 min. The solution was spin-coated on it at 3000 rpm for 10 s, then annealed directly at intended temperature (220–300 °C) for 1 h to proceed to the combustion reaction. Moreover, the procedure was repeated once but spin-coating time extended to 20 s. The 2-layer active channel was achieved with a thickness of ∼30 nm.

Device fabrication

To characterize electrical properties of gate dielectric, a 100 nm-thick circular Al top electrode was deposited onto an aluminum oxide film by thermal evaporation process via a shadow mask (electrode radius was 100 μm) to construct a metal–insulator–metal (MIM) structure. To fabricate TFTs, a bottom gate, top contact structure was used, because it is easy to handle. Moreover, a 100 nm-thick Al source and drain electrodes were deposited on IGZO thin film by thermal evaporation through a shadow mask. The width of the channel (W) was 1500 μm, and the length of the channel (L) was 50 μm. 9 devices of each temperature were fabricated and the average performances were given.

Characterization methods

The contact angles of the solutions were measured with a contact angle microscope. X-ray diffraction analysis of gate dielectrics and active channel was performed by Siemens D5005 X-ray diffractometer using Cu K_α radiation. Surface morphologies of Al₂O₃ and IGZO thin film were characterized by atomic force microscope (AFM), Vecco Nanoscope IIIa. X-ray photoelectron spectroscopy was carried out by VG Scientific ESCALAB 250. The frequency-dependent capacitance of MIM structure was measured by HP 4284A, in a frequency range of 20 Hz to 1 MHz. Breakdown voltage and leakage current of MIM structure were measured by Keithley 4200 SCS Analyzer under ambient conditions. TFTs characteristics were also characterized by Keithley 4200 SCS Analyzer under ambient conditions.

4. Conclusions

In this report, all solution-processed IGZO thin film transistors (TFTs) with Al₂O₃ gate dielectrics were successfully fabricated at low temperature by a low cost method. Using acetylacetone and aqueous ammonia as an additive, aluminum nitrate solution can be spin-coated on highly doped silicon and it easily forms a fully covered continuous film. Good quality amorphous Al₂O₃ gate dielectrics were obtained with a solution process at a low temperature of 220 °C. By utilizing combustion process for IGZO thin film fabricated on the top of Al₂O₃ gate dielectric, a high quality TFTs has been fabricated at a low temperature of 240 °C. Combination of combustion-processed active layers and solution-processed high-k dielectrics provides a feasible approach for low-cost flexible oxide TFTs applications, such as utilizing polyimide substrate after selecting a proper gate electrode.

Acknowledgements

The work is in part supported by Research Grants Council of Hong Kong, particularly, via Grant nos AoE/P-03/08, N_CUHK405/12, and CUHK Group Research Scheme. J. B. Xu would like to thank the National Science Foundation of China for the support, particularly, via Grant no. 61229401.

References

1. S. Masuda, K. Kitamura, Y. Okumura, S. Miyatake, H. Tabata and T. Kawai, J. Appl. Phys., 2003, 93(3), 1624.
2. J.-S. Park, H. Kim and I.-D. Kim, J. Electroceram., 2014, 32(2–3), 117–140.
3. H. Q. Chiang, J. F. Wager, R. L. Hoffman, J. Jeong and D. A. Keszler, Appl. Phys. Lett., 2005, 86(1), 013503.
4. V. Subramanian, T. Bakhishev, D. Redinger and S. K. Volkman, J. Disp. Technol., 2009, 5(12), 525–530.
5. R. John, Rep. Prog. Phys., 2006, 69(2), 327.
6. Y. Su, C. Wang, W. Xie, F. Xie, J. Chen, N. Zhao and J. Xu, ACS Appl. Mater. Interfaces, 2011, 3(12), 4662–4667.
7. K. Song, W. Yang, Y. Jung, S. Jeong and J. Moon, J. Mater. Chem., 2012, 22(39), 21265.
8. P. K. Nayak, M. N. Hedhili, D. Cha and H. N. Alshareef, Appl. Phys. Lett., 2013, 103(3), 033518.
9. M. S. Oh, K. Lee, J. H. Song, B. H. Lee, M. M. Sung, D. K. Hwang and S. Im, J. Electrochem. Soc., 2008, 155(12), H1009.
10. Y. B. Yoo, J. H. Park, K. H. Lee, H. W. Lee, K. M. Song, S. J. Lee and H. K. Baik, J. Mater. Chem. C, 2013, 1(8), 1651.
11. A. H. Chen, L. Y. Liang, H. Z. Zhang, Z. M. Liu, X. J. Ye, Z. Yu and H. T. Cao, Electrochem. Solid-State Lett., 2011, 14(2), H88.
12. S. Jeong, Y. G. Ha, J. Moon, A. Facchetti and T. J. Marks, Adv. Mater., 2010, 22(12), 1346–1350.
13. T. P. Trong, T. Miyasako, L. Jinwang, T. H. T. Cam, S. Inoue, E. Tokumitsu and T. Shimoda, IEEE Trans. Electron Devices, 2013, 60(1), 320–326.
14. K. Nomura, H. Ohta, A. Takagi, T. Kamiya, M. Hirano and H. Hosono, Nature, 2004, 432(7016), 488–492.
15. K. H. Lee, J. H. Park, Y. B. Yoo, W. S. Jang, J. Y. Oh, S. S. Chae, K. J. Moon, J. M. Myoung and H. K. Baik, ACS Appl. Mater. Interfaces, 2013, 5(7), 2585–2592.
16. S. Jeong, J.-Y. Lee, M.-H. Ham, K. Song, J. Moon, Y.-H. Seo, B.-H. Ryu and Y. Choi, Superlattices Microstruct., 2013, 59, 21–28.
17. J. Li, L. Niu, Z. Zheng and F. Yan, Adv. Mater., 2014, 26(31), 5239–5273.
18. F. Yan, M. Zhang and J. Li, Adv. Healthcare Mater., 2014, 3(3), 313–331.
19. H. Bong, W. H. Lee, D. Y. Lee, B. J. Kim, J. H. Cho and K. Cho, Appl. Phys. Lett., 2010, 96(19), 192115.
20. Y. Zhao, G. Dong, L. Duan, J. Qiao, D. Zhang, L. Wang and Y. Qiu, RSC Adv., 2012, 2(12), 5307.
21. W. Xu, H. Wang, L. Ye and J. Xu, J. Mater. Chem. C, 2014, 2, 5389–5396.
22. W. Xu, D. Liu, H. Wang, L. Ye, Q. Miao and J.-B. Xu, Appl. Phys. Lett., 2014, 104(17), 173504.
23. M.-G. Kim, H. S. Kim, Y.-G. Ha, J. He, M. G. Kanatzidis, A. Facchetti and T. J. Marks, J. Am. Chem. Soc., 2010, 132(30), 10352–10364.
24. Y. H. Kim, J. S. Heo, T. H. Kim, S. Park, M. H. Yoon, J. Kim, M. S. Oh, G. R. Yi, Y. Y. Noh and S. K. Park, Nature, 2012, 489(7414), 128–132.
25. J. W. Hennek, M. G. Kim, M. G. Kanatzidis, A. Facchetti and T. J. Marks, J. Am. Chem. Soc., 2012, 134(23), 9593–9596.
26. J. W. Hennek, J. Smith, A. Yan, M. G. Kim, W. Zhao, V. P. Dravid, A. Facchetti and T. J. Marks, J. Am. Chem. Soc., 2013, 135(29), 10729–10741.
27. M. Lorenz, H. von Wenckstern and M. Grundmann, Adv. Mater., 2011, 23(45), 5383–5386.
28. K.-C. Liu, J.-R. Tsai, W.-K. Lin, C.-S. Li and J.-N. Chen, Thin Solid Films, 2011, 519(15), 5110–5113.
29. H. Birey, J. Appl. Phys., 1978, 49(5), 2898.
30. L. Herlogsson, X. Crispin, N. D. Robinson, M. Sandberg, O. J. Hagel, G. Gustafsson and M. Berggren, Adv. Mater., 2007, 19(1), 97–101.
31. G. Wu, H. Zhang, J. Zhou, A. Huang and Q. Wan, J. Mater. Chem. C, 2013, 1(36), 5669–5674.
32. H. Zhang, L. Guo and Q. Wan, J. Mater. Chem. C, 2013, 1(15), 2781–2786.
33. J. H. Park, K. Kim, Y. B. Yoo, S. Y. Park, K.-H. Lim, K. H. Lee, H. K. Baik and Y. S. Kim, J. Mater. Chem. C, 2013, 1(43), 7166–7174.
34. Y. Liu, P. Guan, B. Zhang, M. L. Falk and H. E. Katz, Chem. Mater., 2013, 25(19), 3788–3796.
35. See ESI† at for transfer performances including leakage.
36. M.-G. Kim, Kanatzidis, G. Mercouri, Facchetti, Antonio, Marks and J. Tobin, Nat. Mater., 2011, 10(5), 7.
37. P. Barquinha, A. Pimentel, A. Marques, L. Pereira, R. Martins and E. Fortunato, J. Non-Cryst. Solids, 2006, 352(9–20), 1749–1752.
38. J. H. Jeong, H. W. Yang, J.-S. Park, J. K. Jeong, Y.-G. Mo, H. D. Kim, J. Song and C. S. Hwang, Electrochem. Solid-State Lett., 2008, 11(6), H157.
39. X. Zou, G. Fang, L. Yuan, X. Tong and X. Zhao, Microelectron. Reliab., 2010, 50(7), 954–958.
40. S. Masuda, K. Kitamura, Y. Okumura, S. Miyatake, H. Tabata and T. Kawai, J. Appl. Phys., 2003, 93(3), 1624–1630.
41. S. Y. Cho, E. J. Bae, Y. H. Kang, M. Han and C. Lee, J. Mater. Chem. C, 2014, 2, 5695–5703.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c4ra09077b

---