Hybrid cluster precursors of the LaZrO insulator for transistors: lowering the processing temperature

Solution processing of ternary and multinary amorphous metal oxide insulators at processing temperatures below 250 °C remains challenging. Here, we report that the synthesis of a hybrid

cluster structure, where the metal oxide core is coordinated by ligands and the different metal elements are incorporated into one core, is an effective strategy for the low-temperature

processing of the ternary LaZrO insulator. Solvothermal treatment at 160–180 °C facilitated the development of a cluster structure. From the cluster precursor, high-performance insulating

LaZrO films were obtained at 200 °C under the irradiation of ultraviolet light. The analysis data indicate that the solvothermal treatment led to structural unification of the metal oxide

network and facilitated stabilization of the residual organic ingredients in UV annealing, which both contributed to the improved insulating properties of LaZrO. Together with a

solution-processed channel, we have been able to fabricate LaZrO-based transistors at 200 °C. Though the channel material has not been optimized, the transistor have showed a low gate

leakage current around 10 pA at an operating voltage of 15 V, an on/off ratio of near 106, a field-effect saturation mobility of 0.37 cm2 V−1 s−1, a subthreshold swing factor of 0.61 V

decade−1.

Low-temperature solution processing of oxide thin-film transistors (TFTs) has attracted significant interest because it can allow the use of low-cost, flexible, and transparent plastic

substrates in conjunction with facile solution deposition or printing techniques. Hitherto, several low-temperature methodologies have been developed, namely the “sol-gel on chip” process1,

solution combustion processing2,3,4, ultraviolet (UV) light-assisted (photochemical) annealing5,6,7, as well as the very recently reported methods employing redox reactions8 or aqueous

solutions9. Using these techniques, processing at 250 °C or lower temperatures is possible and processing at as low as 150 °C through UV annealing has been reported.5–67. These advances are

significant, but the processable materials, especially insulators, are limited. Most of these processes have been applied to the semiconducting channel materials in transistors, and only the

UV-annealing and aqueous solution techniques have been applied to alumina, 6, 7, 9 zirconia, 7 and hafnia 7 insulators. These insulators are binary oxides, among which alumina can be doped

with zirconium at an ultimate concentration of a few percent. For practical applications, it is necessary to expand the types of insulating materials processable at low temperatures. In

particular, low-temperature processing routes for ternary or multinary insulating metal oxides remain unexploited and should also be developed. One practical reason is that incorporation of

more elements allows tailoring the properties in a wider range and may lead to finding new functionalities. Compared to binary oxides, ternary and multinary oxides may have more stable

amorphous structures because the arrangement of three or more types of atoms into crystals would be more difficult, as in our LaZrO. Amorphous structures are desirable for low leakage

currents, smooth surfaces, and sharp insulator–channel interfaces, all of which lead to high transistor performances.

For ternary and multinary oxides, the preferential decomposition of one metal-precursor compound over another can result in compositional segregation, even though the compositional

distribution is on the molecular scale in the precursor. This imposes additional difficulties for low-temperature processing. The length scale of compositional heterogeneity can be up to

nanometers, and high temperatures are needed to decrease it10. Compositional heterogeneity significantly decreases the insulating properties of the material11.

We have investigated the solution-phase synthesis of oxide structures containing two or more metals. In our previous studies, we observed that hybrid cluster structures are typically formed

in solution precursors of oxides such as indium oxide12 and ruthenium oxide13. The hybrid clusters have an inorganic core (M-O-M) chelated by organic ligands. For instance, we observed

typical cores having seven indium atoms and three ruthenium atoms, respectively, in the above examples. Here, we propose synthesizing such a cluster core containing two or more metal

elements with a structure similar to that of the desired final oxide, and then, using the core as a building block for film deposition. In this method, the preferential decomposition of one

metal compound over the others, which causes compositional segregation, is not expected to occur because the different metal elements have already been combined into one core and

thermodynamically stabilized. Hence, the decomposition and densification are similar to those occurring in a binary metal oxide system. Compared with nanoparticles, such clusters are

preferred because nanoparticles are relatively large and only compact into porous structures unless sintered at high temperatures. The core of our proposed cluster typically has less than or

around ten metal atoms and is expected to be less than or around one nanometer in size. The previously reported strategies can be used to remove the ligands at low temperatures.

We studied this method for the ternary insulating material LaZrO, which has a relative dielectric constant (~20 or higher, depending on processing, impurities, etc.) around three times

higher than that of alumina, and has resulted in high TFT performance when used as an amorphous gate insulator in our studies12,14,15. In an organic acid solution, a cluster structure

coordinated by carboxylate ligands with six zirconium atoms (Zr6) in the oxide core has been reported16,17. We further synthesized LaZrO clusters by adding La and achieved uniform and stable

clusters under solvothermal conditions. The well-developed clusters obtained after solvothermal treatment can form high-quality insulating films by annealing at 200 °C, with the assistance

of UV light irradiation. The structure of the clusters, as well as their uniformity and stability, not only influences the insulating properties of low-temperature deposited films but also

significantly affects the structures and properties of films deposited at high temperatures, as presented in our previous paper18. In that paper18, we have reported detailed structural

analyses of solutions, gels, and solids, and show how the precursor structure determines the final solid structure after annealing at high temperature. These results also indicate the

importance of detailed investigation into precursor structure and its optimization, which are rarely performed. In the current paper, we focus on the processing of LaZrO films at low

temperatures and demonstrate its application as a gate insulator in transistors fabricated at 200 °C.

In a typical experiment, lanthanum(III) acetate 1.5-hydrate (0.686 g, 99.99%, Kanto Kagaku) and zirconium(IV) butoxide solution (0.959 g, 80 wt. % in 1-butanol, Aldrich) were each dissolved

in appropriate amounts of propionic acid (>99.0%, Kanto Chemical) in capped glass vials with magnetic stirring for 30 min on a hotplate set at 110 °C to produce 0.2 mol/kg La and Zr

solutions (10 g of each). The Zr compound was handled in a dry nitrogen-filled glovebox before being sealed in the vial. After cooling to room temperature, the two solutions were mixed to

obtain LaZrO precursor solutions with La/Zr molar ratio of 3/7 or 1/1. For solvothermal treatment, 10 g of LaZrO precursor solution was sealed in an autoclave with a 50 ml PTFE inner

container (HUT-50, San-Ai Kagaku Co. Ltd.) and heated at 160–180 °C on an autoclave heater (RDV-TMS-50, San-Ai Kagaku Co. Ltd.) for 2–5 h with magnetic stirring. The selection of the

reagents and solvents is decided based on our previous study that this system formed La-Zr clusters that developed toward structural unification after solvothermal treatment. In our

preliminary experiments, we had used nitrate reagents in 2-methoxyethanol for LaZrO and did not find such a feature, and the properties of the resulting transistors were less stable.

To prepare the solution for the semiconducting indium oxide (InO) channel material, indium(III) nitrate hydrate (0.355 g, 99.999%, Aldrich) was dissolved in deionized water (5 g, >99.0%,

Kanto Chemical) in a capped glass vial with stirring at room temperature for 1 h.

All solutions used for film deposition were filtered through a 0.2-μm-pore filter. Pt(200 nm)/SiO2(200 nm)/Si substrates, typically 2 cm × 2 cm in size, were cleaned using oxygen plasma

before deposition. The LaZrO precursor solution was spin-coated onto a substrate at 2000 rpm for 25 s, dried at 100 °C or 150 °C on a metallic hotplate for 5 min, and then annealed at a

substrate temperature of 150–250 °C for a typical time of 10 min under UV irradiation (wavelengths 184.9 (~10%) nm and 253.7 nm (~90%), generated by a low-pressure mercury lamp, with an

intensity of ~ 10 mW cm−2) in an oxygen flow (5 L/min × 2 inlets) using the UV-300H system (Samco Inc.), with O3 produced by an O3 generator in the system. The spacing between the UV lamp

and the samples was around 1 cm. To study the effect of atmosphere, the UV-annealing was also performed in a nitrogen flow (5 L/min × 2 inlets) without turning on the O3 generator under

otherwise same conditions. The spin-coating and annealing cycles were performed up to 5 times to achieve a film thickness of around 120 nm.

For comparison, samples only thermally annealed at 500 °C and 600 °C were prepared. These samples were dried at 250 °C on a metallic hotplate for 5 min after spin-coating. The spin-coating

and drying procedure was performed 5 times for a final film thickness of around 120 nm. The samples were then pre-annealed at 400 °C for 5 min on a ceramic hotplate in air, followed by

annealing at 500 °C or 600 °C for 20 min under an oxygen flow in a rapid thermal annealing furnace.

For the measurement of insulating properties (current–voltage and dielectric constant–frequency relations), Pt top electrodes (100 nm thick and 200–500 μm in diameter) were deposited through

a metal mask using radio-frequency (RF) plasma sputtering at room temperature to form a capacitor structure. Post-annealing was performed at 200 °C for 20 min in air.

TFTs were fabricated using a conventional photolithography patterning process. First, a Pt gate-electrode pattern (100 nm thick) was prepared using a RF plasma sputtering (at room

temperature) and lift-off process on a SiO2/Si substrate. A RF plasma-sputtered Ti adherence layer (10 nm thick) was used between the Pt and the SiO2. Second, the LaZrO insulator was

deposited as described above. Third, the InO channel layer was deposited using spin-coating (3000 rpm, 20 s) followed by drying (100 °C, 5 min) and UV-annealing at 200 °C for 10 min under a

nitrogen flow (UV-300H system, Samco Inc.) with the O3 generator switched off. The use of a nitrogen atmosphere instead of an ozone and oxygen atmosphere was based on our preliminary

experiments on a transistor with an InO(channel)/SiO2(insulator)/p++Si(gate) structure, where nitrogen gave rise to higher properties. Fourth, Pt (100 nm thick)/ITO (5% Sn-doped indium

oxide, ~60 nm) source and drain electrodes were prepared using the sputtering and lift-off process. Fifth, the InO channel was patterned by wet etching through a photolithographically

deposited resist mask. The resist mask was thereafter removed via dissolving in a solvent. Finally, the TFTs were post-annealed at 200 °C for 2 h on a hotplate in air.

UV–visible absorption measurements for the solution samples were performed using a V-630 spectrophotometer (JASCO Corporation) with a 10 mm-path-length quartz cell. Film thicknesses were

evaluated through an ellipsometry (Sopra GES-5E Ellipsometer, SEMILAB Japan K.K.). Film densities were measured using the X-ray reflectivity (XRR) method (X’Pert PRO MRD, PANalytical) for

the LZO/Pt/SiO2/Si samples. The XRR data fitting was performed using the analysis software of the XRR system. Ideal pure oxide compositions of La3Zr7O18.5 and La5Zr5O17.5 were used in the

fitting. The thickness values from ellipsometry analysis were input as initial parameters. Both density and thickness values were fitted. The resulted thickness values were similar to the

ones from ellipsometry analysis. The uncertainty of the obtained density values was around ±0.05 g cm−3. A typical resulted plot of the fitting is presented in Figure S1. X-ray photoelectron

spectroscopy (XPS) analysis was performed using a Kratos AXIS-ULTRA DLD system (Shimadzu). The surface morphologies were investigated through an atomic force microscopy (AFM; NanoNavi,

SII). Elemental compositions of the thin films were obtained from combinational Rutherford backscattering spectrometry, nuclear reaction analysis, and hydrogen forward scattering

spectrometry on Pelletron 3SDH (National Electrostatics Corp.) at Toray Research Center, Inc., Japan, with estimated precisions of ±0.5 atom% for C and H, ±1 atom% for La and Zr, and ±4

atom% for O. Impedance measurements for Pt/LaZrO/Pt capacitor samples were performed using SI 1260 Impedance/Gain-Phase Analyzer (Solartron Analytical). The current–voltage characteristics

of these capacitor samples and the TFT performance were measured using a 4155C Semiconductor Parameter Analyzer (Agilent).

The details have been presented in our previous paper18. A simple description is given here for the convenience of readers. We dissolved lanthanum acetate and zirconium butoxide in propionic

acid below 110 °C. In the absence of La, this system forms the Zr6 cluster16. We used La/Zr atomic ratios of 1/1 (LZ55) and 3/7 (LZ37) based on our observation that Zr-rich compositions led

to higher TFT performance. A simple and convincing evidence for the cluster formation is that the thermal decomposition temperature of the LaZrO precursor (two main exothermic peaks at 327 

°C and 345 °C) is lower than those of both Zr-only (350 °C) and La-only (365 °C) precursors, indicating that the Zr and La components were not simply mechanically mixed.

However, these LaZrO precursors were found to result in films with poor insulating properties when annealed at both low (see below) and high (see ref.18) temperatures. This is attributed to

a low degree of cluster development under the above synthesis conditions. Therefore, we further heated the solutions under solvothermal conditions in an autoclave (AC) at temperatures

(160–180 °C) higher than the boiling temperature of the solvent (propionic acid, b.p. 141.2 °C) in order to facilitate the development of the clusters. This led to clusters of improved

uniformity and stability, as indicated by thermal analysis18. The higher uniformity was also supported by synchrotron X-ray diffraction measurements, which showed that solvothermal treatment

of the precursor solution led to the disappearance of a second phase that was otherwise present in the samples annealed at 500 °C.

As a result of solvothermal treatment, the appearance of the precursor solution changed from colorless to slightly yellowish, indicating that the light absorption range widened into the

visible region. The color became deeper with increasing duration and temperatures of the solvothermal treatment. The UV–visible light absorption of the solutions was measured and the results

(Fig. 1 and Figure S2) showed that the absorption range was significantly widened to include longer wavelengths in the visible region after the solvothermal treatment. The absorption

reached the measurement limit of the detector at 252 nm before solvothermal treatment (without AC or w/o AC), while the measurement limit was reached at 260 nm, 300 nm, 348 nm, and 370 nm

after the treatment at 160 °C for 2 h (AC160-2h), 180 °C for 2 h (AC180-2h), 180 °C for 5 h (AC180-5h), and 180 °C for 12 h (AC180-12h), respectively. Such changes in absorption were

considered to be a result of the structural reorganization of the cluster cores. Hence, the solvothermally treated solutions were found to have a highly enhanced UV absorption ability, which

may facilitate the decomposition of their organic ligands under UV irradiation.

The UV light we applied is in the deep UV range that has energies (647 and 472 in kJ mol−1 for UV light at wavelengths of 184.9 nm and 253.7 nm, respectively) high enough to decompose

typical bonds of impurities in the precursors (e.g., energies in kJ mol−1 for C-C is 347.7, C-O 351.5, O-H 462.8). Thus, deep UV-annealing (without ozone) has been applied for the

low-temperature solution deposition of films of aluminum oxide and zirconium oxide from nitrate or organic precursors.5-,67 On the other hand, in the presence of ozone and O2, active atomic

oxygen (O), with strong oxidation ability, is produced according to the Chapman mechanism (O2 + UV light