The Effect of Anodization Parameters on the Aluminum Oxide Dielectric Layer of Thin-Film Transistors

Summary

Anodization parameters for growth of the aluminum-oxide dielectric layer of zinc-oxide thin-film transistors (TFTs) are varied to determine the effects on the electrical parameter responses. Analysis of variance (ANOVA) is applied to a Plackett-Burman design of experiments (DOE) to determine the manufacturing conditions that result in optimized device performance.

Abstract

Aluminum-oxide (Al₂O₃) is a low cost, easily processable and high dielectric constant insulating material that is particularly appropriate for use as the dielectric layer of thin-film transistors (TFTs). Growth of aluminum-oxide layers from anodization of metallic aluminum films is greatly advantageous when compared to sophisticated processes such as atomic layer deposition (ALD) or deposition methods that demand relatively high temperatures (above 300 °C) such as aqueous combustion or spray-pyrolysis. However, the electrical properties of the transistors are highly dependent on the presence of defects and localized states at the semiconductor/dielectric interface, which are strongly affected by the manufacturing parameters of the anodized dielectric layer. To determine how several fabrication parameters influence the device performance without performing all possible combination of factors, we used a reduced factorial analysis based on a Plackett-Burman design of experiments (DOE). The choice of this DOE permits the use of only 12 experimental runs of combinations of factors (instead of all 256 possibilities) to obtain the optimized device performance. The ranking of the factors by the effect on device responses such as the TFT mobility is possible by applying analysis of variance (ANOVA) to the obtained results.

Introduction

Flexible, printed and large area electronics represent an emerging market that is expected to attract billions of dollars in investments in upcoming years. To achieve the hardware requirements for the new generation of smartphones, flat panel displays and internet-of-things (IoT) devices, there is a huge demand for materials that are lightweight, flexible and with optical transmittance in the visible spectrum without sacrificing speed and high performance. A key point is to find alternatives to amorphous silicon (a-Si) as the active material of the thin-film transistors (TFTs) used in the drive circuits of most of the current active-matrix displays (AMDs). a-Si has low compatibility to flexible and transparent substrates, presents limitations to large-area processing, and has a carrier mobility of about 1 cm²∙V^-1∙s^-1, which cannot meet the needs of resolution and refresh rate for next generation displays. Semiconducting metal oxides (SMOs) such as zinc oxide (ZnO)^1,2,3, indium zinc oxide (IZO)^4,5 and indium gallium zinc oxide (IGZO)^6,7 are good candidates to replace a-Si as the active layer of TFTs because they are highly transparent in the visible spectrum, are compatible to flexible substrates and large area deposition and can achieve mobilities as high as 80 cm²∙V^-1∙s^-1. Moreover, SMOs can be processed in a variety of methods: RF sputtering⁶ , pulsed laser deposition (PLD)⁸, chemical vapor deposition (CVD)⁹, atomic layer deposition (ALD)¹⁰, spin-coating¹¹, ink-jet printing¹² and spray-pyrolysis¹³.

However, few challenges such as the control of intrinsic defects, air/UV stimulated instabilities and formation of semiconductor/dielectric interface localized states still need to be overcome to enable the large-scale manufacturing of circuits comprising SMO-based TFTs. Among the desired characteristics of high performance TFTs, one can mention the low power consumption, low operation voltage, low gate leakage current, threshold voltage stability and wideband frequency operation, which are extremely dependent on the gate dielectrics (and the semiconductor/insulator interface as well). In this sense, high-κ dielectric materials^14,15,16 are particularly interesting since they provide large values of capacitance per unit area and low leakage currents using relatively thin films. Aluminum oxide (Al₂O₃) is a promising material for the TFT dielectric layer since it presents a high dielectric constant (from 8 up to 12), high dielectric strength, high electrical resistivity, high thermal stability and can be processed as extremely thin and uniform films by several different deposition/growth techniques^{15,17,18,19,20,21}. Additionally, aluminum is the third most abundant element in the Earth’s crust, what means that it is easily available and relatively cheap compared to other elements used to produce high-k dielectrics.

Although deposition/growth of Al₂O₃ thin (below 100 nm) films can be successfully attained by techniques such as RF magnetron sputtering, chemical vapor deposition (CVD), atomic layer deposition (ALD), the growth by anodization of a thin metallic Al layer^{17,18,21,22,23,24,25,26} is particularly interesting for flexible electronics owing to its simplicity, low cost, low temperature, and film thickness control in nanometric scale. Besides, anodization has a great potential for roll-to-roll (R2R) processing, which can be easily adapted from processing techniques already being used at industrial level, permitting quick manufacturing upscaling.

Al₂O₃ growth by anodization of metallic Al can be described by the following equations

2Al + 3 / 2 0₂ → Al₂O₃      (1)

2Al + 3H₂O → Al₂O₃ + 3H₂      (2)

where the oxygen is provided by the dissolved oxygen in the electrolyte solution or by the adsorbed molecules at the film surface, whereas the water molecules are promptly available from the electrolyte solution. The anodized film roughness (which affects the TFT mobility due to carrier scattering at the semiconductor/dielectric interface) and the density of localized states at the semiconductor/dielectric interface (which affects the TFT threshold voltage and electrical hysteresis) are strongly dependent on anodization process parameters, to name a few: the water content, the temperature and the pH of the electrolyte^24,27. Other factors related to the Al layer deposition (like evaporation rate and metal thickness) or to post-anodization processes (like annealing) can also influence the electrical performance of fabricated TFTs. The effect of these multiple factors on response parameters can be studied by varying each factor individually while keeping all other factors constant, which is an extremely time-consuming and inefficient task. Design of experiments (DOE), on the other hand, is a statistical method based on the simultaneous variation of multiple parameters, which permits the identification of the most significant factors on a system/device performance response by using a relatively reduced number of experiments²⁸.

Recently, we have used multivariate analysis based on a Plackett-Burman²⁹ DOE to analyze the effects of Al₂O₃ anodization parameters on the performance of sputtered ZnO TFTs¹⁸. The results were used to find the most significant factors for several different response parameters and applied to the optimization of the device performance changing only parameters related to the anodization process of the dielectric layer.

The current work presents the whole protocol for manufacturing TFTs using anodized Al₂O₃ films as gate dielectrics, as well a detailed description for the study of the influence of the multiple anodization parameters on the device electrical performance by using a Plackett-Burman DOE. The significance of the effects on TFT response parameters such as the carrier mobility is determined by performing analysis of variance (ANOVA) to the results obtained from the experiments.

Protocol

The protocol described in the present work is separated into: i) preparation of the electrolytic solution for anodization; ii) substrate cleaning and preparation; iii) anodization process; iv) deposition of the TFT active layer and drain/source electrodes; v) TFT electrical characterization and analysis and vi) application of ANOVA to determine the significance of the manufacturing factors in the TFT mobility.

1. Preparation of the electrolytic solution for anodization

1. Perform all the procedures of the protocol inside a cleanroom or a laminar flow cabinet, to avoid dust or contaminants during the sample preparation.
2. Prepare two solutions of tartaric acid (0.1 M) in different water/ethylene glycol volume ratios (16% and 30%), which will be used as the anodization electrolytic solution. Use the water content in the electrolytic solution as fabrication parameter of the anodized layer.
3. In a 150 mL beaker, dissolve 1.5 g of tartaric acid into 16 mL of deionized water and 84 mL of ethylene glycol to obtain a 16% water electrolyte stock solution. For a 30% water electrolyte stock solution, use 1.5 g of tartaric acid, 30 mL of deionized water and 70 mL of ethylene glycol. Stir both solutions using a magnetic bar for 30 min.
4. Separate about 10-20 mL of ammonium hydroxide (NH₄OH) solution (as purchased, 28 – 30% NH₃ in volume) in a 20 mL beaker to make the rough adjustment of the pH of the electrolytic solution.
5. Prepare 80 mL of a diluted solution (about 2% in volume) from the original NH₄OH solution to make the fine control of the pH of the electrolytic solution.
6. Separate the electrolyte solution into a 150 mL beaker to adjust the pH of the solution.
7. Measure the pH of the electrolytic solution using a bench pH meter. Start pipetting the more concentrated NH₄OH until the pH is close to the desired pH (5 or 6).
8. Pipette the more diluted NH₄OH solution into the electrolytic solution until the pH is set in the desired value. Prepare the electrolyte solutions at pH values of 5 and 6 to study the effect on the anodization process.

2. Substrate cleaning and preparation

1. Use 20 mm x 25 mm glass slides (1.1 mm thick) as substrates.
2. Sonicate the glass slides in a heated (60 °C) alkaline detergent solution (5% in deionized water) for 15 min. Rinse abundantly in deionized water and dry in clean dry air (CDA) or nitrogen.
3. Sonicate the glass slides in acetone (ACS reagent grade or superior) for 5 min. Dry the substrates in CDA or nitrogen.
4. Sonicate the glass slides in isopropanol (ACS reagent grade or superior) for 5 min. Dry the substrates in CDA or nitrogen.
5. Insert the substrates into the chamber of a plasma cleaner, close the lid and evacuate the chamber using a vacuum pump.
6. When the vacuum is achieved, switch on the RF generator at medium power (10.5 W) for 5 min. After plasma cleaning, the substrates are ready for aluminum gate deposition.

3. Aluminum gate electrode evaporation

1. Insert the glass slides into mechanical shadow masks to deposit an aluminum stripe of 25 x 3 mm. This aluminum stripe will be used as the TFT gate electrode and the aluminum oxide layer formed by anodization will be the TFT dielectric layer. Example of shadow mask design for the gate electrode is presented in the supplementary files.
2. Place the substrates with the shadow mask inside the chamber of the thermal evaporating chamber for the aluminum layer deposition. Shut the chamber. Start the chamber evacuation procedure. Wait until the chamber pressure is below 2.0 x 10^-6 mbar to start the thermal evaporation.
3. Deposit the aluminum layer. Use two different thicknesses (60 nm and 200 nm) to evaluate the effect on the dielectric layer. Use two different evaporation rates 5 Å/s and 15 Å/s to study the influence of the Al evaporation rate.
4. Remove the samples from the evaporation chamber after aluminum evaporation.
5. Remove the glass slides with the aluminum stripe from the masks and check if the aluminum layer was properly deposited. The electrode is ready for the anodization process.

4. Anodization process of the aluminum layer

1. Attach two alligator clip connectors in a plastic lid that fits on top of the beaker. This lid can be 3-D printed.
2. Connect one of the clip connectors to the aluminum strip of a glass slide and the other to a gold-plated stainless-steel sheet (0.8 mm thick, 20 x 25 mm). Face both electrodes towards each other with a separating distance of about 2 cm.
3. Use approximately 150 mL of the electrolytic solution (after pH adjustment) in a 150 mL beaker. Use a small magnetic bar to stir the solution during the anodization procedure.
4. Place the beaker on top of a magnetic stirrer with heating. Adjust the temperature to the desired value (40 °C and 60 °C were used in the current paper).
5. Immerse the electrodes in the electrolytic solution by covering the beaker with the plastic lid attached to the clip connectors.
6. Connect the aluminum electrode to the positive output and the golden-plated stainless-steel electrode to the negative output of a current/voltage source and measuring unit (SMU).
7. Calculate the submerged area of the aluminum electrode and apply a constant current equivalent to the desired current density (we used two values 0.45 mA/cm² and 0.65 mA/cm²) and monitor the linear increase of the voltage until the pre-set final value (we used V_F = 30 V and V_F = 40 V).
8. After the final voltage is achieved, switch the SMU from the current source to the voltage source and apply a constant voltage (equal to the final voltage) during a time long enough to the current decrease next to zero (about 5 min). Use a script in Python 2.7 to automatically control the SMU during anodization process. A copy of this script is available in the supplementary files section.
9. Remove the electrodes from the electrolytic solution, rinse abundantly with deionized water, dry with CDA or nitrogen and store the Al/Al₂O₃ glass substrates until use.
10. To observe the effect of annealing on the dielectric layer, anneal the substrates in an oven at 150 °C for 1 h.

5. Deposition of the ZnO Active layer

1. Insert the substrates with the anodized aluminum oxide layer in appropriate mechanical shadow masks for active layer deposition.
2. Place the substrates with the masks inside the chamber of the sputtering system. Use a ZnO (99.9%) sputtering target. Close the chamber and start the evacuating procedure.
3. Adjust the Ar pressure to 1.2 x 10^-2 Torr and the RF power to 75 W and start the ZnO deposition. Control the deposition rate at 0.5 Å/s. Stop the ZnO deposition when the active layer thickness achieves 40 nm.
4. Open the chamber and remove the samples.

6. Drain and source electrodes deposition

1. Insert the samples with the sputtered ZnO layer in appropriate mechanical shadow masks for TFT source/drain electrodes deposition. An appropriate drain and source electrode spacing is 100 µm, with a lateral overlapping of 5 mm. A template of the drain/source mask design is supplied with the supplementary files. In such a configuration, notice that both drain and source electrodes are identical and can be interchangeable without change on the device operation.
2. Place the samples attached to the shadow masks inside the chamber of the thermal evaporating system and start the procedure for aluminum evaporation.
3. Deposit a 100 nm Al layer at a deposition rate of 5 Å/s to obtain the drain/source electrodes on top of the active layer, finishing the TFT manufacture procedure.
4. Remove the TFTs from the evaporation chamber, check the quality of the deposited electrodes and store them protected from light until use.

7. TFT electrical characterization

1. Place the TFTs on a semiconductor probe station or custom sample holder. Connect the gate, drain and source electrodes using spring-probe connectors for electrical contacts.
2. Connect the probes to a two-channel source-measuring unit (recommended Keithley 2612B or similar). Connect the gate electrode to the “high” output/input of channel 1 and the drain (or source) electrode to the “high” output/input of channel 2. Short the “low” output/input terminals of both channels and the source (or drain) electrode, which remained disconnected.
3. Obtain characteristic TFT curves. Obtain the output curve by applying constant voltage bias at the gate (V_g) and sweeping the drain-source voltage (V_DS) and recording the drain-source current (I_DS). Obtain the transfer curve by recording the drain-source current (I_DS) while sweeping the gate voltage (V_g) and maintaining the drain-source voltage (V_DS) constant.
4. Plot the square root of the drain current versus the gate voltage ((I_DS)^1/2 vs. V_g) and obtain the carrier mobility in the saturation regime (µ_s) from the curve slope and the threshold voltage from the x-axis intercept of the linear portion of the curve.
5. If wanted, determine other performance parameters from the transistors curves as described elsewhere¹⁸.

8. ANOVA and influence of design factors on device performance

1. Use a software to set a design of experiment (DOE) based on a Placket-Burman matrix considering 8 fabrication factors. We used Chemoface, which is a free, user-friendly software developed by Federal University of Lavras (UFLA), Brazil³⁰.
2. Use as factors the anodization parameters: i) the thickness of the Al layer; ii) the Al evaporation rate; iii) the water content in the electrolytic solution; iv) the temperature of the electrolyte; v) the pH of the electrolytic solution; vi) the current density during anodization; vii) the annealing temperature and viii) the final voltage of anodization.
3. For each factor, consider two levels, as given by Table 1.
4. Assemble the Plackett-Burman design table aided by the DOE software as given by Table 2.
5. Prepare the TFTs varying the fabrication parameter according to the 12 generated “runs” from Table 2. Each run provides a representative variation of the fabrication factors without the need to perform all 256 (2⁸) possible combinations for a two-level, eight-parameters experiment.
6. Feed the DOE table from the software with the performance data from TFT characterization (e.g., TFT mobility in saturation) following the manufacturing directions of each run.
7. Add as many replicates from different devices using the same fabrication factors to increase the number of degrees of freedom for the analysis.
8. Perform ANOVA from the data and analyze the output to determine which anodizing parameters influence most the TFT performance.

Results

Eight different aluminum oxide layer manufacture parameters were used as the fabrication factors which we used to analyze the influence on the TFT performance. These factors are enumerated in Table 1, where the corresponding “low” (-1) and “high” (+1) values for the two-level factorial DOE are presented.

For simplicity, each manufacturing factor was named by a capital letter (A, B, C, etc.) and the corresponding “low” or “high” l...

Discussion

The anodization process used to obtain the dielectric has a strong influence on the performance of the TFTs fabricated, keeping constant all geometrical parameters and the fabrication parameters of the active. For the TFT mobility, which is one of the most important performance parameters for TFTs, it can vary more than 2 orders of magnitude by changing the manufacturing factors in the range given by Table I. Therefore, the careful control of the anodization parameters is of great importance when fabricating devices comp...

Materials

Name	Company	Catalog Number	Comments
Acetone	LabSynth	A1017	ACS reagent grade
Aluminum (Al) Wire Evaporation	Kurt J. Lesker Company	EVMAL40060	1.5 mm (0.060") Dia.; 1lb; 99.99%
Ammonium hydroxide solution	Sigma Aldrich	338818	ACS reagent, 28.0-30.0% NH3 basis
Chemoface - Software to set a design of experiment (DOE)	Federal University of Lavras (UFLA), Brazil		Free software developed by Federal University of Lavras (UFLA), Brazil - http://www.ufla.br/chemoface/
Cleaning detergent	Sigma Aldrich	Alconox	Alkaline detergent for substrate cleaning
Ethylene glycol	Sigma Aldrich	102466	ReagentPlus, ≥99%
Isopropanol	LabSynth	A1078	ACS reagent grade
Glass substrates	Sigma Aldrich	CLS294775X50	Corning microscope slides, plain
L-(+)-Tartaric acid	Sigma Aldrich	T109	≥99.5%
Mechanical shadow mask for deposition of the sputtered ZnO active layer	Lasertools, Brazil	custom mask	10 mm x 10 mm square.
Mechanical shadow mask for TFT gate electrode	Lasertools, Brazil	custom mask	25 mm long stripe, 3 mm wide.
Mechanical shadow mask for TFT source/drain electrodes	Lasertools, Brazil	custom mask	100 µm stripes, separated by 100 µm gap, overlapping of 5 mm
Plasma cleaner	MTI	PDC-32G	Campact plasma cleaner with vacuum pump
Sputter coating system	HHV	Auto 500	RF sputtering system with thickness and deposition rate control
Stiring plate	Sun Valley	MS300	Stiring plate with heating control
Thermal evaporator	HHV	Auto 306	it has a high precision sensor for measure the thickness and rate of deposition of thin films
Two-channel source-measuring unit	Keithley	2410	Keithley model 2410 or similar/for anodization process
Two-channel source-measuring unit	Keithley	2612B	Dual channel source-measure unit (SMU) for TFT measurements
Ultrasonic bath	Soni-tech	Soni-top 402A	Ultrasonic bath with heating control
Zinc Oxide (ZnO) Sputtering Targets	Kurt J. Lesker Company	EJTZNOX304A3	3.0" Dia. x 0.250" Thick; 99.9%