Area-selective atomic layer deposition of Al2O3 using inkjet-printed inhibition patterns and lift-off process

Area-selective atomic layer deposition (AS-ALD) has been studied as an alternative method for metal oxide ALD film patterning in the microelectronics industry. To perform AS-ALD, area-deactivation or -activation processes should be implemented in advance using selective surface modification through a photography-based lift-off process and printing techniques. This study introduces a novel approach for Al2O3 AS-ALD using inkjet-printed inhibition patterns. ALD-inhibition patterns with two-layer structures of fluorocarbon (FC) thin film and photoresist (PR) were patterned by inkjet printing. Low surface energy FC thin films were used to block the Al2O3 nucleation and growth during the ALD process, and PR was patterned to easily remove ALD-inhibition patterns using a lift-off process. To demonstrate AS-ALD, an Al2O3 thin film approximately 10 nm thick was deposited via an in-house built system with a multiple-slit gas source. The proposed AS-ALD method was evaluated for topological analysis using atomic force microscopy and surface composition analysis using a time of flight secondary ion mass spectrometry after Al2O3 deposition and a lift-off process. Finally, a 6 nm thick Al2O3 film was selectively patterned by the lift-off process using an inkjet-printed 1.28 μm thick FC-covered PR inhibition pattern.


Introduction
Atomic layer deposition (ALD) as a thin film deposition process is implemented by the alternate pulsing of precursor gases and surface reactions between vapors and reactive sites on the substrate [1]. The formation of films is restricted to one atomic layer at a time because surface reactions in the ALD process occur only at reactive sites [2]. Therefore, ALD enables the deposition of films with high material quality, good uniformity, high conformality, and accurate thickness control [3,4].
In the microelectronics industry, area-selective atomic layer deposition (AS-ALD) has recently attracted significant attention as an alternative method for metal oxide ALD [5]. Since ALD depends on surface chemistry, AS-ALD can be achieved by selective chemical modification of the substrate surface, which is in turn achieved by a photography-based lift-off process as well as printing techniques for area-deactivation or -activation. Many research groups have studied the AS-ALD process to achieve metal oxide film patterns of Al 2 O 3 , HfO 2 , TiO 2 , and ZnO [2,[4][5][6][7][8][9]. Self-assembled monolayers (SAMs) such as octadecylphosphonic acid (ODPA) and octadecyltrichlorosilane (ODTS) are used as ALD-blocking layers for selective prevention of nucleation and growth during ALD processes [4,6]. Some of the reported AS-ALD processes use unreactive and thermally stable polymer thin films made of poly (methyl methacrylate) (PMMA) and poly (vinyl pyrrolidone) (PVP) [7,9] as an ALDblocking layer. Al 2 O 3 thin film is widely used as a functional material in the fabrication of various electronic and optoelectronic devices due to its high transparency and refractive index, thermal conductivity, good adhesion to various surfaces, thermal and chemical stability, and high dielectric constant [10][11][12]. Some of its important applications include as gas diffusion barrier layer in the encapsulation of organic light-emitting diode devices [13,14], antireflection coating in solar cells [15], and gate dielectric film in thin-film transistors [16,17]. However, very few studies have reported AS-ALD of Al 2 O 3 using SAMs and polymer thin films or patterning processes constructed using O 2 plasma and acid chemical etching on the metal oxide-coated surface [6,18]. Seo [18]. It has been reported that selective patterning of Al 2 O 3 grown using trimethylaluminum (TMA) and H 2 O for the Si surface is nearly impossible, even if hard mechanical scribing is used to remove Al 2 O 3 from blocking layers such as PMMA, PVP, and plasma-polymerized fluorocarbon (FC) films [19][20][21][22]. Most studies have used AlCl 3 [7,20] or dimethylaluminum isopropoxide [9,23] as the aluminum precursor to demonstrate AS-ALD of Al 2 O 3 without additional chemical etching of metal oxide surface.
Photolithography-based lift-off processes or microcontact printing are generally used to pattern SAMs for AS-ALD [8,[23][24][25][26]. However, since a typical SAM is ∼ 2 nm thick, it is quite challenging to fabricate features thicker than a few nanometers [5]. The increase in film thickness over the height of SAM during ALD leads to lateral film overgrowth [3,5]. Furthermore, it is important to form a defect-free SAM to effectively block the growth of films during ALD [3]. It usually takes more than 24 h to obtain a well-packed SAM by immersion of the sample in solution or by vapor deposition [27]. In the case of polymer films, spin coating and photolithography-based lift-off processes can be used to easily control the thickness of ALD-inhibition patterns and obtain thicker polymer films [7,9,28]. The inkjet printing process was introduced as an alternative patterning method to pattern polymer thin films for AS-ALD [9,20]. This process was transformed into a sophisticated device manufacturing method for low-cost, large-area, and flexible electronics [29]. One of the main advantages of this process is the ability to position the material on planar and non-planar substrates using computer control to produce structures without the need for photomasks [30]. The inkjet printing process can control the thickness of the target materials at the desired position using an overprinting technique. By increasing the number of the inkjet head, various materials can be printed simultaneously and multi-layered patterns can be realized by layer-by-layer printing. The inkjet printing process is less time-consuming to form mask patterns for AS-ALD when compared to the film patterning methods of both SAMs and polymers.
To date, there has been no demonstration of AS-ALD using Al 2 O 3 grown by TMA and H 2 O using inkjet-printed blocking layers on the Si surface. In this study, the inkjet printing process was utilized to form ALD-inhibition patterns with two-layered structures for the AS-ALD process. The ALD-inhibition pattern printing process was developed by using two techniques: overprinting for thickness control and multilayer printing, which are unique advantages of inkjet printing. Al 2 O 3 was chosen as a model metal oxide for selective ALD due to its enormous potential in various fields. FC thin films were inkjet-printed to inhibit Al 2 O 3 nucleation and growth during the ALD process. A thicker photoresist (PR) was inkjet-printed to allow easy removal of ALD-inhibition patterns using a lift-off process. FC, PR, and FC-covered PR (FC/PR) patterns were prepared as ALD-blocking test patterns on Si substrate by inkjet printing. To demonstrate AS-ALD, Al 2 O 3 thin film was deposited via an in-house built system with a multipleslit gas source. The Al 2 O 3 ALD-inhibition performance of the prepared patterns was evaluated for topological analysis using atomic force microscopy (AFM), surface composition analysis using time of flight secondary ion mass spectrometry (TOF-SIMS) after the lift-off process by O 2 plasma, and ultrasonic agitation.

Experimental details
An inkjet printer (DMP2800, Dimatix-Fujifilm Inc.) was used to print the patterns with FC and PR inks. The FC ink (mixture of FC40 and FC722, 3M) was printed to form a low surface energy pattern. FC 40 was used as a solvent, and FC722 as a solute to prepare FC ink. The PR ink was prepared by mixing commercial PR (DSAM-3020, Dongjin Semichem Co.) with ethyl lactate (EL) solvent. FC and PR inks were jetted using 1 and 10 pl cartridges (DMC-11610, Fujifilm Dimatix Inc.), respectively. Si wafer was used as the substrate.
We deposited the Al 2 O 3 thin films via an in-house built roll-to-roll ALD system with a multiple-slit head for gas source supply. The growth rate is determined by the speed of the substrate that is continuously moving relative to the simultaneously delivered precursor sources. The substrate moving speed was maintained at 5 mm/sec. Al 2 O 3 thin films were grown on the substrates at room temperature and under a working pressure of 750 Torr. A total gap of ∼ 250 µm was maintained between the multiple-slit head and the moving substrate. Both the multiple-slit head and substrate were heated to 100°C.
TMA and H 2 O were used as precursors, which were delivered into the multi-slit head with nitrogen gas at flow rates of 700 and 1000 sccm, respectively. H 2 O was purged for 10 s. The growth rate of Al 2 O 3 was ∼ 0.97 Å per cycle. To achieve ∼ 10 nm target thickness, Al 2 O 3 was deposited on Si substrates with printed ALD blocking patterns [31,32]. Figure 1 shows a schematic diagram of the proposed AS-ALD method using the inkjet-printed ALD-inhibition test patterns and lift-off process. The procedure was as follows: (a) FC inkjet printing was used to define the PR pattern width by hydrophobic patterns; (b) PR was printed inside the gap of hydrophobic line patterns; (c) FC removal was performed using O 2 -plasma; (d) FC cover-layer was printed on the PR pattern; (e) Al 2 O 3 ALD was conducted; (f) FC-covered PR inhibition pattern was removed by a lift-off process using an ultrasonicated acetone solution.
The cross-sectional profile and topology of the printed patterns and the Al 2 O 3 layer were analyzed using AFM (NX10, PARKSYSTEMS). Surface composition analysis was performed using TOF-SIMS (ION TOF, TOF.SIMS-5 equipped with a Bi source and sputter gun). The pressure for the TOF-SIMS and preparation chambers was maintained at 3.8 × 10 −9 Torr. The Bi + primary ions were accelerated at 30 keV and 0.5 nA current. A ∼ 90 nm diameter probe beam was rastered over a 150 µm × 150 µm area during data acquisition.

Results and discussion
To investigate the inhibition of Al 2 O 3 film growth during the ALD process using the prepared ink materials, FC, PR, and FC-covered PR (FC/PR) patterns were fabricated on a Si substrate by inkjet printing. The layer structure is depicted in Figure 1. In our previous research [33], we demonstrated AS-ALD using inkjet-printed FC patterns with low surface energy and O 2 plasma. An experiment on droplet jetting stability, print-overlap rate, and substrate temperature was conducted to find the correct printing condition for forming a straight FC line pattern. An FC line pattern with a width of ∼ 40 µm was formed by printing FC dot patterns of ∼ 40 µm diameter with a print-overlap rate of 73% (print drop pitch = 29 µm) at a substrate temperature of 50°C [33].
The printed FC patterns were heated at 110°C for 10 min in an oven. In this study, we also prepared 40 µm-wide FC stripe patterns at intervals of 10 µm on the Si substrate with the same ALD-inhibition pattern design. The FC thin film has a sufficiently low surface energy of 13.04 dynes/cm compared to that of other inhibitory layers such as ODPA, ODTS, PMMA, and PVP thin films, which acts as an inhibitor for AS-ALD [4,6,7,9].
The surface energy of FC thin films was calculated by the Owen-Wendt geometric mean model, and contact angles were measured via a sessile drop method using three probe liquids (CH 3 NO, CH 2 I 2 , H 2 O). The thickness of the FC pattern was about 3 nm at its middle region, while the borderline height of the FC mask line pattern was ∼ 8 nm owing to the coffee-ring effect [33].
The ink-jetting characteristic of the diluted PR solution for PR printing was evaluated by changing the EL-mixing ratio. Table 1 shows the viscosity, surface tension, and ink-jetting performance for the various mixing ratios of the PR and EL solvents. Undiluted PR of 25 cP was not ink-jetted, but it was possible to ink-jet the mixture of PR and EL at mixing ratios of 33 and 50% v/v, resulting in a viscosity of 7.5 cP or less. Figure 2(a-d) shows the PR dot patterns inkjet-printed with a single drop in the hydrophilic tracks. First, the hydrophobic FC ink was printed with the desired gap distance to  form a hydrophilic track. In this step, the width of the PR pattern was decided by the width of the hydrophilic track (Wtrack) between the FC line patterns [34]. The hydrophilic track width decreased at 10-µm intervals from 60 to 30 µm. Lengths of the PR dot patterns in the hydrophilic track were measured and found to be 69, 73, 85, and 96 µm, depending on the hydrophilic track width. Thereafter, to form the PR straight line using the process depicted in Figure 1(a-b), PR ink was continuously printed with a print-overlap rate of 30% of each dot length, as shown in Figure 3. The center thickness of the line pattern increased from 0.1-0.5 µm, as the line width decreased from 60 to 30 µm, as shown in the cross-sectional profile analysis in Figure 3(e). In addition, the coffee-ring effect decreased as the line width of the PR pattern decreased, and its thickness increased. For better understanding, the region of the FC line pattern is indicated by a dotted line in Figure 2(a) and Figures  3(a), and FC line patterns exist on both side regions of the PR patterns. After dipping in water, it was possible to discriminate FC patterns with the help of water spots formed on the non-wettable FC patterns (image not shown). Because the printed FC pattern is transparent, it was hard to directly discriminate the patterns in optical microscopic images.
To increase the thickness of the PR patterns, these were overprinted in the hydrophilic track using the layer-by-layer printing method. Overprinting was performed by a solvent drying process between inkjet printing passes. After overprinting three times while maintaining the line width pre-defined by both hydrophobic FC patterns, straight PR patterns remained welldefined without the bulging phenomenon (Figure 4(a)) [35]. The thickness of the 30 µm width PR line pattern increased from 0.45-1.28 µm depending on the print number (from 1 to 3), without the coffee-ring effect as shown in the cross-sectional profile analysis in Figure  4(b). However, in the case of 40, 50, and 60 µm line widths, the coffee-ring effect was observed on the edge of the printed PR line pattern despite increasing the print number. In particular, for the 60 µm line width, the center thickness was almost unchanged at about 0.1 µm and the height of both side edges increased from ∼ 0.3 to ∼ 0.8 µm (Figure 4(c)). Finally, FC/PR patterns ( Figure  4(d)) were fabricated by printing FC ink onto PR patterns as a cover layer according to the process shown in Figure 1 (a-d).
To demonstrate AS-ALD, Al 2 O 3 thin films approximately 10 nm thick were deposited on the inkjet-printed FC, PR, and FC/PR ALD-inhibition patterns on the Si substrate by inkjet printing. Figure 5 shows the AFM analysis results for a 1 µm × 1 µm scanned area before and after Al 2 O 3 ALD on various surfaces. The root mean square roughness (Rq) of Al 2 O 3 -deposited Si substrate increased to 0.355 nm ( Figure 5(b)) compared to that of bare Si surface, 0.313 nm (Figure 5(a)). Peak-to-valley roughness (R p-v ) of Al 2 O 3 -deposited Si substrate increased from 3.032 nm to 4.923 nm after the ALD process. For the FC surface, R p-v remarkably increased from 2.509 ( Figure 5(c)) to 23.84 nm ( Figure 5(d)). Unlike the Al 2 O 3 -deposited Si surface, many cavities were irregularly distributed on the lumped and undulated Al 2 O 3 -deposited FC surface. Fine grains appeared on the Al 2 O 3 deposited PR surface, as shown in Figure 5(e, f). R q decreased from 1.398-0.300 nm, and R p-v decreased from 9.110-2.725 nm after Al 2 O 3 deposition. For the FC film-covered PR surface, R q abruptly increased from 0.455 ( Figure 5(g)) to 0.860 nm ( Figure  5(h)). Cavities were observed on the lumped and undulating surface, similar to the Al 2 O 3 -deposited FC surface. To summarize the AFM analysis, all surfaces before Al 2 O 3 deposition showed fine-grain spreading topology. Al 2 O 3 deposition showed a tendency to increase Rq and decrease R p-v . In the case of the Si and PR surfaces, surface morphology maintained a fine-grain spreading topology even after Al 2 O 3 deposition. However, FC and FC film-covered PR showed lumped and undulating surfaces, including cavities, along with an abrupt increase of R q .
After Al 2 O 3 ALD on the prepared lift-off patterns, we tried to remove unwanted regions using O 2 plasma or ultrasonic agitation. First, in the experiment for AS-ALD of Al 2 O 3 using FC patterns, removing the unwanted FC patterns with 10 min of O 2 plasma treatment was possible as reported in the previous study. Finally, a 10 µm wide and 6 nm thick Al 2 O 3 pattern was formed on the Si substrate (Figure 6(b)) [33]. In this study, TOF-SIMS analysis was carried out to map residual element information before and after lift-off patterning of Al 2 O 3 thin film and to investigate the inhibition properties of FC patterns during the Al 2 O 3 ALD process. As shown in Figures 6(c) and (d), there was no signal of AlO-before Al 2 O 3 thin film deposition on Si substrate with FC pattern. Although Al 2 O 3 was physically patterned by O 2 plasma as shown in Figure 6(b), AlOwas detected in the two-dimensional TOF-SIMS mapping images after Al 2 O 3 deposition and lift-off patterning (Figure 6(e)). Also, AlOwas detected by depth profile analysis underneath the background blocked by the FC lift-off patterns. This means that FC lift-off patterns could not entirely inhibit Al 2 O 3 growth. Al 2 O 3 thickness was estimated to be ∼ 0.75 nm by fractional analysis of the TOF-SIMS depth intensity for the AlOsignal in the background and pattern regions, as shown in Figure 6(f) . It was confirmed that a very thin Al 2 O 3 layer remained after the lift-off process. In tandem with the AFM, surface profilometry, and TOF-SIMS analyses after lift-off patterning of Al 2 O 3 using the FC pattern, it can be concluded that the Al 2 O 3 layer was porously formed on the FC pattern during the ALD process. Porous Al 2 O 3 was also removed by gaseous lift-off phenomenon during O 2 plasma etching of FC film. With the result of the AFM analysis, porous topology was observed, and many cavities were irregularly distributed on the lumped and undulating surface of the Al 2 O 3 -deposited FC surface. Nevertheless, the discovery of AlO-underneath the FC film through the TOF-SIMS analysis indicates that the FC film cannot completely prevent Al 2 O 3 nucleation during subsequent exposure to TMA and H 2 O. Figure 7 shows optical microscopy images before and after ultrasonic agitation in acetone of Al 2 O 3 -deposited PR pattern for lift-off patterning. The Al 2 O 3 -deposited PR pattern showed a wrinkled surface (Figure 7(b)) compared to the surface prior to the ultrasonic agitation (Figure 7(a)), and the Al 2 O 3 -deposited PR pattern could not be removed even after 15 min of ultrasonic agitation. Further, considering that the Al 2 O 3 deposited PR surface is reasonably smooth with fine grains as seen in the AFM surface analysis, it can be inferred that the dense Al 2 O 3 film eventually grew on the PR pattern surface through subsequent exposure to TMA and H 2 O. Other studies have reported that TMA is physically adsorbed and provides nucleation sites for Al 2 O 3 film growth with a polymer-blocking layer, although TMA does not react chemically [18,19,21]. Figure 8 shows the AFM and TOF-SIMS surface analyses results of the lift-off patterning of Al 2 O 3 using the FC/PR pattern. We prepared two FC/PR patterns  AS-ALD process using the FC/PR pattern can be inferred as follows: the FC film partially blocks Al 2 O 3 nucleation leading to porous Al 2 O 3 film; and second, 1.28 µm-thick PR pattern blocks Al 2 O 3 cluster diffusion to Si surface. Acetone resolves the PR mask pattern underneath the Al 2 O 3 film through cavities of the porous Al 2 O 3 film, leading to the lift-off of undesired Al 2 O 3 film for the patterning process. In general, to make the acetone penetrate the conformal Al 2 O 3 and attack PR below, manual scratching of the Al 2 O 3 surface is necessary [7,20,22]. However, in our study, the Al 2 O 3 film was successfully patterned by a lift-off process using an inkjet-printed FC/PR blocking layer and ultrasonic agitation without manual scratching.

Conclusion
In this study, AS-ALD of Al 2 O 3 was successfully demonstrated using ALD inhibition patterns consisting of a two-layer structure formed by inkjet printing. An FC thin film was coated to inhibit Al 2 O 3 nucleation and growth during the ALD process after thick PR was inkjetprinted to allow easy removal of ALD-inhibition patterns by the lift-off process. The thickness of the inkjet-printed FC pattern was about 3-8 nm owing to a coffee-ring effect. The thickness of the PR line pattern increased from 0.45-1.28 µm for the 30 µm line width by layer-by-layer printing without a coffee-ring effect. To confirm the proposed AS-ALD process, FC, PR, and FC/PR were printed as ALD-inhibition test patterns on Si substrates. Thereafter, a ∼ 10 nm thick Al 2 O 3 thin film was deposited via an in-house built system with a multiple-slit gas source. The Al 2 O 3 film was successfully patterned by a lift-off process using an FC-covered 1.28 µm-thick PR inhibition layer and ultrasonic agitation. In the TOF-SIMS depth profile analysis, AlOwas not detected underneath the FC/PR patterns. By combining the AFM, surface profilometry, and TOF-SIMS analyses results, inferences can be made regarding the developed AS-ALD mechanism. The FC film partially blocked Al 2 O 3 nucleation leading to porous Al 2 O 3 film, while the thick PR pattern additionally blocked Al 2 O 3 cluster diffusion to the Si surface. The sonicated acetone resolves the PR mask pattern underneath the Al 2 O 3 film through cavities of porous Al 2 O 3 , leading to lift off of undesired Al 2 O 3 film for the patterning process. In the case of the FC pattern, although Al 2 O 3 was physically patterned by O 2 plasma, AlOwas detected in the FC pattern-blocked area by TOF-SIMS depth profile analysis. Also, for the PR pattern, the Al 2 O 3 -deposited PR mask could not be removed even after 15 min of ultrasonic agitation. Our study demonstrated an efficient patterning performance of the inkjet printing process to form inhibition patterns for AS-ALD. The developed inkjet printing process successfully patterned multilayer structures with different thicknesses and surface properties by precisely positioning target materials on a substrate using computer control. This process can be used as an effective method for patterning inorganic thin films for low-cost, largearea, flexible electronic components without using a photolithography patterning process.

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

Funding
This study was conducted with the support of the Korea Institute of Industrial Technology as a part of 'Development of Core Technologies for a Working Partner Robot in the Manufacturing Field (kitech EO-22-0009)'. This work was also supported by the National Research Foundation (NRF) of Korea grant funded by the Korea government (MSIT) (No. 2020R1A5A1019649).
Kyung Hyun Choi received his B. S. and M. S. degrees from Pusan National University, Pusan, Korea, in 1985 and 1990, respectively, and his Ph.D. degree from the University of Ottawa, Canada, in 1995. He is a professor of Mechatronics Engineering at Jeju National University in South Korea. His current research interests involve printed electronics and devices such as OLED, OPV, and Memristor.