Mechatronic Spatial Atomic Layer Deposition for Closed‐Loop and Customizable Process Control

A customized atmospheric‐pressure spatial atomic layer deposition (AP‐SALD) system is designed and implemented, which enables mechatronic control of key process parameters, including gap size and parallel alignment. A showerhead depositor delivers precursors to the substrate while linear actuators and capacitance probe sensors actively maintain gap size and parallel alignment through multiple‐axis tilt and closed‐loop feedback control. Digital control of geometric process variables with active monitoring is facilitated with a custom software control package and user interface. AP‐SALD of TiO2 is performed to validate self‐limiting deposition with the system. A novel multi‐axis printing methodology is introduced using x‐y position control to define a customized motion path, which enables an improvement in the thickness uniformity by reducing variations from 8% to 2%. In the future, this mechatronic system will enable experimental tuning of parameters that can inform multi‐physics modeling to gain a deeper understanding of AP‐SALD process tolerances, enabling new pathways for non‐traditional SALD processing that can push the technology towards large‐scale manufacturing.


Introduction
Atomic layer deposition (ALD) is a nanomanufacturing technique that is widely used to control surfaces and interfaces and DOI: 10.1002/admt.202301728 has been commercialized, most notably, by the semiconductor industry.ALD utilizes self-limiting chemical reactions to create thin films with atomic-scale control, enabling conformal coating of high-aspect ratio surfaces. [1,2]The substrate is exposed to alternating precursor pulses (Figure 1), which are separated temporally to deposit films with sub-nanometer precision.However, to effectively separate the precursors with respect to time, a purging step is introduced between the reactants, resulting in relatively slow cycle times.In addition, current ALD equipment relies on vacuum systems that are energy and cost intensive to scale, limit compatibility with substrates that are sensitive to vacuum environments, and are difficult to interface with continuous (as opposed to batch processing) manufacturing lines.As a consequence, while ALD has been commercialized for low-volume, high-value applications, [3] there is limited commercialization in higher-volume, lower-cost applications, such as clean energy technologies, catalysis, and biomedical applications.
[6][7] A relative motion path between the zones and substrate exposes the surface to the alternating precursors, as shown in Figure 1b.Delivery of the ALD reactants in isolated spatial zones that are separated by inert gas regions eliminates the need for time-expensive purging steps.As a result, significantly faster film growth times, in the range of single nm s −1 compared to several hundredths of nm s −1 for conventional ALD, have been demonstrated. [8,9]Additionally, SALD can be performed at atmospheric pressure, which eliminates the need to scale-up vacuum systems, opens a wider range of substrate compatibility, and enables integration into current manufacturing processes.Because of these benefits, AP-SALD holds promise to address the large-scale manufacturing needs of interfacial engineering at the nanoscale.
A technique known as close-proximity SALD uses a depositor to create precursor zones that are separated with inert gas curtains and exhaust lines to isolate the precursors and prevent transport of oxygen from the surroundings.The substrate is brought in close proximity to the depositor and then moved relative to the zones, exposing the substrate to alternating precursors and depositing a film (Figure 1b).Among the stateof-the-art close-proximity AP-SALD systems, multiple designs have been demonstrated for both batch and roll-to-roll [10,11] processes, including rotating disc, [8,12,13] rotating drum, [6,14,15] linear unidirectional, [10,[16][17][18][19][20][21][22][23][24][25][26] and linear multi-directional. [27,28] Notably, AP-SALD has been industrialized for batch passivation of solar cells. [6] Typically, clo-proximity AP-SALD systems have fixed geometric parameters such as the gap size and relative alignment between the depositor and substrate (Figure 1c). While many systems hav been designed to alter these parameters manually, there are few examples where integrated sensors and actuators are used to actively monitor and adjust geometric process parameters in real time. Therefoe, there is limited scientific understanding of the importance of tolerances to these adjustable process parameters.This lack of understanding yields uncertainty in optimized design constraints and bounds on the sample geometry.
The influence of variations of gap size and alignment during close-proximity AP-SALD is particularly interesting since ALD processes have been proposed for a wide range of applications where the bulk substrate is not a planar geometry.For example, anti-corrosive coatings could be used for automobiles, [31] and anti-biofouling coatings could be used for marine vessels. [32]Previous work has investigated the use of AP-SALD on silicon (Si) trenches, [19] Si pillars, [28] textured Si, [29] and nanowires, [33][34][35] but the characteristic length scale of these features is typically much smaller than that of the bulk substrate.A unique solution to enable coating of challenging samples is to design the depositor geometry to match that of a specific sample, as shown by Toldra-Reig et al. [36] Therefore, various depositors have been 3D printed for specific applications, but the complexity of this approach grows greatly as the substrates become larger and require design flexibility.
One could envision an AP-SALD "paint brush" that could coat sizeable surfaces that are too large to fit into a conventional ALD chamber and too complex to be uniformly coated by the stateof-the-art close-proximity systems.Mousa et al. demonstrated a Figure 2. a) Simplified 3D model of system.The depositor is fixed to the top plate, which is used to define its orientation using two linear actuators (L4 and L5) and a pivot point (P).The substrate plate is located below the depositor.The motion of the substrate is controlled by two precision motorized stages (X and Y).b) Three capacitance probe sensors (C1, C2, and C3) are mounted to the depositor and three linear actuators (L1, L2, and L3) are used to define the orientation of the substrate plate.Note that C1 and L1 are on the backside of the system and are not directly visible.c) A photograph of the real system is provided and cross-labeled.
similar concept on a car windshield, using in loco conventional ALD, [37] rather than SALD.The foundation for this approach using close-proximity AP-SALD has been demonstrated with a "pen" that allows localized printing with a line width on the order of millimeters. [27,28]In the future, one could design a variety of close-proximity AP-SALD systems to coat customized geometries over a range of length scales; however, there is limited understanding of the acceptable tolerances for the AP-SALD geometric parameters needed to achieve uniform films, including gap size and relative alignment, and how they are coupled with gas flow parameters, such as flow rates and pressure.Therefore, there is a need to understand the extent to which a sample geometry can vary beneath an AP-SALD depositor and how to monitor and control the process to yield quality, uniform films over large areas.
In this study, a customized AP-SALD system with mechatronic control and in situ sensing of the process parameters has been designed and built.Herein, we discuss the design of this system with closed-loop process control over a range of process variables, most notably gap size, alignment, and relative motion.We then demonstrate and validate our system by depositing titanium dioxide (TiO 2 ) films using titanium(IV) isopropoxide (TTIP) and deionized (DI) water as the metal and oxidant precursors, respectively.Finally, we investigate the effects of multi-axis printing on film uniformity.

System Design
A close-proximity, linear travel, mechatronic AP-SALD system was designed and manufactured with closed-loop control over the key geometric parameters of gap size, alignment, and relative motion (Figure 2).A showerhead depositor and substrate plate form gas/precursor zones for the process region (Figure 2a).Capacitance probe sensors (Figure 2b) measure the gap size and alignment, while linear actuators and a pivot point (Figure 2a,b) are used to maintain the desired geometry by adjusting the orientation of the top plate and substrate plate.Two precision motorized stages control the substrate velocity and positioning (Figure 2a).Independent control of gas flow rates and pressure is facilitated by a fluid control system (not shown, further details in Note S1, Supporting Information).Digital control of process variables with active monitoring is facilitated with a custom software control package and user interface.

Process Region
The process region is defined as the volume of space between the bottom of the depositor and the top of the substrate plate, which is comprised of the various gas zones (Figure 3).To achieve selflimiting, conformal ALD within an AP-SALD system, the process region must be isolated from ambient oxygen and water vapor, while also preventing cross-contamination of the precursor regions.A depositor was designed with a showerhead geometry and mounted to a top plate (Figures 2a and 3a).The input gases are directed through the depositor manifold, which has internal flow channels to distribute the gases to their respective zones (not shown, further details in Note S2, Supporting Information).The pinholes create four distinct zone categories to establish an isolated and defined process region -metal precursor, water precursor, outer nitrogen, and inner nitrogen (Figure 3c).Further details on the gas delivery and heating systems are provided in Note S1, Supporting Information.Exhaust lines remove excess nitrogen gas, unreacted precursors, and reaction The depositor is mounted to the top plate.The substrate plate and its alignment system are mounted to two motorized stages (X and Y) that control the position and velocity of the substrate plate.Vertical scale bar indicates 100 mm.b) Bottom view of the depositor shows the pinholes of the showerhead design which form the gas/precursor zones.Center-to-center pinhole spacings are shown.c) Cross-section view of the depositor shows the gas zones (metal precursor, water precursor, inner nitrogen, and outer nitrogen) and exhaust lines.The target substrate is moved relative to these zones for alternating precursor exposure.
byproducts from the process region.The target substrate is introduced into the process region by mounting it to the substrate plate via a vacuum chuck.
[40][41][42] These process region metrics will be directly influenced by process variables including gas flow rates, exhaust pressure, gap size, alignment, and relative motion.Previous close-proximity systems have demonstrated control over the gas flow rates and exhaust pressures, [10,17,19,22,25,26,43] but few have implemented experimental, closed-loop control over the geometric parameters (e.g., gap size and alignment changes during the deposition process), which motivated the development of the mechatronic AP-SALD system described in this study.
When investigating control of these geometric parameters, three planes are critical to the deposition process (P1, P2, and P3 in Figure 4).P1 is defined as the bottom surface of the depositor (Figure 4) from which the precursors exit the manifold.This plane defines the upper surface of the process region.P2 is the top surface of the substrate plate (Figure 4), which defines the lower surface of the process region.Finally, the precision motorized stages (X and Y in Figures 2 and 4) define the third plane (P3 in Figure 4), which is the plane of motion for the substrate plate.We note that in theory, two of these three planes could be rigidly fixed and assumed to be parallel.However, this would not enable the system to compensate for any manufacturing tolerances or drift arising from uncontrolled mechanical or thermal changes.Furthermore, the ability to study AP-SALD under conditions where all three planes may be misaligned allows for future investigations of more complex system topologies such as macroscopically curved substrates.For example, an AP-SALD "paint brush" depositor could be mounted to a robotic arm in a man-ufacturing plant to dynamically adjust the gap size, alignment, and relative motion on the surface of a large non-planar object such as a vehicle frame.
If the relationships between these three planes of interest are not actively monitored and controlled, cross contamination between the metal and water precursor zones can occur, resulting  P1) is defined as the bottom surface of the depositor, and its rotational orientation can be controlled about the x 1 and y 1 axes.Plane 2 (P2) is defined as the top surface of the substrate plate, and its rotational orientation can be controlled about the x 2 and y 2 axes.Additionally, its translational position can be controlled along the z 2 axis to adjust the gap size (Δz).Plane 3 (P3) is defined as the top surface of the orthogonal motorized stage stack, which can be translated along the x 3 and y 3 axes.
in unintended chemical vapor deposition (CVD) on the substrate surface.Furthermore, within a large-scale manufacturing context, precursors could leak from the process region to the ambient environment posing a safety risk to workers.Our system integrates sensors, actuators, and controls to measure, alter, and maintain desired system geometries, as discussed in the following sections.

Sensors
Within close-proximity SALD systems, the gap size (Δz in Figure 4) is defined as the distance between the depositor and the substrate plate.Within literature, this process parameter is often quoted as a single number or not reported at all, [8,10,[14][15][16][17][18][19][20][21][22][23][25][26][27][28] thus assuming perfect alignment. However the gap size can vary in both space and time if active control of parallel alignment is not ensured.Despite this fact, many systems rely on static or passive control of the gap size.
To this end, the described system incorporates three capacitance probe sensors (C1, C2, and C3 in Figure 2) to measure the distance between the depositor (P1 in Figure 4) and substrate plate (P2).Capacitance probe sensors were chosen because they are rated to operate at elevated temperatures, as the system is heated during operation (further details in Section 5, Experimental Section).While the capacitance probes perform well for this study, there are some requirements to enable their use.As the operating principle requires two electrically conductive planes, the capacitance probe sensors must be electrically isolated from the mounting hardware and depositor.The conductivity requirements of the substrate depend on its geometry.If the substrate is larger than the linear distance between the sensors, the substrate would typically need to be conductive to provide the second plane to form a capacitor.However, if using a sufficiently thin substrate, a large non-conductive substrate could also be used as it would only alter the dielectric constant between the two planes.Furthermore, if the sample is smaller than the distance between the sensors, only the substrate plate must be conductive.
The sensors are mounted on separate sides of the depositor such that they lie within P1 in Figure 4.With three gap size measurements, the plane equation for the substrate plate (P2) relative to the depositor surface (P1) can be calculated using the system of equations shown in Note S3 (Supporting Information).In addition, the plane transformation to achieve parallel alignment of the substrate plate and depositor at a given gap size can be calculated using matrix manipulations and Cramer's rule (further details in Note S3, Supporting Information).These sensors, combined with the control software described in Section 2.4, allow for high-resolution monitoring in both the spatial (< 50 nm) and temporal (< 100 ms) domains.

Actuators
The sensors described above allow for the measurement of the current gap size and alignment.Given these values, one can calculate the needed changes to the system configuration to achieve a desired gap size and alignment during a deposition process.A combination of linear actuators, a pivot point, and precision mo-torized stages are used to control the position and orientation of the depositor and substrate plate during the deposition process.
The substrate plate rests upon three stepper-motor-driven linear actuators (L1, L2, and L3 in Figure 2) which constitute the substrate alignment system.The three points of contact define the substrate plane (P2 in Figure 4).This alignment system controls the orientation of the substrate plate (rotation about x 2 and y 2 in Figure 4) and vertical position (along z 2 in Figure 4).The linear actuators can be driven simultaneously or independently allowing for precise control of both the gap size and relative orientation of planes P1 and P2.
The top plate, to which the depositor is rigidly mounted (Figure 2a), rests upon its own alignment system.Three supports define its position -one passive pivot point and two steppermotor-driven linear actuators (P, L4, and L5 in Figure 2).Controlling the top plate allows for the angular orientation of P1 (rotation about x 1 and y 1 in Figure 4) to be adjusted and controlled relative to P3.
The substrate alignment system is mounted on two orthogonal precision motorized stages to control the x 3 and y 3 position, velocity, and acceleration of the substrate during a deposition process (Figure 3a).The use of a two-axis stage configuration, as opposed to the more common uniaxial motion (along the x-axis), enables more complex substrate paths, as will be discussed in Section 3.2.For further details on the linear actuators and motorized stages, please see Section 5, Experimental Section.

Controls
The sensors and actuators described above provide the inputs and outputs, respectively, to a control software package.To develop an integrated control software, Python 3 was selected, as it allows for customizable user interface design, parallel processing with multi-threading, a wide range of communication protocols, and many open-source controls packages.A graphical user interface (GUI) enables the user to actively control and monitor the various system parameters such as valve actuation, gas flow rates, process temperatures, gap size, alignment, substrate position, and velocity in real time.The following sections demonstrate experimental control with closed-loop feedback over the gap size, alignment, and relative motion of the close-proximity AP-SALD system.

Stationary Alignment Control
As previously discussed, the alignment of the substrate plate (P2) relative to the depositor (P1) is important for maintaining a defined process region.While the motorized stages are held stationary, the substrate alignment system can adjust the substrate position and orientation with real-time measurements from the capacitance probe sensors.This closed-loop feedback allows for automatic alignment of P2 and P1 for a given gap size.An automated example of the process-region alignment can be seen in Figure 5a-b, where the substrate was initially misaligned and then the alignment was corrected to a specified gap size of 850 μm using closed-loop control.As a proxy for the misalignment of two planes, one can compare in Figure 5b the range of the gap sizes measured by the capacitance probes (C1-3) before (Region I) and after (Region III) alignment.Initially, the range of measured gap sizes was 256 μm, indicating non-parallel alignment.After the process-region alignment, this value was reduced to 1.44 μm, demonstrating the effectiveness of the mechatronic system control.The alignment time can vary based on the initial misalignment conditions, but typically is completed within 10 to 15 s.

Path-Dependent Alignment Control
While stationary alignment control is necessary for SALD, it does not guarantee that a consistent gap size is maintained between the depositor and substrate during the deposition process.This is because the depositor (P1) may be misaligned to the plane of motion established by the linear stage (P3).If all three planes (P1, P2, and P3) are not aligned, the average gap size may increase or decrease during a reciprocating printing motion.This is illustrated schematically in Figure 5c, where the average gap size (Δz 1 , Δz 2 ) varies as a function of the translational position of the x-stage, which is responsible for the back-and-forth printing motion.The reason for this variation in gap size with respect to substrate position is that while P1 and P2 were brought into parallel alignment during the stationary alignment process, they are not necessarily parallel to P3 along the path of motion, as discussed in Section 2.1.
Figure 5d shows experimental measurements from the capacitance probe sensors when P1 and P2 have been brought into alignment using the stationary alignment control algorithm described above, but they are not yet aligned to P3.As a result, the range of average gap size along the motion path before performing path-dependent alignment control is approximately 320 μm (left panel of Figure 5d).The angle of the misaligned depositor (P1) relative to the plane of motion (P3) can be calculated using the change in average measured gap size over the known linear distance of travel, resulting in a calculated angle of approximately 0.4 degrees.While this may seem like a small misalignment error, the combination of small gap sizes and large deposition motions can result in crashing of the depositor and substrate, which is unacceptable within a manufacturing process.[40][41][42] To correct for the P1-P3 misalignment, the path-dependent alignment control algorithm follows a two-step procedure.In the first step, the capacitance probe sensors are used to measure the variations in gap size along the motion path.In the second step, the depositor is brought into alignment with the plane of motion established by the linear stages.This alignment procedure is accomplished by first calculating the required angular orientation of the depositor (P1) with respect to the orientation of the plane that defines the motion path (P3).The needed adjustment to bring P1 and P3 into alignment is accomplished by the linear actuators that control the orientation of the top plate (L4 and L5 in Figure 2a) as described in Section 2.3.After bringing P1 and P3 into parallel alignment, the stationary alignment control process is repeated to realign P1 and P2, which ensures that all three planes are parallel throughout the motion path.Figure 5d shows that the range in the average gap size along the motion path after path-dependent alignment is significantly reduced to approximately 4 μm, which corresponds to a misalignment angle of approximately 0.005 degrees.
In summary, the function of path-dependent alignment control is to generate a "map" of relative alignment along the full motion path, and then adjust the actuators in a manner that maintains parallel alignment throughout this motion path.In the current study, all three planes of interest (P1, P2, and P3) are flat, and therefore, a singular configuration of the actuators is sufficient to ensure parallel alignment throughout the full motion path.In the future, having a mechatronic SALD system could also enable non-planar motion paths, for example, coat curved surfaces.The demonstration of path-dependent alignment control in this study represents a first step toward this vision, where the motion path is flat.However, in the future, path-dependent alignment control could be performed by first using sensors to map the variations of substrate topography (P2) along a printing motion path (P3).This could then be used to define a tool path, analogous to CNC control of milling machines, of the depositor motion and alignment (P1).Once the predicted tool path is defined, closedloop control in real time could be further implemented to adjust for deviations in parallel alignment (i.e., drift), which will be discussed in the subsequent section.

Drift Control
The combination of stationary and path-dependent alignment control produces a system with all three planes (P1, P2, and P3 in Figure 4) aligned so that the gap size is consistent during a printing process.However, machining and process tolerances and environmental variables (such as thermal expansion) in the manufacturing setting can cause the gap size to drift during a deposition process.With the mechatronic system described, the gap size and alignment can be monitored and corrected throughout a deposition process with closed-loop feedback.In Figure 6, we show that the average gap size drifted ≈25 μm away from the set point of 800 μm over 600 AP-SALD cycles without active control.With closed-loop control, the gap size was corrected every 20 cycles, resulting in an average gap size that did not vary more than 4 μm from the set point.While the exact drift for a given system will be largely dependent on its design and process conditions, we demonstrate that with closed-loop control of the effects of drift on the gap size and alignment can effectively reduced during an AP-SALD process.

System Demonstration
To demonstrate the ability of the AP-SALD system to deposit films in a self-limiting ALD mode, titanium dioxide (TiO 2 ) films were grown using titanium(IV) isopropoxide (TTIP) and deionized (DI) water as the metal and oxidant precursors, respectively.[46][47][48][49][50][51][52][53] For each deposition, the substrate was moved at a velocity of 20 mm s −1 resulting in a cycle time of 4.6 s.The total process time linearly increases with the number of cycles.A full description of the deposition parameters, including temperatures, flow rates, substrate velocity/position, is provided in Section 5, Experimental Section.
To characterize the film growth rate, ellipsometry measurements were performed at 20 separate locations along the substrate within the deposition region.We note that geometry of the deposited region depends on the range of linear travel of the xstage during the back-and-forth motion of the printer, and therefore the thickness measurements were collected within the deposition region to avoid edge effects from overlapping exposures (further details in Note S4, Supporting Information).The films exhibit linear and saturated growth behavior that is characteristic of well-behaved ALD processes (Figure 7).A linear increase in film thickness with respect to cycle number was demonstrated over 800 cycles (Figure 7a) resulting in a calculated growth per cycle (GPC) of 0.54 Å cycle −1 .Furthermore, the growth per cycle was observed to saturate at 0.51 Å cycle −1 when the precursor flux (which corresponds to the precursor dose in traditional ALD) was increased by increasing the flow rate through the TTIP bubbler (Figure 7b).The measured growth rate of this AP-SALD system lies within the reported range of values in the literature which is ≈0.04-11][52][53]

Film Characterization
To characterize the film stoichiometry, X-ray photoelectron spectroscopy (XPS) was performed on a TiO 2 film that was deposited using 600 AP-SALD cycles, which corresponds to a thickness of 32 nm as measured by ellipsometry (Figure 8).As seen in the inset, the actual titanium (Ti) to oxygen (O) ration is approximately 1:2.The small variation from the ideal is likely due to the known preferential sputtering of O within TiO 2 . [54]The film shows no evidence of pinholes, as there are no silicon (Si) peaks present, even after three minutes of Argon (Ar) sputtering, which corresponds to ≈3 nm of film removed.A small percentage of carbon (C) is observed in the film which is consistent with previously reported TiO 2 films deposited using TTIP. [55]

Multi-Axis Printing
][18][19][20][21][22][23][24][25][26] Recently, a new type of SALD system design has been introduced that allows for multiaxis printing of SALD patterns, allowing for area-selective deposition in the patterned regions. [27,28]Through miniaturization of a close-proximity SALD head, an "SALD pen" can be fabricated, which allows for customizable and free-form patterning of the deposited material on the substrate.However, to our knowledge, in larger-scale SALD depositors, multi-axis printing has not been previously demonstrated.Here, we demonstrate the potential of multi-axis printing to compensate for non-uniformities in the deposited material geometry that may arise as a result of the system design.We note that the depositor design in this study only "prints" material when the substrate is translated along the xaxis.The y-axis motion adjusts the position of the deposited pattern in orthogonal direction but does not actively print new material because of the geometry of the precursor zones.
As shown in Figure 3, our depositor is comprised of a showerhead geometry, which is common in large-area deposition systems as it offers a simple method of distributing the gases over a large substrate area.However, as a result of the pinhole geometry from the showerhead, the localized fluid mechanics within the process region will exhibit spatial variations in the pressure/concentration profile of the precursors as they are transported radially away from the center of a pinhole.As a result of this flow profile, the local precursor concentration will be largest immediately below the pinholes, and concentration gradients will be present in the regions between the pinholes (Figure 9a).When using single-axis reciprocating motion (i.e., along the x 3axis), the deposited films display periodic variations in the film thickness (Figure 9b).This variation in concentrations means that portions of the substrate surface may not be exposed to a sufficient precursor dose to fully saturate the ALD growth, resulting in variations in the film thickness.
One potential strategy to reduce the film non-uniformity associated with the pinhole geometry is to decrease the substrate velocity, which would increase the precursor dose to the substrate within the lower concentration regions.To examine this possibility, an experiment was conducted at a lower substrate velocity.However, the periodic variations in film thickness were Figure 9.A uniaxial printing motion produces a) a precursor concentration gradient as a result of depositor pinhole design, which creates non-uniformity in the film thickness that can be seen b) optically and c) in a thickness line scan.The effects of multi-axis printing motions can be seen by the c) improved uniformity of the precursor gradient, which deposits a more uniform film as seen in e) the thickness line scan and f) optically.Note that the thickness line scans were measured over the same spatial coordinates on both samples.still observed, while negatively impacting process throughput (Note S8, Supporting Information).Therefore, there is a need for alternative approaches to improve the film non-uniformities that can occur during close-proximity SALD.
Because TiO 2 has a relatively high index of refraction (n = ≈2.4), the small changes in the film thickness can be seen optically (Figure 9b).To provide a quantitative measure of the thickness variations, line scans were performed using spectroscopic reflectometry (further details in Section 5, Experimental Section).When single-axis motion was used along the x 3 -axis, the total range of film thickness along the y 3 -axis was from 30.6 to 36.1 nm, resulting in a deviation from the average thickness of ±8.22% (Figure 9c).The correlation between the pinhole spacing (3.175 mm) and the periodicity of the film thickness (≈3.25 mm) was also confirmed.
Because of the mechatronic system design, a multi-axis substrate motion path was enabled, which can be used to compensate for the periodic variations described above.A customized motion path was selected based on the pinhole geometry.Specifically, a "box path" pattern was programmed using the Python control software package, as shown in Figure 9d, which introduced new steps along the y 3 -axis at the end of each linear motion along the x 3 -axis.The magnitude of the y 3 -axis steps was calculated to be half of the period of the pinholes and ribbing features (1.5875 mm), such that substrate surface is more uniformly exposed to a saturating flux of precursors.This can be thought of as step toward the envisioned "paint brush" described in Section 1, where the customized motion path can interweave the contact points between the "bristles" of the brush and the substrate.Holding all other parameters the same, the customized motion path resulted in a film that was visually more uniform (Figure 9f).This improved uniformity was confirmed with a thickness line scan (Figure 9e).The total range of film thickness along the y 3 -axis was from 33.0 to 34.5 nm, resulting in a deviation from the average thickness of ±2.22%, which is a nearly fourfold decrease compared to single-axis printing.
To our knowledge, this is the first demonstration of multiaxis printing using a large-scale showerhead depositor design to achieve more uniform films.Another important factor that would affect this non-uniformity is the specific design of the depositor flow channels.For example, a slotted geometry for the precursor delivery zones with continuous channels would also assist in alleviating the non-uniformity associated with the pinholes.However, avoiding a significant pressure drop along a slot can also become challenging when the length of the slot is large, which requires more complicated flow control within the internal flow channels of the depositor.Furthermore, showerhead designs are commonly used within industrial systems for flow distribution and delivery, making their improved implementation of general interest.This work demonstrates that with mechatronic sensing and control non-uniformities in the film can be addressed, irrespective of the depositor design.
It is important to note that all the results in Section 3 were performed using the more uniform, multi-axis printing mode, to ensure the most idealized version of SALD possible.A comparison of the bulk TiO 2 films with uniaxial and multi-axis printing is in Note S5 (Supporting Information).Furthermore, while a simple "box path" was demonstrated in this study, more complex motion paths could be designed in the future, depending on the specific depositor and substrate geometries.These micro-scale motions could also be combined with more macroscopic patterning in the x-, y-, and z-directions as well as the rotations along these axes, as shown in Figure 4, which would enable future manufacturing with SALD as a "paint brush" on complex surface topologies.

Conclusion
In this work, a mechatronic AP-SALD system was introduced, closed-loop process control was demonstrated, and TiO 2 deposition was performed.Specifically, in this work: • A mechatronic AP-SALD system was designed and implemented with sensors and actuators that actively measure, alter, and maintain desired system geometries This novel mechatronic AP-SALD system uses capacitance probes, linear actuators, and motorized stages to actively measure, alter, and maintain desired system geometries during deposition.The geometry of the process region is directly controlled by three planes -the bottom surface of the showerhead depositor, the top surface of the substrate plate, and the plane of motion established by the motorized stages.The time-dependent geometrical relationships between these three key planes of interest (e.g., gap size and alignment changes during the deposition process) can affect the process region size and shape, which then affects the localized fluid mechanics within this region.
The three capacitance probe sensors actively measure the gap size between the depositor and substrate plate, defining the relative orientation of the planes.Linear actuators adjust both the position of the depositor and substrate plate to alter the gap size and relative orientation of the planes.Two orthogonal precision motorized stages controlled the x and y position, velocity, and acceleration of the substrate during a deposition process, which enables customized motion paths.An integrated control software was developed using Python 3 to actively monitor and control the geometric process parameters in real-time.
• Closed-loop control over the geometric process parameters enables stationary alignment control, path-dependent alignment control, and drift control The sensors and actuators enable closed-loop control over the key geometrical process parameters of gap size and parallel alignment.Stationary alignment control was performed to bring the substrate into parallel alignment with the depositor within ±1 μm of a desired gap size in 10-15 s.Active monitoring of the plane orientations during deposition revealed the need for pathdependent alignment control to bring all three key planes into alignment.Using the sensors, actuators, and closed-loop control, the range of the average gap size during a deposition process of ≈320 μm (≈0.4 degrees) was reduced to ≈4 μm (≈0.005 degrees).After the initial alignment, the orientation of the three key planes can drift during a deposition process, causing the average gap size and orientation to change with time.The mechatronic system effectively limited the drift in the system to less than 4 μm with closed-loop control during the deposition process, compared to ≈25 μm without correction.
• Linear and saturated growth of TiO 2 thin films was demonstrated TiO 2 thin films were deposited and characterized.A linear growth rate was shown to be 0.54 Å cycle −1 and saturated growth of 0.51 Å cycle −1 was demonstrated by varying the TTIP precursor flux to the substrate.XPS was performed to validate the film stoichiometry and continuity.
• A novel multi-axis printing strategy was introduced to improve material uniformity The geometry of the showerhead depositor caused thickness variations in the deposited films when using single-axis substrate motion.A novel multi-axis printing methodology used x-y position control to define a customized motion path, which enabled an improvement in the thickness uniformity.A simple "box path" motion was designed specifically for the depositor geometry and resulted in a more uniform film by reducing variations from 8% to 2%.This work only used rectangular motions, but more complex, customized substrate paths could be designed to further increase uniformity.
In the future, this mechatronic system design will allow tuning of process parameters experimentally, which will enable a deeper understanding of the process-property relationships during SALD.This capability of controlling the process parameters can further inform multi-physics models enabling digital twins.

Experimental Section
Sensor Details: The capacitance probe sensors used (HPB-75A, Capacitec, Inc.) have a resolution of 24.48 nm with a measurement range of 1.3 to 1270 μm over the temperature range of -73 to 871 °C.Further discussion on the capacitance probe sensor calibration procedure is provided in Note S6 (Supporting Information).
Actuator Details: The linear actuators used to control the depositor and substrate plate orientations are stepper-motor-driven linear actuators.The three linear actuators (Haydon Kerk 28H47-2.1-915)for the substrate plate (L1, L2, and L3 in Figure 2b) have a linear resolution of ≈0.2 μm, resulting in an angular resolution of approximately 1.13e-4 degrees and 8.20e-5 degrees about the x 2 and y 2 axes, respectively.The two linear actuators (Standa 8CMA28-10) for the depositor (L4 and L5 in Figure 2a) have a linear resolution of ≈0.08 μm, resulting in an angular resolution of 1.03e-5 degrees and 6.22e-5 degrees about the x 1 and y 1 axes, respectively.Commands for each linear actuator are sent to an Arduino board via the Python controls software, which in turn send the appropriate signals to a stepper motor driver for each linear actuator.Both types of linear actuators were driven with 1/16 th micro-stepping.
The two, orthogonally installed linear stages (Aerotech PRO165LM-200) control the x 3 -and y 3 -axis position, velocity, and acceleration of the substrate during a deposition process (X and Y in Figure 2a).A direct drive motor system was selected for the fast and precise motion over a large range, up to 2 m s −1 , ±8 μm, and 200 mm.The linear speed of the stages enables cycle times on the order of tens of cycles per second and areal throughputs less than 1,280 cm 2 s −1 .In addition, the stage selected had small pitch, roll, and yaw error (8.2 arc sec) for stable substrate motion.
Deposition Parameters: SALD of TiO 2 was performed using titanium(IV) isopropoxide (TTIP, min.98%, Strem Chemicals Inc.) and deionized (DI) water as the metal and oxidant precursors, respectively.The TiO 2 films are deposited on 150 mm test-grade silicon wafers that were positioned on the substrate plate using a customized alignment tool.Across depositions, the substrate was moved consistently at 20 mm s −1 with a travel distance of ±23 mm, resulting in a cycle time of 4.6 s and areal throughput of ≈14 cm 2 s −1 .The substrate plate was held at a consistent temperature of 105 °C while the depositor was heated to 115 °C.The TTIP bubbler was heated to a temperature of 70 °C and the water bubbler and associated tubing was kept at room temperature, approximately 25 °C.A rising temperature gradient from the TTIP bubbler to the depositor was established with four temperature zones at 65, 75, 80, and 95 °C, respectively.
Both stainless-steel bubblers have a flow-through design.The flux of vapor supplied by the bubbler can be controlled by the precursor vapor pressure, which is dictated by the bubbler temperature, and the nitrogen carrier gas flow rate through the bubbler.This flow, called the bubbler flow, is directed through the liquid precursor inside the bubbler.The gas mixture that exits the bubbler is composed of the precursor vapor and nitrogen gas, and is merged with another nitrogen carrier gas flow outside the bubbler.This secondary carrier gas, called the driving flow, enables the user to decouple the precursor flux (defined by the bubbler flow) from the total flow rate (defined by the sum of bubbler and driving flows).Both the bubbler and driving flows for each precursor bubbler are controlled by their own respective mass flow controllers (MFCs), totaling four.A detailed schematic of the bubbler and driving flow configuration is provided in Note S1 (Supporting Information).The bubbler flow rates were 45 and 200 standard cubic centimeters per minute (sccm) and the driving flow rates were 2,200 and 1,000 sccm for the TTIP and H 2 O bubblers, respectively.
The flow for each of the two inner nitrogen zones was supplied by a combined flow of 10,000 sccm.The outer nitrogen zone surrounded the entire process region on the four equal sides (Figure 3).The combined flow for the entire outer nitrogen flow was 25,000 sccm.The exhaust pressure was held at -1.07 kPa, gauge pressure.For all experiments, the gap size was held at 1,000 μm and three-plane alignment was performed unless marked otherwise.Further discussion on the temperature zones, bubblers, and flow components is provided in Note S1, (Supporting Information), along with a full schematic of the fluid control system.
The flow parameters were selected using a combination of experimentation and calculations.Given the linear velocity of 20 mm s −1 , the needed flux of precursor molecules (and therefore bubbler flow rate) was calculated based on the estimated precursor zone width, surface site density, and precursor vapor pressure (Note S7, Supporting Information).This established a minimum flow rate for each bubbler flow.By experimentally varying the bubbler flow rate within the rated range of the MFC, the required flow rate needed to achieve saturated, self-limiting films was determined (Figure 7b).The driving, inner, and outer nitrogen flow rates were empirically determined with the goal of balancing the pressure distribution within the process region to ensure adequate separation of the precursor zones while also preventing air infiltration from the surrounding ambient.
Film Thickness Measurements: The thickness measurements of the TiO 2 films were performed with two techniques -ellipsometry and spectroscopic reflectometry.Ellipsometry was performed using a Film Sense FS-1 Multi-Wavelength Ellipsometer system with a 65 degrees angle of in-cidence, a beam size of 4 mm × 9 mm, and a Cauchy model.Spectroscopic reflectometry was performed using a Nanospec 6100 system with a 25 μm diameter spot size and a film stack for calculation and fitting of Air-TiO 2 -Si.The wavelength dependence of the optical constants (n and k) for the deposited TiO 2 were measured using a Woollam M-2000 Ellipsometer.
XPS Measurements: Measurements were performed using a Kratos Axis Ultra XPS with a monochromated Al K X-ray source (10 mA, 12 kV).The spot size was 700 μm × 300 μm.An electron gun was used to maintain charge neutrality on the surface of each sample.Survey scans (pass energy: 160 eV) were used to quantify the atomic composition of the various samples.Core scans (pass energy: 40 eV) were used to investigate the binding environment of elements in each sample.The binding energies were calibrated to that of adventitious surface carbon (284.8 eV). [56]rgon sputtering was performed for three minutes, which corresponds to a film thickness reduction of ≈3 nm, to remove adventitious carbon on the surface.CasaXPS software was used to analyze the XPS data.

Figure 1 .
Figure 1.Schematic of a) conventional atomic layer deposition with a substrate exposed to temporally separated pulses of precursors A and B between purge steps, each with a characteristic time, b) spatial atomic layer deposition (SALD) with a substrate moving between precursors A and B and inert zones, each with characteristic length, and c) close-proximity spatial ALD, showing the gas/precursor zones formed in the gap between the depositor and substrate and the relative motion that enables alternating precursor exposures.

Figure 3 .
Figure 3. a) Simplified 3D schematic of system showing the process region of gas/precursor zones formed between the depositor and substrate plate.The depositor is mounted to the top plate.The substrate plate and its alignment system are mounted to two motorized stages (X and Y) that control the position and velocity of the substrate plate.Vertical scale bar indicates 100 mm.b) Bottom view of the depositor shows the pinholes of the showerhead design which form the gas/precursor zones.Center-to-center pinhole spacings are shown.c) Cross-section view of the depositor shows the gas zones (metal precursor, water precursor, inner nitrogen, and outer nitrogen) and exhaust lines.The target substrate is moved relative to these zones for alternating precursor exposure.

Figure 4 .
Figure 4. Simplified 3D schematic of system showing the three planes of interest.Plane 1 (P1) is defined as the bottom surface of the depositor, and its rotational orientation can be controlled about the x 1 and y 1 axes.Plane 2 (P2) is defined as the top surface of the substrate plate, and its rotational orientation can be controlled about the x 2 and y 2 axes.Additionally, its translational position can be controlled along the z 2 axis to adjust the gap size (Δz).Plane 3 (P3) is defined as the top surface of the orthogonal motorized stage stack, which can be translated along the x 3 and y 3 axes.

Figure 5 .
Figure 5. Stationary and path-dependent alignment control.a) Schematic of initial (stationary) misalignment before SALD deposition.b) Experimental demonstration of closed-loop control of stationary alignment.c) Schematic of path-dependent misalignment during SALD deposition, where gap size (Δz 1 and Δz 2 ) varies as a function of position.d) Experimental demonstration of the effect of closed-loop path-dependent alignment control on gap size variation during printing motion.

Figure 6 .
Figure 6.The average gap size as a function of AP-SALD cycles to demonstrate the system drift during a deposition process with and without correction.

Figure 7 .
Figure 7. a) Thickness of deposited TiO 2 films as a function of Spatial ALD cycles.Growth per cycle (GPC) is calculated as 0.54 Å cycle −1 via the linear fit of the data points (dashed line).For each deposition, the substrate velocity was maintained at 20 mm s −1 over a linear travel distance of ±23 mm, resulting in a cycle time of 4.6 s. b) Growth per cycle (GPC) as a function of TTIP bubbler flow rate.Saturated growth rate of 0.51 Å cycle −1 is shown with the dashed line.Error bars in both panels represent the standard deviation of 20 independent measurements within the process region.All measurements were taken with ellipsometry (see Section 5, Experimental Section for more details).

Figure 8 .
Figure 8. XPS survey scan of AP-SALD TiO 2 film with an inset that shows the estimated atomic percentages.