Automated Contact Angle Tester (ACAT)

• ACAT is an automated instrument that measures surface wettability by integrating controlled droplet deposition, optical image capture, and robotic handling.
• It employs diverse angle-extraction methods, including polynomial fitting, local geometry reconstruction, and model-based inversion, to handle static, dynamic, and anisotropic wetting scenarios.
• ACAT systems are designed for high-throughput precision metrology and are increasingly integrated into self-driving laboratories with closed-loop optimization workflows.

An Automated Contact Angle Tester (ACAT) is an automated instrument or robotic work cell for measuring surface wettability, typically through static or dynamic contact angles of sessile drops, by integrating controlled droplet deposition, optical acquisition, image analysis, and experiment orchestration. In its most explicit recent form, ACAT is a fully integrated robotic work cell developed for automated surface wettability measurement on 3D-printed materials, combining programmable robotics, precise liquid dispensing, and a modular software–hardware architecture (Burgess et al., 16 Jul 2025). More broadly, the term also covers modular wetting-analysis subsystems embedded in self-driving laboratories, including top-down reflectance systems and high-throughput vision pipelines (Quinn et al., 11 Jan 2026, Nazeri et al., 8 Oct 2025). Across these implementations, the central technical problem is not only automation of droplet handling, but reliable estimation of an angle that is often apparent, anisotropic, history-dependent, and sensitive to optics, geometry, and dynamics (Palmetshofer et al., 2024, Silvestrini et al., 2021).

1. Metrological definition and scope

Contact angle metrology begins with the equilibrium relation for an ideal smooth and chemically homogeneous surface, where Young’s angle $\theta_Y$ is set by interfacial energies: $$\gamma_{LG} \cos \theta_Y = \gamma_{SG} - \gamma_{SL}$$ On flat, unstructured PMMA surfaces, $\theta_{FS}$ was used as an estimate of $\theta_Y$, with untreated PMMA giving $\theta_{FS} \approx 75^\circ \pm 5^\circ$ and fluoropolymerized PMMA giving $\theta_{FS} \approx 120^\circ \pm 5^\circ$ (Palmetshofer et al., 2024). In practical ACAT operation, however, what is usually measured is an apparent contact angle from a 2D optical profile rather than a true local 3D Young angle. On microstructured or curved surfaces, the local interface is distorted by pinning, curvature, gravity, or motion, so the metrological object is often $\theta_{\text{app}}$ or a dynamic contact angle rather than a unique equilibrium scalar (Palmetshofer et al., 2024, Liu et al., 2022).

A common oversimplification is to treat wettability as a single angle attached to a material. This fails in several experimentally important regimes. On structured PMMA surfaces with ramps, pyramids, and staggered cubes, the apparent contact angle can vary by $10\text{–}30^\circ$ or more around the droplet perimeter, and a single side-view angle does not represent the droplet (Palmetshofer et al., 2024). On rough pillared surfaces, the free-energy landscape can contain multiple minima, so the measured angle depends strongly on how the droplet was prepared; the Cassie–Baxter state can have only one free-energy minimum, while wet states can present multiple minima and large hysteresis (Silvestrini et al., 2021). ACAT therefore belongs as much to precision metrology and experimental protocol control as to image processing.

Dynamic measurements further enlarge the scope. On curved wetting surfaces in lattice Boltzmann simulations, a local geometric scheme defines the contact angle from the angle between a reconstructed interface line and the local tangent to the substrate, rather than from global drop height and width; the same paper shows that microscopic equilibrium contact angle is independent of gravity even though global shape changes strongly (Liu et al., 2022). This distinction is fundamental for ACAT implementations that attempt to measure under tilted, curved, oscillating, or rapidly forced conditions.

2. Instrument architectures and subsystem design

Recent ACAT architectures fall into a small number of recurring patterns: the enclosed robotic work cell, the modular gantry-based measurement node, and the self-driving laboratory subsystem. The robotic work-cell implementation described in “Design and Development of an Automated Contact Angle Tester (ACAT)” is organized into three core subsystems: an electrical system including power, control, and safety circuits; a software control system based on a Raspberry Pi and Python; and a mechanical system including a 3-axis Cartesian robot, pneumatic actuation, and a precision liquid dispenser enclosed within a safety-certified frame (Burgess et al., 16 Jul 2025). Its electrical design is explicitly tied to NEC 70, NFPA 79, and UL 508A, and its safety circuit is built around an Allen-Bradley MSR138DP safety relay with dual-channel monitoring of emergency-stop pushbuttons and the door interlock (Burgess et al., 16 Jul 2025).

The same paper defines the canonical work-cell layout for an industrially legible ACAT: a lockable AC enclosure, a low-voltage main control enclosure, an operator station, and a robotic test cell built from 40×40 mm aluminum extrusion with acrylic guarding. Motion is provided by a belt-driven $Y$-axis, a ball-screw $X$-axis, and a pneumatically actuated $$\gamma_{LG} \cos \theta_Y = \gamma_{SG} - \gamma_{SL}$$0-axis. Sample handling is performed with a venturi vacuum gripper, and droplet deposition uses a MY2626 pneumatic dispensing valve supplied by a 12 V diaphragm pump. The target droplet volume is $$\gamma_{LG} \cos \theta_Y = \gamma_{SG} - \gamma_{SL}$$1, and the loading station is indexed for 25 parts per batch (Burgess et al., 16 Jul 2025).

A more modular architecture appears in PANDA-film, where the ACAT functionality is one module within a self-driving polymer-film platform. There the measurement stack is built around a modified CNC gantry with a Process Automation Widget, an OT-2 P300 pipette, a FLIR Grasshopper 3 camera, an Edmund Optics telecentric lens, a NeoPixel ring light, and paired red/blue LEDs for top-down reflectance-based contact-angle inference (Quinn et al., 11 Jan 2026). The same platform adds an electromagnetic capping/decapping system, using a 5 V, 50 N electromagnet and an IR break-beam sensor, to mitigate evaporation during long autonomous campaigns (Quinn et al., 11 Jan 2026).

RAISE represents a third pattern: the contact-angle tester as a subsystem inside a closed-loop formulation laboratory. Its hardware centers on an Opentrons OT-2, a custom multilevel imaging stage, controlled backlighting, and a pick-and-place camera tool that the robot handles like a pipette tip. In that system, the same robot both deposits $$\gamma_{LG} \cos \theta_Y = \gamma_{SG} - \gamma_{SL}$$2 droplets and positions the side-view camera, allowing automated formulation, deposition, imaging, and optimization in a single loop (Nazeri et al., 8 Oct 2025).

Platform	Core hardware	Distinctive measurement mode
ACAT robotic work cell (Burgess et al., 16 Jul 2025)	Cartesian robot, pneumatic $$\gamma_{LG} \cos \theta_Y = \gamma_{SG} - \gamma_{SL}$$3-axis, vacuum gripper, MY2626 dispenser	Side-view contact-angle tester with automated pick-place and $$\gamma_{LG} \cos \theta_Y = \gamma_{SG} - \gamma_{SL}$$4 droplets
PANDA-film wetting module (Quinn et al., 11 Jan 2026)	CNC gantry, OT-2 P300 pipette, telecentric camera, red/blue LEDs	Top-down reflectance with calibrated LED-spot distances
RAISE (Nazeri et al., 8 Oct 2025)	OT-2 liquid handler, pick-and-place camera, flat-panel LED stage	Side-view automated imaging with Bashforth–Adams fitting

These architectures differ in form factor, but they share a common design doctrine: fixed and calibrated optics, deterministic droplet generation, known sample coordinates, safety interlocks, and software that treats wetting measurement as a sequence of machine states rather than as a manual visual task.

3. Automated workflows and angle-extraction algorithms

The physical workflow of an ACAT is typically a cycle of loading, positioning, deposition, imaging, and unloading. In the 2025 ACAT work cell, the process begins with the operator loading up to 25 parts, after which the robot homes its axes, picks a part from the loading station, places it on the contact-angle stage, dispenses a $$\gamma_{LG} \cos \theta_Y = \gamma_{SG} - \gamma_{SL}$$5 water droplet, triggers image acquisition, and finally moves the tested part to an unloading area (Burgess et al., 16 Jul 2025). This baseline workflow is extended in self-driving systems: RAISE adds formulation preparation, replicate scheduling, and optimization logic, while PANDA-film adds z-stack imaging and top-down reflectance calibration (Nazeri et al., 8 Oct 2025, Quinn et al., 11 Jan 2026).

Angle extraction is more heterogeneous than the mechanics. One important side-view strategy is full azimuthal scanning by sample rotation. On anisotropic microstructured PMMA surfaces, a DataPhysics OCA 15EC with an IDS UI-3360CP-M camera, Navitar Zoom 6000 lens, diffuse backlight, and an electronically actuated rotation stage was used to acquire 360 images per droplet at $$\gamma_{LG} \cos \theta_Y = \gamma_{SG} - \gamma_{SL}$$6 angular increments. The sample, not the camera, was rotated, and a manual baseline was defined once per experiment. For each frame, the droplet contour near the baseline was fitted by a polynomial, explicitly avoiding ellipsoidal fitting because pinning-induced 3D distortions made spherical-cap assumptions unreliable; the contact angle was then computed from the tangent to that polynomial at the baseline (Palmetshofer et al., 2024). Relative humidity was actively controlled to 98% RH with ultrasonic atomizers to suppress evaporation during the full 360° scan, and each structure/treatment condition was repeated in 10 experiments (Palmetshofer et al., 2024).

A second strategy is local-geometry reconstruction. In the lattice Boltzmann scheme for curved wetting surfaces, the gas–liquid interface is reconstructed from the iso-density contour

$$\gamma_{LG} \cos \theta_Y = \gamma_{SG} - \gamma_{SL}$$7

two interface points approximately 1 and 2 lattice units from the wall define a local straight-line approximation, and the contact angle is then computed from the angle between this local interface line and the local substrate tangent (Liu et al., 2022). Although derived for simulation, the same logic is explicitly presented as generalizable to physical ACAT design: reconstruct interface from image data, use two nearby points to define local slope, fit the local wall tangent, and measure the angle between them (Liu et al., 2022).

A third strategy is model-based contour fitting. RAISE uses a side-view silhouette acquired against strong backlighting, converts the image to grayscale, applies Gaussian and Sobel filtering, thresholds with Otsu’s method, partitions contours to exclude the reflected droplet below the baseline, and then fits the upper droplet contour using the Bashforth–Adams solution of the Young–Laplace equation (Nazeri et al., 8 Oct 2025). The fit quality is reported by an RMSE between the fitted curve and the detected contour; example values were $$\gamma_{LG} \cos \theta_Y = \gamma_{SG} - \gamma_{SL}$$8 for a water droplet with contact angle $$\gamma_{LG} \cos \theta_Y = \gamma_{SG} - \gamma_{SL}$$9 and $\theta_{FS}$0 for an ethanol droplet with contact angle $\theta_{FS}$1 (Nazeri et al., 8 Oct 2025).

PANDA-film implements a top-down alternative. There the distances between reflected red and blue LED spots, $\theta_{FS}$2 and $\theta_{FS}$3, are extracted from a telecentric top view, and the calibrated contact angle is computed by a quadratic regression: $\theta_{FS}$4 The best-focused image is selected from a 2 mm z-stack acquired in 0.25 mm steps using an LED-aware focus score that combines Laplacian-variance sharpness, local contrast, LED-spot compactness, and geometric consistency. On eight calibration substrates and 32 droplets, this top-down method achieved a mean residual of $\theta_{FS}$5, an RMSE of $\theta_{FS}$6, and a mean absolute error of $\theta_{FS}$7 relative to side-view goniometry (Quinn et al., 11 Jan 2026).

These methods show that ACAT is not tied to one fitting doctrine. Polynomial tangents, local-line/tangent geometry, Bashforth–Adams profile inversion, and top-down reflectance regression are all viable, provided the imaging geometry, calibration, and error model are matched to the substrate and protocol.

4. Anisotropy, curved geometries, and dynamic wetting

The need for automation becomes most evident on substrates where the angle is not a single static observable. On structured surfaces featuring ramps, pyramids, and staggered cubes, all structures caused significant azimuthal variation in $\theta_{FS}$8. High apparent-contact-angle directions corresponded to shorter contact-line diameters, and low-angle directions to longer diameters. For untreated surfaces the angular variations were larger than for polymerized surfaces, and for staggered cubes the untreated angular spread $\theta_{FS}$9 reached about $\theta_Y$0 (Palmetshofer et al., 2024). Ramps exhibited strong asymmetry because pinning at ramp tops on the “high side” generated the highest peaks, while pyramids showed pinning along both structure axes and diagonals, and staggered cubes exhibited preferred pinning along primary groove-like axes with secondary peaks at $\theta_Y$1 and $\theta_Y$2 (Palmetshofer et al., 2024). An ACAT intended for such substrates must therefore resolve angular maps rather than report a single scalar without qualification.

Dynamic wetting in microchannels introduces another class of nontrivial behavior. In curved microchannels fabricated in PMMA and sealed by a TPU membrane, an automated top-view fluorescent image-analysis procedure tracked the moving interface and measured dynamic contact angles for water and a 50 wt% glycerin/water mixture. Interface velocity derived from the automated tracking agreed with theoretical values to within 5%, and dynamic contact-angle errors relative to manual measurements were below 2% (Zhang et al., 2024). The comparison with molecular kinetic theory showed that dynamic wetting could be described well by MKT even in highly curved microchannels, and the measured behavior depended strongly on channel geometry and curvature (Zhang et al., 2024). A practical implication is that ACAT designs for seal-joint reliability or embedded channels need access not only to $\theta_Y$3, but also to interface position, local contact-line velocity, and local capillary number,

$\theta_Y$4

all as functions of space and time (Zhang et al., 2024).

Acceleration introduces a further limitation. In oscillating menisci, traditional interface models were shown to fail at large interface and contact-line accelerations, particularly for low-surface-tension HFE7200, because inertial deformation of the central meniscus was falsely absorbed into apparent changes in contact angle. The proposed inertia-corrected model added an explicit acceleration-dependent pressure term and reduced regression errors substantially; for HFE7200, once that inertial correction was included, the dynamic angle remained broadly compatible with steady $\theta_Y$5 descriptions, whereas for water stick-slip and hysteresis prevented a unique mapping between $\theta_Y$6, $\theta_Y$7, and acceleration (Fiorini et al., 2022). This constrains ACAT operating regimes: under strongly transient forcing, frame rate, interface model, and inference method must be chosen together.

A more general caution concerns prediction. Classical Wenzel and Cassie–Baxter relations assume equilibrium and do not account for the rugged free-energy landscapes of textured substrates. On pillared surfaces, Monte Carlo plus string-method calculations showed that the Cassie–Baxter state had only one free-energy minimum, while wet states could present multiple minima; when surface roughness decreased, the amount of local minima in the free-energy profile increased (Silvestrini et al., 2021). For ACAT interpretation, this means that measured “static” angles on rough surfaces may encode metastability and protocol history rather than a unique equilibrium property.

5. Calibration, uncertainty, and surface free energy

A technically mature ACAT needs metrological traceability. The low-cost goniometer of Schuster et al. provides a concrete template. Its optics were first characterized for pincushion distortion using stacked glass slides viewed in orthogonal orientations; a ShiftN correction factor of $\theta_Y$8 produced nearly equal reference distances in both directions (Schuster et al., 2018). Contact-angle calibration then used four steel hemispheres with theoretical angles of $\theta_Y$9, $\theta_{FS} \approx 75^\circ \pm 5^\circ$0, $\theta_{FS} \approx 75^\circ \pm 5^\circ$1, and $\theta_{FS} \approx 75^\circ \pm 5^\circ$2, establishing a validated range from approximately $\theta_{FS} \approx 75^\circ \pm 5^\circ$3 to $\theta_{FS} \approx 75^\circ \pm 5^\circ$4. Across 10 images per standard and three analysis procedures, the maximum absolute error was $\theta_{FS} \approx 75^\circ \pm 5^\circ$5, the maximum standard deviation was about $\theta_{FS} \approx 75^\circ \pm 5^\circ$6, and relative errors were $\theta_{FS} \approx 75^\circ \pm 5^\circ$7, $\theta_{FS} \approx 75^\circ \pm 5^\circ$8, and $\theta_{FS} \approx 75^\circ \pm 5^\circ$9 for the three procedures, respectively (Schuster et al., 2018). For an ACAT, this is a clear calibration doctrine: geometric standards, repeated measurements, automatic outlier rejection, and periodic performance verification.

The same paper also formalizes uncertainty propagation from contact angle to surface free energy through the Owens–Wendt model. With

$\theta_{FS} \approx 120^\circ \pm 5^\circ$0

and

$\theta_{FS} \approx 120^\circ \pm 5^\circ$1

the linearized variable

$\theta_{FS} \approx 120^\circ \pm 5^\circ$2

leads to a two-liquid regression in which slope and intercept yield $\theta_{FS} \approx 120^\circ \pm 5^\circ$3 and $\theta_{FS} \approx 120^\circ \pm 5^\circ$4, respectively (Schuster et al., 2018). The paper derives closed-form uncertainty expressions for $\theta_{FS} \approx 120^\circ \pm 5^\circ$5, $\theta_{FS} \approx 120^\circ \pm 5^\circ$6, and $\theta_{FS} \approx 120^\circ \pm 5^\circ$7, and shows that uncertainty is minimized when the two testing liquids have very different $\theta_{FS} \approx 120^\circ \pm 5^\circ$8 values and one liquid is apolar. Their practical conclusion is explicit: uncertainty is minimized when the testing liquids are an apolar liquid and water (Schuster et al., 2018).

Measured PTFE and POM results illustrate the attainable performance. For PTFE, the total SFE lay in the range $\theta_{FS} \approx 120^\circ \pm 5^\circ$9 with relative error lower than 5.5%; for POM, the SFE lay in the range $\theta_{\text{app}}$0 with relative error lower than 4.3% (Schuster et al., 2018). The polar component of very nonpolar solids carried much higher relative uncertainty, while the total SFE and dispersive part remained comparatively stable. An ACAT that automates multi-liquid testing should therefore report not only $\theta_{\text{app}}$1, but also propagated uncertainty on $\theta_{\text{app}}$2, $\theta_{\text{app}}$3, and $\theta_{\text{app}}$4.

Profile-based angle inference is not the only route. An alternative method reconstructs the sessile-drop profile by numerically integrating the Young–Laplace equations under gravity from measured drop height $\theta_{\text{app}}$5 and volume $\theta_{\text{app}}$6, using them as constraints to solve for the apex radius of curvature. With this method, pure water on Teflon yielded computed contact angles of $\theta_{\text{app}}$7 across $\theta_{\text{app}}$8 drops, while Lucite gave $\theta_{\text{app}}$9, showing consistency across drop sizes (Behroozi et al., 2018). This suggests a complementary ACAT mode in which global observables such as $10\text{–}30^\circ$0 and $10\text{–}30^\circ$1 substitute for local tangent estimation when edge noise or near-contact-line optical distortion is limiting.

6. Integration into self-driving laboratories and adjacent methods

Recent work situates ACAT within larger autonomous experimentation frameworks. RAISE links liquid formulation, droplet transfer, image capture, and Bayesian optimization in a closed loop. Its throughput is approximately one contact-angle measurement per minute, corresponding to about 30 formulations in 90 minutes with 3 replicates each and a projected 600 measurements in a 10-hour shift (Nazeri et al., 8 Oct 2025). Benchmarking against a Ramé-Hart Model 260 goniometer showed closely matching mean contact angles but lower standard deviations: the RAISE pipeline reduced SD by 20.8% on PDMS, 29.1% on PS, and 61.9% on PTFE, with the PTFE improvement statistically significant ($10\text{–}30^\circ$2) (Nazeri et al., 8 Oct 2025). Here ACAT is no longer merely an instrument; it is a measurement node inside a closed-loop optimizer.

PANDA-film extends the same trend into synthesis-characterization coupling. After electrodeposition of polymer thin films in a 40-well array, the platform dispenses $10\text{–}30^\circ$3 DI water droplets, acquires top-down reflectance images, and computes contact angle using the calibrated LED-spot model. In the validation example, three PAMA replicates gave an average contact angle of $10\text{–}30^\circ$4 (Quinn et al., 11 Jan 2026). The platform is explicitly modular and open-source, with software and hardware designs released for replication (Quinn et al., 11 Jan 2026). This suggests a plausible trajectory in which ACAT modules become standard interchangeable components in SDL ecosystems.

At the same time, ACAT is not the only automated wettability modality. Droplet probe AFM measures interaction forces between a microdroplet and a surface with approximately $10\text{–}30^\circ$5 force resolution and can map topographical and chemical heterogeneities with micron resolution (Daniel et al., 2020). That method is presented as complementary rather than substitutive: conventional contact-angle measurements are described as crude and imprecise for spatially heterogeneous surfaces, while droplet probe AFM provides local force information and micron-scale maps (Daniel et al., 2020). In encyclopedic terms, this marks the boundary of ACAT’s domain. ACAT remains the principal platform for high-throughput, image-based wettability metrology, but its most advanced forms increasingly coexist with local-force, top-down reflectance, azimuthal-scanning, and self-driving optimization frameworks.

The present research landscape therefore defines ACAT less as a single device than as a class of automated wettability instruments. Its mature form combines controlled droplet generation, calibrated optics, explicit uncertainty analysis, and protocol-aware interpretation of apparent and dynamic angles. Its emerging form adds azimuthal mapping, top-down reflectance, model-based inversion, self-driving orchestration, and closed-loop optimization.