A novel approach for the evaluation of ice release performance of coatings using static friction measurements

Butyl methacrylate (BMA), hydroxy ethyl methacrylate (HEMA, 97%), 2,2′-azodi(2-methylbutyronitrile) (AMBN, ≥ 98%) were purchased from Sigma-Aldrich. Xylene (≤ 25% ethyl benzene) was purchased from Brenntag Nordic AS. Tetrahydrofurane (THF, ≥ 99.8%) was purchased from VWR. Isophorone diisocyanate trimer (IPDI trimer in 30 wt% butyl acetate, Desmodur Z 4470 BA) and hexamethylene diisocyanate biuret (HDI biuret in 25 wt% 1-methoxypropylacetate/xylene, 1:1, Desmodur N75 MPA/X) were purchased from Covestro AG. Dioctyltin dilaurate (DOTDL, 95%) was purchased from TIB Chemical AG. All reagents were used as received. Penguard Universal (epoxy primer), Sea Lion tie-coat (silicone tie-coat), and Sea Lion Repulse (condensation cured silicone foul release coating) were supplied by Jotun A/S with technical datasheets and application guides available through www.​jotun.​com. Wacker primer G 790 was purchased from Wacker Chemie AG and Sylgard 184 (Dow Corning) from Sigma-Aldrich, with technical datasheets and application guides available through www.​wacker.​com and www.​sigmaaldrich.​com, respectively.

Chrome steel beads (100Cr6, Ø 6 mm) were purchased from Rotek AS. Aluminum panels (1050A, 1.5 mm thickness) were purchased from Erling Freitag AS. Mylar^® polyester film (19 µm thickness) was purchased from Arcon AS.

Gel permeation chromatography (GPC) was carried out at 30°C on a Malvern Omnisec GPC system with a PL-gel 5 µm Mixed – D column (Polymer Labs), THF as mobile phase and with a flow of 1 mL/min. Absolute molar masses were detected using a low- and right-angle light scattering detector (LALS/RALS) with a 640 nm laser source. Values for dn/dc_w were determined using a differential refractive index detector, which was also used for concentration monitoring. Linear polystyrene standards were used for calibration of the light scattering detectors. Water contact angle (WCA) measurements were carried out on a Krüss DSA100 drop shape analyzer with 2 µL drop size of distilled water at 23°C and 50% relative humidity (RH). Stress–strain curves were obtained using a universal testing machine (UTM) from Testometrics Co. Ltd. at a pull rate of 5 mm/min at 23°C according to ASTM D638-14.25 Static and dynamic friction measurements were carried out on an Anton Paar MCR-302 rheometer fitted with a pin-on-disk T-PID/44 tribology system, H-PTD 200 Peltier Hood and P-PTD200/80/I Peltier cooling plate.

For the 2-component polyurethane (2K PU), a polyacrylic polyol was synthesized. Synthesis was carried out in a 250 mL four-necked round-bottomed flask, equipped with a mechanical stirrer, a condenser and thermocouple, with nitrogen flow throughout the reaction.

Xylene (40.0 g) was added, and the temperature was increased to 90°C. A mixture of BMA (80.0 g, 563.0 mmol), HEMA (20.0 g, 154.0 mmol), AMBN (4.7 g, 24.5 mmol), and xylene (60.0 g) was added dropwise over 2 h. Following addition, the reaction mixture was stirred for a consecutive 30 min, after which AMBN (0.3 g, 1.6 mmol) in xylene (3.0 g) was added as chaser. After 90 min, the reaction was cooled to ambient temperature. The resulting 50 wt% polymer solution was used without further purification. GPC data: weight average molar mass (M_w) = 21,440 g/mol, dispersity (Đ) = 2.24.

For all commercial coatings, application was carried out following the technical datasheets and application guides referred to in the “Materials” section. Examples, including application of the 2K PU coating, are given below, and further details about the materials, coating application, and curing conditions can be found in the supplementary information (Table S1). All coatings were mixed and applied at room temperature and 30–50% RH.

Chrome steel beads were used as substrate for the break-away torque experiments. The beads were cleaned by immersion in xylene, dried, and held into position on a flat surface by using a magnet. Finally, the beads were coated with a primer coat (Wacker primer G 790) using a conventional spray gun (1.6 mm nozzle, no. 510 + aircap, approx. 2 bar inlet pressure, Devilbiss Advance HD). The primer coat was cured for 2 h at ambient temperature. Sylgard 184 base resin and Sylgard 184 hardener were mixed (10:1 parts by weight) using a speedmixer (Speedmixer™ DAC 600.1 FVZ) at 2300 rpm for 1 min. Immediately following mixing, the paint was applied onto the primed steel beads using conventional spray (the same apparatus and settings as used for the primer coat). As reference, a flat aluminum panel was coated alongside the beads with the same distance to the paint nozzle. The beads and the aluminum reference panels were cured at ambient temperature for 24 h followed by 80°C for 2 h. After curing, the beads were held in position by the coating itself. The dry film thicknesses (DFTs) were measured with a coating thickness gauge (Positector 6000) according to standard ISO 2360:2017 on the aluminum reference panels, and the DFT was assumed to be the same for the steel beads and the aluminum reference. By assuming that the paint settles equally on the flat panel and the small part of the bead in contact with the ice, measurement of the DFT of the flat metal panel should equal that of the beads.

The foul release coating, Sea Lion Repulse, was prepared by mixing the base formulation (component A) with crosslinker (component B) and catalyst (component C) (A: B: C = 17: 0.7: 0.75 parts by volume) and applied onto chrome steel beads previously coated with a primer- and tie-coat (Penguard Universal primer and Sea Lion tie-coat, respectively) in a similar manner as above.

The 2-component polyurethane (2K PU) coating was prepared by mixing the polyacrylic polyol synthesized as explained in the "Synthesis of polyacrylic polyol" section (200 g of the crude polymer solution composed of 50 wt% polymer in xylene) with DOTDL (0.02 g, 0.027 mol), followed by mixing with an isocyanate curing agent mixture (18.3 g HDI biuret + 27.3 g IPDI trimer) with an overall isocyanate to hydroxyl molar ratio of 1.1: 1. Immediately after mixing, the coating was applied in a similar manner as described for Sylgard 184.

For the ice shear stress test, aluminum panels (75 × 100 × 1.5 mm) were used as substrate. The panels were lightly abraded with P150 sandpaper and washed with xylene. Coating was applied using a bar-applicator, and the DFT was measured for the individual panels using a coating thickness gauge (Table S2). The panels coated with 2K PU and Sea Lion Repulse were cured at 23°C/50% RH for 1 week prior to ice adhesion testing. The panels coated with Sylgard 184 were cured as described above. The mechanical and surface-properties of the coatings are summarized in Table 1. The relationship between the properties reported in Table 1 is not expected to change significantly within the testing temperature range (− 20°C to 23°C) as no phase transitions occur for the polymeric binders of the coatings.

A rotational rheometer (Anton Paar MCR-302) with a pin-on-disk tribology-fixture with three points of contact to the substrate was employed (depicted in Fig. 1). The rheometer was fitted with a Peltier cooling plate to control the temperature of the ice and a cooling hood to control the temperature of the surroundings. The temperature of the ice and air was kept equal during the test. Three equally coated beads were inserted into the tribology fixture sample holder for each measurement. This was done to avoid variation between measurements due to wear of the coating.

Ice surfaces were premade by pouring 6.5 mL distilled water into silicone molds (d = 60 mm, thickness = 2 mm) and covered with a Mylar^® polyester film to obtain planar surface topography and kept at – 18°C for at least 3 h. A premade ice disk (as shown in Fig. 1) was then transferred to the rheometer, kept at the desired analysis temperature, and allowed to equilibrate for 15 min.

A typical break-away torque experiment, exemplified by run 1 in Table 3, was carried out by first lowering the sample beads onto the ice (− 20°C) and stabilizing the normal force. To ensure the same running conditions for all samples, the sample fixture was rotated at a constant speed and normal force (20 rpm, 1 N) for 15 s (running in step) prior to a dwell time (5 min) at rest with constant normal force (0.3 N). During this step, a liquid boundary layer of water is formed which refreezes upon motion stop during the dwell time. The sliding distance was then recorded by logarithmically increasing the torque from 0.1 mN·m to 100 mN·m with 123 points per decimal while maintaining a constant temperature and normal force.

The Stribeck curves were obtained in a similar manner as the break-away torque without a running in- and dwell time-step. The rotational speed was increased logarithmically from 10⁻⁶ rpm to 10³ rpm with constant normal force and temperature. More details of the test protocol can be found in the supporting information.

Ice shear stress was measured using a UTM fitted with a 50 kg load cell and a climate chamber (temperature range − 40 to 200°C) both from Testometrics Co. Ltd. A custom-made fixture was employed, containing a locking position for the coated aluminum test panels with an off-set in a horizontal direction, deflecting the normal force into a shear force pulling a mold wherein ice had been formed on the coated aluminum panel. The molds were 3D printed with polylactic acid (PLA Extrafill, Fillamentum) with an inner diameter of 20 mm, wall thickness 3 mm, and tracks for mounting the wire 3.5 mm from the substrate.

The mold, placed on the coated aluminum panel, was filled with distilled water and frozen at – 18°C for 24 h. Prior to starting the test, the test setup was equilibrated at – 10°C in the climate chamber for 1 h. Each sample (mold with ice on coated aluminum panel) was moved quickly from freezer to the instrument and allowed to equilibrate for 5 min before starting the test.

The mold (and thus water frozen onto the coating) was pulled by a fixed pulley via a wire (Dubro 4-40 pull–pull) attached to the load cell at a constant speed of 10 mm/min until an adhesive break between the ice and the coating occurred, and the peak force was recorded. See Fig. 2 for a depiction of the setup, example of a shear curve obtained using the method, and the PLA mold.

JMP^® statistical software (SAS Institute) was used for the design setup and analysis of results. Normal force, temperature, coating thickness, and dwell time were defined as the factors and break-away torque as the single response. Response surface methodology (RSM) was used to find interactions between the factors. The ranges in which the parameters were varied are based on previous reported values.20,27,28

The custom design function in JMP^® was used for the design setup with the design space given in Table 2. The coating thickness was handled as a discrete numeric value for practical reasons.

A total of 21 experiments were carried out according to the testing variables given in Table 2. Each experiment was performed twice in a random order to avoid systematic errors. The ambient temperature outside the test chamber was constant at 23°C, and the RH in the test chamber was approximately 100% at all test temperatures given in Table 2.

The measurements of break-away torque between ice and three chemically different commercial coatings are presented in Fig. 3 (conditions given in figure text). The figure illustrates how such measurements can be used to compare ice release properties of coatings. The sliding distance between the fixed ice disk and the rotating specimen holder shows initial small movement in the low torque region, followed by a sudden leap in sliding distance once the torque reaches a critical level. The onset of the rapid movement indicates initiation of motion and is defined as the break-away torque. A different degree of sliding distance prior to the break-away point is observed for the three coatings as an effect of the differences in elastic modulus and thickness of the coatings. The onset point is used as a measure of ice release property of the coatings throughout the work and was found as the intersection of the tangent lines before and after motion was initiated (as shown in Fig. 3). Sea Lion Repulse has a break-away torque around 0.43 ± 0.06 mN·m, indicating superior ice release as compared to Sylgard 184 and 2K PU reaching this point at 5.7 ± 0.4 mN·m and 15.34 ± 1.7 mN·m, respectively (from three measurements). Testing of all three coating systems was carried out with a start torque of 0.1 mN·m. For Sea Lion Repulse, testing was also carried out with a start torque of 0.05 mN·m (used for depiction in Fig. 3) as it turned out that the onset point was close to 0.1 mN·m. Thus, 0.05 mN·m resulted in more precise break-away torque values. Sea Lion Repulse has a different curve development than the two other coatings, with a less sharp transition from stationary to motion because of the low break-away torque and modulus of the coating. Sylgard 184 and the 2K PU are both less elastic and have a significantly higher break-away torque than Sea Lion Repulse, which results in a sharper transition between stationary and motion.

The motivation for using a design of experiments approach was to determine the impact of the measurement parameters on the break-away torque, and to indicate which parameter settings resulting in least variation of repeated measurements. Only the coating system with Sylgard 184 was used in these experiments for which parameter settings, as proposed by the JMP software and explained in the "Experimental design" section, are given in Table 3.

In the same table, break-away torques are given as the average of the two measurements carried out per experimental parameter, and standard deviation (SD) of those measurements. Prior to preparing a predictive model from the results, a jackknife analysis was carried out to determine potential outliers.29 Three measurements were identified as potential outliers and thus excluded from the model, resulting in 39 datapoints.

The standard least squares linear regression in JMP resulted in the following linear regression model:where y is the break-away torque, T the temperature, F the normal force, t the dwell time and CT the coating thickness. The parameter terms (T, F, t and CT) are scaled as shown in equation 2 where the levels H (high level) and L (low level) correspond to the low and the high values of the parameter, e.g., −5°C and − 20°C, respectively, for the temperature-parameter and “parameter” is the actual value.

The summary of fit produces a high coefficient of determination (R²) and low root mean square error (RMSE), indicating that the model is suitable for predicting break-away torque in the design space (Table 4), i.e., within values given in Table 3 for the Sylgard 184 coating system.

Temperature, normal force, freezing time, and coating thickness were all found to be significant factors with a P value < 0.05 (Student’s t-test, null hypothesis is that the true value for each parameter is zero). In Fig. 4, the effect of each significant parameter is plotted as the -log (p-value) visualizing the relative parameter impact on break-away torque.

As directly seen from the effect plot, the applied normal force has the highest contribution on the break-away torque. This can be explained by the increased contact area between the coating and ice when increasing the normal force. The contact area between a sphere and a planar surface can be expressed as follows30where α is the radius of the circular contact area, W the applied normal force, R′ the radii of curvature, and E′ the reduced Young's modulus, which is a function of the Poisson’s ratio and Young's moduli of the contacting bodies. From equation (3), it follows that both increased normal force and reduced Young's modulus will increase the area of contact and the effect plot in Fig. 4 indicates that normal force is the most important parameter to control when measuring break-away torque.

Figure 5 gives a visualization of parameter interactions from the linear regression model (equation 1) as predicted by the JMP software. The y-axis of each square shows how the break-away torque is dependent on the main parameters: temperature, normal force, dwell time, and coating thickness.

The break-away torque is significantly lower at temperatures closer to – 5°C and high normal force than at lower temperatures and high normal force (Figs. 5a and 5d). This indicates more water in the boundary for the temperatures closer to – 5°C than for the experiments carried out at temperatures closer to – 20°C. There is also a significant increase in break-away torque for long dwell times and low temperatures which again indicates more water at the boundary with increasing temperature of the ice (Figs. 5b and 5g). Longer dwell times will allow more of the boundary water to freeze and thus increase the break-away torque.

The break-away torque increases with increasing coating thickness and dwell times (Figs. 5i and 5l). Sylgard 184 has a Young's modulus between 1.32 and 2.97 MPa depending on curing conditions26 and is a soft material with glass transition temperature well below the temperatures used in this test. The increased coating thickness of the relatively soft Sylgard 184 will increase the real contact area to the ice and thus give increased break-away torque. This is in alignment with the findings of Blackford et al.20 that rigid materials (steel) have lower static friction than the lower elastic modulus materials (polymethylmethacrylate) on ice. The bulk modulus of Sylgard 184 on steel substrate (beads) will be affected by the coating thickness. However, it is interesting to note that when measuring ice shear strength on a planar surface of an elastic coating, the ice adhesion is reduced with increased coating thickness because of the deformation of the coating when stress is applied, and ice more easily shed off the surface.28

The reproducibility of the model obtained from the statistical software was evaluated by using replicated points as part of residual error (lack of fit test). Good reproducibility is indicated by p-value = 0.4845 (> 0.05, i.e., there is no significant lack of fit).

The lowest SD is found for measurements with normal force ≤ 1.15 N (Fig. 6). Measurements at normal force > 1.5 N have less reproducibility. The data from which the model is constructed only have few datapoints in this region as some datapoints were identified as outliers, using the jackknife analysis as described above, and thus removed thereby also resulting in lower reproducibility > 1.5 N. The large variation in the experiments carried out at high normal force might be explained by inconsistent amounts of water in the ice-coating boundary. The Hertzian contact pressure cannot be directly applied for the sample specimen as the elastic modulus of the thin coating films will be affected by the steel substrate, though it is evident that increased normal force affects the reproducibility of the measurements.

The linear regression model (equation 1) obtained from the data in Table 3 was validated by a set of separate experiments where the parameters were chosen randomly within the design space. Ninety-two percent of the experiments fit with the predicted values (95% confidence interval encompassing variation from predicted value and individual measurements). As expected, the experiments that did not match the predicted values were those carried out at high normal force. Further details about experimental parameters and break-away torques can be found in Table S3 in the supporting information.

Stribeck curves offer insight into the frictional forces between two surfaces from the boundary lubrication through the hydrodynamic lubrication regime. The frictional forces are usually high because of direct contact between the asperities of two solids at low sliding speeds. Frictional heat and pressure melting at higher speeds form a lubricating layer of liquid water resulting in reduced friction between the surfaces.31 Foul release coatings typically contain some extent of free silicone oils that lubricate the surface. The presence of a lubricant oil in the ice-coating interphase might also result in reduced friction and adhesion, as previously shown for slippery liquid infused porous surfaces.6 The dynamic friction was therefore measured to investigate potential lubrication from liquid water and/or silicone oils.

In Fig. 7, the friction factor µ is plotted as a function of sliding speed under the same conditions for the three coatings. As seen from Fig. 7, the data do indeed confirm that the Sea Lion Repulse coating (the foul release coating) has much lower friction up to 10⁻³ m/s compared to the two other coatings. This corresponds well with the extra lubrication from free silicone. In the high-speed domain (> 10⁻³ m/s), however, a significant increase in friction is observed for the Sea Lion Repulse only.

Mielonen et al.24  observed higher sliding friction for soft materials than for harder materials (3.2 x 10⁻² m/s sliding speed). As Sea Lion Repulse is a much softer material than the two other coatings, this explains the large increase in friction above 10⁻³ m/s sliding speeds. Sylgard 184 demonstrates a change from boundary lubrication to mixed lubrication/hydrodynamic lubrication for sliding speeds above 10⁻³ m/s.

It was only one run per dynamic friction measurement per sample due to wear of the samples. Replicate runs for samples coated via exactly the same approach and DFT showed the same trend. A few trials were conducted prior to the measurements where the test was stopped prior to rapid increase in friction (at approximately 10⁻³ m/s sliding velocity), and samples were evaluated under optical microscope. Up to this point, no macroscopic wear was observed. This also confirmed that no cohesive fracture or adhesive fracture between coating and steel-bead took place at these sliding velocities.

Test conditions in the range that resulted in least variation in the break-away torque for Sylgard 184 (experiments with normal force < 1.15 N in Table 3) were used to compare break-away torque between Sylgard 184, a 2K PU coating and Sea Lion Repulse. The ice adhesion strength was also measured for these coatings by a conventional shear stress method, and Fig. 8 compares the break-away torque (mN·m) and shear strength adhesion (kPa).

As can be seen from the comparison of break-away torque and shear stress curves, the scaled SD of the shear strength measurements are three-fold greater than the scaled SD of the break-away torque measurements (for details, see Table S4 in the supporting information). Importantly, however, the trends of both test methods are similar with Sea Lion Repulse revealing the best ice release properties, followed by Sylgard 184 and the 2K PU coating with highest ice adhesion value. The coatings with low elastic modulus (Sea Lion Repulse and Sylgard 184) show lower ice adhesion in both tests compared to the 2K PU coating. As seen from the WCA results given in Table 1, it is moreover indicated that the former two coatings are more hydrophobic than the latter, which should also result in lower ice adhesion. However, the elasticity modulus and surface roughness must also be considered for a direct comparison.31 The previous literature demonstrates that a large difference in elastic modulus between the ice and the coating, e.g., silicone rubbers, gives a reduction in ice adhesion strength.32,33 This is also the case for the observed static friction measurements in Fig. 8.

The comparison of the developed break-away torque method and a conventional ice shear test for the three commercial coatings showed the same trend for break-away torque and shear strength. However, the uncertainty between Sylgard and the 2K PU coating in the ice shear test is so large that they can appear constant for this test method. For the break-away torque method, the uncertainties are much lower and indicate large difference between those two coatings. Both tests are laboratory screening tests as they measure ice adhesion on bulk-water ice. The static friction method measures the friction force (acting opposing to the shear force applied) necessary to initiate macroscopic motion between the contact-points. In the ice shear stress method, a shear force is applied on the mold of ice, where the obtained value might be affected by the force distribution on the mold itself or throughout the ice. Another obvious difference is the contact area where break-away torque is measured on a relatively small area, whereas the shear strength method measures ice adhesion on relatively large surface areas. The contact between the ice and coating in the developed method is different from ice prepared directly on the substrate, such as bulk-water ice, precipitation ice or in-cloud ice.14