Low–interfacial toughness materials for effective large-scale deicing

See allHide authors and affiliations

Science  26 Apr 2019:
Vol. 364, Issue 6438, pp. 371-375
DOI: 10.1126/science.aav1266

Easy ice removal

The accumulation of ice on a surface can lead to hazardous conditions, such as on the surface of an airplane wing or the side of a tall building. Ice adhesion, even to a surface treated to minimize the bonded force, will usually depend on the amount of surface coverage. Golovin et al. compared strength-limited deicing with toughness-limited deicing. Whereas normal deicing materials focus on minimizing the adhesion strength, the authors show that if a material is designed with low-adhesion toughness, deicing is no longer a function of the coverage area.

Science, this issue p. 371


Ice accretion has adverse effects on a range of commercial and residential activities. The force required to remove ice from a surface is typically considered to scale with the iced area. This imparts a scalability limit to the use of icephobic coatings for structures with large surface areas, such as power lines or ship hulls. We describe a class of materials that exhibit a low interfacial toughness with ice, resulting in systems for which the forces required to remove large areas of ice (a few square centimeters or greater) are both low and independent of the iced area. We further demonstrate that coatings made of such materials allow ice to be shed readily from large areas (~1 square meter) merely by self-weight.

The accretion of ice on surfaces can have a severe detrimental impact on a range of commercial and residential activities (1). Consequently, there has been an effort to create coatings that protect against the buildup of ice. Typically, the efficacy of these coatings has been evaluated by measuring the force, F, to debond a specified area, A, of ice, and defining an ice adhesion strength τice = F/A as the characteristic property for the system (2). The term icephobic is generally used to describe surfaces for which τice < 100 kPa (3, 4), in comparison to structural materials such as aluminum and steel, for which τice > 1000 kPa (2, 5, 6).

Using τice to characterize an interface inevitably requires that the force necessary to remove ice scales with the iced area. Many engineering structures susceptible to icing, such as airplane wings, wind turbine blades (7), and boat hulls (8), have surface areas that can approach thousands of square meters. Consequently, even with extremely icephobic coatings, structures with large surface areas would require prohibitively high forces to detach entire sheets of ice from the surface. In this work, we developed low–interfacial toughness (LIT; interfacial toughness Γ < 1 J/m2) materials for which the force required to remove adhered ice from large areas (few cm2 or greater) is both low and independent of interfacial area.

An interfacial cohesive strength, as represented by the ice adhesion strength, is one way to describe the bonding across an interface (9). A countervailing perspective on fracture (10, 11) is that an interface should be described in terms of its bonding energy (or, more correctly, its toughness). Further, although the work of adhesion is often discussed in connection to ice adhesion, it is the strength that is generally used to describe failure (1, 2, 4). This is true whether the adhesion is viewed in terms of surface energy (3, 12, 13), interfacial cavitation (14), or lubrication (1517).

The two competing perspectives for delamination, strength and toughness, can be rationalized by means of cohesive zone models of fracture (1822). Simple analytical models (23) can be used to demonstrate that the shear strength of the interface, τ^, controls delamination when the length of the interface is relatively small, so that τice=τ^. This is manifested by a spontaneous rupture along the entire interface (movie S1). Conversely, Γ controls delamination when the length of the interface is relatively large. This is manifested by the propagation of an interfacial crack (movie S1). The analysis shows that there is a critical bonded length at which a transition between the two modes of failure occurs, given by Lc=2EiceΓh/τ^2, where Eice is the modulus of ice (~8.5 GPa) of thickness h. In this context, ice is not a ductile material even close to its melting temperature, and treating it as an elastic solid, at the engineering strain rates and time scales relevant here, is a reasonable approach (11) provided that 2EiceΓ/σY2<h<L, where σY is the yield strength of ice. Note that when L > Lc, the force required to delaminate the ice is constant, no matter how large the interface may be.

We first verified the concept that the force required to remove an ice layer reaches an asymptotic value if the interface is long enough. This was done using substrates made from common plastics such as polyethylene, polypropylene, and polystyrene (substrate thickness t = 1.6 mm; see Table 1) without any additional modification. We used a setup similar to those reported previously (2, 14, 24) but instead of using relatively short lengths corresponding to a few millimeters of bonded ice (2, 14, 2426), we designed our apparatus so that much longer interfaces could be evaluated (Fig. 1, C and D, insets) (23). Plots showing the force (per unit width) necessary to detach the ice, F˜ice, as a function of the bonded length, L, are shown in figs. S7 to S12. For the most part, we observed that F˜ice increased proportionally to L only when L was small. Beyond the transition length Lc, no additional force was necessary to dislodge the ice. The corresponding asymptotic force, F˜icecr, can be used to determine the interfacial toughness, Γ, from Γ=(F˜icecr)2/2Eiceh (2729). Specific examples of this behavior are shown in Fig. 1A for ice bonded to four different plastic substrates, each of which has a transition length less than 10 cm.

Table 1 Values for interfacial properties measured between ice and 20 different surfaces.

A comparison is made between 1-mm-thick coatings of icephobic silicones on aluminum, 1.6-mm-thick plastic substrates, and 1- to 2-μm-thick LIT coatings on aluminum [see (23) for fabrication and composition descriptions]. F˜icecr for pure van der Waals interaction was calculated using Γ = 0.1 J/m2 (5, 34) for a 6-mm-thick ice sheet. Data uncertainty denotes 1 standard deviation (N ≥ 5). See (23) for long forms of abbreviations.

View this table:
Fig. 1 Strength- versus toughness-controlled fracture.

(A) The force per unit width required to debond ice from four polymers (each 1.6 mm thick). Up to a critical length Lc, the shear strength of the interface, τ^, controlled the fracture of ice from these systems. However, after Lc, no additional force was necessary to remove the adhered ice. (B) The force per unit width required to debond ice from four silicones (thickness ~1 mm) (23). In all cases the fracture was controlled by the adhesive strength up to L = 20 cm, and no toughness-controlled fracture was observed. (C) The force per unit width required to debond ice from silicone B and polypropylene as a function of interfacial length. For polypropylene (thickness = 1.6 mm) (23), the force increased linearly with the length of ice until Lc = 3.6 cm, after which no additional force was required to remove the accreted ice. For silicone B (thickness ~1 mm), strength always controlled the fracture even up to 100 cm. The inset shows a schematic of the situation being investigated. (D) Data from (C) recast in terms of the apparent shear strength τice, indicating that τice for silicone B is less than that of polypropylene only when L ≤ 50 cm. The inset shows our experimental setup, with 11 pieces of ice of eight different lengths adhered to silicone B. All experiments shown were conducted at –10°C. Error bars denote SD (N ≥ 5).

We repeated this experiment for aluminum substrates coated with different icephobic coatings (t ≈ 1 mm; Fig. 1B). These coatings were all based on polydimethylsiloxane (PDMS) rubber, which has been studied for its low–ice adhesion properties enabled via lubrication (25), interfacial cavitation (14, 24), low surface energy (30, 31), and interfacial slippage (32, 33). In contrast to the plastic substrates, these materials did not exhibit a toughness-controlled regime of delamination within the range of bonded lengths studied.

A low interfacial shear strength does not necessarily imply a low toughness. Thus, a material that debonds from ice more readily if the interface is short does not necessarily debond more readily if the interface is long. This can be seen by comparing the results for polyvinyl chloride (PVC) and polyamide (Fig. 1A) or the results for polypropylene and silicone B (Fig. 1, A and B). The shear strength of the interface between ice and polypropylene can be calculated from the initial slope of the line in Fig. 1A as τ^ = 320 ± 40 kPa. The shear strength of the interface between ice and the silicone B coating is an order of magnitude lower, being equal to τ^ = 29 ± 2 kPa. However, the force for detachment continually increases for silicone B, even out to 100 cm (Fig. 1C). For interfaces longer than 50 cm, the ice is removed more easily from polypropylene (Γ = 1.9 J/m2) than it is from the silicone B coating (Γ > 9 J/m2).

These data can be reexpressed in terms of the apparent ice adhesion strengths for the two interfaces by dividing the force by the initial bonded area (Fig. 1D). As such, the apparent ice adhesion strength, for a length of 100 cm for polypropylene (τice ≈ 12 kPa), was less than half that of the icephobic PDMS (τice ≈ 29 kPa), although the true ice adhesion strength τ^ was an order of magnitude greater (Fig. 1D). Over the past decade, achieving τice < 15 kPa has necessitated the use of either soft rubbers (14, 24) or highly lubricated systems (25, 26, 32), which can suffer from poor durability (14). However, here we show that one can obtain much lower values of apparent ice adhesion strength for large structures by selecting a material with a low toughness.

There are several contributions to interfacial toughness. One is associated with the bonding energy between the ice and the coating. A lower bound on this energy would be ~0.1 J/m2, corresponding to van der Waals interactions (5, 34). An additional contribution could come from localized losses within the coating, associated with the high-stress region at the crack tip. These two effects would be classically considered to be contributing to interfacial toughness. However, if the process of delamination causes deformation of the coating, then the strain energy associated with this deformation must also be considered as a contribution to the effective toughness between the ice and substrate. From a cohesive zone perspective, one can consider the toughness of an interface to be given by the area under the force displacement curve of the entire interface, including the coating (18, 19). Therefore, assuming linear elasticity, this contribution to the toughness can be estimated as Γτ^2t/2G, where G is the shear modulus and t is the thickness of the coating (23) (fig. S6). Consequently, it should be possible to minimize the effective toughness by minimizing the thickness of a polymeric coating.

To investigate this concept, we varied the thickness of PVC films and confirmed that Γ scaled with the coating thickness t (Fig. 2A). Lowering t from 150 μm to 2 μm (23) reduced Γ from ~3 J/m2 to 2 J/m2. Previous work (14, 35) has shown that for icephobic elastomers, τ^t1/2. Therefore, the design of LIT (Γ < 1 J/m2) materials can be substantially different from the design of icephobic materials. Icephobic surfaces can be made more effective as t increases, whereas LIT surfaces become more effective as t decreases.

Fig. 2 Controlling interfacial toughness.

(A) The effect of coating thickness on the effective interfacial toughness between ice and an aluminum substrate coated with plasticized PVC for four different contents of the plasticizer MCT. (B) The force per unit width required to fracture ice from three different thicknesses of PVC plasticized with 50 wt % MCT. Note that for the thickest sample, strength controlled the fracture up to at least L = 20 cm. (C) The force per unit width required to fracture ice from thin (t ≈ 1 to 2 μm) PVC coatings with four different contents of MCT. A toughness-controlled regime of fracture was always observed for lengths less than 20 cm. (D) The effect of plasticizer content on Γ for three different thicknesses of plasticized PVC. All experimental results shown were obtained at –10°C. (E) The force required to fracture ice from the LIT PDMS and LIT PVC systems (thickness ≈ 1 to 2 μm). Even over an interfacial length of 1 m, the necessary force of fracture remained constant beyond Lc. The inset shows our experimental setup, performed in a walk-in freezer at –10°C. Error bars denote SD (N ≥ 5).

To further reduce Γ, we explored the effects of plasticizing PVC with medium-chain triglyceride oil (MCT) (23). Figure 2A shows the general drop in toughness observed with increased plasticization. As shown in Fig. 2B, the additional drop in shear strength associated with the addition of 50 weight percent (wt %) MCT was large enough for the transition length to become too long for the toughness to be measured for the thicker coatings. The general trends among strength, toughness, coating thickness, and level of plasticization can be seen in Fig. 2, C and D.

By optimizing the thickness and plasticizer content within the PVC, we fabricated LIT materials exhibiting Γ as low as 0.27 J/m2 (F˜icecr = 52 ± 7 N/cm). Similarly, by optimizing the thickness and plasticizer content, we fabricated LIT polystyrene (20 wt %) diisodecyl adipate; Table 1 and fig. S18] (23) with Γ ≈ 0.43 J/m2 (F˜icecr = 66 ± 6 N/cm) and LIT PDMS (40 wt % silicone oil; Table 1) that displayed an even lower Γ ≈ 0.12 J/m2 (F˜icecr= 35 ± 4 N/cm).

We coated 1.2-m-long aluminum beams with these LIT PVC and LIT PDMS coatings (with a nominal thickness of t ≈ 1 to 2 μm) (23), and conducted large-scale testing inside a walk-in freezer at –10°C. Figure 2E shows that the force of detachment did not increase for L > Lc, even over 1 m of interfacial length (F˜icecr = 52 ± 7 N/cm for LIT PVC; F˜icecr = 35 ± 4 N/cm for LIT PDMS). These correspond to τice values of <6 kPa and <4 kPa for the LIT PVC and LIT PDMS surfaces, respectively. In contrast, for aluminum coated with an extremely soft, icephobic PDMS rubber (plasticized silicone B, τ^ = 12 kPa), we measured F˜ice = 126 N/cm at L = 100 cm. Additional experiments confirmed that we were not observing delamination (fig. S3) or cohesive failure (fig. S5) of the LIT coatings during ice removal.

For a given ice thickness, there will always be an interfacial length beyond which LIT materials require less force than icephobic materials to remove the adhered ice (fig. S19). As an example, to mimic the deicing of power line cables, we conducted off-center loaded beam tests by flexing 1.2-m-long uncoated and coated (with either icephobic or LIT coatings) aluminum beams with ice adhered on one side (23). The icephobic (silicone B; t ≈ 1 mm) and LIT (silicone B + 40 wt % silicone oil; t ≈ 1 to 2 μm) coatings were fabricated using the same polymer, PDMS, but the icephobic PDMS system exhibited low interfacial strength (Γ > 8.8 J/m2 and τ^ = 30 kPa), whereas the LIT PDMS exhibited low interfacial toughness (Γ = 0.12 J/m2 and τ^ = 115 kPa). Upon flexing, ice fractured cleanly from the LIT PDMS at a low deflection of 2.4 cm from the center of the beam (Fig. 3A). Both the uncoated and icephobic-coated beams displayed limited signs of ice detachment even at an extreme deflection of ~35 cm (movie S2). To mimic the deicing of an airplane via wing tip deflection, we conducted end-loaded cantilever beam tests (fig. S20 and movie S3). The deflection necessary to remove the ice adhered to the LIT coating was an order of magnitude less than that of the icephobic or uncoated surfaces. The LIT coating also enabled the removal of ice from complex geometries, such as an ice cube tray (fig. S17 and movie S4), with little to no force.

Fig. 3 Large-scale testing of LIT materials.

(A) Comparison of uncoated, icephobic, and LIT aluminum beams adhered to a sheet of ice (100 cm × 2.5 cm × 0.8 cm) undergoing off-center load flex tests inside a walk-in freezer held at –20°C (23). Ice fractured from the LIT-coated specimen with a remarkably low apparent ice adhesion strength of 0.39 kPa, whereas ice remained adhered to the uncoated aluminum and icephobic specimens even at severe deflections (movie S2). (B), An aluminum sheet coated with LIT PDMS before, during, and after fracture from a large sheet of ice (0.95 m × 0.95 m × 0.01 m). The weight of the ice sheet alone was sufficient to cause fracture, displaying an exceedingly low apparent ice adhesion strength of 0.09 kPa. A comparison is also made to uncoated aluminum (movie S5).

The isothermal freezing conditions within a freezer (Fig. 1, C and D, and Figs. 2E and 3A) differ from those experienced in Peltier plate–based systems (data shown in Figs. 1 and 2), in which ice is formed via unidirectional cooling from the surface. Ice formation conditions, particularly ambient temperature, can substantially affect the structure and interfacial properties of ice (1, 3638). The similitude of our data for L ≤ 20 cm (Peltier) and L > 20 cm (freezer; see Fig. 1, C and D, and Figs. 2E and 3A) lengths of ice indicated that LIT materials can be effective in shedding ice in different ice formation conditions. Additionally, we measured τ^, F˜icecr, and Γ values for polypropylene and LIT-PDMS at –20°, –10°, and –5°C. The values of these interfacial properties (fig. S4), for both LIT PDMS and polypropylene, appear to be invariant with temperature and the different ice formation conditions within the ranges studied.

To evaluate a third ice formation condition, we coated a 1 m × 1 m aluminum panel with our LIT PDMS and allowed ice to form outdoors at –7°C overnight (23) (Fig. 3B). We observed that the weight of the ice at a thickness of 1 cm, once fully frozen, was enough to completely and cleanly remove the attached ice (movie S5). This yielded τice = 0.09 kPa. Whereas varying icing conditions can lead to tensile cracking and subsequent fragmentation of the ice, as long as the fragmented length remains greater than Lc, LIT materials will remain an effective means of ice removal.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S20

Table S1

References (3947)

Movies S1 to S5

Data S1

References and Notes

  1. See supplementary materials.
Acknowledgments: Funding: We thank K.-H. Kim and the Office of Naval Research for support under grant N00014-12-1-0874; K. Caster and the Air Force Office of Scientific Research for support under grant FA9550-10-1-0523; and NSF and the Nanomanufacturing program for supporting this work through grant 1351412. K.G. was supported by a National Defense Science and Engineering Graduate Fellowship (U.S. Department of Defense) and Natural Sciences and Engineering Research Council of Canada grant RGPIN-2018-04272. Author contributions: K.G. and A.D. designed and performed all experiments and wrote the manuscript; M.D.T. and A.T. conceived the research, designed experiments, and wrote the manuscript. Competing interests: The University of Michigan has applied for a patent based on this technology. A startup company, HygraTek LLC, has licensed this technology from the University of Michigan. A.T. is the chief technology officer of, and has been a paid consultant for, HygraTek LLC. Data and materials availability: All data, including the data used to create the figures, are available in the main text or the supplementary materials.

Stay Connected to Science

Navigate This Article