**Spectroscopic and Structural Analyses of** *Opuntia Robusta* **Mucilage and Its Potential as an Edible Coating**

**Aurea Bernardino-Nicanor 1, José Luis Montañez-Soto 2, Eloy Conde-Barajas 1, María de la Luz Xochilt Negrete-Rodríguez 1, Gerardo Teniente-Martínez 1, Enaim Aída Vargas-León 1, José Mayolo Simitrio Juárez-Goiz 1, Gerardo Acosta-García <sup>1</sup> and Leopoldo González-Cruz 1,\***


Received: 28 November 2018; Accepted: 12 December 2018; Published: 15 December 2018 -

**Abstract:** Mucilage extracted from the parenchymatous and chlorenchymatous tissues of *Opuntia robusta* were obtained using water or ethanol as the extraction solvent. The changes in the different tissues by using different extraction solvents were evaluated via scanning electron microscopy (SEM) and Fourier transform infrared (FT-IR) and Raman spectroscopy; in addition, the effect of mucilage coating on the various quality characteristics of tomato (*Lycopersicum sculentum*) was evaluated. The SEM results showed that the mucilage extracted from the parenchyma had a higher aggregation level that the mucilage extracted from the chlorenchyma. The presence of three characteristic bands of pectic substances in the FT-IR spectra between 1050 and 1120 cm−<sup>1</sup> indicated that the mucilage extracted from the parenchymatous tissue had a higher content of pectic compounds than the mucilage extracted from the chlorenchymatous tissue. It was also observed in the Raman spectra that the level of pectic substances in the mucilage extracted from the parenchymatous was higher than that in the mucilage extracted from the chlorenchymatous tissue. The mucilage extracted from the parenchymatous tissue was more effective as an edible coating than the mucilage extracted from the chlorenchymatous tissue. Tomatoes covered with mucilage showed significantly enhanced firmness and reduced weight loss. The uncoated tomatoes showed higher lycopene content than the coated tomatoes on the 21st day. This study showed that the *Opuntia robusta* tissue and extraction solvent influence mucilage characteristics and that *Opuntia robusta* mucilage is a promising edible coating.

**Keywords:** mucilage; *Opuntia robusta*; shelf life; tomatoes

#### **1. Introduction**

The tomato (*Lycopersicum esculentum*) is a climacteric fruit with a relatively short postharvest life during which softening and textural changes occur. In addition, many components of biochemical pathways involved in pigmentation, carbohydrate metabolism and ethylene biosynthesis are also modified [1]. Postharvest practices can have a significant effect on tomato sugar content; however, physical characteristics are the most important factor for the final consumer, and fresh tomato quality is

determined by appearance, color, firmness and flavor. On the other hand during maturation, the weight loss of the tomato is increased by the moisture effect, solute movement, and water loss [2]. For this reason, diverse studies have been focusing on the control of tomato ripening through of the use of modified atmospheres, temperature and humidity control, hypobaric storage and edible coatings [3,4].

The use of edible coatings has been considered an alternative in response to the increasing demand for fresh and minimally processed tomatoes, enhancing their shelf life. The application of edible coatings in tomatoes is a promising technology since these coatings can act as a vehicle for the incorporation of fungicides and antimicrobial agents [3,5], reduce the rate of color change, and inhibit ethylene production [6]; they can also act as barriers to water loss and gas exchange, enable the controlled release of bioactive compounds, and enhance the antioxidant activity and total phenolic contents [7].

In the development of edible coatings, a broad range of edible biopolymers has been used, such as mucilage, starches, cellulose derivatives, pectin, carrageenan, chitosan/chitin, sucrose, gums, sodium alginate, proteins (animal or plant-based) and lipids [8]. The use of edible coatings has diverse advantages compared with synthetic polymers; edible coatings endow food products with a natural appearance, reduce the environmental impact, and are nontoxic, economical and easily available in the environment [9]. In addition, edible coatings offer biocompatibility and can be used as additive carriers of colorants, flavors, antioxidants or antimicrobials.

The plants of the *Opuntia* genus have high concentration of polysaccharides, principally mucilages that contain arabinose, galactose, galacturonic acid, rhamnose, and xylose [10], on the other hand, Rocchetti et al., [11] reported the presence of high content of β-glucans in the *Opuntia ficus-indica* cladodes, that are capable of forming gels in water, for this reason, the mucilaginous compounds of *Opuntia robusta* have been used in the development of foodstuffs [12]; in addition, it has been observed that the mucilage from nopal also has the ability to form edible coatings [13] and has been used to develop an edible coating that increases the shelf-life of strawberry [14].

However, few studies have been conducted using Raman, FT-IR and scanning electron microscopy (SEM) methods to compare mucilage extracted from the parenchyma and that from the chlorenchyma of *Opuntia robusta* or to examine the effect of modifying the extraction solvent. For this reason, in this study, FT-Raman, FT-IR spectroscopy and scanning electron microscopy (SEM) were used to determine the differences between mucilage extracted from the parenchyma and mucilage extracted from the chlorenchyma of *Opuntia robusta* and the effect of the extraction solvent on the structure of the mucilage. The suitability of *Opuntia robusta* mucilage as an edible coating to extend the shelf life of tomato was also evaluated.

#### **2. Materials and Methods**

#### *2.1. Vegetative Material*

*Opuntia robusta* is a shrubby to tree-like cactus originating in central Mexico that can attain a height of 3 m, is a source of fruits as well as young cladodes used for animal feed. Young cladodes from *Opuntia robusta* Wendl var. *robusta* were obtained from Tulancingo, Hidalgo, Mexico. The cladodes were harvested manually at 12:00 p.m. and selected according to size to normalize the state of maturity: 35 cm (large), 30 cm (width) and 4.5 cm (thickness). The samples were stored at 4 ◦C in a refrigerator (LG, Model GR-452SH, LG electronics, Mexico, Mexico) for up to 24 h until used.

#### *2.2. Obtention of Mucilage*

Chlorenchymatous tissue and spongy parenchymatous tissue were removed from the cladodes using a knife, and both tissues were subsequently cut into approximately 2 cm cubes, packed in PVC bags and stored in the freezer (LG Model GR-452SH) until used.

One hundred g of parenchymatous or chlorenchymatous tissue were placed in a flask (500 mL) containing a stirrer and 100 mL of ethanol or water. The mixture was stirred frequently (2 h; 50 ◦C) and then ground in a high-speed blender for 5 min. The mixture was allowed to stand for 1 h and filtered using Whatman (Maidstone, UK) No. 40 filter paper to obtain a fully clarified mucilaginous extract. The mucilaginous extract obtained from *Opuntia robusta* was dried (50 ◦C) until the complete elimination of moisture in a forced convection drying-oven (Binder, Model FD115-UL, Tuttlingen, Germany). Dried mucilage was milled into a fine powder and sieved through a size 40 mesh (425 μm). The mucilage powder was packaged in 25-g glass bottles and stored at 25 ◦C until used.

#### *2.3. Scanning Electron Microscopy*

Samples of *Opuntia robusta* mucilage powder were examined using a JEOL (JEOL, type EX-1200, Tokyo, Japan) scanning electron microscope fitted with a Kevex Si(Li) X-ray detector (Kevex inc, Newark, DE, USA). The analyses were performed under vacuum at an accelerating voltage of 15 kV. The samples were mounted on double-sided carbon tape and covered with approximately 10 nm of gold using a Denton sputter coater (Denton Vaccum LLC, Moorestown, NJ, USA).

#### *2.4. FT-IR Spectroscopy*

The FT-IR spectra of the *Opuntia robusta* mucilage were acquired on a Perkin Elmer FT-IR spectrophotometer (Perkin Elmer, Inc., Waltham, MA, USA) using potassium bromide (KBr) discs prepared from powered samples mixed with dry KBr. The spectra were recorded (16 scans) in transparent mode at a resolution of 4000 to 400 cm<sup>−</sup>1.

#### *2.5. Raman Spectroscopy*

The Raman measurements were performed on a Perkin-Elmer (Perkin Elmer, Inc., Waltham, MA, USA) 2000R NIR FT-Raman Spectrometer equipped with a Nd:YAG laser emitting at a wavelength of 1064 nm and an InGaAs detector. For these analyses, the 180◦ backscattering refractive geometry was used. The spectrometer was managed using Perkin-Elmer Spectrum software (Version 3.02.00 [2000]). The spectral data for rice bean starch were obtained at a wavenumber resolution of 4 cm−<sup>1</sup> and at a nominal laser power of 500 mW. For each spectrum, 20 scans were accumulated to ensure an acceptable signal-to-noise ratio. All Raman spectra were collected at room temperature.

#### *2.6. Edible Coating Preparation*

Six different treatments were prepared, which differed in the solvent (water or ethanol) and cladode tissue (parenchyma or chlorenchyma) used, for the reconstitution of the mucilage powder (Table 1). For the preparation of the edible coatings, 50 g of the mucilage powder were placed in a flask (500 mL), and then water or ethanol was added to obtain a solution of mucilage with 12% soluble solids; this procedure was carried out for each cladode tissue.



\* Water:Ethanol 50:50 vol.% (relation obtained in previous experiments; data not showed).

#### Edible Coating Application

Prior to the application of edible coatings, the tomatoes were washed with distilled water. The coatings were applied uniformly by brushing on various tomatoes (26 pieces), three times. The tomatoes were stored at 20 ◦C. A control lot of tomatoes (26 pieces) was used (without edible coating).

#### *2.7. Weight Loss During Postharvest Storage*

Weight loss during postharvest storage was determined by subtracting the sample weight from their previous recorded weight and was presented as % of weight loss compared to the initial weight. Weight loss was calculated using the Equation (1).

$$\text{weight loss} \left( \% \right) = \frac{\text{Initial weight (g)} - \text{final weight (g)}}{\text{Initial weight (g)}} \times 100\tag{1}$$

#### *2.8. Texture Measurement*

A penetration test was performed on the skin of the whole fruit using a TA.XT2i texture analyzer (Stable Micro Systems, Surrey, UK) with a 2 mm diameter cylindrical probe. The samples were penetrated to a depth of 6 mm. The speed of the probe was 1.0 mm·s−<sup>1</sup> during the pretest and penetration. The tomatoes were placed in a way that the probe penetrated their equatorial zone.

#### *2.9. Determination of Lycopene*

Fifty g of tomato were ground to a homogeneous puree using a hand-held mixer (Braun Inc., Lynnfield, MA, USA). The puree was mixed with 50 mL of a hexane-acetone-ethanol mixture (50:25:25) and placed in an orbital shaker (Eberbach Corp., Ann Arbor, MI, USA) (200 rpm) for 15 min. Thereafter, 3 mL of distilled water were added, and the sample was shaken for another 15 min and then vacuum-filtered. The combined filtrates were mixed with 50 mL of hexane in a separatory funnel, and 200 mL of double-distilled water were then added; the mixture was allowed to stand for 15 min to enable phase separation. The water–acetone–ethanol phase was discarded; the hexane phase was collected into an amber screw-capped vial and concentrated to dryness with high purity nitrogen (99.99%).

Lycopene was quantified using an external standard, and the absorbance was taken at 503 nm using a spectrophotometer (Thermo Fisher Scientific Inc., Waltham, MA, USA); hexane was used as solvent blank. The results were expressed as mg·kg−<sup>1</sup> of sampler.

#### *2.10. Statistical Analysis*

Quantitative data are expressed as the mean ± standard deviation, and the results were statically analyzed with ANOVA and Tukey's test at the 95% confidence level. SAS software (Statistical Analysis System V. 8.0) was used for the data analysis, and all experimental determinations were performed in triplicate.

#### **3. Results and Discussion**

#### *3.1. Characterization of Mucilage*

#### 3.1.1. Morphology of *Opuntia* Mucilage

Scanning electron microphotographs (SEM) of the mucilage obtained from *Opuntia robusta* are shown in Figure 1. A higher aggregation level of small particles was observed in the mucilage extracted from the parenchyma (Figure 1b) than in the mucilage extracted from the chlorenchyma (Figure 1a). The particles of the mucilage extracted from the chlorenchyma of *Opuntia robusta* are mostly seen as aggregates of irregular shapes and dimensions and are fibrous in nature. Superimposed fibers have been observed in chia mucilage and increase the complexity of the aggregates [15]; apparently, a similar behavior can be observed in the mucilage extracted from the parenchyma of *Opuntia robusta*. On the other hand, the high aggregation observed in the mucilage obtained from *Opuntia robusta* is in concordance with those reported by du Toit et al. [16], indicating that *Opuntia robusta* has a higher fiber content that other *Opuntia* species.

**Figure 1.** SEM micrograph of *Opuntia robusta* mucilage: (**a**) extracted from the chlorenchymatous tissue; (**b**) extracted from the parenchymatous tissue.

3.1.2. FT-IR Analysis of Structural Changes in Mucilage Due to the Extraction Process and the Source

The FT-IR spectra of the mucilage extracted using either water or ethanol are shown in Figure 2. It can be observed that the spectra of the mucilage present a broad, strong signal at approximately 3400 cm−<sup>1</sup> due to the stretching vibration of O–H bonds; it can be observed that in both parenchyma and chlorenchyma tissues, this band was smoother in the spectrum of mucilage extracted with ethanol (Figure 2a,c) than in the spectrum of mucilage extracted with water (Figure 2b,d). Apparently, this difference is due to the formation of more hydrogen bonds in the mucilage by the interaction with ethanol. The best defined band was observed at approximately 2920 cm−<sup>1</sup> for the ethanol-extracted mucilages of both tissues (parenchyma and chlorenchyma), and this band was assigned to the stretching vibration of C–H bonds from pyranose groups [17] or C–H group of the methyl ester of galacturonic acid [18].

**Figure 2.** FT-IR patterns of *Opuntia robusta* mucilage: (**a**) extracted from the parenchyma with ethanol; (**b**) extracted from the parenchyma with water; (**c**) extracted from the chlorenchyma with ethanol; (**d**) extracted from the chlorenchyma with water.

However, some authors have indicated that this band can be attributed to proteins [19,20]; in addition, the water-extracted mucilages of both tissues show that the best defined band near 2870 cm−<sup>1</sup> overlaps with the stretching vibration of the C–H bond from glucose units and with the –CH2 symmetric stretching vibration of proteins derived from carboxylic acids, indicating the presence of protein residues on the mucilage [19,21].

The broad bands observed between 1600 and 1640 cm−<sup>1</sup> correspond to the scissor vibrations of –OH due to bound water [17] or the stretching vibration of the carboxyl group –COOH [19,22,23]. In mucilage extracted with water or ethanol, the presence of a free carboxyl group was observed, represented by the band at approximately 1430 cm−<sup>1</sup> [22]. The best defined band for the mucilage extracted with ethanol was observed at approximately 1320 cm<sup>−</sup>1, indicating the presence of *o*–acetyl groups; other authors have attributed this band to the glycosidic linkage [24]. The three characteristic bands of pectic fractions [22] were observed only in the mucilage extracted from the parenchyma (1120, 1080, and 1050 cm−1); in the mucilage extracted from the chlorenchyma, an overlap was observed at approximately 1085 and 1065 cm<sup>−</sup>1.

3.1.3. Raman Spectroscopy Analysis of Structural Changes in the Mucilage Due to the Extraction Process and the Source

The Raman spectra of the mucilage samples from *Opuntia robusta* are presented in Figure 3. The principal differences observed in the spectra of mucilage extracted with ethanol or water occurred in the principal bands associated with mucilage; on the other hand, differences in relative intensity were also observed between the patterns of Raman spectra of the mucilage extracted with ethanol (Figure 3b,d) and those of the mucilage extracted with water (Figure 3a,c) and between the spectra of the mucilage extracted from the chlorenchyma and that extracted from the parenchyma. In the mucilage extracted from the parenchyma, a well-defined band at 2930 cm−<sup>1</sup> was observed in the sample in comparison with the mucilage extracted from the chlorenchyma, which presented a change in position; in addition, overlapping bands were observed (Figure 3a).

**Figure 3.** FT-Raman patterns of *Opuntia robusta* mucilage: (**a**) extracted from the chlorenchyma with water; (**b**) extracted from the chlorenchyma with ethanol; (**c**) extracted from the parenchyma with water; (**d**) extracted from the parenchyma with ethanol.

Some authors have indicated that the 2930 cm−<sup>1</sup> band is related to the C–H stretching band of pectin, while the overlapping band at 2880 cm−<sup>1</sup> is related to hemicellulose [25]; however, other authors have reported that this band is present in the mucilage of chia seeds [15].

The vibrational bands between 1400 and 1470 cm−<sup>1</sup> are related to the elongation of the CH2 groups of xylose, galactose and arabinose, and the CH3 group of rhamnose; all of these bands were present for the *Opuntia* mucilage [10], and a higher intensity of these bands was observed in the water-extracted mucilage from the chlorenchyma than in the mucilage extracted from the parenchyma or extracted with ethanol; the intensity depends on the degree of crystallinity.

Additionally, the bands observed at 1350, 1270 and 1076 cm−<sup>1</sup> are assigned as C–O–H related modes and probably correspond to residual galacturonic acid; however, some authors have indicated that the band at approximately 1330 cm−<sup>1</sup> is related to the CH2 vibration of α-glucose and that the band at approximately 370 cm−<sup>1</sup> is assigned to the skeletal vibrational mode of δs(C–C) [15].

#### *3.2. Mucilage Used as Edible Coating*

#### 3.2.1. Weight Loss of Tomatoes

The tomatoes coated with parenchyma as an edible coating tended to have smaller weight loss than the control tomatoes (Figure 4). No significant difference in weight loss was observed in the three lots of tomatoes covered with parenchyma (C1, C2 and C3); however, the weight loss of the coated tomatoes was 7 wt.% lower than that of the control tomatoes (Figure 4). On the other hand, for the tomatoes coated with chlorenchyma, the differences between the three mucilage coatings tested (C4, C5 and C6) and the control (Figure 4) were significant. The tomatoes coated with chlorenchyma as an edible coating showed higher weight loss than the tomatoes coated with parenchyma; however, smaller weight loss was observed compared to that of the uncoated tomatoes (Figure 4).

**Figure 4.** Effect of mucilage coating on weight loss of tomatoes during the 21-day storage period (20 ± 2 ◦C): (C1) edible coating sample prepared with parenchyma mucilage + water; (C2) edible coating sample prepared with parenchyma mucilage + ethanol; (C3) edible coating sample prepared with parenchyma mucilage + water:ethanol (50:50 vol.%); (C4) edible coating sample prepared with chlorenchyma mucilage + water; (C5) edible coating sample prepared with chlorenchyma mucilage + ethanol; (C6) edible coating sample prepared with chlorenchyma mucilage + water:ethanol (50:50 vol.%).

The differences between the weight loss of the tomatoes coated with mucilage extracted from the parenchyma and that of tomatoes coated with mucilage extracted from the chlorenchyma could be due to the different compositions of the mucilage sources; however, it can be inferred that either mucilage source (parenchyma or chlorenchyma) formed a semipermeable layer that reduced respiration and transpiration on the tomato surface. It has been observed that coatings confer a physical barrier against O2, CO2, moisture and solute movement, consequently reducing water loss and weight loss [26–28].

Our results are in agreement with the finding of Del-Valle et al. [14], where the mucilage coating from *Opuntia ficus-indica* was effective in extending the shelf-life and other postharvest parameters of quality of strawberry; in other studies, the weight loss of breba fig was strongly influenced by *Opuntia ficus-indica* mucilage coating [29].

#### 3.2.2. Effect on the Firmness of Tomatoes

Tomato firmness decreased significantly during storage in all coated and uncoated tomatoes (Figure 5); however, the coated tomatoes showed resistance against the loss of firmness and maintained texture during the storage period, particularly those tomatoes coated by mucilage extracted from the nopal parenchyma (Figure 5). The reduction in firmness of the uncoated sample was 62% compared to 47% for the tomatoes coated with mucilage extracted from the parenchyma and 60% for the tomatoes coated with mucilage extracted from the chlorenchyma (Figure 5). Similar firmness reduction trends were observed by Allegra [29], who showed that the loss of firmness in tomatoes coated with *Opuntia ficus indica* mucilage was significantly delayed during the storage period.

**Figure 5.** Effect of mucilage coating on the firmness of tomatoes during the 21-day storage period (20 ± 2 ◦C): (C1) edible coating sample prepared with parenchyma mucilage + water; (C2) edible coating sample prepared with parenchyma mucilage + ethanol; (C3) edible coating sample prepared with parenchyma mucilage + water:ethanol (50:50 vol.%); (C4) edible coating sample prepared with chlorenchyma mucilage + water; (C5) edible coating sample prepared with chlorenchyma mucilage + ethanol; (C6) edible coating sample prepared with chlorenchyma mucilage + water:ethanol (50:50 vol.%).

Surface coatings have been found to retain higher firmness, and the positive effect is attributed to the restriction in metabolic activities associated with cell wall-degrading enzymes or to the calcium content in species of the *Opuntia* genus [12,30], since it has been observed that calcium treatment maintains the firmness of peach, fig and strawberry [31–33].

#### 3.2.3. Effect on the Lycopene Content of Tomatoes

The lycopene contents of the tomatoes coated and uncoated with mucilage during the red stage on the 21st day are shown in Table 2. In general, the results are in concordance with those of other studies on lycopene content in tomatoes that indicated that the lycopene content of stored tomatoes is in the range of 78.6 and 324 mg·kg−<sup>1</sup> of tomato on a fresh weight basis and between 1710 and 4550 mg·kg−<sup>1</sup> of tomato on a dry weight basis [34,35], however the extraction method was determinant to the extractable amount of lycopene and purity, reaching in some cases around purity level over 95% [36]. The lycopene content of the uncoated tomatoes on the 21st day of storage was 20% and 6% higher than those of the tomatoes coated with mucilage extracted from the parenchyma and the chlorenchyma, respectively.


**Table 2.** Effect of coating on the lycopene content of tomato on the 21st day of storage.

Our results are in agreement with those of previous reports suggesting that the formation of lycopene depends on the rate of respiration during storage [37,38]; apparently, the mucilage provided a thick barrier against ethylene production and gas exchange between the inner and outer environments and therefore delayed the ripening of the fruit during storage [2].

#### **4. Conclusions**

Differences in morphology between the mucilage extracted from the parenchymatous tissue and the mucilage extracted from the chlorenchymatous tissue were detected. The FT-IR and Raman spectroscopy data showed that the mucilage extracted from the parenchymatous tissue had an apparently higher level of pectic substances than the mucilage extracted from the chlorenchymatous tissue; in addition, changes in the FT-IR and Raman spectra patterns were observed due to the extraction solvent. The scanning electron microscopy (SEM) showed higher aggregation level in the mucilage extracted from the parenchyma than in the mucilage extracted from chlorenchyma. Tomatoes covered with mucilage showed significantly enhanced firmness and reduced weight loss, while the uncoated tomatoes showed higher lycopene content than the coated tomatoes on the 21st day. Finally, this study showed that *Opuntia robusta* mucilage is a promising edible coating and that tissue and extraction solvent influence mucilage characteristics.

**Author Contributions:** Conceptualization, A.B.-N. and G.-C.L.; Formal Analysis, E.C.-B. and M.d.l.L.X.N.-R.; Funding Acquisition, A.B.-N. and G.-C.L.; Investigation, J.L.M.-S. and G.A.-G.; Methodology, E.A.V.-L.; Resources, G.T.-M. and J.M.S.J.-G.; Validation, G.T.-M. and J.M.S.J.-G.; Writing–Original Draft, A.B.-N. and G.-C.L.; Writing–Review & Editing, G.A.-G.; Supervision, A.B.-N. and G.-C.L.

**Funding:** This research was funded by National Technologic of Mexico (TecNM, No. 6405.18-P).

**Acknowledgments:** The authors would like to acknowledge the Autonomous University of Hidalgo State for technical services during this study.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Self-Stratification of Ternary Systems Including a Flame Retardant Liquid Additive**

#### **Agnes Beaugendre 1, Stephanie Degoutin 1, Severine Bellayer 1, Christel Pierlot 2, Sophie Duquesne 1, Mathilde Casetta <sup>1</sup> and Maude Jimenez 1,\***


Received: 3 November 2018; Accepted: 5 December 2018; Published: 6 December 2018

**Abstract:** Particular coating compositions based on incompatible polymer blends can produce coatings having complex layered structures after film formation. The most traditional approaches to their structural modification are the introduction of additives (extenders, inorganic pigments, surface active agents, etc.). As minor additives, some are capable of substantially accelerating the phase separation process with a moderate or negligible influence on the composition equilibrium of solutions. In contrast, in order to be effective, some have to be introduced in significant amounts, thereby substantially changing the resulting distribution of components through the film. Up to now, most of the liquid additives that have been tested destabilized the solutions while impacting the layering process. In this work, two phosphorus based liquid fillers have been introduced (at 2.5 and 5 wt.%) in a partially incompatible polymer blend based on a silicone resin and a curable epoxy resin to fire retard a polycarbonate matrix. Self-stratification was evidenced by microscopic and chemical analyses, flammability by Limiting Oxygen Index (LOI) and UL-94 tests, fire performances by Mass Loss Calorimetry and thermal stability by using a tubular furnace and ThermoGravimetric Analysis. The ternary compositions including 5 wt.% of additives exhibit the best stratification and excellent adhesion onto polycarbonate. Improvements of the fire resistant properties were observed (+7% for the LOI compared to the virgin matrix) when a 200 μm wet thick coating was applied, due to reduced flame propagation and dripping.

**Keywords:** self-stratifying coating; phase separation; incompatibility; solution equilibrium; flame retardancy; epoxy resin; silicon resin; liquid phosphate

#### **1. Introduction**

In the field of coatings science, significant improvements have been noted in properties such as chemical and fire resistance, antifouling, corrosion protection, flexibility and impermeability by using new approaches, such as the application of inorganic/organic hybrid coating systems, self-stratifying coatings, and so on [1–5]. These properties depend on the polymer selected as the bare resin of the formulation. Coatings in the industry are complex multilayer systems that must be designed to possess specific properties: the air-coating interface has to protect the underlying coatings against abrasion, chemical/solvents and UV-light, and to provide excellent impact resistance. The intermediate coating provides color, special effects and properties such as flame retardancy. Finally, the primer promotes adhesion to the substrate and protects against molecular diffusion (and corrosion in the case of a

steel substrate). Each coating has a specific purpose; however, the successive application of multiple coatings is time consuming and leads to increased expenses as well as potential drawbacks such as, for example, poor interfacial adhesion.

During paint application and film formation, a polymer solution simultaneously undergoes mass and heat transfers, surface phenomena and sometimes also chemical reactions, especially in the presence of a two-component system. Phase separation of binders, from an initially constituted homophase solution, can occur under certain conditions leading to self-stratification, that is, to the formation of non-homogeneous in-layer coatings during after film formation (Figure 1). In addition, in the presence of reactive systems, further reactions like oxidation or crosslinking (e.g., two-component epoxies and urethanes) are required to generate a cross-linked film of higher molecular weight. During the reaction, the viscosity, density, and the modulus of elasticity increase due to an increase in the molar mass and crosslinking. Besides interfering with the self-stratifying process, it allows the formation of a harder primer, less flexible and much less susceptible to damage from chemical, weather or UV rays [6,7].

**Figure 1.** Self-layering concept.

Besides polymer/polymer ratio and curing reaction, heterophase polymer structures are controlled by many process parameters in commercial coating compositions. The most important ones are the presence of fillers, the solvent properties and the processing conditions (application method and baking temperature). In particular, the incorporation of inorganic pigments, surface additives (polymer or oligomer surfactants), extenders or other types of fillers often interfere with the layering process. Most of these additives have a rather negative effect on the phase separation between two binders, and do not allow the development of sufficiently interesting properties compared to those of conventional multilayered systems. In many cases, liquid fillers (wetting, levelling, anti-flotting, surface active or dispersing agents) appear to impair the stratification process [8–11]. As an exception, some silicone or fluorine-containing oligomers added in minor amounts were proven to substantially accelerate the phase separation process with a moderate or negligible influence on the solutions equilibrium [9]. No previous study of the addition of a consequent amount (>2 wt.%) of a functional liquid additive in self-stratifying compositions has been reported. In this work, the aim was to design a self-stratifying coating on a polycarbonate substrate, incorporating a liquid flame retardant (FR) additive.

On the past few years, phosphorus-based flame retardants have been increasingly used as an alternative to halogen-based systems due to environmental considerations. This has generated active research in identifying novel phosphorus based FRs and also possible synergistic combinations with other elements, such as nitrogen for example [12]. Some phosphorus-containing compounds, besides their FR effect, have proven to be of great value due to their low toxicity and low smoke-emission during combustion [13]. In particular, arylphosphates have been proven to be suitable halogen-free FRs in blends containing polycarbonate (PC) (particularly in PC/ABS blends) [14,15].

In this work, film forming compositions based on a partially incompatible polymer blend, including a curable epoxy resin and a silicone resin in combination with a phosphorous-based liquid fire retardant filler are considered [16]. Two liquid fire retardant additives were tested: bisphenol-Abis(diphenyl phosphate) (BDP) and resorcinol bis(diphenyl phosphate) (RDP). The objective was to form thin flame retardant films which spontaneously self-layer after application of one single formulation.

#### **2. Experimental**

#### *2.1. Materials*

On the basis of previous results [16], two commercial resins and a curing agent were selected, respectively: an epoxy (Bisphenol-A epoxide from Sigma-Aldrich, St. Louis, MO, USA, equivalent weight: 172–176, 100% solids), a silicone (a phenyl branched resin containing 6% of hydroxyl group from Dow Corning, Seneffe, Belgium) and a polyamine (diethylene triamine (99%), Sigma-Aldrich).

m-Xylene (99%) and butylacetate (BuAc, ≥99.5%) purchased from Sigma-Aldrich, were chosen as solvent and used without any purification.

Two phosphorus based liquid FRs, that is, bisphenol-A bis(diphenyl phosphate), Fyrolflex BDP, 8.9% of phosphorus and resorcinol bis(diphenyl phosphate), Fyrolflex RDP, 10.7% of phosphorus were purchased from ICL-Industrial Products (Amsterdam, The Netherlands) and were incorporated as additives in various amounts (2.5 wt.% and 5 wt.%).

Transparent polycarbonate plates (Lexan, 1mm-thick for microscopic analyses and 3 mm-thick for fire testing) were supplied by Polydis (Ligny Le Chatel, France).

#### *2.2. Coatings Formulation and Application*

Resins were separately dissolved at 30 wt.% in a blend of BuAc:xylene (1:1), mixed and then stirred together at a 1:1 ratio by weight (epoxy:silicone). Liquid additives were incorporated into the epoxy medium for 10 min at 300 rpm before the combination with the second resin. Finally, the curing agent was thoroughly added drop by drop, with respect to the epoxy number, and mixed for 3 min before the application of the thin film by spraying (air pressure of 200 kPa). The nominal wet film thickness was set at 200 μm. Coatings were dried at 20 ◦C for 24 h in an oven and then cured for 2 h at 110 ◦C.

#### *2.3. Characterization of Film Properties*

Microscopic analyses coupled with X-ray mappings were used to detect the potential formation of stratified layers and to determine the film thickness. Fire behavior and thermal stability of the samples were investigated using Limiting Oxygen Index (LOI), UL-94 Mass Loss Calorimetry (MLC) and ThermoGravimetric Analysis (TGA). The specimens and residues obtained were analyzed either by microscopic analyses coupled with X-ray mapping or by using a numerical microscope. To complete the evaluation of the film properties, cross-hatch testing was used to quantify the adhesion.

#### 2.3.1. Microscopic Analyses

Coatings were cut in liquid nitrogen to allow the analysis of their cross-section and the determination of their thickness by Scanning Electron Microscopy with X-Ray mapping (SEM-EDX). Analyses were respectively carried out at 5.0 kV, 20 μA and 13.0 kV, 25 μA using a Hitachi S4700 (Tokyo, Japan) with field emission gun. Carbon metallization was carried out before any characterization.

Numerical pictures of sample residues were performed using a microscope VHX-1000 supplied by Keyence (Osaka, Japan, ×20). The microscope creates a 3D image based on automatically captured images.

#### 2.3.2. Classification of Stratification

The degree of layering was measured following the guidelines developed in the Brite Euram Project [8]. The degree of stratification was classified from 1 to 4 according to the microscopic characterization of the cross-section of the coatings. From this ranking, a type I pattern corresponds to a perfect stratification (e.g., two well distinct and homogeneous layers), a type II pattern to an homogeneous concentration gradient between the two resins through the film thickness, a type III to the formation of spherical particles rich in one of the resins dispersed in a medium enriched with the other resin, and finally a type IV corresponds to the presence of large islands composed of a majority of one of the resins.

#### 2.3.3. Adhesion Testing

Adhesion rating of the coating on polycarbonate was carried out following the ASTM D3359-97 standard [17]. An Elcometer 107 Cross Hatch Cutter was used to perform the test: a pressure-sensitive tape is applied and then removed over cuts made in the film, and according to the specimen obtained, adhesion is rated from 0B to 5B, 0B corresponding to the worst adhesion (no coating left on the substrate) and 5B to the best one (the edges after the test are completely smooth, and none of the squares of the lattice is detached).

#### 2.3.4. Fire Testing

A FFT (Fire Testing Technology) mass loss calorimeter (MLC) was used to perform the experiments following the procedure defined in ASTM E906 [18]. Specimens measuring 100 × <sup>100</sup> × 3 mm<sup>3</sup> were exposed in horizontal orientation to an external flux of 50 kW/m2 and a forced ignition. The selected flux corresponds to the common heat flux generated in fully-developed fires. The equipment can be compared to the one used in oxygen consumption cone calorimetry (ASTM E-1354-90 [19]), except that a thermopile is set in the chimney in order to obtain the heat release rate (HRR) instead of using the oxygen consumption principle. The distance sample-heater was set to a distance of 35 mm from the cone base, on a ceramic backing board. The following parameters were estimated: heat release rate (HRR) as a function of time, time to ignition (TTI), peak of heat release rate (pHRR) and total heat release rate (THR). The experiments were carried out three times to ensure the repeatability of the results, and the data reported are the most representative of the three replicated experiments. Values were found to be reproducible within a relative standard deviation of ±10%.

Two vertical fire tests were used to complete the evaluation of fire performances on barrels (100 × <sup>10</sup> × 3 mm3): Limiting Oxygen Index (LOI) and UL-94 test. LOI, corresponding to the minimum oxygen concentration needed to support the candle like combustion of plastics, was measured using a Fire Testing Technology instrument following the standard oxygen index test ISO 4589-2 [20]. The UL-94 test was performed according to IEC 60695-11-10 standard [21], that is, in a vertical position (the bottom of the sample is ignited with a bunsen burner). This test provides a qualitative classification of the samples, from V0 (the best ranking, when burning is short and there is no dripping of flaming particles) to NC (non-classified, i.e., with a burning of more than 30 s or up to the holding clamps at 100 mm from the ignition point).

#### 2.3.5. Thermal Stability

#### Thermogravimetric Analysis (TGA)

A Discovery TGA from TA Instrument was used to carry out thermogravimetric analyses. 10 mg samples (liquid materials or dry grounded coatings samples which have undergone the same heat treatment as the coatings) were thermally decomposed in alumina crucibles under a nitrogen atmosphere. They underwent an isotherm of 120 min at 50 ◦C for thermal homogeneity, followed by a heating ramp from 50 to 800 ◦C at 20 ◦C·min−1. The nitrogen flow rate was set at 50 mL·min−1. Difference weight loss curves were calculated to detect a potential variation in the thermal stability of the systems, due to the incorporation of the filler (Equation (1)). These curves represent the difference between the experimental TG curve for the mixture (*wexp*(*T*)) and the linear combination of TG curves (*wtheo*(*T*)) for the neat components (Equation (2)) when fillers are incorporated at *x* wt.%.

$$
\Delta w(T) = \left. w\_{\text{exp}}(T) - \left. w\_{\text{theo}}(T) \right| \right. \tag{1}
$$

$$w\_{l\text{neo}}(T) = (1 - \mathbf{x}) \times w\_{resin} + \mathbf{x} \times w\_{filler} \tag{2}$$

where *wresin* and *wfiller* correspond to the weight determined from the experimental TG curves; *x* the weight percentages of the sole filler. If Δ*w*(*T*) < 0, then the experimental weight loss is higher than the theoretical one. This shows that the reactivity and/or interaction between the polymer and the filler leads to a thermal destabilization of the material. If Δ*w*(*T*) > 0, then the system is thermally stabilized.

Heat treatments and characterization of the heat-treated residues were carried out. Heat treatments of pure components or of coatings were carried out in a tubular furnace under nitrogen flow (75 mL·min<sup>−</sup>1) for 3 h at characteristic temperatures selected thanks to TG curves. The collected residues were then analyzed using a numerical microscope.

#### **3. Results and Discussion**

#### *3.1. Stratification Study*

After mixing and before application as a film, the four-component system (epoxy and silicone resins, curing agent and phosphate additive) produced homophase solutions whatever the phosphate concentration. After drying and curing, the formation of separated in-layers coating structure through the film thickness was checked and adhesion properties were evaluated (Table 1).


**Table 1.** Morphology and adhesion properties of the coatings containing or not phosphate additives.

The reference system without any phosphate additive shows perfect stratification (type I pattern) and the best adhesion rating (5B) on polycarbonate: the hydrophobic layer (the silicone medium) migrated to the upper part of the coating whereas the epoxy system was found at the interface with the substrate.

In presence of liquid FR fillers, nice coatings free of bubbles or defaults were obtained in all cases. They even show slight improvement in terms of surface aspect (less rough and glossier) than the reference coating. The 5B adhesion rating is maintained after the incorporation of RDP in the system, for both amounts tested, whereas the addition of BDP slightly impacts the adhesion of the coating (4B rating is obtained).

Unlike the results mentioned in the literature, it appears that neither RDP nor BDP affect the layering process when incorporated at 5 wt.% in the formulation: a type I pattern is obtained with fully separated layers, as observed with the unfilled system (Figure 2).

Both additives migrate to the upper layer of the film, in the silicone medium, meaning that RDP and BDP have a higher affinity with silicone than with epoxy. This migration is rather a positive phenomenon in this case as it allows concentrating the additives' properties (i.e., flame retardant properties) in the upper part of the coating, where it is the most needed in case of fire. With RDP, some inhomogeneities are noticeable along the film near the two resins interface region: some island shaped regions rich in silicone have not totally migrated to the upper layer of the film. A concentration gradient of phosphorus is also detectable through the film thickness, although the major part of RDP has migrated to the upper layer. With BDP, the same behavior can be observed: the silicone resin and the filler are mainly concentrated in the upper part of the coating. However, the concentration gradient of phosphorus is more pronounced through the thickness than with RDP. The difference could be explained by a higher affinity of RDP with the silicone resin, promoting the migration of the additive to

the air interface. However, the concentration gradient is much more noticeable with phosphorus than with silicium: this means that the affinity between the silicone resin and the filler is not strong enough to allow its complete migration toward the top layer. A difference in viscosity could also explain the difference in concentration gradient between BDP and RDP containing formulations. Indeed, BDP is much more viscous than RDP (17,000 CPS [22] compared to 12,450 CPS for RDP [23]. As its viscosity is higher, its migration to the upper layer may be slowed down. It is noteworthy to notice that even if the additive has not enough time to completely migrate to the upper layer of the film, it does not impair the stratification process: the layering of the silicone to the top of the coating remains perfect.

**Figure 2.** EDS X-ray mappings of Silicon and Phosphorus on a cross-section of an epoxy/silicone coating filled with 5 wt.% of (**a**) resorcinol bis(diphenyl phosphate) (RDP) and (**b**) bisphenol-Abis(diphenyl phosphate) (BDP) additives.

When a smaller amount of phosphate additives is introduced in the formulation (2.5 wt.%), evidences of phase separation are found with a concentration gradient of silicone through the thickness. Nevertheless, some inhomogeneities in the matrix are observed (Figure 3), consisting in isolated spherical particles composed of epoxy resin dispersed in the continuous silicon matrix. Furthermore, the phosphorus element is found in the silicone phase whereas the stratification is not complete in those areas. This corroborates its higher affinity with silicone compare to that of epoxy. The phosphate additive destabilizes the interface between the two phases and, in that case, influences the preferential orientation of the phases in the course of film forming process. The dispersion of the filler is portioned in the silicone medium, however interacting in some way in the course of crosslinking reactions. A minimum amount of filler may be needed to allow a homogeneous dispersion into the silicone medium and this would need further investigation to the forces questioned in this process. This morphology corresponds to a type II/type III stratification pattern. However, this incomplete phase separation has no influence on the adhesion properties of the coatings as similar adhesion ratings are obtained with 2.5 wt.% and 5 wt.% of fillers, for both RDP and BDP [24].

**Figure 3.** Scanning Electron Microscopy (SEM) pictures of the cross-section of the ternary system filled with 2.5 wt.% of BDP.

In conclusion, when a low amount of additive is added (2.5 wt.%), self-layering is slightly compromised in some areas of the film. The filler destabilizes the equilibrium between the two solutions, and impairs the formation of two distinct and homogeneous layers. On the contrary, the phosphorus-based fillers tested do not affect the formation of oriented heterophase structures when they are introduced at 5 wt.%: they mostly migrate, with a concentration gradient more or less pronounced, towards the air interface with the silicone resin. In that case, some disparities in the concentration gradient can be observed, which may be due to a difference in viscosity between the two phases. Finally, coating's composition provided nice visual appearance and a high adhesion onto the polycarbonate substrate. It was proven that it is possible to add a substantial amount (5 wt.%) of liquid functional filler in a self-layering formulation without affecting the self-stratification process, the coating aspect and the adhesion. In these formulations, the phosphorus FR additives migrate into the upper silicone layer, which can potentially lead to an effective FR effect of the coating when exposed to fire.

#### *3.2. Flame Retardant Properties*

Virgin PC and PC coated with the unfilled and filled self-stratified compositions were then tested for fire performances (MLC) and flammability (UL-94 and LOI).

HRR curves and characteristic MLC parameters of the systems are given in Figure 4 and Table 2. It is noteworthy that no ignition of the coated samples (filled and unfilled) occurs at 35 kW/m<sup>2</sup> (mild fire scenario), even after 30 min of exposure; whereas the raw PC ignites after 319 s and releases a total heat of 35 MJ/m<sup>2</sup> (with a pHRR of 202 kW/m2, see Supplementary Materials Figure S1). PC is a char-forming polymer. At 50 kW/m2, after its ignition, it melts and forms a char which swells until a height of about eight centimeters is reached. The char then degrades and only ashes remain at the end of the test.

Even at a higher heat flux (50 kW/m2), the coating allows protecting temporarily the PC. MLC results obtained for the coating with the unfilled epoxy/silicone coating evidence an improvement of the fire behavior, as both the pHRR and THR are reduced, respectively by 24% and 21% compared to raw PC. Ignition time is also delayed more than 50 s, which is remarkable with such thickness (55 μm). Ignition occurs only because PC swells under the coating, which thus progressively delaminates from the edges of the plate due to the uprising of the PC char. Once ignited, PC forms a char which rapidly swells up to 8 cm. The char is then consumed and collapses before the flame out takes place. At the end of the test, PC is completely consumed, which also explains the pHRR and THR values obtained.

**Figure 4.** Heat release rate (HRR) curves obtained with uncoated polycarbonate (PC) and coated PC with the self-stratified epoxy/silicone compositions containing (**a**) RDP and (**b**) BDP flame retardant (FR) additives at 2.5 wt.% and 5 wt.%.


**Table 2.** Mass loss calorimeter (MLC) values obtained at 35 and 50 kW/m with uncoated and coated PC with the epoxy/silicone composition with and without FR additives.

<sup>1</sup> Percentages represent the difference compared to the unfilled epoxy/silicone system.

MLC experiments on PC samples coated with only one resin, whether the epoxy resin or the silicone one, have been carried out to try to explain the behavior of PC coated with the unfilled epoxy/silicone blend (Figure 5, Table 3). When the epoxy resin is applied on PC, the behavior during the MLC test is similar to that of pure PC: no delay of the TTI and similar pHRR and THR. Thus, the epoxy resin does not bring any fire retardant effect, which is not surprising as it is composed of Bisphenol A (similarly to PC). On the contrary, the application of a silicone coating on PC leads to a fire retardant effect: slight decrease of pHRR and THR and TTI increase of about 40 s. Thus, the presence of silicone is responsible for the improvement of the fire behavior when the epoxy/silicone coating is applied to PC. However, the epoxy resin contributes to bringing better adherence between the self-stratifying coating and PC. In fact, the silicone coating gives a 3B rating on PC compared to a 5B rating with the self-stratifying coating. Moreover, there is no interlayer adhesion failure between the epoxy and the silicone layers thanks to the use of the self-stratifying coating.

Similar trends are observed when PC is coated with the epoxy/ silicone/ filler mixture, whatever the amount of additive incorporated. However, although the best TTI is obtained when RDP/BDP are incorporated at 2.5 wt.%, fire behavior are similar compared to that of the binary composition. The impact of RDP and BDP addition in terms of flame retardant behavior is thus negligible compared to the FR effect of the silicone coating itself. Finally, BDP seems slightly more efficient than RDP, which is not surprising as at a same ratio, BDP contains more phosphorus than RDP. Indeed, there is 10.7% of P in RDP whereas 8.9% are present in BDP. All the systems were consumed at the end of the test at 50 kW/m2: only residue of silica and ashes remained. The slight thickness difference between the two samples (Table 1) is not responsible for this effect since any correlation is noted between the thickness, the TTI and THR: close THRs are reached by both filled coatings at equal weight percentage whereas a reverse effect is observed on the time before ignition.

**Figure 5.** HRR curves obtained with uncoated PC and coated PC with the self-stratified epoxy/silicone coating, the epoxy coating and the silicone coating alone.

**Table 3.** MLC values obtained at 50 kW/m<sup>2</sup> with uncoated and coated PC with the epoxy/silicone self-stratifying coating, the epoxy and the silicone coating solely.


<sup>1</sup> Percentages represent the difference compared to uncoated PC.

LOI and UL-94 tests were also performed on uncoated and coated PC formulations (Table 4). The epoxy/silicone reference coating does not improve the intrinsic flame retardant properties of PC in the case of vertical burning tests: close LOI values (27 vol.% and 28 vol.% respectively for the raw and coated PC) and a NC rating are obtained for both samples. Up to 28 vol.%, the extinction of virgin PC occurs mainly because of the flaming drops or of the rapid combustion of the material. The unfilled coating allows reducing the combustion speed of the material and prevents the dripping of PC. At UL-94 test, virgin PC is not classified at 3 mm, but close to meet the requirements of V-2 classification: short combustion time (<32 s) and dripping (Table 5).

When the polymer is coated with the filled mixture, the fire behavior is strongly dependent on the amount of fillers used in the coating, whatever the type of additive (BDP or RDP). The addition of 2.5 wt.% of additives does not improve the fire behavior compared to the unfilled system: similar LOI value and same rating at UL-94 test. For these materials, the amount of additives is too low to allow the decrease of the flame spread. With 5 wt.% of additives, although the rating at UL-94 test is the same (NC), the behavior is much different: the combustion time is very short (≤2 s) after the first ignition and remains shorter than the unfilled system after the second ignition (Table 5). The classification is very close to V-1, as with both additives= only one specimen out of the five tested has a t2 afterflame time higher than 30 s (in bold, Table 5). In addition, LOI values are significantly improved: 34 vol.% and 32 vol.% respectively with RDP and BDP fillers. The charring is more pronounced (fast expansion of the foamed structure) and no dripping occurs. Last but not least, samples were more prone for burning during the second ignition with a lower concentration of additive in the coating.

Thus, similar behavior between RDP and BDP can be registered: they are both more efficient when they are introduced at a higher amount (5 wt.%). As with MLC, the slight difference observed between RDP and BDP may be due to the higher amount of phosphorus in the RDP formulation.


**Table 4.** LOI, UL-94 rating for PC, and coated PC with the binary and ternary mixtures.


**Table 5.** Time of the two flaming combustions during UL-94 test.

To conclude, formulations containing 5 wt.% of RDP and BDP exhibit the most promising results: charring is more pronounced and dripping is avoided. At MLC test, fire performances of coated formulations are mainly enhanced through the shift of the TTI toward higher values due to the silicone coating. The influence of the two phosphorus-based additives on the fire properties is not significant compared to the improvements already reached by the unfilled epoxy/silicone system.

Occurrence and efficiency of the systems depend not only on the FR itself, but also on the interactions with the degradation products of the different components of the materials (resins and substrate). To go deeper in the understanding of the mode of action of the phosphorus compounds, TG analyses and difference weight loss calculations were carried out on the silicone resin, containing phosphorus additives or without. During a fire scenario, the silicone will be the resin in contact with the open flame or exposed to the heat source as it is located in the upper layer of the film. Consequently, only the silicone resin was considered.

#### *3.3. Thermal Stability*

The thermal stability of the unfilled system and systems filled with RDP and BDP were investigated using TG analyses (Figures 6 and 7. Weight difference curves between experimental and calculated TG curves when RDP or BDP is added to the resin (10 wt.%) under N2 conditions). The percentage of fillers introduced in the silicone medium was kept constant compared to the epoxy/silicone system as the major part of the filler migrates to the silicone layer (i.e., 10 wt.% of filler in the silicone medium).

The decomposition of the pure silicone film involves a three-step process with a 68 wt.% residual weight at 800 ◦C. The first degradation step may be coupled with the release of the remaining solvents from the resin's preparation and the release of silicone oligomers (11 wt.%) and the second and third steps (overlapped) are correlated with to the release of aromatic compounds (such as benzene and bisphenyl, 21 wt.%) [25]. The total weight loss is only 32% at 800 ◦C, demonstrating the excellent thermal stability of the silicone resin (under TGA conditions). When FR liquid additives are incorporated, an additional degradation step occurs close to the maximum degradation temperature of RDP and BDP (at respectively 398 and 420 ◦C). This leads to a slight destabilization of the system, particularly with BDP between 370 and 480 ◦C (−2 %/◦C, Figure 7). Weight difference curves between experimental and calculated TG curves when RDP or BDP is added to the resin (10 wt.%) under N2 conditions). The residual weights of the additives are low (respectively 4 wt.% and 17 wt.% at 800 ◦C with BDP and RDP), although it does not neither influence significantly the thermal stability of the resins nor favor the formation of a charring structure at high temperature.

**Figure 6.** Comparison of TG and DTG curves of the silicone system with and without (**a**) RDP and (**b**) BDP (10 wt.%) under nitrogen at a heating rate of 20 ◦C/min.

**Figure 7.** Weight difference curves between experimental and calculated TG curves when RDP or BDP is added to the resin (10 wt.%) under N2 conditions.

TG experiments enable us to define characteristic temperatures of degradation. Accordingly, heat treatments were performed at 300 ◦C in a tubular furnace and under a nitrogen flow on the silicone and the silicone/fillers systems (Figure 8). From the numerical pictures, the formation of an expanded foamed structure with small cells, whatever the phosphorus compound used is evidenced: the fillers do not modify the cellular structure. Finally, RDP and BDP have low influence on the thermal stability of the silicone resin; the amount of the remaining residue at 800 ◦C is almost the same, and the degradation process remains unchanged compared to the pure silicone. Based on those observations, the formation of an expanded coating during the test explains, at least partially, the fire protection brought by the paint and the silicone-based coating containing the fillers. The silicone and silicone/filler compositions allow the formation of a protective barrier, thus reducing heat transfers from the external heating source to the substrate.

**Figure 8.** Numerical pictures of residues of (**a**) silicone, and (**b**) silicone/RDP, (**c**) silicone/BDP at 10 wt.% at 300 ◦C.

Visual observations during the horizontal fire tests confirm the higher char yield obtained when phosphorus compounds are incorporated into the systems, even if thermal analyses did not show major improvements in terms of stability. Based on the literature review, phosphorus compounds accumulate in the condensed phase in the char layer, at the surface of the burning specimens in a blend of PC-ABS + RDP [26]. Results suggested that PC undergoes photo-fries rearrangement upon thermal decomposition, and RDP would react with the formed phenolic groups through a trans-esterification mechanism. Kinetic analysis of the thermal decomposition of PC containing RDP supported the proposed mechanism.

In conclusion, the high thermal stability of the silicone resin under inert atmosphere at elevated temperatures has been demonstrated. As this component is at the interface with the air in the self-stratifying coating composition, it will consequently be the first in contact with either the heat source or the flame during the fire tests. According to the results, the silicone resin provides some thermal stability to the system, which may explain the higher time to ignition obtained during MLC experiments. Under elevated temperatures, the silicone coating protects the underlying substrate (and the epoxy resin) from decomposition and creates a barrier between the heat source and the substrate. Moreover, the similar behavior in terms of thermal stability for the silicone resin with and without fillers could explain the close results at MLC test for all the epoxy/silicone coatings.

#### **4. Conclusions**

A properly-selected partially incompatible polymer blend composed of silicone, of a curable epoxy resin and of a liquid functional filler successfully formed a double-layered coating, showing excellent adhesion to the underlying polycarbonate substrate. The binders, dissolved in solvents to produce a two-phase liquid medium during film formation, led to structures with sharply defined layers. The topcoat layer was found to be composed of the silicone resin, and the base layer of the curable epoxy resin. Microscopic analyses demonstrated that phosphorus-based liquid fillers do not impact the stratification process when incorporated at 5 wt.%. However, their migration to the upper layer of the coating (silicone phase) was not always complete: a concentration gradient of additives through the film thickness is obtained, with a higher concentration toward the top of the coating. When fillers were incorporated at a lower amount (2.5 wt.%), incomplete phase separation was observed in some areas of the film, although a high adhesion rating (5B and 4B with RDP and BDP fillers respectively) and nice visual appearance remained.

The two phosphorus-based compounds led to an increase in the LOI value: up to 34 vol.% is obtained with 5 wt.% RDP compared to 27 vol.% for the unfilled system. However, the systems are still non classified at UL-94, even if combustion times with 5 wt.% additives are very close to a V1 rating. For MLC test at 35 kW/m2, the coated samples do not ignite, and at 50 kW/m2, the application of the epoxy/silicone coating allows delaying the TTI, with and without the presence of phosphate additives: improvements are due to the silicone resin and are not significantly modified by the addition of the liquid fire retardants. Synergistic effects between the two liquid fillers are not expected, the level of flame properties reached are mainly dependent on the amount of phosphorus introduced in the system.

Finally, numerous factors are of importance for the formation of two well separated layers. It is known that the efficiency and mechanism of action of phosphorus flame retardants in coating compositions are influenced and can be optimized by modifying them using specific synergists. In that respect, the surface tension of the paint components, their viscosity and the amount introduced in the coating compositions have a decisive role.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-6412/8/12/448/s1, Figure S1: HRR curves obtained with PC and coated PC with the self-stratified epoxy/silicone compositions containing RDP and BDP FR additives at 2.5 wt. % and 5 wt. % (the four curves overlapped).

**Author Contributions:** Conceptualization, A.B., M.J., M.C., S.B., S.D. (Stephanie Degoutin), S.D. (Sophie Duquesne), C.P. Methodology, A.B., M.J., S.B., C.P.; Software, S.B., C.P.; Validation, M.J., M.C., S.B., S.D. (Stephanie Degoutin), S.D. (Sophie Duquesne), C.P.; Investigation, A.B., S.B.; Resources, M.J., M.C., S.B., S.D. (Stephanie Degoutin), S.D. (Sophie Duquesne), C.P.; Data Curation, A.B.; Writing-Original Draft Preparation, A.B.; Writing-Review & Editing, M.J., M.C., S.D. (Stephanie Degoutin), S.D. (Sophie Duquesne), C.P.; Visualization, A.B., M.J., M.C.; Supervision, M.J., M.C. and S.D. (Sophie Duquesne); Project Administration, M.J.; Funding Acquisition, M.J.

**Funding:** This research was funded by the french ANR (Agence Nationale de la Recherche, No. ANR-14-CE27-0010), STIC project.

**Acknowledgments:** We are also thankful to the MATIKEM competitiveness cluster for supporting the project and Lille University for administrative support.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Bio-Inspired Fluorine-Free Self-Cleaning Polymer Coatings**

**Lionel Wasser 1, Sara Dalle Vacche 1,2, Feyza Karasu 1,†, Luca Müller 1, Micaela Castellino 2,3, Alessandra Vitale 2, Roberta Bongiovanni <sup>2</sup> and Yves Leterrier 1,\***


Received: 30 September 2018; Accepted: 26 November 2018; Published: 28 November 2018 -

**Abstract:** Bio-inspired fluorine-free and self-cleaning polymer coatings were developed using a combination of self-assembly and UV-printing processes. Nasturtium and lotus leaves were selected as natural template surfaces. A UV-curable acrylate oligomer and three acrylated siloxane comonomers with different molecular weights were used. The spontaneous migration of the comonomers towards the polymer–air interface was found to be faster for comonomers with higher molecular weight, and enabled to create hydrophobic surfaces with a water contact angle (WCA) of 105◦. The replication fidelity was limited for the nasturtium surface, due to a lack of replication of the sub-micron features. It was accurate for the lotus leaf surface whose hierarchical texture, comprising micropapillae and sub-micron crystalloids, was well reproduced in the acrylate/comonomer material. The WCA of synthetic replica of lotus increased from 144◦ to 152◦ with increasing creep time under pressure to 5 min prior to polymerization. In spite of a water sliding angle above 10◦, the synthetic lotus surface was self-cleaning with water droplets when contaminated with hydrophobic pepper particles, provided that the droplets had some kinetic energy.

**Keywords:** self-cleaning; lotus; nasturtium; siloxane surfactants; acrylates; photopolymerization; UV nanoimprint lithography; PDMS template

#### **1. Introduction**

Natural superhydrophobic self-cleaning surfaces such as the famous lotus leaf consist of an intrinsic hierarchical structure with epithelial cells of sizes in the micrometer range, and sub-micron, low surface energy epicuticular wax crystals [1,2]. Such structures favor trapping of small air pockets at the interface with water droplets, which considerably reduces the contact area between the droplet and the surface, resulting in the reduction of contact angle hysteresis, tilt angle, and adhesive force. The self-cleaning effect is the removal of dirt particles by the (rain) water, which form droplets that do not 'stick' on the surface and run away with the dirt. The development of bio-inspired synthetic self-cleaning surfaces has stimulated a considerable research effort since more than a decade, leading to remarkable results. However, reported surfaces often rely on environmentally-detrimental

and cost-intensive approaches. These are primarily based on fluorinated compounds [3] owing to their low surface energy, and to nanoparticles giving rise to nanopatterned structures: some fluorinated compounds—i.e., long chained fluoroalkylic compounds—create concerns for their biopersistency, while nanoparticles are possibly associated with uncontrolled release issues [4]. In addition, many of the demonstrated methods are hardly scalable to cost-effective, large area surfaces [5,6]. Efforts to create fluorine-free and scalable self-cleaning surfaces have been initiated in the last few years, including superhydrophobic water-based nanoparticulate dispersions [7,8], sprayed waxes dissolved in supercritical CO2 [9], and various micro/nanotextures impregnated with lubricating liquids [10,11]. The present work follows up with the development of low-energy surfaces based on the spontaneous, enthalpy-driven migration of comonomers and resulting segregation at the polymer–air interface [12–16]. It is based on a highly scalable low-pressure, solvent-free, and ambient UV replication process of plant surfaces using a silicone template as demonstrated in a recent work [17]. In this study, one of the main challenges was the reverse migration of the fluorinated surfactant comonomer, from the polymer–air interface back to the bulk, upon contacting the low-surface energy silicone template, and suppression of the superhydrophobic properties. This problem was solved using a flash of UV light to chemically attach the comonomer previously segregated at the polymer surface, prior to UV printing, but this increased the viscosity of the superficial polymer layers, which was detrimental to the low-pressure replication fidelity.

The objective of this work was to produce bio-inspired self-cleaning coatings using siloxane surfactant comonomers as fluorine-free alternatives, and a cost-effective, low-pressure UV printing process. The leaves of two superhydrophobic plants—namely lotus and nasturtium—were selected to create templates. These two plants possess similar epicuticular wax crystals in the form of sub-micron tubules, however they exhibit totally different hierarchical microstructures: the lotus leaf structure is composed of microscale papillae, whereas nasturtium has larger convex epidermal cells [18,19]. Attention was also paid to the influence of the molecular weight and concentration of Si in the surfactant comonomer, and on the reverse migration phenomenon on the superhydrophobic character of synthetic replica of plant surfaces.

#### **2. Materials and Methods**

#### *2.1. Materials*

A hyperbranched polyester acrylate oligomer (CN2302, Sartomer, Colombes, France) was selected owing to its low polymerization shrinkage [20], which warrants a very high replication fidelity [21–23]. It has a theoretical functionality of 16 and a real one of 13, a density of 1.13 g·cm−<sup>3</sup> and a Newtonian viscosity of 0.3 Pa s at 25 ◦C. The photo-initiator was diphenyl(2,4,6-trimethylbenzoyl) phosphine oxide (TPO Esacure, Lamberti, Gallarate, Italy). Three different siloxane comonomers were used, namely an acryloxy-terminated polydimethylsiloxane (PDMSat, ABCR GmbH, Karlsruhe, Germany), and two polyether-modified polysiloxane polyurethane acrylates (PESiUA1 and PESiUA2) whose synthesis and structures are described elsewhere [24]. Table 1 provides molecular weight, number of repeating siloxane units, and amount of Si in the three comonomers.

**Table 1.** Molecular weight, atomic and mass concentrations of Si for the siloxane comonomers.


A total of 12 formulations were prepared with the three comonomers, at concentrations equal to 0.5, 1, 2, and 5 wt %. For each formulation, the hyperbranched acrylate was first mixed with a concentration of 6 wt % of TPO at 75 ◦C and stirred for 15 min using a magnetic stirrer, until the mixture was homogenous. The hyperbranched acrylate + TPO was used as the control material and will be referred to as the acrylate in the following. PDMSat formulations were prepared by mixing selected PDMSat amounts with the acrylate at 40 ◦C for further 15 min. PESiUA1 and PESiUA2 formulations were more difficult to prepare due to the very high viscosity of these two comonomers, which prevented accurate dosage and homogeneous mixing with the acrylate. To overcome this problem, the comonomers were heated for 10 min at 50 ◦C and selected amounts could be mixed with the acrylate and stirred for at least 12 h at ambient temperature. All formulations were transparent and stirred again prior to further processing in order to guarantee a good homogeneity.

#### *2.2. Process Methods*

Flat surfaces were produced in a first step in order to study the migration of siloxane oligomers. 200 μm thick coatings were prepared on glass slides using a doctor blade and photopolymerized immediately, or after selected times to allow monomer migration to the free surface exposed to air. Curing was then performed using a 200 W high-pressure mercury lamp (OmniCure 2000, EXFO, Mississauga, ON, Canada) and a collimator positioned at 12 cm above the sample during 3 min under a UV intensity of 75 mW·cm−<sup>2</sup> at the surface of the sample. The light intensity at the surface of the sample was measured between 230 and 410 nm using a calibrated radiometer (Silver Line, CON-TROL-CURE, Chicago, IL, USA). Notice that for the surfaces photopolymerized under air, the concentration of photoinitiator was high enough to overcome oxygen inhibition, and the cured surfaces were non-sticky and hard. This was further checked by measuring the water contact angle (WCA): Surfaces polymerized under N2 and under air showed the same wettability (see Table 2).


**Table 2.** WCA and WSA of flat and texturized surfaces, varying air exposure time, creep time, and template materials. 1: 50 μL droplets.

Texturized surfaces were produced in a second step. The leaves of two plants were selected, namely lotus (*Nelumbo nucifera*) and nasturtium (*Tropaeolum majus*). The WCA of the lotus and nasturtium leaves, reported in Table 2, were found to be approximately 20◦ lower than previously reported values of 164◦ [25] and 162◦ [26], respectively, due to seasonal factors. The surface of the fresh leaves was used as master and was replicated in the formulations using a UV-nanoimprint lithography process (UVNIL) and an intermediate negative polydimethylsiloxane mold (PDMS, SYLGARD™ 184, Dow, Midland, MI, USA) as detailed in [17]. This soft, vacuum-free and ambient molding technique preserves the delicate biological surfaces from damage and enables to accurately reproduce their nanometer scale features [27,28]. In short, square samples (2 cm × 2 cm) cut from the leaves were fixed onto Petri dishes. PDMS was mixed with hardener (10:1 ratio). The mixture was homogenized

manually for 5–10 min and degassed under a reduced pressure of 50.8 kPa for 5 min and 84.7 kPa for 10 min. It was poured onto the samples and subsequently cured at room temperature for 48 h. A UVNIL tool equipped with independent control of UV exposure and pressure was used to print the composite surface with the PDMS mold. A 200 μm thick layer of the liquid formulations was applied on a glass slide using a doctor blade. The PDMS mold was attached to another glass slide and the PDMS surface was put in contact with the liquid formulation under a controlled pressure of 3 bars. Curing was then performed using the same source and same UV dose as for the flat films. The printed surfaces were finally carefully demolded.

#### *2.3. Characterization Methods*

The water contact angle (WCA) of the polymerized surfaces was measured using a contact angle meter (EasyDrop, Krüss GmbH, Hamburg, Germany) at room temperature, with deionized water and a droplet volume of 10 μL. Four WCA measurements were made on each sample and the values were averaged. The water sliding angle (WSA) of selected surfaces was measured using a tilting support equipped with a protractor. A droplet of water was placed on the surface of the sample in horizontal position and the support was slowly tilted until the drop started to move. Measurements were made at room temperature using deionized water. The volume of the droplets varied from 10 to 100 μL.

The kinetics of the photopolymerization process were analyzed in real time during irradiation, by real-time Fourier transform infrared spectroscopy (RT-FTIR, Thermo-Nicolet 5700 spectrometer, Thermo Fisher, Waltham, MA, USA) on 12 μm thick coatings. The results are reported in the Supplementary Materials (Figure S1). Three formulations were tested: acrylate control, acrylate with 0.5 wt % of PESiUA2 and with 5 wt % of PESiUA2. The coatings were applied on a silicon wafer using a wire-wound Meyer bar; the IR experiments were made immediately or after 120 min exposure to air prior of starting the irradiation. The IR instrument was settled with a resolution of 4 cm−<sup>1</sup> and the acquisition rate was 1 Hz. A 200 W high pressure mercury-xenon lamp (LC8, Hamamatsu, Shizuoka, Japan) was used and the intensity of the UV light was fixed at 66 mW·cm<sup>−</sup>2. The conversion was calculated recording the decrease of the area of the absorption band of the C=C double bonds with time *t*, using the methacrylate peak at 1636 cm−<sup>1</sup> (*P*C=C(*t*)) [29]. The peak area of the C=O double bond at 1726 cm−<sup>1</sup> (*P*C=O) was chosen as the reference. The degree of conversion α was calculated as

$$\alpha = 1 - \frac{P\_{\text{C=C }}\left(t\right) / P\_{\text{C=O}}\left(t\right)}{P\_{\text{C=C }}\left(0\right) / P\_{\text{C=O}}\left(0\right)}\tag{1}$$

The influence of short flashes of UV light on the shear viscosity of selected formulations was determined using oscillatory shear rheometry and a UV-coupling cell (AR2000, TA instruments, New Castle, DE, USA), with a plate–plate geometry of diameter 20 mm and a gap of 500 μm. Samples were tested before, and after being illuminated for short periods. A strain amplitude sweep was performed first at 1 Hz to determine the limit for the linear viscoelastic range. Frequency sweep tests were then performed at a strain amplitude within the linear range, and usually close to 1%. The UV source and light intensity were the same as used for UVNIL experiments (75 mW·cm<sup>−</sup>2). The rheology results are reported in the Supplementary Information.

The topography of the polymer surfaces was observed using a scanning electron microscope (SEM, FEI XLF30-FEG, Philips, Amsterdam, The Netherlands). The SEM was operated in high resolution mode using an acceleration voltage of 5 kV. The working distance was fixed to 10 mm. The samples were coated with a thin carbon layer of approximately 12 nm in order to avoid charging effects.

Chemical surface composition was evaluated by means of a PHI 5000 Versa-Probe scanning X-ray photoelectron spectrometer (XPS, PHI Versaprobe 5000, Physical Electronics, Inc., Chanhassen, MN, USA), with a monochromatic Al Kα source at 1486.6 eV. A spot size of 100 μm was used in order to collect the photoelectron signal for both the high resolution (HR) and the survey spectra. The semi-quantitative atomic compositions and deconvolution procedures were obtained using Multipack 9.7 dedicated software. All core level peak energies were referenced to C 1*s* peak at

284.5 eV (C–C) and the background contribution has been subtracted by means of a Shirley function. Depth profile has been performed using the Ar+ source with a 2 kV ions accelerating voltage, alternate mode with sputter cycle of 30 s each.

#### **3. Results and Discussion**

#### *3.1. Segregation of Siloxane Comonomers towards Polymer–Air Interface*

The influence of the addition of the siloxane comonomers and their concentration on the wettability of the cured polymer surfaces at the air side was investigated by WCA measurements, in view of selecting the comonomer leading to the most hydrophobic surface, and the results are depicted in Figure 1. When photopolymerized onto the glass substrate, the WCA of the pure acrylate was 61◦.

**Figure 1.** WCA vs. concentration of siloxane comonomer of flat surfaces polymerized without air exposure (**a**) and after 120 min of air exposure (**b**).

The addition of the three comonomers led to an increase of the WCA up to a saturation level around 1 wt %, beyond which the concentration increase had a marginal influence. Formulations with PESiUAs were beyond the hydrophobic limit independently of the concentration with WCA values comparable to the value for the PESiUAs homopolymers (104◦). They were lower than the value of 113◦ for the PDMS SYLGARD™ 184 (Table 2), that is close to the maximum WCA of 115.2◦ for a flat surface [30]. In contrast PDMSat formulations only reached the hydrophobic limit after an exposure time of 120 min, although the WCA on PDMSat homopolymers were found to be equal to 97◦.

Comparing the three siloxane comonomers, it is evident that the parameter influencing the surface activity of the comonomers was the length of the siloxane chain, i.e., the number of siloxane units. PDMSat contains only three units, which were not sufficient to impart a hydrophobic character. The dependence of wettability on the concentration and on the length of the apolar moiety of surface active comonomer has been already reported for fluorinated systems [17,31,32].

As detailed in Supplementary Materials (Figure S2), the delay time between coating application and irradiation processes was effective for the surface modification: comonomers diffused to the polymer–air interface, with a diffusion time depending on the siloxane structure. The higher the molecular weight the faster the diffusion, due to the reduced affinity with the bulk [32]: the PESiUA2 diffusion was immediate after coating, while PESiUA1 and PDMSat required longer times to reach the highest WCA values. In summary, PESiUA2 was the best choice to obtain the highest hydrophobicity of photocured films: it allowed obtaining the highest WCA without a delay time. However, the values were far away from the superhydrophobicity threshold (150◦), therefore a change in surface morphology was required besides the modification of the chemical composition of the surface.

#### *3.2. Influence of Texturization with PDMS Templates*

The photopolymer showing the highest WCA for flat surfaces (acrylate + 5 wt % PESiUA2) was used to obtain texturized materials. The acrylate homopolymer was imprinted as reference, and Figure 2 shows the morphology of the synthetic replica of nasturtium and lotus leaves. The nasturtium replica was characterized by entangled island-like epithelial cells with dimensions around 100 μm [26]. The lotus replica exhibited a high density of micropapillae of diameter around 10 μm. A closer look revealed that the sub-micron epicuticular crystal structures were, however, not well replicated, in particular in the case of nasturtium. Such a difference in small-scale structures between the two types of plants was surprising since both natural surfaces are quite similar at this scale, with comparable tubular wax crystal morphologies. A first explanation is related to the possible erosion of the superficial waxes on old leaves [19], which would explain the rather low values of WCA on the nasturtium. In addition, the chemical composition of the lotus and nasturtium waxes differs totally [18], and one additional hypothesis would be that the nasturtium wax tubules partly dissolved in PDMS during the 48 h curing time [33]. Further work would be needed to clarify the observed differences.

**Figure 2.** Electron micrographs of the synthetic replica made of photocured acrylate homopolymer: nasturtium (**A**) and higher magnification inset (**B**); lotus (**C**) and higher magnification inset (**D**).

Table 2 summarizes the WCA data for photopolymers having either flat surfaces or texturized surfaces. The WCA of the synthetic nasturtium surfaces with PESiUA2 comonomer was found to be slightly above 100◦, as for the corresponding flat surface, and slightly higher than that of the synthetic nasturtium surface without comonomer. These values were 50◦ below the superhydrophobic threshold and some 40◦ below the WCA of the fresh nasturtium leaf. This result confirms that the nanosized epicuticular crystal features of the plant, which were not well replicated, are essential to achieve superhydrophobic properties. In contrast, the WCA of the synthetic lotus surfaces with PESiUA2 comonomer was found to be in the range 142◦–144◦, slightly higher than that of the fresh leaf and acrylate lotus surface, and very close to the superhydrophobic limit. In fact, the lotus replica was rather accurate (Figure 2C). This result further indicates the key role of the texture to promote a superhydrophobic state. Nevertheless, the WSA of the synthetic lotus surfaces also reported in Table 2 was rather large and much higher than the WSA of 3◦ of the plant, so that these synthetic replicas, although accurate were not superhydrophobic. Additional WSA data are reported in the Supplementary Materials (Table S1).

Surprisingly, the WCA of the flat surfaces with PESiUA2 comonomer polymerized in contact with a flat PDMS mold was 91◦, i.e., below the values obtained without using the mold and irrespective of air exposure time prior to curing. This may imply that the contact with the PDMS template during the UV printing process led to a reverse migration of the comonomer as was reported for the case of fluorinated moieties [4]. All these measurements reveal the competing influence of PDMS and presence of siloxane monomer on the surface segregation of the latter.

Figure 3 shows XPS scans of the flat surfaces of acrylate and acrylate with PESiUA2 polymerized with and without 120 min of exposure to air prior to polymerization, and the lotus texturized surface of acrylate with PESiUA2 after 120 min of exposure to air. Apart from carbon and oxygen, present in all the surfaces, Si 2*p* and N 1*s* signals were carefully analyzed as fingerprints of the PESiUA2. The relative atomic concentration of these two elements is reported in Table 3. The aim was to first to confirm the migration of the PESiUA2 towards the polymer–air interface, and second to check the presence of this molecule at the surface of the texturized surface. The Si 2*p* and N 1*s* signals were indeed absent in the plain acrylate sample. The migration process was confirmed, with an increase of the superficial concentration of these two elements with time prior to polymerization, which is consistent with the increase of WCA with delay time shown in Figure 1. The concentration of Si at the surface of the lotus-texturized sample was found to be higher than that on the flat samples. Possible explanations are the sensitivity of the XPS measurement to surface roughness, and differences in superficial concentration of PESiUA due to different curing kinetics [34], hence different migration processes between the flat and the texturized materials.

**Figure 3.** (**A**–**D**) XPS Survey scans for the four samples analyzed (see Table 3 for details). C 1*s*,O1*s*, N 1*s*, and Si 2*p* peaks have been highlighted. Depth profiles of C 1*s*,O1*s*,N1*s*, and Si 2*p* of (**E**) Sample C and (**G**) Sample D. Si 2*p* high resolution curves vs. sputter cycles for (**F**) Sample C and (**H**) Sample D.


**Table 3.** Relative atomic concentration of Si and N at the surface of acrylate with and without PESiUA2, for flat and texturized surfaces.

The in-depth distribution of Si in the flat and texturized sample of acrylate with PESiUA2 polymerized after 120 min of exposure to air was further analyzed by depth-profiling using an Ar<sup>+</sup> source and is also shown in Figure 3. Assuming an etching rate of 5.5 nm/min for acrylate as previously determined for a BEMA–PEGMA siloxane enriched copolymer [35], we can see that there is an exponential decay of the Si 2*p* signal in the first 5 nm for the flat sample (Figure 3E), and in the first 10 nm for the texturized sample, beyond which the signal smoothly decreases till 50 nm in depth (Figure 3G). In Figure 3F,H, we have reported the Si 2*p* high resolution curves acquired during depth profiles for the two samples, to better highlight the signal decrease. Again, the differences in depth profiles may result from the sensitivity of the XPS measurement to surface roughness.

The XPS analyses imply that the siloxane comonomers did not fully migrate back into the bulk upon contacting the PDMS surface, in contrast with fluorinated comonomers [4]. To further check this, and totally prevent the trend for reverse migration, flashes of UV light were applied to the free surface of coatings after migration to chemically immobilize the surfactant before UVNIL. Acrylate coatings with 5 wt % of PESiUA2 exposed to air for 120 min were flashed for periods of 0.2 s up to 2 s before UVNIL with the negative PDMS lotus template. The WCA of the flashed, texturized surfaces turned out to be lower than that of surfaces produced without UV flash. As shown in the Supplementary Materials (Figure S3), this reduction of WCA was due to the large increase of the viscosity of the liquid formulation, by more than four orders of magnitude at low shear rates, and emergence of a yield stress behavior for flashes as short as 0.2 s. This considerable thickening compromised the fidelity of the low-pressure replication process. An alternative strategy was therefore tested, based on viscoelastic creep flow.

#### *3.3. Influence of Creep*

The low-pressure UVNIL process is fundamentally based on the viscoelastic creep flow of the resin into the sub-micron topography of the PDMS template, so that the replication is pressure and time-dependent. In order to allow the resin to better fill the mold cavities, a pressure of 3 bars was applied to the resin in contact with the PDMS mold for periods up to 10 min prior to UV curing. Figure 4 shows the influence of creep on the WCA of synthetic lotus surfaces, based on the acrylate with and without 5 wt % of PESiUA2 (SEM images of these surfaces are shown in Figure S4). In both cases, the WCA increased with creep to a maximum value of 152◦ (shown in Figure S5) with PESiUA2 after 5 min and then decreased. Nevertheless, the WSA reported in Table 2 was found to remain rather high and equal to 30◦.

**Figure 4.** WCA of synthetic lotus surfaces based on acrylate and acrylate + 5 wt % PESiUA2 vs. creep time under 3 bars prior to UVNIL.

#### *3.4. Self-Cleaning*

Although the WSA was equal to 30◦ the synthetic acrylate + PESiUA2 lotus surface possessed effective self-cleaning properties. This was demonstrated by letting 10 μL water droplets fall from a height of 1 cm on the surface partly covered by ground pepper, on a glass support tilted at a 10◦ angle. A sequence of three images taken from a video are shown in Figure 5, where the capture of pepper grains by one bouncing droplet is evident (the video is available in the Supplementary Materials). Notice that, due to its hydrophobicity, ground pepper was reported to be more challenging to remove than other hydrophilic particles such as MnO and SiC [8]. The drop velocity upon impact was close to 0.4 m/s, which was high enough for the drop to bounce on the surface as observed (Figure 5b and Video S1 in the Supplementary Materials) [36]. The corresponding kinetic energy of the drop was close to 1 μJ, which again was high enough to overcome the effective interfacial energy *E* close to 0.07 μJ. The latter was roughly estimated as

$$E \sim \text{y}\_{\text{s}} f \pi \text{R}^2 \tag{2}$$

where γsl is the water-polymer interfacial tension, found to be close to 40 mJ/m2 from the measured WCA data, *f* = (cosθ + 1)/(*r*cosθ<sup>0</sup> + 1) is the wet area fraction calculated and found to be equal to 13% using the Cassie–Baxter model [37] (θ and θ<sup>0</sup> represent the WCA of the texturized and flat surfaces, respectively, and *r* is the roughness factor, equal to 3.2 for lotus), and *R* is the radius of the droplet at the impact point, evaluated as 2 mm from the video images. Important to point out is that no such self-cleaning behavior was observed in the case of flat surfaces.

**Figure 5.** Sequence of photographs showing the behavior of a water droplet falling (**a**), bouncing (**b**), and rolling (**c**) on a synthetic acrylate lotus surface with 5 wt % of PESiUA2, printed on a glass support with 1 min of creep under 3 bars. The synthetic lotus surface was contaminated by pepper grains and support was tilted by 10◦. Image (**b**) shows the bouncing droplet with trapped pepper grains, leaving a clean impact trace. Image (**c**) shows the water drop with trapped pepper grains, sticking on the smooth glass support surface after leaving the printed lotus surface.

The behavior of the synthetic lotus surface based on the acrylate with 5 wt % of PESiUA2 to other liquids than water is shown in Figure 6. The difference of contact angle is evident, confirming the hydrophobic character of the surface.

**Figure 6.** Coffee, olive oil, and water droplets from left to right on a synthetic lotus surface based on acrylate with 5 wt % PESiUA2 and 120 min air exposure prior to UVNIL.

#### **4. Conclusions**

Bio-inspired fluorine free synthetic replica of lotus and nasturtium surfaces were developed in view of obtaining self-cleaning properties, based on an energy efficient UV printing process. A UV-curable acrylate oligomer and three acrylated siloxane comonomers with different molecular weights were used. The process combined the self-assembly of the comonomers at the acrylate/air interface and a UVNIL replication step with a PDMS negative replica of the plant surfaces. The morphology and water contact angle of flat and texturized surfaces were systematically analyzed, leading to the following conclusions.

The migration of the comonomers towards the polymer–air interface led to an increase of the hydrophobicity of the surfaces, from 61◦ for the plain acrylate to 105◦ for the acrylate with 5 wt % of PESiUA2 after 30 min of exposure to air prior to photopolymerization. Comonomers with a higher molecular weight migrated faster.

The replication fidelity was excellent for the lotus leaf surface whose hierarchical texture comprising micropapillae and sub-micron crystalloids was well reproduced in the acrylate/ comonomer material. It was less accurate for the nasturtium surface, with a lack of replication of the sub-micron features. The texturization of the surfaces had a large influence on hydrophobicity. The WCA of synthetic replica of lotus and nasturtium with 5 wt % of PESiUA2 was equal to 144◦ and 102◦, respectively. The highest WCA, equal to 152◦ was measured on a synthetic lotus surface with 5 wt % PESiUA2 and 5 min of creep flow under a pressure of 3 bars prior to polymerization. The WSA of this surface was found to be equal to 30◦ due to weak adhesion of water. The surface was nevertheless self-cleaning with water droplets when contaminated with hydrophobic pepper particles, provided that the droplets had some kinetic energy.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-6412/8/12/436/s1, Figure S1: Photoconversion vs. time for HBP and HBP-PESiUA2 (0.5 and 5 wt %) without and with 120 min of air exposure prior to measurements; Figure S2: WCA vs. air exposure time of flat surfaces of acrylate and acrylate-siloxane formulations: (a) PDMSAT; (b) PESiUA1; and (c) PESiUA2; Figure S3: Viscosity of acrylate (a), acrylate + 0.5 wt % PESiUA2 (b), acrylate + 5 wt % PESiUA2 (c) as function of angular frequency, for increasing UV flash duration (0.2−2 s) as indicated; Figure S4: Scanning electron micrographs of the synthetic lotus surfaces, based on the acrylate with 5 wt % PESiUA2, without (a) and with (b) a creep time of 5 min under a pressure of 3 bars prior to photopolymerization; Figure S5: WCA of 151.9◦ measured on a synthetic lotus replica surface with 5 wt % PESiUA2 and 5 min of creep; Table S1: WSA of synthetic lotus surfaces; Video S1: MVI\_8122.mv4.

**Author Contributions:** Conceptualization, S.D.V. and Y.L.; Data Curation, L.W., L.M., and M.C.; Formal Analysis, L.W., S.D.V., F.K., L.M., M.C., A.V., R.B., and Y.L.; Funding Acquisition, Y.L.; Investigation, L.W., S.D.V., F.K., A.V., R.B., and Y.L.; Methodology, R.B.; Supervision, S.D.V., R.B., and Y.L.; Writing—Original Draft Preparation, L.W. and M.C.; Writing—Review & Editing, S.D.V., F.K., L.M., A.V., R.B., and Y.L.

**Funding:** This research was partly funded by EPFL's Integrated Food and Nutrition Center.

**Acknowledgments:** The authors acknowledge the Botanical Garden of Lausanne for the supply of fresh plants.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Coatings* Editorial Office E-mail: coatings@mdpi.com www.mdpi.com/journal/coatings

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18