Ultrafast Processing of Hierarchical Nanotexture for a Transparent Superamphiphobic Coating with Extremely Low Roll‐Off Angle and High Impalement Pressure

Low roll‐off angle, high impalement pressure, and mechanical robustness are key requirements for super‐liquid‐repellent surfaces to realize their potential in applications ranging from gas exchange membranes to protective and self‐cleaning materials. Achieving these properties is still a challenge with superamphiphobic surfaces, which can repel both water and low‐surface‐tension liquids. In addition, fabrication procedures of superamphiphobic surfaces are typically slow and expensive. Here, by making use of liquid flame spray, a silicon dioxide–titanium dioxide nanostructured coating is fabricated at a high velocity up to 0.8 m s−1. After fluorosilanization, the coating is superamphiphobic with excellent transparency and an extremely low roll‐off angle; 10 µL drops of n‐hexadecane roll off the surface at inclination angles even below 1°. Falling drops bounce off when impacting from a height of 50 cm, demonstrating the high impalement pressure of the coating. The extraordinary properties are due to a pronounced hierarchical nanotexture of the coating.

of the applications, such as wind screens, window panes, lenses, or protective goggles, the superamphiphobic coating should be optically transparent, mechanically stable, and capable of repelling liquid impalement even under high hydrostatic pressure or drop impact. [11] The capability of superamphiphobic surfaces to repel low-surface-tension liquids arises from a combination of their overhang, inward curved surface morphology and low-surface-energy chemistry. Overhanging morphology and low surface energy are required to stabilize an air cushion below the drops and maintain the so-called Cassie-Baxter state. [12] An exception to this is a doubly re-entrant micropillar surface introduced by Liu and Kim, [13] where a low-surface-energy chemistry is not needed. The required surface morphology for superamphiphobic materials can be realized with well-defined, mushroom-like micropillars, [2,13,14] or with random, sub-micrometer scale surface textures with overhang curvature. The micropillars are typically fabricated by reactive ion etching. [2,13] Randomly structured superamphiphobic surfaces have a higher potential for scaled up production. Methods to fabricate these surfaces include growth of silicone nanofilaments [3,10] and templating candle soot. [4] Because the morphology required for superamphiphobic surfaces is rather complex, the number of suitable approaches is limited. Most methods are energy-, chemical-, or time-consuming with multiple process steps. [5,7] Furthermore, it is still a challenge to fabricate superamphiphobic surfaces which combine high receding contact angles with high impalement pressure and mechanical robustness.
Spray methods are potential candidates for scaled up fabrication of super liquid-repellent surfaces. [15,16] Particularly, liquid flame spray (LFS) has been applied to produce superhydrophobic surfaces in high-speed roll-to-roll processes. [17][18][19] In LFS, a liquid feedstock is injected and atomized in an oxygenhydrogen flame. Dissolved in the liquid are organometallic precursor molecules. After evaporating and reacting in the flame they form nanoparticles. These nanoparticles are collected on the surface. With the heat from the flame the particles partially sinter together to form a stable, highly porous film. Advantages of LFS are that the deposition process is solvent-free and takes only fractions of seconds as the sample is rapidly moved www.advmat.de www.advancedsciencenews.com through the flame spray even at the velocities of the order of m s −1 . In addition, a broad range of materials including vulnerable biomaterials, such as cellulose-based paper and wood can be coated. [16,20] A certain minimal velocity is required to avoid destroying the substrate.
Here we use LFS to fabricate a superamphiphobic and optically transparent silicon dioxide (SiO 2 )-titanium dioxide (TiO 2 ) nanoparticle coating on glass. Our coating shows minimal solid-liquid interactions for both high-and low-surface-tension, polar and nonpolar liquids. Drops of water and n-hexadecane (10 µL) deposited on the surface easily roll off the coating at inclination angles <1°. To our knowledge, this is the lowest rolloff angle toward hexadecane ever reported. To achieve these superior properties, first, we adjusted the surface morphology by varying the ratio of silicon dioxide and titanium dioxide in the coating. Second, after achieving the right morphology, we applied chemical vapor deposition (CVD) of a 1H,1H,2H,2Hperfluorooctyl-trichlorosilane (97% pure, Sigma-Aldrich) to lower the surface energy. In this way, we left the nanoporous morphology of the coating intact. This is necessary to achieve the superamphiphobic properties-already a 20 nm thick additional layer on top of the nanoparticles hinders liquid repellency.
To synthesize the surfaces by LFS, we use hydrogen (50 L min −1 ) and oxygen (15 L min −1 ) as combustion gases to achieve a turbulent, high temperature flame (>2500 °C), [19] and inject the liquid feedstock, tetraethyl orthosilicate (TEOS, 98% pure, Alfa Aesar) and titanium(IV) isopropoxide (TTIP, 97% pure, Alfa Aesar) dissolved in isopropanol (technical grade, Neste), into the flame through a custom-made spray torch at a feed rate of 12 mL min −1 (the overall Si+Ti atomic concentration in the precursor solution was kept constant at 50 mg mL −1 ), Figure 1. The organometallic precursors react and nucleate in the flame to form nanosized oxide particles. The particles aggregate and are deposited directly on the surface-driven by diffusion and thermophoresis through a boundary layer of air at the substrate [19] -to form a porous coating. More details of LFS method are given elsewhere. [17][18][19] Silicon dioxide and titanium dioxide were selected as coating materials since they are widely used in different coating applications such as painting, [21] cast-, [22] dip-, [23] spray-, [18] and vapor-phase [4] deposition. Titanium dioxide is well-known for its photocatalytic activity. This property can be utilized in self-cleaning coatings [24] and to decompose atmospheric pollutants such as nitrogen oxides (NO x ). [21] We first investigate potential of a pure silicon dioxide coating (Si/Ti ratio = 100/0 wt% in the precursor) to form the overhang morphology. From now on, we call this "Si 100 wt% coating." The coating was synthesized on a smooth glass substrate by injecting TEOS diluted in isopropanol into the upward pointing LFS flame, through which the sample was moved at a velocity of 0.8 m s −1 at the distance of 6 cm from the burner face. Scanning electron microscopy shows that the resulting coating is ≈100 nm thick and is composed of highly sintered, round sub-micrometer scale clusters evenly distributed on the surface (Figure 1a,g). After fluorination the surface shows moderate liquid repellency with apparent static contact angles of 138° for water (surface tension γ = 72.8 mN m −1 ), 118° for ethylene glycol (γ = 47.3 mN m −1 , 99.8% pure, Sigma-Aldrich), and 83° for n-hexadecane (γ = 27.5 mN m −1 , 99% pure, Sigma-Aldrich).
To enhance liquid repellency, we use two approaches ( Figure 1). First, we substitute part of the silicon dioxide by titanium dioxide in the coating by adding TTIP to the precursor solution ( Figure 1, leftmost column). Second, we increase the thickness of the coating by applying 5 LFS coating cycles prior to fluorination (Figure 1, middle column). To gain insight on the individual agglomerate morphology, we collected particles from LFS on transmission electron microscopy grids (Figure 1, rightmost column).
Already 1 wt% addition of titanium atoms with respect to silicon atoms in the precursor (Si/Ti ratio = 99/1 wt%) drastically changes the morphology of the coating. We call this "Si 99 wt% coating," Figure 1c,h. An energy dispersive X ray spectroscopy (EDS/EDX) analysis indicates that the Si/Ti ratio within the coating is 96.2/3.8 wt% ( Figure S1 and Table S1, Supporting Information). The changes become more prominent when the titanium content was increased up to 99 wt% (Si/Ti ratio = 1/99 wt%). We call this "Si 1 wt% coating," Figure 1e,i,j. Silicon dioxide is no longer aggregated in highly sintered, dense clusters. Instead, the coating shows increasing amount of porous, nanosized particle aggregates with overhang structures. EDS/EDX analysis shows that the Si/Ti ratio within the coating is 2.1/97.9 wt%. Titanium dioxide exists mainly as anatase with small fraction, 10-15%, of rutile independently on the Si/Ti ratio. Anatase is known to be photocatalytically more active than rutile. [25] Silicon dioxide remains amorphous. [18] We speculate that these morphological changes are caused by an early nucleation of titanium dioxide in the cooling flame, while silicon dioxide remains still in vapor phase. [18,19] Titanium dioxide particles thus act as nucleation sites for silicon dioxide and facilitate formation of the porous particle aggregates within the coating. Silicon dioxide, which sinters at lower temperature than titanium dioxide, acts as a "binding agent" within the coating and thus enhances its mechanical stability (see discussion with drop impact and sand abrasion experiments).
Then we increase the coating thickness by moving the samples through the flame spray 5 sequential times at intervals of ≈2 s. The growth mechanism of the coating through the boundary layer of air at the surface-driven by thermophoresis and diffusion-induces accumulative growth of large particle aggregates at the surface ( Figure S2, Supporting Information). As a consequence, the height of the surface protrusions and hierarchical roughness of the coating increase (Figure 1b,d,f k; and Figures S3 and S4, Supporting Information). The final height of the surface textures depends on the coating composition. With highly sintered, dense Si 100 wt% coating the highest protrusions reach ≈700 nm after 5 coating cycles ( Figure S3, Supporting Information). With Si 1 wt% coating the highest peaks of the surface texture are ≈700 nm already after the first coating cycle (Figure 1j) and reach a height of at least 7 µm after 5 cycles ( Figure 1k; and Figure S4, Supporting Information).
After a single LFS coating cycle, referred to as "thin coating," the best liquid repellency is given by Si 1 wt% coating. Water drops deposited on the surface adapt a spherical shape with static contact angle >160°. 10 µL sized drops roll off the surface as soon as the substrate is inclined by less than ≈3°. With ethylene glycol and n-hexadecane, the static contact angles www.advmat.de www.advancedsciencenews.com approach 150° but the drops pin to the surface, i.e., roll-off angles are typically >10° ( Table 1).
A coating with superamphiphobic properties was achieved by coupling the two approaches, i.e., by reducing the amount of Si/Ti ratio in the precursor from 100/0 wt% to 1/99 wt% and by increasing the number of coating cycles from 1 to 5, referred to as "thick coating." In this way, we obtained a roll-off angle below 1° for 10 µL n-hexadecane drops. Advancing and receding contact angles were 165° with nonmeasurable contact angle hysteresis when the drop volume was increased and decreased between 15 and 25 µL at the rate of 1 µL s −1 using a standard contact angle goniometer ( Figure S5 and Video S1, Supporting Information).
The superamphiphobic coating developed here shows extremely low interaction with water and even with n-hexadecane. To our knowledge, the lowest roll-off angles reported for n-hexadecane on superamphiphobic surfaces are 2°-5° depending on the surface and the drop size. [2][3][4][5]7,9,10,16] Here, 10 µL drops of n-hexadecane typically rolled off the surface as soon as the goniometer needle tip was detached although the substrate was adjusted in horizontal plane without any apparent inclination angle (Video S2, Supporting Information).  and e,f) Si 1 wt% coating. Insets: the shape of 5 µL water (left) and n-hexadecane (right) drops resting on the respective surfaces. Transmission electron microscopy (TEM) images show different degree of sintering and overhang morphology of the particle aggregates: g) Si 100 wt%, h) Si 99 wt%, and i) Si 1 wt% coating. Side-view SEM images of j) Si 1 wt% thin coating (coated 1 time) and k) Si 1 wt% thick coating (coated 5 times).

www.advmat.de www.advancedsciencenews.com
Our coating consists of random, overhang nanostructures where the diameter of spherical primary particles is ≈10-20 nm. We investigated the effect of texture size on water repellency by growing an additional 20 nm thick silicon dioxide layer on the surface using a gas-phase Stöber-like reaction. [4] After growing the silicon dioxide layer ( Figure S6, Supporting Information) and modifying the surface with the fluorosilane, the antiwetting performance declined. For all coating compositions (Si 100 wt%, Si 99 wt%, and Si 1 wt%) the water contact angle decreased and the roll-off angle increased due to the increased solid-liquid contact area and smoothed out overhangs as compared to the pristine coatings (Table S2, Supporting Information). This underlines the role of nanosized texture in reducing the solid-liquid interactions on the coating. Coupling this nanosized texture with pronounced hierarchical surface roughness supports the air cushion below the liquids and leads to extremely small overall contact area between the coating and the liquids.
We expect that the extreme liquid repellency of the thick Si 1 wt% layer is caused by the hierarchical surface roughness in addition to the overhanging morphology on the 10 nm scale. A simple estimation on a surface consisting of spherical, randomly aggregated particles shows that this combination ensures low penetration depth and wetted contact area of both polar and nonpolar liquids on the solid substrate, Figure 2. Knowing the mean radius of the particles r and the Young contact angle θ on the solid, fluorinated silicon oxide, the penetration depth δ around a single particle on the surface in thermodynamic equilibrium at zero external pressure can be estimated (Figure 2a, Supporting discussion in the Supporting Information) to be δ ≈ 0.58r for water (θ = 115°) and δ ≈ 1.29r for n-hexadecane (θ = 73°). Here, we assumed that the particle is fixed at the bottom. Taking into account the random packing in the porous structure, the number of wetted particles underneath the first layer of particles will increase before reaching θ (Figure 2b). This increase in the wetted contact area is larger for nonpolar liquids than for polar, high-surface tension liquids. That is, overhangs can support the air cushion below low-surface-tension liquids, however, the liquid still wets large surface area and therefore pins to the solid (Supporting discussion, Supporting Information). To reduce the overall wetted area of the solid and thus adhesion of drops, hierarchical roughness needs to be introduced (Figure 2c). Indeed, several surfaces with inherent hierarchical roughness such as paper, [9] wood, [16] and fabrics [7] serve as ideal substrates for randomly structured superamphiphobic coatings where both polar and nonpolar liquid drops can bead up and easily roll off the surface.
To achieve superamphiphobicity, a nanoscopic overhanging structure needs to be combined with roughness on the >1 µm length scale; in the following we call this a hierarchical structure. Such hierarchical structures need a certain minimum coating thickness. Below this minimum thickness, it might not be possible to create a superamphiphobic surface. The overall solid-liquid contact area would become too large.
Adv. Mater. 2018, 30, 1706529 Table 1. Wettability of the liquid-repellent coatings. Apparent static contact angles (CA) and roll-off angles (RA) of 10 µL drops of water, ethylene glycol, and n-hexadecane on coatings with different silicon dioxide content and thickness after chemical vapor deposition of the fluorosilane. "Thin" refers to a single LFS coating cycle. "Thick" refers to 5 subsequent LFS coating cycles. The standard deviations are given by individual contact angle goniometer measurements. Note that contact angles larger than ≈155° cannot reliably be measured using the goniometer technique and thus the real error is larger.  Figure 2. Schematic illustration of wetting of a model surface by water and a nonpolar liquid. The surface consists of spherical particles. Penetration depth δ of the liquid around a single particle with radius r depends on the intrinsic wettability of the material, characterized by the Young contact angle θ. a) Water (large θ) wets small fraction of individual particles within the first particle layer, indicated by the dashed line in (b). A nonpolar liquid (small θ) wets large fraction of individual particles and b) invades from one particle to the other into the texture of the solid until θ is reached at the overhangs. c) Hierarchical roughness of the surface has critically important role in reducing the overall solid-liquid contact area and pinning of lowsurface-tension liquids on randomly structured superamphiphobic surfaces.
www.advmat.de www.advancedsciencenews.com This condition poses a lower limit to the thickness of a superamphiphobic coating on a smooth substrate such as plain glass. Here the coating was not superamphiphobic after one coating cycle when its thickness was ≈700 nm (Figure 1j). However, after 5 coating cycles the coating became superamphiphobic (Figure 1k; and Figure S4, Supporting Information). It is reasonable to assume that with most of the randomly structured superamphiphobic surfaces, the hierarchical surface structure increases with increasing coating thickness. For example, the candle soot templated coating remains superamphiphobic only when the coating is thicker than ≈2 µm. [4,26] We verify optical transparency of our coatings by ultraviolet-visible light transmittance spectroscopy. All thin coatings (coated 1 time) transmit more than 98% as compared to the transmittance through the pristine glass substrate (for wavelengths higher than 500 nm, Figure S7, Supporting Information). Thick coatings (coated 5 times) transmit 97% of the light for Si 100 wt% coating and 79% for Si 1 wt% coating (Figure 3a). High transmittance of light at the visible light spectrum results in good optical transparency (Figure 3b) of the super liquid-repellent coatings (Figure 3c).
In addition to liquid repellency and optical transparency, the impalement resistance of the coating decides about potential applications. How stable is the Cassie-Baxter state before the whole surface texture is wetted by the liquid and the system goes to the so-called Wenzel state? [27] We investigate the impalement resistance of our superamphiphobic coating by letting water drops impact the surface from different heights. Water drops of 15 µL volume (radius R = 1.5 mm) were released from heights of 1-200 cm leading to impact velocities v between 0.4 and 5.4 m s −1 . This approaches the terminal velocity of falling medium-sized rain drops (R < 1 mm). [28] Such an impact velocity and drop radius corresponds to Weber numbers up to We = ρv 2 R/γ ≈ 600. Here, ρ is the density of water = 1 g cm −3 . The drops always rebounded from the surface and no impalement was observed.
To prevent full or partial penetration of the impinging drops, [11] the capillary pressure P C generated within the textures should be higher than the maximal effective hammer pressure P E , which is the upper limit for the pressure the surface can experience during the impact. For drops impacting on horizontal flat surface, one can estimate the maximal hammer pressure [29] Here, C is the sound velocity (for water C = 1497 m s −1 ). For an impact velocity v = 5.4 m s −1 and water the hammer pressure can be estimated to be 1.6 MPa. The maximal capillary pressure developed within the particle-like surface texture to prevent the impalement can be estimated from [30] γ θ ≈ 2 sin 2 C 2 2 adv P r d (2) Here, d is the mean distance between protrusions, r is the radius of the constituting particles, and θ adv = 119° is the advancing contact angle of water on a smooth fluorosilane coated silicon oxide. When the hammer pressure P E exceeds P C , one expects the Cassie-Baxter state to collapse. With roughly r = 5-10 nm and setting P E = P C we get a required maximal spacing of protrusions = 18-26 nm. On our coating, the smallest pores between the particles and their aggregates can fulfill this criterion (Figure 1f,i). It is expected that the liquid will penetrate in between the largest protrusions on the surface. These protrusions dampen the impact and relief the pressure experienced by the surface in between them. For comparison, on a rectangular array of the fluorosilane modified SU8 micropillars with solid area fraction of 0.06 (5 µm side wall, 10 µm height, 15 µm spacing, fabricated by photolithography [31] ) impalement of impacting 15 µL water drops occurred already at the hammer pressure of ≈240 kPa (release height = 3 cm, impact v = 0.8 m s −1 , We = 12).
The superamphiphobic coating also repels impalement of impacting n-hexadecane drops (R = 1 mm, ρ = 0.773 g cm −3 , Figure 4). At 1 cm release height the 5 µL drop bounced 4 times before settling down at the surface (v = 0.4 m s −1 , We = 5.4, Video S3, Supporting Information). At 10 cm release height the drop bounced 6 times before settling down (v = 1.4 m s −1 , We = 55, Video S4, Supporting Information). The n-hexadecane drops even rebound when released from a height of 50 cm. The corresponding impact velocity for the drop was 3 m s −1 (We = 250). Above the release height of 50 cm, the kinetic energy becomes large as compared to the surface tension causing that the rim of the drop breaks up and many satellite drops are generated. Because of the breakdown of the drop, at higher release heights it becomes difficult to reliably determine whether the drop partially impales the coating. Calculating the pressure experienced by the surface during impact of the n-hexadecane drop with Equation (1)   www.advmat.de www.advancedsciencenews.com in n-hexadecane = 1339 m s −1 [32] we get P E = 650 kPa for the impact from the height of 50 cm. Assuming that this pressure is balanced by the capillary pressure (Equation (2)) a protrusion spacing of maximal d = 13-18 nm is allowed when θ adv = 77° is the advancing contact angle of n-hexadecane on a smooth fluorosilane coated silicon oxide. This is in the same order of magnitude that we got for the maximal spacing with water.
The mechanical stability of the superamphiphobic coating (Table S3, Supporting Information) was tested by letting 15 µL water drops impact on the surface from the release height of 200 cm at v = 5.4 m s −1 . The sample was tilted by 10° to ensure rapid removal of the impinging drops. The coating could withstand at least 20 000 drop impacts (90 impacts min −1 ) by completely allowing the impinging drops to bounce off the surface. After the experiment, roll-off angle of 10 µL water drops at the impacted area was 13°. The nanotexture was partially damaged and n-hexadecane drops started to pin to the impacted surface ( Figure S8, Supporting Information). With increasing silicon dioxide content the mechanical stability of the coating increased against impacting drops. After 20 000 drop impacts on both the Si 100 wt% and Si 99 wt% coating the roll-off angle of 10 µL water drops remained at 3°-5°. That is, the surfaces remained superhydrophobic after the exposure to the impacting drops.
To further test the robustness of the superamphiphobic coating (Si 1 wt%, coated 5 times), in particular the adhesion of the coating to the glass substrate, we exposed the surface to steam and continuous water flush. Therefore, water was heated in a beaker on a hot plate at 150 °C and the sample was placed face down 5 cm above the water surface for 1 h. In a second set of experiments, the surface was rinsed with Milli-Q water flow at v = 1.7 m s −1 for 1 h. After both treatments, the roll-off angle for 10 µL water drops remained unchanged, i.e., <1°, which indicates good adhesion of the coating. Additionally, the adhesion of the coating was tested by adhering and peeling off an adhesive tape (Scotch Magic), applied with the pressure =2.5 kN m −2 for 60 s. After the tape test, the coating maintained low roll-off angle for both water and n-hexadecane, <1° and 6° ± 2°, respectively.
Abrasion by impacting sand particles can locally damage the superamphiphobic coating (Si 1 wt%, coated 5 times). After impacting the surface with 100-200 µm diameter sand grains from the height of 2 cm for 10 s (5 g of sand; impact v = 0.63 m s −1 ; sample was adjusted at an angle of 45°), n-hexadecane drops pinned to the surface. However, water drops kept the high static contact angle = 155° and roll-off angle = 25° after the abrasion. Although the impact of the sand particles damaged the top part of the protrusions, the measurements indicate that the adhesion of the coating to the glass substrate was not compromised, i.e., it exceeded the cohesive strength of the coating. Si 100 wt% coating (coated 5 times) remained superhydrophobic after the sand abrasion test and maintained a low roll-off angle = 2° for 10 µL water drops.
Temperature-stability of the superamphiphobic coating (Si 1 wt%, coated 5 times) was investigated between −200 °C and +500 °C. Delamination of the coating was not observed even after freezing the sample in liquid nitrogen for ≈30 s ( Figure S9, Supporting Information). Roll-off angles for water and n-hexadecane remained <1° and 10° ± 5°, respectively. Heating the sample in an oven at 500 °C for 3 h degraded the  www.advmat.de www.advancedsciencenews.com fluorosilane layer. After reapplying the fluorosilane, the coating recovered its high static contact angle >160° and low roll-off angle <1° for both water and n-hexadecane.
Photosensitivity of the superamphiphobic coating (Si 1 wt%, coated 5 times) was investigated by illuminating the surface with UV-A light (2.3 ± 0.3 mW cm −2 ) up to 4 h. After 40 min of illumination, 10 µL drops of n-hexadecane pinned to the surface (Table S4, Supporting Information) because of photodegradation of the fluorosilane coating on top. To delay this, before the fluorination we encapsulated the coating with a ≈3 nm thick silicon dioxide shell by applying a gas-phase Stöber-like reaction for 4 h. Such a thin passivation layer did not reduce the superamphiphobic properties of the coating. Moreover, the coating remained superamphiphobic even after the UV-A illumination of 4 h: roll-off angles for 10 µL water, and n-hexadecane drops were <1° and 4° ± 2°, respectively.
Robustness of the air cushion on the superamphiphobic coating (Si 1 wt%, coated 5 times) under prolonged contact with n-hexadecane was investigated by letting 10 µL drops to rest on the surface for 30 min. The static contact angle remained unchanged within the experimental error ( Figure S10, Supporting Information). After the period of 30 min, roll-off angle for the n-hexadecane drops was 10° ± 2°, proving the stability of the air cushion.
In summary, we introduce an up-scalable method to fabricate optically transparent superamphiphobic surfaces with low drop adhesion and high impalement resistance against both high-and low-surface-tension liquids. With LFS and by mixing Si and Ti precursors, surfaces can be fabricated with high apparent contact angles and low roll-off angles below ≈1° even for n-hexadecane. To achieve an ultralow drop adhesion for nonpolar liquids, the superamphiphobic surface needs to fulfill the following criteria: (1) low-surface-energy chemistry, (2) nanoscale, overhang surface structures, (3) hierarchical roughness, and (4) sub-micrometer scale pore size to increase the critical impalement pressure. We show that increasing the Ti content or increasing the number of coating cycles increases porosity, thickness, and hierarchical structure of the coating. Both measures improve the superamphiphobic properties. On the other hand, optical transparency of the coating decreases and better mechanical stability is achieved with a high Si content. The Si/Ti ratio needs to be optimized depending on the specific requirements for the coating.

Experimental Section
Surface Modification: Prior to CVD with the fluorosilane the samples were activated by oxygen plasma (Femto low-pressure plasma system, Diener electronic, Germany) at 300 W for 10 min. The samples were placed in a desiccator together with 100 µL of the fluorosilane and the pressure was reduced to ≈200 mbar for 2 h. After the CVD, the samples were placed in a vacuum oven at 60 °C for 2 h to remove unreacted silane. To investigate the effect of the size of surface textures on the wetting properties of the coatings, a gas-phase Stöber-like reaction was applied for selected samples (Si 100 wt%, Si 99 wt%, Si 1 wt%, coated 1 time). The samples were placed in a desiccator together with ammonia (3 mL) and TEOS (3 mL) at atmospheric pressure and room temperature for 24 h. This resulted in growth of an additional 20 nm thick porous silicon dioxide shell around the particles. [4] After the silicon dioxide growth, the samples were sintered in an oven at 500 °C for