The bleached softwood kraft pulps (BSKP) with a moisture content of 8% were provided by Guangzhou Paper Mill. 3-Glycidyloxypropyltrimethoxysilane (GPS), (3-Aminopropyl) tri-ethoxy-silane (AMS), Silicon dioxide (30 nm), N-(3-Dimethylaminopropyl)-N’-ethyl-carbodiimide hydrochloride (EDC), N- Hydroxy-succinimide (NHS) and 1H, 1H, 2H, 2H-perfluorooctyltriethoxysilane (PFOTES) were all purchased from Aladdin Chemistry Co., Ltd. (Guangzhou, China). Tetraethyl orthosilicate (TEOS), ammonium hydroxide (30% in water), and all other chemicals were purchased by Tianjin Chemical Co., Ltd. (Tianjin, China).

BSKP (10 g, dry weight) was dispersed into 1 L of deionized water with continuous mechanical stirring at 300 rpms, followed by addition of NaBr (0.1 g/g d.w.p.) and TEMPO (0.015 g/g d.w.p.) [25] . The reaction was started by addition of NaClO solution (12%, 6, 8 and 10 mmol/g d.w.p.) at 25˚C. pH of the reaction was maintained at 10 ± 0.2 by addition of 0.5 M NaOH until no further decrease. Finally, the pulp was washed with deionized water and filtered by a Büchner funnel with a filter cloth (400-mesh) to obtain TEMPO-oxidized cellulose (OC) suspension at a concentration of 0.5 wt%.

TEMPO-oxidized cellulose membranes (MNOCM) were prepared by vacuum filtering 50 ml of TEMP-oxidized cellulose suspension on a 0.45 μm filter membrane and dried at room temperature.

A variety of sizes of silica nanoparticles were obtained by polymerization of TEOS using the Stöber method [26] . Typically, a fixed volume of TEOS was added drop wise to a flask under magnetic stirring containing ammonia solution, water, and ethanol (see Table 1) followed by addition of 5 mL of APS and heating at 70˚C for 6 h. The resulting silica nanoparticles were separated by centrifugation and washed by ethanol in triplicate. Finally, the obtained particle was dried in vacuo at 50˚C overnight.

2 g of silica nanoparticles of 30 nm diameter were dispersed into 70 mL anhydrous toluene and ultrasonicated for 10 mins. Subsequently, 7 mL of GPS was

added drop wise with mechanical stirring, and was kept at 126˚C under N₂ atmosphere for 3 h. The resulting silica nanoparticles were separated by centrifugation and washed with toluene in triplicate. Finally, the epoxy-functionalized silica nanoparticles were dried in vacuo at 50˚C overnight.

Micro/nano surface architecture was obtained by inter-grafting the two sizes of nanoparticles as obtained from the LbL assembly process. The first step was grafting amino-SiO₂ NPs onto the MNOCM through carbodiimide-mediated coupling of the amino group from amino-SiO₂ NPs to the carboxyl group of the TEMPO-oxidized cellulose membranes. First, 0.02 g of AS was dispersed into 8 mL of ethanol and ultrasonicated for 10 mins. Subsequently, 0.1 g EDC, 0.1 g NHS, and MNOCM were added and kept at room temperature for 4 h. After reaction, the MNOCM@SiO₂-NH₂ was removed and washed with ethanol for three times.

The second step was reacting MNOCM@SiO₂-NH₂ and epoxy-SiO₂ NPs. 0.02 g of epoxy-SiO₂ NPs by dispersing into 8 mL of ethanol and ultrasonicating for 10 mins. Then the MNOCM@SiO₂-NH₂ was immersed in the solution at 40˚C for 12 h to afford surface 1 (denoted as 1 layer). Several cycles were repeated to afford 2 to 5 additional layers.

The as-prepared MNOCMs were treated with PFOTES by chemical vapor phase deposition [9] . The samples were placed in a 250 mL metal bottle containing two small open bottles holding 200 μL of PFOTES and 1 mL of deionized water. The metal bottle was then sealed and heated inside an 80˚C oven for 6 h. Subsequently, the surface-treated samples were taken out and placed in a vacuum oven for more than 1 h at 60˚C to remove unreacted silane.

The size distribution of SiO₂ NPs was obtained from TEM images. Samples were dispersed in ethanol (0.02% w/v) and deposited on a carbon-coated copper grid. The images were observed by JEM 2100 transmission electron microscope at an accelerating voltage of 200 kV.

The MNOCM were first sectioned into pieces. These pieces were fixed on sample holders and coated with Au for SEM (Merlin, Zeiss, Germany) at an operating voltage of 10 kV in secondary electron mode.

ATR-FT-IR spectra were recorded on a VERTEX 70 infrared spectrum instrument (Bruker Corporation, Germany). The scanning range was 500 - 4000 cm⁻¹.

Water contact angle (WCA) measurements were carried out on an OCA20 Micro (Data physics Instruments) to evaluate hydrophobicities by using a droplet (5 μL) of deionized water as an indicator at room temperature. The values reported were taken after the contact angle equilibrated.

In general, the pre-eminent physical factors that contribute to the phenomenon referred to as superhydrophobicity are (1) roughness or surface texture and (2) attenuated (low) surface energy [29] . The silica nanoparticles to induce the micro/nano texturized structure on the MNOCM surface were prepared by the Stöber procedure which is characterized by catalyzed hydrolysis and condensation of TEOS under alkaline conditions [12] [30] (Figure 1(a)). It is noteworthy that in the process of TEOS hydrolysis, the ammonia content, alcohol dosage, temperature and the time of the reaction influence the sizes of silica nanoparticles.

A variety of multi-sized silica nanoparticles were obtained by changing the reaction conditions (Table 1). The formation of a stable interaction between the MNOCM surface and silica nanoparticles was accomplished by grafting these silica nanoparticle via their active sites. In the present work, AMS-functionalized silica nanoparticles (amino-SiO₂ NPs) and GPS functionalized silica nanoparticles (epoxy-SiO₂ NPs) were prepared by hydrolytic condensation of AMS or GPS with SiO₂, respectively (Figure 1(b)). The AMS or GPS was hydrolyzed and grafted onto the silica nanoparticles surface to afford surface reactive amino groups or epoxy groups.

Morphologies and nanoparticle sizes of the as-prepared SiO₂ NPs are shown in Figure 2. The TEMs showed that the sizes of the obtained silica nanoparticles increased after sequential changes in reaction conditions. The SiO₂ nanoparticles presented spherical shapes with an average diameter of ~30 nm, ~70 nm, ~260 nm and ~500 nm. In Figure 3, all the silica nanoparticles exhibited strong bands at ~800 cm⁻¹ and ~1100 cm⁻¹ corresponding to the Si-O-Si asymmetric and symmetric stretching vibrations, respectively [31] . The peaks at ~3421 cm⁻¹ and

1630 cm⁻¹ belong to O-H stretching and distorting vibrations on silica particles, respectively [32] , whereas the peak at ~955 cm⁻¹ corresponded to the bending vibration of Si-OH [11] .

When the surface silica nanoparticles were grafted subsequently AMS or GPS, new peak appeared at ~2945 cm⁻¹, corresponds to the C-H stretching of -CH₂ groups [11] , which confirmed that AMS or GPS had been successfully grafted because both groups contain -CH₂-. Moreover, the peaks (blue line) at ~1538 cm⁻¹ (N-H bending) and ~1400 cm⁻¹ (C-H bending) further proved amino groups existed in the modified silica particles. The color of these modified particles changed blue-purple within a few min when added into 5% ninhydrin aqueous solution. This reaction is the typical method to identify the amino group existence.

In general, the rough micro/nano surface texture of the final MNOCM was created by covalent LbL assemblies of alternating large silica amino-NPs and smaller silica epoxy-functionalized-NPs. Figure 4 showed the scheme of reaction. TEMPO-mediated oxidation was carried out to produce varying levels of surface carboxyl on cellulose. The acid enriched sites were exploited to chemically anchor amino groups on the amino-SiO₂ NPs by activation with EDC and NHS (Figure 3). Additionally, free amino groups can react with the epoxy groups of epoxy-SiO₂ NPs by ring opening to afford layer 1 [2] . Subsequently, free epoxy groups can react with amino-SiO₂ NPs again. These reactions can be repeated as many times as desired to form n layers. Finally, samples were treated with PFOTES through vapor phase deposition to provide a surface capping layer as stated previously (vide supra).

ATR-FT-IR (Figure 5) showed the ATR-FT-IR spectra of MNOCM and MNOCM@SiO₂-NH₂. The two samples exhibited strong absorbing peaks at ~3335 cm⁻¹ due to O-H stretching of cellulose molecules. The peak at ~2920 cm⁻¹ was attributed to C-H stretching in -CH₂ or -CH₃ groups because the silica particles grafted -CH₂- groups. Meanwhile, the peaks at ~1736 cm⁻¹ and 1600 cm⁻¹ could be attributed to the -C=O and sodium carboxyl groups (-COONa), respectively [33] . Compared with MNOCM, MNOCM@SiO₂-NH₂ had new peaks at 1698 cm⁻¹, 1544 cm⁻¹ and 1279 cm⁻¹, corresponding to the C=O stretching vibration of -NHCO- (amide I), N-H bending (amide) and amide III, respectively [34] [35] [36] [37] . Moreover, two broad and enhanced peaks at 1050 cm⁻¹ and 1029 cm⁻¹ were characteristic of the overlapping absorption peaks between Si-O-Si and C-O-C [38] . The information confirmed that the amino-SiO₂ NPs were successfully grafted onto the MNOCM surface through amino-carboxyl amidation [39] .

Furthermore, the carboxyl content may influence grafting of amino-SiO₂ NPs. Therefore, cellulose with different carboxyl contents (0.95, 1.24, and 1.28 mmol/g o.d.p) were prepared. It was found that the carboxyl content affected on the grafting of amino-SiO₂ NPs showed in Figure 6. For the MNOCM with a carboxyl content 0.95 mmol/g o.d.p (Figure 6(a)), we can clearly observe the surface was a little corrugated and smooth without any silica. Similar results can be observed in MNOCM with a carboxyl content 1.24 mmol/g o.d.p and 1.28 mmol/g o.d.p. For MNOCM@ SiO₂-NH₂ with a carboxyl content 0.95 mmol/g o.d.p (Figure 6(b)), the particles do not homogeneously over the surface and some places were exposed. The surface of MNOCM@ SiO₂-NH₂ were more covered by particles when the carboxyl content of MNOCM was 1.24 mmol/g o.d.p (Figure 6(c)), however, some areas were still exposed without particles. When

the MNOCM with a higher carboxyl content (1.28 mmol/g o.d.p, Figure 6(d)), more nanoparticles adhered to the surface randomly, almost completely covered. It revealed that the amount of silica on cellulose membranes surface increased with increasing carboxyl content.

To provide roughness to the MNOCM surface, 260 nm amino-SiO₂ NPs were grafted followed by covalent LbL assembly of 30 nm SiO₂ NPs. This was repeated several times to form n layers of particles (260 nm particles + 30 nm particles represented one layer). The SEM images in Figures 7(a)-7(e) showed the surface morphology of the MNOCM having a varying number of layers. The particles were randomly assembled indicating different degrees of surface roughness. For 1 layer (Figure 7(a)), the MNOCM surface was mainly covered with 260 nm particles and only a small amount of 30 nm particles, the average roughness is around 260 nm. The roughness increased visually after 2 layers (Figure 7(b)). Figure 7(c) and Figure 7(d) showed the particle distribution in

the 3 and 4 layers to form a rougher surface. Not surprisingly, after final assembly, a heavily rough surface was formed. The degree of roughness changed with each succeeding layer allowing for an elegant means of control of surface roughness through succeeding nanoparticles coverage [12] [40] . Interestingly, assembly of 30 nm SiO₂ NPs on the surface of 260 nm SiO₂ NPs allowed for the development of a raspberry-like structure (Figure 7(e), Figure 7(f)) very evident in the 5 layers. The roughness of the surface after assembling from 1 layer to 5 layers doesn’t increase linearly with the growth of layers because the small nanoparticles may plug the valley between particles. The suitable layers may provide best roughness structure, but not the biggest roughness.

The samples were subsequently capped with PFOTES by CVD and tested for wettability (Figure 7 insets). Not surprisingly, owing to its polyhydroxylic nature, pure OCM can be completely wetted by water. However, even after the deposition of one layer of NPs, the contact angle was 133˚ (Figure 7(a)-inset). The pure OCM without particle covering is of course not superhydrophobic even capping with PFOTES. After deposition of 2 - 5 layers of particles (Figures 7(b)-7(d) insets), the water contact angles increased to 142˚, 143˚, 148˚, and ultimately 151˚, respectively, with a slide angle of 4˚ (Figure 8), demonstrating the attainment of a superhydrophobic surface. Such a change in contact angles indicated that the hydrophobicity of MNOCM surface displayed a hysteresis associated with LbL. Modifying the layers of nanoparticles can therefore effectively hydrophobize MNOCM.

The EDX of MNOCM after treatment with PFOTES is presented in Figure 9. Only carbon, oxygen, fluorine, silicon and sodium elements were detected. The carbon and oxygen were of course from MNOCM, silica, and PFOTES. Sodium was the counterion from the carboxyl groups to demonstrate that a fraction of the surface neat carboxylic groups remained carboxylates, silicon originated from silica and PFOTES, whereas fluorine was exclusively from PFOTES. These results provided further confirmation that silica NPs and PFOTES were on the MNOCM surface.

The sizes of amino-SiO₂ NPs were varied for the covalent layer-by-layer assembly on the MNOCM surface to interrogate how particles sizes affect hydrophobicity (Table 2). Although, the 70 nm and 260 nm amino-SiO₂ NPs showed superhydrophobicity, applying 500 nm (~twice the second sample size) amino-SiO₂, the water contact angle was attenuated to 140˚ ± 2˚. Therefore, the hydrophobicity of the raspberry-like structure for the silica NP coated surfaces is not only a function of the numbers of layers, but also of the size of the silica nanoparticles [1] . Surface roughness is likely more critical than surface chemical composition in determining hydrophobicity of such surfaces [1] [41] .

Self-cleaning is a property that is typically associated with superhydrophobicity as evidenced in the self-cleaning of mud from lotus leaves. Therefore, the self-cleaning of superhydrophobic MNOCM was evaluated according to a simple construct shown in Figure 10. The superhydrophobic-MNOCM films were attached to a glass slide onto which a dry slurry of activated carbon was placed to simulate contaminants (Figure 10(a)). Upon the disposition of water to the surface, the intrinsic surface tension of water on such an ultralow energy surface caused it to roll along down the incline due to gravity. During its roll, the contaminants on the surface were adsorbed (Figures 10(b)-10(e)), leaving a clean surface without any ostensible trace of contaminants (Figure 10(f)).