Hydrophobisation of wood surfaces by combining liquid flame spray (LFS) and plasma treatment: dynamic wetting properties

Introduction

Wood has a hydrophilic nature, and the resulting moisture sorption in outdoor applications leads to biological degradation (Rowell 2012). A hydrophobic protective coating on the surface is one of the approaches to reduce water up-take (Zanini et al. 2008), which entails an improved decay resistance and better dimensional stability (Podgorski et al. 2000, 2002; Mahltig et al. 2008; Manolache et al. 2008; Wang et al. 2011a; Poaty et al. 2013). Both impregnation of wood (Rowell 1982) and coating with varnish and paint (Probst et al. 1997) are the most common wood-protecting techniques. In both cases, the natural colour of the wood is changing as a side effect.

Superhydrophobic surfaces not only show water repellency but also antifouling and self-cleaning effects (Callies and Quere 2005; Ma and Hill 2006; Hsieh et al. 2008; Cao et al. 2009; Banerjee et al. 2011). The superhydrophobicity can be achieved by lowering the surface energy and modifying the surface roughness. The fabrication of multi-scale hierarchical surface structures in the micro- and nano-scale also can enhance the hydrophobicity (Wang et al. 2011a). Topographies in combination with sub-micron and nano-scale roughness are more water repellent than a micro-scale structure alone (Teisala et al. 2012).

Hydrophobic wood surfaces can be created by different techniques such as the sol-gel method (Wang et al. 2011a,b; Gholamiyan et al. 2016), plasma etching and polymerisation (Denes et al. 1999; Denes and Young 1999; Podgorski et al. 2002; Bente et al. 2004; Avramidis et al. 2009; Levasseur et al. 2012; Poaty et al. 2013), thermal and hydrothermal treatment (Fu et al. 2012; Wang et al. 2012, 2015), covalent grafting by more step procedures (Sèbe and Brook 2001; Mamiński et al. 2009; Liu et al. 2015), deposition of nanoparticles (Wang et al. 2011c, 2013; Shupe et al. 2012; Lu et al. 2014; Gao et al. 2016), and various layer-by-layer assembly methods (Rao et al. 2016).

Cold plasma treatment in the dry state is a simple and low energy consumption approach in this context. It allows the deposition and polymerisation of a wide range of monomers at temperatures which are slightly above room temperature, while the final characteristics of the layer can conveniently be controlled by the plasma polymerisation parameters. This technique is versatile (Denes and Young 1999), for instance it is suitable to improve hydrophilicity and wettability (Podgorski et al. 2000; Rehn and Viöl 2003; Busnel et al. 2010), to increase coating adhesion or hydrophobicity (Podgorski et al. 2002; Zanini et al. 2008; Poaty et al. 2013). The plasma polymerisation affects only the outer layer of the wood surface (Mai and Militz 2004). Fluorine-containing (e.g. CF₄ and C₆F₁₄) and silicone-containing compounds (e.g. HMDSO) are often plasma polymerised in order to create a thin, hard and transparent hydrophobic layer containing highly-cross-linked and branched macromolecules (Denes et al. 1999). Podgorski et al. 2002 studied plasma modified polymer layers of fluorine-containing monomers, which were found to be more hydrophobic than layers based on HMDSO. Zanini et al. (2008) found that treating the wood surface consecutively first with HMDSO plasma and then SF₆ plasma enhanced the hydrophobic character more compared to HMDSO plasma treatment alone. Usually, the static contact angle (CA_stat) is around 130°.

The present work deals with the creation of submicron- and nano-scale roughness of wood surfaces brought about with a thermal aerosol based technique, called liquid flame spray (LFS) (Teisala et al. 2010; Mäkelä et al. 2011). LFS followed by plasma polymerisation leads to surfaces with CAs >130°. In the course of LFS, nanosized metal oxides particles are generated and deposited on the surface. The literature describes the fabrication of superhydrophobic and transparent layers on paperboards by LFS (Teisala et al. 2010, 2012; Tuominen et al. 2012). This approach is also attracktive because the original colour aesthetic appearance of the materials is not changed. In our previous studies (Sedighi Moghaddam et al. 2013, 2016), a multicycle Wilhelmy plate (mWP) method has been developed, which is based on a wood veneer is immersed and withdrawn from a liquid several times, while the force acting on the plate is continuously measured by a microbalance. It was demonstrated that this approach can be utilised to study the dynamics of sorption, swelling and dimensional stability of the porous materials wood.

In focus of the present study is the investigation of the effects LFS/plasma treatment by means of the mWP method. The expectation is that the dynamic contact angles (CA_dyn) obtained will characterise the created surfaces better than the hitherto applied CA_stat obtained by the sessile drop technique. Hydrophobic and superhydrophobic surfaces can be certainly better described by CA_dyn and CA hysteresis (multi-cycle measurements). The focus of the scientific goals is a better understanding of the multi-scale roughness on the robustness of the wetting behavior of surfaces obtained by LFS/plasma treatment.

Materials and methods

Scots pine (Pinus sylvestris L.) sapwood blocks (120×30×7 mm³ in the T, L, and R directions, respectively) were cut from larger kiln dried boards with a band saw and hand veneer saw. To prevent end-grain sorption, one cross-section of the block was sealed by polyurethane lacquer. Veneers with dimensions of approx. 30×7×1 mm³ (L×R×T, respectively) were then prepared by splitting the blocks along the fibre direction and from the same series of annual rings with a wood chisel. The wood veneers were wrapped in aluminium foil prior to further treatment. Double sided polished silicon wafers (Siegert Wafer GmbH, Aachen, Germany) with a thin silica surface layer served as reference substrates. These were cut to dimensions of 30×10×0.52 mm³ from silicon wafer discs, and cleaned by rinsing with water and analytical grade ethanol several times followed by drying under nitrogen stream. The wafers were finally cleaned in an 18 W air plasma cleaner (PDC-3XG, Harrick, Ithaca, USA) for 5 min.

Nano-sized metal oxide particles (TiO₂) were deposited on the wood veneers (with a MC of 6–9%) by the liquid flame spray (LFS) method. To this purpose, a single nozzle type burner processing device developed at the Tampere University of Technology (Tampere, Finland) was applied. A detailed description of the LFS process is provided by Teisala et al. (2010), Mäkelä et al. (2011) and Stepien et al. (2011). The liquid precursor titanium tetraisopropoxide (97%, Aldrich, Sant Louis, USA) was diluted in isopropanol. The concentration of the precursor solution was 50 mg of atomic metal per ml. The precursor solutions together with the combustion gases (hydrogen and oxygen) were then provided to a specially designed burner for creating nanoparticles to be deposited onto the wood veneers. The flow rates of H₂ and O₂ were 50 and 15 l min^-1, respectively. The feed rate was 6 ml min^-1 and the deposition time was 10 s. LFS-generated nanoparticles were introduced indirectly onto the substrate by a tailor-made flow tube to avoid undesired effects of heat.

Plasma polymer deposition was carried out in an in-house constructed plasma reactor at SP, Technical Research Institute of Sweden, Stockholm (Cho et al. 1990; Pykönen 2010). The plasma reactor contains a glass vessel connected to a double-stage rotary vacuum pump (Leybold-Heraeus D 65 B, Colonge, Germany). Two externally wrapped, capacitively coupled, copper electrode bands were powered by a 13.56 MHz radio-frequency generator (ENI, Model ACG-3) connected to an automatic matching network (ENI, Model MW-5D). The nanoparticle-treated veneers were placed in the reactor followed by evacuating the chamber to a pressure below 1.3 Pa before introducing the monomer from the top of the reactor [hexamethyldisiloxane (HMDSO, ≥99.5%, Fluka, Steinheim, Germany) and perfluorohexane (PFH, ≥98%, Apollo Scientific, Manchester, UK)]. The plasma parameters were selected to obtain the same polymer thickness in both cases. The discharge power and pressure during the treatment were 150 W and 3.3 Pa for HMDSO, and 40 W and 18 Pa for PFH. The inlet flows were 5 and 40 ml min^-1, respectively for HMDSO and PFH. The treatment time was 5 min in both cases. A schematic diagram of the combined LFS and plasma polymerisation on the substrate is presented in Figure 1.

Figure 1:

Schematic illustration of the wood surface before and after deposition of nanoparticles, and after deposition of the top plasma polymer layer. The size of the particles and the plasma polymer layer are not drawn to scale.

XPS analysis was performed in an AXIS Ultra^DLD X-ray photoelectron spectrometer (Kratos Analytical, Manchester, UK) equipped with a monochromatic Al K_α X-ray source operated at 150 W. For each sample type, two plates with approximate dimensions of 7×7×1 mm³ were prepared and the XPS analysis was carried out on one position (analysis area: 700×300 μm²) per plate. Survey spectra (lowly resolved at pass energy 160 eV) and detailed spectra (medium resolved at pass energy 80 eV) were recorded. The chemical shifts of the carbon signal, corresponding to different functional groups, were quantified by curve-fitting high-resolved carbon 1s spectra (highly resolved at pass energy 20 eV). XPS peaks were quantified with a linear background based on a combined Gaussian and Lorentzian type peak shape (70% and 30%). The relative sensitivity factors provided by the manufacturer were employed (C 1s=0.278, O 1s=0.780, Si 2p=0.328, and F 1s=1.000).

An atomic force microscope (AFM) Dimension Icon (Bruker, USA) was employed. Images were recorded in tapping mode in air with antimony (n) doped silicon rectangular cantilevers with a nominal tip radius of 8 nm (MPP-13120, Bruker). The NanoScope Analysis software (Bruker) was utilised. The thickness of the plasma polymer layers were measured based on an area of the silica substrates masked with a 10% (wt) solution of poly(d,l-lactide-co-glycolide) (Resomer^® RG 755 S, Evonik, Darmstadt, Germany) in acetone, and then the same plasma polymer layer was applied on the masked substrate. The mask was removed by a scalpel from the surface and the substrate was then analysed again by recording 10×10 μm² AFM tapping mode images with the same cantilever type. Finally, the “bearing analysis” feature in the NanoScope Analysis software was applied for measurement of thickness. The bearing analysis shows how much of a surface lies above or below a given height. The height histogram extracted for this feature gives two characteristic peaks corresponding to the substrate (silica) and the surface modified layer (plasma polymer) heights. By subtracting the lower value from the upper, the plasma polymer thickness was obtained.

A scanning electron microscope (SEM), FEI Quanta FEG 250 (Hillsboro, USA) was utilised. To avoid mechanically induced microcracks, sample preparation was performed by UV laser ablation (Wålinder et al. 2009). From the surface modified veneers, sections with the size of approx. 7×7×1 mm³ (L×R×T, respectively) were cut with a 248 nm laser. The instrument works in a low vacuum mode (LV-SEM), thus sputtering of the specimen surfaces is not needed. For the mWP method a sample in the form of a plate was immersed into and withdrawn from a probe liquid for several cycles, and the force acting on the plate was measured simultaneously (Sedighi Moghaddam et al. 2013). For porous and hygroscopic samples such as wood, the measured force F can be written as:

where P is the wetted perimeter of the plate, γ the surface tension of the probe liquid, θ the liquid-solid-air contact angle (CA), ρ the probe liquid density, A the cross-sectional area of the plate, h the immersion depth, g the gravitational constant and F_w(t) the force due to sorption of the liquid at time t. The method provides the advancing and receding contact angle (θ_A and θ_R), sorption and swelling, dimensional stability and answers the question, how extractives dissolution affects wettability properties. The unmodified wood surface becomes completely wetted by water after the first immersion, and thus only the advancing CA during the first immersion has a non-zero value. By modifying the wood surface, non-zero values of the advancing and receding CA could be obtained for the other cycles as well. For swellable materials as unmodified veneers, the sample perimeter is not constant during the measurement. In addition, the surface tension of the liquid (γ) may change with time/cycle as a result of extractives migration and dissolution in the liquid (Sedighi Moghaddam et al. 2013). However, for the hydrophobised samples in focus these quantities are close to a constant as a result of a very low level of water up-take, and no change in water surface tension was observed after the measurements. θ_A and θ_R can be obtained by linear regression of the corresponding curves extrapolated to zero depth (h=0) in the mWP plot (Figure 5), assuming that P and γ are constant, and Eq. 1 can be simplified to:

where F_f,n-1 and F_f,n are the final forces after cycle n-1 and n, respectively, due to the sorption, which is the same as the term F_w(t) in Eq. 1.

The two CAs at cycle n,θ_A,n and θ_R,n, can be calculated as:

Sorption is determined by linear regression of the final force, F_f, to zero depth for each cycle. The percentage of water mass up-take, relative the weight of the dry veneer, after cycle n, l_n, is calculated as:

where m_i is the initial mass of the of oven dried plate (veneer).

A Sigma 70 tensiometer from KSV Instruments (Helsinki, Finland) was utilised for wettability measurements. A 20-cycle WP measurement was performed with the veneers and ultrapure water (resistivity >18 MΩ cm), with an immersion and withdrawal velocity of 12 mm min^-1. The veneer was immersed to a depth of 10 mm and then withdrawn to 5 mm above the liquid surface. The veneers were after completion of the 20^th cycle conditioned to ca. the same weight as before the immersion and then immersed in n-octane (≥99% from Alfa Aesar, Karlsruhe, Germany) to obtain precisely their perimeters (Sedighi Moghaddam et al. 2013). The surface tension of water and octane were 72.0±0.2 mN m^-1 and 21.4±0.1 mN m^-1, respectively, based on ten measurements. Measurements were performed at 22–23°C and 35±5% RH.

Results and discussion

Surface characterisation by AFM

The thickness of the HMDSO and PFH plasma polymer layers on silica substrates was measured to be 29±6 nm and 31±2 nm, respectively. The root-mean-square (RMS) surface roughness and mean tapping phase angle are provided in Table 3 for the bare silica surface, and the Si-HMDSO- and Si-PFH-treated silica surfaces. Clearly, both the bare silica and the plasma polymer treated silica surfaces are smooth and thus are suitable references to the rough modified wood surfaces. The large change in phase angles (Table 3) is due to the softness of the plasma polymer layer.

Table 3

Surface roughness and phase angle of reference samples (silicon wafers) obtained from AFM height and phase images, respectively.

Surface	RMS	Mean
	Roughness (nm)	Phase angle (°)
Bare silica	0.2	94.4
Si-HMDSO	0.7	18.5
Si-PFH	0.3	17.6

AFM height and phase images of bare and modified wood surfaces are illustrated in Figure 2. Obviously, when the plasma polymer is directly applied on the wood surface (Figure 2b and c some of the smaller morphological features of the native wood are no longer detectable (Figure 2a). Moreover, for TiO₂ treated samples, the PFH plasma polymer layer follows the surface features more truly than in case of HMDSO plasma coating (compare Figure 2b with c and d with e). Thus, the HMDSO plasma polymer fills the space between small surface features, proving less fissured profiles than a PFH plasma polymer layer. This is most clearly observed for LFS treated surfaces (compare Figure 2d with e). One should bear in mind that mainly the nano-scale structure of the coatings is different but their micro-scale morphology (see SEM section) originating from the high roughness of the wood surface itself is more similar.

Figure 2:

AFM height (left) and phase (right) images of (a) bare wood (b) wood-HMDSO, (c) wood-PFH, (d) wood-TiO₂-HMDSO and (e) wood-TiO₂-PFH.

Surface characterisation by SEM

SEM images of TiO₂-PFH-treated wood and bare wood samples are presented in Figure 3. The image in Figure 3a illustrates the high roughness of the wood surface, which is in order of 100 μm. Comparing the images of surface modified samples (Figure 3b) with the image of the bare wood (Figure 3c) shows the deposition of nanoparticles and plasma polymer on the top surfaces.

Figure 3:

SEM images of a TiO₂-PFH-treated wood sample; (a) radial-tangential plane and (b) longitudinal-radial plane and (c) SEM image of bare wood from longitudinal-radial plane.

Multicycle Wilhelmy plate (mWP) wetting

The data collected during 20-cycle WP measurements for different selected samples are presented in Figure 4. As explained in the experimental part, the θ_A and θ_R of each cycle are evaluated and the veneers are fully withdrawn from the water for 50 s between two wetting cycles. Untreated silicon with a silica top layer has a very low CA (Figure 4a) which remains unchanged throughout the 20 cycles. PFH plasma treated surfaces are significantly more hydrophobic with relatively low hysteresis (Figure 4b) while for the silica-PFH system, θ_A is stable over the time and the θ_R decreases slightly.

Figure 4:

Force per perimeter length as a function of sample position for 20-cycle Wilhelmy plate measurements for different sample surfaces. The advancing and receding contact angle data for the first cycle (θ_A,1 and θ_R,1) are presented for all samples.

As seen in Figure 4c, the untreated wood surface has initially a weakly hydrophilic surface (θ_A,1=61°). The θ_R and all following θ_A values are zero due to the complete surface wetting. The force detected after removal of the sample from the liquid increases by each cycle due to water sorption as discussed by Sedighi Moghaddam et al. (2013). The wood-TiO₂ surface (Figure 4d) is initially more hydrophobic (θ_A,1=110°) but for all cycles θ_R and for all subsequent cycles θ_A is zero. Also in this case, significant water sorption occurs because of the porous nature of the nanoparticle layer. High θ_A and θ_R can be achieved by treating the surface with a plasma polymer (Figure 4e), but a clear hysteresis is observable. When the plasma polymer is added on the top of the LFS layer resulting in higher hydrophobicity (Figure 4f), a relatively small hysteresis is. The advancing and receding curves measured during the different cycles are similar indicating an extended wetting stability. As seen in Figure 4e and f, the advancing curves are even in contrast to the receding curves. This was noticed for all wood-plasma and wood-nanoparticle-plasma samples and is most probably related to pinning of receding water (n.b. the jagged receding curves) due to combined hydrophobicity and surface roughness. The presence of the plasma polymer significantly reduces water sorption, as seen by the small change in final force observed in Figure 4e and f compared to the significantly larger changes observed in Figure 4c and d.

The θ_A, θ_R and hysteresis determined during the first and the last cycle (20^th cycle) for the different surfaces are summarised in Table 4. The CA stability can be evaluated by the CA changes after 20 cycles. On the one hand, samples with only plasma polymer layer (wood-PFH and wood-HMDSO) show very high hysteresis compared to the surfaces which have both LFS and a plasma polymer layers. Thus, modification of the highly rough surface with hydrophobic polymers leads to a surface with a high θ_A and comparatively low θ_R related to the CA results on the smooth silica-HMDSO surface. In contrast, adding the multi-scale roughness to the surfaces with nanoparticles provided by the LFS method, results in surfaces with higher θ_A and θ_R and lower hysteresis.

Table 4

Advancing and receding CAs and hysteresis results for the first cycle and cycle 20 for different samples (W is for wood).

Sample	Initial CA (°) (1st cycle)			CA (°) at cycle 20			γ (cosθ20-cosθ1) (mnm-1)		Water uptake, l20(%)a
	θA	Surface	Hyster.	θA	θR	Hyster.	Advan.	Reced.
W	65±7	≈0	65±7	≈0	≈0				48.5±6.8
W-TiO2	107±3	≈0	107±3	≈0	≈0				31.8±6.0
W-PFH	144±10	80±8	64±13	136±5	72±12	65±9	6.5±3.5	9.7±4.3	7.8±1.5
W-HMDSO	145±5	88±5	57±4	141±4	75±9	68±8	3.0±1.6	16.1±5.4	7.6±1.7
W-TiO2-PFH	159±9	134±10	26±4	159±10	127±8	29±6	≈0	6.7±3.4	3.9±0.7
W-TiO2-HMDSO	160±6	128±18	31±11	161±7	127±19	35±11	1.4±0.7	1.0±1.4	4.3±1.2
Si-PFH	112±1	99±1	13±1	111±1	99±1	13±1	0.8±0.7	1.2±1.2	≈0
Si-HMDSO	106±1	90±0	16±1	104±1	88±1	15±1	-3.5±1.1	-3.3±0.7	≈0
Silica	13±1	11±2	2±2	12±1	11±1	1±1	0.3±0.3	0.2±0.2	≈0

^aFrom Eq. 6.

The PFH treated surfaces show slightly higher θ_R (lower hysteresis) compared to the HMDSO treated ones which are consistent with the fact that fluorocarbon is more hydrophobic than siloxanes. In summary, wood-TiO₂-PFH surfaces revealed the highest θ_A and θ_R and the lowest hysteresis. This is explained by the topography, showing roughness features on multiple length scales, and the highly non-polar character of fluorocarbon. Hydrophobisation of wood by plasma polymerisation alone leads in most cases to a CA_stat ≤130° (Denes et al. 1999; Podgorski et al. 2002; Zanini et al. 2008; Poaty et al. 2013). However, Bente et al. (2004) reported CAs around 145° on layers obtained by plasma polymerisation of silanes. Higher CA with a low hysteresis was attainable by combining a plasma polymerisation with an LFS deposition that enhances the roughness over smaller length scales than that provided by the natural wood surface.

Table 4 also reveals that multiscale rough surfaces (e.g. wood-TiO₂-PFH) have more stable θ_A and θ_R over the time than a microscale rough surface (e.g. wood-PFH). In the former case, θ_A is not changed and θ_R lowered only in the range of 1–8° after the 20^th cycle. In contrast, for the micro-scale rough surfaces, both θ_A and θ_R diminished as a function of the cycle numbers (i.e. of the time). It means that adding a nano-scale roughness to a micro-scale hydrophobic wood surface contributes to the stability of wetting properties. Accordingly, the mWP method is well suited for determination of CA_dyn data including the hysteresis and the stability of the obtained effects.

The evaluation of the data by Eq. 6 is presented in Figure 5. It is clear that both PFH and HMDSO type plasma polymerisation reduce the water sorption. The benefit of the LFS layer is small, which can be attributed to its porous character.

Figure 5:

Water up-take as a function of cycle number for unmodified and modified samples, (a) PFH treated veneers; (b) HMDSO treated veneers.

Conclusions

A thin, transparent and superhydrophobic layer on wood surface was achieved by combining plasma polymerisation and LFS. Modified wood surfaces with two different scales of roughness were prepared, i.e. (a) in micro-scale by means of plasma only and (b) multi-scale by combining LFS and plasma. The effect of detailed surface chemistry on the wetting properties were investigated based on two types of plasma polymer layers (PFH and HMDSO). The changes of advancing and receding contact angles (CA) were measured by the mWP method and in this way the dynamic wettability was studied over repeated cycles on surface modified samples. The LFS method in combination with plasma polymerisation has a high capability for maximising water repellence. The modified surfaces show CAs >150° with relative low hysteresis, while the surfaces treated only with plasma revealed a CA around 130° with high hysteresis. Thus, the multiscale roughness increased the hydrophobicity as well as the forced wetting durability compared to microscale roughness alone. According to AFM images, the HMDSO plasma polymer provides larger morphological features than the PFH plasma do. Shortly, CA hysteresis is higher for surfaces treated with HMDSO plasma than that treated with PFH plasma.

Supplementary material

XPS high resolution C1s spectra.