Vacuum infusion as a novel method to determine wood permeability

This study aims to propose a novel method, vacuum infusion process, to measure the longitudinal permeability of wood. The vacuum infusion method uses a vacuum bag sealed over the fibrous material, with a vacuum inlet and a vacuum outlet. It can be performed on top of any flat surface, and its process is relatively swift. Six different woods (Pinus elliottii, Araucaria angustifolia, Ochroma pyramidale, Cedrela fissilis, Tectona grandis, and Eucalyptus grandis) and three different fluids (water, soybean oil, and furfuryl alcohol) were selected for the study. After preliminary evaluations of morphology, chemical characteristics, density, porosity, contact angle and capillary pressure, three woods and two fluids were selected for the actual permeability measurements. The highest permeability was obtained for the Ochroma pyramidale wood, being 0.45–7.49 × 10–11 m2. This wood was 58–88% and 18–62% more permeable than the Pinus elliottii and Eucalyptus grandis woods, respectively. The fluid was found to have some influence on the experiment and therefore must be carefully selected. The difference in permeability of the woods was attributed to morphological characteristics, especially the presence of axial vessels, which are 60% larger for Ochroma pyramidale wood compared to Eucalyptus grandis wood, while Pinus elliottii has no vessels. The amount of voids in all woods, nevertheless, was similar, as well as the evaluated chemical characteristics and structural anatomical elements (tracheids and/or fibers). In all, the determination of apparent permeability using the vacuum infusion process is practical and with good accuracy, yielding results similar to those from other methods in the literature.


Introduction
Permeability is the measure of a fluid (gas or liquid) flow through a porous medium in response to a pressure gradient (Hansmann et al. 2002). The wood permeability plays an important role in impregnation rate and retention of the impregnated substances, as well as the consequent changes in other properties (Tarmian et al. 2020). Wood permeability depends on some of its physical and chemical characteristics and may vary according to both genus and species. From a morphological point of view, the vessels (in hardwoods) and tracheids (in softwoods) have a direct influence on wood permeability. Furthermore, regarding chemical features, the content and composition of extractive compounds may be considered as main factors, that affect wood permeability.
The wood vessels are the main liquid conducting tissues in hardwoods. These anatomical elements are responsible for 90% of the transported water and their content may vary from 10 to 60% in relation to the total area in the transverse plane (Bajpai 2018). Furthermore, diameter, clustering form, and perforation type of these vascular elements also affect wood permeability (Mahdian et al. 2020;Rezende et al. 1 3 2018) evaluated the water permeability of two eucalypt woods and observed that the wood with the smallest vessel diameter (Eucalyptus dunnii wood) showed a mean longitudinal permeability 10% smaller than the same property of the wood with the largest vessel diameter (Eucalyptus grandis wood). Rogério et al. (2010) compared the longitudinal water permeability of other two eucalypt woods (Eucalyptus grandis and Corymbia citriodora) and concluded that the permeability of the Eucalyptus grandis wood was about 20% higher due to the average diameter of its vessels, which was 50% larger. Emaminasab et al. (2015) affirmed that the effectiveness of the vessels in conducting liquids is due to their large diameter and the presence of perforations that axially connect them to each other.
Regarding softwoods, the tracheids are the main fluidconducting anatomical elements and account for about 90% of the overall morphological constitution in volume (Brändström 2001). Thus, tracheids differ from vascular and other axial anatomical elements due to their small diameters and absent perforations (i.e. they are not axially interconnected) (Emaminasab et al. 2015). Although they have no axial perforations, the tracheids are laterally interconnected by small openings called radial pits, which vary in diameter from 0.2 to 20 μm (Zabel and Morrell 2020). In this regard, Lehringer et al. (2009) stated that the presence of obstructed pits, usually caused by internal pressure differences during wood drying processes is a limiting factor for permeability in softwoods.
This obstruction process, known as tylosis, occurs in some ring and diffuse porous wood species when embolisms arise in vessels either during drought or in response to wounding (Barnett 2004). This is an irreversible process since it involves the formation of hydrogen bonds between the torus/margo and the pit opening. When the sapstream is broken, this balance is destroyed and there is a net osmotic flow of water into the parenchyma cells and the resulting pressure increase causes the pit membranes to swell into the vessel lumen (Barnett 2004). Thus, the formation of tylosis impairs the passage of fluids within the conductive elements of the wood. Furthermore, at a certain age, after the cells become inactive for conducting liquids, the sapwood from some trees progressively becomes heartwood, from inside to outside, and then oils, resins, gums and phenolic compounds are formed in this central region of the trunk. With the increase in the extractives content and the decrease in diameter of tracheids or fibres, the heartwood also becomes less permeable (Gartner and Meinzer 2005). Brito et al. (2019) evaluated the gas permeability of eucalypt wood and reported that the sapwood was two times more permeable than the heartwood due to the obstruction of intravascular pits from the heartwood by resinous substances. In another study, Rogério et al. (2010) evaluated longitudinal water permeability of Eucalyptus grandis and Eucalyptus citriodora woods and observed that their heartwoods were 10-15 times less permeable than their respective sapwoods.
Several methods have been developed for measuring wood permeability (Ahmed and Chun 2009;Bufalino et al. 2013;Taghiyari et al. 2014;Ai et al. 2017;Rezende et al. 2018). Most of these studies evaluated the longitudinal wood permeability instead of the other directions (ca. radial and tangential) since the permeability in this direction is 10-15 times higher than the others and, because of that, typical movements of fluids inside the wood are vertical, either bottom-to-top or top-to-bottom (Hansmann et al. 2002;Lehringer et al. 2009). Ai et al. (2017), for example, developed a complex and costly system that consisted of two identical rigid stainless-steel tanks connected in series to a sample holder, a vacuum valve, actuators and pressure gauges attached at each end. The pressure vessels were also connected to a vacuum pump at one end and to the fluid reservoir at the other end. The fluid passed through the wood, and the permeability was measured based on the pressure difference between inlet and outlet. Ahmed and Chun (2009) proposed a simple process for measuring wood permeability. The authors inserted a small prismatic Samanea saman wood sample into a Petri dish and poured a dye solution (10 g of safranin diluted in 500 mL of 50% ethyl alcohol solution). The infiltration speed of the safranin solution was measured using a microscope aided by a software for image processing. This method was found unreliable due to the difficulty in monitoring the flow front justified by the heterogeneous flow inside the wood.
More recently, Leggate et al. (2019) measured the longitudinal permeability of Pinus elliottii and Pinus caribaea woods. The samples were sealed on their lateral surfaces with epoxy resin and then impregnated with water at a constant pressure of 4.20 MPa. Wood permeability was measured using a Porolux 1000 equipment (commonly used industrially), which consists of a sample holder placed on an analytical scale, which is connected to a pressure vessel and a manometer monitored by a computer. This method, however, has high costs associated with the device, and the benchtop equipment could not be easily moved, restricting its use.
Nonetheless, the most efficient method published before was developed by Taghiyari et al. (2014). In this equipment, two milli-second precision electronic clocks allow a high accuracy fluid movement capture. This apparatus consists in a changeable glass-made cylinder column with a variable length (from 30 cm to 2 m) (Esmailpour et al. 2019). In all, the permeability tests already designed for analyzing wood parts are limited by the required sample shapes and the low measurement accuracy.
The vacuum infusion process, commonly used as a manufacturing method for polymeric composites, is proposed in this work as a novel way to measure wood permeability. In this process, the sample is placed on a smooth and rigid surface and then sealed with the aid of a polymeric film and a sealing tape (Alms et al. 2010). Channels for vacuum application and fluid injection are positioned and a vacuum pump is used to evacuate the air from the mold and sample. The vacuum drives the impregnation of the porous material, simultaneously flowing in the plane and through the thickness due to the pressure gradient applied between inlet and outlet ends (Summerscales and Searle 2005). A camera can be used to monitor the resin flow, allowing calculation of the permeability based on Darcy's law.
The infusion process has high versatility, the shape of the sample can vary, the material cost is low, the system is simple and easily transported, and requires an easy operation, being common in the composites industry. The vacuum infusion process is relatively recent (around 2 decades ago), but has been successfully used for measuring permeability of fibrous mats and fabrics, which are used as reinforcements in composites (Li et al. 2015;Aitomäki et al. 2016;Yun et al. 2017;da Silva et al. 2020). For instance, da Silva et al. 2020) determined the in-plane permeability of R-glass and aramid-based fabrics positioned in different stacking sequences by vacuum infusion. They adjusted a constant pressure of -100 kPa between inlet and outlet ends and reported values ranging from 2.95 × 10 -11 to 8.27 × 10 -11 m 2 . In this context, this work introduces a novel method, vacuum infusion, to measure the longitudinal wood permeability.

Raw material
1 mm thick sapwood veneers cut by rotary peeling from six wood species, namely Pinus elliottii, Araucaria angustifolia, Cedrela fissilis, Tectona grandis, Eucalyptus grandis and Ochroma pyramidale, were acquired from the following Brazilian companies: Ecofolhas (Pinheiros/SP), ArtBalsa (Florianopolis/SC) and Léo Madeiras (Curitiba/PR). This material was cut from juvenile (10-15 years old) forests located in South America. These species were chosen due to both the great differences in anatomical features and commercial availability between each other.
Three impregnating fluids were used in the experiments: (i) deionized water, a polar liquid commonly used for measuring permeability in woods, with a density of 1.005 g/cm³ and dynamic viscosity of 1.05 cP; (ii) soybean oil (purchased from Klemm, Santa Cruz do Sul/Brazil), an inexpensive, non-toxic, and non-polar liquid, with a density of 0.899 g/cm³ and dynamic viscosity of 74.92 cP; (iii) furfuryl alcohol (purchased from Sigma Aldrich, São Paulo/Brazil), a resin commonly used to impregnate solid woods, with a density of 1.136 g/cm³ and dynamic viscosity of 5.0 cP.
Specific gravity was considered as the simple ratio between the mass and volume for each fluid. The dynamic viscosity was determined in triplicate for each liquid using a Brookfield viscometer (#3 spindle) at room temperature (~ 25 ºC).

Preliminary characterization
Oven-dried (at 103 °C for 24 h) wood fragments cut from the veneers were prepared (Tappi T257 2012) and characterized via wet chemical analyses to obtain moisture (Tappi 2007), ashes (Tappi T 211 om-02 2002), ethanol-toluene extractives (Tappi T 204 cm-97 2012), acid-insoluble lignin (Tappi T222 Om-02 2011), and holocellulose (remaining mass up to 100%) contents. All these results were obtained for three samples per species. Anatomical properties were measured for three 13 μm-thick veneer samples per species, which were prepared on a microtome (MRP2015). They were firstly dyed with an alcoholic solution of malachite green and then morphologically analysed using an optical microscope adjusted to a magnification of 25 times and a free software called ImageJ. Apparent (green) density (ρ a ), basic (air-dry) density (ρ b ), moisture content (MC), and porosity (Ø) were evaluated for five samples per species. The latter is shown in Eq. 1 described by Siau (1984). A digital caliper (resolution of 0.01 mm) and an analytical scale (resolution of 0.001 g) were used to determine the dimensions and mass, respectively.
The chemical affinity of each fluid/wood combination was assessed by tangential surface wetting of one sample per species by measuring the apparent contact angle at six different times from 5 to 120 s after contact of a 50 µl droplet using a DSA2995 equipment (Kruss ® brand) according to the sessile drop method. Capillary pressure (P c ) was determined for two samples per species based on the methodology described in Amico and Lekakou (2002). For that, a wooden veneer measuring 250 × 50 × 1 mm 3 was fixed to a metal rod and immersed 10 mm deep in the liquid (Fig. 1).
The test was ended when the liquid flow had stabilized, in height and weight (around 72 h), which was determined using a ruler and an analytical scale (resolution of 0.01 g). After equilibrium, the capillary pressure (P c , in Pa) was calculated according to Eq. 2.

Apparent permeability evaluation
Three different species (Pinus elliottii, Ochroma pyramidale, and Eucalyptus grandis) were selected based on their dissimilarity in terms of the previous physical and chemical results. Moreover, the permeability tests for soybean oil and furfuryl alcohol were successfully carried out at a -1.0 MPa vacuum, which yielded a too fast liquid movement when the water was selected as infiltrating liquid. Two samples for each liquid/wood combination were studied. This impaired the permeability measurement in this case and can be attributed to the small viscosity and high polarity of the deionized water. Because of that, the permeability experiment was later performed at -0.5 MPa vacuum for the water.
The permeability experiment was performed in duplicate at room temperature (~ 25 °C), for each wood/liquid combination. For that, two 120 mm-long plastic spiroducts were connected to the inlet and outlet hoses, forming a rectangular injection area of 120 × 300 mm 2 on a smooth surface. At one end, the inlet hose was connected to a 2 L beaker, in which the fluid was previously poured. At the other end, the outlet hose was attached to a pressure pot, connected to a vacuum pump.
A rectangular wooden veneer with dimensions of 120 × 300 × 1 mm 3 was placed between the spiroducts and a vacuum bag was placed on top to seal the injection area with the aid of a tacky tape. After that, the vacuum pump was adjusted to a constant pressure of -92 kPa yielding a predominantly unidirectional flow front, which was monitored using a digital camera. Transverse marks, 30 mm apart, were made on the wood veneer to allow easy flow-front observation. Figure 2 illustrates the vacuum infusion system. Some of the consumables, i.e., hoses, spiroducts, connections and valves, were reused, minimizing disposal and the associated environmental impact.
The ImageJ software was used to obtain the flowfront position from the photographs taken (a minimum of 20 for each experiment). After collecting the data, a graph of squared flow front position × time was plotted and linear regression was applied, considering a minimum r 2 of 0.95. From the slope of this line, it was possible to calculate apparent in-plane permeability (κ, in m 2 ) based on Eq. 3, reported in Rudd et al. (1993). This equation for rectilinear flow under constant pressure is based on Darcy's law, which states that the flow velocity is proportional to the pressure gradient (Caglar et al. 2018).
where: ε is preform porosity, is fluid viscosity (cP), P o is injection pressure (92 kPa), X 2 ff is the flow front position (m), and t is the infiltration time (s).
All data, except the chemical analyses, capillary pressure and contact angle results, were subjected to ANOVA analysis of variance. Homogeneity of variances and data normality were verified using Shapiro-Wilk tests. Whenever the null hypothesis was rejected, General Linear Model (GLM) tests were used to compare the means. All statistical analyses were implemented at a significance level of 5%. Figure 3 shows the chemical characteristics of the studied woods. The highest holocellulose content was observed for Eucalyptus grandis wood (69.51%), followed by Cedrela fissilis wood (69.02%) and Ochroma pyramidale wood (69.02%). The highest lignin content was observed for Tectona grandis wood (38.11%), followed by Araucaria angustifolia wood (34.90%) and Pinus elliottiii wood (30.27%). High extractives contents were found for Ochroma pyramidale wood (6.35%), Tectona grandis wood (5.25%) and Cedrela fissilis wood (4.90%). These results can be attributed to the wood nature since the hardwoods presented higher holocellulose and extractives contents, while the softwoods presented higher lignin contents, except for the Tectona grandis wood. These results are similar to those reported in the literature, i.e., 60-70% range for holocellulose content, 20.5-38.5% range for lignin content, and 1-10% range for extractives content (Severiano et al. 2010;Esteves et al. 2013;Borrega et al. 2015;Lengowski et al. 2020;Acosta et al. 2021b). Figure 4 shows the cross-sectional images of the studied woods. The Pinus elliottii wood shows some resin canals, which are elongated, tube-shaped intercellular spaces surrounded by epithelial cells.

Results and discussion
The optical micrographs also confirmed the large differences in terms of diameter of the main liquid conducting anatomical elements from softwoods (ca. tracheids) and hardwoods (ca. vessels), i.e., the latter were 10-20 times wider. Table 1 shows anatomical characteristics acquired from the aforementioned optical micrographs. Among the studied softwoods, Araucaria angustifolia showed the largest tracheid diameter (Ø F ) and lumen diameter (Ø LF ), and the Cedrela fissilis wood was the hardwood with the widest vessels (Ø V ), in a decreasing order, followed by Ochroma pyramidale, Eucalyptus grandis and Tectona grandis. Regarding the radial ray width, the Ochroma pyramidale wood stood out compared to the other woods. The presence of vessels in the hardwoods shown here is an important anatomical difference between hardwoods and softwoods. Softwoods have a simpler morphological structure, which can be explained by the biological evolution of these plants, being botanically classified as primitive vegetables (Bajpai 2018). Table 2 shows the apparent density, basic density, moisture content and porosity results for the studied woods. The Tectona grandis wood showed the highest apparent and basic densities, justified by the small diameter of its anatomical elements, especially vessels and fiber lumens, and the consequent high lignin content since it is most likely located in the middle lamella between fibers. The Ochroma pyramidale wood presented the smallest density, which is attributed to its wide fibers and vessels, as well as its high porosity. The values found in this study are in agreement with those reported in the literature for all woods, 0.15-0.60 g/cm 3 for apparent density and 0.11-0.55 g/cm 3 for basic density (Valério et al. 2008;Trevisan et al. 2016;Mahdian et al. 2020;Acosta et al. 2021a).
Among the studied woods, Tectona grandis and Eucalyptus grandis woods presented the largest water contact angles (Fig. 5a). These two woods were followed by Cedrela fissilis wood, Araucaria angustifolia wood, Pinus elliottii wood, and Ochroma pyramidale wood in a decreasing order of water contact angle. On the other hand, all woods presented more similar contact angle results when soybean oil or furfuryl alcohol were used (Fig. 5b, c), with a slightly higher value again for Tectona grandis and Eucalyptus grandis with soybean oil. In all cases, the contact angles for all liquids decreased up to 10-15 s, mostly stabilizing after that. Previous studies reported significant correlations of contact angle values and chemical (Kishino and Nakano 2004;Rossi et al. 2012), anatomical (Piao et al. 2010;Oberhofnerová and Pánek 2016), and other physical properties (Amorim et al. 2013), which did not occur in the present study. Figure 6 shows the height of the flow front in the capillary experiments for the studied woods. The values progressively increased until about 72 h, with little change after that until 160 h. These same three woods stood out in terms of both the soybean oil and furfuryl alcohol capillary pressures. In Fig. 3 Chemical properties of the woods the case of the softwoods, these results can be ascribed to their high tracheid length and tracheid diameter. Ahmed and Chun (2011) stated that the capillary pressure of a softwood is inversely proportional to its tracheid lumen diameter, which corroborates the findings of the current work.
On the other hand, the Ochroma pyramidale wood presented high capillary pressure probably due to its high porosity, which indicates that this physical property is a better permeability indicator than vessel diameter or length. In hardwoods, the porosity is dependent on both anatomical properties and the presence of some anatomical elements, especially vessels, which explains the obtained result.
For water as the infiltrating fluid, the Pinus elliottii wood presented the highest value (Table 3), followed by Ochroma  pyramidale wood and Araucaria angustifolia wood. For soybean oil as the infiltrating fluid, the Ochroma pyramidale wood presented the highest value, followed by Pinus elliottii wood and Araucaria angustifolia wood. For furfuryl alcohol as the infiltrating fluid, the Pinus elliottii wood presented the highest value, followed by Ochroma pyramidale wood and Araucaria angustifolia wood. The capillary pressure results did not show a direct correlation with the previous contact angle results, which suggests that it is probably affected by bulk anatomical and chemical characteristics and not only surface ones.
Due to their differences (anatomical and chemical), only three wood species were selected for permeability studies. The Ochroma pyramidale wood was chosen due to its large vessel diameter and high porosity, and Eucalyptus grandis due to its large availability in Brazil. Among the softwoods, Pinus elliottii was selected since Araucaria    angustifolia is a native wood in Brazil and there are restrictive laws regarding its exploitation. Figure 7 shows the flow front position (X ff ) and Fig. 8 shows the squared flow front position squared (X ff 2 ) as a function of the injection time data for the selected woods. The Ochroma pyramidale wood showed the fastest flow front for both liquids, which can be partly explained by its significantly higher porosity compared to the other woods. Although the infiltration is affected by capillary pressure, this is a comparatively small driving force in the permeability experiment and indeed it did not show a direct correlation.
Compared to furfuryl alcohol, the soybean oil was approximately 5 times slower to permeate the wood, which is due to its higher viscosity (Zhang and Cai 2008). The lower chemical affinity with the wood, since these two compounds have a high degree of polarities, may also be partly responsible. Besides, the pattern of the flow front evolution for the soybean oil seems to be clearer than that of the furfuryl alcohol. In that context, Trindade et al. (2019) affirmed that, when the fluid flows more slowly, the determination of permeability is more accurate, depending on the balance between injection pressure and capillary pressure. Figure 9 shows the calculated mean of longitudinal permeability values. Compared to the literature, the used methodology showed a smaller variability (maximum CV of 22.25%), which suggests that this method is more precise than those found in the literature (Table 4). The results obtained for the water were similar to those related to the other fluids, although they are not comparable since they were measured at different pressure levels. This also indicates that other infiltrating fluids could be applied to measure longitudinal wood permeability by vacuum infusion.
The reported CV values ranged from 25 to 55%, depending on the selected wood species, infiltrating fluid and measurement apparatus. The high variation in longitudinal wood permeability has already been reported in the literature and could be attributed to the inhomogeneous wood morphology, including diameter, distribution and grouping of porous anatomical elements. Nevertheless, it may be seen that the values reported in the literature are in the same range as   Rogério et al. (2010) those found in the present study, considering the similarity between the viscosities of FA and water.

Conclusion
In this study, longitudinal wood permeability was determined by a novel methodology based on vacuum infusion, which has been used to measure longitudinal permeability of fibrous reinforcements used in polymer composites. Two fluids and three woods were selected for the longitudinal permeability study after preliminary chemical, physical and anatomical analyses. Contact angle and capillary pressure results did not directly correlate to each other, and neither with the longitudinal wood permeability. Two of the three highest capillary pressures were obtained for softwoods, which indicates a high influence of the tracheid diameter on this property.
Although the infiltration is affected by capillary pressure, this is a minor driving force in the longitudinal permeability experiment carried out and indeed it did not show a direct effect on longitudinal permeability. It was observed that the most porous woods were the most permeable ones, especially the Ochroma pyramidale wood.
In all, the vacuum infusion method proved to be a reliable process for measuring wood longitudinal permeability. The values found for permeability agreed with the literature, however, the method used in the present study presented comparatively lower coefficient of variation than those reported. Thus, this methodology proves to be an accurate and practical alternative, able to be easily reproduced.
Further studies may include radial and tangential wood permeability. Moreover, variable veneer thicknesses and refractory woods may be addressed in the coming studies of the research group.