Water vapour sorption behaviour of thermally modified wood

Introduction

Although the properties of thermally modified wood have been extensively studied, relatively little has been reported on the water vapour sorption isotherms until very recently (Hill et al. 2012b) and the effect of multiple sorption cycles upon the sorption isotherm had not been studied at all. It is known that densified wood undergoes a relaxation process under humid or wet conditions, caused by the release of internal stresses and one purpose of this investigation was to examine whether there would be any change in the sorption isotherms or sorption kinetics between the first and subsequent sorption cycles. The study was also designed to determine whether this relaxation behaviour characteristic of densified wood could be reduced by thermal modification.

It has recently been shown that the water sorption kinetics behaviour of small samples of wood, modified wood, natural fibres and cellulose is non-Fickian and actually obeys a parallel exponential kinetics (PEK) sorption model (Hill et al. 2010a, b, c, 2011, 2012a, b; Jalaludin et al. 2010a, b, c; Xie et al. 2010a, b, 2011a, b; Hill and Xie 2011). The PEK model has a double exponential form (equation (1)) (1) where MC is the moisture content at time t of exposure of the sample to a constant relative humidity (RH) and MC₀ is the moisture content of the sample at time zero. The sorption kinetic curve is composed of two exponential terms which represent a fast process {MC₁[1–exp(−t/t₁)]} and a slow {MC₂[1–exp(−t/t₂)]} process having characteristic times of t₁ and t₂ respectively. The terms MC₁ and MC₂ are the moisture contents at infinite time associated with the fast and slow processes respectively. It has been demonstrated that the PEK model fits the sorption kinetics data for small wood and plant fibre samples with a very high degree of accuracy. It has been argued that the sorption kinetics is determined by the rate of swelling (relaxation) of the cellular matrix. In this model, the two processes are considered to be represented by a pair of series-coupled Kelvin–Voigt elements (Fig. 1).

OPEN IN VIEWER

Figure 1

Two Kelvin–Voigt elements in series

The form of the fast and slow component of the PEK equation is identical with that describing the dynamic response of a Kelvin–Voigt element when subjected to an instantaneous stress increase σ₀ (2) where ϵ is the strain at time t, E is the elastic modulus and φ is a time constant which is defined as the ratio η/E, where η is the viscosity. In the present case, there is a change in atmospheric RH which leads to a response in the wood cell wall. The maximum swelling pressure (Π – which is here equivalent to σ₀) that is exerted on an elastic gel when the surrounding water vapour pressure is raised from an initial value p_i to final value p_f is given by the following equation (3) where ρ is the density, M is the molecular weight of water, R is the gas constant and T is the isotherm temperature in kelvin. In the present situation, the strain of the system is assumed to be equivalent to the volume change of the cell wall as a result of water vapour adsorption or desorption. This volume change is assumed to be linearly related to the change in the mass fraction of the water present in the cell wall. The adsorbed water vapour molecules exert a pressure within the cell wall leading to dimensional change, which is considered equivalent to the extension of the spring in the Kelvin–Voigt model. The PEK model has been criticised recently on the grounds that if not all of the kinetic curve is used for the analysis then different results are obtained for the fitting parameters (Driemeir et al. 2012). It is for this reason that we set the dm/dt sorption criteria, in order that each kinetic curve is comparable. In the study of Dreimeir et al. (2012), they set the instrument to change RH after 60 min sorption time (180 min for the first sorption step), irrespective of how close to equilibrium the sample was meaning that the kinetic curves cannot be directly compared with one another.

Experimental

The wood material used in this study was the clear sapwood of kiln dried Scots pine (Pinus sylvesteris L.) obtained from southeast Finland with an average density (at RH 65%, 20°C) of 0·59 g cm⁻³. The material was conditioned for at least 6 months before use. Four specimens with dimensions of 145 mm (longitudinal), 95 mm (tangential) and 12 mm (radial) were cut from the same board. One specimen was thermally modified, another was densified and the third specimen was densified and subsequently thermally modified. In addition, one control specimen was left untreated. All samples were re-conditioned at 65% RH and 20°C for 3 weeks after treatment.

Densification was performed in an open-press system at 150°C for 1 h, followed by a 3 h cooling period until the press temperature was below 100°C. Perforated metal plates were used to facilitate the escape of steam generated in the specimens during densification. The specimens were compressed at a pressure of 6 MPa in the radial direction to a target thickness of 5 mm. After densification, the thickness of the wood samples was measured and found to be 5·38 mm, indicating spring-back of 0·38 mm after treatment.

For the thermal modification process, the specimens were first oven dried for 24 h at 103°C. Subsequent thermal modification was carried out in a bespoke treatment chamber and began at a temperature of 120°C without steam injection. A constant temperature of 120°C was maintained for 0·5 h; thereafter the temperature was increased to 200°C over a period of 0·5 h. At the same time saturated steam was injected into the chamber. The treatment at 200°C was carried out for 3 h, with continuous steam injection. A mass loss of 3·4% due to thermal modification was recorded in the non-densified sample and 4·3% in the densified sample. No noticeable spring-back occurred in the densified specimen following thermal modification.

For the sorption studies, small wood samples of 5·0±1·0 mg were removed from the samples for the water vapour sorption analyses using a scalpel. Analysis was performed using a surface measurement system (London, UK) intrinsic dynamic vapour sorption apparatus. The sorption cycle started at 0% RH and increased in 5% RH steps to a maximum of 95% RH. The desorption cycle employed a reverse sequence. Three sorption cycles were run. All other experimental details are exactly as described previously (e.g. Hill et al. 2012a, b).

Results and discussion

Figure 2 shows the sorption isotherms over three sorption cycles for the unmodified, densified, thermally modified and thermally modified plus densified samples. Densification reduces the equilibrium moisture content (EMC) of the material, as does thermal modification. The isotherm for the wood that was densified and then thermally modified was almost identical with that of the thermally modified wood. The original intention of running multiple consecutive sorption cycles was to investigate whether or not there was any change in sorption behaviour as a result of relaxation phenomena (an annealing effect). This arises as a result of plasticisation of the cell wall polymeric network by sorbed water. For this reason, three sorption cycles were performed and the isotherms were determined, with the results from this study shown in Fig. 3. There is essentially no difference in the shape of the sorption isotherms for the unmodified wood over the three consecutive sorption cycles (Fig. 3a). With the densified wood (Fig. 3b) there is a significant difference between the first and second sorption cycles, which is attributed to an annealing effect due to plasticisation of the cell wall at high RH. This behaviour was also noted with the thermally modified plus densified wood (Fig. 3d). In both these cases, the change in the isotherms could (potentially) be attributed to a stress relaxation process at high cell wall moisture contents. However, the same behaviour is also noted with thermally modified wood (Fig. 3c), an entirely unexpected result. In all cases, the change occurs on the adsorption cycle and involves an increase in the EMC from the first to second sorption cycle. The reason for this behaviour is not known, but may involve an increase in sorption site accessibility between the first and subsequent sorption cycles. This comes about as a result of plasticisation of the cell wall macromolecular matrix network. However, it should be noted that all the samples had been pre-conditioned at 65% RH for 3 weeks and that therefore exposure to RH values in excess of this is required in order to facilitate the annealing effect. Further studies are required to examine where this relaxation RH point may occur. This finding has important implications for the study of sorption behaviour of wood and thermally modified wood in particular. Its importance cannot be over-emphasised. Clearly, it is necessary to perform more than one sorption cycle in order to obtain a reproducible isotherm. Furthermore, the sorption behaviour is strongly dependent upon the sorption history of the wood and will depend very much upon the use to which the modified wood is put.

OPEN IN VIEWER

Figure 2

Comparison of sorption isotherms for first sorption cycle for unmodified (U), densified (D), thermally modified (TM) and thermally modified plus densified (TMD) wood

OPEN IN VIEWER

Figure 3

Changes in sorption isotherms over three consecutive cycles for a unmodified, b densified, c thermally modified and d densified plus thermally modified wood

All of the modifications resulted in an increase in hysteresis when compared to the unmodified wood for cycle 1 (Fig. 4). The phenomenon of sorption hysteresis in wood can be explained by consideration of the sorption process in a glassy polymer matrix (Hill et al. 2009). During the adsorption process, nanopores are created within the cell wall matrix, whereas these collapse during the desorption process. The rate of response of the matrix to the ingress or egress of sorbent molecules is dependent upon the mobility of the constituent molecules. Below the glass transition temperature T_g, the matrix is unable to respond instantaneously to the movement of the sorbent molecules and consequently the opening and closing of the nanopores in response is delayed somewhat. As a result, the adsorption and desorption processes take place in a material that is in different states. Based upon such a model, it would be predicted that there is an increase in sorption hysteresis as the distance between the isotherm temperature and the T_g increases. If the isotherm temperature remains constant, then an increase in T_g will result in an increase in hysteresis. Thermal modification degrades the hemicelluloses in the cell wall and is additionally thought to increase the level of cross-linking in the lignin matrix (Hill 2006), both of which result in an increase in T_g. Although the densified sample was not subjected to such high temperatures as the thermally modifies samples, some degradation of the cell wall polymers will have taken place at 150°C. It is perhaps of interest to note that the absolute hysteresis values of the modified wood are similar, irrespective of the type of modification and despite there being significant differences between the isotherms of the densified wood and the thermally modified wood. However, hysteresis associated with the isotherms of the modified wood decreases dramatically between the first and second sorption cycles and to a lesser extent between the second and third cycles. This reduction in hysteresis occurs due to the annealing effect associated with the first sorption cycle, yet the chemical changes to the cell wall polymers brought about through thermal modification remain the same. Based upon the relaxation model for hysteresis, this reduction would imply that the T_g of the matrix has now been reduced as a result of exposure of the wood to high RH; this suggests that a high humidity post-treatment step after thermal modification might provide benefits in terms of wood properties.

OPEN IN VIEWER

Figure 4

Variation in absolute hysteresis over three sorption cycles for unmodified Scots pine and modified wood

It has recently been argued that the sorption kinetic processes are also controlled by molecular relaxation phenomena and that there should be a link between sorption hysteresis and kinetics (Hill and Xie 2011). The kinetic curves were analysed using the PEK model and the curve fitting parameters used to calculate the internal matrix modulus of elasticity and viscosity of the cell wall. The variation in modulus associated with the fast process for the different wood samples is shown in Fig. 5. A number of features can be observed in these plots. There is a reduction in modulus from 20–40 GPa at low cell wall moisture content, to somewhere of the order of a few hundred MPa at high cell wall moisture content. This large decrease can be explained as being due to plasticisation of the cell wall matrix macromolecules by sorbed water. However, such a large reduction in modulus is not observed when the mechanical properties of wood are examined at high moisture contents and it is therefore necessary to consider whether this behaviour is realistic. At lower RH, the high values of the modulus indicate that there is presumably a contribution from the microfibrils, whereas at the upper end of the hygroscopic range, this contribution is absent and the sorption rate is only apparently determined by plasticised matrix. What is being determined in these experiments is the rate limiting step of the sorption process and it is proposed that polymeric relaxation processes are the ‘bottleneck’. The reduction in modulus may indicate that there is a considerable degree of hydrogen bond breaking and a much greater amount of molecular mobility in the matrix than at lower cell wall moisture contents, leading to a decoupling of any contributions from the microfibrils. An important consideration when discussing these results is that the mechanical properties of the cell wall are being determined by the application of a stress within the cell wall matrix rather than externally. Clearly, at this stage, the approach is speculative and there is a need to undertake independent experiments to verify the conclusions. What form such experiments might take is an open question at this time; it is possible that nanoindentation may be a possible approach, but even here the experiments still use an externally applied stress to determine the material properties. In all cases, the modulus values associated with the fast adsorption process are higher than those found for desorption and there is more variation in the fast process modulus values between sorption cycles.

OPEN IN VIEWER

Figure 5

Difference in modulus associated with fast adsorption process for unmodified and modified wood

Wood modification affects the magnitude of the fast modulus associated with adsorption, with the thermally modified plus densified wood exhibiting the highest values. This is also the case for the desorption modulus values, although the differences are smaller. Similar behaviour is found with the variation in slow process modulus. Modification of the wood also affects the adsorption modulus values, which are larger for the thermally modified and thermally modified plus densified samples, but the desorption modulus is scarcely affected. The trends in modulus are consistent with the effects of moderate thermal modification which is often claimed to increase the modulus of elasticity of wood. An increase in modulus is associated with a lower EMC value and the decrease in EMC in the isotherms is divided between the two processes under adsorption conditions, but predominantly with the fast process with desorption. There is a general trend for the modulus values to decrease with each sorption cycle, where such differences occur. This could perhaps be associated with an annealing process taking place within the cell wall matrix. The fast process viscosity is insensitive to wood treatment and also shows little difference between adsorption and desorption as well as between cycles. This is in marked contrast to the behaviour noted with the slow process viscosity (Fig. 6) shown here for densified wood. The viscosity associated with the slow process is always higher under desorption conditions compared to adsorption, whereas the opposite is the case with the fast process. Thermal modification results in an increase in slow process viscosity, an observation consistent with the known effects associated with destruction of the hemicelluloses due to thermal modification.

OPEN IN VIEWER

Figure 6

Comparison of viscosity associated with fast and slow processes under conditions of adsorption and desorption for densified wood

Conclusion

Densification, thermal modification and a combination of those treatments reduce the hygroscopicity of wood, but all result in an increase in the hysteresis between the adsorption and desorption branches of the isotherm for the first sorption cycle. Changes were noted in the sorption isotherms when the wood samples were subjected to three sorption cycles. With unmodified wood, these differences were minor and restricted to the upper part of the hygroscopic range, but with the other wood samples, there was a large change recorded in the adsorption branch between the first and subsequent cycles, which resulted in a reduction in hysteresis, in some cases to values lower than that found for the unmodified wood. The sorption kinetics behaviour was analysed using the PEK model and the parameters from the curve fitting procedure were used as input into Kelvin–Voigt series-coupled elements, thus allowing calculations to be made for the moduli and viscosities associated with the fast and slow sorption processes. The moduli associated with the sorption processes are affected by wood treatments, with the modulus increasing in association with the thermal modifications especially. The fast process viscosity is little affected and the main changes due to wood modification are observed with the slow process viscosity.