Variability in water vapour sorption isotherm in Japanese Larch ( Larix kaempferi Lamb.) – earlywood and latewood influences

Introduction

This paper reports on an investigation of the sorption properties of larch, which is part of a larger project investigating the physical properties of UK grown larch. In terms of total area, the UK has one of the largest larch resources in Europe outside of Russia (Bergstedt and Lyck 2003). In the UK, larch timber is mainly used for external cladding; however, much of this material is Siberian larch (Larix siberica Ledeb) which comes from old growth stands, has narrow annual rings and a large mature heartwood zone. While there is increasing use of home grown larch for external cladding, it is still an undervalued commercial species in the UK in terms of potential timber utilisation. The majority of UK grown larch timber is used for pallet production, which is a comparatively low value market. In other parts of Europe and North America, larch timber is used in applications where a combination of high strength, durability, or hardness is required; for example bridges and heavy structures. It is also used in indoor applications where aesthetics are important, e.g. ceilings, walls, exposed roof structures, stairs and floors, and as exterior cladding due to its relatively good natural durability, which often eliminates the need for preservative treatment. The UK resource is under considerable threat from Phytophthora ramorum with more than two million infected trees felled since 2009. The relatively low value markets that larch timber is traditionally destined for are already saturated. The overall aim of the project is to develop a better understanding of the performance of larch timber and develop predictive models which can be used to aid the management of larch stands to maximise its value. This paper concentrates on investigations into the sorption behaviour of Larix kaempfera.

There is extensive literature concerned with the relationship of wood with atmospheric water vapour. It is well established that the water vapour sorption isotherm is influenced by the nature of the substrate, with variations reported for different wood species (Popper and Niemz 2009). It is also known that the presence of extractives in wood can influence the sorption isotherm (Popper et al. 2007) and that there are differences in sorption behavior between heartwood and sapwood (Ball et al. 2001; Obataya et al. 2006). Experiments have also revealed differences in hygroscopicity between juvenile and mature wood (Lenth and Kamke 2001; Majka and Olek 2008). However, there are very few studies investigating the influence that earlywood and latewood have upon the sorption behavior and in particular whether there is any influence due to the location of the wood in the stem. One reason for this is undoubtedly because many previous studies have used the method where wood samples are in an atmosphere where the relative humidity (RH) is controlled by saturated salt solutions and it is necessary to have large samples spanning several growth rings for the gravimetric determinations. In separating earlywood from latewood, very small samples are often obtained, making measurements very difficult. Despite the challenges of conducting such experiments there have been some studies where the influence of ring position and earlywood and latewood on the sorption behaviour has been reported. Neimsuwan et al. (2008) used saturated salt solutions to control the RH and passed a constant airflow from the humidity chambers over Pinus taeda L. wood samples, which were suspended by a hang-down wire to a contact angle measurement balance. In this way it was possible to determine both the sorption isotherm and the sorption kinetics. It was found that the sorption rate of the outer tree rings was generally higher than the sorption rate exhibited by the inner rings and the initial sorption rate of earlywood was always higher than that of the latewood. The sorption isotherms were analysed using the Hailwood Horrobin model, with differences in the fitting parameters being reported, but no direct comparison of the experimental isotherms given in the paper. Sargent et al. (2008, 2010) constructed a dynamic sorption apparatus by attaching small wood samples to piezoelectric transducers and used this equipment (called a Dynamic Sorption Platform) to examine the sorption behaviour of latewood and earlywood from the sapwood and heartwood of radiata pine. The wood specimens showed differences in the rate of sorption depending on the origin of the wood, but the EMC was unaffected. Apparent diffusion coefficients for earlywood in heartwood and sapwood were similar, but latewood specimens showed higher diffusion coefficients than earlywood and were different in the heartwood and sapwood.

Differences in the fibre saturation point (FSP) between the earlywood and latewood of Douglas fir and aspen, as determined by solute exclusion, have been reported by Ahlgren et al. (1972). These differences were removed by pre-extracting the samples with solvent. However, the values for the FSP obtained were considerably higher (54–82%) than are usually reported (around 40%) using this technique, casting doubt on the observations. Kärenlampi et al. (2005) investigated the response of 50 mg microtomed samples of earlywood and latewood spruce sapwood to changes in RH in a humidity chamber, recording mass changes twice a day. It was found that the EMC of latewood samples was slightly higher than earlywood and it was also reported that the EMC at any given RH did not change with repeated wetting and drying cycles.

With the advent of commercially available dynamic vapour sorption (DVS) apparatus, it has become possible to routinely make determinations of water vapour sorption on isotherms on very small samples (typically 10 mg), making it very easy to examine possible differences in properties between earlywood and latewood. Sharratt et al. (2010) used a DVS instrument to determine the sorption behaviour of earlywood and latewood samples of Scots pine (Pinus sylvestris L.). No difference in the sorption isotherm was found, but it was reported that that earlywood samples responded faster to changes in atmospheric RH compared to latewood. A commercial DVS apparatus was also used in a study of the effect of ring position on the sorption kinetics and sorption isotherm of Sitka spruce [Picea sitchensis (Bong.) Carr.]. In this work Hill et al. (2011) found no trends in the sorption isotherm related to ring number, nor did earlywood and latewood show any differences in the isotherm. However, differences in sorption kinetics were again found that related to wood type, with slower sorption being observed for latewood.

The purpose of the present study was to conduct an analysis of the sorption properties of wood taken from Japanese larch (Larix kaempferi Lamb.) and determine if ring position and earlywood/latewood had any influence upon the sorption behaviour. Such investigations are of importance in the understanding of wood distortion during drying, the stability of clear coatings on an exposed wood surface, etc. In the course of conducting this study, anomalous behaviour was found with the properties of the latewood compared to the sapwood and it is this that is reported upon and discussed herein. As part of the wider study into the physical properties of larch, samples were also sent to Innventia in Sweden for analysis using Siliviscan equipment. The SilviScan instrument was developed by Robert Evans at CSIRO in Melbourne as a rapid automated assessment tool for measuring radial variation in wood density, microfibril angle (MFA), cross-sectional dimensions of tracheid cell walls, modulus of elasticity, earlywood and latewood content and properties (Evans et al. 1995; Evans 1999). MFA and density data are reported in this publication and a possible link to the sorption behaviour discussed.

Experimental

Wood samples were cut from a larch tree from a mature stand in central Scotland. A 20 cm thick disc was taken at 1·5 m height and kiln dried at a wet bulb temperature of 50°C. The 20 cm discs were later cut to 2–3 cm thickness on a WoodMizer (WoodMizer UK). A section was cut from the north face (pith to bark) of the disc with cross-sectional dimensions of approximately 15×15 mm and glued to a sample holder. When set, the sample holder was positioned so that it could be automatically passed through the first of 2 sets of twin circular fine saw blades for the 2 mm radial-tangential cut. This 2 mm radial strip was then placed laterally in another sample holder and passed through the second twin-blade saw with the distance between the blades preset to 7 mm. The samples were now of the correct dimensions for SilviScan analysis. The samples were mounted on a sample holder and automatically passed through the optical scanning microscope lens and images were recorded using a CCD camera. From the images, the SilviScan software calculates mean tracheid dimensions and cell wall thickness from the image at 25 μm intervals. The orientation of the annual rings across the sample (isopycnic angle) is also determined at this stage. The samples were then scanned using X-ray densitometry to determine the radial density profile at 25 μm intervals. Using the isopycnic angle information obtained by the optical scanner, the samples were rotated around the radial-longitudinal axis to align each annual ring with the X-ray beam. The radiographic images obtained were related to the volumetric density of each sample to account for varying attenuation coefficients between samples. For the final stage, each sample holder was transferred to the X-ray scanning diffractometer. Analysis of the diffraction profiles provides information about the orientation and distribution of the cellulose microfibrils in the S2 layer of the secondary call wall. Samples for SilviScan and DVS measurements were not subjected to any extraction procedure.

A second, adjacent set of samples were cut and prepared from the same north-facing block for the water sorption experiments. Earlywood and latewood samples were very carefully removed from this slice to obtain samples from growth rings at year 4, 28, 39, 47 (i.e. ring 4 is closest to the pith). Samples of 10 mg±1 mg were used for the experiments (it is important to note that sample to sample weight variations should be no more than 20%). One replicate was used for each sorption run. Isotherm analyses were performed using a Surface Measurement Systems ‘Intrinsic’ Dynamic Vapour Sorption apparatus. This is a very useful means of generating accurate sorption isotherms at different temperatures and using a range of pre-set RH values. The apparatus contains a measurement pan suspended from a Cahn ultra-sensitive microbalance, capable of measuring changes in sample mass as low as 1 part in 10 million. The instrument is claimed by the manufacturers to have excellent long term stability and no drift in mass values was found when the measurement pan did not contain sample material. The sample holder is connected to the microbalance by a hanging wire sitting in the measurement chamber, which is located in a thermostatically controlled cabinet, and through which there is a constant flow of nitrogen gas into which is mixed nitrogen containing a pre-set amount of water vapour. A run started at zero percent RH and increased in 5% RH steps up to a maximum of 95% RH, before decreasing to zero percent RH in 5% RH steps. Humidity and temperature probes are located in close proximity to the sample holder providing direct measurement of these parameters. It was found that the temperature and humidity values were very stable during the tests, although both the RH and temperature were seldom exactly at the pre-set values and it was necessary to read the actual RH and temperature values at each adsorption and desorption stage from the output data spreadsheets. The sample material was placed onto a pre-cleaned sample pan and carefully placed on the hang down wire connected to the microbalance, the sample chamber was then closed. Before a run commences, the instrument produces a ‘drying curve’ with the RH set at zero percent until the sample weight is stable, according to the pre-set dm/dt criterion. Two full sorption cycles were performed. The instrument maintained a sample at a constant RH until the weight change per minute (dm/dt) reached a value lower than 0·002% per minute for a ten minute period; a criterion that from previous experiments had been shown to give a sample MC within less that 0·1% of the equilibrium value. A full description of the apparatus and the methodology has been reported (Hill et al. 2009; Hill et al. 2010).

Results and discussion

The DVS instrument begins an isotherm run by initially passing dry nitrogen over the sample until the equilibrium criterion is reached. This produces a drying curve. When the rate of change of the sample mass achieves the programmed equilibrium criterion, the balance is automatically tared and this value is used as the dry mass of the sample for all subsequent calculations of the moisture content. This means that the sample is not entirely moisture-free after the drying run, but in practice, the slight error that this introduces does not usually compromise the data quality. Similarly, when a sample approaches equilibrium at any given RH, the transition to the next RH is based upon the pre-programmed equilibrium criterion. Many previous experiments over the past decade have established that the mass change criterion (dm/dt) of less than 0·002% MC per minute over a ten minute period is satisfactory and produces isotherms with EMC values within 0·1% of the ‘true’ equilibrium value (Hill et al. 2009; Hill et al. 2012a). Pre-drying the sample in an oven at 103°C will inevitably cause some damage to the material and can produce artefacts in the isotherm. Pre-drying by placing over a desiccant (such as phosphorus pentoxide) is no guarantee that the sample is fully dry, since the ‘constant’ mass criterion which is used as a definition of a water free sample may not be true. Finally, the act of transferring the sample to the DVS instrument results in exposure to atmospheric moisture, meaning that a ‘dry’ sample is not available before the experiment starts. It is necessary to establish that the samples show reproducible behaviour and for this reason it is now becoming more common to perform two sequential isotherm runs. Thermal modification of wood has been shown to exhibit changed sorption behaviour between the first sorption cycle from that measured in subsequent cycles (Mitchell et al. 1953; Obataya et al. 2000; Obataya and Tomita 2002; Hill et al. 2012b). The behaviour has also been noted with historic wood samples, although this appears to depend upon sample history (Popescu and Hill 2013). However, variation in the isotherms for ‘unaged’ wood samples has not been observed and reproducibility between isotherms is usually assumed. The exception to this is the first desorption from a ‘green’ state compared to subsequent isotherms (Hill et al. 2009).

Figure 1 shows a comparison of the sorption isotherms (cycle 1) of earlywood and latewood samples taken from growth rings 4, 28, 37 and 49. The isotherms have been superimposed in order to show whether there are any differences between the growth rings. Although there is some variation in the isotherms of the different earlywood samples, this does not correlate with ring position (e.g. rings 4 and 47 show higher EMC values in the desorption step). These differences are associated with the observed slight variation in isotherms between different samples, rather than representing any significant inherent variation in properties (Hill et al. 2009). This contrasts markedly with the isotherms exhibited by the latewood samples for the same growth rings. There is a reduction in EMC for rings 28, 37 and 49 compared with ring 4. However closer inspection of the data also reveals that for the isotherms associated with rings 28, 37 and 49, negative EMC values are recorded when the desorption branch returns to 0% RH. Furthermore, the adsorption branches of the isotherm for these rings did not show the usual sigmoidal form associated with wood (i.e. IUPAC type II behaviour), which is seen with the earlywood and latewood ring 4. Rather, the isotherm shows a continuous upward curve (convex), characteristic of an IUPAC type III isotherm. The explanation for this is due to the presence of residual water remaining in the cell wall after the drying curve prior to the first sorption cycle. The negative values obtained at zero percent RH on desorption in the first cycle indicate that the sample was not fully dried during the initial process where the drying curve was produced. Accordingly, the latewood samples were subjected to an additional sorption cycle in order to determine whether the isotherm would now return to zero EMC, or still continue to exhibit negative MC values on completing the sorption cycle. The results of this study are reproduced in Fig. 2. This shows the result of two sequential isotherm cycles. In the case of rings 4 and 28, the second isotherm exhibits IUPAC Type II behaviour and the desorption cycle returns to 0% EMC, but with rings 39 and 47 the isotherm still returns to negative MC values on the second cycle, although to a lesser extent than was observed with the first sorption cycle. When using the standard drying conditions with the usual mass change criteria (i.e. dm/dt = 0·002% over a ten minute period) which have proved satisfactory for all other wood species tested to date, this leads to the presence of residual water in the cell wall of the latewood specimens. In order to remove this residual water it is necessary to perform more than two sorption cycles. This effect is not observed with the earlywood and has not previously been observed with any other wood species studied in this laboratory. Running the initial drying curve for longer periods of time prior to an isotherm run would not result in any improvement in this situation, since the curve is asymptotically moving towards a false ‘zero’ moisture content. Indeed, this suggests that the standard criterion for zero moisture content (drying to constant weight) should be treated with some caution.

OPEN IN VIEWER

Figure 1

Differences in sorption isotherms for earlywood and latewood for different growth rings at diameter breast height of larch wood for first sorption cycle (Ring 4 is close to pith)

OPEN IN VIEWER

Figure 2

Differences in sorption isotherms of latewood of larch for different growth rings for first and second sorption cycles

It is suggested that the reason may lie with the way in which the pathways from the interior to the surface of the cell wall are connected and how these connections are disrupted as the cell wall dries out. The swollen cell wall of wood contains transient nanopores which exist between the matrix macromolecules and which contain the sorbed water molecules (Hill et al. 2005). The interconnections between these nanopores allow for communication between the surface and interior of the cell wall. If nanopores collapse in a critical part of the communication pathway, then water can no longer escape to the surface. Such a situation can be modelled using percolation theory, where a network can only communicate between two spatially separated points above what is termed the ‘percolation threshold’ (Hill and Hillier 1999). However, since the formation of nanopores is reversible (unless hornification occurs) there is the possibility of the network reforming on the next adsorption cycle, possibly with different interconnections. A pathway for escape of the water molecules may then become available on the next drying cycle. This observation has important implications for the assumption that drying to constant weight implies absence of water in the cell wall. It has been known for some time that the cell wall nanoporosity (as determined from non-freezing water in dynamic scanning calorimetry) varies between earlywood and latewood (Kärenlampi et al. 2005). It has also been demonstrated how the resistance to the transport of water vapour increases markedly from the earlywood to the latewood cell wall with a considerably reduced porosity associated with the latewood (Derome et al. 2012). A reduced nanopore volume in the cell wall would lead to a lower probability of interconnections and hence increased resistance to transport of water molecules, particularly as MC is reduced. This hypothesis remains speculative at present, but does explain the observations.

The sorption isotherms of the latewood samples in the second sorption cycle are compared with the earlywood isotherms from the same growth rings in Fig. 3. The sorption isotherms for rings 39 and 47 exhibit EMC values substantially lower than those associated with the earlywood, even when it is taken into account that some slight adjustment upward is necessary to compensate for the non-zero MC at the start of the isotherm run. The earlywood and latewood isotherms for rings 4 and 39 are very nearly superimposable. These large differences in the sorption isotherm for the later growth rings would also indicate differences in swelling behaviour leading to stresses developing at the earlywood–latewood boundary if coatings are applied. This important finding indicates the need for further work investigating the possible differences in sorption behaviour between growth rings and earlywood and latewood.

OPEN IN VIEWER

Figure 3

Differences in sorption isotherms for earlywood and latewood of larch for different growth rings for second sorption cycle

In order to provide a possible explanation for the difference in behaviour between the different growth rings, the wood was also studied using SilviScan. Adjacent sample slices were studied using Silviscan equipment in order to determine the mean microfibrillar angle of each growth ring as well as the density of the earlywood and latewood in each growth ring. The apparatus first determines the average radial and tangential fibre width and cell wall dimensions are determined at 25 μm intervals using optical image analysis. X-ray densitometry is then used to determine the radial density profile (also at 25 μm intervals), then average MFA is determined at 5 mm intervals using a scanning X-ray diffractometer which estimates mean MFA from the resultant diffraction profiles. The Silviscan-3 software can then calculate the measured properties for each annual ring in the sample. The results from this study are reproduced in Fig. 4. The change in MFA with growth ring position is shown in Fig. 4a, indicating a reduction in MFA until approximately ring 15, indicating that there is juvenile wood below the age of 15 years. Hence only ring 4 is composed of juvenile wood. According to Karlman et al. (2005), the age set to demarcate the boundary between juvenile and mature wood in Larix species is set at 15–20 years. Karlman et al. used X-ray densitometry to investigate the changes in wood density in L. kaempferi, reporting that there was lower density up to annual ring 20 and used this as a criterion for identifying the juvenile wood. However, inspection of their data reveals an asymptotic increase in ring density with age, with no obvious demarcation at ring 20. Unfortunately their study did not differentiate between earlywood and latewood. The data reported herein (Fig. 4b) shows that the increase in ring density with age is almost entirely due to changes in latewood and not earlywood, however although there is a trend, there is considerable variation in density from ring to ring. Visual inspection showed that all of the rings except ring 47 were located within the heartwood region. No obvious relationship exists between the wood density of the annual ring and the sorption behaviour, although there may possibly be a link with MFA. Without further studies, it is not clear whether the changes in the isotherm observed here are restricted to larch or are a more widespread phenomenon. The role of the often high percentage of water-soluble arabinogalactans in larch was not investigated in this study and may perhaps be linked to the behaviour reported here. Further studies of this nature are clearly required.

OPEN IN VIEWER

Figure 4

Changes in mean microfibril angle (MFA) and density with growth ring position, as determined by Silviscan

Conclusion

This study has shown substantial differences in the sorption isotherm in latewood samples from later growth rings between the first and second sorption cycles. Earlywood samples did not show such differences in behaviour and nor did latewood samples from the earlier growth rings (nearer the pith). Latewood samples from later growth rings also exhibited reduced hygroscopicity when compared to adjacent earlywood growth rings. The position of the wood in the tree can have a major influence upon the sorption isotherm. It is recommended that a minimum of two full sorption cycles is conducted when isotherms are determined. This study has shown clearly that the location of wood in the tree can have a significant influence on the sorption behaviour and indicates the need for more studies of this nature. Very little work has been performed investigating this phenomenon and a much larger study investigating the influence on the sorption isotherm of the position of the wood within the tree, as well as extending this work to other species is required. Different sorption behaviour in the annual rings and between earlywood and latewood would also be reflected in variations in swelling behaviour and the development of shear forces at the earlywood–latewood boundary. This type of behaviour would explain observations showing that premature failure of clear coatings on wood often occurs in this region.

Acknowledgements

JR wishes to acknowledge financial support from Forest Research UK and support from COST Action FP0904 for funding for a short term scientific mission to Innventia in Sweden.