Save

Mitigation of cellular collapse during drying of Eucalyptus nitens wood using supercritical CO2 dewatering

In: IAWA Journal
Authors:
Hamish Pearson Scion, 49 Sala Street, Private Bag 3020, Rotorua, New Zealand

Search for other papers by Hamish Pearson in
Current site
Google Scholar
PubMed
Close
https://orcid.org/0000-0002-7019-669X
,
Lloyd Donaldson Scion, 49 Sala Street, Private Bag 3020, Rotorua, New Zealand

Search for other papers by Lloyd Donaldson in
Current site
Google Scholar
PubMed
Close
https://orcid.org/0000-0002-3845-4177
, and
Mark Kimberley Scion, 49 Sala Street, Private Bag 3020, Rotorua, New Zealand

Search for other papers by Mark Kimberley in
Current site
Google Scholar
PubMed
Close
https://orcid.org/0000-0003-2890-0871
Open Access

Summary

Removal of lumen water by dewatering using supercritical CO2 offers an alternative method to mitigate cellular collapse in susceptible hardwoods compared to conventional timber drying methods. The anatomy of Eucalyptus nitens was quantitatively measured by light microscopy, SEM and micro-CT to provide an understanding of the mechanism of collapse during drying. These measurements were then used to recalibrate a previously developed fluid-dynamics model to predict E. nitens vessel dewatering and develop a dewatering treatment strategy for collapse mitigation. The lumens of E. nitens were from fibres (58.5% cross-section) and vessels (10.0% cross-section) with mean diameters of 8 and 142 μm, respectively. Micro-CT measurements revealed that the vessels were empty after treatment with a supercritical CO2 dewatering schedule optimised for softwood. However, the fibres remained full and this led to significant collapse during subsequent oven drying. Based on this information, a two-phase dewatering schedule was developed to include removal of fibre lumen water. Results showed that 90% of collapse could be mitigated to a change in external volume of only 3.9% provided the green moisture content was lowered to 70% before oven drying. The predicted effective diffusion coefficient of CO2 in E. nitens was comparable to Pinus radiata and they showed similar anatomical tortuosity and porosity resistance in their hydrofluidic networks. Collapse mitigation using supercritical CO2 could be combined with extraction of desirable sap components, post-dewatering drying, preservative treatment, and mechanical forming. These processes may be achieved in a single supercritical plant and apply to most anatomically similar hardwoods.

Introduction

Collapse is a type of cellular distortion that can occur in wood. Typical indications are crushed, flattened, fractured or buckled cells which occur when cell-wall strength and stiffness are compromised by load forces which exceed the cellular compressive strength (Bodig & Jayne 1982; Chafe 1985; Blakemore 2011). Collapse may be observed externally from corrugations or excessive shrinkage, but it can easily become integrated with other forms of distortion such as checking, and present as internal voids or honeycomb (Ross 2010). This can lead to collapse being defined in different ways and cause difficulty in interpreting the results from different studies (Chafe 1985). Evaporative wood drying can lead to collapse when compressive drying stresses are created in the core of a board as the exposed outside shell dries and shrinks in advance (Skaar 1988). However, the principal cause of collapse arises from liquid tension, when elevated tensile capillary forces occur as liquid menisci retreat into the filled lumens of cells whose cell walls are above the fibre saturation point (Kauman 1964a, b; Siau 1995). Wood collapse can occur in softwoods or hardwoods but typically occurs in wood of low permeability (Siau 1995). It is therefore highly prevalent amongst hardwoods such as those within the genus Eucalyptus (Kauman 1964a, b; Wilkes 1988; Bariska 1992; Wu et al. 2010; Yang & Liu 2018), particularly E. nitens ((H. Deane & Maiden) Maiden) (Milne 1991; Kube & Raymond 2005; Blackburn et al. 2010), and members of the ‘Ash’ group eucalypts which includes E. regnans (F. Muell.) (Kininmonth 1971; Chafe 1987; Blakemore & Northway 2009).

Eucalypts are endemic to Australia, but their collapse is a significant issue for the wood processing industry there even when combined with steam reconditioning schedules that aim to recover cells from collapse (Kauman 1964a, b; Chafe 1986; Blakemore & Northway 2009). Interest also exists in New Zealand for eucalypts as an alternative solid wood species to Pinus radiata (D. Don) (Shelbourne et al. 2002) although most of the crop is currently used for pulp rather than solid wood, and the current (2021) area of eucalypts is only 1.3% of the plantation forest estate compared to 89.7% P. radiata (F.O.A. 2021). This compares to a worldwide plantation area of 26% Eucalyptus species which is the second-largest genus after Pinus at 46% (F.O.A. 2021). E. nitens is stronger and has a higher elastic modulus than P. radiata (Kininmonth & Whitehouse 1991; McKinley et al. 2002) and outperforms the growth rates of other eucalypts in New Zealand (Low & Shelbourne 1999). If recovery could be improved through the reduction of collapse without the need for reconditioning, the increased yield could improve the economic viability of Eucalyptus species in New Zealand and help diversify the species composition of New Zealand plantations.

Collapse mitigation techniques generally fall into three categories, aiming to reduce surface tension or increase nucleation, to increase permeability, or increase cell-wall stiffness (Blakemore & Northway 2009). Sap extraction using nucleated supercritical CO2 dewatering (Aggarwal et al. 2019; Asafu-Adjage et al. 2021) followed by oven drying has been proven to work well with P. radiata softwood (Franich et al. 2014; Dawson & Pearson 2017). However, the same dewatering treatment schedule has been found to yield relatively poor moisture removal and result in collapse after drying for impermeable hardwoods (Liu et al. 2022), especially E. nitens (Dawson & Pearson 2017). The primary reason for this is thought to be due to anatomical differences. This study aimed firstly to understand these differences from an anatomical point of view. Secondly, based on this understanding the study aimed to develop supercritical CO2 dewatering treatments which mitigate drying collapse in E. nitens. Thirdly, a previously developed predictive computational fluid dynamics (CFD) dewatering model (Pearson et al. 2019b) was used to confirm the dewatering mechanism and quantify the tortuosity and porosity of E. nitens compared with P. radiata using the CO2 diffusion coefficient. Although the study focussed on collapse in E. nitens as a representative of difficult-to-dry hardwoods, the results are likely to apply to hardwoods in general.

Materials and methods

Anatomical assessment

The anatomy of E. nitens was assessed to establish its likely dewatering behaviour as a hardwood compared to P. radiata softwood. The relative transverse cross-sectional areas of open (porous hydrofluidic lumen network) vessels compared to closed (sealed lumens) fibres were also required for the estimation of dewatered bulk sap flow and CO2 diffusion using a previously developed microfluidic CFD dewatering model (Pearson et al. 2019a, b).

Specimen preparation

Fifteen-year-old E. nitens was sourced from a single plantation-grown stand from Nelson Forests Limited (formerly Weyerhaeuser Limited) and located in Golden Downs Forest (lat. 41°24′S, long. 172°48′E) in the Nelson region of the South Island of New Zealand. Pith to bark increment cores were removed at a height of 1300 mm above the ground for basic density analysis. One, north-facing, pith-to-bark strip was then removed from each of thirteen randomly selected trees. Each 50 mm longitudinal (l) × 50 mm radial (r) strip was removed at a height of 6300 mm from the base of freshly felled and check-free logs. The specimens were immediately wrapped in plastic before being transported under refrigerated conditions to Scion, Rotorua, New Zealand. The specimen strips were then stored in 98% (w/w) formalin aceto-alcohol for a minimum of one month before sectioning with a sledge microtome. Transverse sections for microscopy were removed at every growth ring (start of earlywood to end of latewood) from four strips and at growth ring numbers (starting from the pith) two, five, ten and the maximum growth ring, for the remaining nine strips.

SEM and confocal microscopy

Confocal fluorescence microscopy and scanning electron microscopy (SEM) measurements were performed at Scion using a Leica TCS NT confocal microscope and a Cambridge Stereoscan 240 SEM. Anatomical measurements, including radial & tangential lumen diameter, vessels per unit area, vessel area, and vessel percent, were measured using scanning electron microscopy. Radial and tangential lumen diameter were measured on 20 vessels using the on-screen cursor system. The number of vessels per unit area was determined from counts on 10 fields of view (at 95× magnification with a field of view of 1.21 mm2). Vessel area and vessel area percent were calculated using the above parameters assuming an elliptical shape. Mean vessel and fibre lumen diameter were calculated by assuming a circular cross-section of the equivalent area to each ellipse. Fibre dimensions were determined from sections stained with safranin using confocal microscopy and digital image analysis, as described by Donaldson & Lausberg (1998), with modifications to magnification (20× objective with 2× digital zoom) and thresholding (mean brightness recalibrated for E. nitens). Sections were mounted in immersion oil after oven drying and confocal fluorescence images were collected using 568 nm excitation and 580–620 nm emission. Fibre dimensions included wall thickness, radial and tangential lumen diameter, coarseness, and external fibre perimeter. Two fields of view were imaged for each sample and 50 fibres were measured from each image. Tension wood was assessed on the same sections using widefield fluorescence to identify gelatinous fibres. An anatomical description was developed using conventional light microscopy. Additional cross-sectional images of tracheid bordered pits in P. radiata and fibre pits in E. nitens were acquired using confocal microscopy and autofluorescence (Leica SP5) using 488 and 561 nm excitation, and 500–550 and 570–700 nm emission.

Data were statistically analysed and averaged across specimen growth rings to give the mean vessel and fibre pore diameter, wall thickness, and vessel lumen area fraction along with basic density and green moisture content (MC).

Supercritical CO2 dewatering

Specimen preparation

Fifteen-year-old E. nitens was sourced from a single plantation-grown stand from Southwood Export Limited, Invercargill, New Zealand and located in Longwoods Forest (lat. 46°14′S, long. 167°59′E) in the Southland region of the South Island of New Zealand. Logistical constraints meant that the E. nitens for dewatering assessment had to be sourced from a different stand to the trees used for anatomical assessment, but average basic density was measured to ensure comparable material was selected. One check-free log, measuring 1200 mm long and taken at a height of 1300 mm above the ground, was removed from each of nine randomly selected trees. After harvesting, each log was immediately end sealed with wax, wrapped in plastic, and transported under refrigerated conditions to Scion, Rotorua, New Zealand, for experimental preparation. Four boards, each measuring 1200 mm × 25 mm × 50 mm in the longitudinal, radial and tangential (l, r, t) directions respectively, were removed from the same relative position of four equal quadrants ( r × t face) of the outer layer of heartwood of each log. Each board was then cut in half lengthwise to yield eight boards measuring 25 mm × 25 mm × 1200 mm. Three clear-wood specimens were then cross-cut from the middle of each board. The specimen in the centre (length: 150 mm) was used for measuring green moisture content and basic density, while the other two (length: 200 mm) were used for dewatering experiments. All specimens were randomised for experimentation to yield a total of 144 matched specimens, measuring 25 mm × 25 mm × 200 mm (r, t, l), for dewatering and distortion assessment. Each specimen was wrapped in plastic and refrigerated at 4°C before being dimensionally measured and weighed before treatment. Basic density and green MC wood quality properties were averaged using SAS statistical software.

Apparatus

Supercritical CO2 wood dewatering was performed in a 0.5 litre treatment plant that was designed to operate at pressures from atmospheric to 30 MPa, and temperatures from 20 to 90°C (Pearson et al. 2019b). Dewatering involved increasing the CO2 pressure from atmospheric to a maximum pressure (P) and holding for a period (t), at a constant temperature (T), before being linearly released with respect to a pressure-release-time (Rt) and holding at atmospheric pressure for a final hold time (fht), if required depending on the experimental design. The first pressure hold-time (t) was treated as a separate variable called the initial-pressure-hold-time ( t i ). The plant was designed to treat one green wood specimen at a time and released sap was collected from the bottom of the plant once per cycle when atmospheric pressure was reached. Specimen weights and dimensions were immediately recorded before dewatering (green state), immediately after dewatering, and after oven drying to give initial and dewatered MCs.

Experimental

Dewatering was performed using two experiments (Table 1). The first design designated as Phase-1 dewatering was a full 34 factorial experiment to measure the effect of P, T, t i , and Rt. Phase-1 dewatering involved cycling the supercritical CO2 pressure between atmospheric and a maximum for a total of ten pressure cycles and used one randomised specimen for each factorial combination. Replicated runs were performed for one combination of the factorial design termed the base-case schedule, with P = 20  MPa, T = 50 C, t i = t = 2  min and Rt = 5 min, using three randomly selected specimens from each tree to determine experimental variability between and within trees. The base-case schedule was previously designed for rapid dewatering of the relatively open porous network of softwood tracheids associated with P. radiata sapwood (Pearson et al. 2019a) and significantly reduced collapse after oven drying when the final dewatered MC was approximately 40% (Dawson & Pearson 2017).

Table 1.
Table 1.

Experimental designs and control variable levels for base-case treatment, and the Phase-1- and Phase-2-dewatering designs.

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10101

The second dewatering experiment called Phase-2 dewatering tested treatments designed to further reduce the MC of E. nitens by focussing on the reduction of fibre moisture without pressure rupture (Schneider et al. 2005; Gething et al. 2013). Phase-2 dewatering required the base-case treatment, described above, to have been performed immediately beforehand and used pressure hold times of t = 7.5, 15, 30, 60 min, at a constant maximum pressure of 20 MPa and temperature of 50°C. This compared to a constant pressure hold time for cycles 2–10 in Phase-1-dewatering of 2 min. The pressure-release-time, Rt, for Phase-2-dewatering was calculated as a function of the pressure hold time t according to Eq. 1, and a final hold time, fht, was calculated using Eq. 2. Phase-2 dewatering was performed for a total of five pressure cycles after the ten base-case pressure cycles had been completed and used 3 randomly selected specimen replicates from one of each of three trees (Nos. 1, 5, 9) for each treatment. The statistical analysis of Phase-2 included base-case results from Phase-1 but only for specimens from the same trees.
(1) Rt = 3.5 + t i / 5
where Rt = pressure release time (min); t i = initial pressure hold-time (min)
(2) fht = 20 Rt
where fht = final hold time (min); Rt = pressure release time.

The weight and dimensions of all dewatered specimens were recorded immediately after treatment and oven drying to calculate MC as a function of dewatering pressure cycle. After oven drying the weights were recorded again and each specimen was coated in paraffin wax for collapse distortion assessment.

Micro-CT X-ray tomography

Micro-CT X-ray tomography was used to confirm the location of water in E. nitens after Phase-1 base-case dewatering by allowing differentiation between morphology of different density. The denser components such as water and wood appeared lighter, compared to the less dense elements such as air in empty vessels, which appeared darker. Two randomly selected E. nitens specimens were selected from the dewatering specimen set from the same tree and tree height. The first specimen was left green whilst the second specimen was dewatered with the base-case treatment schedule. After dewatering both specimens were wrapped in plastic and sealed in refrigerated airtight containers before immediate transportation for tomographic analysis using a Bruker-Micro-CT Skyscan scanner, model 1172, with a 50 mm field of view (cooled CDD coupled by fibre optic to scintillator) and 12-bit, 10 megapixel (4000 × 2300) camera, located at the University of Otago, Dunedin, New Zealand. Specimens were weighed before and after treatment and finally, oven-dried for wood quality assessment.

Before imaging, each specimen was reduced to 60 mm (l) × 25 mm (r) × 25 mm (t) by removing equal portions from each end. Cross-sectional images were rendered as average projections of 50 tomographic slices at locations near the centre of the specimens.

Dewatering modelling

Phase-1 regression modelling

Data from the Phase-1 E. nitens trial were used to develop regression models for predicting the MC at each pressure cycle of the dewatering process across the factorial treatments of the experimental design. The same universal dewatering equation developed for P. radiata (Pearson et al. 2019a) was used. Each measurement of MC was scaled using Eq. 3 to range between zero (fully dewatered), and unity (green). The minimum MC for E. nitens was assumed to be equal to the experimental limit obtained for P. radiata in the earlier study (MCmin = 40.1%; Pearson et al. 2019a). The fibre saturation point (FSP) for the conditions used in this study was calculated to be 29.0% (Ross 2010).
(3) MC sc , i j = ( MC i j MC min ) / ( MC i 0 MC min )
where MCsc,ij = scaled MC for specimen i after j pressure cycles; MCij = MC of specimen i after j pressure cycles (%); MCi0 = initial (green) MC of specimen i (%); MCmin = the minimum MC (%)
The universal dewatering equation (Eq. 4) is a modified form of the Chapman-Richards (CR) growth curve (Richards 1959) using the experimental control variables of greatest significance. It included a slope parameter a, and a shape parameter b, which were expressed as functions of the experimental factors with specific coefficients (Eqs 5 and 6).
(4) MC sc , i j = 1 ( 1 e ( ( a + R i ) × j ) ) b + e i j
where,
(5) a = α 0 ( P / 20 ) α 1
and,
(6) b = β 0 + β 1 ( ln ( t i ) ln ( 2 ) ) + β 2 ( T 50 ) + β 3 ( ln ( Rt ) ln ( 5 ) )
Where MCsc,ij = scaled MC of specimen i after j pressure cycles ( 0 j 10); R i are random effects for each specimen assumed to be independently and identically distributed (iid) normal variables with zero mean; e i j are error terms assumed to be iid normal with zero mean; P = maximum pressure (MPa); T = temperature (°C); t i = initial pressure hold-time (min); Rt = pressure release time (min); α0, α1, β0, β1, β2 and β3 are fitted model coefficients.

Parameter estimates were obtained for E. nitens by fitting the equation as a nonlinear mixed model using the SAS Version 9.4 NLMIXED procedure. Contrasting the parameter estimates for E. nitens with those obtained for P. radiata provided a comparison of the dewatering behaviour of a hardwood with a softwood.

Phase-2 regression modelling

Measurements of MC for each of the first ten pressure cycles (Phase-1 base-case treatment) and the fifteenth cycle (end of Phase-2 treatment) were analysed to test the effect of the Phase-2 treatments on dewatering behaviour. The data used in this analysis included all Phase-1 base-case results for specimens from the same three trees used for the Phase-2 trial in addition to the Phase-2 specimens. The following model was used in the analysis:
(7) MC sc , k i j = 1 ( 1 e ( ( a + R k i ) × j ) ) b + I j T k + e k i j
where MCsc,kij = scaled MC for the ith specimen in treatment k after j pressure cycles; a and b are fitted model parameters; T k is a fitted parameter representing the effect of Phase-2 treatment k, I j is an indicator variable with value 1 when j = 15, and 0 otherwise; R k i is a random specimen effect assumed iid normal with zero mean; e k i j are error terms assumed iid normal with zero mean. The model was fitted using the SAS NLMIXED procedure. The treatment effects T k represent the difference in MC of Phase-2 treatment k in comparison to the Phase-1 base-case treatment extrapolated to fifteen pressure cycles. Differences between Phase-2 treatment effects T k were tested using the least significant difference (LSD) test.

CFD modelling

Phase-1 E. nitens dewatering was predicted using a previously developed two-dimensional (2D) Ansys Fluent 2019 microfluidic CFD dewatering model (Pearson et al. 2019a, b). The CFD model was designed for the prediction of dewatering from open hydrofluidic networks such as capillary tubes, porous ceramics and softwood tracheids, and assumed that the flow of water through the hydrofluidic conducting cells of wood could be compared to flow through capillaries (Zimmermann 1983). The model had previously been used to describe the dewatering of green P. radiata sapwood tracheid cells treated with a schedule of six pressure cycles for ceiling pressures of P = 5, 10, 20 MPa at a constant T = 50 C, t i = t = 2  min and Rt = 5 min. A capillary geometry of equivalent length and diameter to the capillary path length and mean pore size used with P. radiata (Pearson et al. 2019a) was required along with an effective CO2 diffusion coefficient (Deff) to compensate for the presence of porosity and tortuosity which must be found experimentally for complex materials such as wood (Shen & Chen 2007; Cussler 2014). The mean path length was measured perpendicular to the grain direction in the direction of rays and served to act as a characteristic length to define the scale of the system and was required for use with dimensionless number comparisons as with the Péclet number (Pearson et al. 2019b).

A similar CFD modelling procedure was followed for the prediction of E. nitens dewatering. A capillary path length (half the minimum specimen dimension) and mean pore size was obtained from geometric and anatomical measurements and a parametric study was used to evaluate Deff by minimising the error between experimental and predicted dewatering results. However, the CFD model was originally designed to model the dewatering of an open capillary network such as softwood tracheids or hardwood vessels rather than hardwood libriform fibres (Jane 1956). Therefore, for the relatively fast dewatering schedule involving hold times of t i = t = 2  min and release time of Rt = 5 min, the scaled MC was assumed to only apply to vessels. The relative amount, and preference to dewater, of vessels compared to fibres was confirmed from the anatomical and X-ray tomography method described earlier. All other CFD model parameters remained the same for predicting the dewatering of E. nitens compared to P. radiata.

Distortion assessment after oven drying

Free-shrinkage

Twelve control specimens were randomly selected from the dewatering specimen set and solvent dried from green for measurement of free-shrinkage. Solvent drying significantly reduces water surface tension (σ) and therefore distortion after moisture loss can be related to free-shrinkage alone, if unwanted stress gradients are avoided (Hansmann et al. 2002). Each green specimen was weighed, and its breadth, width and length dimensions were measured before being immersed in 96% (w/w) ethanol. All specimens were held at a room temperature of approximately 21°C for three days before the ethanol solution was replaced. After eight solution replacements, the specimens were stored in a vacuum chamber with silica gel desiccant at 0.05 MPa and 21°C before being weighed daily until the maximum percentage change in weight for any specimen was less than 0.25%. After alcohol drying the specimens were re-weighed and their dimensions re-measured before being coated in paraffin wax and external volume measured by water displacement followed by internal area and perimeter measurement measured by image analysis as described below.

Collapse

Collapse was calculated as the normalised post-treatment dimensional change of green wood after subtraction of predicted free shrinkage. Free-shrinkage was subtracted because it is unavoidable when bound water is removed from wood cell walls but other forms of distortion such as collapse may be completely absent. External volumetric specimen collapse was calculated from water volume displacement. After volumetric measurement, each specimen was cross-cut into two specimens measuring 25 mm× 25 mm× 100 mm, and an image was taken of one freshly cut face at a resolution of 400 dpi. From the image, measurements were made to establish the total cross-sectional area of wood and collapse voids respectively. Cross-sectional area measurements were made using Digital Optics V + + image analysis software. Relationships between collapse distortion and MC were statistically analysed using logistic regressions fitted using the SAS LOGISTIC procedure.

Table 2.
Table 2.

Mean E. nitens vessel and fibre diameters, vessel lumen area fractions and fibre and vessel wall thickness for selected growth rings.

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10101

Results

Anatomy

The mean basic density from tree increment cores for the anatomy assessment specimens was 489.5 kg/m3 (standard deviation (SD) 32.1 kg/m3) with a range of 440–539 kg/m3. The mean cell-wall thickness was 1.9 μm (SD 0.33 μm) and the mean lumen diameters were 142.0 μm (SD 30.65 μm) for vessels, and 7.8 μm (SD 1.33 μm) for fibres. The mean vessel and fibre cross-sectional lumen area fractions were 0.100 and 0.585, respectively, with the remaining area representing cell-wall material. The results fall within the general range of values for eucalypts (Wilkes 1988). Anatomical measurements are summarised in Table 2. The following trends were observed:

  1. Vessel lumen area, and radial and tangential vessel lumen diameter tended to increase from pith to bark.
  2. Vessels per mm2 tended to decline from pith to bark.
  3. Vessel eccentricity showed little variation.
  4. Area percentage of vessels showed variation from ring to ring in some trees with no trend from pith to bark.
  5. Fibre wall thickness varied in proportion to ring width and proportion of latewood with no trend from pith to bark.
  6. Radial and tangential lumen diameter of fibres varied from ring to ring with no trend from pith to bark.

Between-tree variation was most apparent in factors related to wood density including vessel lumen area, radial and tangential vessel lumen diameter, the area percentage of vessels, fibre wall thickness, and radial and tangential fibre lumen diameter.

Anatomical features that are likely to affect water movement during wood drying include the presence of tyloses in heartwood vessels, predominantly solitary vessels, the presence of vasicentric tracheids, and occlusion of all pits with extractives (Fig. 1). While free water in the hydrofluidic diffuse-porous vessel network may be relatively mobile, free water in fibres diffuses relatively slowly through cell walls during drying due to the small size of fibre pits and their occlusion with extractives.

Fig. 1.
Fig. 1.

(A–F) Anatomy and pitting of P. radiata and E. nitens using confocal fluorescence. (A) P. radiata sapwood/earlywood. (B) P. radiata aspirated bordered pit (arrow) (C) E. nitens sapwood. (D) E. nitens intervascular pits (arrows), V = vessel. (E) E. nitens heartwood, T = tyloses. (F) E. nitens fibre-fibre pits (arrows), F = fibre, R = ray. Green fluorescence originates from lignin while crimson fluorescence originates from extractives. Scale bars (A, C, E) = 100 μm, (B, D, F) = 10 μm.

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10101

Experimental

Phase-1 dewatering

The mean basic density and green MC for all dewatering specimens was 386.6 kg/m3 (SD 21.7 kg/m3 with range 340–513 kg/m3) and 166.2% (SD 15.9% with range 121–199%) respectively. The mean basic density was less than that recorded for the anatomical specimens by a difference of 102.9 kg/m3. However, both sets of specimens were considered to be experimentally comparable because all specimens used in this study lay within the general range reported previously for E. nitens (McKimm et al. 1988; Milne 1991; Lausberg et al. 1995; Shelbourne et al. 2002). A general reduction in density has been observed for more southern grown plantation wood and attributed to increased rainfall (Shelbourne et al. 2002). Therefore, it is likely that wood sourced from Golden Downs, which has a relatively low rainfall exhibited greater than average basic density compared to wood sourced from Southland which has a higher relative rainfall.

Phase-1 dewatering results were used to develop a regression model for predicting MC (Eq. 4) (Table 3). The same form of the CR growth curve used for P. radiata (Pearson et al. 2019a) (coefficient of determination (R2) = 84.3%) applied to E. nitens (R2 = 59.0%). This showed that the dewatering behaviour of E. nitens was more variable than P. radiata and correlated strongly to tree source ( P = 0.085) but not to the relative circumferential log position or height at which specimens were removed ( P = 1).

Table 3.
Table 3.

Parameter estimates, standard errors, t-values and p-values for nonlinear mixed models describing P. radiata and E. nitens dewatering using the universal dewatering model (Eq. 4).

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10101

The amount of moisture remaining in E. nitens after 10 Phase-1 dewatering pressure cycles, was significantly greater than that for P. radiata (Pearson et al. 2019a) for all equivalent dewatering schedules. This was primarily attributed to the retention of lumen water in fibres which were regarded as being sealed because the pit membranes were occluded with resin (Fig. 1D, F), thus decreasing their permeability, especially to moisture movement during dewatering.

The final predicted MC for E. nitens after the Phase-1 base-case treatment schedule was MCsc = 0.700 (CR model) compared to the previously reported value for P. radiata of MCsc = 0.007 (CR model) (Pearson et al. 2019a). This showed that a significant amount of moisture remained in E. nitens even after dewatering with the most efficient softwood treatment schedule for P. radiata which had 96.2% of the water removed after only seven pressure cycles.

Fig. 2.
Fig. 2.

Transverse micro-CT images of the central portion of 25 mm (r) × 25 mm (t) E. nitens specimens. The lighter areas relate to areas of increased density. (A) green E. nitens (MC = 178.8%). (B) E. nitens after dewatering with base-case schedule (green MC = 176.2%; dewatered MC = 132.4% and dewatered MCsc = 0.678). Scale bars = 4 mm.

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10101

The location of the remaining moisture in E. nitens after dewatering with the base-case schedule was confirmed using micro-CT (Fig. 2). The specimen in Fig. 2B was dewatered from an initial MC of 176.2% (MCsc = 1) to a final MC of 132.4% (MCsc = 0.678) and the remaining moisture was assumed by mass balance to be located within the fibres since the vessels were empty. This compared to a matched green specimen with an MC of 178.8% (Fig. 2A) which showed almost all vessels to still be full of water. The total number of vessels in Fig. 2B, counted by area extrapolation, was similar to the expected number calculated from anatomical vessel lumen diameter and area fraction and differed by less than 4%. Fig. 2B shows that the relatively open hydrofluidic network of E. nitens vessels could be effectively dewatered in a similar manner to softwood tracheids, but that removal of fibre moisture requires fluid transfer across the much less permeable fibre cell walls.

The single greatest Phase-1 dewatering E. nitens moisture reduction of MCsc = 0.341 was achieved over ten pressure cycles with the experimental control variables all set to their maximum values. However, the treatment schedule took over 12 hours to complete compared to a minimum MCsc of 0.007 for P. radiata which took approximately one hour to complete after six pressure cycles. Therefore, rapid dewatering of E. nitens was best achieved by a schedule to firstly dewater the vessels using a Phase-1 base-case approach, followed immediately by Phase-2 dewatering designed to access moisture sealed within the fibres.

Phase-2 dewatering

Phase-2 dewatering treatments were designed to investigate moisture reduction from the sealed fibre lumens of E. nitens that had been quantified anatomically and observed to contain residual moisture after Phase-1 vessel dewatering. The fibre lumen was dewatered by firstly allowing CO2 gas to diffuse across the fibre cell walls into the fibres, followed secondly by allowing moisture to diffuse across the cell walls out of the fibres during pressure release. Two control variables, pressure and time, were varied to reduce fibre lumen moisture without cell rupture from excess pressure differential by reducing the magnitude of dynamic pressure change. This meant that CO2 and moisture could pass through the cell walls under a combination of bulk flow and diffusion (Siau 1995).

Fig. 3.
Fig. 3.

Mean experimental scaled MC as a function of pressure cycle number for matched Phase-1 base-case, and all phase-2, specimens. The dotted line denotes Phase-1 CR model for specific matched specimens. Error bars denote 95% CI.

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10101

Mean Phase-2 dewatering MC results are shown in Fig. 3 and indicate significant reductions in MCsc compared to Phase-1 dewatering extrapolated to fifteen pressure cycles using specific matched specimens. The final minimum Phase-2 MCsc was 0.243 for a pressure hold time of t = 30 min. The LSD test revealed no significant difference for MCsc between t = 30  min or 60 min showing a minimum moisture limit may have been reached (Table 4). All Phase-2 final mean scaled MCs were lower, and took less treatment time, compared to even the best performing Phase-1 treatment schedule. Longer Phase-1 treatment schedules may have included increased dewatering from increased diffusion times, and drying from CO2 gas exchange, but the magnitude of this phenomenon is not known.

CFD modelling

Based upon the anatomical imaging results, E. nitens vessel dewatering was predicted using the CFD model after calibration using a mean internal capillary diameter of 142 μm and scaling the experimental CR model to only include the vessel lumen water. This was achieved by assuming the relative amounts of vessel lumens and cell walls from the anatomical assessment were fully saturated and the fibre lumens contained only enough moisture (approx. 75% full) for the green MC and basic density to equal the experimental mean of MCmin = 139.8%. After scaling, the mean experimental MC was predicted to reduce from 166.2% (MCsc = 1) to 139.8% (MCsc = 0.001). The base-case dewatering schedule was found to be effectively complete after seven pressure cycles (Fig. 4B) but still dewatered at a slower rate than P. radiata (Fig. 4A) probably due to increased pit resistance, especially between vasicentric tracheids and vessel ends (Zimmermann 1983).

Table 4.
Table 4.

Scaled moisture content statistical analysis results for Phase-2 dewatering at 15 pressure cycles.

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10101

Fig. 4.
Fig. 4.

Experimental (CR model) and predicted (CFD model) scaled dewatering MC as a function of pressure cycle number for (A) P. radiata (Pearson et al. 2019a), and (B), E. nitens, at maximum treatment pressures P = 5, 10, 20 MPa and a constant temperature T = 50 C, initial and subsequent pressure hold times t i = t = 2  min and pressure release time Rt = 5 min.

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10101

As with P. radiata, the effect of tortuosity and porosity on the effective CO2 diffusion coefficient in E. nitens decreased as conditions approached supercritical. This meant that tortuosity and porosity of morphological pathways for gas and fluid flow became more important at lower pressures and temperatures when the cell-wall material offered greater diffusive resistance to CO2 diffusion. The behaviour of the effective (Deff), and normalised diffusion coefficients (Deff/D) of CO2 in the lumen moisture of E. nitens followed an almost identical trend to previous results for P. radiata and reduced from a value of Deff = 3.21 × 10−9 m2/s (Deff/ D = unity) at 20 MPa to Deff = 4.59 × 10−11 m2/s (Deff/ D = 0.014) at 5 MPa (Fig. 5). This showed that the tortuosity and porosity of the hydrofluidic pathways of E. nitens, as described by a lumped parameter approach in the effective diffusion coefficient, was similar to P. radiata, especially under supercritical CO2 conditions (Pearson et al. 2019a).

Fig. 5.
Fig. 5.

Effective (Deff), and normalised (Deff/D), diffusion coefficient for CO2 in wood moisture as a function of ceiling pressure P, for E. nitens compared to previously published data for P. radiata and a straight capillary tube (Pearson et al. 2019a).

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10101

Collapse distortion

The mean volumetric free-shrinkage for the twelve alcohol-dried specimens was 9.8% (SD 1.44%), with mean central cross-sectional area shrinkage of 8.6% (SD 0.96%) and mean cross-sectional length shrinkage of 4.4% (SD 0.50%).

Internal and external collapse as a function of dewatered MC for all Phase-1 and -2 dewatering specimens, after oven drying and pure shrinkage compensation, are shown in Fig. 6. Logistic regression of the non-linear relationships revealed strong negative correlations between both external volume (deviance explained = 78.5) (Fig. 6A), and internal cross-sectional area (deviance explained = 80.5) (Fig. 6B) as functions of dewatered MC.

Fig. 6.
Fig. 6.

External volume (A), cross-sectional wood area (B) and cross-sectional void area (C) collapse ratios compared to pure shrinkage specimen dimensions, after oven drying and as a function of dewatered moisture content for all Phase-1 and -2 dewatered specimens. Solid line equals regression curve of best fit. Dotted lines equal 95% CI of the regression curve.

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10101

The mean external volume of E. nitens was reduced by 33% due to collapse after oven drying from a green MC of 180%, but the volume reduction was only 3.9% after oven drying from a dewatered MC of 70% (Fig. 6A). Similarly, the cross-sectional wood area (excluding voids) was reduced after oven drying from a green MC of 180% by 40% compared to 3.9% when oven-dried from 70% MC. This meant that a reduction in the dewatered MC of E. nitens from green to 70% eliminated approximately 90% of both oven-dried volumetric and internal wood area collapse (Fig. 6A, B).

There was a slight positive correlation between the internal collapse void area and dewatered MC (deviance explained = 13.5%) (Fig. 6C). This indicates that the greater the amount of external volumetric and cross-sectional wood collapse, the greater the amount of internal collapse voids which cannot be visualised externally. Conversely, specimens with minimal external collapse after drying are likely to have minimal internal voids. This can be seen in Fig. 7 which shows cross-sections of the greatest-, and lowest-, twelve dewatered specimens after ranking for the magnitude of volumetric change after oven drying.

Fig. 7.
Fig. 7.

Middle cross-sections of (A) greatest-12, and (B) lowest-12 external volume change after dewatering followed by oven drying. The relative log source of each specimen is shown with Phase-1 dewatered specimens in blue and Phase-2 dewatered specimens in red. Scale: each set (B) specimen approx. 25 mm (t) × 25 mm (r).

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10101

Specimens with the least collapse generally exhibited internal checks in the form of narrow fracture lines following a straight path parallel to the radial growth direction (Fig. 7B). The checking pattern was attributed to anisotropic shrinkage from oven drying because the tangential wood growth direction experiences greater anisotropic shrinkage compared to the longitudinal or radial directions (Bodig & Jayne 1982; Blakemore 2011).

Discussion

The amount of moisture remaining in E. nitens heartwood after dewatering was significantly greater than that for P. radiata sapwood (Pearson et al. 2019a) due to anatomical differences between the two species. The main anatomical difference was the presence of hardwood fibres in E. nitens which were assumed to be sealed because the fibre pit apertures and pit membranes were occluded with extractives (arrows in Fig. 1F) which likely decreased their permeability (Siau 1995; Carlquist 2001; Donaldson et al. 2018). Even though fibres can contain moisture, their main function is structural support and they are not naturally part of the hydrofluidic transport system of hardwoods (Skaar 1988; Siau 1995; Carlquist 2001). However, the permeability of fibres is difficult to isolate and measure (Zauer et al. 2014; Donaldson et al. 2018). The principal modes of moisture transport in wood are bulk flow, as described by Darcy’s law, and diffusion as described by Fick’s law (Siau 1995). Darcy’s law permeability measurements have commonly focussed on bulk flow of sap in green wood, which follows the natural hydrofluidic pathways of a living tree such as occurs in hardwood vessels and rays (Siau 1995; Perré & Turner 2001; Hansmann et al. 2002). For supercritical CO2 dewatering of hardwood fibres, flow normal to the cell-wall surface is of primary importance and permeability models typically assume negligible cell-wall permeabilities (Siau 1995). Palin & Petty (1981) measured the permeability of Picea abies ((L.) H.Karst.) heartwood by applying an osmotic gradient after the lumens were filled with paraffin wax and found the cell-wall permeabilities parallel to the surface to be so low that they were orders of magnitude lower than those measured for untreated wood. Similarly, Stamm (1960) found the water diffusion coefficient in the longitudinal direction to be two to three times that in the transverse direction in several wood species. Estimation of moisture transfer through a fibre pit membrane is complex because it involves both diffusion and bulk flow with a pressure differential from CO2 gas as a driving force. For this reason, empirical studies to measure the overall mass transfer coefficient and effective diffusion coefficient are recommended (Cussler 2014).

A second anatomical difference between hardwoods and softwoods was the presence of hardwood tyloses which generally increase in number and size for older heartwood and may affect dewatering efficiency due to vessel occlusion (Siau 1995; Carlquist 2001). However, their presence is physiological rather than phylogenetic in nature and the prediction of their occurrence is imprecise (Carlquist 2001). Even with tyloses present, it is likely that the open hydrofluidic vascular network of vessels and vasicentric tracheids (Fig. 1D) generally remain open with pit membranes (arrows in Fig. 1D and 1F) still acting as a major source of flow resistance (Carlquist 2001).

Anatomical analysis was pivotal in the development of a two-stage experimental treatment to dewater the moisture contained in the fibres of E. nitens. Dewatering efficiency of the vessel network of E. nitens was compared to P. radiata using a CFD dewatering model which confirmed a similar hydrofluidic resistance in terms of lumped-parameter tortuosity and porosity.

In this study, the mean volumetric collapse after dewatering and oven drying (Fig. 6A) was 5% when the initial dewatered MC was 86%, but this increased to 22% at an MC of 155%. By comparison, Chafe (1985) observed a relatively reduced level of volumetric collapse (10%) for E. nitens at MC = 155%, after oven drying from green followed by steam reconditioning to 16–20% MC. Although pure shrinkage was not included in Chafe’s calculation of collapse, the primary reasons for less collapse were attributed to different specimen volume, shape and grain direction. Generally, reconditioning is thought to recover only 50% of collapse (Yang & Liu 2018).

Specimens with the greatest amount of collapse from this study typically displayed an hourglass shape along the 200 mm l-axis, with greater collapse occurring in the middle of each specimen compared to the ends. This was attributed to preferential moisture loss near vessel termination ends (rt faces) compared to in the cross-grain direction (lr and lt faces) through rays (Carlquist 2001; Pfautsch et al. 2015). However, collapse is a complex phenomenon and comparative studies are required if different geometries are to be fully understood.

Narrow and radially oriented checks occurred within oven-dried specimens that exhibited the least collapse. The checks were caused by cross-grain anisotropic distortion and were primarily associated with wood rays because they are a structural point of weakness in bulk wood when stressed in tension perpendicular to their long axis (Kauman 1964a, b; Bodner et al. 1997). Ray separation helps to relieve cross-grain stress after fracture initiation between adjacent areas of cells with differing strength such as earlywood latewood bands, cellular junctions with rays, and around larger and relatively weaker vessels in E. nitens (Boutelje 1962; Simpson 1991; Bodner et al. 1997; Ross 2010; Donaldson 2011; Valenzuela et al. 2012; Rebolledo et al. 2013). This phenomenon was observed by light microscopy on several selected specimens where matching ray segments frequently occurred on opposite sides of the internal surfaces of both checks and collapse voids.

Checks were not observed in specimens with elevated amounts of collapse because the relative strain type was either insignificant or was not present after stress relief had taken place. However, distortion during drying is complex and dependent on the type of drying schedule applied and the morphology of the material involved. Whether stress relief presents as collapse, checking, or a combination depends on the mechanisms of anisotropic distortion during drying (Boutelje 1962; Bodig & Jayne 1982; Redman et al. 2016). Distortion can be related to many variables such as wood species, wood quality, cellular morphology and genetics as well as wood history, and the drying schedule employed. Studies to relate micromechanical behaviour with these variables invariably produce results that are either specific to a study or involve decreased accuracy when applied to a wider range of conditions (Booker 1994; Donaldson 1997; Danvind & Synnergren 2001; Murata et al. 2001; Kang & Lee 2002; Kube & Raymond 2005; Wu et al. 2006; Badel & Perré 2007; Hamilton et al. 2009). Furthermore, distortion strains can be separated into several types such as free-shrinkage, elastic, creep, mechanosorptive and thermal. The relative magnitudes of each strain type and their incidence history can have a dramatic effect on fracture locations and stress-strain patterns. Comparisons between studies is difficult if the different strains cannot be successfully isolated (Pearson et al. 2012, 2015).

Oven drying at low dew points represents an extreme form of drying that is usually associated with defects, such as radial checking, and is not a typical industrial drying regime (Simpson 1991). It represents the fastest possible drying schedule and was used in this study to induce measurable changes. Milder drying schedules can reduce checking, especially in difficult-to-dry species (Kauman 1964a, b; Innes 1996; Salin 2010; Wu et al. 2010; Yang & Liu 2018). However, dewatered wood is unique because dewatering disrupts the traditional Fick’s law drying front which causes reverse case hardening (Siau 1995; Blakemore 2011). The process creates a matrix of empty lumen drying pathways of least resistance early in the dewatering process. A similar theory has been proposed by Salin (2006a, b) who acknowledged the capillary network of wood and modelled drying based upon percolation theory. Based on this work, a completely novel drying schedule is recommended for minimisation of checking, and to limit the strain type to the equivalent of pure shrinkage which leads to the least distortion possible after green cells have lost their bound water.

The dewatering process is robust for E. nitens because 90% of collapse distortion can be mitigated after oven drying, provided supercritical CO2 dewatering is employed to sufficiently reduce the MC beforehand. The methods available to mitigate collapse described by Blakemore & Northway (2009), including supercritical CO2 dewatering, have all been described as economically expensive. However, the degree to which the dewatering process can mitigate collapse in E. nitens now makes the process more attractive for use with hardwoods as well as softwoods. Moreover, the general dewatering process (Dawson & Pearson 2017) has the unique possibility of combining collapse mitigation with supercritical extraction of desirable sap components, a post-dewatering drying schedule aimed to also mitigate checking, preservative treatment, and mechanical forming all within the same plant.

Conclusions

This study has found that collapse of the hardwood species E. nitens during drying can be significantly reduced to a volumetric change of 3.9% if the MC is reduced to 70% by supercritical CO2 dewatering prior to oven drying. This equates to a reduction in collapse of 90% compared with oven drying with no dewatering from an initial MC of up to approximately 190% (Fig. 6). A two-phase dewatering schedule is required to achieve an MC of 70%. The first phase typically dewaters E. nitens to an MC of 140.5% by acting on the open hydrofluidic porous network of vessels. The second stage reduces the MC to 70% by acting on the closed fibres.

Predictive CFD dewatering modelling revealed that the anatomical lumped tortuosity and porosity parameter of the open hydrofluidic vessel network of E. nitens was similar to that for P. radiata, even though the open vessel pathways of E. nitens were restricted to only 10.0% of the total transverse cross-sectional area.

The supercritical CO2 dewatering process developed to mitigate collapse for E. nitens would apply to other anatomically similar hardwoods and could be combined in the same plant with extraction of desirable sap components, a post dewatering drying schedule aimed to mitigate checking, preservative treatment, and mechanical forming, to improve the process economics.

*

Corresponding author; email: hamish.pearson@scionresearch.com

Acknowledgements

The authors would like to acknowledge the following: Ian McElroy (Scion) for engineering support; Steve MacIntosh and Graeme Manley for providing E. nitens; Andrew McNaughton (University of Otago) for performing the micro-CT scans. Experimental dewatering assistance was provided at Scion by Bruce Davy (Phase-1 and -2) and Bernard Dawson (Phase-1). The authors are also grateful to Rosie Sargent, Alankar Vaidya, Doug Gaunt and Elspeth MacRae (Scion) for reviewing the paper. The New Zealand Ministry of Business, Innovation and Employment provided funding for this study via the Strategic Science Investment Fund (contract No. C04X1703).

References

  • Aggarwal S, Johnson S, Hakovirta M, Sastri B, Banerjee S. 2019. Removal of water and extractives from softwood with supercritical carbon dioxide. Ind. Eng. Chem. Res. 58: 31703174. DOI: 10.1021/acs.iecr.8b05939.

    • Search Google Scholar
    • Export Citation
  • Asafu-Adjaye O, Via B, Sastri B, Banerjee S. 2021. Mechanism of dewatering porous structures with supercritical carbon dioxide. Case Stud. Chem. Environ. Eng. 4: 13. DOI: 10.1016/j.cscee.2021.100128.

    • Search Google Scholar
    • Export Citation
  • Badel E, Perré P. 2007. The shrinkage of oak predicted from its anatomical pattern: validation of a cognitive model. Trees 21: 111120. DOI: 10.1007/s00468-006-0105-z.

    • Search Google Scholar
    • Export Citation
  • Bariska M. 1992. Collapse phenomena in eucalypts. Wood Sci. Technol. 26: 165179.

  • Blackburn D, Hamilton M, Harwood C, Innes T, Potts B, Williams D. 2010. Stiffness and checking of Eucalyptus nitens sawn boards: genetic variation and potential for genetic improvement. Tree Genet. Genom. 6: 757765. DOI: 10.1007/s11295-010-0289-7.