Influence of habitat, density, lignin structure, and extraction treatment on thermal-softening properties of water-swollen wood: a study of 87 wood specimens

2.1 Pretreatment of samples

A total of 15 softwood and 72 hardwood specimens were obtained from the xylarium of the Forestry and Forest Products Research Institute (TWTw). The annual rainfall data were obtained from the Japan Meteorological Agency (https://www.data.jma.go.jp/gmd/risk/obsdl/) and the World Weather Online (https://www.worldweatheronline.com/), based on the registered latitude and longitude information. For specimens without latitude information, the latitude data was estimated from the name of the registered region or representative point in the country (Table 1).

For DMA measurement, each wood sample was cut in the longitudinal direction into an end-grain plate (1 mm thickness). Subsequently, the strip-shaped DMA samples (3.2 mm (tangential) × 30 mm (radial)) and square-shaped weight measurement samples (15 mm (tangential) × 15 mm (radial)) were cut from the plate (Figure 1). Half of the DMA samples were extracted for 6 h with ethanol:benzene (1:2, v/v) in a Soxhlet extractor to remove the resin, fat, wax, and soluble tannin. The extracted DMA samples were air-dried at room temperature and freeze-dried for 6 h to completely volatilize the liquid. The weight measurement samples were first treated in the same manner as the extracted DMA samples. Next, they were boiled for 10 min and cooled naturally to room temperature in water. This process was performed two times according to the temperature rise and fall program of the DMA samples. The weight loss rate due to the extraction treatment was calculated by weighing the oven-dried (105 °C) samples (Figure 2). The density of all the samples was obtained from the dimensions and weight of the oven-dried weight measurement samples before extraction. The weight loss due to extraction and density were calculated from the average value measured from two or three weight measurement samples that could be taken from each specimen.

Figure 1:

Sample dimensions used for measurements.

Figure 2:

Sample preparation procedures and temperature program for dynamic viscoelastic measurement.

2.2 Dynamic viscoelasticity measurement

A forced-vibration dynamic viscoelastometer (DMA 42E-SF 28 Artemis, NETZSCH Japan K.K., Kanagawa, Japan) was used to measure the temperature-dependent, dynamic viscoelastic properties of the unextracted and extracted water-swollen samples. The dynamic viscoelastic properties of wood changed due to the drying history or cooling rate, even for the same samples (Furuta et al. 2001, 2008a, b; Kojiro et al. 2008). Therefore, to unify the heating history immediately before viscoelastic measurement for all the specimens, this study recorded the data in the second temperature rise after cooling them from 95 °C to 5 °C at 1 °C/min (Figure 2). The data for temperatures up to 98 °C were obtained originally, which is very close to the boiling point of water. Since some samples contained noise in the temperature range higher than 95 °C, DMA data up to 95 °C, which could be stably measured, was analyzed to compare the results. During the viscoelastic measurement, the span was 10 mm in radial direction (Figure 1) and a 0.1 Hz sine wave with a tension displacement amplitude of 5 µm was applied to the sample. In the preliminary experiments conducted on several wood samples, the offset loads in the measured temperature range, when the displacement amplitude was controlled at 5 µm, were within the linear region of the stress-strain line. DMA measurement was performed one time for each sample because the preliminary experiments confirmed that the same temperature program would produce nearly identical measurement results. As an example of the measured DMA data, the results of the temperature dependence of tan δ before and after extraction for softwood and hardwood samples are shown in Figure 3. The peak top of tan δ was read as the peak temperature.

Figure 3:

Examples of DMA data for untreated and extracted samples.

2.3 Estimation of the syringyl ratio

The extracted DMA samples, which were used for the dynamic viscoelastic measurements, were analyzed using an IR spectrometer (FT/IR-4700, JASCO Corporation, Tokyo, Japan). Fine wood meal was prepared from the middle length portion of dried DMA samples, using a glass file, mixed at 1 % with KBr and pressed into a disc. Each spectrum was based on 64 scans with a wavenumber range from 4000 to 500 cm⁻¹. Following Huang et al. (2012), the areas under the IR peak at 1595 cm⁻¹ (range: 1605–1574 cm⁻¹, A ₁₅₉₅) and 1509 cm⁻¹ (range: 1535–1492 cm⁻¹, A ₁₅₀₉) were measured. Huang et al. (2012) presented a linear regression equation with a correlation coefficient of 0.98 to describe the relationship between the syringyl ratio (syringyl/(syringyl+guaiacyl)) of lignin aromatics, measured by alkaline nitrobenzene oxidation analysis, and log (peak area ratio of A ₁₅₉₅/A ₁₅₀₉). In the present study, the same regression equation was used to determine the syringyl ratio of all the specimens from their IR spectra. The peak areas were measured three times for each sample, and after confirming that there were no outliers, area ratios were calculated from the data that showed the clearest spectrum with the least noise.

3.1 Relationship between thermal-softening temperature and the growing environment of the wood

In this study, the peak temperature of tan δ is defined as the thermal-softening temperature of the sample. In this paper, the peak temperatures of tan δ were read for samples for which the peak was observed in the range of 95 °C. Samples for which no peak was observed at 95 °C are indicated as 95+ in Table 1, and the legend is marked in red in the figure. The tan δ values for the 95+ samples were read at 95 °C and indicated in Table 1. Figure 4 shows the relationship between the peak temperature of and the value of tan δ. The peak temperatures of tan δ were slightly different between the extracted and unextracted samples (Figure 4a and b, respectively). Six softwood samples were identified in which the extraction process changed the peak temperature to 95 °C or higher. Table 1 shows that the peak temperature of tan δ in softwoods changed to higher temperatures after extraction. However, some hardwood samples showed a decrease in the peak temperature of tan δ due to extraction. Although the thermal-softening temperatures changed by a maximum of 6 °C due to extraction, the temperature range and the direction of increase or decrease varied among individual samples. No uniform trend was observed among the wood species and no correlation was observed between the peak temperature and the value of tan δ in Figure 4. The peak temperature of tan δ is generally lower for hardwood and higher for softwood, as reported in previous studies (Furuta et al. 2001, 2008a, b). In hardwood species, the peak temperature of tan δ in the radial-, ring-, and non-porous samples was in a temperature range from 70 to 80 °C; however, the temperature range was wider for diffuse-porous samples. The two diffuse-porous samples (both were Ulin (Eusideroxylon zwageri), which is a tropical hardwood) showed higher tan δ values than the others. Table 1 lists the peak temperature of tan δ for unextracted specimens, showing some variation within the same wood species.

Figure 4:

Relationship between loss tangent (tan δ) and peak temperature of tan δ in water-saturated radial samples measured at 0.1 Hz. (a) Untreated samples and (b) extracted samples. The symbol for hardwood species indicates the vessel arrangement type. “Non” = vesselless. Samples collected from tropical climates are presented in green. Samples for which the peak temperature of tan δ could not be observed by 95 °C are presented in red.

Previous studies have reported that among hardwood species, tropical species show higher thermal-softening temperatures (Furuta et al. 2010) and are rich in guaiacyl units (Wu et al. 1992). If both the physical properties and chemical structures observed in tropical wood species are determined by their habitat, the regional climate may exert an influence on the thermal-softening temperature. To confirm this possibility, the relationship between the peak temperature of tan δ in untreated specimens and the latitude and annual rainfall at the collection site is plotted in Figure 5.

Figure 5:

Relationship between habitat and peak temperature of tan δ of water-saturated, untreated radial specimens measured at 0.1 Hz. The symbol for hardwood species indicates the vessel arrangement type. “Non” = vesselless. Samples collected from tropical climates are presented in green. Samples for which the peak temperature of tan δ could not be observed by 95 °C are presented in red.

The softwood specimens showed high peak temperatures of tan δ, regardless of latitude. Among the hardwood specimens, the radial-porous and diffuse-porous wood cover a wide latitude range. Certain diffuse-porous specimens collected from near the equator exhibited high peak temperatures compared to softwood. This distribution of the peak temperature of tan δ was similar to the high peak temperatures of tan δ previously observed in tropical hardwood species (Furuta et al. 2001, 2008a, b). However, this study suggests that the latitude does not necessarily correspond to the thermal-softening temperature among various wood species.

Specifically, in softwood and hardwood except diffuse-porous wood, no relationship was observed between the annual rainfall and peak temperature of tan δ (Figure 5b). Among the diffuse-porous specimens, there was only a weak positive correlation between the two parameters. However, this correlation could be enhanced in certain tropical diffuse-porous wood, because many areas with high rainfall are located near the equator (see the clustering of diffuse-porous specimens near 0° latitude in Figure 5a). Overall, neither the latitude nor annual precipitation were related to the thermal-softening temperatures of wood in this study.

3.2 Relationship between thermal-softening property and density

In this section, the thermal-softening properties of extracted samples are presented to discuss the relationship between the wood cell wall constituents and the thermal-softening properties. Figure 6a–c plot the storage elastic modulus (E′) at 30 °C, loss elastic modulus (E″) at 30 °C, and peak temperatures of tan δ versus the wood density, respectively. A positive correlation was observed between E′ and density in softwood, ring-porous hardwood, non-porous hardwood, and diffuse-porous hardwood, all following the same linear relationship. However, radial-porous wood was distributed at slightly lower locations and did not display a clear positive correlation with density. The relationship between E″ and density showed a positive correlation when the density was 1000 kg/m³ or less. The distribution patterns of E″ and E′ were nearly the same. Figure 6a and b presents three Ulin specimens with the highest densities (1000 kg/m³ or more). Their E′ and E″ values deviate from the linear distribution of the other specimens. While E′ and E″ show a positive correlation with wood density, the peak temperature of tan δ shows no such relationship (Figure 6c). Thus, the wood species exhibited a wide range of thermal-softening temperatures, regardless of their density.

Figure 6:

Relationship between data from dynamic viscoelastic measurements at 0.1 Hz and the density of extracted radial specimens in water-saturated conditions. (a) Storage elastic modulus (E′) at 30 °C, (b) loss elastic modulus (E″) at 30 °C, and (c) peak temperatures of tan δ. The symbol for hardwood species indicates the vessel arrangement type. “Non” = vesselless. Samples collected from tropical climates are presented in green. Samples for which the peak temperature of tan δ could not be observed by 95 °C are presented in red.

3.3 Relationship between thermal-softening property and lignin structure

To investigate the relationship between thermal-softening temperature and lignin structure, the syringyl ratio of lignin aromatics was estimated from the peak area ratio of the IR spectrum. Since the sample used for the IR measurement had undergone a heating history during DMA measurement, there was a possibility that the syringyl ratio would be slightly different from that of the untreated sample. However, this experiment prioritized the comparison of the results with the same sample.

Understanding the partial structure of lignin is useful for inferring its structural characteristics. The β-O-4 and phenyl-coumarin structures are the specific partial structures responsible for lignin’s linear chain, while the biphenyl (5-5 bond) and diaryl ether structures (4-O-5 structure) are responsible for its branch chain. The composition ratio of the partial structures changes according to the aromatic nucleus during lignin production. The main reason for this is that the aromatic nuclei of guaiacyl can be coupled at the 5-position to form fused structures, such as β-5, 5-5, and 4-O-5, whereas the aromatic nuclei of syringyl cannot be coupled alone at the 5-position due to the methoxy group, which acts as the substitute (Stewart et al. 2009). Akiyama et al. (2005) investigated the relationship between the partial structure of lignin and the syringyl ratio of lignin aromatics in various wood species and found that the biphenyl structure content decreased quadratically with an increase in the syringyl ratio. Based on previous research and this study’s results, it can be concluded that wood species with a low syringyl ratio have a dense lignin structure, with several condensed structures derived from the guaiacyl unit, whereas wood species with a high syringyl ratio have a sparse lignin structure, with several linear partial structures.

In Figure 7, the syringyl ratio is distributed in a wide range from 0 to 0.8 (approximately), and almost 0 in softwood and tropical hardwood species. Since softwood has no syringyl aromatic nuclei, the syringyl ratio should be exactly 0. However, the syringyl ratio was estimated using a calibration curve with an error of about ±0.2. Tropical hardwood samples showed a wide range of syringyl ratios, indicating that these samples are not necessarily rich in the guaiacyl unit. Nevertheless, several samples showed particularly low syringyl ratios. Hardwoods with syringyl ratios near 0 possessed a high peak temperature of tan δ compared to softwoods. Thus, the syringyl ratio and peak temperature of tan δ display a negative correlation. Based on the interpretation of the partial structure of lignin, the peak temperature of tan δ is considered to be higher for wood species with denser lignin structures and lower for those with sparser lignin structures. It is known that in polymers, the glass transition temperature (T _g) shifts to higher values as the crosslink density increases (Nielsen 1962). In wood, the peak temperature of tan δ, derived from lignin, could similarly depend on the lignin’s condensation structure in each wood species. Although this study’s results were obtained by using the calibration curves of IR spectra, they supported the correlation shown by Olson and Salmén (1997) between thermal-softening temperature and methoxyl group content.

Figure 7:

Relationship between peak temperature of tan δ measured at 0.1 Hz and syringyl ratio of extracted radial specimens. The symbol for hardwood species indicates the vessel arrangement type. “Non” = vesselless. Samples collected from tropical climates are presented in green. Samples for which the peak temperature of tan δ could not be observed by 95 °C are presented in red.