1. Introduction

Wood is a natural material widely used for its versatility and strength in construction. There are significant variations in wood properties (e.g., density, moisture content, grain angle, crystallinity and micro fibril angle) between species and even among the same species. From the viewpoint of quality assurance in industrial processes, nondestructive monitoring and control of the physical, mechanical and chemical properties of wood are strongly desired.

Interactions with electromagnetic radiation at various wavelengths are preferable for the rapid, non-destructive and non-contact monitoring of wood quality. X-ray densitometer is a reliable method for the monitoring of wood density with good spatial resolution and transparency [1]. The accurate determination of wood density, grain angle and moisture content using microwave frequency has been reported [2]. Recently, the potential of the terahertz (THz) wavelength range for monitoring wood quality has been demonstrated [3–5]. These techniques have high potential for wood products because of their good transmission through wood.

Near-infrared (NIR) reflectance spectroscopy, which is mainly a tool for the analysis of chemical composition, is receiving ever-increasing attention as a means to monitor properties of wood. Many studies showing that the chemical, physical and mechanical properties of wood can be predicted by NIR reflectance spectroscopy with the aid of statistical methods (i.e., chemometrics) have been reported [6–8]. However, such chemometric NIR approaches have some disadvantages. First, the spectral contribution of the light absorption and scattering phenomena cannot be explained independently. Second, the construction of a calibration model requires a considerable amount of data, which is usually not transferable among instruments. Additionally, the light scattering contribution to NIR spectra is significant because wood is a complex and highly scattering material due to its cellular structure. The structure of softwood mainly consists of tracheid, arrayed along the longitudinal direction, although hardwood has various structures (e.g., tracheids, vessels, libriform wood fibers, ray cells). For the construction of robust calibrations for wood properties by NIR spectroscopy, it is important to independently evaluate the spectral contribution due to light absorption (absorption resulting from harmonics or overtones of the fundamental absorptions of molecular vibrations of cellulose, hemicellulose, lignin, extractives and water) and light scattering (mainly due to the cellular structure).

Time-of-flight (TOF) spectroscopy, commonly utilized in biomedical fields, has emerged as a technique to characterize the absorption and scattering effects in highly scattering media. Patterson et al. developed a model based on the diffusion approximation of radiative transfer, which yielded an analytical expression for pulse shape in terms of the interaction with a homogeneous slab, for the determination of optical properties [i.e., absorption coefficient (μ_a) and reduced scattering coefficient (μ´_s)] in tissue [9]. μ´_s is defined as

The first studies on wood material using TOF spectroscopy were reported by Tsuchikawa et al. [10–12]. They mainly investigated the relationships between time-resolved profiles (i.e., peak intensity, time delay, full width at half maximum) and wood properties. Recently, the application of time-resolved diffuse optical spectroscopy for the determination of optical properties of wood was reported by European groups [13–16]. D’Andrea et al. determined μ_a and μ´_s in the wavelength range of 700-1040 nm for two kinds of wood species treated in different ways (dry, wet and degraded) by time-resolved reflectance spectroscopy with two different orientations of the optical fiber and obtained many interesting results [14]. It was reported, for all cases, that the reduced scattering coefficient (μ´_s = 10 – 100 cm⁻¹) was much larger than the absorption coefficient (μ_a = 0.05 – 1.00 cm⁻¹). Scattering spectra were constant over the wavelength ranges. They also found that μ´_s strongly depend on the wood species (Silver fir and sweet Chestnut wood). μ´_s of wet wood was much smaller than that of dried wood due to the refractive index matching effect between the wood cell wall substance and water in the pores. D’Andrea et al. also evaluated the moisture content (MC) of wood using the absorption coefficient and found a good relationship between MC and the absorption coefficient at a specific wavelength [15]. Kienle et al. fully investigated the origin of scattering in wood by comparing the experimentally measured light propagation on the microstructure of Silver fir with a simulation modeled by the Monte Carlo method [13]. They determined μ´_s (for wet wood μ´_s = 1.79 mm⁻¹, for dried wood μ´_s = 6.68 mm⁻¹) due to tracheids by solving Maxwell’s equation and μ_{s iso}, which is the scattering coefficient due to all other scattering media (pits, ray cells, rough border between the lumen and wood cell wall substance), calculated by fitting between the measured and simulated light propagation.

In past research dealing with wood material using TOF spectroscopy, the wood samples were regarded as a homogenous material. However, in practice, wood has complex, heterogeneous and anisotropic cellular structures. Therefore, the scattering coefficient highly depended on the wood species because the cellular structure, which caused multiple light reflections, differed significantly between wood species. In particular, hardwood species have various cell arrangements, i.e., ring-porous, diffuse-porous, radial-porous and figured-porous. For the construction of robust calibrations for wood properties by NIR spectroscopy, the determination of the true μ_a and μ´_s values of wood cell wall substance itself is necessary. In this study, we determine the true μ_a and μ´_s values of wood cell wall substance. We expect that the μ_a and μ´_s values of the cell wall substance itself are identical or similar between species because the density of wood cell wall is about 1.4 - 1.5 g cm⁻³ regardless of species (i.e., wood density depends on the pore volume in wood) [17]. Since the density is identical, the factor affecting the optical properties is thought to be the concentration of the three main polymers comprising the cell wall (cellulose, hemicellulose and lignin). However, at the wavelength used in this study (846 nm), there was no specific absorption band, as shown by Hans et al. which implied that the relative concentration of the three major polymers in the cell wall does not strongly affect μ_a [18].

Todoruk et al. reported a curious experiment measuring the THz time domain transmission spectra of liquid saturated woods with refractive index spanning that of the cell wall substance in order to determine the source of birefringence observed in the THz region [4]. In their study, samples were saturated with hexane, toluene or quinolone to minimize the form birefringence resulting from periodic cellular structure. They quantitatively evaluated the contribution of form birefringence and intrinsic birefringence (birefringence resulting from microscopic structure consisting of crystalline and polycrystalline regions). Large contributions of both form and intrinsic birefringence were reported for maple and fir, whereas in aspen, the intrinsic birefringence contribution dominated. We conducted experiments of the same kind to evaluate the scattering due only to wood cell wall substance. The determination of μ´_s due to wood cell wall substance will lead to the robust evaluation of cellular structure of wood by spectroscopic techniques.

2. Material and method

2.1. Samples

The experimental procedure is depicted in Fig. 1. Douglas fir (Pseudotsuga menziessi) was cut into wafers approximately 3 mm thick (tangential) x 60 mm (longitudinal) x 90 mm (radial). 12 samples were prepared and dried in a 105 °C oven for 24 hours to remove moisture. The samples were saturated with hexane, toluene or quinolone to reduce the multiple reflections of NIR light on the boundary between pore-cell wall substance in wood. Chemical structures, densities, refractive indices and absorption coefficients at 846 nm of organic solvents are summarized in Table 1. Transmission measurements of these organic solvents were performed with a quartz cell of 10 mm path length using a UV-VIS spectrometer at 846 nm to determine the absorption coefficient. The refractive indices of hexane, toluene and quinolone are 1.388, 1.4969 and 1.625, respectively. The refractive index for wood cell wall substance (n_cw = 1.55) falls within this range [13].

 

Fig. 1 Experimental procedure.

Download Full Size | PDF

Table 1. Summary of Saturation Liquids.

View Table

Oven-dried samples were submerged in the liquid under vacuum conditions until the mass stabilized and samples were fully saturated (12 hours maximum). Four samples were submerged in each organic solvent. To ensure the saturation, the weight of wood samples was compared to the estimated saturation weight calculated from the volume fraction of air, wood cell wall substance and volume of wood samples. Volume fraction of air was calculated as

2.2 TOF-NIR measurement

The TOF-NIR measurement system is mainly composed of a picosecond light pulser with a wavelength of 846 nm (PLP-10, Hamamatsu Photonics Co., Hamamatsu, Japan). A pulse width of 70 ps was used and the beam diameter on the sample was approximately 1.5 mm. The time variation of the transmitted radiation intensity was recorded using a streak camera (C5680: 10.3 ps time resolution, 200 – 900 nm effective wavelength range, Hamamatsu Photonics Co., Hamamatsu, Japan) and a CCD camera (C9300: Hamamatsu Photonics Co., Hamamatsu, Japan). The wood sample was placed at a distance of 135 mm from the laser source. The distance between the light pulser and the streak camera was 360 mm. The instrument response function (IRF) was measured with a couple of neutral density filters with a transmission ratio of about 2% at 846 nm (FND-30C02-0.1 Sigma Koki Co., Tokyo, Japan) placed directly in front of the camera slit. Time-resolved profiles (TRP) were obtained from the center of oven-dried and liquid-saturated wood samples on the radial face with a time range of 2 ns. Photon counting was performed for 60 s.

2.3 TRP analysis

TRP processing was conducted using Matlab (MathWorks, Inc., MA, USA). The measured TRP can be expressed as a convolution between the IRF and true photons of the time-resolved distribution (TRD) as Eq. (4):

3. Result and discussion

3.1. TRP distribution of oven dried and liquid saturated wood samples

Figure 2 shows the TRP obtained for liquid-saturated (hexane: line with triangles; toluene: solid line; quinoline: line with circles) and oven dried wood (dashed line). While organic solvent absorbs more NIR light than air, a higher intensity of transmitted light for liquid-saturated wood samples could be observed than in the oven dried samples. An increase in the transmission intensity of 830 nm continuous-wave laser radiation with increasing moisture content in wood was also reported by Tsuchikawa et al. [21]. D’Andrea et al. compared the μ´_s between dried and wet woods (Silber fir; Avies alba and Sweet Chestnunt; Castanea sativa) in the wavelength range of 700 – 1050 nm [14]. For both wood species, they found a strong decrease in μ´_s for wet wood with respect to dried samples, confirmed with both perpendicular and parallel optical fiber alignment with regard to wood fiber grain. They attributed the decrease in μ´_s to the presence of water inside the wood pores, which reduces the refractive index mismatch between the interior and exterior of the fiber. Furthermore, Kienle et al. estimated the light scattering in the tracheids in Silber fir using Maxwell’s equations for a plane wave, which is incident onto an infinitely long cylinder, assuming that the long parallel cylinder has a diameter of 30 μm with a wall thickness of 3 μm [13]. They obtained anisotropy factors of g = 0.959 and g = 0.850, and scattering coefficients μ_s = 43.8 mm⁻¹ and μ_s = 44.6 mm⁻¹. Thus, μ´_s = 1.79 mm⁻¹ and μ´_s = 6.68 mm⁻¹ for wet and dried wood, respectively.

 

Fig. 2 TRP obtained from liquid saturated wood samples.

Download Full Size | PDF

The higher transmitted light intensity observed in this study for liquid-saturated wood is thought to be due to the index-matching effect of organic liquid in wood pores, which reduces the reflection of light on the boundary between wood cell wall substance (n ≈1.55) and liquid (n ≈1.5), allowing more light to reach the detector. As wood dries, its pores fill with air, resulting in a larger difference in the refractive indices at the boundary (the refractive indices of wood cell wall substance and air are n ≈1.55 and n ≈1.0). Multiple reflections of light on the repetitive wood cell wall substance-air boundary result in less transmitted light intensity.

Figure 3 shows the average path length of liquid saturated and oven-dried wood, where the points are averaged values calculated from 4 samples. The error bars shows the standard deviation within 4 samples as a function of index refractive value of filler materials. When the refractive index of the filler liquid matches the refractive index of the wood cell wall substance, the average path length reaches a minimum as a result of the index-matching effect. The minimum path length occurs when toluene is saturated into wood, since the refractive index of toluene is very close to the refractive index of wood cell wall substance, 1.55. The path length for the toluene saturated wood is about 12.7 mm, which is about 4 times thicker than the sample.

 

Fig. 3 Relationship between the refractive index of the saturation liquids and the average path length of the TRP.

Download Full Size | PDF

3.2. Absorption and reduced scattering coefficient of wood cell wall substance in wood

μ_a and μ´_s values are plotted as a function of the refractive index of pore filler in Fig. 4, where the points are averaged values calculated from 4 samples with air, hexane, toluene and quinolone as the pore fillers. Error bars indicate the standard deviation values computed from 4 samples for each organic solvent. The acquired value of μ_a for oven dried Douglas fir in this study [Fig. 4(a): μ_a = 0.008 ± 0.003 mm⁻¹] was similar to the values reported by Hans et al. (0.011 ± 0.002 mm⁻¹ and 0.008 ± 0.001 mm⁻¹ for air dried hinoki cypress and Japanese larch at 846 nm) [18] and D’Andrea et al. (about 0.005 mm⁻¹ for air-dried Silver fir at 840 nm for perpendicular alignment of optical fiber with regards to wood grain angle) [14]. Absorption bands of NIR spectra of air-dried wood are assigned to cellulose, hemicellulose lignin, extractives and bound water. However, at the wavelength used in this study (846 nm), there were no specific absorption bands and the values were likely independent of moisture content, as reported by D’Andrea [15]. This is the reason for the similarity in the μ_a values for oven-dried wood and air-dried wood samples reported by Hans et al. and D’Andrea et al. Hans et al. reported a linear relationship between μ_a at a wavelength of 846 nm and air-dried densities of 9 species of wood (including both soft and hardwood) with a determination coefficient of 0.56 and a root-mean-square error of 0.047 g cm⁻³. This linear relationship can be explained by the fact that the increase in wood density results in an increase in the volume fraction of wood cell wall substance that absorbs NIR light. The linear relationship also implies that the absorption coefficient at 846 nm is identical or similar among all species of wood used in their study. The μ_a value of liquid-saturated wood samples are higher compared to oven-dried wood samples because the μ_a values of hexane, toluene and quinoline are significantly higher than that of air, as shown in Table 1.

 

Fig. 4 Relationship between the refractive indices of the saturation liquids and (a) the absorption coefficient (μ_a) and (b) the reduced scattering coefficient (μ´_s) measured by TOF-NIRS.

Download Full Size | PDF

The μ´_s values of oven-dried wood samples were higher than those of liquid-saturated wood samples, as shown in Fig. 4 (b). The acquired μ´_s value for oven dried Douglas fir in this study was μ´_s = 16.8 ± 0.8 mm⁻¹, although Hans et al. reported μ´_s = 15.3 ± 0.3 mm⁻¹ and μ´_s = 17.4 ± 0.9 mm⁻¹ for air-dried hinoki cypress and Japanese larch at 846 nm [18] and D’Andrea et al. reported about μ´_s = 10 mm⁻¹ for air-dried Silver fir at 840 nm [15]. The light scattering in wood is complex. It is caused by: (1) the interaction of light with small particles with different refractive indices compared to surrounding materials and (2) a refractive index mismatch between wood cell wall substance and lumens. Since the type of cells (i.e., tracheid, ray cells, vessel), shape of cells (i.e., between type of cells, between latewood and earlywood), and alignment of vessels (i.e., ring-porous, diffuse-porous, radial-porous and figured-porous) differ widely between wood species, the scattering coefficient values should also differ between wood species. Furthermore, (3) scattering at the rough border between lumen of the tracheid and wood cell wall substance [10] and (4) scattering due to the birefringent nature of the cellulose microfibril or spatial distribution of chemical components in wood cell wall substances should be taken into account.

In this research, the minimum μ´_s was found when toluene was saturated into wood since the refractive index of toluene is very close to the refractive index of wood cell wall substance, 1.55. In the toluene saturated wood sample, light scattering (1) at small particle, (2) between lumen and wood cell wall substance for all cell types (tracheid, ray cells, vessel) and (3) at the rough border are minimized as a result of the index-matching effect of toluene in wood pores. Since the light scattering by toluene itself might be negligible, the value of μ´_s = 4.97 ± 0.48 mm⁻¹ should be due only to the (4) scattering due to wood cell wall substance.

We calculated the optical parameters for unit volume, assuming that μ_a and μ´_s are proportional to volume, by dividing the μ_a and μ´_s by volume fraction of wood cell wall substance (oven dried density = 0.38-0.41 g cm⁻³ corresponding to the volume fraction of wood cell wall substance = 0.25-0.27) yielding μ_{a_ws} = 0.030 mm⁻¹ and μ´_{s_ws} = 18.4 mm⁻¹.

3.3. Light propagation simulation using the Monte Carlo method

Measured light propagation by TOF-NIRs was compared to simulations carried out using the Monte Carlo method [22] using the true optical parameters of cell wall substance determined in this study. We modeled wood cell structure (right side of Fig. 5) based on the real cell wall thickness and cell size by the optical microscopic observations (left side of Fig. 5). The values μ_{a_ws} = 0.030 mm⁻¹ and μ´_{s_ws} = 18.4 mm⁻¹ were set for the pixels corresponding to the cell wall and μ_{a_ws} = 0 mm⁻¹ and μ´_{s_ws} = 0 mm⁻¹ were set for the pixels corresponding to air. We performed Monte Carlo simulations similar to those carried out by Wang et al. [23]. Detailed descriptions of the Monte Carlo simulations can be found in the literature [24]. For the analysis of light propagation in wood, light was treated as a set of photons. Photons that traveled a specific distant L are scattered toward the direction of (θ, Φ) and attenuated to the specific ratio of μ´_s/(μ_a + μ´_s). L, θ and Φ are defined as

 

Fig. 5 Model of cell wall structure in wood.

Download Full Size | PDF

The number of incident photons was set to 10⁶. The convolution between the simulated time-resolved transmittance spatially integrated over the back side surface of the wood and IRF is shown in Fig. 6 as a black solid line with measured TRP (dashed line). The simulated and experimental TRPs were max-normalized. There was good agreement between the Monte Carlo data and the observed TRP. However, there were considerable differences between measured and simulated pulses at the falling edge (i.e., 0.6-1.0 ns). This may originate from the uncertainty due to the finite number of photons. However, the time-resolved profiles such as peak intensity, time delay, and full width at half maximum coincided well between measured and predicted values, which suggests the validity of the optical parameters of wood cell wall substance calculated in this study.

 

Fig. 6 Measured (dashed line) and simulated (solid line) TRP (sample thickness: 1.03 mm)

Download Full Size | PDF

This research is the first attempt to calculate the optical parameters of wood cell wall substance itself. Scattering by 1) small particle, (2) lumen and wood cell wall substance border for all cell types (tracheid, ray cells, vessel) and (3) rough border might depend on the wood species since the alignment of vessel (i.e., ring-porous, diffuse-porous, radial-porous and figured-porous) differs widely between species. However, we expect that the μ_a and μ´_s values of wood cell wall substance might be identical or similar between species as explained in the introduction section. The true values of μ_a and μ´_s of wood cell wall substance were obtained by a nondestructive evaluation of wood cell wall structure (i.e., ring-porous, diffuse-porous, radial-porous and figured-porous) by transmission measurement of NIR laser spectroscopy. To confirm our methods, further measurements of the μ_a and μ´_s values for many more wood species is needed.

4. Conclusions

Absorption coefficients and reduced scattering coefficients of cell wall substance in Douglas fir at 846 nm were determined via time-resolved NIR transmission spectroscopy. Time-resolved transmittance of liquid saturated wood samples was fitted to the diffusion approximation equation to calculate μ_a and μ´_s. μ´_s exhibits a minimum when toluene is saturated into wood because the refractive index of toluene is close to the refractive index of wood cell wall substance. In the toluene saturated wood sample, scattering at small particle, between lumen and wood cell wall substance for all cell types (tracheid, ray cells, vessel) and at the rough border are minimized as a result of the index-matching effect of toluene in wood pores. The optical parameters of wood cell wall substance calculated taking into account the volume fraction of wood cell wall substance were μ_a = 0.030 mm⁻¹ and μ´_s = 18.4 mm⁻¹. We observed good agreement between Monte Carlo simulations using optical parameters of wood cell wall substance and measured time-resolved transmittance profiles.