1. INTRODUCTION

For gas sensing and the concentration evaluation of trace gases using tunable diode laser absorption spectroscopy (TDLAS), multipass or long path absorption gas cells are commonly employed in order to increase the light–gas interaction distance and thus increase the detected absorption signal. Conventional multipass gas cell designs include the Pfund cell [1], the White cell [2], and the Herriot cell [3]. These absorption cells are all based on the alignment of mirrors which enable multiple passes of the light. A recent development in the field involves making compact multipass absorption gas cells, for example, placing a series of lenses in a circular pattern or using spherically abbreviated lenses, usually involving precise alignment [4,5]. An alternative approach to make compact multipass gas sensors, instead of using mirrors, is to utilize the multiple scattering found in porous materials. It has been demonstrated that by selecting materials with high scattering coefficients and high porosities it is possible to obtain significantly longer absorption path lengths compared to the thickness of the scattering material itself, which is on the order of 5–15 mm [6,7].

The technique, based on TDLAS, probes the weak absorption of the free gas molecules enclosed in the pores of porous media to obtain information about the gas content. The method is based on the fact that the gas absorption profile is about 10,000 times narrower than that of the surrounding solid-state materials. This specific application of TDLAS in scattering media goes under the name gas in scattering media absorption spectroscopy (GASMAS) [8] and has previously been employed to investigate optical properties in wood, such as porosity [9], gas diffusion processes [10], and wood drying [11,12]. The conventional method of measuring gas diffusion, both in the laboratory and the field, is to extract gas, using gas syringes, from a sample volume and then analyze it with gas chromatography [13]. The disadvantage of having to store the extracted gas in vials for postanalysis is avoided by gas sensors such as fluorescence [14] and semiconductor sensors [15]. In this work, a multipass oxygen gas sensor probe, based on the GASMAS technique, is developed, consisting of a laser-coupled optical fiber, a ceramic material, and a photodiode. It is here applied to gas diffusion studies of opaque materials. Previous gas diffusion studies using GASMAS have been done by monitoring the transmitted absorption signal of molecular oxygen in wood samples with slab geometry [10]. However, due to substantial light extinction of highly dense and scattering materials, the transmitted absorption signal can sometimes be difficult to extract even for samples of millimeter thickness. This is especially true for dark wood materials, of which archeological waterlogged wood is a special example. Although reflectance measurements can in principle collect more photons, the gas absorption signal is still rather weak since the light–gas interaction distance is often much shorter compared to the transmission measurement scheme. For large-volume samples, for example, in situ wood material studies, as in the example on the Swedish 17th-century warship VASA [16], the transmission measurements are even more difficult. For such difficult cases, a different approach is demonstrated here, where a miniaturized random-scattering multipass gas cell is inserted into holes in the sample, which can increase the gas absorption path length as well as the detected light intensity [6,17]. The gas exchange between the hole and the ambient environment through the wood material can be observed from the absorption signal of the gas enclosed in the hole of the wood sample.

The objective of this work is to first develop a random multipass gas sensor based on the GASMAS technique, where the experimental setup is described first (Section 2.A) followed by a description of the random multipass gas sensor (Sections 2.B and 2.C), and then a study of the gas diffusion processes in wood materials, which is covered in Section 4. Extremely difficult to examine because of being optically dark, oak samples are also examined to investigate the possibility of in situ measurements at the VASA ship, which has oak as its main construction material and where oxidation of sulphur inside the wood presently endangers the continued preservation of the ship [18].

2. METHODS AND MATERIALS

A. Experimental Setup

The experimental setup is shown in Fig. 1. A distributed feedback semiconductor diode laser (0760-0040-DFB-1, Toptica) with a lasing wavelength close to 760 nm and mounted on a thermoelectric cooler (TCLDM3, Thorlabs) was employed to probe one of the absorption lines in the A-band of molecular oxygen. The laser was controlled through a combined temperature and current laser driver (ITC 502, Thorlabs). The wavelength of the diode laser is scanned by a triangular signal at 25 Hz and simultaneously modulated by a sinusoidal wave at 11,025 kHz through the injection current of the laser, in order to apply the wavelength modulation spectroscopy (WMS) technique for sensitivity enhancement. Because of the modulation, oscillating signals of a derivative type are generated, which can be detected at different harmonic frequencies of the modulation as recorded with a lock-in amplifier [19]. As the light propagates through the sample, a photodiode is used to detect the scattered laser light from the random-scattering multipass gas cell such as our piece of ceramic sample [Fig. 1(a)]. The signal from the photodiode is then sent to a current-to-voltage amplifier (DLPCA-100) before being sampled by a data acquisition (DAQ) card.

 

Fig. 1. (a) Experimental schematic: the wavelength-modulated light from a diode laser is guided, in the first configuration (1), to the ceramic material of the probe through an optical fiber, and then the scattered light is detected in transmission geometry by a photodiode. The signal is then being amplified and sampled by a DAQ card. For the second configuration (2), light is delivered to the probe volume through an optical fiber, while the scattered light is collected by a second optical fiber and then transmitted to a photodiode. (b) Illustration of the ceramic that is here used as a random multipass gas cell. The laser light is guided through an optical fiber connected to the ceramic material. A silicon photodiode detects the transmitted light, and the signal is then sent to the DAQ board. (c) Illustration of the probe without the housing.

Download Full Size | PDF

The sampled signal is, in this case, processed by performing digital lock-in detection based on Fourier analysis [20] on twice the wavelength modulation frequency, $2f$, which corresponds to a broadened second derivative of the absorption profile. The obtained $2f$ absorption signal is then used for evaluation of the absorption imprint due to the molecular oxygen, which is proportional to the product of the oxygen concentration and the light–gas interaction distance. Since the path length through the gas does not change during the diffusion process as long as the optical properties of the sample volume do not change, the variation of the absorption signal is then a direct measure of the changes in oxygen concentration and thus the gas diffusion. The gas diffusion is studied by flushing nitrogen into the wood sample at a set distance and observing the oxygen absorption signal in the probe volume. A flowmeter (McMillan Co., 110-Flo meter) is utilized to read out the flow of nitrogen, which was kept at a read-out level of 2.5 l/min. During the measurement, the probe volume was sealed carefully to prevent direct gas exchange between the probe volume and the ambient air.

The presence of optical interference or etalon fringes will distort the WMS signal. Fringes arise from reflections between optical components but can also be observed from solid samples, for instance, the wood materials used in this study. To suppress the interference fringes, small vibrational motors are employed to vibrate the optical components; therefore the interference will be suppressed when signal averaging is performed.

B. Random Multipass Gas Cells

A random multipass gas sensor compatible with the experimental setup is developed using a piece of ceramic attached between an optical transmitting fiber and a photodiode [Fig. 1(b)]. The ceramic material is mounted in a metallic housing to increase stability and robustness, with openings to enable interaction with the surrounding atmosphere. The ceramic material has a porosity of 75% [21] and is shaped to a diameter of around 9 mm and a thickness of 5 mm to fit into the metallic case which has an approximate wall thickness of 1 mm. Good characteristics of ceramic materials, or porous media in general, when employed as multipass gas cells are high porosity, high scattering, low absorption in the bulk material, and a fast gas exchange. It has also been shown that for porous materials such as ceramics, the absorption path length is proportional to the square of the thickness [22]. However, the maximal possible thickness is limited because the transmitted light though the ceramic decreases due to the scattering and absorption of the light.

As a comparison to the random multipass gas cell, measurements were performed when the laser light was focused through a lens to an optical fiber with the transmitting end located in the probe volume of the wood material, and a second optical fiber is used to collect backscattered light. Both studies were conducted as is described in the experimental setup (Section 2.A).

C. Data Analysis

To obtain information on the gas content without knowledge on the absorption path length, a so-called equivalent mean absorption path length ($L_{\mathrm{eq}}$) can be used. This value corresponds to the distance that the light would have to travel in ambient air to experience the same fractional absorption as in the scattering medium. It is evaluated by comparing the measured $2f$ signal ($S_{2f,\text{abs}}$) to a $2f$ signal obtained from a reference measurement ($S_{2f,\text{ref}}$) performed in ambient air at a known distance ($L_{\text{ref}}$):

D. Materials

Wood samples originating from oak, mahogany, and spruce, with dimensions of $10\mathrm{cm}\times 10\mathrm{cm}\times 5\mathrm{cm}$ were prepared by drilling holes with a diameter of around 11–12 mm and a depth of 50 mm. The samples are heartwood materials. The edge-to-edge distances between the holes were 10 and 15 mm. Wood species can very coarsely be divided into two different types: softwood and hardwood. Spruce is classified as a softwood, while oak and mahogany are hardwoods. Spruce, part of the softwoods and the conifer species, is mostly built up by wood cells called tracheids [23], which function to give structural support and to transport nutrients. Hardwoods can be further divided depending on their pore distribution. Mahogany, which is classified as a diffuse-porous hardwood, has a more homogenous distribution of its pores, whereas oak is a ring-porous hardwood which has a pore distribution that decreases over one season, that is, between two annual rings. Photos of the surfaces of the wood samples are shown in Fig. 2. The densities of the wood samples were calculated from the ratio between weight and total volume of the wood samples, namely, ${\rho }_{\text{mahogany}}=0.54\pm 0.03\mathrm{kg}/{\mathrm{dm}}^3$, ${\rho }_{\text{spruce}}=0.43\pm 0.02\mathrm{kg}/{\mathrm{dm}}^3$, and ${\rho }_{\text{oak}}=0.78\pm 0.04\mathrm{kg}/{\mathrm{dm}}^3$.

 

Fig. 2. Cross section of the wood samples. From left to right: oak (ring-porous hardwood), spruce (softwood), and mahogany (diffuse-porous hardwood).

Download Full Size | PDF

3. PERFORMANCE OF THE PROBE

A. Signal-to-Noise Ration

The diffusion properties of the ceramic material in the probe have to be tested before performing gas diffusion studies in wood samples. The signal-to-noise ratio (SNR) for the ceramic probe was evaluated at oxygen concentrations of the ambient value of 20.9% and at 0.0%. The experimental results are shown in Fig. 3, and the SNR was found to be around 60, implying a concentration sensitivity of 0.3% when averaging over 20 s.

 

Fig. 3. $2f$ WMS signal measured at oxygen concentrations at 20.9% and 0%, respectively, using the random multipass gas cell.

Download Full Size | PDF

B. Gas Diffusion in the Random-Scattering Multipass Gas Cell

Another important aspect of the gas cell is the intrinsic response time of the ceramic material, which puts a limit on the temporal resolution of the measurement, and a study was performed by putting the gas sensor together with a tube connected to a nitrogen gas cylinder into a plastic bag and then sealing it. Measurements were initiated in ambient air conditions with recordings every 20 s, corresponding to 500 wavelength scans; then the bag was flushed with nitrogen through the tube until the oxygen signal in the probe volume was no longer observable. Thereafter, the bag was removed so that the ambient oxygen could reinvade into the ceramic sample. The diffusion time of the ceramic was estimated to be around 30 s according to Fig. 4, and the absorption path length of the ceramic was 130 mm, giving a path length enhancement of 26.

 

Fig. 4. Here the gas diffusion in the ceramic material is monitored. Initially, the ceramic is exposed to ambient air conditions, and then nitrogen is flushed until the oxygen signal is no longer observable. Thereafter, the ceramic is again exposed to atmospheric conditions, the oxygen concentration restores to the ambient level, and the value of $L_{\mathrm{eq}}$ also comes back to the original level.

Download Full Size | PDF

4. WOOD DIFFUSION STUDIES

For each wood species, oak, pine, and mahogany, monitoring of gas diffusion was performed for two different geometries (Fig. 5) with the probe–gas tube distances of 10 and 15 mm, while the probe was not moved. The measurement is started by observing the oxygen absorption signal for about 5–10 min without any nitrogen flushing to achieve a stable base line, that is, there is no gas diffusion for the first few minutes. These data are then used as a reference when calibrating the mean absorption path length in order to obtain values of the oxygen concentration (Section 2.C). The gas absorption signal is continuously recorded when the gas valve is turned on, and the gas flow is monitored by the flow meter.

 

Fig. 5. Directional dependence of wood. Here two cases are defined. Longitudinal (L) and radial (R) directions with regard to the tree lines are represented by the lines in the figure.

Download Full Size | PDF

To investigate the influence of the flushing mechanism on the measurements, a spruce sample was studied for different gas flows (Fig. 6) at a fixed diffusion distance of 10 mm; this showed no significant difference in the diffusion time. Thus, it can be argued that most of the gas transport occurs through the wood, and no direct nitrogen flow through the possibly imperfect sealing, which would be exposed to an ambient low-oxygen environment. This also implied that the gas diffusion process is not influenced by the pressure of the gas injection.

 

Fig. 6. Comparison of the gas diffusion in spruce for different flow settings for flushing of nitrogen at a distance of 10 mm from the probe.

Download Full Size | PDF

When measuring on the oak sample, gas diffusions for distances of both 10 and 15 mm were close to identical with ${\tau }_{10\mathrm{mm}}=91\pm 6\mathrm{s}$ and ${\tau }_{15\mathrm{mm}}=99\pm 6\mathrm{s}$, respectively [Fig. 7(a)]. Similar values were also observed in the equilibrium oxygen concentration, with ${\alpha }_{15\mathrm{mm}}=10.2\pm 0.1\%$ and ${\alpha }_{10\mathrm{mm}}=10.0\pm 0.1\%$. Measurements performed in the radial direction to the annual rings showed no observable change in the oxygen signal, indicating no or very small gas diffusion in this direction. The main conclusion from the measurements is that the annual rings are the main barriers for gas diffusion, and thus the gas diffusion dominates in the longitudinal direction, which is in good agreement with findings by Sorz and Hietz [24]. This is in analogy to the water transport from the roots to the leaves in the sapwood of trees. Another interesting observation is that the weak gas diffusion in the radial direction could be very helpful for the trees to prevent bacteria growth in the heartwood, which is basically dead tissue and gives the strength of trees. From another point of view, the heartwood does not consume energy and thus oxygen. As has been pointed out by Wiedenhoeft [25], “There is no property of wood that is not fundamentally derived from the fact that wood is formed to meet the needs of living trees.” To quantify gas transport in wood, several wood properties have to be taken into account, such as the pore diameter, the pore distribution, the fiber structure, and the cell composition. In this work, we do not discuss these aspects in details.

 

Fig. 7. Gas diffusion measured in (a) oak, (b) spruce, and (c) mahogany for longitudinal and radial geometries.

Download Full Size | PDF

Performing the same procedure on the spruce samples yielded the time constants ${\tau }_{10\mathrm{mm}}=97\pm 4\mathrm{s}$ and ${\tau }_{15\mathrm{mm}}=122\pm 5\mathrm{s}$ with equilibrium oxygen concentrations of ${\alpha }_{10\mathrm{mm}}=8.7\pm 0.1\%$ and ${\alpha }_{15\mathrm{mm}}=10.6\pm 0.1\%$ [Fig. 7(b)]. Here the diffusion longitudinally at a distance of 10 mm is similar to that of the oak, but a noticeable increase in diffusion time is observed for the 15 mm diffusion distance. A small change is observed in the radial direction, which reconfirms the conclusion drawn previously that barriers in wood materials, such as the annual rings, are preventing gas from diffusing into the wood. The time constant for mahogany was evaluated to ${\tau }_{10\mathrm{mm}}=303\pm 8\mathrm{s}$, with ${\alpha }_{10\mathrm{mm}}=5.4\pm 0.1\%$, and ${\tau }_{15\mathrm{mm}}=256\pm 36\mathrm{s}$, with ${\alpha }_{15\mathrm{mm}}=16.8\pm 0.1\%$ [Fig. 7(c)]. Again, no obvious concentration variation is observed in the radial direction. In general, the time constants for the mahogany sample are much larger than for the other samples, and we also observe a shorter diffusion time for a gas diffusion distance of 15 mm compared to the 10 mm distance in mahogany. However, the equilibrium oxygen concentration for the 10 mm distance in longitudinal geometry is much larger than the 15 mm distance, and a possible explanation for the shorter gas diffusion time could be small cracks in the wood material.

The difference in the equilibrium oxygen concentration, $\alpha$, observed for the diffusion curves in Fig. 7 is mainly due to competing diffusion of the nitrogen flow to the probe volume and the gas exchange between the probe volume and the surrounding oxygen-containing wood as well as with the ambient air due to imperfect sealing; the net balance of these diffusion rates is then observed as a nonzero value of the oxygen concentration.

To understand the improvement of using a random-scattering multipass gas cell in this particular application, gas diffusion processes monitored by fibers are also recorded, as shown in Fig. 8(a). As can be seen, similar gas diffusion processes are observed for the spruce sample, while the diffusion curves are significantly noisier. When comparing the $2f$ absorption signals between the measurements performed by employing only fibers and by utilizing a multipass gas cell, it can be concluded that the SNR has been improved significantly. Moreover, the $2f$ absorption signals are distorted because of the fringes originating from the interference in the probe volume—the hole of the wood sample. In particular, it is actually not possible to monitor the gas diffusion process for the oak sample due to heavy light absorption and interference fringes. An additional advantage of using the multipass gas cell is that the retrieved signal strength is then independent of the optical properties of the samples studied.

 

Fig. 8. Gas diffusion of spruce (a) for samples with the longitudinal and radial diffusion geometry, using fibers only. For spruce in a longitudinal diffusion geometry, diffusion times are ${\tau }_{10\mathrm{mm}}=222\mathrm{s}$ and ${\tau }_{15\mathrm{mm}}=228\mathrm{s}$. No obvious change in equivalent path length is observed in radial geometry. The diffusion times are much longer than those found previously using the multipass gas cell. This difference in diffusion time could be due to difficulties with the sealing method, resulting in different gas exchange rates between the probe volume and ambient air during the measurement. (b) Comparison between the $2f$ WMS signals obtained for measurements in spruce at atmospheric oxygen levels for the cases with and without probe.

Download Full Size | PDF

5. CONCLUSION

We have here demonstrated how an optical probe based on GASMAS can be employed to study the gas diffusion processes in wood materials. A random multipass gas cell to make the measurements independent of the optical properties of the samples bulk material was successfully employed. The probe, including the random multipass gas cell, a photodiode, and a laser coupled optical fiber, was inserted into the wood material, and the gas diffusion was studied by flushing nitrogen into the wood at a fixed distance from the probe. When only using the optical fibers in a probe, the signal strength varies between different samples, and a reduced SNR is obtained in samples that are optically dark. In addition, the ceramic introduced less prominent interference fringes than otherwise observed.

We could observe in the study that the densest wood material (oak) showed a fast diffusion compared to the other studied wood materials, which had a lower density. A higher density does not per se imply a faster diffusion, but instead the pore structure and the openness of the pores are important in characterizing the diffusion mechanisms in wood materials.

We recognize that a more complete implementation for quantitative gas diffusion measurements requires a robust and reliable sealing mechanism. Such measures could be to make measurements much deeper down in the sample with adequate sealing behind (e.g., an inflatable balloon), so that the diffusion distance, that is, the distance between the probe and the flushing volume, is much shorter than the diffusion distance from the probe to the atmosphere. Different measurement geometries of the probe position relative the gas flushing will be taken into consideration for future studies of gas diffusion. The advantage of having measurements geometries such as that in Fig. 9 is a more simple diffusion path, which reduces the complexity of the diffusion model.

 

Fig. 9. Sketch over measurement geometries in consideration for future studies together with a sealing mechanism, such as a balloon, behind the probe.

Download Full Size | PDF