Introduction

Torrefaction is a pre-treatment method, where the handling, logistics, and combustion properties of different biomasses, including wood, can be improved. The end product can be combusted directly or refined into pellets, biogas, or bio-oil. Any type of biomass can be used as feedstock. The process is executed by heating the material at 200–300 °C, in absence of oxygen and at ambient pressure [1]. If wood is used, an intermediate between raw wood and charcoal is obtained. During torrefaction, thermal degradation of the wood’s components takes place. The end product has reduced tenacity and volatile content, and increased heating value and homogeneity. An important aspect in respect to the fuel properties is the reduction of hygroscopicity. Most of the adsorption takes place in the polar hydroxyl (OH) groups of hemicelluloses and amorphous cellulose. These components are mostly volatilised in pyrolysis and thus the amount of available OH groups is reduced [1, 2]. Further heat-induced reactions, such as cross-linking and formation of non-polar compounds, also play a role [1, 3–5]. In addition to the degradation of hydrophilic structures, the accessibility of OH groups has an important effect on the water sorption. Up to 65 % of cellulose is crystalline [6], structurally blocking the access of water into available binding sites in the glucose molecules of cellulose. The crystallinity of cellulose further increases in pyrolysis [5], leaving less OH groups available and therefore in part reduces the hygroscopicity.

Exposure to deuterated water vapour can be used to estimate the content of accessible OH groups. Sorption of deuterium oxide can be compared to sorption of ordinary water. The OH groups are converted into OD groups and the accessibility can be directly measured through increase in mass. This can be accomplished by the use of dynamic vapour sorption apparatus (DVS) that can provide very accurate information on the sorption properties [7].

Though some of the properties of torrefied biomass have been studied extensively, the sorption properties still need to be investigated with respect to the logistics of the material [8, 9]. The material’s interaction with water is crucial to the economics of the supply chain, most importantly transport and storage. Sorption isotherms for the same material as used in this study were reported previously [10] but only adsorption was considered. In this study, both adsorption and desorption and also the differences between first and second cycle will be recorded, as well as the sorption hysteresis. To our knowledge, neither the effect of more than one cycle nor the hysteresis has previously been investigated specifically for torrefied wood. To gain a more thorough understanding on the effect of the pre-treatment on sorption, accessibility of the hydroxyl groups and changes in the specific surface area, as well as the particle size distribution will be reported. The clustering behaviour of water molecules will also be discussed.

Materials and methods

The material consisted of one mature Norway spruce (Picea abies Karst.) and several young downy birches (Betula pubescens Ehrh.), felled in Helsinki, Finland. The wood (stem and thick branches) was cut into approximately 5 × 5 × 5 cm blocks without removing the bark. The pyrolysis was executed in an indirectly heated pilot-scale reactor managed by Biosampo project in Kouvola Region Vocational College (KSAO). The material (~6 % moisture content, MC) was torrefied at 220, 260, 300, and charred at 450 °C in 25 kg batches. The holding time at peak temperature was 3 h, which was relatively long but necessary in ensuring thorough conduction of heat within the blocks. After 3 h, heating was switched off and the reactor was left to cool before extracting the samples. Representative bark-free sample blocks were ground with a Wiley mill. Sorption, hysteresis, and the accessibility of the material were measured with a dynamic vapour sorption apparatus (DVS; Surface Measurement Systems Ltd, London, UK). A more detailed description of the apparatus can be found in [11–13]. 20 ± 1 mg of powdered material was weighed onto the sample pan. The change in sample mass was measured by a microbalance, located in a chamber with constant flow of dry nitrogen mixed with a flow of nitrogen and water vapour. A temperature of 24–25 °C was retained in the chamber. Relative humidity (RH) was increased stepwise from 0 to 95 % (5 % steps), with the RH remaining constant until change in sample mass was below 0.002 % min−1 over a 10-min period. After constant mass, the RH was raised to the next step. After constant mass at 95 %, the RH was lowered stepwise back to zero. The first sorption and desorption cycle was immediately followed with a second sequence under identical conditions. Absolute hysteresis was calculated as the difference between EMC at desorption and EMC at adsorption.

The amount of accessible hydroxyl groups was measured through mass increase of sample during deuterium exchange in DVS, where water vapour was substituted for deuterium vapour. Similar measurements have been conducted before [7, 14–16]. Altogether 8 cycles were conducted in a similar manner as in the sorption experiment. The amount of accessible OH groups was then calculated according to Eq. 1,

$$ A\, = \,\frac{{m_{\text{f}} \, - \,m_{\text{i}} }}{{m_{\text{i}} }}\, \times \,1000\; \left( {\text{mol/kg}} \right), $$
(1)

where A is the amount of accessible OH groups per dry mass of the sample, m i is the dry mass of sample before deuteration, and m f the mass of sample after 8 cycles. The atomic mass difference between deuterium (2H) and protium (1H) is approximated as 1 g/mol.

The specific surface area was determined with DVS using the BET method provided as a function of the software. The instrument can perform the measurements in room conditions and with small sample amounts and the results are comparable to the traditional BET method usually performed at −196 °C. The function is based on determination of the amount of gas (water vapour in this case) adsorbed on the surface of the solid, from which the amount of the gas corresponding to a monomolecular layer on the surface is calculated. Changes in the particle size distribution were investigated with Mastersizer 2000 (Ver. 5.60; Malvern Instruments LTD, Malvern, UK). The particle size measurements were repeated in triplicate.

The clustering behaviour of the sorbed water molecules was determined using the Zimm–Lundberg (Z–L) method [17, 18], which has been applied to sorption on wood only twice before [19, 20]. One of the advantages of using such an analysis is that it only requires the sorption data from the isotherm as inputs. The Z–L clustering function is defined as follows:

$$ \frac{{G_{\text{ww}} }}{{v_{\text{w}} }} = \left( {1\, - \,\emptyset_{\text{w}} } \right)\left[ {\frac{{\partial \left( {\frac{{a_{\text{w}} }}{{\emptyset_{\text{w}} }}} \right)}}{{\partial a_{\text{w}} }}} \right]_{P,T} \, - \,1, $$
(2)

where G ww /v w is the cluster integral, Ø w is the volume fraction of water in the substrate, and a w is the water activity. The average number of water molecules in a cluster, or the mean cluster size (MCS) can be calculated from the following equation [21]:

$$ {\text{MCS}}\, = \,1\, + \,\frac{{\emptyset_{\text{w}} G_{\text{ww}} }}{{v_{\text{w}} }}. $$
(3)

Results

In a previous experiment [22] it was determined that the pyrolysis (torrefaction and charring combined) increased the amount of carbon (percentages of mass) in both wood species, from around 50 % of untreated wood to up to 91 % of wood pyrolysed at 450 °C. Figure 1 shows the sorption isotherms for spruce (Fig. 1a, b) and birch (Fig. 1c–e) for the first and second sorption cycles. The shape of the isotherms changes with increasing severity of pyrolysis. With birch, a distinct change occurs already at the lowest treatment temperature of 220 °C (Fig. 1d) and the hysteresis becomes very small (>0.24 %; Fig. 2d). A very small hysteresis loop can still be seen, even after torrefaction at 450 °C. At 260 and 450 °C, the adsorption of spruce was slightly smaller than that of birch wood.

Fig. 1
figure 1

Adsorption isotherms of spruce (a, b) and birch (c–e) at 0–95 % RH measured for two cycles of adsorption–desorption

Full size image
Fig. 2
figure 2

Hysteresis of spruce (a, b) and birch (c–e) at 0–95 % RH measured for two cycles of adsorption–desorption

Full size image

Hygroscopicity of wood in the adsorption branch of the isotherm is reduced by repeated humid–dry cycles [11], and such changes, although small, can be seen especially with spruce treated at 220 °C. The differences between the two cycles disappear almost entirely at temperatures above 220 °C. This may be related to the increasing rigidity of the wood sample, though more cycles would probably give more information on the matter. The pyrolysis decreased hysteresis of all samples, and the hysteresis of birch was smaller than that of spruce at all temperatures (Fig. 2a–e). Absolute hysteresis decreased from 2.4 for the untreated sample to 0.07 for the charcoal on both tree species (average of values at 0–95 % RH).

The accessibility measured by deuteration of available OH groups (Fig. 3) first decreased faster for birch than spruce, but at 300–450 °C the accessibility of spruce was lower than that of birch. Overall, the accessibility of spruce was slightly smaller with a decrease of up to 88 % compared to 81 % of birch at 450 °C.

Fig. 3
figure 3

Changes in OH accessibility after pyrolysis at 220–450 °C, a spruce, b birch

Full size image

The clustering behaviour of water molecules was analysed using the Z–L approach and the results are presented in Fig. 4. The data indicates that water clusters apparently begin to form in the cell wall at a RH of around 75 % in the untreated wood samples and also with the spruce sample modified at 220 °C. With thermal modification the water is much more distributed in the material until the very top of the hygroscopic range studied. However, at a treatment temperature of 450 °C, there is a distinct change in behaviour with clustering apparently occurring in the mid hygroscopic range before decreasing again.

Fig. 4
figure 4

Clustering of sorbed water molecules in birch and spruce determined using the Zimm–Lundberg approach

Full size image

The specific surface area determined by the BET function of the DVS decreased along with increasing treatment temperature with both tree species: from 166 to 90 m2/g (birch; untreated to 450 °C), and from 183 to 87 m2/g (spruce; untreated to 300 °C). Over the range of 5–35 % RH, the fit of the model was excellent (r 2 > 0.998) for samples treated at 0–300 °C, but linearity of model was lost with samples treated at 450 °C.

The particle size distributions (Fig. 5a–b) were quite similar for untreated birch and spruce, but for treated samples there were clear differences. The size distribution of spruce wood particles was narrower than that of birch wood, about 6–640 and 5–1200 µm, respectively. The pyrolysis clearly increased the share of small particles of spruce wood (>100 μm), whereas for birch the majority of particles were over 50–100 µm in size, although the share of small particles also increased when compared to the untreated reference.

Fig. 5
figure 5

The particle size distribution of spruce (a) and birch (b) before and after pyrolysis

Full size image

Discussion

When wood undergoes pyrolysis, the structural components undergo thermochemical degradation and volatilise. The anatomical and chemical characters of soft- and hardwoods differ, but their sorption behaviour is fundamentally similar [23] with hemicelluloses being the most hygroscopic components in both woods. Hardwood hemicelluloses are largely composed of glucuronoxylan, which degrades more extensively at elevated temperatures than softwood hemicelluloses [24]. Deterioration of hemicelluloses begins around 180 °C [3], but the actual pyrolysis peak for hemicelluloses is situated between 200 and 300 °C [25]. The effect of faster degradation of xylan can be seen in the sorption isotherm of birch torrefied at 220 °C, as well as in the more extensive reduction of OH accessibility of birch wood at the lower treatment temperatures. With spruce wood, the isotherm of the sample treated at 220 °C still exhibits the characteristic sigmoid shape (IUPAC Type II) associated with cellulosic materials [26, 27], after which the isotherms resemble more combinations of Type I and Type V: Type I isotherms are given by microporous solids and Type V is typical for wood charcoal, where the final adsorption is limited by fine rigid pores that fill up before saturation vapour pressure is reached [26, 28]. The samples treated at 450 °C follow this type of curve. The isotherms seem somewhat comparable to ones obtained with DVS in other cases [13, 15, 29], although manufacturing conditions and feedstocks are different.

Compared to labile hemicelluloses, cellulose has higher thermal stability. This is usually depicted by relatively small changes in the content of glucose in relation to other polysaccharides, as cellulose pyrolysis takes place at 325 and 340 °C for spruce and birch, respectively [25]. In cellulose, each monomeric unit contains three OH groups, but approximately 50–65 % of cellulose is crystalline and inaccessible to water molecules [6, 23]. The degradation of lignin takes place over a wide temperature range and the relative mass proportion of lignin increases with elevated temperature [2, 30, 31]. Due to its aromatic structure, lignin is more hydrophobic than the holocellulose fraction, but lignin may be able to deform to accommodate water [32] and thus affect sorption. Because of the extensive volatilisation of the polysaccharide fraction, the char residue left over from pyrolysis at 450 °C is mainly composed of lignin [33].

Sorption hysteresis was reduced in all samples. Mild thermal treatment is reported to increase hysteresis [13, 29], but as the severity of the treatment increases, hysteresis also decreases due to an overall decrease in hygroscopicity [14]. The phenomenon of hysteresis is linked to the history of the sample where the starting point affects the result. The occurrence of hysteresis in the sorption isotherm of wood has been associated with the changes in molecular conformation of the cell wall matrix during the sorption process and associated with the dimensional changes in the substrate. The phenomenon has been explained in the polymer literature for quite some time ago [34], but the theory explaining this behaviour has only relatively recently been introduced to explain hysteresis in wood- and plant-derived fibres [32]. The model considers the creation and the destruction of nanopores in the cell wall matrix during the adsorption and desorption processes, respectively. Crucial to the model explaining the hysteresis phenomenon is the consideration of the matrix relaxation processes associated with adsorption and desorption occurring below the glass transition temperature (T g) of the material. Due to a lack of free volume surrounding the relaxing elements of the polymeric matrix, the creation of the cell wall micropores during adsorption and their collapse during desorption is kinetically hindered. This results in adsorption and desorption taking place on a substrate that is in two distinctly different physical states, giving rise to hysteresis. The further the isotherm temperature from the T g, the greater will be the hysteresis. This explains the decrease of hysteresis observed with Sitka spruce as the isotherm temperature is increased [11] and the disappearance of hysteresis in guar gum films when the T g is reached [35]. As the samples are pyrolysed, the dimensional changes associated with sorption become less important and this leads to a decrease in hysteresis; however at the same time, the increasing rigidity of the sample would be expected to result in an increase in hysteresis due to the associated increase in T g. The loss of hysteresis at higher treatment temperatures can be explained by the increased rigidity of the substrate, meaning that the micropore creation/destruction mechanism is no longer operative. Nonetheless, in activated charcoals, hysteresis is still observed due to the presence of (rigid) micropores in the structure [15]. The lack of hysteresis in the thermally treated samples in this study indicates that the thermal exposure has not resulted in the creation of any substantial level of microporosity in the now rigid materials [36], or that the internal porosity is inaccessible to water vapour, contrasting markedly with activated charcoals.

The higher EMCs of birch at higher relative humidities may be explained by the creation of voids following the faster structural degradation, although the presence of capillary-condensed water below saturation vapour pressure has been argued to be negligible in unmodified wood [37, 38]. In the case of thermally degraded wood, there is the opportunity for newly created voids to allow for capillary condensation. The voids could present more available sites for water molecules to occupy, although it has been shown that the OH accessibility does not necessarily correlate with EMC [17]. The newly created voids are presumably of a size that is larger than required for hysteresis to occur (see the argument above). In a study by Nakano and Miyazaki [4], a peak in hygroscopicity was reported at 350 °C, after which it decreased again. The explanation was that the hygroscopicity of wood is not dictated by accessibility alone, but also has to do with the microstructural complexity (accessible sites) and formation of hydrophobic groups. Also the behaviour of water brings about difficulties when drawing conclusions on the obtained results. Water adsorbs to active sites presented by the OH groups, but as the energies of sites differ, water molecules often preferentially adsorb to already hydrogen bound molecules and form clusters. The molecules may then bridge over spaces, a behaviour that is unique to water [39, 40]. The clustering could provide an explanation to the discrepancies between EMCs and accessibility. According to the Z–L model, clusters start forming only at the higher end of the hygroscopic range. The clustering behaviour changes at a treatment temperature of 450 °C, where the clustering apparently occurs in the mid hygroscopic range before decreasing again. This behaviour is not readily explainable and might point to a breakdown of the model for what is now essentially a rigid material. The result is not what would be expected for the formation of clusters in the cell wall and could only be explained by considerable expansion of the substrate at high RH, which is not possible. Therefore, it seems that the Z–L function cannot be applied in such circumstances. Although potentially a useful analytical tool, the Z–L function has been criticised when applied to sorption on glassy polymers because the theoretical basis for the function assumes thermodynamic equilibrium. According to [21] when sorption behaviour of glassy polymers is investigated below the glass transition temperature (T g), difficulties can arise due to the polymer matrix being in a non-equilibrium state. In order to determine the applicability of the Z–L approach, Davis and Elabd [21] used Fourier transform infrared (FTIR) spectroscopy to directly determine water association in several glassy polymers. They concluded that the Z–L function tends to underestimate the extent of water clustering in the sorbent. It is clear from that study that the results of the Z–L function need to be treated with caution when dealing with glassy polymers below T g. This is a situation that applies to wood. Further work is needed to clarify the situation and to determine the applicability of the method. It has also been suggested that the oxygen and carbon dioxide complexes chemisorbed on the surfaces of charcoal can fix water, providing additional sites [41]. At low treatment temperatures the swelling of wood may also contribute to opening more sites, and under the pressure exerted by water vapour, formation of microcracks may be possible.

Wood is considered to be a macro- and mesoporous substance and thermal treatment increases the porosity [10, 42]. The existing pores enlarge, open, widen, and coalesce with neighbouring pores [43–45]. As temperatures increase to and above 300 °C, micropores start developing rapidly [45]. The BET model is used to approximate the surface area of the samples and holds validity over the range of 5–35 % RH. Within this range, the data are considered acceptable, if the R-squared value is above 0.995 [44]. The model provides excellent fit with macro- and mesoporous materials, but cannot be used with microporous materials [27, 44, 46]. The surface area of wood often increases in vacuum pyrolysis due to micropore development [47], but the BET surface area does not take micropores into account. The breakdown of the model at 450 °C may have some link to the anomalous behaviour observed with the Z–L analysis. However, it must be borne in mind that the BET method is based upon various assumptions such as zero interaction between sorbent molecules, non-swelling substrate, and a very regular formation of monolayer followed by multilayers. This type of behaviour is far from the truth in the case of sorption onto the wood internal surface. The BET results can therefore be questioned, but this type of measurement is very commonly reported in the literature and we include the values for comparison purposes. Possible micropores may also be blocked by re-condensation of degradation products [47], or destroyed due to matrix collapse. However, in practice the microporosity of torrefied and charred wood does not have much significance, as water molecules may not be able to penetrate the smallest pores. Adsorption and accessibility increase with smaller particle size, but in presence of moisture the particles can agglomerate and block the access of adsorptive gas to pores on the surfaces [48–50]. Access to pores may also be blocked by re-condensation of degradation products on particle surfaces and inside pores, as well as matrix collapse during the pyrolysis process.

From a practical point of view, the most important outcome is the reduced hygroscopicity and accessibility. Microporosity may increase, but in logistics the volatilisation of wood components and the subsequent enlargement of inner volume that can hold more water originating from precipitation and ground contact holds the most significance. The relationship of water and pyrolysed wood is a combination of complex interactions with accessibility, hydrophobicity, particle size, and surface area. The smaller particle size increases accessibility, which is simultaneously decreased by removal of available sites due to chemical and physical alterations. The increased share of smaller particles, re-condensation of degradation products, and matrix collapse may block access to pores. As hardwoods are more reactive, they degrade faster than softwoods, but the increased porosity may lead to larger EMC. The compositional changes may have an effect in the handling and storage properties of pyrolysed wood, especially through capillary absorption in enlarged pores and increased reactive surface area due to decreasing particle size.