A few studies report the analysis of oxygen limitation. However, the use of dissolved oxygen in enzymatic fuel cells is one of the main limitations of these systems due to low concentration (~ 0.2 mM) and low diffusion coefficient of oxygen (1.97 10-5 cm2 s-1 at 20 °C) (Barton, 2005). Usually, an exponential decay in the availability of oxygen at the cathode is observed along the length of microchannel (Bedekar et al., 2008). As a result, the oxygen flux is very low, limiting the generation of current. In the middle portion of the channel, the oxygen remains unconsumed and can still diffuse to the electrode. However, due to pressure-driven convective flow, the oxygen is still not consumed. To increase the availability of oxygen at the cathode surface, one strategy consists of designing a branched — microchannel configuration with several electrodes that allows periodically full contact of the electrolyte with the electrodes (Bedekar et al., 2008).

Another promising approach developed until now and only for fuel cells (methanol or acid formic/O2), is to incorporate cathodes that access the surrounding air with higher diffusivity and O2 concentration. Porous gas diffusion electrodes allow gaseous reactants to pass. Devices were developed either with flow-through porous electrodes able to increase delivered power densities with near complete fuel utilization (Kjeang et al., 2008), or with an air-breathing porous cathode structure (Jayashree et al., 2005). In the latter case, the use of air breathing cathode showed that the rate of oxygen reduction is enhanced with corresponding increase in current densities. However this approach requires precise control of pore size of the electrode to maintain constant rate of air delivery to the cathode.

These are used to modify the structure and dimensions of macroscopic and microscopic raw materials, and also their chemical composition. They have the effect of solubilizing the hemicellulose, reducing the crystallinity, and increasing the available surface area and porosity of the substrate. An effective pretreatment must meet following requirements: — it must increase sugar formation or facilitate the subsequent formation of sugars during hydrolysis, preventing any degradation or the loss of carbohydrates, and avoiding the formation of byproducts capable of inhibiting the subsequent processes of hydrolysis and fermentation, all at a competitive cost (Balat et al., 2008).

Pretreatments are particularly essential before enzymatic hydrolysis and may be of various types, i. e. physical, chemical, biological, steam explosion, and ammonia fiber explosion (AFEX).

Physical pretreatments may or may not be mechanical. The mechanical physical pretreatments include milling and grinding, that not only reduce the substrate, but also increase its surface area to volume ratio, thus making the cellulose easier to convert during hydrolysis. "Ball milling" could also be used to reduce the crystallinity of the cellulose, but this practice is not only very expensive, but also takes a long time (nearly a week) to complete, so it is hardly practicable on an industrial scale. The non-mechanical pretreatments feature a combination of high-power internal and external forces that decompose the lignocellulose.

Chemical pretreatments are used mainly to reduce the crystalline content of the cellulose. Using this type of pretreatment poses plant-related problems, however, since all the structural materials have to be capable of withstanding the severe working conditions imposed by the chemical agents.

The chemical pretreatments most often used are an alkaline treatment to delignify and solubilize the glycan, and an NaOH treatment that dissolves the lignocellulose biomass, destroying its lignin structure. Pretreatment with diluted sulfuric acid is also very important but this poses serious problems if it is associated with diluted acid hydrolysis, because the hydrolyzed end products become scarcely fermentable.

Other chemical pretreatments include: pretreatment with hydrogen peroxide (H2O2), which exploits oxidative delignification to separate and solubilize the lignin, and dissolve the lignocellulose matrices, thereby increasing the enzymatic digestibility of the mass; pretreatment with ozone, which degrades the lignin polymers; and pretreatment with liquid hot water (LHW), which is applied mainly to alfalfa. It was demonstrated (Laser et al., 2002) that, in ideal conditions, this method is as effective as diluted acid hydrolysis, without the need to use any acid or create any products of neutralization).

Biological pretreatments involve the use of enzymes, which are already useful in industrial processes on timber waste, in the processing of pulp and scraps. Several microorganisms studied years ago are the enzymes produced by the basidiomycetes Pleurotus ostreatus: these enzymes are homologous proteins characterized by different specifications, depending on which phenols are substituted. Another fungus in the basidiomycetes class that is effective in delignification is the Phanerochaete chrysosporium (Palmieri et al., 1997).

In the steam explosion process, saturated steam is used at very high temperatures and pressures to break up the chemical bonds in the cellulose, hemicellulose and lignin in order to break down the fibers and hydrolyze the biomass. The process consists in delivering steam under high pressure into a sealed chamber containing the lignocellulose material, then reducing the pressure and thus making the steam and matrix expand, and obtaining its explosive decompression through an orifice, which disrupts the cellular structure of the substrate, breaking up the acetyl groups of the hemicellulose. In some cases (e. g. Angiosperm), it is preferable to use acid catalysts, such as H2SO2 or SO2, to make the cellulose-rich components more accessible to the enzymes. SO2 gas is better able to attack the fibers (Shevchenko et al., 2000), but its use makes it necessary to carefully consider the working conditions in which the steam explosion takes place. In fact, it becomes necessary to find the best compromise between a strong enzymatic hydrolysis (obtainable in very severe conditions) and a good recovery of the components containing hemicellulose, that are in the form of monomeric sugars (which demand much less severe conditions) (Silverstein et al., 2007). That is why a severity indicator has been developed (Overend & Chornet, 1987), which correlates pretreatment temperatures and times, assuming that the pretreatment obeys Arrhenius’s equation and has first-order kinetics. The indicator R0 is:

where t is the duration of the pretreatment (min), Tr is the reaction temperature (°C), Tb is the baseline temperature (100°C) and the constant 14.75 is the conventional activation energy, assuming that the whole conversion is of the first order. If the version with sulfuric acid is being used, then the severity parameter, called M0 in this case, is slightly modified:

where C is the chemical concentration (wt%) and n is an arbitrary constant (Chum et al., 1990).

Ammonia fiber/freeze explosion (AFEX) pretreatment involves the use of liquid ammonia and steam explosion: in this process, the previously-humidified lignocellulose material is placed in a vessel under pressure with liquid NH3 in proportions of 1-2 kg NH3/kg of dried biomass. This method is very effective for non-woody materials such as bagasse and newspaper, but less so in the case of "soft" wooden materials. This system does not release any sugars directly, but it does make the polymers (hemicellulose and cellulose) easier for the enzymes to attack. The ammonia can also be replaced with carbon dioxide because the latter is relatively less costly and also because the alcohol waste product contains traces of pollutants that would thus derive only from the lignin.

The most promising pretreatments for farming waste are AFEX and LHW, while pretreatment with steam affords a high output of sugars from both farming waste and forest waste.

Lignocellulosic biomass is a readily available feedstock that can be purchased yearly from forest and agricultural operations. Forest residues comprise unusable trunk sections, limbs and tops. Typical composition of these residues is similar to that of common wood chips shown in Table 1. We define, as agricultural residues, the non-edible part of the plant which is let on the field after the harvest and the latter are usually composed of straw and stalk. It also comprises the parts of the cultivated plants that are thrown out after industrial processes. A specific example of such biomass includes but is not limited to corn cobs. Agricultural residues are the most probable feedstock that will be the original source for the production of ethanol from lignocellulsic materials due to their availability, their quantity and their proximity to the existing grain to ethanol platform. Non conventional plantation crops ( i. e. ‘energy crops’ ) are also to be considered as feed for biorefineries. Most of these crops have not reach industrial scale production (in North America) but an increasing amount of information has been published during the last few years about their chemical composition. Pricing for this biomass has been evaluated to 100-120$ per (dry basis; prices courtesy of CRB Innovations) metric ton but it tends to decrease because of a reduced use of fertilizers and the utilisation of marginal lands instead of high value agricultural land. The characteristics which make energy crops, especially perennial grasses, attractive for ethanol production, are the high amount of cellulose and hemicellulose as well as, under certain restriction, the favorable environmental impact.

Lignocellulosic biomass is composed of cellulose, hemicelluloses, lignin, extractives and ashes. The quantities of each fraction are detailed below for a large range of lignocellulosic materials including agricultural residues, energy crops and forest residues which are divided into leafy hardwoods and coniferous families (Table 1). Cellulose is the principal constituent of lignocellulosic plants representing 30-50 wt% of its composition. It is a polymer composed of D-glucose. Contrarily to cellulose, hemicelluloses are a heterogeneous polymer principally composed of pentoses (P-D-xylose, a-L-arabinose), hexoses (P-D — mannose, P-D-glucose, a-D-galactose) and/or uronic acids (a-D-glucuronic, a-D-4-O — methylgalacturonic, a-D-galacturonic acids). Among them, xylans and glucomannans are the most common compounds. Hemicelluloses represent 15-35 wt% of the plant. We can define lignin as a relatively hydrophobic amorphous polymer which consists of phenylpropane units. This macromolecule occurs primarily between the fibre cells, acting as a cementing material and giving the wood its rigidity and its impact resistance. It is always associated with hemicelluloses through carbon-carbon and ether linkages (Xu et al., 2008) and can be classified following 2 major classes (Gibbs and Thimann, 1958): (1) the guaiacyl which includes most of the lignins of softwoods (gymnosperms), (2) the guaiacyl-syringyl which comprises the lignins of hardwoods (angiosperms) and the lignins of grasses (non woody or herbaceous crops) (angiosperms). Extractives are composed of resins, fats and fatty acids, phenolics, phytosterols, terpenes, salts and minerals. This fraction is not used for the production of ethanol, and, for obvious reasons, neither are the ashes. The latter is defined as the residue remaining after total combustion. It is composed of elements such as silicon, aluminum, calcium, magnesium, potassium, and sodium. Typically, the amount of each fraction can also differ within a single biological species (following the environment: soil composition, water supply and weather patterns) and also during the growth of the plant making the quantification of sugar present in the holocellulose (sum of hemicelluloses and cellulose) difficult to specify.

The valorization of the lignocellulosic materials and more particularly of the carbohydrates (composing the holocellulose) into ethanol is made possible through their fermentation. Each plant has different composition (Table 1) but, as detailed before, contains the same major compounds. All the biomasses show comparable characteristics (Table 2) with the following order by quantity: Glucan>Xylan>Mannan-Galactan-Arabinan, except for the coniferous forest residues which show a high amount of mannan, which leads to a shift between the glucan and the xylan. Table 2 also shows an average of the 6 carbons sugars (C6) which can be fermented by most common yeast (including but not limited to S. cerevisiae) to give ethanol. Agricultural residues, energy crops and leafy forest residues present high averages of 41.5, 46.6 ans; 48.2 wt% respectively. Furthermore, coniferous forest residues could potentially produce more ethanol as they possess a very high amount of C6 (56 %) sugars which can be explained by a high amount of mannose (about 10 % more than the other species).

North America produced 46% of the world biofuels in 2008 (IEA, 2009)and the R &D efforts on second generation biofuels have been widely orientated toward the production of ethanol. From the results in Table 2, it is possible to estimate the ethanol production directly from C6 sugars using S. cerevisiae. Production of ethanol from C5 sugar was not taken into account in our study as these sugars require the use of special microorganisms. C5 sugars, although hard to ferment to ethanol could be converted into other value added products as ethyl levulinate (considered as part of the extended P-fuel pool) by successive dehydration, reduction and ethanolysis. Table 3 shows a comparison between the actual possible production of ethanol (not operational) using the forest residues, the agricultural residues and the unexploited forest biomass available in Quebec, Canada and North America versus the operational ethanol production from energy crop (first generation of biofuel) and the consumption of gasoline. Only 25 % of the forest and agricultural residues have been taken into account for ethanol production estimate as the rest of the biomass is already dedicated for others purposes. In the case of the unexploited forest, our study presents a result based on the forest zone which can be used without causing damage on biodiversity (Ministere des Resources Naturelles et de la Faune MNRF, Quebec, 2009).

Table 1. Chemical composition of various lignocellulosic materials

In the case of forest residues, we can assume that for 1 m3 of roundwood exploited, 0.6 m3 of residual biomass is left behind (Smeets & Faaij, 2007). In the province of Quebec, forest residues have been estimated to 6.9 millions of tons per year (Goyette & Boucher, 2009). Thus, the production of ethanol from glucose fermentation can be determinated assuming that the average of this sugar in such materials is about 52.4% (calculated from the Table 2) and that the maximum yield is equal to 0.51g of ethanol per g of glucose. Thus, 584 millions liters of ethanol could be produced in Quebec. To put such a value in perspective, consumption of refined petroleum in Quebec reached 9 billion liters in 2007 (Ministeres de Ressources Naturelles et de la Faune, MNRF, Quebec, 2009). The production of ethanol from forest residues is sufficient to reach the objective fixed by the government (5 vol% in gasoline in 2012) since it represents 6.5 vol%. The North American consumption of gasoline for transport was estimated in 2008 at 518 739 millions liters with respectively 479 243 millions liters for the United States of America and 39 496 millions liters for Canada (IEA energy statistic, 2010). More ethanol can be produced by using agricultural residues, energy crops and unexploited forest zone. In Quebec, the latter represents 14,100,000 m3 or 5.6 millions tons assuming an average density of 400kg/ m3. Thus, 1 638 millions liters of ethanol can be produced per year. Quebec will be able to replace 24.7 vol% of its gasoline by ethanol just by using exploited forest zone and forest residues, thus using only residual

a(Lee, 1997); b(S0rensen et al., 2008); c(Ballesteros et al., 2004); d(Wang et al. 2008); e(Sassner et al., 2008); f(Zhu & Pan, 2010); g(Zhu et al., 2009); h(Youngblood et al., 2009); i(Pettersen, 1984); jMesured in our laboratory

Table 2. Details of carbohydrates and lignin amounts present in various lignocellulosic materials

Table 3. Comparison between the actual possible production of ethanol from C6 sugars contained in the lignocellulosic biomass, the operational ethanol production from energy crop and the gasoline consumption

In this estimation of the possible production of ethanol, marginal lands have not been taken into account. The potential of surplus land for the cultivation of energy crops like willows, poplars, miscanthus, switchgrass, panic, reed canary grass (second generation of biofuel) considerably depends on the regions. Numerous constraints exist for the implementation of new energy crops, making the estimation of biofuel production very approximative. The first one is food competition. In fact, in major countries, populations are growing, consequently increasing the food demand therefore reducing surplus land, which overall limits the additionnal production of energy crops. Among the others constraints we can mention, the water shortages, the implementation of indigeneous species (could be a probleme for biodiversity), the type of plant, improvement of agricultural system (allowing the cultivation of other land), etc. Depending on the scenario, disparate results are obtained; bioenergy could reach between 39EJ to 204 EJ in 2050, furthermore, Smeets et al. 2007 show that the biggest energy apport will be done by dedicated energy crops with 20-174 EJ of biomass against 6-11 EJ of agricultural residues and 6 EJ of forestry residues.

More than replacing gasoline coming from fossil ressources, the employement of biofuels in well defined conditions can contribute to reduce the Greenhouse Gas (GHG) emissions. The GHG balance varies significantly following the choice of biomass, the technology employed throughout the full « fuel cycle » from biomass production to final fuel consumption, the caracteristic of the land and climate, the crop management, etc. Thus, the choice of biomass is essential. As for ethanol production, the potential of lignocellulosic biomass to reduce GHG is comprised between 60 and 120 % and it is comparable to the high diminution of GHG observed with sugar cane (90%). In comparison, production of ethanol from wheat grain brings a lesser gain of 20 to 50% (IEA, 2004). The reduction, especially for lignocellulosic biomass, is due to the compostion of plant itself, to fertilizer loading and to the efficiency of vehicles. The high reduction of carbon dioxide emissions in the case of lignocellulosic biomass (cellulose to ethanol) essentially comes from the use of the other part of the plant (mainly lignin) as a source of energy for the process. However the previous
estimations do not take into account the modifications affecting the lands. In fact, the GHG balance of the second-generation biofuels are closely related to the land use change (LUC) and the indirect use change (ILUC) which could in certain case conduct to a negative GHG balance. When a prior land-use like forest is replaced by culture for biofuel production, a direct land-use change occurs which can change the carbon stock of that land. This aspect has been widely studied and factors of changing balance can be found in the literature. However, the changes on GHG are induced by ILUC (takes place when land use change implies the displacement of the previous activity on another land). Thus, the replacement of sparsely vegetated and certain grass land by energy crops could generate a positive effect on GHG and in the mean time participate to the stockage of carbon in soil. Contrarily, the estimation made by Farrel and O’hare (2008) on the ILUC GHG emission shows that if the actual crops of soybean are used for the production of ethanol, the result could lead to the expansion of soybean for food into forests and will conduct to more emission than the use of fossil ressources (6 times more). These cultivations can also have several positive or negative impacts on soil, water and biodiversity. ILUC are normally less important if residues are used as feedstock since there is no need for additional land to be cultivated.

The second-biofuel generation, and in particular in the case of lignocellulosic ethanol production, should start with a sustainable development of agricultural and forestry residues which are at that time very interesting in terms of productivity (as it was shown on the Table 3) and environment. Even if some species contains much more C6 sugars like coniferous and notably loblolly Pine, the use of a wide range of biomass genotypes is advised and needed. All the species presented show an interest for ethanol production. In fact, the use of just the best species would be catastrophic for biodiversity. In general, lignocellulosic biomass is a promising source of fuel as it is shown on the figure 1. As an example, in about ten years the production of biofuel via the lignocellulosic biomass in the US could almost reach the same production as from the other sources of biofuel.

160

140

a>

120

100 ro

80 60 40

T3

о

20 0

2012 2014 2016 2018 2020 2022

Year

Fig. 1. Biofuel Mandate from lignocellulosic materials in the United States Renewable Fuels Standard

3. Fractionation

The starting materials used for alkali-catalyzed transesterification of glycerides must meet certain specifications. The presence of water during alkali catalyzed transesterification causes a partial reaction change to saponification, which produces soap. For that reason, the glycerides and alcohol must be substantially anhydrous (Wright et al., 1944). A small amount of soap favors the consumption of catalyst and reduces the catalytic efficiency, as well as causing an increase in viscosity, the formation of gels, and difficulty in achieving separation of glycerol. Ma et al. (1998) suggested that the free fatty acid content of the refined oil should be as low as possible, below 0.5%, and Feuge & Grose (1949) also stressed the importance of oils being dry and free of free fatty acids.

The use of alkali catalysts in the transesterification of used cooking oil is somewhat limited because the FFA in used cooking oil reacts with the most common alkaline catalysts (NaOH, KOH, and CHsONa) and forms soap. Because water makes the reaction partially change to saponification, the alkali catalyst is consumed in producing soap and reduces catalyst efficiency. The soap causes an increase in viscosity, formation of gels which reduces ester yield and makes the separation of glycerol difficult. These two problems not withstanding, literature is replete with studies on the transesterification of waste cooking oil using alkaline catalyst (Marchetti & Errazu, 2010).

Course of fermentations carried out with C. acetobutylicum DSM 1731 and milled corn as substrate was similar to that referred for C. acetobutylicum ATCC 824 (Lee S. Y. et al., 2008) i. e. it was characterized by distinct metabolic phases, reutilization of acids during solventogenesis and development of hydrogen that peaked during acidogenesis. According to Johnson et al., (1997), C. acetobutylicum DSM 1731 showed 96% DNA sequence similarity with C. acetobutylicum ATCC 824. The so-called acid crash i. e. the state when the fermentation finished in acidogenic step was sometimes observed from unclear reason, using this strain and milled corn as substrate (Rychtera et al., 2010). Unfortunately, intracellular level of formic acid was not determined and therefore it was not proved or disproved whether acid crash in these cases was also caused by formic acid (Wang et al., 2011).

The strain C. beijerinckii CCM 6218 should be identical with the strain C. beijerinckii ATCC 17795 according to data of Czech Collection of Microorganisms. Surprisingly, if the strain C. beijerinckii ATCC 17795 was tested for butanol production using molasses cultivation medium (Shaheen et al., 2000), both yield and maximum butanol production was low, 10% and 6.1 g. L-1, respectively. In addition this strain together with C. pasteurianum NRRL B-598 showed different fermentation pattern in comparison with C. pasteurianum NRRL B-598 and butanol production initiation started during exponential growth phase. The strain also metabolized substrate, saccharose, faster than both other tested strains what was reflected in higher productivity of butanol.

The strain Clostridium pasteurianum NRRL B-598 used in this study differed significantly in some physiological traits from both the species characteristics published in Bergey’s Manual of Systematic Bacteriology (Rainey et al., 2009). Although strains of the species C. pasteurianum are known rather as acetic and butyric acids or hydrogen producers (Rainey et al., 2009; Heyndrickx et al., 1991), the strain C. pasteurianum NRRL B-598 was cited in US Patent No 4539293 as butanol producing when used in mixture with further acidogenic strain e. g. C. butylicum. Unfortunately precise cultivation conditions, yields, solvents concentrations and other data are not available in the mentioned patent.

2.1 Effect oftemperature on viscosity of bio-oils from different feedstocks

The viscosities of bio-oils produced from different feedstocks were measured at 20, 40, 50, 60, 80, and 100°C and are shown in Fig. 3. In general, the viscosity of bio-oil at 20°C was higher than that of 100°C, irrespective of the feedstocks and with or without catalyst. This observation was in agreement with viscosity of bio-oil from pine wood chips (Thangalazhy — Gopakumar et al., 2010). Bio-oils used in this study had a higher viscosity at lower shear rate and the viscosity decreased exponentially at higher shear rate (>10/s) and similar result was reported for bio-oil from pine wood chips (Thangalazhy-Gopakumar et al., 2010). All the bio-oils used in this study showed non-Newtonian behavior as evident from Fig. 1. As mentioned in the introduction section, viscosity plays an important role in atomization through influencing inertial and aerodynamic instabilities. The Sauter mean diameter (SMD) of spray increases with viscosity for Newtonian fluid, whereas elasticity or shear thinning behavior of non-Newtonian fluid would affect the SMD. Thus, it is important to examine the non-Newtonian or viscoelastic nature of bio-oil since it may exhibit these effects during the application. According to Lu et al (2009b), most of the bio-oils behave as Newtonian fluids at temperatures lower than 80°C, whereas all the bio-oils used in this study even at high temperature (100°C) behaved as non-Newtonian fluid. A prevalent shear thinning behavior was observed at 50 and 80°C by Tzanetakis et al (2008) and similar behaviors were observed for all the bio-oils irrespective of the type of process (batch or continuous),with or without catalyst and kinds of feedstocks.

The viscosity of bio-oil between 20, 40, and 80°C were statistically different for all the feedstocks. Although the viscosities of bio-oil from corn cob were different at 20°C, the differences vanished at 80°C. In general, the viscosity of bio-oils produced from different feedstocks decreased with an increase in temperature. Similar trends were reported for bio­oils produced from different feedstocks such as softwood bark (Boucher et al., 2000a, b), sugarcane bagasse (Garcia-Perez et al., 2002), rice husk (Zhang et al., 2006), switchgrass (Boateng et al., 2007), corn stover (Yu et al., 2007), hardwood (Tzanetakis et al., 2008), pine and oak wood and bark (Ingram et al., 2008), pine wood chips (Thangalazhy-Gopakumar et al., 2010), and rice husk (Ji-Lu & Yong-Pong, 2010). When temperature was increased from 20 to 40°C, viscosity of bio-oil from canola showed a minimum decrease of 9% and bio-oil from corn cob 1 showed a maximum decrease of 25%. A further increase in temperature to 80°C resulted in viscosity decrease of 26 and 52%, respectively, for the bio-oil produced from canola and corn cob 1. Bio-oil derived from hardwood showed a similar behavior; however, the decrease was seven fold (Tzanetakis et al., 2008). The bio-oil viscosity measured at 40°C in this study was ten-fold lower than the viscosity (0.02 Pa. s) of the bio-oil produced from (heterotrophic) microalgae (Miao & Wu, 2004). The viscosity of bio-oils produced from different feedstocks though MAP was lower than the light fuel viscosity of 4 cSt (Mohan et al., 2006), the heavy fuel oil viscosity of 50 cSt (Czernik and Bridgwater, 2004), the US #4 fuel oil viscosity of 5.5-24 cSt (Oasmaa et al., 2009), commercial automotive #2 diesel viscosity of 2-4.5 cSt (Islam et al., 2010), diesel viscosity of 0.011 Pa s (Thangalazhy-Gopakumar et al.,

> 100C ■ 80C 60C «50C «40C	20C
Canola
50 100 150 200 250 > 100C ■ 80C ■ 60C «50C «40C 20C
11 Comcobl—і 200 250
50	100	150
Fig. 3. Viscosity of bio-oils produced from different feedstocks at indicated temperatures

2010), and was higher than JP4 viscosity of 0.88 cSt (Chiaramonti et al., 2007) and gasoline viscosity of 0.006 Pa s (Thangalazhy-Gopakumar et al., 2010) at a temperature of 40°C. Considering the viscosity criteria (15 cSt at 35-45°C and 21.5 cSt at 30°C) presented by researchers (Pootakham & Kumar 2010a; Islam et al., 2010) for loading/handling and pipe transportation, the bio-oils from different feedstocks produced through MAP can be easy to load using existing petroleum loading equipments and easy to transport through pipe also. According to ASTM burner fuel standard, the bio-oil can have a maximum viscosity of 125 cSt at 40°C without filtering (Oasmaa et al., 2009). Considering this limit, the viscosity of the bio-oils used in this study had a much low viscosity and these bio-oils can be used as burner fuel.

Czernik and Bridgwater (2004) reported that the viscosity of bio-oil produced from wood would vary between 40 and 100 cP at 50°C, whereas the viscosity of the heavy fuel oil is 180 cP. As evident from Fig. 3, the viscosity of bio-oils produced from different feedstocks through batch or continuous MAP with or without catalyst had a significantly lower viscosity (1.5-2.2 cP) than the viscosity values reported by Czernik and Bridgwater (2004) and the viscosity of bio-oil (33 cP) from hardwood at 50°C (Tzanetakis et al., 2008). The viscosity of bio-oils used in this study was lower than that of bio-oil from sugarcane bagasse (12.1-28 cSt) measured at 60 and 80°C (Das et al., 2004). This result indicates that the bio-oils produced through MAP can be easily atomized. One possible reason for low viscosity of the bio-oils in this study was the absence of agitation and fluidization in MAP resulted in a clearer bio-oil (free from fine char/particles) than that of conventional pyrolysis.

The catalytic hydrotreatment reactions were carried out at three process severity levels, a mild hydrogenation at either 175 or 225 oC, a mild hydrodeoxygenation (HDO) at 225 — 275 oC and a deep hydrodeoxygenation. For the latter, samples from the mild HDO were first allowed to phase separate completely, after which the organic fraction (containing about 3 wt.% water) was treated at temperatures ranging from 350 oC in the first two reactor segments, to 400 oC in the last two.

1.1.1.2 Visual appearances of liquid phase after reaction

The catalytic hydrotreatment reaction at 175 oC resulted in a single phase oil with a visual appearance close to that of the original feed. Thus, at this temperature, phase separation does not occur. This may be related to the limited production of water at this temperature. The product has a considerable sweeter smell/odor than the original pyrolysis oil. The mild hydrogenation at 225 oC gives two liquid phases, an organic and a water rich phase. The water phase has a higher density than the aqueous phase. A similar situation was observed for experiments at higher process severities (mild HDO), see Figure 5 for details. The second stage HDO product oil has even a lower density than the aqueous phase.

The organic product yields for the various process severities are given in Figure 6. Here, the severity is expressed in terms of hydrogen consumption, and high severity is associated with high hydrogen consumption. The yield is a clear function of the temperature. A drop in the yield to about 40% is observed at about 200 oC due to the occurrence of phase separation and transfer of part of the carbon and oxygen to the aqueous phase. A further slight reduction in yield is observed at higher severities, presumably due to gasification reactions and further net transfer of components from the organic to aqueous phase.

Oxygen contents of the product oils are a function of the process severity, see Figure 6 for details. Phase separation between 175 and 225 oC results in a dramatic drop in the oxygen content. This is due to the loss of water and the transfer of very polar highly oxygenated components to the aqueous phase. At the highest severity, the oxygen content is about 15%, compared to about 40% for the original pyrolysis oil.

The hydrogen consumption ranges between 65 and 250 Nm3/1 pyrolysis oil. Higher process severities lead to higher hydrogen uptakes (Figure 6).

A useful representation to assess the changes in the elemental composition of the product oils at various process severities is a van Krevelen diagram. Here, the ratio between O/C and H/ C of the products are plotted together in a single diagram. In Figure 7, a typical plot is provided for selected literature data on pyrolysis oil hydroprocessing (Elliott, 2007; Venderbosch et al., 2010) and our results with Ru/ C at different severities. Presented here are data points from e. g.:

— wood and pyrolysis oil, and for the four cases referred to in this paper (HPTT, hydroprocessing at 175 and 225 oC, Mild HDO and 2nd stage HDO);

— A selection of data points derived from literature studies (Baldauf et al. 2007; Churin et al., 1988; Conti, 1997; Diebold, 2002; Kaiser 1997; Samolada et al., 1998). Some of these data are derived from various oils from a variety of resources and processed in different reactors, different catalysts and at different conditions.

The plot also contains curves to represent the changes taking place in elemental composition during hydroprocessing, a theoretical curve for the dehydration of pyrolysis oil, and trend lines for the thermal (HPTT) route and hydroprocessing routes based upon the experimental data points.

Based on our work on the Ru/C catalysts and supported by the literature points in Figure 7, several reaction pathways can be distinguished:

a. Essentially repolymerisation of the pyrolysis oil (no catalyst, no hydrogen, ‘HPTT’);

b. Merely hydrogenation of the pyrolysis oil at mild conditions (up to 250oC, with catalyst and hydrogen, referred to as mild hydrogenation),

c. Dehydration of the oil at temperatures near 250-275 oC, and

d. Hydroprocessing of pyrolysis oil at temperatures up to 400 oC

Upon thermal treatment, the principal reactions are rejection of oxygen as water. Some CO2 and CO is released as well, which shifts the trend line to slightly higher H/ C ratios (but decarboxylation / decarbonylation is limited to approx. 10 wt. % of the feed). A high conversion (i. e. at high temperatures and residence times) eventually leads to a hydrogen — depleted solid material (and probably similar to conventional carbonisation processes, charcoal).

О

To obtain a liquid product with a higher H/C ratio, additional hydrogen is thus required. This path is shown in Figure 7 and includes the mild hydroprocessing step, at around 175 oC (no phase separation) and 225 oC (phase separation), followed by further hydrodeoxygenation (and hydrocracking).

Jones (1960) expressed the density-pressure data of compacted powder in the form of equation 1.

ln(p) = mln(P) + b (1)

where, p is bulk density of compact powder mixture, kg/m3, P is applied compressive pressure, MPa; m and b are model constants.

The constants b and m are determined from the intercept and slope, respectively, of the extrapolated linear region of the plot of ln(p) vs ln(P). The constant m has been shown to be
equal to the reciprocal of the mean yield pressure required to induce plastic deformation (York and Pilpel, 1973). A large m value (low yield pressure) indicates the onset of plastic deformation at relatively low pressure, thus, the material is more compressible.

Kinetic parameters are evaluated using equation (5) at a heating rate of 10, 30 and 50 °С/min. Figures 6-9 shows the plot of lnpi against 1/T at different mass fraction reacted (a) for cellulose, lignin, EFB and PS. The higher lnpi values shows high heating rate (50 °C/min) followed by 30 and 10 °C/ min. Kinetic parameters evaluated at each a are given in the Tables 4-7. For pure cellulose and lignin, the kinetic parameters are determined at a=0.1 to 0.8 and a=0.1 to 0.6, respectively. EFB and PS kinetic parameters are evaluated at a=0.1 to 0.7. Among all samples, pure lignin produced highest residual fraction and hence kinetic parameter determined up to a=0.6. The correlation coefficients (R2) determined are higher than 0.991 for all cases.

The kinetic parameters for EFB, PS, pure cellulose and lignin are determined as listed in the Table 8 and compared with experimental works reported. The average activation energy of cellulose is 160 kJ/mol and pre-exponential factor is 1.2*1013 m-1 for first order kinetic model. Several researchers found that first order kinetics fit well for cellulose decomposition reaction (Varhegyi et al., 1997; Gronli et al., 1999; Hu et al., 2007).

Activation energy evaluated for cellulose in the present study is in good agreement with the work reported by Zhang et al. (2009) for a first order kinetic model. Nevertheless, the value is comparatively low corresponding to values reported by Varhegyi et al. (1997), Gronli et al. (1999) and Yang et al. (2004). The reason may be different source of cellulose and different method followed by the authors. Secondly, this may be due to relatively high heating rates used in the present study. Meanwhile, Hu et al. (2007) reported high activation energy (233 kJ/mol) using Flynn-Wall-Ozawa method at low heating rates of 2.5, 5, 10 °С/ min. Gronli et al. (1999) observed the effect of different heating rates on activation energy for cellulose and found low activation energy at high heating rates. Similarly, much lower activation energies were found by Milosavljevic et al. (1995) at high heating rates. The same effect was also observed for pre-exponential factor. This is due to the heat transfer limitation between sample particles and the surroundings at high heating rates. Varhegyi et al. (1997) suggested utilization of low sample mass to minimize the heat transfer limitations at high heating rates.

Pure lignin is decomposed with average activation energy of 175 kJ/mol and pre­exponential factor of 2.06×1015 m-1 for first order kinetics model. These kinetic parameters are in good agreement with work reported by Murugan et al. (2008).

EFB and PS are decomposed with the average activation energy of 151 and 199 kJ/mol and reaction order of 5 and 5.3, respectively. The corresponding average pre-exponential factors evaluated for EFB and PS are 1.4×1015 and 1.1×1017 m-1, respectively. These values are somehow larger than those reported by Yang et al. (2004) for first order reaction kinetics. Luangkiattikhun et al. (2008) and Guo & Lua, (2001) observed lower activation energy and pre-exponential factor based on single-step nth order and first order kinetic model for PS. However, in these studies, comparatively high values were obtained using two step kinetic models.

Based on kinetic parameters, it is easy to decomposed EFB as compared to PS, pure cellulose and lignin. The order of decomposition from fast to slow is EFB > cellulose > lignin > PS.

4. Conclusion

A biomass decomposition study has been carried out to investigate different breakdown region and kinetic parameter evaluation for EFB and PS. As major components of biomass, pure cellulose and lignin decomposition kinetics were also studied. TG and DTG curves were studied in detail to understand the major decomposition region in EFB, PS, pure cellulose and lignin. The kinetic parameters of EFB and PS are found to be higher compared to reported values in the literatures. This difference may be due to the different methods for kinetic parameter determination and relatively high heating rates used in the present study. Based on the kinetic parameters, PS was difficult to decompose as compared to EFB. The possible reason is may be relatively high lignin content present in PS. This high lignin content was also responsible for low decomposition rate of PS as compared to EFB. Pure lignin had the lowest decomposition rate among all the species. Moreover, lignin content in PS was decomposed at high temperature as compared to EFB based on higher heating rates.

5. Acknowledgments

The author thanks the Petroleum Research Fund (PRF) of PETRONAS and Universiti Teknologi PETRONS for their financial support.

The energy consumption of chipping, transportation and drying is as follows:

a. Chipping: The energy consumption of the chipping process is due to electricity and diesel. The specific units of energy consumption are 13.6 kWh/material-t (122.4 MJ/material-t) and 1.23 L-diesel/material-t (43.7 MJ/material-t), respectively (Hashimoto et al., 2000).

b. Transportation: The chopped biomass materials are delivered to the plant by 10 ton diesel trucks. CO2 emissions and/or energy intensities on a given transportation run would be affected by the weight of biomass materials. That is, the weight of which the materials can be carried is restricted to bulk density. We measured the bulk density (=0.14 t/m3) in the atmosphere. The bulk density is dependent upon the moisture content. Thus, assuming that the bulk density is at a moisture content of 15 wt. % (p15 ), the bulk density pMC at any moisture content (MC wt.%) is

_ 0.85

PMC _ 1-MC P15

Next, the loading platform of 10t-trucks is to be approximately 24.7 m3 (Suri-Keikaku Co. Ltd., 2005). Consequently, even a truck with 10 ton’s volume cannot always carry that in full weight. Here, CO2 emissions and/or energy intensities are assumed to be due to the fuel consumption of truck, which is indicated as a function of the loading rate of weight. That is, using the loading rate of X, the fuel consumption rate of a 10t — truck fFC (X) is

fFC (X) _ aX + b (14)

where, a(=714 g-CO2/km) and b(=508 g-CO2/km) are constants on the fuel consumption of the truck (Dowaki et al., 2008b).

The definition of the loading rate of X is as follows: Assuming that the plant scale is Ps dry-t/ d, and that the annual operating time is 300 days, the annual material balance on the feed materials is 300Ps t-dry/yr. Since the throughput per year at MC wt.% is 300Ps/(1 — MC), the total number of transportation by 10 t trucks at MC wt.% (Nmat) is

Where, [a] is represented as the maximum integer, so as not to exceed a. Thus, the average loading rate of a 10 t-truck (Xave) is

Providing the average loading rate, and multiplying fFC (X) by the transportation distance and the specific CO2 emissions or the energy consumption of diesel, we can estimate CO2 emissions or fuel consumption in the transportation sub-process. In this paper, the transportation distance is the range between 5 (Distmin) and 50 km (Distmax ), because the wooden materials in Japan are distributed widely. That is, it is assumed that the feed materials are collected within a radius of 50 km.

c. Drying: Next, on the sub-process of drying, the energy consumption was estimated under the condition that the moisture content of feed materials would decrease to 20 wt. %. Here, assuming that the initial moisture contents are from 20 (MCmin) to 50 wt. % ( MCmax the raw materials are dried by a boiler. Also, the auxiliary power of a pump in a boiler is included in the energy consumption of the sub-process. The operational specification of a wood-chip dryer (boiler) is the energy efficiency of 80 %, and the auxiliary power of a pump of 0.195 kWh/t-water (1.75 MJ/t-water). Note that the moisture content of feedstock can be reduced by the hot exhausted gas to some extent.

d. Monte Carlo simulation on the uncertainties: As the above, in this paper, we estimated CO2 emissions and/ or energy intensities, considering the uncertainties of the transportation distance and the moisture content. In this paper, the following two uncertainties of the distance and the moisture content were considered by the Monte Carlo simulation.

That is, the uncertainties on the transportation distance ( Dist km) and the moisture content (MC wt.%) are represented by uniform random numbers Rndi between 0 and 1 in Eqs. (17) and (18). Note that Rnd1 and Rnd2 are independent and identically distributed.

Dist = Ch’s^in + Rnd2 (Distmax — Distmn ) (17)

MC = MCmin + Rnd1 (MCmax — MCmin ) (18)

An iteration count in the simulation was executed up to 10,000. The range within a 95 % significant level was adopted as the uncertain data on the distance and the moisture content, in order to estimate CO2 emissions. In this case, the gross distributions on CO2 emissions would be normal distributions.