2.1 Raw materials

The biomass used in this study was beech wood. The chips used were provided by the company Mallo wood (Reguisheim, France). Beech is a hardwood species, which is widespread in Europe. The wood was obtained as saw dust and it was used without any further grinding for pyrolysis, the particle size ranged between 0.1 and 1 mm with a mean size close to 0.3 mm. Beech wood char was obtained by heating the different samples in a tubular furnace from room temperature to 800 °C at 10 °C/min under argon at a flow rate of 25 NL/h. The sample was kept at this temperature for 1 h before cooling down to the room temperature. The final char particles were similar in size to the pristine wood.

2.2 Preparation and characterization of metal-impregnated biomass and blank samples

The metallic precursors used for impregnation were potassium nitrate (KNO₃) and silicon dioxide (SiO₂). The sample was impregnated with potassium to obtain 0.5 wt%, 1 wt% and 2 wt% K-impregnated samples and with silicon to obtain 0.5 wt%, 1 wt% and 2 wt% Si-impregnated samples. These impregnated chars were called K0.5, K1.0, K2.0 and Si0.5, Si1.0, Si2.0, respectively.

The choice of potassium nitrate for the impregnation is explained by the fact that it provides no additional carbon atoms and therefore no change in the carbon balance at the end of conversion. Also, nitrate can be easily decomposed by a simple increase of temperature. On the other hand, the impregnation with Si was carried out with colloidal silica Ludox 40% wt. This is a suspension of silicon in water. The choice of this precursor was made to prevent any further intrusion of inorganic elements in the composition of biomass. Indeed, by using sodium, calcium or potassium silicate, there would have been inorganic intrusion in the sample of biomass studied and the effect of silicon could not be investigated correctly.

The impregnation with inorganics was inspired by the methodology used in the study by Hwang et al. [24]. However, no acidic treatment was performed in order to avoid any damage of the structure of the char which could influence char reactivity. The impregnation was carried out on the char obtained after heat treatment of the beech sample. A mixture containing 8 mL of anhydrous absolute ethanol and 2 mL of distilled water was prepared and a defined amount of the precursor was added. 500 mg of the char was immersed in the mixture. Each mixture was then placed on a magnetic stirrer at room temperature until evaporation of the solvents. After stirring, the sample was neither filtered nor dried in an oven. The impregnated char was recovered to be characterized. A witness sample was prepared by following the same protocol in an impregnating solution containing no precursor.

2.3 X-ray fluorescence

The amounts of inorganic elements in raw and impregnated samples were measured by using an X-ray fluorescence spectrometer in order to investigate the content in inorganic elements in chars and to verify if they had been correctly impregnated on chars. The spectrometer was of type PHILIPS MAGIX series with a ferryman PW2540 sample changer.

2.4 Textural characterization of raw and impregnated chars

Surface areas were determined from nitrogen adsorption manometry at 77K and isotherms were obtained using an ASAP 2420 apparatus (Micromeritics). Each adsorption measurement was out-gassed overnight at 300 °C prior to analysis. The BET model was applied to the N₂ adsorption data at relative pressures of 0.01–0.05. The sample pore volume was obtained from the amount of N₂ adsorbed at P/P⁰ = 0.95.

2.5 SEM analyses

Scanning electron microscopy (SEM) was used to investigate the sample morphology. SEM was coupled with EDX (energy dispersive X-ray spectroscopy). The microscope was of type SEM XL30 FEG. The software used for EDX analysis was Oxford INCA Energy with a detector Si (Li). The analyses were done in low vacuum mode without gold coating of the samples.

2.6 Char reactivity measurements

Char samples reactivity was investigated by thermogravimetric analysis (TGA) using a thermoanalyzer SENSYS evo TG-DSC (SETARAM). About 8 mg of char was placed in an alumina crucible of 0.1 mL and the temperature was measured by using a thermocouple placed near the sample. The sample was heated from 25 °C to 800 °C with a heating rate of 24 °C/min under N₂ and then, after stabilization of temperature and mass, the gas was switched to a mixture of 80% N₂ and 20% CO₂ at a 12 NL/h flow rate. Due to the particle size, no diffusion limitation was observed inside the bed.

From TGA measurements, the conversion level at a given time t was obtained according to the equation:

3.1 Composition and texture of impregnated samples

Fig. 1 presents the results obtained after X-ray fluorescence analysis of the raw char and the impregnated samples with K and Si. As expected for wood (see for instance Vassilev, 2010 [25] and Dupont, 2010 [26]), Fe, Al, P and Mn were found in low quantities in the pristine and impregnated samples while Mg and Ca were found in significant concentration. The concentrations of these six elements were not significantly modified by the different impregnation processes. The data showed that the saturation in the sample impregnated with potassium was reached. Indeed, the K content of the impregnated sample at K1.0 and K2.0 was substantially equal to a concentration of 0.4 wt%. This may be due to the saturation of the surface of the char because it has been covered by a monolayer of catalysts [14]. Saturation could also be explained by the microporous texture of the pristine char sample. The nitrogen adsorption isotherm of beech char was of type 1 (Fig. 2), and showed that the samples exhibit very narrow micropores below 1 nm. It is shown in Table 1 that for K1.0 and K2.0 samples, the major fraction of the pores is plugged, and the surface area decreases from 262 to 30 m²/g.

For samples impregnated with silicon, a clear difference was observed between the contents of Si at different levels of impregnation indicating that saturation was not reached for this element. The textural properties presented in Table 1 clearly show that the porosity of chars was only slightly affected by the impregnation process. This allows us to make the amount of silicon added to the pristine beech char vary in significant proportions. Then it could be assumed that the silicon content was the only property modified in these chars compared with the pristine beech char.

3.2 Dispersion of elements in the char structure

As shown in Fig. 3, as expected for wood char (reference), macropores are clearly visible. The surface area of the char is 263 m²/g, which means that the walls of these macropores exhibit some microporosity. EDX mapping of raw char confirms that potassium is well dispersed on the entire matrix. This is not the case for silicon which is concentrated in small micrometer-sized inclusions. Calcium was another catalytic element present in high amounts in the pristine char. The EDX map showed that, as for silicon, calcium was located in micrometer-sized particles, probably as calcium carbonate or as calcium hydroxide. It must be noticed that oxygen mapping showed higher concentrations on the locations of calcium and silicon particles (data not shown). This confirms that calcium and silicon elements are in oxidized form in the pristine char.

After impregnation with potassium, the macrostructure of char was not significantly modified (Fig. 4). Potassium is still well dispersed in char although its concentration has increased from 0.22 to 0.33 wt%. From this EDX analysis and the evaluation of textural parameters after impregnation by KNO₃, we can conclude that potassium forms a layer on the external surface of the char which plugs the micropores of the carbon material.

A similar analysis was done on the char impregnated with silicon Si-1 (Fig. 5). The silicon is also well dispersed on the external surface of the char, but micrometer-sized particles can be detected. These particles should be related to the pristine silicon content of the char (see Table 1). The SEM image in Fig. 5 shows a major difference between K-chars and Si-chars: in the latter case silicon is under the form of very small particles which are dispersed in the macropores of the char. These particles have a size of a few micrometers and do not plug the micropores of the carbon material. This fact explains that the surface area does not significantly decrease after impregnation. As for the pristine char, the analysis of oxygen showed that silicon was present as silica.

A comparison of the pristine char (Fig. 3) and the impregnated ones (Figs. 4 and 5) shows that we are able to obtain a dispersion of K and Si which is very similar to the natural one with our impregnation process. It has been observed nevertheless that beside dispersion, the mobility of added catalysts may be lower than that of naturally occurring catalysts [27]. This may lead to differences between pristine and impregnated chars since the mobility of potassium is a key point for its catalytic activity and for its release at high temperatures.

3.3 Catalytic effects of metal on char CO₂ gasification

For the pristine beech char, the profile of reaction rate as a function of conversion exhibits a shape which was already observed in previous studies on CO₂ gasification under isothermal conditions [14]. The reactivity increases up to 50% of conversion due to an increase of accessible surface area, while the subsequent decrease can be attributed to the collapse of pores or the deactivation of potential catalysts. K and Si are present in the pristine char, and the K/Si ratio is 2.0, as shown in Table 1. Hognon et al. [22] found that for K/Si values higher than 1.0, an activation occurred at high conversion ratios during steam gasification of biomass chars. This shows that the quantitative influences of catalysts and inhibitors on gasification reactions with CO₂ and H₂O are different.

The curves of gasification rate r_t versus conversion for samples impregnated with K and Si are shown in Figs. 6 and 7, respectively. As shown in Fig. 6, for samples impregnated with potassium two successive behaviors were observed:

At low conversion (X_t < 60%), no influence of the potassium concentration could be observed on the rate of the reaction in the range of potassium concentrations tested. Indeed, the difference of reactivity between samples with high K content and those with lower content is not significant. As in many similar experiments, an increase of the reaction rate is observed when the conversion increases up to 50%, which is related to the development of open porosity inside the char structure. Then, the increase of external surface area leads to an increase of reaction rate. The values of r_t are similar for all samples although the BET surface areas are far lower for samples K1.0 and K2.0 compared to the two other chars (raw and K0.5). This behavior shows that the internal surface is not accessible for CO₂, the nitrogen adsorption isotherms were of type 1 and they showed indeed that the micropore size is mainly below 1 nm. Then, we can assume that diffusion limitations would occur for raw and K0.5 samples and only the external surface area is used for gasification. The plugging of micropores for chars K1.0 and K2.0 has then no influence on the gasification rate. This diffusion limitation for CO₂ gasification of biomass chars was already observed for some particle sizes and gasification conditions [28]. In our case, it is strongly related to the very small size of the micropores, which are mainly ultra-micropores (i.e. below 0.8 nm).

At high conversion (X_t > 60%), reaction rates of the samples are significantly different depending on the potassium content of the sample. For the two samples with the lowest potassium content (raw and K0.5), the reaction rate decreases with conversion. This behavior was observed in many cases and may be related to the collapse of pores for X_t > 50% which would result in a decrease of the external surface area. For the two samples with the highest potassium content (K1.0 and K2.0), the catalytic effect would lead to a continuous increase of reaction rate until the conversion ratio reaches 95%. Several publications have related the reactivity of a set of different biomass to an alkali index, but the correlations are complex since many elements found in the samples may have activating or inhibiting influences [15,16,18]. In our case, only the potassium content was modified in the chemical analysis of the different chars tested, and a correlation was observed between the potassium content and the char reactivity.

The effect of Si content on the rate of the gasification reaction was also examined in Fig. 7. It could be seen that whatever the silicon content, the global shape of the gasification profile was similar to the one of pristine beech char. Two different behaviors could be observed for low and high conversion ratios:

• • For X_t < 60%, the influence of Si concentration on char reactivity is low. This may be related to the fact that the dispersion of silicon inside the char structure is different from the one of catalysts like potassium and calcium. Therefore, reactions between these elements would not be favored.
• • For X_t > 60%, the major inhibiting influence of Si could be observed, even for the Si0.5 sample. It must be noticed that even for this char, the K/Si mass ratio is far lower (0.7) than that for the beech char (2.0). As pointed out in previous studies, this kind of catalyst/inhibitor parameter leads to good correlations with biomass char reactivity [21,23]. The observed results on the effect of Si on the reactivity of chars under CO₂ are in line with what has been mentioned in the literature in a recent study on the reactivity of chars under H₂O by Hognon et al. [22]. The authors explained this effect by assuming the encapsulation of K by Si.

The theory of encapsulation of alkaline catalysts by Si during the combustion phase was also discussed in the study of Sippula et al. [29]. The mechanism of this phenomenon can be the formation of alkali metal silicates by chemisorption of alkali by silicates. The addition of potassium to form silicate is one of the mechanisms which lead to slagging in biomass fired boilers [27]. This phenomenon is strongly related to the high mobility of potassium which allows it to come in contact and react with less mobile silica.

It must also be noticed that char reactivity also depends on its calcium content, and it was also observed that this element also tends to bind to silicon to form silicate [30]. The influence of this element on char reactivity may be complex: it could also participate in the decrease of char reactivity with increasing Si content through the capture of calcium. Calcium can also limit the silica available for potassium capture, thus limiting this decrease of reactivity [31].

By combining all the samples used in this study, a general tendency for the gasification rate as a function of X_t and K/Si ratio can be obtained. Fig. 8 shows the reactivity for conversion higher than or equal to 50%. The values below 50% were not shown since only small variations were observed. A more general correlation should include the variation of textural parameters from one sample to another. However, we have shown, in our case, that diffusion limitations occurring in narrow micropores hid this influence. Therefore, a direct comparison of reactivities is meaningful. Although the reactivities for X_t = 90% exhibit a larger spreading, it could be seen that for K/Si below 3 (in mass), reactivity strongly decreased with increasing conversion, while a significant increase could be observed for K/Si > 3.