Determining the kinetics of sunflower hulls using dilute acid pretreatment in the production of xylose and furfural

2.1 Source of sunflower hulls

The raw sunflower hulls were obtained from Dahlgren & Company, Inc. (Crookston, MN, USA). The sunflower seeds are passed through the seed mill where seeds open up. To separate the mixture of seeds, hulls were dropped in to water. The hulls will float on the water due to density difference and are removed easily. The separated hulls were air dried. The size of sunflower hulls was approximately 6–8 mm. Moisture content of the raw sunflower hulls was determined by oven drying at 105°C for 12 h. It was around 5 wt%.

2.2 Compositional analysis

The major hull components measured as dry wt% were cellulose (34.1±1.1), lignin (25.3±0.9), xylan (21.2±1.5), galactan (3.5±0.6), arabinan (0.5±0.1) and extractives (13.2±2.5), which were primarily composed of waxes, proteins, nitrates and nitrites. The amount of ash present in sunflower hulls was approximately (0.4±0.1) wt%. The initial xylan percentage corresponds to 4.2±0.8 g for 21 g of hulls loading for pretreatment.

2.3 Pretreatment of hulls

The hull pretreatment was performed in a 300 ml internal volume jacketed batch reactor manufactured by Autoclave Engineers, Erie, PA, USA. The reactor was made of Hastelloy C-276 to mitigate acid corrosion at high temperatures. Dry biomass (21 g, 10% w/w) was added to the sulfuric acid solution (prepared in 0.5, 1.0 and 2.0% w/w concentrations from deionized water and sulfuric acid). The heating source was saturated steam drawn into the external jacket of the reactor by a three-way valve. The agitation speed was maintained at a constant 60 rpm throughout the reaction. The reactor heating rate was 35±3°C/min. After the desired temperature was achieved, the reaction time commenced; the temperature in the reactor was then maintained constant. When the allotted reaction time was reached, the reactor was cooled by passing tap water into the external jacket. Once the reactor was cooled below 40°C, the reaction slurry in the reactor was discharged and collected in a polyethylene bottle for analysis. The temperature data from the reactor were recorded with PicoLog software throughout the reaction time. All the experiments were duplicated. More detailed information regarding the reactor and process setup is published elsewhere [11].

The pretreatment reaction conditions were selected to approximate those that the current National Renewable Energy Laboratory (NREL) process designs for biochemical conversion of lignocellulosic biomass to ethanol [11]. According to the report, the described NREL process design uses a temperature of 158°C in a dilute-acid pretreatment batch reactor. Large amounts of poorly fermentable oligomers are formed at lower reaction temperatures and acid concentrations, whereas significant further degradation of C₅ products usually occurs under higher severity conditions. Hence, we have chosen to perform this study between 140°C and 160°C; the operational conditions are listed in Table 1. The acid concentration was varied between 0.5 wt% and 2.0 wt%. Each pretreatment experiment was performed up to a maximum reaction time of 30 min. The liquid hydrolyzate samples were withdrawn every 5 min.

2.4 Analytical procedures

Pretreated slurry samples were vacuum-filtered and collected as liquid hydrolyzates and solid substrates. The liquid hydrolyzate samples were analyzed for xylose and furfural production. This analysis was performed based using the NREL analytical procedures (NREL/TP-510-42623). The quantitative analysis for determining monosaccharides present in liquid hydrolyzates was performed by Agilent 1200 HPLC with a 300×7.8 mm Transgenomic CHO-Pb column (Omaha, NE, USA). All samples were replicated during HPLC analysis in order to obtain precision. The mobile phase used for analysis was deionized water with a flow rate of 0.6 ml/min [12]. Prior to the analysis of pretreated hydrolyzate samples, a set of calibration standards was run to validate the HPLC refractive index detector (RID). The concentrations of standards ranged from 0.5 to 18 g/l. In addition, an internal sugar recovery standard with a concentration of 4.0 g/l was run frequently (every 8 injections) to test for the peak height and RID validity. The standard solutions and sugar recovery standard solution consisted of D-(+) glucose, D-(+) xylose, D-(+) galactose, L-(+) arabinose and D-(+) mannose.

Furfural was analyzed using an Agilent 1200 HPLC with a Phenomenex Rezex RFQ 100×7.8 mm column (Torrance, CA, USA). The 0.01N sulfuric acid mobile phase with a flow rate of 1.0 ml/min was used for analysis [12]. The calibration standards for furfural were obtained from Absolute Standards, Inc. (Hamden, CT, USA).

The xylan retained in the pretreated solid substrate of the biomass was measured based on the published protocol for analysis of structural carbohydrates and lignin in the pretreated substrate (NREL/TP-510-42618).

2.5 Kinetic model and data analysis

The models repeat used in the literature are based on the assumption of pseudo-homogeneous irreversible first-order reactions. The first successful model was proposed by Saeman [13], designed originally for sulfuric acid pretreatment [14]. This model is applied to hydrolysis of polymeric hemicellulose and is based on Eq. (1).

where k₁ is the rate of the monosaccharide generation reaction and k₂ is the rate of the decomposition reaction (min^-1). The polymer in Eq. (1) is xylan, the monomer is xylose and the decomposition product is furfural. Solving the differential equations, the following model predicts the concentration of monomers and degradation products:

Since

where P_o denotes the initial xylan, P is the xylan left over after the pretreatment, M is the xylose monomer and D is furfural in the given application of Eq. (1); brackets reflect concentrations measured in wt%.

In modeling hulls’ hydrolysis, we ignored the occurrence of xylose oligomers, i.e., shorter chain xylan polymers that are usually formed prior to xylose monomers from xylan. The reason for not incorporating xylose oligomers into modeling is that models tend to overestimate the xylose oligomers formation [15]. Moreover, the amount of oligomer concentration in the liquid hydrolyzate samples was very low, as shown in Table 2.

Eqs. 2–5 were fitted with the corresponding data sets for each of the applied temperature/acidity conditions using the Lavenberg-Marquardt non-linear curve fitting method in Mathcad 15 (Needham, MA, USA). The fitted parameters are the two kinetic constants involved in the proposed kinetic mechanism. The kinetic constants, k_i, are functions of absolute temperature and acid concentration according to the following Arrhenius expression [16]:

where T is absolute temperature (K), C is the acid concentration (wt%), A_o is a pre-exponential factor (min^-1), n_i is a dimensionless acid concentration exponent, E_i is the Arrhenius activation energy (kJ/mol) and R is the universal gas constant (8.3143×10^-3 kJ/mol-K).

3.1 Effect of acid concentration and reaction temperature on xylose and furfural yields

Table 3 lists the concentration of xylose produced from xylan as a function of time. The dynamics of xylan hydrolysis is similar to that observed with other feedstocks; namely, the xylose concentration increases and then declines, with a concomitant increase of the furfural concentration as evident from Table 4 [14].

From the Table 3 data it is evident that the amount of xylose observed was rather low at lower reaction temperatures and acid concentrations, even at longer reaction times. Higher xylose yields were obtained at higher acid concentrations, even at lower temperatures. The effect of acid concentration was thus found to be more pronounced than that of reaction temperature.

Table 4 lists the furfural concentrations recovered as a result of xylose chemical dehydration. Low furfural yields were obtained at lower acid concentrations, even at higher reaction temperatures. However, using higher acid concentrations led to higher furfural yields. Two other observations concerning threshold xylose concentrations can be made based on Table 4 data. First, furfural formation was observed only when the xylose concentration exceeded 5.8±0.7 g/l that corresponds to 24.0±1.1 wt%. Second, the xylose concentration never exceeded 13.7±1.1 g/l corresponding to a 57.1±0.7 wt% yield, followed by a decline with a concomitant furfural formation.

3.2 Model justification

Figure 1 shows the experimental data along with the simulation produced using the time-dependent expressions of Eqs. (2)–(5). The best fitted kinetic constants k₁ and k₂ for the proposed model [Eq. (1)] are listed in Table 5. The kinetic constants were higher for xylose monomer than for furfural formation as expected, because a similar pattern (faster xylose formation followed by its slower decomposition to furfural) was observed earlier for any other crop considered [15]. The rate increased for both xylose and furfural formation with the increase of reaction temperature and acid concentration, which was also expected based on the literature analysis [17].

Figure 1

Model and experimental data for xylan, xylose and furfural for sun hulls: (A) pretreated at 140°C at 0.5 wt%, 1.25 wt% and 2.0 wt% acid concentrations for hulls; (B) pretreated at 150°C at 0.5 wt%, 1.25 wt% and 2.0 wt%; and (C) pretreated at 160°C at 0.5 wt%, 1.25 wt% and 2.0 wt%.

From Figure 1 it is evident that the fitted parameters predicted the experimental data reasonably well. The only exception was the furfural formation at 160°C. As can be seen in Figure 1C, the model under-predicted the furfural formation at this highest temperature used, particularly for the highest acid concentration, 2.0 wt%. These effects can be explained by subsequent reactions of furfural decomposition, which are more pronounced at the highest severity conditions [18–20].

To further justify the model used, the observed reaction orders, n_i (Table 6) were replaced with the kinetically relevant integers (0, 1, 2) in Eq. (6); then, the model was run with these artificially set values. As a result of this treatment, the model lost its predictive power; furthermore, in most of the cases, the activation energies obtained in such a way turned out to be negative, thus contradicting the experimentally observed trend (Table 5).

Model testing is often performed by comparing the R² values obtained by least square fitting; however, exponential kinetic data may be skewed as a result of linearization. Hence an F-test was performed instead, comparing the experimental data with those generated by the theoretical model by varying the pre-exponential factor, activation energy and dimensionless reaction order [21]. Table 7 suggests that the experimental data fitted the model accurately, as the sums of squared errors (SSE) values were low. The differences between the experimental rate coefficients and those generated by the model were low as the sets passed the F-test (F>F_critical). Thus, the model applied can be deemed adequate, despite the inherent heterogeneity of the system used.

3.3 Reasons for a relative recalcitrance of sunflower hulls

The amount of initial xylan present in hulls is 21.2 as evident from Table 8. The comparison of runs conducted under varied conditions shows that the amount of xylan hydrolyzed was lower whenever the acid concentrations were lower. More than 50 wt% of the initial xylan was still retained in hulls at a 0.5 wt% acid concentration at 140°C. By contrast, almost 80 wt% of the initial xylan was hydrolyzed at 160°C with a 2 wt% acid concentration. Compared to other lignocellulosic biomasses, such as forage sorghum, kenaf and sunn hemp, pretreated under similar conditions, xylan hydrolysis was significantly less pronounced for hulls [17, 22].

This difference could be due to a unique cell wall structure specific for sunflower hulls. It consists of a black pigmented layer with a high wax content. Besides this specific layer, the sunflower hull cell wall features a higher lignin content compared to most of the other crops, as seen from Table 8. The presence of a wax layer at the surface of hulls is to protect the seeds against mold by repelling water [23]. This wax/lignin barrier may hinder the access of hydronium ions to xylan, resulting in both a lower effective reaction order on the acid, n₁, and higher activation energy for xylan hydrolysis. As a result, the process occurs under lower effective acid concentrations than set by the bulk acid concentration. Corroborating this assumption, the values of n_i and E_i obtained in this study are similar to those obtained for other crops, such as aspen, balsam, or switch grass, at lower acid concentrations [15].

To obtain higher xylose yields during acid pretreatment, one practical recommendation would thus be subjecting hulls to a prior ethanol or other organic solvent extraction to dissolve the waxes. An alternative would be a de-lignification prior to xylan hydrolysis, e.g., an alkaline pretreatment.