NC State
Wang, S., You, T., Xu, F., Chen, J., and Yang, G. (2015). "Optimization of [Amim]Cl pretreatment conditions for maximum glucose recovery from hybrid Pennisetum by response surface methodology," BioRes. 10(4), 7021-7037.


Because of a complex chemical ultra-structure of lignocellulosic biomass, pretreatment is a necessary step for its conversion into bio-ethanol. In the present study, pretreatment conditions using the ionic liquid (IL) 1-allyl-3-methylimidazolium chloride ([Amim]Cl) were optimized for a relatively new model energy crop, hybrid Pennisetum (P. americanum × P. purpureum) to maximize the yield of fermentable sugars (glucose). The design of experiment programs employed a central composite design (CCD), with variables of temperature (102 to 187 °C), retention time (0.5 to 5.5 h), and solids loading (2 to 15 wt%). These factors were further optimized using response surface methodology (RSM). The proposed quadratic model to predict the glucose recovery from hybrid Pennisetum was verified by variance analysis (ANOVA). The model displayed high F and R2 values, indicating that it could be successfully used to identify the relationship among the independent variables studied. A maximum glucose recovery of 72.2% was found with temperature conditions of 139 °C, 2.97 h retention time, and 9.1 wt% solids loading.

Full Article

Optimization of [Amim]Cl Pretreatment Conditions for Maximum Glucose Recovery from Hybrid Pennisetum by Response Surface Methodology

Shengdan Wang,Tingting You,b Feng Xu,a,* Jiachuan Chen,a,* and Guihua Yang a

Because of a complex chemical ultra-structure of lignocellulosic biomass, pretreatment is a necessary step for its conversion into bio-ethanol. In the present study, pretreatment conditions using the ionic liquid (IL) 1-allyl-3-methylimidazolium chloride ([Amim]Cl) were optimized for a relatively new model energy crop, hybrid Pennisetum (P. americanum × P. purpureum) to maximize the yield of fermentable sugars (glucose). The design of experiment programs employed a central composite design (CCD), with variables of temperature (102 to 187 °C), retention time (0.5 to 5.5 h), and solids loading (2 to 15 wt%). These factors were further optimized usingresponse surface methodology (RSM). The proposed quadratic model to predict the glucose recovery from hybrid Pennisetum was verified by variance analysis (ANOVA). The model displayed high F and R2values, indicating that it could be successfully used to identify the relationship among the independent variables studied. A maximum glucose recovery of 72.2% was found with temperature conditions of 139 °C, 2.97 h retention time, and 9.1 wt% solids loading.

Keywords: Hybrid Pennisetum; Pretreatment; Ionic liquids; Optimization; Response surface methodology; Central Composite Design

Contact information: a: Key Lab of Pulp & Paper Science and Technology of Education Ministry of China; b: Beijing Key Laboratory of Lignocellulosic Chemistry, Beijing Forestry University, Beijing, China; *Corresponding author:;


Recently the miss-match between ever-growing energy demands and diminishing fossil fuel reserves has become increasingly prominent (Luque et al. 2008). Among the sustainable energy alternatives to fossil fuels, bio-fuels produced from biomass have been considered to be pre-eminent. Relative to traditional grain materials, lignocellulosic biomass (e.g., corn stalk, wheat straw, and switchgrass) is regarded as one of the most promising materials for the production of fuels (Jorgensen et al. 2007). As a relatively new candidate biomass, the herbaceous energy grass hybrid Pennisetum has attracted increasing research because of its various advantages, such as resistance to salinity and drought, suitability for marginal land quality, high productivity, low nutritional requirements, environmental benefits, and multipurpose uses (Bai et al. 1994; Chen et al. 2014).

Hybrid Pennisetum is one of the most prolific renewable cellulose resources, containing 36.04±1% cellulose, 20.11±1% hemicelluloses, and a relatively low content of lignin, 8.89±1% (Fan et al. 2012). The dissolution and subsequent hydrolysis of the cellulose are basic pre-requisites to produce bio-ethanol (Tadesse and Luque 2011). However, cellulose in the form of micro-fibrils with radii of 2 to 3 nm are enclosed within a rigid protective sheath of lignin and hemicelluloses to form a cell wall (Tadesse and Luque 2011; Brandt et al. 2013) that is intransigent to saccharification and hydrolysis efforts. This is a natural circumstance because the rigid protective sheath and complex structure of lignocellulosic biomass make it resistant to biological and chemical degradation (Himmel et al. 2007). Therefore, efficient pretreatment before saccharification is key to unlocking its intrinsic recalcitrance by increasing porosity to accelerate enzyme accessibility (Zhang et al. 2009).

Pretreatment is a primary stage in an efficient and economic conversion of lignocellulosic biomass into bioethanol. Quite a number of physical, biological, and chemical methods have been proposed (Yang and Wyman 2008). However, various issues remain. For instance, physical pretreatment suffers from relatively poor performance, high-energy requirements, and high capital costs (Hendriks and Zeeman 2009), whereas a controllable and sufficiently rapid system has not yet been found for biological pretreatment (Chandra et al. 2007). To date, chemical pretreatment is considered to be the most promising option (Yang and Wyman 2008). A few chemical solvents show solubility for cellulose, such as dimethyl sulfoxide/tetra-N-butyl ammonium fluoride; N,N-dimethyl acetamide/lithium chloride; and N,N-dimethyl formamide/nitrous tetraoxide (Wang et al. 2012). However, all of them have drawbacks, such as poisonous gas production, high toxicity, difficulty in down-stream recovery, and unrecyclability (Zheng et al. 2009). Compared to traditional cellulose solvents, ionic liquids (ILs) have low melting points, low volatility and toxicity, high thermal and chemical stability, and an environmentally benign nature. Therefore, they have attracted increasing research for the dissolution and subsequent hydrolysis of lignocellulosic biomass (Zhao et al. 2009; Sun et al. 2011). For example, 1-allyl-3-methylimidazolium chloride ([Amin]cl), 1-butyl-3-methylimidazolium chloride ([Bmim]Cl), and 1-ethyl-3-methylimidazolium acetate ([Emim]Ac) are able to effectively dissolve cellulose (Feng and Chen 2008). With the addition of an anti-solvent (water, ethanol, or acetone), the dissolved cellulose can be easily regenerated (Zhu et al. 2006). The regenerated cellulose exhibits significantly increased porosity and reduced crystallinity, which enhances its digestibility (Kuo and Lee 2009). Among different ILs, [Amim]Cl, which possesses an asymmetrical structure, shows better solubility for cellulose samples than ILs having symmetrical structures (Liu et al. 2012).

An efficient and economic pretreatment process is affected by factors such as pretreatment temperature, solids loading, retention time, particle size, biomass type, and moisture (Badgujar and Bhanage 2015). These factors have a complex interaction with each other; therefore, optimization of the pretreatment process is important for improving the performance of the hydrolysis of biomass. The traditional ‘one-factor-at-a-time approach’ is time-consuming. Furthermore, the interactions between independent variables are not considered. Response surface methodology (RSM) is a statistical technique for modeling and analysis of the interactions of designated independent variables on final response (Kleijnen 2008). It has been widely used in fields such as food process operations (Altan et al. 2008), new product development, biotechnology, media composition, and bio-processing (Xin and Saka 2008). Many types of lignocellulosic biomass, including sugarcane bagasse, wheat straw, oil palm frond, and sago waste, have been pretreated by ILs and optimized by RSM to achieve efficient and economic conversion of lignocellulosic biomass into fermentable sugars (Fu and Mazza 2011; Yoon et al. 2012; Lee et al. 2013). For IL pretreatment optimization, temperature, retention time, and solids loading are considered to have significant effects on biomass digestibility (Tan et al. 2011; Fu and Mazza 2011). However, the parameters of pretreatment processing of hybrid Pennisetum with [Amim]Cl have not been optimized, and are thus the focus of this study.

The objectives of the current study were as follows: (1) to evaluate and optimize the effect of [Amim]Cl pretreatment parameters (temperature, solids loading, and retention time) to obtain the maximum glucose recovery using RSM and a factorial central composite design (CCD); and (2) to develop empirical predictive models for glucose recovery from hybrid Pennisetum under [Amim]Cl pretreatment.


Materials and Chemicals

Hybrid Pennisetum was collected from an experimental field of the Beijing Academy of Agricultural Sciences (China), containing 33.64% cellulose, 18.05% hemicelluloses, and 11.03% lignin. It was milled with a FZ120 plant shredder (Truelab, Shanghai, China) and then sieved to 40 to 60 mesh. A comprehensive extraction process with toluene/ethanol (2:1, v/v) was performed in a Blst-250SQ Soxhlet apparatus (Bilon, Shanghai, China) for 5 h and oven-dried at 45 °C for 24 h. The extractives-free samples were stored at 4 °C in a sealed bag. The ionic liquid 1-allyl-3-methylimidazolium-chloride ([Amim]Cl) was purchased from the Lanzhou Institute of Chemical Physics (Lanzhou, China). Commercial cellulase (Celluclast 1.5 L) from Tricoderma reesei was purchased from Novozymes (China). All chemicals used were obtained from Sigma Chemical Co. (Beijing, China).


Design of experiments

Response surface methodology (RSM) is a collection of mathematical and statistical technique for designing experiments, developing regression models, analyzing effects of variables, and optimizing response curves (Ruangmee and Sangwichien 2013). In this study, the effect of three independent variables on the response was investigated using RSM and CCD. Three independent variables, i.e., pretreatment temperature (X1, 102 to 187 °C), retention time (X2, 0.5 to 5.5 h), and solids loading (X3, 2 to 15 wt%), were studied at five levels with three repetitions at the central point and two replicates at the axial and factorial points (Table 1). For each independent variable, the highest and lowest levels were selected based on the results achieved in preliminary tests. The central composite rotatable designed (CCD) matrix was generated with the aid of Design Expert 8.0.6 software (STAT-EASE Inc., Minneapolis, MN).

Table 1. Coded and Decoded Values for Each Variable of the Central Composite Rotatable Design

A total of 20 experiment designs were randomly performed and consisted of eight trials for factorial design, six trials for axial points, and six trials for repetitions of the central point. Three replicate confirmation experiments were conducted using the complete experimental design matrix shown in Table 2.

Table 2. Experimental Design Matrix Showing both Coded and Actual Values of Variables, as well as Observed and Predicted Responses

Ionic liquid pretreatment

The first step of the pretreatment process was dissolution of material by [Amim]Cl. Various solids loading amounts of hybrid Pennisetum(2 to 15 wt%) were prepared by mixing the shredded hybrid Pennisetum powder with [Amim]Cl in screw-capped Synthware tubes (Synthware, Beijing, China). The pretreatment was performed in a preheated oil bath with a stirring rate of 450 rpm at various temperatures and reaction times. The conditions for biomass dissolution are given in Table 2. At the end of the reaction time, deionized water (90 °C) was added as an anti-solvent to the hybrid Pennisetum/[Amim]Cl slurry with an agitation rate of 250 rpm for 1 h to regenerate biomass. The mixture was centrifuged at 5000 rpm (Sigma, Beijing, China) to separate the solid (regenerated biomass) and liquid (IL and water) phases. The precipitate (regenerated biomass) was washed thoroughly with hot deionized water several times until a colorless supernatant was obtained. It has been reported that ILs are not detected in the colorless supernatant using Fourier transform infrared (FTIR) measurement (Arora et al. 2010). The precipitate (regenerated biomass) was freeze-dried to avoid the recrystallization of cellulose and collected for subsequent enzymatic hydrolysis (Cao et al. 2014).

Enzymatic hydrolysis

The enzymatic hydrolysis reaction was carried out in 25-mL stoppered conical flasks. First, 0.2 g of pretreated sample was suspended in 10 mL of 50 mM citrate acid-sodium citrate buffer (pH 4.8) at an enzyme (Celluclast 1.5 L, activity 84.5 FPU/g) loading of 20 FPU/g substrate. The flasks were incubated at 50 °C in an air bath shaking incubator (ZWY-2102C, Shanghai, China) at 150 rpm for 72 h. After hydrolysis, the samples were kept in oven at 105 °C for 5 min to inactivate the enzyme.

The glucose concentrations from the enzymatic hydrolysates of pretreated samples were quantitatively analyzed by HPLC (Agilent 1200, USA) with a Bio-Rad Aminex HPX-87H analytical column and a refractive index detector. The mobile phase was 5 mM sulfuric acid solution with a volume flow rate of 0.6 mL/min. The column temperature was set at 35 °C. The temperature of RID was 50 °C, and the injection volume was 10 µL.

Under the optimal conditions, the yield of pretreatment samples relative raw materials was 59.32%. The yield of glucose was defined as the percentage of glucose produced from pretreated samples, which was calculated using Eq. 1 as proposed by Watanabe (2010).


Statistical analysis

The results of the enzymatic hydrolysis experiments were analyzed using RSM (Design Expert 8.0.6 software). The experimental data were fit using a second-order polynomial quadratic equation to evaluate the effects of the variables in terms of linear, quadratic, and interactions on the response (glucose recovery), and determine the optimum process conditions. The polynomial quadratic model for the response, as shown in Eq. (2), was employed,


where y is the predicted response, βis the constant coefficient, βi is the ith linear coefficient, βii is the quadratic coefficient, βij is the ijthinteraction term, andXi and Xj are the independent variables studied.

Each coefficient in Eq. 2 was computed and the interaction effects of the process variables on the response were obtained. The significance of the coefficients was checked by variance analysis (ANOVA). The quality of fitness of the polynomial quadratic model was expressed by the coefficient of determination, R2, and the statistical significance was determined by the F test. Using the same program, three-dimensional plots were drawn to illustrate the effects of the interactions of pretreatment variables on glucose recovery. Optimization (maximizing glucose recoveries) of the fitted polynomial was determined by numerical optimization in Design Expert 8.0.6. Three additional replicates were performed to verify the validity of the optimum values predicted by the program.


Model Development

The experimental data shown in Table 2 were fitted into the quadratic polynomial equations to predict the responses and determine the regression coefficients of the polynomial equation. The following models describe the glucose recovery in terms of coded factors (Eq. 3), and actual factors (Eq. 4); where A is the pretreatment temperature (°C), B is the retention time (h), and C is the solids loading (wt%). Similar equations have been reported for solvent-ionic liquid [Bmim]Cl pretreatment of palm (Tan et al. 2011), aqueous ionic liquid pretreatment of wheat straw (Fu and Mazza 2011), alkali-extrusion pretreatment of big bluestem (Karunanithy and Muthukumarappan 2011), alkaline peroxide pretreatment of wheat straw (Qi et al. 2009), and alkali-extrusion pretreatment of rice husk (Bazargan et al. 2015).

Glucose recovery = 71.45 – 7.26 * A – 0.69 * B + 0.60 * C – 0.70* A * B +1.28 * A * C – 3.15* B * C – 17.39 * A2 – 12.89 * B2 – 13.15 * C2 (3)

Glucose recovery= -595.27 + 7.72 * Temperature + 41.33 * Time+14.67 * Solid loading +1.86e2 * Temperature* Time+1.27e* Temperature * Solid loading – 0.525* Time * Solid Loading – 0.027887 * Temperature2 –5.73* Retention time2 –0.82119 * Solid loading2 (4)

The significance of the developed models was analyzed by ANOVA and is shown in Table 3. The ‘Prob > F’ values were used to check the significance of the effects of each independent variable and quadratic and interaction term on the response. As indicated in Table 3, the ‘F’ values of each model term suggest that A, BC, A2, B2, and C2 had significant effects (P < 0.05) on the glucose recovery from hybrid Pennisetum from the [Amim]Cl pretreatment. It has been reported that A, C, B2, C2, AB, and AC had significant effects on the glucose recovery from palm under [Bmim]Cl pretreatment (Tan et al. 2011). However, the significance of independent variable C was not observed in this study. The effects of B and C terms on glucose recovery were not significant because their ‘Prob > F’ values are greater than 0.1. Although not significant, those factors could not be eliminated from the response model to maintain model hierarchy because the quadratic effects of B and C are significant (Tan et al. 2011).

The effects of the three independent variables on glucose recovery were analyzed using the regression equation. As observed from Eq. 4, pretreatment temperature, retention time, and solids loading all had a positive influence. However, all of the quadratic terms had a negative influence on glucose recovery from hybrid Pennisetum under [Amim]Cl pretreatment. This phenomenon showed that when the three variables were high, the quadratic terms dominated the glucose recovery. These findings did not agree well with previous studies, which indicated that some quadratic terms have a positive influence on glucose recovery (Fu and Mazza 2011; Tan et al. 2011). The reasons for this difference might be that a larger range of independent variables was used in this study or because of the different solvation properties between ILs and substrates (Xie et al. 2012). The magnitude of the coefficient of pretreatment temperature was also lower than that of the retention time and solid loading (Eq. 4). From the coefficent perspective, the effect of pretreatment temperature on glucose recovery was not as significant as retention time and solid loading. But, as can be seen from Table 3, the effect of pretreatment temperature (A) on glucose recovery was more significant than retention time (B) and solid loading (C). This contradictory result was reasonable because the pretreatment temperature value was higher than the retention time value and solids loading value, in Eq. 4.Similar results were reported for oil palm frond (Tan et al. 2011).

Effects of the Interactions of Pretreatment Variables on Glucose Recovery

Along with linear and quadratic terms, interactional terms also contributed to glucose recovery. The effects of the interactions of pretreatment variables on the production of glucose were analyzed by RSM. Three-dimensional response surface and contour plots were generated to illustrate the interactions between any two variables. This was done by evaluating two variables at a time while fixing the remaining variable at a central level, as shown in Fig. 1(a)–(c).

Effect of pretreatment temperature and retention time

The effect of the interaction between pretreatment temperature and retention time on glucose recovery in enzymatic hydrolysis with a constant solids loading of 9 wt% is shown in Fig. 1(a). As the retention time and pretreatment temperature was increased from 0.5 to 4.5 h and 110 to 170 °C, respectively, the glucose recovery increased almost linearly, from 11.4% to 75.2%. Additionally, under an optimal retention time of 2.97 h, the glucose recovery increased from 47.8% to 72.2% with increasing pretreatment temperature. The glucose recovery exceeded 65% when the temperature was increased to 125 °C. It can be inferred that when the pretreatment temperature was higher than 125 °C, cellulose from hybrid Pennisetum was easily dissolved by [Amim]Cl and then hydrolyzed to glucose. This was consistent with previous studies in that the swelling and dissolution of cellulose in ILs are temperature-dependent (Xu et al. 2010). Thus, a higher pretreatment temperature may accelerate the rate of heat transfer between ILs and cellulose to destabilize the H-bonding network and enhance cellulose dissolution by solvation (Kilpeläinen et al. 2007; Xu et al. 2010; Badgujar and Bhanage 2015). Similar results showed that the glucose yield of bagasse pith in cellulase hydrolysis for 72 h had not yet reached its peak when the pretreatment temperature was 80 °C, while it gradually increased as pretreatment temperature increased to 120 °C (Wang et al. 2015). However, the opposite trend was observed when the reaction temperature was over the optimum temperature of 139 °C, which may have been due to simultaneous carbohydrate degradation when the temperature was too high (Swatloski et al. 2002; Sun et al. 2009; Brandt et al. 2010; Wang et al. 2011;).

In addition, the glucose recovery increased substantially from 37.5% to 72.2% when the retention time was increased from 0.5 h to 2.97 hat the optimal pretreatment temperature (139 °C). This result indicated that extended retention time of the biomass material with ILs can boost dissolution by efficient mass transfer/diffusion (Zavrel et al. 2009; Andanson et al. 2014). However, glucose recovery followed a slowly decreasing trend when retention time was prolonged over 2.97 h, which might be due to excessive retention time. A similar finding for excessive retention time on the pretreatment of lignocellulosic palm using [Bmim]Cl was also reported (Tan et al. 2011).

Effect of pretreatment temperature and solids loading

The interaction effect of pretreatment temperature and solids loading on the response observed by RSM is shown in Fig. 1(b), at a fixed retention time of 3 h. With a maximum solids loading of 15 wt%, the glucose recovery increased slowly from 16.6% to only 30%, as the pretreatment temperature increased. High solids loading does not only lead to heat and mass transfer limitation and nonuniformity of the system, but also promotes the formation of inhibitory compounds, which will inhibit subsequent cellulose hydrolysis (Kootstra et al. 2009; Noureddini and Byun 2010). Moreover, the same trend was observed when the solids loading was reduced to 3 wt%. It revealed that lower solids loading did not enhance the final glucose recovery, a result consistent with a previous study (Tan et al. 2011). The glucose recovery increased to a maximum and then decreased as the solids loading and pretreatment temperature increased from 3 to 15 wt% and 110 to 170 °C, respectively. The maximum glucose recovery (72.2%) was obtained under the conditions of a solids loading of 9.06 wt% and pretreatment temperature of 139 °C.

With the optimal pretreatment temperature (139 °C), glucose recovery increased rapidly as the solids loading was increased from 3 to 9 wt%. Similarly, Tan et al. (2011) found that with a solids loading of 10 wt%, the glucose recovery increased substantially as pretreatment temperature increased. This revealed that higher solids loading did enhance the final glucose recovery. However, it has been reported that lower biomass loadings could accelerate cellulose dissolution (Sun et al. 2009; Wang et al. 2011). This difference might reveal no obvious correlation between the IL pretreatment conditions for maximum glucose recovery and IL dissolving conditions for maximum cellulose dissolution. The inconsistency might be due to the degradation of cellulose in ILs with low solid loading, difference in the characteristics of the biomass, or solvation properties of ILs (Badgujar and Bhanage 2015).

Effect of retention time and solids loading

The effect of interaction between retention time and solids loading on glucose recovery is shown in Fig. 1(c). As illustrated, the extremely low predicted value was obtained at low solids loading and short retention time. With an increase in retention time from 0.5 to 2.9 h, the predicted value of glucose increased almost linearly from -0.9% to 71.4% as the solids loading increased from 3 to 9 wt%. In practice, a negative value should not be found. In this three-dimensional response surface, a negative value demonstrated that the pretreatment conditions of low solids loading and short retention time were undesirable. Similar results were reported by Fu and Mazza (2011), which was a significant comprehensive prediction of pretreatment parameters for glucose recovery.

Thus, with optimal retention time, appropriately raising the solids loading can increase the economic feasibility of the [Amim]Cl pretreatment of hybrid Pennisetum and also has the potential to reduce reactor size during pretreatment. Increasing solids loading also leads to a more concentrated product stream, thereby increasing the efficiency of downstream processing while decreasing the operating costs (Kootstra et al. 2009).

Response Surface Model Evaluation

The responses of the predicted values and observed values along with coded and actual variables are presented in Table 2. The relationship between the observed values and predicted values is shown in Fig. 2.The predicted values of the response from the developed model agreed well with the observed ones in the range of the operating variables. The similarities of the predicted and observed responses reflect the goodness of fit.

The ANOVA for the statistical significance of the quadratic model is shown in Table 3. A model is deemed significant if its ‘Prob > F’ value (p-value) is smaller than 0.05 and its F-value is relatively high, suggesting only a 5% chance that a ‘Model F-value’ could occur due to confounders (Ruangmee and Sangwichien 2013). In this model, the F-value of 65 was higher when compared to that of similar models, such as Ruangmee and Sangwichien (2013) and Tan et al. (2011), with F-values of 36.76 and 13.3, respectively. This indicated that the model was highly significant. Moreover, the ‘Prob > F’ value was less than 0.0001, which implied that the regression equation was sufficient to explain the response. The lack of fit confirms the failure of the model to represent data in the experimental area at points, which are not included in the quadratic equation. The insignificant lack of fit further proves the goodness of the model in representing the reactions.

Fig. 1. Interaction effect of two independent variables on glucose recovery (other factors fixed at the center point: 145 °C, 3 h, and 9 wt%)

Fig. 2. Predicted glucose yield vs. actual glucose yield

Furthermore, the ANOVA for the fitness of the quadratic model is presented in Table 4. The coefficient of determination (R2) is always between 0 and 1. The closer to 1.0 the Ris, the stronger the model is and the better it predicts the response value. Generally, a quadratic model with an R2 value higher than 0.90 is considered to have a very high correlation (Haaland 1989). As can be seen from Table 4, the R2value of the model was 0.9832, which further suggested that the quadratic polynomial equations were suitable to adequately represent the relationships among the independent variables. However, a high R2 value does not always mean that the fitness of the quadratic model is good. This is because R2 will increase with the addition of a variable, regardless of whether the additional variable is significant or not (Xin and Saka 2008). Hence, the coefficient of adjusted determination (Adj-R2) and the coefficient of predicted determination (Pred-R2) were calculated to check the adequacy of the model. The Adj-Rrectifies the Rvalue for the sample size and for the number of terms. The Adj-R2 is markedly smaller than R2 if there are many variables in the model and a sample size that is not too large. In this model, the high value of Adj-R2 (0.9681) implies a high significance of the model. The Pred-R(0.9194) is in reasonable agreement with the adjusted determination coefficient, which is also satisfactory to verify the fitness of the model.

Another measure used to evaluate the precision degree of the model is the coefficient of variation (CV), represented as a percentage. It is the ratio of standard error to the mean values. In general, the CV should not be greater than 10% (Linko et al. 1984; Cadoche and Lopez 1989). Consequently, the value of CV, as shown in Table 4, revealed that the experiments conducted are precise and reliable. “Adeq Precision” measures the ratio of signal to noise. A larger ratio denotes better prediction; as a general rule, a ratio greater than 4 is desirable. In this regression model, the ratio of 22.661 suggests an adequate signal, which implies that the model can be used to navigate the design space (Liu et al. 2010). The residuals from the least squares are a significant tool for estimating the model adequacy. Additionally, the normal probability was checked by plotting the normal probability plot of the internally studentized residuals (Li et al. 2007). The nonlinear internally studentized residuals not only indicate that the model is against underlying assumptions, but they also suggest that the error terms are not normally distributed (Alireza et al. 2015).

As shown in Fig. 3, a linear normal distribution of the observed data was verified. The plot of residuals versus the predicted response is shown in Fig. 4. The residual plots of the model exhibited randomly scattered data. This result indicates the adequacy of the quadratic models and good predictions of the response (Mu et al. 2006).

Fig. 3. Normal probability plot of residual errors for glucose yield

Fig. 4. Plot of internally studentized residuals vs. predicted response

Table 3. ANOVA Table for the Quadratic ModelTable 4. Fitness of Quadratic Models

Optimization and Model Validation

Optimization has been defined as enhancing a system’s performance to obtain the maximum benefit (Bezerra et al. 2008). Design Expert 8.0.6 software was used to predict the optimal conditions for the [Amim]Cl pretreatment. In the pretreatment process, an optimum yield of 72.2% of glucose was predicted using the quadratic model at pretreatment conditions of 2.97 h, 139 °C, and 9.06 wt% solids loading. To validate the optimum conditions, three replicate confirmation experiments were conducted, and the results are shown in Table 5.

Table 5. Verification Experiments at Optimum Conditions

An average yield of 70.2% of glucose was obtained, which was consistent with the predicted values. These data show that the experimental values obtained were in agreement with the values calculated from the quadratic model developed using the software. Therefore, the current model is useful for predicting glucose recovery from hybrid Pennisetum under [Amim]Cl pretreatment.


  1. [Amim]Cl pretreatment parameters were optimized for enhanced glucose recovery from the underutilized biomass hybrid Pennisetum by the proposed quadratic model. Under the temperature of 139 °C, retention time of 2.97 h, and solid loading of 9.06 wt-%, an optimal glucose recovery of 72.21% was obtained, which was confirmed by validation.
  2. Statistical analyses revealed that the pretreatment temperature had a more significant effect on sugar recovery than retention time and solid loading.
  3. The interactional effects of retention time and solid loading were greater than the interactional effects of temperature and solid loading on glucose recovery.
  4. Compared to the interaction of retention time and temperature, that of temperature and solid loading has greater effect on glucose recovery.


The authors are grateful for the support of the Chinese Ministry of Education (113014 A) and the National Science Foundation of China (Grant No. 31270627, 31370580, 31470602).


Altan, A., McCarthy, K. L., and Maskan, M. (2008). “Twin-screw extrusion of barley–grape pomace blends: Extrudate characteristics and determination of optimum processing conditions,” Journal of Food Engineering 89(1), 24-32. DOI: 10.1016/j.jfoodeng.2008.03.025

Alireza, B., Majid B., and Gordon, M. (2015). “Optimization of rice husk pretreatment for energy production,” Renewable Energy 77, 512-520. DOI: 10.1016/j.renene.2014.11.072

Andanson, J., Bordes, E., Devémy, J., Leroux, F., Pádua, A. A. H., Costa Gomes, M. F. (2014). “Understanding the role of co-solvents in the dissolution of cellulose in ionic liquids,” Green Chemistry 16(5), 2528-2538. DOI: 10.1039/C3GC42244E

Arora, R., Manisseri, C., Li, C., Ong, M. D., Scheller, H. V., Vogel, K., Simmons, B. A., and Singh, S. et al. (2010). “Monitoring and analyzing process streams towards understanding ionic liquid pretreatment of switchgrass (Panicum virgatum L.),” Bioenergy Research 3(2), 134-145. DOI: 10.1007/s12155-010-9087-1

Badgujar, K. C., and Bhanage, B. M. (2015). “Factors governing dissolution process of lignocellulosic biomass in ionic liquid: Current status, overview and challenges,” Bioresource Technology 178, 2-18. DOI: 10.1016/j.biortech.2014.09.138

Bai, S. J., Yang, Y. S., Gu, G. R., and Zhou, W. X. (1994). “The utilization of Pennisetum hybrid in agricultural arca of south Jiangsu,” Grassland of China 4, 33-35.

Bazargan, A., Bazargan, M., and McKay, G. (2015). “Optimization of rice husk pretreatment for energy production,” Renewable Energy 77, 512-520. DOI: 10.1016/j.renene.2014.11.072

Bezerra, M. A., Santelli, R. E., Oliveira, E. P., Villar, L. S., and Escaleira, L. A. (2008). “Response surface methodology (RSM) as a tool for optimization in analytical chemistry,” Talanta 76(5), 965-977. DOI: 10.1016 /j.talanta.2008.05.019

Brandt, A., Hallett, J. P., Leak, D. J., Murphy, R. J., and Welton, T. (2010). “The effect of the ionic liquid anion in the pretreatment of pine wood chips,” Green Chemistry 12(4), 672-679. DOI: 10.1039/B918787A

Brandt, A., Gräsvik, J., Halletta, J. P., and Welton, T. (2013). “Deconstruction of lignocellulosic biomass with ionic liquids,” Green Chemistry 15(3), 550-583. DOI: 10.1039/C2GC36364J

Cadoche, L., and Lopez, G. D. (1989). “Assessment of size reduction as a preliminary step in the production of ethanol from lignocellulosic wastes,” Biological Wastes 30(2), 153-157. DOI: 10.1016/0269-7483(89)90069-4

Cao, X., Peng, X., Sun, S., Zhong, L., Wang, S., Lu, F., and Sun, R. (2014). “Impact of regeneration process on the crystalline structure and enzymatic hydrolysis of cellulose obtained from ionic liquid,” Carbohydrate Polymers 111, 400-403. DOI: 10.1016/j.carbpol.2014.05.004

Chandra, R., Bura, R., Mabee, W., Berlin, A., Pan, X., and Saddler, J. (2007). “Substrate pretreatment: The key to effective enzymatic hydrolysis of lignocellulosics?,” Biofuels 108, 67-93. DOI: 10.1007/10_2007_064

Chen, M., Liu, Y., Sui, N., Yuan, F., and Wang, B. S. (2014). “The bioenergy plant hybrid Pennisetum improves chilling resistance by enhancing the unsaturation of membrane lipid under low temperature,” Applied Mechanics and Materials 521, 81-87. DOI: 10.4028/

Fan, X. F., Hou, X. C., Zhu, Y., and Wu, J. Y. (2012). “Biomass yield and quality of hybrid Pennisetum as energy crop,” Chinese Journal of Grassland 34(1), 48-52.

Feng, L., and Chen, Z. L. (2008). “Research progress on dissolution and functional modification of cellulose in ionic liquids,” Journal of Molecular Liquids 142(1), 1-5. DOI: 10.1016/j.molliq.2008.06.007

Fu, D., and Mazza, G. (2011). “Optimization of processing conditions for the pretreatment of wheat straw using aqueous ionic liquid,” Bioresource Technology 102(17), 8003-8010. DOI: 10.1016/j.biortech.2011.06.023

Haaland, P. D. (1989). “Separating signals from the noise,” in:Experimental Design in Biotechnology, Haaland, P. D. (ed.), Marcel Dekker, New York, pp. 61.

Hendriks, A. T. W. M., and Zeeman, G. (2009). “Pretreatments to enhance the digestibility of lignocellulosic biomass,” Bioresource Technology 100(1), 10-18. DOI: 10.1016/j.biortech.2008.05.027

Himmel, M., Ding, S., Johnson, D., Adney, W., Nimlos, M., Brady, J., and Foust, T. (2007). “Biomass recalcitrance: Engineering plants and enzymes for biofuels production,” Science 315(5813), 804-807. DOI: 10.1126/science.1137016

Jørgensen, H., Kristensen, J. B., and Felby, C. (2007). “Enzymatic conversion of lignocellulose into fermentable sugars: Challenges and opportunities,” Biofuels, Bioproducts and Biorefining 1(2), 119-134. DOI: 10.1002/bbb.4

Karunanithy, C., and Muthukumarappan, K. (2010). “Effect of extruder parameters and moisture content of switchgrass, prairie cord grass on sugar recovery from enzymatic hydrolysis,” Applied Biochemistry and Biotechnology 162, 1785-1803. DOI: 10.1007/s12010-010-8959-3

Karunanithy, C., and Muthukumarappan, K. (2011). “Optimization of alkali, big bluestem particle size, and extruder parameters for maximum enzymatic sugar recovery using response surface methodology,” BioResources 6(1), 762-790. DOI: 10.15376/biores.6.1.762-790

Kilpeläinen, I., Xie, H., King, A., Granstrom, M., Heikkinen, S., and Argyropoulos, D. S. (2007). “Dissolution of wood in ionic liquids,” Journal of Agricultural and Food Chemistry 55(22), 9142-9148. DOI: 10.1021/jf071692e

Kleijnen, J. P. (2008). “Response surface methodology for constrained simulation optimization: An overview,” Simulation Modelling Practice and Theory 16(1), 50-64. DOI: 10.1016/j.simpat.2007.10.001

Kootstra, A. M. J., Beeftink, H. H., Scott, E. L., and Sanders, J. P. M. (2009). “Comparison of dilute mineral and organic acid pretreatment for enzymatic hydrolysis of wheat straw,” Biochemical Engineering Journal 46(2), 126-131. DOI: 10.1016/j.bej.2009.04.020

Kuo, C., and Lee, C. (2009). “Enhancement of enzymatic saccharification of cellulose by cellulose dissolution pretreatments,” Carbohydrate Polymers 77(1), 41-46. DOI: 10.1016/j.carbpol.2008.12.003

Li, Y., Cui, F., Liu, Z., Xu, Y., and Zhao, H. (2007). “Improvement of xylanase production by Penicillium oxalicum ZH-30 using response surface methodology,” Enzyme and Microbial Technology 40(5), 1381-1388. DOI: 10.1016/j.enzmictec.2006.10.015

Li, Q., He, Y., Xian, M., Jun, G., Xu, X., Yang, J., and Li, L. Z. (2009). “Improving enzymatic hydrolysis of wheat straw using ionic liquid 1-ethyl-3-methylimidazolium diethyl phosphate pretreatment,” Bioresource Technology 100, 3570-3575. DOI: 10.1016/j.biortech.2009.02.040

Li, B., Asikkala, J., Filpponen, I., and Argyropoulos, D. S. (2010). “Factors affecting wood dissolution and regeneration of ionic liquids,” Industrial and Engineering Chemistry Research 49(5), 2477-2484. DOI: 10.1021/ie901560p

Li, W. Z., Ju, M. T., Wang, Y. N., Liu, L., Jiang, Y. (2013). “Separation and recovery of cellulose from Zoysia japonica by 1-allyl-3-methylimidazolium chloride,” Carbohydrate Polymers 92(1), 228-235. DOI: 10.1016/j.carbpol.2012.09.075

Linko, P., Mattson, C., Linko, Y. Y., and Antila, J. (1984). “Production of flat bread by continuous extrusion cooking from high α-amylase flours,” Journal of Cereal Science 2(1), 43-51. DOI: 10.1016/S0733-5210(84)80007-7

Liu, Q., Cheng, K. K., Zhang, J. A., Li, J. P., and Wang, G. H. (2010). “Statistical optimization of recycled-paper enzymatic hydrolysis for simultaneous saccharification and fermentation via central composite design,” Applied Biochemistry and Biotechnology 160(2), 604-612. DOI: 10.1007/s12010-008-8446-2

Liu, D. T., Xia, K. F., Cai, W. H., Yang, R. D., Wang, L. Q., and Wang, B. (2012). “Investigations about dissolution of cellulose in the 1-allyl-3-alkylimidazolium chloride ionic liquids,” Carbohydrate Polymers 87(2), 1058-1064. DOI: 10.1016/j.carbpol.2011.08.026

Lee, K. M., Ngoh, G. C., and Chua, A. S. M. (2013). “Process optimization and performance evaluation on sequential ionic liquid dissolution–solid acid saccharification of sago waste,” Bioresource Technology 130, 1-7. DOI: 10.1016/j.biortech.2012.11.124

Luque, R., Herrero-Davila, L., Campelo, J. M., Clark, J. H., Hidalgo, J. M., Luna, D., Marinas, J. M., and Romero, A. A. (2008). “Biofuels: A technological perspective,” Energy and Environmental Science 1(5), 542-564. DOI: 10.1039/B807094F

Mu, Y., Wang, G., and Yu, H. Q. (2006). “Response surface methodological analysis on biohydrogen production by enriched anaerobic cultures,” Enzyme and Microbial Technology 38(7), 905-913. DOI: 10.1016/j.enzmictec.2005.08.016

Noureddini, H., and Byun, J. (2010). “Dilute-acid pretreatment of distillers’ grains and corn fiber,” Bioresource Technology 101(3), 1060-1067. DOI: 10.1016/j.biortech.2009.08.094

Qi, B., Chen, X., Shen, F., Su, Y., and Wan, Y. (2009). “Optimization of enzymatic hydrolysis of wheat straw pretreated by alkaline peroxide using response surface methodology,” Industrial and Engineering Chemistry Research 48(15), 7346-7353. DOI: 10.1021/ie8016863

Ruangmee, A., and Sangwichien, C. (2013). “Response surface optimization of enzymatic hydrolysis of narrow-leaf cattail for bioethanol production,” Energy Conversion and Management 73, 381-388. DOI: 10.1016/j.enconman.2013.05.035

Sun, N., Rahman, M., Qin, Y., Maxim, M. L., Rodríguez, H., and Rogers, R. D. (2009). “Complete dissolution and partial delignification of wood in the ionic liquid 1-ethyl-3-methylimidazolium acetate,” Green Chemistry 11(5), 646-655. DOI: 10.1039/B822702K

Sun, N., Rodríguez, H., Rahman, M., and Rogers, R. D. (2011). “Where are ionic liquid strategies most suited in the pursuit of chemicals and energy from lignocellulosic biomass?,” Chemical Communications 47(5), 1405-1421. DOI: 10.1039/C0CC03990J

Swatloski, R. P., Spear, S. K., Holbrey, J. D., and Rogers, R. D. (2002). “Dissolution of cellose with ionic liquids,” Journal of the American Chemical Society 124(18), 4974-4975. DOI: 10.1021/ja025790m

Tadesse, H., and Luque, R. (2011). “Advances on biomass pretreatment using ionic liquids: An overview,” Energy and Environmental Science 4(10), 3913-3929. DOI: 10.1039/C0EE00667J

Tan, H. T., Lee, K. T., and Mohamed, A. R. (2011). “Pretreatment of lignocellulosic palm biomass using a solvent-ionic liquid [Bmim]Cl for glucose recovery: An optimisation study using response surface methodology,” Carbohydrate Polymers 83(4), 1862-1868. DOI: 10.1016/j.carbpol.2010.10.052

Wang, X., Li, H., Cao, Y., and Tang, Q. (2011). “Cellulose extraction from wood chip in an ionic liquid 1-allyl-3-methylimidazolium chloride (AmimCl),” Bioresource Technology 102(17), 7959-7965. DOI: 10.1016/j.biortech.2011.05.064

Wang, G., Zhang, S., Xu, W., Qi, W., Yan, Y., and Xu, Q. (2015). “Efficient saccharification by pretreatment of bagasse pith with ionic liquid and acid solutions simultaneously,” Energy Conversion and Management 89(1), 120-126. DOI: 10.1016/j.enconman.2014.09.029

Wang, H., Gurau, G., and Rogers, R. D. (2012). “Ionic liquid processing of cellulose,” Chemical Society Reviews 41(4), 1519-1537. DOI: 10.1039/C2CS15311D

Watanabe, H. (2010). “The study of factors influencing the depolymerisation of cellulose using a solid catalyst in ionic liquids,” Carbohydrate Polymers 80(4), 1168-1171. DOI: 10.1016/j.carbpol.2010.01.039

Xie, H., Liu, W., and Zhao, Z. K. (2012). “Lignocellulose pretreatment by ionic liquids: A promising start point for bio-energy production,” in: Biomass Conversion, C. Baskar, S. Baskar, and Dhillon, R. S. (eds.), Springer, Berlin, pp. 123-144. DOI: 10.1007/978-3-642-28418-2_3

Xin, L., and Saka, S. (2008). “Optimization of Japanese beech hydrolysis treated with batch hot compressed water by response surface methodology,” International Journal of Agricultural and Biological Engineering (IJABE) 1(2), 39-45.

Xu, A., Wang, J., and Wang, H. (2010). “Effects of anionic structure and lithium salts addition on the dissolution of cellulose in 1-butyl-3-methylimidazolium-based ionic liquid solvent systems,” Green Chemistry 12(2), 268-275. DOI: 10.1039/B916882F

Yang, B., and Wyman, C. E. (2008). “Pretreatment: The key to unlocking low cost cellulosic ethanol,” Biofuels, Bioproducts and Biorefining 2(1), 26-40. DOI: 10.1002/bbb.49

Yoon, L. W., Ang, T. N., Ngoh, G. C., and Chua, A. S. M. (2012). “Regression analysis on ionic liquid pretreatment of sugarcane bagasse and assessment of structural changes,” Biomass & Bioenergy 36, 160-169. DOI: 10.1016/j.biombioe.2011.10.033

Zavrel, M., Bross, D., Funke, M., Büchs, J., and Spiess, A. C. (2009). “High-throughput screening for ionic liquids dissolving (ligno-) cellulose,” Bioresource Technology 100(9), 2580-2587. DOI: 10.1016/j.biortech.2008.11.052

Zhang, Y. H., Berson, E., Sarkanen, S., and Dale, B. E. (2009). “Pretreatment and biomass recalcitrance: Fundamentals and progress,” Applied Biochemistry and Biotechnology 153(1-3), 80-83. DOI: 10.1007/s12010-009-8610-3

Zhao, H., Jones, C. L., Baker, G. A., Xia, S., Olubajo, O., and Person, V. N. (2009). “Regenerating cellulose from ionic liquids for an accelerated enzymatic hydrolysis,” Journal of Biotechnology 139(1), 47-54. DOI: 10.1016/j.jbiotec.2008.08.009

Zheng, Y., Pan, Z., and Zhang, R. (2009). “Overview of biomass pretreatment for cellulosic ethanol production,” International Journal of Agricultural and Biological Engineering (IJABE) 2(3), 51-68

Zhu, S., Wu, Y., Chen, Q., Yu, Z., Wang, C., Jin, S., Ding, Y., and Wu, G. et al. (2006). “Dissolution of cellulose with ionic liquids and its application: A mini-review,” Green Chemistry 8(4), 325-327. DOI: 10.1039/B601395C

Article submitted: June 15, 2015; Peer review completed: August 24, 2015; Revised version received and accepted: August 25, 2015; Published: September 1, 2015.

DOI: 10.15376/biores.10.4.7021-7037