论文部分内容阅读
Abstract： Phosphotungstic acid （H3PW12O40， HPW）， a kind of solid acid， is widely used for hydrolyzing cellulose to prepare microcrystalline cellulose （MCC）. MCC is usually used in food， synthetic leather， chemical and pharmaceutical industries. The use of response surface methodology （RSM） can help avoid the random error caused by single factor experimental design， reduce test times and cost， and improve quality. The RSM was used in this study to determine the following optimal process conditions： H+ molar quantity， 31 mmol/L； reaction temperature， 93℃； reaction time， 2 h； and solid to liquid ratio， 1∶38. Under these conditions， the crystallinity of MCC was 77.4%. Thus， the use of RSM allows the preparation of MCC with higher performance and increased crystallinity.
Keywords： cellulose hydrolysis； phosphotungstic acid； response surface； microcrystalline cellulose； crystallinity
1 Introduction
With the increasing shortage for fossil fuels， biomass resources are attracting growing interest as they are low cost and renewable and can be used to produce high-value-added products[1]. Many studies have focused on preparing microcrystalline cellulose （MCC） using chemicals on biomass resources[2-3]. MCC is obtained by using natural cellulose hydrolyzed with a dilute acid. The hydrolyzed cellulose is then ground to a crystalline powder to obtain the leveling-off degree of polymerization （LODP）[4]. MCC is colorless， tasteless， and insoluble in both water and organic solvents[5]； its particle size ranges from 20 mm to 80 mm and the LODP is 15～375. MCC can be prepared by acid hydrolysis， oxidation， enzymatic hydrolysis， or a combination of several methods[6-10]. However， conventional acids such as HCl and H2SO4 are disadvantageous for hydrolyzing cellulose， because their use involves complex recycling processes， and they also cause corrosion of equipment and environmental pollution. Therefore， it is necessary to find new acids that can replace such conventional acids for hydrolyzing the cellulose.
Many papers have reported that solid heteropoly acids possess high activity， thermal stability， and selectivity， among other characteristics[11]. Among heteropoly acids， phosphotungstic acid （HPW） has been proven to be the strongest Br?nsted acid to release protons （H+） and break the β-1，4-glycosidic bonds in cellulose[12]. In this study， HPW was used to hydrolyze cellulose. Because the single factor experimental design suffers from randomness， discontinuity， and other such problems， the authors attempted to seek a more effective design to optimize the process parameters and to improve the accuracy of the experimental results. Response surface methodology （RSM）， also called the regression design method， involves the use of multiple regression analyses of two regression methods to fit the relationship between the factors and the response values. It can also be used to analyze and optimize the regression equation and to obtain the optimal process conditions for the multivariable and statistical method of optimizing the response value[13].
Therefore， in this study， the authors utilized the RSM to optimize the amount of HPW （expressed as the H+ molar quantity）， reaction temperature， reaction time， and solid to liquid ratio for designing a cellulose hydrolysis reaction scheme by using HPW. The aim is to achieve higher-performance MCC with increased crystallinity.
2 Experimental
2.1 Materials
The raw material used in the experiment was eucalyptus kraft dissolving pulp， sourced from Shandong province， China. The crystallinity of the dissolving pulp is 56.5%. Phosphotungstic acid （H3PW12O40， HPW）， ethanol， and diethyl ether were bought from Sinopharm Chemical Reagent Co.， Ltd. （Guangzhou， China）. All of the chemicals were analytically pure and used as received.
2.2 Preparation of MCC
The reaction was carried out in a 250 mL three-necked flask equipped with a stirrer； the speed was set to 120 r/min. The flask was kept in a thermostatic water bath. Then， 0.5 g of the dissolving pulp （oven-dried）， HPW （7～12 mmol/L）， and deionized water were added at a temperature range of 70～100℃. The reaction time ranged from 1 h to 3 h. After hydrolysis， the product was first filtered by a glass filter （G4， filter bore diameter was in the range of 5～15 ?m）， and then successively washed with deionized water and ethanol until it reached a neutral pH value. The resulting products were stored for further processing.
2.3 Methods
2.3.1 Establishment of the response surface method
Through the previous single factor experiment[14]， the following optimum technological conditions for hydrolyzing cellulose by HPW could be obtained： H+ molar quantity， 30 mmol/L； reaction temperature， 90℃； reaction time， 2.0 h； and solid to liquid ratio， 1：40. Using the central composite design （CCD）， H+ molar quantity （A）， reaction temperature （B）， reaction time （C）， and solid to liquid ratio （D） were chosen as the four independent variables. The crystallinity of MCC （R） was defined as response value to optimize the experimental process parameters and determine the maximum value of crystallization under the optimum conditions. The experimental factors and level encoding are listed in Table 1. According to Table 1， using Design-Expert 7 software （Stat-Ease Inc.， 7.1， Minneapolis， MN， USA） to design 4 factors 5 levels and a total of 26 experiments and the order of the experiments was randomly disrupted for reducing the error.
Table 1 Factors and levels coding table
Factors Level
-2 （-a） -1 0 1 2 （+a）
A/（mmol·L-1） 27.0 28.5 30.0 31.5 33.0
B/℃ 80 85 90 95 100
C/h 1.50 1.75 2.00 2.25 2.50
D 1∶30 1∶35 1∶40 1∶45 1∶50
2.3.2 Analytical methods
The reliability of the regression model was determined by variance analysis using Design-Expert 7 software.
The surface morphologies of the raw material pulp and MCC were characterized by using an environmental scanning electron microscope （ESEM， Q45， FEI， Hillsboro， USA）. The samples were treated for gold sputtering using an ion sputtering apparatus （ETD， COXEM， Daejeon， Korea） and analyzed at 20 kV.
The particle size distribution of raw material pulp and MCC were determined by using a BT-9300H laser particle size analyzer （Dandong， China）. The samples were dispersed in distilled water and ultrasonically treated for 1 min.
The crystalline structures of the raw material pulp and MCC were investigated by using the Japanese Neo Confucianism X-ray diffraction （XRD） spectrometer （Rigaku， Tokyo， Japan）， which used a Ni-filtered Cu-Ka radiation source （λ=0.1518 nm）. The diffraction angle 2θ scan range was 10°～40°and the scanning speed was 2°/min. The crystallinity （CI） of MCC was measured using Eq.（1）[15].
（1）
Where Fa and Fc refer to the area of crystalline and amorphous regions， respectively.
The thermal stability of the samples was analyzed by thermogravimetric analysis （TGA， STA449 F3， Bruker， Germany）. The heating temperature was increased from 25℃ to 600℃ at a heating rate of 10℃/min.
3 Results and discussion
3.1 Experimental results of RSM
RSM was used to design the experimental scheme.
Table 2 Experimental results by RSM
Serial
number Level Response value
（Crystallinity）
A B C D Actual
value R1/% Predicted
value R2/%
1 0 0 0 0 76.29 76.44
2 1 1 1 1 74.02 74.62
3 -1 1 1 -1 74.91 73.67
4 -1 1 -1 -1 72.59 73.17
5 0 2 0 0 75.41 75.36
6 0 0 0 0 76.50 76.44
7 1 -1 1 -1 73.40 73.86
8 0 0 0 -2 71.71 71.98 9 1 1 -1 -1 73.01 74.01
10 1 -1 -1 -1 71.94 71.11
11 -1 -1 1 -1 70.53 70.77
12 1 1 1 -1 77.32 76.76
13 2 0 0 0 77.78 77.38
14 1 1 -1 1 73.69 74.16
15 -1 1 1 1 73.31 73.59
16 1 -1 1 1 70.86 69.88
17 -1 -1 -1 1 69.51 70.64
18 0 0 2 0 70.61 71.44
19 0 0 0 2 70.44 70.21
20 0 0 -2 0 71.26 70.48
21 0 -2 0 0 67.62 67.71
22 -1 -1 -1 -1 70.50 70.26
23 -1 -1 1 1 69.35 68.85
24 -1 1 -1 1 76.46 75.38
25 -2 0 0 0 75.16 75.51
26 1 -1 -1 1 68.92 69.41
Experimental results obtained under different conditions are shown in Table 2.
3.2 Establishment and analysis of regression model
The regression equation was established on the basis of the data given in Table 2. Design-Expert 7 software was used to carry out two multivariate linear regression fitting. After optimization， the response value （crystallinity） for the independent variables （four factors） was obtained. The regression equation was as follows （Coded representation）：
R=76.44+0.47A+1.91B+0.24C–0.44D+0.56AC– 0.52AD+0.46BD–0.57CD–1.23B2–1.37C2 –1.34D2（2）
The results of variance regression analysis obtained Table 3 Variance analysis of regression model
Variance Sum of
squares Liberty Mean
square F P Conspicuousness
Model 187.29 11 17.03 22.69 <0.0001 Significant
A 5.26 1 5.26 7.01 0.0191
B 87.71 1 87.71 116.87 <0.0001
C 1.39 1 1.39 1.85 0.1947
D 4.70 1 4.70 6.26 0.0253
AC 5.06 1 5.06 6.75 0.0211
AD 4.28 1 4.28 5.71 0.0315
BD 3.40 1 3.40 4.54 0.0514
CD 5.24 1 5.24 6.99 0.0193
B2 33.10 1 33.10 44.11 <0.0001
C2 41.39 1 41.39 55.15 <0.0001
D2 39.30 1 39.30 52.37 <0.0001
Residual 10.51 14 0.75 — —
Lack of fit 10.48 13 0.81 36.58 0.1288 Not significant
Net error 0.02 1 0.02 — —
The sum 197.80 25 — — —
Notes： Homogeneity test of variance P （Prob.>F） could be used to determine the significant impact of the various factors that affected the results of the test. When P<0.0001， the difference was extremely significant； when P<0.05， the difference was significant. In the test of variance analysis， Response surface variation coefficient （C.V.）=1.19%， R2=95%， Adj. R2=91%， Pred. R2=81%， Signal-to-noise ratio Adeq. Precision = 16.42.
by response surface optimization are shown in Table 3.
In general， when the determination coefficient R2 of the regression model was closer to 100%， the dependent and independent variables showed a better linear correlation. In this model， R2 = 95%， which showed that the model could explain the 95% change of response value， owing to the use of four variables A， B， C， and D. C.V. was 1.19%， which showed that the experimental results had a high precision， and the experimental operation was reliable[16]. Fig.1 shows a straight-line plot between the predicted and experimental results， which means that this model can be used because the experimental results and the model predictions were in a high degree of coincidence. Moreover， from Table 3， it is clear that the size of the F value does not exert a very obvious effect on the molar quantity of H+ and hence on crystallinity of MCC. If the influence of the H+ molar quantity （A） on the crystallinity of MCC is not considered， then the other three factors influence the crystallinity of MCC in the following order of significance of degree： reaction temperature （B）， solid to liquid ratio （D）， and reaction time （C）. Fig.1 Actual value and predicted value of Crystallinity
For the residual analysis of the regression model， the residual normal graph was closer to a straight line， and the regression model fit better. Fig.2 showed that these residuals lie on a straight line and that the dispersity was small which illustrate that the above model was reasonable. Meanwhile， the plot with different colors used in Fig.1 and Fig.2 was based on changes of crystallinity.
Fig.2 Standardized residual and normal probability of crystallinity
3.3 Response surface and contour analysis
A 3D picture of the response surface and the corresponding contour map can be used to directly show the effect of interactions between the two factors on the response value. The following graphics were all of the variables A， B， C， and D and the influence of any two factors on the crystallinity of MCC. If any two factors were set to 0 levels， the interaction between the other two factors that influences the response value R on the 3D response surface and the contour were studied. Combined with Table 3 analysis， the P values of AC， AD， BC and BD were small， which showed that the response value was more significant. The response surface and contour lines of the four items were analyzed， and the results are shown in Fig.3～Fig.6.
Fig.3 showed that when the reaction temperature was fixed at 90℃ and the solid to liquid ratio was 1∶40， the crystallinity of MCC increased slightly along with an increase in the H+ molar quantity. When the reaction time was increased， the crystallinity of MCC first increased and reached the maximum value at about 2 h and then decreased. In addition， if the shape of the contour lines were not round， the interaction between the H+ molar quantity and the reaction time was judged to be significant. Based on the results of a previous study[14]， it is clear that the H+ molar quantity is the most important factor. According to the contour lines， density changes could be used to investigate whether the reaction time exerted more influence on the crystallinity of MCC than the H+ molar quantity. If the reaction time was further extended， it could improve the crystallinity of MCC. However， if the time was extended too long， it not only destroyed the cellulose crystalline region but also produced many by-products such as glucose， fructose， cellobiose， and cellotriose. In addition， formic acid， acetic acid， and other such acids can be generated during hydrolysis[17-18]. Although the content of the by-products was higher， the crystallinity of the MCC was lower. As Fig.3 showed， when the H+ molar quantity was 31.5 mmol/L and the reaction time was 2 h， the maximum value of crystallinity of MCC reached was 76.6%. If the reaction temperature was fixed at 90℃ and the reaction time of 2 h was used， then the increasing of the H+ molar quantity slightly increased the crystallinity of MCC （Fig.4）. This is consistent with the results shown in Fig.3. As the solid to liquid ratio was increased， the crystallinity of MCC first increased and then decreased； when the solid to liquid ratio was 1∶38， the crystallinity of MCC reached the maximum value. From the shape of the contour， it could be seen that the interaction between the H+ molar quantity and the solid to liquid ratio exerted a significant impact on the crystallinity of MCC. The contour line density showed that the solid to liquid ratio exerted more influence than the H+ molar quantity on the crystallinity of MCC. This was because when the solid to liquid ratio was too low， the HPW could not fully come into contact with the cellulose material； therefore， the reaction was not complete. If the solid to liquid ratio was too high， the HPW could not remove the amorphous regions of cellulose as much as possible， which decreased the crystallinity of MCC. The crystallinity of MCC could reach the maximum value of 76.7% when the H+ molar quantity was 31.5 mmol/L and the solid to liquid ratio was 1∶38.
When the H+ molar quantity was fixed at 30 mmol/L and
a reaction time of 2 h was used， along with the increasing of the reaction temperature and the solid to liquid ratio， the crystallinity of MCC first increased and then decreased， as shown in Fig.5. Because the shape of the contour lines was almost circular， this interaction had a lesser significant effect on the crystallinity response value than the above interaction term of AD and AC. The change of the contour line density showed that the reaction temperature exerted more influence on the crystallinity of MCC than the solid to liquid ratio. In general， increased temperature could intensify the reaction， which made the HPW come into full contact with cellulose， especially when amorphous region degraded and MCC crystallinity improved. However， if the temperature was too high， then the HPW would also hydrolyze the cellulose amorphous region. As a result， the crystalline area would inevitably experience a certain degree of destruction， which decreased the crystallinity of MCC. In Fig.5， the effect of the solid to liquid ratio on the crystallinity of MCC was consistent with the result shown in Fig.4. Therefore， by fixing the other two factors and maintaining the reaction temperature at 93℃ and the solid to liquid ratio to 1∶38， MCC showed the best crystallinity value of 76.2%. When the H+ molar quantity was fixed at 30 mmol/L and the reaction temperature was maintained at 90℃， then extending the reaction time and increasing the amount of water would first increase the crystallinity of MCC and then decrease it （Fig.6）. From the shape， size， and color depth of the contour， it is clear that this interaction was less significant to the crystallinity of MCC compared to the above three interaction terms. When the contour line density was changed， the solid to liquid ratio exerted more influence on the crystallinity of MCC than the reaction time. The effect of each factor on the crystallinity of MCC was consistent with the above analysis. When the reaction time was 2 h and the solid to liquid ratio was 1∶38， the best crystallinity value of MCC was greater than 75.9%.
The effect of each factor on the crystallinity of MCC and the response value significance is shown in Table 3， and the influence of various interaction terms is shown in Fig.3～Fig.6. The factors influenced the crystallinity of MCC in the following order： H+ molar quantity （A）， reaction temperature （B）， solid to liquid ratio （D）， and reaction time （C）.
3.4 Experimental validation of regression model
Based on the above experimental results， it can be expected to achieve a higher crystallinity of MCC through the optimization of the model. The process conditions for MCC were considered by modifying the actual situation and the convenience of the experimental operation. Under these conditions， three parallel experiments were performed and the average of the measured values of the crystallinity of MCC was obtained. The results are shown in Table 4.
Table 4 Optimal factors and results of crystallinity
Parameter H+ molar
quantity
/（mmol·L-1） Reaction temperature
/℃ Reaction
time
/h Solid to liquid ratio Crystallinity /%
Predicted value 31.4 93.1 2.1 1∶38 77.8
Actual value 31.0 93.0 2.0 1∶38 77.4
3.5 SEM observation
The morphologies of the raw material pulp and MCC were investigated by SEM （Fig.7a and Fig.7b）. Though treated by HPW， the ribbon shape of the raw material pulp broke and degraded， as shown in Fig.7b， where the MCC became short and rod-like. In addition， the particle size distribution and the particle size of the raw material pulp and MCC are shown in Fig.7c. The fiber length of MCC was smaller than that of the raw material pulp， which was consistent with the SEM results. 3.6 XRD analysis
The XRD patterns of raw material pulp and MCC are shown in Fig.8. They exhibited similar diffractograms， where the peaks at 2θ=14.9°， 16.5°，22.5°， and 34.7° accorded with the 1-10， 110， 200， and 040 planes， respectively. According to previovus studies[4，19]， these typical peaks corresponded to cellulose I β structure and were used for determining the crystalline area. Compared with the raw material pulp， the crystallinity of MCC increased remarkably from 56.4% to 77.4%. The structure of MCC did not change after hydrolyzation. The peak at 2θ=22.5°became sharper， indicating that the MCC had a perfect crystal lattice in contrast to the raw material pulp， which resulted in MCC with a higher crystallinity[20-21].
Fig.8 XRD of raw material pulp and MCC
3.7 Thermal stability analysis
The thermogravimetric analysis （TGA） curves for raw material pulp and MCC are shown in Fig.9. The raw material pulp began to decompose at 245℃， and the MCC probably initially degraded at 270℃. When the temperature exceeded 350℃， weight loss occurred because the thermal stability of the material is related to its crystalline property. The higher the crystallinity， the better the thermal stability， and the higher the temperature required for degradation[22]. Hence， the crystallinity of MCC is higher than that of the raw material pulp. The results were consistent with the XRD analysis.
Fig.9 TGA thermograms of raw material pulp and MCC
4 Conclusions
In this paper， the single factor experimental results were optimized by response surface methodology. The analysis of variance showed that this model was available， and the significance of each factor on the crystallinity of MCC was explored. Finally， the MCC with higher crystallinity was obtained under the optimized reaction condition. Specific conclusions were as follows.
4.1 The variance analysis showed that the model could correctly predict the experimental results. Each factor and the interaction between the two factors exerted different effects on the response value. Each factor affected the crystallinity of MCC in the following order by degree of significance： H+ molar quantity， reaction temperature， solid to liquid ratio， and reaction time.
4.2 The optimum experimental conditions for preparing high-crystallinity MCC by HPW hydrolysis were as follows： H+ molar quantity， 31 mmol/L； reaction temperature， 93℃； reaction time， 2 h； and solid to liquid ratio， 1∶38. Under these conditions， the maximum value of crystallinity of MCC was 77.4%. 4.3 The SEM images showed the changes between the raw material pulp and MCC. The XRD showed that they had the same structure. TGA showed that MCC had higher crystallinity than the raw material pulp.
Acknowledgments
This work was supported by the National Key Research and Development Program of China （2017YFB0307900）， the Key Research and Development Project of Shaanxi Province （2017ZDXM-SF-090） and the State Key Laboratory of Donghua University （NO. LK1601）.
References
[1] Rodrigues F G， Monteiro D S， Cdas M， et al. Synthesis and characterization of cellulose acetate produced from recycled newspaper[J]. Carbohydrate Polymers， 2008， 73（1）： 74-82.
[2] Araki J， Wada M， Kuga S， et al. Flow properties of microcrystalline cellulose suspension prepared by acid treatment of native cellulose[J]. Colloids & Surfaces A： Physicochemical and Engineering Aspects， 1998， 142（1）： 75-82.
[3] El-sakhawy M， Hassan M L. Physical and mechanical properties of microcrystalline cellulose prepared from agricultural residues[J]. Carbohydrate Polymers， 2007， 67（1）： 1-10.
[4] Adel A M， El-Wahab Z H A， Ibrahim A A， et al. Characterization of microcrystalline cellulose prepared from lignocellulosic materials. Part II： Physicochemical properties[J]. Carbohydrate Polymers， 2011， 83（2）： 676-687.
[5] Wang D， Shang S B， Song Z Q， et al. Evaluation of microcrystalline cellulose prepared from kenaf fibers[J]. Journal of Industrial and Engineering Chemistry， 2010， 16（1）： 152-156.
[6] Kalita R D， Nath Y， Ochubiojo M E， et al. Extraction and characterization of microcrystalline cellulose from fodder grass； Setaria glauca （L） P. Beauv， and its potential as a drug delivery vehicle for isoniazid， a first line antituberculosis drug[J]. Colloids and Surfaces B： Biointerfaces， 2013， 108（4）： 85-89.
[7] Zhang M Y， Qiang D D， Li J B， et al. Changes of Structure and Morphological Characteristics of Cellulose during Preparation of Microcrystalline Cellulose[J]. China Pulp & Paper， 2016， 35（6）： 28-32.
[8] Ciolacu D. On the supramolecular structure of cellulose allomorphs after enzymatic degradation[J]. Journal of Optoelectronics and Advanced Materials， 2007， 9（4）： 1033-1037.
[9] Agblevor F A， Ibrahim M M， El-Zawawy W K. Coupled acid and enzyme mediated production of microcrystalline cellulose from corn cob and cotton gin waste[J]. Cellulose， 2007， 14（3）： 247-256.
[10] Tang Y J， Shen X C， Zhang J H， et al. Extraction of cellulose nano-crystals from old corrugated container fiber using phosphoric acid and enzymatic hydrolysis followed by sonication[J]. Carbohydrate Polymers， 2015， 125： 360-366. [11] Deng W P， Zhang Q H， Wang Y. Polyoxometalates as efficient catalysts for transformations of cellulose into platform chemicals[J]. Dalton Transactions， 2012， 41（33）： 9817-9831.
[12] Siró I， Plackett D. Microfibrillated cellulose and new nanocomposite materials： a review[J]. Cellulose， 2010， 17（3）： 459-494.
[13] Adel A M， El-Shinnawy N A. Hypolipidemic applications of microcrystalline cellulose composite synthesized from different agricultural residues[J]. International Journal of Biological Macromolecules， 2012， 51（5）： 1091-1102.
[14] Qiang D D， Zhang M Y， Li J B， et al. Selective hydrolysis of cellulose for the preparation of microcrystalline cellulose by phosphotungstic acid[J]. Cellulose， 2016， 23（2）： 1199-1207.
[15] Park S， Baker J O， Himmel M E， et al. Cellulose crystallinity index： measurement techniques and their impact on interpreting cellulase performance[J]. Biotechnology for Biofuels， 2010， 3（1）： 1-10.
[16] Duret X， Fredon E， Masson E， et al. Optimization of acid pretreatment in order to increase the phenolic content of Picea abies bark by surface response methodology[J]. BioResources， 2013， 8（1）： 1258-1273.
[17] Ma Y H， Ji W Q， Zhu X， et al. Effect of extremely low AlCl3 on hydrolysis of cellulose in high temperature liquid water[J]. Biomass Bioenergy， 2012， 39（39）： 106-111.
[18] Zhuang X S， Wang S R， Yuan Z H， et al. Research on biomass hydrolysis under extremely low acids by HPLC[J]. Acta Energiae Solaris Sinica， 2007， 28（11）： 1239-1243.
[19] French A D. Idealized powder diffraction patterns for cellulose polymorphs[J]. Cellulose， 2014， 21（2）： 885-896.
[20] Shimizu K， Furukawa H， Kobayashi N， et al. Effects of Bronsted and Lewis acidities on activity and selectivity of heteropolyacid-based catalysts for hydrolysis for hydrolysis of cellobiose and cellulose[J]. Green Chemistry， 2009， 11（10）： 1627-1632.
[21] Lu Q L， Tang L R， Lin F C， et al. Preparation and characterization of cellulose nanocrystals via ultrasonication-assisted FeCl3-catalyzed hydrolysis[J]. Cellulose， 2014， 21（5）： 3497-3506.
[22] Yang P Y， Kokot S. Thermal analysis of different cellulosic fabrics[J]. Journal of Applied Polymer Science， 1996， 60（8）： 1137-1146.
Keywords： cellulose hydrolysis； phosphotungstic acid； response surface； microcrystalline cellulose； crystallinity
1 Introduction
With the increasing shortage for fossil fuels， biomass resources are attracting growing interest as they are low cost and renewable and can be used to produce high-value-added products[1]. Many studies have focused on preparing microcrystalline cellulose （MCC） using chemicals on biomass resources[2-3]. MCC is obtained by using natural cellulose hydrolyzed with a dilute acid. The hydrolyzed cellulose is then ground to a crystalline powder to obtain the leveling-off degree of polymerization （LODP）[4]. MCC is colorless， tasteless， and insoluble in both water and organic solvents[5]； its particle size ranges from 20 mm to 80 mm and the LODP is 15～375. MCC can be prepared by acid hydrolysis， oxidation， enzymatic hydrolysis， or a combination of several methods[6-10]. However， conventional acids such as HCl and H2SO4 are disadvantageous for hydrolyzing cellulose， because their use involves complex recycling processes， and they also cause corrosion of equipment and environmental pollution. Therefore， it is necessary to find new acids that can replace such conventional acids for hydrolyzing the cellulose.
Many papers have reported that solid heteropoly acids possess high activity， thermal stability， and selectivity， among other characteristics[11]. Among heteropoly acids， phosphotungstic acid （HPW） has been proven to be the strongest Br?nsted acid to release protons （H+） and break the β-1，4-glycosidic bonds in cellulose[12]. In this study， HPW was used to hydrolyze cellulose. Because the single factor experimental design suffers from randomness， discontinuity， and other such problems， the authors attempted to seek a more effective design to optimize the process parameters and to improve the accuracy of the experimental results. Response surface methodology （RSM）， also called the regression design method， involves the use of multiple regression analyses of two regression methods to fit the relationship between the factors and the response values. It can also be used to analyze and optimize the regression equation and to obtain the optimal process conditions for the multivariable and statistical method of optimizing the response value[13].
Therefore， in this study， the authors utilized the RSM to optimize the amount of HPW （expressed as the H+ molar quantity）， reaction temperature， reaction time， and solid to liquid ratio for designing a cellulose hydrolysis reaction scheme by using HPW. The aim is to achieve higher-performance MCC with increased crystallinity.
2 Experimental
2.1 Materials
The raw material used in the experiment was eucalyptus kraft dissolving pulp， sourced from Shandong province， China. The crystallinity of the dissolving pulp is 56.5%. Phosphotungstic acid （H3PW12O40， HPW）， ethanol， and diethyl ether were bought from Sinopharm Chemical Reagent Co.， Ltd. （Guangzhou， China）. All of the chemicals were analytically pure and used as received.
2.2 Preparation of MCC
The reaction was carried out in a 250 mL three-necked flask equipped with a stirrer； the speed was set to 120 r/min. The flask was kept in a thermostatic water bath. Then， 0.5 g of the dissolving pulp （oven-dried）， HPW （7～12 mmol/L）， and deionized water were added at a temperature range of 70～100℃. The reaction time ranged from 1 h to 3 h. After hydrolysis， the product was first filtered by a glass filter （G4， filter bore diameter was in the range of 5～15 ?m）， and then successively washed with deionized water and ethanol until it reached a neutral pH value. The resulting products were stored for further processing.
2.3 Methods
2.3.1 Establishment of the response surface method
Through the previous single factor experiment[14]， the following optimum technological conditions for hydrolyzing cellulose by HPW could be obtained： H+ molar quantity， 30 mmol/L； reaction temperature， 90℃； reaction time， 2.0 h； and solid to liquid ratio， 1：40. Using the central composite design （CCD）， H+ molar quantity （A）， reaction temperature （B）， reaction time （C）， and solid to liquid ratio （D） were chosen as the four independent variables. The crystallinity of MCC （R） was defined as response value to optimize the experimental process parameters and determine the maximum value of crystallization under the optimum conditions. The experimental factors and level encoding are listed in Table 1. According to Table 1， using Design-Expert 7 software （Stat-Ease Inc.， 7.1， Minneapolis， MN， USA） to design 4 factors 5 levels and a total of 26 experiments and the order of the experiments was randomly disrupted for reducing the error.
Table 1 Factors and levels coding table
Factors Level
-2 （-a） -1 0 1 2 （+a）
A/（mmol·L-1） 27.0 28.5 30.0 31.5 33.0
B/℃ 80 85 90 95 100
C/h 1.50 1.75 2.00 2.25 2.50
D 1∶30 1∶35 1∶40 1∶45 1∶50
2.3.2 Analytical methods
The reliability of the regression model was determined by variance analysis using Design-Expert 7 software.
The surface morphologies of the raw material pulp and MCC were characterized by using an environmental scanning electron microscope （ESEM， Q45， FEI， Hillsboro， USA）. The samples were treated for gold sputtering using an ion sputtering apparatus （ETD， COXEM， Daejeon， Korea） and analyzed at 20 kV.
The particle size distribution of raw material pulp and MCC were determined by using a BT-9300H laser particle size analyzer （Dandong， China）. The samples were dispersed in distilled water and ultrasonically treated for 1 min.
The crystalline structures of the raw material pulp and MCC were investigated by using the Japanese Neo Confucianism X-ray diffraction （XRD） spectrometer （Rigaku， Tokyo， Japan）， which used a Ni-filtered Cu-Ka radiation source （λ=0.1518 nm）. The diffraction angle 2θ scan range was 10°～40°and the scanning speed was 2°/min. The crystallinity （CI） of MCC was measured using Eq.（1）[15].
（1）
Where Fa and Fc refer to the area of crystalline and amorphous regions， respectively.
The thermal stability of the samples was analyzed by thermogravimetric analysis （TGA， STA449 F3， Bruker， Germany）. The heating temperature was increased from 25℃ to 600℃ at a heating rate of 10℃/min.
3 Results and discussion
3.1 Experimental results of RSM
RSM was used to design the experimental scheme.
Table 2 Experimental results by RSM
Serial
number Level Response value
（Crystallinity）
A B C D Actual
value R1/% Predicted
value R2/%
1 0 0 0 0 76.29 76.44
2 1 1 1 1 74.02 74.62
3 -1 1 1 -1 74.91 73.67
4 -1 1 -1 -1 72.59 73.17
5 0 2 0 0 75.41 75.36
6 0 0 0 0 76.50 76.44
7 1 -1 1 -1 73.40 73.86
8 0 0 0 -2 71.71 71.98 9 1 1 -1 -1 73.01 74.01
10 1 -1 -1 -1 71.94 71.11
11 -1 -1 1 -1 70.53 70.77
12 1 1 1 -1 77.32 76.76
13 2 0 0 0 77.78 77.38
14 1 1 -1 1 73.69 74.16
15 -1 1 1 1 73.31 73.59
16 1 -1 1 1 70.86 69.88
17 -1 -1 -1 1 69.51 70.64
18 0 0 2 0 70.61 71.44
19 0 0 0 2 70.44 70.21
20 0 0 -2 0 71.26 70.48
21 0 -2 0 0 67.62 67.71
22 -1 -1 -1 -1 70.50 70.26
23 -1 -1 1 1 69.35 68.85
24 -1 1 -1 1 76.46 75.38
25 -2 0 0 0 75.16 75.51
26 1 -1 -1 1 68.92 69.41
Experimental results obtained under different conditions are shown in Table 2.
3.2 Establishment and analysis of regression model
The regression equation was established on the basis of the data given in Table 2. Design-Expert 7 software was used to carry out two multivariate linear regression fitting. After optimization， the response value （crystallinity） for the independent variables （four factors） was obtained. The regression equation was as follows （Coded representation）：
R=76.44+0.47A+1.91B+0.24C–0.44D+0.56AC– 0.52AD+0.46BD–0.57CD–1.23B2–1.37C2 –1.34D2（2）
The results of variance regression analysis obtained Table 3 Variance analysis of regression model
Variance Sum of
squares Liberty Mean
square F P Conspicuousness
Model 187.29 11 17.03 22.69 <0.0001 Significant
A 5.26 1 5.26 7.01 0.0191
B 87.71 1 87.71 116.87 <0.0001
C 1.39 1 1.39 1.85 0.1947
D 4.70 1 4.70 6.26 0.0253
AC 5.06 1 5.06 6.75 0.0211
AD 4.28 1 4.28 5.71 0.0315
BD 3.40 1 3.40 4.54 0.0514
CD 5.24 1 5.24 6.99 0.0193
B2 33.10 1 33.10 44.11 <0.0001
C2 41.39 1 41.39 55.15 <0.0001
D2 39.30 1 39.30 52.37 <0.0001
Residual 10.51 14 0.75 — —
Lack of fit 10.48 13 0.81 36.58 0.1288 Not significant
Net error 0.02 1 0.02 — —
The sum 197.80 25 — — —
Notes： Homogeneity test of variance P （Prob.>F） could be used to determine the significant impact of the various factors that affected the results of the test. When P<0.0001， the difference was extremely significant； when P<0.05， the difference was significant. In the test of variance analysis， Response surface variation coefficient （C.V.）=1.19%， R2=95%， Adj. R2=91%， Pred. R2=81%， Signal-to-noise ratio Adeq. Precision = 16.42.
by response surface optimization are shown in Table 3.
In general， when the determination coefficient R2 of the regression model was closer to 100%， the dependent and independent variables showed a better linear correlation. In this model， R2 = 95%， which showed that the model could explain the 95% change of response value， owing to the use of four variables A， B， C， and D. C.V. was 1.19%， which showed that the experimental results had a high precision， and the experimental operation was reliable[16]. Fig.1 shows a straight-line plot between the predicted and experimental results， which means that this model can be used because the experimental results and the model predictions were in a high degree of coincidence. Moreover， from Table 3， it is clear that the size of the F value does not exert a very obvious effect on the molar quantity of H+ and hence on crystallinity of MCC. If the influence of the H+ molar quantity （A） on the crystallinity of MCC is not considered， then the other three factors influence the crystallinity of MCC in the following order of significance of degree： reaction temperature （B）， solid to liquid ratio （D）， and reaction time （C）. Fig.1 Actual value and predicted value of Crystallinity
For the residual analysis of the regression model， the residual normal graph was closer to a straight line， and the regression model fit better. Fig.2 showed that these residuals lie on a straight line and that the dispersity was small which illustrate that the above model was reasonable. Meanwhile， the plot with different colors used in Fig.1 and Fig.2 was based on changes of crystallinity.
Fig.2 Standardized residual and normal probability of crystallinity
3.3 Response surface and contour analysis
A 3D picture of the response surface and the corresponding contour map can be used to directly show the effect of interactions between the two factors on the response value. The following graphics were all of the variables A， B， C， and D and the influence of any two factors on the crystallinity of MCC. If any two factors were set to 0 levels， the interaction between the other two factors that influences the response value R on the 3D response surface and the contour were studied. Combined with Table 3 analysis， the P values of AC， AD， BC and BD were small， which showed that the response value was more significant. The response surface and contour lines of the four items were analyzed， and the results are shown in Fig.3～Fig.6.
Fig.3 showed that when the reaction temperature was fixed at 90℃ and the solid to liquid ratio was 1∶40， the crystallinity of MCC increased slightly along with an increase in the H+ molar quantity. When the reaction time was increased， the crystallinity of MCC first increased and reached the maximum value at about 2 h and then decreased. In addition， if the shape of the contour lines were not round， the interaction between the H+ molar quantity and the reaction time was judged to be significant. Based on the results of a previous study[14]， it is clear that the H+ molar quantity is the most important factor. According to the contour lines， density changes could be used to investigate whether the reaction time exerted more influence on the crystallinity of MCC than the H+ molar quantity. If the reaction time was further extended， it could improve the crystallinity of MCC. However， if the time was extended too long， it not only destroyed the cellulose crystalline region but also produced many by-products such as glucose， fructose， cellobiose， and cellotriose. In addition， formic acid， acetic acid， and other such acids can be generated during hydrolysis[17-18]. Although the content of the by-products was higher， the crystallinity of the MCC was lower. As Fig.3 showed， when the H+ molar quantity was 31.5 mmol/L and the reaction time was 2 h， the maximum value of crystallinity of MCC reached was 76.6%. If the reaction temperature was fixed at 90℃ and the reaction time of 2 h was used， then the increasing of the H+ molar quantity slightly increased the crystallinity of MCC （Fig.4）. This is consistent with the results shown in Fig.3. As the solid to liquid ratio was increased， the crystallinity of MCC first increased and then decreased； when the solid to liquid ratio was 1∶38， the crystallinity of MCC reached the maximum value. From the shape of the contour， it could be seen that the interaction between the H+ molar quantity and the solid to liquid ratio exerted a significant impact on the crystallinity of MCC. The contour line density showed that the solid to liquid ratio exerted more influence than the H+ molar quantity on the crystallinity of MCC. This was because when the solid to liquid ratio was too low， the HPW could not fully come into contact with the cellulose material； therefore， the reaction was not complete. If the solid to liquid ratio was too high， the HPW could not remove the amorphous regions of cellulose as much as possible， which decreased the crystallinity of MCC. The crystallinity of MCC could reach the maximum value of 76.7% when the H+ molar quantity was 31.5 mmol/L and the solid to liquid ratio was 1∶38.
When the H+ molar quantity was fixed at 30 mmol/L and
a reaction time of 2 h was used， along with the increasing of the reaction temperature and the solid to liquid ratio， the crystallinity of MCC first increased and then decreased， as shown in Fig.5. Because the shape of the contour lines was almost circular， this interaction had a lesser significant effect on the crystallinity response value than the above interaction term of AD and AC. The change of the contour line density showed that the reaction temperature exerted more influence on the crystallinity of MCC than the solid to liquid ratio. In general， increased temperature could intensify the reaction， which made the HPW come into full contact with cellulose， especially when amorphous region degraded and MCC crystallinity improved. However， if the temperature was too high， then the HPW would also hydrolyze the cellulose amorphous region. As a result， the crystalline area would inevitably experience a certain degree of destruction， which decreased the crystallinity of MCC. In Fig.5， the effect of the solid to liquid ratio on the crystallinity of MCC was consistent with the result shown in Fig.4. Therefore， by fixing the other two factors and maintaining the reaction temperature at 93℃ and the solid to liquid ratio to 1∶38， MCC showed the best crystallinity value of 76.2%. When the H+ molar quantity was fixed at 30 mmol/L and the reaction temperature was maintained at 90℃， then extending the reaction time and increasing the amount of water would first increase the crystallinity of MCC and then decrease it （Fig.6）. From the shape， size， and color depth of the contour， it is clear that this interaction was less significant to the crystallinity of MCC compared to the above three interaction terms. When the contour line density was changed， the solid to liquid ratio exerted more influence on the crystallinity of MCC than the reaction time. The effect of each factor on the crystallinity of MCC was consistent with the above analysis. When the reaction time was 2 h and the solid to liquid ratio was 1∶38， the best crystallinity value of MCC was greater than 75.9%.
The effect of each factor on the crystallinity of MCC and the response value significance is shown in Table 3， and the influence of various interaction terms is shown in Fig.3～Fig.6. The factors influenced the crystallinity of MCC in the following order： H+ molar quantity （A）， reaction temperature （B）， solid to liquid ratio （D）， and reaction time （C）.
3.4 Experimental validation of regression model
Based on the above experimental results， it can be expected to achieve a higher crystallinity of MCC through the optimization of the model. The process conditions for MCC were considered by modifying the actual situation and the convenience of the experimental operation. Under these conditions， three parallel experiments were performed and the average of the measured values of the crystallinity of MCC was obtained. The results are shown in Table 4.
Table 4 Optimal factors and results of crystallinity
Parameter H+ molar
quantity
/（mmol·L-1） Reaction temperature
/℃ Reaction
time
/h Solid to liquid ratio Crystallinity /%
Predicted value 31.4 93.1 2.1 1∶38 77.8
Actual value 31.0 93.0 2.0 1∶38 77.4
3.5 SEM observation
The morphologies of the raw material pulp and MCC were investigated by SEM （Fig.7a and Fig.7b）. Though treated by HPW， the ribbon shape of the raw material pulp broke and degraded， as shown in Fig.7b， where the MCC became short and rod-like. In addition， the particle size distribution and the particle size of the raw material pulp and MCC are shown in Fig.7c. The fiber length of MCC was smaller than that of the raw material pulp， which was consistent with the SEM results. 3.6 XRD analysis
The XRD patterns of raw material pulp and MCC are shown in Fig.8. They exhibited similar diffractograms， where the peaks at 2θ=14.9°， 16.5°，22.5°， and 34.7° accorded with the 1-10， 110， 200， and 040 planes， respectively. According to previovus studies[4，19]， these typical peaks corresponded to cellulose I β structure and were used for determining the crystalline area. Compared with the raw material pulp， the crystallinity of MCC increased remarkably from 56.4% to 77.4%. The structure of MCC did not change after hydrolyzation. The peak at 2θ=22.5°became sharper， indicating that the MCC had a perfect crystal lattice in contrast to the raw material pulp， which resulted in MCC with a higher crystallinity[20-21].
Fig.8 XRD of raw material pulp and MCC
3.7 Thermal stability analysis
The thermogravimetric analysis （TGA） curves for raw material pulp and MCC are shown in Fig.9. The raw material pulp began to decompose at 245℃， and the MCC probably initially degraded at 270℃. When the temperature exceeded 350℃， weight loss occurred because the thermal stability of the material is related to its crystalline property. The higher the crystallinity， the better the thermal stability， and the higher the temperature required for degradation[22]. Hence， the crystallinity of MCC is higher than that of the raw material pulp. The results were consistent with the XRD analysis.
Fig.9 TGA thermograms of raw material pulp and MCC
4 Conclusions
In this paper， the single factor experimental results were optimized by response surface methodology. The analysis of variance showed that this model was available， and the significance of each factor on the crystallinity of MCC was explored. Finally， the MCC with higher crystallinity was obtained under the optimized reaction condition. Specific conclusions were as follows.
4.1 The variance analysis showed that the model could correctly predict the experimental results. Each factor and the interaction between the two factors exerted different effects on the response value. Each factor affected the crystallinity of MCC in the following order by degree of significance： H+ molar quantity， reaction temperature， solid to liquid ratio， and reaction time.
4.2 The optimum experimental conditions for preparing high-crystallinity MCC by HPW hydrolysis were as follows： H+ molar quantity， 31 mmol/L； reaction temperature， 93℃； reaction time， 2 h； and solid to liquid ratio， 1∶38. Under these conditions， the maximum value of crystallinity of MCC was 77.4%. 4.3 The SEM images showed the changes between the raw material pulp and MCC. The XRD showed that they had the same structure. TGA showed that MCC had higher crystallinity than the raw material pulp.
Acknowledgments
This work was supported by the National Key Research and Development Program of China （2017YFB0307900）， the Key Research and Development Project of Shaanxi Province （2017ZDXM-SF-090） and the State Key Laboratory of Donghua University （NO. LK1601）.
References
[1] Rodrigues F G， Monteiro D S， Cdas M， et al. Synthesis and characterization of cellulose acetate produced from recycled newspaper[J]. Carbohydrate Polymers， 2008， 73（1）： 74-82.
[2] Araki J， Wada M， Kuga S， et al. Flow properties of microcrystalline cellulose suspension prepared by acid treatment of native cellulose[J]. Colloids & Surfaces A： Physicochemical and Engineering Aspects， 1998， 142（1）： 75-82.
[3] El-sakhawy M， Hassan M L. Physical and mechanical properties of microcrystalline cellulose prepared from agricultural residues[J]. Carbohydrate Polymers， 2007， 67（1）： 1-10.
[4] Adel A M， El-Wahab Z H A， Ibrahim A A， et al. Characterization of microcrystalline cellulose prepared from lignocellulosic materials. Part II： Physicochemical properties[J]. Carbohydrate Polymers， 2011， 83（2）： 676-687.
[5] Wang D， Shang S B， Song Z Q， et al. Evaluation of microcrystalline cellulose prepared from kenaf fibers[J]. Journal of Industrial and Engineering Chemistry， 2010， 16（1）： 152-156.
[6] Kalita R D， Nath Y， Ochubiojo M E， et al. Extraction and characterization of microcrystalline cellulose from fodder grass； Setaria glauca （L） P. Beauv， and its potential as a drug delivery vehicle for isoniazid， a first line antituberculosis drug[J]. Colloids and Surfaces B： Biointerfaces， 2013， 108（4）： 85-89.
[7] Zhang M Y， Qiang D D， Li J B， et al. Changes of Structure and Morphological Characteristics of Cellulose during Preparation of Microcrystalline Cellulose[J]. China Pulp & Paper， 2016， 35（6）： 28-32.
[8] Ciolacu D. On the supramolecular structure of cellulose allomorphs after enzymatic degradation[J]. Journal of Optoelectronics and Advanced Materials， 2007， 9（4）： 1033-1037.
[9] Agblevor F A， Ibrahim M M， El-Zawawy W K. Coupled acid and enzyme mediated production of microcrystalline cellulose from corn cob and cotton gin waste[J]. Cellulose， 2007， 14（3）： 247-256.
[10] Tang Y J， Shen X C， Zhang J H， et al. Extraction of cellulose nano-crystals from old corrugated container fiber using phosphoric acid and enzymatic hydrolysis followed by sonication[J]. Carbohydrate Polymers， 2015， 125： 360-366. [11] Deng W P， Zhang Q H， Wang Y. Polyoxometalates as efficient catalysts for transformations of cellulose into platform chemicals[J]. Dalton Transactions， 2012， 41（33）： 9817-9831.
[12] Siró I， Plackett D. Microfibrillated cellulose and new nanocomposite materials： a review[J]. Cellulose， 2010， 17（3）： 459-494.
[13] Adel A M， El-Shinnawy N A. Hypolipidemic applications of microcrystalline cellulose composite synthesized from different agricultural residues[J]. International Journal of Biological Macromolecules， 2012， 51（5）： 1091-1102.
[14] Qiang D D， Zhang M Y， Li J B， et al. Selective hydrolysis of cellulose for the preparation of microcrystalline cellulose by phosphotungstic acid[J]. Cellulose， 2016， 23（2）： 1199-1207.
[15] Park S， Baker J O， Himmel M E， et al. Cellulose crystallinity index： measurement techniques and their impact on interpreting cellulase performance[J]. Biotechnology for Biofuels， 2010， 3（1）： 1-10.
[16] Duret X， Fredon E， Masson E， et al. Optimization of acid pretreatment in order to increase the phenolic content of Picea abies bark by surface response methodology[J]. BioResources， 2013， 8（1）： 1258-1273.
[17] Ma Y H， Ji W Q， Zhu X， et al. Effect of extremely low AlCl3 on hydrolysis of cellulose in high temperature liquid water[J]. Biomass Bioenergy， 2012， 39（39）： 106-111.
[18] Zhuang X S， Wang S R， Yuan Z H， et al. Research on biomass hydrolysis under extremely low acids by HPLC[J]. Acta Energiae Solaris Sinica， 2007， 28（11）： 1239-1243.
[19] French A D. Idealized powder diffraction patterns for cellulose polymorphs[J]. Cellulose， 2014， 21（2）： 885-896.
[20] Shimizu K， Furukawa H， Kobayashi N， et al. Effects of Bronsted and Lewis acidities on activity and selectivity of heteropolyacid-based catalysts for hydrolysis for hydrolysis of cellobiose and cellulose[J]. Green Chemistry， 2009， 11（10）： 1627-1632.
[21] Lu Q L， Tang L R， Lin F C， et al. Preparation and characterization of cellulose nanocrystals via ultrasonication-assisted FeCl3-catalyzed hydrolysis[J]. Cellulose， 2014， 21（5）： 3497-3506.
[22] Yang P Y， Kokot S. Thermal analysis of different cellulosic fabrics[J]. Journal of Applied Polymer Science， 1996， 60（8）： 1137-1146.