3.1. Physicochemical properties of synthesized catalysts
3.1.1. Textural properties
N2 adsorption-desorption isotherms and pore size distribution of the synthetized catalysts are shown in Fig. 2A and 2B, respectively. The isotherm corresponding to the Z-Ex catalyst exhibits an abrupt change in slope at relative pressures bellow 0.10 and a linear region from 0.10 to 0.50 before reaching a plateau. This kind of isotherm is situated between types I and IV, according to the IUPAC classification [85]. The pore size distribution covers a wide and heterogeneous range centered at 19 Å, indicating the contribution of micropores, supermicropores and mesopores [86]. A similar type of isotherm is observed for Z-EC material. In contrast, Z-Cal catalyst reports a well-defined IV type isotherm with an accentuated H2-type hysteresis loop at relative pressures from 0.40 to 0.65. This suggests the presence of a disordered structure with wormhole-like and interconnected mesopores [87, 88]. The pore size distribution shows a broad band between 20 and 60 Å with a maximum at 37 Å.
Table 2 summarizes the textural properties calculated from isotherms, such as specific surface area (SBET), average pore diameter (DP), total (VP) and micropore volume (VµP). The Z-Ex catalyst reported the highest BET surface area and total pore volume, reaching values of 323 m2 g− 1 and 0.21 cm3 g− 1, respectively. Z-Ex also obtained the smallest average pore diameter (DP) and a considerable micropore volume of 0.12 cm3 g− 1. The high specific surface area exhibit by the Z-Ex catalyst could be due to a strong hydrogen bonding interaction between the zirconium propoxide molecules and the aggregated entities of Pluronic P123 in isopropanol during hydrolysis-condensation process [89].
Table 2
Textural and crystalline properties of synthesized catalysts
Catalyst
|
SBETa (m2 g-1)
|
VPb (cm3 g-1)
|
VµPc (cm3 g-1)
|
DPd (Å)
|
Dhkle (Å)
|
Z-Ex
|
323
|
0.21
|
0.12
|
19
|
-
|
Z-Cal
|
108
|
0.09
|
0.00
|
37
|
68
|
Z-EC
|
118
|
0.07
|
0.04
|
23
|
97
|
a Specific surface area determined by BET method. |
b Pore volume determined by single point at P/P0 = 0.975. |
c Micropore volume determined by t-plot method. |
d Average pore diameter determined by BJH method on adsorption branch. |
e Crystalline average size calculated by Debye-Scherrer equation from XRD diffractograms. |
On the other hand, calcination and extraction-calcination methods significantly reduce the surface area with respect to Z-Ex catalyst, by 66.6% and 63.5%, respectively. In the same way, a decrease in pore volume is manifested, being 0.09 cm3 g− 1 for Z-Cal and 0.07 cm3 g− 1 for Z-EC. It is noted that thermal treatment causes micropores collapse during crystallization and increases the average pore diameter, with the consequent reduction of specific surface area. This effect was more evident with direct calcination, thus Z-Cal exhibited the largest DP and null micropore volume was determined.
It is highlight that the refluxing extraction method allows the template removal avoiding the destruction of micropore structure and preserving the favorable textural properties, such as the high specific surface area and pore volume.
3.1.2. Crystalline properties
X-Ray diffraction measurements were performed to analyze the crystalline structures formed after template removal procedures. Figure 3 shows the diffractograms obtained for synthesized catalysts. Extracted solid (Z-Ex) shows a weak broad peak at 2θ range from 20 to 40 degrees and centered at about 30 degree. Deshmane et al. [90] attributed this peak to extremely small tetragonal phase nanocrystallites formed on hydrous zirconia.
On the other hand, the thermally treated catalysts at 400°C (Z-Cal and Z-EC) exhibit peaks at 2θ values of 30.3°, 34.8°, 35.3°, 50.4°, 50.7°, 59.6°, 60.2° and 62.9° corresponding to diffraction planes (101), (002), (110), (112), (200), (103), (211) and (202) of pure tetragonal ZrO2 [91]. This result is consistent with those reported in literature [71, 74, 92], in which metastable tetragonal zirconia phase was obtained at relatively low temperatures (400–600°C) by sol-gel method. The stability of this phase is attained by controlling the synthesis parameters of hydrous zirconia during sol-gel procedure. Then, zirconia can transform into the thermodynamically stable monoclinic phase upon calcination above 600°C [90].
The crystalline size was calculated by Debye-Scherrer equation (Eq. 1) by using the most intense peak located at 2θ value of 30.3 ° and corresponding to (101) diffraction plane. As shown in Table 2, average crystallite sizes of 68 Å and 97 Å were determined for Z-Cal and Z-EC catalysts, respectively. The difference in crystallite size between both methods may be a consequence of partial template removal prior to calcination, which produces more thermally unstable structures against crystal growth.
3.1.3. Acidic properties
The acidic properties of heterogeneous catalysts such as, amount, density and strength of acid sites are key parameters for the selective hydrothermal conversion of sugars. Therefore, acidic properties of synthesized catalysts were determined to explain their catalytic performance. Two acidity measurement techniques were carried out: temperature programed desorption of pyridine (TPD-Py) and potentiometric titration with n-butylamine (PTB). These procedures are performed under very different environments. The TPD-Py experiments were done under inert gas phase at temperatures between 150°C and 850°C. On the other hand, PTB was performed in liquid phase at room temperature [81, 93]. Unlike TPD-Py, very weak and/or thermosensitive acid sites can be measured under these milder conditions.
The total amounts of acid sites determined by both techniques are compared in Table 3. In all synthesized catalysts, the values determined by PTB resulted higher than those determined by TPD-Py. The Z-Ex catalyst exhibits the highest amount of acid sites according to both techniques, being 359 and 340 µmol g− 1measured by PTB and TPD-Py methods, respectively. These values differ by only 5.3%, indicating that there is a very small amount of acid sites that are not detected by TPD-Py because they are excessively weak or eliminated at temperatures below 150°C, which is the minimum desorption temperature in TPD-Py experiments. In contrast, the total amounts of acid sites determined in thermally treated materials (Z-Cal and Z-EC) are much lower than those reported for Z-Ex, and resulted considerably different between both measurement techniques. Since these solids were treated at temperatures up to 400°C, the existence of thermosensitive acid sites is not probably. Thus, the number of acid sites not detected by TPD-Py but quantified by PTB could be very weak acid sites.
Table 3
Total amount and density of acid sites measured by different techniques
Catalyst
|
TPD-Pya
|
PTBb
|
Amount (µmol g− 1)
|
Densityc (µmol m− 2)
|
Amount (µmol g− 1)
|
Densityc (µmol m− 2)
|
Z-Ex
|
340
|
1.053
|
359
|
1.111
|
Z-Cal
|
15
|
0.139
|
49
|
0.454
|
Z-EC
|
17
|
0.157
|
34
|
0.288
|
a Temperature-programmed desorption of pyridine (150–850°C). |
b Potentiometric titration with n-butylamine. |
c Ratio between amount of acid sites and BET specific surface area. |
Due to the significant differences in specific surface area between the catalysts, acid sites density was used for a more appropriate comparison, as seen in Table 3. Z-Ex catalyst presented an acid sites density of 1.111 µmol m− 2 measured by PTB, while Z-Cal and Z-EC catalysts showed an abrupt reduction to 0.454 and 0.288 µmol m− 2, respectively. According to TPD-Py, this difference was even more notably, resulting the acid sites density in Z-Ex catalyst more than 7.5 times higher than Z-Cal and 6.7 higher than Z-EC.
There is not an extensively difference in amount or density of acid sites between both thermally treated materials, while the extracted catalyst stood out showing enhanced acidic properties. To explain the different acidic properties in the synthesized catalysts, it is important to consider that the thermal treatment at 400°C transforms hydrous zirconia to tetragonal phase zirconia. The condensation of adjacent Zr-OH groups from hydrous zirconia and their elimination as water, allow the crystallization and the consolidation of the ZrO2 structure [90, 94]. This structure formation removes most of the acid sites on catalyst surface, which are Lewis acid sites formed by framework defects and Brønsted acid sites originated by chemically adsorbed water [31, 58].
3.1.4. Acidic strength distribution
The potentiometric titration curves for Z-Ex, Z-Cal and Z-EC catalysts are shown in Fig. 4. Total and partial amount of acid sites were calculated by the first derivative method. The first derivative curves and deconvoluted peaks are presented in Fig. S2 (Online Resource). Since the catalysts contain acid sites of different nature and, consequently, different acidic strength, several local maximum points along the curves were detected [81, 82]. Each peak center value correspond to a partial amount of acid sites and its strength is directly related to the potential difference (Ei) reached [95–97]. The acid sites were classified according to acidic strength as weak, moderate and strong, corresponding to a potential difference range of Ei < 10 mV, 10 mV < Ei < 60 mV and Ei > 60 mV, respectively. The acidic strength distribution in the synthesized catalysts was summarized in Table 4.
Table 4
Acid sites strength classification by potentiometric titration with n-butylamine (PTB) measures
Catalyst
|
Max Ei
|
Strength and amount of acid sites (µmol g− 1)
|
Weak (Ei < 10 mV)
|
Moderate (10 < Ei < 60 mV)
|
Strong (Ei > 60 mV)
|
Total
|
Z-Ex
|
183
|
188
|
82
|
89
|
359
|
Z-Cal
|
82
|
32
|
8
|
9
|
49
|
Z-EC
|
90
|
21
|
3
|
10
|
34
|
The acid sites distribution in Z-Ex catalyst was 52.4% weak, 22.8% moderate and 24.8% strong. The thermally treated catalysts had a higher proportion of weak acid sites and did not show substantial differences between them. The resultant distribution was 65.3% weak, 16.3% moderate and 18.4% strong for Z-Cal, and 61.8% weak, 8.8% moderate and 29.4% strong for Z-EC. In the same way, the initial electrode potential values reached were strongly different between the extracted and the thermally treated catalysts, being 183, 82 and 90 mV for Z-Ex, Z-Cal and Z-EC, respectively. The initial electrode potential determines the maximum strength of the acid sites [95]. Indeed, the high value attained for Z-Ex indicates that strong acid sites could be classified as very strong according to the criteria adopted by other authors (Ei > 100 mV) [96, 98, 99].
Figure 5 shows the TPD-Py curves for Z-Ex (Fig. 5A), Z-Cal (Fig. 5B) and Z-EC (Fig. 5C) catalysts. Analogous to PTB method, several peaks were observed at different desorption temperatures (Td), corresponding to acid sites with different strength. In this case, the acid sites strength was classified according to desorption temperature range of 150°C < Td < 350°C, 350°C < Td < 500°C and Td < 500°C, in weak, moderate and strong, respectively. The acidic strength distributions obtained by TPD-Py are shown in Table 5.
Table 5
Acid sites strength classification by pyridine temperature-programmed desorption (TPD) measures
Catalyst
|
Strength and amount of acid sites (µmol g− 1)
|
Weak (150 < T < 350°C)
|
Moderate (350 < T < 500°C)
|
Strong (T > 500°C)
|
Total
|
Z-Ex
|
257
|
64
|
19
|
340
|
Z-Cal
|
2
|
8
|
5
|
15
|
Z-EC
|
3
|
10
|
4
|
17
|
The Z-Ex catalyst presented an acid strength distribution concentrated in weak acid sites, representing the 75.6%, in addition to 18.8% moderate and 5.6% strong acidity. Regarding the acidic strength distributions in thermally treated catalysts, there is a major contribution of moderate and strong acidity respect to those measured by PTB. However, the low total amounts of acid sites achieved could indicate that Z-Cal and Z-EC catalysts contain undetected very weak acid sites.
The total amount of acid sites measured in Z-Ex catalyst resulted approximately the same by both assessment methods. In contrast, the acidic strength distribution was very dissimilar. To interpret these results, it is assumed that the acidic strength scales adopted for both techniques are compatible. In general terms, the sum of moderate and strong acid sites measured with PTB method, i.e., all acid sites reaching an electrode potential above 10 mV, was 171 µmol g− 1. This value was twice that determined by TPD-Py. Within this group, strong acidity contribution was 4.7 times larger. The remarkable difference could indicate that the Z-Ex catalyst has strong but thermosensitive acid sites. According to the TPD-Py method, these acid sites could be misclassified as weak because are removed at a temperature bellow 350°C. As shown in Fig. 5A, there is a very intense peak centered at 315°C that contributes to 242 µmol g− 1 acid sites. Whereas the PTB method allows them to be detected within moderate and strong acidic strength ranges.
3.2. Catalytic performance
Synthesized zirconia-based catalysts were evaluated in the hydrothermal conversion of a multiple sugars solution whose was representative of de-starched wheat bran hydrolysate (1.81 g L− 1 glucose, 6.23 g L− 1 xylose and 3.13 g L− 1 arabinose). In order to select the most promising catalyst, the operating conditions were adopted according to those reported in the literature research for metal oxides in similar reactions [29, 54, 55]. In general, temperatures between 160 and 200°C, reaction times ranging 30 to 240 min, pressures between 25 and 35 bar, and catalyst/reactants mass ratios in the range of 0.25 to 0.80 g g− 1 were used. In this work, the catalytic evaluation was carried out under moderate conditions, at 180°C, 90 min, 30 bar and using a catalyst/reactants mass ratio of 0.35 g g− 1. Subsequently, the reaction conditions were optimized by means of an experimental design (Section 3.3).
Figure 6A shows individual and total sugars conversion and Fig. 6B the molar yields to furans (F), lactic (LA), glycolic (GA), formic (FA) and acetic (AA) acids, being the main products. A difference in catalytic behavior between non-calcined (Z-Ex) and thermally treated (Z-Cal and Z-EC) catalysts was clearly identified. As can be seen in Fig. 6A, the catalyst subjected to refluxing extraction (Z-Ex) achieved the highest total conversion, being 90.5%. On the other hand, the calcined catalysts, Z-Cal and Z-EC, reached a lower total conversion of 66.3% and 58.8%, respectively. When analyzing the conversion of sugars individually, it is observed that the difference between the solids was more pronounced in glucose conversion, being almost double with Z-Ex (79.1%) respect to thermally treated ones (43.9% and 40.5%). Moreover, xylose was the most converted sugar in the three cases, followed by arabinose and finally glucose, in agreement with their relative composition in the reactant solution.
On the other hand, Fig. 6B shows that the Z-Ex catalyst presented the highest molar yield towards LA, reaching a value of 47.8%, notably being the main product. In contrast, the Z-Cal and Z-EC catalysts hardly reached 16.9% and 15.2% LA molar yield, respectively. Instead, furanic compounds were the main products obtained with them. Z-Cal catalyst reached the highest molar yield towards furans (25.9%).
Regarding the organic acids by-products, glycolic, formic and acetic acids were obtained with Z-Ex catalyst, reaching molar yields of 15.8%, 12.1% and 4.1%, respectively. Other compounds detected among the reaction products, but not shown in Fig. 6B, were glyceraldehyde (GH) and levulinic acid (LeA), achieving low molar yields of 1.6% and 2.8%, respectively. With the thermally treated catalysts the by-products molar yields were lower and almost identical between both, averaging 12.2% of GA, 5.2% of FA, 1.9% of AA, 0.8% of GH and 0.5% of LeA. The triose GH was mainly obtained through the retro-aldol reaction, as seen in Fig. 1. In addition, GA and AA are probably generated by GH degradation via oxidation and decarboxylation reactions. On the other hand, FA and LeA production indicates the occurrence of furans degradation reactions.
It is important to note, that no significant differences were found in conversion and products yields between Z-Cal and Z-EC catalysts. In contrast, Z-Ex exhibited very high total conversion and a significantly higher LA molar yield. This indicates that thermal treatment, either by direct calcination or post-extraction, does not improve catalytic performance towards LA. These results are consistent with those obtained in acidity assessment and could be explained with the notably differences in amount, density and strength of acid sites observed (Section 3.1). As reported in numerous works [27, 32, 55], Lewis acidity is required in isomerization and retro-aldol condensation of sugars, while Brønsted acid sites are involved in dehydration step of glyceraldehyde and DHA towards pyruvaldehyde, a key intermediate in LA formation (Fig. 1). In this case, in Z-Ex catalyst, weak and moderate acid sites play a role in isomerization and retro-aldol reactions, while strong acidity highly promotes the dehydration steps. The latter includes the trioses conversion, as well as furanic compounds formation and subsequent degradation.
The synthesized catalysts performances were compared under identical reaction conditions with other reference catalysts. A commercial zirconium dioxide (m-ZrO2) with pure monoclinic phase and a low surface area of 6.9 m2 g− 1 was used as heterogeneous catalyst. In addition, an alkaline solution of NaOH 0.05 M (pH = 12) was evaluated as homogeneous catalyst. Finally, the non-catalyzed hydrothermal conversion was carried out. Table 6presents the results obtained for sugars conversion and selectivities towards the main products.
Table 6
Catalytic performance comparison of hydrothermal conversion of sugars solution
Catalysta
|
X
(total conversion %)
|
SPb (selectivity %)
|
LA
|
F
|
GA
|
FA
|
AA
|
LeA
|
GH
|
Non-catalyzed
|
34.7
|
5.4
|
71.0
|
6.9
|
7.2
|
3.3
|
0
|
2.0
|
m-ZrO2c
|
34.9
|
8.5
|
61.0
|
8.3
|
1.1
|
1.3
|
0.6
|
2.6
|
NaOH 0.05 M
|
93.7
|
25.8
|
6.3
|
27.1
|
16.2
|
16.8
|
1.2
|
1.5
|
Z-Ex
|
90.5
|
52.8
|
22.2
|
17.4
|
13.4
|
4.6
|
3.1
|
1.8
|
Z-C
|
66.3
|
25.5
|
39.1
|
18.8
|
7.1
|
2.9
|
0.5
|
1.1
|
Z-EC
|
58.8
|
26.0
|
40.8
|
20.2
|
9.6
|
3.2
|
1.0
|
1.4
|
a Reactant solution: glucose 1.18 g L− 1, xylose 6.23 g L− 1 and arabinose 3.13 g L− 1. Reaction conditions: temperature = 180°C; reaction time = 90 min, catalyst/reactants mass ratio = 0.35 g g− 1, initial pressure = 30 bar N2. |
b % Molar selectivities of LA: lactic acid, F: total furans, GA: glycolic, FA: formic, AA: acetic and LeA: levulinic acids, GH: glyceraldehyde. |
c Commercial monoclinic phase zirconium dioxide. |
The non-catalyzed hydrothermal reaction reached a conversion of 34.7% and was quite selective towards furfural and HMF production, with a selectivity of 71.0%. LA production was almost negligible, reaching a selectivity of 5.4%. This behavior was similar to that found with m-ZrO2 catalyst, resulting in a low conversion (34.9%) and selectivities of 61.0% and 8.5% towards furans and LA, respectively. These results indicated that monoclinic zirconium dioxide, under these conditions, did not have a substantial catalytic behavior in the reaction. On the other hand, the alkaline hydrothermal reaction reached a high conversion (93.7%) and the selectivity towards LA hardly reached 25.8%. A high production of organic acids, such as GA, FA and AA acids with selectivities of 27.1%, 16.2% and 16.8%, respectively, was obtained. Alkaline hydrothermal conversion is known to be acceptably selective towards LA production, however, high NaOH concentrations (> 1 M) and higher temperatures (> 200°C) are required to reach a selectivity above 50% [55].
On the other side, the Z-Ex catalyst was the most selective towards LA with a value of 52.8%, while the selectivity towards furans was almost 3 times lower than non-catalyzed reaction. A considerably high FA a LeA production is observed, with selectivities of 13.4% and 3.1% respectively. This indicates that higher amount of furans were indeed obtained, but that they were subsequently degraded due to the catalyst acidity. The LA selectivities attained with Z-C and Z-EC catalysts, were similar to those obtained in alkaline hydrothermal conversion, being 25.5% and 26.0%, respectively. However, the main products resulted to be furans, with selectivities of 39.1% and 40.8%, respectively.
The catalytic performance results obtained with the Z-Ex catalyst are promising compared to those reported in literature, where LA yields in the range of 30–45% are achieved with simple metal oxides [53–55]. It is important to highlight that the evaluation conditions in this work were a moderate temperature and a relatively short reaction time, while several authors reported temperatures as high as 200°C and reaction times up to 6 h. In this case, the use of homogeneous co-catalysts, such as NaOH, was not necessary to reach a conversion greater than 90 %. Inaddition, a multiple sugars solution was used as feedstock for the reaction, instead of a simple xylose or glucose solution. The complexity of the chemical reactions involved and the formation of various by-products make even more difficult to achieve a higher LA selectivity.
3.3. Statistical study results
The Z-Ex was the most outstanding synthesized catalyst, as it obtained the highest surface area, pore volume and amount of acid sites. These physicochemical properties make it an effective catalyst for hydrothermal conversion of sugars, as proved in the catalytic performance reaching a relative high LA yield. All comparison reaction tests were performed under moderate conditions based on literature research, therefore, the optimal operating variables for this specific catalyst and feedstock had to be found to maximize the LA production. In order to optimize the operating variables for Z-Ex catalyst, an experimental design was performed and detailed studied through the response surface methodology (RSM).
3.3.1. Response surface methodology (RSM)
The response surface methodology is a sequential procedure based on several statistical and mathematical techniques in order to find a suitable approximation of the function between a response and a set of independents variables [64]. A simplified schematic of the method is shown in Fig. S3 (Online Resource). The first step consist on planning and executing the appropriate experimental design to collect the input data. Preliminary conclusions about how the variables affect the response can be drawn with graphical analysis. Box-plots, factors interaction plots and simple XY plots are the available tools for the data interpretation. Subsequently, a function model is proposed, and its coefficients (effects) estimated with multiple linear regression fit (MLRF) by means of least squares method [100]. For a relatively bounded region, linear, quadratic and interactional terms, can approximate complex functions with insignificant error. Higher order terms can also be added if the experimental data volume allow it. To obtain the final regression models, an iterative method is used, starting with a full second-order model and then adding or removing effects until all statistical parameters specifications be accomplished. These statistical parameters, on which the rationality of the model depends, are the probability value (p-value), the lack of fit (LOF) and the multiple factors determination coefficient (R2).
The statistical significance of the effects in MLRF method was decided based on their corresponding standard errors and using the t-student distribution as the contrasting hypothesis. The coefficients accepted in the fitting model must obtain a p-value less than 0.05, indicating an influence on the response with a confidence above 95%. Non-significates effects, i.e., with p-value greater than 0.05, are removed from the model and the MLRF is recalculated. Additionally, the relevance of the model is supported with an Analysis of Variance (ANOVA). The degrees of freedom and the sums of mean squared errors are the statistical attributes calculated to perform the ANOVA. The estimated coefficients are grouped in linear, interactional, quadratic and cubic, and their statistical significance is decided using the Fischer-Snedecor statistic distribution (F) as hypothesis contrasting. Again, a p-value less than 0.05 must be found. Another p-value is also estimated for the LOF by building an F-value between the residuals variance and the pure experimental error. In contrast, LOF is not expected to be significant for a reliability model. The R2 is another vital parameter to determine the goodness of fit of the model. An R2 greater than 0.95 indicates that more than 95% of the data variability is explained by the fitting model.
The final verification of the model is carried out with several statistical test provided by the software. The residues analysis allows detecting the presence of outliers, deviations in normality of errors and inconsistency in the variance. The Cook’s distance method is performed to rule out the existence of influential data. Once the model is fully accepted, response surfaces and contour lines plots are drawn as a function of the variable factors. Mathematical methods are used to locate the stationary points (maximum or minimum) in which all the partial derivatives of the empirical equation are equal to zero. If the range selected to model the response is outside the area of maximum performance, the fastest ascent trajectory method can be used to redesign the experiment on the appropriate ranges.
3.3.2. Experimental design
All reaction tests results for the experimental design are presented in Table S1 (Online Resource). Of the total runs, 27 correspond to factorial points and 3 are replicates in central point located at A = 180°C, B = 90 min, C = 0.35 g g− 1 and D = 30 N2 bar. The replicates in the center of the design are necessary to obtain a variance estimation of the pure experimental error and calculate the lack-of-fit.
Although the aim of this statistical study is maximizing the LA production, the behavior of by-products formation should not be neglected. As shown in Table S1 (Online Resource), the output response factors were the individual (XGl, XXy and XAr) and total sugars conversion (X), lactic acid (YLA) and main by-products molar yields, which were furfural (YFu), HMF (YHMF), glyceraldehyde (YGH), glycolic (YGA), formic (YFA), acetic (YAA) and levulinic acids (YLeA). For a more simple analysis, the sum of furfural and HMF molar yields were included as furans molar yield (YF) in the global response.
The variable factors chosen for the experimental design were temperature (A), reaction time (B), catalysts/reactants mass ratio (C) and initial pressure (D) (Table 1). From previous literature, it has been observed that temperature, reaction time, catalyst amount and reactant concentration were the most influential operating parameters in sugars hydrothermal conversion. In this work, reactants concentration was not selected as a variable factor because it represents the real composition of a wheat bran hydrolysate. At industrial scale, diluting or concentrating the feedstock solution would result in extra production costs. Instead, the catalyst/reactants mass ratio was selected as a variable factor, keeping the sugars concentration constant.
Regarding the factors levels for the design, temperature was varied in a range from 140°C to 200°C. Some authors reported higher operating temperatures ranging between 200°C and 300°C, at short reaction times [35, 36]. However, this high temperature range was avoided because it was desired to work under moderate operating conditions. In the same way, the maximum reaction time was limited to 4 h [48, 51]. The system initial pressure was included among the variable factors. It is known that the self-dissociation constant of subcritical and supercritical water is affected by the system temperature and pressure. Nevertheless, no analysis of pressure influence on this specific reaction has been found. The initial pressure range selected was from 10 bar to 50 bar in order to operate in subcritical water conditions. The minimum initial pressure was set at 10 bar to avoid water evaporation that modifies the concentration of sugars in the liquid phase. In addition, further evaporation at lower pressures can cause sugars precipitation and form undesirable low temperature combustion products.
Within the 30 experimental runs, the highest LA yield of 57.4% with a total conversion of 97.5% was located at the design point A = 180°C, B = 240 min, C = 0.5 g g− 1 and D = 50 bar. It is highlighted that at the highest design temperature (200°C), LA molar yields were lower and slightly varied from 40.2% in 30 min, to 47.9% in 240 min.
3.3.3. Graphical analysis
Before continuing with the statistical study, it is important to graphically analyze the data obtained in order to facilitate the fit model proposal. Figure 7 shows the grouped averages of total (Fig. 7A) and individual (Fig. 7B) sugars conversion, LA molar yield (Fig. 7C) and total furans molar yield (Fig. 7D) as a function of adopted temperature levels. In addition, the lower and upper limits of the central quantiles of data are included. The vertical scatter of data at each level indicates the variability of the response due to the effect of reaction time and catalyst mass. However, the trends of average curves show that temperature is clearly the most influential factor in the three responses analyzed. The variability of data at 200°C is small in total conversion and LA yield responses, because maximum values are reached in short periods of time and the catalyst amount loses influence. On the other hand, it can be concluded that the optimal conditions to maximize LA yield are between 180°C and 200°C.
As shown in Fig. 7B, the grouped average conversion curves of xylose and arabinose as a function of temperature are quite similar. Nevertheless, glucose is notably more difficult to convert than pentoses at temperatures below 200°C. This difference was probably due to its low concentration in reactant solution relative to the other sugars, making competition for active sites unfavorable.
In a different way, it was observed that initial pressure is not an influential factor. The median values of all pressure levels are statistically equal as can be seen in a box-plot (Fig. S4, Online Resource). Furthermore, initial pressure does not have an interaction with any other variable factor, for example, catalyst/reactants mass ratio. This effect can be visualized in an interaction plot (Fig. S5A, Online Resource). It is highlighted that initial pressure did not seem to be significant for any of studied responses. Probably, the selected range does not allow appreciating a relevant difference in acidity contribution of pressurized subcritical water. On the other hand, temperature and catalyst/reactants mass ratio maintain a strong joint influence on total conversion of sugars (Fig. S5B, Online Resource).
3.3.4. Models fitting by Multiple Linear Regression Fit (MLRF)
The results obtained from experimental runs were fitted to different models by multiple linear regression fit (MRLF). These fitting models established a relation between the responses under study, total conversion (X), LA molar yield (YLA) and total furans molar yield (YF), with the variable factors. The final regression models were obtained by an iterative method until all statistical parameters specifications such as, p-values, lack-of-fit (LOF) and R2 coefficient, were accomplished.
Tables S2, S3 and S4 (Online Resource) compile the estimated coefficients with the MLRF method in the final fit models for conversion, LA yield and furans yield, respectively. All coefficients accepted in the fitting model obtained a p-value greater than 0.05, indicating an influence on the response with a confidence greater than 95%. The empirical equations containing the estimated coefficients are shown below:
\(X=88.92+12.40A+5.39B+3.36C-2.86AB-2.79{A}^{2}-1.10{B}^{2}\) (Eq. 8)
\({Y}_{LA}=46.72+14.70A+8.70B+2.48C-8.22{A}^{2}-2.08{B}^{2}-1.40{AB}^{2}-1.06{A}^{2}B-4.17{A}^{3}\) (Eq. 9)
\({Y}_{F}=19.58+2.10A-0.59B-0.92C-2.27AB-1.22AC+0.51A{B}^{2}-3.00{A}^{2}-0.70{A}^{3}\) (Eq. 10)
Where, X: sugars total conversion (%), YLA: LA molar yield (%), YF: total furans molar yield (%), A: temperature, B: reaction time and C: catalyst/reactants mass ratio. Variable factors are dimensionless coded.
The results showed that none of the responses is significantly dependent on the initial pressure (D). The fitted model of total conversion contains linear, interactional and second order terms as shown in Eq. 8. In contrast, the models obtained for LA and furans molar yields needed additional third order terms (Eqs. 9 and 10). It is also be noted that all three fitting models contains a positive linear coefficient for temperature effect (A). However, the linear terms for reaction time (B) and catalyst/reactants mass ratio (C) are positive for conversion and LA yield, but negative for the furans yield.
The other statistical parameters involved in model acceptability are presented in Tables 7, 8 and 9. The degrees of freedom, sum of squared errors, sum of mean squared errors and residuals variance resulting from ANOVA are shown. A p-value less than 0.05 was found for each group of coefficients. Additionally, the p-value for the lack of fit resulted greater than 0.05 for all cases indicating that cannot be considered significant. The estimated R2for the fitted models of total conversion, LA and furans yields were 0.9845, 0.9708 and 0.9506, respectively.
Table 7
ANOVA and statistics summary of total sugars conversion fitted model
Effects
|
Deg. of Freedom
|
Sum of squares
|
Mean squares
|
F value
|
P (> F)
|
Linear
|
3
|
8467
|
2822,2
|
451,1
|
2,2E-16
|
Interaction
|
1
|
375
|
375,2
|
60,0
|
7,5E-08
|
Quadratic
|
2
|
287
|
143,7
|
23,0
|
3,3E-06
|
Cubic
|
-
|
-
|
-
|
-
|
-
|
Residuals
|
23
|
144
|
6,3
|
-
|
-
|
Pure error
|
3
|
11
|
3,8
|
-
|
-
|
Lack of fit
|
20
|
133
|
6,6
|
1,8
|
3,6E-01
|
|
|
|
Multiple R2
|
0,9845
|
Table 8
ANOVA and statistics summary of lactic acid molar yield fitted model
Effects
|
Deg. of Freedom
|
Sum of squares
|
Mean squares
|
F value
|
P (> F)
|
Linear
|
3
|
4563
|
1521,0
|
199,6
|
1,4E-15
|
Interaction
|
2
|
196
|
98,0
|
12,9
|
2,2E-04
|
Quadratic
|
2
|
359
|
179,5
|
23,6
|
4,3E-06
|
Cubic
|
1
|
216
|
216,0
|
28,4
|
2,8E-05
|
Residuals
|
21
|
160
|
7,6
|
-
|
-
|
Pure error
|
3
|
4
|
1,3
|
-
|
-
|
Lack of fit
|
18
|
156
|
8,7
|
6,9
|
6,8E-02
|
|
|
|
Multiple R2
|
0,9708
|
Table 9
ANOVA and statistics summary of total furans molar yield fitted model
Effects
|
Deg. of Freedom
|
Sum of squares
|
Mean squares
|
F value
|
P (> F)
|
Linear
|
3
|
384
|
127,9
|
38,0
|
1,1E-08
|
Interaction
|
3
|
159
|
52,9
|
15,7
|
1,4E-05
|
Quadratic
|
1
|
133
|
133,3
|
39,6
|
3,1E-06
|
Cubic
|
1
|
16
|
16,2
|
4,8
|
4,0E-02
|
Residuals
|
21
|
71
|
3,4
|
-
|
-
|
Pure error
|
3
|
6
|
1,8
|
-
|
-
|
Lack of fit
|
18
|
65
|
3,6
|
2,0
|
3,2E-01
|
|
|
|
Multiple R2
|
0,9506
|
The verification of the final fitting models was performed with various statistical tests provided by the software. No outliers, influential data, deviations in normality of errors or inconsistency in the variance were detected. In addition, predicted values from the fit models were compared with the experimental data. The predicted conversion, LA and furans yields were found to be in a good agreement with the experimental values (Fig. S6, Online Resource). Only slight deviations from the data were detected for the furans yield model, which also obtained the lowest R2 value and the highest lack of fit of the three models. This probably be since the response studied includes both furfural and HMF molar yields.
3.3.5. Total sugars conversion model
Response surface plots are useful to analyze the behavior of the response as a function of the influencing variables, as well as to identify maximum yield points. These plots were obtained for each fitted model by means of their corresponding equations (Eqs. 8, 9 and 10).
Figure 8A shows the response surfaces for total sugars conversion as a function of temperature (A) and reaction time (B), parametric on catalyst/reactants mass ratio (C). The variables were decoded and are presented in their original units. Each color represents an invariable catalyst mass surface, being blue C = 0.175, green C = 0.35 and red C = 0.5 g g− 1.
The influence of temperature and reaction time is evident, since a strong linear increase and a slight deceleration of the curvature are perceived. Additionally, a positive influence of increasing catalyst mass is observed. For a C equal to 0.5, a complete conversion zone was situated at temperatures above 190°C and reaction times above 100 min, as summarized in Table 10. The complete conversion region can be seen more clearly in Fig. 9A, where the contour lines are shown as a function of A and B, for the plane section describing by a constant C factor equal to 0.5.
Table 10
Optimized variable factors for maximum predicted responses found by RSM
Response
|
Max value predicted
|
Optimized variable factorsa
|
A (°C)
|
B (min)
|
C (g g-1)
|
Lactic acid molar yield %
|
61.9
|
189
|
208
|
0.500
|
Total sugars conversion %
|
100.0
|
> 190
|
> 100
|
0.500
|
Furans molar yield %
|
24.6
|
200
|
30
|
0.175
|
a A: temperature, B: reaction time, C: catalyst/reactants mass ratio. |
3.3.6. LA molar yield model
Figure 8B shows the LA molar yield surfaces obtained with the respective fitted model equation. It is observed that the yield reaches a maximum at some point within the ranges of time and temperature studied. The influence of the catalyst/reactants mass ratio is positive, and the highest yield of LA was obtained for a value of C = 0.5 g g− 1. The optimum temperature and reaction time was found using the partial derivative method. The maximum yield point was located at A = 189°C and B = 208 min, and corresponds to a LA yield value of 61.9% as shown in Table 10. The contour lines for a constant C = 0.5 g g− 1 allows identifying this point more clearly, as shown in Fig. 9B.
To verify the prediction accuracy of the fitting model, an additional experiment was carried out under the optimal conditions for maximum LA yield. In Table 11, the results predicted by the model equations and those obtained in the experimental run at A = 189°C, B = 208 min, C = 0.5 g g− 1and D = 10 bar, are compared. Experimentally, a complete sugars conversion and a yield of 61.2 ± 0.5% towards LA was achieved. The resultant LA molar yield corresponds to an error of only 1.1% respect to the fitted model prediction. Regarding the furans production, the experimental run achieved a 15.8 ± 0.3% molar yield, which implies an error of 3.2% respect to the predicted value. The low prediction errors, confirms that the fitted models of conversion, LA and furans molar yields, predict with good accuracy the responses within the reaction conditions ranges analyzed. Moreover, the RSM allowed increasing the LA yield by optimizing the operating parameters.
Table 11
Predicted and experimental results at optimized reaction conditions for maximizing LA molar yield
Responsea
|
X
|
YLA
|
YF
|
Predicted %b
|
100.0
|
61.9
|
16.3
|
Experimental %c
|
100.0
|
61.2 ± 0.5
|
15.8 ± 0.3
|
Error %
|
0.0
|
1.1
|
3.2
|
a Total sugars conversion (X), lactic acid molar yield (YLA) and total furans molar yield (YF). |
b Operating conditions: temperature (A) = 189°C, reaction time (B) = 208 min, catalyst/reactants mass ratio (C) = 0.5 g g− 1, initial pressure (D) = 10 bar. |
c Experimental runs were made by duplicate. |
The catalyst effect is less significant than reaction temperature and time. At the optimized conditions point, an increase from C = 0.175 to 0.5 g g− 1 implies an increase of only 4.9% yield points. However, as observed in Section 3.2, there is almost not LA formation from sugars in the absence of catalyst. On the other hand, the temperature is the most significant factor. At a constant temperature of 160°C, the LA molar yield increase from 22.2% in 30 min to 45.2% in the maximum reaction time (240 min), achieving an increase of 23.1% points. At 200°C, a yield of 40.4% is reached with the shortest design reaction time. However, along the whole reaction time range, LA yield increment is smaller and does not exceed 15.3% points. The maximum molar yield is reached at 156 min corresponding to 55.7%. This behavior is probably due to the contribution of side reactions. As show in Fig. 1, the direct dehydration of sugars towards furans competes with the target reaction mechanism, the retro-aldol condensation to obtain LA, since it takes place during non-catalyzed hydrothermal conversion at high temperatures and is strongly promoted by the presence of acid sites.
3.3.7. Furanic compounds model
The response surfaces for the furans molar yield model were plotted from the respective equation (Eq. 10). Figure 8C shows the graph obtained as a function of A, B and C factors. It is observed that the furans yield increases along time at low temperatures, but above a certain temperature value, it decreases due to degradation reactions. This effect can be observed in the contour lines presented in Fig. 9C at a constant C factor of 0.5 g g− 1. The maximum furans molar yield was predicted at the highest design temperature (200°C), the lowest C ratio (0.175 g g− 1) and the shortest reaction time (30 min), corresponding to a 24.6% yield. The experimental run under identical operating conditions allowed reaching a 25.9% molar yield, which implies an absolute prediction error of 5.28%, as observed in Table 11.
On the other hand, it can be seen that the influence of the catalyst mass is positive at temperatures below 165°C and becomes negative at higher temperatures. This effect can be seen more clearly in the contour lines presented in Fig. 12. At a temperature of 140°C (Fig. 9D), the furans yield increase with C factor. Then, exactly at 165°C, the response does not depend on the catalyst, but only on the reaction time (Fig. 9E). At higher temperatures, such as the represented at 200°C (Fig. 9F), the molar yield decreases with increasing catalyst mass. This occurs because the catalyst contains strong acid sites and promotes furans degradation reactions, such as rehydration to formic and levulinic acids, and polymerization to humins.
3.3.8. Multiple responses analysis of degradation products
Due to the complexity of the reactions involved in the hydrothermal conversion of a multiple sugars solution, a large number of secondary products were obtained. Since the main objective of this work is to find an optimal operating point to maximize the LA production, by studying the behavior of all reaction products it is possible to find potential improvements outside the established limits of operating conditions. This was carried out by using multi-response analysis of LA, intermediates and by-products.
Using the same iterative procedure for multiple linear regression fit (MLRF), molar yields fitting models were obtained for the rest of quantified by-products. The fit model equations for GH, GA, FA, LeA and AA molar yields as response factors are listed below:
![](data:image/png;base64,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)
Where YGH: glyceraldehyde, YGA: glycolic, YFA: formic, YLeA: levulinic and YAA: acetic acids molar yields (%); A: temperature, B: reaction time and C: catalyst/reactants mass ratio in coded dimensionless units.
Figure 10 shows the evolution of LA and all the by-products molar yields as a function of temperature, predicted with the fit model equations at constant B = 150 min and C = 0.5 g g− 1. The variables were decoded and shown in original units. The constant B value was selected because it allows comparing together the behavior of reactants, intermediates and products along the temperature. It can be observed that the maximum yield of LA occurs at around 190°C. At this temperature, GH is almost exhausted decreasing to 0.5% molar yield and GA, which reaches the highest yield at 166°C, is partially degraded. On the other hand, it can be seen that furans molar yield reaches a maximum at 174°C and strongly decreases upon higher temperatures due to degradation reactions. In this way, the degradation products molar yields (FA, AA and LeA) proportionally increase until reach their maximum levels at 200°C.
Direct dehydration towards furanic compounds and subsequent degradation reactions to FA and LeA are promoted by high temperatures. In addition, both reactions are strongly promoted by the presence of strong acid sites. In the same way, AA can be produced by several routes and it comes mainly from the oxidation and decarboxylation of aldehydes, such as GH or pyruvaldehyde. On the other hand, GH is a product of retrol-aldol condensation of both pentoses and hexoses and an intermediate in LA obtaining. The 2-carbon glycolaldehyde is inevitably produced during retro-aldol reaction of pentoses, thus the oxidation product derived (GA) is obtained as a side reaction.
In order to integrate the minor by-products to the statistical study and have a comprehensive analysis of the complete reaction system modeled, all the response factors of products molar yields were classified according to the reaction pathway. On the one hand, LA, GA and GH were grouped as retro-aldol products (YRA). On the other hand, FA, AA and LeA were grouped as degradation products (YDG). Dehydration products, i.e. the sum of furfural and HMF molar yields were maintained as total furans response (YF). Finally, the grouped fit models were superposed on a multiple contour lines graph, shown in Fig. 11. The black contour lines represent the retro-aldol products (YRA), the white contour lines are the total furans (YF), and the colored contour surfaces represent the degradation products (YDG) molar yields.
It is highlight that the calculated optimum operating condition for maximum LA molar yield (189°C and 208 min) is in the zone of maximum degradation products. It can be seen that furans achieve a maximum yield rapidly at high temperatures and with increasing reaction time are transformed to large amounts of degradation products. The furans are formed and rapidly consumed by LeA, FA, AA or humins production, displacing the reaction equilibrium towards side reactions. Because of that, higher molar yields of LA cannot be obtained at temperatures above 190°C. If it is desired to expand the operating conditions limits of the experimental design, it can be noted that working at temperatures between 160 and 170°C and longer reaction times may allow enhanced LA yields with less con-generation of secondary degradation products.