Qingjun Zhang,Youguang Ma,2,Xigang Yuan,2,Aiwu Zeng,2,*

1 State Key Laboratory of Chemical Engineering,School of Chemical Engineering and Technology,Tianjin University,Tianjin 300350,China

2 Chemical Engineering Research Center,Collaborative Innovative Center of Chemical Science and Engineering (Tianjin),Tianjin 300350,China

Keywords:Carboxylation Solvent-free medium Molten salt Optimization Response surface methodology

ABSTRACT A feasible synthesis route is developed for achieving the direct carboxylation of thiophene and CO2 in a relatively mild solvent-free carboxylate-assisted carbonate (semi) molten state.The effects of reaction factors on the carboxylate yield are investigated in the preliminary screening experiments,and the phase behavior analysis of the reaction medium is detected through the thermal characterization analysis of insitu high temperature X-ray diffraction measurement (in-situ XRD).The application of response surface methodology (RSM) based on the Box-Behnken design (BBD) is conducted to investigate the effect of the reaction parameters,such as reaction temperature,carbonate proportion,CO2 pressure and thiophene amount,on the product yield.The regressed second-order polynomial model equation well correlates all the independent variables.The analysis of variance (ANOVA) results reveal that the quadratic effect of reaction temperature is the most effective parameter in this carboxylation reaction owing to it’s the highest contribution to the sum of square (30.18%).The optimum reaction conditions for maximum product yield are the reaction temperature of 287°C,carbonate proportion of 32.20%,CO2 pressure of 1.0MPa and thiophene amount of 9.35 mmol.Operating under these selected experimental conditions,a high product yield (50.98%) can be achieved.

1.Introduction

There is a large amount of coking thiophene in the traditional refining process in China,while it has not been reasonably used,since the vast majority of thiophene in the production process is simply removed or destroyed as the sulfur impurities through the pickling or hydrogenation means,resulting in the inefficient use of this resource and the significantly environmental pollution.Besides,thiophene is similar in nature to benzene and can be substituted for benzene in many applications.Through comprehensive analysis of the added value of thiophene downstream products,it can be found that the thiophene carboxylic acids,especially 2-thiophenecarboxylic acid and 2,5-thiophenedicarboxylic acid,are of great research significance in the application value and substitution value.

Taking the generation of 2-thiophenecarboxylic acid as an elucidating example,there are three mainly conventional synthesis paths including the formylation [1],acetylation [2] and Grignard reagent [3] methods.These traditional synthesis pathways have some disadvantages,such as multiple reaction steps leading to the lower atomic efficiency,the application of toxic reagents and a large amount of waste discharges resulting in the environment pollution.As a result,a new feasible approach,direct carboxylation of thiophene with CO2,should be necessarily designed from the dual perspectives of green synthesis and atomic efficiency,taking the comprehensive consideration of the progress of the carbon fixation in recent years and the successful application of some aromatic arenes,such as toluene and phenol,carboxylation reactions.

Of particular importance is that,given that CO2is a rich,nontoxic and renewable C1 source,there is also an economic and environmental way that directly converts CO2into the high-value carboxylic acids [4,5].Although there are several restrictions that need to be overcome properly for this C-H carboxylation route,while it has attracted some attentions in the aromatic arenes [6].Note that it is difficult to achieve the direct C-H carboxylation under mild conditions owing to the kinetic and thermodynamic stability of CO2and relatively inactive aromatic compounds [6,7].Therefore,some specific limiting reagents are used to realize this process,such as Lewis acid [8-15],carbene metal complexes[16,17] or alkali-metaltert-butoxides [18-22].However,the majority of these metal additions are overpriced,non-renewable,or unstable in the air.As a result,a more efficient and environmentally friendly synthesis approach using the benign and regenerable reagents should be devised.

The C-H carboxylation is also feasibly realized through the cleavage of C-H bond(s) of aromatic arenes in the solvent or solvent-free environment leading to the formation of the reactive carbon-centered nucleophileviathe acid-base chemistry strategy.One of the most representative is the strong base carbonate induced the C-H carboxylation of aromatic compounds [23-26].However,this carboxylation strategy is ineffective once the pKavalue of aromatic substrate exceeds 27 [27].Carboxylation occurs for aromatic substrate with the very weak acidic C-H bond(s) or high bond dissociation energy (BDE) of aromatic C-H bond(s),such as furoate and benzoate,by removing proton(s) to form the carbanion that attacks the electrophile CO2in the molten salt-CO2interface [24,28].Apart from these solid reaction mixtures,the less acidic liquid-phase benzene [27] could also be converted into carboxylic acids in the presence of cesium carbonate with the poor selectivity,but the additional co-salt cesium isobutyrate is indispensable to generate the necessary molten phase [24].However,there has extremely high reaction temperature (340-380 °C) and pressure (7 MPa) in this system and the combined yield of carboxylic acids is relatively poor (about 90 μmol·g-1catalyst(0.84%on the basis of benzene))owing to the chemical inertness and low solubility of benzene in the salt medium.

Another problem that needs to be faced directly is how to adjust the operating parameters to optimize the performance of the process after examining the process feasibility.There are two classifies.One is the classical optimization,which is to change only one factor at a time to measure its effect,which requires many experiments and is time-consuming.Note that the combined effects and interactions between variables are not considered in this method.However,another method,the response surface methodology(RSM),overcomes the flaws of the classical optimization since the corresponding combined effects of process variables are considered in the optimization [29].RSM is one of the experimental design techniques that is used for statistical modeling as well as process optimization,which is an effective method for investigating the effect of parameters and their interactions on the response of interest [30].Many studies have shown that the Box-Behnken design(BBD),based on the RSM,is a superior experimental design that may be utilized for any process optimization[31-33].RSM has,of course,been used to optimize the operating variables of a variety of processes.However,there is no evidence of the use of RSM in the direct carboxylation of aromatic compounds with CO2in the open literatures.In the present study,we report for the first time an optimization approach for realizing the direct carboxylation reaction of thiophene and CO2.

In this paper,we develop a feasible route to achieve the direct carboxylation of thiophene and CO2in a relatively mild solventfree carboxylate-assisted carbonate (semi) molten salt.The effects of various reaction factors are explored in the preliminary screening experiments,and the corresponding results serve as the reference for the RSM optimization.The combination of carbonate and carboxylate co-salt exhibits the synergetic effect in the carboxylation reaction,and the cesium pivalate-assisted Cs2CO3system affords the better product yield among these co-salts.The phase behavior analysis of the reaction medium is detected through the thermal characterization technique.Furthermore,to further determine the optimal reaction conditions,the optimization experiments for carbonate-promoted direct thiophene carboxylation with CO2are performed by taking the application of Box-Behnken design as the optimization tool for RSM to investigate the effects of the process parameters on the product yields taking the reaction temperature,carbonate proportion,CO2pressure and substrate amount as the four independent variables.

2.Materials and Methods

2.1.Reagents

Thiophene (99%),cesium or potassium carbonate (99.9%),cesium or potassium pivalate (＞97%),deuterium oxide (99%),potassium isobutyrate (＞98%),and cesium or potassium acetate(99.9%) were obtained from Aladdin;carbon dioxide (99.9%) was purchased from Tianjin Liufang Industrial Gas Distribution Co.,ltd;cesium oxalate (98%) was purchased from Jinjinle Chemical Co.,Ltd;cesium formate (98%) was supplied by Energy Chemical;anhydrous sodium tartrate (99.0%) was purchased from Jiaxing Sicheng Chemical Co.,Ltd;3-(trimethylsilyl)-1-propane sulfonic acid sodium salt (DSS,97%),potassium pivalate (＞98%) were purchased from Macklin.Thiophene was dried over 4A sieves before using,and other purchased reagents were used without further treatment.

2.2.Experimental procedure

The specifically calculated mixed salts and thiophene were charged successively into 50 ml reactor (316 L) equipped with quartz tube.The sealed reactor was backfilled through CO2three times,and reached the specific pressure through filling CO2gas reagent.This mixture was heated to the certain reaction temperature through using intelligent heating mantle and maintained the setting temperature for certain hours.Subsequently,this reactor was depressurized carefully after being cooled to ambient temperature.The unreacted thiophene was removed from the autoclave,and the remaining product mixtures were dissolved with 1 ml D2O,and then filtered through 0.22 μm MCE syringe filter to remove the insoluble materials.The corresponding carboxylate yield was calculated by integrating the1H NMR peaks taking the anhydrous sodium tartrate as internal standard.The reacting carboxylate product was also verified through13C NMR with the added DSS as internal standard.

2.3.Characterization techniques

Different cesium carboxylate products were identified and the corresponding product yields were analyzed through the1H NMR and13C NMR.1H NMR and13C NMR signals were recorded on the VARIAN INOVA 500 MHz and AVANCE III 400 MHz spectrometers,respectively.1H NMR chemical shifts were assigned relative to the residual proton signal (D2O,δ=4.79) and13C NMR chemical shifts were referenced with added DSS (δ=0).The relaxing delay and number of scans were assigned as 8 s and 64 in the1H NMR spectra.

The phase transition of cesium salt mixtures could also be collected under vacuum throughin-situhigh temperature X-ray diffraction measurement (Bruker D8 advance) equipped with the reaction chamber using the Cu radiation(Kα=0.154 nm).The sample was heated from 100 °C to 280 °C with the ramp rate of 10 °C·min-1,and each diffraction data was collected in the 20 °C steps.Each XRD pattern was scanned over the 2θ range from 5°to 50° with the step size of 0.02° and the scan speed of 0.083 (°)·s-1.

2.4.Experimental designs

2.4.1.Single-factor-experiment for selection of variables

We devised a feasible synthesis route (Fig.1) of achieving the direct carboxylation of thiophene and CO2in a relatively mild solvent-free carboxylate-assisted cesium carbonate (semi) molten state.In this carboxylation reaction,there were many reaction parameters that can affect the result of the reaction,such as reaction temperature(T),substrate amount(nthio),CO2pressure(PCO2),amount of the carbonate and co-salt(nsalts),reaction time,and the combinational effects of the carbonate and carboxylate co-salts.In order to determine the experimental factors and their levels required in the following response surface optimization calculations,we conducted the corresponding single-factor experiments,shown in Section 3.2,to determine what experimental factors to choose and their corresponding optimal optimization intervals.As observed in these screening experiments that the reaction effect of the cesium carbonate system assisted by the co-salt cesium pivalate (CsOPiv) was the best.The effects of the reaction time and mixed salts amount on the reaction outcome were also insignificant among these reaction conditions,and these two variables were maintained at constant values in the RSM experimental tests.There were four factors being considered as independent variables,such as reaction temperature (X1),carbonate proportion(X2),CO2pressure (X3) and the amount of thiophene (X4),and the range of these parameters was also determined by these screening experiments.

2.4.2.RSM analysis based on Box-Behnken design for optimization process parameters

Based on the single-factor experiment results,it was found that the factors affecting the product yield of the direct carboxylation of thiophene and CO2included reaction temperature,combination of carbonate and carboxylate,carbonate proportion in mixed salt,CO2pressure,thiophene amount,mixed salt amount and reaction time.As demonstrated from the single-factor experimental results that the reaction effect of the cesium carbonate-cesium pivalate (CsOPiv)system was the best,therefore,the response surface optimization method was used to determine the optimal condition in direct carboxylation reaction of thiophene and CO2mediated by the cesium carbonate-cesium pivalate system.In addition,in the process of optimizing the experimental design,the reaction time was taken as a fixed variable,because with the extension of the reaction time,the content of decomposed substances in the system increased(taking the 2 mmol thiophene reaction as an example,it was found that when the reaction time was increased from 2 h to 3 h,the amount of decomposed substances in the system increased by 3 times).Besides,there was little difference in the product yield between the reaction time of 3 h and 2 h (taking the reaction of 2 mmol thiophene as an example,when the reaction time was increased from 2 h to 3 h,the product yield increased from 8.35%to 9.53%).Therefore,considering the experimental time and reaction effect comprehensively,we set the reaction time as 2 h in the response surface optimization test.

The optimization experiments for carbonate-promoted direct thiophene carboxylation with CO2were performed by taking the application of Box-Behnken design (BBD) as the optimization tool for RSM to investigate the effects of the process parameters on the product yields.Based on the results of the screening tests,a three-level-four-factor Box-Behnken design was employed.Enumerated in Table 1 illustrated that all of the variables were fixed at three levels (-1,0,and +1) based on the single-factorexperiment studies,with reaction temperatureX1(260,280,and 300 °C),carbonate proportionX2(20%,40%,and 60% (mol)),CO2pressureX3(0.6,0.8,and 1.0 MPa),and thiophene amountX4(6,8,and 10 mmol).

Table 1 Range and levels of parameters in BBD experimental design

There were a total of 28 runs,available from the Design Expert,with 24 factorial points and four center points(Eq.(1)),and Table 2 showed the experimental BBD adopted in this present study.Following the experiments,the correlation between the predicted response and the process parameters was expressed using a second-order polynomial regression model equation (Eq.(2)),where Y denoted the response variable,XiandXjdenoted independent variables,andkdenoted the number of independent variables(k=4).The regression coefficients of β0,βi,βii,and βijwere,respectively,for the intercept,linear,quadratic and interaction terms.Then the statistical analysis of variance (ANOVA) was performed to determine the fitness,suitability,and significance of this regression model.And the goodness of this second-order model equation to the response was assessed using the regression coefficient (R2).

Table 2 Summary reaction results for thiophene carboxylation as the function of temperature and salts

3.Results and Discussion

3.1.Characterization

To determine whether the reaction was going on at the molten salt interface,the phase behavior of the Cs2CO3-CsOPiv medium was measured through thein-situhigh temperature X-ray diffraction (in-situXRD).The intensities of crystalline diffraction peaks were collected at the temperature of 100-280 °C (Fig.2).The diffraction intensity of crystalline CsOPiv was weakened distinctly relative to the Cs2CO3peaks at 120-160 °C,and then was not detectable at 180°C in the XRD pattern,although diffraction peaks of crystalline Cs2CO3remained in the measured temperature region.The disappearance of the CsOPiv diffraction peaks was assigned to the eutectic melting behavior and the characterization results revealed that this reaction was also proceeded in the presence of the molten eutectic phase and solid Cs2CO3phase.

Fig.1.Possible Cs2CO3-mediated direct carboxylation process.

Fig.2. In-situ XRD pattern of CsOPiv-Cs2CO3 system.

3.2.Preliminary experiments for selection of variables

Direct thiophene carboxylation reaction was performed in the solvent-free mixed (semi) molten salt containing the carbonate M2CO3(M=Cs,K and Na)and the correspondingly co-salt carboxylate (Cs or K).It is now well established that carboxylate salts,especially acetate and pivalate,can be used to promote the stoichiometric and catalytic C-H activation effects in the transition metal catalysis.Initially taking the cesium acetate as an illustrating example (Table 2),the maximum carboxylate yield (4.98%) was reached at the temperature of 300°C(Entry 4).The decreased carboxylate yield was observed when the temperature exceeded 300 °C (Entries 1-5).Note that there was a slight difference in the reaction effect when simply changing the assisted carboxylate of cesium acetate to potassium acetate owing to the difference in alkalinity between these cesium and potassium salts (Entry 4vsEntry 10).Among these alternatively assisted carboxylate bases,the improvement effect for carboxylation reaction changed with the variation of the assisted carboxylate salts.The less basicity of their corresponding conjugate bases resulted in the relatively weaker reaction ability in comparison with the cesium acetateassisted Cs2CO3system when taking the cesium formate (Entry 6) and oxalate (Entry 7) as the assisted carboxylate salts.There was no doubt that the cesium pivalate system was more basic than the cesium acetate system,so its reaction effect was better(Entries 8-9),namely,when the reaction temperature was 280°C or 300°C,the product yield of the cesium pivalate-cesium carbonate system was 2.7 times or 1.4 times that of the cesium acetate-cesium carbonate system.In addition,the same trend was found for the pivalate,that the potassium salt was less reactive than the cesium salt(Entry 8vsEntry 13).Besides,when the auxiliary base remained unchanged,the reaction effect of potassium carbonate (Entries 14-16) and sodium carbonate (Entries 17-18) was,respectively,also less than that of cesium carbonate since the former was less basic than the latter.

Enumerated in Table 2 showed that the cesium pivalate was the most effective among these alternative carboxylates (Entry 8).To further understand the thiophene C-H carboxylation progress,the effects of the reaction factors,such as CO2pressure,substrate amount,and carbonate proportion,etc.,on the reaction results were assessed.

For the direct cesium carbonate-assisted thiophene carboxylation process,the effect of the reaction temperature was crucial.The carboxylation reaction was carried out in molten salt,shown in Fig.2,and the temperature had a great influence on whether the cesium carbonate-carboxylate mixed salt system can form a uniform molten environment.The reaction effects were assessed at the temperature range of 240-320 °C for 2 h after filling 0.8 MPa CO2at the ambient temperature(Fig.3(a)).The maximum carboxylate yield (8.33%) was reached at the temperature of 280°C,in which the 255.87 μmol·g-1carboxylates were generated as the selectivity of thiophene-2-carboxylate(51.39%)was also larger than that(22.54%)in the acetate system.The total carboxylate yield was then decreased as the reaction temperature continued to elevate.

Fig.3.The effects of reaction factors on the yield and selectivity in the Cs2CO3/CsOPiv system.The effects of various factors on product yields and selectivity in the Cs2CO3/CsOPiv system.(a) reaction condition: 2 mmol substrate,40% carbonate,0.8MPa CO2 initial pressure,3 mmol total cesium salts and 2 h;(b) reaction condition: 2 mmol substrate,280°C,0.8 MPa CO2 initial pressure,3 mmol cesium salts and 2 h;(c)reaction condition:2 mmol substrate,280°C,40%carbonate,3 mmol cesium salts and 2 h;(d)reaction condition: 280 °C,0.8 MPa CO2 initial pressure,40% carbonate,3 mmol cesium salts and 2 h;(e) reaction condition: 2 mmol substrate,280 °C,0.8 MPa CO2 initial pressure,40% carbonate and 2 h;(f) reaction condition: 2 mmol substrate,280 °C,0.8 MPa CO2 initial pressure,40% carbonate,3 mmol cesium salts.

The proportion of cesium carbonate in the mixed salt was a very important factor for this carboxylation reaction.As is well-known that the most critical step in the direct carboxylation of thiophene and CO2was the base cesium carbonate induced substrate deprotonation step to complete the acid-basic chemistry reaction.Therefore,the amount of cesium carbonate determined the product yield and selectivity;besides,the reaction was carried out in the molten phase(shown in Fig.1),and when the mixed salt system contained more carbonate,due to the high melting point characteristic of carbonate (793 °C),less molten phase was generated in the system,thus affecting the reaction effect.That is to say,in the mixed salt system,only the ‘‘synergistic” effect of cesium carbonate and cosalt can help the carboxylation reaction process.Fig.3(b) illustrated the results of the effect of carbonate proportion on the product yield and selectivity.When the proportion of cesium carbonate was 0,namely,there was only cesium pivalate in the system (melting point was 344-348 °C),the cesium pivalate was not melted in this reaction temperature(280°C),and product yield was only 0.05%.The ratio of monocarboxylic acid to dicarboxylic acid in the product was 6:1.Compared with the pure cesium acetate process (0.15%),as found that the reaction effect of the pure cesium pivalate system was weaker(0.05%),not because the basicity of cesium acetate was stronger than that of cesium pivalate,but because cesium pivalate did not form a molten phase.Therefore,this fully illustrated the characteristics of the reaction process,namely,the basic state and the molten state existed at the same time,and neither was indispensable.When the proportion of cesium carbonate increased to 40% (mol),the product yield reached the maximum (8.33%).At this time,the amount of monocarboxylic acid and dicarboxylic acid in the product was approximately equal.When the cesium carbonate content continued to increase,the product yield gradually decreased.This was because when the high melting point cesium carbonate in the mixed salt gradually increased,the molten phase generated in the system decreased,which made the reaction environment smaller,hence,the relative contact probability between the substrate and the eutectic molten salt became smaller,so the reaction effect became worse.This was consistent with the rule that the reaction was carried out in the mixed salt molten phase.Therefore,in this system,the optimal proportion of cesium carbonate was 40% (mol).

During the carboxylation reaction,the effect of CO2on the system had two sides.On the one hand,the thermal decomposition of carboxylic acid products and mixed salts can be inhibited by charging an appropriate amount of CO2into the reactor.On the other hand,excess CO2will react with cesium carbonate and byproduct water in the mixed salt system to form cesium bicarbonate,which made a part of the carbonate in the system be sequestered in the form of bicarbonate,which reduced the basic strength of the system,resulting in the deterioration of the reaction effect.Therefore,in order to ensure the effective progress of the reaction,it was very necessary to determine the amount of CO2charged in the system.The effect of CO2pressure on the carboxylate yield was explored in the 0.2-1.0 MPa pressure range(Fig.3(c)).The carboxylate yield reached the 81.26 μmol·g-1of Cs2CO3at the combined yield of 2.65%with the total reaction pressure of approximately 0.68 MPa when introducing 0.2 MPa CO2at ambient temperature.Optimal carboxylate yield was obtained when the CO2pressure was 0.8 MPa,and the further yield enhancement was not observed with the elevated pressure because the carbonate could be sequestered in the form of bicarbonate due to the large amount of CO2[21,34].

When the thiophene amount was increased from 2 mmol to 8 mmol,product yield of carboxylate gradually increased from 8.31%to 42.38%,shown in Fig.3(d),and the maximum yield of carboxylate was 1301.92 μmol·g-1.This trend suggested that the strongly basic nature of alkali metal salts may not be fully released at relatively low substrate amounts.However,when the substrate thiophene amount continued to increase to 10 mmol,product yield dropped to 39.14%,which may be due to the excessive thiophene atmosphere on the one hand reducing the probability of the carbanion contacting CO2,and on the other hand,the presence of the by-product water made the part of the cesium carbonate sequestered,which reduced the reaction effect,while this was still better than the product yield (21.79%) in the cesium carbonatecesium acetate system.This reaction result can directly present that the deprotonation ability of cesium carbonate-cesium pivalate system was stronger than that of cesium carbonate-cesium acetate system.In addition,with the increase of the content of the substrate thiophene,the products generated by the reaction gradually tended to be dicarboxylic acids,and the proportion increased from 50% to about 77%.

In this synthesis route,the amount of mixed salt also had a great influence on the carboxylation reaction process.One of the conditions for the reaction in this path was that the mixed salt formed a molten phase,and the amount of mixed salt added to the system determined the amount of molten phase formed in the system.If the amount of molten phase generated was relatively more,the probability of thiophene and CO2contacting with the molten phase was greater,which will further contribute to the formation of products.In addition,taking the consideration of the equipment size in exploring the influence of this factor on the reaction effect,the maximum amount of mixed salt was limited to 3 mmol.In addition,when exploring the influence of this factor on the reaction effect,it can also be judged according to the results whether the specific role of the mixed salt in the reaction process was to act as a catalyst or participate in the reaction as the stoichiometric reagent.Therefore,it was necessary to explore the influence of mixed salt content on reactivity in this reaction process.And the corresponding results were shown in Fig.3(e).With the increase of the mixed salt content,the product yield first increased and then decreased,and the optimum mixed salt content in this process was 2.5 mmol.The product yield was 9.47%,and the ratio of monocarboxylic acid to dicarboxylic acid was 1:0.94.Except for the reaction results with a mixed salt content of 2.5 mmol,on the whole,the product yield also increased approximately linearly with the increase of mixed salt content.This may also indicate that the carboxylation reaction was stoichiometric in nature.Of particular importance was that,in the whole process,the effect of mixed salt amount on product selectivity was relatively small(taking monocarboxylic acid as an example,the selectivity was relatively stable at about 50%).

During the carboxylation reaction,the effect of reaction time on the system also had two sides.On the one hand,appropriately increasing the contact time between the reactants and the molten salt will increase the formation of reaction products;On the other hand,with the extension of the reaction time,more water will be produced in the reaction system,which not only promoted the dissociation of the reactants,thereby slowing down the reaction rate,and even the generated water will dissolve a part of the cesium carbonate Cs2CO3,and the yield of the target product was reduced.Therefore,in order to ensure the effective progress of the reaction,it was very necessary to determine the appropriate reaction time in the system.Fig.3(f)showed the effect of reaction time on the product yield and selectivity.When the reaction time was prolonged from 1 h to 3 h,the product yield increased from 5.36% to 9.65%;when the reaction time was prolonged,it gradually decreased from 9.65% to 6.16%.Note that in the whole process,the effect of reaction time on product selectivity was relatively small (take monocarboxylic acid as an example,its selectivity was relatively stable at about 50%).Therefore,based on the above results,the optimal reaction time for this reaction process was 3 h.

3.3.Statistical analysis

There were twenty-eight batch experiments,enumerated in the Box-Behnken matrix,to explore the individual and combined effects of reaction parameters on product yield under the various combinations of process variables.Depicted in Table 3 showed the corresponding experimental design results for the product yield,monocarboxylate and dicarboxylate productivities obtained at a reaction time of 2 h and a mixed salt amount of 3 mmol.

Based on the product yield (Y) results,a second-order polynomial regression model correlation between the independent process variables (based on the actual level of factors) and the response was obtained using the Design-Expert software,as shown in Eq.(3).

where the factors ofX1,X2,X3andX4represented the independent variables of reaction temperature,carbonate proportion,CO2initial pressure and thiophene amount,respectively;and all the terms of=1-4,i＜j）were for the interaction effects for the allindependent variables;and the variables ofwere,respectively,for the main effects of the parameter reaction temperature,carbonate proportion,CO2pressure and thiophene amount.

This responseYincreased as the number of factors with positive coefficients increases,and decreased as the number of factors with negative coefficients increases.The plus or minus sign before the interaction item denoted the synergistic and antagonistic effects,in both [35,36].

To verify the fitness and significance of this response surface model,an ANOVA based on the BBD was performed using Design Expert software.The corresponding ANOVA result for the response of product yield was shown in Table 4.Note that a large F-value with a smallP-value ((i.e.,P＜0.05)) indicated that the model was statistically significant when analyzing ANOVA results.Enumerated in Table 4 illustrated that this quadratic model was significant,with a confidence level of greater than 95%and anF-value of 20.73,since the corresponding calculatedP-value was less than 0.0001.Besides,the Lack of Fit test,along with three crucial model correlation coefficients,can be used to determine whether this chosen model adequately described the observed data or that amore complex model should be used.TheP-value for lack-of-fit(0.7943 ＞0.05) in this study indicated that it was not statistically significant when compared to the pure error.As a result,this model can take an adequate prediction that corresponded to the response values.Of course,there were also three very important indicators to demonstrate whether this model fitted well in the whole regression regions being studied,namely,the correlation coefficient (R2),adjusted correlation coefficient (R2(adj.)) and predicted correlation coefficient(R2(pred.)).The larger the correlation coefficientR2,the better fit of the observed and the predicted values.If the correlation coefficientR2(adj.) andR2(pred.) were bothlarge and near (＜0.2),the regression model can adequately explain the process [29].Note that these three indicators were,respectively,for 95.71%,91.09% and 81.35%,and the corresponding difference for the adjusted and predicted correlation coefficients was 0.1680,lowering than 0.2,which indicated that the model was significant enough to demonstrate the compatibility of the model predicted values with the experimental values.

Table 3 Experimental results of Box-Behnken designs for product yield

Table 4 Analysis of variance for the second-order model

As also indicated in Table 4 that the total contribution of the linear effect on product yield was 46.49%,among which the CO2pressure (X3) had the most important influence on product yield,accounting for 15.81%.The effect of carbonate proportion (X2)and thiophene amount (X4) on the yield was approximately the same,12.18% and 11.06%,respectively.The reaction temperature factor (X1) had the least influence on product yield (7.44%) among these four independent factors.Moreover,there had 47.45%contribution to the product yield in the quadratic effects.The quadratic effect of reaction temperature () had the most significant effect with the contribution of 30.18%.Note that this term was the most important factor affecting the product yield among all the factors.The second important factor for impacting the product yield in quadratic effect was the carbonate proportion () term of 11.90%,which was very close to the percentage in the linear effect.The other two quadratic terms were not significant.Similarly,in the interaction effects,there had the 2.29%contribution to the product yield.The interaction（X1X3）between the reaction temperature and CO2pressure had the most contribution (1.57%) and was the significant,while the others were found to be insignificant.Therefore,the combined effects among these process parameters in the current study were highly significant,demonstrating the successful application of RSM in identifying these effects in depth.Using Pareto graphic of effects(Fig.4),we can also visually identify the important effects and compare the relative magnitudes of various effects.Also,the largest effect can be found to be the quadratic term of reaction temperature since it extended the furthest.And all variables showing significant levels were located to the right of the red dotted line.Note that the degree of influence of different significant factors on the yield of the product was in order:the quadratic effect of reaction temperature(30.18%),CO2pressure(15.81%),carbonate proportion (12.18%),the quadratic effect of carbonate proportion (11.90%),thiophene content (11.06%),reaction temperature (7.44%),the quadratic effect of thiophene amount(4.11%),and the interaction effect of reaction temperature and CO2pressure (1.57%).

The refined version of the response surface model was obtained as shown in Eq.(4).The corresponding correlation coefficients were enumerated in Table 4,which indicated that the accuracy of the predicted value had improved.

Table 5 Selection of optimal condition for maximum product yield

Fig.4.(a) The standardized Pareto chart for product yield;(b) The normal plot of standardized effect.(1bar=0.1MPa)

In addition to the above factors to explain the accuracy of this statistical model,on the other hand,the reliability of the model can also be verified by studying the characteristics of the residuals.The residuals of a well fitted model often had three characteristics[37]of normality,homogeneity of variance and independence,and their corresponding analysis plots were shown in Fig.5.The plot of normal % probability vs internally studentized residuals was shown in Fig.5(a),which demonstrated that it fitted the normal distribution very well.And it also showed that no response transformation was required,and there was no obvious problem with normality.Fig.5(b)gived the plot of internally studentized residuals and predicted product yield.It was symmetrical from top to bottom and was distributed around the zero value.The properties of the distribution did not change as the predicted value rised,indicating that the homogeneity and independent of data variance requirements were met.This also meaned that the response variable did not need to be transformed.As also observed in Fig.5(c),the actual and predicted plot for product yield,that the predicted responses well satisfied the original observations.The last graphic,Fig.5(d),illustrated that the residuals were randomly distributed around the centerline,which can verify the hypothesis that the residuals were not correlated with each other.In general,the model was adequate for characterizing this direct carboxylation process.

3.4.Effect of process variables on the product yield in direct carboxylation reaction

As observed from the results of analysis of variance on this statistical model that the process variables and their combined effects had a great influence on the response of product yield.The threedimensional (3D) response surface and matched contour map of the interaction between the test factors can be generated using the regressed quadratic equation model based on RSM,and the influence of the interaction of the other two factors on the response when some factors remain their corresponding central values can be investigated.The corresponding 3D surface and contour plots for the combined effects of two factors on the product yield were shown in Figs.6 and 7.

Fig.5.Residual plots for product yield.

Fig.6.3D surface and contour plots for the combined effects of (a) temperature and carbonate proportion,(b) temperature and CO2 pressure,and (c) temperature and thiophene amount on the product yield.

Fig.6 presented the response surface diagrams and their equivalent contour plots between the reaction temperature and the other three independent variables.In Fig.6(a),the combination effect of temperature and carbonate proportion,product yield increased with the increase of temperature up to a certain value and then decreased with the further increase the temperature over the entire range of carbonate proportion(20%-60%),and this trend was also consistent with the variable carbonate ratio,which implied that that there had extreme point.Besides,as can be seen intuitively from the contour map in Fig.6(a)that the product yield was greater than or equal to 35%in the elliptical region where the carbonate ratio was less than 45% at the reaction temperature of 275-292 °C.It can be seen from the contour map in Fig.6(b) that the interaction between reaction temperature and carbon dioxide pressure was very significant under the condition that other factors remained unchanged.As obtained from the response surface plot,product yield had the maximum under the appropriate reaction temperature and CO2pressure.The maximum occurs in the region where the reaction temperature was 275 to 295 °C,and the CO2pressure was not less than 0.85 MPa.Fig.6(c) provided the combined effects of reaction temperature and thiophene amount on the product yield with the other two factors remaining their central points.There also had the extreme point in the region where the thiophene amount was greater than 7.5 mmol at the reaction temperature of 275-295 °C.Both the surface and contour plots in Fig.7(a)-(c) indicated that the higher product yield was obtained at a higher CO2pressure and higher thiophene amount,but the relatively lower carbonate proportion.

Fig.7.3D surface and contour plots for the combined effects of (a) carbonate proportion and CO2 pressure,(b) carbonate proportion and thiophene amount,and (c) CO2 pressure and thiophene amount on the product yield.

3.5.Optimization and confirmation test

Analyzing the above surface and contour plots showed that there was extreme point in the model and the range corresponding to the optimal process parameters.The optimal conditions were determined by maximizing the product yield using the Design Expert and Minitab.The optimal reaction conditions were enumerated in Table 5,namely,the reaction temperature was 287 °C,the proportion of cesium carbonate in the mixed salt was 32.20%,the CO2pressure was 1.0 MPa,and the substrate thiophene content was 9.35 mmol.And the predicted product yield response value at this time was 47.14%.

In order to test the reliability of the response surface method,it was generally necessary to verify the theoretical optimal conditions obtained by the RSM optimization and the corresponding results were also enumerated in Table 5.Two verification experiments were carried out under the reaction conditions given by the model.The experimental results showed that the average yield of the product during the carboxylation reaction was 48.65%.Compared with the theoretical value of the product yield given by the model,the relative error was 3.20%,which verified the validity of the model prediction results.At this time,the selectivity of dicarboxylic acid in the product was 70.26%,and the yields of monocarboxylate and dicarboxylate were 465.80 μmol·g-1and 1100.50 μmol·g-1,respectively.Therefore,the Box-Behnken response surface design method can be effectively used to optimize the reaction conditions of the direct carboxylation of thiophene and CO2.

4.Conclusions

Direct carboxylation of thiophene with CO2was carried out in the presence of mixed cesium carbonate and carboxylate salts.The carboxylate salt could create the synergistic effect on the carbonate-promoted carboxylation and the reaction effect varied with the carboxylates owing to their different deprotonation abilities.As concluded from the phase behavior analysis that the reaction was proceeding at the (semi) molten salt interface.The pivalate-assisted Cs2CO3-promoted C-H carboxylation reaction had the best reaction effect in the preliminary screening experiments.To further determine the optimized operation parameters in this carboxylation reaction,the optimization based on BBD using response surface methodology was then performed.A quadratic model equation was established with the high correlation coefficient,and the ANOVA results indicated that the main effect of reaction temperature had a significant effect on the response of product yield.The optimum reaction conditions were as follows:reaction temperature of 287°C,carbonate ratio of 32.20%,CO2initial pressure of 1.0 MPa,and the thiophene amount of 9.35 mmol.The experimental results showed that the average yield of the product during the carboxylation reaction was 48.65%,and the selectivity of dicarboxylic acid in the product was 70.26%,and the yields of monocarboxylate and dicarboxylate were 465.80 μmol·g-1and 1100.50 μmol·g-1,respectively.And the corresponding results of the model validation experiments agreed with the predicted value.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.