Dehydration by Pervaporation of an Organic Solution for the Direct Synthesis of Diethyl Carbonate

Abstract

Pervaporation has been a central subject in the research community within the scope of the further development of energy- and cost-efficient alternatives to conventional liquid–liquid separation technologies. The potential eligibility of four commercial membranes (ZEBREX ZX0, PERVAP^TM 4155-80, PERVAP^TM 4100, PERVAP^TM 4101) for use in an integrated dehydration application of a diethyl carbonate/water/ethanol mixture by pervaporation was assessed experimentally. The impact of feed concentration, operating temperature, pressure, and sweep gas flow rate on membrane separation performance, including permeation flux, permeate quality, selectivity, and permeance, was thoroughly investigated. Applying the ZX0 membrane delivered the best qualities of all tested membranes of the permeate stream, with a water concentration of mostly >98%. In comparing the water flux, the ZX0 membrane remained reasonably competitive with the polymer membranes. Furthermore, the sweep gas volume flow rate and the operating temperature were identified as influencing the flux significantly but not the product composition. At the same time, the feed concentration of water also influenced the water purity within the permeate. The experiments were monitored with a partial least squares model, allowing a quick assessment of obtained samples while delivering accurate results.

1. Introduction

After several decades of comparatively high raw material security in industrialized countries, issues such as the procurement of raw materials are once again becoming the central focus of researchers and businesses. Significantly changed regulations in industry, natural disasters, and the shift in political interests result in a stronger emphasis on realignment towards a more sustainable raw material supply chain [1]. In the course of this, industrial residues are also being re-evaluated for further use, as are conventional technologies. In particular, separating liquid–liquid mixtures, often carried out by distillation, is considered energy-intensive. This can frequently be replaced entirely by pervaporation (PV) membrane technology. Pervaporation represents a separation technology in which a liquid mixture (feed) is brought into contact with one side of a membrane. The fraction that passes through the membrane (permeate) is withdrawn as a low-pressure vapor (vacuum) or with a gaseous compound (sweep gas), diluting the product on the permeate side. Major influencing parameters of the transport of permeating compounds are the chemical potential difference of the compounds on both sides of the membrane, as well as interactions between the permeate’s and the membrane’s selective layer, along with its mobility within the membrane matrix [2]. The comparatively low demands and increasing possibilities for a compact design of pervaporation open up new opportunities for recovering valuable liquid materials and new options for integrating the separation of product mixtures [3].

Further, pervaporation is more suitable for the separation of compounds that would be thermally decomposed if distilled [4]. This makes it potentially eligible to recover many organic compounds, such as organic carbonates, from various applications.

Organic carbonates are used in the chemical industry to produce resins, lithium-ion batteries, plastics, pharmaceuticals, and other materials. There is evidence of their use as a solvent and as a fuel additive. Their properties as hardly toxic aprotic substances with higher boiling points make them versatile substances for use in the chemical industry [5,6,7]. Among multiple current opportunities under which organic carbonates are synthesized, the most common ones rely on rather problematic conditions (toxicity, high energy demand). Recent years have brought about novel approaches to producing organic carbonates. Nevertheless, these synthesis routes still bear certain drawbacks (limited catalytic activity, adverse thermodynamics) before they become commercially attractive. Diethyl carbonate (C₅H₁₀O₃ or also DEC) is increasingly becoming of great interest among known organic carbonates since direct synthesis from sustainable raw materials (CO₂, bio-EtOH) is possible.

(1)$2C_2{H}_{5}OH+C{O}_2\rightleftharpoons C_5{H}_{10}{O}_3+H_{2}O$

The reaction is carried out at elevated temperatures of 130–160 °C and pressures between 20–60 bar. The presence of water especially seems to drastically inhibit the reaction [8]. Membranes could potentially be considered a highly matching solution to remove water from the system and reduce the effect of water presence. Therefore, the dehydration of the reaction mixture is a promising and preferable approach. One of this work’s primary interests is the assessment of the eligibility of membrane technology to dehydrate a ternary mixture consisting of DEC/water/EtOH. The suitability of FTIR-ATR measurements assessed with partial least squares regression (PLS) for fast process monitoring is also discussed.

1.1. Dehydration with Membranes—State of the Art

DEC forms an azeotrope with water at a molar ratio of 1:3 (DEC/water) [9]. Therefore, distillation is not the most efficient method to recover the desired purified product. Hinchliffe and Porter reported the potential eligibility of inorganic membranes to separate non-desired products of a reaction with a membrane back in 2000, while being competitive with distillation [10]. A novel zwitterionic polyamide membrane was developed by Zhang et al. [11] to dehydrate organic solvents mainly consisting of ethanol, with a relatively high flux of 4380 g/m²h and selectivities of 3870 (${\alpha }_{H_{2}O/EtOH}$) when fed with a water/ethanol mixture containing 90 wt% ethanol, yielding better qualities than 99.7% water purity. Increasing the temperature and feed water content likewise results in rising fluxes. No information, however, has been provided regarding the stability of elevated pressures and temperatures exceeding 76 °C [11]. Fujiki et al. [12] suspect limited levels of stability for organic membranes and, therefore, applied a microporous TiO₂–SiO₂–organic-chelating ligand composite membrane to dehydrate an IPA/water mixture containing 90 wt% IPA. Fluxes surpassing 1500 g/m²h have been recorded for temperatures of 60 °C. As expected, a rising trend was observed for the flux with increasing temperature, while selectivities also rose alongside the rising flux until 70 °C [12].

As indicated in Section 1.2, the dehydration of the solution is preferred over the selective separation of ethanol or DEC for process enhancement reasons. Further, the availability of organic carbonate selective membranes is also limited compared to dehydrative membranes. One reason behind this could be a generic membrane selectivity design based on molecular sizes, resulting in lower diffusion of larger molecules, in which organic carbonates are generally compared to water. Wang et al. [13] engineered a nano-silica PDMS composite membrane to separate DMC from methanol in 2011. Even though a relatively high flux was achieved, yielding more than 700 g/m²h of membrane flux, the selectivity (${\alpha }_{MeOH/DMC}$) for this configuration remained comparably low, just under 4 [13]. Another approach was presented by Číhal et al. by deploying a polymer of intrinsic microporosity (PIM-1) membrane for MeOH/DMC separation, achieving slightly higher selectivities of 5 but with significantly higher permeances with values up to 1.82 mol/m²h∗Pa than other solutions operated under sweep gas [14]. The earliest entry in the literature discussing the actual reactive conditions in which DEC would be formed was reported by Dibenedetto et al. [15] by testing several esterification-enhancing catalysts and a process loop including a pervaporation unit prior to the reaction. The consecutive dehydration of the reaction was identified as a central factor in boosting the reaction, which is why merging water separation and the catalyst into a single entity of a reactive membrane was proposed. Obstacles such as the DEC loss into the permeate for tested polymeric membranes have been reported. Additionally, the issue of a significant DEC loss due to the CO₂ content in the feed seems to be unsolved. However, it is advantageous that the operating temperatures of the membranes were comparably high and very close to the DEC formation temperature within the reactor [15]. Décultot et al. [16] also applied pervaporation to dehydrate the ternary DEC/water/EtOH mixture with a PERVAP^TM 4100 polymeric membrane, reducing the water content below 0.33% in the retentate. Further, it is stated that a DEC content of 0–15% seemingly has no effect on the water flux. The ideal selectivities ${\alpha }_{H_{2}O/DEC}$ and ${\alpha }_{H_{2}O/EtOH}$ have a range of 800–1630, which is tremendously high in comparison to other tested membranes. However, the total fluxes have a range of 8–14 g/m²h and barely exceed purities of 80% for water in some cases. A temperature of 87 °C was the highest applied temperature, which is low compared to the synthesis conditions of DEC [16]. As already mentioned, the data on the dehydration of ternary systems and the separation of alcohol/organic–carbonate mixtures, including DEC, is very limited. However, DEC is expected to have similarities in terms of synthesis and separation methods like DMC. Thus, it is assumed to show a similar behavior to some pervaporation applications for DMC, of which further dehydration applications are reported in the literature [17,18,19,20].

1.2. Partial Least Squares Regression

In spectroscopy, partial least squares regression (PLS regression) is used to predict chemical or physical properties (e.g., concentration, viscosity, density, etc.) from spectra after calibration. Therefore, PLS has become a standard method in chemometrics and is primarily used in infrared, Raman, and fluorescence spectroscopy. Some recent example applications in ATR-FTIR spectroscopy for the quantification and qualification/authentication of chemical/biological substances or physical properties are as follows:

• Comparison of Raman and attenuated total reflectance (ATR) infrared spectroscopy for water quantification in a natural deep eutectic solvent [21];

• Application of ATR-FTIR spectroscopy along with regression modeling for the detection of adulteration of virgin coconut oil with paraffin oil [22];

• ATR-FTIR spectroscopy and chemometric techniques for the determination of polymer solution viscosity in the presence of SiO₂ nanoparticles and salinity [23];

• Further applications can be found in [24,25,26,27,28,29,30].

PLS can be described as an extension of principal component regression (PCR) or a combination of principal component analysis (PCA) and multiple regression. The main difference between PLS and other multivariate methods is the inclusion of the structure of the Y-data (variables to be predicted, e.g., concentrations) during the determination of the principal components for the X-data (predictor variables, e.g., spectra). This procedure increases the relation between the spectral data (X-data) and the analyte data (Y-data), resulting in slightly rotated principal components, termed PLS components. A good description of the mathematical background behind PLS can be found in [31].

2. Materials and Methods

2.1. Pervaporation System Setup

Figure 1 depicts a flow diagram of the membrane pervaporation system used. The specific components, valves, and sensors outlined in the flow diagram are listed in Table 1. For comprehensive details regarding the system’s design, construction, and testing, please refer to the master’s theses by G. Greilinger [32] and M. Annerl [33].

Several feed mixtures with various concentrations consisting of water (deionized), ethanol (99.9%, denatured with toluene, obtained from AustrAlco, Spillern, Austria), and DEC (99.9%, purchased from Carl Roth, Karlsruhe, Germany) were acquired and prepared for the experiments and standards. Preliminary tests were carried out to investigate the suitability of the design. The applied parameters are presented in Table 2. The data and parameters for the experimental campaign of the membranes are presented and discussed from Section 3.2.

Approximately 1 L of a defined mixture consisting of DEC, ethanol, and water (feed mixture) was filled into the tank (B1). Subsequently, the feed was pumped through a heat exchanger (W1) to reach the operating temperature. A water bath provided the operating temperature of the membrane module connected to a tube wrapped around the module. Tank B1 was pressurized through a feed line by introducing CO₂ after a consistent operating temperature was reached. Following this, a continuous carrier gas flow (N₂) on the permeate side of the membrane module was released under atmospheric pressure. The permeate was separated from the carrier gas flow by condensation with subsequent collection in a separator (F1), which maintained a temperature range between −15 °C and −20 °C due to a salted iced water mixture. This temperature range was selected because of the anticipation of a high water content in the permeate. It is sufficiently low to separate water without causing freezing or condensation in the pipes of the carrier gas. A Liebig condenser (W2) with a collecting flask was also used in high permeate flow cases. The volume flow of the circulating mixture was adjusted to a flow of 1.5–1.7 L/min.

Extraction of samples from the circulating mixture occurred by opening valve V7, and for the permeate by simultaneously emptying the separator F1 and (if present) collection of the content in the flask (W2) of the Liebig condenser. The retentate of the PV membrane process was recycled back to the feed tank, thus continuously changing the concentration actually fed to the membrane. It will be referred to as “circulate” in the following sections for better understanding.

Prior to each test, the system was rinsed with high-purity ethanol (99.9%) to remove residues. The predefined test mixture was then filled into the system and operated for at least half an hour. This was to free the system from local concentration differences caused by residues from dead volumes from preliminary tests and to achieve optimum homogenization. Furthermore, this procedure was used to test the system for the stability of the test parameters.

Figure 2a showcases the tubular module integrated into the membrane pervaporation system. The module consists of three main components, with the ZEBREX ZX0 (DeltaMem) tubular ceramic membrane featuring a module length of 20 cm and a non-disclosed selective layer measuring 0.00565 m². Alongside the heating hoses of the water bath (W1) wrapped around the casing, the module was also insulated with foam. Figure 2b displays the flat module integrated into the membrane pervaporation system. Again, assembled on a supporting structure, the most crucial element is the flat-sheet polymer membrane with an active membrane diameter of 9 cm, positioned between the upper and lower parts of the module. The feed inlet and outlet occur in the upper part, while the sweep gas inlet and the permeate-carrier gas mixture outlet occur in the lower part.

Heating for the flat module was conducted similarly to that of the tube module, utilizing a water hose and insulation with foam. The polymer membranes (PERVAP^TM 4100, PERVAP^TM 4101, PERVAP^TM 4155-80) were purchased from DeltaMem (Allschwil, Switzerland). The membranes were each preserved in the feed mixture for 24 h prior to the installation to condition them appropriately.

2.2. Analytics

The permeate and circulating mixture were analyzed using a gas chromatograph (GC-2010, Shimadzu, Kyōto, Japan) equipped with an AOC-5000 autosampler, a flame ionization detector (GC-FID), and an RTX Volatiles capillary column (60 m length, 0.53 mm inner diameter, film thickness 2 µm). The sample compositions were evaluated based on standards (>100) consisting of water, ethanol, and DEC in varying concentrations. For the quantitative analysis of water in the feed and permeate samples, volumetric Karl Fischer titration (KFT, Karl Fischer titration) was employed. The double analyses were conducted using the Eco KF Titrator (Metrohm, Herisau, Switzerland). Additionally, all samples were analyzed using an FTIR-ATR spectrometer (Vertex 70, Bruker, Billerica, MA, USA) with PLS modeling.

2.3. PLS

In this work, PLS was used to create a calibration model for the prediction of the ethanol, diethyl carbonate, and water concentrations in a defined mixture (the feed), which was analyzed with an ATR-FTIR spectroscopy device. For this purpose, a MATLAB (2023b version) algorithm was created, which applies the plsregress function from the statistics and machine learning toolbox [34]. The algorithm behind the plsregress function provided by MATLAB is based on SIMPLS [35] and automatically performs a k-fold cross-validation [36].

To obtain the calibration model, a set of standard mixtures was created in the following concentration ranges: 0–100 wt% ethanol, 0–100 wt% DEC, and 0–1.5 wt% water. For each measurement, a wavenumber of 400–4000 cm⁻¹ was taken into account, and 16 spectra were subsequently recorded and averaged, and the blank value was measured for every 4 measurements. The raw spectra (without smoothing, baseline correction, or derivation of the spectra) were used for further calibration.

The optimum number of PLS components was determined via mean squared error (MSE), which was calculated from models with a number of PLS components ranging 1–12. This step is necessary to prevent underfitting or overfitting [31]. Additionally, during the calculation of these models, ethanol and DEC were evaluated together, while water was evaluated separately to achieve a better predictive performance. The MSE was determined for all models and applied to the number of PLS components (see Figure 3). If the concentration is predicted via PLS regression, the MSE is generally defined by

(2)$MSE={\displaystyle \frac{1}{n}}\times \sum _{i=0}^n{(y_i-{\hat{y}}_i)}^2,$

where $n$ is the number of elements in the concentration matrix $\mathbf{Y}$, $y_i$ is each element of $\mathbf{Y}$, and ${\hat{y}}_i$ is the predicted concentration by the PLS regression model [31].

For the PLS regression model of DEC and ethanol (model 1), a PLS component number of 3 was selected, and for water (model 2), a number of 7 was selected. The two actual calibration models were created using these numbers of PLS components.

The plsregress function provided by MATLAB was used to determine the regression coefficients for these two PLS regression models. The regression coefficients are essential for the prediction of an unknown sample. The MATLAB program of the PLS regression model of DEC and ethanol (model 1) can be found in the Supplementary Materials.

Furthermore, to examine the accuracy of the PLS regression models, the coefficients of determination $R^2$ and the standard error $SE$ were calculated for each model. If the concentration is predicted via PLS regression, $R^2$ and $SE$ are defined as follows:

(3)$R^2={\displaystyle \frac{\sum _{i=0}^n{\left({\hat{y}}_i-\overline{y}\right)}^2}{\sum _{i=0}^n{\left(y_i-\overline{y}\right)}^2}}$

(4)$SE=\sqrt{{\displaystyle \frac{\left(y_i-{\hat{y}}_i-BIAS\right)}{n-1}}}$

where $n$ is the number of elements in the concentration matrix $\mathbf{Y}$, $y_i$ is each element of $\mathbf{Y}$, $\overline{y}$ is the average of all elements from $\mathbf{Y}$, ${\hat{y}}_i$ is the predicted concentration by the PLS regression model, and $BIAS=\sum _{i=0}^n\left(y_i-{\hat{y}}_i\right)/n$ [31].

3. Results

3.1. PLS Model

The two PLS regression models are sufficiently accurate for the application since all $R^2$ values are >0.99, and the $SE$ of the organic components (model 1) is <0.5 wt% and of water (model 2) is <0.05 wt% (see Table 3).

Figure 4a,c illustrates the residuals of each standard mixture. These diagrams provide a good overview of the absolute deviations between the actual mass percentages of the standard mixtures and the mass percentages predicted by the PLS regression model. The standard error $SE$ summarizes this observation in a single quantity.

In Figure 4b,d, the actual mass percentages of the standard mixtures are plotted against the mass percentages predicted by the PLS regression model. It can be seen that there is a strong correlation between these two values for both models. This property is summarized by the coefficient of determination $R^2$.

3.2. Comparison of Tested Membranes

For a better understanding of the system, a more extensive concentration range of the components—water, ethanol, and minimally also DEC—was covered in the preliminary tests. Any minimal limitations concerning the applied methodology (leakage of the system, temperature drop at the measuring points used, pressure loss over the test period) were successfully corrected before the start of the test campaigns (V1–V17), which will be discussed in the following sections.

Four membranes were tested at different feed pressures to evaluate their eligibility for use in an integrated dehydration process for organic mixtures. In each experiment (V1–V16), mixtures consisting of 1% of each of H₂O and DEC, and 98% ethanol were fed into the equipment and homogenized in the system. An overview of the resulting experimental data from these assessments is provided in Table 4.

First, it becomes apparent that the concentration of the circulate differs from the concentration fed to the system. This can be explained by the fact that, despite the preparation and rinsing processes, dead volumes remain in the system, marginally shifting the overall concentrations in the system after homogenizing the circulating mixture. This observation also applies to all subsequent experiments and is considered negligible.

While the tubular membrane module design allowed pressures up to 5 bar, the flat-sheet models were tested at 1 and 3 bar. The tubular membrane module’s permeate stream had high water purity levels of >99%, while this quality remains unmatched by the polymer membranes. Only the PERVAP^TM 4101 membrane delivers a permeate stream of mainly water, with more than 79%.

The water flux is almost the only constituent of total fluxes for the ceramic membrane experiments. The 4155-80 and 4100 polymer membranes deliver comparably high total fluxes of more than 250 g/m²h, resulting in comparably large mass streams. Even though the permeance of water is within a similar magnitude for the two mentioned membranes, as for the ceramic one, it is already highly implied that both membranes are rather unsuitable for the desired application since large DEC and ethanol fluxes have also been observed, resulting in a comparably lower selectivity. Even though the water flux is the main constituent for the PERVAP^TM 4101 membrane, it shares some drawbacks with the other tested polymeric membranes and is still not competitive enough regarding the selectivity compared to the ceramic membrane. Interestingly, the application by Décultot et al. [16] of the same membrane module (PERVAP^TM 4100) delivers lower fluxes but higher selectivities for water, even though the mixtures submitted to the membrane are comparable. However, the experiments conducted by Décultot et al. were operated under a vacuum. The driving force of this work’s setup was enhanced by the sweep gas of N₂ on the permeate side and pressurization with carbon dioxide in the circulating system. The comparatively poor quality of the permeate could also be potentially caused by effects such as membrane swelling due to the combined presence of CO₂ and relatively high temperature and pressure. Therefore, particular emphasis was placed on the ZEBREX ZX0 membrane.

3.3. Effect of Feed Water Content on Dehydration

Several experiments with the ZEBREX ZX0 membrane were carried out to assess the membrane’s behavior on the water flux in dependency on the composition of the feed. The results of a long-term experiment carried out for 16 h with a water feed concentration of 0.7% are illustrated in Figure 5. The rest of the feed comprised about 1% DEC and >98% ethanol.

The flux of water decreases on average, as Figure 5a portrays, throughout the test period. Furthermore, Figure 5b demonstrates a correlation between the permeate flow of water and the water content in the feed, illustrating that the water flux decreases as the water content in the feed decreases. A closer look at Figure 5c reveals an almost linear decrease in the circulating water concentration with passing time, depleting the water content in the circulate close to 0.2 wt% after 16.5 h of experimental running time. Further trend development was not assessed, but this data does not reveal a diminishment of the decreasing trend. The permeate was identified to be mainly water (Figure 5d) with a purity of mostly >98 wt%. Despite the trend, significant fluctuations in the measured values can be observed in Figure 5a,b,d. Due to the low sample mass of the permeate, slight fluctuations occurred in the measurements. These fluctuations increase by scaling up to the units shown (Figure 5), which makes the fluctuations appear large. Despite the comparatively poor $R^2$, a trend can nevertheless be recognized. This is particularly evident in the constant decrease in the water concentration of the circulating mixture, as shown in Figure 5c.

3.4. Effect of $T_C$ on Dehydration

Since pressure and temperature are usual parameters that influence the flux, temperature variation experiments were performed, as shown in in Table 5.

A closer look at the H₂O fluxes and the permeate water content in Table 5 gives further insight into the behavior of the ceramic membrane. The flow of the organic part of the permeate rises while the percentage share of H₂O stays rather consistent. The total flux mainly consists of water. However, the absolute volume seems to level off somewhere between 98 and 99%. Considering the feed water content, a decreasing water flux with a reduced feed water amount is expected. The experiments carried out at 75 and 85 °C, however, debunk this theory by delivering consistently rising fluxes with rising temperature $T_C$. The last experiment carried out with even less H₂O in the feed and at a temperature of 98 °C, yields a purity of 99.98% of H₂O, delivering a comparatively tremendous ideal selectivity $\alpha$ of >200,000. This also marks the end of a seemingly decreasing $\alpha$ in dependence on rising temperature. Further elevation of the feed temperature is expected to result in even higher fluxes of H₂O but possibly slightly decreased selectivities due to the enhanced diffusion of the other components as well. Nevertheless, an assessment of the membrane’s performance was conducted in the following section.

3.5. Effect of $P_C$ on Dehydration

Several experiments were carried out to observe the influence of temperature on the PV separation using the tubular ceramic membrane. The results of these experiments are documented in Table 6.

The initial concentrations are not precisely the same but close enough for the purposes of the experiments. A positive correlation between the transmembrane flux and the increased pressure difference was observed. At first glance, no particular correlation could be observed for the purity of the permeate. The permeate quality is highest in the test conducted at 3 bar, with 99.31%. The test conducted at 1 bar showed the lowest value of the entire test campaign for the ceramic ZEBREX ZX0 membrane, at 94.48%. Considering the minimal sample quantity (<0.1 mL), an outlier favoring lower water purity cannot be ruled out.

3.6. Effect of $V_s$ on Dehydration

The effect of the sweep gas flow rate on the separation was observed, and the data are summarized in Table 7.

It has to be mentioned that the feed concentrations of the experiments are not perfectly equivalent. Nonetheless, with rising sweep gas flow rates of N₂, a positive trend can be observed in the permeate concentrations and the flux. The purities of the permeates are consistently at least as high as 99%, while the flux for a sweep gas flow of 1.4 L/min delivered the highest flux of all experiments using the ceramic ZX0 membrane. While the permeate quality remained nearly unchanged, the flux decreased to less than one-third by cutting the sweep gas flow to half.

This indicates that dissolving water molecules on the selective layer of the membrane and diffusing through the membrane matrix is still faster than the desorption process on the permeate side. A high water concentration on the membrane permeate side surface could support this assumption. Conversely, doubling the sweep gas flow rate increased the flux by 50%.

3.7. Effect of Increased DEC Concentration in the Circulating Mixture on Dehydration

In order to gain an impression of the behavior of the membrane at a higher DEC concentration, the initial concentration of the circulating mixture was changed for experiment V17 (parameter). In this experiment, 8.89% DEC, 0.30% H₂O, and 90.81% ETOH were used and homogenized within the system, as in the other experiments (V1–V16). After the initial homogenization time, 8.92% DEC and 0.33% water were measured at the beginning of the experiment. Figure 6 shows the water concentration in the circulating mixture and in the permeate.

The water concentration in the circulating mixture decreases depending on the duration of the experiment. Despite the initially very low water concentration, the mixture was still dehydrated to 0.27% water concentration. Due to the very low permeate sample quantity, the water concentrations of only two samples could be determined, which, in both cases, matched the qualities of the other tests well, with >96%. The transmembrane flux was determined to be 10.6 g/m²h after 3 h and 8.6 g/m²h after a further 2 h and 15 min. Although the test time was somewhat shorter, it can still be observed that the continuously decreasing water concentration in the circulate also reduces the transmembrane flux. The determined permeance of 1.03 × 10⁻⁶ also fits into the overall context compared with the other experiments’ results. In addition, this experiment shows that selective separation of water from the mixture ($\alpha =5839$) is nevertheless possible with an increased DEC concentration.

4. Conclusions

Ternary DEC/water/EtOH mixtures were introduced into a pervaporation setup fordehydration. The removal of water is considered a crucial element in the direct synthesis of DEC, as water drastically slows down the reaction. The literature has reported challenges in the operation of pervaporation in combination with the direct synthesis due to stability issues of membranes regarding the applicable temperature, pressure, and CO₂ presence. The tubular ceramic membrane could be a possible solution for this demand, primarily since it operates close to DEC’s synthesis temperature (>130 °C). Ceramic membranes typically do not suffer as frequently from swelling as polymeric membranes. Further, the presence of CO₂ was primarily below the detection limit in the permeate, while the circulating feed was saturated with CO₂. The polymeric membranes PERVAP^TM 4155-88 and 4100 provided the highest total fluxes and simultaneously the most significant losses in DEC and ethanol, resulting in comparatively low selectivities. Therefore, these membranes are not considered eligible for the initial separative instance of dehydration. The third polymeric membrane, PERVAP^TM 4101, showed superior dehydration characteristics ($\alpha$ > 120) compared to those of the other tested polymeric membranes ($\alpha$ < 18). Nevertheless, the losses of DEC and ethanol into the permeate were still too significant to be neglected. Therefore, all polymeric membranes are expected to perform insufficiently in terms of permeate quality within a direct synthesis process of DEC. However, the PERVAP^TM 4101 membrane could be considered in a possible second stage of permeate purification. Further tests would be necessary in this case. The tubular ZEBREX ZX0 membrane shows the best overall dehydration characteristics, with selectivities of at least 3800. The selectivity exceeded 10,000 in most cases, delivering permeate streams with mostly >98 wt% water. Parameters such as temperature, pressure, and sweep gas velocity significantly impacted the flux. Increasing the DEC content in the system had no noticeable effect on the permeate quality. Adapting the operating temperature to the proposed temperature for DEC synthesis may even enhance the flux of water but may also result in a reduced quality of the permeate. Nevertheless, increasing the pressure did not result in the same performance as altering the other parameters. The overall fluxes were observed to be comparatively low compared to other dehydration applications but higher than those with the ternary system composed of DEC/H₂O/EtOH. The ceramic membrane could be a suitable option to dehydrate the observed quaternary (DEC/H₂O/EtOH/CO₂) mixture. However, it has not been fully assessed whether the membrane could dehydrate the solution with sufficiently high flux. At the same time, monitoring the process using FTIR-ATR analyses is a time-saving approach, as using the PLS model provides relatively immediate and sufficiently accurate assessments of the samples. Accordingly, using the PLS model is preferable to using competing methods such as the KFT and GC.

Author Contributions

Conceptualization, M.T.-K., F.K., C.J. and M.H.; methodology, K.A., M.A., G.G. and C.J.; software, M.A. and G.G.; validation, K.A., M.A. and C.J.; formal analysis, K.A., M.A. and C.J.; investigation, K.A., M.A., G.G., M.T.-K., F.K., C.J. and M.H.; resources, M.T.-K., F.K., C.J. and M.H.; data curation, K.A., M.A. and G.G.; writing—original draft preparation, K.A. and M.A.; writing—review and editing, K.A., M.A., G.G., M.T.-K., F.K., C.J. and M.H.; visualization, K.A. and M.A.; supervision, K.A., M.T.-K., C.J. and M.H.; project administration, M.T.-K. and M.H.; funding acquisition, M.T.-K., F.K. and M.H. All authors have read and agreed to the published version of the manuscript.

Data Availability Statement

The data recorded in this study are available upon request from the corresponding author.

Acknowledgments

The authors are highly grateful for Markus Pekovits’ technical support on the electrical installations.

Conflicts of Interest

Author Gerhard Greilinger, Magdalena Teufner-Kabas and Florian Kabas were employed by the company kleinkraft OG. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Abbreviations

ATR	attenuated total reflection
CCU	carbon capture and utilization
CeO2	cerium oxide
CO2	carbon dioxide
DEC	diethyl carbonate
DMC	dimethyl carbonate
EtOH	ethanol
FID	flame ionization detector
FTIR	Fourier-transform infrared spectroscopy
GC	gas chromatograph
GHG	greenhouse gas
H2O	water
IPA	Isopropanol
KFT	Karl Fischer titration
MeOH	methanol
MSE	mean squared error
PLS	partial least squares
SE	standard error
V	experiment
${\mathit{J}}_{\mathit{i}}$	flux of component $\mathit{i}$, $\mathit{g}\mathbf{/}{\mathit{m}}^{\mathbf{2}}\mathit{h}$
${\mathit{J}}_{\mathit{t}\mathit{o}\mathit{t}\mathit{a}\mathit{l}}$	total flux, $\mathit{g}\mathbf{/}{\mathit{m}}^{\mathbf{2}}\mathit{h}$
$\mathit{n}$	number of elements in $\mathit{Y}$
${\mathit{m}}_{\mathit{i}}$	mass of component $\mathit{i}$, $\mathit{g}$
${\mathit{M}}_{\mathit{i}}$	molar mass of component $\mathit{i}$, $\mathit{g}\mathbf{/}\mathit{m}\mathit{o}\mathit{l}$
${\mathit{P}}_{\mathit{c}}$	feed pressure, $\mathit{b}\mathit{a}\mathit{r}$
${\mathit{Q}}_{\mathit{i}}$	permeance of component $\mathit{i}$, $\mathit{m}\mathit{o}\mathit{l}\mathbf{/}\mathbf{(}\mathit{P}\mathit{a}\mathbf{\ast }{\mathit{m}}^{\mathbf{2}}\mathit{h}\mathbf{)}$
${\mathit{R}}^{\mathbf{2}}$	coefficient of determination
${\mathit{T}}_{\mathit{c}}$	feed temperature, $\mathbf{°}\mathit{C}$
${\mathit{V}}_{\mathit{s}}$	volumetric sweep gas flow, $\mathit{L}\mathbf{/}\mathit{m}\mathit{i}\mathit{n}$
${\mathit{w}}_{\mathit{c}\mathbf{,}\mathit{i}}$	$\mathit{w}\mathit{t}\mathbf{\%}$ of component $\mathit{i}$ in circulate
${\mathit{w}}_{\mathit{p}\mathbf{,}\mathit{i}}$	$\mathit{w}\mathit{t}\mathbf{\%}$ of component $\mathit{i}$ in permeate
${\mathit{y}}_{\mathit{i}}$	element $\mathit{i}$ of matrix $\mathit{Y}$
${\overline{\mathit{y}}}_{\mathit{i}}$	average of all elements of $\mathit{Y}$
${\hat{\mathit{y}}}_{\mathit{i}}$	prediction of element $\mathit{i}$ in $\mathit{Y}$
$\mathit{Y}$	concentration matrix
$\mathit{\alpha }$	ideal selectivity of H2O over ethanol

Footnotes

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Figures and Tables

Figure 1. Flow diagram of the pervaporation equipment [33].

Figure 2. Design of the tubular membrane module (a) and the flat-sheet module (b) (adapted from [33]).

Figure 3. Mean squared error against the number of partial least squares components. ((a) Model 1) Models for ethanol and diethyl carbonate. ((b) Model 2) Models for water.

Figure 4. Residual and correlation plots for both PLS regression models: (a,b): model 1 and (c,d): model 2.

Figure 5. Illustration of the flux over time (a) and in dependency on the feed water content (b) as well as the circulates (c) and the permeates (d) water content using the ZEBREX ZX0 membrane.

Figure 6. Experiment with increased DEC and decreased water concentration in the circulate of the tubular ZEBREX ZX0 ceramic membrane at 5 bar and 85 °C.

Table 1

List of applied sensors.

	Abbreviation	Measured Value	Measured Medium
pressure	PIC1	$P$	CO2
	PIC2	$P_S$	sweep gas
	PR1	$P_F$	feed
temperature	TR1	$T_F$	feed
	TIR1	$T$	membrane module
volumetric flow	FR1	$V_F$	feed
	FRC1	$V_S$	sweep gas
relative humidity	MIR1/MIR2	$\phi$	permeate
volumetric concentration of CO2	QR1	$v_{p,C{O}_2}$	permeate

Table 2

Overview of applied preliminary tests of the pervaporation setup using the ceramic ZEBREX ZX0 membrane.

Name	${\mathit{w}}_{\mathit{c},\mathit{i}}$			${\mathit{P}}_{\mathit{c}}$	${\mathit{T}}_{\mathit{b}}$	Duration	${\mathit{N}}_2$—Sweep Gas Flow
	$\mathbf{\left[}\mathit{w}\mathit{t}\mathbf{\%}\mathbf{\right]}$			$\mathbf{\left[}\mathit{b}\mathit{a}\mathit{r}\mathbf{\right]}$	$\mathbf{\left[}°\mathit{C}\mathbf{\right]}$	$\mathbf{\left[}\mathit{h}\mathbf{\right]}$	$\mathbf{\left[}\mathit{L}/\mathit{m}\mathit{i}\mathit{n}\mathbf{\right]}$
	DEC	H2O	EtOH
PV1	0	0.00	100.00	3	80	2.00	1.5
PV2	0	6.20	93.80	3	80	1.53	1.5
PV3	0.79	0.99	98.22	7	80	5.33	0.5
PV4	1.21	0.00	98.79	6	85	4.30	0.7
PV5	1.12	0.24	98.64	6	85	4.80	0.7
PV6	0.13	1.15	99.58	6	85	5.00	0.7
PV7	1.14	0.00	98.86	6	85	5.60	0.7
PV8	0.25	3.78	99.46	6	65–85	7.50	0.7
PV9	0.88	3.70	95.42	3–6	75	7.50	0.7
PV10	3.37	0.89	95.74	5.7	75	7.50	0.35–1.4
PV11	1.31	0.00	98.69	3	85	6.75	0.7
PV12	0.18	1.06	98.76	3	85	5.50	0.7

Table 3

Coefficient of determination and standard error.

Model	Analyte	${\mathit{R}}^2$	$\mathit{S}\mathit{E}$ [wt%]
1	DEC	0.9999	±0.2379
	EtOH	0.9996	±0.4302
2	water	0.9945	±0.0330

Table 4

Experimental data of tested membranes at 98 °C.

Name	Membrane	${\mathit{P}}_{\mathit{c}}$	${\mathit{w}}_{\mathit{c},\mathit{i}}$			${\mathit{w}}_{\mathit{p},\mathit{i}}$	${\mathit{J}}_{\mathit{i}}$		$\mathit{\alpha }$	${\mathit{Q}}_{{\mathit{H}}_2\mathit{O}}$
		$\mathbf{\left[}\mathit{b}\mathit{a}\mathit{r}\mathbf{\right]}$	$\mathbf{\left[}\mathit{w}\mathit{t}\mathbf{\%}\mathbf{\right]}$			$\mathbf{\left[}\mathit{w}\mathit{t}\mathbf{\%}\mathbf{\right]}$	$\mathbf{\left[}\mathit{g}/{\mathit{m}}^2\mathit{h}\mathbf{\right]}$		$\mathbf{\left[}-\mathbf{\right]}$	$\mathbf{\left[}\mathit{m}\mathit{o}\mathit{l}/\mathit{P}\mathit{a}\mathbf{\ast }{\mathit{m}}^2\mathbf{\ast }\mathit{h}\mathbf{\right]}$
			DEC	H2O	EtOH	H2O	H2O	Total
V1	ZEBREXTM ZX0 at 85 °C	5	0.90	0.98	98.70	99.6	30.5	30.7	>13,000	3.39 × 10−6
V2	PERVAPTM	1	0.96	1.01	98.63	12.3	32.5	268.7	13–15	1.80 × 10−5
V3	4155-80	3	0.99	0.96	98.60	12	45.7	377.5	11–17	8.45 × 10−6
V4	PERVAPTM	1	1.04	0.95	98.55	5.1	30.6	599.1	5–6	1.70 × 10−5
V5	4100	3	1.07	0.94	98.52	4.7	24.4	523.3	5–6	4.52 × 10−6
V6	PERVAPTM 4101	3	1.31	1.08	98.28	73.9	11.8	16.0	120–180	2.19 × 10−6

Table 5

Overview of temperature variation experiments of the tubular ZEBREX ZX0 ceramic membrane at 3 bar.

Name	${\mathit{T}}_{\mathit{c}}$	${\mathit{w}}_{\mathit{c},{\mathit{H}}_2\mathit{O}}$	${\mathit{w}}_{\mathit{p},{\mathit{H}}_2\mathit{O}}$	${\mathit{J}}_{{\mathit{H}}_2\mathit{O}}$	${\mathit{J}}_{\mathit{t}\mathit{o}\mathit{t}\mathit{a}\mathit{l}}$	$\mathit{\alpha }$	${\mathit{Q}}_{{\mathit{H}}_2\mathit{O}}$
	$\mathbf{\left[}°\mathit{C}\mathbf{\right]}$	$\mathbf{\left[}\mathit{w}\mathit{t}\mathbf{\%}\mathbf{\right]}$	$\mathbf{\left[}\mathit{w}\mathit{t}\mathbf{\%}\mathbf{\right]}$	$\mathbf{\left[}\mathit{g}/{\mathit{m}}^2\mathit{h}\mathbf{\right]}$	$\mathbf{\left[}\mathit{g}/{\mathit{m}}^2\mathit{h}\mathbf{\right]}$	$\mathbf{\left[}-\mathbf{\right]}$	$\mathbf{\left[}\mathit{m}\mathit{o}\mathit{l}/\mathit{P}\mathit{a}\mathbf{\ast }{\mathit{m}}^2\mathbf{\ast }\mathit{h}\mathbf{\right]}$
V7	65	0.67	98.38	16.3	16.5	9990	1.80 × 10−6
V8	75	0.80	98.53	19.2	19.5	5068	2.13 × 10−6
V9	85	0.72	98.39	30.5	31.0	3800	3.38 × 10−6
V10	98	0.43	99.98	13.3	13.3	208,755	1.47 × 10−6

Table 6

Overview of temperature variation experiments of the tubular ZEBREX ZX0 ceramic membrane at 75 °C.

Name	${\mathit{P}}_{\mathit{c}}$	${\mathit{w}}_{\mathit{c},{\mathit{H}}_2\mathit{O}}$	${\mathit{w}}_{\mathit{p},{\mathit{H}}_2\mathit{O}}$	${\mathit{J}}_{{\mathit{H}}_2\mathit{O}}$	${\mathit{J}}_{\mathit{t}\mathit{o}\mathit{t}\mathit{a}\mathit{l}}$	$\mathit{\alpha }$	${\mathit{Q}}_{{\mathit{H}}_2\mathit{O}}$
	$\mathbf{\left[}\mathit{b}\mathit{a}\mathit{r}\mathbf{\right]}$	$\mathbf{\left[}\mathit{w}\mathit{t}\mathbf{\%}\mathbf{\right]}$	$\mathbf{\left[}\mathit{w}\mathit{t}\mathbf{\%}\mathbf{\right]}$	$\mathbf{\left[}\mathit{g}/{\mathit{m}}^2\mathit{h}\mathbf{\right]}$	$\mathbf{\left[}\mathit{g}/{\mathit{m}}^2\mathit{h}\mathbf{\right]}$	$\mathbf{\left[}-\mathbf{\right]}$	$\mathbf{\left[}\mathit{m}\mathit{o}\mathit{l}/\mathit{P}\mathit{a}\mathbf{\ast }{\mathit{m}}^2\mathbf{\ast }\mathit{h}\mathbf{\right]}$
V11	1	0.80	94.48	10.9	11.5	8005	6.03 × 10−6
V12	3	0.72	99.31	12.7	12.7	180,984	2.34 × 10−6
V13	5	0.77	98.41	19.9	20.2	86,036	2.20 × 10−6

Table 7

Overview of sweep gas variation experiments of the tubular ZEBREX ZX0 ceramic membrane at 5 bar.

Name	${\mathit{V}}_{\mathit{s}}$	${\mathit{w}}_{\mathit{c},{\mathit{H}}_2\mathit{O}}$	${\mathit{w}}_{\mathit{p},{\mathit{H}}_2\mathit{O}}$	${\mathit{J}}_{{\mathit{H}}_2\mathit{O}}$	${\mathit{J}}_{\mathit{t}\mathit{o}\mathit{t}\mathit{a}\mathit{l}}$	$\mathit{\alpha }$	${\mathit{Q}}_{{\mathit{H}}_2\mathit{O}}$
	$\mathbf{\left[}\mathit{L}/\mathit{m}\mathit{i}\mathit{n}\mathbf{\right]}$	$\mathbf{\left[}\mathit{w}\mathit{t}\mathbf{\%}\mathbf{\right]}$	$\mathbf{\left[}\mathit{w}\mathit{t}\mathbf{\%}\mathbf{\right]}$	$\mathbf{\left[}\mathit{g}/{\mathit{m}}^2\mathit{h}\mathbf{\right]}$	$\mathbf{\left[}\mathit{g}/{\mathit{m}}^2\mathit{h}\mathbf{\right]}$	$\mathbf{\left[}-\mathbf{\right]}$	$\mathbf{\left[}\mathit{m}\mathit{o}\mathit{l}/\mathit{P}\mathit{a}\mathbf{\ast }{\mathit{m}}^2\mathbf{\ast }\mathit{h}\mathbf{\right]}$
V14	0.35	0.86	99.43	9.7	9.7	37,290	1.07 × 10−6
V15	0.70	0.98	99.55	30.5	30.5	13,708	3.39 × 10−6
V16	1.40	0.90	99.77	45.9	46.0	59,724	5.09 × 10−6

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/separations11100289/s1, The source code for the developed pls model is submitted with the manuscript.

References

1. Kanike, U.K. View of Factors Disrupting Supply Chain Management in Manufacturing Industries. J. Supply Chain Manag. Sci.; 2023; 4, pp. 1-24. [DOI: https://dx.doi.org/10.18757/jscms.2023.6986]

2. Liu, G.; Jin, W. Pervaporation Membrane Materials: Recent Trends and Perspectives. J. Membr. Sci.; 2021; 636, 119557. [DOI: https://dx.doi.org/10.1016/j.memsci.2021.119557]

3. Norkobilov, A.; Gorri, D.; Ortiz, I. Comparative Study of Conventional, Reactive-Distillation and Pervaporation Integrated Hybrid Process for Ethyl Tert-Butyl Ether Production. Chem. Eng. Process. Process Intensif.; 2017; 122, pp. 434-446. [DOI: https://dx.doi.org/10.1016/j.cep.2017.07.003]

4. Khoo, Y.S.; Tjong, T.C.; Chew, J.W.; Hu, X. Techniques for Recovery and Recycling of Ionic Liquids: A Review. Sci. Total Environ.; 2024; 922, 171238. [DOI: https://dx.doi.org/10.1016/j.scitotenv.2024.171238]

5. Zhang, H.; Liu, H.-B.; Yue, J.-M. Organic Carbonates from Natural Sources. Chem. Rev.; 2014; 114, pp. 883-898. [DOI: https://dx.doi.org/10.1021/cr300430e]

6. Zou, B.; Hu, C. Halogen-Free Processes for Organic Carbonate Synthesis from CO₂. Curr. Opin. Green Sustain. Chem.; 2017; 3, pp. 11-16. [DOI: https://dx.doi.org/10.1016/j.cogsc.2016.10.007]

7. Tabanelli, T.; Bonincontro, D.; Albonetti, S.; Cavani, F. Conversion of CO₂ to Valuable Chemicals: Organic Carbonate as Green Candidates for the Replacement of Noxious Reactants. Studies in Surface Science and Catalysis; Elsevier: Amsterdam, The Netherlands, 2019; Volume 178, pp. 125-144. ISBN 978-0-444-64127-4

8. Giram, G.G.; Bokade, V.V.; Darbha, S. Direct Synthesis of Diethyl Carbonate from Ethanol and Carbon Dioxide over Ceria Catalysts. New J. Chem.; 2018; 42, pp. 17546-17552. [DOI: https://dx.doi.org/10.1039/C8NJ04090G]

9. Ripoll, J.D.; Mejía, S.M.; Mills, M.J.L.; Villa, A.L. Understanding the Azeotropic Diethyl Carbonate-Water Mixture by Structural and Energetic Characterization of DEC(H₂O)(n) Heteroclusters. J. Mol. Model.; 2015; 21, 93. [DOI: https://dx.doi.org/10.1007/s00894-015-2593-5]

10. Hinchliffe, A.B.; Porter, K.E. A Comparison of Membrane Separation and Distillation. Chem. Eng. Res. Des.; 2000; 78, pp. 255-268. [DOI: https://dx.doi.org/10.1205/026387600527121]

11. Zhang, Y.; Liu, M.; Wu, Y.; Zhao, J.; Zhou, S.; Gu, P. Zwitterionic Polyamide Membranes via In-Situ Interfacial Polymerization Modification for Efficient Pervaporation Dehydration. Sep. Purif. Technol.; 2024; 333, 125847. [DOI: https://dx.doi.org/10.1016/j.seppur.2023.125847]

12. Fujiki, T.; Kaji, M.; Tamamizu, Y.; Yasunari, R.; Nakagawa, K.; Kitagawa, T.; Okamoto, Y.; Matsuoka, A.; Kamio, E.; Matsuyama, H. et al. Pervaporation Dehydration of an Isopropanol Aqueous Solution Using Microporous TiO₂-SiO₂-OCL (Organic Chelating Ligand) Composite Membranes Prepared under Different Firing Conditions. Sep. Purif. Technol.; 2024; 337, 126249. [DOI: https://dx.doi.org/10.1016/j.seppur.2023.126249]

13. Wang, L.; Han, X.; Li, J.; Zhan, X.; Chen, J. Hydrophobic Nano-Silica/Polydimethylsiloxane Membrane for Dimethylcarbonate–Methanol Separation via Pervaporation. Chem. Eng. J.; 2011; 171, pp. 1035-1044. [DOI: https://dx.doi.org/10.1016/j.cej.2011.04.058]

14. Číhal, P.; Vopička, O.; Durďáková, T.-M.; Budd, P.M.; Harrison, W.; Friess, K. Pervaporation and Vapour Permeation of Methanol-Dimethyl Carbonate Mixtures through PIM-1 Membranes. Sep. Purif. Technol.; 2019; 217, pp. 206-214. [DOI: https://dx.doi.org/10.1016/j.seppur.2019.02.023]

15. Dibenedetto, A.; Aresta, M.; Angelini, A.; Ethiraj, J.; Aresta, B.M. Synthesis, Characterization, and Use of Nb^V/Ce^IV-Mixed Oxides in the Direct Carboxylation of Ethanol by Using Pervaporation Membranes for Water Removal. Chem. Eur. J.; 2012; 18, pp. 10324-10334. [DOI: https://dx.doi.org/10.1002/chem.201201561]

16. Décultot, M.; Ledoux, A.; Fournier-Salaün, M.C.; Estel, L. Organic Carbonates Synthesis Improved by Pervaporation for CO₂ Utilisation. Green Process. Synth.; 2019; 8, pp. 496-506. [DOI: https://dx.doi.org/10.1515/gps-2019-0018]

17. Won, W.; Feng, X.; Lawless, D. Pervaporation with Chitosan Membranes: Separation of Dimethyl Carbonate/Methanol/Water Mixtures. J. Membr. Sci.; 2002; 209, pp. 493-508. [DOI: https://dx.doi.org/10.1016/S0376-7388(02)00367-8]

18. Dong, G.; Nagasawa, H.; Yu, L.; Wang, Q.; Yamamoto, K.; Ohshita, J.; Kanezashi, M.; Tsuru, T. Pervaporation Removal of Methanol from Methanol/Organic Azeotropes Using Organosilica Membranes: Experimental and Modeling. J. Membr. Sci.; 2020; 610, 118284. [DOI: https://dx.doi.org/10.1016/j.memsci.2020.118284]

19. Xu, X.; Van Eygen, G.; Molina-Fernández, C.; Nikolaeva, D.; Depasse, Y.; Chergaoui, S.; Hartanto, Y.; Van der Bruggen, B.; Coutinho, J.A.P.; Buekenhoudt, A. et al. Evaluation of Task-Specific Ionic Liquids Applied in Pervaporation Membranes: Experimental and COSMO-RS Studies. J. Membr. Sci.; 2023; 670, 121350. [DOI: https://dx.doi.org/10.1016/j.memsci.2023.121350]

20. Kim, D.; Lee, M.; Shin, Y.; Lee, J.; Lee, J.W. Direct Production of Diethyl Carbonate from Ethylene Carbonate and Ethanol by Energy-Efficient Intensification of Reaction and Separation. Chem. Eng. Process. Process Intensif.; 2023; 192, 109519. [DOI: https://dx.doi.org/10.1016/j.cep.2023.109519]

21. Elderderi, S.; Wils, L.; Leman-Loubière, C.; Henry, S.; Byrne, H.J.; Chourpa, I.; Munnier, E.; Elbashir, A.A.; Boudesocque-Delaye, L.; Bonnier, F. Comparison of Raman and Attenuated Total Reflectance (ATR) Infrared Spectroscopy for Water Quantification in Natural Deep Eutectic Solvent. Anal. Bioanal. Chem.; 2021; 413, pp. 4785-4799. [DOI: https://dx.doi.org/10.1007/s00216-021-03432-2]

22. Amit,; Jamwal, R.; Kumari, S.; Dhaulaniya, A.S.; Balan, B.; Singh, D.K. Application of ATR-FTIR Spectroscopy along with Regression Modelling for the Detection of Adulteration of Virgin Coconut Oil with Paraffin Oil. LWT; 2020; 118, 108754. [DOI: https://dx.doi.org/10.1016/j.lwt.2019.108754]

23. Mohammadi, M.; Khorrami, M.K.; Ghasemzadeh, H. ATR-FTIR Spectroscopy and Chemometric Techniques for Determination of Polymer Solution Viscosity in the Presence of SiO₂ Nanoparticle and Salinity. Spectrochim. Acta A Mol. Biomol. Spectrosc.; 2019; 220, 117049. [DOI: https://dx.doi.org/10.1016/j.saa.2019.04.041] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/31141782]

24. Anjos, O.; Santos, A.J.A.; Dias, T.; Estevinho, L.M. Application of FTIR-ATR Spectroscopy on the Bee Pollen Characterization. J. Apic. Res.; 2017; 56, pp. 210-218. [DOI: https://dx.doi.org/10.1080/00218839.2017.1289657]

25. Shi, H.; Yu, P. Comparison of Grating-Based near-Infrared (NIR) and Fourier Transform Mid-Infrared (ATR-FT/MIR) Spectroscopy Based on Spectral Preprocessing and Wavelength Selection for the Determination of Crude Protein and Moisture Content in Wheat. Food Control; 2017; 82, pp. 57-65. [DOI: https://dx.doi.org/10.1016/j.foodcont.2017.06.015]

26. Oleszko, A.; Hartwich, J.; Wójtowicz, A.; Gąsior-Głogowska, M.; Huras, H.; Komorowska, M. Comparison of FTIR-ATR and Raman Spectroscopy in Determination of VLDL Triglycerides in Blood Serum with PLS Regression. Spectrochim. Acta A Mol. Biomol. Spectrosc.; 2017; 183, pp. 239-246. [DOI: https://dx.doi.org/10.1016/j.saa.2017.04.020] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/28454077]

27. Mukherjee, S.; Martínez-González, J.A.; Dowling, D.P.; Gowen, A.A. Predictive Modelling of the Water Contact Angle of Surfaces Using Attenuated Total Reflection—Fourier Transform Infrared (ATR-FTIR) Chemical Imaging and Partial Least Squares Regression (PLSR). Analyst; 2018; 143, pp. 3729-3740. [DOI: https://dx.doi.org/10.1039/C8AN00320C]

28. Madsen, R.B.; Anastasakis, K.; Biller, P.; Glasius, M. Rapid Determination of Water, Total Acid Number, and Phenolic Content in Bio-Crude from Hydrothermal Liquefaction of Biomass Using FT-IR. Energy Fuels; 2018; 32, pp. 7660-7669. [DOI: https://dx.doi.org/10.1021/acs.energyfuels.8b01208]

29. Barmpalexis, P.; Karagianni, A.; Nikolakakis, I.; Kachrimanis, K. Artificial Neural Networks (ANNs) and Partial Least Squares (PLS) Regression in the Quantitative Analysis of Cocrystal Formulations by Raman and ATR-FTIR Spectroscopy. J. Pharm. Biomed. Anal.; 2018; 158, pp. 214-224. [DOI: https://dx.doi.org/10.1016/j.jpba.2018.06.004]

30. Nespeca, M.G.; Hatanaka, R.R.; Flumignan, D.L.; De Oliveira, J.E. Rapid and Simultaneous Prediction of Eight Diesel Quality Parameters through ATR-FTIR Analysis. J. Anal. Methods Chem.; 2018; 2018, 1795624. [DOI: https://dx.doi.org/10.1155/2018/1795624]

31. Kessler, W. Multivariate Datenanalyse für Die Pharma-, Bio- und Prozessanalytik: Ein Lehrbuch; Wiley-VCH: Weinheim, Germany, 2006; ISBN 978-3-527-61003-7

32. Greilinger, G. Construction of a Membrane Pervaporation Unit for the Dehydration of an Organic Reaction Mixture. Master’s Thesis; Technische Universität Wien: Vienna, Austria, 2024.

33. Annerl, M. Optimierung Der Diethylcarbonatsynthese Aus CO₂ Und Ethanol Durch Pervaporation. Master’s Thesis; Technische Universität Wien: Vienna, Austria, 2024; [DOI: https://dx.doi.org/10.34726/HSS.2024.93303]

34. Partial Least-Squares (PLS) Regression—MATLAB Plsregress—MathWorks Deutschland. Available online: https://de.mathworks.com/help/stats/plsregress.html (accessed on 8 March 2024).

35. De Jong, S. SIMPLS: An Alternative Approach to Partial Least Squares Regression. Chemom. Intell. Lab. Syst.; 1993; 18, pp. 251-263. [DOI: https://dx.doi.org/10.1016/0169-7439(93)85002-X]

36. Xu, Q.S.; Liang, Y.Z. Monte Carlo Cross Validation. Chemom. Intell. Lab. Syst.; 2001; 56, pp. 1-11. [DOI: https://dx.doi.org/10.1016/S0169-7439(00)00122-2]