Experimental investigation of bioethanol liquid phase dehydration using natural clinoptilolite

Graphical abstract

Abstract

An experimental study of bioethanol adsorption on natural Iranian clinoptilolite was carried out. Dynamic breakthrough curves were used to investigate the best adsorption conditions in bioethanol liquid phase. A laboratory setup was designed and fabricated for this purpose. In order to find the best operating conditions, the effect of liquid pressure, temperature and flow rate on breakthrough curves and consequently, maximum ethanol uptake by adsorbent were studied. The effects of different variables on final bioethanol concentration were investigated using Response Surface Methodology (RSM). The results showed that by working at optimum condition, feed with 96% (v/v) initial ethanol concentration could be purified up to 99.9% (v/v). In addition, the process was modeled using Box–Behnken model and optimum operational conditions to reach 99.9% for final ethanol concentration were found equal to 10.7 °C, 4.9 bar and 8 mL/min for liquid temperature, pressure and flow rate, respectively. Therefore, the selected natural Iranian clinoptilolite was found to be a promising adsorbent material for bioethanol dehydration process.

Introduction

Fuel grade bioethanol is one of the widely used alternative for fossil fuels or gasoline additive [1], [2]. In bioethanol–gasoline mixture, the presence of even a very small amount of water in bioethanol is unfavorable and leads to a two phase mixture [3], [4], [5]. The bioethanol dehydration is a process to eliminate water from bioethanol–water mixture up to 99.6% (V/V). There are several methods including azeotropic distillation [6], [7], [8], extractive distillation [9], [10], pervaporation with membranes [11], [12], [13] and adsorption using adsorbents [3], [5], [14], [15], [16], [17], [18], that are being used for water elimination to overcome the ethanol–gasoline mixing problem. The azeotropic distillation and extractive distillation are too expensive process [9], [19]. Literatures show that extractive distillation is more complex due to the design and process application and articles on energy consumption and cost, during recent years confirm that this method has high performance but needs further studies on energy consumption [20]. Conventional extractive distillation is energy consumption process because of using reboilers and condensors. Different refine processes were used to improve conventional extractive distillation such as heat-pump-assisted extractive distillation for bioethanol purification [21], Ethanol dehydration via azeotropic distillation with gasoline fraction mixtures as entrainers [20] and Control comparison of conventional and thermally coupled ternary extractive distillation processes [22]. In addition, although the pervaporation is a new generation in separation technology, it has industrial installation limitations. The adsorption by selective porous adsorbents is a common high performance method in bioethanol dehydration. Many studies have focused on different types of water adsorbents including biobased adsorbents namely natural corncobs, natural and activated palm stone and oak [3], [23], [24], Calcium Carbide [25], calcium chloride and lime [26], silica gel [27], cellulose and lignocellulose based (bleached wood pulp, oak sawdust and kenaf core) [14], [28], Aluminas and γ-alumina [29], Starch-Based Adsorbents [5], [24], [30] and different types of molecular sieves especially the zeolites [16], [31], [32], [33], [34]. Finding appropriate, effective and cheap adsorbent material is a way to reduce the final bioethanol production costs. The zeolites with porous structures and selectivity properties can let water molecules to penetrate inside pore volumes of hydrophilic adsorbents and separate ethanol–water mixture. The natural zeolites and clays such as clinoptilolite [17], [35], [36], [37], chabazite and phillipsite [38] are plentiful material in nature with hydrophilic properties suitable for ethanol–water separation. For instance, it has been shown that the clinoptilolite water adsorption capacity is more than 50% of water adsorption capacity of 3A zeolite [39]. In various previous studies, the parameters influencing ethanol–water separation such as temperature [39], system pressure [40], adsorption heat [41] and particle size [39] have been investigated. As a lot of industrial separation processes based on adsorption mechanism are carried out in liquid phase [42], using mesoporous adsorbents such as clinoptilolite is highly recommended to adsorb the big molecules in liquid phase [43]. Although there are some studies on using clinoptilolite as a adsorbent for purification of ethanol in liquid phase [35], [37], the effect of operational conditions has not well understood. So, we aim to use of Iranian clinoptilolite in both batch and continuous adsorption processes to separate the water contents from water/ethanol mixture which is usual product of biofuel production. In this research work, the Iranian natural clinoptilolite is presented as a cheap water adsorbent media to separate the water from hydrous bioethanol in a fixed bed setup. Furthermore, the optimum operational conditions have been found both experimentally and theoretically.

Material and methods

The deionised water and ethanol were purchased from Bidestan Co. (Qazvin-Iran). The natural clinoptilolite used in this research work was purchased from Afrazand Co. (East Semnan–Iran). The chemical analysis showed the high content of K⁺ and Na⁺. The zeolite was approximately 65 wt.% pure in clinoptilolite. The composition of the material based on X-ray fluorescence (XRF) (Model: Philips PW 2404) analysis was 71.159 wt.% SiO₂, 11.335 wt.% Al₂O₃, 0.936 wt.% Fe₂O_3, 0.807 wt.% CaO, 0.478 wt.% MgO, 3.064 wt.% Na₂O, 4.48 wt.% K₂O, 0.164 wt.% TiO₂ and 0.847 wt.% SO₃. Loss of ignition (LOI) is 6.23. The bulk density was calculated and it was found 820 kg m⁻³ (1–2 mm particle size). The silica modulus (molar ratio) of the sample was η = SiO₂/Al₂O₃ = 6.26.

The pore structure and surface area of Iranian clinoptilolite were characterized by N₂ adsorption–desorption isotherm at 77 K which has been illustrated in Fig.1a. Nitrogen adsorption was carried out using Belsorp mini II (Bel Japan). Before the experiments, the sample was dried to be degassed at 25 °C for 5 h and vacuum. The adsorption isotherm has hysteresis loop along with a relative pressure from 0.4 to 0.99. This isotherm is type I, which is typical property for mesoporous materials [44]. As it has been presented in Table 1, the BET surface area (a_BET), total pore volume (V_t), (from the last point of isotherm at a relative pressure of 0.99), micropore volume (V_m) and mean pore size have been calculated using Brunauer–Emmett–Teller (BET) method. The Barrett–Joyner–Halenda (BJH) pore size distribution of the Iranian clinoptilolite sample was calculated based on the adsorption data. As it can be seen in Fig.1b, the majority of pores have the radius size of less than 10 nm with mean pore diameter of 26.47 nm based on BJH method which has good agreement with what has been calculated from BET method [45] (See Table 1).

Fig. 1.

(a) Adsorption/desorption of N₂ gas on Iranian Clinoptilolite sample at T = 77 K. (b) Pore size distribution based on BJH method.

Table 1.

Porosity characterization of Iranian clinoptilolite.

BET method				BJH method
aBET (m2/g)	Vm (cm3/g)	Vt (cm3/g)	Dmean (nm)	aBET (m2/g)	Vm (cm3/g)	Dmean (nm)
14.394	3.3071	0.094766	26.334	15.4	0.094327	26.334

A stainless steel column with 4 cm diameter and 55 cm height was designed and fabricated to regenerate samples using high temperature and vacuum pressure. Regeneration column consists of three heating elements with a heating rate of approximately 20 °C/min and indicators. Three thermocouples provide the required feedback for an on/off temperature controlling system. Vacuum gage is used for indicating the column vacuum pressure. A cooling setup – condensers and cooling water circulator – collects regeneration liquid. Afterward, regenerated zeolite is cooled to ambient temperature in desiccator. Regeneration operation is completed in 0.6 bar vacuum pressure and 300 °C for 50 min.

Static adsorption isotherms (batch)

Water removal from water/ethanol mixture by natural clinoptilolite in batch condition was examined for different initial concentrations of water. The experiments were carried out at ambient temperature (20 °C) and static conditions in a thermo-stated laboratory scale adsorption vessel, with an initial liquid weight of 100 g. The ethanol–water mixture at different concentrations was applied as adsorptive and 60 g zeolite and the contact time of 24 h was selected for experiments. The water concentration in feed was varied between 50 and 363 kg m⁻³ (kg of water in feed to feed volume). The Langmuir and Freundlich isotherm models were used for description of the adsorption process (Eqs. (1), (2)):

$$q_e=\frac{k_l·C_e·q_m}{1+k_l·C_e}$$ (1)

$$q_e=k_f{C}_e^{1/n}$$ (2)

where q_e is the amount of solute adsorbed per unit weight of solid (kmol/kg), C_e is equilibrium concentration of water remaining in solution (kg/m³), q_m is maximum adsorption capacity (kmol/kg) and k_f and k_l are Freundlich and Langmuir constants (m³/kg), respectively. 1/n is a measure of intensity of adsorption. The higher the 1/n value, the more favorable is the adsorption. q_e is calculated from equation as follows (Eq. (3)):

$$q_{e(\text{kmol}/\text{kg})}=\frac{C_{0-}{C}_e}{C_{\mathit{zeo}}}$$ (3)

Because of using Temperature Swing Adsorption process (TSA) for adsorbents regeneration (high temperature and low pressure), it was necessary to find temperature dependent isotherm. An Extended Langmuir isotherm was used to find adsorption dependence with temperature. For this, q_e and C_e values in different temperature between 10 and 70 °C were found and temperature dependent equation was expressed as follows:

$$q_e=\frac{k_1{k}_2\mathit{exp}\left(\frac{k_3}{T}\right)C_e}{1+k_2\mathit{exp}\left(\frac{k_3}{T}\right)C_e}$$ (4)

where k₁ is q_m (kmol/kg), k₂ is k_el (m³/kg) and k₃ is ΔH/R (°K). ΔH and R are enthalpy changes and gas constant, respectively.

Dynamic adsorption (continuous)

An apparatus with packed bed adsorption column was designed for dynamic standard experiments. The schematic of the designed apparatus is shown in Scheme 1. The column was designed based on the Yamamoto’s set-up dimensions for liquid phase adsorption [16]. The retention time in the set-up was determined to be 21.2 min and for this research work this value was assumed as 25 min. According to the maximum flow rate of 14 mL/min, a stainless steel column with 4 cm diameter and 40 cm height was designed and fabricated. Its dimension ensured good flow distribution since the bed internal diameter was at least 10 times as much as the particle size and its length was at least 100 times as much as the particle size [40]. A jacket of cooling water with 4 cm thickness was connected to circulator and surrounded the main column to fulfill the isotherm conditions. Two pressure and temperature sensors were located on both inlet and outlet of the column for monitoring pressure drop and changes in system temperature. A copper coil was used in initial bioethanol container to control the initial bioethanol temperature. Scheme 1 illustrates the experimental setup.

Scheme 1.

Schematic of experimental apparatus used for bioethanol dehydration. (1. Initial Container, 2. Liquid Pump, 3. Adsorption Column, 4. Cooling Circulator, 5. Temperature Sensor, 6. Pressure Sensor, 7. Flowmeter and 8. Final Container.)

Two pressure sensors in bottom and top of the column show the pressure and pressure drop. Using circulator and temperature sensors, initial and final bioethanol temperatures were controlled. The bioethanol concentration in initial container was constant and it was 96%. The final bioethanol concentration leaves the top valve and is shown by the Portable Density Meter of Anton DMA 35. A water flow meter (calibrated for ethanol v/v concentration) was used for adjustment of bioethanol flow rate and it is one of the main experimental parameters. Before carrying out the experiments, the adsorbent samples were treated by thermal regeneration for elimination of water from the adsorbent pores. The procedure was completed by putting samples in furnace for 2 h at 300 °C and then it was cooled to ambient temperature in desiccator. Natural zeolites with the HEU (Heulandite) framework are divided into two distinct classes based on Si/Al ratio. Those with Si/Al of less than 4 are known as heulandite and those with Si/Al greater than 4 are known as clinoptilolite or silica-rich heulandite. The key difference in these materials is those with Si/Al of less than 4 are not thermally stable to calcination above 350 °C [46] and High silica Clinoptilolite is thermally stable to temperatures in excess of 500 °C [47].

The initial tank was filled with 96% (v/v) ethanol–water mixture. The bioethanol entered from the bottom of the column which was packed with a mass of natural zeolite. When the mixture leaved the column, the pressure, temperature and flow rate were controlled. The data collected and concentration were obtained every minute by Density Meter. All of the experiments were organized by RSM to find the optimal operational conditions. This was done by Design-Expert 7 software.

Results and discussion

Static isotherm models

To determine the model to be used to describe the adsorption for an adsorbent–adsorbate isotherm experiments were carried out. Initial concentration was varied from 50 to 363 kg m⁻³ for all the experiments at different temperature. Raman spectroscopy test was used to determine water–ethanol adsorption on natural clinoptilolite. The Raman spectra were collected at the Spectroscopy Laboratory, Atomic and Molecular Group, Physics Department, Tarbiat Modares University by using a Thermo Nicolet Almega dispersive micro-Raman scattering spectrometer. Results showed that the main peak in water Raman spectrum is for stretching O–H bond around 3000–3400 cm⁻¹ and it is obvious in Fig. 2 that there is a strong peak in this area after the treatment of zeolite by mixture of water and ethanol; hence, it could be concluded that only water is adsorbed and the amount of adsorbed ethanol is negligible.

Fig. 2.

Raman spectrum of natural clinoptilolite before and after water adsorption.

By applying the linearization of both Langmuir and Freundlich models, it has been indicated that data could be well described by both models (Figs. S1–S5 in Supplementary information data). Linear form of experimental data for Langmuir and Freundlich isotherms at different temperatures was shown in Supplementary details (Figs. S6–S15). Table 2 shows Langmuir and Freundlich isotherm parameters in different temperatures varied between 283 and 343 K. It is obvious that Langmuir isotherm has a better correlation than Freundlich at low temperatures near ambient. Fortunately, working at low temperature is desirable for industrial adsorption process and based on data presented in Table 2, it is obvious that decreasing in temperature causes increase in final uptake. Hence, Extended Langmuir isotherm was selected as a temperature dependent isotherm equation to describe the temperature behavior in adsorption process that could also be used for any future process simulation or scale-up and design the industrial plant. Eq. (5) shows the Extended Langmuir model that describes the static data. Fig. 3 shows the experimental data and Extended Langmuir model at different temperatures. Furthermore, Fig.4a and b shows the amount of uptake as a function of temperature and initial concentration for both experimental data and Extended Langmuir model, respectively that are well matched together.

$$\mathit{qe}=\frac{0.0076\times 0.0001244\mathit{exp}\left(\frac{1386.5}{T}\right)\mathit{Ce}}{1+0.0001244\mathit{exp}\left(\frac{1386.5}{T}\right)\mathit{Ce}}$$ (5)

Table 2.

Langmuir and Freundlich parameters in different temperatures.

Temperature (K)	Uptake (kmol/kg)	Langmuir			Freundlich
kl (m3/kg)	qm (kmol/kg)	R2	kf (kmol/kg)	1/n	R2
283	0.00746	0.0179	0.00746	0.986	0.00066	0.412	0.973
298	0.00701	0.0156	0.00701	0.986	0.00051	0.442	0.984
313	0.00689	0.0125	0.00689	0.991	0.00041	0.47	0.987
328	0.00662	0.0113	0.00662	0.971	0.00033	0.498	0.979
343	0.00628	0.0106	0.00629	0.974	0.000264	0.529	0.992

Fig. 3.

The Extended Langmuir isotherms in different temperatures.

Fig. 4.

Experimental data (a) and Extended Langmuir isotherms (b) in different temperatures.

Breakthrough curves based on dynamic study (Continuous)

In order to understand the effect of parameters including flow rate, pressure and temperature on the breakthrough curves and finding the optimum operating condition, the variation of each parameter was studied. The breakthrough curves in Fig. 5 show the effects of pressure, temperature and flow rate on breakthrough point. Fig.5a shows that most of the mass transfer and adsorption takes place at the moment that the fluid first comes in contact with the inlet of the bed and by increasing the flow rate, there is a decrease in time required for saturation of adsorbent and breakthrough point shifts to the right side by increasing the flow rate. Increasing pressure from 1 to 5 bar sat defined temperature (T = 293 K) also enhances the saturation time of adsorbent while there is a slight increase in adsorption capacity (Fig.5b). Reducing the temperature changes breakthrough point and increases adsorption, the breakthrough point shifts to the right side and the maximum adsorption is obtained.

Fig. 5.

Breakthrough curves at (a) different flow rates and constant T = 288 K and P = 3 bar, (b) different pressures and constant F = 14 mL/min T = 293 K, and (c) different temperatures and constant F = 10 mL/min and P = 1 bar.

To find the total adsorption capacity (q) from breakthrough curve, the area above the curve divided into the total area (above and under the curve) results in adsorption percentage [16]. Hence, according to Eq. (6), total adsorption capacity could be calculated from feed loading and adsorption percentage.

$$q=\frac{F{\rho }_f}{w}{C}_0{\int }_0^{t_b}\left(1-\frac{C_t}{C_0}\right)\mathit{dt}$$ (6)

The total adsorption capacity q in Fig.5a for flow rates 6, 10 and 14 mL/min (in 3 bar pressure and 288 K) was obtained 1.75, 1.52 and 1.26 mol/kg_zeo, respectively. Also in Fig.5b, for column pressures 1, 3 and 5 bar (14 mL/min and 293 K), q values were calculated 1.5, 1.74 and 1.97 mol/kg_zeo respectively. According to Fig.5c, for temperatures 283, 288 and 293 K and 10 mL/min flow rate and 1 bar pressure, q values obtained were 1.13, 1.31 and 1.44 mol/kg_zeo, respectively.

The response surface methodology

The response surface methodology (RSM) is a collection of statistical and mathematical techniques for obtaining empirical models. Understanding the effect of parameters and optimization is the common advantage of the RSM application. Table 3 shows the coded level for each parameter. For the three parameters, the Box–Behnken Method was used. In this method, each factor or independent variable is placed at one of the three equally spaced values, usually coded as −1, 0 and +1. In this research work, the temperature, pressure and flow rate are independent variables and the final bioethanol concentration is the responses. Table 4 shows the experimental design used in this study.

Table 3.

The three level factors used in the Box–Behnken design.

Coded factors	Corresponding parameters	Coded levels
−1	0	1
Corresponding values
A	Pressure (bar)	1	3	5
B	Flow rate (mL/min)	6	10	14
C	Temperature (°C)	10	15	20

Table 4.

The experimental design used in this research work.

Pattern	A (pressure)	B (flow)	C (temperature)	Final concentration
000	3	10	15	99.3
000	3	10	15	99.2
000	3	10	15	99.3
++0	5	14	15	99.3
+−0	5	6	15	99.7
−0+	1	10	20	98.4
−+0	1	14	15	98.7
0++	3	14	20	99.1
000	3	10	15	99.4
−−0	1	6	15	98.9
+0−	5	10	10	99.9
0−−	3	6	10	99.6
−0−	1	10	10	98.9
0−+	3	6	20	99.0
000	3	10	15	99.3
+0+	5	10	20	99.5
0+−	3	14	10	99.3

The two linear and Quadratic models were used for the concentration modeling in Box–Behnken method. Fig. 6 illustrates the different variables to response diagrams in linear model for concentration response. An analysis of variance for linear model is shown in Table 5. The Model F-value of 36.39 implies the model significant. The values of Prob > F less than 0.05 indicate that the model terms are significant and the parameter has a significant effect on the response. In this case, A (pressure), B (flowrate) and C (temperature) are significant model terms. The values greater than 0.10 indicate that the model terms are not significant. The prediction expression is given as follows:

$$\text{Concentration}=99.43+0.2125\times P-0.025\times F-0.04\times T$$ (7)

Fig. 6.

The different variables to response diagrams in linear model, (a) effects of A and C on response, (b) effects of B and C on response, and (c) effects of B and A on response.

Table 5.

Analysis of concentration variance for ANOVA table by linear model.

Source	Sum of squares	df	Mean square	F value	P-value Prob > F
Model	1.84	3	0.61	36.39	<0.0001
A-pressure	1.45	1	1.45	85.50	<0.0001
B-flow rate	0.080	1	0.080	4.73	0.0486
C-temperature	0.32	1	0.32	18.93	0.0008
Residual	0.22	13	0.017
Lack of fit	0.20	9	0.022	4.44	0.0826
Error	0.020	4	5.0E−3

According to Fig. 6 increasing the pressure causes an increase in maximum final bioethanol concentration. For the low temperatures, the effect of pressure on maximum bioethanol final concentration was intensified. It can be seen that there is a positive relation between the pressure and maximum bioethanol final concentration for a given temperature. Decreasing the temperature causes the maximum bioethanol concentration to increase. Reducing the temperature in liquid affects the adsorbent surface and makes water molecules enter the adsorbent pores. The temperature control in liquid phase adsorption process is more effective and it is simpler than the other two operational parameters and the obtained results show that the temperature has more influence on bioethanol final concentration. The flow rate decrease causes an increase in maximum bioethanol concentration. Reducing the flow rate makes more retention time and hence creates a better contact between the adsorbent and solute.

Fig. 7 shows the relationship between the predicted and actual data line. The R² value of 0.89 indicated that the actual and the predicted data had a relatively good correlation in linear model.

Fig. 7.

The predicted vs. actual data in linear model.

Analysis of variance for quadratic model is shown in Table 6. The Model F-value of 18.11 implies that the model is significant. The values of Prob > F less than 0.05 indicate that the model terms are significant and have a significant effect on the response. In this case, A (temperature), B (flowrate) and C (Temperature) are significant model terms. The values greater than 0.1 indicate that the model terms are not significant. The prediction expression in quadratic model is given as follows (Eq. (8)):

$$\text{Conc}=99.637+0.3875\times P-0.050\times F-0.0750\times T-0.00625\times \mathit{PF}+0.005\times \mathit{PT}+0.005\times \mathit{FT}-0.03125P^2-0.00015F^2-0.001T^2$$ (8)

Table 6.

ANOVA in quadratic model for concentration.

Source	Sum of squares	df	Mean square	F value	P-value Prob > F
Model	1.98	9	0.22	18.11	0.0005
A-pressure	1.45	1	1.45	119.00	<0.0001
B-flow rate	0.080	1	0.080	6.59	0.0372
C-temperature	0.32	1	0.32	26.35	0.0013
AB	0.010	1	0.010	0.82	0.3943
AC	0.010	1	0.010	0.82	0.3943
BC	0.040	1	0.040	3.29	0.1124
A2	0.066	1	0.066	5.42	0.0528
B2	2.632E−3	1	2.632E−3	0.22	0.6557
C2	2.632E−3	1	2.632E−3	0.22	0.6557
Residual	0.085	7	0.012
Lack of fit	0.065	3	0.022	4.33	0.0953
Error	0.020	4	5.0E−3

Fig. 8 shows the independent variables to response diagrams in quadratic model. Fig. S16 illustrates the relationship between the predicted and actual data line. The R² value of 0.92 shows a relatively good correlation between predicted and actual data. The optimization as a point of view of concentration shows that at 10.7 °C liquid temperature, 4.9 bar pressure and 8 mL/min liquid flow rate, the best response was 99.9% bioethanol final concentrations.

Fig. 8.

The independent variables to response diagrams in quadratic model, (a) effects of A and C on response, (b) effects of B and C on response and (c) effects of B and A on response.

Conclusions

In this research work, Iranian natural clinoptilolite was used to dehydrate hydrous ethanol. Results showed that in optimum operating conditions, bioethanol final concentration can reach to 99.9% and above. Static and dynamic studies were done and adsorption isotherms were obtained and experimental data were well described by Extended-Langmuir isotherm. The effects of operating parameters such as temperature, pressure and flow rate were investigated on final ethanol concentration and the adsorption process was optimized. In Box–Behnken analysis, the linear and quadratic models could be successfully applied for description of dynamic process. Results showed that the selected natural clinoptilolite could be used as a favorable adsorbent in bioethanol drying without any pretreatment processes.

Conflict of Interests

The authors have declared no conflict of interests.

Compliance with Ethics Requirements

This article does not contain any studies with human or animal subjects.

Appendix A. Supplementary material

Supplementary data 1

Supplementary material contains Figs. S1–S16.