Extractive distillation:Advances in conceptual design,solvent selection,and separation strategies☆

Shirui Sun,Liping Lü,Ao Yang,Shun'an Wei,Weifeng Shen,*

1 School of Chemistry and Chemical Engineering,Chongqing University,Chongqing 400044,China

2 School of Chemistry and Chemical Engineering,Yangtze Normal University,Chongqing,Fuling 408100,China

3 Collaborative Innovation Center for Green Development in Wuling Moutain Area,Yangtze Normal University,China

Keywords:Extractive distillation Separation strategies Solvent selection Conceptual design

A B S T R A C T Extractive distillation(ED)is one of the most promising approaches for the separation of the azeotropic or closeboiling mixtures in the chemical industry.The purpose of this paper is to provide a broad overview of the recent development of key aspects in the ED process involving conceptual design,solvent selection,and separation strategies.To obtain the minimum entrainer feed flow rate and reflux ratio for the ED process,the conceptual design of azeotropic mixture separation based on a topological analysis via thermodynamic feasibility insights involving residue curve maps,univolatility lines,and unidistribution curves is presented.The method is applicable to arbitrary multicomponent mixtures and allows direct screening of design alternatives.The determination of a suitable solvent is one of the key steps to ensure an effective and economical ED process.Candidate entrainers can be obtained from heuristics or literature studies while computer aided molecular design(CAMD)has superiority in efficiency and reliability.To achieve optimized extractive distillation systems,a brief review of evaluation method for both entrainer design and selection through CAMD is presented.Extractive distillation can be operated either in continuous extractive distillation(CED)or batch extractive distillation(BED),and both modes have been well-studied depending on the advantages in flexibility and low capital costs.To improve the energy efficiency,several configurations and technological alternatives can be used for both CED and BED depending on strategies and main azeotropic feeds.The challenge and chance of the further ED development involving screening the best potential solvents and exploring the energy-intensive separation strategies are discussed aiming at promoting the industrial application of this environmentally friendly separation technique.

1.Introduction

Distillation is one of the most widely used separation technology of liquid mixtures in various chemical and pharmaceutical industries[1].In most separation system,molecular interactions often can cause the predominant nonideality in liquid phase and tend to form maximum or minimum azeotrope.However,only adopting conventional distillation hardly achieve the separation of azeotropic mixtures that is the most challenging tasks in chemical industries.To overcome this problem, special distillation including pressure-swing distillation [2-6],extractive distillation [7-11] and azeotropic distillation [12-14] has been proposed and are well-described in open literatures.Extractive distillation(ED)is one of the most attractive approaches for separating the azeotropic, close-boiling or low relative volatility mixtures due to the superiority of easy operability and controllability.To break the distillation boundary and achieve the separation of the binary azeotrope,an entrainer which alters the relative volatilities of the different components in the mixture are required for ED system. Compared with azeotropic distillation,ED has benefit in that there is no new azeotrope formed and the original component with the greatest volatility separates out as the top product.A typical process of the extractive distillation features an extractive column and a regeneration column.Generally, entrainer is fed to the upper part of the extractive column and withdrawn as a bottom liquid product with one of the components. The bottom liquid then is taken to a second regeneration column,in which the other product is removed from the distillate and the entrainer recycles back to the extractive column. Some of main studied topics of ED including column with all possible configurations;process operation polices and strategy;process design,synthesis,and optimization;determination of separation sequencing;entrainer design and selection;feasibility studies[15-17].In this review,we intend to provide a broad overview of the recent development of key aspects in the ED process involving separation strategies,solvent selection,and conceptual design.

2.Conceptual Design Based on Thermodynamics Analysis

2.1.Residue curve map

The residue curve map (RCM)describes liquid composition trajectories remaining in a simple distillation with respect to time, and it plays a significant role in the analysis of distillation feasibility and the conceptual design of conventional multicomponent separation processes. The concept of RCM was first defined by Schreinemakers[18] and developed by the pioneering work of Doherty and Perkins[19]. The study of the topological properties of RCM based on the theory of differential equations is summarized by Kiva et al. [20] and Hilmen et al.[22].The simple RCM can be defined by the following differential equation:

where h is a dimensionless time describing the relative loss of the liquid in the still-pot; xiand yiare the mole fractions of the component i in the liquid phase and in vapor phase, respectively. A simplest residue curve map is illustrated in Fig. 1a. The trajectories of the various residue curves have an obvious directional feature which is represented by arrows. Normally, RCM must follow the two rules:first, it cannot intersect each other; second, it always moves along the boiling temperature surface in the direction of increasing temperature. For nonreactive RCM, there are three singular points: unstable node, stable node, and saddle point (Fig. 1b), relying on the sign of the eigenvalues related to the residue curve equation. As shown in Fig.1b,all the residue curves extend from an unstable node,terminate in a stable node, and both approach and depart a saddle. The residue curves move away from the unstable node to the stable node with increasing temperatures.Some residue curves move away from a saddle point with decreasing temperatures and others with increasing temperatures.

2.2.Unidistribution and relative volatility

The distribution coefficient and relative volatility are well-known characteristics of the vapor-liquid equilibrium(VLE).The distribution coefficient Kithat featured the distribution of component i between the vapor and liquid phases in equilibrium is defined by Eq.(2).

The vapor is enriched with component i if Ki＞1,and is impoverished with component i if Ki＜1. The unidistribution curve is defined as that the distribution coefficient of the given component i equals unity Ki=1.The ratio of the distribution coefficient of components i and j is defined as the relative volatility seeing in Eq.(3).

The relative volatility features the ability of component i to transfer(evaporate)into the vapor phase compared to the ability of component j. Component i is more volatile than component j if αij＞1, and less volatile if αij＜1.The isovolatility lines represent the characteristics the relative volatility, and are the key of a general feasibility criterion to infer which component is an attainable product and what the related column configuration is.Then the system of univolatility lines where the isovolatility curve αij=1 was proposed[18],which separate the composition triangle into regions with an obvious order of volatility of the i and j components.For the mixture of ideal or nearly ideal,the relative volatilities are nearly constant in the whole composition space.However,for nonideal,especially azeotropic mixtures where the composition dependence can be complex,the situation seems to be different.

The function of unidistribution and univolatility line diagrams is used to describe the VLE diagrams and represent the topologic feature of the simple phase transformation trajectories. As shown in Fig. 2,unidistribution lines are named using the component where they started(e.g.,a in class 1.0-1a),while univolatility lines are never started from pure component, and they are named by two letters (e.g., a in class 1.0-1a). The point of pure component i may (or may not) give rise to a unidistribution line.The existence of a binary azeotrope or a ternary azeotrope gives rise to two or three unidistribution,respectively.It is obvious that the volatility order of the components cannot coincide with their boiling point temperatures in the whole composition space.The diagram of unidistribution lines is used as a main tool for analysis of tangential azeotrope and biazeotrope [20]. Unidistribution lines divide the concentration simplex into two regions such that Ki＞1 in one of them and Ki＜1 in the other,where component i behaves as a low boiling component in the first region and as a high boiling component in the second[21,23].The target of unidistribution lines is to determine the regions of higher concentration for a given component.The purpose of unidistribution and univolatility line diagrams is to sketch volatility order regions and assess the feasible structures involving two aspects:giving possible products and offering information related to possible limitations of entrainer feeds.

Fig.1.Features of(a)1.0-1a residue curve map and(b)singular points.

Fig.2.Unidistribution and univolatility line diagrams for the most probable classes of ternary mixtures according to Reshetov's statistics.

2.3.Recent advances

The thermodynamic topological analysis is based on the relationship between the VLE of a mixture and the behavior of open evaporation residue curves for ternary mixtures.Only via a RCM to determine the ease of separation is difficult because relative volatility information is not presented.Based on knowledge of the thermodynamic properties of RCM and the information of the univolatility and unidistribution curves location,the general feasibility criterion for extractive distillation under infinite reflux ratio is proposed by Rodriguez Donis et al.[24-27].The number of feasible VLE diagram structures is limited by topological and thermo-dynamical constraints.Serafimov et al.[28-30]presented a complete classification of these feasible structures for ternary mixtures, and this classification is improved by Serafimov in 1996 [31].Serafimov's classification results in the 26 classes of ternary VLE diagrams are presented in Fig. 3. The completion and extension of thermodynamic insight on Serafimov's classes: 1.0-1a, 1.0-1b, 1.0-2(one azeotrope),2.0-1,2.0-2a,2.0-2b and 2.0-2c(with two azeotropes)to other processes was well-studied.

The conceptual design of ED is a challenge because of the great many of process parameters(e.g.,the reflux ratio,the tray numbers of the feed and entrainer,and the ratio of entrainer and feed).Thus,a systematic method for the conceptual design of ED has been paid more and more attention because of the complexity of“trial and error”in the process of finding optimal splits and operation parameters.The conceptual design of ED depends on the knowledge of RCM topology and the calculation of univolatility curves to analyze the volatility order of the mixture.

Fig.3.Azeotropic ternary mixture:most probable topological systems selected from Serafimov's classes[28].

The application of ternary maps and RCM analysis to study the feasible for separating azeotropic mixture in ED sequence is explored by Doherty and Caldarola[32].Knapp and Doherty[33]proposed a systemic mathematical model to investigate the feasible of their proposed theory method by specified conditions(i.e.,feed composition,desired product purity,and the ratio of entrainer and feed)to define a given split.Brüggemann and Marquardt[34]proposed that the minimum entrainer flow rate and the reflux ratio for the given feed and split could be determined via a shortcut method based on a non-linear analysis of the ED.What needs to be pointed out is that this method is suitable for ternary mixtures with binary feed having a minimum-azeotrope and high-boiling entrainers.

Lelkes et al.[35]is the first one to propose a method involving the calculation of feasible profiles for different column sections and the determination of the most important operational parameters(e.g.,number of stages,flow rate of entrainer,and reflux ratio),to assessment the feasibility of extractive distillation in a batch rectifier.Based on the method presented by Lelkes et al.,Rodríguez Donis et al.[24,26,27]further developed the method to study the feasibility of the extractive distillation in the batch operation mode by isovolatility lines.Shen et al.[16,36]extended thermodynamic insights on batch extractive distillation to continuous operation with both high- and low-boiling entrainers based on the knowledge of RCM.Their results indicate that the feasible product and feasible ranges of the operating parameters reflux ratio and the ratio entrainer and feed flow rate for continuous and batch operation can assessed via the occurrence and location of univolatility curves.A robust and reliable method based on the analysis of the mode infinitely sharp splits(ISS mode)to search and identify the feasible splits for ED is presented by Petlyuka et al.[37].

3.Solvent Selection

The effectiveness of an ED process depends on the selection of extractive solvent which can change the relative volatilities of the original of the mixture.The selection of solvent which makes the design of ED more complex is a challenging issue.However,the selection of suitable solvent which is indispensable in the separation of the azeotropic or close-boiling mixtures is a very demanding task.The selection criteria of entrainer are related to the important aspects:first,thermodynamics including selectivity and boiling point;second,process operation involving the optimal ratio of entrainer and feed flow rate,low corrosion and price,high thermal stability,and low molar volume in ED process[38].The foundation of several methodologies for solvent selection is one-way interaction parameters, indicating that interactions of the components to be separated with the solvent.

For ED,the selectivity or relative volatility at infinite dilution always be adopted as the standard to evaluating the suitability of solvent[10,39,40].The selectivity Sij∞and the relative volatility αij∞at infinite dilution are respectively given by the following Eqs.(4)and(5):

where γi∞is activity coefficient that related to N(the total number of mole of solvent)at a given time t.Pi0and Pj0represent the pure component vapor pressure of i and j,respectively.Aioand Ajoare the original peak area of the solute of component i and j, respectively; Aiand Ajrepresent the peak area at intervals of the time t of component i and j,respectively.It is obvious that the selectivity Sij∞is coincide with relative volatility αij∞in the process of evaluating the suitability of solvent.To simplify, it is better to choose αij∞as a criteria. Furthermore, the data of αij∞is more accurate than γi∞because it not involving N and t which may bring some extra errors in the experiment. Thus if the peak area A of the solute at different time is known,αij∞is able to be calculated according to Eq.(5)in which only the peak area A is required.Another proposed the criteria to assess the suitability of an entrainer is the capacity[41],which is determined by:

where j represents the solute.The smaller the value of the activity coefficient γi∞,the stronger the interactions are between component j and the entrainer,which results in a larger capacity C∞j,entrainer.

The search of candidate solvents for a given separation process is a major issue in ED process design, and it is cumbersome to choose the optimal solvent from thousands of different substances for a given system just through experiments.Many study exhibit examples of the efforts to provide methodologies for solvent selection. A screening procedure for solvents which is fast and evaluates solvents ahead of equipment design based on isovolatility,equivolatility and binary VLE diagrams has been proposed by Luyben and Chien[1].Kossack et al.[42]developed another entrainer selection approach that is a framework linked to the application of an optimization approach for the process design based on solvent properties and a rectification body method.However,there are some approach to preselect the possible solvents by calculation, such as Pierotti-Deal-Derr method [43,44],Parachor method[45,46],Weimer-Prausnitz method[42,47],and the computer-aided molecular design(CAMD).Among them,CAMD based on the UNIFAC group contribution is both the most important and the recent development [48]. More importantly, it has replaced other methods in the process of screening solvents since the quantitative calculation of salt effect is complex,and the thermodynamic method is incomplete.

CAMD is essential the inverse of property prediction by group contribution.It is proposed to find a combination of structural groups and satisfying the property specification based on a set of desirable properties.Fig.4 shows an approach of screening for solvents in extractive distillation considering quantitative risk analysis.To test and match the molecule with specified objective molecular properties, a lot of CAMD approaches based on the concept of molecular generation were being developed in the past[49-52].The methodologies for the screening of solution may be well-established based on either software tools[53,54]or a mathematic programming mixed integer nonlinear programming with a specific solution technique (e.g., the support grid methodology and the branch-and-bound algorithm).Kossack and coworkers[38,55]proposed a process design frameworks for ED starting from the solvent selection stage until reaching the column optimization stage, which integrates the CAMD application to the design of ED. Lek-utaiwan et al. [56] proposed a practical methodology for the design and optimization of ED with the help of CAMD approach, and proved that the experimental verification and the property parameter determination were to be necessary to achieve a successful and reliable design.Shen et al.[10]proposed a systematic approach to design of two-column extractive distillation for the separation of azeotropes with heavy entrainers, and they use the five important properties as a standard to evaluate entrainer.The Fuzzy logic and develop membership functions are adopted to calculate attribute values of selected properties. They also presented an indicator “total score”defined as the summation of weighted attribute values to evaluate entrainer.

In most cases, more than one solvent can be used. Therefore, a screening is necessary because only one alternative must be chosen for the specified question.At this point,factors such as price and source should be considered.

4.Separation Strategies

ED is an efficient and widely used approach for the separation of azeotropic and close-boiling mixtures in petroleum and chemical industries.A typical continuous ED process is described in Fig.5a,which includes an extractive distillation column and a solvent regeneration column.

The azeotropic(A-B)and makeup entrainer(E)mixture with recycle entrainer fed to the extractive column.The extractive column differs from the convention distillation column in that an extractive section is added to the usual stripping and rectifying sections.The distillate of extractive column is light product A,and the bottom is mixture B-E then fed to the entrainer regeneration column.Product B is distillate from regeneration column and the bottom entrainer after cooled to a temperature preset in the extractive column recycles the extractive column.ED process needs a makeup to compensate entrainer losses along with products.Fig.5b shows the typical continuous extractive distillation(CED)process with light entrainer.

ED can be operated either in batch or in continuous mode.For batch extractive distillation(BED),the entrainer must be fed continuously at some tray of the column or into the still during the whole operation.Thus,BED is a semi batch process,that is to say,as the main feed FABis loaded initially in the boiler,while the entrainer is fed continuously at a higher tray.Finishing the extractive task,the entrainer regeneration task is operated at the same column. The column configurations for BED processes with intermediate entrainer are descried in Fig.6.

Fig.4.Approach of screening for solvents in ED considering quantitative risk analysis.

The application of BED via a pilot-plant column and a computer model for comparison was first proposed by Yatim et al. [57]. They present and simulated a four-step operating procedure including dynamic material and energy balances for BED of acetone and methanol mixtures. However, the liquid holdup was assumed constant and vapor holdup was negligible.

Düssel and Stichlmair[58]investigated a systematical presentation of BED for separating binary azeotropic mixtures,in which the batch extractive distillation as a hybrid process(distillation and absorption).The first rigorous optimization case study of BED was presented by Low and S?rensen[59],and they explored the wide range of degrees of freedom available(e.g.,feed placement and stream configuration).Lang et al.[60]investigated the application of a light entrainer for batch extractive distillation.BED in a rectifier and in a stripper column has been summarized by Stéger et al.[61]and Varga et al.[62],respectively.

Compared with continuous distillation,batch distillation has the advantages in simplicity of operation,flexibility,and lower capital cost,resulting in that it has received increasing attention in the last few years.However,some attempts were carried out to improve the batch configurations and make batch distillation more efficient.

Fig.5.Flowsheet of typical extractive distillation with(a)heavy entrainer and(b)light entrainer.

A shortcut method based on a general differential model was extended for the heterogeneous extractive distillation to calculate the rectifying,extractive,and stripping composition profiles of each column configuration.The general configuration for the heterogeneous distillation column is shown in Fig.7.The feasibility of heterogeneous extractive distillation process in a continuous column considering several feed point strategies for the entrainer recycle stream and for the main azeotropic feed have been investigated by Rodriguez-Donis et al.[63].In the homogeneous extractive distillation process, the entrainer can be fed at the bottom,top, or intermediate position.However,as shown in the Fig.8,there exist seven different configurations for the continuous heterogeneous extractive distillation[34].The position of the entrainer recycles stream returns to the heteroazeotropic column is the key to feasibility of heterogeneous extractive distillation process.Thus,a critical analysis of heterogeneous extractive distillation process with different entrainer recycle strategies has been proposed by Doherty and co-workers[64,65].To better design the heterogeneous extractive distillation, the application of a shortcut design method based on the rectification body method (RBM) is investigated by Marquardt and co-workers [34,66,67]. Yi et al. [68] investigated an energy-efficient heterogeneous extractive distillation system for cyclohexane-cyclohexene system using ethylene glycol as heavy entrainer,and their results show that the proposed heterogeneous extractive distillation system with a process-to-process heat exchanger has the predominant benefit of the lowest total annual cost(TAC).

ED is widely used in a number of processes, but it is still energy intensive [69]. To improve the energy efficiency on convention ED,an innovative solution ED is using advanced process intensification and integration techniques(e.g.,heat-integrated distillation columns,dividing-wall columns(DWCs)and heat-pump-assisted distillation).Heat integration is a useful method to significantly reduce the energy consumption compared to conventional process with no integration.The method for the recovery of aromatics from pyrolysis gasoline with NMP as entrainer is studied by Abushwireb et al.[70],and the results indicates that the heat-integrated extractive distillation system was the best candidate.You et al.[71]proposed a double-effect heat integration and mechanical heat pump technique for the ED process of the acetonemethanol binary azeotropic system with water as entrainer,and compared those processes from the economical view and environmental aspect. Gu et al. [72] proposed a three double-effect heatintegration processes for the separating tetrahydrofuran-water binary azeotropic mixture with dimethyl sulfoxide as an entrainer,which are employed under atmospheric and a reduced pressure for the first time to further improve the energy efficiency.

Fig.6.Column configurations for batch extractive distillation processes with intermediate entrainer feeding(a)Direct split in a rectifying column and(b)Indirect split in a stripping column.

Recently,the use of dividing wall columns has drawn considerable attention. The feasible of novel distillation alternatives based on extractive and azeotropic DWC for enhanced bioethanol dehydration is investigated by Kiss and Suszwalak [73]. A constrained stochastic multi-objective optimization technique for the extractive dividingwall column(EDWC)is proposed by Bravo-Bravo et al.[74],and their results show the influence of the main variables on the complex ED process.Luo et al.[69]investigated a novel heat-pump-assisted extractive distillation process taking place in DWCs for the separation of the azeotropic behavior of the ethanol-water mixture, and proved that the proposed process can make significant decrease in TAC when considering the requirements for a compressor and use of electricity.Lately,Yang[75]et al.carried out a study on the process intensification for ED and attempted to investigate an energy-saving EDWC process with heat integration for the separation of heterogeneous system methanol/toluene/water with multi-azeotrope,involving thermodynamic feasible insights via RCM to determine separation constraints,global optimization based on a proposed complete process optimization model,and a dynamic control through Aspen Dynamics simulator to better maintain product purities.The proposed process is illustrated in Fig.9.

Fig.7.The general configuration for the heterogeneous distillation column.

Fig.8.Configurations for the heterogeneous distillation column considering all possibilities for both the entrainer recycle and the main azeotropic feed.

Fig.9.An energy-saving extractive dividing-wall column(EDWC)process with heat integration for separating heterogeneous mixtures methanol/toluene/water.

5.Conclusions

A comprehensive description of the current research developments for extractive distillation(ED)from separation strategies,solvent selection,and conceptual design is reviewed in this paper.Because of the high energy consumption of the distillation technology,several separation strategies and technological alternatives considering the design sustainability of ED process for both continuous and batch have been investigate. As the solvent selection is the key of ED process, more attention should be paid on the selection of the potential solvent.The computer-aided molecular design (CAMD), as the most competitive method,has been largely used for the screening of liquid solvents,and the application on the solid salt is very few.Thus,with the development of CAMD,more studies would be concerned with the ionic liquids which are a kind of special solvent in ED. More importantly, CAMD should reduce the inaccuracy caused by the lack of appropriate interaction parameters for the property prediction.The target of conceptual design in ED is the key parameters(e.g.,reflux ratio,boilup ratio,and the ratio of entrainer and feed flowrate),relating well-known design tools:residue curve map analysis,univolatility and unidistribution curves,and some topological character analysis, and conceptual design can give a clear guidance on the more efficient design of the ED process. The studies performed to date show a diverse field of current popular ED design methodology research.Aiming at promoting the industrial application of this environmentally friendly separation technique,It is imperative that the further ED development involving screening the best potential solvents and exploring the energy-intensive separation strategies can be used to replace the traditional technologies.