In Depth Look at Extractive Distillation

1. Distillation for azeotropic mixture

Distillation is the most widely used separation technique in the chemical and petroleum industry. However, not all liquid mixture are amenable to ordinary fractional distillation. When the components of the system have low relative volatilities (1.00 < a < 1.05), separation becomes difficult and expansive because a large number of trays are required and, usually, a high reflux ratio as well. Both equipment and utilities costs increase markedly and the operation can become uneconomical. If the system forms azeotropes, as in a benzene and cyclohexane system, a different problem arises, - the azeotropic composition limit the separation, and for a better separation this azeotrope must be bypassed in some way. At low to moderate pressure, with the assumption of ideal-gas model for the vapor phase, the vapor-liquid phase equilibrium (VLE) of many mixture can be adequately describe by the following Modified Raoult’s Law:

When g_i = 1, the mixture is said to be ideal Equation 1 simplifies to Raoult’s Law. Nonideal mixtures   (g_i ¹ 1) can exhibit either positive (g_i > 1) or negative deviations (g_i < 1) from Raoult’s Law. In many highly nonideal mixtures these deviations become so large that the pressure-composition (P-x, y) and temperature-composition (T-x, y) phase diagrams exhibit a minimum or maximum azeotrope point. In the context of the T-x, y phase diagram, these points are called the minimum boiling azeotrope (where the boiling temperature of the azeotrope is less than that of the pure component) or maximum boiling azeotrope (the boiling temperature of the azeotrope is higher than that of the pure components). About 90% of the known azeotropes are of the minimum variety. At these minimum and maximum boiling azeotrope, the liquid phase and its equilibrium vapor phase have the same composition, i.e.,

x_i = y_i for i = 1, …, c (2)

Two main types of azeotropes exist, i.e. the homogeneous azeotrope, where a single liquid phase is in the equilibrium with a vapor phase; and the heterogeneous azeotropes, where the overall liquid composition which form two liquid phases, is identical to the vapor composition.

Most methods of distilling azeotropes and low relative volatility mixtures rely on the addition of specially chosen chemicals to facilitate the separation. The selection of the separating agent will be discussed later.

The five methods for separating azeotropic mixtures are:

1. Extractive distillation and homogeneous azeotropic distillation where the liquid separating agent is completely miscible.
2. Heterogeneous azeotropic distillation, or more commonly, azeotropic distillation where the liquid separating agent, called the entrainer, forms one or more azeotropes with the other components in the mixture and causes two liquid phases to exist over a wide range of compositions. This immiscibility is the key to making the distillation sequence work.
3. Distillation using ionic salts. The salts dissociates in the liquid mixture and alters the relative volatilities sufficiently that the separation become possible.
4. Pressure-swing distillation where a series of column operating at different pressures are used to separate binary azeotropes which change appreciably in composition over a moderate pressure range or where a separating agent which forms a pressure-sensitive azeotrope is added to separate a pressure-insensitive azeotrope.
5. Reactive distillation where the separating agent reacts preferentially and reversibly with one of the azeotropic constitutes. The reaction product is then distilled from the nonreacting components and the reaction is reversed to recover the initial component.

2. Residue Curve Maps

The most simple form of distillation, called simple distillation, is a process in which a muticomponent liquid mixture is slowly boiled in an open pot and the vapors are continuously removed as they form. At any instant in time the vapor is in equilibrium with the liquid remaining on the still. Because the vapor is always richer in the more volatile components than the liquid, the liquid composition changes continuously with time, becoming more and more concentrated in the least volatile species. A simple distillation residue curve is a graph showing how the composition of the liquid residue curves on the pot changes over time. A residue curve map is a collection of the liquid residue curves originating from different initial compositions. Residue curve maps contain the same information as phase diagrams, but represent this information in a way that is more useful for understanding how to synthesize a distillation sequence to separate a mixture.

All of the residue curves originate at the light (lowest boiling) pure component in a region, move towards the intermediate boiling component, and end at the heavy (highest boiling) pure component in the same region. The lowest temperature nodes are termed as unstable nodes (UN), as all trajectories leave from them; while the highest temperature points in the region are termed stable nodes (SN), as all trajectories ultimately reach them. The point that the trajectories approach from one direction and end in a different direction (as always is the point of intermediate boiling component) are termed saddle point (S). Residue curve that divide the composition space into different distillation regions are called distillation boundaries. A better understanding of the residue curve map may be view in Fig. 1. Notice that the trajectories move from the lowest temperature component towards the highest.

Fig. 1 Residue curve map for a ternary mixture with a distillation boundary running from pure component D to the binary azeotrope C.

Residue curve maps would be of limited usefulness if they could only be generated experimentally. Fortunately that is not the case. Using various references, the simple distillation process can be described by the set of equations:

Research studies have also been done to determine the relationship between the number of nodes (stable and unstable) and saddle points one can have in a legitimately drawn ternary residue plot. The equation is based on topological arguments. One form for this equation is:

4(N₃ - S₃) + 2(N₂ - S₂) + (N₁ - S₁) = 1

Where:
N_i = number of nodes (stable and unstable) involving i species

S_i = number of saddles involving i species

Many different residue curve maps are possible when azeotropes are present. Ternary mixtures containing only one azeotrope may exhibit six possible residue curve maps that differ by the binary pair forming the azeotrope and by whether the azeotrope is minimum or maximum boiling.

Even though the simple distillation process has no practical use as a method for separating mixtures, simple distillation residue curve maps have extremely useful applications, such as:

1. Testing of the consistency of experimental azeotropic data;
2. Predict the order and content of the cuts in batch distillation;
3. In distillation, to check whether the given mixture is separable by distillation, identification of the entrainers / solvents, prediction of attainable product compositions, qualitative prediction of composition profile shape, and synthesis of the corresponding distillation column.

By identifying the limiting separation achievable by distillation, residue curve maps are also useful in synthesizing separation sequences combining distillation with other methods.

3. Homogeneous Azeotropic Distillation

The most general definition of homogeneous azeotropic distillation is the separation of any single liquid-phase mixture containing one or more azeotropes into the desired pure component or azeotropic products by continuous distillation. Thus, in addition to azeotropic mixtures which require the addition of a miscible separating agent in order to be separated, homogeneous azeotropic distillation also includes self-entrained mixtures that can be separated without the addition of a separating agent.

The first step in the synthesis of a homogeneous azeotropic distillation sequence is to determine the separation objective. Sometimes it is desirable to recover all of the constituents in the mixture as pure components other times it is sufficient to recover only some of the pure components as product. In our example, we would like to recover the cyclohexane product at 90% purity and recycle the separating agent back to the initial separating column for further use.

The second step is to sketch the residue curve map for the mixture to be separated. The residue curve map allows one to determine whether the goal can be reached and if so how to reach it, or the goal needs to be redefined.

Distillation boundaries for continuous distillation are approximated by simple distillation boundaries. It is a good approximation for mixtures with nearly simple distillation boundaries. For a separation to be feasible by distillation, the simple distillation boundary should not be crossed, i.e. the distillate and bottom composition should lie in the same distillation region. A more detail calculation method involving the composition will be discuss in the later section.

In the most common situation, a separating agent is added to separate a minimum boiling binary azeotrope into its two constituent pure components by homogeneous azeotropic distillation. Michael F. D. and Jeffrey P. K. presented seven of the most favorable residue curve maps for this task. Of the seven, the map representing extractive distillation is by far the most common and the most important. Its corresponding residue curve and column sequences are shown in Fig. 2 below.

4. Extractive distillation

Extractive distillation is defined as distillation in the present of a miscible, high boiling, relatively nonvolatile component, the solvent, that forms no azeotropes with the other components in the mixture. It is widely used in the chemical and petrochemical industries for separating azeotropic, close-boiling, and others low relative volatility mixture.

Extractive distillation works because the solvent is specially chosen to interact differently with the components of the original mixture, thereby altering their relative volatilities. Because these interactions occur predominantly in the liquid phase , the solvent is continuously added near the top of the extractive distillation column so that an appreciable amount is present in the liquid phase on all of the trays below. The mixture to be separated is added through second feed point further down the column. In the extractive column, the component having the greater volatility, not necessarily the component having the lowest boiling point, is taken overhead as a relatively pure distillate. The other component leaves with the solvent via the column bottoms. The solvent is separated from the remaining components in a second distillation column and then recycled back to the first column.

Fig. 2 Extractive distillation with a heavy solvent
which introduce no new azeotrope for a minimum boiling
azeotrope. In some case, B can come off the top of the first column.

There are several industrial application for homogeneous azeotropic distillation listed in the Encyclopedia of Separation Technology by Michael F. D., Jeffrey P. K.

Extractive distillations can be divided into three categories, each correspond to the different residue curve maps, the minimum boiling azeotropes, maximum boiling azeotropes and the nonazeotrope mixtures. Since our benzene-cyclohexane mixture to be separated is of the second type of mixture, i.e. the minimum boiling azeotrope, we will focus our attention on column sequencing this type of azeotropic separation method in the following section.

As in azeotropic distillation, design of extractive distillation system will also requires significant preliminary work including:

• Choosing the solvent
• Developing or finding necessary data, such as azeotropic condition or residue curve
• Preliminary screening
• Computer simulation
• Small scale testing

For our example, we will consider the first four steps.

5. Solvent screening and selection

     5.1 Solvent criteria

One of the most important steps in developing a successful (economical) extractive distillation sequence is selecting a good solvent. Approaches to the selection of an extractive distillation solvent are discussed by Berg, Ewell et al. ^, and Tassions. In general, selection criteria include the following :

1. Should enhance significantly the natural relative volatility of the key component.
2. Should not require an excessive ratio of solvent to nonsolvent (because of cost of handling in the column and auxiliary equipment.
3. Should remain soluble in the feed components and should not lead to the formation of two phase.
4. Should be easily separable from the bottom product.
5. Should be inexpensive and readily available.
6. Should be stable at the temperature of the distillation and solvent separation.
7. Should be nonreactive with the components in the feed mixture.
8. Should have a low latent heat.
9. Should be noncorrosive and nontoxic.

Naturally no single solvent or solvent mixture satisfy all the criteria, and compromises must be reached.

        5.2 Solvent screening

Perry's handbook serve as a good reference for the solvent selection procedure, which can be thought of as a two-step process, i.e.:

         5.2.1  First level: Broad screening by functional group or chemical family

1. Homologous series: Select candidate solvent from the high boiling homologous series of both light and heavy key components.
2. Robins Chart: Select candidate solvents from groups in the Robbins Chart (part of the chart is shown in Table 3) that tend to give positive (or no) deviations from Raoult's law for the key component desire in the distillate and negative (or no) deviations for the other key.
3. Hydrogen-bonding characteristic: are likely to cause the formation of hydrogen bonds with the key component to be removed in the bottoms, or disruption of hydrogen bonds with the key to be removed in the distillate. Formation and disruption of hydrogen bonds are often associated with strong negative and positive deviations, respectively from Raoult's Law.
4. Polarity characteristic: Select candidate solvents from chemical groups that tend to show higher polarity than one key component or lower polarity than the other key.

         5.2.2  Identification of individual candidate solvents

1. Boiling point characteristic: Select only candidate solvents that boil at least 30-40^oC above the key components to ensure that the solvent is relatively nonvolatile and remains largely in the liquid phase. With this boiling point difference, the solvent should also not form azeotropes with the other components.
2. Selectivity at the infinite dilution: Rank the candidate solvents according to their selectivity at infinite dilution.
3. Experimental measurement of relative volatility: Rank the candidate solvents by the increase in relative volatility caused by the addition of the solvent.

Residue curve maps are of limited usefulness at the preliminary screening stage because there is usually insufficient information available to sketch the them, but they are valuable and should be sketched or calculated as part of the second stage of the solvent selection.

        6   The scenario of our distillation process

        6.1  The azeotropic condition

For our example which deals with the azeotropic mixture formed between benzene and cyclohexane, we have chosen extractive distillation (one of the homogeneous azeotropic distillation methods). The reason of choosing this method is due to the availability of information regarding this separation technique and its tendency to operate more efficiently, i.e. in separating and recycling the separating agent. A brief discussion of the process is given below.

After the mixture exited as the bottom product of the flash unit, it contains mostly our desire product of cyclohexane and also a significant amount of unreacted benzene, which is to be recycled back to the reactor for further conversion. Our main goal now is to further separate the remaining components in the mixture. As cyclohexane and benzene have been encounter most of the remaining composition with the mole % of 44.86 and 54.848 respectively (Table 1), we will consider this to be a binary mixture in our further discussion.

From the process flowsheeting, we would like to operate the distillation column at the pressure of 150 kPa. At this condition, cyclohexane and benzene will have boiling points of 94.34^oC and 93.49^oC respectively (Fig. 3). This is a typical case where conventional distillation would struggle to perform the separation of this type of close boiling mixture. Thus, a special type of distillation technique, i.e. extractive distillation has been chosen in order to purify the desire product, i.e. cyclohexane to our desired purity of 99.3%.

As can be shown from Fig. 3, this binary composition will form a minimum boiling, homogeneous azeotrope at the temperature of 91^oC and the corresponding composition at this point will be 45.5 mole % for cyclohexane and 55.5 mole % for benzene (Fig. 4).

Table 1 Properties of Stream 26 (bottom product of Flash Drum (F13)

          6.2  Solvent selection for benzene-cyclohexane binary mixture

In order to perform a successive extractive distillation, a solvent needs to be chosen to "break" the azeotrope that forms at the operating pressure of the distillation column. Recommended solvent for the benzene-cyclohaxane mixture from the literature^,,, is aniline, with a solvent to feed ratio (S/F) of 4, which will shift the azeotropic point toward the corner of the high-boiling component cyclohexane, and the equilibrium curve of the original components fall below the diagonal (Fig. 5).

As was stated in the above section, the primary goal of solvent selection is to identify a group of feasible solvents to perform a good separation. The desired product, i.e. cyclohexane should have a purity of above 99% to meet the market standard. Aniline was the first solvent that had been put to the simulator to be tried out, as it is of the same homologous group as benzene. As can be shown from the result in Table 2, this solvent will produce the desire production rate of 150 with the solvent flow rate of 3500, i.e. a S/F ratio of 9.85. However, the product purity can only reach 70.08% and this does not meet our product specification. As a result, other solvent may have to be researched to perform the desire separation.

We will have to perform the solvent selection criteria as stated in the preceding section. At the column pressure of 150 kPa, cyclohexane and benzene boil at 94.34^oC and 93.49^oC respectively and form a minimum-boiling azeotrope at 91^oC. The natural volatility of the system is benzene > cyclohexane, so the favored solvents most likely will be those that cause the benzene to be recovered in the distillate. However, in order to get a better quality of product, we would like to recover cyclohexane as the distillate rather than from the bottom stream. Thus, solvent to be chosen should give positive deviations from Raoult's law for cyclohexane and negative (or zero) deviation for benzene.

Table 2 Result from computer simulation using aniline as solvent.

Table 3 Part of Robbins Chart taken from Perry's Chemical Handbook.

Turning to the Robbins Chart (Table 3), we note that solvents that may cause the positive deviation for cyclohexane (Class 12) and negative (or zero) to benzene (Class 11) came from the groups of 4, 7, 8 and 9, which consist of polyol, amine and ether. We further consider the solubility, the hydrogen bonding effect, and also the homologous characteristic of the solvent with the corresponding components in the feed mixture. As few candidate solvents that had been put to the computer simulation, included phenol (homologous to benzene), 1,2-benzenediol (homologous to benzene, with -OH group that will produce hydrogen bonding), 1,3-butanediol (with -OH group that will produce hydrogen bonding), and also 1,2-propanediol (same characteristic as with 1,3-butanediol). 1,2-propanediol (often known as propylene glycol), seem to give the most promising results compared to the other solvents. This result may be caused from the high solubility of benzene in this solvent and the hydrogen bonding that were formed between the two constituents. Simulation result of this solvent can be view in Table 4.

        6.3  Construction of the residue curve

Equation 3 and 4 were used to sketch the corresponding residue curve for the three species. From the above information, we know that these species have boiling points at 94.34 (cyclohexane), 93.49 (benzene) and 200.35^oC (propylene glycol) at the pressure of 150 kPa, and an azeotrope that boils at 91^oC between the two more volatile species. As were shown from Fig. 6 and Fig. 7 there were no new azeotropes formed between the solvent 1,2-propanediol respectively with the another two component in the feed.

Table 3 Result from computer simulation using 1,2-propanediol as solvent.

We then start to sketch our residue curve map by sketching the triangular diagram in Fig. 8, and placing the arrows pointing from the lower to higher temperatures around the edge. The corner points for benzene and cyclohexane are single species point, and both are unstable nodes - all residue curves leave. The corner point for propylene glycol is a single species point which is a stable node - all residue curve enter. All three are nodes; none are saddles, thus;

N₁ = 3 and S₁ = 0

We then further assume that there will be no ternary azeotrope been form among the three constituents, i.e.,

N₃ = S₃ = 0

The remaining steps here require the identification of the only binary azeotrope that form between benzene and cyclohexane, to be either a node or a saddle point. From equation 4:

4(0-0) + 2(N₂ - S₂) + (3-0) = 1
2(N₂ - S₂) = -2
N₂ - S₂ = -1

Thus, the only way we can satisfy the above equation is letting N₂ = 0 and S₂ = 1, i.e. the binary azeotrope is a saddle point, which directs the trajectories in another direction.

        6.4 Column operation

The extractive distillation unit of this cyclohexane production plant consists of two distillation columns (Fig. 10), which we can easily classify as direct sequence columns. The first column acts as an extractive column where the solvent is introduced at the second stage of the column, so that it will be present throughout the column and exits with the bottoms. As were stated above, the solvent alters the natural volatility of the binary mixture by forming hydrogen bonds with benzene and allowing it to be recovered as the bottom product.

The bottom product of the first column will then fed to the second column, i.e. the solvent recovery column, to undergo the normal distillation to separate both the components for further usage, i.e. benzene being recycled to the reactor for further conversion while solvent to the first column for reuse. The main operation parameter of the distillation unit is shown in Table 4.

Table 4 Distillation unit summary

Fig. 10 Extractive distillation unit for cyclohexane production plant

References:

J. M. Smith, H. C. Van Ness, M. M. Abbott, 1996, Introduction to Chemical Engineering Thermodynamics, McGraw Hill, p.449.

Douglas M. Ruthven (Ed.), 1997, Encyclopedia of Separation Technology, Vol. 1, John Wiley, [Distillation, Azeotropic and Extractive by Michael F. D., Jeffrey P. K.]

L. T. Biegler, E. I. Grossmann, Arthur W. Westerberg, 1997, Systematic Methods of Chemical Process Design, Prentice Hall, [Separating Azeotropic Mixture]

Abid.

Green, Perry, 1997, Perry’s Chemical Engineers’ Handbook, 7^th Edition, McGraw Hill,. [Section 13-78 Enhanced Distillation by J. D. Seader, Jeffrey J. S., Scott D. B.]

Philip A. Schweitzer, 1988, Handbook of Separation Technique for Chemical Engineers, 2^nd Edition, McGraw Hill, [Continuous Distillation: Separation of Multicomponent Mixture by Edward C. R., John E. M.]

James R. F., Distillation.

Robert F. G., 1972, Extractive and Azeotropic Distillation, [Rapid screening of Extractive Distillation Solvent. Predictive and Experimental Techniques by Tassios P. D.]

Green, Perry, 1997, Perry’s Chemical Engineers’ Handbook, 7^th Edition, McGraw Hill,. [Section 13-79 Enhanced Distillation by J. D. Seader, Jeffrey J. S., Scott D. B.]

James R. Fair, Distillation [Special Distillation]

Kith-Othmer, 1965, Encyclopedia of Chemical Technology, 2^nd Edition, Vol. 6, John Wiley, [Cyclohexane by James J.Kirk]

Lee L. E., 1999, Plant Design Project Cyclohexane Production, UTM, {Chapter 7 Discussion].

**This article was graciously submitted by Dominic Foo Chwan Yee for publication at www.cheresources.com.   The author can be reached for questions/comments at cyfoo98@pd.jaring.my