Thesis.dvi

Preface

This thesis is submitted in partial fulfillment of the requirements for the degree doktor ingeniør
at the Norwegian University of Science and Technology (NTNU). The majority of the work has
been carried out during the period from 1994 to 1999 at the Department of Chemical Engineering,
NTNU with Professor Sigurd Skogestad as main advisor. Since 1997, Dr. Valeri N. Kiva from
Karpov Institute of Physical Chemistry in Moscow has been a co-advisor of the thesis.

I spent the academic year 1995-96 in the group of Professor Michael F. Doherty and Professor
Michael F. Malone at the Department of Chemical Engineering at the University of Massachusetts
in Amherst. The following academic year 1996-97, Dr. Valeri N. Kiva was a visiting researcher
in our group at NTNU. I had one year maternity leave in 1999-2000.

The work is a part of an industrial project on solvent recovery in fine and specialty chemical in-
dustries in cooperation with Borregaard, Nycomed Imaging, Norsk Hydro and Dyno Industries.
The project has been organized through the Fine and Specialty Chemicals and PROSMAT pro-
grams of the Research Council of Norway. The efforts of the industrial participants is greatly
appreciated. My visits to the plants and production facilities were very fortunate. Special thanks
to Thor Alvik at Borregaard and Joachim Schmidt at Nycomed Imaging for their contributions.

I would like to thank my advisor Professor Sigurd Skogestad for his invaluable advice and con-
tributions to this work. His insights and high standards have definitely helped to shape this work.
It is a pleasure to have an advisor being so joyful in his work and creative in his research.

I am in particular indebted to Dr. Valeri N. Kiva for his significant impact on the course of this
work. His insight to the field of azeotropic distillation and to practical distillation operation has
been most valuable. Many achievements of his research are included in this thesis. Thanks also
for being a man of strong character. Our discussions have been most amusing and enlightening.

My thanks goes to Professor Michael F. Doherty and Professor Michael F. Malone for a great and
inspiring year in their group at the University of Massachusetts in Amherst, introducing me to the
many aspects of azeotropic distillation and nonideal thermodynamics.

Thanks to former and present colleagues at NTNU for making an enjoyable working environment.
Stewart Clark, Stathis, Hilde, Lucie and Jan are thanked for proofreading parts of the thesis.
Ingvild for being my confidant. I acknowledge the assistance of Dr. Bella Alukhanova in drawing
the figures for the survey on azeotropic phase equilibrium diagrams (Chapter 3) in this thesis.

Finally, I want to express gratitude to my parents, sister and grandparents for their unconditional
support and encouragement. My warmest thanks goes to my husband Erik for putting up with me
during what must have been demanding times. Without his immense support and patience this
work would not have resulted. Our beloved daughter Tone was born in June 1999 and the thesis
completed in November 2000, beginning a new era in our lives.

Financial support from the Research Council of Norway, Borregaard, Nycomed Imaging, Norsk
Hydro, Dyno Industries, PROST and the Fulbright Foundation is gratefully acknowledged.

iii

Contents

Abstract

Preface

Introduction

1.1

Motivation and industrial relevance . . . . . . . . . . . . . . . . . . . . . . . . .

1.2

Thesis overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.3

The reader’s guide to the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . .

Principles and Tools for Analysis

Methods of Azeotropic Mixtures Separation: General Considerations

2.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.2

Azeotropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.2.1

What is azeotropy? . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.2.2

Vapor-liquid phase equilibrium, nonideality and azeotropy . . . . . . . .

2.2.3

Nonideality and separation by distillation . . . . . . . . . . . . . . . . .

2.2.4

Physical phenomena leading to nonideality and azeotropy . . . . . . . .

2.3

General separation processes . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.3.1

Membrane-distillation hybrids . . . . . . . . . . . . . . . . . . . . . . .

2.3.2

Pressure-swing distillation . . . . . . . . . . . . . . . . . . . . . . . . .

2.3.3

Entrainer-addition distillation methods . . . . . . . . . . . . . . . . . . .

2.3.4

Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.4

Entrainer-addition distillation methods . . . . . . . . . . . . . . . . . . . . . . .

2.4.1

Feasibility analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.4.2

Feasibility analysis for azeotropic mixtures . . . . . . . . . . . . . . . .

2.4.3

Homogeneous azeotropic distillation . . . . . . . . . . . . . . . . . . . .

2.4.4

Heteroazeotropic distillation . . . . . . . . . . . . . . . . . . . . . . . .

2.4.5

Extractive distillation . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.4.6

Entrainer selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.5

Summary and conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Azeotropic Phase Equilibrium Diagrams: A Survey

3.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.2

Thermodynamic basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.3

Isotherm maps of ternary mixtures . . . . . . . . . . . . . . . . . . . . . . . . .

3.4

Open evaporation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.4.1

Residue curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.4.2

Properties of residue curves’ singular points . . . . . . . . . . . . . . . .

3.4.3

Rule of azeotropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.4.4

Structure of residue curve maps . . . . . . . . . . . . . . . . . . . . . .

3.4.5

Separatrices of residue curves and flexure of the boiling temperature surface 65

3.4.6

Structure of distillate curve maps

. . . . . . . . . . . . . . . . . . . . .

3.5

Other simple equilibrium phase transformations . . . . . . . . . . . . . . . . . .

3.5.1

Open condensation . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.5.2

Repeated equilibrium phase mapping and distillation lines . . . . . . . .

3.5.3

Relationship between residue curves, distillation lines and condensation
curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.6

Heterogeneous mixtures

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.6.1

Simple phase transformations in heterogeneous mixtures . . . . . . . . .

3.6.2

Examples of simple equilibrium phase transformation maps . . . . . . .

3.7

Ternary VLE diagrams. Classification, occurrence and structure determination . .

3.7.1

Classifications of ternary VLE diagrams . . . . . . . . . . . . . . . . . .

3.7.2

Completeness of classifications

. . . . . . . . . . . . . . . . . . . . . .

3.7.3

Occurrence of predicted structures . . . . . . . . . . . . . . . . . . . . .

3.7.4

Determination of the structure . . . . . . . . . . . . . . . . . . . . . . .

3.8

Unidistribution and univolatility lines

. . . . . . . . . . . . . . . . . . . . . . .

3.8.1

Distribution coefficient and relative volatility . . . . . . . . . . . . . . .

3.8.2

Univolatility and unidistribution line diagrams

. . . . . . . . . . . . . .

3.8.3

Structure of unidistribution and univolatility line diagrams . . . . . . . . 103

3.9

Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

Definitions of Terms

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

On the Topology of Ternary VLE Diagrams: Elementary Cells

123

4.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123

4.2

Classification of ternary VLE diagrams

. . . . . . . . . . . . . . . . . . . . . . 124

4.3

Elementary cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126

4.4

Occurrence in nature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

4.5

Use of elementary cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131

4.5.1

The album

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131

4.5.2

Multiple steady states . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132

4.6

Pseudo-component subsystem . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

4.7

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134

Studies on Batch Distillation Operation

137

Holdup Effect in Residue Curve Map Prediction of Batch Distillation Products

139

5.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

5.2

Model and data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

5.3

Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

5.4

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150

Appendix 5A

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151

Batch Distillation Columns and Operating Modes: Comparison

153

6.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

6.2

Batch distillation column configurations . . . . . . . . . . . . . . . . . . . . . . 154

6.2.1

Regular and inverted columns . . . . . . . . . . . . . . . . . . . . . . . 154

6.2.2

Middle vessel and multivessel columns . . . . . . . . . . . . . . . . . . 154

6.3

Operating policies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157

6.3.1

Open operation (conventional) . . . . . . . . . . . . . . . . . . . . . . . 158

6.3.2

Closed operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158

6.3.3

Cyclic operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159

6.3.4

Semicontinuous operation . . . . . . . . . . . . . . . . . . . . . . . . . 159

6.4

Comparison of cyclic and closed middle vessel columns . . . . . . . . . . . . . . 159

6.4.1

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161

6.4.2

Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162

6.4.3

Comparison with integrated continuous distillation . . . . . . . . . . . . 163

6.4.4

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164

Closed Multivessel Batch Distillation of Ternary Azeotropic Mixtures

169

7.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169

7.2

Columns and operating modes for ternary separation . . . . . . . . . . . . . . . 171

7.3

VLE diagrams and steady state composition profiles . . . . . . . . . . . . . . . . 173

7.4

Mixtures studied

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174

7.5

Mathematical model

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176

7.6

Simulation results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177

7.6.1

Results on ternary zeotropic mixture (Cell I)

. . . . . . . . . . . . . . . 179

7.6.2

Results on ternary azeotropic mixture (Cell II)

. . . . . . . . . . . . . . 183

7.6.3

Results on ternary azeotropic mixture with one distillation boundary . . . 194

7.7

Discussion and suggestions for further research . . . . . . . . . . . . . . . . . . 200

7.8

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202

Appendix 7A

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203

Extractive Batch Distillation: Column Schemes and Operation

207

8.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207

8.2

Mixtures studied

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211

8.3

Mathematical model and assumptions . . . . . . . . . . . . . . . . . . . . . . . 213

8.4

Extractive batch distillation in the conventional column . . . . . . . . . . . . . . 214

8.4.1

Process description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214

8.4.2

Simulation results for System 1 (ethanol - water - ethylene glycol) . . . . 215

8.4.3

Simulation results for System 2 (acetone - methanol - water) . . . . . . . 219

8.4.4

Summary: Extractive batch distillation . . . . . . . . . . . . . . . . . . . 226

8.5

Extractive batch distillation in the middle vessel column

. . . . . . . . . . . . . 226

8.5.1

Process description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226

8.5.2

Simulation results with simultaneous separation (Scheme I)

. . . . . . . 229

8.5.3

Summary: Extractive middle vessel batch distillation . . . . . . . . . . . 233

8.6

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233

Appendix 8A

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234

Control Strategies for Extractive Middle Vessel Batch Distillation

237

9.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237

9.2

Important considerations for control . . . . . . . . . . . . . . . . . . . . . . . . 239

9.3

Two-point composition control . . . . . . . . . . . . . . . . . . . . . . . . . . . 240

9.4

Simulation model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242

9.5

Closed loop simulation results . . . . . . . . . . . . . . . . . . . . . . . . . . . 242

Chapter 1. Introduction

All liquid mixtures have forces of intermolecular attraction. That is why they form liquids and
not gases. The molecular interactions when two or more components are mixed may cause the
mixture to form certain “inseparable” compositions where the vapor and liquid compositions at
equilibrium are equal within a given pressure and temperature range. These specific mixture
compositions are called azeotropes. Azeotropy is not a rare phenomena.

Azeotropy plays an important role in vapor-liquid equilibrium separation processes such as dis-
tillation, similar to eutetics and peretetics in liquid-solid systems such as crystallization. To be
able to develop separation processes for azeotropic mixtures, there is a need for insight into the
fundamental phenomena of azeotropic phase equilibria. The vapor-liquid envelope of the equi-
librium temperature surfaces defines the feasible operating region in which any real distillation
process must operate. The existence of azeotropes complicates the structure of this operating
envelope, and the resulting distillation behavior of multicomponent azeotropic mixtures may be
very complex. The main interest behind azeotropic phase equilibrium diagrams is to reveal all
the physiochemical restrictions imposed on the separation process by the nature of the mixture
in question. Diagram analysis provides important insight for the optimal choice of separation
method and the design and synthesis of azeotropic distillation separation systems. Thus, there
is an incentive to look more carefully into the field of vapor-liquid equilibrium diagrams and the
so-called thermodynamic topological analysis of azeotropic mixtures.

Tools for analysis: graphical methods

Nonideal and azeotropic phase equilibrium diagrams like residue curve maps are sometimes nick-
named the McCabe-Thiele of azeotropic distillation (Kunesh et al., 1995), indicating its im-
portance as a graphical design tool in distillation separation of azeotropic mixtures. The term
analysis-driven synthesis is used by Westerberg and Wahnschafft (1996) to emphasize that the
real work in synthesis of nonideal separation sequences is to establish enough information to
begin the synthesis process. Such pre-synthesis analysis involves determination of the nonideal
behavior of the mixture (finding azeotropes, discovering liquid-liquid behavior, pressure depen-
dencies, distillation behavior and so forth). Graphical representations of the vapor-liquid equilib-
rium with material balance lines, operating lines and distillation composition trajectories drawn
into the diagram are commonly used for such analysis.

The power of graphical representation is it gives insight and understanding. Human beings have
the ability to identify problems and see patterns, and extract information from these. On a graph
one can see the effects of changes clearly and often sources of a particular problem becomes
obvious.

Using azeotropic phase equilibrium diagrams (like residue curve maps), we can immediately
identify infeasible specifications and alternative distillation sequences can be proposed without
trial and error. The shapes of the distillation lines (or residue curves) are very useful in the design
of real finite reflux distillation columns since they give a qualitative picture of the composition
profiles in actual staged distillation columns (Wasylkiewicz et al., 1999a).

In retrofit operations, graphical VLE diagrams can help to understand the behavior of the azeotropic
system when attempting to change operating conditions, and can be used to ensure that it is

1.1 Motivation and industrial relevance

thermodynamically possible to move from the initial state to the final state (Wasylkiewicz et
al., 1999a). By analyzing the residue curve map topology, troubleshooting process difficul-
ties, reasons for controllability and purity problems can be quickly identified (Wasylkiewicz et
al., 1999a).

Solvent recovery in fine- and specialty chemical industries

As waste regulations are becoming more stringent, the economical benefits of having recovery
systems that allow for recycling and reuse of the solvents within the process are significant. Re-
covery and reuse of organic solvents is generally practiced because of increased solvent costs
and potential solvent shortages and because disposal often results in violation of air-, water- or
land-pollution regulations.

The work presented in this paper is connected to an industrial project on solvent recovery in four
Norwegian fine- and specialty chemical plants. The production facilities under consideration pro-
cess high-value products at moderate to low volumes. The process feed stocks often vary both in
size and quality from one day to another, which favors flexible processing equipment. The exist-
ing equipment is mainly conventional batch distillation columns. The solvent recovery typically
involves purification of low-boiling organic solvents that form minimum-boiling homoazeotropes
with water. Azeotropy and the highly nonideal behavior during distillation complicates the re-
covery of the solvents. As a contradiction to this, high purities above azeotrope composition
are usually required in order to reuse the waste solvents in the bulk synthesis operations. Often
solvent mixtures that cannot be fully recovered end up as toxic waste crossing the plant boundary.

“Designing equipment for such [azeotropic] separations is a real challenge to the
distillation engineer. Often, sufficient data are not available for complete calcula-
tion so that the process must be developed in the laboratory, or, at least, assump-
tions checked. Combinations of continuous and batch distillations, decanting, mul-
tistage extraction, and chemical treatment may often be combined in a successful
process.”

Drew as quoted by Schweitzer (1997)

The thesis focus is on various batch distillation configurations for separating azeotropic mixtures.
Interest in batch processes has increased with the growing importance of specialty chemicals,
characterized by high-value, low-capacity, short-term production and strongly nonideal liquid
mixtures. These high value-added products often require high purity and this makes the separation
step an important part of the process. The use of solvents is unavoidable to many processes in the
fine- and special chemical industries.

“An understanding of azeotropes is desirable for two reasons. First, they often occur
in distillation and make a given separation impossible by ordinary distillation. Sec-
ondly, they may be utilized to separate mixtures not ordinarily separable by straight
distillation.”

Ewell et al. (1944)

This statement clearly sets the perspective in which the work presented in this thesis has been
conducted. Dealing with solvent recovery distillation systems this inevitable includes azeotropic

Chapter 1. Introduction

mixtures. Further, some of the established and very powerful separation methods for solvent mix-
tures, even those solvent mixtures that do not form azeotropes, utilize the formation of azeotropes
to increase the effectiveness of the distillation separation.

1.2

Thesis overview

The thesis contents is twofold as the title indicates. First, general aspects of azeotropic mixtures
separation and tools for the analysis of their thermodynamic behavior during distillation are con-
sidered. Second, extensions to batch distillation operation are made, which include simulation
studies of multivessel and extractive batch distillation with emphasis on operation and control.

The thesis is organized in three parts: Part I (Chapter 2 - 4) comprises theory on separation meth-
ods for azeotropic mixtures and structural properties of ternary phase equilibrium diagrams, Part
II (Chapter 5 - 9) involves studies on azeotropic batch distillation and Part III (Chapter 10) gives
practical implications for an industrial project connected to the thesis and overall conclusions.
The thesis chapter-by-chapter:

Part I

Principles and Tools for Analysis

Chapter 2 gives an introduction to the phenomenon of azeotropy and the various methods by
which azeotropic mixtures can be separated. The possibilities to separate azeotropic mixtures by
entrainer-addition distillation, and in particular extractive distillation, are considered.

Chapter 3 comprises a survey of vapor-liquid equilibrium (VLE) diagrams which reveal the basic
thermodynamic phenomena of nonideal and azeotropic mixtures. The survey is a result of a
cooperation performed in parallel with more specific studies on azeotropic batch distillation. The
main contributor of the survey is Valeri N. Kiva.

Chapter 4 presents the novel concept of elementary topological cells of ternary VLE diagrams
and discuss the occurrence of the thermodynamic feasible structures in nature. Some uses of the
elementary cells in azeotropic distillation presynthesis analysis are included.

Part II

Studies on Batch Distillation Operation

Chapter 5 includes a study on the effect of holdup in azeotropic batch distillation product predic-
tion based on residue curve map analysis.

Chapter 6 gives an introduction and overview of various batch distillation columns and modes of
operation, along with a comparative study on two alternative operating policies for the multivessel
batch distillation column. This chapter only considers ideal (constant relative volatility) mixtures.

Chapter 7 considers the direct application of VLE diagram tools in closed multivessel batch
distillation analysis. The chapter includes dynamic simulation of the cyclic batch column and the
closed middle vessel column for the separation of ternary azeotropic mixtures. An indirect level
control strategy is implemented that eliminates the need for pre-calculated vessel holdups. VLE

Chapter 2. Methods of Azeotropic Mixtures Separation: General Considerations

Separation of a liquid mixture by distillation is dependent on the fact that even when a liquid is
partially vaporized, the vapor and liquid compositions differ. The vapor phase becomes enriched
in the more volatile components and is depleted in the less volatile components with respect to
its equilibrium liquid phase. By segregating the phases and repeating the partial vaporization, it
is often possible to achieve the desired degree of separation.

An azeotrope cannot be separated by ordinary distillation since no enrichment of the vapor phase
occurs at this point. Therefore, in most cases, azeotropic mixtures require special methods to
facilitate their separation. Such methods utilize a mass separating agent other than energy that
causes or enhances a selective mass transfer of the azeotrope-forming components. This agent
might be a membrane material for pervaporation or an entrainer for extractive distillation. Extrac-
tive and heteroazeotropic distillation are the most common methods, and are described in more
detail in a separate section in this chapter.

2.2

Azeotropy

2.2.1

What is azeotropy?

The term azeotrope means “nonboiling by any means” (Greek: a - non, zeo - boil, tropos -
way/mean), and denotes a mixture of two or more components where the equilibrium vapor and
liquid compositions are equal at a given pressure and temperature. More specifically, the vapor
has the same composition as the liquid and the mixture boils at a temperature other than that of
the pure components’ boiling points. Azeotropes have sometimes been mistaken for single com-
ponents because they boil at a constant temperature. However, for an azeotrope a variation in
pressure changes not only the boiling temperature, but also the composition of the mixture, and
this easily distinguishes it from a pure component.

The term azeotropy was introduced by Wade and Merriman in 1911 to designate mixtures charac-
terized by a minimum or a maximum in the vapor pressure under isothermal conditions, or, equiv-
alently, with an extremal point in the boiling temperature at constant pressure (Swietoslawski,
1963; Malesinkski, 1965). Since then the term has been used for liquid systems forming one or
several azeotropes. The mixture whose composition corresponds to an extremal point is called
an azeotrope. If at the equilibrium temperature the liquid mixture is homogeneous, the azeotrope
is a homoazeotrope. If the vapor phase coexists with two liquid phases, it is a heteroazeotrope.
Systems which do not form azeotropes are called zeotropic (Swietoslawski, 1963). A brief history
of the first observations of azeotropy is given in, for example, Gmehling et al. (1994).

The defining conditions of an azeotropic mixture and the physical phenomena leading to nonide-
ality and azeotropy are questions that are addressed in more detail in the next sections.

2.2 Azeotropy

2.2.2

Vapor-liquid phase equilibrium, nonideality and azeotropy

At low to moderate pressures and temperatures away from the critical point, the vapor-liquid
phase equilibrium for a multicomponent mixture may be expressed as:

;

)

sat

);

:::;

(2.1)

where

and

are the vapor and liquid compositions of component

, respectively,

and

are

the system pressure and temperature,

is the activity coefficient of component

in the liquid

phase, and

sat

is the saturated vapor pressure of component

. The activity coefficient

a measure of the nonideality of a mixture and changes both with temperature and composition.
When

, the mixture is said to be ideal and Equation (8.2) simplifies to Raoult’s law:

sat

);

:::;

(2.2)

Nonideal mixtures exhibit positive (

) or negative (

) deviations from Raoult’s law.

If these deviations become so large that the vapor pressure exhibits an extremal point at constant
temperature, or, equivalently, an extremal point in the boiling temperature at constant pressure,
the mixture is azeotropic. At azeotropic points, the liquid phase and its equilibrium vapor phase
have the same composition

, and the condensation and boiling temperature curves are tan-

gential with zero slope (see Figure 2.1a). If the positive deviations are sufficiently large (

azeo

Vapor

azeo

Liquid

Liquid - Liquid

(a)

(b)

Figure 2.1: Azeotropes may be homogeneous or heterogeneous. Binary phase diagram with min-
ima in the condensation and boiling temperatures at constant pressure (positive deviation): (a)
minimum-boiling homoazeotrope; (b) heteroazeotrope.

typically), phase splitting may occur and form a heteroazeotrope where the vapor phase is in equi-
librium with two liquid phases (see Figure 2.1b). In the heteroazeotropic point the overall liquid
composition

is equal to the vapor composition, and the vapor and liquid temperature sur-

faces are tangential with zero slope, but the three coexisting phases have distinct compositions. In
ternary and multicomponent systems saddle azeotropes may occur, which Swietoslawski (1963)
call positive-negative azeotropes. The first reported ternary saddle azeotrope was the acetone,

Chapter 2. Methods of Azeotropic Mixtures Separation: General Considerations

chloroform and methanol mixture shown in Figure 2.2, which was examined by Ewell and Welch
(1945). Saddle azeotropic mixtures exhibit a hyperbolic point in the temperature isobar which is

61.3 oC

Chloroform

56.4 oC

Temperature C

55.5 oC

53.6 oC

65.5 oC

57.6 oC

64.5 oC

Methanol

Acetone

Figure 2.2: Complex azeotropic behavior of a ternary mixture. Boiling temperature surface as a
function of liquid composition for the mixture of acetone, chloroform and methanol forming three
binary azeotropes and one ternary saddle azeotrope at atmospheric pressure.

neither a minimum nor a maximum in either boiling temperature or vapor pressure (see Figure
2.2, saddle point at 57.6

C). The corresponding residue curve diagram is shown in Figure 2.3.

57.6 oC

55.5 oC

65.5 oC

53.6 oC

64.5 oC

Methanol

61.3 oC

Acetone

56.4 oC

Chloroform

Figure 2.3: Residue curve map for the mixture in Figure 2.2. The residue curves’ singular points
are indicated by

(stable node),

(unstable node) and

(saddle), and boundaries by bold lines.

2.2 Azeotropy

A mixture is said to be tangentially azeotropic if the boiling temperature isobar exhibits an ex-
tremal point located at the pure component edges, i.e., at zero concentration for one or sev-
eral of the components in the mixture (Swietoslawski, 1963). A mixture is said to be tangen-
tially zeotropic if the boiling temperature at constant pressure exhibits a nearly extreme point
at zero concentration for one or several of the components. This phenomenon is also called
a tangent pinch (Doherty and Knapp, 1993). Such mixtures are of great practical importance
(Serafimov, 1987). Examples of common tangentially zeotropic mixtures are acetic acid and wa-
ter and acetone and water. The binary phase diagram for the acetone and water mixture is given
in Figure 2.4. The mixture exhibits a tangent pinch at the acetone edge.

2.2.3

Nonideality and separation by distillation

One measure of the degree of enrichment, or, the ease of separation, is the relative volatility
between pair of the components, derived from Equation (8.2):

sat

(2.3)

The relative volatility of most mixtures changes with temperature, pressure and composition. The
more

deviates from unity, the easier it is to separate component

from component

. At an

azeotropic point, the relative volatility of the azeotrope-forming components equals one (

)

and it is impossible to further enrich the vapor. Thus azeotropes can never be separated into pure
components by ordinary distillation. However, heteroazeotropes may be separated into its two
liquid phases by other methods such as decantation. Similarly, any mixture where the relative
volatilities are close to unity is difficult to separate by ordinary distillation since little enrichment
occurs with each partial vaporization step. Typically, conventional distillation becomes uneco-
nomical when

0:95

1:05

(there are exceptions), since a high reflux ratio and a large

number of theoretical equilibrium stages are required (Van Winkle, 1967). The special meth-
ods used to facilitate the separation of azeotropic mixtures may also be applied to close-boiling
zeotropic mixtures.

Note that ternary and multicomponent mixtures may have a locus in the composition space where
the relative volatility between a pair of the components is equal to unity without forming any
azeotropes. The presence of such univolatility lines does not hinder separation of the mixture by
ordinary distillation. It only means that the relative volatility order of the components changes
within the composition space and this is common for nonideal mixtures. Therefore, the terms
“light”, “intermediate” and “heavy” that are often used to indicate the relative volatility order
of the pure components in ternary mixtures have limited meaning for nonideal and azeotropic
mixtures.

Chapter 2. Methods of Azeotropic Mixtures Separation: General Considerations

y-x diagram for ACETONE / WATER

LIQUID MOLEFRAC ACETONE

0.2

0.4

0.6

0.8

0.2

0.4

0.6

0.8

VAPOR MOLEFRAC ACETONE

Liquid molefraction of acetone

Vapor molefraction of acetone

(a)

T-xy diagram for ACETONE / WATER

MOLEFRAC ACETONE

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

100

105

Temperature C

Temperature C

Molefraction of acetone

(b)

Figure 2.4: Tangent pinch is a common phenomenon. Binary phase diagram for acetone and
water mixture at atmospheric pressure: (a) vapor and liquid equilibrium compositions; (b) con-
densation and boiling temperatures as a function of composition.

2.2.4

Physical phenomena leading to nonideality and azeotropy

Why are azeotropes formed? Azeotropy is merely a characteristic of the highly nonlinear phase
equilibria of mixtures with strong molecular interactions. Azeotropes are formed due to differ-
ences in intermolecular forces of attraction among the mixture components (hydrogen bounding

2.2 Azeotropy

and others). The particular deviation from ideality is determined by the balance between the phys-
iochemical forces between identical and different components. Briefly, we have three groups:

1. Positive deviation from Raoult’s law: The components “dislike” each other. The attraction

between identical molecules (A-A and B-B) is stronger than between different molecules
(A-B). This may cause the formation of a minimum-boiling azeotrope and heterogeneity.

2. Negative deviation from Raoult’s law: The components “like” each other. The attraction

between different molecules (A-B) is the strongest. This may cause the formation of a
maximum-boiling azeotrope.

3. Ideal mixture obeys Raoult’s law: The components have similar physiochemical properties.

The intermolecular forces between identical and different molecules (A-A, B-B and A-B)
are equal.

For simplicity, the above explanation is given for binary mixtures. The main difference for ternary
and multicomponent mixtures is that an azeotropic point is not necessarily an absolute extreme
(minimum or maximum point) of the boiling temperature at isobaric condition, but it may be a
local extreme (saddle).

The tendency of a mixture to form an azeotrope depends on two factors: (i) the difference in the
pure component boiling points, and (ii) the degree of nonideality (Horsley, 1973; King, 1980).
The closer the boiling points of the pure components and the less ideal mixture, the greater the
likelihood of an azeotrope. A heuristic rule is given in Perry and Chilton (1973) that azeotropes
occur infrequently between compounds whose boiling points differ by more than about 30

C. An

important exception to this rule is heteroazeotropic mixtures where the components may have a
large difference in the boiling points of the pure components, and still exhibit strong nonideality
and immiscibility regions. In heterogeneous mixtures the different components may even repel
each other (Group 1 in the above list). This is why only minimum-boiling heteroazeotropes occur
in nature.

Swietoslawski (1963) emphasizes that most mixtures of organic compounds form nonideal sys-
tems. The presence of some specific groups, particularly polar groups (oxygen, nitrogen, chlo-
rine and fluorine), often results in the formation of azeotropes. Roughly half of the 18 800
binary systems collected in Gmehling et al. (1994) are azeotropic. More than 90 % of the
known azeotropes show positive azeotropy, that is, minimum-boiling azeotropes in the binary
case (Lecat, 1949; Rousseau, 1987; Gmehling et al., 1994). Further, more than 80 % of these are
homoazeotropic (Gmehling et al., 1994).

Ideal or nearly ideal behavior is encountered in mixtures formed by closely related chemical
substances such as heptane and hexane, or butyl alcohol and hexyl alcohol. However, there are
exceptions to this heuristic rule as well. If there is a point in the composition space in which
the vapor pressures of the two components are equal at a given temperature (in the region of
the normal boiling-points of the pure components), this may correspond to an azeotropic point
which is called a Bancroft point, illustrated in Figure 2.5. This is referred to as Bancroft’s rule
(Malesinkski, 1965). The intermolecular forces in the mixture may be similar, but the mixture still

2.3 General separation processes

separation is most commonly done by heteroazeotropic distillation using cyclohexane or benzene
as entrainer, but applications of distillation coupled with membranes or adsorption are also seen
(Fair, 1987). There is also the possibility of using liquid extraction coupled with a simple flash
concentration step.

In the following we outline two promising methods that do not depend on the introduction of an
additional liquid component to the system, that is, membrane-distillation hybrids and pressure-
swing distillation, before we focus on the more utilized entrainer-addition distillation methods.
The section is summed up in a collective discussion of the advantages and disadvantages of these
alternative separation processes.

2.3.1

Membrane-distillation hybrids

The separation of liquid and gas mixtures with membranes as separating agents is an emerging
separation operation. Industrial applications were greatly accelerated in the 1980s (Seader and
Henley, 1998). The membrane itself acts as the mass separating agent, preferentially absorbing
and diffusing one of the azeotrope-forming components. The feed mixture is partially separated
by means of a semipermeable barrier (the membrane) into a retentate, that part of the feed that
does not pass through the membrane (i.e., is retained), and, a permeate, that part of the feed that
does pass through the membrane (Seader and Henley, 1998).

Pervaporation

The most commonly used membrane technology for liquid azeotropic mixtures is pervaporation
with low pressure on the permeate side of the membrane so as to evaporate the permeate (Pettersen
and Lien, 1995; Seader and Henley, 1998). A composite membrane is used that is selective for
one of the azeotrope constituents. Vapor permeation is a similar technology, where the feed en-
ters as saturated vapor. This is specially used in combination with distillation processes, where
saturated vapor streams are readily available (Pettersen and Lien, 1995). One important limitation
of vapor permeation, however, (which does not exist for pervaporation) is that the vapor perme-
ation fluxes are very sensitive to superheating. Even small degrees of superheating can result in
a significant flux decrease (Rautenbach and Vier, 1995). Pervaporation is reported to account for
3.6 % of the total membrane separation applications in chemical and pharmaceutical production
(Knauf et al., 1998). Pervaporation is considered to be expensive both in terms of investment and
processing costs (at high throughput) as discussed later. Because of its high selectivity, however,
pervaporation is of interest in cases where conventional separation processes either fail or result
in a high specific energy consumption or high investment costs. The most important category of
separation tasks for which pervaporation is promising, is in fact mixtures with a homoazeotrope
or close-boiling characteristics (Rautenbach and Albrecht, 1989).

One often distinguishes between two different types of pervaporation-distillation hybrids: (a)
where the task of the pervaporation (or vapor permeation) stage is limited to azeotrope separation
only, and (b) where the pervaporation stage (membrane unit) is designed for azeotrope separation
and for achieving the product specifications. Figure 2.6 illustrates typical configurations for these

Chapter 2. Methods of Azeotropic Mixtures Separation: General Considerations

two processes. The azeotropic mixture withdrawn from the top of the columns (

) is supplied

pure 1

Retentate

pure 2

Permeate

pure 1

pure 2

Retentate

Permeate

(a)

(b)

Figure 2.6: Membrane-distillation hybrids. Separation of a binary homoazeotropic mixture: (a)
pervaporation unit between two distillation columns; (b) distillation column augmented with a
vapor permeation unit where the retentate has sufficient purity to be a final product.

to the membrane modules. There are many other membrane-distillation hybrid configurations.
The key to an efficient and economic membrane separation process is the membrane. It must
have good permeability (high mass-transfer flux), high selectivity, stability, freedom from fouling
(defect-free), and a long life (two or more years) (Seader and Henley, 1998). The membranes
used to separate organic solvent and water mixtures are mostly organic polymers, but also ce-
ramic inorganic membranes are seen (Li and Xu, 1998). Polyvinyl alcohol membranes provide a
high driving force just in the ethanol-water azeotrope composition region, and have a high selec-
tivity for water (Guerreri, 1992). Pervaporation membrane materials suitable for various mixture
systems can be found in the “Membrane Handbook” (Flemming and Slater, 1992). Pervapora-
tion is applicable to separations over the entire composition range, but the phase change requires
considerable more energy than other pressure driven membrane techniques like reverse osmosis
(Goldblatt and Gooding, 1986).

The main industrial applications of pervaporation separation processes include (Rautenbach and
Albrecht, 1989; Seader and Henley, 1998):

removal of water from organic solvents (e.g., dehydration of alcohols, ketones and esters);

removal of organics from water (e.g., removal of volatile organic compounds (VOCs),
toluene and trichloroethylene from wastewater); and

separation of organic-organic azeotropes (e.g., benzene-cyclohexane) and isomers.

2.3 General separation processes

An illustration of an industrial process for dehydration of ethanol using membrane pervapora-
tion in combination with a continuous distillation column is given in Figure 2.7. By integrating

4 mole %

Feed

Retentate

60 mole %

EtOH

99.5 mole %

EtOH

Permeate 40-50 mole %

Pure water

500 ppm EtOH

EtOH

Figure 2.7: Industrial ethanol dehydration using two pervaporation modules and one distillation
column (Goldblatt and Gooding, 1986).

the two separation techniques, the separation task is switched among the technologies in such a
way that each operates in the region of the composition space where it is most effective (Pressly
and Ng, 1998). Goldblatt and Gooding (1986) performed experiments and simulations to evalu-
ate the pervaporation-distillation hybrid process in Figure 2.7. Their study indicates a reduction
of production costs by 50 % over conventional heteroazeotropic distillation for dehydration of
ethanol using benzene as entrainer. They concluded that high selectivity and moderate flux mem-
branes are necessary for the system to be economically attractive compared to the conventional
heteroazeotropic distillation process. Guerreri (1992) reports about 25 % energy savings for this
hybrid process compared to traditional heteroazeotropic distillation process.

Pervaporation is best applied when the feed solution is diluted in the main permeant (Seader
and Henley, 1998). Permeate fluxes decrease significantly with decreasing feed concentrations.
There is a trade-off between the permeate flux and selectivity. Membrane modules can be ar-
ranged in series or in parallel cascades in order to increase the obtainable degree of separation.
A number of membrane stages may be needed, with a small amount of permeant produced per
stage and re-heating of the retentate between stages (Seader and Henley, 1998). The example of
ethanol dehydration has an azeotrope concentration that favors the membrane pervaporation pro-
cess, since only a single-stage pervaporation with a moderate specific membrane area is needed
for the azeotrope separation. A question that still remains to be answered is the performance of
such membrane separation systems for other binary homoazeotropes that have less concentrated
azeotrope compositions.

Pervaporation-distillation hybrids are commercially used for the dehydration of isopropanol too.
Rautenbach and Vier (1995) considered a combination of batch distillation and vapor perme-
ation for removal of isopropanol from water. The constant distillate composition policy was

Chapter 2. Methods of Azeotropic Mixtures Separation: General Considerations

implemented, thus the distillate product flowrate decreases during the operation. Therefore, the
membrane area of the subsequent vapor permeation unit is determined by the distillate product
flowrate at the beginning of the production period. Rautenbach and Vier (1995) found this hybrid
pervaporation-batch distillation process for separating isopropanol and water superior to extrac-
tive distillation using ethylene glycol as the entrainer. However, no attempt was made to compare
the process with the conventional process of heteroazeotropic distillation-decanter hybrid.

2.3.2

Pressure-swing distillation

Pressure changes can have a large effect on the vapor-liquid equilibrium compositions of azeotropic
mixtures and thereby affect the possibilities to separate the mixture by ordinary distillation. By
increasing or decreasing the operating pressure in individual columns we can move distillation
boundaries in the composition space or even make azeotropes appear or disappear (or transform
into heteroazeotropes). For some mixtures, a simple change in pressure can result in a significant
change in the azeotrope composition and enable a complete separation by pressure-swing distil-
lation illustrated in Figure 2.8. A binary homoazeotropic mixture is introduced as feed

to the

pure B

(a)

(b)

pure A

Temperature

= F

High pressure

Low pressure

Figure 2.8: Pressure-swing distillation: (a) temperature-composition diagram for a minimum-
boiling binary azeotrope that is sensitive to changes in pressure; (b) distillation sequence.

low-pressure column. The bottom product from this

column is relatively pure A, whereas the

overhead is an azeotrope with

. This azeotrope is fed to the high-pressure column, which pro-

duces relatively pure B

in the bottom and an azeotrope with composition

in the overhead.

This azeotrope is recycled into the feed of the low-pressure column. The smaller the change
in azeotropic composition with pressure, the larger is the recycle in Figure 2.8b. According to
Smith (1995), a change in azeotropic composition of at least 5 percent with a change in pres-
sure is usually required. Examples of industrial mixtures that can be separated by pressure-swing
distillation are given by Schweitzer (1979). For example, the tetrahydrofuran-water azeotrope

2.3 General separation processes

may be separated by using two columns operated at

= 1 atm and

= 8 atm. Our exam-

ple of the ethanol-water azeotrope is not considered to be sufficiently pressure sensitive for the
pressure-swing distillation process to be competitive. However, it is interesting to mention that a
hybrid system of pressure-swing adsorption (PSA) and distillation is reported to be a competitive
alternative to the conventional methods (Humphrey and Seibert, 1992).

Although changing the vapor-liquid equilibrium (VLE) properties of an azeotropic mixture by
purely physical means (pressure or temperature changes) is an attractive and definite possibility,
it is generally not an option and such processes are often uneconomical (Van Winkle, 1967).

2.3.3

Entrainer-addition distillation methods

Physiochemical changes of the VLE behavior of an azeotropic mixture by the addition of an extra-
neous liquid component offers a number of possibilities. We name the mixture to be separated as
the original mixture, and the added component that facilitates the separation the entrainer. For the
purpose of ease of visualization, we limit our considerations to binary homoazeotropic original
mixtures and one-component entrainers. The same separation techniques apply to multicompo-
nent mixtures where the key components form characteristic mixtures falling into the categories
discussed, and the entrainer may be a mixture of components. We distinguish between three dif-
ferent conventional entrainer-addition based distillation methods depending on the properties and
role of the entrainer and the organization (scheme) of the process:

Homogeneous azeotropic distillation (ordinary distillation of homoazeotropic mixtures):

The entrainer is completely miscible with the components of the original mixture. It
may form homoazeotropes with the original mixture components.

The distillation is carried out in a conventional single-feed column.

Heteroazeotropic distillation (decanter-distillation hybrids that involve heteroazeotropes):

The entrainer forms a heteroazeotrope with at least one of the original mixture com-
ponents.

The distillation is carried out in a combined column and decanter system.

Extractive distillation:

The entrainer has a boiling-point that is substantially higher than the original mixture
components and is selective to one of the components.

The distillation is carried out in a two-feed column where the entrainer is introduced
above the original mixture feed point.

The main part of the entrainer is removed as bottom product.

We actually prefer an even broader definition of extractive distillation, as discussed later, that
includes the symmetrical process of separating maximum-boiling azeotropes using a low-boiling

Chapter 2. Methods of Azeotropic Mixtures Separation: General Considerations

entrainer (re-extractive distillation), and combined heteroazeotropic and extractive distillation
schemes (heteroextractive distillation). There are other more exotic methods that may also be
called entrainer-addition distillations, including:

Reactive distillation: The entrainer reacts preferentially and reversibly with one of the original

mixture components. The reaction product is distilled out from the non-reacting component
and the reaction is reversed to recover the initial component. The distillation and reaction
is usually carried out in one column (catalytic distillation).

Chemical drying (chemical action and distillation): The volatility of one of the original mixture

components is reduced by chemical means. An example is dehydration by hydrate forma-
tion. Solid sodium hydroxide may be used as an entrainer to remove water from tethra
hydrofuran (THF). The entrainer and water forms a 35-50 % sodium hydroxide solution
containing very little THF (Schweitzer, 1997).

Distillation in the presence of salts: The entrainer (salt) dissociates in the mixture and alters

the relative volatilities sufficiently so that the separation becomes possible. A salt added to
an azeotropic liquid mixture will reduce the vapor pressure of the component in which
it is more soluble. Thus extractive distillation can be applied using a salt solution as
the entrainer. An example is the dehydration of ethanol using potassium acetate solution
(Furter, 1968).

Obviously, the various methods may be combined and modified. For example, an entrainer
may form a pressure-sensitive azeotrope that makes it possible to utilize entrainer-addition and
pressure-swing distillation. The three conventional entrainer-addition distillation methods out-
lined in this section; homogeneous azeotropic distillation, heteroazeotropic distillation and ex-
tractive distillation, are described in more detail in Section 2.4.

2.3.4

Discussion

In the overview of separation processes given by Smith (1995), the basic advantages of distillation
are listed as: potential for high throughput, any feed concentration, and high purity. Because of
these advantages compared with other thermal separation processes, distillation is used in 90 %
of cases for the separation of binary and multicomponent liquid mixtures (Gmehling et al., 1994).
The distillation technique itself is a mature technology, that is, we have confidence in its design,
operation and control, in contrast to other promising technologies such as membranes. On the
negative side, distillation tends to use large amounts of energy (low thermodynamic efficiency)
and the introduction of an additional liquid entrainer to the mixture system causes additional
complexities. Distillation is particularly energy intensive when the heat of vaporization of the
mixture to be separated is high, such as for water and alcohols, and it is for these particular
separation tasks that pervaporation-distillation hybrids are said to be competitive.

Augmenting distillation with, for example, pervaporation can simplify the overall process struc-
ture, reduce the energy consumption and avoid entrainers (H¨ommerich and Rautenbach, 1998).

2.3 General separation processes

The main advantage of using membranes as mass separating agents is that the selectivity is in-
dependent of the vapor-liquid equilibria. For example a pervaporation unit can be used for the
azeotrope separation only. This allows the distillation column to be designed for less stringent op-
erating conditions, while the subsequent pervaporation stage pushes the separation to completion.
But, pervaporation and vapor permeation are expensive because of relatively low permeate fluxes
and the usually low condensation temperature (Rautenbach and Vier, 1995). In case of pervapora-
tion, processing costs are further increased by the required re-heating energy and the integration
of heat exchangers into the process. Industrial applications of pervaporation have generally been
limited to moderate volumes, because the prices of the membrane modules tend to be high for
large capacities. While the capital investments for distillation as a function of capacity scales
according to the “six-tenths power rule”, membranes tend to scale linearly with capacity (Kunesh
et al., 1995). Thus, distillation often has a distinct economic advantage at large throughput.

The fact that many of the alternative separation technologies to distillation can only handle low
throughput may not be a limitation. For example, pervaporation-distillation hybrids may be well
suited in small to moderate volume production and batchwise operation, such as in pharmaceutical
and specialty chemical industries. Dehydration of organic solvents by a pervaporation-batch dis-
tillation hybrid has significant advantages when batch units are required for subsequent treatment
of a variety of spent solvents (Rautenbach et al., 1992). Membrane units only handle relatively
pure feeds, that is, liquid mixtures dilute in one component and without heavy contaminants like
salts and tar. Solvent recovery from “mother liquors” may require pretreatment (e.g. filtration and
distillation) before the membrane unit.

Membrane units have little flexibility to variations in feed composition. Distillation, and in par-
ticular batch distillation, is based on robust and flexible equipment that can handle a wide range
of compositions and frequent changes of the feed mixture components. In addition, distillation
is able to achieve high-purity of the separation products. Many of the alternatives to distillation
only carry out partial separation and cannot produce pure end-products. Thus, several of the other
alternative separation methods can be used only in combination with distillation, while distillation
itself can be used as a stand-alone operation.

Let us go back to the commercial example of ethanol dehydration. The conventional method of
heteroazeotropic distillation using cyclohexane or benzene as entrainer produces absolute ethanol
(

99.5 mole %). The entrainer forms a heteroazeotrope that is easily recovered by decantation.

Still, some of the entrainer (about 50 ppm, typically) remains in the ethanol product, precluding
its use in pharmaceutical, medical, food or beverage products. Alternatively, the ethanol-water
azeotrope may be separated by molecular sieves (adsorption) or membrane pervaporation. In
combination with ordinary distillation these hybrid processes produce absolute ethanol without
other contaminants than the initial water. The advantages that are highlighted are that these pro-
cesses generate safer products and give significant energy savings. Sometimes the membrane
material must be regenerated using chemical solvents and the same added aspects as for entrainer-
addition distillations applies.

Bergdorf (1991) studied several membrane-distillation hybrids for dehydration of azeotrope-
forming solvents. Interestingly, a combined system of pervaporation and pressure-swing distilla-

Chapter 2. Methods of Azeotropic Mixtures Separation: General Considerations

tion was found to be economically favorable in one case. No particular separation technique was
concluded to be more promising than others, as Bergdorf (1991) formulated “each dehydration
process finds its range of application”.

It is important to consider the separation sequence as a whole, including the regeneration of
the separating agent. Exploiting pressure sensitivity should usually be considered before the
use of an extraneous liquid component. In the choice between an entrainer that forms a new
homoazeotrope, and an entrainer that forms a heteroazeotrope, the latter is normally preferred
since a heteroazeotrope can be separated simply by decantation in combination with distillation.
Occasionally, a component that already exists in the process (or production site) can be used as
an entrainer. These options should be investigated first. In pharmaceutical production, a common
requirement is that the products must be completely entrainer-free. As mentioned, the latter may
be unavoidable in entrainer-addition distillation methods and causes such methods to be entirely
excluded from consideration. This issue may even be the main incentive to look more carefully
into alternatives to the conventional entrainer-addition distillation methods.

Brown (1963) compared the economics of nine separation processes for the separation of acetic
acid and water (a zeotropic close-boiling mixture). These included ordinary distillation, het-
eroazeotropic distillation, extraction with low-boiling, and extraction with high-boiling entrain-
ers. They concluded that the most economical process varied with the concentration of the orig-
inal mixture feed composition. Extraction processes gave the greatest return for 2 to 60 mole %
feeds and azeotropic distillation processes greater return at above 75 mole % feed concentration.
According to Gmehling et al. (1994) a general comparison of extractive and heteroazeotropic
distillation favors the latter because extractive distillation has the disadvantage that the entrainer
must be vaporized and a higher energy requirement is therefore necessary. Another industrial mix-
ture example is tetrahydrofuran and water which form a minimum-boiling azeotrope. Chang and
Shih (1989) compared heteroazeotropic distillation using n-pentane as entrainer to the pressure-
swing distillation scheme, which is the conventional method for separation of this mixture. The
proposed n-pentane heteroazeotropic distillation scheme was found to be competitive with the
low/high pressure distillation scheme, but with “no technical or economical edge”, that is, with
slightly higher investment costs and a lower operating cost.

According to Tassios (1972), azeotropic distillation separation methods like extractive- and het-
eroazeotropic distillation have not been used as frequently as they should in industry, even if the
often used reasons of high investment and high operating costs are usually weak when one does
an in-depth study. Roche [in Tassios (1972)] claims that the real reason lies with the time and
money required to obtain a satisfactory process design. Today, simple tools based on graphical
analysis of vapor-liquid equilibrium diagram structures are established providing a rapid method
for evaluating the possibilities to separate a given azeotropic mixture, choose a suitable separation
(distillation) method and process scheme and predict the achievable distillation products composi-
tions. Together with more advanced distillation synthesis tools and simulation software, this dra-
matically change the way distillation processes for complex nonideal mixtures can be analyzed,
understood and designed. Significant improvements over conventional process designs have been
shown using these relatively new tools (Wahnschafft, 1997; Serafimov and Frolkova, 1997).

2.4 Entrainer-addition distillation methods

2.4

Entrainer-addition distillation methods

Given an original binary mixture with one homoazeotrope we search for an entrainer that forms
a ternary mixture that is separable, either by conventional distillation or by the special distillation
methods outlined in Section 2.3. The distillation behavior of ternary azeotropic mixtures may
be very complex and the decision whether the resulting mixture is separable by distillation is
not straightforward. Therefore, we need a tool for the evaluation of the possibilities to separate
ternary azeotropic mixtures at the presynthesis stage.

2.4.1

Feasibility analysis

Ternary VLE diagrams, such as distillation line map and residue curve map, provide a graphical
tool for the prediction of feasible distillation product compositions. Here, we only consider the
simple

analysis for such prediction (Stichlmair, 1988), that is, the infinite reflux bounding

of the feasible product composition regions with a finite number of equilibrium stages. From
this quick graphical method we can immediately identify the limited separations achievable by
distillation.

For an introduction to graphical methods for azeotropic distillation feasibility analysis and sepa-
ration synthesis the reader is referred to Westerberg and Wahnschafft (1996) and Stichlmair and
Fair (1998). It is desirable to depict the regions in which the distillate and bottom products are
confined. Triangular diagrams of distillation lines or residue curves are particularly suited to
identify feasible products at the limiting operating condition of infinite and total reflux (that is,
the reflux ratio

L=D

in Figure 2.10 is very high or equal to infinity). The distillation lines

coincide exactly with the infinite reflux steady state composition profiles of equilibrium staged
distillation columns, that is, the operating line at infinite reflux ratio. The residue curves coin-
cide exactly with the infinite reflux steady state liquid composition profiles of packed distillation
columns when all the resistance to mass transfer is in the vapor phase (perfectly mixed liquid
phase) (Serafimov et al., 1973; P¨ollmann and Blass, 1994). For a feasible separation, the material
balances have to be fulfilled:

That is, the distillate and bottom product compositions must lie on the straight material balance
line through

, and, due to the assumption of infinite reflux, on the same distillation line. The

resulting (infinite reflux) feasible distillation product regions are illustrated in Figure 2.9. The
regions are bounded by the distillation line through the feed and the two material balance lines
for pure distillate and pure bottom product, respectively. These two material balance lines corre-
spond to the direct and indirect sharp splits (with an infinite number of equilibrium stages). The
distillation line through the feed represents the separation with minimum number of equilibrium
stages. Note in particular that it is impossible to achieve a saddle (

) as a pure product in a

Chapter 2. Methods of Azeotropic Mixtures Separation: General Considerations

Material balance lines

through feed

Distillation line

direct split

indirect split

Figure 2.9: Feasible product regions for a ternary zeotropic mixture with feed

determined by

infinite reflux analysis. The possible distillate

and bottom

product compositions (shaded)

are limited by the distillation line through the feed

, and by the straight material balance line

through

and the initial and final points of the distillation lines, respectively. Pure components

indicated by

(stable node),

(unstable node) and

(saddle). (The diagram corresponds to

Class 0.0-1 (Cell I), see Chapter 4).

single-feed zeotropic distillation column. The nodes (

and

) determines the potential distillate

and bottom products for a given feed (for each column in a separation sequence).

2.4.2

Feasibility analysis for azeotropic mixtures

Unlike zeotropic distillation, the product compositions of azeotropic distillation depend on the
feed composition and its position relative to distillation line (or residue curve) boundaries. As a
first approximation, we assume that the composition trajectories of the distillation column operat-
ing at finite reflux cannot cross the infinite reflux distillation boundaries. Then, the feasible prod-
uct regions for azeotropic mixtures are limited by distillation lines boundaries too, if any. How-
ever, small deviations occur for finite reflux. Advanced approaches for the prediction of feasible
product compositions at real operating conditions have been developed (Petlyuk, 1978; Nikolaev
et al., 1979; Kiva et al., 1993; Wahnschafft et al., 1992; P¨ollmann and Blass, 1994; Castillo et
al., 1998; Thong et al., 2000). These approaches give some refinements of the feasible product
regions even for zeotropic mixtures (additional product compositions reachable at finite reflux),
and additional possibilities for certain types of azeotropic diagrams. But, the main conclusions are
the same as for the simple infinite reflux approach; that is, the possibilities of separating (ternary)
azeotropic mixtures by ordinary distillation are very limited. An important exception is mixtures
with curved distillation line boundaries which can be crossed at finite reflux if the feed is located
in the concave region of the boundary (Marchenko et al., 1992; Wahnschafft et al., 1992).

2.4 Entrainer-addition distillation methods

2.4.3

Homogeneous azeotropic distillation

Here we consider ordinary distillation of ternary mixtures with at least one binary homoazeotrope
(that is, the one that stems form the original mixture). The entrainer may or may not form new
azeotropes in the system. The only criteria is that the resulting ternary system should form a VLE
diagram structure that is promising for separation. This statement may seem obvious, but we find
it necessary to emphasize that no other limitations or criteria are necessary to pose at this stage.
The entrainer selection criteria given by several authors such as Van Winkle (1967), Doherty and
Caldarola (1985) and Stichlmair and Fair (1998) are discussed later. The distillation is carried out
in a conventional single-feed column as illustrated in Figure 2.10. The entrainer

is introduced

Entrainer

Distillate

Bottom

Original mixture

Figure 2.10: Homogeneous azeotropic distillation in a single-feed column.

to the original mixture feed

to form a ternary mixture feed

. The separation sequence alter-

natives are similar to the various schemes of ordinary zeotropic distillation of ternary mixtures
(direct, indirect, prefractionator and so forth).

We distinguish between two broad cases of the entrainer properties: (1) the entrainer is zeotropic,
and (2) the entrainer is azeotropic. Our presentation here does not consider all the resulting
feasible diagram structures, but we give selected examples to illustrate the ideas.

If the entrainer does not form any new azeotropes (zeotropic entrainer) the resulting ternary mix-
tures are shown in Figure 2.11. We refer to Chapter 4 for an overview of the classifications
of feasible VLE diagrams for ternary azeotropic mixtures. The terms “light”, “intermediate”
and “heavy” entrainer are used here in reference to the boiling temperature of the pure com-
ponents that forms the mixture. The relative volatility of the components in the mixture vary
within the composition diagram and may be identified by univolatility lines and K-ordered re-
gions (Chapter 3). Note the symmetry between the “antipodal diagrams” of minimum-boiling
binary azeotropes and maximum-boiling binary azeotropes. For simplicity, we only consider an

Chapter 2. Methods of Azeotropic Mixtures Separation: General Considerations

E (intermediate)

E (heavy)

E (light)

E (intermediate)

E (light)

E (heavy)

Class 1.0-1a

Class 1.0-2

Class 1.0-1b

min

max

Figure 2.11: Zeotropic entrainer. Structures of the resulting ternary VLE diagrams of the original
binary azeotropic mixture of components 1 and 2, and the entrainer E. The singular points are
indicated by

stable node,

unstable node and

saddle.

original mixture with a minimum-boiling azeotrope. The same arguments applies to the oppo-
site case of maximum-boiling binary azeotrope, but with inverted schemes for the distillation
separation (and some additional aspects due to the different volatility behavior in the column).

For a quick analysis of the possibilities and limitations of the distillation of the resulting ternary
mixtures (original mixture + entrainer) we use the infinite reflux approach:

Heavy zeotropic entrainer: The analysis illustrated in Figure 2.12 shows that it is not possible

to separate the original mixture into pure components 1 and 2 (both saddles) by a sequence
of ordinary single-feed distillation columns, and the azeotrope

is an inevitable prod-

uct. A more detailed (finite reflux) analysis gives the same conclusion (Wahnschafft and
Westerberg, 1993). (However, the special method of extractive distillation may be used to
separate mixtures of this structure in a two-feed column scheme as presented later.)

Intermediate zeotropic entrainer: The analysis illustrated in Figure 2.13 shows that we can

achieve pure components 1 and 2 and E by ordinary single-feed distillation columns de-
pending on the feed composition (and operating conditions), even though one of the orig-
inal mixture components is a saddle. An intermediate zeotropic entrainer thus seems very
promising. However, there is a possibility for multiple steady states for certain operating
conditions, which may cause operational problems, and care must be taken when designing

2.4 Entrainer-addition distillation methods

Figure 2.12: Feasible product regions for mixture with heavy entrainer. It is impossible to obtain
pure components 1 and 2 in a sequence of single-feed distillation columns. (Corresponds to Class
1.0-1a, Cell II.)

Figure 2.13: Feasible product regions for mixture with intermediate entrainer. Pure components
1 and 2 (and E for specific feed compositions) may be obtained. (Corresponds to Class 1.0-1b,
Cell III.)

such processes (Laroche et al., 1992b; Bekiaris et al., 1993; G¨uttinger and Morari, 1996;
Kiva and Alukhanova, 1997). Furthermore, finding an entrainer that forms this diagram
structure is difficult as its occurrence in nature is limited. In addition, because of the re-
quirement that the entrainer has a boiling-point between the original mixture components,
the relative volatilities in the resulting mixture are usually close to unity and the separation
often uneconomical. However, the often referred mixture of acetone-heptane-benzene is an
example for which this process is economical (Stichlmair and Fair, 1998).

Light zeotropic entrainer: The analysis is illustrated in Figure 2.14. This diagram class has a

distillation boundary that splits the composition space into two distillation regions. The
original mixture components 1 and 2 are located in two different distillation regions and if
the boundary is linear the separation is impossible. If we view the diagram as a combination
of two cells, we see that the feasible regions (Figure 2.14 left) are as showed in Figure
2.9. However, for curved boundaries the answer is not so obvious (Figure 2.14 right).

Chapter 2. Methods of Azeotropic Mixtures Separation: General Considerations

00000000

11111111

Feed region for

at finite reflux

sharp 1/2E split

Figure 2.14: Feasible product regions for mixture with light entrainer. Some possibility for ob-
taining pure components 1 and 2. (Corresponds to Class 1.0-2 consisting of two Cell I’s.)

Complete separation may be achieved by a scheme with recycle for feed compositions
located in the concave region of the distillation line boundary (Wilson et al., 1955; Balashov
et al., 1970; Kiva and Serafimov, 1976b; Doherty and Caldarola, 1985; Stichlmair, 1988;
Wahnschafft et al., 1992; Wahnschafft et al., 1993). This scheme may be economical for the
separation of a maximum-boiling azeotrope using a entrainer (antipodal diagram) where the
heavy entrainer recycle does not have to be evaporated. Detailed analysis show that there
is a feed region (striped) where the sharp separation 1/2E can be achieved at finite reflux
ratio (Marchenko et al., 1992; Kiva et al., 1993; Wahnschafft et al., 1992; Wahnschafft et
al., 1993). For feeds located in this narrow region, pure components 1 and 2 can be achieved
through a sequence of ordinary single-feed distillation columns (that is, the distillation line
boundary is crossed by the given distillation trajectory at finite reflux). In practice, however,
it is difficult to find an entrainer that gives sufficient curvature of the boundary and large
enough relative volatilities.

If the entrainer forms new azeotropes, there are numerous possible diagrams structures of the
resulting ternary mixture. Selected examples are shown in Figure 2.15. Analysis of the feasible
product regions of all possible ternary VLE diagram structures with an evaluation of the possi-
bilities to separate each structure is beyond the scope of this thesis. For a quick estimate of the
possibilities to separate resulting ternary diagrams with several distillation line regions, we pro-
pose to use the concept of elementary topological cells presented in Chapter 4. The idea is that,
based on our knowledge about the simple diagrams given above, we can qualitatively predict the
behavior of other more complex diagrams by looking at them as combinations of these simple
structures. For example (from Figure 2.15):

Class 2.0-2a: This diagram includes the promising Cell III (with the characteristic U-shaped

distillation lines) adjoining the original mixture edge. That is, the lower region of this
diagram is similar to the Class 1.0-1b in Figure 2.13, where the new azeotrope

acts

as an “intermediate” entrainer. The separation into pure components 1 and 2 is possible
(Foucher et al., 1991).

2.4 Entrainer-addition distillation methods

2.0-2a

2.0-2b

2.0-2c

3.0-2

3.0-1a

III

2.0-1

Figure 2.15: Azeotropic entrainer. Selected structures of the resulting ternary VLE diagrams of
the original mixture 1 and 2, and the entrainer E. Elementary cells indicated by I, II, III and IV.

Class 2.0-2b: This diagram may be viewed as a combination of two subsystems: Cell I (pseudo

ternary zeotropic mixture similar to the structure in Figure 2.9) and Cell II (similar to the
Class 1.0-1a in Figure 2.12). Separation into pure components 1 and 2 is not possible due to
the distillation boundary between the components 1 and 2, unless the boundary is (highly)
curved. Then, the scheme with recycle can be implemented where the azeotrope

acts

as the entrainer (Stichlmair and Fair, 1998).

Class 3.0-1a: The same arguments as for Class 2.0-2a applies.

Class 3.0-2: The same arguments as for Class 2.0-2b applies.

Doherty and Caldarola (1985) give the condition that the separation is possible if both the original
mixture components 1 and 2 belong to the same distillation line region (as is the case for the
Classes 2.0-1, 2.0-2a and 3.0-1a in Figure 2.15). An important exception is the crossing of curved
distillation line boundaries (as might be a possibility for the Classes 2.0-2b and 3.0-2 in Figure
2.15). Here, the condition given by Stichlmair and Fair (1998) is more appropriate; that is, the
separation is possible if the points of the original mixture components 1 and 2 are nodes of two
neighboring regions (with curved boundary).

Even though some of the resulting structures with an azeotropic entrainer are promising, the
overall conclusion is that the possibilities of separating azeotropic mixtures in ordinary single-

Chapter 2. Methods of Azeotropic Mixtures Separation: General Considerations

feed columns are limited and therefore special distillation methods needs to be applied.

2.4.4

Heteroazeotropic distillation

Heteroazeotropic (VLLE) distillation, often referred to as azeotropic distillation (but we do not
recommend this nomenclature), involves the formation of a heteroazeotrope (or the use of an
existing heteroazeotrope) to effect the desired separation. The crucial difference compared to
homoazeotropic distillation is that a heteroazeotrope can be separated easily by decantation. This
also makes the process scheme simpler. First, for illustration purposed, we consider in Figure
2.16 a particular example of heteroazeotropic distillation where the original mixture is ternary
mixture of Class 1.0-2 with a saddle heteroazeotrope that splits the diagram into two distillation
region. The distillation boundary is “crossed” by the liquid-liquid phase splitting (decantation)

Pure 3

Pure 1

1, 2, 3

1.0-2

"Pure" 2

Figure 2.16: Heteroazeotropic distillation with side-stream decanter.

of the heteroazeotrope, and the separation into the three original mixture components is feasible.
An example mixture of this type is acetone - water - 1-butanol (Stichlmair and Fair, 1998).

Given that the original mixture is binary and includes a homoazeotrope, there are many possible
VLLE diagrams that result from adding an entrainer that forms a binary heteroazeotrope with
one of the original mixture components, or, a ternary heteroazeotrope. The most common struc-
ture is Class 3.1-2 with three binary minimum-boiling azeotropes and a ternary heterogeneous
azeotrope. The entrainer may be mixed with the original feed or introduced to the column as a
separate second feed. If the entrainer has a volatility near that of the feed, it is usually mixed with
the feed to form a single stream. If volatility is below that of the feed, the entrainer is usually
added above the feed stage (Schweitzer, 1997). A typical heteroazeotropic distillation process
scheme is shown in Figure 2.17. Relatively pure component 2 is removed as the bottom product.
The ternary heteroazeotrope is distilled overhead. Part of the condensate

is refluxed to the

2.4 Entrainer-addition distillation methods

Entrainer makeup

"Pure" 1

Entrainer-phase

Original mixture

3.1-2

Pure 2

Figure 2.17: A more common heteroazeotropic distillation scheme with distillate decanter.

column, if necessary, to achieve a satisfactory separation of the heteroazeotrope, and the remain-
der is separated into two liquid phases in the decanter. One liquid phase contains the bulk of the
entrainer (and component 2) which is returned as reflux. The other liquid phase contains the bulk
of component 1. If this layer contains a significant amount of entrainer, then it may be fed to
an additional column to separate and recycle the entrainer and produce pure 1. Promising VLE
diagrams of the resulting ternary system (original mixture + entrainer) are structures where the
heteroazeotrope formed by the entrainer is an unstable node (

). The diagram is then split into

at least two distillation regions. The original mixture components 1 and 2 are stable nodes (

)

of these regions. The feed composition is located in one region and one end of the liquid-liquid
tie-line is located in the other.

Ethanol dehydration: The earliest example of heteroazeotropic distillation is the dehydration of
ethanol using benzene as entrainer that was presented by Young in 1902 (Pratt, 1967). Ethanol (2)
- water (1) - benzene (E) forms a ternary VLE diagram of Class 3.1-2 in Figure 2.17. Pure (an-
hydrous) ethanol is produced in the bottom. The ternary heteroazeotrope is removed as distillate
and separated into an organic-phase containing only 1 mole % water and a water-phase containing
36 mole % water in the decanter. The organic-phase is recycled to the column. The water-phase
is sent to recovery columns for removal of water and recovery of benzene and ethanol. Today,
cyclohexane is a more common entrainer for ethanol dehydration by heteroazeotropic distillation.

An interesting example, given by Wilson et al. (1955), is the dehydration of hydrazine, which
form a maximum-boiling homoazeotrope with water

, by using aniline as an entrainer. Aniline

is a “heavy” entrainer that forms a binary heteroazeotrope with water

. The resulting ternary

VLLE diagram is of Class 2.0-2a with a curved distillation boundary as illustrated in Figure 2.18.
The previously described scheme with recycle (moving the feed composition of Column 1 into
the concave region of the curved distillation boundary) was implemented.

Chapter 2. Methods of Azeotropic Mixtures Separation: General Considerations

Column 3

Column 2

Column 1

Decanter

Aniline

Water

Aniline

Hydrazine

Hydrazine (2)

Water (1)

Aniline (E)

Class 2.0-2a

Figure 2.18: Heteroazeotropic distillation scheme with recycle (Wilson et al., 1955).

For more details about heteroazeotropic distillation processes in general the reader is referred to
the literature, e.g. Doherty and coworkers (1990a; 1990b; 1990c), ?, and Widagdo and Seider and
coworkers (1992; 1996).

The possible entrainers for heteroazeotropic distillation are limited, but if one can be found then
it is usually preferred compared to an entrainer for extractive distillation.

It is generally possible to find a heteroazeotropic entrainer when the original mixture con-

2.4 Entrainer-addition distillation methods

sists of hydrophilic and hydrophobic organic components (such as ethanol and water);

It is very difficult to find an entrainer when the original mixture consists of only hydrophilic
(or only hydrophobic) components (such as benzene and cyclohexane).

2.4.5

Extractive distillation

Extractive distillation is the oldest known method for separating azeotropic mixtures. Benedict
and Rubin (1945) give the following definition of extractive distillation:

“Distillation in the presence of a substance which is relatively non-volatile compared
to the components to be separated and which, therefore, is charged continuously near
the top of the distilling column so that an appreciable concentration is maintained
on all plates of the column.”

The basic principle is that the extractive entrainer interacts differently with the components of
the original mixture and thereby alters their relative volatility. These interactions occur predomi-
nantly in the liquid phase and the entrainer is introduced above the original mixture feed point to
ensure that the entrainer remains in appreciable concentration in the liquid phase in the column
section below.

The definition of extractive distillation given above is much broader than the traditional one
found in most textbooks and encyclopedia (Hoffman, 1964; Van Winkle, 1967; Perry, 1997).
A common restriction on the “extractive distillation” is that the entrainer should not introduce
new azeotropes in the system, that is, no distillation boundaries (Stichlmair et al., 1989; Doherty
and Knapp, 1993; Gmheling and M¨ollmann, 1998). However, this restriction seems unnecessary.
The entrainer may well form new azeotropes and this may even be preferably as in the process of
heteroextractive distillation where the entrainer forms a heteroazeotrope with one of the original
mixture components and is miscible and selective to the other. Furthermore, the extractive distilla-
tion scheme may operate within one distillation region of the ternary mixture composition space,
and still achieve a desirable separation (in terms of the original mixture product specifications).

Let us first consider the simplest example of extractive distillation where the original mixture
is a binary minimum-boiling homoazeotrope and the entrainer is miscible and zeotropic. The
main conditions that a candidate extractive entrainer must satisfy in order to make the separation
feasible are:

First condition:

The resulting ternary system of the binary minimum-boiling azeotrope and the entrainer
should form a VLE diagram (or cell) of the structural type shown in Figure 2.19.

Chapter 2. Methods of Azeotropic Mixtures Separation: General Considerations

Residue Curves Map

METHANOL

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

ACETONE

METHANOL

WATER

Univolatility line

= 1

12E

21E

Figure 2.19: Promising VLE diagram for extractive distillation (Class 1.0-1a, Cell II). The uni-
volatility line

=1 indicates the change in volatility order of the mixture components.

Second condition:

The entrainer must cause a substantial change in the relative volatility between the azeotrope-
forming components 1 and 2. Typically, the relative volatility

at infinite dilution in the

entrainer should be larger than 2.

Third condition:

The entrainer must have a higher boiling-point than the original mixture components. Typ-
ically,

1;2

. The relative volatility between the entrainer and the compo-

nent that is recovered in the distillate should preferably be large (typically,

3:0

The larger the boiling-point difference between the entrainer and the distillate compo-
nent, the less reflux and less number of trays needed in the rectifying section (typically,

1;2

150

The most important feature of the diagram in Figure 2.19 is the shape of the residue curves near
the entrainer stable node (circled). The presence of a univolatility line deforms the residue curves
according to the entrainer selectivity (here causing S-shaped residue curves). The shape of the
residue curves in Figure 2.19 indicates that the entrainer has strongest affinity to component 2.
The entrainer must be miscible with the component removed in the bottom product, but there are
no special requirements to the miscibility with the other component removed as distillate (that is,
the entrainer can be miscible, partly miscible or immiscible with the distillate component).

2.4 Entrainer-addition distillation methods

Process scheme

An illustration of extractive distillation is shown in Figure 2.20. Extractive distillation differs

(b)

Extractive distillation

Entrainer recovery

Pure 2

Pure E

(a)

Pure 1

Figure 2.20: Extractive distillation: (a) composition profile (bold line), material balance line
(dashed) and distillation lines (dotted); (b) extractive distillation column sequence.

distinctly from conventional homogeneous azeotropic distillation in the organization of the flows
and the resulting material balance equations (operating line). The crucial element is the extractive
section with countercurrent flow of volatile and non-volatile components. A greater entrainer
flowrate generally yields better separation, but this increases the energy demand in both columns
and increases the reboiler temperature in the extractive column (Smith, 1995).

In extractive distillation a saddle is obtained as a distillate product which is “against” the dis-
tillation lines (residue curves) (see the composition profile in Figure 2.20). This is the main
characteristic of the extractive distillation process. Wahnschafft and Westerberg (1993), Knapp
and Doherty (1994) and Bauer and Stichlmair (1995) presents extractive distillation composition
profiles for the system of acetone, methanol and water (Class 1.0-1a), and geometric representa-
tions of the extractive distillation process are given in a series of papers by Doherty and coworkers
(Levy and Doherty, 1986; Knapp and Doherty, 1990; Knapp and Doherty, 1994). The liquid com-
position profile along the column shown in Figure 2.20 turns off the binary edge between the
distillate product (1) and the entrainer (E). The imbalance of mixing due to the entrainer feed

causes a shift in the composition profile and the profile moves across the distillation lines (or

residue curves) in the extractive section.

For the mixture given in Figure 2.19, the minimum-boiling azeotrope is an unstable node in
the whole composition triangle. The pure components 1 and 2 are both saddle points (but the
component 1 is an unstable node of the binary edge between component 1 and the entrainer).
The univolatility line of the azeotrope constituents, that is, the locus of points in the composition
space where

, extends from the azeotrope

to the binary edge between component

Chapter 2. Methods of Azeotropic Mixtures Separation: General Considerations

1 and the entrainer. Thus, component 1 is recovered in the distillate since it is the most volatile
component in the rectifying section (Laroche et al., 1993). Component 2 is removed as bottom
product together with the entrainer. Thus, the complete extractive distillation scheme requires a
sequence of two columns, one extractive distillation column and one ordinary column (or stripper)
to separate component 2 from the entrainer.

High reflux may be harmful in extractive distillation because it weakens the extractive effect
(decreases the entrainer concentration in the extractive column section). Previous studies on
continuous extractive distillation have shown that there is a maximum as well as a minimum
reflux ratio to make the separation scheme feasible (Andersen et al., 1995; Laroche et al., 1993).
Increased reflux ratio moves the distillate composition from the saddle (component 1) to the
unstable node (

) in Figure 2.19.

The same extractive distillation scheme can in principle be applied to the symmetrical system of
separating a maximum-boiling azeotrope using a light entrainer (antipodal diagram Class 1.0-1a),
and this process is referred to as re-extractive distillation. The light entrainer feed is here intro-
duced below the original mixture feed point. However, this system is not as beneficial because the
entrainer interactions predominantly occur in the liquid phase and with a volatile entrainer this
effect is more difficult to obtain than in the conventional scheme.

According to Smith (1995), extractive distillation is more useful than heteroazeotropic distillation
because the process does not depend on the formation of an azeotrope and thus a greater choice
of mass separating agent is, in principle, possible. Although it is correct that it is easier to find
suitable entrainers for extractive distillation, this is not because “the process does not depend
on the formation of an azeotrope”. High selectivity is one of the most desirable features of an
entrainer, and this is generally related to a high degree of nonideality (molecular interaction).
Formation of an azeotrope is not necessarily a disadvantage for the extractive distillation scheme.

Heteroextractive distillation

If a new heteroazeotrope forms between the extractive entrainer and the component of the original
mixture that is to be removed as distillate, then both heteroazeotropic and the extractive distilla-
tion can be combined into a single process as illustrated in Figure 2.21. The heteroazeotrope

(saddle) is removed overhead and separated by condensation and decantation, while excess of the
entrainer is removed in the bottom product together with the one of the original mixture compo-
nent that it extracts (here: component 1). The extractive process works within one distillation line
(residue curve) region (Cell II) of the complex diagram, and the distillation boundaries are crossed
by the liquid-liquid phase splitting in the heterogeneous region (Wijesinghe, 1985). This is in fact
a very attractive separation method because the entrainer is double-effective. Example mixtures
of the type given in Figure 2.21 are methyl ethyl ketone (2-butanone) - water - cyclohexanone;
isopropanol - water - isobutanol; acetone - methylene chloride (CH

) - water; methanol - car-

bon tetrachloride (CCl

) - water. In the screening for entrainer such candidates are often ruled

out, e.g. by the entrainer selection criteria given by Doherty and coworkers (1985; 1991; 1993)
and Seader and Henley (1998). Thus, many of the entrainer selection criteria are too limiting,
and one should not narrow the search for a new separation system by using conditions that apply
for “old” processes, but rather see the possibilities and limitations that lie within each system in

2.4 Entrainer-addition distillation methods

1E2

21E

12E

pure 2

Entrainer phase

1 + E

2.0-2b

Figure 2.21: Heteroextractive distillation (Wijesinghe, 1985).

question.

Another interesting example, given by Petlyuk and Danilov (2000), is the separation of acetone
and chloroform, which form a maximum-boiling homoazeotrope, by using water as an entrainer.
Water is relatively high-boiling and selective to acetone. In addition, an immiscibility region and
a binary heteroazeotrope

and a ternary heteroazeotrope

12E

are formed as illustrated in

Figure 2.22. Note that there is a change in volatility order of the components within the region

Acetone + Water

Water-phase

Water

Chloroform (2)

Acetone (1)

Water (3)

Acetone

Chloroform

321

123

132

321

213

123

Water

Chloroform

2.1-3a

231

Figure 2.22: Heteroextractive distillation (Petlyuk and Danilov, 2000).

where the feed compositions are located. The

univolatility line indicates this change. If the

Chapter 2. Methods of Azeotropic Mixtures Separation: General Considerations

summary feed

is located in the region with the 213 volatility order of the components, the

azeotrope

is removed as distillate product (not the saddle azeotrope

123

) and split into pure

2 and E by condensation and decantation.

In conclusion, one should be careful about limiting oneself when considering extractive distilla-
tion (“no distillation boundaries”) and also careful about using established selection criteria for
entrainers.

2.4.6

Entrainer selection

The task of entrainer searching should be based on thermodynamic and physical insights (e.g.
about the intermolecular forces and the VL(L)E diagram structure).

Ewell et al. (1944) studied the relationship between hydrogen bonding and azeotrope forma-
tion and classified entrainers into groups according to their molecular interactions. From this
they developed guidelines to identify chemical classes suitable as entrainers for heteroazeotropic
and extractive distillation. These guidelines were discussed by Berg (1969), who further devel-
oped a classification of organic and inorganic mixtures and used information about the molecular
structure to point out promising entrainers for extractive and heteroazeotropic distillation. Berg
(1969) states that successful entrainers for extractive distillation are highly hydrogen-bonded liq-
uids (such as water, amino alcohols, amides, phenols alcohols and organic acids).

The effectiveness of the entrainer is based on its ability to modify the relative volatility of the
original mixture. Various approaches for this evaluation are given, for example by Van Winkle
(1967):

Extractive distillation: A promising entrainer forms a complex or hydrogen bond with one of

the components

in the original mixture, resulting in a decrease in the vapor pressure of

component

such that

decreases. If this reduces

to 0.8 (or

to 1.2 or greater),

the separation is greatly enhanced.

Heteroazeotropic distillation: A promising entrainer “breaks” a complex formed between the

components 1 and 2, or between 1 and 1, resulting in an increase in the vapor pressure of
component

such that

increases.

Van Winkle (1967) emphasizes that the whole process system must be considered in the selection
of an effective entrainer and that the recovery process must be included in the evaluation. A
candidate entrainer might have a very good selectivity, but it might be difficult to recover in the
second column. Therefore, the final selection must be made based on an economic evaluation.

Molecular design of an entrainer may even be pursued (see for example Brignole et al. (1986) and
Pretel et al. (1994)). Group contribution methods are used to synthesize entrainers with specific
physical properties (Furtzer, 1994; Gmheling and M¨ollmann, 1998). Search for specific compo-
nent properties in computerized data banks may be performed, where activity coefficients at infi-

2.5 Summary and conclusions

nite dilution are helpful measures of the entrainer selectivity. Expert systems to identify candidate
entrainers have been developed with heuristic rules implemented into knowledge databases.

2.5

Summary and conclusions

Separation of homogeneous azeotropic mixtures is a topic of great practical and industrial inter-
est. Most liquid mixtures of organic components form nonideal systems. The presence of some
specific groups, particularly polar groups (oxygen, nitrogen, chlorine and fluorine), often results
in the formation of azeotropes. Binary homoazeotropes (at least the ones that are not pressure
sensitive) are impossible to separate by ordinary distillation, but may be effectively separated by
membrane-distillation hybrids or distillation methods where a third component is added to the
system. Analysis of ternary VL(L)E diagrams is an efficient tool to predict feasible separations
for entrainer-addition distillation processes. Simplest approaches are sufficient if used properly.
We distinguish between three different entrainer-addition distillation methods depending on the
properties of the entrainer and the organization of the column flows and the following conclusions
can be made:

Homogeneous azeotropic distillation (ordinary distillation of homoazeotropic mixtures)

Only a few VLE diagram structures of the resulting ternary system of original mixture and
entrainer are possible to separate by ordinary distillation (sequence of single-feed columns).
The separation schemes are complex and may be very energy intensive, or may have mul-
tiplicities.

Heteroazeotropic distillation (decanter-distillation hybrids involving heteroazeotropes)

Several VLLE diagram structures that involves one or more heteroazeotropes are possible
to separate by ordinary distillation combined with decantation. The separation schemes are
fairly simple, but the range of feasible entrainers is rather limited.

Extractive distillation

Extractive distillation is the most general distillation method for separating homoazeotropic
mixtures. The scheme is not sensitive to the type of original mixture as there is a broad
range of feasible entrainers. The process is not limited to homogeneous entrainers, and
heteroextractive distillation-decanter hybrids may be applied with success.

Insights into to the thermodynamic behavior of azeotropic mixtures is fundamental for the devel-
opment of separation system involving azeotropic mixtures. This is an incentive to consider the
powerful tool of graphical VLE representation in more detail.

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

called azeotropes and mixtures with this phenomena are called azeotropic mixtures. The highly
nonlinear vapor-liquid equilibrium behavior of azeotropic mixtures complicates the prediction of
feasible separation sequences further, and an azeotrope itself cannot be separated by ordinary
distillation since no enrichment of the vapor phase occurs at this point. Usually, special methods
are required.

Azeotropic distillation is defined as distillation that involves components which form azeotropes.
Generally, there are two cases of azeotropic distillation: (i) when the original mixture to be sep-
arated is an azeotropic mixture, and, (ii) when an azeotropic mixture is formed deliberately by
adding one or more azeotrope-forming components to the original mixture. In the first case we
have to find a way to separate the azeotropic mixture and obtain the desired product specifications
and recovery. In the second case, in addition, we have to select an azeotrope-forming compo-
nent (called entrainer) that is effective for the desired separation and easily recovered afterwards.
In either case we should establish the options before analyzing them in detail. For this purpose
we need a tool to qualitatively predict the feasible separations for multicomponent azeotropic
mixtures. The tool is known as thermodynamic topological analysis of distillation (in Russian
literature) or residue curve (or distillation line) map analysis. It provides an efficient way for
preliminary analysis of nonideal distillation problems and presynthesis of separation sequences.

Thermodynamic topological analysis

The thermodynamic topological analysis is based on the classical works of Schreinemakers (1901b;
1901c; 1902) and Ostwald (1902), where the relationship between the vapor-liquid equilibrium
of a mixture and the behavior of open evaporation residue curves for ternary mixtures was es-
tablished. Although open evaporation (a single vaporization step with no reflux) itself is not of
much industrial interest in the 21st century, it nevertheless forms an important basis for the un-
derstanding of distillation (a sequence of partial vaporization steps with reflux). The findings by
Schreinemakers did not receive recognition until the 1940s when Reinders and de Minjer (1940)
and Ewell and Welch (1945) showed the possibility to use this approach to predict the behav-
ior of batch distillation. These achievements stimulated subsequent investigations by Haase and
Lang (1949; 1950a; 1950b; 1951) and Bushmakin and Kish (1957a; 1957) on the structure of
phase equilibrium diagrams and its connection with batch distillation behavior. In 1958, Gurikov
(1958) formulated the ”Rule of azeotropy” and proposed a classification of ternary mixtures based
on their thermodynamic topological structures (residue curve map analysis).

In the late 1960s, Zharov (1967; 1968c) gave a more rigorous mathematical foundation of the
residue curve map analysis and expanded it to multicomponent mixtures. During the same period,
Serafimov (1968a; 1968d) proposed to use structural information of VLE diagrams to predict
feasible separations in continuous distillation. The results of Zharov and Serafimov initiated
a large effort in Russia (USSR) on the development and application of qualitative analysis of
multicomponent nonideal and azeotropic distillation. One reason for the large interest in this
”pen-and-paper” approach in Russia in the 1960s and 70s was the fact that during these years
the Russian chemical industry expanded whereas there was a shortage of computers and limited
possibilities for numerical computation.

3.1 Introduction

The contribution of Doherty and Perkins and co-workers (1978a; 1978b; 1979b; 1979a; 1985b;
1985a; 1985; 1985) formed the basis for a renewed interest in this subject also in the West. The
reason for this renewed interest was the realization that, in spite of (or maybe because of) the great
advances in vapor-liquid equilibrium calculations and simulations, there was a need for simpler
tools to understand the fundamental limitations and possibilities in distillation of azeotropic mix-
tures. These tools are today well-established, and residue curve map analysis is included in the
main recent encyclopedia in chemical engineering (Perry, 1997; Stichlmair, 1988; Doherty and
Knapp, 1993). Four review papers have been published during the last few years (P¨ollmann and
Blass, 1994; Fien and Liu, 1994; Widagdo and Seider, 1996; Westerberg and Wahnschafft, 1996).
Furthermore, the topic is included in recent textbooks (Biegler et al., 1997; Seader and Hen-
ley, 1998; Stichlmair and Fair, 1998) as a fundamental part of chemical engineering.

Nevertheless, the field is still rather bewildering and we believe that many of the recent results
have not been put into proper perspective and compared to earlier work, especially to work in the
Russian literature. The main objective of this paper is therefore to present a broad survey of the
field, which includes references also to papers that are less recognized in the English-language
literature.

The theory of thermodynamic topological analysis (TTA) of distillation can be divided into two
parts.

1. The first considers in detail the feasible structures of VLE diagrams (including their classi-

fication). The main concept is that there is a unique and relatively simple correspondence
between the vapor-liquid equilibrium (VLE) characteristics of a given mixture and the path
of equilibrium phase transformations such as residue curves and distillation lines.

2. The second part of the TTA is directed to the prediction of feasible separations in distilla-

tion, for which the main concepts are:

There is a unique correspondence between the state of a distillation column at extreme
operating parameters (infinite flows or infinite stages) and the composition trajectories
of simple phase transformations (e.g., residue curves). Consequently, these column
states can be determined from specific VLE characteristics of the mixture to be sepa-
rated depending on the feed composition of the mixture.

The state of a distillation column at real operating parameters can be qualitatively
determined based on the column states at extreme operating parameters, that is, from
the composition trajectories of the simple equilibrium phase transformations such as
residue curve maps.

As a result, the product composition areas for all combinations of the operating pa-
rameters can be predicted qualitatively based on information about the simple phase
transformation map (e.g., residue curve map) and, thus, the feasible separations and
separation sequences can be determined.

Therefore, analysis of VLE diagrams is the starting point for the prediction of feasible separations
by distillation. It allows us to determine the thermodynamic possibilities and limitations of the

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

separation caused by the nature of the mixture. After determining the feasible separations, we can
synthesize the alternative separation sequences that should be subject to further investigation and
comparison in order to choose the optimal one. Such an approach is called analysis-driven distil-
lation synthesis (Westerberg and Wahnschafft, 1996). As a result of the analysis it may turn out
that the mixture cannot be separated by conventional distillation. Then, special distillation meth-
ods may be employed. TTA (or residue curve map analysis) also provides a tool for evaluation of
these separation techniques. In particular, TTA is very useful for the screening of entrainers for
heteroazeotropic and extractive distillation.

Comment on terminology

There are many alternative and even conflicting key terms in the distillation literature. Table 3.1
gives the correspondence between four central terms in English-language, Russian and German
literature. Note in particular that in many Russian publications “distillation lines” means residue
curves, “c-lines” means distillation lines and “node” means vector. One reason for this incom-
patible terminology is due to the historical fact that the term distillation was originally used to
denote simple distillation (open evaporation, Rayleigh distillation). Schreinemakers (1901b) and
succeeding researchers used the term “distillation lines” as a short term for simple distillation
liquid composition trajectories, which was later termed residue curves by Doherty and Perkins
(1978a). Today, the term distillation has become a common, simplified term for distillation with
the use of reflux (rectification, fractional distillation). Stichlmair (1988) uses the term distillation
line to denote the total reflux composition profile in an equilibrium staged distillation column.
The term “node” is just a result of poor English language translation. In the present text we use
the terms that are broadly accepted in English-language distillation publications. See section on
Definitions of Terms for more detail.

Table 3.1: Correspondence between terms in English-, Russian- and German-language literature

Current terms commonly

Equivalent terms mainly

Example of reference

found in English-language

found in Russian- and

where this term is used

literature (and in this paper)

German-language literature

tie-line

Zharov and Serafimov (1975)

equilibrium vector

node line

Serafimov (1996)

mapping vector

Widagdo and Seider (1996)

connecting line (c-line)

Zharov (1968c)

distillation line

chain of conjugated tie-lines

Zharov (1968c)

(continuous and discrete)

tie-line curve

Westerberg (1997)

equilibrium rectification line

Pelkonen et al. (1997)

distillation line of the residue

Schreinemakers (1901c)

residue curve

Ostwald (1902)

distillation line

Serafimov (1968b)

distillation line of the vapor

Schreinemakers (1901b)

distillate curve

vapor line

Bushmakin and Kish (1957a)

boil-off curve

Fidkowski et al. (1993)

3.1 Introduction

Structure of the survey

This survey is devoted to the first part of thermodynamic topological analysis, that is, to the
characteristics of vapor-liquid equilibrium diagrams. The thermodynamic basis of vapor-liquid
equilibrium (VLE) and ways to graphically represent the VLE of a mixture are briefly given in
Section 3.2. The various representations of the VLE for ternary mixtures are considered in de-
tail in the subsequent sections: isotherm maps (Section 3.3); residue curve maps and evaporation
distillate curve maps (Section 3.4); open condensation curve maps and distillation line maps (Sec-
tion 3.5). These sections also include the structures of these maps and the relationship between
the different representations. Some peculiarities of the simple equilibrium phase transformation
maps for heterogeneous mixtures are considered in Section 3.6.

Section 3.7 is devoted to the classification of VLE diagrams for ternary mixtures and related
issues such as completeness of classification, occurrence of the various structure classes in na-
ture, and more. Determination of the VLE diagram structures based on incomplete information
is discussed. In Section 3.8, unidistribution and univolatility line diagrams and their role in the
determination of the geometry (pathway) of the residue curve and distillation line maps are con-
sidered. Finally, the overall conclusions are given in Section 3.9.

Contributions

Fundamental theory on structural properties of VLE diagrams for azeotropic mixtures are re-
viewed. In addition, some new results and concepts are presented for the first time. In particular:

A simple rule on residue curve regions is formulated: the composition space splits into the
same number of residue curve regions as there are repeated nodes in the system. For exam-
ple, if the residue curve system has two stable nodes and one unstable node, the number of
nodes of the same type is two, and the residue curve map is split into two regions.

The relative location of the region boundaries (separatrices) of residue curves, open con-
densation curves and distillation lines in the composition space are presented.

The thermodynamical meaning of distillation lines are emphasized.

Peculiarities of residue curves for heterogeneous mixtures are examined. Although the
topology of the residue curve maps generally do not differ from homogeneous mixtures,
and can be determined in a similar way from the shape of the boiling temperature surface,
there are some peculiarities of the ”inner” topology in heterogeneous mixtures.

A table of correspondence between the different classifications of ternary VLE diagrams is
given. The classifications by Gurikov (1958), Serafimov (1970b), Zharov and Serafimov
(1975) and Matsuyama and Nishimura (1977) are included.

The question of the existence of all the classified structures is evaluated. Although a struc-
ture may be thermodynamically and topologically feasible, its occurrence is determined by

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

the probability of certain combinations of molecular interactions. We present data on the
reported occurrence of ternary mixtures, and reveal factors that limit the natural occurrence
of mixtures with certain VLE diagram structures.

The problem of indeterminacy in predicting ternary VLE diagram structures based on in-
complete information (binary or experimental data only) is critically considered.

Feasible patterns of the VLE functions for binary mixtures represented by the equilibrium
line, distribution coefficients and relative volatility depending on the molecular interactions
are presented.

The structures of unidistribution and univolatility line diagrams for ternary mixtures of
different classes and types, and the relation to the shape of simple equilibrium phase trans-
formation trajectories such as residue curves and distillation lines, are considered in detail.
We show that a combined diagram of these unitylines can successfully be used as a char-
acteristic of the VLE. These maps are easy to generate and a lot of information can be
retrieved from them.

Rules on the location of the inflection point curves in the composition space are given.

3.2

Thermodynamic basis

The fundamental laws of thermodynamics and phase equilibrium is all that is needed to derive
the results presented in this paper, and in this section we briefly review the thermodynamic basis.
From thermodynamics, the compositions of liquid and vapor in phase equilibrium are determined
by the vapor-liquid equilibrium (VLE) condition which may be expressed as:

;

(3.1)

where

and

are the liquid and vapor compositions, and

and

are the system pressure and

temperature, respectively.

and

are not independent at the equilibrium state since:

i=1

(3.2)

For example,

;

]

may be determined as a function of

;

. Thus, at isobaric conditions

we may use only the liquid composition

as the independent variable, and we may write (“E-

mapping”):

(x);

(x)

)

(3.3)

where

is the equilibrium mapping function that assigns a composition in the liquid phase to

the corresponding equilibrium vapor phase composition, and

and

are the mixture boiling

3.2 Thermodynamic basis

temperature (bubble-point) and condensation temperature (dew-point), respectively. Equivalently,
we may write (“C-mapping”):

);

)

(x)

(3.4)

where

is the inverse equilibrium mapping function that assigns a composition in the

vapor phase to the corresponding equilibrium liquid phase composition.

For any liquid composition

, there is a point

;

]

on the boiling temperature surface

(x)

and

a corresponding point

;

]

on the condensation temperature surface

)

that are connected

by an equilibrium vector (also called a vapor-liquid tie-line). The projection of this equilibrium
vector onto the composition space represents the equilibrium mapping vector

, that is, the

graph of the function

. In this sense, the condensation temperature surface

)

simply an equilibrium

-mapping of the boiling temperature surface

(x)

. The two tempera-

ture surfaces merge at the points of pure components where

for component

and

for all other components

. According to Gibbs-Konovalov Law formulated in

the 1880s (Prigogine and Defay, 1954; Tester and Modell, 1997), the existence of an extremum
of the boiling temperature function

(x)

leads to an extremum of the condensation tempera-

ture function

)

. At the extremal points, the liquid and its equilibrium vapor compositions

are equal and the temperature surfaces are in contact. The existence of such extrema is called
azeotropy, and the corresponding compositions where

are called azeotropes. Mixtures

that form azeotropes are called azeotropic, and mixtures that do not form azeotropes are called
zeotropic.

The existence of azeotropes complicates the shape of the boiling and condensation temperature
surfaces and the structure of the vapor-liquid envelope between them, and the equilibrium map-
ping functions. The envelope of the equilibrium temperature surfaces defines the feasible operat-
ing region in

;

space in which any real distillation process must operate. This motivates

a more careful analysis of the VLE behavior. For the prediction of feasible separations upon dis-
tillation of azeotropic mixtures we need a qualitative characterization of the VLE, preferable a
graphical representation.

Graphical VLE representation

The possibility to graphically represent the VLE depends on the number of components in the
mixture. In a mixture of

components, the composition space is (

)-dimensional because

the sum of molefractions must be equal to unity. For binary mixtures the composition space is
one-dimensional. Graphical representations of the VLE for the most common types of binary
mixtures are presented in Figure 3.1. The left part of Figure 3.1 shows a combined graph of the
boiling and condensation temperatures and the vapor-liquid equilibrium phase mapping, which
gives a complete representation of the VLE. In addition, the right part gives the equilibrium phase
mapping alone. Each of these diagrams uniquely characterizes the type of the mixture (zeotropic
or azeotropic, minimum- or maximum-boiling etc.), and, what is more important, we do not need
the exact trajectories of the functions to make this characterization. It is sufficient to know the

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

Liquid

Vapor

Liquid

Liquid Liquid

Vapor

Liquid

Vapor

x ;y

Figure 3.1: Graphical representations of the VLE for the most common types of binary mixtures
at constant pressure: a) zeotropic; b) minimum-boiling homoazeotrope; c) minimum-boiling het-
eroazeotrope; d) maximum-boiling azeotrope. Left: boiling temperature

and condensation

temperature

and the equilibrium mapping vectors in

;

space. Right:

relation-

ship (equilibrium line).

boiling points of the pure components and azeotropes, if any, to determine the type of the mixture
and to qualitatively predict the distillation behavior for various feed compositions of the mixture.

For ternary mixtures the composition space is a two-dimensional normalized space and its graph-
ical representation is an equilateral or rectangular triangle. The VLE functions are surfaces in a
three-dimensional prism. The complete representation of the VLE is the vapor-liquid envelope

3.2 Thermodynamic basis

of the boiling and condensation temperature surfaces, and the set of equilibrium vectors between
them, as illustrated in Figure 3.2a. The equilibrium mapping function itself can be represented by
pair of the surfaces

(x)

and

(x)

in the prism

, as illustrated in Figure 3.2b. It is difficult

x =1

y =f(x)

T ( )

x ; y

; y

y'''

x'''

X ;

y (

y ( )

Figure 3.2: Graphical representation of the VLE for a ternary zeotropic mixture: a) boiling and
condensation temperature surfaces and the connecting equilibrium vectors; b) equilibrium vector
surfaces.

to interpret three-dimensional surfaces in a two-dimensional representation, thus other represen-
tations are preferred for ternary mixtures. The most straightforward approach is to use contour
plots of the surfaces similar to topographic maps with isolines of constant values of the functions
to be represented. Such isoline maps, specially isotherm map of the boiling temperature, play
an important role in the qualitative analysis of VLE diagrams. Isotherm maps are considered in
Section 3.3.

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

The dependence of the equilibrium vapor composition on the liquid composition given by the
equilibrium

-mapping function can be graphically represented by a field of equilibrium vec-

tors in the composition plane. Such vector fields usually appear rather chaotic (see, for example,
the equilibrium vector fields for ternary mixtures in Gmehling’s ”Vapor-Liquid Equilibrium Data
Collection”). They do not give a clear graphical representation of the VLE. In addition, this rep-
resentation has the disadvantage of being discrete and requires a great number of (experimental)
data to draw it.

An alternative method to graphically represent the VLE for a given mixture is by using simple
equilibrium phase transformations such as open evaporation (simple distillation), open condensa-
tion and repeated equilibrium phase mapping. The liquid and vapor composition changes during
these transformations can be represented by trajectories in the composition space. The pathway
of these trajectories depends on the vapor-liquid equilibrium only, and in turn characterizes the
VLE. This is the topic of Sections 3.4 and 3.5.

For multicomponent mixtures (

) the composition space is (

-1)-dimensional and its graph-

ical image is a polyhedron (tetrahedron for

, pentahedron for

and so on). In the

case of quaternary mixtures it is possible to graphically represent the VLE functions in a com-
position tetrahedron, but it may be complicated to interpret. If the number of components is
more than four, graphical representation is difficult. However, a multicomponent mixture may be
divided into ternary subsystems of components and pseudo-components. The ternary VLE sub-
diagrams are evaluated as a whole. Alternatively, a multicomponent mixture may be represented
by a pseudo-ternary or quaternary system of three or four of its key components. Obviously, VLE
information may be used without graphical representation too.

In conclusion, most of the studies on the qualitative analysis of VLE are focused on ternary
mixtures where graphical representations are readily visible and can be used for the prediction of
distillation behavior.

3.3

Isotherm maps of ternary mixtures

Throughout this paper we assume constant pressure. Isotherms are contour lines of constant tem-
perature (also called isotherm-isobars). Isotherm maps as a way to represent the system of VLE
functions Equations (3.3) and (3.4) were first used by Schreinemakers (1901b). Isotherms of the
boiling temperature surface

(x)

and the isotherms of the condensation temperature surface

)

were later called liquid isotherms and vapor isotherms, respectively, by Reinders and

de Minjer (1940). The liquid compositions of a liquid isotherm are connected with the vapor
compositions of the corresponding vapor isotherm (at the same temperature) by the equilibrium
vectors in the composition space

, see Figure 3.3. Such a diagram is a complete graphical rep-

resentation of the VLE system Equation (3.4). Schreinemakers (1901c) studied complete graphi-
cal representations of the VLE system Equation (3.3) for various combinations of a given binary
azeotropic mixture with a third component that forms additional binary and ternary azeotropes.
He also showed the analogy between an isobar map at constant temperature and the isotherm

3.3 Isotherm maps of ternary mixtures

Vapor

Liquid

; y

x ; y

T=const

T ( )

Figure 3.3: Liquid and vapor isotherms and the equilibrium vectors that connects them for a
ternary zeotropic mixture with pure component boiling points T

, T

and T

at constant pressure

(Schreinemakers, 1901b).

maps at constant pressure. Isotherm maps are presented in Figure 3.4 as they are given in the
original publication by Schreinemakers (1901b). From this study, Schreinemakers established
that:

1. Ternary mixtures may exhibit distinctly different isotherm maps depending on the exis-

tence of azeotropes, their type (binary or ternary, minimum-, maximum- or intermediate-
boiling), their location, and the relative order of the boiling points of the pure components
and azeotropes.

2. A binary azeotrope that is an extremal point of the binary mixture temperature functions, is

not necessary a global temperature extreme for the ternary mixture as a whole, thus being
a saddle point.

3. In the same way, a ternary azeotrope is not necessary a global temperature extreme for the

ternary mixture, thus being a saddle point (minimax) of the temperature surface.

The examples of isotherm maps given in Figure 3.4 are only for purposes of illustration. In
practice, it is difficult to draw such combined diagrams of liquid and vapor isotherms (at constant
pressure) because an isotherm of one phase may be (and usually is) intersected by isotherms of the
other phase at other temperatures. Equilibrium vector field diagrams are also confusing since the
equilibrium vector projections in the composition space intersect each other. The more detailed
diagram we want to draw, the less understandable it becomes. Isotherm maps, for example as a
representation of the boiling temperature function, are more readable. But, with the exception
of binary mixtures, it is difficult to immediately understand the distillation behavior of ternary

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

80 C

90 C

100 C

80 C

90 C

70 C

75 C

85 C

65 C

70 C

90 C

100 C

75 C

90 C

80 C

90 C

75 C

70 C

78 C

90 C

80 C

75 C

85 C

Figure 3.4: Liquid and vapor isotherms and equilibrium vectors in the composition space at con-
stant system pressure: a,b,c,d) examples from Schreinemakers (1901b); e,f) additional examples.

mixtures from the liquid isotherm map without special knowledge. Schreinemakers demonstrated
the severity of the problem, and it has taken the efforts of many investigators to find a proper way
of characterizing vapor-liquid equilibrium diagrams.

The key to a qualitative characterization of the VLE lies with the analysis of simple equilibrium
phase transformation processes, and above all of open evaporation (simple distillation) which is
the topic of the next section.

3.4 Open evaporation

3.4

Open evaporation

3.4.1

Residue curves

Open evaporation, also known as simple distillation or Rayleigh distillation, is batch distillation
with one equilibrium stage where the vapor formed is continuously removed so that the vapor at
any instant is in equilibrium with the still-pot liquid (residue). Schreinemakers (1901c) considered
the instantaneous mass balance equation of this process:

d(Lx

)

Ldx

)

(3.5)

where

[mol] is the amount of the residue liquid in the still-pot,

[mol] is the amount vapor

evaporated,

and

are the mole fractions of the component

in the still-pot liquid and in the

vapor, respectively. Schreinemakers called the liquid composition trajectory of simple distillation
a ”distillation line”. The present term used in English is residue curve (Doherty and Perkins,
1978a). Schreinemakers proposed for ternary mixtures to consider the ratio of still-pot liquid
components:

)

(3.6)

Integrating Equation (3.6) from any initial composition

will generate a residue curve. It is

evident from Equation (3.6) that the residue curve paths are governed by the VLE solely, that is,
by the set of functions in system Equation (3.3), and, in turn, can characterize the VLE of any
given mixture. We also see from Equation (3.6) that the equilibrium vector in the composition
space

is tangent to the line of still-pot liquid composition change at any point

Schreinemakers established that the interior of the composition space is filled in with residue
curves. The points of components and azeotropes are the isolated residue curves, and the edges
of the composition space between the singular points are residue curves too. He analyzed the
relation between the position of the equilibrium vectors at the liquid isotherms and the path of the
residue curves. This connection is presented in Figure 3.5. Based on this, he derived what in the
Russian literature is referred to as Schreinemakers’ rules:

1. a residue curve always moves along the boiling temperature surface in the direction of

increasing temperature and cannot intersect twice with the same liquid isotherm;

2. residue curves cannot intersect each other.

The path of the residue curves can be determined from the isotherm map, but the intersection angle
between the liquid isotherm and the residue curve does not depend on the shape of the isotherm but
rather on the direction of the equilibrium vector at this point. Hence, the path of residue curves
can only be qualitatively determined from the isotherm map. A residue curve map is a more
legible representation of the VLE than isotherms because from this diagram we can immediately

3.4 Open evaporation

90 C

80 C

90 C

80 C

100 C

70 C

85 C

75 C

85 C

70 C

65 C

78 C

90 C

Figure 3.6: Residue curve maps for the mixtures whose isotherm maps are presented in Figure
3.4: a,b,c,d) examples from Schreinemakers (1901b); e,f) additional examples.

3.4.2

Properties of residue curves’ singular points

Haase (1949) studied the behavior (paths) of the residue curves near the vertices of the composi-
tion triangle and the azeotropic points (i.e., near the residue curves’ singular points). In particular,
he found that a pure component vertex is an initial or a final point (node) of the residue curves if
the liquid boiling temperature increases or decreases (near it) by the movement from the vertex
along both adjoining edges. At other singular points (saddles), the residue curves have a hyper-
bolic shape in the vicinity of the vertex if the liquid boiling point temperature increases at the
movement from the vertex along one of the edges and decreases at the movement along another
edge. This observations are known in Russian literature as Haase’s rule. In summary, the residue

3.4 Open evaporation

autonomous differential equations:

)

:::;

(3.7)

Here

is a dimensionless time describing the relative loss of the liquid in the still-pot;

. Note that Zharov and Serafimov use opposite sign of the Equation (3.7) according to their

definition of

, and thus the sign of the eigenvalues are opposite in their thermodynamic topo-

logical analyses. Fortunately, the results on the properties of the residue curves’ singular points
are independent of this choice of preserving the minus sign in Equation (3.7) or not. By solving
Equation (3.7) together with the Van der Waals-Storonkin condition of the coexistence of phases
(a generalization of Van der Waals’ equation of two-component two-phase system and Clausius’
equation for one-component system for the general case of multicomponent multiphase system),
Zharov proved that the Schreinemakers’ rules are valid also for multicomponent mixtures. The
right-hand side of Equation (3.7) is equal to zero at the points of pure components and azeotropes,
and the system has no other singular points. From the theory of differential equations it is known
that the type of a singular point depends on the signs of the eigenvalues of Equation (3.7) in the
vicinity of the singular point. From this, Zharov showed that the singular point of Equation (3.7)
can only be a generalized node or a generalized saddle. The signs of the eigenvalues of Equation
(3.7) can also be determined from the signs of the derivatives

=Æ

in the vicinity of the singu-

lar point. The eigenvalue is negative (positive) if the liquid boiling point temperature decreases
(increases) as we move from the singular point in the direction of the eigenvalue:

If the temperature decreases in all directions (all eigenvalues are negative), the singular
point is a stable node (

If the temperature increases in all directions (all eigenvalues are positive), the singular point
is an unstable node (

If the temperature decreases in some directions and increases in other directions (eigenval-
ues of different signs), the singular point is a saddle (

Thus, the character of the behavior of the residue curves in the vicinity of a singular point uniquely
depends on the shape of the boiling temperature surface near this point.

The ten possible types of singular points for ternary mixtures given by Zharov and Serafimov
(1975) are presented in Figure 3.8. The arrows point in direction of increasing temperature,
or equivalently, in the direction of increasing

. Similar diagrams are also presented by Mat-

suyama and Nishimura (1977) and by Doherty and Caldarola (1985). Whereas Figure 3.7 is
determined from the direction of increasing temperature moving along the boiling temperature
surface

(x)

, Figure 3.8 is determined based on stability properties of the singular points.

For ternary mixtures, the type of singular points of the residue curves of any dimension (one-
component, binary or ternary) is characterized by two eigenvalues. The singular point is a stable
node if both eigenvalues are negative (Figure 3.8a, b, c left triangles SN). The singular point is an
unstable node if both eigenvalues are positive (Figure 3.8a, b, c right triangles UN). If one of the
eigenvalues is negative and the other is positive, the singular point is a saddle (Figure 3.8d, e, f).

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

Figure 3.8: Types of singular points for ternary mixtures indicated by

stable node (SN),

unstable node (UN), and

saddle (S) (Zharov and Serafimov, 1975). Arrows point in the direction

of increasing boiling temperature, and the letters (a, b, c, d, e, f) show the correspondence to
Figure 3.7.

The twenty types of singular points for quaternary mixtures that are given in Zharov and Serafi-
mov (1975) are presented in Figure 3.9. The sign of the topological indices are valid according
to the sign of Equation (3.7) given in the present text. For quaternary mixtures, the type of the
residue curves’ singular points of any dimension (one-component, binary, ternary or quaternary)
is characterized by three eigenvalues. The singular point is a stable node if all the three eigen-
values are negative (Figure 3.9a, e, k, q). The singular point is an unstable node if all three
eigenvalues are positive (Figure 3.9b, f, l, r). The singular point is a saddle if one eigenvalue is
positive and the other eigenvalues are negative (Figure 3.9c, g, i, m, o, s), or, if one eigenvalue
is negative and the other eigenvalues are positive (Figure 3.9d, h, j, n, p, t). Only elementary
singular points are presented in Figure 3.8 and 3.9.

In summary, the type of singular points (nodes and saddles) of the residue curves can be deter-
mined from the direction of increasing boiling temperature in the vicinity of these point.

3.4.3

Rule of azeotropy

The boiling temperature surface, in which simple distillation residue curves belong, is a contin-
uous surface. It follows from topological theory that a combination of singular points of such
surfaces corresponds to certain topological restrictions. Using Poincare’s theory of topological
properties of continuous surfaces, Gurikov (1958) showed that for ternary simple distillation the
combination of the singular points of different type always satisfies the rule:

(3.8)

where

)

is the number of ternary nodes (saddles),

)

is the number of binary nodes

(saddles), and

is the number of pure component nodes. All thermodynamically and topo-

3.4 Open evaporation

Figure 3.9: Types of singular points for quaternary mixtures indicated by

stable node SN,

unstable node UN, and

saddle S, and their corresponding topological indices (Zharov and

Serafimov 1975). Arrows point in the direction of increasing boiling temperature.

logically feasible combinations of the singular points of ternary residue curve maps must obey
this ”Rule of azeotropy”. Zharov (1969a) extended this consideration to multicomponent mix-
tures. Using the concept of topological index (theorems by Kronecker and Hoppf) he obtained
the generalized rule of azeotropy for

-component mixtures:

k =1

)

(

(3.9)

where

)

is the number of

-component nodes with positive (negative) index,

)

is the number of

-component saddles with positive (negative) index. The index of the singular

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

point can be determined from the signs of the eigenvalues of Equation (3.7). An index is positive
(negative) if the number of negative eigenvalues is even (odd). Gurikov’s Equation (3.8) is the
specific case of Equation (3.9) for ternary mixtures. From the topological analysis of the singular
points given in Section 3.4.2 we have for ternary mixtures that all nodes (stable or unstable, one-
component, binary or ternary) have an index

and all saddles have an index

. The numbers

of negative nodes and positive saddles are equal to zero. Thus, for ternary mixtures Equation (3.9)
has the form:

8(N

)

4(N

)

2(N

)

(3.10)

Excluding all the indices and taking into account that

we obtain Gurikov’s rule

of azeotropy given by Equation (3.8). For quaternary mixtures, the unstable and stable nodes
have different indices, and the saddles can be positive or negative. Referring to Figure 3.9, where
the different types of singular points are given for a quaternary mixture, we can determine the
topological indices given in Table 3.2.

Table 3.2: Topological indices for quaternary mixtures

Case in

Number of negative

Type of

Topological

Nomination

Figure 3.9

eigenvalues

singular point

index

a, e, k, q

stable node

-1

b, f, l, r

unstable node

c, g, i, m, o, s

saddle

d, h, j, n, p, t

saddle

-1

Another approach to the analysis of topological restriction was independently performed by Ser-
afimov (1968d; 1968b; 1968c; 1969a; 1969b; 1969c). The results of both authors were sum-
marized in Serafimov et al. (1971) and Zharov and Serafimov (1975) and recently presented in
Serafimov (1996).

In conclusion, the rule of azeotropy simply means that a definite combination of singular points
of the various types must lead to a definite system of residue curves. It is used in the topological
classification of VLE diagrams that we present in Section 3.7.

3.4.4

Structure of residue curve maps

Based on the rule of azeotropy and the postulate that residue curves must be closed (that is,
originate and end in nodes), it has been found that if we know the type of all singular points (i.e.,
the direction of the residue curves near all points of pure components and azeotropes), then we can
qualitatively construct the diagram of residue curves (Serafimov, 1968d; Zharov and Serafimov,
1975; Doherty, 1985). Some examples of qualitatively residue curve map construction are given
in Figure 3.10 for various cases of azeotropy of a ternary mixture of component 1 (light), 2
(intermediate) and 3 (heavy), where the boiling points are

. We can see that in

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

are the singular residue curves that divide the composition space into these regions. Only saddle
separatrices interior to the composition space can be residue curve boundaries.

Bushmakin and Molodenko (1957) and Doherty and Perkins (1978a) proposed to distinguish
between stable separatrices (boundaries going from a saddle to a stable node), and unstable sep-
aratrices (boundaries going from an unstable node to a saddle). In Figure 3.10 we can see the
unstable separatrices (Cases 2 and 4, separatrices 1O and AO in Case 5), and the stable separa-
trices (Case 3, separatrices O2 and O3 in Case 5). Bushmakin and Molodenko (1957) assumed
(correctly) that the composition triangle is split into residue curve regions only when there is at
least one saddle binary azeotrope, or, in other words, when there is at least one separatrix inte-
rior to the composition triangle (necessary condition). However, we can see by the example of
Case 4 in Figure 3.10 where we only have one residue curve region in spite of the existence of
an unstable separatrix, that the existence of a binary saddle (A) and its separatrix (OA) is not a
sufficient condition for the splitting of the composition space into several residue curve regions.
This may cause confusion (see, for example, Safrit (1996)), but the ”contradiction” has a simple
answer. Strictly speaking, the separatrix itself is a simple distillation region and the existence of
the binary saddle leads to the splitting of the interior of the composition triangle into two ”re-
gions” with distinct final points: (1) the separatrix itself and (2) the set of the liquid compositions
belonging to all other non-singular residue curves. Here, the separatrix is not a boundary because
it does not correspond to the adopted definition of a boundary (it does not divide the family of the
residue curves into different bundles).

Thus, a boundary is a separatrix but a separatrix is not necessarily a boundary. We can formulate
the general rule:

The composition space is split into several residue curve regions if there are several unstable
or several stable nodes (necessary condition).

The number of residue curve regions equals the total number of repeated nodes, that is,
nodes of the same type (stable or unstable). For example, if we have three stable nodes and
one unstable node the total number of repeated nodes is three and we have three residue
curve regions. If we have two stable nodes and two unstable nodes the total number of
repeated nodes is four and we have four residue curve regions.

Here, a residue curve region is defined as the set of residue curves with common initial and final
points. Doherty and Perkins (1978a) proposed another definition of residue curve regions as the
set of liquid compositions that belong to the residue curves with a common stable node. Thus, any
unstable separatrix is a residue curve region boundary, and only unstable separatrices can be the
boundaries of residue curve regions. We can see from Figure 3.10 that both statements are invalid.
In Case 4 there is an unstable separatrix which is not a residue curve boundary. Furthermore, Case
3 has two residue curve regions separated by a stable separatrix boundary. The definitions were
corrected by Doherty et al. (1985; 1993), but the error appeared again in Rev (1992) and in the
review paper by Fien and Liu (1994).

It should be stressed that we cannot construct an exact map of residue curves for a given mixture
without knowing the (exact) location and shape of the separatrices (in addition to knowing the

3.4 Open evaporation

initial and terminal points), and the exact location and shape of ordinary residue curves. Thus, we
hereafter distinguish between the residue curve map, sketch of the residue curve map and structure
of the residue curve map. The first term means the exact representation of the residue curves and
must be constructed by integration of the simple distillation residue curve Equation (3.7). The
term ”sketch” means the qualitative diagram of residue curves representing the sign of curvature
(convex/concave) and the existence (or nonexistence) of inflection points (qualitatively), but not
the exact location of the residue curves. This is the focus in Section 3.7.4. Finally, the term
”structure of the residue curve map” is the simplest graphical representation that characterizes
the VLE and simple distillation of a given mixture (Serafimov, 1968d). The structural diagram
represent the topology of the residue curve map as given by the residue curves’ singular points
(nodes and saddles), the residue curves going along the edges of the composition triangle, and
the saddles separatrices located interior to the composition space (Serafimov, 1968d), that is,
only the the singular residue curves. For example, the dashed residue curves in Figure 3.10 are
unnecessary for the characterization of the structure.

As mentioned earlier, using information about the shape of the boiling temperature surface near
the singular points we can only determine the structure of the residue curve map. Since the
residue curves follow the direction of increasing boiling temperature, the structure of the residue
curve map is also the structure of the boiling temperature surface. Representing the VLE diagram
structure by an oriented graph of the boiling temperature increase allows us to determine the
structure of the diagrams for multicomponent mixtures. Algorithms for this were proposed by
Vitman and Zharov (1971b; 1971a) and Babich (1972). These oriented graphs may be visualized
for quaternary mixtures, but not readily if the number of components is more than four. Instead,
the graphs can be represented in matrix form and the determination of the VLE diagram structure
can be computerized. Algorithms for computer-aided determination of multicomponent VLE
diagrams structures were developed by Petlyuk et al. (1975b; 1975a; 1977a; 1977c; 1977b; 1978),
and are also given by Julka and Doherty (1993), Rooks et al. (1998) and Safrit and Westerberg
(1997a) and others.

3.4.5

Separatrices of residue curves and flexure of the boiling temperature
surface

A surface can be characterized by its topology. As just noted the topology of the residue curve
diagram and the boiling temperature surface coincide. That is, their stable nodes (peaks), unstable
nodes (pits) and saddle points are the same. Thus, it may appear that the stable separatrices of
a residue curve diagram are coincident with the projections of the locus of the ridges in the
boiling temperature surface onto the composition simplex, and the unstable separatrices with the
valleys. This was claimed by Reinders and de Minjer (1940) based on the intuitive assumption that
residue curves could not cross the ridges or valleys in the boiling temperature function

(x)

since this would violate the Schreinemakers’ rules. Opposed to this, Haase (1949) stated that the
location of the flexures (ridges and valleys) in the boiling temperature surface cannot coincide
with the boundaries of the simple distillation regions. Haase was right, however, this point was
not immediately recognized.

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

Based on experiments, Bushmakin and Kish (1957a) argued that for the mixture methyl acetate-
chloroform-methanol the boundaries of simple distillation regions run along the projections of
the flexures in the boiling temperature surface on the plane of the composition triangle ”within
the limits of the experimental accuracy”. However, they did not describe how they localized the
flexures, nor did they give a quantitative comparison of the path of the boiling temperature surface
flexures and the simple distillation region boundaries. It should be noted that Bushmakin and Kish
did not vigorously claim the identity of the boundaries with the flexures in the boiling temperature
surface in the general case, but wrote about the correspondence between them (Bushmakin and
Molodenko, 1957). Nevertheless, these papers supported the hypothesis by Reinders and De
Minjer resulting in its added confidence. Numerous attempts were subsequently made to prove the
identity of the simple distillation boundaries with the flexures in the boiling temperature surface.
The interest in this problem is understandable. Although a separatrix is the singular line of the
residue curve family, its equation is not distinct from the equation of the ordinary residue curves.
Under integration of Equation (3.7), the separatrix can be localized only as a line which arrives at

-vicinity of a saddle at

, or, starts from

-vicinity of a saddle at

. In other words,

the localization is only approximate. If the separatrix had the specific property that it coincides
with the flexure of the boiling temperature function it could be exactly localized in a simple way.
Various ways to determine the flexure (ridges and valleys) in the boiling temperature surface have
been proposed:

As the locus of extremal boiling temperatures when moving along straight lines parallel to
the edges of the composition triangle (see, for example, Doherty and Perkins, 1978a);

As the separatrix of the steepest gradient lines (Malenko, 1970; Kogan and Kafarov, 1977a;
Kogan and Kafarov, 1977b; Van Dongen and Doherty, 1984);

As the locus of liquid composition points with maximum isotherm curvature (Boyarinov
et al., 1974). Path of the largest gradient on the boiling point surface (Stichlmair and
Fair, 1998);

As the locus of liquid composition points where the equilibrium vector is co-linear with the
temperature gradient (Sobolev et al. (1980)).

However, as it was demonstrated by Van Dongen and Doherty (1984) and Rev (1992), these flex-
ures do not coincide with the separatrices of residue curves. This finally confirmed the conclusion
made by Haase as early as 1949.

This conclusion is not surprising since the pathways of the residue curves do not only depend
on the boiling temperature surface, but also on the direction of the equilibrium vectors between
the phases. The direction of the equilibrium vectors are not necessarily the same as that of the
temperature gradient.

Kiva and Serafimov (1973) showed that residue curves crossing the valley or ridges of the boil-
ing temperature surface do not violate Schreinemakers’ rules since the residue curve ” [...] rises
along the slope, intersects the fold [ridge] at an acute angle, and, [...] continues to rise, moving

3.5 Other simple equilibrium phase transformations

along the opposite slope of the curved fold [ridge] to its highest point”. Even when a residue
curve crosses a ridge or valley, the residue temperature is continuously increasing. Similar con-
siderations are given in more detail by Rev (1992).

In conclusion, although the structure of a residue curve map can be determined solely from the
shape of the boiling temperature surface, the (exact) location of the residue curve separatrices
cannot be determined from these data.

3.4.6

Structure of distillate curve maps

The vapor composition trajectory during open evaporation (simple distillation) can be represented
graphically by the ”vapor line” (Schreinemakers, 1901c), later named a distillate curve by Do-
herty and Perkins (1978a). Each distillate curve is connected to a residue curve and moves along
the condensation temperature surface

)

. The distillate curves are going through the vapor

ends of the equilibrium vectors that are tangent to the corresponding residue curve. The projec-
tion of the distillate curve in the composition space is always positioned on the convex side of
the corresponding residue curve projection. The less the curvature of the residue curve, the less
is the gap between it and the corresponding distillate curve. The vapor condensation tempera-
ture is equal to the liquid boiling temperature at any instant and increases monotonically during
the open equilibrium evaporation process. The system of distillate curves has the same singular
points as the system of residue curves, and the singular point type is the same as the type of the
corresponding singular point of the residue curves (Haase, 1950b; Storonkin, 1967).

3.5

Other simple equilibrium phase transformations

3.5.1

Open condensation

Open condensation is another equilibrium phase process that solely depends on the VLE (no
design or operating parameters) and can in turn characterize it. Open condensation, or simple
condensation, is a hypothetical process where drops of condensate are continuously removed
from a bulk of vapor in such a way that the composition of this liquid drop of condensate is in
equilibrium with the still-pot vapor composition, at any instant. This process is symmetrical to
the simple distillation process (Kiva and Serafimov, 1973c). It may appear that the open conden-
sation curves are directed opposite to the simple distillation residue curves, but this is not correct.
Only the structure is the same, but they do not coincide in general. It is evident from Figure 3.11,
where simple distillation residue curves and open condensation curves going through points

and

of an arbitrary equilibrium vector are given. Nevertheless, open condensation and open

evaporation (simple distillation) are symmetrical processes with isomorphus composition trajec-
tory maps for antipodal VLE diagrams (diagrams with inverted signs of the nodes and opposite
directed trajectories).

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

Figure 3.11: Vapor and liquid composition trajectories of open evaporation and open condensa-
tion going through the points

and

of an equilibrium vector in the composition space.

3.5.2

Repeated equilibrium phase mapping and distillation lines

Zharov (1968c; 1969b) proposed the use of repeated equilibrium

-mapping as a characterization

of the VLE:

;

:::

;

(3.11)

Zharov named the sequence

a ”chain of conjugated tie-lines [equilibrium vectors]”, today

often referred to as tie-line curves or discrete distillation lines. The repeated equilibrium phase
mapping is illustrated in Figure 3.12. The chain describes the stepwise movement along both
the boiling and condensation surfaces similar to the McCabe-Thiele approach for binary mixtures
at total reflux. Starting with a given liquid composition

, the equilibrium relationship (

mapping) determines the corresponding equilibrium vapor composition

. The vapor is totally

condensed and, thus, the vapor and liquid compositions are equal at each step,

, which

gives the initial point

of the next step and so forth. The VLE describes movement horizontally

(

constant) in Figure 3.12. Projection of a sequence of equilibrium phase mappings onto the

composition plane presents a discrete line. When there is a number of repeated mappings, the
sequence strives to the component or azeotropic point that has a minimum boiling temperature.
The sequence will follow the direction of decreasing boiling (condensation) temperature, that is,
it has the same direction as the composition trajectories of open condensation. Choosing several
different initial points

, we will obtain a set of equilibrium vector chains. With the proper

choice of the initial points we obtain a map of the chains of conjugated equilibrium vectors in the
composition space that characterizes the equilibrium mapping function

. The set of

these equilibrium vector chains is an organized selection from the equilibrium vector field. Zharov
proposed to consider smooth ”connecting lines” instead of the discrete chains of conjugated
equilibrium vectors. The term connecting line (“c-line”) means a smooth line going in such a
way that the vapor-liquid equilibrium vector at an arbitrary point is a chord of this line. As result,
any connecting line is a common arc of infinite equilibrium mappings as illustrated in Figure
3.13. Today, these lines are commonly named distillation lines. For homogeneous mixtures
the distillation lines (connecting lines) cannot intersect each other (in contrast to the chains of

3.5 Other simple equilibrium phase transformations

x ;

; y

T ( )

Figure 3.12: Equilibrium distillation line (also named tie-line curve) is a chain of conjugated
vapor-liquid equilibrium vectors in the composition space (a sequence of repeated equilibrium
phase mapping).

Figure 3.13: Distillation line (connecting line) and tie-line curves (chains of conjugated vapor-
liquid equilibrium vectors).

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

conjugated vectors in the composition space as shown in Figure 3.13) and qualitative differential
equation theory can be used for their analysis. Based on this analysis, Zharov showed that the
properties of distillation lines near the pure component and azeotropic points are identical with
those of the residue curves, but the signs of the eigenvalues and indices of the singular points are
inverted according to their opposite direction (which is just a matter of definition). As a result,
the structure of the distillation line map (and of the conjugated vector chain map too) can be
determined from the shape of the boiling temperature surface in the same way as the structure of
the (simple distillation) residue curve map.

Zharov (1968c) noted that the discrete distillation line coincides with the composition profile of
an equilibrium staged distillation column at total reflux. Based on this coincidence, Stichlmair
(1988; 1989) uses the name distillation lines for the connection lines of conjugated vapor-liquid
equilibrium vectors. Wahnschafft et al. (1992) named the discrete chain of conjugated equilibrium
vectors a tie-line curve. These terms are in common use in English-language literature, and
we will consequently use them in the present text. Wahnschafft et al. (1992) and Widagdo and
Seider (1996) define a tie-line curve and a distillation line as the liquid composition profile of
a distillation column at total reflux. Wahnschafft et al. noted that these trajectories ”are in
principle dependent of tray efficiency”. This is true with regard to a column composition profile.
However, they concluded that tie-line curves and distillation lines cannot be uniquely defined in
a thermodynamic sense. This is not true. Discrete distillation lines (tie-line curves) are defined
uniquely from the VLE solely, regardless of its connection with the column composition profile.
Tie-line curves are equivalent to residue curves as characteristics of the VLE in a thermodynamic
sense. Stichlmair (1988) underlines that the (discrete) distillation lines give information about
the lengths of the vapor-liquid equilibrium vectors which can characterize not only the qualitative
course of distillation, but the distillation separability as well. The larger the distance between
the points on the distillation line, the easier is the separation. Residue curves do not bring this
information.

Discrete distillation lines (tie-line curves) are readily constructed for any given mixture by com-
putation if the mathematical description of the VLE is available. This computation is quicker than
the computation of residue curves. However, the exact construction of the distillation lines is more
ambiguous because the equation of this line is not determined (not mathematically uniquely de-
fined) and various algorithms of its construction, for example as given by P¨ollmann et al. (1996),
will not be simpler than the integration procedure for the construction of the residue curves. How-
ever, the determination of the structure of the distillation line maps does not differ from the residue
curve map structures. Along with

-fold and infinity

-mappings, we can consider

-fold and

infinite

-mappings, that is, the sequence

;

:::

;

(3.12)

We name this chain of conjugated “vapor-to-liquid” vectors a reverse discrete distillation line
(reverse tie-line curve). It is apparent that the chains of conjugated vectors

and their smooth

connecting lines will follow the direction of increasing boiling (condensation) temperature. Thus,
they have the same direction as residue curves of open evaporation. For homogeneous mixtures
the reverse distillation lines are direct inversions of distillation lines. In other words, distillation

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

3.6.1

Simple phase transformations in heterogeneous mixtures

The specific features of vapor-liquid-liquid phase equilibria lead to peculiarities of the simple
phase transformations for heterogeneous mixtures. It has been shown by Haase (1950a) that the
Equation (3.6) of open evaporation residue curves for ternary mixtures can be applied to heteroge-
neous mixtures referring to the overall liquid composition. The tangents to the residue curves are
the vapor-liquid equilibrium vectors in the homogeneous region. In the heterogeneous region, the
tangents to the residue curves are the vapor-liquid equilibrium vectors in the point of the overall
liquid composition. Storonkin and Smirnova (1963) proved that the residue curves move continu-
ously in the direction of the increasing boiling temperature and cross the binodal curves without a
gap or a fracture. Timofeev et al. (1970) proved that all the conclusions concerning the behavior
of open evaporation residue curves are valid also for ternary and multicomponent heterogeneous
mixtures.

The main feature that distinguishes heterogeneous mixtures from homogeneous mixtures is that
a heterogeneous azeotrope (hereafter called a heteroazeotrope) is either an unstable node or a
saddle. It cannot be a stable node since heteroazeotropes cannot be maximum-boiling azeotropes.
This is easily explained by physical reasoning: heterogeneity (liquid-liquid phase splitting) and
minimum-boiling azeotropes occurs when the different components of the mixture repel each
other, whereas maximum-boiling azeotrope occurs when the components attract each other.

The relationship between isotherms and residue curves for ternary heterogeneous mixtures pre-
sented by Timofeev et al. (1970) is shown in Figure 3.17. Unlike homogeneous mixtures, the
isotherm passing through the point of a heteroazeotrope is always an ”isotherm of final longi-
tude”. Bushmakin’s definition of the connection between an isotherm and the patterns of the
residue curves (Section 3.4.2) should be reformulated as follows:

(a) A heteroazeotropic point is a nodal point if the isotherm passing through this point is an

”isolated” isotherm surrounded by the closed isotherms or by the isotherms which end at
the edges of the composition triangle (see Figure 3.17c and d).

(b) A heteroazeotropic point is a saddle point if the isotherm passing through this point is a

closed isotherm or an isotherm which ends at the edges of the composition triangle (see
Figure 3.17a and b).

Thus, the main relationship between the structure of the boiling temperature surface and the
structure of the residue curve map applies to both homogeneous and heterogeneous mixtures.
Pham and Doherty (1990b) arrived at the same conclusions. The topology of a residue curve map
of a heterogeneous mixture does not differ from that of a homogeneous mixture with the same set
of singular points. However, distillate curves do not satisfy the uniqueness condition in the whole
composition space. For the part of a residue curve in the heterogeneous region, the corresponding
distillate curve moves along the singular vapor line. As a result, the distillate curves partially
coincide with each other and with the singular vapor line (Storonkin and Smirnova, 1963; Pham
and Doherty, 1990b). This means that the behavior of the distillate curves will be somewhat
different from that of the residue curves for the same mixture, namely:

3.6 Heterogeneous mixtures

binodal curve

isotherm

residue curve

Saddle

Unstable node

Figure 3.17: Patterns of isotherms and residue curves near heterogeneous azeotropic points (Tim-
ofeev 1970).

(1) there is a singular line in addition to the pure component and azeotropic points;

(2) if the composition space is split into residue curve regions and the boundary is related to the

existence of the singular vapor line (this is not necessary), then the boundary of the distillate
regions coincide partially or completely with the singular vapor line, and the coinciding part
of both lines will be a manifold of contact for adjacent regions;

(3) the system of distillate curves can include a nonelementary singular point in addition to the

component and azeotropic points.

Nevertheless, the global structure of the distillate curve map (initial and final points and splitting
into regions) will be identical to the residue curve map.

3.6.2

Examples of simple equilibrium phase transformation maps

As an example, let us consider a mixture of Serafimov’s class 2.0-2b (Section 3.7.1). Three
types of heterogeneity for this structure are presented in Figure 3.18. When the heterogeneous
region includes both the unstable node

and the binary saddle

, the singular vapor line

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

Type A

Type B

Type C

binodal curve
liquid-liquid tie-line

singular vapor line
vapor-liquid equilibrium vector for plait point

Figure 3.18: Three possible types of heterogeneity for a mixture with the boiling temperature
structure of Serafimov’s class 2.0-2b. The singular points are indicated by

stable node,

unsta-

ble node, and

saddle.

moves from the unstable node to the saddle inside the heterogeneous region (Type A). When the
heterogeneous region only includes the unstable node

, the singular vapor line moves from

the unstable node in the direction of increasing temperature and has the end point

inside the

heterogeneous region (Type B). When the heterogeneous region only includes the saddle

, the

singular vapor line moves from the saddle in the direction of decreasing temperature and has the
end point

outside the heterogeneous region (Type C).

The structures of the residue curve maps for these cases are presented in Figure 3.19. In all the
cases the residue curve maps have the same structure, though the shapes of the residue curves will
differ from one another. In all cases the residue curve boundary is located at the concave side of
the singular vapor line.

The structures of the distillate curve maps are presented in Figure 3.20. For the mixture of Type
A all equilibrium vectors that extend from the residue curve separatrix terminate at the singular
vapor line. Hence, the singular vapor line is a separatrix (region boundary) of the distillate curves.

For the mixture of Type B, the equilibrium vectors arrive at the singular vapor line for the part
of the residue curve separatrix that is located in the heterogeneous region (that is, from the point

to the point of intersection of the separatrix with the binodal curve

). Hence, the part of

the singular vapor line from

will coincide with the boundary of the distillate regions,

where point

is the end point of the equilibrium vector which is tangential to the separatrix of

the residue curves at the point

For the mixture of Type C, the equilibrium vectors arrive at the singular vapor line for the part
of the residue curve separatrix that is located in the heterogeneous region (that is, from the point

to the point of intersection of the separatrix with the binodal curve

). Hence, the part of

the singular vapor line from

will coincide with the boundary of the distillate regions,

where point

is the end point of the equilibrium vector which is tangential to the separatrix of

3.6 Heterogeneous mixtures

Type A

Type B

Type C

binodal curve
singular vapor line

residue curve boundary

residue curves

Figure 3.19: Residue curve maps for the three types of heterogeneity given in Figure 3.18.

the residue curves at the point

. For both diagrams of Types B and C, point

is a point of

bifurcation of the singular vapor line and the separatrix of the distillate curves.

One can see that, unlike the residue curves, the behavior of distillate curves depends not only on
the structure of the boiling temperature surface but on the location of the binodal curve as well.
The peculiarities of the inner topology of the distillate curve maps include:

(1) distillate curves may coincide with each other along the singular vapor line;

(2) the coinciding part of singular vapor line and the distillate curve boundary is a manifold of

contact for adjacent regions;

(3) the system of distillate curves can include a nonelementary singular point in addition to the

3.7 Ternary VLE diagrams. Classification, occurrence and structure determination

3.7

Ternary VLE diagrams. Classification, occurrence and
structure determination

3.7.1

Classifications of ternary VLE diagrams

From the previous discussions we conclude that the structure of simple phase transformations
maps is determined, in general, by the shape of the boiling temperature surface for both homo-
geneous and heterogeneous mixtures. As a consequence, knowledge of the shape of the boiling
temperature surface alone permits us to characterize, distinguish and classify any ternary mixture.
Since the feasible structures must satisfy the rule of azeotropy (Section 3.4.3), this rule may be
used to classify ternary VLE diagrams. This classification was first proposed by Gurikov (1958).
Gurikov considers ternary mixtures with no more than one binary azeotrope for each binary pair
of the components and no more than one ternary azeotrope (no biazeotropy), i.e.,

and

. He denotes the total number of binary azeotropes as being

where

Substitution into Equation (3.8) gives:

(3.13)

Both

and

can take the values 0, 1, 2 or 3. From Equation (3.13) we see that the quantities

and

are bound to be of equal parity. If

is an even number (0 or 2), then

is also even.

is an odd number (1 or 3), then

is also an odd number. Gurikov uses the quantity

as a

classification parameter. Specifying the value of

he considers feasible variants of the structure

of residue curve maps for each of the two corresponding values of

. A structure is considered

feasible if it does not violate Schreinemakers’ rules (Section 3.4.1) and the condition of closed
sets of the residue curves (i.e. extending from and terminating in pure component and azeotropic
singular points). From this analysis Gurikov revealed 22 feasible topological structures of residue
curve maps (or boiling temperature surfaces).

Serafimov (1968d; 1970b) extended the work of Gurikov and used the total number of binary
azeotropes M and the number of ternary azeotropes T as classification parameters. Serafimov’s
classification denotes a structure class by the symbol ”M.T” where M can take the values 0, 1, 2
or 3 and T can take the values 0 or 1 (i.e. same assumption as Gurikov that the ternary mixture
does not exhibit biazeotropy). These classes are further divided into types and subtypes denoted
by a number and a letter. As a result of this detailed analysis, four more feasible topological
structures, not found by Gurikov, were revealed. Thus Serafimov’s classification includes 26
classes of feasible topological structures of VLE diagrams for ternary mixtures. The foundation
of this work is also presented more recently (Serafimov, 1996).

Both the classifications of Gurikov and Serafimov consider topological structures and thus do
not distinguish between antipodal (exact opposite) structures since they have the same topology.
Thus, the above classifications include ternary mixtures with opposite signs of the singular points
and opposite direction of the residue curves (antipodal diagrams). Serafimov’s classification is
presented graphically in Figure 3.21. The transition from one antipode to the other (e.g. changing
from minimum- to maximum-boiling azeotropes) can be made by simply changing the signs of

3.7 Ternary VLE diagrams. Classification, occurrence and structure determination

the nodes and inverting the direction of the arrows. As discussed in Sections 3.3 to 3.5, all
the representations of the VLE (isotherm map, residue curve map, vector field of equilibrium
vectors in the composition space) are related and are equally capable of classifying the mixtures.
Feasible structures of residue curve maps, isotherm maps and vector fields for ternary mixtures
are presented by Serafimov (1971; 1973).

The classification of ternary mixtures may be refined by distinguishing between antipodes inside
each structure class, based on the reasoning that ”minimum- and maximum-boiling azeotropes
have dissimilar physical nature and dissimilar behavior during distillation” (Zharov and Serafi-
mov, 1975). This refined classification includes a total of 49 types of feasible VLE diagrams
(p. 96-98 in Zharov and Serafimov, 1975). An even more detailed classification is proposed by
Matsuyama and Nishimura (1977). This classification is founded on the same principles, but
the diagrams are further distinguished according to the relative location of the binary azeotropic
points on the edges of the composition triangle. The components are ranked in the order of
their boiling temperature (“light”, “intermediate” and “heavy”). The classification includes 113
diagram classes of which 87 are presented graphically by Doherty and Caldarola (1985). Nev-
ertheless, among these 113 classes there are still only the 26 topologically distinct structures of
Serafimov. Matsuyama and Nishimura’s classes are denoted by three digits according to the ex-
istence and type of binary azeotropes at the edges 12, 23 and 13. The digits can take values 0
(no azeotrope), 1 (minimum-boiling azeotrope, unstable node), 2 (minimum-boiling azeotrope,
saddle), 3 (maximum-boiling azeotrope, stable node) or 4 (maximum-boiling azeotrope, saddle).
When a ternary azeotrope exists, a letter follows the three-digit classification code: m (minimum-
boiling ternary azeotrope, unstable node), M (maximum-boiling ternary azeotrope, stable node)
and S (intermediate-boiling ternary azeotrope, saddle).

In this connection we want to emphasize that the relative volatility of the components in an
azeotropic mixture change within the composition space. The terms ”light”, ”intermediate” and
”heavy” component have little meaning in general for nonideal and azeotropic mixtures. This
distinction made in Matsuyama and Nishimura’s classification does not give any additional in-
formation. Actually, it is sometimes ambiguous, as some of Matsuyama and Nishimura’s classes
where a ternary saddle azeotrope is present have two or even three possible topological struc-
tures. For example, code 112-S can be either of Serafimov’s topological class 3.1-3a or 3.1-3b
depending on whether there is a saddle - saddle separatrix or not. This ambiguity was also pointed
out by Foucher et al. (1991) who recommend an extension of the Matsuyama and Nishimura’s
classification code name in these cases. This issue is further discussed in Section 3.7.4.

The relationship between the different classifications is presented in Table 4.2 where the classes
are ordered according to Serafimov’s classification (1970b) with increasing number of azeotropes
occurring in the ternary mixture. Hereafter we use Serafimov’s (1970b) nomenclature for the
topological classes and Zharov and Serafimov’s (1975) nomenclature for the antipodal structure
types (referred to as the ZS-type).

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

Table 3.3: Correspondence between different classifications of ternary VLE diagrams

Gurikov (1958)

Serafimov (1970)

ZS-type

Matsuyama & Nishimura (1977)

22 classes

26 classes

49 classes

113 classes

000

1.0-1a

100

030

1.0-1b

001

003

1.0-2

020

400

1.1-1a

200-m

040-M

1.1-1b

002-m

004-M

1.1-2

010-S

300-S

2.0-1

031, 103, 130

2.0-2a

023, 320

401, 410

2.0-2b

11a

102, 120, 021

21a

043, 430, 403

2.0-2c

11b

201, 210, 012

21b

034, 304, 340

2.1-1

032-m, 230-m, 203-m

041-M, 140-M, 104-M

2.1-2a

16a

420-m, 402-m

16b

024-m, 420-M

2.1-2b

022-m, 220-m, 202-m

044-M, 440-M, 404-M

10a

2.1-3a

013-S, 310-S, 301-S

10b

2.1-3b

011-S, 110-S, 101-S

033-S, 330-S, 303-S

14a

3.0-1a

411

323

14b

3.0-1b

123, 321, 132, 213, 312, 231

413, 314, 431, 341, 134, 143

3.0-2

212, 122, 221

434, 344, 443

3.1-1a

27b

421-m, 412-m

32b

423-M, 324-M

3.1-1b

232-m, 223-m, 322-m

414-M, 441-M, 144-M

3.1-1c

27a

142-M, 241-M, 124-M, 214-M, 421-M, 412-M

32a

423-m, 324-m, 432-m, 342-m, 234-m, 243-m

3.1-2

222-m

444-M

3.1-3a

25a

121-S, 112-S, 211-S

38a

343-S, 334-S, 433-S

3.1-3b

25b

121-S, 112-S

38b

343-S, 334-S, 433-S

3.1-4

131-S, 113-S, 311-S

133-S, 313-S, 331-S

ZS-type refers to the refined classification on antipodal structures by Zharov and Serafimov (1975)

3.7 Ternary VLE diagrams. Classification, occurrence and structure determination

3.7.2

Completeness of classifications

The classifications presented above do not include (by assumption) biazeotropic mixtures, i.e.
mixtures where two binary azeotropes exist for one or more of the binary constituents or where
there exists two ternary azeotropes. Kogan (1971) suggested that the rule of azeotropy is not
valid for biazeotropic mixtures. However, this is not true, because the rule of azeotropy is a
topological rule based on the combination of singular points of different types for continuous
surfaces (without a gap), and its validity does not depend on the number of singular points of one
element of the surface. This fact has been proved by Komarova et al. (1974) and by Matsuyama
and Nishimura (1977), and is also discussed in Serafimov (1996). An example of a hypothetical
biazeotropic mixture is given in Figure 3.22. The set of singular points in the mixture is

Figure 3.22: The rule of azeotropy also applies to biazeotropic mixtures: Hypothetical mixture
with two ternary azeotropes and two binary azeotropes for one of the binary pairs.

and

, which satisfies the rule of azeotropy.

The classifications do not take into account structures with nonelementary singular points. The
existence of biazeotropy and nonelementary singular points of the residue curves is closely con-
nected to tangential azeotropy. The structure of residue curve maps and the rule of azeotropy for
such mixtures are considered in the papers by Serafimov (1971b; 1971a; 1971c; 1971d; 1974),
Doherty and Perkins (1979a), Kushner et al. (1992) and Serafimov et al. (1996). It is beyond the
scope of this survey to review these papers, though they touch upon interesting aspects. For ex-
ample, structures with biazeotropic elements and nonelementary singular points are encountered
in the transition from one topology to another with pressure (or temperature) change. The fact
that the classifications do not include such transient structures, which usually exist over only very
small pressure (temperature) intervals, does not limit their practical use.

3.7.3

Occurrence of predicted structures

Experimental VLE data indicates the natural occurrence of ternary mixtures for at least 16 of
Serafimov’s 26 topological classes, and 28 of the 49 antipodal types given by Zharov and Ser-
afimov (1975). Serafimov (1968d) analyzed the occurrence of different types of VLE diagrams

3.7 Ternary VLE diagrams. Classification, occurrence and structure determination

Table 3.4: Occurrence of classes of ternary VLE diagrams found in published mixture data

No.

Serafimov’s

Occurrence

ZS-type of

Set of Azeotropes

Occurrence

Class

Serafimov

Antipodal

Reshetov

(until 1968)

Structure

(1965-1988)

zeotropic

244

1.0-1a

min

283

max

1.0-1b

min

max

1.0-2

min

max

1.1-1a

None

min + min A

None

max + max A

None

1.1-1b

None

min + min A

None

max + max A

None

1.1-2

min + S

max + S

2.0-1

min + max

2.0-2a

None

min + max

max + min

2.0-2b

11a

min + min

280

21a

max + max

2.0-2c

11b

min + min

21b

max + max

2.1-1

None

min + max + min A

None

min + max + max A

None

2.1-2a

None

16a

min + max + min A

None

16b

min + max + max A

None

2.1-2b

min + min + min A

max + max + max A

None

2.1-3a

min + max + S

2.1-3b

min + min + S

max + max + S

3.0-1a

None

min + min + max

None

max + max + min

None

3.0-1b

None

min + min + max

max + max + min

3.0-2

min + min + min

114

max + max + max

None

3.1-1a

27b

min + min + max + min A

None

32b

max + max + min + max A

None

3.1-1b

None

min + min + max + min A

None

max + max + min + max A

None

3.1-1c

None

27a

min + min + max + max A

None

32a

max + max + min + min A

None

3.1-2

171

min + min + min + min A

355

max + max + max + max A

None

3.1-3a

None

25a

min + min + min + S

None

38a

max + max + max + S

None

3.1-3b

None

25b

min + min + min + S

None

38b

max + max + max + S

None

3.1-4

min + min + max + S

max + max + min + S

min - minimum-boiling binary azeotrope, max - maximum-boiling binary azeotrope
A

- ternary node azeotrope, S

- ternary saddle azeotrope

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

Baburina et al. (1983; 1988) claimed that some types of the VLE diagrams are thermodynamically
inconsistent, namely Serafimov’s classes 1.0-1a, 1.0-1b, 2.1-1, 3.0-1a, 2.0-2a, 3.0-1a, 3.0-1b, 3.1-
3a and 3.1-3b. Thus, according to Baburina, the number of feasible classes of topological VLE
diagram structures is only 15 rather than 26. This statement was based on the reasoning that:
(a) mixtures of these classes were not reported in Serafimov’s statistics (Serafimov, 1968d); (b)
these diagrams cannot be described by the Wilson activity coefficient equation; (c) theoretical
analysis by the authors confirms the inconsistency of these diagrams. All these arguments appear
to be incorrect: (a) real mixtures representing the classes 1.0-1a, 1.0-1b, 2.0-2a, and 3.0-1b are
found as shown in the statistics given above; (b) Zhvanetskij et al. (1993) demonstrated that all
the 26 classes of VLE diagrams can be simulated by the Wilson activity coefficient equation;
(c) it has been shown that the theoretical analysis by Baburina et al. is not valid (Reshetov et
al., 1990; Zhvanetskij et al., 1993). Later, accepting Baburina’s claims, P¨ollmann and Blass
(1994) and P¨ollmann et al. (1996) proposed to reduce the number of ternary VLE diagrams to
consider only ”physically meaningful” structures. The list of ”physically meaningful” structures
in P¨ollmann and Blass (1994) includes 19 structures of Serafimov’s classes 0, 1.0-1a, 1.0-1b, 1.0-
2, 1.1-2, 2.0-1, 2.0-2a, 2.0-2b, 2.0-2c, 2.1-2b, 2.1-3a, 2.1-3b, 3.0-1a, 3.-1b, 3.0-2, 3.1-1b, 3.1-2,
3.1-3a, 3.1-4.

However, in our opinion, it is impossible in principle to state that some classes of ternary mix-
tures cannot exist in nature or are “physically meaningless” because all the structures predicted
by the rule of azeotropy are thermodynamically and topologically consistent by definition. We
can only discuss the probability of the existence of some types of the VLE diagram structure. It
is well-known that binary maximum-boiling azeotropes are less abundant than minimum-boiling
azeotropes. According to Lecat (1949) the ratio of the minimum-boiling versus maximum-boiling
azeotropes that occur in nature is about 9 to 1, and the statistics of Reshetov confirms this approx-
imate rule, see Figure 3.24. In particular, no ternary mixture with three binary maximum-boiling
azeotropes has been found among the selection of 1365 mixtures, and no ternary maximum-
boiling azeotrope has been found. As a result, even for the topological structures where existence
is beyond question, the occurrence of antipodes with maximum-boiling azeotropes is much less
common than that of antipodes with minimum-boiling azeotropes, see Table 3.5.

Table 3.5: Occurrence of antipodes of the most common topological classes of ternary VLE dia-
grams

Serafimov’s

Reshetov’s Statistics (1965-1988)

Class

min. azeotrope(s)

max. azeotrope(s)

1.0-1a

283

1.0-2

2.0-2b

280

2.1-2b

None

We propose to consider some conditions that limit the natural occurrence of certain VLE diagram
structures. First, we find for maximum-boiling azeotropes:

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

Table 3.6: Improbable types of topological classes of ternary VLE diagrams

Serafimov’s

Matsuyama and Nishimura’s

Probability

Limiting

Class

type

Class

Factors

1.1-1a

200-m

low

040-M

improbable

2A + 1B

1.1-1b

020-m

low

004-M

improbable

2A + 1B

2.1-1

032-m, 230-m, 203-m

low

041-M, 140-M, 104-M

improbable

2B + 1B

2.1-2a

16a

420-m, 402-m

low

16b

024-m, 420-M

improbable

2A + 1B

2.1-2b

044-M, 440-M, 404-M

improbable

3.1-1a

27b

421-m, 412-m

low

2A + 2B

32b

423-M, 324-M

improbable

2A + 2B + 1B

3.1-1b

232-m, 223-m, 322-m

low

414-M, 441-M, 144-M

improbable

2B + 1B

3.1-1c

27a

142-M, 241-M, 124-M, 214-M, 421-M, 412-M

improbable

2A + 2B

32a

423-m, 324-m, 423-m, 342-m, 234-m, 243-m

low

2A + 2B + 1B

3.1-2

444-M

improbable

1A + 1B

3.1-3a

25a

121-S, 112-S, 211-S

low

38a

343-S, 334-S, 433-S

very low

2C + 1A

3.1-3b

25b

121-S, 112-S

low

38b

343-S, 334-S, 433-S

very low

2C + 1A

ZS-type refers to the refined classification on antipodal structures by Zharov and Serafimov (1975)

cases. The mixtures in the collection by Reshetov are ranked in the order of decreasing occurrence
in Table 3.7. Most common VLE diagram structures are given in Group A. Group B structures
are also common, and Group C can be considered as rare structures.

3.7.4

Determination of the structure

The question of whether it is possible to determine the VLE diagram structure of a mixture solely
from the boiling temperatures of the pure components and azeotropes has been discussed since
the first publications in this field. It was claimed by Gurikov (1958) and later confirmed by Serafi-
mov (1968a), and Zharov and Serafimov (1975) that, in general, pure component and azeotropic
boiling temperature data only are not sufficient to construct the diagram uniquely and that ad-
ditional information about the direction of increasing boiling temperature near some singular
points is needed. It has been proposed using ebulliometric measurements to obtain this addi-
tional information experimentally (Serafimov, 1968a; Zharov and Serafimov, 1975). Obviously,
a mathematical description of the VLE may be used for this purpose too.

3.7 Ternary VLE diagrams. Classification, occurrence and structure determination

Table 3.7: Identified structures of ternary VLE diagrams among real mixtures according to
Reshetov’s statistics (1965 to 1988)

Structure of Ternary VLE Diagram

No.

Serafimov’s

Matsuyama & Nishimura’s

Occur

Class

type

Class

Group A

3.1-2

222-m

26.0

5 %

1.0-1a

3a, 7a

100, 030

21.6

2.0-2b

11a, 21a

102, 120, 021, 043, 430, 403

21.0

1.0-2

4, 8

020, 400

8.5

3.0-2

212, 122, 221

8.4

Group B

2.1-2b

022-m, 220-m, 202-m

4.0

1 - 5 %

3.1-4

30, 35

131-S, 113-S, 311-S, 133-S, 313-S, 331-S

3.3

2.1-3a

013-S, 310-S, 301-S

2.7

1.1-2

5, 9

010-S, 300-S

1.2

Group C

2.0-2c

11b, 21b

201, 210, 012, 034, 304, 340

0.9

1 %

3.0-1b

123, 321, 132, 213, 312, 231

0.88

2.0-1

031, 103, 130

0.66

13, 14

2.0-2a

17, 18

023, 320, 401, 410

0.36

13, 14

1.0-1b

3b, 7b

001, 003

0.36

2.1-3b

12, 22

011-S, 110-S, 101-S, 033-S330-S, 303-S

0.22

ZS-type refers to the refined classification on antipodal structures by Zharov and Serafimov (1975)

Doherty (1985) stated that the residue curve maps can be uniquely and explicitly determined
from the components and azeotropes boiling points only. This concept is used by Bernot (1990),
Foucher et al. (1991) and Peterson and Partin (1997). However, this assumption is not theoreti-
cally founded and is not valid in general. On the other hand, as noted by Foucher et al. (1991)
and as we show below, in most cases the simple ranking of singular points (components and
azeotropes) in order of increasing boiling points is sufficient to construct the structure of the
residue curve map. In specific cases we have indeterminacy (Foucher et al., 1991), but this is re-
ally a result of the lack of data required for a unique solution. The cases where different structures
of VLE diagrams correspond to the same order of the boiling temperatures of the singular points
are presented in Figures 3.25 and 3.26.

In Cases 1 and 2 the indetermination occurs for one topological class (antipodes). Case 1 is
described by Foucher et al. (1991). For all the cases in Figure 3.25 the real structure of the given
mixture can be easily determined experimentally. Depending on the direction of the increasing
temperature b or c near the singular point in the triangle given in column a) the mixture has a
structure as given in column b) or c). A more complex situation presented in Figure 3.26 is also
described by Foucher et al. (1991) and named “global” indeterminacy. The real structure cannot
be identified from additional investigation of the boiling temperature surface, and computing the
actual residue curve map is required. Fortunately, this type of structure has low probability (no
mixtures reported yet, see Table 3.6).

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

Case 1

Case 2

Case 3

2.0 - 2 a ZS 18

3.0 - 1a

ZS 29

3.1 - 1a

ZS 27b

2.0 - 2 a ZS 17

3.0 - 1b ZS 18

3.1 - 1b

ZS 26

Figure 3.25: Indeterminacy of the residue curve map structure for a ternary mixture with the
given order of the components and azeotropic temperature points: a) the given set of temperature
ordered singular points; b,c) feasible structure alternatives.

There are some cases where identification of the structure of VLE diagram can be made with
a high degree of probability from the data on the components and binary azeotropes. These
cases are presented in Table 3.8. In all other cases, binary data are not sufficient for the precise
identification of the structure. By way of example let us consider the mixtures with two and three
binary minimum-boiling azeotropes where we know all singular points at the edges of triangle
(Figure 3.27 a, cases 1 and 2). Three feasible structures are shown in Figure 3.27 (b, c, d for each
case). All feasible alternatives are likely to occur and it is very risky to assign a structure to the
given mixture based on information about the singular points at the edges of triangle only.

This is also pointed out by Westerberg (1997): Generally, we cannot construct a unique diagram if

3.7 Ternary VLE diagrams. Classification, occurrence and structure determination

3.1 - 3a

3.1 - 3b

Figure 3.26: “Global” indeterminacy of the residue curve map structure for a given ranking of
the components and azeotropic points (according to Foucher, 1991): a) the given ranking of the
singular points; b,c) feasible structure alternatives.

Table 3.8: Window of prediction of ternary VLE diagram structures from binary data only

Singular points

on the edges

Structure Alternatives

Most probable structure

12-23-13

(Matsuyama’s

Serafimov’s

ZS-

Serafimov’s

ZS-

Serafimov’s

ZS-

Comment

nomination)

Class

type

Class

type

class

type

100

1.0-1a

1.1-1a

1.0-1a

most probable

030

1.0-1a

1.1-1a

1.0-1a

almost certain

001

1.0-1b

1.1-1b

1.0-1b

most probable

003

1.0-1b

1.1-1b

1.0-1b

almost certain

103 ( or 130,

2.0-1

2.1-1

2.0-1

most probable

or 031)

2.1-1

we only know the existence of the azeotropes and their boiling point temperatures. If we know the
temperature and also the nature of all the pure component and azeotrope points (type of saddle,
stable node, unstable node), then we can sketch a unique diagram.

It has been stated by Matsuyama and Nishimura (1977), Yamakita et al. (1983) and Foucher et
al. (1991) that the rule of azeotropy can be used to check the consistency of azeotropic data. It
is appropriate to emphasize that this only applies to the consistency of the experimentally deter-
mined ternary azeotropic points. Furthermore, only when the boiling temperature of this point
lies between the temperatures of the singular point of a separatrix or a line between an unstable
node and a stable node, can an erroneous ternary azeotrope be singled out and excluded. If the
experimentally determined singular point has a boiling point temperature outside these intervals
it cannot be excluded on the basis of the rule of azeotropy. This is demonstrated in Figure 3.28.

The issues concerning the identification of the VLE diagram structure on the basis of incomplete

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

2.0 - 2b

3.0 - 2

2.1 - 2b

3.1 - 2

2.1 - 3b

3.1 - 3b

Case 1

Case 2

Figure 3.27: Feasible alternatives of the residue curve map structure for the given ranking of
the singular points at the edges of the composition triangle: a) the given set of ranking of the
components and azeotropic points at the edges of the composition triangle; b) structure without
ternary azeotrope; c) structure with a ternary unstable node azeotrope; d) structure with a ternary
saddle azeotrope.

information (and indeterminacy and inconsistency of such identification) were of great signifi-
cance before the mathematical modeling of VLE became widely applicable. Today we can obtain
an appropriate mathematical description of the ternary VLE, find all singular points and construct
the diagram of residue curves or distillation lines based on data for the binary constituents. It
is also possible to predict the binary parameters from the characteristics of the pure components
(e.g., by means of the group contribution models). VLE diagrams obtained by modeling are deter-
mined and consistent by definition (but they do not necessarily represent the behavior of the real
mixtures properly). The real issue today is the accuracy of the model description. The description
of a ternary mixture can be wrong even with reliable binary data because the ternary azeotrope
can be found (or not found) mistakenly by the model. For example, modeling the VLE for the
mixture of n-butanol, n-butyl acetate and water by means of the Wilson and NRTL activity coeffi-
cient equations may give different topological structures of the diagrams depending on the binary
parameters sets that were chosen, see Figure 3.29. The parameters for the model can be readily
corrected if experimental data on the ternary azeotrope are available. However, topological or
thermodynamic considerations do not help to select the correct model in such a situation if only
binary data are available.

3.7 Ternary VLE diagrams. Classification, occurrence and structure determination

T < T < T

T < T

T < T < T

Case 1

Case 2

Case 3

Real

Infeasible

Feasible

Figure 3.28: Consistency test based on the rule of azeotropy may fail to detect an erroneous exper-
imentally determined ternary azeotrope (

): a) real structure; b) infeasible structure (existence

of the false ternary azeotrope is excluded); c) feasible structure (existence of the false ternary
azeotrope obeys the rule of azeotropy and cannot be excluded).

3.8 Unidistribution and univolatility lines

3.8

Unidistribution and univolatility lines

3.8.1

Distribution coefficient and relative volatility

The distribution coefficient and relative volatility are well-known characteristics of the vapor-
liquid equilibrium. The distribution coefficient

is defined by:

(3.14)

As the name implies,

characterizes the distribution of component

between the vapor and liq-

uid phases in equilibrium. The vapor is enriched with component

and is impoverished

with component

for

compared to the liquid. The ratio of the distribution coefficients of

components

and

gives the relative volatility of these components, usually denoted by

(3.15)

The relative volatility characterizes the ability of component

to transfer (evaporate) into the va-

por phase compared to the ability of component

. Component

is more volatile than component

, and less volatile if

. For ideal and nearly ideal mixtures, the relative volatili-

ties for all pair of components are nearly constant in the whole composition space. The situation
is different for nonideal and in particular azeotropic mixtures where the composition dependence
can be complex. The qualitative characteristics of the distribution coefficient and relative volatil-
ity functions are also considered in a general form by the ”pen-and-paper” approach that is typical
for the thermodynamic topological analysis.

Serafimov (1970a) considered the behavior of these functions for binary mixtures. Based on
Serafimov’s approach and the paper by Kushner et al. (1992) we present the feasible patterns
of the distribution coefficient functions

(x)

and

(x)

for binary mixtures according to the

feasible paths of the equilibrium line

(x)

in Figure 3.30. The distribution coefficients of the

azeotrope-forming components are equal to unity in the points of the pure components and the
azeotrope. One of these distribution coefficient functions has an extremum when the equilibrium
line (

relationship) has an inflection point, and both distribution coefficient functions have

extrema when there is an azeotrope in the mixture. The relative volatility function

(x)

inter-

sect the line

at the azeotrope points(s). The transitions from one type of diagram to another

is caused by the change of the molecular interactions of the components resulting in positive or
negative deviations from ideality given by Raoult’s law. The particular deviation from Raoult’s
law is determined by the balance between the physiochemical forces between identical and differ-
ent components in the mixture, and we can make the following qualitative distinction for binary
mixtures of component 1 and 2:

Positive deviation: The components 1 and 2 do not like each other. The attraction between iden-

tical molecules is stronger than between different molecules, which may cause a minimum-
boiling azeotrope.

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

Negative deviation: The components 1 and 2 like each other. The attraction between identical

molecules is the strongest, which may cause a maximum-boiling azeotrope.

In conclusion, the composition dependence of the distribution coefficients are qualitative and
quantitative characteristics of the VLE for the given mixture. The patterns of these functions de-
termines not only the class of binary mixture (zeotropic, minimum- or maximum-boiling azeotrope,
or biazeotropic), but the individual behavior of the given mixture as well, as will be shown in the
following section.

3.8.2

Univolatility and unidistribution line diagrams

The composition dependence of the distribution coefficients of a ternary mixture of components
1, 2 and 3 can be represented by three surfaces

(x)

and

(x)

. Serafimov (1970a)

proposed to consider the system of unidistribution lines in the composition space where the dis-
tribution coefficient of the given component

equals unity

(x)

The dynamic system of open evaporation residue curves Equation (3.7) can be represented as:

)

:::;

(3.16)

The singular points of this system are related to the unidistribution lines, because:

(a) in the point of pure component

(b) in the point of binary azeotrope A

Thus, the existence of a binary azeotrope gives rise to two unidistribution lines (

and

), and the existence of a ternary azeotrope gives rise to three unidistribution lines (

). The point of pure component

may (or may not) give rise to an unidistribution

line of component

. Serafimov (1970b; 1970a; 1973) showed that a given residue curve map

corresponds to given feasible diagrams of unidistribution lines.

In a similar way, the relative volatility function can be represented by isovolatility lines, that is,
contour lines in the composition plane of the three surfaces

(x)

for all pairs of the compo-

nents. It is difficult to to interpret all three isovolatility lines (

) in the same diagram.

Serafimov et al. (1972) propose to use the system of univolatility lines where

. However,

isovolatility lines for other values than unity are useful to evaluate the influence of a component

on the relative volatility of the other two components

given by

, in the search for an entrainer

for extractive distillation (Laroche et al., 1993; Wahnschafft and Westerberg, 1993).

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

We use the term univolatility lines to distinguish between the lines where

from other

isovolatility lines where

. It is evident that the point of a binary azeotrope Az

gives

rise to an

-univolatility line and that the point of a ternary azeotrope gives rise to the three

univolatility lines (

, and

). Serafimov et al. (1972) showed that in

addition there are univolatility lines which are not connected with the azeotropic points, and that
such lines can occur even in zeotropic mixtures.

Analysis of feasible diagrams of unidistribution and univolatility lines is given by Serafimov
(1970b; 1970a; 1972). The main aim of this work was to consider feasible structures of the
residue curve maps in more detail, and in fact exactly this approach helps to construct more
refined classification of ternary diagrams. Later, the diagrams of unidistribution lines were used
as a main tool for analysis of tangential azeotropy and biazeotropy (Serafimov, 1971b; Serafimov,
1971a; Serafimov, 1971c; Serafimov, 1971d; Kushner et al., 1992; Serafimov et al., 1996). The
diagrams of univolatility lines were used for the same purpose by Zhvanetskij and Reshetov et
al. (Zhvanetskij et al., 1988; Zhvanetskij et al., 1989; Zhvanetskij et al., 1993; Reshetov et al.,
1990; Sluchenkov et al., 1990). In addition, Zhvanetskij et al. (1988) noted that univolatility lines
split the composition triangle into regions of certain order of volatility of components which they
named “regions of

-ranking”, or, “K-ordered regions”. The feasible structures of the diagrams

of unidistribution and univolatility lines given by Serafimov are considered in Section 3.8.3 for
the purpose of sketching the simple phase transformation trajectories such as residue curves and
distillation lines.

The unidistribution (

) and univolatility (

) lines play the role as ”road signs” for

the pathway of the simple phase transformation composition trajectories (such as residue curves
and distillation lines) Kiva and Serafimov (1973):

1. Along the unidistribution line (

) all equilibrium vectors in the composition space

are parallel to the edge

of the composition triangle (see Figure 3.31).

2. Along the univolatility line (

) at a point

, the equilibrium vector

lies at the

secant going from the vertex

through this point (see Figure 3.32).

3. At the intersection of a residue curve with the unidistribution line (

) there is an

extrema of the composition

(

) (see Figure 3.31).

4. If there are no univolatility lines, there are no inflection points of the simple phase transfor-

mation trajectories (see Figure 3.31).

5. The univolatility line

gives rise to an inflection point on the residue curves (and the

other simple phase transformation trajectories) if there is no

- or

-unidistribution lines

along the residue curve between the univolatility line in question and the stable or unstable
nodes of the residue curves. This rule is illustrated in Figure 3.32 for a zeotropic mixture
of components 1-2-3 with two univolatility lines (

and

). It can be seen

that there is no unidistribution line between the unstable node (1) and the

-univolatility

line, and there is an inflection point on the residue curves between the unstable node and

3.8 Unidistribution and univolatility lines

103

tribution and univolatility lines permits us to construct quantitatively the sketch of the map of
the simple phase transformations. It not only characterizes the topology of the maps but their
geometry as well. The unidistribution and univolatility lines can be readily determined numer-
ically (by simulation) based on a VLE model of the given mixture (see for example Bogdanov
and Kiva (1977)) and their computation is much easier than computation of the residue curves or
(especially) the distillation lines.

3.8.3

Structure of unidistribution and univolatility line diagrams

The unidistribution and univolatility lines can be easily located quantitatively by computation of
the VLE for the given mixture (see, for example, Bogdanov and Kiva (1977)), and diagrams for
the most common structures of ternary VLE diagrams according to Reshetov’s statistics as given
in Table 3.7, Section 3.7.3, are presented in Figure 3.38.

The theory for finding such feasible structures of unidistribution line diagrams was developed
by Serafimov (1970b; 1970a) and presented in Serafimov et al. (1973). Feasible structures of
univolatility line diagrams were considered by Serafimov et al. (1972).

To understand the ideas better, let each binary composition diagram present a side of the unfolded
triangle formed by the three components in ternary mixtures as shown in Figure 3.39. Various
combinations of binary constituents and shapes of the

(x)

surfaces lead to various structures

of the ternary diagrams.

Example 1. Zeotropic mixture 1-2-3

a) Let us assume that all three binary constituents are mixtures of Type 1. Unfolding the prism,

we draw the curves

(x)

at the edges of the prism and determine the values

i(j

)

(infinite

dilution of

) at the edges of the prism (Figure 3.39). Here:

2(1)

3(1)

1(2)

3(2)

1(3)

2(3)

Assuming that the functions

i(j

are monotonous, we construct their feasible trajectories

at the edges of the prism. We can see that line

2(13)

intersects line

and there

are only two points with the value

at the edges of the triangle. These points are

connected by the

-unidistribution line. There are no univolatility lines and, consequently,

there are no inflection points at the residue curves or distillation lines. A sketch of the
diagram is made based on the location of unidistribution line. This cell of residue curves
(distillation lines) is C-shaped.

106

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

b) Let us assume that the binary mixture 1-3 is the mixture of Type 2’, i.e. it becomes more

negative or less positive than the mixture 2-3 (Figure 3.40).

Figure 3.40: Univolatility and unidistribution line diagram and phase portrait of the residue
curve map for a zeotropic mixture 1-2-3 where the binary constituents 1-3 are a mixture of Type
2’ (S-shaped residue curves).

Then there is

2(3)

1(3)

, and there are points of intersection of the lines

(x)

and

(x)

at the edges 13 and 23. As a result, the

-univolatility line connects

edges 13 and 23 of the composition triangle (Figure 3.40), and there is not a unidistribution
line between this line and the stable node of residue curves. This univolatility line gives
rise to inflection points of residue curves and distillation lines. A sketch of the diagram
is determined based on the position of unidistribution and univolatility lines. This cell of
residue curves (distillation lines) is S-shaped.

c) Let the binary mixture 1-3 be a mixture of Type 2, i.e. it becomes more positive than the

mixture 12. In the similar way we find that there is an

-univolatility line connecting

edges 12 and 13 and no unidistribution lines between it and the unstable node of the residue
curves (Figure 3.41). This univolatility line gives rise to inflection points of residue curves
and distillation lines. A sketch of the diagram is determined basing on the position of

108

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

section points and other between the univolatility line and the Apex 1. These lines have a
specific

-shape and other lines are

-shaped.

It can be seen from these examples that the single unidistribution line going from the saddle point
is an inherent characteristic of zeotropic mixtures (Serafimov’s Class 0). Univolatility lines can
optionally exist if the mixture is nonideal, and there is a diversity of the diagrams of univolatility
lines caused by various combinations of the deviation from ideality in the binary constituents.

Detailed analysis of all feasible variants of the univolatility lines for ternary zeotropic mixtures
was made later by Zhvanetskij et al., 1988, and 33 variants of these diagrams were revealed.
Tough the occurrence of most of these types is questioned, all these types are possible theoreti-
cally.

Example 2. Mixture 1-2-3 with minimum-boiling azeotrope

(Serafimov’s Class 1.0-1a)

a) The mixture 1-2 in this ternary mixture must be the mixture of Type 3. Let both mixtures 2-3

and 1-3 be mixtures of Type 1. The diagram of unidistribution and univolatility lines and
the sketch of the RC (DL) map are presented in Figure 3.42 a. The -univolatility line goes
from the point of azeotrope to the edge 13. The lines of the simple phase transformations
have inflection points between the univolatility line and Apex 3. A part of the lines in this
cell are C-shaped, ant the other lines are S-shaped.

b) Let the mixture 13 be a mixture of Type 2’, i.e., it is more negative (or less positive) than

mixture 2-3. Here the -univolatility line goes from the point of azeotrope to edge 23. The
lines of the simple phase transformations have another sign of curvature (Figure 3.42 b).

Figure 3.42: Unidistribution and univolatility line diagram and the sketch of residue curves for a
mixture 1-2-3 with a minimum-boiling azeotrope

for various types of the binary constituents:

a) both mixtures 1-3 and 2-3 are mixtures of Type 1; b) mixture 1-3 is a mixture of Type 2.

Let us consider now various trajectories of the functions

(x)

at edge

of the prism. In

the case where mixtures 1-3 and 2-3 are the mixtures of Type 1, the only intersection of

-lines

3.8 Unidistribution and univolatility lines

109

corresponds to the point of binary azeotrope, and it leads to the structure presented in Figure 3.42
a. Other cases are presented in Figure 3.43.

Figure 3.43: Distribution coefficient trajectories

(x)

at the edge

for a mixture 1-2-3

with minimum-boiling azeotrope 1-2: a) Mixture 2-3 is a mixture of Type 2; b) both mixtures 2-3
and 1-3 are the mixtures of Type 2; c) additional positive deviations at the ternary mixture.

In Case (a), the mixture 2-3 becomes more positive (Type 2), and the line

3(12)

intersects twice

in succession with the line

(12)

. It leads to the appearance of two points

at edge 1-2.

The lines of the simple phase transformation have two inflection points when they intersect twice
in succession with the

-univolatility line and there is a local deformation of the residue curves

or distillation lines. The corresponding diagram is presented in Figure 3.44a.

In Case (b), both mixtures 1-3 and 2-3 are mixtures of Type 2 (more positive). Line

3(12)

intersects twice in succession with the line

(12)

, and intersects twice in succession with the

line

(12)

. It leads to the appearance of two points

and

two points at the edge

1-2. The lines of simple phase transformation have two inflection points where they intersect
these arc-wise univolatility lines, and the map has two local deformations (Figure 3.44 b).

Both these diagrams are diagrams of Serafimov’s Class 1.0-1a but they have more complex ge-
ometry in the simple phase transformation lines than the mixtures represented in Figure 3.42,
corresponding to a more complex diagram of univolatility lines.

It can be seen that the inherent characteristic of the diagram of Class 1.0-1a is just existence of
two unidistribution lines going from the azeotrope point to tops 1 and 2 and the

-univolatility

line going between them to one of sides 13 or 23. The movement of the

-univolatility line and

the existence of other univolatility lines are the specific characteristics of the given mixture.

Let us consider the situation where both mixtures 13 and 23 are mixtures of Type 2, but the
positive deviations in the ternary mixture are so strong that line

3(12)

has a maximum with the

value

(Figure 3.43 c). Here we have in addition two points at edge 12. This changes the

diagram of unidistribution and univolatility lines totally and, consequently, changes the RC (DL)
map (Figure 3.44 c). We can see that strengthening of positive deviations leads to appearance of

3.9 Conclusions

117

3.9

Conclusions

The main conclusions that can be reached from the literature on qualitative analysis of the VLE
diagrams and the additional investigations presented in this survey are:

Vapor-liquid equilibrium can be characterized by various means; first and foremost by simple

phase transformation maps, among which open evaporation residue curve maps and dis-
tillation line maps have received broad acceptance. Note that distillation lines (discrete
and continuous) have a well-defined thermodynamical meaning. These lines are rigorously
defined by the VLE and, in turn, characterize it, as well as the residue curves.

The VLE can also be qualitatively characterized by unidistribution line diagrams and uni-
volatility line diagrams. Univolatility line diagrams give additional information about the
regions of the composition space with different volatility order of the components. The
combined diagram of unidistribution and univolatility lines characterizes the shape (geom-
etry) of the trajectories of simple phase transformations.

We recommend that a distinction is made between the topological structure of simple phase

transformation maps (i.e., the set of singular points and the splitting into regions of the
simple phase transformation trajectories), the sketch of the map (i.e., the qualitative pattern
of the trajectories with representation of their shape), and the exact map of the simple phase
transformation trajectories (residue curves or distillation lines).

The structures of VLE diagrams of azeotropic mixtures are limited by rigorous topological con-

straints. The structures for ternary mixtures without biazeotropy are classified into 26 pos-
sible topological structures. All types of the classified diagrams are topologically and ther-
modynamically feasible, but their occurrence in nature is determined by the probability of
certain combinations of molecular interactions. Data on the natural occurrence of various
mixture classes permit us to exclude rare or improbable diagrams from consideration in the
investigation of the correlation between VLE diagram structures and feasible separations
upon distillation.

Usually the (topological) structure of the VLE diagram for a given mixture can be uniquely

determined based solely on information of the existence and boiling point temperature of
all singular points (pure components and azeotropes). It should be noted however, that there
are some cases of indeterminate diagram structure.

The (topological) structure of the VLE diagram identifies the class of a given mixture, but not

its specific characteristic like the shape of the singular and ordinary simple phase transfor-
mation trajectories.

A sketch of the simple phase transformation map (residue curve or distillation line map), in-

cluding the specific features of a given mixture (i.e., the shape of the trajectories), can be
determined from the localization of the unidistribution and univolatility lines. These lines
give complete information about the curvature (path) of singular and ordinary trajectories
of the simple phase transformation.

118

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

Generating an exact map including the localization of simple phase transformation trajectory

boundaries (separatrices) requires numerical calculation. Often, only the boundaries are
located by appropriate simulation rather than by generating a complete trajectory map.

It is reasonable to use the combined diagram of unidistribution and univolatility lines as a repre-

sentative qualitative characteristic of the VLE.

Definitions of Terms

This section contains an alphabetic listing with definitions of central terms as they are used in this
text. In addition, Table 3.1 in the introductory pages gives an overview (quick reference) of the
different terms used for key concepts in the azeotropic distillation literature and may be helpful
in particular when reading Russian publications that are translated into English.

Antipode - exact opposite, but topological equivalent structure.

Azeotrope - mixture whose equilibrium vapor and liquid compositions are equal at the given

system temperature and pressure. Mathematically: a singularity in the VLE manifold that
does not correspond to a pure component.

Azeotropic distillation - distillation that involves azeotropic mixtures. Note that this is a much

broader definition than traditionally used in the distillation literature where the term is fre-
quently used as an abbreviation of heteroazeotropic distillation.

Azeotropic mixture - mixture that forms one or several azeotrope(s).

Biazeotropic mixture - mixture that forms two binary azeotropes for a binary pair or two ternary

azeotropes.

Composition space - for an

-component mixture the composition space contains all the

dimensional vectors of

where the elements sum to unity.

Composition trajectory - the steady state or instantaneous composition profile of a distillation

column, or the trajectory of composition change during batch distillation.

Condensate curve - open condensation liquid composition trajectory.

Condensation curve - open condensation vapor composition trajectory.

Conjugate - an element of a mathematical group that is equal to a given element of the group

multiplied on the right by another element and on the left by the inverse of the latter ele-
ment. [Britannica]

Distillate curve - composition trajectory of the vapor during open evaporation.

3.9 Conclusions

119

Distillation - the process of separating a liquid mixture by successive evaporation and conden-

sation.

Distillation boundary - locus of compositions that cannot be crossed by the given distillation

process. The existence and location of such boundaries depend critically on the type of
distillation in question. Distillation boundaries are thermodynamic barriers for multicom-
ponent distillations in the same way that azeotropes are barriers to binary distillation. They
split the composition space into distillation regions.

Distillation line - the set of points

for which the vapor composition in equilibrium

also

lies on the same line in the composition space (Zharov and Serafimov, 1975). A discrete
distillation line is drawn as a piecewise linear curve through a sequence of conjugated
vapor-liquid equilibrium vectors in the composition space. Continuous distillation lines
interpolate the same sequence of equilibrium points and are not mathematically unique.

Distillation line map - diagram that shows distillation lines for different initial conditions for a

given mixture in the composition space.

Distillation region - a subspace of the composition space defined by a family of distillation

trajectories that join a stable and unstable node.

Distribution coefficient - ratio of mole fractions in vapor and liquid phase for the given species.

Commonly referred to as vapor-liquid equilibrium ratio or K-value.

Entrainer - material (liquid component) that acts as a mass separating agent.

Equilibrium vector - composition vector between two phases in equilibrium, also named tie-

line.

Equilibrium staged distillation - staged distillation column where vapor-liquid equilibrium is

assumed between the phases on all trays.

Heteroazeotrope - azeotrope where the vapor phase coexists with two liquid phases.

Heterogeneous mixture - mixture with more than one liquid phase.

Homoazeotrope - azeotrope where the vapor phase coexists with one liquid phase.

Homogeneous mixture - mixture with a single liquid (and vapor) phase.

Ideal mixture - mixture that obeys Raoult’s law (or Henry’s law), i.e., the activity coefficients

for all components of the liquid are equal to unity. Ideal mixtures are a special case of
zeotropic mixtures.

Inflection point - root of the second order derivative of a given function, point of change of

function curvature.

Inflection point curve - locus of inflection points.

120

Chapter 3. Azeotropic Phase Equilibrium Diagrams: A Survey

Injective - being a one-to-one mathematical function.

Isodistribution line - locus of points in the composition space where the distribution coefficient

for a particular component is constant.

Isomorphic - being of identical or similar form, shape, or structure. Antipodal topological VLE

diagrams are isomorphous.

Isotherm - locus of points in the composition space where the temperature for a given mixture

at vapor-liquid equilibrium is constant. Liquid isotherms are lines of constant value of the
boiling temperature and vapor isotherms are lines of constant value of the condensation
temperature.

Isotherm map - diagram that shows isotherms for a given mixture in the composition space.

Contour plot of one or both of the vapor-liquid equilibrium temperature surfaces for a given
mixture.

Isovolatility line - locus of points in the composition space where the relative volatility of a pair

of components is constant.

K-value - see distribution coefficient.

Mapping - to assign (as a set or element) in a mathematical correspondence.

Mapping function - function that gives the mathematical correspondence between two sets of

elements.

Monoazeotropic mixture - mixture that forms only one azeotrope at each element of the com-

position simplex (only one binary azeotrope for each binary pair in an azeotropic mixture).

Node (mathematically) - critical point with all paths either approaching (stable) or departing

(unstable).

Nonideal mixture - mixture that exhibits deviations from Raoult’s law (or Henry’s law). Non-

ideal mixtures may be zeotropic or azeotropic.

Ockham’s Razor - principle stated by English philosopher William of Ockham (1285-1347/49)

that the simplest of competing theories should be preferred to the more complex, and that
entities are not to be multiplied beyond necessity.

Open condensation - condensation of a vapor phase where the liquid formed is removed con-

tinuously.

Open distillation - distillation with input and/or output mass streams.

Open evaporation - evaporation of a liquid phase where the vapor formed is removed continu-

ously. Also referred to as simple distillation or Rayleigh distillation.

Open evaporation distillate curve - see distillate curve.

3.9 Conclusions

121

Relative volatility - the ratio of distribution coefficients for pair of components in a mixture.

Residue curve - composition trajectory of the residue liquid in the still during open equilib-

rium evaporation. Sometimes referred to as “distillation line” in Russian and old German-
language literature.

Residue curve map - diagram that shows residue curves for different initial still compositions

for a given mixture in the composition space.

Residue curve region - set of liquid compositions in the composition space that belong to a

family of residue curves with common initial and final points.

Saddle (mathematically) - singular point with finitely many paths both approaching and depart-

ing. Only separatrices extends or terminate in the saddle point. All other paths have a
hyperbolic course in the vicinity of the singular point.

Separatrix - locus of bifurcation points of differential equations, i.e., points at which an in-

finitesimal perturbation causes at least one of the integration endpoints to change. A sepa-
ratrix may be stable or unstable dependent on whether solutions in its vicinity approach or
depart as the value of the independent (integration) parameter goes to infinity.

Simple distillation - see open evaporation.

Simple distillation region - see residue curve region.

Simple distillation residue curve - see residue curve.

Simple equilibrium phase transformation - process where the composition of one phase

changes according to its phases equilibrium and where the other phase is continuously
removed. Open evaporation and open condensation are such processes.

Simplex - spatial configuration of (

-1) dimensions determined by

points in a space of dimen-

sion equal to or greater than (

-1). A triangle together with its interior determined by its

three vertices is a two-dimensional simplex in the plane.

Singular point - root to the first order derivative of a function. The singular points of the

residue curve Equations (3.7) are given by the solution of

. Azeotropes and pure

components are singular points of Equations (3.7).

Tie-line - see equilibrium vector.

Tie-line curve - sequence of conjugated vapor-liquid equilibrium vectors (tie-lines) in the com-

position space. Here referred to as a discrete distillation line, see distillation line.

Topology - a branch of mathematics concerned with those properties of geometric configurations

(as point sets) which are unaltered by elastic deformations (as a stretching or a twisting).

TTA - thermodynamic topological analysis.

124

Chapter 4. On the Topology of Ternary VLE Diagrams: Elementary Cells

4.2

Classification of ternary VLE diagrams

The number of feasible VLE diagram structures is limited by topological and thermodynamical
constraints. A complete classification of these feasible structures for ternary mixtures was given
by Serafimov (1970b) and Serafimov et al. (1971; 1973), and is more recently presented by
Serafimov (1996). Serafimov’s classification results in the 26 classes of ternary VLE diagrams
presented in Figure 4.1. The classification is based on the simplifying and common assump-
tion that there exists no more than one binary azeotrope for each binary pair and no more than
one ternary azeotrope. The classification of Serafimov considers topological structures and thus
does not distinguish between antipodal diagrams (switching of minimum- and maximum-boiling
azeotropes) since they have the same topology. Transition from one antipode to the other can
be made by simply changing the sign of the nodes and inverting the direction of the arrows of
increasing boiling temperature.

Gurikov (1958) was the first to derive the rule of azeotropy and propose a thermodynamic topolog-
ical classification of ternary mixtures. However, his classification was incomplete, and Serafimov
(1968d) revealed four additional feasible structures and established the 26 topological classes
presented in Figure 4.1. Later, Serafimov’s classification of ternary mixtures was refined by dis-
tinguishing between antipodes inside each structure class, based on the reasoning that ”minimum-
and maximum-boiling azeotropes have dissimilar physical nature and dissimilar behavior during
distillation” (Zharov and Serafimov, 1975). This refined classification includes a total of 49
types of ternary VLE diagrams. An even more detailed classification is proposed by Matsuyama
and Nishimura (1977) who also rank the components in the order of their boiling temperatures
(”light”, ”intermediate” and ”heavy”). This classification includes 113 diagram classes of which
87 are presented graphically by Doherty and Caldarola (1985). Nevertheless, among these 113
classes there are only the 26 topologically distinct structures of Serafimov. Actually, the classifi-
cation of Matsuyama and Nishimura adds some ambiguity as some of their classes with a ternary
saddle azeotrope have two or three possible topological structures. For example, their classifica-
tion code 112-S can be either of Serafimov’s class 3.1-3a or 3.1-3b depending on whether there
exists a saddle - saddle separatrix or not. This is also pointed out by Foucher et al. (1991) who
recommend an extension of the Matsuyama and Nishimura’s classification code name in these
cases.

In this paper, we use Serafimov’s (1970b) classification of the topological classes and Zharov
and Serafimov’s (1975) refinement of the antipodal structure types (referred to as the ZS-type).
The relationships between the classifications of Gurikov (1958), Serafimov (1970b), Zharov and
Serafimov (1975) and Matsuyama and Nishimura (1977) are presented in Table 4.2. The table is
useful when relating publications where the different classifications are used.

126

Chapter 4. On the Topology of Ternary VLE Diagrams: Elementary Cells

4.3

Elementary cells

There is a great diversity of VLE diagrams for ternary mixtures caused by the variety in physi-
cal properties of the components and their molecular interaction. As mentioned, if we consider
only topological differences there are 26 distinct types as presented in Figure 4.1. Furthermore,
there is a far greater diversity in possible shapes (geometry) of the simple phase transformation
trajectories such as residue curves and distillation lines. It is possible to reduce this complexity
to a combination of a few topological building blocks (“elementary cells”) and some basic inter-
nal structures (shapes of the simple phase transformation trajectories). We use residue curves to
represent the simple phase transformation trajectories in this paper.

We find that among Serafimov’s 26 topological classes there are eight elementary topological
cells (denoted I, II, III, IV, II’, III’, IV’ and V) that constitute all the ternary diagrams, where
a cell is defined as one residue curve region taken with its boundaries (that is, a subspace of the
composition space constrained by residue curve boundaries, if any, and the composition simplex).
From these eight elementary cells we may construct all the 26 diagrams as shown in Figure 4.2.
Each cell has one unstable and one stable node and some set of saddle points. There are four
“primary” diagrams where the composition triangle consists of a single cell (one residue curve
region; Serafimov’s class 0.0-1, 1.0-1a, 1.0-1b and 2.0-1), and these elementary cells (denoted I,
II, III, IV) are also the only so far reported for naturally occurring mixtures. Cells II’, III’ and IV’
are modifications of the primary cells II, III and IV, respectively (with internal nodes), for which
there are no reported physical mixtures. Cell V only occurs as an element in Serafimov’s class
3.1-1b, and also for this class there is so far no physical mixture reported.

The four primary diagrams and the corresponding elementary cells I, II, III, IV are shown in
Figure 4.3. Each elementary cell is characterized by a certain set and order of the singular points
(nodes and saddles) along the contour of its border (that is, by its topology). Accordingly we can
name them as given in Figure 4.3.

There is an important difference between an elementary cell of the “primary” diagrams (one
residue curve region) and an elementary cell incorporated into more complex diagrams. If an
elementary cell is a primary diagram, the saddle point is a pure component point. The borders
of the cell are the edges of the composition triangle (linear). The (stable and unstable) nodes are
pure component points or points of binary azeotropes. If an elementary cell is a constituent of
a complex diagram, at least one of its saddles is a binary or a ternary azeotrope, and, therefore,
at least one of the borders of the cell is a residue curve boundary (curved) showed by the solid
thick lines in Figure 4.2. One of the nodes can be a point of a ternary azeotrope. In general, the
composition space is broken into several residue curve regions (cells) if there are more than one
unstable node or more than one stable node.

Despite these differences, a single elementary cell and an elementary cell incorporated into a
complex diagram are topological equivalent. Note, however, that inside similar topological cells
there may be various shapes of the residue curves (simple phase transformation trajectories).

4.4 Occurrence in nature

129

for 1609 ternary mixtures (in which 1365 are azeotropic) based on thermodynamic data that were
published during the period from 1965 to 1988. To the best of our knowledge we have not found
any other publications that address this issue. The occurrence of the various classes as reported
by Serafimov and Reshetov are given in Table 4.1. (Reshetov’s statistics are also presented in
Figure 4.2). Note that Reshetov (1998) found that the three mixtures reported for class 3.1-1a in
the statistics by Serafimov (1968d) actually was of another structure class. From Reshetov’s data
we see that only 16 of Serafimov’s 26 classes are reported to occur in nature. If we also differ-
entiate between minimum- and maximum-boiling azeotropes (ZS-type), then we find that 27 of
the 48 classes are reported. The distribution reported in these studies does not necessarily reflect
the real occurrence in nature. The azeotropic data selection is small and occasional. Moreover,
the distribution can be distorted compared to the unknown natural distribution since the published
components data are results of deliberate search for entrainers for specific industrial separation
problems. Nevertheless, these data are interesting and can be used for some deductions:

Serafimov’s class 3.1-2 with three minimum-boiling binary azeotropes and one minimum-
boiling ternary azeotrope has the largest number of reported mixtures. About 26 % of the
1365 ternary azeotropic mixtures in the study by Reshetov are of this class.

Elementary cells I and II covers more than 90 % of all the reported ternary azeotropic
mixtures. The three most common structures are Serafimov’s class 1.0-1a (21.6 %), 2.0-2b
(21.0 %) and 3.1-2 (26.0 %). Among these three classes only cells I and II occur.

Ternary azeotropes are common in nature. About 40 % (37.9 %) of the reported ternary
mixtures have a ternary azeotrope. These are Serafimov’s class 1.1-2, 2.1-2b, 2.1-3a, 2.1-
3b, 3.1-2 and 3.1-4.

For each of the reported classes, one may find a great number of (similar) mixtures in nature.
However, for those classes with no reported mixtures, and with a structure that requires a set of
molecular interactions that are unlikely, we expect that few if any real mixtures will be found. The
experimental data for ternary azeotropic mixtures used for the above statistics is rather limited.
However, the experimental data for binary azeotropic pairs is considerably more extensive. For
example, Gmehling et al. (1994) has compiled data for 18 800 binary systems (involving about 1
700 components). From this we can estimate the behavior of a large number of ternary mixtures
combinations from a VLE model, but to our knowledge no systematic effort has yet been done.

It is well-known that binary maximum-boiling azeotropes are less abundant than minimum-
boiling azeotropes. According to Lecat (1949) the ratio of minimum-boiling versus maximum-
boiling azeotropes that occurs in nature is about 9 to 1. The statistics of Reshetov confirm this
heuristic rule. Thus, the more maximum-boiling binary azeotropes that are included in the ternary
mixture, the less is the probability of its occurrence. This is clearly demonstrated in Table 4.1.
In particular, no ternary mixture with three binary maximum-boiling azeotropes has been found
among Reshetov’s selection of 1365 mixtures, and no ternary maximum-boiling azeotrope has
been found. As a result, even for the topological structures where the existence is beyond ques-
tion, the occurrence of antipodes with maximum-boiling azeotropes is much less than that of
antipodes with minimum-boiling azeotropes.

130

Chapter 4. On the Topology of Ternary VLE Diagrams: Elementary Cells

Table 4.1: Occurrence of ternary VLE diagram structures found in published mixture data

Occurrence

Serafimov’s class

Serafimov

ZS-type

Set of azeotropes

Reshetov

(until 1968)

(1965-1988)

0.0-1

zeotropic

244

1.0-1a

min

283

max

1.0-1b

min

max

1.0-2

min

max

1.1-1a

None

min + min Az

None

max + max Az

None

1.1-1b

None

min + min Az

None

max + max Az

None

1.1-2

min + S

max + S

2.0-1

min + max

2.0-2a

None

min + max

max + min

2.0-2b

11a

min + min

280

21a

max + max

2.0-2c

11b

min + min

21b

max + max

2.1-1

None

min + max + min Az

None

min + max + max Az

None

2.1-2a

None

16a

min + max + min Az

None

16b

min + max + max Az

None

2.1-2b

min + min + min Az

max + max + max Az

None

2.1-3a

min + max + S

2.1-3b

min + min + S

max + max + S

3.0-1a

None

min + min + max

None

max + max + min

None

3.0-1b

None

min + min + max

max + max + min

3.0-2

min + min + min

114

max + max + max

None

3.1-1a

27b

min + min + max + min Az

None

32b

max + max + min + max Az

None

3.1-1b

None

min + min + max + min Az

None

max + max + min + max Az

None

3.1-1c

None

27a

min + min + max + max Az

None

32a

max + max + min + min Az

None

3.1-2

171

min + min + min + min Az

355

max + max + max + max Az

None

3.1-3a

None

25a

min + min + min + S

None

38a

max + max + max + S

None

3.1-3b

None

25b

min + min + min + S

None

38b

max + max + max + S

None

3.1-4

min + min + max + S

max + max + min + S

min - minimum boiling binary azeotrope; max - maximum boiling binary azeotrope;

- ternary node azeotrope; S

- ternary saddle azeotrope.

4.5 Use of elementary cells

131

P¨ollmann and Blass (1994) propose to reduce the number of ternary VLE diagrams by consid-
ering only ”physically meaningful” structures. Their list includes 19 of Serafimov’s 26 classes,
excluding 1.1-1a, 1.1-1b, 2.1-1, 2.1-2a, 3.1-1a, 3.1-1c, and 3.1-3b. However, it is impossible in
principle to state that some classes of ternary mixtures cannot exist in nature or are “physically
meaningless”, because all the structures in Figure 4.2 are thermodynamically and topologically
feasible. We can only discuss the probability of the existence of some types of the VLE diagram
structures.

4.5

Use of elementary cells

The concept of elementary topological cells is a simplification which primarily is made to re-
duce the number of structures of ternary VLE diagrams. It is useful for preliminary analysis
of azeotropic distillation (presynthesis). Based on the knowledge of the distillation behavior of
azeotropic mixtures of the primary diagrams elementary cells in Figure 4.3, or combinations of
these, we also have information about what behavior to expect from other more complex mixtures.
It is important to recognize that each real mixture has its own specific thermodynamic character-
istics and should therefore be analyzed in detail in a later step of the separation synthesis.

The concept of elementary cells is even more important when using the classification of Mat-
suyama and Nishimura with its 113 classes (less surveyable).

4.5.1

The album

One goal underlying the classification of ternary mixtures is to have a complete album of possible
VLE diagram structures with their corresponding scheme of separation by distillation. But this is
not established knowledge. Prediction of feasible distillation product compositions for even some
of the simplest diagram structures is still under development. Furthermore, methods or separation
schemes to separate all classes of (ternary) azeotropic mixtures are not established. One reason is
that there are many possible structures with deformation of the simple phase transformation paths
due to regions with different volatility order within the composition space, making this an almost
impossible task. Instead, we propose to consider selected VLE diagrams and specific mixtures of
these diagrams in detail:

1. Zeotropic mixture, ideal and nonideal with univolatility line(s)

(Serafimov’s class 0.0-1: cell I);

2. Mixture with one separatrix (one binary azeotrope)

(Serafimov’s class 1.0-2: combination of two cell I’s);

3. Mixture without separatrix, but with one binary azeotrope

(Serafimov’s class 1.0-1b: cell III);

132

Chapter 4. On the Topology of Ternary VLE Diagrams: Elementary Cells

4. Mixture without separatrix, but with two binary azeotropes nodes

(Serafimov’s class 2.0-1: cell IV).

An illustration of a ternary mixture with one binary saddle azeotrope (Class 1.0-2) consisting
of two elementary cell I’s is given in Figure 4.4. The left cell has “C-shaped” residue curves,

S-shape

C-shape

= 1

Figure 4.4: Ternary azeotropic mixture with one binary saddle azeotrope (Class 1.0-2 consisting
of two Cell I’s). Residue curves; unidistribution (solid) and univolatility (dash-dotted) lines are
given.

and the right cell has “S-shaped” residue curves (inflection) caused by the univolatility line

Although both cells are of type I, this difference in shape may have a large effect on the actual
separation process. For details on the internal structure (“shape”) of ternary VLE diagrams, the
reader is referred to Chapter 3 and Reshetov et al. (1999).

4.5.2

Multiple steady states

As an example of the use of elementary cells we may consider the possibility for multiple steady
states in (homogeneous) azeotropic distillation. Such multiplicities may lead to problems in col-
umn operation and control, as well as problems in column design and simulation. When two or
more multiple steady states exists for the same inputs it is possible that, for some disturbance,
the column profile jumps from the desirable (in terms of product specifications) to an undesir-
able steady state. However, such catastrophic jumps may be avoided by proper control of the
column, and the separation schemes for such mixtures may well be feasible and economical. The
possibility of multiple steady states at infinite efficiency of the distillation (infinite reflux, infinite
theoretical equilibrium trays, and

from 0 to 1) was first noted by Balashov et al. (1970).

They considered a mixture of Serafimov’s class 1.0-1b (primary diagram of elementary cell III
with U-shaped residue curves) with a binary maximum-boiling azeotrope (ZS-type 7b), and found
that for these mixtures it is feasible to have multiple products for the same value of the parameter

. Later, Petlyuk and Avet’yan (1971) analyzed this issue in more detail (including bifurca-

tion analysis). This analysis is also included in the textbook by Petlyuk and Serafimov (1983),
where Petlyuk writes “the existence of more than one saddle along a distillation line for sharp

4.6 Pseudo-component subsystem

133

(infinite reflux, infinite column) separation going from an unstable node to a stable node leads to
multiplicity of the separation products for the same feed composition and the same value of pa-
rameter

”. The geometrical considerations given by Petlyuk and coworkers are also found

in Serafimov et al. (1971). This is a sufficient, but not a necessary condition for multiplicity of the
mixture. Bekiaris et al. (1993) give a similar condition, roughly that the existence of two or more
neighboring saddles may lead to output multiplicity. Bekiaris et al. (1993) also identifies other
structural characteristics that may induce multiple steady states, such as highly curved distillation
boundaries (“pseudosaddles”), as for the mixture of acetone-chloroform-methanol.

The elementary cells that may lead to output multiplicity are III, III’, IV and IV’. From this we can
predict the possibility of multiple steady states (sufficient condition) for any given mixture that
is caused by these structural characteristics. From Figure 4.2 we see that 14 of the 26 diagrams
include these elementary cells (1.0-1b, 1.1-1a, 1.1-1b, 2.0-1, 2.0-2a, 2.0-2c, 2.1-1, 2.1-2a, 2.1-2b,
3.0-1b, 3.1-1a, 3.1-1b, 3.1-1c, 3.1-3a). From the statistics by Reshetov we find that about 7 % of
the reported VLE diagram structures include these cells. Serafimov’s class 2.1-2b, which includes
cell IV, is relatively common (4 %), and this class may give multiplicities for feeds in the region
with U-shaped residue curves.

The significance of multiplicities for column operation have been studied by Morari and cowork-
ers (Laroche et al., 1992a; Laroche et al., 1992b; Bekiaris et al., 1993; Bekiaris and Morari, 1996;
G¨uttinger and Morari, 1996; G¨uttinger and Morari, 1997). Mixtures with one binary minimum-
boiling azeotrope of Serafimov’s class 1.0-1b (primary diagram of cell III with U-shaped residue
curves) has been the focus, in particular the mixture of acetone - heptane - benzene. However,
from the occurrence statistics we see that Serafimov’s class 1.0-1b is not very common.

Bekiaris and Morari (1996) found multiple steady states for the mixture ethanol - ethyl propanoate
- toluene (Serafimov’s class 2.0-2a, ZS-type 18, which includes cells I and III) for a specific feed
region. The specific feed region corresponds exactly to elementary cell III.

4.6

Pseudo-component subsystem

An idea related to the concept of elementary cells is the concept of pseudo-components. This
approach is known to many engineers working with azeotropic distillation. Vogelpohl (1999)
proposes to analyze real azeotropic mixtures as subsystems approximated by zeotropic mixtures
where each azeotrope is represented by a pseudo-component. A residue curve region with

singular points (pure component and azeotropes) is thus represented as a

-component zeotropic

mixture by assuming constant relative volatilities between the real components and the pseudo-
components (azeotropes). Ideal distillation lines are calculated for each subsystem. However,
this strong simplification has major pitfalls. The ideal distillation lines (or residue curves) diverge
from the real (exact) ones as azeotropic mixtures necessarily have univolatility lines that deform
the simple phase transformation trajectories and cause S-,

-, and even more complex shapes

(internal structures) of the distillation lines (residue curves). For example, a ternary azeotropic
mixture of Serafimov’s class 1.0-2 (Figure 4.4) may be considered to be a quaternary system with

4.7 Conclusion

135

Table 4.2: Relationship between different classifications of ternary VLE diagrams

Gurikov (1958)

Serafimov (1970)

ZS-type

Matsuyama & Nishimura (1977)

Zeotropic

0.0-1

000

One binary

1.0-1a

100

azeotrope

030

1.0-1b

001

003

1.0-2

020

400

One binary

1.1-1a

200-m

and one

040-M

ternary

1.1-1b

002-m

azeotrope

004-M

1.1-2

010-S

300-S

Two binary

2.0-1

031, 103, 130

azeotropes

2.0-2a

023, 320

401, 410

2.0-2b

11a

102, 120, 021

21a

043, 430, 403

2.0-2c

11b

201, 210, 012

21b

034, 304, 340

Two binary

2.1-1

032-m, 230-m, 203-m

and one

041-M, 140-M, 104-M

ternary

2.1-2a

16a

420-m, 402-m

azeotrope

16b

024-m, 420-M

2.1-2b

022-m, 220-m, 202-m

044-M, 440-M, 404-M

10a

2.1-3a

013-S, 310-S, 301-S

10b

2.1-3b

011-S, 110-S, 101-S

033-S, 330-S, 303-S

Three binary

14a

3.0-1a

411

azeotropes

323

14b

3.0-1b

123, 321, 132, 213, 312, 231

413, 314, 431, 341, 134, 143

3.0-2

212, 122, 221

434, 344, 443

Three binary

3.1-1a

27b

421-m, 412-m

and one

32b

423-M, 324-M

ternary

3.1-1b

232-m, 223-m, 322-m

azeotrope

414-M, 441-M, 144-M

3.1-1c

27a

142-M, 241-M, 124-M, 214-M, 421-M, 412-M

32a

423-m, 324-m, 432-m, 342-m, 234-m, 243-m

3.1-2

222-m

444-M

3.1-3a

25a

121-S, 112-S, 211-S

38a

343-S, 334-S, 433-S

3.1-3b

25b

121-S, 112-S

38b

343-S, 334-S, 433-S

3.1-4

131-S, 113-S, 311-S

133-S, 313-S, 331-S

ZS-type refers to the refined classification of the 49 antipodal structures by Zharov and Serafimov (1975), p. 96-98

140

Chapter 5. Holdup Effect in Residue Curve Map Prediction of Batch Distillation Products

composition of the remaining liquid (reboiler residue) moves rectilinearly in the opposite direc-
tion from the invariant components. From this, it is possible to predict the sequence of constant
boiling vapor distillates which appears overhead, directly from the residue curve map without
solving any equation.

The use of residue curve maps for synthesis and design of batch distillation has been further
explored in a number of more recent papers (Van Dongen and Doherty, 1985a; Bernot et al., 1990;
Bernot et al., 1991; Ahmad and Barton, 1994; Ahmad and Barton, 1996; Ahmad, 1997; Ahmad et
al., 1998; Safrit, 1996; Safrit and Westerberg, 1997a; Safrit and Westerberg, 1997c). We refer to
the literature mentioned for a detailed description of the application of residue curve map analysis
to batch distillation.

In design studies of batch distillation, holdup on the trays is usually neglected, resulting in the
simple material balance:

d(H

)

(5.1)

where

is the liquid composition in the reboiler and

is the vapor composition in the distil-

late. Here

(5.2)

and by introducing the dimensionless time

, describing the relative loss of the liquid in the

reboiler, we get:

(5.3)

Equation (5.3) tells us that the distillate and reboiler compositions lie on a straight line that is
tangent to the change of the reboiler composition (i.e., the batch distillation reboiler composition
trajectory). The equation differs from the (open evaporation) residue curve equation in that

is not in equilibrium with

(

will be in equilibrium with the liquid in the condenser with

composition

In conventional batch distillation with a high reflux and a large number of equilibrium trays, it
is possible to divide the composition diagram into batch distillation regions defined as the set
of compositions that leads to the same set of product cuts (Ewell and Welch, 1945). For any
initial feed composition taken in a given region, leads to the same set of distillate cuts. The
batch distillation regions are similar to the residue curve regions, but are further subdivided by
straight “material balance” line boundaries (dashed lines in Figure 5.3). Ewell and Welch (1945)
differentiate between these straight line boundaries and the curved boundaries (separatrices) by
the fact that the straight lines cannot be crossed during batch distillation. The batch distillation
composition trajectories may “cross” a residue curve boundary as this is not a boundary for high
reflux batch distillation.

5.1 Introduction

141

There are two limiting conditions:

: batch distillation reboiler composition trajectory coincides with the residue curve.

: batch distillation reboiler composition trajectory coincides with the distillation line.

The distillation line boundary is located from the concave side of the curved residue curve bound-
ary in Figure 5.3 (see Chapter 3, Section 3.5.3).

Bernot (1990) used residue curve analysis to predict the product distribution in batch distilla-
tion using a simple dynamic model of the batch column with no holdup on the trays or in the
condenser, i.e. based on constant molar overflow (CMO) and quasi steady state (QSS) assump-
tions. However, there is always some holdup in a real column, and models of batch distillation
which ignore column holdup, such as Equation (5.3), may in some cases lead to significant errors
(Robinson, 1970). The overall material balance including column holdup may be written:

)

(5.4)

The amount of column holdup

varies with the column design (diameter, trays, packing

material etc.). A typical value of the total liquid holdup is about 5-10 % of the total column
volume for staged columns (Skogestad, 1992) and 2-4 % for packed columns. In our simulations
the amount of holdup was varied to see the effect on the distillate composition path (product
sequence). We considered the following column holdups as a percentage of the initial feed charge:
2.1 % (low), 4.1 % (normal) and 10.5 % (high). According to King (1980) one can employ the
quasi steady state (QSS) Equation (5.3) for batch distillation when the column holdup is less than
5 % of the charge to the reboiler (still).

Referring to the early work of Pigford and Rose (during the 1940-50s), Perry and Chilton (1973)
summarize the following two compensating effects of column holdup in batch distillation:

1. During startup, the reboiler composition is depleted in the more volatile component while

the column holdup is established. Thus, the separation becomes more difficult and the dis-
tillate composition is less “pure” than would be expected from neglecting column holdup.

2. Column holdup presents an inertia effect, which slows the composition dynamics and usu-

ally improves the separation.

Later, several authors have studied the column (liquid) holdup in batch distillation (e.g. Sørensen
(1994) and Mujtaba and Macchietto (1998)). The main conclusion is that the column holdup has
a large impact on the separation efficiency for difficult separations, and for separations with a
small amount of the light component in the initial feed (for regular batch distillation).

Our objective is to study the effect of column holdup in residue curve map prediction of batch
distillation products and off-cuts. We will show that the distillate product sequence may differ
from that found from an analysis neglecting holdup.

142

Chapter 5. Holdup Effect in Residue Curve Map Prediction of Batch Distillation Products

5.2

Model and data

A conventional batch distillation column is shown in Figure 5.1.

j+1

j-1

j+2

Reflux

Distillate

j+1

Accumulator

Reboiler

.
.

Condenser

Condenser Drum

j+1

Figure 5.1: Holdup on the trays in a conventional batch distillation column.

The assumptions and mathematical model equations are the same as used for the simulations
presented in Chapter 7 (Appendix 7A) assuming constant molar stage holdup:

(5.5)

This assumption is partly justified by the fact that the dominant composition dynamics are much
slower that the flow dynamics and nearly unaffected by the flow dynamics (Skogestad, 1992), and
it may therefore be used to get good estimates of the dominant response.

We introduce a dimensionless “warped time” (Van Dongen and Doherty, 1985a):

(5.6)

where

represents the distillate flowrate (denoted

in Appendix 7A) and

is the instanta-

neous liquid holdup in the reboiler.

5.3 Results and discussion

145

5.3

Results and discussion

In batch distillation of an azeotropic mixture several kinds of cuts can be obtained. We define
the “azeotropic cuts” as those containing an azeotrope and the “pure component cuts” as those
producing a (nearly) pure component.

We here focus on feeds located in the batch distillation region 5 in Figure 5.3 (

[0:3;

0:4;

0:3]

which from a residue curve map analysis gives a counterintuitive distillate sequence due to the
highly curved boundary of this region referred to as “distillation anomaly” (Van Dongen and
Doherty, 1985b; Bernot et al., 1990):

1. Azeotropic cut

(

53:3

)

2. Azeotropic cut

123

(

57:6

)

3. Azeotropic cut

(

53:3

)

4. Pure component 1 (acetone) cut (

56:4

)

5. Azeotropic cut

(

65:5

)

The “anomaly” is seen from the fact that the distillate temperature decreases as we go from cut

123

(

57:6

) and back to cut

(

53:3

). It is caused by the highly curved residue curve

boundary between batch distillation regions 2 and 5 (see Figure 5.3).

Note that this set of “product cuts” is mostly of theoretical interest as it is unlikely that one would
want to obtain four “unpure” azeotropic products in real operation. Nevertheless, our question is
whether introducing holdup on the trays will alter this sequence of cuts predicted by the residue
curve map analysis. We consider three cases of total liquid holdup in the column section as a
percentage of the initial feed charge to the reboiler: 2.1 % (low), 4.1 % (normal) and 10.5 %
(high).

The column specifications used in the simulations for the limiting conditions of high reflux and
large number of trays are given in Table 5.1.

Table 5.1: Column data for the limiting conditions of high reflux and large number of trays.

Total no. of trays

(excl. reboiler)

Reflux ratio

L=D

200

Initial reboiler holdup

10:000

kmol

Condenser holdup

specified in Table 5.2

Tray holdup

specified in Table 5.2

Boilup, vapor flow (constant)

5:0

kmol/h

146

Chapter 5. Holdup Effect in Residue Curve Map Prediction of Batch Distillation Products

The resulting distillate and reboiler composition trajectories are shown in Figures 5.4 - 5.6. The
corresponding product sequences are summarized in Table 5.2.

Table 5.2: Results of batch distillation with column holdup given as a percentage of the initial
feed charge to the reboiler.

Holdup

Sequence of product cuts

Case 1:

1. Azeotropic cut

2.1 % (low)

0:005

kmol

2. Azeotropic cut

123

3. Azeotropic cut

4. “Pure” acetone cut
5. Azeotropic cut

Case 2:

1. Azeotropic cut

4.1 % (normal)

0:01

kmol

2. Azeotropic cut

123

3. Azeotropic cut

4. Azeotropic cut

Case 3:

1. Azeotropic cut

10.5 % (high)

0:025

kmol

2. Azeotropic cut

123

3. Azeotropic cut

4. Azeotropic cut

Holdup on each stage and in the condenser

We see that the expected sequence of distillate products is obtained when the holdup is low, but
that the pure acetone product cut disappears when the holdup is medium and large. We also note
that the composition trajectories “cross” the residue curve boundary (stable separatrix), but not
the distillation line boundary (not shown).

5.4 Conclusion

151

Appendix 5A

Table 5.3: Thermodynamic data for the mixture acetone-water-chloroform (from Mayflower
[presently: HYSYS Conceptual Design Application

]).

Parameters for the Antoine Equation

Component

[

mol

]

1) Methanol

23.49989

3643.3136

-33.434

40.73

2) Acetone

21.3099

2801.53

-42.875

74.04

3) Chloroform

20.865

2696.900

-46.160

80.67

Binary Interaction Parameters for the Wilson Equation

161:8813

351:1964

583:1054

484:3856

1760:6741

28:8819

154

Chapter 6. Batch Distillation Columns and Operating Modes: Comparison

000 liters capacity are quite common according to Drew (Schweitzer, 1997). For batch times
exceeding about 12 hours, time losses due to startup and shutdown are not severe.

Batch distillation is generally more expensive than its continuous counterparts in terms of costs
per unit of product (Rousseau, 1987). However, for multicomponent mixtures with a large number
of components it may be less expensive to use batch distillation since only one column is required
for any number of products. A single batch column can separate a

-component mixture whereas

(

-1) continuous distillation columns would be required.

Batch distillation is also preferred when there are special operational conditions that makes con-
tinuous operation difficult, for example, processes that requires very long residence times, mate-
rials that are difficult to pump and processes where the equipment needs frequent cleaning.

6.2

Batch distillation column configurations

There are many possible configurations for batch distillation columns. The simplest example is
evaporation (without reflux), often called Rayleigh distillation. Of more practical importance is
batch distillation with reflux as described in the following.

6.2.1

Regular and inverted columns

In the regular batch column (also called the batch rectifier) illustrated in Figure 6.1a, the feed
vessel is located at the bottom of a rectifying column and product fractions are sequentially re-
moved overhead as distillate. In the inverted column (also called the batch stripper) illustrated in
Figure 6.1b, the feed vessel is located at the top of a stripping column and the product fractions
are sequentially removed in the bottom. For more detail about batch distillation in general we
refer to, for example, Diwekar (1995) and Perry (1997).

6.2.2

Middle vessel and multivessel columns

The combination of the regular and inverted batch columns connected by an intermediate vessel
(holdup tank), later named the middle vessel column, was proposed by Robinson and Gilliland
(1950) and is illustrated in Figure 6.1c. There are two principally different applications for the
middle vessel batch distillation column: (1) binary mixtures separation, and (2) ternary or mul-
ticomponent mixtures separation. In both cases, the mixture is usually charged to the middle
vessel initially, but it may also be distributed along the column or charged to the reboiler. For the
first application, the binary separation is performed simultaneously in the rectifying and stripping
sections. The light component is removed as a distillate product and the other component as a
bottoms product. The amount in the middle vessel decreases steadily during the distillation run.
This closely resembles continuous distillation. In the second application, which is the focus in
this thesis, the intermediate component(s) may be accumulated in the middle vessel whereas light

156

Chapter 6. Batch Distillation Columns and Operating Modes: Comparison

In 1976, Devyatikh and Churbanov [as referenced by Davidyan et al. (1994)] described separation
of ternary mixtures by the middle vessel batch distillation column.

In a series of papers, Davidyan, Kiva and Platonov (1991b; 1991a; 1992b; 1992a; 1993) presented
the use of batch distillation columns with a middle vessel for separating binary and ternary mix-
tures. Based on this work, Davidyan et al. (1994) presented an analytical analysis of the dynamic
behavior of the middle vessel batch distillation column for separating multicomponent mixtures.
The authors point out several advantages and new opportunities for the middle vessel configura-
tion as compared to conventional batch distillation column designs (Davidyan et al., 1994):

A binary mixture can be separated into pure components under finite reboil and
reflux ratio and, in some cases, faster than using conventional types of batch
distillation columns.

For ternary separation one can always choose parameters, like reflux and reboil
ratios, such that the composition in the middle vessel tends to either one of the
three components [e.g. the intermediate-boiling component for ideal mixtures].

A multicomponent mixture can be separated into heavy, intermediate and light
fractions simultaneously, so one can remove light and heavy impurities from
the mixture.

Meski and Morari (1995) analyzed the behavior of the middle vessel configuration for binary and
ternary mixtures at total reflux and minimum reflux conditions. They confirm the previous results
by Hasebe et al. (1992) that the middle vessel batch distillation column always perform better
than the conventional batch distillation design. They also conclude that it is optimal to operate
the column under constant reflux- and reboil ratios in the case of binary separation, keeping the
composition in the middle vessel constant at the initial feed composition with gradual decreasing
holdup of the middle vessel. Barolo and Guarise et al. (1996a; 1996b; 1996) address issues
on design and operation of the middle vessel batch distillation column for separating ternary
mixtures, partly based on experience from operating a continuous distillation pilot plant in a
batch mode.

In the multivessel column (also called multieffect column by Hasebe et al. (1995)) several batch
distillation columns are coupled by vapor and liquid streams with vessels (holdup tanks) between,
but with only one reboiler and one condenser. One objective is to minimize the energy consump-
tion, and this is analogous to integrated (multieffect) continuous columns like the Petlyuk column.
The Petlyuk column configuration is a fully thermally coupled distillation column arrangement
where two or more columns are linked together through vapor and liquid streams without reboil-
ers or condensers between the columns or simply one column with a dividing wall inside one
column shell. Such complex column arrangements are shown to offer large potential savings in
energy compared with sequences of simple columns for multicomponent continuous distillation
separation systems.

Several authors have presented comparative studies of conventional batch distillation columns
and the middle vessel batch configuration. Hasebe et al. (1992) compared the performance of

6.3 Operating policies

157

the middle vessel batch distillation column with conventional batch distillation for separating
ideal ternary mixtures. The middle vessel batch distillation column was shown to have the best
separation performance. The idea of using a middle vessel was later extended to a multivessel
batch column configuration with total reflux operation for separating more than three components
(Hasebe et al., 1995). They claim that the multivessel batch distillation column is potentially more
energy efficient than the corresponding sequence of continuous distillation columns for mixtures
with more than five components. However, their comparison was rather unfair since there was
no energy integration for the continuous columns. Hasebe et al. (1997) gave optimization results
on optimal vessel holdup policy in the closed multivessel column for ideal mixtures in terms of
minimum batch time. They proposed to charge the feed to the reboiler and gradually increase the
holdups of the other vessels up to pre-calculated (optimized) values and then run in total reflux
mode until the desired product purities are reached in each vessel.

Skogestad et al. (1997) proposed a simple feedback control strategy for operation of the closed
multivessel batch distillation column where the reflux flow out of each vessel is manipulated to
control the temperature at some location in the column section below. This indirectly adjust the
vessel holdups and there is no need for level control. The feasibility of this strategy was demon-
strated by simulations for an ideal quaternary mixture separated in a column with four vessels,
and later experimental verification was given for the zeotropic mixture of methanol, ethanol,
n-propanol and n-butanol (Wittgens and Skogestad, 1997). Up to 50 % reduction in energy con-
sumption was found separating a quaternary mixture using the closed multivessel column com-
pared to conventional batch distillation Wittgens and Skogestad (1998). Furlonge et al. (1999)
studied optimal (open) operation of the multivessel batch distillation column for quaternary ideal
mixtures and also found up to 50 % time savings compared with conventional batch distillation.

6.3

Operating policies

Determining the best way of running a batch distillation column is important in order to improve
the efficiency of energy consumption and the product quality, to reduce production time and to
reduce waste cuts. Sørensen (1997) found that time savings of up to 70 % can be made by using
alternative operating procedures.

Batch distillation is usually operated with production cycles consisting of charging, heat-up, equi-
libration, product draw-off, slop cut (off-cut removal), product draw-off, cool-down, dumping
and cleanup. The size of the off-cuts (if any) depends on the sharpness of separation between the
products. The sharpness of separation depends on holdup, relative volatility, reflux and number
of plates or amount of packing. Intermediate cuts are often recycled.

In this study we consider unconventional batch distillation operations and columns such as the
middle vessel column and extractive batch distillation. Thus, we need to define some basics for
the various ways of operating such columns.

158

Chapter 6. Batch Distillation Columns and Operating Modes: Comparison

6.3.1

Open operation (conventional)

There are various modes of operation of the different batch distillation columns. The conventional
operating policies for regular batch distillation columns are:

1. Constant reflux ratio (variable distillate composition)

2. Constant distillate composition (variable reflux ratio)

3. Optimal reflux ratio (usually variable reflux ratio)

The constant distillate composition policy (2) is usually implemented using feedback control with
either reflux flow or distillate flow as the manipulated variable. The reflux ratio policies (1 and 3)
are open loop operations, that is, predefined values are used without feed back from the process.

Constant reflux with varying distillate composition

With the constant reflux ratio policy, the reflux ratio is maintained constant throughout the batch.
It is the simplest policy to implement, but less efficient than the other variable reflux ratio policies.

If a batch distillation is to be run at constant reflux, then it follows that the distillate product
composition must change with time.

Constant distillate composition with variable reflux

With the constant distillate composition policy, the product is usually withdrawn at the maximum
possible rate consistent with there being sufficient reflux to maintain the product composition at
the desired value. As the batch progresses, the composition of the product tends to deteriorate and
the reflux ratio is increased gradually by reducing the product withdrawal rate, until the desired
composition can no longer be maintained. At this point the product is diverted to another receiver
and an intermediate cut (off-cut) is withdrawn usually at a constant high reflux ratio.

This procedure is repeated until all the product fractions have been removed. If the next batch
contains the same mixture components, the off-cuts are usually returned to the still together with
fresh charge.

6.3.2

Closed operation

The total reflux operation (no product withdrawal) is normally used during startup equilibration,
but it may be a mode of operation as well. Total reflux operation is used to get the best possible
separation, without regarding the length of time required for the revaporization of the reflux.

6.4 Comparison of cyclic and closed middle vessel columns

159

This operating mode is particularly useful when separating close boiling mixtures since it ob-
viates the necessity for keeping a close watch on the temperature and adjusting the reflux ratio
accordingly (Pratt, 1967). The temperature feedback approach of Skogestad et al. (1997) for
the multivessel column may also be successfully used for the closed operation of conventional
(two-vessel) distillation.

6.3.3

Cyclic operation

In cyclic operation (e.g. Sørensen and coworkers (1994; 1997; 1999)) there is a series of closed
operations with product withdrawal between each. Sørensen and Skogestad (1994) and Sørensen
and Prenzler (1997) stidued the optimal operation of the cyclic policy and comapred it to con-
ventional operating policies. The cyclic policy was found to be favorable for difficult separations
where small amounts of light product is to be recovered. More than 30 % reduction of the batch
time were found.

6.3.4

Semicontinuous operation

The middle vessel batch column operation where the feed is charged to the middle vessel which
is gradually emptied during operation, closely resembles continuous distillation. Sometimes, the
middle vessel column is operated with re-charging of the middle vessel during operation and a
more appropriate term for this operating mode may be semicontinuous operation because the feed
mixture is continuously supplied during parts of the batch operation.

Obviously, if the middle vessel does not have entering streams the operation is continuous (within
the time frame of emptying the feed vessel).

6.4

Comparison of cyclic and closed middle vessel columns

It is widely assumed that multieffect column configurations like the Petlyuk column in continuous
distillation and the multivessel column in batch distillation are most suited when the ideal ternary
mixture feed is rich in the intermediate-boiling component and when the product specifications
and relative volatilities are uniform (see for example Lestak and Collins (1997)). In this study we
examine the validity of this statement.

Hasebe et al. (1992) compared conventional batch distillation with open middle vessel batch
column for removal of light and heavy impurities from a ternary mixture feed rich in intermediate-
boiling component. They concluded that the middle vessel column is “effectively used when the
heavy impurity is easy to remove compared with the light impurity“.

In this study, we compare the batch time for the cyclic column and the closed middle vessel
column for separating ternary ideal mixtures, illustrated in Figure 6.2. The effects of mixture

160

Chapter 6. Batch Distillation Columns and Operating Modes: Comparison

volatilities, feed composition and feed charge location are investigated. Both columns are oper-
ated with indirect level control.

The underlying assumption is that there are economical benefits of minimizing the batch time,
either because this frees the equipment for other separations, or, because the energy consumption
(given constant heat input) is reduced. To minimize batch time the column is therefore operated
at maximum boilup (reboiler capacity). Thus, in this study the vapor flow is constant and equal
in all column sections. The ratio of the vapor flow relative to the total initial charge (

) is

a measure on how many times the feed charge is reboiled. In this study this is once per hour, i.e.

hr.

The larger the value of the relative volatility between a pair of components in a mixture

, the

easier it is to separate component

from component

. We have investigated theoretical ternary

mixtures of components 1-2-3 where the relative volatilities are set constant and equal for each
binary pair. Three values of the relative volatilities are considered: 1.2 (low), 2.0 (medium) and
5.3 (high).

The theoretical minimum batch time is given for the limiting case of infinite column sections,
which in this study is approximated by a very large number of equilibrium stages (with negligible
holdups). The operation is terminated when the compositions of the liquid in all the vessels satisfy
the specifications for the final products, that is, 99.0 mol % purity.

Reboiler

Condenser

Middle Vessel

Condenser

Reboiler

(b)

(a)

Figure 6.2: Closed batch distillation column alternatives for ternary mixtures separation: (a)
cyclic column (sequential separation); (b) closed middle vessel column (simultaneous separa-
tion).

6.4 Comparison of cyclic and closed middle vessel columns

161

6.4.1

Results

The resulting time consumption and separation performance of the cyclic two-vessel column (in-
direct split) for various initial feed compositions are given in Table 6.1. The corresponding time
consumption of the closed middle vessel column with difference feed charge locations (reboiler,
middle vessel, distributed according to the given feed composition, equally distributed in the ves-
sels) are given in Table 6.2, and the corresponding separation performances are given in Table
6.3. In most cases the middle vessel column is better than cyclic operation.

Ideally, we should have compared the middle vessel column with the best of the cyclic column
sequences (indirect or direct split). Furthermore, a measure such as the degree of difficulty of
separation proposed by Christensen and Jørgensen (1987) would have been better than relative
volatility to characterize the batch distillation column separations task. Nevertheless, the results
show an interesting tendency and the conclusions give a guideline for further investigations.

Time savings with the middle vessel column compared to cyclic operation are more signif-
icant for difficult separations (low

) for which savings up to about 50 % were obtained.

Time savings with the middle vessel column:

Difficult separation (

1:2

): up to about 50 %

Medium difficult separation (

2:0

): up to about 40 %

(saving varied from negative, that is more time consumption, to 46.5 %)

Easy separation (

5:3

): up to about 10 %

Largest savings were found for the following feed compositions:

Difficult separation (

1:2

): low in intermediate component

Medium difficult separation (

2:0

): low in intermediate component

Easy separation (

5:3

): rich in heavy component, but no pronounced difference

For difficult separations (low relative volatility), the middle vessel column is always better
than the direct split in the cyclic column, however the time savings are not very large for
feeds rich in light component.

For medium difficult separations (intermediate relative volatility), the benefits of the middle
vessel column are low for feeds rich in light component and feeds low in heavy component,
but there is a large time saving for feeds rich in heavy component.

For easy separations (high relative volatility), the middle vessel column has no pronounced
benefit, except large time savings for feeds rich in heavy component.

Least savings was found for feed compositions rich in intermediate.

162

Chapter 6. Batch Distillation Columns and Operating Modes: Comparison

The location of the initial feed charge affects the time consumption:

For difficult separations (low relative volatility), the best location depends on the feed com-
position, but the difference between the best and the worst cases is not very large (about
5-10 %). The reason that the location of the feed charge is not important to the batch time,
is that the initial period becomes small relative to the total batch time for low

mixtures

(that is, the impact of this “initial distribution” effect becomes less significant).

For medium difficult separations (intermediate relative volatility), the best location was
dependent on the feed composition. For an equimolar feed, feed charged to the reboiler
showed the best performance (minimum time). Interestingly, charging the feed to the inter-
mediate vessel was worst in all cases of feed composition.

For easy separations (high relative volatility), we found large time saving for feed charged
to the reboiler instead of to the middle vessel (in average 32.5 %).

6.4.2

Discussion

Meski et al. (1998) compared the middle vessel batch distillation column (open operation) with
the regular and inverted batch columns for binary and ternary ideal mixtures. They considered
both the direct and indirect split, and combinations of both in these columns and found that for
ternary mixtures:

The middle vessel column was best for feeds low in intermediate component.

The batch time increases sharply as the product specification of the intermediate component
become tighter (that is, it is time consuming to obtain high purity intermediate component
in the middle vessel column).

The middle vessel column was better than the direct split in the regular batch column for
feeds low in the intermediate component;

The regular and conventional columns were best for feeds rich intermediate component.

Our results on the cyclic and closed operation confirm and refine these findings, which contradicts
the results found by Hasebe et al. (1992).

Meski et al. (1998) conclude by proposing heuristic rules, similar to continuous distillation sepa-
ration sequencing, like “remove the most plentiful component first” and “easy separation first”.

Sørensen (1994) found for binary ideal mixtures that the cyclic operating policy is favorable for
difficult separations (with small amounts recovered) and feeds low in light component. Further,
Sørensen (1994) found that that feed low in light component is best separated in the inverted
column. This result was also confirmed by Meski et al. (1998). For ternary mixtures, Meski et
al. (1998) found as mentioned above that regular batch distillation (direct sequence) was best for

6.4 Comparison of cyclic and closed middle vessel columns

163

feeds rich in light component and low in heavy, and that the inverted column (indirect sequence)
was best for feeds rich in heavy component and low in light component.

For difficult separations with low relative volatility, the actual time to achieve the (high purity)
product specifications is closer to the time to reach steady state (“equilibrium time”) than for easy
separations (Sørensen and Skogestad, 1996). This further favors closed operation for mixtures
with low relative volatilities.

Our results also confirms the time savings found by Wittgens and Skogestad (1998) and Furlonge
et al. (1999) for ideal quaternary mixtures.

Note that the conclusions made in this study only hold if the number of stages in each column
section is sufficient for the separations. If the mixtures to be separated requires more stages than
the column actually have, the cyclic two-vessel column is better than closed three-vessel column.
In general, the closed multivessel column is only meaningful if number of stages is sufficient
for the purity specifications. The effect of liquid column holdup was not included in our study,
but Mujtaba and Macchietto (1998) found that the minimum batch time for conventional binary
batch distillation (regular columns) increased from a few percent up to 15-20 % depending on the
difficulty of separation. The effect was larger for easier separations.

In practice, the time consumption for the cyclic two-vessel column will be even higher because
of the time used to discharge the condenser vessel between the cycles. In addition, distillation
columns are usually operated with a safety margin, that is, the actual product compositions are
higher than the specifications to cope with process upsets and other problems (Humphrey and
Seibert, 1992). Thus, the cyclic column will in practice consume extra time since each cycle of
the process will be operated longer than the final time reported in this study. This added time
consumption will be less for the closed three-vessel column since the separation is done in one
operation (that is, one only needs to run the process once). This is an additional advantage in
terms of time (energy) consumption for the multivessel column.

6.4.3

Comparison with integrated continuous distillation

Table 6.4 gives the minimum energy requirement for ordinary continuous distillation (direct and
indirect sequence) and Petlyuk column (minimum boilup with infinite number of equilibrium
stages, sharp split analytical expressions) (Halvorsen and Skogestad, 1999). The energy con-
sumption

min

for continuous distillation can be compared directly with

for batch

distillation as it in both cases represents the numer of times the feed needs to be reboiled. In our
case

which means that the time consumption

[hr] for batch distillation (batch

time) can be compared directly with

min

. As expected, the energy consumption in continu-

ous Petlyuk distillation (

min

in Table 6.4) is significantly lower than for middle vessel batch

distillation (minimum batch time in Table 6.2). We find that

min

is less than

0:4

in most

cases. Moreover, interestingly, we note that there are similar trends for the Petlyuk column and
the middle vessel column: they perform best for difficult separations (low

) and for feeds low

in the intermediate-boiling component (but for high

they are best for feeds rich in the heavy

6.4 Comparison of cyclic and closed middle vessel columns

165

Table 6.1: Time consumption and separation performance in cyclic two-vessel batch distillation
column for various constant relative volatilities and initial feed compositions

Low relative volatility mixture

Time [hr]

Impurities %

1:2

total

cycle 1

cycle 2

in B/D

Feed 0

equimolar

22.79

13.15

9.64

1.0/1.0/0.0

Feed 1

rich in light component

16.21

13.41

2.80

1.0/0.2/0.0

Feed 2

rich in intermediate component

41.82

11.44

30.38

0.0/1.0/0.0

Feed 3

rich in heavy component

26.30

11.70

14.60

0.4/1.0/0.0

Feed 4

low in light component

23.89

10.87

13.02

1.0/1.0/0.0

Feed 5

low in intermediate component

20.20

11.48

8.72

0.6/1.0/0.1

Feed 6

low in heavy component

32.57

14.16

18.41

0.0/1.0/0.0

Average

(for all given feed compositions)

26.25

Medium relative volatility mixture

Time [hr]

Impurities %

2:0

total

cycle 1

cycle 2

in B/D

Feed 0

equimolar

4.91

2.71

2.20

1.0/1.0/0.0

Feed 1

rich in light component

3.13

2.47

0.66

1.0/1.0/0.0

Feed 2

rich in intermediate component

4.60

2.33

2.27

1.0/1.0/0.0

Feed 3

rich in heavy component

9.15

1.92

7.23

0.0/1.0/0.0

Feed 4

low in light component

5.04

2.08

2.96

1.0/1.0/0.0

Feed 5

low in intermediate component

7.05

2.34

4.17

0.0/1.0/0.0

Feed 6

low in heavy component

4.65

3.13

1.52

1.0/1.0/0.0

Average

(for all given feed compositions)

5.50

High relative volatility mixture

Time [hr]

Impurities %

5:3

total

cycle 1

cycle 2

in B/D

Feed 0

equimolar

1.75

0.90

0.85

1.0/1.0/0.0

Feed 1

rich in light component

2.41

1.69

0.72

0.0/1.0/0.0

Feed 2

rich in intermediate component

2.28

0.60

1.68

1.0/1.0/0.0

Feed 3

rich in heavy component

2.13

0.37

1.76

0.0/1.0/0.0

Feed 4

low in light component

1.66

0.50

1.16

1.0/1.0/0.0

Feed 5

low in intermediate component

1.12

1.03

1.09

0.0/1.0/0.0

Feed 6

low in heavy component

2.14

1.17

0.97

1.0/1.0/0.0

Average

(for all given feed compositions)

1.93

Feed 1: [0.80 0.10 0.10], Feed 2: [0.10 0.80 0.10], Feed 3: [0.10 0.10 0.80], Feed 4: [0.10 0.45 0.45],

Feed 5: [0.45 0.10 0.45], Feed 6: [0.45 0.45 0.10]

166

Chapter 6. Batch Distillation Columns and Operating Modes: Comparison

Table 6.2: Time consumption in closed middle vessel batch distillation column for various con-
stant relative volatilities, initial feed charge locations and initial feed compositions

Low relative volatility mixture

Time consumption [hr] with feed charge in

1:2

reboiler

middle

according

equally

vessel

to feed

distributed

Feed 0

equimolar

12.63

13.24

13.22

Feed 1

rich in light component

14.25

14.28

13.25

13.41

Feed 2

rich in intermediate component

16.85

16.21

16.34

16.65

Feed 3

rich in heavy component

11.12

11.79

11.12

11.25

Feed 4

low in light component

13.31

15.32

14.07

14.69

Feed 5

low in intermediate component

8.23

8.13

7.27

Feed 6

low in heavy component

15.81

15.60

14.07

14.75

Average

(for all given feed compositions)

13.17

13.52

12.76

13.03

Medium relative volatility mixture

Time consumption [hr] with feed charge in

2:0

reboiler

middle

according

equally

vessel

to feed

distributed

Feed 0

equimolar

3.23

4.31

4.05

Feed 1

rich in light component

2.90

2.96

4.45

3.48

Feed 2

rich in intermediate component

3.39

5.37

5.32

5.04

Feed 3

rich in heavy component

2.33

2.48

2.28

1.95

Feed 4

low in light component

3.09

4.86

4.49

4.61

Feed 5

low in intermediate component

2.19

2.36

2.27

2.48

Feed 6

low in heavy component

4.72

5.22

4.44

4.71

Feed 7

rich in light, low in intermediate

2.30

2.98

3.74

2.59

Feed 8

rich in light, low in heavy

4.23

4.47

4.67

3.92

Feed 9

rich in heavy, low in intermediate

2.28

2.26

1.96

2.13

Feed 10

rich in heavy, low in light

3.06

3.95

3.15

3.66

Feed 11

rich in intermediate, low in light

2.91

5.44

5.30

5.19

Feed 12

rich in intermediate, low in heavy

4.74

5.64

5.28

5.19

Average

(for all given feed compositions)

3.18

4.02

3.95

3.77

High relative volatility mixture

Time consumption [hr] with feed charge in

5:3

reboiler

middle

according

equally

vessel

to feed

distributed

Feed 0

equimolar

1.92

2.61

2.45

Feed 1

rich in light component

2.13

2.12

2.90

2.08

Feed 2

rich in intermediate component

1.82

3.29

2.90

Feed 3

rich in heavy component

0.69

2.01

0.70

1.39

Feed 4

low in light component

1.29

2.88

2.64

2.68

Feed 5

low in intermediate component

1.42

1.51

1.96

1.73

Feed 6

low in heavy component

2.63

3.21

2.70

2.89

Average

(for all given feed compositions)

1.70

2.52

2.32

2.30

Feed 1: [0.80 0.10 0.10], Feed 2: [0.10 0.80 0.10], Feed 3: [0.10 0.10 0.80], Feed 4: [0.10 0.45 0.45],

Feed 5: [0.45 0.10 0.45], Feed 6: [0.45 0.45 0.10], Feed 7: [0.60 0.10 0.30], Feed 8: [0.60 0.30 0.10],

Feed 9: [0.30 0.10 0.60], Feed 10: [0.10 0.30 0.60], Feed 11: [0.10 0.60 0.30], Feed 12: [0.30 0.60 0.10]

6.4 Comparison of cyclic and closed middle vessel columns

167

Table 6.3: Separation performance of closed middle vessel batch distillation column correspond-
ing to results in Table 6.2

Low relative volatility mixture

Impurities % in B/M/D with feed charge in

1:2

reboiler

middle

according

equally

vessel

to feed

distributed

Feed 0

equimolar

0.1/1.0/0.0

0.0/1.0/0.0

Feed 1

rich in light component

0.0/0.0/1.0

0.0/0.2/1.0

0.0/0.1/1.0

Feed 2

rich in intermediate component

0.0/1.0/0.0

Feed 3

rich in heavy component

1.0/0.1/0.0

1.0/0.0/0.0

1.0/0.1/0.0

Feed 4

low in light component

1.0/0.2/0.0

0.0/1.0/0.0

0.1/1.0/0.0

0.0/1.0/0.0

Feed 5

low in intermediate component

1.0/0.4/0.4

0.5/0.0/1.0

0.4/1.0/0.6

0.2/1.0/0.3

Feed 6

low in heavy component

0.0/1.0/0.0

0.0/1.0/0.3

0.0/1.0/0.1

Medium relative volatility mixture

Impurities % in B/M/D with feed charge in

2:0

reboiler

middle

according

equally

vessel

to feed

distributed

Feed 0

equimolar

0.2/1.0/0.0

0.0/1.0/0.0

0.0/1.0/0.1

Feed 1

rich in light component

0.0/1.0/0.0

0.0/0.4/1.0

Feed 2

rich in intermediate component

0.0/1.0/0.0

Feed 3

rich in heavy component

1.0/0.0/0.0

0.0/1.0/0.0

1.0/0.0/0.0

0.6/1.0/0.0

Feed 4

low in light component

1.0/0.4/0.0

0.0/1.0/0.0

Feed 5

low in intermediate component

1.0/0.4/0.0

0.0/1.0/0.0

0.0/0.9/1.0

0.0/1.0/0.9

Feed 6

low in heavy component

0.0/1.0/0.0

0.0/1.0/0.4

0.0/1.0/0.2

Feed 7

rich in light, low in intermediate

0.4/1.0/0.0

0.0/0.7/1.0

0.0/1.0/0.0

Feed 8

rich in light, low in heavy

0.0/1.0/0.0

0.0/0.2/1.0

0.0/1.0/0.8

Feed 9

rich in heavy, low in intermediate

1.0/0.1/0.0

0.0/1.0/0.0

1.0/1.0/0.5

0.3/1.0/0.5

Feed 10

rich in heavy, low in light

1.0/0.1/0.0

0.0/1.0/0.0

0.4/1.0/0.0

0.1/1.0/0.0

Feed 11

rich in intermediate, low in light

1.0/0.9/0.0

0.0/1.0/0.0

Feed 12

rich in intermediate, low in heavy

0.0/1.0/0.0

High relative volatility mixture

Impurities % in B/M/D with feed charge in

5:3

reboiler

middle

according

equally

vessel

to feed

distributed

Feed 0

equimolar

0.0/1.0/0.0

0.0/1.0/0.1

Feed 1

rich in light component

0.0/1.0/0.0

0.0/0.5/1.0

Feed 2

rich in intermediate component

1.0/0.3/0.0

0.0/1.0/0.0

1.0/0.1/0.0

Feed 3

rich in heavy component

1.0/0.1/0.0

0.0/0.1/0.0

0.8/1.0/0.0

0.0/1.0/0.0

Feed 4

low in light component

1.0/0.3/0.0

0.0/1.0/0.0

Feed 5

low in intermediate component

0.0/1.0/0.0

0.0/1.0/0.9

Feed 6

low in heavy component

0.0/1.0/0.0

0.0/1.0/0.3

0.0/1.0/0.1

Feed 1: [0.80 0.10 0.10], Feed 2: [0.10 0.80 0.10], Feed 3: [0.10 0.10 0.80], Feed 4: [0.10 0.45 0.45],

Feed 5: [0.45 0.10 0.45], Feed 6: [0.45 0.45 0.10], Feed 7: [0.60 0.10 0.30], Feed 8: [0.60 0.30 0.10],

Feed 9: [0.30 0.10 0.60], Feed 10: [0.10 0.30 0.60], Feed 11: [0.10 0.60 0.30], Feed 12: [0.30 0.60 0.10]

168

Chapter 6. Batch Distillation Columns and Operating Modes: Comparison

Table 6.4: Minimum energy in continuous and Petlyuk distillation columns for various constant
relative volatilities and initial feed compositions

. Minimum boilup with infinite stages columns.

Low relative volatility mixture

min

1:2

Direct

Indirect

Petlyuk

Savings

sequence

column

Feed 0

equimolar

7.87

7.83

4.49

42.6

Feed 1

rich in light component

6.45

8.00

5.35

17.1

Feed 2

rich in intermediate component

10.11

5.61

44.5

Feed 3

rich in heavy component

7.31

5.68

4.68

17.6

Feed 4

low in light component

8.74

7.93

5.18

34.7

Feed 5

low in intermediate component

6.49

6.38

3.64

42.9

Feed 6

low in heavy component

8.30

9.08

5.10

38.6

Average

(for all given feed compositions)

7.90

7.86

4.86

34.0

Medium relative volatility mixture

min

2:0

Direct

Indirect

Petlyuk

Savings

sequence

column

Feed 0

equimolar

2.07

2.03

1.37

32.8

Feed 1

rich in light component

2.01

2.21

1.71

15.0

Feed 2

rich in intermediate component

2.73

1.83

33.0

Feed 3

rich in heavy component

1.52

1.24

1.04

16.1

Feed 4

low in light component

2.13

2.00

1.45

27.5

Feed 5

low in intermediate component

1.70

1.60

1.05

34.4

Feed 6

low in heavy component

2.37

2.47

1.57

33.6

Feed 7

rich in light, low in intermediate

1.82

1.85

1.32

27.4

Feed 8

rich in light, low in heavy

2.21

2.36

1.51

31.6

Feed 9

rich in heavy, low in intermediate

1.60

1.40

1.00

28.6

Feed 10

rich in heavy, low in light

1.86

1.68

1.28

23.8

Feed 11

rich in intermediate, low in light

2.39

2.32

1.62

30.2

Feed 12

rich in intermediate, low in heavy

2.52

2.58

1.68

33.3

Average

(for all given feed compositions)

2.07

2.04

1.42

28.3

High relative volatility mixture

min

5:3

Direct

Indirect

Petlyuk

Savings

sequence

column

Feed 0

equimolar

0.98

0.96

0.81

16.1

Feed 1

rich in light component

1.16

1.17

1.01

12.7

Feed 2

rich in intermediate component

1.32

1.11

15.9

Feed 3

rich in heavy component

0.48

0.41

0.36

11.3

Feed 4

low in light component

0.90

0.88

0.75

14.6

Feed 5

low in intermediate component

0.81

0.76

0.63

16.9

Feed 6

low in heavy component

1.24

1.25

1.04

16.4

Average

(for all given feed compositions)

0.98

0.96

0.82

14.8

Feed 1: [0.80 0.10 0.10], Feed 2: [0.10 0.80 0.10], Feed 3: [0.10 0.10 0.80], Feed 4: [0.10 0.45 0.45],

Feed 5: [0.45 0.10 0.45], Feed 6: [0.45 0.45 0.10], Feed 7: [0.60 0.10 0.30], Feed 8: [0.60 0.30 0.10],

Feed 9: [0.30 0.10 0.60], Feed 10: [0.10 0.30 0.60], Feed 11: [0.10 0.60 0.30], Feed 12: [0.30 0.60 0.10]

170

Chapter 7. Closed Multivessel Batch Distillation of Ternary Azeotropic Mixtures

Conventional operation of multicomponent batch distillation columns requires timely switching
between products and off-cuts and often a variable reflux ratio policy to ensure the wanted product
purities. This may require a lot of attention from the operators and a complicated control system.
Closed operation of multivessel batch distillation columns with the simple indirect level control
strategy proposed by Skogestad et al. (1997) eliminates these difficulties and the products are
separated in an energy efficient, stand-alone operation.

By closed multivessel batch distillation we mean a “total reflux” operation where the products
are accumulated in vessels along the column (including the reboiler and condenser vessel). There
are no input or output streams, but the internal liquid reflux flows may vary during the operation
causing accumulation or depletion of mass in the vessels. Thus, closed system is a more correct
term than “total reflux” since we generally only achieve total reflux as time approaches infinity,
i.e., at steady state. We use the term multivessel column as a generalized configuration of all batch
distillation columns. In this context the regular batch distillation operation (also called the batch
rectifier), where the feed is charged to the reboiler and the products are withdrawn as distillate
during operation, is a two-vessel column. Correspondingly, open operation of the inverted batch
distillation column (batch stripper) and the middle vessel batch column (a combined system of
a batch rectifier and a batch stripper where distillate and bottoms products are withdrawn from
the column ends) are both two-vessel columns. However, the middle vessel column where the
products are accumulated in the condenser, middle vessel and reboiler is a three-vessel column.

Relation to previous work

Simple total reflux operation until equilibrium followed by total distillate draw-off has been used
in laboratory batch distillation for a long time (see for example Perry and Chilton (1963)). A re-
peated sequence of such total reflux operations is called the cyclic operating policy and is used for
separation of multicomponent mixtures (Perry and Chilton, 1973). Treybal (1970) and Bortolini
and Guarise (1970) independently investigated closed total reflux operation of the conventional
batch distillation column with a condenser vessel (i.e., a two-vessel column) for separation of
ideal binary mixtures. In both studies, total liquid reflux with pre-fixed condenser vessel holdup
was suggested. This requires exact knowledge of the initial feed composition and charge to
the column in order to pre-calculate the vessel holdups so that the wanted product purities are
achieved at steady state. Kiva et al. (1988) extended this idea to separation of multicomponent
and azeotropic mixtures in the two-vessel batch distillation column operated in a sequence of to-
tal reflux operations where the accumulated condenser (or reboiler) vessel product is discharged
between each closed operation, i.e., the cyclic operating policy. In addition to general analytical
considerations, experimental results on the ternary azeotropic mixture of acetone, isopropanol
and water were reported. Sørensen and Skogestad (1994) and Sørensen (1999) found that the
cyclic operating policy for separation of multicomponent mixtures in the two-vessel column re-
quires significantly less batch time for some ideal mixtures as compared to the conventional open
operating strategies. Closed operation of the middle vessel configuration for separation of ternary
mixtures (i.e., a three-vessel column) was proposed by Bortolini and Guarise (1970). A further
generalization is the multivessel column for separation of multicomponent mixtures suggested by
Hasebe et al. (1995) which they call “multieffect batch distillation system”. In this configura-

7.2 Columns and operating modes for ternary separation

171

tion it is possible to separate a multicomponent mixture simultaneously in one closed operation.
Skogestad et al. (1997) proposed a novel feedback control strategy for closed multivessel batch
distillation columns (described in Chapter 6). The feasibility of this strategy was demonstrated
by simulations for an ideal quaternary mixture separated in a column with four vessels, and later
experimental verification was given for the zeotropic mixture of methanol, ethanol, n-propanol
and n-butanol (Wittgens and Skogestad, 1997).

In this study we extend some of the above considerations to azeotropic mixtures. Note that closed
multivessel batch distillation does not offer a method of azeotropic mixtures separation, but only
a simple way to operate the batch distillation process for multicomponent mixtures in general.
However, new aspects on the operation and control of the closed multivessel batch distillation
column are introduced when azeotropes are involved.

7.2

Columns and operating modes for ternary separation

There are basically two column alternatives for separation of ternary mixtures in the closed mul-
tivessel batch distillation scheme. One is the cyclic operation of the regular (two-vessel) batch
distillation column shown in Figure 7.1 (referred to as the cyclic column or cyclic operation in the
following), where the products are separated one at the time in a sequence of closed operations.
Three alternative splits of an ideal ternary mixture may be performed: the direct split 1/23; the in-

Condenser

Reboiler

Cycle 2

Cycle 1

Figure 7.1: Cyclic two-vessel batch distillation of ternary zeotropic mixture. Sequential sepa-
ration of all three components (direct sharp split illustrated). Steady state composition profiles
(solid line with open circles along the edges of the composition triangle) and material balance
line (dashed).

direct split 12/3; and the distributed split 12/23 where each vessel product must be re-processed in
order to achieve all three components as pure products. In this study, we only consider the direct

172

Chapter 7. Closed Multivessel Batch Distillation of Ternary Azeotropic Mixtures

(not necessarily sharp) split illustrated in Figure 7.1. The distillate product in the condenser after
the first cycle (with composition

) is discharged from the condenser vessel. The second cycle

is then mainly a binary split between the remaining components into a second distillate product
(with composition

) and a bottom product (with composition

) in the reboiler. Column

holdup may require the removal of an off-cut fraction between the cycles in order to achieve the
desired product purities.

There are at least two advantages with the cyclic column compared to the regular open batch
distillation operation where the products are continuously withdrawn overhead, one at a time.
First, the closed operation is simpler since there is a definite distinction between the product
change-overs (i.e., the condenser is discharged by on/off distillate withdrawal) and it is easier
to assure the product qualities. Second, the “total reflux at steady state” operation utilizes the
maximum attainable separation in the column (at least for ideal and nearly ideal mixtures) and is
especially advantageous for high purity distillations.

The other closed multivessel column alternative for ternary mixtures separation is the closed mid-
dle vessel (three-vessel) column shown in Figure 7.2 (referred to as middle vessel column in the
following), where all three components (or azeotropes) are separated simultaneously during one

+ 1

Reboiler

Condenser

Middle Vessel

Figure 7.2: Closed middle vessel (three-vessel) batch distillation of ternary zeotropic mixture.
Simultaneous separation of all three components. Steady state composition profile (solid line
with open circles) and material balance lines (dashed).

closed operation. The three vessel products (with compositions

and

) are discharged

from the vessels at the end of the batch operation. The main advantage of the closed vessel config-

7.3 VLE diagrams and steady state composition profiles

173

uration is the simplicity of operation since no product change-overs are required during operation.
In addition, it may imply energy savings as compared to both the regular open batch column and
the cyclic column due to the multieffect (mass and heat integrated) nature of the process. Fur-
ther, there is no need to remove an off-cut fraction as may be the case in the cyclic column for
separating ternary mixtures. However, saddles (component 2 in Figure 7.2) are difficult to obtain
as pure products and a large number of equilibrium stages may be required to obtain the product
specifications of such intermediate fractions.

Some loss of products due to column holdup is unavoidable in batch distillation in general. To
avoid that liquid holdup in the column sections contaminates the vessel products after the distilla-
tion is completed, there must be a shut-off valve on the liquid reflux into each vessel. This applies
to both cyclic and closed middle column operation.

7.3

VLE diagrams and steady state composition profiles

There are various ways to graphically represent the VLE behavior of a mixture and its relation-
ship to the distillation process. McCabe-Thiele diagrams for binary systems are probably the most
known graphical tool. For ternary and multicomponent mixtures, residue curve maps and distilla-
tion line maps are two closely related VLE representations that are commonly used, particularly
in nonideal and azeotropic distillation analysis. A distillation line is defined as a connection line
through a sequence of equilibrium mapping vectors in the composition space (Zharov, 1968c;
Stichlmair, 1988) and corresponds to the liquid (or vapor) composition profile at total reflux for
equilibrium staged columns. The closely related residue curves (Schreinemakers, 1901c; Ost-
wald, 1902; Zharov, 1967; Serafimov, 1968a; Doherty and Perkins, 1978a) coincide exactly with
total reflux packed distillation column profiles when all resistance to mass transfer is in the va-
por phase (Serafimov et al., 1973; P¨ollmann and Blass, 1994). This make distillation line maps
and residue curve maps particularly suited to identify feasible products at the limiting operating
condition of total reflux. In the closed operating regime, the correspondence between the VLE
diagram and the steady state of the closed multivessel batch distillation column is simple and
straightforward:

The steady state vessel product compositions of a closed multivessel batch distillation
column are connected by a distillation line with the given number of equilibrium
trays. In addition the material balances must be satisfied.

For the closed middle vessel (three-vessel) batch distillation column shown in Figure 7.2, the
steady state material balances become (in the case of zero or negligible column holdup):

(7.1)

(7.2)

which gives

)

(7.3)

174

Chapter 7. Closed Multivessel Batch Distillation of Ternary Azeotropic Mixtures

The same equations applies to the steady state of each cycle in the cyclic conventional (two-vessel)
column shown in Figure 7.1 by simply setting the middle vessel holdup

equal to zero. From

Equation (7.3) we see that the feed composition is a linear combination of the vessel products. The
vessel holdups must obey the so-called lever rule. The material balance is visualized by straight
dashed lines in Figure 7.1 and 7.2 that connect the vessel products with the feed composition.
The steady state compositions in the condenser, middle vessel and reboiler (

and

respectively) must lie on the same distillation line given by the vapor-liquid equilibrium mapping:

(7.4)

where

is the equilibrium mapping function (see Chapter 3). Consequently, distillation lines

may be used to (graphically or analytically) determine the number of equilibrium trays needed
for a given separation to convert

(or vice versa) given as the multiple of the vapor-liquid

equilibrium phase mappings (Kiva et al., 1988) in the same fashion as Fenske’s equation. Thus, in
this study the distillation line map analysis is not only a qualitative tool but also a quantitative tool.
Note that if

is specified, the number of equilibrium trays required is infinite,

min

. The number of equilibrium trays determines the achievable separation in the column and the

steady state vessel holdups determines the purity of the product fractions (Kiva et al., 1988).

In summary, the feasible region of vessel compositions in the middle vessel column are enclosed
by the feed distillation line and the borders of the current feed distillation region and by the
material balance triangle connecting the unstable node, the stable node and one of the saddles of
the current feed distillation region. In the cyclic column, the material balance line that represents
the total reflux steady state separation performed in each cycle is a chord of the distillation line
joining the unstable node and the stable node of the current feed distillation region.

7.4

Mixtures studied

Ternary mixtures of elementary cells I and II are considered, represented by the mixtures given in
Figure 7.3 and 7.4. In addition, a mixture with a distillation line boundary (separatrix) that con-
sists of two cell I’s is considered, represented by the mixture given in Figure 7.5. The elementary
cell concept is described in Chapter 4.

The simplest example of Cell I is Serafimov’s class 0.0-1. A mixture that represent this class
is the zeotropic system of methanol, ethanol and 1-propanol shown in Figure 7.3. Note that
the distillation lines in Figure 7.3 are drawn as smooth curves (spline interpolation) through the
chains of conjugated equilibrium vectors according to the definition by Stichlmair (1988). For the
other example mixtures presented in this paper, the distillation lines are drawn with solid straight
tie-lines between each point of conjugation. Liquid isotherms (dotted lines) are contour plots of
the boiling temperature surface.

The simplest example of Cell II is Serafimov’s class 1.0-1a with one binary azeotrope and where

7.6 Simulation results

177

control from the column section below

j;m

j;sp

)

(7.5)

where

is the bias term,

is the proportional gain,

j;m

is the measured temperature at some

tray

in the column section below the actual vessel and

j;sp

is the temperature set-point. Note

that there is no explicit level control, rather the holdup in each vessel is adjusted indirectly to meet
the temperature specifications. As an initial choice of the proportional gain we used

(7.6)

where

is the nominal vapor flow (in this study kept constant) and

is the boiling point

difference between the two main components (or azeotropes) being separated in that section. The
controller bias was set equal to the vapor flow

so that total reflux was achieved at steady

state. This eliminates the need for using integral action to avoid steady state offset. In practical
implementation of the indirect level control strategy, a PI-controller would be used since there
are no direct measurement of the vapor flowrate in the column (Wittgens and Skogestad, 1998).
The temperature measurement was taken on the middle tray in the column section since this
temperature is more sensitive to composition changes than it is in, for example, the vessel or top
tray that are dominated by the product component boiling point. For simplicity, the temperature
set-point to the controller was set as the average boiling temperature of the two main components
(or azeotropes) being separated in the given section (i.e., assuming 50-50 mole percent of the
components on the middle tray in the column section). Alternatively, they can be obtained by
steady state calculations to get a desired separation, or they may be optimized as a function of
time. Other controller parameters such as the controller gain and the localization of the tray
temperature measurement may also be optimized.

The dynamic models were implemented in the equation-oriented software package SpeedUp
(1993), with user-written code for the nonideal thermodynamic VLE relationships. Thermo-
dynamic data for the various example mixtures presented in this study are given in Table 7.19
(Appendix A7).

7.6

Simulation results

In addition to some dynamic column composition profiles, we present the steady state values and
composition profiles which would be achieved in the closed multivessel columns under consid-
eration if we were to let the batch time approach infinity. Of course, in practice we want the
batch time to be as short as possible to minimize energy consumption and processing time and we
would terminate the batch operation when the product specifications are met or the improvement
in purity is small. Nevertheless, the steady state values are interesting since they give the achiev-
able separation for a given case and have a simple connection to the VLE diagram of the mixture
given by the distillation lines.

178

Chapter 7. Closed Multivessel Batch Distillation of Ternary Azeotropic Mixtures

The column specifications and initial conditions for the simulations of the cyclic and middle
vessel columns are given in Table 7.1.

Table 7.1: Columns data and initial conditions

Cyclic two-vessel batch distillation column

Total no. of trays

(excl. reboiler)

Total initial charge

5:385

kmol

Initial condenser holdup

0:035

kmol

Initial reboiler holdup

5:300

kmol

Tray holdups (constant)

1=300

kmol

Vapor flow (constant)

kmol/h

Closed middle vessel batch distillation column

Total no. of trays

(excl. reboiler)

No. of trays per section

Total initial charge

5:385

kmol

Initial condenser holdup

0:035

kmol

Initial middle vessel holdup

5:000

kmol

Initial reboiler holdup

0:250

kmol

Tray holdups (constant)

1=300

kmol

Vapor flow (constant)

kmol/h

In this study, the vapor flow (boilup) in the column sections is kept constant. The achievable
separation is not dependent on the boilup, but the operating time to reach steady state is. Gen-
erally, batch distillation columns are usually operated at maximum boilup (reboiler capacity) to
achieve maximum throughput (productivity) and minimize the batch time. The ratio of the vapor
flow relative to the total initial charge (

) is a measure on how many times the feed charge

is reboiled (this study: about once per hour) and is the same in both the cyclic column and the
middle vessel column as given in Table 7.1. All simulations are at atmospheric pressure. Ini-
tial column compositions are equal to that of the feed mixture in all the simulations presented in
this paper. Mostly equimolar feeds are studied and none of the mixture components are present
in small proportions in the initial feed (relatively to the total column holdup). The influence of
column holdup is not significant in this study. The liquid holdup in the column sections is less
than 1 % of the total initial charge (e.g., about 2 % of the total charge is distributed on the trays
in the middle vessel column). The temperature measurements for the indirect level controllers
are on tray 8 in each column section (trays numbered from the top and down) in both the cyclic
conventional column and the closed middle vessel column shown in Figures 7.1 and 7.2.

The desired product quality is at least 95 mole % in all the vessels. However, the number of
column trays used in the simulations are set larger than the minimum number of equilibrium trays
required in order to achieve the product specification. This is done to avoid the situation where
the number of equilibrium trays in the columns limit the achievable separation (according to the
given product specification) since we did not want to study the effect of limiting number of trays

7.6 Simulation results

179

in this paper. The recovery of component

in an arbitrary vessel is given by the equation

o;i

100%

(7.7)

For the simulation results presented in the Figures given next, the steady state column composition
profiles are given by solid lines with open circles for each equilibrium tray. The temperature set-
points are in general set to the mean value of pair of the desired products boiling points and are
visualized by the corresponding liquid isotherms as dotted lines.

7.6.1

Results on ternary zeotropic mixture (Cell I)

In this section we consider separation of the ternary zeotropic mixture of methanol, ethanol and
1-propanol which is given as an example mixture of elementary cell I (Serafimov’s class 0.0-1) in
Figure 7.3. The initial feed composition and controller parameters for both the cyclic conventional
(two-vessel) and closed middle vessel (three-vessel) column are given in Table 7.2 where the
parameter number indicates the Cycles 1 and 2, or, column section 1 and 2, respectively. The
temperature set-points are the mean values of pair of the pure component boiling points.

Table 7.2: Initial feed composition and controller parameters for methanol, ethanol and 1-
propanol

[1=3;

1=3;

1=3]

c;1

0:36

kmol/h

sp;1

71:5

c;2

0:26

kmol/h

sp;2

88:1

For the middle vessel column, the equimolar feed was mainly charged to the middle vessel ini-
tially as given in Table 7.1. The dynamic development of the composition profile along with the
resulting steady state composition profile are given in Figure 7.6. The steady state values of the
vessel holdups and compositions and the product recovery are given in Table 7.3. For ternary
zeotropic mixtures the whole composition space is encapsulated by the material balance triangle
and is a feasible product region. The composition in the middle vessel was slowly changed by
the gradual depletion of methanol and 1-propanol. High purity of methanol in the condenser and
1-propanol in the reboiler were achieved rapidly. The steady state values of the purities and vessel
holdups are independent of the initial feed charge distribution. However, if the initial feed was
mainly charged to the reboiler instead of to the middle vessel, for example, the dynamic devel-
opment of the column composition profile would differ. This has an impact on the time to reach
steady state in the column (i.e., approach to equilibrium) which is not an issue we address in this
study. The effect of initial feed charge distribution on the time consumption of closed middle
vessel batch distillation of ternary ideal mixtures is evaluated in Chapter 6 where the operation
is terminated when a given product specification is reached in all three vessels (not until steady
state as in this study).

180

Chapter 7. Closed Multivessel Batch Distillation of Ternary Azeotropic Mixtures

Methanol

64.5 oC

71.5 oC

Tsp1

88.1 oC

Tsp2

97.8 oC

1-Propanol

Ethanol

78.4 oC

Figure 7.6: Middle vessel column composition profiles for an equimolar feed of methanol, ethanol
and 1-propanol: Instantaneous profiles at 6 minutes and 2 hours (dash-dotted lines) and steady
state profile (solid line with open circles along the edges of the composition triangle).

Table 7.3: Steady state data for closed middle vessel batch distillation of methanol, ethanol and
1-propanol illustrated in Figure 7.6.

Condenser

Middle

Reboiler

Holdup [kmol]

1.757

1.765

1.763

ethanol

0.987

0.021

0.000

thanol

0.013

0.975

0.001

opanol

0.000

0.004

0.999

Recovery %

96.6

95.9

98.1

For the cyclic column, the equimolar feed was mainly charged to the reboiler initially as given in
Table 7.1. The resulting steady state composition profiles for each cycle are given in Figure 7.7.
The corresponding steady state values of the vessel holdups and compositions and the product
recovery are given in Table 7.4. After the first cycle, the condenser product was removed by
emptying the condenser. Some residue of the first product component (methanol) still remains
in the condenser and in the column section and this will contaminate the products of the second
cycle. Therefore, an off-cut fraction was removed between the two cycles. This was done by
closed operation with indirect level control where an off-cut fraction equal to the total column
holdup in the condenser was accumulated in the condenser using the same controller parameters
as for the second cycle. There are several (other) ways to remove the column holdup off-cut
between the cycles. In practice, one would probably “blow-off” an off-cut fraction (i.e., open
operation with no reflux) at the same time as the condenser vessel product of the first cycle is

182

Chapter 7. Closed Multivessel Batch Distillation of Ternary Azeotropic Mixtures

The feed of the second cycle (

in Figure 7.7b) is the composition in the reboiler residue. In

addition, the liquid holdup in the column and condenser vessel should be included but the contri-
bution to the overall feed composition is small due to the small column holdup in this study.

Table 7.4: Steady state data for cyclic batch distillation of methanol, ethanol and 1-propanol
illustrated in Figure 7.7.

Cycle 1

Cycle 2

Condenser

Off-Cut

Condenser

Reboiler

Holdup [kmol]

1.773

0.050

1.753

1.763

ethanol

0.987

0.501

0.014

0.000

thanol

0.013

0.498

0.982

0.001

opanol

0.000

0.001

0.004

0.999

Recovery %

97.3

95.9

98.1

Both column alternatives show about the same separation performance for this specific simula-
tion example. The cyclic column gives a slightly higher recovery of the first condenser product
(methanol) and a higher purity but the same recovery of the second condenser product (ethanol)
as compared to the middle vessel column. That is, a high purity saddle product (ethanol) is easier
to obtain in the cyclic column. Furthermore, the removal of an off-cut fraction between the cycles
(containing both methanol and ethanol) does not result in less recovery of the products.

In conclusion, for zeotropic mixtures separation, there seems to be not additional benefits of the
closed middle vessel column (for the given number of equilibrium stages) compared to the cyclic
operation of the regular batch column in terms of feasibility and separation performance. By
“additional” we here mean in addition to the energy or time savings and ease of operation which
were discussed in Chapter 6.

Our analysis and simulation results confirm that the assumption of constant relative volatility
used in the study by Wittgens and Skogestad (1997) for the mixture methanol, ethanol and 1-
propanol is acceptable and that the simple temperature set-points selection described in Section
7.5 (average boiling temperature of pair of the components) is satisfactory.

7.6 Simulation results

183

7.6.2

Results on ternary azeotropic mixture (Cell II)

In this section we consider the ternary azeotropic mixture of acetone, methanol and water which
is given as an example mixture of elementary cell II (Serafimov’s class 1.0-1a) in Figure 7.4. For
mixtures of this class the composition space (or cell) is divided into two feed regions with different
sets of possible steady state vessel products. This is illustrated by a straight line going from
the azeotrope point to the zeotropic component vertex (water) in Figure 7.8 (that is, the secant
between the opposite directed nodes in Cell II). Four different cases of initial feed composition
and controller parameters for both the cyclic and the middle vessel column simulations are given
in Table 7.5 where the parameter number indicates Cycle 1 and 2, or, section 1 and 2, respectively.

Table 7.5: Initial feed composition and controller parameters for acetone, methanol and water

Case 1:

[1=3;

1=3;

1=3]

c;1

0:56

kmol/h

sp;1

60:0

c;2

0:14

kmol/h

sp;2

82:3

Case 1’:

[0:6;

0:1;

0:3]

c;1

0:56

kmol/h

sp;1

60:0

c;2

0:14

kmol/h

sp;2

82:3

Case 2:

[0:6;

0:1;

0:3]

c;1

0:56

kmol/h

sp;1

56:0

c;2

0:11

kmol/h

sp;2

78:2

Case 3:

[1=3;

1=3;

1=3]

c;1

0:56

kmol/h

sp;1

56:0

c;2

0:11

kmol/h

sp;2

78:2

Case 4:

[0:6;

0:1;

0:3]

c;1

0:56

kmol/h

sp;1

55:8

c;2

0:11

kmol/h

sp;2

75:0

The steady state composition profiles along the middle vessel column are given for an initial feed
in both feed regions in Figure 7.8 (Cases 1 and 2). The corresponding steady state values of the
vessel holdups and compositions and the product recovery are given in Tables 7.8 and 7.7.

Table 7.6: Steady state data for closed middle vessel batch distillation of acetone, methanol and
water (Case 1) illustrated in Figure 7.8a.

Condenser

Middle

Reboiler

Holdup [kmol]

2.448

1.072

1.765

Acetone

0.727

0.001

0.000

ethanol

0.273

0.998

0.000

ater

0.000

0.001

1.000

Recovery %

59.6

98.3

7.6 Simulation results

185

Table 7.7: Steady state data for closed middle vessel batch distillation of acetone, methanol and
water (Case 2) illustrated in Figure 7.8b.

Condenser

Middle

Reboiler

Holdup [kmol]

3.295

0.415

1.575

Acetone

0.840

0.961

0.000

ethanol

0.160

0.017

0.000

ater

0.000

0.022

1.000

Recovery %

22.2

87.7

If the feed is shifted from the lower feed region (Case 1) to the upper feed region without changing
the temperature set-points (Case 1’) there exists no steady state as illustrated in Figure 7.9. The
indirect level controllers attempt to achieve products outside the material balance triangle of this
feed region and the middle vessel runs empty. By simply changing the temperature set-points to

Water

Methanol

Acetone

56.4 oC

55.5 oC

64.5 oC

100.0 oC

60.0 oC

Tsp1

Tsp2

82.3 oC

Figure 7.9: Infeasible temperature-set points for the middle vessel column with feed in the upper
feed region (unsteady state composition profile at 1 hour and 50 minutes). The middle vessel runs
empty (Case 1’).

the mean boiling-point values of pair of the feasible products for this feed region (i.e., azeotrope,
acetone and water for the upper feed region, Case 2), we achieve the steady state composition
profile given in Figure 7.8b. Note that these temperature set-points (Case 2) are feasible for feeds
in the lower feed region since the temperature specifications are feasible for this feed region as
well (Case 3). The resulting steady state composition profile is given in Figure 7.10. In this
manner, robust values of the temperature set-points may be chosen by inspection of the isotherms
of the VLE diagram (isotherm map) if one expect variations in the feed composition of a mixture
of this type (cell II). If the feed composition lies directly on the secant that divides the two feed
regions, the middle vessel column performs a “binary” separation between the azeotrope and the

186

Chapter 7. Closed Multivessel Batch Distillation of Ternary Azeotropic Mixtures

56.4 oC

Acetone

55.5 oC

56.0 oC

Tsp1

64.5 oC

Methanol

Water

100.0 oC

78.2 oC

Tsp2

Figure 7.10: Middle vessel column steady state composition profile for an equimolar feed of
acetone, methanol and water (Case 3).

water and we achieve zero holdup in the middle vessel. Each feed region of the composition
triangle may be interpreted as a local Cell I, one with C-shaped distillation lines and the other
with S-shaped distillation lines for this particular mixture.

Table 7.8: Steady state data for closed middle vessel batch distillation of acetone, methanol and
water (Case 3) illustrated in Figure 7.10.

Condenser

Middle

Reboiler

Holdup [kmol]

2.285

1.234

1.766

Acetone

0.764

0.020

0.000

ethanol

0.236

0.979

0.000

ater

0.000

0.001

1.000

Recovery %

67.3

98.4

For feeds located in the lower feed region, the temperature set-points in Case 3 result in a better
separation performance (higher recovery but less purity of methanol) as compared to the temper-
ature set-points in Case 1 because the distribution of the given number of equilibrium stages for
each separation in the column sections is more favorable (split Az-methanol and split methanol-
water). For feed located in the upper feed region, the separation is improved as compared to Case
2 by changing to the temperature set-points given in Case 4 for the same reasons. The steady state
composition profile for Case 4 is given in Figure 7.11 and the corresponding steady state values
of the vessel holdups and compositions and the product recovery are given in Table 7.9. Even
this small change in the temperature set-points results in higher recoveries of both acetone and

7.6 Simulation results

187

100.0 oC

Water

64.5 oC

Methanol

55.5 oC

56.4 oC

75.0 oC

Acetone

Tsp2

Tsp1

55.8 oC

Figure 7.11: Middle vessel column steady state composition profile for an equimolar feed of
acetone, methanol and water (Case 4).

water (as compared to Case 2). This demonstrates the importance of the proper choice of temper-
ature set-points for the indirect level control strategy to obtain a specified separation (purity and
recovery). The sensitivity is less if the number of equilibrium trays is significantly larger than
the minimum number of trays required for the separation. For nonideal and azeotropic mixtures
of Cell I and II (S-shape) it is not recommended to set the temperature set-points as the average
boiling-point temperature of pair of the products for the given feed region.

Table 7.9: Steady state data for closed middle vessel batch distillation of acetone, methanol and
water (Case 4) illustrated in Figure 7.11.

Condenser

Middle

Reboiler

Holdup [kmol]

2.788

0.930

1.567

Acetone

0.819

0.949

0.000

ethanol

0.181

0.031

0.000

ater

0.000

0.020

1.000

Recovery %

27.3

97.0

The steady state composition profiles for each cycle in the cyclic column for Case 1, 2, 3 and 4
in Table 7.5 are given in Figure 7.12, 7.13, 7.14 and 7.15. The corresponding steady state values
for the vessel holdups and compositions and the product recovery are given in Tables 7.10, 7.11,
7.12 and 7.13. For feed in the lower feed region, the temperature set-points in Case 3 are better
than the temperature set-points in Case 1 for the cyclic batch distillation process (higher recovery
of methanol). For feed in the upper feed region, the temperature set-points in Case 4 is better than
the temperature set-points in Case 2 (significant higher recovery of acetone). This is due to the

190

Chapter 7. Closed Multivessel Batch Distillation of Ternary Azeotropic Mixtures

Table 7.10: Steady state data for cyclic batch distillation of acetone, methanol and water (Case
1) illustrated in Figure 7.12.

Cycle 1

Cycle 2

Condenser

Off-Cut

Condenser

Reboiler

Holdup [kmol]

2.448

0.050

1.075

1.765

Acetone

0.727

0.306

0.001

0.000

ethanol

0.273

0.694

0.997

0.000

ater

0.000

0.001

1.000

Recovery %

59.7

98.3

Table 7.11: Steady state data for cyclic batch distillation of acetone, methanol and water (Case
2) illustrated in Figure 7.13.

Cycle 1

Cycle 2

Condenser

Off-Cut

Condenser

Reboiler

Holdup [kmol]

2.754

0.050

0.967

1.568

Acetone

0.832

0.897

0.907

0.000

ethanol

0.167

0.093

0.077

0.000

ater

0.001

0.010

0.016

1.000

Recovery %

27.1

97.1

Table 7.12: Steady state data for cyclic batch distillation of acetone, methanol and water (Case
3) illustrated in Figure 7.14.

Cycle 1

Cycle 2

Condenser

Off-Cut

Condenser

Reboiler

Holdup [kmol]

2.309

0.050

1.213

1.766

Acetone

0.764

0.495

0.006

0.000

ethanol

0.236

0.505

0.993

0.000

ater

0.000

0.001

1.000

Recovery %

67.1

98.4

Table 7.13: Steady state data for cyclic batch distillation of acetone, methanol and water (Case
4) illustrated in Figure 7.15.

Cycle 1

Cycle 2

Condenser

Off-Cut

Condenser

Reboiler

Holdup [kmol]

1.825

0.050

1.904

1.560

Acetone

0.814

0.869

0.886

0.000

ethanol

0.186

0.124

0.101

0.000

ater

0.000

0.007

0.014

1.000

Recovery %

52.2

96.6

194

Chapter 7. Closed Multivessel Batch Distillation of Ternary Azeotropic Mixtures

7.6.3

Results on ternary azeotropic mixture with one distillation boundary

In this section we consider the ternary azeotropic mixture of acetone, chloroform and benzene
which is given as an example mixture where a distillation line boundary splits the VLE diagram
into two elementary cell I’s (Serafimov’s class 1.0-2) shown in Figure 7.5. This example is inter-
esting because it clearly demonstrates an important difference between the closed middle vessel
column separation and the cyclic column separation in terms of separation feasibility (feasible
vessel products). Three different cases of initial feed composition and controller parameters for
both the cyclic column and the middle vessel column were considered and the results are given in
Table 7.14. The parameter number indicates Cycle 1 and 2, or, column section 1 and 2, respec-
tively.

Table 7.14: Initial feed composition and controller parameters for acetone, chloroform and ben-
zene

Case 1:

[1=3;

1=3;

1=3]

c;1

0:63

kmol/h

sp;1

60:4

c;2

0:32

kmol/h

sp;2

72:3

Case 2:

[0:15;

0:05;

0:80]

c;1

0:63

kmol/h

sp;1

60:4

c;2

0:32

kmol/h

sp;2

72:3

The steady state composition profile in the middle vessel column for Case 1 and 2 is given in Fig-
ure 7.16 and 7.17. The corresponding steady state values for the vessel holdups and compositions
and the product recovery are given in Table 7.15 and 7.16.

In the closed middle vessel column the three vessel products are bounded within one distillation
line region (cell) and it is not possible to cross a distillation line boundary. The distillation behav-
ior of this class of mixture within one cell of the composition triangle is similar to the behavior of
other mixtures of elementary cell I, as for example the mixture of methanol, ethanol, 1-propanol
considered in the previous section. The azeotrope forms a saddle of the cell I and acts like a
“pseudo-component”. The purity of the vessel products for the same given temperature set-points
is independent of the initial feed composition within one distillation line region (cell). This con-
firms the flexibility of the indirect level control strategy based on temperature feedback control as
stated by Skogestad et al. (1997).

196

Chapter 7. Closed Multivessel Batch Distillation of Ternary Azeotropic Mixtures

Table 7.15: Steady state data for closed middle vessel batch distillation of acetone, chloroform
and benzene (Case 1) illustrated in Figure 7.16.

Condenser

Middle

Reboiler

Holdup [kmol]

1.670

2.990

1.625

Acetone

0.999

0.362

0.000

hlor

0.0005

0.581

0.018

enz

ene

0.0005

0.057

0.982

Recovery %

92.9

88.9

Table 7.16: Steady state data for closed middle vessel batch distillation of acetone, chloroform
and benzene (Case 2) illustrated in Figure 7.17.

Condenser

Middle

Reboiler

Holdup [kmol]

0.665

0.279

4.341

Acetone

0.999

0.363

0.000

hlor

0.0005

0.581

0.018

enz

ene

0.0005

0.056

0.982

Recovery %

82.2

99.0

7.6 Simulation results

197

The steady state composition profile for each cycle in the cyclic column for Case 1 and 2 are
given in Figure 7.18 and 7.19. The corresponding steady state values for the vessel holdups and
compositions and the product recovery are given in Table 7.17 and 7.18. In both cases, the first
condenser product (acetone) is discharged after the first cycle and an off-cut fraction equal to the
column holdup is taken out between the cycles.

80.1 oC

Benzene

Acetone

56.4 oC

64.4 oC

Chloroform

61.3 oC

Tsp1

60.4 oC

(a)

Cycle 1

80.1 oC

Benzene

Acetone

56.4 oC

64.4 oC

Chloroform

61.3 oC

72.3 oC

Tsp2

(b)

Cycle 2

Figure 7.18: Cyclic operation steady state composition profiles for an equimolar initial feed of
acetone, chloroform and benzene (Case 1:

[1=3;

1=3;

1=3]

[0:425;

0:420;

0:155]

198

Chapter 7. Closed Multivessel Batch Distillation of Ternary Azeotropic Mixtures

Table 7.17: Steady state data for cyclic batch distillation of acetone, chloroform and benzene
(Case 1) illustrated in Figure 7.18.

Cycle 1

Cycle 2

Condenser

Off-Cut

Condenser

Reboiler

Holdup [kmol]

1.102

0.050

2.514

1.672

Acetone

0.997

0.734

0.261

0.000

hlor

0.000

0.235

0.690

0.018

enz

ene

0.002

0.032

0.048

0.982

Recovery %

61.0

91.5

For the benzene-rich initial feed (Case 2) given in Figure 7.19, the first cycle bottoms product,
which becomes the (reboiler) feed of the second cycle, lies on the binary edge of the composi-
tion triangle between chloroform and benzene. Thus, the second cycle of the cyclic conventional
batch distillation process achieves chloroform as condenser product even if the initial feed to the
process lies in the other distillation line region (cell). Note however that only small amounts of
this second condenser product are recovered since the column feed of the second cycle mainly
contains benzene. This is of theoretical interest only and included to illustrate the principal dif-
ference between the cyclic conventional batch distillation process and the closed middle vessel
distillation in terms of feasible vessel products.

Comparing the separation in the closed middle vessel column with the separation in the cyclic
column, we see that for Case 1 the middle vessel column has better performance since it achieves a
higher acetone recovery even though less benzene recovery. Both columns achieves the azeotrope
(saddle) as the intermediate product fraction. For Case 2, however, the separation in the cyclic
column is much better since it achieves a higher acetone recovery, the same benzene recovery
and, in addition, some chloroform is recovered (no azeotrope fraction).

Table 7.18: Steady state data for cyclic batch distillation of acetone, chloroform and benzene
(Case 2) illustrated in Figure 7.19.

Cycle 1

Cycle 2

Condenser

Off-Cut

Condenser

Reboiler

Holdup [kmol]

0.768

0.050

0.170

4.349

Acetone

0.992

0.589

0.089

0.000

hlor

0.000

0.340

0.884

0.019

enz

ene

0.007

0.071

0.026

0.981

Recovery %

93.9

55.8

99.0

In the cyclic column only pair of the vessel products are bounded within one distillation line
region and the column may operate in two different distillation line regions (cells) dependent
on the column feed composition for each cycle. Due to the curved distillation line boundary
of this particular mixture, the cyclic operation is able to cross the distillation line boundary and

7.6 Simulation results

199

80.1 oC

Benzene

Acetone

56.4 oC

64.4 oC

Chloroform

61.3 oC

60.4 oC

Tsp1

(a)

Cycle 1

80.1 oC

Benzene

Acetone

56.4 oC

64.4 oC

Chloroform

61.3 oC

Tsp2

72.3 oC

(b)

Cycle 2

Figure 7.19:

Cyclic operation steady state composition profiles for a benzene-rich ini-

tial feed of acetone, chloroform and benzene (Case 2:

[0:15;

0:05;

0:80]

[0:946;

0:052;

0:002]

)).

separate the mixture into its pure components. When the feed lies within a certain region of the
composition space (i.e., in the area between the convex side of the distillation line boundary and
the straight line going from the benzene node to the azeotrope saddle), the material balance line
may cross the curved distillation line boundary as shown in Figure 7.18 (b). This is known from

7.8 Conclusion

203

Appendix 7A

The dynamic model of the closed multivessel batch distillation column (with two and three ves-
sels) used in the simulations presented in this paper is based on the following assumptions:

Staged distillation column sections (trays numbered from the top to the bottom)

Perfect mixing and VLE on all trays and in all vessels

Negligible vapor holdup

Constant molar liquid holdups on all trays (i.e., neglecting liquid dynamics)

Constant molar vapor flows (i.e., simplified energy balance)

Constant pressure

Total condenser

Ideal behavior in the vapor phase

Note that the vapor flow does not pass through the intermediate vessels in our model of the
multivessel column. Thus, the total number of equilibrium trays the closed middle vessel column
shown in Figure 7.2 is

(inclusive the reboiler).

Material balances over

total condenser:

;

(7.8)

d(H

)

(7.9)

tray

::;

in column section 1:

;

(7.10)

(7.11)

middle vessel:

;

(7.12)

d(H

)

(7.13)

tray

::;

in column section 2:

;

(7.14)

(7.15)

204

Chapter 7. Closed Multivessel Batch Distillation of Ternary Azeotropic Mixtures

total reboiler:

;

(7.16)

d(H

)

(7.17)

The cyclic conventional (two-vessel) column shown in Figure 7.1 has only one column section
and is described by Equations (7.8), (7.9), (7.10), (7.11), (7.16) and (7.17).

Indirect level control

The liquid reflux out of the condenser and middle vessel (

and

, respectively) are given

by the proportional controller in Eq. 7.5. In addition, we do not allow a vessel to be completely
empty. If a vessel holdup decreases towards zero the liquid reflux out of that particular vessel
is decreased, or, in the limit, turned off to accumulate material in the vessel. To achieve a soft
“turn off” the liquid reflux determined by the controller is modified by an exponential function
(Furlonge et al., 1999):

exp

min

(7.18)

where

is the manipulated liquid reflux (i.e.,

and

in Equations (7.8)-(7.9) and Equations

(7.12)-(7.13),

is the controller-determined liquid reflux given by Equation (7.5),

is the liquid

holdup in the current vessel and

min

is a selected minimum value of the liquid holdup.

Vapor-liquid equilibrium for component

::;

;

)

sat

)

(7.19)

where

1 atm in all the simulations presented in this article and

i=1

(7.20)

The vapor pressures were calculated by the Antoine equation:

sat

(7.21)

where

sat

is in

mmH

and

. The liquid activity coefficients were calculated by the

Wilson equation:

k =1

k i

k j

;

(7.22)

exp

(7.23)

7.8 Conclusion

205

where

is in

=mol

cal =mol

and

. We use parameters for the Antoine and

Wilson equations and liquid molar volumes reported in Gmehling et al. (1977-1990), as listed in
Table 7.19.

Table 7.19: Thermodynamic data for the example mixtures

Class 0.0-1

Parameters for the Antoine Equation

[

mol

]

(1) Methanol

8.08097

1582.271

239.726

40.73

(2) Ethanol

8.11220

1592.864

226.184

58.68

(3) 1-Propanol

8.37895

1788.020

227.438

75.14

Binary Interaction Parameters for the Wilson Equation

135:8113

397:4862

132:0576

102:6706

265:4397

40:8745

Class 1.0-1a

Parameters for the Antoine Equation

[

mol

]

(1) Acetone

7.63132

1566.69

273.419

74.05

(2) Methanol

8.08097

1582.271

239.726

40.73

(3) Water

8.01767

1715.70

234.268

18.01

Binary Interaction Parameters for the Wilson Equation

161:8813

489:3727

583:1054

19:2547

1422:849

554:0494

Class 1.0-2

Parameters for the Antoine Equation

[

mol

]

(1) Benzene

6.87987

1196.760

219.161

89.41

(2) Chloroform

6.95465

1170.966

226.232

80.67

(3) Acetone

7.11714

1210.595

229.664

74.05

Binary Interaction Parameters for the Wilson Equation

49:6010

34:7648

161:8065

485:5470

240:8594

28:4509

From Gmehling et al. (1977-1990), VLE Data Collection, DECHEMA series, at atmospheric pressure.

208

Chapter 8. Extractive Batch Distillation: Column Schemes and Operation

Section

Extractive

Section

Rectifying

Reboiler

Condenser

Condenser Drum

Accumulators

Initial charge of

azeotropic mixture

Entrainer Feed

Figure 8.1: Conventional extractive batch distillation column.

pers, Yatim and Lang and coworkers (1993; 1994; 1995) gave experimental as well as dynamic
simulation results on the operation of extractive distillation in the conventional batch column.

The column scheme in Figure 8.1 has two main disadvantages caused by accumulation of the
entrainer in the reboiler: (i) operational problems due to filling, and (ii) decreased separation
efficiency during operation. One potential way to overcome these problems is to use the middle
vessel column shown in Figure 8.2, which closely resembles continuous extractive distillation
except that the operation is batchwise (dynamic) and that the separation may be performed in one
column rather than in a column sequence. The middle vessel implementation is potentially more
energy efficient than the conventional scheme due to the combination of a rectifying-extraction
section and a stripping section in one column.

Safrit et al. (1995; 1997b) studied extractive batch distillation in the middle vessel column for the
separation of acetone and methanol using water as entrainer. Water was continuously removed as
a bottom product and recycled to the process. They compared the separation performance of this
process with the conventional extractive batch distillation column. The extractive middle vessel
distillation column showed better performance with regard to energy consumption. The study
also showed that the performance of the separation is sensitive to the entrainer feed flowrate,
as known from continuous operation, and to the switching times between fractions in the batch
operation. The constant product composition policy were considered, achieving a desired purity
in the distillate and bottom product. They proposed an algorithm to “steer” the middle vessel

210

Chapter 8. Extractive Batch Distillation: Column Schemes and Operation

the reflux ratio in order to achieve the same separation results. K¨ohler et al. (1995) provided an
interesting industrial example of combined heteroazeotropic and extractive distillation performed
in a conventional batch column and decanter.

Warter and Stichlmair (1999) studied extractive batch distillation in both the conventional column
and the middle vessel column. The mixture system of ethanol and water using ethylene glycol as
entrainer was considered. The middle vessel column was found to be superior to the conventional
batch extractive distillation column in terms of energy demand. Furthermore, they suggested a
modification of the middle vessel column arrangement where the liquid stream from the extractive
section bypasses the middle vessel and enters directly the stripping section (see the liquid bypass
configuration in Figure 8.17). This configuration had better performance in terms of separation
performance and energy demand. They were able to achieve simultaneous ethanol and ethylene
glycol production and pure water in the middle vessel at the end of operation. This is explained
by the fact that ethylene glycol has very low volatility and thus remains almost completely in the
liquid phase. Thus, negligible amounts of ethylene glycol enters the middle vessel in the vapor
stream from the stripping section in this modification of the column.

Dynamic optimization of the complete extractive batch distillation process was reported by Mu-
jtaba (1999). For the mixture acetone-methanol-water the best performance was found when the
entrainer was introduced as a semicontinuous feed to the column section and not as a batch charge
to the reboiler. They optimized piecewise constant reflux ratio and entrainer feed flowrate for each
operating step (minimum time and maximum profit objective functions), but did not compare the
results with previous work. Milani (1999) considered batch extractive distillation of acetone and
methanol where the water entrainer was charged directly to the reboiler. Large quantities of the
entrainer was needed in order to achieve high purity acetone in the distillate.

Previous studies on continuous extractive distillation have shown that there is a maximum as well
as a minimum reflux ratio (L

/D) to make the separation feasible (Andersen et al., 1995; Laroche

et al., 1993). High reflux (high internal flows) may be harmful in extractive distillation because it
decreases the entrainer concentration in the extractive column section. If the desired purity of the
distillate product requires high reflux one may have to purify the distillate fraction in a subsequent
operation. Similarly, there is a maximum as well as a minimum entrainer feed ratio to make the
separation feasible. Thus, there is a trade-off between the entrainer feed flowrate and the reflux
and reboil ratios.

The entrainer load to the column is an important variable in effective separation. A very rough
rule of thumb for continuous extractive distillation is to use 1 to 4 moles of entrainer per mole
of feed (Schweitzer, 1997). However, the actual value must be determined by experiment or by
computer simulation (and possibly optimization).

In this paper two main configurations are considered: (1) conventional extractive batch distillation
column (also called an extractive batch rectifier); and (2) middle vessel extractive batch distilla-
tion column (an extractive batch rectifier and a batch stripper combined with a middle vessel).
The columns are operated “open-loop”, that is, with no feedback control of compositions (or tem-
peratures). These results are then rather sensitive to the operating parameters such as entrainer
flowrate and reflux. The values were found by trial-and-error and are by no means optimal.

8.2 Mixtures studied

211

8.2

Mixtures studied

The following two mixture systems are studied in this paper:

System 1: ethanol (1) and water (2) + ethylene glycol entrainer (E); (easy separation)

System 2: acetone (1) and methanol (2) + water entrainer (E); (more difficult separation)

System 1 (ethanol - water - ethylene glycol)

Extractive distillation of ethanol and water using ethylene glycol as an entrainer (System 1) is a
classical example in the distillation literature. The mixture was chosen mainly because it is easy
to find comparable examples in the distillation literature. This separation also represents a typical
separation task in the fine- and specialty chemical industries. The residue curve map is shown
in Figure 8.3. Ethanol and water form a minimum-boiling azeotrope rich in ethanol. The binary

Residue Curves Map

WATER

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

ETHANOL

WATER

ETG

100.0 C

197.4

78.1

78.3

Ethanol (1)

Ethylene glycol (E)

Water (2)

Figure 8.3: Residue curve map for the ternary azeotropic mixture ethanol, water and ethylene
glycol (System 1) at 1 atm (from

ASPEN PLUS

azeotropic composition is 89.3 mole % ethanol at 1 atm (Horsley, 1973). By adding ethylene
glycol, the relative volatility between ethanol and water becomes larger than one and separa-
tion into pure components by distillation is possible. Note the large boiling-point differences
between the ethylene glycol entrainer (E) and the azeotrope constituents ethanol (1) and water
(2):

119:1

and

97:4

. Hence, the separation of ethanol and water from

ethylene glycol by distillation is relatively easy and requires only a few number of equilibrium
trays and low reflux. Due to the very large relative volatility between ethanol and ethylene glycol

212

Chapter 8. Extractive Batch Distillation: Column Schemes and Operation

it is sufficient with only one or a few equilibrium trays in the rectifying section of the extractive
distillation column.

System 2 (acetone - methanol - water)

The importance and effect of the rectifying section and the reflux ratio are better illustrated by
the acetone and methanol separation using water as entrainer (System 2). This is the standard
industrial method for separating acetone from methanol using water as entrainer. Water has a
greater affinity to methanol, the more polar of the pair, so the relative volatility of methanol is
decreased. This allows the acetone to be distilled overhead. The residue curve map is shown in
Figure 8.4. Acetone and methanol form a binary minimum-boiling azeotrope of about 78.5 mole

Residue Curves Map

METHANOL

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

ACETONE

METHANOL

WATER

64.5 C

56.4

55.5 C

100.0

Acetone (1)

Methanol (2)

Water (E)

Figure 8.4: Residue curve map for the ternary azeotropic mixture acetone, methanol and water
(System 2) at 1 atm (from

ASPEN Plus

% acetone. Water is miscible and zeotropic in the whole composition space. This system also has
relatively large boiling-point differences between the water entrainer (E) and the azeotrope con-
stituents acetone (1) and methanol (2):

43:9

and

35:5

. However, the

binary system of acetone and water forms a tangentially zeotropic mixture (tangent pinch). This
is shown in Figure 8.5 where we see that at high concentrations of acetone, the relative volatility
between acetone and water is close to unity. Thus, nearly all the equilibrium stages required to
separate acetone from water are due to the last purification part of high purity acetone. Even with
a high reflux ratio and a large number of stages the distillate acetone will be contaminated with a
small amount of water. Requiring acetone of high purity may, as discussed earlier, cause opera-
tional problems since high reflux may be harmful in extractive distillation. This problem can be
circumvented by relaxing the water impurity demand in the acetone distillate product and instead
purify the distillate in a separate step.

8.4 Extractive batch distillation in the conventional column

215

8.4.2

Simulation results for System 1 (ethanol - water - ethylene glycol)

The original mixture feed composition is 65 mol% ethanol (1) and 35 mol% water (2). Ethylene
glycol (E) is used as entrainer. The column specifications and initial conditions for the simulations
are given in Table 8.1. The total liquid holdup in the column is less than 2 % of the total feed
charge. The first distillate product (ethanol) is collected in the accumulator with an overall purity
of 99.5 mole % ethanol and 0.5 mole % water. Ethylene glycol of 99.99 mole % purity is obtained
in the reboiler at the end of operation.

Table 8.1: Column data and initial conditions for extractive batch distillation of ethanol and
water using ethylene glycol as entrainer.

Total no. of trays

(excl. reboiler)

No. of trays per section

Total initial charge

5:135

kmol

Condenser holdup (constant)

0:035

kmol

Initial reboiler holdup

5:000

kmol

Tray holdup (constant)

0:005

kmol

Boilup, vapor flow (constant)

5:0

kmol/h

The results from five selected case studies are summarized in Table 8.2. The recovery of ethanol
in Step 2 (1st cut) is seen to vary between 83.1 % and 93.0 %. Simulation results for Case 1 are
shown in Figures 8.6 and 8.7.

Table 8.2: Operating conditions and simulation results for extractive batch distillation of ethanol
and water using ethylene glycol as entrainer.

Effect studied

Operating conditions

Recovery

Case 1:

Entrainer flowrate,

2:5

kmol/hr

ethanol

93:0

base case

Reflux ratio,

1:5

Entrainer-to-vapor ratio,

0:5

2:35

Case 2:

Entrainer flowrate,

2:5

kmol/hr

ethanol

84:3

decrease R

Reflux ratio,

1:0

Entrainer-to-vapor ratio,

0:5

2:40

Case 3:

Entrainer flowrate,

3:0

kmol/hr

ethanol

84:8

decrease R

Reflux ratio,

1:0

increase E

Entrainer-to-vapor ratio,

0:6

2:54

Case 4:

Entrainer flowrate,

2:0

kmol/hr

ethanol

83:1

decrease R

Reflux ratio,

1:0

decrease E

Entrainer-to-vapor ratio,

0:4

2:52

Case 5:

Entrainer flowrate,

3:0

kmol/hr

ethanol

92:8

increase E

Reflux ratio,

1:5

Entrainer-to-vapor ratio,

0:6

2:50

216

Chapter 8. Extractive Batch Distillation: Column Schemes and Operation

0.5

1.5

2.5

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Water

Ethylene glycol

2nd cut

Ethanol

Off-cut

Time [hr]

Liquid molefraction in distillate

1st cut

Entrainer feeding

Figure 8.6: Batch extractive distillation separating ethanol and water using ethylene glycol as
entrainer: Distillate product composition trajectory. Entrainer feeding is stopped at water break-
through (

1:56

hours). Ethanol recovery 93.0 % (Case 1).

Ethylene glycol

Water

78.3 oC

197.4 oC

100.0 oC

Ethanol

composition profile

Reboiler composition

78.1 oC

trajectory

Instantaneous column

(t=40 min)

(t=0)

(t=3 hr)

Figure 8.7: Liquid composition profile in the column at 40 minutes (1st cut) and reboiler compo-
sition change with time for the complete process corresponding to Figure 8.6 (Case 1).

Figure 8.6 shows the composition of the distillate as a function of time. The first distillate fraction
(1st cut) is almost pure ethanol. Once the ethanol is exhausted in the reboiler still, there is a
breakthrough of water in the distillate and entrainer feeding is stopped. The off-cut fraction is
a mixture of ethanol and water. The second distillate fraction (2nd cut) is almost pure water.
The column essentially separates a binary mixture of water and ethylene glycol during Step 4

8.4 Extractive batch distillation in the conventional column

217

(2nd cut). To obtain pure ethylene glycol in the reboiler at the end of operation (for reuse), we
removed a second off-cut fraction of water and ethylene glycol as distillate (Step 5). Further, pure
ethylene glycol could have been produced as a third distillate fraction (Step 6).

The size (amount of material) of the off-cuts depends on the sharpness of the separation between
the products, which is dependent on the column holdup, relative volatility, reflux and number of
theoretical trays.

Figure 8.7 shows the reboiler composition change during the operation. The initial change is
almost linear because the column parameters and operating conditions give a constant distillate
composition during the 1st cut (Step 2). Once the ethanol has been removed from the system the
composition trajectories move along the edge of the composition diagram that belong to these
components.

Figure 8.7 also shows a snapshot of the column profile at a time instant during the period with
entrainer feeding (Step 2: 1st cut). It is similar to the profile in the rectifying and extractive
sections of a continuous extractive distillation column (see Chapter 2). The distillate is nearly
pure ethanol. The rectifying section of the column essentially separates ethanol from ethylene
glycol during this step of the process. The profile lies on the binary edge between ethanol and
ethylene glycol in the composition diagram. The profile in the extractive section is the part that
goes inside the composition triangle in Figure 8.7. There are essentially three effects going on in
this section: (i) mixing of the entrainer and the original feed, (ii) countercurrent flow of streams
with different compositions, and, (iii) separation between ethanol and water (in the presence of
the ethylene glycol entrainer). If the entrainer load is too low, i.e. insufficient separation of the
azeotrope-forming components, the rectifying profile does not lie on the binary edge but inside
the composition triangle.

For the given column parameters and operating conditions (entrainer load, internal flows) the
separation between ethanol and water (in the presence of ethylene glycol) in the extractive section
is complete. Therefore the distillate trajectory reaches the unstable node (ethanol) and follows the
edges of the composition triangle during Step 2 (1st cut).

Effect of reflux and entrainer load

From Table 8.2 we see that the ethanol recovery depends mostly on the reflux ratio. It seems
like the reflux ratio mainly affects the amount of off-cut (Step 3). In our study, the reflux ratio
was kept constant during the whole operation although the separation (recovery) would probably
improve if the reflux ratio was increased after Step 2.

In Figure 8.8, we show graphically the recovery of ethanol as a function of the ratio between
entrainer and vapor flow (E/V). We see that increased entrainer load improves the ethanol recov-
ery. The improvement is larger at low reflux ratios, and there seem to be a maximum entrainer
feed flowrate E

, for which an increase in the entrainer load does not improve the separation

efficiency in the extractive section. (Obviously, there is also an E

min

to for which the separation

in the extractive section is insufficient to meet the product specifications).

8.4 Extractive batch distillation in the conventional column

219

8.4.3

Simulation results for System 2 (acetone - methanol - water)

The original feed composition is 50 mol% acetone and 50 mol% methanol. Pure water is used as
the entrainer. The column specifications and initial conditions are given in Table 8.3.

Table 8.3: Column data and initial conditions for the extractive batch distillation column sepa-
rating acetone and methanol using water as entrainer.

Total no. of trays

(excl. reboiler)

No. of trays per section

Total initial charge

5:385

kmol

Condenser holdup (constant)

0:035

kmol

Initial reboiler holdup

5:283

kmol

Tray holdup (constant)

1=300

kmol

Boilup, vapor flow (constant)

5:0

kmol/h

The existence of a tangent pinch for the binary pair acetone and water (see Figure 8.5) makes
it difficult to achieve water-free acetone (from simulations we learned that it was difficult to get
more than 98.5 mole % acetone), but otherwise the separation between acetone-water is easy.
Thus, we relax the water impurity specification of the acetone product, and use the following
specifications:

1st cut: 96.0 mole % acetone with maximum 0.5-1.0 mol% methanol impurity;

2nd cut: 99.0 mol% methanol with maximum 1.0 mol% acetone impurity;

Entrainer recovery: 99.9 mol% water with maximum 0.1 mol% methanol impurity.

Startup procedures

During startup (Step 1) the column was operated at total reflux for 15 minutes. This time period
was chosen somewhat arbitrary. Two alternative startup procedures were compared to see which
one that yields the best overall separation results:

Startup procedure 1: Total reflux operation without entrainer feeding

Startup procedure 2: Total reflux operation with entrainer feeding

Startup procedure 2 was found to be the best. Here we build up both the column composition
profile and the purity of distillate before any product removal. This avoids an off-spec distillate
fraction at the beginning of the 1st cut (Step 2).

220

Chapter 8. Extractive Batch Distillation: Column Schemes and Operation

There is also the third option of skipping the total reflux operation and starting directly with Step 2
(finite reflux with entrainer feeding), but then there will be a greater loss of ethanol and methanol
because of the need to remove an off-cut before the 1st cut.

In all simulations the column is operated with a constant entrainer flowrate until there is a methanol
breakthrough in the acetone distillate product (or more precisely; when there is 96.0 mole % ace-
tone in the accumulated distillate product). Then, ordinary distillation with a constant reflux ratio
(equal to the ratio used during the entrainer feeding period) is performed.

We report five selected case studies where we use values of E and R which meet the purity
specifications (trial-and-error approach). These values are by no means optimal for the process.
Safrit and Westerberg (1997b) formulated this into an optimization problem (variables E and
R and profit function), but there are no studies reported on how to implement such an optimal
operating policy.

The results from the dynamic simulations are summarized in Table 8.4 (Cases 1-5), and more
detailed results are shown in Figures 8.9-8.14 (Cases 2, 3 and 5).

Table 8.4: Operating conditions and simulation results for extractive batch distillation of acetone
and methanol using water as entrainer.

Effect studied

Operating conditions

Recovery

Case 1:

Entrainer flowrate,

3:5

kmol/hr

acetone

83:9

startup 1

Reflux ratio,

Entrainer-to-vapor ratio,

0:7

2:35

Case 2:

Entrainer flowrate,

3:5

kmol/hr

acetone

85:6

startup 2

Reflux ratio,

Entrainer-to-vapor ratio,

0:7

2:40

Case 3:

Entrainer flowrate,

6:0

kmol/hr

acetone

90:6

increase E

Reflux ratio,

startup 2

Entrainer-to-vapor ratio,

1:2

2:54

Case 4:

Entrainer flowrate,

8:0

kmol/hr

acetone

90:0

increase E

Reflux ratio,

startup 2

Entrainer-to-vapor ratio,

1:2

2:52

Case 5:

Entrainer flowrate,

10:0

kmol/hr

acetone

89:1

further increase E

Reflux ratio,

startup 2

Entrainer-to-vapor ratio,

2:0

2:50

8.4 Extractive batch distillation in the conventional column

221

composition profile

Instantaneous column

Reboiler composition

trajectory

56.4 oC

Methanol

64.5 oC

Water

100.0 oC

55.5 oC

Acetone

at 1 hour

Figure 8.9: Extractive batch distillation of acetone and methanol using water as entrainer. Liquid
composition profile in the column at 1 hour and reboiler composition trajectory for the complete
process (Case 2 in Table 8.4).

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Liquid molefraction in distillate

Time [hr]

1st-cut

Off-cut

Methanol

Acetone

Water

2nd-cut

Off-cut

Figure 8.10: Distillate composition trajectory corresponding to Figure 8.9. Entrainer feeding
stopped at methanol breakthrough in acetone product (

2:40

hours). Acetone recovery 85.6

% (Case 2).

222

Chapter 8. Extractive Batch Distillation: Column Schemes and Operation

composition profile

Instantaneous column

56.4 oC

Methanol

64.5 oC

Water

100.0 oC

55.5 oC

Acetone

Reboiler composition

trajectory

at 1 hour

Figure 8.11: Increased entrainer load improves acetone recovery: Extractive batch distillation
of acetone and methanol using water as entrainer. Liquid composition profile in the column at 1
hour and reboiler composition trajectory for the complete process (Case 3).

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Acetone

Water

1st-cut

Time [hr]

Off-cut

Methanol

Liquid molefraction in distillate

2nd-cut

Off-cut

Figure 8.12: Increased entrainer load improves acetone recovery: Distillate composition tra-
jectory corresponding to Figure 8.11. Entrainer feeding stopped at methanol breakthrough in
acetone product (

2:54

hours). Acetone recovery 90.6 % (Case 3).

8.4 Extractive batch distillation in the conventional column

223

Reboiler composition

trajectory

composition profile

Instantaneous column

56.4 oC

Methanol

64.5 oC

Water

100.0 oC

55.5 oC

Acetone

at 1 hour

Figure 8.13: An even further increase in entrainer load reduces acetone recovery: Extractive
batch distillation of acetone and methanol using water as entrainer. Liquid composition profile
in the column at 1 hour and reboiler composition trajectory for the complete process (Case 5).

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Water

Time [hr]

Liquid molefraction in distillate

Methanol

Acetone

1st-cut

Off-cut 2nd-cut Off-cut

Figure 8.14: An even further increase in entrainer load reduces acetone recovery: Distillate
composition trajectory corresponding to Figure 8.13. Entrainer feeding stopped at methanol
breakthrough in acetone product (

2:50

hours). Acetone recovery 89.1 % (Case 5).

224

Chapter 8. Extractive Batch Distillation: Column Schemes and Operation

Degradation of column profile

In extractive distillation using a conventional batch column, the reboiler residue is gradually en-
riched in water during the extractive distillation period (Step 2), as illustrated in Figure 8.15.
This causes a “degradation” of the column profile which decreases the efficiency of the sepa-

Methanol

64.5 oC

55.5 oC

Reboiler composition

trajectory

Acetone

56.4 oC

Water

100.0 oC

Instantaneous column

during 1st-cut

composition profiles

Breakthrough of methanol

Figure 8.15: Degradation of column profile during 1st cut: Liquid composition profiles at

=12

min,

=30 min,

=1 hr,

=2 hr and

=2.5 hr. Entrainer feeding stopped at methanol break-

through in distillate product (

=2.5 hr) (Case 3).

ration in the extractive section. This is different from continuous extractive distillation where
the strongly curved profile in the extractive section is maintained (steady-state operation). The
original mixture of acetone and methanol is marked by F

in Figure 8.15. The resulting com-

position in the reboiler at the end of Step 1 (with startup procedure 2) is marked by B

(

[0:387;

0:396;

0:217]

). During the extractive distillation step (Step 2), the reboiler product gets

enriched in the water entrainer, marked by B

, B

, ... , B

. When the reboiler composition has

reached B

(at 2.5 hours in Figure 8.15), the instantaneous composition profile along the column

is shifted and a binary fraction of acetone and methanol is achieved in the distillate. When all
of the acetone has been removed as distillate product, the composition of the reboiler reaches the
binary edge of water and methanol and the column essentially performs a binary separation of
methanol and water.

226

Chapter 8. Extractive Batch Distillation: Column Schemes and Operation

8.4.4

Summary: Extractive batch distillation

The simulation results can be summarized as follows:

Increased entrainer load gives better separation of the azeotrope-forming components 1 and
2 in the extractive section (here: less methanol impurity in the acetone distillate product).

Increased reflux gives better separation of the entrainer and the azeotrope-forming compo-
nent 1 in the rectifying section (here: less water in the acetone distillate product).

These two effects need to be balanced as increased reflux results in less relative entrainer
load. For large entrainer flowrates there was no further advantage in terms of separation
and a large amount of the entrainer was accumulated in the reboiler.

8.5

Extractive batch distillation in the middle vessel column

8.5.1

Process description

Extractive distillation in a middle vessel batch column is illustrated in Figure 8.2. The original
mixture to be separated may be initially charged to the middle vessel or distributed to both the
reboiler and the middle vessel. The middle vessel configuration allows for simultaneous removal
of distillate and bottom product. In addition, the middle vessel may be gradually purified to
retrieve a product at the end of operation. The configuration consists of three column sections:
a rectifying section, an extractive section and a stripping section connected by a middle vessel
(holdup tank). The rectifying section serves to separate acetone (1) from water (E), the extractive
section to extract methanol (2) from the original feed mixture with the help of water (E), and the
stripping section to strip acetone (1) out of the bottom product (possibly E).

The operating procedure for the middle vessel extractive batch distillation process used in this
study is:

Step 1 Startup: Total reflux (and reboil) operation to build up the column composition profile.

Step 2 1st cut: Production of the first distillate product (component 1) and removal of bottom

product. Finite reflux with continuous entrainer feeding. (Step 2 is the focus of this paper.)

Step 3 Off-cut: Removal of an off-cut fraction (either as distillate or bottom product or both) to

purify component 2 in the middle vessel. Finite reflux and reboil without entrainer feeding.

Step 4 2nd cut: Production of second distillate or bottom product.

Finite reflux without entrainer feeding.

228

Chapter 8. Extractive Batch Distillation: Column Schemes and Operation

Separation schemes

There are two principally different ways of operating the extractive middle vessel column:

Scheme I: Simultaneously producing acetone (1) and water (E) with methanol (2) accumu-

lated in the middle vessel;

Scheme II: Producing acetone (1) distillate only and removing a fraction of methanol (2) and

water (E) in the bottom (2 + E) to be separated in a second operation.

These alternative separation schemes are illustrated in Figure 8.18.

2,E

emptying middle vessel

II. Sequential separation:

recover 2 in middle vessel

ternary split 1/2/E

direct split 1/2,E and 2/E

I. Simultaneous separation:

Figure 8.18: Separation schemes for the separation of an original binary mixture 1 and 2 using E
as entrainer in the extractive middle vessel batch distillation column.

Scheme I is attractive since it requires only a single operation. However, it may not be optimal in
terms of product recovery and minimum batch time.

Scheme II is very similar to continuous extractive distillation scheme with a direct sequence of
two columns.

The main problem with the basic and vapor bypass configurations is that entrainer must accu-
mulate in the middle vessel during Step 2 (entrainer feeding). This follows from a component
balance for entrainer in the middle vessel. For the “best” case where we neglect the volatility of
the entrainer in the basic configuration, we get:

)

(8.1)

where

is the molefraction of entrainer in the middle vessel. If

, then the amount of

entrainer accumulated will increase steadily. If

, then the amount will increase until we

reach a value where

, i.e. where

. (This simple analysis which gives

the “unavoidable” accumulation of entrainer in the middle vessel may be used to conclude that
some earlier claims in the literature about simultaneous separation into pure products in the basic
configuration must be incorrect.)

8.5 Extractive batch distillation in the middle vessel column

229

8.5.2

Simulation results with simultaneous separation (Scheme I)

The original mixture feed composition is 50 mol% acetone and 50 mol% methanol. Pure water is
used as the entrainer. The column specifications and initial conditions are given in Table 8.5.

Table 8.5: Column data and initial conditions for the middle vessel extractive batch distillation
column separating acetone and methanol using water as entrainer.

Total no. of trays

(excl. reboiler)

No. of trays per section

Total initial charge

5:385

kmol

Condenser holdup (constant)

0:035

kmol

Initial middle vessel holdup

5:125

kmol

Reboiler holdup (constant)

0:035

kmol

Tray holdup (constant)

1=300

kmol

Boilup, vapor flow (constant)

5:0

kmol/h

The operating conditions are chosen to be comparable with the parameters used for the conven-
tional batch extractive distillation presented in the previous section (Case 3 in Table 8.4). The
values are by no means optimal for the operation of the extractive middle vessel batch distillation
column. The product specifications are the same.

The results from dynamic simulation of selected cases are summarized in Table 8.6 where we see
that the liquid bypass configuration is superior in terms of acetone recovery.

Table 8.6: Operating conditions and simulation results for middle vessel extractive batch distil-
lation of acetone and methanol using water as entrainer.

Configuration

Operating conditions

Recovery

Liquid bypass

Entrainer flowrate,

6:0

kmol/hr

acetone

93:3

Reflux ratio,

Entrainer-to-vapor ratio,

1:2

1:0

kmol/hr and

6:0

kmol/hr

2:62

Basic

Entrainer flowrate,

6:0

kmol/hr

acetone

90:3

Reflux ratio,

Entrainer-to-vapor ratio,

1:2

7:5

kmol/hr and

2:5

kmol/hr

2:53

Vapor bypass

Entrainer flowrate,

6:0

kmol/hr

acetone

46:4

Reflux ratio,

Entrainer-to-vapor ratio,

1:2

7:5

kmol/hr and

2:5

kmol/hr

1:30

230

Chapter 8. Extractive Batch Distillation: Column Schemes and Operation

For the liquid bypass configuration we used the following procedure (see also Table 8.6):

Step 1 Startup: with bottom removal: R=

, E=B=6.0 kmol/hr, L

=0;

Step 2 Distillate 1st cut: R=4, E=B=6.0 kmol/hr, L

=D=1.0 kmol/hr (until

2:62

hr);

Step 3 Distillate off-cut: R=4, E=0, L

=2.0 kmol/hr, B=1.0 kmol/hr (until

2:95

hr);

Step 4 Only bottom product removal: R=

, E=0, L

=1.0 kmol/hr, B=1.0 kmol/hr (until

4:84

hr);

Note that we remove a bottom product during startup (because the level in the reboiler was con-
stant). This results in some loss of acetone in the bottom flow (about 3 % of the initial methanol
charge). In Step 2 we used

and

, that is, the amount in the middle vessel

decreases according to what is removed as distillate.

The results for the liquid bypass configuration are illustrated in Figures 8.19 - 8.22. We were
unable to obtain simultaneous production of acetone distillate and pure water entrainer bottom
product.

For the basic and vapor bypass configurations it is even more difficult to choose reasonable values
for the operating parameters. We used

M,min

=V) and tried several values. In addition

to the two cases shown in Table 8.6, simulations with

kmol/hr (B=E) yield 82.4 %

acetone recovery with the basic configuration and 68.4 % acetone recovery with the vapor bypass
configuration.

We conclude that simultaneous separation (Scheme I) is difficult to achieve. However, if we
allow for a separate step to purify the bottom product (Scheme II) then the purity specifications
are easily achieved.

8.5 Extractive batch distillation in the middle vessel column

231

composition profile

Instantaneous column

56.4 oC

Methanol

64.5 oC

Water

100.0 oC

55.5 oC

Acetone

at 1 hour

Middle vessel

composition trajectory

end of Step 2

Figure 8.19: Extractive middle vessel batch distillation with liquid bypass configuration attempt-
ing simultaneous separation (Scheme I). Operating conditions given in Table 8.6.

0.5

1.5

2.5

3.5

4.5

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Time [hr]

Liquid molefraction in distillate

1st cut

Acetone

Off-cut

Time [hr]

Figure 8.20: Distillate composition trajectory for liquid bypass configuration corresponding to
Figure 8.19. Entrainer feeding stopped at methanol breakthrough in acetone product (

2:62

hr).

232

Chapter 8. Extractive Batch Distillation: Column Schemes and Operation

0.5

1.5

2.5

3.5

4.5

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Liquid molefraction in middle vessel

Acetone

Time [hr]

Water

Methanol

Figure 8.21: Middle vessel composition trajectory for liquid bypass configuration corresponding
to Figure 8.19.

0.5

1.5

2.5

3.5

4.5

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Time [hr]

Liquid molefraction in bottom

Water

Methanol

Figure 8.22: Bottom product composition trajectory for liquid bypass configuration correspond-
ing to Figure 8.19.

234

Chapter 8. Extractive Batch Distillation: Column Schemes and Operation

Appendix 8A

The vapor- liquid equilibrium calculations are based on the following equations:

;

)

sat

)

and

i=1

where

1 atm in all the simulations presented in this paper. The vapor pressures were calcu-

lated by the Antoine equation:

sat

where

sat

is in

and

. The liquid activity coefficients were calculated by the Wilson

equation:

k =1

k i

k j

exp

where

is in

=mol

cal =mol

and

. Parameters for the Antoine and Wilson

equations for the example mixtures used in the simulations presented in this article are given in
Table 8.7.

8.6 Conclusion

235

Table 8.7: Thermodynamic Data for the Example Mixtures

Parameters for the Antoine Equation

[

mol

]

(1) Ethanol

23.5807

3673.81

-46.681

58.69

(2) Water

23.2256

3835.18

-45.343

18.01

(3) Ethylene glycol

25.1431

6022.18

-28.250

55.92

Binary Interaction Parameters for the Wilson Equation

393:1971

129:2043

926:263

1266:0109

1539:4142

1265:7398

Parameters for the Antoine Equation

[

mol

]

(1) Acetone

21.3009

2801.53

-42.875

74.04

(2) Methanol

23.4999

3643.31

-33.434

40.73

(3) Water

23.2256

3835.18

-45.343

18.01

Binary Interaction Parameters for the Wilson Equation

161:8813

489:3727

583:1054

19:2547

1422:849

554:0494

From Gmehling et al. (1977-1990), VLE Data Collection, DECHEMA series, at atmo-

spheric pressure.

238

Chapter 9. Control Strategies for Extractive Middle Vessel Batch Distillation

Condenser

Reboiler

Liquid bypass modification

Middle vessel

Extractive

Section

Stripping

Section

Entrainer feed

Rectifying

Section

Figure 9.1: Extractive middle vessel batch distillation column. To the right is the liquid bypass
configuration proposed by Warter et al. (1999).

open loop studies have identified operating parameters that influence the separation performance.
Amongst these are the entrainer feed flowrate (most important), reflux ratio, entrainer feed purity
and temperature. Column design parameters like entrainer feed tray location and middle vessel
configuration are also found to be important.

Hilmen et al. (1997) were the first to propose a control strategy for extractive batch distillation in
the middle vessel column. In the present paper, these considerations are extended to include the
liquid bypass configuration proposed by Warter and Stichlmair (1999), which was shown to have
the the best separation performance for extractive distillation in the middle vessel batch column.

Recently, Phimister and Seider (2000b) proposed a DB-control configuration for extractive dis-
tillation in the middle vessel column (where the distillate flow is used to control the distillate
purity and the bottom flow is used to control the bottom purity by single loop P-controllers and
the entrainer flowrate is constant). However, they used the vapor bypass configuration (that is,

9.2 Important considerations for control

239

only liquid flows through the middle vessel) which was found to be inferior in Chapter 8.

Farschman and Diwekar (1996), Barolo et al. (1998) and Phimister and Seider (2000a) evaluated
various control configurations for the middle vessel batch distillation column separating ideal
ternary mixtures. Farschman and Diwekar (1996) use the ratio of the vapor flow in the rectifying
section to the vapor flow in the stripping section as a control parameter (this ratio

was proposed

by Davidyan et al. (1994) as discussed in Chapter 6). They showed that the middle vessel provides
a decoupling of the two-end control loops. This middle vessel decoupling effect between the
attached column sections was also pointed out by Barolo et al. (1996a) and Barolo et al. (1996b).

An indirect level control strategy was proposed for the multivessel batch distillation column
(Skogestad et al., 1997), and our proposed control strategy is similar to this.

The literature on control of extractive distillation in continuous columns is extensive and was
reviewed by Jacobsen et al. (1991). Andersen et al. (1995) considered integrated design and
control of continuous extractive distillation columns. They show that a simpler control situation
can be obtained by increasing the entrainer feed flow above its economically optimum.

9.2

Important considerations for control

An analysis of the middle vessel batch distillation columns reveals some important considerations
for control:

1. The entrainer feed flowrate determines the separation efficiency between the original mix-

ture components 1 and 2 in the extractive section. There is a minimum entrainer feed
flowrate necessary to “break” the azeotrope

2. High reflux “weakens” the effect of the entrainer (E). There is a minimum and a maximum

reflux ratio in order to achieve acceptable separation between component 1 and E in the
rectifying section (1/E split) and between component 1 and 2 in the extractive section (1/2
split with help of E).

3. Entrainer may accumulate in the feed vessel (reboiler or middle vessel).

4. The middle vessel decouples the extractive section and the stripping section. However, this

decoupling is less with the liquid bypass configuration.

5. Because the entrainer dominates the liquid phase in the extractive section during the en-

trainer feeding period, it is difficult to use temperature measurements to infer compositions
in this section.

6. Temperature is not a unique indicator of the composition for multicomponent mixtures.

This may cause difficulties in using temperature measurements to infer composition even
in the stripping section of the extractive middle vessel column.

240

Chapter 9. Control Strategies for Extractive Middle Vessel Batch Distillation

7. The liquid load (internal liquid flow) is higher in the stripping section than in the rectifying

section and, for some cases, in the extractive section. This may cause difficulties in achiev-
ing high purity of the bottom product (entrainer recovery). It may be necessary to have a
higher vapor flowrate in the stripping section of the column.

In continuous extractive distillation the entrainer-to-feed flow ratio (E/F) is usually held con-
stant, although some authors propose to keep the entrainer-to-distillate flow ratio (E/D) constant
(Westerberg and Wahnschafft, 1996). In most continuous extractive distillation processes, the
entrainer flowrate (E) is not used for control. However, due to the transient behavior of batch
operation it seems more pertinent to use the entrainer flowrate (E) for composition control in the
extractive batch distillation column.

In order to minimize the time consumption of the batch distillation, maximum boilup (constant
vapor flow in all three column sections) was used in the simulations presented in this study.

9.3

Two-point composition control

We propose the following feedback control strategy for the extractive middle vessel batch distil-
lation column (see Figure 9.2) during the entrainer feeding period:

Liquid reflux (L

) is used to control the 1/E split in the rectifying section;

Entrainer feed flowrate (E) is used to control the 1/2 split in the extractive section;

Liquid flow out of middle vessel (L

) is used to control the 2/E split in the stripping section;

Distillate flowrate (D) is used to control the level in condenser drum;

Bottom flowrate (B) is used to control the level in reboiler.

Note that it may be possible to change the “pairings”, but the above proposal gives the most
“direct” effect in terms of composition control. Due to lack of composition measurements, we
propose to use the reflux (L

) to control the column temperature near the middle of the rectifying

section. (Alternatively, the reflux ratio (R

D) may be held constant within its allowed range

(minimum and maximum value.) Similarly, the liquid flow from the middle vessel (L

) is used

to control the temperature in the stripping section below (

). This also indirectly adjusts the

middle vessel holdup and there is no need for level control (same strategy as used in Chapter 7).
We do not recommend using a temperature measurement for adjusting the entrainer flowrate since
the temperature in the extractive section is dominated by the entrainer and thus insensitive to the
split between component 1 and 2. We therefore propose to measure the split between 1 and 2 by
measuring the amount of component 2 (impurity) in the distillate (

). An alternative is to use

a more direct composition measurement of the methanol in the extractive section, but this is less
practical.

242

Chapter 9. Control Strategies for Extractive Middle Vessel Batch Distillation

9.4

Simulation model

The mathematical model is based on the same assumptions as the model described in Chapter 8.
The model with PI control was implemented in SpeedUp (1993) and the input file is included in
Appendix A in the back of the thesis.

9.5

Closed loop simulation results

In this study we consider the mixture acetone (1) - methanol (2) - water (E) (System 2 in Chapter
8). The original mixture feed composition was 50 mol% acetone and 50 mol% methanol. Pure
water was used as entrainer. The column specifications and initial conditions for the simulations
are given in Table 9.1.

Table 9.1: Column data and initial conditions for the middle vessel extractive batch distillation
column separating acetone and methanol using water as entrainer.

Total no. of trays

(excl. reboiler)

No. of trays per section

Total initial charge

5:385

kmol

Condenser holdup (constant)

0:035

kmol

Initial middle vessel holdup

5:125

kmol

Reboiler holdup (constant)

0:035

kmol

Tray holdup (constant)

1=300

kmol

We used the same product specifications as in Chapter 8:

Acetone: 96.0 mol % acetone (maximum 0.1 mol % methanol and 3 mol % water impurity);

Methanol: 99.0 mol % methanol (maximum 1.0 mol % acetone impurity);

Entrainer recovery: 99.9 mol % water (maximum 0.1 mol % methanol impurity).

For all the simulations in this study we use constant vapor flow in the stripping section (

and constant vapor flow in the extractive and rectifying sections (

). The extractive section

determines the minimum vapor flow in the column

min

. We have considered two cases:

Equal vapor flows:

kmol/hr;

Different vapor flows:

kmol/hr and

kmol/hr

(with cooling in the middle vessel).

The cooling introduces an extra degree of freedom which may be necessary to achieve a pure
bottom product.

9.5 Closed loop simulation results

243

The control parameters for the PI-controllers used in the simulations are given in Table 9.2.

Table 9.2: Control parameters used in simulations of extractive middle vessel batch distillation
with liquid bypass configuration in Figure 9.2.

CC-loop:

1000

0:25

2;sp

0:001

max

kmol/hr

TC-loop:

0:25

0:15

2;sp

specified in Table 9.3,

M,min

Step 1 Startup: The column is operated at total reflux (D=B=0) and L

=0 with constant entrainer

flowrate

kmol/hr in 15 minutes (that is, the middle vessel and reboiler holdup

increase). After the startup period the reboiler holdup is pure water with

1:035

kmol and the middle vessel holdup is

5:715

kmol.

Step 2 Extractive distillation operation: The controllers are switched on (manipulating the en-

trainer flowrate E and the liquid flow out of the middle vessel L

) in order to drive the

distillate composition and bottom entrainer composition to their desired values. The en-
trainer feeding is stopped when the impurity of the distillate product exceeds a pre-specified
maximum value (at time

Step 3 Ordinary distillation: The column is operated without entrainer feed as an ordinary middle

vessel batch distillation column with feedback control to separate the residual material in
the system. The setpoints

2;sp

and

2;sp

in Table 9.2 are changed accordingly.

The results of the closed loop extractive distillation operation (Step 2) are summarized in Table
9.3. The two different separation schemes described in Chapter 8 (I: simultaneous and II: sequen-
tial) were implemented by changing the temperature set-point in the stripping section

2;sp

(that

determines the 2/E split). We only show the results for the extractive distillation operation (Step
2) as this is the crucial part of the separation. Step 3 is ordinary distillation separating methanol
and water in the middle vessel column.

For the simultaneous separation (Scheme I), we found that the vapor flow must be higher in the
stripping section than in extractive and rectifying sections in order to obtain the specified water
purity in the bottom product (Case 1: acetone recovery 86.0 %). Operation with equal vapor flows
and parameters in Case 2 did not obtain pure water in the bottom product during the whole Step
2 (see Figure 9.12b). In both cases, there must be a period without entrainer feeding where the
middle vessel product (methanol) is purified by ordinary distillation, but this could possibly have
been avoided by using the alternative liquid bypass configuration shown in Figure 9.2.

From Figure 9.6 we see that the entrainer flowrate is almost constant during Step 2: The entrainer
load is high initially (E

max

), then almost constant at about 5.2 kmol/hr (moderately increasing)

with a rapid increase towards the end of the extractive distillation period (E

max

). This indicates

244

Chapter 9. Control Strategies for Extractive Middle Vessel Batch Distillation

Table 9.3: Operating conditions and simulation results for control of extractive middle vessel
batch distillation with liquid bypass configuration.

Separation scheme

Operating conditions

Recovery

Case 1:

Reflux ratio,

acetone

86:0

I: simultaneous

2sp

77:0

different vapor flows

kmol/hr,

kmol/hr

2:41

Case 2:

Reflux ratio,

acetone

86:7

I: attempt simultaneous

2;sp

77:0

equal vapor flows

kmol/hr,

kmol/hr

2:43

Case 3:

Reflux ratio,

acetone

99:3

II: sequential

2;sp

70:0

equal vapor flows

kmol/hr,

kmol/hr

2:79

Case 4:

Reflux ratio,

acetone

99:4

II: sequential

2;sp

70:0

different vapor flows

kmol/hr,

kmol/hr

2:79

Startup period (15 min):

kmol/hr,

that it may be acceptable with constant entrainer feed flowrate for batchwise extractive distillation
in the middle vessel column (similar to continuous columns).

From Figure 9.7 we see that the liquid flow out of the middle vessel increases to a value twice
the distillate flow (Case 1: D= 1 kmol/hr). The middle vessel holdup is decreased during the
whole extractive distillation operation (Step 2). The bottom flowrate (B) is governed by varying
entrainer feed flowrate (proportionally by mass balance).

For the sequential separation (Scheme II), the separation between 2 and E in the stripping section
is not important (that is, a mixture of 2 and E is removed in the bottom similar to the continuous
extractive distillation scheme) and the vapor flow may be equal in all sections. In Case 4, the
entrainer flowrate reaches saturation E

max

=10 kmol/hr. The entrainer does practically not accu-

mulate in the middle vessel even though the entrainer load is higher (see Figure 9.17) and the
acetone recovery is greatly improved (Case 4: acetone recovery 99.4 %).

246

Chapter 9. Control Strategies for Extractive Middle Vessel Batch Distillation

0.5

1.5

2.5

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Time [hr]

Acetone

Methanol

Water

Liquid molefraction in distillate

(a)

Distillate product composition

0.5

1.5

2.5

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Time [hr]

Acetone

Methanol

Water

Liquid molefraction in bottom

(b)

Bottom product composition

Figure 9.5: Distillate and bottom product composition trajectories corresponding to Figure 9.3.
Simultaneous production of acetone and water during the whole entrainer feeding period (Step
2) of the extractive distillation process (Case 1).

250

Chapter 9. Control Strategies for Extractive Middle Vessel Batch Distillation

0.5

1.5

2.5

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Time [hr]

Acetone

Methanol

Water

Liquid molefraction in distillate

(a)

Distillate product composition

0.5

1.5

2.5

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Time [hr]

Acetone

Methanol

Water

Liquid molefraction in bottom

(b)

Bottom product composition

Figure 9.12: Distillate and bottom product composition trajectories corresponding to Figure 9.10.
The separation in the stripping section is insufficient, that is, the bottom product is not pure water
during the whole entrainer feeding period (Case 2).

252

Chapter 9. Control Strategies for Extractive Middle Vessel Batch Distillation

0.5

1.5

2.5

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Time [hr]

Acetone

Methanol

Water

Liquid molefraction in distillate

(a)

Distillate product composition

0.5

1.5

2.5

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Time [hr]

Acetone

Methanol

Water

Liquid molefraction in bottom

(b)

Bottom product composition

Figure 9.15: Distillate and bottom product composition trajectories corresponding to Figure 9.13.
Production of acetone distillate product and a fraction of methanol and water in the bottom
product (Case 3).

260

Chapter 10. Practical Implications, Conclusions and Suggestions for Future Work

Complete separation of multicomponent azeotropic mixtures may be very difficult and sometimes
an azeotrope fraction is an acceptable product. Examples of such mixtures from industry are
THF (tetrahydrofuran) -MEK (methylethylketone) - toluene; methanol - ethanol - propanol; and
benzene - toluene - xylene (Schweitzer, 1997). In the worst cases, it may only be possible to take
fractions by boiling-point range, reusing as much as possible and incinerating or selling the rest.
Having said this, there are “azeotrope-breaking” processes that separate these mixtures into pure
components, and some of these processes are developed in particular for solvent recovery in fine-
and specialty chemical industries.

Specifications for recovered solvents which normally can be met economically are listed by Drew
(1975) are: water content of solvent; 0.05-0.5 %, cross mixing of solvents; 0.5-5%, and organic
content of waste water; 10 to 100 parts/million. Usually, this means the recovered solvent will
meet manufacturer’s specifications for industrial-solvent grades.

How to approach and analyze the separation task

There is no standard procedure by which separation processes are chosen. However, there are
certain guidelines that are usually followed (Rousseau, 1987):

1. Define the problem. Usually, this entails establishing product purity and recovery specifi-

cations.

2. Determine which separation methods are capable of accomplishing the separation.

The general approach for the development of a liquid mixture separation process is illustrated in
Figure 10.1. The first task is often to assess the vapor-liquid-liquid equilibrium behavior of the

Modeling and
simulation,

experiments

Generate

alternatives

process

Properties

Analysis

What is feasible?

Identifying alternatives

Separation system process developement

Figure 10.1: Development of separation systems.

species to be separated (Westerberg and Wahnschafft, 1996). Experimental data is the only assure
way to predict the behavior, and many such data are currently available. Horsley (1973), for ex-
ample, has compiled a listing of known azeotropes. Available data for ternary or higher mixtures

10.1 Practical implications

261

over the concentration range of interest are seldom adequate (Stichlmair and Fair, 1998). West-
erberg and Wahnschafft (1996) give a way to determine if the multicomponent mixture displays
highly nonideal behavior (from flash calculations) based on binary pair infinite dilution K-values
(see also Chapter 3, Section 3.8). The hydrogen-bonding guidelines given by Berg (1969) based
on specific molecular groups are also useful for predicting deviation from ideality.

If the mixture displays azeotropic or heterogeneous behavior, the separation systems will be very
different from the distillation systems for fairly ideal mixtures (Westerberg, 1997).

Generally, if the mixture in question is heterogeneous or multiphase, phase separation of the
existing phases should be carried out before any homogeneous separation, as recommended by
Smith (1995). For example, pretreatment by filtration or decantation may be adequate if the given
mixture is contaminated with solids and tars.

The next step after this property assesment is to generate separation process alternatives. This
may be done by “property difference analysis” described by, for example, Jaksland and Gani
(1995; 1997). The property differences of the mixture components indicate what separation unit
operation that is most tangible for the given separation task.

A binary homoazeotropic mixture is inherently impossible to separate by ordinary distillation at
the system pressure. Unless pressure variation can be utilized, which is the first thing to eval-
uate, special methods are required, such as heteroazeotropic or extractive distillation or hybrid
processes that combine membranes and distillation, for example. For these the entrainer-addition
distillation methods we need to search for candidate entrainers that gives (ternary) VLE diagrams
that are promising for these separation schemes (Chapter 2).

If the mixture has more than three components, we need to analyze each binary pair and also the
ternary combinations of the components. This gives us useful information on how to separate the
mixture as a whole.

To determine the mixture’s VL(L)E behavior there are many commercial software packages avail-
able, for example DISTILL from HyproTech and SPLIT from AspenTech. The physical proper-
ties can be obtained from databanks or by physical property estimation packages. From plots of
the phase behavior such as T-x,y diagrams for each binary pair in the mixture, we can see whether
the mixture displays azeotropic and/or heterogeneous behavior. Azeotropic data handbooks (e.g.
Horsley (1973) and Gmehling et al. (1994)) which often provide azeotrope composition and boil-
ing point at given pressures are also useful in determining whether there are any azeotropes ap-
pearing. A step-by-step approach with examples is given by Westerberg and Wahnschafft (1996)
and Westerberg (1997).

Phase diagram analysis such as plot of distillation lines or residue curves on a composition di-
agram is an excellent tool for gaining insights into the complex behavior of nonideal ternary
mixtures. Some uses for industrial mixtures are illustrated by, for example, Wasylkiewicz et al.
(1999b) and in Chapter 2 of this thesis.

262

Chapter 10. Practical Implications, Conclusions and Suggestions for Future Work

From the binary VL(L)E information we recommend drawing composition triangles (either by
hand or by using the mentioned software packages) which include:

location and boiling temperature of the pure components and azeotropes;

sketch of residue curves or distillation lines near the singular points with arrows indicating
increasing temperature.

The ternary mixture has one of the 26 possible topological VLE diagram structures given in
Figure 3.21. See also Table 3.8 for a quick look-up on possible structure classes based solely on
the binary information. In most cases the topological structure can be determined based solely
on this information (see Chapter 3). Usually the existence of azeotropes split the VLE diagram
into distillation regions that limit the possibility of separating the mixture by ordinary distillation.
One should:

localize the feed composition of the mixture in question;

as a first approximation use the

approach or

approach described in Chapter 2

to predict the product areas for this feed.

If separation into (relatively) pure components is not possible by ordinary distillation, then we
need to consider special methods for separating the mixture. Chapter 2 gives an overview of
various “azeotrope-breaking” methods and industrial processes.

10.2

Conclusions

The thesis offers a critical evaluation of azeotropic distillation theory and puts the large number
of publications into perspective. Useful knowledge from the Russian distillation research com-
munity has been included through the cooperation with the thesis co-advisor Dr. Valeri N. Kiva.

The main contributions of this thesis are summarized below:

Chapter 2: Gives an introduction and overview of azeotropic mixtures separation, and illustrates
the use of azeotropic phase equilibrium diagrams for the development of entrainer-addition dis-
tillation systems. The powerful process of combined heteroazeotropic and extractive distillation
is given as an example that is normally excluded by entrainer selection criteria given in recent
textbooks and encyclopedia.

Chapter 3: Provides a survey on azeotropic phase equilibrium diagrams with critical evaluation
of the published literature in this field of research. The specific contributions are given in the
introductory pages of the chapter.

Chapter 4: Presents the simplifying concept of elementary topological cells of ternary VLE
diagrams. This abstraction of the structural VLE diagrams greatly reduces the number structures

10.3 Suggestions for future work

263

that need to be analyzed in order to reveal the qualitative characteristics of any ternary azeotropic
mixture. Statistics on the physical existence (occurrence) of azeotropic mixtures are also included.

Chapter 5: Examines the effect of column holdup in residue curve map prediction of batch
distillation products and off-cuts. Even though column holdup was found to affect the distillate
product sequence, the quick analysis based on residue curve maps and simple models of batch
distillation neglecting tray holdup can be recommended if used with care.

Chapter 6: Gives an introduction to batch distillation in general, and provides a comparison of
the time consumption of the cyclic and the closed middle vessel batch distillation for separating
ternary ideal mixtures. Time savings of up to 50 % were found for the middle vessel column for
difficult (low relative volatility) separations. The middle vessel column performs best for feeds
which are rich in light and heavy components as opposed to some previous claims in the literature.

Chapter 7: Determines the use VLE diagram information, such as distillation lines and isotherms,
as a graphical tool for operation and control considerations in cyclic and closed middle vessel
batch distillation of ternary azeotropic mixtures.

Chapter 8: Extends extractive distillation insight to batch distillation operation in the conven-
tional and middle vessel batch columns. The influence of column design and extractive distillation
operating parameters on the separation performance has been determined. The middle vessel col-
umn with the liquid bypass configuration was found to have the best performance.

Chapter 9: Presents a novel control strategy for extractive batch distillation in the middle vessel
column with the liquid bypass configuration. Tight composition control and simultaneous produc-
tion of distillate product and entrainer bottom product are obtained during the entrainer feeding
period, but an additional step to purify the middle vessel product was found to be unavoidable. A
refinement of the liquid bypass configuration that may avoid entrainer accumulation in the middle
vessel is proposed.

10.3

Suggestions for future work

Some directions for future work may include:

Further extension of continuous azeotropic distillation insights to batch operation. In particu-

lar, the close resemblance between continuous distillation and the middle vessel column
makes the extension for this batch column operation easier. Recently, the research groups
of Stichlmair and coworkers (TU Munich) and Barton and coworkers (MiT Boston) have
pursued this topic (see reference list).

Analysis of the batch distillation behavior for mixtures of Cell III (and Cell IV) and possible

multiplicity (see Chapter 7) remains an unsolved problem.

Heteroextractive batch distillation is a powerful method of separating azeotropic mixtures that

is almost unexplored. We propose to investigate a heteroazeotropic or combined het-

270

References

Barolo, M., G.B. Guarise, N. Ribon, S.A. Rienzi, A. Trotta and S. Macchietto (1996a). Some

Issues in the Design and Operation of a Batch Distillation Column with a Middle Vessel.
Comp. Chem. Engng. 20(Suppl.), S37–S42.

Barolo, M., G.B. Guarise, S.A. Rienzi, A. Trotta and S. Macchietto (1996b). Running Batch

Distillation in a Column with a Middle Vessel. Ind. Eng. Chem. Res. 35(12), 4612–4618.

Barolo, M., G.B. Guarise, S.A. Rienzi and A. Trotta (1998). Understanding the Dynamics of a

Batch Distillation Column with a Middle Vessel. ESCAPE-8, Brugge, Belgium, May 24-27,
1998, Supplement to Comp. Chem. Engng. 22(Supplement), S37–S44.

Bauer, M.H. and J. Stichlmair (1995). Synthesis and Optimization of Distillation Sequences for

the Separation of Azeotropic Mixtures. Comp. Chem. Eng. 19, S15–S20.

Bekiaris, N. and M. Morari (1996). Multiple Steady States in Distillation:

1=1

Predictions,

Extensions, and Implications for Design, Synthesis, and Simulation. Ind. End. Chem. Res.
35, 4264–4280.

Bekiaris, N., G.A. Meski, G.A. Raudu and M. Morari (1993). Multiple Steady States in Homoge-

neous Azeotropic Distillation. Ind. Eng. Chem. Res. 32(9), 2023–2038.

Benedict, M. and D.R. Rubin (1945). Extractive and Azeotropic Distillation I. Theoretical As-

pects. Trans. Am. Inst. Chem. Eng. 41, 353–370.

Berg, L. (1969). Azeotropic and Extractive Distillation: Selecting the Agent for Distillation.

Chem. Eng. Prog. 65(9), 52–57.

Bergdorf, J. (1991). Case Study of Solvent Dehydration in Hybrid Processes with and without Per-

vaporation. Proc. Fifth International Conference on Pervaporation Processes in the Chemi-
cal Industry. Heidelberg (GE), 362–382.

Bernot, C. (1990). Design and Synthesis of Multicomponent Batch Distillation. PhD thesis. Uni-

versity of Massachusetts, Amherst, USA.

Bernot, C., M.F. Doherty and M.F. Malone (1990). Patterns of Composition Change in Multicom-

ponent Batch Distillation. Chem. Eng. Sci. 45,(5), 1207–1221.

Bernot, C., M.F. Doherty and M.F. Malone (1991). Feasibility and Separation Sequencing in

Multicomponent Batch Distillation. Chem. Eng. Sci. 46(5/6), 1311–1326.

Biegler, L.T., I.E. Grossmann and Westerberg A.W. (1997). Systematic Methods of Chemical

Process Design. Prentice Hall, NJ.

Bogdanov, V.S. and V.N. Kiva (1977). Localization of Single [Unity] K- and

-Lines in Analysis

of Liquid-Vapor Phase Diagrams. Russ. J. Phys. Chem. 51(6), 796–798 [1349–1352].

Bortolini, P. and G.B. Guarise (1970). A New Method of Batch Distillation (in Italian).

Quad. Ing. Chim. Ital. 6, 1–9.

References

271

Bossen, B.S., S.B. Jørgensen and R. Gani (1993). Simulation, Design, and Analysis of Azeotropic

Distillation Operations. Ind. Eng. Chem. Res. 32(4), 620–633.

Boyarinov, A.I., V.N. Vetokhin, T.N. Gartman, V.V. Kafarov and D.N. Motyl’ (1974). Calcu-

lation of the Limits of Distillation Regions [Calculation of Residue Curve Boundaries].
Russ. J. Phys. Chem. 48(2), 299–302.

Brignole, E.A., S. Bottini and R. Gani (1986). A Strategy for the Design and Selection of Solvents

for Separation Processes. Fluid Phase Equilibria 29, 125–132.

Brown, W.D. (1963). Economics of Recovering Acetic Acid. Chem. Eng. Prog. 59(10), 65–68.

Bushmakin, I.N. and I. Kish (1957a). Isobaric Liquid-Vapor Equilibrium in a Ternary System with

an Azeotrope of the Saddlepoint Type. Russ. J. Appl. Chem. 30(2), 205–215 [200–211].

Bushmakin, I.N. and I.N. Kish (1957b). ”Separating Lines” of Distillation and Rectification

of Ternary Systems [Residue Curve Boundaries and Distillation of Ternary Mixtures].
Russ. J. Appl. Chem. 30(4), 595–600 [561–567].

Bushmakin, I.N. and P.Ya. Molodenko (1957). The Choice of Entrainer for Azeotropic Recti-

fication (in Russian). Vestnik of Leningrad State University, Ser. Physics and Chemistry
10, 68–92.

Castillo, F.J.L., D.Y.C. Thong and G.P. Towler (1998). Homogeneous Azeotropic Distillation.

1. Design Procedure for Single-Feed Column at Nontotal Reflux. Ind. Eng. Chem. Res.
37, 987–997.

Chang, T. and T.T. Shih (1989). Development of an Alternative Distillation Scheme for Purifica-

tion of Tetrahydrofuran. Fluid Phase Equilibria 52, 161–168.

Cheong, W. and P.I Barton (1999a). Azeotropic Distillation in a Middle Vessel Batch Column. 1.

Model Formulation and Linear Separation Boundaries. Ind. Eng. Chem. Res. 38(4), 1504–
1530.

Cheong, W. and P.I Barton (1999b). Azeotropic Distillation in a Middle Vessel Batch Column. 2.

Nonlinear Separation Boundaries. Ind. Eng. Chem. Res. 38(4), 1531–1548.

Cheong, W. and P.I Barton (1999c). Azeotropic Distillation in a Middle Vessel Batch Column. 3.

Model Validation. Ind. Eng. Chem. Res. 38(4), 1549–1564.

Christensen, F.M. and S.B. Jørgensen (1987). Optimal Control of Binary Batch Distillation with

Recycled Waste Cut. Chem. Engng. J. 34, 57–64.

Davidyan, A.G., V.N. Kiva and V.M. Platonov (1991a). Batch Rectification in a Two-Section

Column with a Central Reservoir [Middle Vessel] for Identical Vapor Flows in the Sections.
Theor. Found. Chem. Eng. 25(6), 771–782.

272

References

Davidyan, A.G., V.N. Kiva and V.M. Platonov (1991b). Separation of Binary Mixtures in

Two-Section Column with Central Reservoir [Middle Vessel]. Theor. Found. Chem. Eng.
25(4), 467–475.

Davidyan, A.G., V.N. Kiva and V.M. Platonov (1992a). Batch Fractionation of Multi-

component Mixtures in Two-Section Column with Central Reservoir [Middle Vessel].
Theor. Found. Chem. Eng. 26(4), 467–477.

Davidyan, A.G., V.N. Kiva and V.M. Platonov (1992b). Batch-Type Fractionation in Two-Section

Column with Central Reservoir [Middle Vessel] and Different Vapor Flow Rates in Sections.
Theor. Found. Chem. Eng. 26(2), 163–172.

Davidyan, A.G., V.N. Kiva and V.M. Platonov (1993). The Minimal-Reflux Regime in a Rec-

tification Column with a Middle Reservoir [Middle Vessel]. Theor. Found. Chem. Eng.
27(4), 373–380.

Davidyan, A.G., V.N. Kiva, G.A. Meski and M. Morari (1994). Batch Distillation with a Middle

Vessel. Chem. Eng. Sci. 49(18), 3033–3051.

Diwekar, U.M. (1995). Batch Distillation. Simulation, Optimal Design and Control. Series in

Chem. and Mech. Eng., Taylor & Francis.

Doherty, M.F. (1985). The Presynthesis Problem for Homogeneous Azeotropic Distillation Has a

Unique Explicit Solution. Chem. Eng. Sci. 40(10), 1885–1889.

Doherty, M.F. and G.A. Caldarola (1985). Design and Synthesis of Homogenous Azeotropic

Distillations 3. The Sequencing of Columns for Azeotropic and Extractive Distillation.
Ind. Eng. Chem. Fundam. 24(4), 474–485.

Doherty, M.F. and J.D. Perkins (1978a). On the Dynamics of Distillation Processes I. The

Simple Distillation of Multicomponent Non-Reacting, Homogeneous Liquid Mixtures.
Chem. Eng. Sci. 33, 281–301.

Doherty, M.F. and J.D. Perkins (1978b). On the Dynamics of Distillation Processes II. The Simple

Distillation of Model Solutions. Chem. Eng. Sci. 33, 569–578.

Doherty, M.F. and J.D. Perkins (1979a). On the Dynamics of Distillation Processes III. The Topo-

logical Structure of Ternary Residue Curve Maps. Chem. Eng. Sci. 34, 1401–1414.

Doherty, M.F. and J.D. Perkins (1979b). The Behaviour of Multicomponent Azeotropic Distilla-

tion Processes. Inst. Chem. E. Symp. Ser. 56, 4.2/21–4.2/48.

Doherty, M.F. and J.F. Knapp (1993). Distillation, Azeotropic and Extractive. Vol. 8. fourth ed..

Kirk-Othmer Enc. Chem. Tech.

Drew, J.W. (1975). Design for Solvent Recovery. Chem. Eng. Prog. 71(2), 92–99.

Drew, J.W. (1997). Solvent Recovery. Handbook of Separation Techniques for Chemical Engi-

neers.

References

273

D¨ussel, R. and J. Stichlmair (1995). Separation of Azeotropic Mixtures by Batch Distillation

Using an Entrainer. Comp. Chem. Eng. 19(Suppl.), S113–S118.

Esbjerg, K., T.R. Andersen, D. M¨uller, W. Marquardt and S.B. Jørgensen (1998). Static Multiplic-

ities in Heterogeneous Azeotropic Distillation Sequences. DYCOPS-5, 5th IFAC Symposium
on Dynamics and Control of Process Systems, Corfu, Greece, June 8-10, 1998, 393–398.

Ewell, R.H. and L.M. Welch (1945). Rectification in Ternary Systems Containing Binary

Azeotropes. Ind. Eng. Chem. Res. 37(12), 1224–1231.

Ewell, R.H., J.M. Harrison and L. Berg (1944). Azeotropic Distillation. Ind. Eng. Chem. Res.

36(10), 871–875.

Fair, J.R. (1987). Distillation: Whither, Not Whether. I. Chem. E. Symposium Series 104, A613–

A627.

Farschman, C.A. and U. Diwekar (1996). Dual Composition in a Novel Batch Distillation Col-

umn. AIChE Annual Meeting, Chicago, November, 1996.

Fidkowski, Z.T., M.F. Doherty and M.F. Malone (1993). Feasibility of Separations for Distillation

of Nonideal Ternary Mixtures. AIChE J. 39(8), 1303–1321.

Fien, G.-J.A.F. and Y.A. Liu (1994). Heuristic Synthesis and Shortcut Design of Separation Pro-

cesses Using Residue Curve Maps: A Review. Ind. Eng. Res. 33, 2505–2522.

Flemming, H.L. and C.S. Slater (1992). Membrane Handbook. Chapman & Hall, London.

Foucher, E.R., M.F. Doherty and M.F. Malone (1991). Automatic Screening of Entrainers for

Homogeneous Azeotropic Distillation. Ind. Eng. Chem. Res. 30, 760–772.

Furlonge, H.I., C.C. Pantelides and E. Sørensen (1999). Optimal Operation of Multivessel Batch

Distillation Columns. AIChE J. 45(4), 781–801.

Furter, W.F. (1968). Salt Effect in Distillation: A Technical Review. Chem. Eng.

Furtzer, I.A. (1994). Synthesis of Entrainers in Heteroazeotropic Distillation Systems. Canadian

J. Chem. Eng. 72, 358–364.

Gmehling, J., J. Menke, J. Krafczyk and K. Fischer (1994). Azeotropic Data, Part I and Part II.

VCH-Publishers, Weinheim, New York.

Gmehling, J., U. Onken and W. Arlt (1977-1990). Vapor-Liquid Equilibrium Data Collection.

Vol. I/1-8. DECHEMA Chemsitry Data Series. Frankfurt/Main, Germany.

Gmheling, J. and C. M¨ollmann (1998). Synthesis of Distillation Processes Using Thermodynamic

Models and the Dortmund Data Bank. Ind. Eng. Chem. Res. 37(8), 3112–3123.

Goldblatt, M.E. and C.H. Gooding (1986). An Engineering Analysis of Membrane-Aided Distil-

lation. AIChE Symp. Series 82(248), 51–69.

274

References

Guerreri, G. (1992). Membrane Alcohol Separation Process - Integrated Pervaporation and Frac-

tional Distillation. Trans. IChemE. 70(A), 501–508.

Gurikov, Y.V. (1958). Structure of the Vapor-Liquid Equilibrium Diagrams of Ternary Homoge-

neous Solutions (in Russian). Russ. J. Phys. Chem. 32(9), 1980–1996. Abstract in English.

G¨uttinger, T.E. and M. Morari (1996). Multiple Steady States in Homogeneous Separation Se-

quences. Ind. Eng. Chem. Res. 35(12), 4597–4611.

G¨uttinger, T.E. and M. Morari (1997). Predicting Multiple Steady States in Distillation: Singu-

larity Analysis and Reactive Systems. PSE ’97-ESCAPE-7, 25-29 May, 1997, Trondheim,
Norway, Supplement to Computers Chem. Engng. 21, S995–S1001. See C97-2.

Haase, R. (1949). Zur Thermodynamik fl¨ussiger Dreistoffgemishe (in German). Z. Naturforsch.

4a, 342–352.

Haase, R. (1950a). Verdampfungsgleichgewichte von Mehrstoffgemishen: VI. Ternare Systeme

mit Mischungsl¨ucke (in German). Z. Naturforsch 5a(2), 109–124.

Haase, R. (1950b). Verdampfungsgleichgewichte von Mehrstoffgemishen:

VII. Ternaren

Azeotropen Punkte (in German). Z. Phys. Chemie 195, 362–385.

Haase,

R. and H. Lang (1951). Rektifikationslinien Tern¨arer Systeme (in German).

Chem. Ing. Tech. 23(13), 313–336.

Halvorsen, I.J. and S. Skogestad (1999). Use of Short-Cut Methods to Analyse Optimal Oper-

ation of Petlyuk Distillation Columns. PRES’99, 2nd Conference on Process Integration,
Modelling and Optimization for Energy Saving and Pollution Reduction, May 31 - June 2,
1999, Budapest, Hungary.

Hasebe, S., B. Abdul Aziz, I. Hashimoto and T. Watanabe (1992). Optimal Design and Opera-

tion of Complex Batch Distillation Column. Proc. IFAC Workshop on Interactions Between
Process Design and Process Control (Pergamon Press, Ed.). Oxford (UK), 177–182.

Hasebe, S., M. Noda and I. Hashimoto (1997). Optimal Operation Policy for Multi-Effect Batch

Distillaiton System. PSE’97-ESCAPE-7, 25-29 May 1997, Trondheim, Norway, Suppl. to
Comp. Chem. Eng. 21, S1221–S1226.

Hasebe, S., T. Kurooka and I. Hashimoto (1995). Comparison of the Separation Performances of

a Multi-Effect Batch Distillation System and a Continuous Distillation System. Proc. IFAC
Symp. DYCORD+’95, Elsingnore, Denmark, published by Pergamon Press, 249–254.

Hilmen, E.K., S. Skogestad, M.F. Doherty and M.F. Malone (1997). Intergrated Design, Opera-

tion and Control of Batch Extractive Distillation with a Middle Vessel. Presented at AIChE
Annual Meeting, Los Angeles, CA, November, 1997.

Hoffman, E.J. (1964). Azeotropic and Extractive Distillation. Vol. IV. Intersci. Lib. Chem. En-

gng. and Processing, John Wiley

Sons, Inc.

References

275

H¨ommerich, U. and R. Rautenbach (1998). Design and Optimization of Combined Pervapora-

tion/Distillation Processes for the Production of MTBE. J. Membrane Sci. 146, 53–64.

Horsley, L.H. (1973). Azeotropic Data - III. Advances in Chemistry Series 116, Ed. R.F. Gould,

American Chemistry Society, Washington, DC.

Humphrey and Seibert (1992). New Horizons. Chem. Eng., 86–98.

Jacobsen, E.W., L. Laroche and M. Morari (1991). Robust Control of Homogeneous Azeotropic

Distillation Columns. AIChE J. 37(12), 1810–1824.

Jaksland, C.A., G. Hytoft and R. Gani (1997). Computer-Aided Process Design and Optimization

with Novel Separation Units. Appl. Thermal Eng. 17(8-10), 973–980.

Jaksland, C.A., R. Gani and K.M. Lien (1995). Separation Process Design and Synthesis Based

on Thermodynamic Insights. Chem. Eng. Sci. 50(3), 511–530.

Julka, V. and M.F. Doherty (1993). Geometric Nonlinear Analysis of Multicomponent Nonideal

Distillation: A Simple Computer-Aided Design Procedure. Chem. Eng. Sci., 1367–1391.

King, C.J. (1980). Separation Processes. 2nd ed.. Mc-Graw Hill.

Kiva, V.N. and B.M. Alukhanova (1997). Multiple Steady States of Distillation and Its Realiza-

tion. PSE ’97-ESCAPE-7, 25-29 May, 1997, Trondheim, Norway, Supplement to Computers
Chem. Engng. 21, S541–S546.

Kiva, V.N. and L.A. Serafimov (1973). Rules that Govern the Composition Trajectories of Simple

Distillation of Ternary Mixtures [Non-local Rules of the Movement of Process Lines for
Simple Distillation in Ternary Systems]. Russ. J. Phys. Chem. 47(3), 638–642.

Kiva, V.N. and L.A. Serafimov (1973c). Batch Distillation of Ternary Mixtures at Constant Reflux

Ratio, II. Analysis of the Dynamic System of the Distillate and the Splitting into Regions of
Distillation. Russ. J. Phys. Chem. 47(3), 634–637.

Kiva, V.N. and L.A. Serafimov (1975). Structure of Phase Vapor-Liquid Diagrams of Ternary

Mixtures and Rectification Boundaries. CHISA’75, Prague, 1975, Lecture Summaries H-
2, 26.

Kiva, V.N. and L.A. Serafimov (1976a). Localization of Boundaries for Batch Distillation of

Ternary Mixtures. Russ. J. Phys. Chem. 50(10), 2481–2483.

Kiva, V.N. and L.A. Serafimov (1976b). Regions of Continuous Distillation of Ternary Mix-

tures and Location of Their Boundaries I. Conditions of Total Reflux. Russ. J. Phys. Chem.
50(4), 526–528 [882–885].

Kiva, V.N., I.M. Marchenco and Yu.N. Garber (1993). Possible Compositions of the Products of

Ternary–Mixture Rectification with a Binary Saddle. Theor. Found. Chem. Eng. 27(4), 336–
342.

276

References

Kiva, V.N., V.V. Starostin, V.A. Cherepakhina and V.S. Timofeev (1988). Separation of Mixtures

by Rectification Redistribution. Soviet Chemical Industry 20(9), 45–47.

Kjensgjord, E. (1998). Modelling and Optimisation of Batch Extractive Distillation. Norwegian

University of Science and Technology: Diploma work.

Knapp, J.P. and M.F. Doherty (1990). Thermal Integration of Homogeneous Azeotropic Distilla-

tion Sequences. AIChE J. 36(7), 969–984.

Knapp, J.P. and M.F. Doherty (1994). Minimum Entrainer Flow for Extractive Distillation: a

Bifurcation Theoretic Approach. AIChE J. 40(2), 243–268.

Knauf, R., U. Meyer-Blumenroth and J. Semel (1998). Membrane Processes in the Chemical

Industry [in German]. Chem. Ing. Tech. 70, 1265–1270.

Kogan, L.V. and V.V. Kafarov (1977a). Determination of the Boundaries of Rectification Regions.

Communication I.. Russ. J. Appl. Chem. 50(2), 302–305 [323–326].

Kogan, L.V. and V.V. Kafarov (1977b). Determination of the Boundaries of Rectification Regions.

Communication II.. Russ. J. Appl. Chem. 50(2), 306–309 [326–329].

Kogan, V.B. (1971). Azeotropic and Extractive Distillation (in Russian). Vol. 2nd edition. Chem-

istry Publishing Co., Leningrad.

K¨ohler, J., H. Haverkamp and N. Schadler (1995). Discontinuous Rectification of Azeotropic

Mixtures with Use of Auxiliary Substances (in German). Chem. Ing. Tech. 67(8), 967–971.

Komarova, L.F., L.A. Serafimov and Yu.N. Garber (1974). Classification of Ternary Mixtures

with Biazeotropic Constituents. Russ. J. Phys. Chem. 48(6), 1391–1396 [817–818].

Kunesh, J.G., H.Z. Kister, M.J. Lockett and J.R. Fair (1995). Distillation: Still Towering Over

Other Options. Chem. Eng. Prog. 18(Oct.), 43–54.

Kushner, T.M., G.V. Shul’ga and L.A. Serafimov (1992). Evolution of VLE diagrams of Bi-

azeotropic Binary Mixtures with Variation of Pressure. In Paper Collection ”Problems of
thermodynamics of heterogeneous systems and the theory of interfacial effects”, Issue 9,
St.Petersburg University, St.Petersburg, 64–81.

Lang, P., H. Yatim, P. Moszkowicz and M. Otterbein (1994). Batch Extractive Distillation Under

Constant Reflux Ratio. Comp. Chem. Eng. 18(11/12), 1057–1069.

Lang, P., Z. Lelkes, P. Moszkowicz, M. Otterbein and H. Yatim (1995). Different Operational

Policies for the Batch Extractive Distillation. Comp. Chem. Engng. 19(Suppl.), S645–S650.

Laroche, L., N. Bekiaris, H.W. Andersen and M. Morari (1992a). Homogeneous Azeotropic Dis-

tillation: Separability and Flowsheet Synthesis. Ind. Eng. Chem. Res. 31(9), 2190–2209.

Laroche, L., N. Bekiaris, H.W. Andersen and M. Morari (1992b). The Curious Behavior of

Homogeneous Azeotropic Distillation - Implications for Entrainer Selection. AIChE J.
38(9), 1309–1328.

References

277

Laroche, L., N. Bekiaris, H.W. Andersen and M. Morari (1993). Homogeneous Azeotropic Dis-

tillation: Comparing Entrainers. Can. J. Chem. Eng. 69(Dec.), 1302–1319.

Lecat, M. (1949). Tables Az´eotropiques, Az´eotropes Binaries Orthobares. Brussels.

Lelkes, Z., P. Lang and M. Otterbein (1998b). Feasibility and Sequencing Studies for Ho-

moazeotropic Distillation in a Batch Rectifier with Continuous Entrainer Feeding. Com-
put. Chem. Eng. 22(Suppl.), S653–S656.

Lelkes, Z., P. Lang, B. Benadda and P. Moszkowicz (1998a). Feasibility of Extractive Distillation

in a Batch Rectifier. AIChE J. 44(4), 810–822.

Lelkes, Z., P. Lang, P. Moszkowicz, B. Benadda and M. Otterbein (1998c). Batch Extractive

Distillation: The Process and the Operational Policies. Chem. Eng. Sci. 53(7), 1331–1348.

Lestak, F. and C. Collins (1997). Advanced Distillation Saves. Chem. Engng. July, 72–76.

Levy, S.G. and M.F. Doherty (1986). Design and Synthesis of Homogeneous Azeotropic Distilla-

tions. 4. Minimum Reflux Calculations for Multiple-Feed Columns. Ind. Eng. Chem. Fun-
dam. 25, 269–279.

Levy, S.G., D.B. Van Dongen and M.F. Doherty (1985). Design and Synthesis of Homogeneous

Azeotropic Distillations. 2. Minimum Reflux Calculations for Nonideal and Azeotropic
Columns. Ind. Eng. Chem. Fundam. 24(4), 463–474.

Li, W. and X. Xu (1998). Separation of Acetone/Water by Modified

-Alumina Membrane via a

New Method. J. Membrane Sci. 149, 21–27.

Malenko, Yu.I. (1970). Physicochemical Analysis of Distillation Diagrams. II. Quaternary Mix-

tures. Russ. J. Phys. Chem. 44(7), 916–918 [1631–1636].

Malesinkski, W. (1965). Azeotropy and Other Theoretical Problems of Vapor-Liquid Equilibrium.

Interscience, New York.

Marchenko, I.M., V.N. Kiva and Yu.N. Garber (1992). Technological Aspects of Tie-Line Trans-

ferability in the Distillation of Class 1.0-2 Systems. Russ. J. Appl. Chem. 65(2), 336–340
[279–283].

Matsuyama, H. and H. Nishimura (1977). Topological and Thermodynamic Classification of

Ternary Vapor-Liquid Equilibria. J. Chem. Eng. Jpn. 10(3), 181–187.

Meski, G.A. and M. Morari (1995). Design and Operation of a Batch Distillation with a Middle

Vessel. Comp. Chem. Eng. 19(Suppl.), S597–S602.

Meski, G.A., M. Han, N. Bekiaris and M. Morari (1998). Optimality for Batch Distillation Con-

figurations. IFAC Symposium DYCOPS-5, Corfu, Greece, 8-19 June 1998, 387–392.

Milani, S.M. (1999). Optimization of Solvent Feed Rate for Maximum Recovery of High Purity

Top Product in Batch Extractive Distillation. Trans. IChemE. 77(Part A), 469–470.

278

References

Mujtaba, I.M. (1999). Optimization of Batch Extractive Distillation Processes for Separating

Close Boiling and Azeotropic Mixtures. Trans. IChemE. 77(Part A), 588–596.

Mujtaba, I.M. and S. Macchietto (1998). Holdup Issues in Batch Distillation - Binary Mixtures.

Chem. Engng. Sci. 53(14), 2519–2530.

M¨uller, E. (1988). Liquid-Liquid Extraction. Vol. B3. Ullmann’s Enc. Ind. Chem.

Nikolaev, E.S., V.N. Kiva, A.S. Mozzhukhin and L.A. Serafimov (1979). Use of Functional Oper-

ators to Determine Continuous Distillation Regions. Theor. Found. Chem. Eng. 13(4), 493–
498.

Ognisty,

T.P.

(1995).

Analyze

Distillation

Columns

With

Thermodynamics.

Chem. Eng. Prog., 40–46.

Ostwald, W. (1902). Lehrbuch der Allgemeinen Chemie. Verwandtschaftslehre. Erste Teil (In Ger-

man). Verlag von Wilhelm Engelmann, Leipzig, Germany.

Pelkonen, S., R. Kaesemann and A. G´orak (1997). Distillation Lines for Multicomponent Sepa-

ration in Packed Columns: Theory and Comparison with Experiment. Ind. Eng. Chem. Res.
36, 5392–5398.

Perry, R.H. (1997). Perry’s Chemical Engineers’ Handbook. Chem. Engng. Series. 7th ed.. Green,

D.W. and Maloney, J.O. (Eds.), McGraw Hill, Inc., New York.

Perry, R.H. and C.H. Chilton (1963). Chemical Engineers’ Handbook. Chem. Engng. Series. 4th

ed.. McGraw Hill, Inc., New York.

Perry, R.H. and C.H. Chilton (1973). Chemical Engineers’ Handbook. Chem. Engng. Series. 5th

ed.. McGraw Hill, Inc., New York.

Peterson, E.J. and L.R. Partin (1997). Temperature Sequences for Categorizing All Ternary Dis-

tillation Boundary Maps. Ind. Eng. Chem. Res. 36(5), 1799–1811.

Petlyuk, F.B. (1978). Rectification of Zeotropic, Azeotropic and Continuous Mixtures in Simple

and Complex Infinite Columns with Finite Reflux. Theor. Found. Chem. Eng. 12, 671–678.

Petlyuk, F.B. and L.A. Serafimov (1983). Multicomponent Distillation. Theory and Design (in

Russian) [Mnogokomponentnaya Rektifikatsiya: Teriyo I. Raschet]. Chemistry Publishing
Co, Moscow.

Petlyuk, F.B. and R.Yu. Danilov (2000). Synthesis of Separation Flowsheets for Multicomponent

Azeotropic Mixtures on the Basis of the Distillation Theory. Presynthesis: Prediction of
Feasible Product Compositions. Theor. Found. Chem. Eng. 34(3), 265–285.

Petlyuk, F.B. and V.S. Avet’yan (1971). Investigation of the Rectification of Three-Component

Mixtures with Infinite Reflux. Theor. Found. Chem. Eng. 5(4), 499–507.

References

279

Petlyuk, F.B., V.Ya. Kievski and L.A. Serafimov (1975a). Thermodynamic and Topologic Anal-

ysis of the Phase Diagram of Polyazeotropic Mixtures. I. Algorithm for Construction of
Structural Graphs for Azeotropic Ternary Mixtures. Russ. J. Phys. Chem. 69(12), 3105–
3108.

Petlyuk, F.B., V.Ya. Kievski and L.A. Serafimov (1975b). Thermodynamic and Topologic

Analysis of the Phase Diagram of Polyazeotropic Mixtures. I. Definition of Distilla-
tion Regions Using a Computer [Computer-Aided Determination of Distillation Regions].
Russ. J. Phys. Chem. 69(12), 3102–3104.

Petlyuk, F.B., V.Ya. Kievski and L.A. Serafimov (1977a). Thermodynamic-Topological Analysis

of VLE Diagrams for Multicomponent Mixtures. III. Algorithm of Structure Graph Drawing
for Four-Component Mixtures. Russ. J. Phys. Chem. 71(2), 315–318.

Petlyuk, F.B., V.Ya. Kievski and L.A. Serafimov (1978). Thermodynamic-Topological Analysis of

VLE Diagrams for Multicomponent Mixtures. IV. Use of the Structure Matrix for Analysis
of Feasible Separations of the Wood Pyrolysis Product. Russ. J. Phys. Chem. 72(5), 1145–
1148.

Petlyuk, F.B., V.Ya. Kievskii and L.A. Serafimov (1977b). A Combined Thermodynamic and

Topological Analysis of Phase Equilibrium Diagrams for Polyazeotropic Systems - V. The
Use of the Phase Equilibrium Model in Combined Thermodynamic and Topological Analy-
sis. Russ. J. Phys. Chem. 51, 338–340.

Petlyuk, F.B., V.Ya.Kievski and L.A. Serafimov (1977c). Method for the Isolation of the

Regions of Rectification of Polyazeotropic Mixtures Using an Electronic Computer
[Computer-Aided Determination of Distillation Regions for Multicomponent Mixtures].
Theor. Found. Chem. Eng. 11(1), 3–10.

Pettersen, T. and K. Lien (1995). Design of Hybrid Distillation and Vapor Permeation Processes.

J. Membrane Sci. 99, 21–03.

Pham, H.N. and M.F. Doherty (1990a). Design and Synthesis of Heterogeneous Azeotrope

Distillation-I. Heterogeneous Phase Diagrams. Chem. Eng. Sci. 45(7), 1823–1836.

Pham, H.N. and M.F. Doherty (1990b). Design and Synthesis of Heterogeneous Azeotrope

Distillation-II. Residue Curve Maps. Chem. Eng. Sci. 45(7), 1837–1844.

Pham, H.N. and M.F. Doherty (1990c). Design and Synthesis of Heterogeneous Azeotrope

Distillation-III. Column Sequences. Chem. Eng. Sci, 45(7), 1845–1854.

Phimister, J.R. and W.D. Seider (2000a). Distillate-Bottoms Control of Middle-Vessel Distillation

Columns. Ind. Eng. Chem. Res. 39, 1840–1849.

Phimister, J.R. and W.D. Seider (2000b). Semicontinuous, Middle-Vessel, Extractive Distillation.

Comp. Chem. Engng. 24, 879–885.

280

References

P¨ollmann, P. and E. Blass (1994). Best Products of Homogeneous Azeotropic Distillations. Gas

Sep. Purif. 8(4), 194–228.

P¨ollmann, P., M.H. Bauer and E. Blass (1996). Investigation of Vapour-Liquid Equilibrium of

Non-Ideal Multicomponent Systems. Gas Sep. Purif. 10(4), 225–241.

Pratt, H.R.C. (1967). Countercurrent Separation Processes. Elsevier Publishing Company. Ams-

terdam.

Pressly, T.G. and K.M. Ng (1998). A Break-Even Analysis of Distillation-Membrane Hybrids.

AIChE. J. 44(1), 93–105.

Pretel, E.J., P.A. Lopez, S.B. Bottini and E.A. Brignole (1994). Computer-Aided Molecular De-

sign of Solvent for Separation Processes. AIChE. J. 40(8), 1349–1360.

Prigogine, I. and R. Defay (1954). Chemical Thermodynamics. Longmans & Green, London.

Rautenbach, R. and J. Vier (1995). Design and Analysis of Combined Distillation/Pervaporation

Processes. Proc. Seventh International Conference on Pervaporation Processes in the Chem-
ical Industry, Reno (NV), 70–85.

Rautenbach, R. and R. Albrecht (1989). Membrane Processes. John Wiley, New York.

Rautenbach, R., S. Klatt and J. Vier (1992). State of the Art of Pervaporation - 10 Years of Indus-

trial PV. Proc. 6th International Conference on Pervaporation Processes in the Chemical
Industry, Ottawa, 2–15.

Reinders, W. and C.H. de Minjer (1940). Vapour-Liquid Equilibria in Ternary Systems I. The

Course of the Distillation Lines. Rec. Trav. Chem. 59, 207–230.

Reinders, W. and C.H. DeMinjer (1940a). Vapour-Liquid Equilibria in Ternary Systems. II. The

System Acetone-Chloroform-Benzene. Rec. Trav. Chem. 59, 369–391.

Reinders, W. and C.H. DeMinjer (1940b). Vapour-Liquid Equilibria in Ternary Systems.

III. The Course of the Distillation Lines in the System Acetone-Chloroform-Benzene.
Rec. Trav. Chem. 59, 392–406.

Reinders, W. and C.H. DeMinjer (1947a). Vapour-Liquid Equilibria in Ternary Systems. IV. The

System Water-Acetone-Thrichloroethene. Rec. Trav. Chem. 66(9/10), 552–563.

Reinders, W. and C.H. DeMinjer (1947b). Vapour-Liquid Equilibria in Ternary Systems. V. The

System Water-Formic Acid-Metaxylene. Rec. Trav. Chem. 66(9/10), 564–572.

Reinders, W. and C.H. DeMinjer (1947c). Vapour-Liquid Equilibria in Ternary Systems. VI. The

System Water-Acetone-Chloroform. Rec. Trav. Chem. 66(9/10), 573–604.

Reshetov, S.A. (1998). Private communications. Karpov Institute of Physical Chemistry, Moscow.

References

281

Reshetov, S.A., I.B. Zhvanetskij and E.V. Orlova (1999). Modeling of Continuous Distil-

lation of Ternary Mixtures with Phase Diagrams Containing One-Side Unity

-Lines.

Theor. Found. Chem. Eng. 33(2), 149–155.

Reshetov, S.A., V.Yu. Sluchenkov, V.S. Ryzhova and I.B. Zhvanetskij (1990). Diagrams of

K-Ordering Regions with an Arbitrary Number of Unitary

-Lines [Univolatility Lines].

Russ. J. Phys. Chem. 64(9), 1344–1347 [2498–2503].

Rev, E. (1992). Crossing of Valleys, Ridges, and Simple Boundaries by Distillation in Homoge-

neous Ternary Mixtures. Ind. Eng. Chem. Res. 31(3), 893–901.

Robinson, C.S. and E.R. Gilliland (1950). Elements of Fractional Distillation. Vol. 4. McGraw

Hill, Inc., New York.

Robinson,

E.R. (1970). Optimal control of an industrial batch distillation column.

Chem. Eng. Sci., 921–928.

Rooks, R.E., V. Julka, M.F. Doherty and M.F. Malone (1998). Structure of Distillation Regions

for Multicomponent Azeotropic Mixtures. AIChE Journal 44(6), 1382–1391.

Rousseau, R.W. (1987). Handbook of Separation Process Technology. Book; Wiley, New York.

Safrit, B.T. (1996). Synthesis of Azeotropic Batch Distillation Separation Systems. PhD thesis.

Carnegie Mellon University, Pittsburgh, USA.

Safrit, B.T. and A.W. Westerberg (1997a). Algorithm for Generating the Distillation Regions for

Azeotropic Multicomponent Mixtures. Ind. Eng. Chem. Res. 36, 1827–1840.

Safrit, B.T. and A.W. Westerberg (1997b). Improved Operational Policies for Batch Extractive

Distillation Columns. Ind. Eng. Chem. Res. 36, 436–443.

Safrit, B.T. and A.W. Westerberg (1997c). Synthesis of Azeotropic Batch Distillation Systems.

Ind. Eng. Chem. Res. 36, 1841–1854.

Safrit, B.T., A.W. Westerberg, U. Diwekar and O.M. Wahnschafft (1995). Extending Contin-

uous Conventional and Extractive Distillation Feasibility Insights to Batch Distillation.
Ind. Eng. Chem. Res. 34, 3257–3264.

Schreinemakers, F.A.H. (1901a). Dampfdrucke Tern¨arer Gemische. Theoretischer Teil: Dritte

Abhandlung (In German). Z. Phys. Chem. 36(6), 710–740.

Schreinemakers, F.A.H. (1901b). Dampfdrucke tern¨arer Gemische. Theoretischer Teil: Erste Ab-

handlung (In German). Z. Phys. Chem. 36(3), 257–289.

Schreinemakers, F.A.H. (1901c). Dampfdrucke Tern¨arer Gemische. Theoretischer Teil: Zweite

Abhandlung (In German). Z. Phys. Chem. 36(4), 413–449.

Schreinemakers, F.A.H. (1902). Einige Bemerkungen ¨uber Dampfdrucke Tern¨arer Gemische (In

German). Z. Phys. Chem. 43, 671–685.

282

References

Schweitzer, P.A. (1979). Handbook of Separation Techniques for Chemical Engineers. Book;

McGraw-Hill. Section 1.3.

Schweitzer, P.A. (1997). Handbook of Separation Techniques for Chemical Engineers. Book;

McGraw-Hill.

Seader, J.D. and E.J. Henley (1998). Separation Process Principles. John Wiley & Sons, Inc.,

New York.

Serafimov, L.A. (1968a). Separation Technology of Azeotropic Mixtures (in Russian). Book:

Azeotropy and Polyazeotropy, Chapter XX1 (additional chapter in Russian edition), W.

Swie¸tosławski (Ed.), Chemistry Publishing Co., Moscow, 186–224.

Serafimov, L.A. (1968b). The Azeotropic Rule and the Classification of Multicomponent Mix-

tures. II. The Form [Behavior] of Distillation Lines [Residue Curves] Near Four-Component
Singular Points. Russ. J. Phys. Chem. 42(1), 130–131 [248–252].

Serafimov, L.A. (1968c). The Azeotropic Rule and the Classification of Multicomponent Mix-

tures. III. Distribution of Singular Points in the Phase Diagram for Liquid-Vapor Equilib-
rium in Four-Component Mixtures. Russ. J. Phys. Chem. 42(1), 132–135 [252–256].

Serafimov, L.A. (1968d). Theoretical Principles of Distillation Sequences Design and Synthesis

for Nonideal Multicomponent Mixtures (in Russian). PhD thesis. Lomonosov Institute of
Fine Chemical Technology, Moscow.

Serafimov, L.A. (1969a). The Azeotropic Rule and the Classification of Multicomponent Mix-

tures. IV. Pricipal Equations for the Calculation of Liquid-Vapor Phase Equilibrium Dia-
grams for Four-Component Mixtures. Russ. J. Phys. Chem. 43(3), 621–624.

Serafimov, L.A. (1969b). The Azeotropic Rule and the Classification of Multicomponent Mix-

tures V. Analysis of Liquid-Vapor Phase Equilibrium Diagrams for Quaternary Mixtures.
Russ. J. Phys. Chem. 43(5), 749–751 [1343–1346].

Serafimov, L.A. (1969c). The Azeotropic Rule and the Classification of Multicomponent Mixtures

VI. n-Component Mixtures. Russ. J. Phys. Chem. 43(7), 981–983.

Serafimov, L.A. (1970a). General Principles of the Course of Unidistribution Lines in Vapor-

Liquid Equilibrium Diagrams of Ternary Mixtures. Physical-Chemical Foundations of Rec-
tification”, Collection of papers by Moscow Lomonosov Institute of Fine Chemical Technol-
ogy, MITChT, Moscow, 20–30.

Serafimov, L.A. (1970b). The Azeotropic Rule and the Classification of Multicomponent Mix-

tures VII. Diagrams for Ternary Mixtures. Russ. J. Phys. Chem. 44(4), 567–571 [1021–
1027].

Serafimov, L.A. (1971a). The Azeotropic Rule and the Classification of Multicomponent Mix-

tures. IX. Tangential Azeotropy and Correlation Between the General Relation Between
Singular Points of Various Types. Russ. J. Phys. Chem. 45(6), 831–833 [1473–1476].

References

283

Serafimov, L.A. (1971b). The Azeotropic Rule and the Classification of Multicomponent

Mixtures. VIII. The General Relations of Tangential Azeotropy. Russ. J. Phys. Chem.
45(5), 638–642 [1140–1147].

Serafimov, L.A. (1971c). The Azeotropic Rule and the Classification of Multicomponent Mix-

tures. X. Doubly Tangential Azeotropes. Russ. J. Phys. Chem. 45(6), 918–921 [1620–1625].

Serafimov, L.A. (1971d). The Azeotropic Rule and the Classification of Multicomponent Mix-

tures. XI. Tangential Azeotropy for Three-Component Systems and Chains of Topological
Structures. Russ. J. Phys. Chem. 45(10), 1388–1389 [2448–2450].

Serafimov, L.A. (1987). Thermodynamic Topological Analysis and Separation of Multicompo-

nent Polyazeotropic Mixtures. Theor. Found. Chem. Eng. 21(1), 44–54.

Serafimov, L.A. (1996). Thermodynamic and Topological Analysis of Liquid-Vapor Phase Equi-

librium Diagrams and Problems of Rectification of Multicomponent Mixtures. Book: Math-
ematical Methods in Contemporary Chemistry, S.I. Kuchanov (Ed.) Gordon and Breach
Publishers, Amsterdam, 557–605.

Serafimov, L.A. and A.K. Frolkova (1997). Fundamental Principle of Concentration-Field Redis-

tribution Between Separation Regions as a Basis for the Design of Technological Systems.
Theor. Found. Chem. Eng. 31(2), 159–166.

Serafimov, L.A., L.F. Komoarova and Yu.N. Garber (1974). Classification of VLE Diagrams for

Ternary Mixtures with Biazeotropic Constituents. Russ. J. Phys. Chem. 48(6), 1391.

Serafimov, L.A., T.M. Kushner and T.V. Cheluskina (1996). Thermodynamic-Topological Analy-

sis of VLE Diagrams for Ternary Mixtures with Two Ternary Azeotropes. In Papers Collec-
tion: Problems of Thermodynamics of Heterogeneous Systems and the Theory of Interfacial
Effects”, Issue 10, St.Petersburg University, St.Petersburg, 26–55.

Serafimov, L.A., V.S. Timofeyev and M.I. Balashov (1973). Rectification of Multicomponent

Mixtures II. Local and General Characteristics of the Trajectories of Rectification Processes
at Infinite Reflux Ratio. Acta Chemica Acad. Sci. Hung. 75(2), 193–211.

Serafimov, L.A., V.T. Zharov and V.S. Timofeyev (1971). Rectification of Multicomponent Mix-

tures I. Topological Analysis of Liquid-Vapor Phase Equilibrium Diagrams. Acta Chemica
Acad. Sci. Hung. 69(4), 383–396.

Serafimov, L.A., Yu.E. Gol’berg, T.A. Vitman and V.N. Kiva (1972). Properties of Univolatil-

ity Sets and Its Location in the Concentration Space (in Russian). In book: ”Chemistry.
Collection of the Scientific Works of Ivanovo Energetic Institute”, Ivanovo-Vladimir,Issue
14, 166–179.

Skogestad, S. (1992). Dynamics and Control of Distillation Columns – A Critical Survey. IFAC

Symposium Dycord+ ’92, Maryland, Apr. 27-29, 1992.

284

References

Skogestad, S., B. Wittgens, R. Litto and E. Sørensen (1997). Multivessel Batch Distillation.

AIChE J. 43(4), 971–978.

Sluchenkov, V.Yu., S.A. Reshetov and I.B. Zhvanetskij (1990). Unit

-Lines in Three-Component

Systems with a Ternary Azeotrope [Univolatility Lines in Ternary Mixtures with Ternary
Azeotropes]. Russ. J. Phys. Chem. 64(5), 737–739 [1381–1384].

Smith, R. (1995). Chemical Process Design. McGraw-Hill, Inc.

Sobolev, D.M., I.L. Shul’gin, V.A. Lovchikov, Yu.I. Malenko and P.G. Romankov (1980). Bound-

aries of Distillation Regions. Russ. J. Phys. Chem. 53(11), 2566–2567.

Sørensen, E. (1994). Studies on Optimal Operation and Control of Batch Distillation Columns.

1994:62. Norwegian Institute of Technology.

Sørensen, E. (1997). Alternative Ways of Operating a Batch Distillation Column. Distillation

and Absorption ’97, IChemE Symposium, Maastricht, The Netherlands, 8-10 September,
1997, 643–652.

Sørensen, E. (1999). A Cyclic Operating Policy for Batch Distillation - Theory and Practice.

Comp. Chem. Eng. 23(4-5), 533–542.

Sørensen, E. and M. Prenzler (1997). A Cyclic Operating Policy for Batch Distillation - Theory

and Practice. PSE ’97-ESCAPE-7, 25-29 May, 1997, Trondheim, Norway, Supplement to
Computers Chem. Engng. 21, S1215–S1220.

Sørensen, E. and S. Skogestad (1994). Optimal Operating Policies of Batch Distillation with

Emphasis on the Cyclic Operating Policy. PSE’94, 5th International Symposium on Process
Systems Engineering, 30 May - 3 June 1994, 449–456.

Sørensen, E. and S. Skogestad (1996). Comparison of Inverted and Regular Batch Distillation.

Chem. Eng. Sci. 51(22), 4949–4962.

SpeedUp (1993). SPEEDUP User Manual. Aspen Technologies, Prosys Technology Ltd.

Stichlmair, J. (1988). Distillation and Rectification. Vol. B3. Ullmann’s Enc. Ind. Chem.

Stichlmair, J. and J.R. Fair (1998). Distillation: Principles and Practices. Wiley, New York.

Stichlmair, J.G., J.-R. Fair and J.L. Bravo (1989). Separation of Azeotropic Mixtures via En-

hanced Distillation. Chem. Eng. Prog. 85(1), 63–69.

Storonkin, A.V. (1967). Thermodynamics of Heterogeneous Systems (in Russian). Leningrad

University Publishing Dept, Leningrad.

Storonkin, A.V. and N.A. Smirnova (1963). The Thermodynamics of Multicomponent Het-

erogeneous Systems. VI. Distillation Curves in the Miscibility Gap of Ternary Systems.
Russ. J. Phys. Chem. 37(3), 601–607.

References

285

Sulzer (1998). Sultzer Technical Review. Webpages.

Swietoslawski, W. (1963). Azeotropy and Polyazeotropy. Oxford Pergamon Press.

Tassios, D.P. (1972). Preface. Extractive and Azeotropic Distillation. Advances in Chemistry Se-

ries 115. American Chemistry Society, Washington DC.

Tester, J.W. and M. Modell (1997). Thermodynamics and Its Applications. Third Ed., Prentice

Hall, NJ.

Thong, D.Y.-C., F.J.L. Castillo and G.P. Towler (2000). Distillation Design and Retrofit Using

Stage-Composition Lines. Chem. Eng. Sci. 55, 625–640.

Timofeev, V.S, L.A. Serafimov and V.V. Beregovyh (1970). Properties of the Vapor-Liquid Equi-

librium Diagrams of Heterogeneous Mixtures (in Russian). ”Physiochemical Foundations
of Rectification”, Collection of papers by Moscow Lomonosov Institute of Fine Chemical
Technology, MITChT, Moscow, 30–39.

Treybal, R.E. (1963). Liquid-Liquid Extraction. McGraw-Hill, New York.

Treybal, R.E. (1970). A Simple Method of Batch Distillation. Chem. Eng., 95–98.

Van Dongen, D.B. and M.F. Doherty (1984). On the Dynamics of Distillation Processes V. The

Topology of the Boiling Temperature Surface and Its Relation to Azeotropic Distillation.
Chem. Eng. Sci. 39(5), 883–892.

Van Dongen, D.B. and M.F. Doherty (1985a). Design and Synthesis of Homogenous Azeotropic

Distillations 1. Problem Formulation for a Single Column. Ind. Eng. Chem. Fund.
24(4), 454–463.

Van Dongen, D.B. and M.F. Doherty (1985b). On the Dynamics of Distillation Processes VI.

Batch Distillation. Chem. Eng. Sci. 40(11), 2087–2093.

Van Winkle, M. (1967). Distillation. Chem. Eng. Series. McGraw Hill, Inc.

Vitman, T.A. and V.T. Zharov (1971a). Vapor-Liquid Equilibrium in the System Cyclohexane-

Acetone-Isopropyl Alcohol-Benzene-Toluene. Russ. J. Phys. Chem. 45(1), 79–80 [147–
149].

Vitman, T.A. and V.T. Zharov (1971b). Vapor-Liquid Equilibrium in the System Cyclohexane-

Acetone-Isopropyl Alcohol-Toluene. Russ. J. Phys. Chem. 45(1), 78–79 [145–147].

Vogelpohl, A. (1999). On the Relation between Ideal and Real Systems in Ternary Distillation.

Trans. IChemE 77(9), 487–492.

Wahnschafft, O.M. (1997). Advanced Distillation Synthesis Techniques for Nonideal Mixtures

are Making Headway in Industrial Applications. Distillation and Absorption ’97, IChemE
Symposium, Maastricht, The Netherlands, 8-10 September, 1997, 613–623.

286

References

Wahnschafft, O.M. and A.W. Westerberg (1993). The Product Composition Regions of

Azeotropic Distillation Columns, 2. Separability in Two-Feed Column and Entrainer Se-
lection. Ind. Eng. Chem. Res. 32, 1108–1120.

Wahnschafft, O.M., J.-P. le Rudulier and A.W. Westerberg (1993). A Problem Decompo-

sition Approach for the Synthesis of Complex Separation Processes with Recycles.
Ind. Eng. Chem. Res. 32(6), 1121–1141.

Wahnschafft, O.M., J.W. Koehler, E. Blass and A.W. Westerberg (1992). The Product Com-

position Regions of Single-Feed Azeotropic Distillation Columns. Ind. Eng. Chem.
31(10), 2345–2362.

Walas, S.M. (1985). Phase Equilibria in Chemical Engineering. Butterworth Publishers.

Warter, M. and J. Stichlmair (1999). Batchwise Extractive Distillation in a Column with a Middle

Vessel. Comp. Chem. Eng. Suppl., S915–S918.

Warter, M. and J. Stichlmair (2000). Batch Distillation of Azeotropic Mixtures in a Column

with a Middle Vessel. European Symposium on Computer Aided Process Engineering - 10,
ESCAPE-10, Florence, Italy, May 7-10 2000, 691–696.

Warter, M., R. D

ssel and J. Stichlmair (1997). To the Separation of Azeotropic Mixtures by

Batchwise Extractive Distillation. Presented at Dist.& Abs.’97, Maastricht, The Netherlands,
8-10 September, 1997.

Wasylkiewicz, S.K., L.C. Kobylka and F.J.L. Castillo (1999a). Design and Synthesis of Distilla-

tion Sequences for the Separation of Multicomponent Azeotropic Mixtures. III Sympozjum
Destylacja, Absorpcja i Ekstrakcja Szklarska Poreba, 13-15 wrzesien 1999, 83–91.

Wasylkiewicz, S.K., L.C. Kobylka and M.A. Satyro (1999b). Designing Azeotropic Distillation

Columns. Chem. Eng., 80–85.

Westerberg, A.W. (1997). Separating Azeotropic Mixtures. Systematic Methods of Chemical

Process Design, L.T. Biegler, I.E. Grossmann and A.W. Westerberg (Eds.), Prentice Hall,
NJ, 455–494.

Westerberg, A.W. and O.M. Wahnschafft (1996). Synthesis of Distillation-Based Separation Sys-

tems. Advances in Chemical Engineering, J.L. Anderson (Ed.), New York: Academic Press
23, 63–170.

Widagdo, S. and W.D. Seider (1996). Journal Review: Azeotropic Distillation. AIChE J.

42(1), 96–128.

Widagdo, S., W.D. Seider and D.H. Sebastian (1992). Dynamic Analysis of Heterogeneous

Azeotropic Distillation. AIChE J. 38(8), 1229–1242.

Wijesinghe, A.M.J.C. (1985). Development of Industrial Complexes of Special Rectification

Techniques for Solvent Recovery (in Russian). PhD thesis. M.V. Lomonosov Institute of
Fine Chemical Engineering, Moscow, USSR.

References

287

Wilson, R.Q., W.H. Mink, H.P. Munger and J.W. Clegg (1955). Dehydration of Hydrazine by

Azeotropic Distillation. AIChE J. 1(2), 220–224.

Wittgens, B. and S. Skogestad (1997). Multivessel Batch Distillation - Experimental Verifica-

tion. Proc. Distillation & Absorption’97, Maastricht, September, IChemE Symposium Series
142, 239–248.

Wittgens, B. and S. Skogestad (1998). Multivessel Batch Distillation - Potential Energy Savings.

IFAC Symposium DYCOPS-5, Corfu, Greece, 8-19 June 1998, 515–520.

Wittgens, B., R. Litto, E. Sørensen and S. Skogestad (1996). Total Reflux Operation of Multives-

sel Batch Distillation. Comp. Chem. Engng. S20, S1041–S1046.

Yamakita, Y., J. Shiozaki and H. Matsuyama (1983). Consistency Test of Ternary Azeotropic

Data by Use of Simple Distillation. J. Chem. Eng. Jpn. 16(2), 145–146.

Yatim, H., P. Moszkowicz and P. Otterbein, M.and Lang (1993). Dynamic Simulation of a Batch

Extractive Distillation Process. Comp. Chem. Eng. 17(Suppl.), S57–S62.

Zharov,

V.T. (1967). Free Evaporation of Homogeneous Multicomponent Solutions.

Russ. J. Phys. Chem. 41(11), 1539–1543.

Zharov, V.T. (1968a). Evaporation of Homogeneous Multicomponent Solutions III. Behavior of

Distillation Lines [Residue Curves] Near Singular Points. Russ. J. Appl. Chem. 42(2), 195–
198 [366–372].

Zharov, V.T. (1968b). Free Evaporation of Homogeneous Multicomponent Solutions II. Four-

Component Systems. Russ. J. Appl. Chem. 42(1), 58–61 [116–122].

Zharov, V.T. (1968c). Phase Transformations [Distillation Lines] and Rectification of Multicom-

ponent Solutions. Russ. J. Appl. Chem. 41(12), 2530–2536.

Zharov, V.T. (1969a). Non-Local Relations in Vapor-Liquid Equilibrium Diagrams for Multicom-

ponent Systems. Russ. J. Phys. Chem. 43(11), 1563–1567 [2784–2791].

Zharov, V.T. (1969b). Phase Representations [Distillation Lines] and Rectification of Many-

Component Solutions - II. Russ. J. Appl. Chem. 42(1), 94–98.

Zharov, V.T. and A.V. Storonkin (1969). Local Rules in the Vicinity of a Multicomponent

Azeotrope. Russ. J. Phys. Chem. 43(5), 628–631 [1126–1131].

Zharov, V.T and L.A. Serafimov (1975). Physicochemical Foundations of Simple Distillation and

Rectification (in Russian). Chemistry Publishing Co., Lenningrad.

Zhvanetskij, I.B., S.A. Reshetov and V.Yu. Sluchenkov (1988). Classification of the K-Order

Regions on the Distillation Line Diagram [Residue Curve Maps] for a Ternary Zeotropic
System. Russ. J. Phys. Chem. 62(7), 996–998 [1944–1947].