Separation system synthesis liquid mixtures separations
Ind. Eng. Chem. Res. 1990,29,421-432 421
Separation System Synthesis: A Knowledge-Based Approach. 1. Liquid
Scott D. Barnicki and James R. Fair*
Department of Chemical Engineering, The University of Texas at Austin, Austin, Texas 78712-1062
A description is given for a task-oriented, or problem decomposition, approach to the selection and
sequencing of methods for separating multicomponent liquid mixtures. The design knowledge of
the expert is organized into a structured query system, the separation synthesis hierarchy (SSH).
This hierarchy divides the overall separation synthesis problem into subproblems or “tasks”. These
tasks can be solved essentially independently from each other. Each task consists of a series of ordered
heuristics based on pure component properties and on process characteristics. In its current im-
plementation, SSH is limited to the sequencing of multicomponent mixture separations using eight
industrially significant separation methods: simple distillation, azeotropic/extractive distillation,
liquid-liquid extraction, stripping, adsorption, membrane permeation, and crystallization.
During the past 15years, considerable effort has been
expended on developing systematic methods for the se-
quencing of distillationcolumns. Evolutionaryand ordered
heuristic methods have been notably successful for this
type of space search problem and require relatively little
expert design knowledge (Nishida et al., 1981; Kelley,
Although distillation is the mainstay of the separation
industry, a considerablenumber of situations exist in which
distillation is a poor choice. The more general industrial
problem of separation synthesis, using a number of dif-
ferent separation methods, has received little attention.
Such a knowledge-intensive problem is not suited to so-
lution solely by the ordered heuristic methods developed
This paper describes a task-oriented, or problem de-
composition, approach to the selection and sequencing of
separation methods for multicomponent liquid mixtures.
The design expert’s knowledge is organized into a struc-
tured query system, the separation synthesis hierarchy
(SSH). This hierarchy divides the overall separation
synthesis problem into subproblems or “tasks”. Each task
can be solved essentially independently from the other
tasks. The separation synthesis hierarchy presented here
is being developed explicitly for implementation in a
knowledge-based expert system, the separation synthesis
advisor (SSAD). SSAD is currently in the prototype stage
of development. In its current implementation, SSAD is
limited to the preliminary sequencing of multicomponent
liquid mixtures using one of the following methods: (1)
simple distillation, (2) azeotropic/extractive distillation,
(3) liquid-liquid extraction, (4) stripping, (5) adsorption,
(6) membrane permeation, and (7) melt crystallization.
In this work, methods requiring an extraneous substance
to effect the separation are called mass separating agent
(MSA) processes. All methods on the list above except
simple distillation and melt crystallization (and sometimes
azeotropic distillation) are MSA processes. Both simple
distillation and melt crystallization require only the ad-
dition or removal of energy. Mass separating agent pro-
cesses are further divided into methods requiring physical
solvents or entrainers (PSE processes-azeotropic/ex-
tractive distillation, liquid-liquid extraction, and strip-
ping), and methods requiring solid-phase agents (SPA
processes-adsorption and membrane permeation).
The term azeotropic distillation is commonly used to
refer to two different types of fractionation involving
azeotropes. The first type relies on the azeotrope(s) in-
herently present in the mixture to effect the separation;
only the addition of energy is required. The second type
of azeotropic distillation is a PSE process. An extraneous
substance, called an entrainer, which forms an azeotrope
with one or more componentsis added to the mixture. The
fractionation of the resulting azeotrope(s) achieves the
The development of an expert system for the synthesis
of separation sequences is an interdisciplinary endeavor,
combining aspects of both chemical engineering and ar-
tificial intelligence (AI). These two diverse fields con-
tribute very different, but deeply interrelated, perspectives
to the separation synthesis problem. In broad terms, the
chemical engineering separation synthesis problem for
liquid mixtures can be stated as follows:
Given (1)an n-component liquid mixture, (2) physical
property data on the mixture, (3) product specifications,
and (4) a portfolio of potential separation techniques, find
the method(s) and sequence(s) of separations that (1)
produce the desired products with the desired purities, (2)
result in minimum separation costs, and (3) result in a
limited number of feasible, reliable process designs.
The synthesis of separation sequences is a classical
chemicalengineering design problem. Such work has been
done successfully for decades. However, due to the in-
herent uniqueness and complexity of each new design
problem, a comprehensive and systematic approach to
process synthesis has remained elusive;process design still
resides in the domain of the “expert”. As such, the fol-
lowing questions remain largely unanswered:
(1j What knowledge is needed to determine which sep-
aration techniques should be used and in what order they
should be accomplished?
(2) How does an expert organize this knowledge to make
The synthesis of separation sequences encompasses
three basic categories of AI problem types: (1)space
search, (2) selection, and (3) design.
The space search problem arises from the need to ef-
ficiently and systematically explore the potential separa-
tion sequences. Parallel to the sequencing is the selection
of a separation method for a given split in a multicom-
ponent mixture. The need for short-cut process modeling
and economic evaluation bring into play design problems.
Moreover,the situation is further complicated by the need
to manipulate a large data base of physical/chemical
The important questions from an AI/ knowledge engi-
neering viewpoint are (1) can the searchselection-design
problems be decoupled or decomposed into tractable
subproblems and (2) what is the most effective way to
represent and structure the separation design knowledge
01990 American Chemical Society
422 Ind. Eng. Chem. Res., Vol. 29. No. 3, 1990
s m m Selector
Figure 1. Separation synthesis hierarchy.
for use in an expert system environment.
Task-Oriented Expert System Design
In the past, a number of highly successful rule-based
expert systems (e.g., DENDRAL, (Feigenbaum et al., 1971),
and MYCIN (Buchanan and Shortliffe, 1984)) have been
constructed. In a rule-based system, the knowledge base
and the inference mechanisms are typically separate from
each other. The rules themselves often do not explicitly
indicate the order in which they should be used. Large-
scale rule-based systems usually resort to metarules, an
implicit grouping of rules. Metarules guide the problem
solving locally, by allowing specific rules to be used only
under certain circumstances. The separation of knowledge
and inference mechanisms promotes general, domain-in-
dependent programming but does not take advantage of
the inherent structure of many problems.
The task-oriented approach to expert system design
represents a strategy of explicit knowledge organization
(Chandrasekaran, 1986). This method is based on the
(1)A complex problem can be decomposed in terms of
“generic” problem types or “tasks”. A large problem may
be composed of scores of interrelated tasks.
( 2 ) The domain knowledge is available to encode into
blocks of knowledge, each of which solves a single task.
(3) The tasks can be built into a structured hierarchy
which solves the overall complex problem.
A problem decomposed in this manner can be thought
of as a group of “specialists” each working on a separate
task. Higher level “managers”ensure that the hierarchy
of specialists works toward resolution of the overall
problem. Tasks at the upper levels of the hierarchy are
more abstract in nature, while those at the lower levels are
more concrete. This behavior is reflected in the expert who
focuses on broader issues in the problem and delays con-
sideration of the low level details until much later. The
task-oriented method has proven useful in malfunction
diagnosis (Davis et al., 1987;Ramesh et al., 1988), equip-
ment design (Myers et al., 1988),and equipment selection
problems (Gandikota, 1988) in chemical engineering.
The Separation Synthesis Hierarchy
The key to the task-oriented approach is problem de-
composition and knowledge structuring. Expert process
engineers are able to select and combine successfully in-
dependent process steps into a coherent problem solution.
Clearly an extensive body of information on separation
processes is available, albeit much of it in a form unsuitable
for direct coding into tasks. The separation synthesis
hierarchy represents our approach to problem decompo-
sition and knowledge organization for separation synthesis.
The hierarchy emulates the approach that an expert
process engineer follows. It is based on interviews with
expert designers and supplemented by information from
Figure 1 presents the complete selection/sequencing
hierarchy in its present form. Each block represents a
clearly defined and essentially independent subtask of the
overall separation selection and desequencing problem.
The SSH consists of three types of task specialists. Each
type of specialist deals with a specific task type: (1)
manager-separation sequencing; (2) selector-separation
method selection, MSA selection; (3)designer-separation
Previously published heuristic methods have dealt with
a simplified version of one of these blocks, the liquid split
manager. These methods deal almost exclusively with split
Ind. Eng. Chem. Res., Vol. 29, No. 3, 1990 423
INPUT MIXTUREOne of the heuristics that is repeatedly referred to in
the literature (e.g., King, 1980;Rudd, 1973)states that the
method of separation should be chosen first. In terms of
the concepts used here, the heuristic states that all se-
lection tasks should be done first, (i.e., all selector spe-
cialists should be at the top of the separation hierarchy).
In most cases, this has meant that distillation is assumed
to be the best method for all separations.
The use of the method selection heuristic, in principle]
greatly reduces the magnitude of the remaining separation
synthesis problem. By eliminating the selection problem
all at once, one is left with only a split sequencing problem.
In other words,the selection and sequencingproblems can
be completely decoupled.
However, we have found the method selection heuristic
to be too restrictive. The selection and sequencing prob-
lems cannot be completely decoupled in this manner.
Although one can gain some early insight into the most
favorable separation method(s) for a given split, the final
choice cannot be made until much later. This is especially
true for methods requiring mass separation agents. A
judgement on separation method cannot be made until a
list of potential solvents or adsorbents is available. In turn,
the choice of solvent/adsorbent is influenced by the com-
position of the mixture to which it is to be added. Thus,
the method selection problem is dependent on both the
solvent/adsorbent selection task and the split sequencing
problem. The separation synthesis hierarchy reflects this
observation; selection and sequencingtasks are distributed
The form of the hierarchy is guided by two principles.
First of all,calculations are done as little as possible. Most
decisions in the upper levels of the hierarchy are based
solely on qualitative relationships. Detailed quantitative
information is used primarily for final comparisons at the
level of the designer specialists.
The second principle is that distillation is the bench-
mark separation method to which all other methods must
be compared. Distillation should always be the first me-
thod considered for any separation. Moreover, when other
methods give comparable results to distillation, the relia-
bility and efficiency of distillation make it the likely choice.
This is reflected in the hierarchy by the continued com-
parison to distillation. The following sections describe in
more detail the structure of the tasks needed for the
preliminary analysis of liquid mixtures.
Phase Separation Selector
At the highest level of the hierarchy is the phase sepa-
ration selector (Figure 2). The phase separation selector
uses equilibriumdata and normal boiling point informatior.
to determine whether a liquid or gas separation system (or
possibly both) is necessary. There are two purposes of this
The first purpose of the task is to divide the input
stream into substreams of low volatility components and
of high volatility components reducing the magnitude of
the sequencing problem. Although some components may
distribute, two smaller, independent sequencing problems
are created. These reduced problems will typically require
considerably less effort to solve. Removal of one compo-
nent from a multicomponent mixture will generally reduce
the number of possible separation sequences by an order
of magnitude or more.
The second purpose of the task is to reduce the method
selection problem. For gaseous mixtures, the number of
potential separations is reduced to only four: absorption,
adsorption, membrane permeation] and cryogenic distil-
lation. Similarly for liquids, one need only consider simple
Rank component8by 1
normal boilingpoints I
' three groups based on
1 normal boiling points
Tbp>50 C ,,.',.a'C c T < 50'C~ Tbpc-XI C
All components I
G o lo
by adjacent reldwe
NO All componenta
maxlmm separationbaween '
O..a + more I key components upulda+ In@
voI~111eG-L componema vomlk 0.L sompononta
Figure 2. Phase separation selector.
distillation, extractive/azeotropic distillation, liquid-liquid
extraction, adsorption, membrane permeation, stripping,
A simple example illustrates the utility of the phase
separation task. Thompson and King (1972) developed
an equation relating the number of components, N , to be
separated by M potential separation methods to the
number of possible sequences, S:
For a 6-component mixture using the 10 potential sep-
aration methods mentioned above, there are 4200000
possible separation sequences. Now assume four compo-
nents appear in each of the liquid and gas substreams (i.e.,
two nonkey components distribute to both the liquid and
gas). Considering four potential separation methods for
the gas mixture and eight methods for the liquid stream,
the number of possible sequences is 320 and 2560, re-
spectively. Thus, for this case, the number of possible
sequences is reduced by 99.9%.
The grouping of components into liquid and gas sub-
streams is based on the relationship of the normal boiling
point of a component to the pressure needed to perform
a separation of that component by distillation. Theoret-
ically, distillation can be used over the entire range that
vapor and liquid phases coexist (i.e., from the freezing
point to the critical point). However, in practice, distil-
lation at extremes of temperature and pressure are often
prohibitively expensive. At these limits, other separation
methods compete favorably with distillation.
424 Ind. Eng. Chem. Res., Vol. 29, No. 3, 1990
Table I. Distillation Conditions LIQUID MIXTURE
component normal bp pressure range, condenser
type_-group range, atm
gas T R p > -20 P > 25 refrigeration
gas-liquid -56 < TBP< 0.0 14.5 < P < 25 partial
0.0 < T B p < 50.0 P < 14.5 total
liquid 7 u p > 50 P < 14.5 total
Gases are taken as those components with normal
boiling points less than -20 "C. Distillations of such
components typically require high pressures (greater than
25 atm) and refrigeration. Components that can be con-
densed by cooling water (normal boiling point of 50 "C or
more) are considered to be liquids. Distillation pressures
are usually less than 14.5atm and total condensers can be
used. Table I summarizes the distillation conditions for
gases and liquids.
Components with normal boiling points between -20 and
50 "C require further evaluation. These gas-liquid tran-
sition components may require either partial or total
condensers with distillation pressures between 25 and 14.5
atm. At this point in the decision process, one cannot
make a clear judgement on the appropriate separation
method for transition compounds. It may be best to
condense these components so as to use liquid separation
methods. On the other hand, gas separations may be more
With the grouping of the components identified, the
next step is to calculate adjacent relative volatilities at the
input mixture temperature and pressure. In most cases,
there will be at least one large adjacent relative volatility
value between two components in the gas-liquid transition
region. This will certainly be true if there are no transition
compounds;the relative volatility between the least volatile
gas and most volatile liquid will undoubtedly be large. The
components with the largest adjacent relative volatilities
in the gas-liquid transition region are chosen as the key
The mixture is divided into gas and liquid streams by
a simple equilibrium flash. The flash is conducted at an
appropriate temperature and pressure so that the split
between the key components is reasonably sharp. (See
Example 2: Purification of Acetic Acid.) The liquids with
some gas-liquid transition compounds proceed to the liq-
uid split manager. Similarly, the gases go to the gas split
Liquid Split Manager
The next phase in the synthesis process involves a
preliminary effort at split sequencing. The sequencing
method emphasizesthe use of distillation for as many splits
as possible and the early use of distillation. The primary
purpose of the liquid split manager (LSM) is to make the
best distillation sequence possible out of those separations
where simple distillation is the favored method. Separa-
tions that require mass separating agent processes or
crystallization are deferred to a lower level manager. The
four-step procedure is outlined below (see also Figure 3).
(1)Identify product streams and product specifications.
This ensures that no unnecessary separations are done.
(2) Rank components in order of decreasing adjacent
relative volatilities. Relative volatility gives a strong in-
dication of the ease of separation and the favorability of
(3) Identify all azeotropic mixtures that may interfere
with product specification. Azeotropes require special
processing considerations and should be dealt with when
I Rank components by
adjacent relative volatilities
l Identify products and 1
1 product stream speclfications
-I Identify all known azeotropes, I
1 potential azeotropes rf information ,IS lacking 1
Iorder speclfied by heuristics
l Gas-Liquid Azeotropic Split
Selector 'Zeotropic Split
1 Repeat until all
Figure 3. Liquid split manager.
further information is available.
(4) Perform splits in the order specified by a set of
sequential heuristics. Each potential split is evaluated by
one of the mixture selector specialists (see next section,
Zeotropic/ Azeotropic Mixture Selectors). If simple dis-
tillation is the favored method, then the separation is
performed, and the resulting substreams are analyzed
further by the LSM. If simple distillation is inappropriate,
the separation is not performed, but other potential sep-
aration methods are identified. The next split specified
by the LSM is now checked for the applicability of simple
The LSM is guided by the assumption that all simple
distillations should be performed first. This is based on
the premise that simple distillation, when suitable, is the
easiest and most reliable method for multicomponent
separations. The presence of nonkey components tends
to complicate the design of MSA processes and crystal-
lizers. Moreover, as mentioned previously, the removal of
a component from the mixture reduces the number of
possible sequences by an order of magnitude or more.
Azeotropicseparations are typically difficult to perform.
They should be performed in the absence of other com-
ponents if possible. It is important to identify these
mixtures as early as possible. When data are available,
the azeotropes can be easily identified. However,for cases
when incomplete information is available, the potential of
azeotrope formation can still be determined. The following
set of five questions, in decreasing order of certainty, are
used to indicate the likelihood of azeotropes. An affirm-
ative response indicates unlikelihood. In other words, an
answer of yes to question 1is a stronger indication that
azeotropes are not present than an answer of yes to
Ind. Eng. Chem. Res., Vol. 29, No. 3, 1990 425
(1)Are the components homologous or isomers of the
(2) Is the difference in normal boiling points greater than
(3) Are the components members of chemical families
(4) Are the carbon numbers of the compounds greater
(5)Is the ratio of vapor pressures less than the infinite
same chemical family?
unlikely to form azeotropes?
dilution activity coefficient?
p?%/p!&< %K (2)
(This is a semiquantitative relationship, based on the as-
sumptions that the binary solution is regular and the ac-
tivity coefficient curves are symmetric (Martin, 1975)).
Once the azeotropes have been identified, a list of or-
dered heuristics is used to obtain a preliminary split se-
quencing. The list of heuristics is based on the work of
Nadgir and Liu (1983). Their list has been modified to
account for azeotropes. The heuristics are applied se-
quentially. If a heuristic is inapplicable, the next one on
the list is considered.
(1)Remove corrosive and hazardous materials first.
(2) Remove reactive components first.
(3) Perform separations between azeotropes last.
Azeotropic separations tend to be difficult, and they should
be done in the absence of other components.
(4) Perform difficult zeotropic (nonazeotropic) separa-
tions last, but before azeotropic separations. This is a
modification of the heuristic of Rudd et al. (1973)and King
(1980) stating that separations of low relative volatilities
should be done in the absence of other components.
(5) Remove components in order of decreasing per-
centage of the feed. If the relative volatility is reasonable,
a component that is a large fraction of the feed should be
removed first to decrease the size of later separation
(6) Favor 50/50 splits. If feed percentages do not vary
widely, favor sequences that give equimolar product and
residue streams provided the relative volatility is reason-
(7) All things being equal, perform the separation with
the smallest coefficient of difficulty of separation (CDS)
first (Nath and Motard, 1981). The CDS quantifies the
last three heuristics:
The first term is the number of minimum stages for
distillation. The second and third terms penalize uneven
distributions and overly large distillates. In essence, the
CDS is a measure of the applicability of distillation.
It must be emphasized that the split sequence specified
at this point is preliminary. The LMS determines the best
sequence for the separations that can be done by simple
distillation. Separations requiring MSA methods or
crystallization are identified. Sequencing of these sepa-
rations is done at a lower level of the hierarchy.
Zeotropic/Azeotropic Mixture Selectors
For each split selected by the LSM,one must determine
a list of potential separation methods. This task is ac-
complished by one of three mixture selector specialists.
The mixture selectors do not indicate a ranking of sepa-
NO Arefhe componems YES
1 temperature sensnwe 7 +---
~ -- -*BULK Is IhlS a DILUTE IS this a BULK
-- bulk ordilute - bulk or dilute -separation7
reparaion 7 I
i T t
SEE SEE SEE
Figure 7: TemperalurwFigure 8: Zwlroplc Figure 6: Dilute Swn8nlve Sapiratlons
MlXture .Bulk. Tmnrmrlture Sepiritlons
Figure 4. Zeotropic mixture selector.
ration methods from most favored to least favored but
rather an unordered list of all possible processes.
The zeotropic selector is used for separations between
nonazeotropic (zeotropic)components (Figures 4 and 6-8).
The azeotropic selector is used for separations between
azeotropic components (Figures 5-7 and 9).
The gas-liquid transition selector determines whether
a group of components identified as gas-liquid transition
components by the phase separation selector should be
condensed or vaporized (Le., determines whether gas or
liquid separation methods should be used). This task will
be described in a future paper.
Qualitative information is still quite useful at this level
of analysis. The mixture selectors employ criteria based
on pure component data, process characteristics, and
whether azeotropes are present. The results of these sim-
ple comparisons generally reduce the number of potential
separation methods to four or less.
(A)ComponentProperties. (1) Relative Volatility.
The relative volatility, a, between two components indi-
cates the ease of separation by simple distillation. For a
> 1.5,simple distillation is generally the most economical
process (seebelow, Process Characteristics. (1)Separation
Type, for a possible exception to this rule). If a < 1.1,
distillation requires high refluxes and large numbers of
stages. For these cases, distillation is ruled out. For the
large gray area (1.1< a < 1.5),other separation methods
may be competitive with simple distillation. No firm
judgment can be made by these qualitative comparisons.
(2) Slope of Vapor Pressure Curve. If the slopes of
the vapor pressure curves of two components differ sig-
nificantly within an acceptable temperature and pressure
range, then the relative volatility can generally be altered,
possibly to greater than 1.5. The “acceptable” temperature
and pressure ranges will depend on available heating
medium and cooling water temperatures and on the tem-
perature sensitivity of the materials being processed.
(3) Freezing Point Differences. Feasible crystalli-
zation processes typically require 20-30 “C differences in
pure component freezing points. In addition, the freezing
points should be at or above ambient temperatures if the
added expense of refrigeration is to be avoided.
(4) Chemical Family Similarity. Selective physical
solvents for PSE processes will achieve separations only
for chemically dissimilar components. Homologues of
similar size and isomers in the same chemical family
generally cannot be separated by PSE methods. Com-
pounds of close molecular weight and shape in the same
chemical family tend to exhibit similar physical properties
and thus similar selectivity and solubility in solvents.
As the size and shape differences increase, the physical
properties may differ considerably, even for homologues.
Typically, the boiling points of compounds of largely
varying sizes in the same chemical family will be suffi-
ciently different to allow the use of simple distillation.
However,when simple distillation cannot be used for other
reasons (e.g.,temperature sensitivity of the compounds),
PSE processes should not be eliminated as potential sep-
aration methods for chemically similar compounds of
widely varying sizes.
The effectiveness of a given membrane for a separation
depends both on the diffusivity and solubility of the
various components in the membrane. Solubility can be
related roughly to the interaction between the functional
groups in the membrane material and those of the com-
ponents to be separated. The differences in solubilities
of two given components will be significant only if the
components themselves contain different functional
groups. Thus, membrane permeation may be a feasible
separation method if the components to be separated are
in different chemical families.
(5) Structure and Size Characteristics. Membrane
permeation based on diffusion effects and molecular sieve
adsorption both require structural and/or size differences
between components to be separated. The effect of
structure and size on selectivity can be especially dramatic
for adsorption using zeolites and carbon molecular sieves.
Certain sizes and shapes of molecules may be excluded
completely from the micropores of the adsorbent due to
the extremely narrow distribution of pore sizes. A number
of industrially important bulk adsorptive separations are
based on this molecular sieving effect, notably Union
Carbide’s IsoSiv processes (Cusher, 1986) and certain
Sorbex processes of UOP (Mowry, 1986). Even if the size
and structural differencesare insignificant,adsorption may
still be a feasible alternative if polarities vary.
(6) Polarity Differences. Commercial adsorbents can
be divided into polar and nonpolar types. Polar adsor-
bents, such as silica gel, activated alumina, and zeolites,
tend to bind the polar compounds in a mixture more
strongly. Nonpolar adsorbents, such as activated carbon,
are more useful for removing less polar materials from a
mixture of more polar compounds. For both polar and
nonpolar adsorbents, higher selectivity is achieved when
there is a large difference in polarity between the desired
adsorbates and the unadsorbed liquid. However, adsorp-
tion may still be a viable option if polarities are similar
when size and structural differences are large.
(7) Boiling Point Range. The boiling range of the
component to be separated may indicate the favored me-
thod. For example, stripping is favored for separations of
low boilers. Liquid-liquid extraction and extractive dis-
tillation are better for high boilers.
(8) Temperature Sensitivity. Some components may
decompose or react unfavorably at the temperature needed
for distillation. Moreover, the freezing point of a com-
ponent may be too high for distillation to be carried out
at an acceptable temperature and pressure. Simple, ex-
tractive, or azeotropic distillation cannot be used for these
(B)Process Characteristics. (1) Separation Type
(Bulk or Dilute). As the ratio of distillate to bottoms
moves away from unity, other separation methods compete
more favorably with distillation. In general, a dilute dis-
tillation is uneconomical. A separation is considered dilute
when the total distillate or bottoms of a potential distil-
lation operation is less than 5% of the feed.
In addition, a large distillate-to-bottoms (D/B) ratio has
a greater effect on the econimics of a distillation than a
small D/B ratio. Mixtures composed of mostly low-value,
low-boiling components to be separated from a small
Ind. Eng. Chem. Res., Vol. 29, No. 3, 1990
amount (less than 10-1570) of a low-value, high-boiling
component require large amounts of energy to vaporize the
85-9070 of the feed that will appear in the distillate. All
forms of distillation (simple, extractive, and azeotropic)
can be eliminated as potential methods for dilute sepa-
During the past 10years, adsorption has gained a place
as a bulk separation method in addition to its continued
use as a dilute purification tool. Union Carbide’s vapor-
phase IsoSiv processes and UOP’s liquid-phase Sorbex
technology have proven economical for the separation of
what are considered here as liquid compounds (see phase
separation selector for the definition of liquid compounds).
Thus, adsorption is a potential method for both dilute and
Membrane permeation can generally be used only for
dilute liquid mixtures. No bulk liquid separationsare done
commercially. Melt crystallization is limited to bulk sep-
arations. The low reliability and low recovery typically
associated with crystallization processes make its use as
a dilute purification tool unfeasible. Liquid-liquid ex-
traction and stripping can be used for either dilute or bulk
separations if an appropriate solvent can be found.
(2) Purity. In practice, both simple distillation and
crystallization can achieve high-purity separations (99+YO
pure). The purity of the products obtained by PSE pro-
cesses depends to a large extent on the solvent chosen.
However, in principal, PSE processes can achieve high-
purity separations if a selective solvent can be found.
Adsorption is much the same. If a selective adsorbent can
be found, high purity is possible.
Membrane permeation, on the other hand, tends to give
only an incremental increase in purity with each passage
through the membrane. As a result, a high-purity product
will not result from membrane permeation unless one re-
sorts to a multistage scheme. Depending on the selectivity
of the membrane, typically at least four stages are needed
to achieve greater than 90% purity, with a correspondingly
low recovery rate. Thus, if a high purity is essential,
membrane permeation can be eliminated as a potential
(3) Recovery. Recovery is defined here as the degree
of separation obtained between product streams. In other
words, a high recovery separation results in two high-purity
products. As is the case with purity, simple distillation
and PSE processes (with a selective solvent) can achieve
high recovery separations. Adsorption recovery can be high
for both bulk and dilute solutions, depending on the ad-
sorbent. Bulk adsorptive separations using IsoSiv or
Sorbex technology are claimed to have recoveries of
95-98% (Mowry, 1986; Cusher, 1986).
Melt crystallization recovery is limited in practice by the
presence of eutectic points. In all crystallization opera-
tions, only one pure component crystal can be obtained
at a time. For simple systems, a second component will
not crystallize until all of the first component is removed
from solution. However, if the system in question forms
a eutectic, the second component will begin to simulta-
neously crystallize at some intermediate composition (see
Walas (1985)for a review of solid-phase thermodynamics).
Although the two crystals can sometimes be separated by
density differences, this is usually not a reasonable in-
dustrial option. Thus, the eutectic point represents a
practical limit on the recovery of crystallization processes.
The maximum fractional recovery,R, of a component can
be related to the eutectic point composition:
Ind. Eng. Chem. Res., Vol. 29, No. 3, 1990 427
Table 11. Special Processing Situations
favored method condition
a dilute solution (between 1% and 5%) of a high boiling, polar compound; distillation would require
vaporization of large amounts of the feed
a dilute solution (<5%) of a low-boiling component; if a highly selective solvent is available, stripping may
compete favorably with distillation
a low concentration (<10-15%) of the component which forms a minimum boiling azeotrope with the
entrainer; in this case, the minimum boiling azeotrope will go overhead; a low concentration of this
component reduces the vapor load
extractive distillation a close boiling mixture in which the product is the less volatile component; the extractive agent, which is
introduced at the top of the column, will alter the relative volatility throughout the column
molecular sieve adsorption when both polarity and size/differences are large, adsorbent selectivity is enhanced by the presence of
molecular sieving and polarity effects
Table 111. Physical Properties of Xylene Isomers and Ethylbenzene
m-xylene o-xylene p-xylene ethylbenzene
composition, mol % 43 23 19 15
normal freezing pt, K
normal bp, K
dipole moment, D
slope of vapor pressure curve,
max diameter, A
where xfc is the feed mole fraction of the crystallizing
component and x,, is the eutectic mole fraction of the
crystallizing component. In addition, the eutectic com-
position determines which component can be obtained as
a pure crystal for a given feed composition. The compo-
nent for which xfe> x,, crystallizes first.
Membrane permeation is typically an enrichment pro-
cess. The relative ratio of component mole fractions is
shifted, but neither the permeate nor the residual liquid
is highly pure. Thus, if high recovery is essential, mem-
brane permeation and melt crystallization (depending on
the eutectic point) can be eliminated as potential separa-
(C) Azeotropic Separations. Azeotropic separations
require additional analysis (see Figure 5). Systems con-
taining homogeneous azeotropes cannot be separated by
simple distillation unless the azeotrope composition varies
with pressure. When the composition changes at least
4-570 over a nominal change of total pressure, then it is
possible to use two-column simple distillation schemes
(Smith, 1963). If the azeotrope composition is not pressure
sensitive, then simple distillation can be immediately
eliminated as a potential separation method.
As is apparent from Figures 6-9, a number of situations
exist in which several separation methods feasibly could
be used for a given split. The mixture selectors determines
which separation methods may be feasible for a given split.
Only those methods that are clearly inappropriate are
For example, assume that two key components are in
different chemical families, indicating that PSE processes
should be considered as potential separation methods.
However, after a more detailed analysis, a solvent cannot
be found with a high enough selectivityfcapacity to make
liquid-liquid extraction competitive. In this case, liquid-
liquid extraction can be eliminated as a potential method,
although in theory it could be used.
Analysis at this highly qualitative level cannot indicate
with further certainty whether one method is favored over
another. Although not conclusive,some methods are more
appropriate for certain special situations. A list of these
special considerations is given in Table 11. The principles
explored here are further illustrated by two examples of
NO Are the components YES
temperame sensme 7
-.a__ ----Y_---. .
BULK Isthisa DILUTE Isthis a BULK
- bulk or diilite -- bulk or d w e -- -
separation 7 1
'I 1 1
Mixture. Bulk. Temperature
Flgure 0: kzwtroplc Flgure 6: Dllule Flgure 7: Temperature
Separatlons Sensnlve Separatlons
Figure 5. Azeotropic mixture selector.
industrially significant separation problems.
Example 1: Xylene Isomer Purification
Xylene isomers and ethylbenzene are important raw
materials in the plastics industry (Debreczeni, 1977). The
typical composition and physical properties of the isomer
product stream from a naphtha reformer are given in Table
According to eq 1,for a 4-component mixture with 10
potential separation methods, the number of possible se-
quences is 5000. Examination of the normal boiling points
revealsthat all components can be considered liquids. This
eliminates the gas separation methods, reducing the num-
ber of possible sequences to 2560. The complete mixture
goes to the liquid splits manager.
For this example, it is assumed that the products are
four 99% pure streams of one component each. No
azeotropes are present in the mixture nor are there any
gas-liquid transition components. Therefore, all splits can
be handled by the zeotropic mixture specialist (ZMS).
Calculated values of adjacent relative volatilities for this
mixture at 1atm and 450 K are q2= 1.60, = 1.06,and
a31= 1.02,where 1is m-xylene, 2 is o-xylene,3 is p-xylene,
4 is ethylbenzene.
The first round of sequencing can now be accomplished
by application of the list of ordered heuristics as follows:
heuristic 1, not applicable; heuristic 2, not applicable;
428 Ind. Eng. Chem. Res., Vol. 29, No. 3, 1990
SEPARATIONIs the dltference Conslder
In polarities large 7 METHOD(S)
Follow ALL of the
branches. not Inany
Are the component.
oi stmllar slze/shape 7
Are tho components ConcIder
In Ihe same LIOUID-LIOUID EXTRACTION
Figure 6. Dilute separations.
Ar. lh. eomponea.
of .Imllar r1i.l.h.p. ?
FollouALL d Ih.
bfanehn .no1Inm y
I" fnrlnp pal"1. *In.ll?
Figure 7. Temperature-sensitive separations.
heuristic 3, not applicable;heuristic 4, the relative vola-
tilities between ethylbenzene and p-xylene (4-3) and be-
tween p-xylene and m-xylene (3-1) are very small; these
separations should be done last; heuristic 5, m-xylene is
the largest component in the feed, but the low relative
volatility overrides this consideration; heuristic 6, a 50-50
split cannot be accomplished with these relative volatilities;
heuristic 7 ,by comparing the coefficients of difficulty of
separation,the favored first split is between o-xylene and
Ind. Eng. Chem. Res., Vol. 29, No. 3, 1990 429
Is Ih. ielaI1Y.
- - -..../" O l ~ l l l l l y1.10 andktw""1.50 7 ~ O I S T I L L A T I O N
Are Ih. canpownls LIQUIC-LIQUID EXTR*CTION
vol.lIlly > 1" OISTILLATIONUaeSIMPLE ,/f i s kch.mk.1 Iamlly 7 UEOTROPICEXTRICTNECa*ld'rD l S n U T l O NO l S n U T l O H - ~ ..._.._--------.__--__
-_ 1~ ,
Follow ALL al vr
b1anch.i. no1 In my
necnsmry w w Ellmlnml.
MOLECULAR SIEVE ADSORPTION
I , ,
,',' ,;' I I
__-- ,' ,
._--..--Consider ._..~.--Are lh. eomponenls hIhe dm.imc.
01 slmll., .Iz.lihap. 7 Inpoiarnin I~,W 7 ADSORPTION
MOLECULAR SIEVE .._._.._.--------
Figure 8. Zeotropic mixture: bulk, temperature-insensitiveseparations.
U W I D SPLIT
I. Ih. .*~110p10compo.llla
P I . Y Y I 1 ..n.m. 7
Ar. Ih. components LIOUIDLIOUID EXTRACTION
EXTPACTIVEDISTILLATION - -~----.... -
b 2.COLUMN SIMPLE
Is Ih. dlflrrmc. C o n * m ...... ~ ~ ~ .......-...,in poimmi.. I.I~.7 ADSORPTION
Follow ALL Z the
branch.. .not 10 any
Are Ih. components
01 SImllar .Izmh.p 7
b lh. ,.I.U".
rd.!lllP+ > I . = ?
A 4 Consldrr
*.pa p,n.ur. SYW..
dm.r *ipnme.ntty 7
I. !he dmeianc.
I"I..llng pol"!. .m.ll?
b e w e n 1 10.nd1.507 2-COLUMN SIMPLE
Figure 9. Azeotropic mixture: bulk, temperature-insensitive separations.
For the m-xylene-o-xylene separation, the analysis by
the ZMS is relatively uncomplicated. This is a bulk sep-
aration of temperature-insensitive compounds with a
relative volatility greater than 1.5. Simple distillation can
be used. The bottoms will contain o-xylene;the distillate
will contain the remaining three components.
Reapplication of the heuristics to the remaining mixture
reveals that the ethylbenzene-p-xylene split should be
done next. For this to be true, simple distillation must be
the favored separation method. Proceeding through the
ZMS, one determines that simple distillation is uncom-
petitive ( a C 1.1and the slopes of the vapor pressure
curves are similar). Similarly, for the meta-para separa-
tion, simple distillation is not the clear-cut choice. At this
point, one must determine what other methods have po-
tential for the desired separations.
Referringto Figure 8,for a bulk, temperature-insensitive
mixture, one must answer all the questions on each of the
430 Ind. Eng. Chem. Res., Vol. 29, No. 3, 1990
Ethylbenzene ~ CRYSTALLIUTION
p Xylene sp't' c
Separaoon bym Xfle-e
5PL1T1 SIMPLE DISTILLATION
o.Xyiene -PRODUCT 1
Figure 10. Xylene isomer purification.
four branches to determine the favored separation me-
thod(s). Since the relative volatilities for both the meta-
para and p-xylene-ethylbenzene splits are less than 1.1,
simple distillation can be eliminated as a potential method.
All three compounds can be considered to be in the same
chemical family (alkyl-substituted benzenes) and are of
very similar size and shape. Most likely, a solvent cannot
be found that will selectively separate one of these com-
ponents from the other two. Thus, liquid-liquid extrac-
tion, azeotropic/extractive distillation, and stripping can
With a less than 1-8,size difference between the three
components, adsorption based solely on molecular sieving
effects is infeasible. However, the polarity differences as
measured by the dipole moments are large. Between p-
xylene and m-xylene, the difference is 0.4 D. Between
m-xylene and ethylbenzene, the difference is 0.18 D. In
addition, p-xylene is completely nonpolar (dipole moment
is 0.0 D). Therefore, adsorption should be considered as
a potential separation method for both splits.
In this mixture, the freezing points of the components
differ considerably (47 K between m-ethylbenzene and 61.2
K between meta-para). Egan and Luthy (1955) showed
that the binary system of m-/p-xylene forms an eutectic
at 15mol 73 p-xylene. The maximum fractional recovery
for the feed conditions is 0.67 (from eq 4). Such a low
recovery does not allow one to meet the product specifi-
cations (99% pure streams of one component each).
Consequently, melt crystallization cannot be used for the
meta-para split as the problem is stated here.
No information is available on ethylbenzene-xylene
eutectics. For this preliminary analysis, one can assume
that the desired recovery can be achieved. Thus, crys-
tallization can be considered as a potential method for the
The analysis possible by the current SSH is now com-
plete. The next step is to determine a list of candidate
adsorbents for the ethylbenzene-m-xylene and m-xylene-
p-xylene splits. Once the adsorbent list is available, one
can compare the favorability of adsorption to crystalliza-
tion. Although the synthesis problem is far from complete,
this simple structured analysis has reduced the number
of potential sequences by 99.970,from 5000 to 4. Figure
10 summarizes the results.
Example 2: Purification of Acetic Acid
Direct oxidation of n-butane to produce acetic acid has
been practiced in the United States since the early 1950's.
A typical reactor effluent from oxidation over a manga-
nese(II1) catalyst is shown in Table IV (Prengle and Ba-
rona, 1970). For this example, it will be assumed that the
Table IV. Acetic Acid Product Mixturea
comDonent mol % comuonent mol 9i
acetic acid 30.5 methanol 1.5
formic acid 11.5 ethanol 3.5
formaldehyde 0.5 acetaldehyde 0.5
acetone 1.0 water 50.0
a Source: Prengle and Barona, 1970.
CO 73.0 CO2 27.0
Table V. Boiling Points of Acetic Acid and Bmroducts
component normal bp, K grouping
Table VI. Liquid-Phase Component Properties
formic acid 46.0
acetic acid 60.0
freezing pt, K
products of interest are pure acetic acid and pure formic
At the level of the phase separator selector, the com-
ponents are first ordered by normal boiling points as shown
in Table V. Calculation of relative volatilities shows that
there is a clear separation point between acetaldehyde and
acetone. A flash separates the mixture into a vapor phase
consisting of the gases, gas-liquid transition components,
some acetone and methanol, and traces of other compo-
nents, plus a liquid phase consisting of the liquid compo-
nents. Table VI lists the physical properties of the liq-
The liquid split manager examines the liquid mixture
for the possibility of azeotrope formation. A number of
azeotropes are present in this mixture (see Figure 11
(Horsley, 197311,but only the binary azeotrope of formic
acid-water and the ternary azeotrope of formic acid-
water-acetic acid may interfere with product specifications.
The ternary azeotrope is pressure sensitive and is not
present at atmospheric conditions. The binary azeotrope,
on the other hand, is not pressure sensitive.
Examining the list of heuristics reveals that acetic acid
should be separated first because of its corrosiveness. Since
the ternary azeotrope is not a problem, the separation of
acetic acid is analyzed by the zeotropic mixture selector.
Acetic acid is not temperature sensitive and is present in
the mixture at bulk concentrations. The relative volatility
between acetic acid and water is fairly low at 1.21. The
slopes of the vapor pressure curves are similar; the relative
volatility cannot be altered. The polarities and sizes are
similar, eliminating adsorption. The freezing points are
too close for crystallization (16 "C), but the components
Acetic Acid 1
MethylEthyl Ketone 5
-Q Ivpe ComDoritlon
maximum P 6% waterA 10765
B 55 5 minimum 12 wb methanol
c 635 minimum 70 6% mathano1
D 74 0 minimum 39 wb ethanol
E 734 minimum 11 3?4water
78 17 minimum 4 wb water
698 ternF$&23 4% water 45 7% FormicAcid
Figure 11. Azeotropes in acetic acid product mixture.
are chemically dissimilar. Therefore, one of five methods
could potentially be used for this separation, simple dis-
tillation, or one of the PSE methods. Of these five,
stripping is inappropriate because the product is not a
dilute, low boiler. One cannot discriminate further be-
tween azeotropic, extractive, and simple distillation or
liquid-liquid extraction until a list of potential solvents
The separation of formic acid is handled by the azeo-
tropic mixture selector because of the formic acid-water
azeotrope. This azeotrope cannot be eliminated by
changing pressure, ruling out simple distillation. As with
the acetic acid separation, this is a bulk, temperature-in-
sensitive separation. Moreover, the freezing point dif-
ference is small (9 "C) and the polarities and sizes are
similar. Thus, the potential separation method is quickly
reduced to only PSE processes. Again, the remaining
methods can only be evaluated after a list of solvents is
Although the problem is not complete, the number of
potential separation sequences has been considerably re-
duced. The results are summarized in Figure 12.
This paper has presented a qualitative, task-oriented
approach to the separation synthesis problem. The overall
synthesis problem is decomposed into a series of essentially
independent sequencing and selection subproblem tasks.
The tasks are organized into a structured hierarchy, the
separation synthesis hierarchy, based on the approach
followed by expert process designers. Four the subtasks,
the phase separation selector, the liquid split manager, and
the zeotropicfazeotropic mixture selectors, are described
The tasks described here represent a qualitative method
of rapidly reducingthe magnitude of the overall separation
synthesis problem. The current version of the SSH is not
a completesolution to the synthesis problem. Considerable
work is required on solvent selection, gas separations, and
Ind. Eng. Chem. Res., Vol.
. . .
29, No. 3, 1990 431
Figure 12. Acetic acid purification.
sequencing of MSA processes. These topics willbe covered
in future papers. In spite of these limitations, the number
of possible separation sequences can be reduced by 90%
or more for most liquid mixture cases.
This work was supported in part by a grant from the
Exxon Education Foundation. Support was also provided
by the Separations Research Program at The University
of Texas at Austin. We appreciate very much the gener-
B = bottoms flow rate
D = distillate flow rate
M = number of potential separation methods
N = number of components in a multicomponent mixture
P = vapor pressure, total pressure
S = number of possible separation sequences
Xi= mole fraction of component i
a = relative volatility
y = activity coefficient
m = infinite dilution
sat = saturation pressure
LK = light key
HK = heavy key
AI = artificial intelligence
CDS = coefficient of difficulty of separation
432 Ind. Eng. Chem. Res. 1990,29, 432-436
LSM = liquid split manager
MSA = mass separating agents (MSA processes include all
PSE and SPA processes)
SSAD = separation synthesis advisor
SSH = separation synthesis hierarchy
ZMS = zeotropic mixture selector
PSE = physical solvents/entrainers (PSE processes include
azeotropic/extractive distillation,liquid-liquid extraction,
SPA = solid-phaseagents (SPAprocesses include adsorption
and membrane permeation)
Buchanan, B. G.; Shortliffe, E. H. Rule-Based Expert Systems;
Addison-Wesley: New York, 1984.
Chandrasekaran,B. Generic Tasks in Knowledge-Based Reasoning:
High-Level Building Blocks for Expert System Design. IEEE
Expert 1986, Fall, 23.
Cusher, N. A. UCC IsoSiv Process. In Handbook of Petroleum
Refining Processes; Meyers. R. A,, Ed.; McGraw-Hill: New York.
Davis, J. F.; Shum, S.K.; Chandrasekaran, B.; Punch, W. F., 111.A
Task-Oriented Approach to Malfunction Diagnosis in Complex
Processing Plants. Presented at NSF-AAAI Workshop in AI in
Process Engineering, Columbia University, New York, March
Debreczeni, E. J. Future Supply and Demand for Basic Petrochem-
icals. Chem. Eng. 1977, 84 (June 6), 135.
Egan, C. J.; Luthy, R. V.Separation of Xylenes: Selective Solid
Compound Formation with Carbon Tetrachloride. Ind. Eng.
Chem. 1955; 17, 250.
Feigenbaum, E. A.; Ruchanan, B. G.; Lederberg, J. On Generality
and Problem Solving: A Case Study Involving the DENDRAL
Program. Mach. Intelligence 1971, 6, 165.
Gandikota, M. S. Expert Systems for Selection Problem Solving
Using Classification and Critique. Masters Thesis, The Ohio
State University. Columbus, 1988.
Horsley, L. H. Azeotropic Data-HI; ACS Advances in Chemistry
Series 116;American Chemical Society: Washington, DC, 1973.
Kelley, R. M. General Process Considerations. In Handbook of
Separation Process Technology; Rousseau, R. W., Ed.; John
Wiley: New York, 1987; Chapter 4.
King, C. -7. Separation Processes, 2nd ed.; McGraw-Hill: New I’ork,
Martin, G. Q. Guide to PredictingAzeotropes. Hydrocarbon Process.
1975, 54 (111,241.
Mowry, J. R. UOP Sorbex Separations Technology. In Handbook
of Petroleum Refining Processes; Meyers, R. A,, Ed.; McGraw-
Hill: New York, 1986.
Myers, D. R.; Davis, J. F.; Herman, D. J. Still: An Expert System
for Distillation Design. Computers in Chemical Engineering The
Ohio State University: Columbus, 1988;Vol. 12, Nos. 9 and 10,
Nadgir, V. M.; Liu, Y. A. Studies in Chemical Process Design and
Synthesis, Part 5. AIChE J. 1983, 29 (6), 926.
Nath, R.; Motard, R. L. Evolutionary Synthesis of Separation Pro-
cesses. AIChE J. 1981, 27 (41, 578.
Nishida, N.; Stephanopoulos, G.; Westerbert, A. W. A Review of
Process Synthesis. AIChE J. 1981, 27 (31, 321.
Prengle, W. H.; Barona, N. Make Petrochemicals by Liquid Phase
Oxidation. Hydrocarbon Process. 1970, 49 (31, 106.
Ramesh, T. S.;Shum, S. K.; Davis, J. F. A Structured Framework
for Efficient Problem Solving Diagnostic Expert Systems. Com-
put. Chem. Engl. 1988, 12, 891.
Rudd, D.; Powers, G.; Sirola,J. J. Process Synthesis: Prentice Hall:
Englewood Cliffs, NJ, 1973.
Smith, B. D. Design of Equilibrium Stage Processes; McGraw-Hill:
New York, 1963.
Thompson, R. W.; King, C. J. Systematic Synthesis of Separation
Schemes. AIChE J. 1972, 28 (5), 941.
Walas, S. M. Phase Equilibria in Chemical Engineering; Butter-
worths: Stoneham, MA, 1985.
Received for review June 2, 1989
Revised manuscript received October 17, 1989
Accepted November 20. 1989
Waste Lubricating Oil Rerefining by Extraction-Flocculation. 2. A
Method To Formulate Efficient Composite Solvents
M. Alves dos Reis* and M. Silva Jeronimo
Faculdade de Engenharia, Departamento de Engenharia Quimica, Rua dos Bragas, 4099 Porto Codex,
A method to design efficient composite solvents is described. The method consists of selecting one
of the components miscible with base oil as the “basic component”. The hydrocarbons and butanone
are possible examples of basic components. The other component or solution of components is
globally treated as the “polar addition”. This is, for example, an alcohol, a ketone, or a solution
of two or more of these compounds. A ternary diagram of waste oil/basic component/polar addition,
where the phase envelope and the curves of constant sludge removal are plotted, summarizes all
information necessary to select the best solvent composition. In all cases studied, addition of 1-3
g/L KOH to the alcohols has increased the sludge and/or additive removal from waste and virgin
oils. Using this method, we have concluded that solvents based on n-hexane and 2-propanol with
3 g/L KOH are very efficient. The weight composition 0.25 waste oil, 0.20 n-hexane, 0.55 2-propanol
is proposed for industrial use.
Treatment of waste oils with polar solvents may be an
interesting alternative to the classical sulfuric acid treat-
ment. This process was named extraction-flocculation by
Reis (1982)because the solvent dissolvesthe base oil and
simultaneously promotes the fast flocculation of the un-
desirable impurities. Since the sludge produced in this
*To whom correspondence should be addressed.
process is organic, it may be mixed with liquid fuels and
burned or, better, find more noble applications. For ex-
ample, it may be used as a component of offset inks (Reis
and Jeronimo, 1982). This seems to overcome the major
problem of the sulfuric acid treatment: the production of
an acid sludge, which is a source of pollution and causes
very difficult disposal problems.
In our previous paper (Reis and Jeronimo, 1988)it was
shown that the flocculating action and subsequent sludge
removal promoted by the one-component solvents studied
0 1990 American Chemical Society