CP302 Vapour-Liquid

Separation Processes

Lecture Notes
2018-19

Dr Iain Burns

1
 CP302 Section A: Vapour-Liquid Separation Processes

Contents
1 Vapour-liquid equilibrium for binary systems _____________________________________ 4
1.1 Composition diagrams for simple binary mixtures ____________________________________ 5
1.1.1 Calculating composition and x-y diagrams __________________________________________________ 6
Dilute solutions ______________________________________________________________________________ 9
1.1.2 Composition diagrams for azeotropes ____________________________________________________ 10
1.2 Volatility ____________________________________________________________________ 11
2 Continuous Multi-Stage Distillation ____________________________________________ 13
2.1 Continuous single-stage separation_______________________________________________ 13
2.2 Continuous multi-stage distillation _______________________________________________ 15
2.2.1 The ideal model of multi-stage distillation ________________________________________________ 16
2.2.2 McCabe-Thiele diagrams ______________________________________________________________ 22
2.2.3 Minimum number of equilibrium stages __________________________________________________ 25
2.2.4 Minimum reflux______________________________________________________________________ 26
2.2.5 Optimal Reflux _______________________________________________________________________ 27
2.2.6 Feed location ________________________________________________________________________ 27
2.2.7 Summary of the McCabe-Thiele method __________________________________________________ 28
2.2.8 Murphree’s tray efficiency _____________________________________________________________ 29
2.2.9 Overall tray efficiency _________________________________________________________________ 29
2.2.10 Mechanical details of trays __________________________________________________________ 30
2.3 Azeotropic and extractive distillation _____________________________________________ 30
2.4 Operating pressure ____________________________________________________________ 31
3 Batch multi-stage distillation _________________________________________________ 32
3.1.1 Constant top product composition ______________________________________________________ 34
3.1.2 Constant reflux ratio __________________________________________________________________ 35

4 Mass transfer across a vapour-liquid interface ___________________________________ 37

4.1 The two-film model ___________________________________________________________ 37
4.1.1 Application of Fick’s First Law ___________________________________________________________ 38
4.1.2 Mass transfer coefficients _____________________________________________________________ 39
4.1.3 Overall Mass transfer coefficients _______________________________________________________ 42

5 Continuous contact distillation ________________________________________________ 45

5.1.1 Properties of packing _________________________________________________________________ 51

6 Gas absorption and Stripping _________________________________________________ 52

Mole ratios _________________________________________________________________________________ 53
xy diagram for Absorption / Stripping ___________________________________________________________ 53
6.1 Continuous single-stage absorption/stripping ______________________________________ 54
6.2 Continuous multi-stage absorption/stripping _______________________________________ 55
6.3 Continuous contact absorption/stripping __________________________________________ 59
6.3.1 Dilute solutions ______________________________________________________________________ 62
6.4 Operating pressure ____________________________________________________________ 62
7 Evaporation _______________________________________________________________ 63
7.1 When does a liquid boil? _______________________________________________________ 63
7.1.1 Boiling-point rise _____________________________________________________________________ 64
7.2 Continuous single-stage evaporation _____________________________________________ 64
Energy balance on condensing steam____________________________________________________________ 65

2
 CP302 Section A: Vapour-Liquid Separation Processes

Material and energy balances on single-stage evaporator ___________________________________________ 65

Heat transfer area ___________________________________________________________________________ 66
7.3 Continuous multi-stage evaporation ______________________________________________ 66
7.3.1 Forward feed ________________________________________________________________________ 66
7.3.2 Backward feed _______________________________________________________________________ 68
7.3.3 Parallel feed_________________________________________________________________________ 68
7.4 Comparison of single and multi-stage evaporators __________________________________ 69
7.5 Further operational considerations _______________________________________________ 70

3
 CP302 Vapour-Liquid Separation Processes

1 Vapour-liquid equilibrium for binary systems

Many separation processes (such as distillation, absorption/stripping and evaporation)
require an understanding of vapour-liquid equilibrium. This subject is not new to you and
has been covered in other modules. You already know how to do a ‘flash calculation’,
which involves the consideration of phase equilibrium.

In a closed system, the number of moles of each component is fixed, and the state of the
system is specified by temperature, T, pressure, P, and mole-fractions xi and yi (where the
index i is used to refer to different components).

The system can be either all vapour, or all liquid, or there can be both vapour and liquid if
the conditions are within the 2-phase region. When we have vapour-liquid coexistence,
the two phases usually have different compositions, giving the potential to use this as a
way of separating mixtures.

For a binary system we have x1 + x2 = 1 (and y1+y2=1).

If we heat a liquid held at constant pressure (e.g. by a piston as illustrated in Figure 1), it
will eventually reach its ‘bubble point’, the temperature at which vapour first appears.
Likewise, if a vapour is cooled at constant pressure, it will reach its ‘dew point’, when the
first droplet of liquid is condensed.

Continuing to consider a binary system we can work in terms of one of the more volatile
component and drop the subscripts. e.g. for a binary mixture of methanol and water,
xmethanol is called ‘x’ and ymethanol is called ‘y’.

Figure 1

The behaviour of the system can be plotted on a two-dimensional graph if pressure or

temperature is held constant.

4
 CP302 Vapour-Liquid Separation Processes

1.1 Composition diagrams for simple binary mixtures

P
liquid
• Pressure-composition diagram
(T fixed)
phase
2-
gas
0 1
Figure 2a
x

T
gas

• Temperature-composition diagram 2-phase

(P fixed)

liquid
0
x 1
Figure 2b

• x-y diagram
(P or T fixed, but not both)

0 1
Figure 2c x

In Figure 2c, x and y refer to the mole fraction of the more volatile component in the liquid
and gas phases, respectively.

Composition diagrams, and x-y diagrams in particular, are crucial for design of separation
processes - we need to understand how they originate from the principles of phase
equilibrium.

5
 CP302 Vapour-Liquid Separation Processes

1.1.1 Calculating composition and x-y diagrams

Our task for vapour-liquid equilibrium (VLE) in a binary system is either:

1. find y and P given x and T

OR
2. find y and T given x and P

At equilibrium, by definition, temperature, T, pressure, P, and chemical

potential, µ, are constant throughout a system. By equating the vapour
and liquid chemical potentials for each component of the system (c.f.
Atkins, Physical Chemistry), we obtain for an ideal solution,

Pyi = Pi* xi

This is familiar to us as ‘Raoult’s Law’. Gas

(P,T,y)
Pi* is the pure component vapour pressure of component i at temperature
T. Thus, Raoult’s Law allows us to estimate the partial pressures exerted
by a solution based on the pure component properties of its
constituents. As an example, the pure component vapour pressure of Liquid
water can be looked up in ‘steam tables’. (P,T,x)
Figure 3
An ideal solution is one in which there is negligible interaction (attraction
and repulsion) between the molecules.

Unfortunately, the ideal solution assumption is not sufficient to capture the behaviour of
many common mixtures (e.g. alcohols in water).

The non-ideality is described by the activity coefficient, γi, for component i, which is
dependent on attractive and repulsive interactions of molecules in solution and is a
function of composition. Thus, equating the chemical potentials of vapour and liquid at
equilibrium:

Pi*γ i
Pyi = Pi* xiγ i or: yi = xi (1)
P

We need to find Pi* and γ i so that we can relate the compositions of the vapour and liquid
phases.

6
 CP302 Vapour-Liquid Separation Processes

1.1.1.1 Estimating pure component vapour pressure

Pure component vapour pressure can be found from a pure fluid ‘equation of state’ (EOS),
such as the van der Waals, Peng Robinson etc.

Alternatively, there are engineering equations for the saturation pressure, such as the
Antoine equation, which we will use here, e.g.

B
Antoine’s equation: lnP * = A − (2)
T +C

Here, A, B and C are empirical constants that are different for each pure fluid and T is the
temperature. The values of A, B and C depend on the pressure and temperature units
being used, and the base of the logarithm, which must therefore always be specified.
Various textbooks (e.g. McCabe, Smith, Harriot, Unit Operations of Chemical Engineering)
contain tables of Antoine coefficients.

1.1.1.2 Estimating activity coefficients

For an ideal liquid mixture we have γi = 1 for each component. Then (1) becomes

Pi *
Raoult’s Law: yi = xi (3)
P

Unfortunately, most liquid mixtures are not close to ideal.

There are several engineering equations for estimating activity coefficients, for example:

Margules equation: ln γ i = x 2j (Aij + 2 x i (A ji − Aij )) (4)

where Aij and Aji (i ≠ j) are dimensionless Margules coefficients that are different for each
binary mixture.

The Margules coefficients are in principle a function of temperature and pressure but they
can fortunately be approximated as constant over sufficiently small ranges of P and T.

The Margules equation is valid only for vapour-liquid equilibria of binary mixtures.

More complex and versatile equations for activity coefficients include the van Laar, Wilson,
Non-Random Two Liquid (NRTL), and the Universal Quasi Chemical (UniQuaC) equations
(note that software such as Aspen gives the user a choice of methods to use for
calculating phase equilibrium). But we will not use these equations here.

7
 CP302 Vapour-Liquid Separation Processes

1.1.1.3 Calculating the temperature

We need to determine T and y given x and P by applying equations (1), (5) and (4). Note
that this would have to be done repeatedly to build up an x-y diagram.

If x and P are given, T can be found iteratively as follows:

1. Calculate γ for each component from (4) (remember, γ is approximately constant
only over a sufficiently small range of P and T).
2. Estimate T.
3. Calculate P* for each component from (2).
4. Calculate yi for each component from (1).
5. If y1 + y2 ≠ 1 then adjust T and repeat from 3. If y1 + y2 < 1 then T is too low and
vice-versa.

The x-y diagram in Figure 4 was generated in precisely this way using:
Antoine coefficients (P in mmHg, T in K, log to base e):
Methanol: A = 18.588, B = 3626.6, C = -34.29
Water: A = 18.304, B = 3816.4, C = -46.13
Margules coeficients:
Awm = 0.6174, Amw = 0.7279

x-y diagram for methanol/water at 1 atm

1
ym (molefraction of methanol in vapour)

0.8

0.6

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1
xm (molefraction of methanol in liquid)

Figure 4

On the other hand, if x and T are given, P can be found straight away by combining
equation (4) with x1 + x2 = 1 and y1 + y2 = 1 to give

P = Py 1 + Py 2 = P1* x1γ 1 + P2* (1 − x1 )γ 2 (5)

And once P is found, all the other variables can also be calculated.

8
 CP302 Vapour-Liquid Separation Processes

Dilute solutions

Note, for a dilute component in a liquid mixture (a solute), the activities can be
approximated based on xi≈0 and xj≈1.

Substituting into Margules equation for this dilute situation:

ln γ i = x 2j (Aij + 2 xi (A ji − Aij )) = Aij

γ i = exp(Aij )
ln γ j = xi2 (A ji + 2 x j (Aij − A ji )) = 0
γ j =1

This means the partial pressure of the solvent (j) is equal to its pure component vapour
pressure at temperature T. In the dilute limit, the solute molecules undergo no interaction
with each other since they are so scarce, meaning that the activity coefficient, γi, is
independent of composition for xi→0.

Together with (1) this gives:

Henry’s law: Pi = Pyi = H ij xi (6)

where Hij (Hij =Pi* γi ) is Henry’s Law constant for component i

Notice also that the pressure of the system in the dilute limit is the pure component vapour
pressure of the solvent at T, i.e. P=Pj*(T). e.g. a very dilute solution of methanol in water in
VLE at 1 atm must have a temperature of approximately 100 ºC.

By examining the x-y diagram for methanol and water, shown in Fig 4, we see that Henry’s
Law approximation becomes inaccurate when the methanol mole fraction exceeds a few
percent.

9
 CP302 Vapour-Liquid Separation Processes

1.1.2 Composition diagrams for azeotropes

An azeotrope is a fluid mixture that has liquid and vapour of the same composition at
equilibrium. There are many examples of azeotropes: one is ethanol and water; another is
propanol and water.

The azeotrope is a consequence of highly non-ideal solution behaviour resulting from

strong interactions (attraction or repulsion) between molecules in solution.

Figure 5 shows the x-y diagram for a mixture of propanol and water. Its azeotropic point
occurs at xp = 0.42 (the azeotropic point for ethanol/water mixtures occurs at xe = 0.89).
Figure 5 shows how temperature varies with composition for this mixture. In this case the
mixture is said to be a minimum boiling point azeotrope. Maximum boiling point azeotropes
are also possible.

Azeotropes cause difficulties for designers of separation processes because such binary
mixtures cannot be distilled ‘through’ the azeotropic point. The reasons for this will become
clearer when we consider distillation in detail. Special distillation process tricks are needed
to ‘break’ the azeotrope in this case, which can involve the addition of a third component.

x-y diagram for propanol/water at 1 atm

1
yp (molefraction of propanol in vapour)

0.8

0.6

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1
x p (molefraction of propanol in liquid)

Figure 5

10
 CP302 Vapour-Liquid Separation Processes

Temperature/composition diagram for

propanol/water at 1 atm
102

100

98
Temperature (oC)

96

94

92

90

88

86
0 0.2 0.4 0.6 0.8 1
mole fraction (liquid (x) and vapour (y))

Figure 6

1.2 Volatility
Vapour-liquid separation processes are dependent on a difference in composition between
the vapour and the liquid at equilibrium. One way of characterising this difference is via a
parameter called ‘volatility’, Ki, defined as:

yi
Ki = (7)
xi

Clearly, a significant difference in volatilities will make a binary mixture easier to separate.
The

The relative volatility, αij, is simply:

Ki
α ij = (8)
Kj

Figure 7 shows the volatilities and relative volatility for a methanol/water mixture at 1
atmosphere. This figure can be calculated from the data in Figure 4. Mixtures like this are
separated by distillation.

Absorption/stripping and evaporation are used when one of the components in the liquid is
effectively non-volatile.

11
 CP302 Vapour-Liquid Separation Processes

volatility for methanol/water at 1 atm

8 8

7 7

6 6

αmw (dashed line)

5 5
K (full lines)

4 4

3 methanol 3

2 2

1 water 1

0 0
0 0.2 0.4 0.6 0.8 1

xm (molefraction of methanol in liquid)

Figure 7

12
 CP302 Vapour-Liquid Separation Processes

2 Continuous Multi-Stage Distillation

We now know enough about vapour-liquid equilibrium to understand multi-stage distillation
processes.

Distillation is used to separate mixtures where at least two of the components are present
(in non-negligible amounts) in both the vapour and the liquid.

We will consider four types of distillation process:

• Continuous single stage: we will learn how to predict output streams given a
continuous input to a single equilibrium stage.
• Continuous multi-stage: a multi-stage process consists of a column containing
stacked ‘plates’ (stages). Input, in the form of a continuous ‘feed’ stream, is
separated to produce two continuous output streams, the ‘top’ and ‘bottom’
products. We will base our analysis on an idealised model of this system.
• Batch multi-stage: similar to continuous multi-stage above, except that the
process only produces a ‘top’ product rich in the most volatile component – no
bottom product is produced and there is no feed stream. Batch distillation is a
dynamic process (not steady-state).
• Continuous contact distillation where distillation occurs in a ‘packed’ tower.
Whereas the first three ‘stage’ processes above require only an understanding
of vapour-liquid phase equilibrium, to analyse this ‘continuous contact’
process we will also need to apply our knowledge of mass transfer across a
vapour-liquid interface.

When designing distillation columns, we want to understand:

• In the case of multi-stage distillation (continuous or batch), how many stages
are required?
• In the case of continuous contact distillation, how high does the packed
column need to be?

2.1 Continuous single-stage separation

You have already studied separation via a single equilibrium stage, sometimes called
‘flash separation’. Consider a vessel fed with F moles/s at overall composition xf (Figure
11). Our task is to calculate the compositions (x and y) of the exiting liquid and vapour
phases, which are assumed to be in equilibrium at pressure P.

V,y
F,xf P
in equilibrium
L,x

Figure 11 heat

13
 CP302 Vapour-Liquid Separation Processes

Material balances (total and component) give: Fx f = Vy + ( F − V ) x , which when rearranged

gives the ‘operating line’

Fx f (F − V )
y= − x (9)
V V

The operating line is a relationship between the liquid and vapour compositions (x and y)
resulting from the material balance.

The vapour and liquid compositions are at equilibrium, giving us a second equation y=f(x).
So we can use the x-y diagram at pressure P. Figure 12 is an x-y diagram for
ethanol/water at 1 atm.

We can draw the operating line on the graph (note: we need to know V (or L) to draw the
operating line). The intersection of the two lines is the solution of the two simultaneous
equations. Thus, we have determined the compositions of the vapour and liquid leaving
the separator.

An alternate way to pose the problem would be if the temperature and pressure in the
separator are known. Then x and y can be determined from vapour-liquid equilibrium and
a material balance (Eqn 29) can be used to find V and L.
single-stage distillation of ethanol/water at 1 atm
1

0.9
ye (molefraction of ethanol in vapour)

0.8
Fx f / V
0.7
slope = (V - F) / V
0.6

0.5

0.4

0.3

0.2
xf
0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

xe (molefraction of ethanol in liquid)

Figure 12

14
 CP302 Vapour-Liquid Separation Processes

2.2 Continuous multi-stage distillation

Multi-stage distillation (Figure 13) can achieve much better separation than a single
equilibrium stage.

A ‘plate’ distillation tower works as follows. Feed can be entered, and product withdrawn,
from any stage. Heat is provided by the reboiler, which generates vapour. The vapour
rises up through perforations in the trays, while liquid flows across the trays and down the
downcomers.

Vapour exiting the top stage is condensed by the condenser. Some of it is returned to the
top plate as liquid ‘reflux’ while the rest exits as top product. The plates have weirs to
maintain a level of liquid on the surface. Some liquid leaves the reboiler as bottom product.

The liquid and vapour compositions progressively change as we move up the column. At
each stage, some of the less volatile component in the vapour condenses and a
corresponding molar amount of the more volatile component evaporates.

The temperature rises as we move from the top to the bottom of the column.

To simplify matters we will only consider columns where:

• Only one feed stream is entered at one of the ideal plates.
• Product is withdrawn from the condenser and the reboiler only.

heat

condenser

top reflux
product

feed

bottom still
product

Figure 13 heat

(Note: this figure shows a column with an internal reboiler. Many distillation columns have
external reboilers but the analysis we will perform here applies equally in either case).

15
 CP302 Vapour-Liquid Separation Processes

2.2.1 The ideal model of multi-stage distillation

Real distillation towers are very complex - they can be modelled with sophisticated
software. Process simulation packages such as Aspen can help to do this, if used
correctly. We will study an ideal model to understand the fundamental principles.
No new concepts are needed to go from the ideal model to reality: only more complex
mass and heat balance equations and equilibrium relations are needed. We will restrict our
treatment to binary mixtures. This special case is sometimes called the McCabe-Thiele
model.

For specified feed and product compositions, feed rate, and operating pressure, we will
see that the only other significant design parameters are the ‘reflux ratio’, R, which
controls the liquid reflux rate, and the feed location. This information is enough to
determine the number of ideal stages needed.

We will find that:

• For total reflux (R = ∞), the number of plates is minimised but operating costs are
infinite in this case.
• There is a minimum possible reflux ratio (R = Rmin) for which the number of plates is
infinite and hence capital costs are infinite.

2.2.1.1 Model assumptions

We will make use of the following assumptions:
• Pressure is approximated to be constant throughout the tower. This allows us to
use the same x-y diagram for all stages. In a real tower there will be a small
pressure gradient.
• We will assume that the only heat input is from the reboiler, and the only heat
loss is from the condenser. So our column is assumed to be perfectly lagged.
This is quite an accurate approximation for a well-insulated column.
• We will consider steady-state operation for continuous distillation.
• We will assume that equilibrium is reached on each plate (each stage), and also
in the reboiler. This is often the weakest approximation and great effort is made
in plate design to reach equilibrium as closely as possible.
• We will assume the enthalpy for each phase is constant on the vapour-liquid
phase boundary. This means that all saturated vapour in the column has the
same molar enthalpy, hv, regardless of composition. The same applies to the
liquid molar enthalpy, hl. This is, of course, not necessarily accurate, but it
simplifies the analysis and often gives reasonable results.

16
 CP302 Vapour-Liquid Separation Processes

2.2.1.2 Overall performance

heat

D, xd

F, xf

W, xw

heat

Figure 14
First, consider the overall inputs and outputs to the system (Figure 14). F, D and W are
flow rates, usually expressed as moles per second. The compositions are expressed as
mole fractions. Total and component balances for the input and output streams gives:

F = D +W ; Fx f = Dx d + Wxw (10)

which, after rearranging, gives:

 x − xd   x − xf 
W = F  f  ; D = F  w  (11)
x
 w − x d   xw − xd 

Sometimes, the ‘recovery’ is specified.

Dx d
rec = (12)
Fx f

There are n ideal equilibrium plates, the top plate is labelled n, then n - 1 etc to 1, which is
the reboiler.

2.2.1.3 Material balance over the condenser

Vapour, with composition yn, rising from the top plate is condensed to liquid, of which D
becomes top product and RD is returned to the top plate as reflux (Figure 15).

heat

D xd
condenser

Vn yn Ln+1 xn+1
Figure 15

17
 CP302 Vapour-Liquid Separation Processes

A material balance (with xn+1 = xd) gives

Vn = Ln +1 + D = D(R + 1) and Vn y n = D(R + 1) x d (13)

So we can re-draw this stage as in Figure .

heat

D xd
condenser

D(R+1) xd DR xd
Figure 16

2.2.1.4 Ideal equilibrium plate

Consider the ith ideal equilibrium plate (Figure 17). A material balance gives

Vi + Li = Vi −1 + Li +1 and Vi y i + Li x i = Vi −1 y i −1 + Li +1 x i +1 (14)

So we can relate yi to xi+1 if we know Vi and Li+1.

Figure 17

To calculate Vi and Li+1 we must examine the energy balance across an ideal plate.
Because there is no heat transfer or work done, this is really the same as an enthalpy
balance (Figure 18).

Vi hi,v Li+1 hi+1,l

ith equilibrium
plate
Figure 18 Vi-1 hi-1,v Li hi,l

Let the molar enthalpy of vapour and liquid streams be hi,v and hi,l respectively. Under
steady-state conditions we must have

Vi hi ,v + Li hi ,l = Vi −1hi −1,v + Li +1hi +1,l (15)

18
 CP302 Vapour-Liquid Separation Processes

Assuming that the enthalpy is constant on the vapour-liquid phase boundary, i.e. hi ,v = hi +1,v
and hi ,l = hi +1,l (see Figure 19), gives

hv (Vi − Vi −1 ) = hl (Li +1 − Li ) (16)

Figure 19
Substituting in (14) then gives

hv (Vi − Vi −1 ) = hl (Vi − Vi −1 ) (17)

Since hv ≠ hl, this is satisfied only if Vi = Vi-1, and hence from (16) Li = Li+1.

This tells us that the vapour and liquid flow rates above the feed point are constant (but
with varying compositions); this is known as ‘constant molar overflow’.

Since we know the vapour reaching the condenser is D(R + 1), and the liquid reflux is DR,
we automatically know Vi and Li for all plates above the feed point. So for an ideal
equilibrium plate above the feed point we can re-draw Figure 17 like this:

D(R+1) yi DR xi+1
ith equilibrium
plate

Figure 19 D(R+1) yi-1 DR xi

Because we now know Vi and Li+1 we can find a relationship between yi and xi+1 by
performing a material balance between stage i and the top of the column:

D(R + 1)y i = DRx i +1 + Dx d

which can be rearranged to give:

xd R
yi = + x i +1 (18)
R +1 R +1

This allows us to define an ‘operating line’ for plates above the feed point.

19
 CP302 Vapour-Liquid Separation Processes

The liquid and vapour leaving each stage have been assumed to be at equilibrium:

xi = f (y i ) (19)

Together, the operating line and the equilibrium line can be represented on a diagram to
understand how the ‘rectification’ stage, i.e. the portion of the tower above the feed point,
works (Figure 20). We draw the rectification operating line by starting from point (xd, xd)
and drawing a line with gradient (R+1).

Starting at the distillate (xn+1, yn) = (xd, xd) we can:

1. Use the equilibrium relationship to find xn from yn

2. Use the operating line to get yn-1 from xn
3. Use the equilibrium relationship to get (xn-1, yn-1)
4. Use the operating line to get (xn-1, yn-1)
5. etc, etc, until we reach the feed plate.

rectification of ethanol/water at 1 atm

1
ye (molefraction of ethanol in vapour)

0.8

xn, yn
xn-1, yn-1 xd, yn
0.6
xn, yn-1

etc xn-1, yn-2

0.4

rectification
operating line
0.2

0
0 0.2 0.4 0.6 0.8 1

xe (molefraction of ethanol in liquid)

Figure 20

To draw the ‘stripping’ section on this diagram, i.e. that portion of the tower below the feed
point, we first need to consider the feed point itself.

2.2.1.5 Feed plate

A feed plate is identical to an ideal equilibrium plate, except that we have some feed as
input, which can be either vapour, or liquid, or both at temperature Tf (Figure 21).

20
 CP302 Vapour-Liquid Separation Processes

D(R+1), yi DR, xi+1

F, xf ith equilibrium
and feed plate

Vs, yi-1 Ls, xi

Figure 21

Material and enthalpy balances yield

D(R + 1) + Ls = DR + Vs + F
Vs = D(R + 1) + Ls − DR − F (20)

D(R + 1)hv + Ls hl = DRhl + Vs hv + Fhf (21)

As before, we have assumed fixed enthalpy on the vapour-liquid phase envelope, i.e. hi,v =
hi+1,v and hi,l = hi+1,l, so (21) becomes

hv (D (R + 1) − Vs ) = hl (DR − Ls ) + Fhf (22)

Substituting in (20) gives

hv (DR − Ls + F ) = hl (DR − Ls ) + Fhf

DR(hv − hl ) + F (hv − hf ) = Ls (hv − hl )

 h − hf 
⇒ Ls = DR + F  v 
 hv − hl 

⇒ Ls = Lr + Fq (23)

where

(hv − hf )
q= (24)
(hv − hl )

21
 CP302 Vapour-Liquid Separation Processes

So this relates the liquid flow rate below the feed plate to that above it. Substituting into
(20) gives the vapour stream below the feed plate

Vs = Vr + F(q − 1) (25)

So we can re-draw the feed plate stage as in Figure 22.

D(R+1), yi DR, xi+1

F, xf ith equilibrium
and feed plate

Figure 22 D(R+1)+F(q-1), yi-1 DR+Fq, xi

See how the feed is divided between the liquid and vapour streams going to the
equilibrium plate below.

Now that we know Vs and Ls, we can determine the operating line for the stripping section
by performing a material balance between stage i and the bottom of the column:

yi −1[D(R +1) + F(q −1)] = (DR+ Fq)xi −Wxw

(DR+ Fq) W
yi −1 = xi − xw (26)
D(R +1) + F(q −1) D(R +1) + F(q −1)

2.2.2 McCabe-Thiele diagrams

We can also represent the stripping operating line on a diagram to understand how the
whole distillation tower works. For our specialized system of a binary mixture in a
distillation tower with one feed, the complete diagram is called a McCabe-Thiele diagram
(Figure 23).

The number of ideal equilibrium stages, n, is the number of points on the equilibrium
curve. Note that the reboiler is an equilibrium stage, so the number of ideal plates is n-1.

In the ethanol/water example above the final point (xw, yw) has exceeded the design
specification. However, we cannot have ‘fractions’ of an ideal stage, nor can we fail to
meet the design specification, so we are left with a distillation tower that has exceeded its
specification.

22
 CP302 Vapour-Liquid Separation Processes

We could draw the stripping operating line by starting at point (xw,xw) and drawing a line
(DR + Fq )
with gradient but there is a more straightforward approach based on the
D(R + 1) + F (q − 1)
‘q-line’.

We see that the slopes of both operating lines depend on the reflux ratio, R.

distillation of ethanol/water at 1 atm

1
ye (molefraction of ethanol in vapour)

0.8
x5, xd
x4, y4 xd, xd
x3, y3 x5, y4
x2, y2 x4, y3
0.6
x3, y2
x1, y1
x2, y1 rectification
0.4 operating line

q-line
0.2 stripping
operating line
xw, yw x1, yw
0
xw xf xd
0 0.2 0.4 0.6 0.8 1

Figure 23 xe (molefraction of ethanol in liquid)

2.2.2.1 The q-line

The q-line defines the locus of intersection of the operating lines for the complete range of
R. So let us find this intersection. The operating lines intersect at (xq, yq) where:

Vr y q − Lr x q = Dx d
(27)
Vs y q − Ls x q = −Wxw

Vr and Lr are the vapour and liquid flow rates above the feed plate. Vs and Ls are those
below. Combining these equations gives

y q (Vr − Vs ) − x q (Lr − Ls ) = Dxd + Wxw (28)

We know that Vs = Vr +F(q - 1) and Ls = Lr +Fq. Inserting these in (28) gives

y q F (q − 1) − x q Fq = −Fx f (29)
i.e.

q x
y q = xq − f (30)
q −1 q −1

This defines a line with slope q / (q -1) that intersects the y = x line when xq = yq = xf.

23
 CP302 Vapour-Liquid Separation Processes

So to draw the q-line, we start from the point (xf,xf) and draw a line with gradient q/(q-1).

There are 5 distinct scenarios for q = (hv − hf ) (hv − hl ) . They are:

Feed composition Temperature Enthalpy q slope

superheated vapour Tf >> Ti hf > hv q<0 +ive
saturated vapour Tf > Ti hf = hv q=0 0
mixture of vapour and liquid Tf ~ Ti hv < hf < hl 0<q<1 -ive
liquid at bubble point Tf < Ti hf = hl q=1 ∞
liquid below bubble point Tf << Ti hf < hl q>1 +ive

Table 3

This information can be nicely represented on a diagram, as shown in Figure 24.

sub-cooled
liquid at liquid
bubble pt
y vapour
and liquid

vapour at
dew pt

Superheated
vapour

Figure 24 x
So we can construct a McCabe-Thiele diagram as follows:
• Draw the two product points, (xd, xd) and (xw,xw).
• Draw the rectification operating line, which starts at (xd, xd) and has slope
R /(R + 1).
• Draw the q-line, which starts at (xf, xf) and has slope q /(q - 1).
• Draw the stripping operating line between (xw, xw) and the intersection of the
q-line and the rectification operating line.
• Step-off the ideal equilibrium stages, starting at (xd, xd), switching operating
lines at their intersection.

24
 CP302 Vapour-Liquid Separation Processes

2.2.3 Minimum number of equilibrium stages

It is worthwhile to consider limiting cases of column operation. The maximum value of
reflux ratio, R, is ∞, i.e. total reflux. At total reflux (Figure 25):
• There is no product, i.e. D = 0, hence W = F = 0.
• Since F = 0, Vr = Vs = Lr = Ls.
• The slope of the operating lines is 1, so they lie on the y = x line.
• The number of equilibrium stages is minimised.

distillation of ethanol/water at 1 atm

1 and at total reflux
ye (molefraction of ethanol in vapour)

0.8

0.6

0.4

0.2

0
xw xf xd
0 0.2 0.4 0.6 0.8 1
Figure 25 xe (molefraction of ethanol in liquid)

We can therefore determine the minimum number of ideal plates, nmin graphically as
shown in Figure 25, by stepping-off the stages to the line y=x. In the case of this example,
nmin=4.

As an aside, note that some textbooks refer to an alternative way to estimate nmin, known
as Fenske’s method. A limitation, however, is that the method is an approximation based
on an average (geometric mean) of the relative volatility. For this reason, we concentrate
instead on graphical explanation above, which provides a neater illustration of the
principle.

Clearly it wouldn’t be practical to operate a column with the minimum number of stages
since it can only achieve the required separation at total reflux, R=∞. We need to establish
a way of choosing the reflux ratio.

25
 CP302 Vapour-Liquid Separation Processes

2.2.4 Minimum reflux

Let’s consider the opposite limiting case. As R is reduced from ∞ the slope of the
rectification operating line reduces. Eventually, as R is reduced, the rectification operating
line will intersect the equilibrium curve at, or before, its intersection with the q-line. When
this happens the number of equilibriums stages, n = ∞ (Figure 26).
distillation of methanol/water at 1 atm
1
and at R min
ym (molefraction of methanol in vapour)

0.8

0.6

0.4

0.2

0
Figure 26 0 0.2 0.4 0.6 0.8 1
xm (molefraction of methanol in liquid)

We can determine the minimum reflux Rmin by drawing the rectification operating line to
intersect the point where the q-line cuts the equilibrium line (as shown in Figure 26), and
reading off the gradient of this line (remember, this can be done by finding where it would
cross the y-axis).

Note that for some mixtures, e.g. ethanol/water at 1 atmosphere, Rmin does not always
occur at the intersection of the q-line with the equilibrium curve, because of the shape of
the equilibrium curve (Figure 27).
distillation of ethanol/water at 1 atm
1 and at Rmin
ye (molefraction of ethanol in vapour)

0.8
intersection with
equilibrium curve

0.6

0.4

0.2

0
0
xw xf 0.5 xd 1
xe (molefraction of ethanol in liquid)
Figure 27

26
 CP302 Vapour-Liquid Separation Processes

2.2.5 Optimal Reflux

Total reflux leads to a minimum number of ideal plates, but is associated with infinite
operating costs since the reboiler duty must also be infinite. Minimum reflux lead to an
infinite number of ideal plates, hence capital costs are infinite, although operating costs are
low because less heat is needed by the boiler to drive the distillation tower.

Somewhere between these two extremes lies the optimal number of ideal plates.
However, we cannot thoroughly quantify the optimal reflux without some cost information,
which is beyond the scope of this course. Nevertheless, under ‘normal’ conditions it is
often found that the optimal reflux ratio happens to be between 1.2 and 1.5 times the
minimum reflux ratio, which is therefore a good starting point in a preliminary design.

Even if the distillation tower is optimised within its specification, this might not be optimal
overall for the whole process plant, i.e. the specification of the distillation tower should also
be optimised. This will require a cost model of the whole plant.

2.2.6 Feed location

The most efficient place to input the feed is at the point where the operating lines (and the
q-line) cross. This can be seen by inspection (Figure 28 a and b). This is a result of the
following fact; it is always more efficient to ‘be on the operating line’ that is furthest from
the equilibrium curve.
distillation of methanol/water at 1 atm
1 with sub-optimal feed location
a
ym (molefraction of methanol in vapour)

0.8

0.6

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1
xm (molefraction of methanol in liquid)

distillation of methanol/water at 1 atm

1 with optimal feed location
b
ym (molefraction of methanol in vapour)

0.8

0.6

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1
Figure 28
xm (molefraction of methanol in liquid)

27
 CP302 Vapour-Liquid Separation Processes

2.2.7 Summary of the McCabe-Thiele method

D, xd
 x − xf 
D = F  w 
 xw − xd 
Vr = D(R + 1) Lr = DR
W =F−D
…
F, x f
F, zf
q=
(hv − hf )
(hv − hl )
…

Vs = Vr + F(q - 1) Ls = Lr + Fq Dx d
Recovery =
Fx f
W, xw
Figure 29

distillation of methanol/water at 1 atm

1 with optimal feed location
ym (molefraction of methanol in vapour)

0.8

0.6
Slope = Lr / Vr

0.4

Slope = q / (q - 1)

0.2
Slope = Ls / Vs

0
0 xw 0.2 0.4
xf 0.6 0.8 xd 1
xm (molefraction of methanol in liquid)

Figure 30

28
 CP302 Vapour-Liquid Separation Processes

2.2.8 Murphree’s tray efficiency

The McCabe-Thiele approach employs several approximations. Usually the worst

approximation is that equilibrium is achieved on each plate. To make theory fit reality, an
empirical adjustment can be made to the McCabe-Thiele diagram based on the concept of
Murphree’s tray efficiency (Figure 31).

If the tray efficiency is quoted at 50%, then a new ‘effective equilibrium curve’ is drawn half
way between the old one and the operating lines. What does ‘half way’ mean? If tray
efficiencies are based on the;
• vapour phase then it means half of the vertical distance between the
equilibrium curve and the operating lines
• liquid phase then it means half of the horizontal distance between the
equilibrium curve and the operating lines
The lower the efficiency, then the closer the effective equilibrium curve and the operating
lines become.

distillation of methanol/water at 1 atm

1 with Murphree's efficiency curve
ym (molefraction of methanol in vapour)

0.8

0.6

0.4

0.2

0
0
xw 0.2 0.4
xf 0.6 0.8
xd 1
xm (molefraction of methanol in liquid)
Figure 31

2.2.9 Overall tray efficiency

This is simply the ratio of the estimated number of ideal equilibrium plates to the actual
number of plates.

Number of ideal plates

E0 = (31)
Number of actual plates

29
 CP302 Vapour-Liquid Separation Processes

2.2.10 Mechanical details of trays

Vapour enters a tray from below through holes in the bottom of the tray and forms bubbles
in the liquid above. So vapour is dispersed in the liquid. These holes can be open (sieve
trays), fitted with flaps that close if the gas pressure is too low (valve trays), or fitted with
‘bubble caps’.

Liquid enters a tray from above via a ‘down-comer’. It flows across a tray, and is
permeated by uprising bubbles of vapour. If the liquid is of sufficient depth it flows over a
weir and down the next down-comer.

Tray efficiency depends on the extent to which vapour and liquid have reached
equilibrium. This is influenced by the gas-liquid interfacial area (the number and size of
bubbles) and the lifetime of a bubble in the liquid (which in turn is influenced by the density
and viscosity of the liquid, turbulence, and liquid depth) relative to the rate of mass transfer
across the interface (which is influenced by viscosity, turbulence, and gas-liquid
concentration differences, i.e. relative volatility).

Increasing the gas flow rate will result in increased turbulence, but too high a gas flow rate
may result in frothing, foaming and entrainment of liquid in the gas, i.e. liquid could be
carried by the gas flow backwards to the tray above (back-mixing is to be avoided).

2.3 Azeotropic and extractive distillation

Azeotropic mixtures cannot be distilled ‘through’ the azeotropic point because vapour and
liquid have the same composition at this point. We find a similar problem with mixtures that
have low relative volatilities, i.e. when the equilibrium curve nearly lies on the diagonal in
an x-y diagram.

To see how these mixtures, where the equilibrium curve is close to or touching the
diagonal, can be dealt with consider the equilibrium relation:

Pi * xi γ i
yi = (32)
P

This tells us that at fixed pressure we can change the equilibrium relationship if we can
change the activity coefficients, γi. We might also be tempted to change P, but remember
that at fixed xi, T and hence Pi*, will also have to change, with the net result being that yi
may change very little.

However, the activity coefficients can be altered significantly by addition of a third

component. To be useful, this third component (C) will be much less volatile than the other
components (A and B), and will act on the other two components in different ways,
resulting in a clear change in equilibrium behaviour.

A distillation of the ternary mixture will separate the original components, with the bottom
product being rich in C. Because the volatility of C is very different it can be separated
(from either A or B) using another distillation tower. C is then recycled for use in the first
distillation tower.

30
 CP302 Vapour-Liquid Separation Processes

Multi-component distillation in beyond the scope of this course, so we will not consider
these processes further.

AB
B
BC

Figure 44

2.4 Operating pressure

The choice of operating pressure will be made on economic grounds. There are several
factors to consider;
• Operation at pressures outside the range of 1-10 bar will tend to significantly
increase capital and operating costs because of pumping, strengthening
vessels and pipework etc.
• For non-ideal mixtures the relative volatility can change with pressure. So the
design (number of plates etc.), and hence the overall cost, of the column
might be different at different operating pressures.
• The temperature profile of the tower will be different for different operating
pressures. Some fluid mixtures might be sensitive to temperature, so to
maintain product quality the design specification might constrain the
maximum temperature, and hence the maximum pressure. Ultimately, the
tower might have to run under vacuum. An example of this is the Acrylic
Acid Process (c.f. Turton book), which you may have seen in the Process
Design class.

31
 CP302 Vapour-Liquid Separation Processes

3 Batch multi-stage distillation

With batch (multi-stage) distillation there is no continuous feed and no bottom product.
The still is filled at the start and heat is provided. All the vapour leaving the top of the
column is condensed and a part is returned as reflux, while the remainder is distillate (top
product).

heat

condenser Reasons to use batch instead of

continuous multi-stage distillation?
top reflux
product • Convenience
• Easy engineering
• Versatility
• Tradition
• May accompany batch
reactor

still
Figure 32
heat

Thus, this is not a steady-state process. The compositions of the liquid in the still and of
the distillate will change continuously during the distillation process.

However, to model batch distillation we will assume that the rate at which the conditions
change is slow enough to justify the pseudo-steady approximation. We have already
encountered this approximation in our study of mass transfer.

By making the pseudo-steady approximation we can use what we have learnt from
continuous multi-stage distillation and apply it immediately.

Distillate is collected in a storage vessel. Here we need to adopt a nomenclature to

distinguish the molar amount of distillate collected, D (with composition xd) from the
distillate molar flow rate, D (whose composition is xd). Similarly, the molar amount of liquid
in the still is defined as W and its composition is xw.

It’s important to note that W is NOT a flow rate and xw is NOT the concentration of a
stream. Instead, W is a total amount, and xw is its composition.

32
 CP302 Vapour-Liquid Separation Processes

During a batch distillation process the number of ideal plates is fixed, so for a particular xw
the product quality (xd) is controlled by R.

D, xd

…
Vr = D(R + 1) Lr = DR

W, xw

Figure 33
batch distillation of methanol/water at 1 atm
1 and 3 ideal plates
ym (molefraction of methanol in vapour)

0.8

0.6

0.4

Slope = Lr / Vr

0.2

0
0 xw 0.2 0.4 0.6 0.8 xd 1
xm (molefraction of methanol in liquid)

Figure 34

During the distillation process xw will decrease, as the more volatile component is
evaporated off. There are two ways of operating a batch distillation process;
• Constant xd ⇒ varying R.
• Constant R ⇒ varying xd.
We will examine each option in turn.

The molar amount of distillate collected during a given interval of time is as follows:

t1

D1 − D 0 = ∫ D(t ′) dt ′ (33)
t0

where D(t’) is the distillate flowrate at time, t’.

33
 CP302 Vapour-Liquid Separation Processes

Total and component balances performed between times t0 and t1, where D0 = 0, give:

W 0 − W 1 = D1 ; W 0 x w 0 − W 1 x w 1 = D1 x d1 (34)

Hence:

 x − xw1 
D1 = W 0  w 0  (35)
 x d1 − x w1 

3.1.1 Constant top product composition

R is varied to keep xd constant. We can calculate this variation of R as follows:
• For several values of R:
Find xw(R) from a McCabe-Thiele diagram for fixed n and xd.
Since xd1 = xd1 when the top product concentration is constant, we can
calculate D1(R) from (35).

1 batch distillation of methanol/water at 1 atm

and 3 ideal plates and x d fixed
ym (molefraction of methanol in vapour)

0.8

0.6

0.4

0.2

0
0 xw1 xw0 0.2 0.4 0.6 0.8
xd 1

xm (molefraction of methanol in liquid)

Figure 35

So we can predict how R should be varied to keep the top product concentration constant.
We only need to know what was in the still to start with and what distillate composition is
required.

From R(D) we can also calculate the amount of heat required for this process. First,
calculate the amount of vapour evaporated from the still:

t1 t1 D
dD 1

V r 1 = ∫ Vr (t ′) dt ′ = ∫t dt ′ (R( D ) + 1 ) d t ′ = ∫0 R( D ) d D + D1 (36)
t0 0

So we can find Vr(D) by integrating R(D). This is effectively a graphical integration since
we’ve evaluated D for a few selected values of R.

34
 CP302 Vapour-Liquid Separation Processes

Figure 36

The total heat, Q, needed to generate this much vapour is then ∆HvapVr1, where ∆Hvap is
the molar latent heat of vaporization (which we have assumed is independent of
concentration).

i.e. ∆Hvap = hv - hl

3.1.2 Constant reflux ratio

At constant reflux ratio (R), the distillate composition, xd, and hence xd, will vary. Usually,
the design specification will limit operation to a minimum value of xd. So, we need to know
how xd varies as the separation proceeds.

Constant R means that the operating line has a constant gradient. This provides a way to
relate the composition of the distillate stream, xd, based on the composition of the liquid in
the still, xw.

Figure 37

35
 CP302 Vapour-Liquid Separation Processes

By rearranging the material balance equation for the column, we find:

W 0( x w 0 − x w 1 )
x d1 = + x w1 (37)
D1

So we need to calculate D as a function of xw. A material balance (total and component)

gives:

Ddt = d D = −dW
(38)
x d Ddt = −d( x w W ) = −W d x w − x w dW

Putting these together gives:

− xd dW = −W d x w − x w dW

⇒ W d x w = dW (xd − x w )

W 0 − D1 xw 1
dW d xw
∫
W0
W
= ∫
xw 0
xd − x w
(39)

xw 1
W − D1 d xw
⇒ ln 0
W0
= ∫ (x
xw 0 d − xw )

Note, xd is NOT a constant but is changing as xw falls with time.

Now, do the following;

• For a wide range of values of xw1:
find xd1(xw1) from a McCabe-Thiele diagram for fixed n and R.
Perform the integral above (graphically) to find D1 for the selected
values of xw1.
The values of D1 and xw1 can be inserted into the material balance
(Equation (37) to find xd(D).

So now we know when to stop a batch process at constant reflux in order to achieve a
distillate of specified composition. We only need to know what was in the still to start with
(i.e. W0 and xw0).

One again, we can calculate the amount of heat required for this process. First, calculate
the amount of vapour evaporated from the still

t1 t1 D
dD 1

V r 1 = ∫Vr (t ′) dt ′ = ∫ (R + 1) dt ′ = (R + 1) ∫ d D = D1(R + 1) (40)

t0 t0
dt ′ 0

The total heat, Q, needed to generate this much vapour is Q = ∆Hvap Vr1.

36
 CP302 Vapour-Liquid Separation Processes

4 Mass transfer across a vapour-liquid interface

An understanding of vapour-liquid equilibrium is sufficient to design ‘staged’ processes
where the engineering equipment is built from discrete stages – i.e. a trayed tower.

But we also want to design ‘continuous contact’ processes in the form of packed
columns. We will meet both these kinds of process later.

To model continuous contact processes we need an understanding of mass-transfer

across a vapour-liquid interface. So, we need to;
• Develop a model for mass transfer across a gas-liquid interface.
• Apply Fick’s laws of diffusion to this model.
• Interpret the results in terms of mass transfer coefficients.

4.1 The two-film model

Consider a binary mixture in vapour-liquid equilibrium – one example is methanol/water,
which we’ve seen recently.
• Temperature and pressure are equal throughout the system but there is a gradient
in chemical potential, which drives mass transfer from one phase to the other.
• Equilibrium is reached very rapidly the interface, just as we saw for diffusion in
solids.
• We can deal with mass transfer in the two phases separately by working in terms of
concentrations as we have done extensively already.

A representation of the two-film model is shown in Figure 8.

• The interface is taken to be planar, and
• Transport through the bulk occurs mostly via convection, which is a fast process.
• The concentration gradient is therefore confined to narrow layers either side of the
interface, in which mass transfer occurs via diffusion, which is a slow process.
• The concentrations in these ‘laminar boundary layers’ vary smoothly with
distance from the interface, but there is a jump at the interface.

The rate of mass transfer from one bulk phase to the other is therefore controlled by
diffusion through the two laminar boundary layers.

Finally, it is assumed that the system is at ‘steady-state’, i.e. that the concentration profile
is fixed.

The corresponding chemical potential profiles are also sketched below. Note how there
are no jumps in these profiles. Material always flows from higher to lower chemical
potential.

We should apply our understanding of diffusion to determine the molar flux.

37
 CP302 Vapour-Liquid Separation Processes

µ µAl

µAv

µBv
V L

µBl

C
C*Al CAl

C*Bl
CBl

CAv C*Av
CBv C*Bv

y or x
y*
y x
x*

z
Figure 8

4.1.1 Application of Fick’s First Law

We’ve already seen how Fick’s First Law relates the diffusive flux to the concentration
gradient. By adding the convective flux, we obtained an expression for the overall molar
flux. This is the general diffusion equation (equation 2.10 in Part 1 of the module, which
is now very familiar).

dCi
Ni = Nxi − Di (41)
dz

We can apply this equation to the laminar boundary layers of liquid and vapour separately.
We can use equilibrium data to relate the vapour and liquid concentrations at the interface.

38
 CP302 Vapour-Liquid Separation Processes

4.1.2 Mass transfer coefficients

Mass transfer coefficients are often used in engineering, just like heat transfer coefficients.
The mass transfer coefficient (k) is defined via Ni =ki(xi1 – xi2). This is useful because we
may not know the thickness of the laminar boundary layer but we can measure mass
transfer coefficients empirically for a process.

We want to calculate the molar flux based on the boundary conditions, i.e. the
concentrations in the bulk and at the interface. So we need to
• apply the general diffusion equation to a single film
• integrate it to see how the flux depends on boundary conditions
• interpret the results in terms of mass transfer coefficients
There are two separate and important cases;
• Equimolar counter-diffusion
• Diffusion of A through stagnant B
Equimolar counter-diffusion is relevant to distillation where both components (of a
binary mixture) are volatile, and hence both diffuse. Diffusion of A through stagnant B
describes the situation in absorption and stripping where only one of the components is
present in both phases, and hence only this component diffuses. We shall consider these
two situations in turn.

We have no need for ‘diffusion with reaction’ here since we are not considering separation
processes involving chemical reactions.

4.1.2.1 Equimolar counterdiffusion

In this situation NA = -NB, so N = 0 and thus,

dC i
N i = −D i (42)
dz

Rearranging, and using mole fractions, gives

λ x i*

∫ N dz = − ∫ D Cdx ′
0
i
xi
i i (43)

where λ is the width of the film (i.e. the laminar boundary layer), and xi* is the mole fraction
of component i in the liquid at the interface.

The flux Ni must be the same throughout the film, since the system is at steady state.
Following the methodology from Part 1, we can rewrite this as:

x i*

λN i = −Di C ∫ dx ′i (44)
xi

39
 CP302 Vapour-Liquid Separation Processes

which is easily integrated to give

Di C
Ni =
λ
(x i − x i* ) (45)

Now we can write down the mass transfer coefficient, which we have already defined via,
Ni =ki(xi1 – xi2). The mass transfer coefficient relates the flux to a driving force (mole
fraction difference in our case).

Heat transfer coefficients have the same form (they relate heat flux to temperature
differences), as does Ohm’s Law for the flow of electrical current through a resistor (where
1/R, which is the conductance, plays the role of a transfer coefficient).

So according to (45), the mass transfer coefficient is k=DC/λ (this is exact for a perfect
gas). In other words, the film’s resistance to mass transfer is λ/(DC). This makes sense
since a thicker laminar boundary layer should present a larger resistance to mass transfer,
etc.

Mass transfer coefficients can be defined for concentration driving forces expressed
differently, including concentration differences and pressure differences (for ideal gases).
In this case we can write

Di C
Ni =
λ
(x i )
− x i* = k i ( x i − x i* )

Di (46)
= (C i − C i* ) = k i′ (C i − C i* )
λ
Di
= ( )
P − Pi * = k i′′( Pi − Pi * )
λ RT i

Here, ki is a ‘film mass transfer coefficient’ for component i (we will use kGi and kLi for gas
and liquid films respectively).

So now all we need to know are the mole fractions in both bulk phases, the equilibrium
relationship at the interface, and the film mass transfer coefficients, so that we can
calculate the flux across the gas-liquid interface.

4.1.2.2 Diffusion of A through stagnant B

Now let’s do the same for this situation where NB = 0. Then the general diffusion equation
becomes:

dC A
N A = N A x A − DA (47)
dz

which can be rearranged to give

D AC dx A
NA = − (48)
1 − x A dz

40
 CP302 Vapour-Liquid Separation Processes

and thus,

x A*
D AC dx ′A
λ x∫A (1 − x ′A )
NA = − (49)

Performing the integration gives

DAC  1 − x A* 
NA = ln  = k A ( x A − x A* )
 (50)
λ  1− x A 

where the mass transfer coefficient is now:

D AC  1 − x *A 
kA = ln   (51)
λ x A − x *A  1 − x A 
( )
which is clearly dependent on the mole fractions in the bulk and at the interface.

However, for a dilute solution where xA ~ 0 and x*A~0, we have

 1 − x A* 
ln  ~ x A − x A*
 (52)
 1− x A 

So the equations for ‘Diffusion of A through stagnant B’ are the same as those for
‘Equimolar counter-diffusion’ when A is a dilute solute.

Use of equilibrium relationship

Interfacial concentrations are not easy to measure (because the interface itself is
microscopic), so we need to develop equations that do not need them.

Because we are now considering both liquid and gas films simultaneously, x and y will
refer to liquid and gas phase mole fractions, and subscripts L and G will refer to other
liquid and gas phase quantities respectively.

Since we’re considering steady-state processes, we must have:

N i = k Li ( x i* − x i ) = k Gi ( y i − y i* ) (53)

This equation can be solved if the film mass transfer coefficients, the bulk concentrations,
and the equilibrium relationship y i* = f ( x i* ) are all known, since this gives us three
equations and three unknowns.

41
 CP302 Vapour-Liquid Separation Processes

4.1.3 Overall Mass transfer coefficients

An alternate approach is to use the concept of ‘overall mass transfer coefficients’, by

analogy to ‘overall heat transfer coefficients’.

We replace the unknown interfacial ‘*’ quantities by re-writing (53) as

N i = K Li ( x i0 − x i ) = k Li ( x i* − x i ) = kGi ( y i − y i* ) = K Gi ( y i − y i0 ) (54)

where K is an overall mass transfer coefficient and superscript ‘0’ indicates the mole
fraction that would be in equilibrium with the bulk mole fraction of the other phase
(so yi0 is in equilibrium with xi and xi0 is in equilibrium with yi – see Figure ).

x y

B
x
x*
x0
y0
y* y

Figure 9

Slope = -kL/kG
y
(x,y)
(x0,y)
(x*,y*)
(x,y0)

x
Figure 10

42
 CP302 Vapour-Liquid Separation Processes

The mole fractions, xi and yi, are usually known, and hence xi0 and yi0 can be calculated
from an equilibrium relationship, e.g. y i0 = f ( x i ) or an x-y diagram. This allows the molar
flux to be calculated if an overall mass transfer coefficient has been measured.

Overall mass transfer coefficients can be related to film mass transfer coefficients as
follows. First rearrange (54) to give

1 1 yi − yi0 + yi* − yi* 1  yi* − yi0 

= = 1 + 
K Gi k Gi yi − yi* k Gi  yi − yi*  (55)

which with

k Gi ( y i − y i* ) = k Li ( x i* − x i )

gives

1 1 1 y i* − y i0
= + (56)
K Gi k Gi k Li x i* − x i

Now, if we approximate the x-y diagram to be a straight line between (xi*,yi*) and (xi,yi0) we
have

y i* = mG x i* + c
(57)
y i0 = mG x i + c

which on substitution into (56) gives

1 1 mGi
= + (58)
K Gi k Gi k Li

i.e. the overall mass transfer coefficient is simply related to the film coefficients and the
gradient of the x-y diagram at xi. There is a similar equation for the liquid phase overall
mass transfer coefficient

1 1 1
= + (59)
KLi kLi mLikGi

where mLi is the gradient of the straight line between (xi*,yi*) and (xi0,yi). Note, when
component i is dilute we find mLi = mGi, = m and hence KLi = mKGi.

Local film coefficients, kG and kL, can be measured by measuring overall coefficients under
specific circumstances:
• From (58) we see that when kLi is very large, kGi is approximately equal to KGi,
which can be measured experimentally.
• This situation is termed ‘gas phase rate limited’ – all the resistance to mass transfer
is provided by the gas film.
• So we should put our gas phase of interest in contact with liquid phase chosen for
its low resistance to mass transfer, and measure the overall coefficient.

43
 CP302 Vapour-Liquid Separation Processes

Similarly, we see from 60 that kLi is approximately equal to KLi when kGi is very large.
• This situation is termed ‘liquid phase rate limited’ – all the resistance to mass
transfer is provided by the liquid film.
So we should put our liquid phase of interest in contact with a gas phase chosen for its low
resistance to mass transfer, and measure the overall coefficient.

44
 CP302 Vapour-Liquid Separation Processes

5 Continuous contact distillation

This mode of operation is almost identical to continuous multi-stage distillation, except
here distillation occurs in packed towers. So we have a reboiler and a condenser but there
are no plates/trays (i.e. no discrete stages).

Instead, the tower is packed with pebble-sized solid objects, providing the surface over
which the liquid flows. The packing material sits on a perforated distributor plate located
near the bottom of the column.

heat

condenser

top reflux
product

feed

bottom still
product

heat Figure 38

This table lists some of the main advantages and disadvantages of packed columns
compared to multi-stage columns.

Advantages of staged columns Advantages of packed columns

Easy side-stream withdrawal Corrosion resistant ceramic packing
Easy cleaning/replacement of trays Low pressure drop (thus suited to vacuum
use)
Suited to high vapour flow rates – i.e. large Suited to low vapour flowrates – i.e. small
column diameter column diameter
Can accommodate low ratio of liquid to Can accommodate high ratio of liquid to
vapour flowrates (low R) vapour flowrates (high R)

45
 CP302 Vapour-Liquid Separation Processes

Analysis of this process parallels that of multi-stage distillation. Let’s just remind ourselves
of the assumptions, approximations and results of this analysis.

Our model of a continuous contact tower assumes a binary fluid mixture in a tower with
one feed, and top and bottom products. We make the following approximations;
• Constant pressure throughout the column.
• No heat losses.
• Steady-state.
• The molar enthalpies of saturated vapour (hv) and saturated liquid (hl) are each
constant.

Our assumption about equilibrium is different for packed (continuous contact) columns.
There are no stages. Equilibrium is reached at the vapour-liquid interface only. The bulk
gas and bulk liquid phases are not in equilibrium. Again we use an x-y diagram with an
equilibrium line describing the binary system at constant pressure.

Likewise, we can draw operating lines for the rectification and stripping sections (as well
as the q-line) for any given reflux ratio. The operating lines represent the conditions in the
bulk of the liquid and vapour at a given point in the column.

Just as for multi-stage distillation we find that there is an optimum reflux ratio located
between the minimum ratio and total reflux.

D, xd

 x − xf 
Vr = D(R + 1) Lr = DR D = F  w 
 xw − xd 

F, xf W =F−D

(h i ,v − hf )
q=
Vs = Vr + F(q - 1) Ls = Lr + Fq (hi ,v − hi ,l )

Dx d
Re cov ery =
W, xw Fx f
Figure 39

However, unlike multi-stage distillation, there are no discrete stages to step off. Instead we
need to calculate the height of the packed section of the column. This analysis is based on
our understanding of mass transfer across a vapour-liquid interface.

46
 CP302 Vapour-Liquid Separation Processes

Figure 40
As both components of the binary mixture are volatile, they are transferred across the
interface in opposite directions. The more volatile component moves from liquid to vapour,
while the less volatile component moves from vapour to liquid.

The compositions of the liquid (x) and vapour (y) are continuously varying, as a function
of vertical position (z). We assume there are no radial concentration gradients.

Consider a differential element (i.e. a thin slice) of the distillation tower with thickness, dz.
We have seen previously that V and L are constant for the sections of the column above
and below the feed. It follows that for each mole of component A passing from liquid to
vapour, there is one mole of component B passing from vapour to liquid. This corresponds
to equimolar counterdiffusion.

V (y+dy) L (x+dx)

NAS
dz NBS

Figure 41 Vy Lx

From a material balance for the most volatile component we see that:

N AsAdz + Vy = V ( y + dy )
N AsAdz + Lx = L( x + dx )

N AsAdz = Vdy = Ldx = NB sAdz (61)

47
 CP302 Vapour-Liquid Separation Processes

where s is the vapour-liquid interfacial area per unit volume of the column (depends on the
packing material) and a is the cross-sectional area of the column. Thus sAdz is the
interfacial area for mass transfer within the differential element.

From previously, we know that for equimolar counterdiffusion:

N A = k GA ( y * − y ) = k LA ( x − x * ) = K GA ( y 0 − y ) = K LA ( x − x 0 ) (62)

Putting (61) and (62) together gives

Vdy Ldx
= k GA ( y * − y ) = k LA ( x − x * ) = K GA ( y 0 − y ) = K LA ( x − x 0 ) = (63)
sadz sadz

Thus, we have a differential equation which we can solve to find the height of a packed
section of column.

By convention, some of the variables are grouped together by introducing the ‘height of a
theoretical unit’ (HTU), which is defined as hGA = V /(sak GA ) etc.

Integration of (63) gives:

z1 + LR y ( z1 + LR ) x ( z1 + LR ) y ( z1 + LR ) x ( z1 + LR )
hGA hLA H GA H LA
∫z dz = y (∫z dy
) ( y *
− y )
= ∫ dx
x ( z ) ( x − x *
)
= ∫ dy (y
y ( z1 )
0
− y)
= ∫ dx ( x − x
x ( z1 )
0
)
(64)
1 1 1

For equimolar counterdiffusion we can make the approximation that all the mass transfer
coefficients are independent of concentration (from mass transfer section).

This means the HTU coefficients are also independent of concentration, so they can be
taken outside the integrals.

So, for the rectification section, within which V and L are constant, we have:

yd xd y
d d x
dy dx dy dx
∫ LA ∫ GA ∫ LA ∫
R R R R
LR = hGA *
= h *
= H 0
= H (65)
y ( zf )
(y − y) x ( zf )
(x − x ) y ( zf )
(y − y) x ( zf )
( x − x0 )

Similarly, for the stripping section we have:

y ( zf ) f x( z ) f y( z ) f x( z )
dy dx dy dx
LS = h ∫ S
GA *
= hLA ∫
S
*
= HGA ∫
S
0
= H LA ∫
S
(66)
y ( z1 )
(y − y) x ( z1 )
(x − x ) y ( z1 )
(y − y) x ( z1 )
( x − x0 )

(Some further comment is needed on the limits of integration – we’ll return to this shortly)

Each of these equations can be written as

L = HTU × NTU (67)

where NTU is the ‘number of transfer units’, which does not need to be an integer.

Thus, the total tower height is simply LT = Ls + LR.

48
 CP302 Vapour-Liquid Separation Processes

So, you see that we have 4 ways of calculating the height of the tower depending on
which mass transfer coefficients we know.
Method Mass transfer HTU Independent Dependent
coefficient variable variable
Gas film kGA hGA y y*
Liquid film kLA hLA x x*
Gas overall KGA HGA y y0
Liquid overall KLA HLA x x0
To calculate the integrals in (65) and (66) we need to know the dependent variable as a
function of the independent variable (e.g. y*(y)). Figure 42 illustrates the various choices.

The operating lines (which describe mass balance) define y(x) throughout the tower, i.e.

xd R
y= + x (68)
R +1 R +1

for the rectification section, and

(DR + Fq ) W
y= x− xw (69)
D(R + 1) + F (q − 1) D(R + 1) + F (q − 1)

for the stripping section.

y0 and x0 are easily found; y0 is the gas phase mole fraction in equilibrium with x, while x0
is the liquid phase in equilibrium with y.
distillation of methanol/water at 1 atm
1
x, y0
slope = -k L / k G x *, y *
ym (molefraction of methanol in vapour)

0.8 x, y
x,y

0.6

0.4 x 0, y x,y

0.2

0
0 0.2 0.4 0.6 0.8 1
xm (molefraction of methanol in liquid)

Figure 42

To determine y* and x* we rearrange (62) to get

kL y * − y
− = (70)
kG x * − x

i.e. the slope of the line joining (x, y) to (x*, y*) is –kL / kG.
49
 CP302 Vapour-Liquid Separation Processes

The integrals in (65) and (66) can be calculated (e.g. graphically) provided we know the
integration limits.

We can find x(zf) and y(zf) (where zf is the height at which feed is entered) by noting that
the feed location should be chosen so that we are always on the lower of the two
operating lines. Thus, the feed is injected where the composition in the column is defined
by the intersection of the operating lines.

To find y(z1) at the bottom of the stripping section we need to realise that there will be a
vapour rising from the reboiler is in equilibrium with the liquid of composition xw (i.e. the
reboiler acts as an equilibrium stage just as we have seen previously). We can find the
liquid composition at the bottom of the packed column by stepping across to the stripping
operating line.

distillation of methanol/water at 1 atm

1
ym (molefraction of methanol in vapour)

0.8

y(zf)
0.6

y(z1)
0.4

0.2

x(z1) x(zf)
0
0
xw 0.2 0.4
xf 0.6 0.8
xd 1
xm (molefraction of methanol in liquid)

Figure 43

50
 CP302 Vapour-Liquid Separation Processes

5.1.1 Properties of packing

Packing material has a high specific surface area, s (surface area to volume ratio). The
liquid is distributed as a film on the surface of the packing and trickles downwards from the
top of the column under gravity. Vapour enters at the bottom of the tower and flows
upwards through voids in the packing material and through the spaces between the
packing. So liquid and gas phases are largely continuous.

The packing material should have large voids. The dimension of the voids must be larger
than the typical liquid film thickness (a few mm). Otherwise the liquid will simply fill up the
pore and it will not contribute to the gas-liquid interfacial area.

The packing material should be chemically inert, have reasonable structural strength and
be low cost. It can be placed (or dumped) in the tower randomly, or be formed into blocks
that can be carefully stacked in the tower (structured packing).

Each individual packing ‘pebble’ is generally only a few centimetres in size, and modern
packing is manufactured into rounded shapes such as rings, cylinders or saddles.

51
 CP302 Vapour-Liquid Separation Processes

6 Gas absorption and Stripping

Because gas absorption and gas stripping are very similar processes, we will consider
them together.
• Gas absorption is used to separate a soluble component of a gas mixture by
dissolving it in a liquid (solvent).
• Gas stripping, the reverse process, removes a volatile component from a
liquid by passing gas through it.
(Note: absorption is sometimes also known as ‘scrubbing’ and absorption columns as
‘scrubbers’)

We will consider three process designs:

• Continuous single-stage absorption/stripping: this is similar to continuous
single-stage distillation.
• Continuous multi-stage absorption/stripping: this is similar to continuous
multi-stage distillation.
• Continuous contact absorption/stripping: this is similar to continuous contact
distillation.

Therefore, much of the analysis we have already developed for distillation will also apply
here. There are only

In each gas absorption/stripping process we have at least 3 components:

• A: the component of interest that we want to absorb/strip
• B: very high boiling-point, non-volatile solvent
• C: very low boiling-point, insoluble gas
A and B exist as a mixture in the liquid phase, while A and C exist as a mixture in the
vapour phase (Figure 45 shows a simple gas absorption process).

AC vapour
C
A
B liquid AB
Figure 45

For gas absorption, the ideal B component will have the following properties:
• Non-volatile – the aim is to remove impurities from the gas
• High solute solubility – for more efficient absorption
• Low viscosity – better plate efficiency; low pumping costs
• Chemically stable – to avoid production of impurities
• Safe: i.e. non-toxic, non-flammable
• Non-corrosive – to reduce maintenance costs
• Low freezing point – to reduce maintenance costs and improve operating
range
• Low cost
52
 CP302 Vapour-Liquid Separation Processes

Mole ratios
Because in an absorption/stripping process there is always an inert component in both the
gas and liquid phases (in the liquid the solvent (B) is non-volatile, in the gas phase one or
more of the gas components (C) are insoluble), we should write flow rates in terms of
these inert components, i.e. we should use mole ratios:

xA xA
XA = = (71)
xB 1 − x A

Hence, it follows that:

XA
xA = (72)
1+ X A

So, the flow rate of volatile A, VA, can be expressed in terms of the flow rate of inert B, VB,

VA = x AVT = X A x BVT = X AVB (73)

which is very useful because VB is constant and hence operating lines expressed in
terms of mole-ratios will be straight lines.

xy diagram for Absorption / Stripping

Because in a gas absorption/stripping process we have 3 components (volatile component
of interest plus insoluble gas plus non-volatile solvent), i.e. a ternary mixture, we have one
more degree of freedom compared to the binary mixtures we’ve been considering up to
now.

This extra degree of freedom allows us to draw an equilibrium diagram for A where both
temperature and pressure are fixed.
X-Y diagram for carbon dioxide (A), nitrogen (C) and
methanol (B) at 223 K and 50 bar
0.3
YCO2 (mole ratio in vapour)

0.2

0.1

0
0 0.2 0.4 0.6 0.8 1

XCO2 (mole ratio in liquid)

Figure 46

53
 CP302 Vapour-Liquid Separation Processes

To understand gas absorption/stripping processes we will use the same methodology as

with distillation, except:
• We will use mole ratios
• No enthalpy balance is required since we already know that the flowrates of
non-volatile solvent and the insoluble gas are fixed throughout the column
• The process has fixed pressure and temperature
The assumption of constant temperature means we are neglecting the heat of solution. In
some cases this will lead to reasonable results and it avoids the need for a much more
involved design methodology.

6.1 Continuous single-stage absorption/stripping

Consider a vessel with input and output streams for both liquid and vapour. The output
streams are in equilibrium with each other and the temperature is fixed (note – the input
streams are not in equilibrium).

The vapour and liquid flow rates are now written in terms of the inert gas and liquid
components. Because the transfer of these components from one phase to the other is
insignificant, their output flow rates are taken to be identical to their input flow rates
(usually a very good approximation). We now switch to a shorthand notation where flow
rates refer to the inert components and mole ratios refer to the volatile component of
interest.

V, Y1 V,Y2
not in P, T
equilibrium in equilibrium
L, X1 L, X2

Figure 47 heat

54
 CP302 Vapour-Liquid Separation Processes

Continuous single-stage absorption for carbon dioxide

(A), nitrogen (C) and methanol (B) at 223 K and 50 bar
0.3
Y (mole ratio CO2 in vapour)

Y1
slope = - L / V

0.2

Y2
0.1

0
0 X1 0.2 X2 0.4 0.6 0.8 1

X (mole ratio CO2 in liquid)

Figure 48

The mole ratios correspond to the component of interest, which is absorbed from the
vapour phase into the liquid. A material balance gives

L
Y2 = ( X 1 − X 2 ) + Y1 (74)
V

The output streams must satisfy this material balance as well as the conditions for
equilibrium, i.e. they correspond to the intersection of the equilibrium curve with the
operating line defined by (74).

6.2 Continuous multi-stage absorption/stripping

Continuous multi-stage absorption/stripping is much like continuous multi-stage distillation:
• Multi-stage absorption/stripping can achieve much better separation than
single-stage
• It occurs in absorption/stripping columns that consist of equilibrium plates
However, there are some differences:
• The mixtures in absorption/stripping towers are at least ternary
• Liquid is entered at the top
• Vapour is entered at the bottom
• There is no reflux
• There is no reboiler or condenser

55
 CP302 Vapour-Liquid Separation Processes

vapour out

liquid in

liquid out

vapour in
Figure 49

In gas absorption/stripping the liquid/gas that is output is separated from component A

(e.g. by distillation), and is then used again to provide the inert liquid/gas input. So the
liquid/gas input is usually not pure and has a small amount of A in it.

To analyse absorption/stripping we can use a method that is similar to the McCabe-Thiele

method for distillation. A mass balance for component A, between stage i and either the
bottom or the top of the column, gives the operating line:

L L
Yi = Y0 + ( X i +1 − X1 ) = Yn + ( X i +1 − X n +1 ) (75)
V V

V, Yn L, Xn+1
…

V, Yi L, Xi

V, Y0 L, X1
Figure 50

56
 CP302 Vapour-Liquid Separation Processes

This tells us how the concentrations of vapour going up from a plate, and the liquid coming
down to it, are related. If we draw this operating line on a plot together with the equilibrium
curve then we can step-off the stages as follows.
• For gas absorption (Figure 52):
First, draw in the point corresponding to Xn+1, Yn, and the line
corresponding to constant Y0. These values are determined by the
composition of the gas to be treated, the target gas purity and the
solvent purity.
Then draw the operating line starting at Xn+1, Yn with slope L / V. Its
intersection with the line Y=Y0 defines X1.
Now step-off the equilibrium plates.
It is unlikely that this can be done exactly without adjusting L / V.

Continuous multi-stage absorption for carbon dioxide (A),

nitrogen (C) and methanol (B) at 223 K and 50 bar
0.3
Y (mole ratio CO2 in vapour)

Y0

0.2 slope = L / V

0.1
Y5

0
0 X6 0.2 0.4 X1 0.6 0.8 1

X (mole ratio CO2 in liquid)

Figure 52

• For gas stripping (Figure 53):

First, draw in the point corresponding to X1, Y0, and the line
corresponding to constant Xn+1. These values are determined by the
composition of liquid to be treated, the target purity of the liquid and
the purity of the inlet gas.
Then draw the operating line starting at X1, Y0 with slope L / V. Its
intersection with the line X=Xn+1 defines Yn.
Now step-off the equilibrium plates.
It is unlikely that this can be done exactly without adjusting L / V.

57
 CP302 Vapour-Liquid Separation Processes

Continuous multi-stage stripping for carbon dioxide (A),

nitrogen and methanol (B) at 223 K and 50 bar
0.3
Y (mole ratio in vapour)

0.2

Y4

0.1 slope = L / V

Y0
0
0 X1 0.2 0.4 X5 0.6 0.8 1
X (mole ratio in liquid)

Figure 53

See how important the flow rate ratio, L / V, is in these processes.

If either of these flow rates, L or V, is adjusted so that the operating line touches the
equilibrium line then we will need an infinite number of equilibrium plates. These limiting
flow rates depend on the design specification and the shape of the equilibrium curve.

When the equilibrium line is straight and goes through the origin, e.g. at low mole ratios,
we can find the number of equilibrium plates analytically using an expression known as a
Kremser equation. You will find reference to these equations in standard textbooks on
separation processes (e.g. McCabe, Smith, Harriot, Unit Operations of Chemical
Engineering) but we mention them here only in passing since they illustrate little about the
underlying principles.

58
 CP302 Vapour-Liquid Separation Processes

6.3 Continuous contact absorption/stripping

This process is almost identical to continuous multi-stage absorption/stripping, except here
absorption occurs in packed towers (and also sometimes in spray towers).
V, Y2 L, X2

V, Y1 L, X1
Figure 54
Analysis of this process parallels that of continuous contact distillation. In the following only
absorption will be considered, although the model and equations for stripping are identical.

Our model of a continuous contact absorption tower has top and bottom feeds and
products. We make the following approximations:
• Constant pressure and temperature throughout the tower.
• Steady-state process.
• Equilibrium is reached at the interface.
By analysing mass balances we can derive equations for the equilibrium curve and
operating line just as we have already done for multi-stage absorption columns.

To estimate the tower height we consider mass transfer across the gas-liquid interface.
The component of interest transfers from the gas to the liquid; all other components are
inert. This corresponds to diffusion of A through stagnant B.

Consider a section of the absorption tower. From this we see that

− N A S = VdY = LdX (76)

V (Y+dY) L (X+dX)

dz NAS

Figure 55 VY LX

59
 CP302 Vapour-Liquid Separation Processes

From section 3 we know that for diffusion of A through stagnant B from gas to liquid we
have

N A = k GA ( y − y * ) = k LA ( x * − x ) = K GA ( y − y 0 ) = K LA ( x 0 − x ) (77)

Putting (76) and (77) together gives

VdY LdX
= k GA ( y * − y ) = k LA ( x − x * ) = K GA ( y 0 − y ) = K LA ( x − x 0 ) = (78)
sadz sadz

where adz is the volume of the differential element (a is the cross sectional area of the
column) and s is the interfacial area per unit volume of tower (which depends on the
packing material). Integration gives

LT Y ( LT ) X ( LT ) Y ( LT ) X ( LT )
hGA hLA H H LA
∫0 dz = Y ∫( 0dY
) (y * − y )
= ∫ dX
X (0) (x − x * )
= ∫ dY 0 GA =
Y (0) (y − y ) ∫ dX ( x − x
X (0)
0
)
(79)

where hGA = V / sakGA etc.

Now there are two complications compared with continuous contact distillation.
• We need to write the integrands in terms of Y and X, rather than y and x
• The mass transfer coefficients are functions of concentration (since we are
dealing with diffusion of A through stagnant B)
Let’s deal with these in turn.

First, from the definition of mole ratio we can use the substitutions

YA XA
y= and x= (80)
1 + YA 1+ XA

Second, because the mass transfer coefficients are functions of concentration (see (51)
and (52)) the HTU coefficients cannot immediately be taken outside of the integral.
However, the variation of these coefficients with concentration is small provided the
concentration range is sufficiently small.
1 − y * 
To illustrate this point, see Figure 56. Here the variation of ln
1 − y
 y − y * with y is ( )
 
drawn for the example case when y* = 0.9 y. This shows that in this case yLM is reasonably
constant over a small interval of y (between 0.1 and 0.2 for example). If this is the case,
then the HTU coefficients can be approximated to be constant. We will use this
approximation.

60
 CP302 Vapour-Liquid Separation Processes

1 − y * 
ln   (y − y*)
 1− y 

Figure 56
Resolving these complications gives the tower height as

Y ( LT ) X ( LT )
(1 + Y * )(1 + Y ) (1 + X )(1 + X * )
LT = hGA ∫ dY = hLA ∫ dX
Y (0 ) (Y * − Y ) X (0 ) (X − X * )
y ( LT ) X ( LT )
(81)
(1 + Y 0 )(1 + Y ) (1 + X )(1 + X 0 )
= HGA ∫ dY = H LA ∫ dX
Y ( 0) (Y 0 − Y ) X ( 0) (X − X 0 )

So, you see that once again we have 4 ways of calculating the height of the tower
depending on which mass transfer coefficients we have.

The operating line defines Y(X) throughout the tower, i.e.

L
Y = Y1 + (X − X 1 ) (82)
V

Figure 57 illustrates how to obtain the dependent variable as a function of the independent
variable (e.g. Y*(Y)). Y0 and X0 are easily found: Y0 is the gas phase mole ratio in
equilibrium with X, while X0 is the liquid phase mole ratio in equilibrium with Y. Y* and X*
can be found iteratively from (77) if y*(x*) is known.
Continuous contact absorption for carbon dioxide (A),
nitrogen and methanol (B) at 223 K and 50 bar
0.2
Y (mole ratio in vapour)

Y1
X, Y

0.1 X, Y
X, Y0
X*, Y*
X, Y
Y2 X 0, Y

0
0 X2 0.2 X1 0.4
X (mole ratio in liquid)

Figure 57

61
 CP302 Vapour-Liquid Separation Processes

6.3.1 Dilute solutions

For dilute solutions the complications discussed in relation to (79) are no longer relevant
because
• From (71) we have Y ~ y and X ~ x
• The mass transfer coefficients are no longer functions of concentration
So (81) becomes:

y ( LT ) x ( LT ) y ( LT ) x ( LT )
dy dx dy dx
LT = hGA ∫ *
y (0 ) ( y − y )
= hLA ∫x(0) ( x − x * ) = HGA ∫ 0
y (0 ) ( y − y )
= H LA ∫ 0
x (0 ) ( x − x )
(83)

Furthermore, we can use Henry’s Law (6) to obtain:

y 0 = mx and y = mx 0 (84)
and the operating line (82) becomes:
L
y = y1 + ( x − x1 ) (85)
V
An important consequence of this is that it is easier to determine y* and x* from x and y
since -kLA/kGA=(y-y*)/(x-x*), i.e. the line connecting the points (x,y) and (x*,y*) has the
gradient -kLA/kGA, just as for distillation. Therefore, for example, y* can be found for various
values of y, including the limits y(0) and y(LT), and the integration can be performed
graphically.

6.4 Operating pressure

It is clear from Figures 52 and 53 that the operating line should always lie above the
equilibrium line for absorption, and below the equilibrium line for stripping. If these lines
were to touch then an infinite number of plates (or infinite tower height) would be required.

A design specification will usually constrain the operating line to some degree, i.e. for
absorption the concentrations of the input liquid and vapour streams and the output vapour
stream are often specified. So, to optimise the process we can modify the equilibrium line.
The optimal process will be the most economic one.

The gas composition is influenced by pressure and temperature. Because absorption and
stripping processes involve ternary systems, the extra degree of freedom allows us to vary
pressure and temperature independently.

Sometimes the pressure is varied to optimise the process. Increasing the pressure leads
to a downward shift of the equilibrium line. So, increasing the pressure is advantageous for
gas absorption, while decreasing the pressure is advantageous for gas stripping. It follows
that there is a minimum allowable pressure for absorption and a maximum allowable
pressure for stripping. At these pressures the operating and equilibrium lines touch.

62
 CP302 Vapour-Liquid Separation Processes

7 Evaporation
Evaporators boil a liquid to evaporate off the volatile component. The liquid can be either a
solution, where the solvent is to be evaporated and the solute is non-volatile, or a ‘slurry’
consisting of liquid and suspended particulate solids. Since the solute (or suspended solid)
is non-volatile, the vapour generated by evaporation is pure.

Evaporation can be used either to concentrate a dilute solution (e.g. to concentrate fruit
juice), to purify a solvent (e.g. to desalinate water), or to treat a slurry. We will consider:
• Single-stage continuous evaporators
• Multi-stage continuous evaporators
Multi-stage are operationally more efficient (they require less steam) than single-stage
evaporators, although they are associated with higher capital cost.

• A single-stage evaporator consists of a boiling vessel in which evaporation

takes place. Heat is provided via a tube bundle immersed in the liquid. The
heat transfer area required and the flowrate of steam needed for heating are
important questions for evaporator design.

7.1 When does a liquid boil?

Industrial evaporators involve the separation of a volatile solvent from a non-volatile solute
or suspended solid. The liquid in the evaporator is usually at its boiling point. We therefore
need to understand the boiling process and the factors that influence it.

Boiling occurs when the vapour pressure of a liquid component is equal to the total
pressure of the system, P. In an open vessel, where vapour may escape so the total
pressure remains constant, boiling can be sustained for as long as there is liquid to boil.

For a ideal gas we have from (1)

Pi = Pi * x i γ i (86)

where Pi = Pyi is the partial pressure of component i. For a solution where the solute is
considered non-volatile, we need only consider the partial pressure of the solvent, so
boiling occurs when

P = Ps* x s γ s = Ps* (1 − x )γ s (87)

where Ps* is the pure component vapour pressure of the solvent at temperature T, x is the
mole-fraction of the solute and γs is the activity coefficient of the solvent in the mixture.
Note that ys = 1, since only the solvent is volatile.

For a dilute solution or an ideal solution this becomes:

P = Ps* (T )(1 − x ) (88)

63
 CP302 Vapour-Liquid Separation Processes

For a liquid with suspended solids the mole fraction of solute is zero, i.e. the solids have
no effect on the properties of the liquid because they are not dissolved within the liquid.

7.1.1 Boiling-point rise

Suppose that we add some salt to a pan of water. It is now a mixture and so xw < 1.

From (86) we see this means the partial pressure drops. So boiling will occur at a higher
temperature than for pure water. This effect is known as the ‘boiling point rise’.

The boiling point rise (BPR) is defined as the difference in the boiling temperatures of the
mixed and pure systems at the same pressure.
For the pure system we have: P = Ps* (T0 ) , where the bracket shows that Ps* is a function
of T. For a mixture of the solvent with a non-volatile solute we have equation (87), which is
written here to show the explicit dependence on temperature

P = Ps* (T1 )(1 − x )γ s (89)

So the boiling point rise is just T1 – T0.

For a liquid with suspended solids the mole fraction of solute, x, is zero, i.e. the solids
have no effect on the properties of the liquid because they are not dissolved within the
liquid. Suspended solids do not cause a boiling point rise.

7.2 Continuous single-stage evaporation

Consider the single-stage evaporator in Figure 58. Steam enters the condenser at TS and
leaves at TC. It provides heat to the boiling vessel. Feed enters the vessel at a flow rate F,
with composition xF (with respect to the solute) and temperature TF. Pressure, P, is
maintained by withdrawing liquid at flow rate L and composition xL and pure solvent gas at
flow rate G.
G, T

F, xF, TF L, xL, T

S, TS Q S, TC
Figure 58

We make several assumptions and approximations:

• The solute is assumed to be non-volatile.
• The vapour is in equilibrium with the liquid product
• The process is perfectly insulated (so the only heat flow occurs from
condensing steam to boiling liquid).
• The molar enthalpy of the solution is not a function of concentration.

64
 CP302 Vapour-Liquid Separation Processes

• Specific heat capacities are independent of temperature (and of pressure)

over the ranges of interest.
• The solvent is water.
Energy balance on condensing steam
We start our analysis with an energy balance over the condensing steam, which provides
heat, Q:

Shs = Q + Shc (90)

where hs and hc are the molar enthalpies of the steam and condensate respectively. We
can define enthalpies with respect to a reference condition, i.e. for a liquid:

hL = hTref + cP ,L (T − Tref ) (91)

while for a gas:

hG = hTref + ∆Hvap + cP ,G (T − Tref ) (92)

where Tref is an arbitrary reference temperature, and ∆Hvap is the heat of vaporization at
Tref. We shall use Tref = 0 ºC for convenience. Substituting into (90) gives:

S( ∆Hvap + cP ,G (Ts − Tref ) + hTref ) = Q + S(cP ,L (Tc − Tref ) + hTref ) (93)

which can be re-arranged to give the following (where temperatures are expressed in ºC).

Q = S (∆Hvap + cP ,GTS − cP ,LTC ) (94)

So the heat transferred, Q, can be calculated provided we know the temperature of the
inflowing steam and outflowing condensate.

Material and energy balances on single-stage evaporator

At steady state we have a material balance

F =G+L ; Fx F = Lx L (95)

A design specification will usually constrain F, xF TF and xL. So L and then G are easily
found. An enthalpy balance across the boiling vessel gives

FhF + Q = GhG + LhL (96)

where hF, hG and hL are the molar enthalpies of the feed, gas and liquid streams
respectively, and Q is the heat transferred. Substituting in the definitions of liquid and gas
enthalpy gives (with Tref = 0):

F (cP ,LTF + hTref ) + Q = G (∆Hvap + cP ,GT + hTref ) + L(cP ,LT + hTref ) (97)

Combining this expression with the material balance (95) gives:

65
 CP302 Vapour-Liquid Separation Processes

Fc P ,LTF + Q = G( ∆Hvap + c P ,GT ) + Lc P ,LT (98)

We know F, G and L from the material balance. We also know T from the calculation of
the boiling point rise at pressure, P. Therefore the enthalpy balance on the evaporator can
be used to determine the duty, Q. The enthalpy balance on the condensing steam then
allows us to calculate the flowrate of steam required, S.

Heat transfer area

We can now find the area of the heat exchanger from Q and T by considering the rate of
heat transfer. We assume that the steam supplied is not significantly superheated and the
condensate is not significantly sub-cooled. Thus, the steam and condensate temperatures
are roughly equal, TS≈TC.

Q = UA(TS − T ) (99)

where U is the overall heat transfer coefficient and A is the heat transfer area. There is no
need to use a log-mean temperature difference here since the steam inside the tubes is at
a uniform temperature, TS, and the boiling liquid in the evaporator is also at a uniform
temperature, T.

7.3 Continuous multi-stage evaporation

In a multi-stage, or ‘multi-effect’, system the gas produced from one stage is used as a
heat transfer fluid to boil the liquid in the next stage. Therefore steam is only required to
heat the first stage. This saves energy, so multi-effect evaporators are more efficient
than single-stage systems.

This can only happen if the boiling temperatures of the stages are not equal, which in turn
requires the pressures to be unequal.

Generally, we find a temperature profile: TS > T1 > T2 > … > Tn, and hence PS > P1 > P2 >
… > Pn, where there are n stages and subscript S indicates steam.

We will consider forward feed, backward feed and parallel feed arrangements.

7.3.1 Forward feed

In a forward feed system the liquid output from the ith stage is input to the i+1th stage, and
the gas output of the ith stage is used to heat the i+1th stage, as shown in Figure 59.

P1 P2
F, xF, TF L1, x1, T1 L2, x2, T2
Q1 Q2
S, TS G1, T1 G2, T2

S, Tsat1 G1, T1-BPR

Figure 59

66
 CP302 Vapour-Liquid Separation Processes

The condensate leaving the 1st stage will be at the same temperature as the steam
entering (i.e. Tsat1=Ts), since we have considered the steam supplied to be saturated and
assumed sub-cooling to be negligible.

For all other stages, steam generated in the preceding stage enters at Ti-1, and
condensate leaves at a lower temperature, where the temperature difference corresponds
to the BPR of stage i-1. This assumes there is no sub-cooling.

Although only two stages are shown in Figure 59, any number of stages can be connected
in this way. Just as with a single stage, a design specification will usually consist of feed
stream flow rate, composition and temperature, and product composition. Some
parameters of the system need to be adjusted to achieve the desired output concentration
from the given feed conditions.

We are generally interested in calculating the heat transfer area of each stage and the
steam requirement. As with the single stage system, the analysis is based on
consideration of mass and energy balances. We will develop equations for each stage,
and then link the stages together. When dealing with many effects in series, this can result
in large systems of equations to be solved simultaneously: a computer can be used to do
this.

We start by considering an energy balance on the condensing steam:

S (∆Hvap + cP ,GTS ) = Q1 + ScP ,LTs ; 1st stage

(100)
Gi −1 (∆Hvap + cP ,GTi −1 ) = Qi + Gi −1cP ,L (Ti −1 − BPRi −1 ) ; i ≠ 1

The heat transferred to the boiling liquid, Qi, must also satisfy an equation for the rate of
heat transfer:

Q1 = U1 A1(Ts − T1 ) ; 1 st stage
(101)
Qi = U i Ai ((Ti −1 − BPRi −1 ) − Ti ) ; i ≠1

An energy balance over the evaporator gives:

FcP ,LTF + Q1 = L1cP ,LT1 + G1 (∆Hvap + cP ,GT1 ) ; 1st stage

(102)
Li −1cP ,LTi −1 + Qi = Li cP ,LTi + Gi ( ∆Hvap + cP ,GTi ) ; i ≠ 1

Total and component material balances define the mole fraction of the exiting liquid and
the gas flow rate:

Fx F = L1 x1 ; 1 st stage
(103)
Li −1 x i −1 = Li x i ; i ≠1

F = L1 + G1 ; 1 st stage
(104)
Li −1 = Li + Gi ; i ≠1

Therefore we have 5 equations for each stage of the multi-effect evaporator.

67
 CP302 Vapour-Liquid Separation Processes

If Ai is known for each effect, then we can solve the simultaneous equations for the
following parameters for each stage: Qi, Ti, Li and Gi, as well as xi for all but the last stage
and S. This amounts to 5n unknowns, where n is the number of effects.

Often, each stage of the multi-effect evaporator is specified to be identical mechanically,

therefore having the same heat transfer area, Ai.

This forward-feed arrangement is useful for solutions where solubility of the solute
increases with decreasing temperature (which is an untypical situation).

7.3.2 Backward feed

In a backward feed system the liquid output from stage i is input to stage i-1, and the gas
output of the stage i is used to heat the stage i+1, as shown below.

Steam is used in the first stage only, feed enters the last stage, and pumps are used to
overcome the increasing pressure as liquid is pumped backwards from the last to the first
stage.

This arrangement has the advantage that the stage with the highest temperature (the first
stage) also has the highest concentration. The increased temperature may serve to
counteract the tendency of higher concentration to cause increased viscosity, thus
improving heat transfer.

Another advantage of the backward feed arrangement is that solubility usually rises with
increased temperature and here the most concentrated solution is found in the stage with
highest temperature.

Pi
Li, xi, Ti Li+1, xi+1, Ti+1
Qi
Gi-1, Ti-1 Gi, Ti

Gi-1, Ti-1-BPRi-1
Figure 60

This arrangement is somewhat more difficult to analyse because the liquid and gas flows
are in opposite directions. However, this process can still be understood in terms of
material and enthalpy balances, just as we have done for forward-feed systems.

7.3.3 Parallel feed

As shown in Figure 61, each stage is provided with a separate feed. These separate feeds
can be identical (the main feed is just divided among the stages) or quite different if a
range of solutions, perhaps with different solvents, are to be evaporated.

This arrangement is useful if crystallisation causes problems with pumping liquid

from stage to stage.

68
 CP302 Vapour-Liquid Separation Processes

Analysis of this arrangement is similar to the forward-feed arrangement but Li-1, Ti-1 and xi-1
should be replaced by Fi, TFi, and xFi respectively.

Pi
Fi, xFi, TFi Li, xi, Ti
Qi
Gi-1, Ti-1 Gi, Ti
Figure 61
Gi-1, Ti-1-BPRi-1

7.4 Comparison of single and multi-stage evaporators

The steam economy, Se, of an evaporation system is defined as the ratio of the gas
evaporated to the steam consumed, i.e.

∑G
i
i
Se = (105)
S

We can perform a simple analysis of steam consumption by using the fact that the heat
required to vaporise liquid is much more than that required to heat liquid or gas, i.e. the
heat of vaporisation dominates heat transfer. With this approximation the equations for
stage i of a forward-feed, multi-effect evaporator become:

1. Qi ≈ Gi −1∆Hvap
2. Ti = TSout − Qi U i Ai
Li −1∆Hvap − Qi
3. Li ≈
∆Hvap
4. x i = Li −1 x i −1 / Li
5. Gi = Li −1 − Li (106)

Putting steps 5, 3 and 1 together gives Gi ≈ Gi-1, i.e. the amount of vapour produced
from each stage is roughly the same. So we can conclude that Gi ≈ G0 ≈ S. Using the
same analytical process, it turns out that the same is true for backward-feed and parallel-
feed evaporators as well. Essentially, Se ≈ nS / S = n.

69
 CP302 Vapour-Liquid Separation Processes

Further, from steps 1 and 2 we find:

S∆Hvap ≈ Qi = U i Ai (TSout − Ti ) (107)

Once again, this result holds for all multi-effect arrangements. So for every arrangement,
the total heat transferred, Qtot, is

Qtot = ∑ Qi = UA∑ (TSout − Ti ) ≈ nS∆Hvap (108)

i i

where U and A are taken to be identical for each stage. Re-arranging, and performing the
summation gives the steam duty

UA  n −1

S≈ Ts − Tn − ∑ BPRi  (109)
n∆Hvap  i =1 

For a single stage evaporator this gives

UA
S1 ≈ (Ts − T ) (110)
∆Hvap

By comparing (109) and (110) we see that if steam is provided at the same temperature
(Ts), if U and A are identical, and if the temperature of the single stage equals that of the
final stage in the multi-stage process (T = Tn), then, ignoring boiling point rise, the single-
stage evaporator will use n times as much steam as the multi-stage process.

However, for a multi-effect process as boiling point rise increases then for the same steam
duty the output temperature of the last stage, Tn, must fall. Because Tn cannot be lower
than the freezing temperature of the solvent, the maximum number of stages, and hence
the maximum efficiency, must fall as the boiling point rise increases. This makes the
importance of boiling point rise clear.

As the number of effects/stages increases the capital cost will increase but the steam duty
will decrease. Optimisation is required to determine the most economic number of effects
for a given separation.

7.5 Further operational considerations

Sometimes it is useful to operate an evaporator at low pressure (under vacuum). This
reduces the temperature of each stage, which has some advantages. In particular, some
solutions are heat sensitive (e.g. fruit juice) and so the quality of product is improved if the
process is operated at lower temperatures. Nevertheless, increased capital and operating
costs are associated with operating under vacuum. The reduced temperature also tends
to increase the liquid viscosity, which reduces the heat transfer coefficient.

70