Multi-Component

Distillation
 Introduction to Multicomponent Distillation

Most of the distillation processes deal with multicomponent mixtures.

Multicomponent phase behaviour is much more complex than that

for the binary mixtures

Rigorous design requires computers

Short cut methods exist to outline the scope and limitations of a

particular process.
Continued

Objective of any distillation process is to recover pure products.

In case of multi component mixtures, we may be interested in

one, two or more than two components in pure form.
D, xD,1

F One distillation column can be optimized to separate one pair

zF,1 of volatile components.
zF,2

B, xB,1
 How many distillation columns are required ?

1 Alternative?

1
1, 2
II F 2
2,3,4 3
F 3,4
zF,1 2
zF,2 I 4
zF,3
zF,4

3, 4 III
Separation of C components requires (C-
1) distillation columns.

4
 Unlike in binary distillation, fixing mole fraction of one of the components
in a product does not fix the mole fraction of other components.

On the other hand fixing compositions of all the components in the

distillate and the bottoms product, makes almost impossible to meet
specifications exactly.

We can specify two fractional recoveries:

FR1 = (DxD,1)/(FzF,1) = 0.95
FR2 = (BxB,2)/(FzF,2) = 0.98

If the feed contains more than two volatile components, we cannot specify the
recoveries of the additional components.
Key components

In practice we usually choose two components separation of which serves

as an good indication that a desired degree of separation is achieved.

These two components are called key components:

- light key
- heavy key

There are different strategies to select these key components:

Choosing two components that are next to each other

on the relative volatility.

Based on distribution of component.

 Distributed and undistributed components

Components that are present in both the distillate and

the bottoms product are called distributed components.

The key components are always distributed components.

Components with negligible concentration (<10-6) in one

of the products are called undistributed.

A B C D E G

light non-distributed components key heavy non-distributed components

(will end up in the overhead product) components (will end up in bottoms product)
 Each column is designed to separate two components of adjacent relative
volatility. These components are the keys.
All other components are non-keys.

component a designation
1 1.5 Light non-key (LNK) assume exclusively in distillate
design for 2 1.4 Light key (LK) specify recovery in distillate
separation 3 1.3 Heavy key (HK) specify recovery in bottoms
4 1.2 Heavy non-key (HNK)
assume exclusively in bottoms
5 1.0 Heavy non-key (HNK)

Distributions in the distillate and bottoms streams are specified for the two key
components.
If we assume that the non-keys do not distribute, the overall mass balance is
easily solved.
 Overall mass balance
with non-distribution of non-keys
If non-keys do not distribute, then
LNK: Dxi,D = Fzi,F
HNK: Bxi,B = Fzi,F

Define fi, di, and bi as molar flow rates of component i in feed, distillate and
bottoms, respectively.
Ex.: 100 mol/h of an equimolar mixture of four components.
Assume 80 % recovery of LK in D, and 80 % recovery of HK in B.
component i zF,i fi = F zF,i di xD,i = di/D bi xB,i = bi/B
LNK 0.25 25 25 0.50 0 0
LK 0.25 25 20 0.40 5 0.10
HK 0.25 25 5 0.10 20 0.40
HNK 0.25 25 0 0 25 0.50
di = D = 50 xD,i = 1 bi = B = 50 xB,i = 1
Hengstebeck and Geddes Equation:

Exact distribution of non - key components in top product and bottom product is not possible
At the starting of design.

But the same can be determined approximately by Hengstebeck and Geddes equation,

di
log A C log a i
bi

Where,
di = moles of ith component in distillate
bi = moles of ith component in residue
i = relative volatility of ith component with respect to heavy key component
A, C = constants
Ex.6. A bubble cap fractionating column consisting of 12 plates working at an average
efficiency of 75% is being used to distill 1000 kg/hr of aqueous methanol entering the tower.
The feed consists of 3/4th vapor and 1/4th liquid. The feed, overhead product and bottom
product are 50 mole %, 90 mole % and 10 mole % methanol respectively. A total condenser
is provided. The reflux is sent at its saturation temperature. If the reflux ratio is 1.7 times
the minimum check whether the column available is satisfactory. VLE DATA is given below:

X 8 10 20 30 40 50 70 80
y 36.5 41.8 57.9 66.5 72.9 77.9 87 95.8
The calculation of the number of equilibrium stages in multi component distillation:

Method Valid for Multi component Distillation (MD):

Methods available for MD can be broadly divided into two groups:

1. Short Cut Methods: FUG method, Smith Brinkley Method etc.

2. Rigorous Method:
Lewis Matheson Method
Thiele Geddes Method
Relaxation Methods
Tridiagonal Matrix Method
Linear Algebra Method etc.

Short Cut Methods:

an estimation of the minimum number of trays
the estimation of minimum reflux rate and
number of stages at finite reflux

A widely used approximate method is commonly referred to as the

Fenske-Underwood-Gilliland method
 FUG shortcut methods
for multicomponent distillation

Suitable for preliminary designs

Three-part F-U-G method:

1. Use Fenske equation to estimate Nmin (at total reflux)

2. Use Underwood equations to estimate Rmin (with N = )
3. Use Gilliland correlation to estimate Nactual

And for Nfeed one can also use Kirkbride equation

 Operating at total reflux

designate the keys A and B, and assume constant aAB

yi,1 according to the definition of equilibrium:
1
yA xA
= a AB
xi,0
y B N+1 xB N+1
at total reflux, the operating line is y = x
yi,j+1 = xi,j so yi,N+1 = xi,N
xA xA
= a AB
yi,N+1 xB N xB N+1
N
move up to stage N:
xi,N
N+1 yA xA xA
= a AB = a AB
2

y B N xB N xB N+1
continue to the top of the column
 Estimating Nmin
xA y A Nmin xA
at the top of the column = = a AB
xB 0 y B 1 xB N+1

x xA
solve for Nmin Fenske ln A


equation xB 0 xB N+1
Nmin =
lna AB

can also write in terms of fractional recoveries (FR)

FR 1- FRB
ln A
FR
1- FR A distillate B bottoms

Nmin =
ln a AB
 Other uses of the Fenske equation
Once Nmin is known,
check non-distribution of non-keys, or estimate their distribution:

a CB
N min

FRC,distillate =
FRB
1- FR + a Nmin
CB

B bottoms

estimate optimum feed stage at total reflux:

x zA
ln A
z
x
B 0 B F
NF,min =
lna AB
Minimum reflux ratio, using Underwoods method:

Where,
A, B, C, etc., are the volatilities with respect to the least volatile component,
and
is the root of equation 1, which lies between the values of A and B.
 Gilliland Correlation
We have Nmin (Fenske) and Rmin (Underwood).
R is often specified as a multiple of Rmin, e.g., R = 1.5 Rmin

y=

x=

Later Molokanov et al. (1972) represented the Gilliland correlation as,

1+54.4 x x-1
0.5
11+117.2x x
y = 1- e
 Optimum Feed Stage

Obtain Nmin and NF,min from Fenske eqn

Obtain N from Gilliland correlation

Assume relative position of the feed stage D

is the same as it was at total reflux: D
NF,min NF F F
=
Nmin N NF,min NF

Remember! Nmin and N include partial

reboiler (and partial condensor).

Works well for symmetrical columns B

(feed stage close to middle).
B
 Kirkbride equation
Better feed stage estimation for
unsymmetrical columns:
F D

0.206
Nrectifying z x 2
B
= HK LK ,B
Nstripping zLK xHK ,D D
F
D

where Nrect and Nstrip do not include the

feed stage: B
F
N = Nrect + Nstrip + 1

B
Example 7: A feed 100 kmoles /h of saturated liquid containing 10 mole % LNK, 55 mole %
LK, and 35 mole % HK and is to be separated in a distillation column. The reflux ratio is 1.2
times of the minimum. It is desired to have 99.5 % recovery of the light key in the distillate.
The mole fraction of the light key in the distillate should be 0.75. Equilibrium data: LNK = 4.0,
LK = 1.0,HK = 0.75.
Find
(i) Minimum number of stages required by Fenske method.
(ii) Minimum reflux ratio by Underwood method.
(iii) Number of ideal stages at R = 1.2 Rmin by Gilliland method.
(iv) Also find the number of ideal stages at rectifying section and the stripping section at the
operating reflux ratio and location of feed stage.
SOLUTION:

K
Using following equation,
x xA
ln A
x
x
B 0 B N+1
Nmin =
lna AB

Nmin =
(ii) To find the minimum reflux at the condition of saturated liquid, q = 1,
Using following equation:

Gives = 0.83

Rmin= 2.47

(iii) Now using the Gilliland correlation to determine number of

ideal stages at R = 1.2 Rmin =2.97

We can get N = 48.89

(iv) Using Kirkbride Equation

NR =

NS =
Lewis Matheson Method, Rigorous Method for Multicomponent Distillation

Method assumes the constant molar overflow

To use this method following information must be specified.

1. Feed flow rate, conditions and compositions

2. Product compositions or distribution of key components.
3. Reflux ratio
4. One product flow
5. Column pressure
6. Assumed values for the distribution of the non-key components

Method involves:

1. If the composition of the liquid on any plate is known, then the composition of the vapour
in equilibrium is calculated using relative volatilities of the individual components.

2. The composition of the liquid on the plate above is then found by using an operating
equation.
If a mixture of components A, B, C, D, has mole fractions xA, xB, xC, xD, in liquid and yA, yB,
yC, yD, in the vapour, then,

(1)
Alternatively,

n
xi

i 1
a ihk xi
yi
(1) =>
 Multi Component Binary System

a1 x1 ax
y1 y
a i xi 1 a - 1x
Ex.8. A mixture of ortho, meta, and para-mono nitro toluenes containing 60, 4, and 36 mole per
cent respectively of the three isomers is to be continuously distilled to give a top product of 98
moleper cent ortho, and the bottom is to contain 12.5 mole per cent ortho. The mixture is to be
distilled at a temperature of 410 K requiring a pressure in the boiler of about 6.0 kN/m2. If a reflux
ratio of 5 is used, how many ideal plates will be required and what will be the approximate
compositions of the product streams? The volatility of ortho relative to the para isomer may be
taken as 1.70 and of the meta as 1.16 over the temperature range of 380415 K.

Solution:
For 100 kmol of feed with D and W kmol of product and bottoms, respectively and xdo and xwo the
mole fraction of the ortho in the distillate and bottoms, then an overall material balance gives:

100 = D + W
As a first estimate, it is supposed that the distillate contains 0.6 mole per cent meta and
1.4 mole per cent para:
Ortho:
 And above the feed plate:

Now, starting the calculation from bottom:

The composition of the vapour in the still is found from the relation

yso = oxso/ xs .
Therefore, yso = 0.191

1.7
1.16
1
Ex. 9.
 Pinch points in binary distillation
Recall for binary distillation: Binary composition profile

Rmin xD xD

pinch
point
zF
x
plateau of
zF constant
composition

xB
Using Rmin generates
a pinch point at the condensor feed reboiler
xB feed stage stage number

near the pinch point, composition changes little from stage to stage
- passing streams are very close to equilibrium
- no change in temperature between stages
 Behavior in ternary systems

in the presence of a third component whose Ternary composition profile

composition does change through the LK-HK-HNK
1
feed stage, the pinch point moves stripping
rectifying

Consider LK
LK is the most volatile component in the
xi
system on every stage
there is almost no LK at the reboiler stage

Consider HNK
HNK is the least volatile component in the 0
C F R
system on every stage
stage number
almost all HNK is found at the reboiler stage
finite HNK at feed stage, drops off rapidly above
 Behavior of HK

Ternary composition profile

LK-HK-HNK
1
Consider HK:
pinch
HK behavior is most complex point
above feed stage, distillation is almost binary

~binary
(HK-LK), with HK the less volatile component xi

~binary
near reboiler, distillation is almost binary
(HNK-HK), but HK is the more volatile
component
HNK goes through a maximum in the stripping 0
C F R
section, creating a pinch point there
stage number
 Pinch points in multicomponent distillation

pinch pinch
F F F
LK, HK, LNK, LK, LNK, LK,
HNK pinch HK HK, HNK pinch

LK-HK-HNK system LNK-LK-HK system LNK-LK-HK-HNK

has a pinch point in has a pinch point in has pinch points in
the stripping section the rectifying section both sections

Graphical methods are not helpful in finding these pinch points.

 Second Underwood equation
CMB in rectifying section at minimum reflux: Vmin y i,j +1 = Lmin xi,j + Dxi,D
at the pinch point: xi,j = xi,j +1
VLE: y i,j +1 = K i xi,j +1
relative volatility: a i = K i Kref y i,j +1 = a i Kref xi,j +1

Lmin Lmin
Vmin y i,j +1 = y i,j +1 + Dxi,D Vmin y i,j +1 1- = Dxi,D
a i Kref a iVminK ref

a i Dxi,D Lmin
Vmin y i,j +1 = define: f =
Lmin VminK ref
ai -
VminK ref

a i Dxi,D
2nd Underwood equation: Vmin = Vmin y i,j +1 =
i i ai - f
 First Underwood equation
a i Bxi,B
in the stripping section: -Vmin = Vmin y i,k +1 =
i i ai - f

assume f =f
a i Dxi,D a i Bxi,B
DVfeed = Vmin -Vmin = +
i a
i - f a i
- f
a i Fzi,F
1st Underwood equation: DVfeed =
a i - f
i

feed quality definition: DVfeed = F(1- q)

searching for:
with C components, the 1st Underwood equation has C roots, one between
each adjacent pair of a-values, and one between 0 and the smallest a-value.
We need the root aHK < < aLK
binary system: use quadratic equation to solve for
trial-and-error: use (aHK + aLK) as initial guess
 Estimating Rmin
Onceis known, obtain Vmin using the 2nd Underwood equation.

a i Dxi,D
Vmin = Vmin y i,j +1 =
i i ai - f

To get xi,D values for non-keys:

assume non-distribution
OR
assume distribution is the same at minimum reflux and total reflux
i.e., use Fenske equation

Lmin = Vmin D

Rmin = Lmin/D
 Sandwich components
At the beginning of this section, we specified that the key components should
have adjacent relative volatilities. What if there is a non-key with intermediate
volatility (i.e., between the two key components)?

This component is called a sandwich component.

component a designation
1 1.5 Light non-key (LNK) D
2 1.4 Light key (LK) F
3 1.3 Sandwich non-key S
4 1.2 Heavy key (HK)
5 1.0 Heavy non-key (HNK)

Sandwich components tend to concentrate in the

middle of the column, and can cause flooding even B
when present at trace concentrations.

If unavoidable, use side-streams to withdraw them.