Department of Chemical Engineering

College of Engineering
University of the Philippines
Diliman, Quezon City
1st Semester A.Y. 2020-2021

ChemE 132
Separation Processes I

MODULE 06
Mass Transfer Coefficients

RECALL: From the discussions in ChemE 130, when a stream of fluid with an average velocity
ū is flowing past a solid surface, the non-slip condition that compels the fluid molecules at the
solid-fluid interface to remain stagnant (u = 0) creates a velocity gradient that drives the
momentum transfer from the bulk fluid to the surface. The velocity gradient is more prominent
at the region very close to the surface of the solid, called the boundary layer.

The boundary layer acts as an interphase: an intermediate

phase between the bulk fluid and the solid. Thus, while it is
part of the fluid phase, its properties and behavior are
different from those of the bulk fluid region due to its close
proximity to the solid surface. The distinction between the
bulk fluid and the boundary layer is defined by the boundary
layer thickness δ, which was arbitrarily set to be the region
where u = 0.99ub = 0.99ū.

Similarly, the occurrence of diffusion of species A from the

solid surface to the flowing fluid species B is affected by this
difference in fluid behavior at the mass boundary layer and
at the bulk fluid.

While there is a distinct concentration gradient along the boundary layer thickness, the
transport of mass near and within the bulk fluid is more rapid due to the presence of eddy
currents that can transport a larger quantity of the solute at a given time.

To account for the increased diffusion rate due to the eddies, Fick’s law is expressed as
dc
J*A =− (DAB + ε M ) A , where ε M = mass eddy diffusivity
dz
 D + εM 
Solving the differential equation gives J*A =−  AB  ( c A 2 − c A1 )
 z 2 − z1 

The problem with using this solution is that both the mass eddy diffusivity and the diffusion
path length ∆z is very difficult to determine experimentally.

1
Thus, for convenience, the variables are lumped into a single coefficient of ∆cA, called the
mass transfer coefficient:

−k′c ( c A 2 − c A1 )
J*A =

Because Fick’s law is expressed in a number of ways, the mass transfer coefficient also
changes according to the associated driving force. Listed below is a summary of the
expressions for the mass transfer equations using liquid and gas phase coefficients:

gradient liquid phase gas phase

concentration −k′c ( c A 2 − c A1 ) −k′c ( c A 2 − c A1 )

−k′L ( c A 2 − c A1 )
partial pressure −k′G ( p A 2 − p A1 )

mole fraction −k′x ( x A 2 − x A1 ) −k′y ( y A 2 − y A1 )

The subscript ‘c’ denotes a gradient expressed in terms of concentration and is applicable to
both liquid and gaseous phases.

The subscripts ‘L’ and ‘G’ denotes a coefficient applicable to liquid-side diffusion and gas-side
diffusion, respectively. These coefficients are used when analyzing mass transfer across a
liquid-gas boundary.

The subscripts ‘x’ and ‘y’ denotes a gradient expressed in terms of liquid mole fractions and
gaseous mole fractions, respectively.

It is important to remember that the units of the mass transfer coefficients are such that the
unit of the molar flux is always in mol/s⋅m2 (or equivalent).

EXAMPLE 01. Given the mass transfer equation

−k′c ( c A 2 − c A1 )
NA =

What is the unit of the mass transfer coefficient if cA is in mol⋅m-3?

mol  mol  m
2 [ ] c
Solution: The corresponding equation for units is = k′ 3 
→
= k′c [ ]s
s⋅m m 

m
Therefore, the unit of the coefficient is
s

which means that k′c = diffusion velocity

2
 dc A c A
− (DAB + ε M )
For the molar flux of A: NA = + (NA + NB )
dz ct

p dy A
− (DAB + ε M )
And for ideal gases, NA = + y A (NA + NB )
RT dz

The convective term can be treated using the two special cases:

1. Equimolar counter-diffusion. The convective term cancels out, leaving

p dy A  D + εM  p
− (DAB + ε M )
NA = → NA =
−  AB  ( y A 2 − y A1 )
RT dz  z 2 − z1  RT

DAB + ε M k′ p
But k′c
=→ − c ( y A 2 − y A1 )
NA =
z 2 − z1 RT

−k ′y ( y A 2 − y A1 )
From the table of coefficients: NA =

k′cp  mol 
Therefore, k′y = in
RT  s ⋅ m2 


2. Diffusion through a stagnant fluid. Since NB = 0, the equation becomes

− (DAB + ε M )
NA =
p dy A
+ y A (NA ) → NA =
(D + ε M ) p dy A
− AB
RT dz 1 − y A RT dz

 D + εM  p  1− yA2 
Solving the differential equation: NA =  AB  ln  
 z 2 − z1  RT  1 − y A1 

 D + ε M  p y A 2 − y A1
or NA = −  AB 
 z 2 − z1  RT (1 − y A )lm

 D + εM  p k′y
but  AB  k′y
=→ NA =
− ( y A 2 − y A1 )
 z 2 − z1  RT yB,lm

−k y ( y A 2 − y A1 )
or NA =

mol
Like k′y , the unit of k y is .
s ⋅ m2

3
EXAMPLE 02. A large volume of pure gas B at 2 atm is flowing over a surface from which
species A is vaporizing from a liquid layer at 298 K. The vapor pressure of A at 298 K is 0.20
atm, and the mass transfer coefficient k’y has been estimated to be 6.78×10-5 kmol/s⋅m2.
Calculate molar vaporization flux of A, and the mass transfer coefficients ky and kG.

Assume:
1. Steady state mass transfer
2. 1D transport along the direction normal to the surface of the liquid layer
3. Ideal gas behavior
4. Constant properties since no temperature change
5. Gas species B is insoluble in liquid layer

Since B is insoluble to the liquid, it can be considered as “stagnant”. Thus, the molar flux
equation for A is NA =−k y ( y A 2 − y A1 ) .

To solve for ky, the mole fractions of B at the boundary conditions are used:

0.20 atm
at z = z1, p A = pvap
A → y A1 = = 0.10 → yB1 = 0.90
2 atm
at z = z 2 , p A = 0 → yA2 = 0 → yB2 = 1.0

1.0 − 0.90
=
Solving for the mean mole fraction: yB,lm = 0.9491
 1.0 
ln  
 0.90 
kmol
k′y 6.78 × 10−5
Therefore,=
ky = s ⋅ m2 → k = 7.144 × 10 −5 kmol
s ⋅ m2
y
yB,lm 0 .9 4 9 1

−k y ( y A 2 − y A1 )
Solving for the molar flux: NA =

kmol
=NA −7.144 × 10−5 ( 0 − 0.10 )
s ⋅ m2

kmol
NA = 7.144 × 10 −6
s ⋅ m2

−k y ( y A 2 − y A1 ) =
To solve for kG: NA = −k G ( p A 2 − p A1 ) but p A= p ⋅ y A

→ − k y ( y A 2 − y A1 ) =−k Gp ( y A 2 − y A1 )

kmol
ky 7.144 × 10 −5 kmol
Therefore, k= = s ⋅ m2 → k G = 3.525 × 10
−10
G
p  1.01325 × 10 Pa 
5
s ⋅ m2 ⋅ Pa
( 2.0 atm )  
 1 atm 

4
MASS TRANSFER CORRELATIONS
Like the diffusion coefficient, the value of the mass transfer coefficient can be either measured
experimentally or estimated using empirical correlations. Many empirical correlations have
been developed to analyze the flow of fluid in a conduit accompanied by the occurrence of
mass transfer, and such correlations take into account several factors that contribute to the
convective mass transfer:

- For the fluid flow, the pertinent variables are fluid density (ρ), average or bulk velocity
(u), fluid, and the diameter of the conduit or pipe (d).

- For the mass transfer, the pertinent variables are the mass diffusivity (DAB), the mass
transfer coefficient (kc’), and also the diameter of the pipe (for the transfer area).

The following is the summary list of the pertinent variables and their respective fundamental
units (M for mass, L for length/distance, and T for time):

variable unit
density (ρ) M
L3
bulk velocity (u) L
T
viscosity (µ) M
L⋅T
diameter (d) L
diffusivity (DAB) L2
T
mass transfer coefficient (kc’) L
T

Because there are too many variables to consider, a systematic approach must be adopted to
establish the correlations between these parameters. One way of establishing the correlation
is to apply the Buckingham-Pi Theorem: the number of dimensionless groups (or pi groups)
is equal to the number of variables less the number of fundamental units associated with these
variables, and one of these pi groups is going to be expressed as a function of the other pi
groups. Thus, for the mass transfer correlation:

no. of pi groups = 6 − 3 = 3 and so π 1 = f (π 2 ,π 3 )

In setting up these dimensionless groups, it is a standard practice to keep three (3) variables
the same/always present in all the groups:

1. The geometric variable, which will provide the unit of length. For this case, the
variable assigned is the pipe diameter.

2. The kinematic variable, which will provide the unit of time without any associated
quantity or mass. For this case, the variable assigned is the mass diffusivity.

5
 3. The dynamic variable, which will provide the unit of quantity or mass. For this case,
the variable assigned in the fluid density.

The rest of the variables will be distributed among the 3 pi groups: π 1 = f (DAB , ρ , d, k′c )
π 2 = f (DAB , ρ, d, µ )
π 3 = f (DAB , ρ, d, u )

Since all the groups should be dimensionless, the 3 common variables in each group are
raised to a power such that all the units will cancel out:

π 1 = (DAB ) ( ρ ) ( d) k′c
a1 a2 a3

π 2 = (DAB ) ( ρ ) ( d) µ
a4 a5 a6

π 3 = (DAB ) ( ρ ) ( d)
a7 a8 a9
u

a1
 L2   M 
a2
L
Examining π1: 1 =    3  (L )
a3
T
 T  L   

The sum of the powers for each fundamental unit must be equal to zero:

L : 0 = 2a1 − 3a2 + a3 + 1
M : 0 = a2
T: 0=−a1 − 1
 a1   −1
Solving the systems of linear equations gives a2  =  0 
a3   1 

Substituting the solved values give π 1 = (DAB ) ( ρ ) ( d) k′c

−1 0 1

k′c d
π1
or= = NSh , where NSh is the Sherwood number
DAB

The Sherwood number is the ratio of diffusive mass transfer resistance to the convective
mass transfer resistance. It is also the mass transfer equivalent of heat transfer’s Nusselt
number, which is used to estimate the value of the convective heat transfer coefficient.

Examining π2 yields π 2 = (DAB ) ( ρ ) ( d)

−1 −1
µ
0

µ
=
or π2 = NSc , where NSc is the Schmidt number
ρDAB

The Schmidt number is the ratio of momentum diffusivity to mass diffusivity.

6
 d⋅u
Examining π3 yields π 3 = (DAB ) ( ρ ) ( d) u
−1
or π 3 =
0 1

DAB

By itself it does not appear familiar; however, if π3 is divided by π2:

 d⋅u 
 
π3  DAB  π d⋅u⋅ ρ
= =
or 3 = NRe , where NRe is the Reynolds number
π2  µ  π2 µ
 
 ρDAB 

Thus, to estimate or predict the value of the mass transfer coefficient, an empirical
correlation for the Sherwood number as a function of the Reynolds number and the
Schmidt number must be employed: NSh = f (NRe , NSc )

There are many reported empirical correlations for the Sherwood number for different types
of mass transfer systems. Because they are empirical models, their applicability is limited to
the nature of the experimental setup used to generate the mass transfer data; however, in the
absence of any established models, existing correlations may be used with caution (for
example, a correlation for diffusion in a spherical catalyst may be applied with some acceptable
errors to analyze mass transfer in non-spherical particles).

Here are three examples of mass transfer models that have been reported:

1. For mass transfer in flow parallel to flat plates:

k′cL
=
NSh = 0.664NRe,L 0.5
N1/3
Sc (NRe,L < 15,000)
DAB

2. For mass transfer for flow past a single sphere:

k′ d
NSh= c p= 2 + 0.6NRe,p
0.5
N1/3
Sc

(Frossling correlation)
DAB

3. For mass transfer in turbulent flow to pipe wall:

k′c d
=
NSh = 0.332NRe 0.5 1/3
NSc
DAB

Other correlations may be found in reference textbooks and published research articles.

REFERENCES:
1. Geankoplis, C. J., Hersel, A. H., and Lepek, D. H. (2018) Transport Processes and
Separation Process Principles 5th Ed., London, UK: Pearson Higher Education.
2. Welty, J. R., Rorrer, G. L., and Foster, D. G. (2019) Fundamentals of Momentum, Heat,
and Mass Transfer 7th Ed., NJ: John Wiley and Sons, Inc.