Chemical Engineering Science 57 (2002) 4691 4696

www.elsevier.com/locate/ces

Modeling in chemical engineering

Octave Levenspiel
Department of Chemical Engineering, Oregon State University, Corvallis, OR 97331, USA

Abstract
In its 90 year life what has chemical engineering (ChE) contributed to society? Firstly, we have invented and developed processes to
create new materials, more gently and more e-ciently, so as to make life easier for all.
Secondly, ChE has changed our accepted concepts and our ways of thinking in science and technology. Here modeling stands out as
the primary development. Let us consider this.
? 2002 Elsevier Science Ltd. All rights reserved.

1. Chemical engineerings $10 and $100 ow models

2. The boundary layer concept

I suspect that most authors put especial thought into the

1rst sentence of a book because it sets the tone for what is to
follow. In this light Denbighs monograph (Denbigh, 1951)
starts with the following sentence:

How is the velocity of a >uid aDected as it >ows close

to a solid surface? Throughout the years many scientists
considered this phenomenon. These studies culminated in
the KarmanPrandtl u vs. y relationship, see von Karman
(1934),



y 0 =
u


+ 5:56;
= 5:75 log

0 =

In science it is always necessary to abstract from the complexity of the real world, and in its place to substitute a
more or less idealized situation that is more amenable to
analysis.
This statement applies directly to chemical engineering, because each advancing step in its concepts frequently starts
with an idealization which involves the creation of a new
and simpli1ed model of the world around us. The acceptance of such a model changes our world view.
Often a number of models vie for acceptance. Should we
favor rigor or simplicity, exactness or usefulness, the $10 or
$100 model? We will look at:

boundary layer theory,

heat transfer and h,
mass transfer and kg and k ,
chemical reactors, RTD, tracer technology, and
the troublesome >uidized bed.

We may call the 20th century chemical engineerings modeling century. This talk considers this whole development.

Tel.: +1-541-737-3618; fax: +1-541-737-46-00.

E-mail address: [email protected] (O. Levenspiel).

where
y0
u=



when u 11:84

0

which is shown in Fig. 1a.

However, Prandtl (1904) proposed a simpler model for
the velocity pro1le: thus a linear velocity change with distance from the surface, or
du
= constant and u = 0 at y = 0
dy
meeting nonviscous >ow further away from the surface; in
eDect, two diDerent types of >uid patched together. This is
shown in Fig. 1b.
This combination of viscous and nonviscous >uid seemed
absurd at that time, but as we shall see, it has been found to
be fabulously e-cient and useful for aeronautics, chemical
and other branches of engineering.
Let me repeat a few words about the creator of the laminar
layer model, words written by his most illustrious student,
Theodore von Karman, director of the Guggenheim Aeronautical Laboratory at the California Institute of Technology
(von Karman, 1954).

0009-2509/02/$ - see front matter ? 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 0 9 - 2 5 0 9 ( 0 2 ) 0 0 2 8 0 - 4

4692

O. Levenspiel / Chemical Engineering Science 57 (2002) 4691 4696

change beyond, as shown in Fig. 2b. Thus, close to the cold

surface he said



replace the  dT
KT
by

Ky across
changing : : : dy close to
the surface

the 1lm

in which case the Fourier heat transfer equation becomes



KT
k
q = kA
=
AKT:
across
the
Ky
Ky
whole 1lm

Calling k=Ky = h he 1nally got

Fig. 1. Velocity pro1le near a wall: (a) actual pro1le, and (b) simpli1ed
pro1le.

q = hAKT;
where h is called the heat transfer coe-cient.
The acceptance and use of this concept led to explosive
research. Thousands of research studies were reported, and
are still being reported today, on h for

Fig. 2. Temperature by a wall: (a) actual pro1le, and (b) simpli1ed pro1le.

Prandtl (1875 1953) was an engineer in training. His

control of mathematical methods and tricks was limited;
many of his collaborators and followers surpassed him
in solving di-cult mathematical problems. However he
had a unique ability to describe physical phenomena in
relatively simple terms, to distill the essence of a situation
and to drop the unessentials. His greatest contribution is
in boundary layer theory.
So here we have two models for >ow close to a surface,
the more precise $100 KarmanPrandtl equation and the
simple $10 laminar layer or 1lm model. We will see that
chemical engineering embraced the $10 model and made it
a cornerstone for its developments in the 20th century.
3. Film model for heat transfer
Twelve years after Prandtl charted the path with his simple >ow model, W. K. Lewis of M.I.T. (Lewis, 1916), following the ideas of Newton, Fourier, and other early workers, adopted the simple 1lm model to represent the rate of
heat transfer from a hot >uid >owing past a cold surface.
This means that instead of considering a changing temperature gradient away from the cold surface, Fig. 2a, he took
a straight line gradient in the 1lm, but with no additional

>ow inside and outside of pipes,

condensing >uids,
boiling >uids,
natural convection and
two phase systems of all types of gases, liquids, and solids.

The resulting correlations were all made in terms of dimensionless groups, examples of which are shown below.
For the inside pipe wall (turbulent >ow):

 0:7 
hd
d
= 0:023
1+
k
L

Nusselt

d
1 + 3:5
dcoil

coiled pipe

entrance
Re0:8

Pr 1=3

0:14

w

temperature
dependency

1=2
Single particle : hD
Pr 1=3 ,
k = 2 + 0:6 Re
Natural convection : hL=k = A[Gr Pr]B ,

Condensation
hundreds of studies
Boiling
: : : etc:
hundreds of correlations:
Radiation

4. Film model for mass transfer

It took another 7 years before Whitman (1923) considered
applying the 1lm model to mass transfer from a >uid to its
bounding surface. With an approach similar to that used by
Lewis he replaced the changing concentration pro1le of a
component A, Fig. 3a, by a linear concentration pro1le as

O. Levenspiel / Chemical Engineering Science 57 (2002) 4691 4696

4693

Fig. 4. Counter>ow heat interchange, plug >ow idealization.

Fig. 3. Mass transfer: (a) actual pro1le, and (b) simpli1ed pro1le.

shown in Fig. 3b. In symbols the absorption rate by the solid,

or removal rate of A from the >uid is


dCA

N A = DA
(mol A removed=s):
dy at surface
With the linear model


D
KCA

=
AKCA = kg A(KCA ) across the
N A = DA
Ky
Ky
whole 1lm

Fig. 5. Countercurrent >ow mass transfer: (a) plug >ow model, and (b)
axial dispersion model.

where kg (or k ) is called the mass transfer coe-cient.

Again, dimensionless groups are involved in the correlations for the various mass transfer situations, as shown below (Perry & Green, 1984).
For gas absorption in columns 1lled with Raschig rings,
and Berl saddles.
2=3

Gg dp 0:36 
kg
)
For the gas phase :
=1:195(
.
Gg
g (1 )
D g
0:45  0:5

G dp
k dp

=25:1
.
For the liquid phase :
D
D
 (1 )

patternsplug >ow, mixed >ow, laminar >ow, crosscurrent

>ow, etc. However, we should be aware that the logarithmic
mean is only correct and should only be used to represent
cocurrent and countercurrent plug >ow. For all other >ow
patterns it is incorrect, often very incorrect to use it.
For mass transfer of a component A from a gas to a liquid
we have a similar development, an overall mass transfer
coe-cient Kg , which is de1ned as


Henry s law constant
1
1
HA
=
+
; HA =
:
Kg
kg
k
for A

5. Design consequences
These concepts of h for heat transfer and k for mass
transfer became the heart of design methods for heat exchangers and for absorption and extraction equipment.
For heat transfer from one >uid through a wall to a second
>uid this leads to the overall heat transfer coe-cient U given
as
1
1
1
+
=
U
h1
h2
and for a heat exchanger for counter (or parallel)->ow of
two >uids a $10 >ow model assumes plug >ow of the two
>uids, as shown in Fig. 4. For this model the performance
expression is
q = UA(KT )m ;

where KT = Ti ti ;

with the logarithmic mean KT is

KT1 KT2
:
KTm =
ln(KT1 =KT2 )
It should be noted that the logarithmic mean driving force
is often used to represent design of all sorts of contacting

In a packed bed tower, with countercurrent >ow of gas and

liquid, the design method, universally used today, is based
on the crude $10 model for the >ow of gas and of liquid,
the plug >ow model, Fig. 5a. To be more realistic, to assume dispersed plug >ow, the operating line would look
diDerent. This model is illustrated in Fig. 5b. This more realistic approach, this $100 >ow model, was championed by
Vermeulen et al. (1966), but it had problems:
It required a complex computer solution.
We did not have reliable values for the extent of dispersion, or deviation from plug >ow for gas and for liquid.
Every time you pack a column you get a diDerent >ow
behavior.
So this $100 approach has been lost in history.
For heat exchanger design we get similar >ow complications if we try to account for deviations from plug >ow.
6. Reections on the 'rst 50 years of chemical engineering
Prandtls simple model was taken up by Lewis for heat
transfer, and by Whitman for mass transfer, and it led

4694

O. Levenspiel / Chemical Engineering Science 57 (2002) 4691 4696

ultimately to the core of the unit operations. Our profession

has really blossomed as a consequence of this simple concept. Can you imagine what heat and mass transfer studies
would be like without h and k? A desert.
7. Chemical reactors
In the 1rst half of this century two >ow models dominated the design of >ow reactors, plug /ow and mixed /ow.
Fig. 6 shows, for the reaction given, if a plug >ow reactor
would need a length of 1 m, then for mixed >ow you would
need a length of 1000 m to produce the same product.
It would be unwise to only use these >ow models, since
real reactors behave somewhere between these two extremes. How do we deal with this? The $100 approach says
evaluate the velocity 1eld within the reactor, or better still,
the $1000 approach says evaluate the three-dimensional
>uctuating velocity 1eld, and then use your computer to
tell you what would happen; in essence, use computational
>uid dynamics. What an ugly procedure!
In so complicated a world what should we do? It was
the genius of Danckwerts (1953) who proposed a ridiculously simple >ow model to tell how a vessel would act as
a chemical reactor. He said, introduce a pulse of tracer into
the >uid entering the reactor and see when it leaves. This
exit concentrationtime curve is called the residence time
distribution curve, or RTD curve, see Fig. 7.
This information tells how the reactor would behave
exactly for linear reaction kinetics, and as a close approximation for more complex reaction kinetics. How wonderfully little information is needed! No need to measure what
is happening within the reactor.
This proposal of Danckwerts led to an explosion of research, and to the development of all sorts of models to represent these response curves. Fig. 8 shows one class of these
models, the one to represent the >ow in pipes and in other
long narrow vessels. Lots of research, lots of publications
and lots of Ph.D. theses resulted from this concept, mainly
in the late 1950s, the 1960s and the 1970s.
The study of the RTD of >owing >uids, and its consequences is called tracer technology. Chemical engineers and
medical researchers are those who are most interested in this
subject, to the chemical engineer to represent the behavior
of reactors, to the medical doctor to represent the movement
and distribution of >uids and drugs in the body, to diagnose
disease, etc.
8. Fluid sciences
Let us brie>y summarize the science and technology of
>uid >ow.
In the 19th century two completely diDerent approaches
were used to study the >ow of >uids. First there was the
theoretical study called hydrodynamics, which dealt with

Fig. 6. Volumes needed for the two ideal >ow patterns: (a) plug >ow
model, and (b) mixed >ow model.

Fig. 7. Key to the tracer method for determining the >ow pattern in
vessels and reactors.

the ideal frictionless >uid (that imaginary stuD). This was

a highly mathematical exercise dealing with what is called
irrotational >ow, velocity potentials and stream functions.
On the other hand, you had hydraulics developed by civil
engineers, who amassed mountains of tables, all obtained
from experiments on pressure drop and head loss of >uids
>owing in all sorts of open and closed channelsmade of
concrete, fresh wood, slimy old wood, etc.
Re>ecting on the situation at the beginning of the 20th
century, Prandtl (see Tietjens, 1934) said
Hydrodynamics has little signi1cance for the engineer because of the great mathematical knowledge required for
an understanding of it and the negligible possibility of applying its results. Therefore engineers put their trust in the
mass of empirical data collectively known as the science
of hydraulics.
Prandtl (1904) was the genius who patched together these
completely diDerent disciplines with his simple boundary
layer model. The result is modern /uid mechanics. The left
side of Fig. 9 shows the pioneers in this development, the
types of problems solved, the terms used, and so on. I call
this Type 1 problem of >uid >ow.
Then in 1952 Danckwerts introduced a completely diDerent type of study of >uids which I call tracer technology.

O. Levenspiel / Chemical Engineering Science 57 (2002) 4691 4696

4695

9.1. The LHHW models

Since the beginning of the century chemists have used
the LangmuirHinshelwood models to represent the rates
of chemical reactions. The chemical engineer, following
the lead of Hougen and Watsons 1947 book (Hougen &
Watson, 1947a), built on this approach developing what
is called the LangmuirHinshelwoodHougenWatson
(LHHW) models.
These models were cumbersome to use, but were based
on the mechanism of action of molecules, such as:

Fig. 8. The dispersed plug >ow model is used to represent many reactors
and vessel types.

adsorption of reactants onto active sites on the solid surface,

reactions of attached molecules
with adjacent attached molecules (dual site)
with free molecules (single site)
by decomposition alone
desorption of product molecules from the surface.
For a typical equation, say for a 1rst-order reversible reaction, A R, dual-site mechanism, surface reaction controlling, no product in the feed, the performance equation for
plug >ow of gas through the packed bed catalytic reactor is:
 
(1=% + KA nA0 + KI nI 0 )2
C W
=
% F
(1 + 1=K)
2(1=% + KA nA0 + KI nI 0 )(KR KA )nA0
(1 + 1=K)2



nA0
(KR KA )n2A0
ln
+
(1 + 1=K)3
nA0 (1 + 1=K)x

+
Fig. 9. The two distinctly diDerent types of >uid mechanics problems.

The right side of Fig. 9 shows the names of the pioneers and
the terms used in this type of study. I call this Type 2 problem of >uid >ow. Both these types of problems started with a
simple approximation of the complex real world. Note how
completely diDerent are these two branches of >uid >ow. It
is interesting to note that there are dozens upon dozens of
books on Type 1 problem, as well as courses in every academic institution in the world on this type of problem. However, there is not a single book in print today devoted to the
Type 2 problem even though it is important in the study of
chemical reactors, in physiology, in oceanography, in dealing with the >ow of ground waters, rivers, and oceans. How
curious.
Also note that the Type 2 branch of the subject was developed as a $10 approximation of what is a very complicated
mathematical model, one which involves stochastic computational >uid dynamics.

9. Models for chemical reactions

There are two broad classes of models for chemical reactions, the LHHW and the CRE. Let us discuss these.

(KR KA )2 x2
2(1 + 1=K)

2(1=% + KA nA0 + KI nI 0 )(KR KA )x

(1 + 1=K)

(KR KA )2 nA0 x
:
(1 + 1=K)2

(1)

This is a rather awkward complicated equation.

One of the two major problems with this approach is that
many possible mechanisms can and do 1t any set of data. For
example, Hougen and Watson (1947b) show that 18 diDerent
mechanisms can be proposed by the simple reaction:
C8 H16 + H2 = C8 H18 :
codimer

Sad to say, not even with the most detailed experimental

program has anyone been able to choose among these models.
Since these models extrapolate diDerently, this means that
one cannot predict what to expect in new conditions so your
choice of this or that mechanism is arbitrary and may not
represent reality.

4696

O. Levenspiel / Chemical Engineering Science 57 (2002) 4691 4696

The second problem with the LHHW models is that they

completely ignore possible heat and mass transfer resistances such as:

1lm mass transfer from gas to particle surface,

pore diDusion in the particle,
1lm heat transfer,
nonisothermal particles.

Now the chemist deliberately chooses conditions where he

can study chemical kinetics free from physical resistances.
However, the chemical engineer, dealing with reactor design, has to consider these contributions.
9.2. The CRE models
In 1956 a new approach was introduced, called chemical
reaction engineering (1958). This approach used a simple
nonmechanistic based kinetic form (often nth order) combined with heat and mass transfer eDects (Thiele modulus,
PraterWeisz nonisothermal criteria).
As an example, for the reaction leading to Eq. (1) the
CRE approach gives
rA

So it is with the 1lm model and the boundary layer idea.

Certainly, earlier works were reported in this area, for example: Newton (1701), Biot (1809), Peclet (1844), Reynolds
(1874), Stanton (1877), and Nernst (1904).
But Prandtls 1904 article (Prandtl, 1904) changed our
thinking in engineering. He was our Columbus.
Let me give you a picture of what engineering was like
early in the century. Until 1912, ASME had only recorded
eight papers in all areas of heat transfer in the previous 32
years of its history, not one of which was on 1lms, boundary
layers or h (Layton & Leinhard, 1988). The very 1rst
book on engineering heat transfer was the well known 1933
volume by McAdams (1933). Thus, these concepts are all
relatively recent and all part of this centurys creations.
May I end up by suggesting the following modeling strategy: always start by trying the simplest model and then only
add complexity to the extent needed. This is the $10 approach, or as Einstein said,
Keep things as simple as possible, but not simpler.
References

= k CA ;

where , the eDectiveness factor, depends on the nonisothermal Thiele modulus. This relationship is based on values for
the particle size, L;
the eDective diDusion coe-cient of gas in catalyst pores,
D;
the thermal diDusivity of the solid, k;
the heat of reaction, KHr .
The CRE approach was found to be simpler and more general, and so found favor and was widely accepted by the
profession. It is the class of model used today.
10. Creators of models
Who discovered America? The Egyptian reed boat explorers, the Mongol wanderers, Lief Ericsson and his Viking
bands or Christopher Columbus? The earlier discoveries
were isolated events which were not followed up on and
were forgotten by history. But Columbus was diDerent. It
was used, it changed societys thinking and action, so we
credit him with the discovery.

Chemical Reaction Engineering (1958). The 6rst symposium of the

European federation of chemical engineers. Chemical Engineering
Science, 8, 1200.
Danckwerts, P. V. (1953). Chemical Engineering Science, 2, 1.
Denbigh, K. (1951). The thermodynamics of the steady state. Methuens
monographs on chemical subjects. London.
Hougen, O. A., & Watson, K. M. (1947a). Chemical process principles,
Part III (pp. 929) (Eq. (b)). New York: Wiley.
Hougen, O. A., & Watson, K. M. (1947b). Chemical process principles,
Part III (pp. 943958). New York: Wiley.
Layton, E. T., & Leinhard, J. H. (Eds). (1988). History of heat transfer.
New York: ASME.
Lewis, W. K. (1916). Industrial and Engineering Chemistry, 8, 825.
McAdams, W. H. (1933). Heat transfer. New York: McGraw-Hill.
Perry, J. H., & Green, D. W. (1984). Chemical engineers handbook (6th
ed) (pp. 1114). New York: McGraw-Hill.
Prandtl, L. (1904). On >uid motions with very small friction (in German).
Third International Mathematical Congress. Heidelberg (pp. 484
491).
Tietjens, O. G. (1934). Fundamentals of hydro- and aeromechanics
(p. 3). New York: Dover Publications.
Vermeulen, T., et al. (1966). Chemical Engineering Progress, 62, 95.
von Karman, T. (1934). Journal of Aeronautical Science, 1, 120.
von Karman, T. (1954). Aerodynamics, topics in light of their historical
development. Ithaca, NY: Cornell University Press.
Whitman, W. G. (1923). Chemical and Metallurgical Engineering,
24, 147.