Professional Documents
Culture Documents
In spite of the advances made by absolute kinetic rate theories, the ultimate determination of the
kinetic rate form and the evaluation of the appropriate kinetic parameters has to be based on
experimental results. Only when the rate form has been confirmed in the laboratory and the rate
parameters evaluated, preferably at conditions close to those contemplated for the large scale process,
can an engineer use with some confidence the rate and its parameters for design purposes or for
predictions of events in the atmosphere.
The question then arises, if the reaction rates have to be determined by experiments of what help
are the kinetic theories to chemical or environmental engineers? The knowledge of reaction mechanism
leads to postulated rate forms. It is always much easier to check a postulated rate form, find out whether
the experimental data confirm it in its entirety or indicate that a limiting case is sufficient, than to find
what rate form experimental data conform to without the prior knowledge of that form. In other words,
it is clear that if one knows the expected rate form one can plan well the experiments, and minimize the
number of necessary experiments, in order to confirm the postulated rate expression and determine its
parameters. Without "a priori" knowledge of the rate expression more experimentation and more work
is necessary in order to extract all the information. At the same time prediction of kinetic constants and
activation energies from transition state theory helps in setting up the expected upper limits on the
kinetic constants and in assessing the temperature sensitivity of reaction. This also helps in planning
properly the experiments.
In order to understand how rate expressions are evaluated from laboratory experimental
measurements it is instructive to consider first what types of experiments are usually possible and in
what environments and under what conditions are they done. We will restrict ourselves here to
experiments performed in homogeneous systems.
10.1 Reaction environments and conditions for determination of reaction rates in homogeneous
systems
Typically laboratory experiments can be performed in one of the reactor types described below.
Each of the reactor types can be operated (more or less successfully) under isothermal, adiabatic, or non-
isothermal conditions.
Under isothermal experimental conditions the temperature of any point of the reaction mixture in the
reactor should be the same and constant (equal to the desired experimental temperature) at all times
during the run.
Adiabatic experimental conditions are achieved when the reaction mixture does not exchange any heat
with the surroundings (other than sensible heat of the inflow and outflow stream in flow reactors) and no
work is done on or by the reaction mixture (other than PV work of the fluid entering and leaving the
vessel in flow reactors).
a. either when a temperature program with respect to time or position in the reactor is established
b. or when the reactor establishes its own temperature profile in time and space while exchanging
heat with the surroundings.
From the point of view of determination of kinetic rates, isothermal conditions are preferred.
Kinetic data for simple reactions with well defined stoichiometry can also be obtained from adiabatic
runs, but interpretation of non-isothermal runs is usually extremely tedious and is to be avoided.
The typical reactor types in which experiments are performed are:
sampler
sampler
V = const P = const
A typical batch reactor V const is a vessel of constant volume (i.e., a flask, autoclave, etc.)
into which the reactants are charged at the beginning of the run. The reactor is equipped either with a
cooling/heating coil or jacket, or is well insulated, and can be run isothermally (or close to it) or
adiabatically. A mixer provides for vigorous agitation of the reaction system. The progress of reaction
can be monitored by taking samples of the reaction mixture in specified time intervals and analyzing
their composition, i.e. concentrations of certain components are observed as a function of time. In case
of gas phase reactions, which proceed with the change in the number of moles, the change of the overall
pressure in the system can be monitored in time and tied to reaction progress in case of single reactions.
Another type of batch reactor, which is infrequently used in practice to generate rate data but
which illustrates an important concept to be used later, is the constant pressure batch system where the
volume of the reaction mixture may change in time (i.e. in case of gas phase reactions which proceed
with the change in the total number of moles).
sampling
In the case of a semibatch reactor some of the reactants are charged at the beginning of the run
while one or more reactants are added continuously throughout the run. Again, the reactor can be run
isothermally or adiabatically, and sampling of the reaction mixture is performed in time in order to
monitor the progress of reaction. This reactor type may be quite useful when one is trying to determine
the reaction order with respect to say reactant B and constantly keeps adding reactant A in large
excess. It is also a convenient device in complex reaction systems to study the effect of the order of
reactant addition on selectivity and yield etc.
10.1.3 CONTINUOUS FLOW REACTOR:
PFR
CFSTR
Reaction rates and rate expressions can also be determined in continuous flow systems run at
steady state. Two basic types of continuous flow reactors are: the ideal plug flow reactor (PFR) and the
ideal continuous flow stirred tank reaction (CFSTR).
aA bB dD; (1)
or
j 1
j Aj 0 (1a)
N Ao N A N Bo N B N D N Do
; (2)
a b d
or
N j N jo
X (2a)
j
it is sufficient to monitor only one variable of the system, i.e. one concentration, or any property of the
system such as its pressure, viscosity, refactive index, etc. that is directly and uniquely related to its
composition. The rate of reaction r T , C A , CB , C D r T , C j is a function of temperature and
composition.
rA rB r
r D (3)
a b d
Suppose we run the reaction in a batch reactor from time 0 to time t and monitor concentration CA as a
function of time.
Amount of A
N Ao moles (5)
charged in the reactor
Amount of A
N A t moles (6)
left unreacted at time t
evaluated at t
By definition
moles of A reacted
ar
unit volume unit time
Then
moles of A reacted
a r T , C dV
V
j
unit time
(moles reacted per unit volume and unit time summed up over the whole volume of the mixture)
Amount of A t
a
reacted from time 0 to time t
r T , C dV dt
o V
j (7)
t
N Ao N A t a r T , C dV dt
o V
j 0 (8)
Cv VT o Cv VT t a U r
r T , C j dV dt
o V
t
UAh T Tc dt 0 (9a)
o
- density of the reaction mixture kg m3
Cv - specific heat of the reaction mixture J kg C
V - reaction mixture volume m3
U r - heat of reaction J mol
U - overall heat transfer coefficient J m 2 s C
Ah - area for heat transfer m 2
Clearly the equations simplify if the temperature is kept constant T Tc in an isothermal system and
if the volume of the reaction mixture is constant.
r Tc , C j dt 0
t
N Ao N A t aV (10)
o
dN A
aV r Tc , C j 0 (11)
dt
1 dN A
or ar C j , Tc (12)
V dt
rA C j , Tc
NA
Now if the volume of the reaction mixture is constant CA
V
and the concentrations of all other components can be related to C A
N B N Bo bX (13)
N Ao N A
X (13a)
a
X C CA
If V cons tan t Ao (14)
V a
b
CB CBo C Ao CA (15)
a
d
CD CDo C Ao CA (16)
a
dC A
rA C A , Tc (17)
dt
dC A
Since we monitor C A as a function of time, the derivative t gives us the value of the rate at that
dt
particular time, i.e. at the concentration C A present at time t .
CA CAo
slope 1 = rate at CA1
In order to get the reaction order with respect to reactant A we can use the method of excess, i.e. take
initial N Bo N Ao 1 so that CB stays practically constant during reaction. In order to be able to
neglect the effect of product D on the rate we can use the method of initial rates, i.e. measure the rates
only while very little product is present.
Let us briefly consider the material balance for a continuous flow reactor for the same reaction at
isothermal conditions and at steady state
Rate of
Rate of molar Rate of molar disappearance
0 (18)
input of A output of A by reaction in
the reactor
a r C A , Tc dV 0
FAo FA
mol mol V (18a)
mol
hr hr lit
lit hr
In terms of conversion:
FA FAo 1 xA (19)
FAo x A a r C A , Tc dV 0 (20)
V
The integral is necessary in the PFR since the concentration of A changes as one moves down the
reactor length and, therefore, so does the rate.
However, in a CFSTR, by assumption, the mixture is perfectly mixed and at uniform composition.
Therefore, every element of the reaction mixture reacts at the same rate and we have ( q - volumetric
flow rate of reaction mixture at the exit, qo - volumetric flow rate of entrance):
F Ao x A CAo qo CA q
rA CA , Tc
V V
concentration in
the reactor and
at the same time (21)
in the exit
Clearly CFSTR is a convenient contacting device for direct measurement of reaction rates (no
derivatives needed). A rate at a different C A can be found by increasing or decreasing qo .
From the above we have seen that PFR operates as an integral reactor, i.e. overall conversion is
given by the integral of the rate over the whole reactor volume.
Sometimes the PFR is used as a differential reactor, i.e. overall conversion is kept very low so that
the change in concentration is very small from entrance to exit. Then the mean value theorem can be
used.
FAo x A ar C Am , Tc V 0 (22)
C Am mean value of concentration C Aexit C Am C Ao . Now the rate is obtained directly but the
small changes in concentration are difficult to measure accurately.
The CFSTR in contrast can give very large changes in C A between inlet and outlet and still the rate is
obtained directly.
For a reaction aA bB dD gG find the reaction rate, i.e. find the rate's dependence on
concentration and temperature.
The experiments are selected to be performed at isothermal conditions so that the rate's
dependence on concentrations is established first. Once the rate form is known, isothermal runs can be
repeated at different temperature levels so that the dependence of the specific rate constant(s) on
temperature can be established.
Let us assume further that these hypothetical experiments are performed in an enclosed batch
vessel so that the volume of the reaction mixture is constant V const .
N Ao N A N Bo N B N D N Do N N Go
G X (2)
a b d g
or for a general single reaction j Aj 0 (1a)
N j N jo
X (2a)
j
N Ao NA X
can be extended by dividing it by V and since C Ao , C A etc., while is the
V V V
extent per unit volume we get
C Ao C A C CB C CDo C CGo
Bo D G (23)
a b d g
C j C jo
or (23a)
j
Let us that suppose we have charged the reactor with N Ao , N Bo moles of A and B and no D and G . A
balance on A gives
dN A
ar C j V (24)
dt
dC A
rA ar C j (25)
dt
The rate is in general a function of concentration of all species. Let us assume that we expect to find the
rate of the form
r k f C A CB kb CD CG (26)
E f RT
where k f k fo e and kb kbo e Eb RT
. We have assumed the rate constants to obey the
Arrhenius form but since we operate at constant temperature (isothermal runs) both k f , kb are constants
to be determined from the run as well as the powers , , , .
Using the stoichiometric relationship we could eliminate all other concentrations in terms of C A .
b In General:
CB CBo C Ao C A
a
j
CD CDo
d
C Ao C A C j C jo Cl Clo (27)
a l
g
CG CGo C Ao C A
a
Thus we would have
b dg dC A
a k f C A CBo C Ao C A kb C Ao C A (28)
a a2
dt
this follows nice since
CDo CGo 0 were assumed
At t 0 C A C Ao (28a)
Clearly, we could monitor the concentration of A, C A , in time continuously, or as is more often the case
dC A
discretely. We can try to evaluate the slope of C A vs. t line, , at various values of C A .
dt
CA CAo
dCA
slope at CAi
dt
CAi
dC A
Plugging the C A and corresponding values in the above equation we would get a large set of
dt
nonlinear equations from which we would have to determine a great number of parameters, namely six
(6) all together: , , , , k f , kb .
This method of attack due to large errors hidden in experimental data does not seem too fruitful. Instead
C CA
we can decide to run the experiment only up to low conversions xA Ao . In this situation we
C Ao
can neglect the rate of the reverse reaction - we are using the method of initial rates.
b dC A
a k f CA
CBo a C Ao CA (29)
dt
Now we would have to determine only three parameters from the experimental run:
, , and k f .
In addition, we can decide, if experimentally feasible, to make the job even easier and use reactant
C b
B in large excess over the stoichiometrically required amount i.e. Bo .
C Ao a
The mass balance can be rewritten in the following form:
CBo b dC A
a k f C A C Ao xA (30)
C Ao a dt
CBo b b b C Ao C A
but since xA 1 (low conversion) and we can safely ignore xA in
C Ao a a a C Ao
CBo
comparison to
C Ao
a k f CBo
C
A
dC A
dt
a k 1f C A
(31)
and we have used the method of excess in addition to the method of initial rates. Now we have only two
parameters to determine: and k 1f . This seems a reasonable task. We can afterwards proceed to repeat
the runs at different temperatures, still measuring rates at small conversions and using B in excess, in
order to determine E f from the Arrhenius plot. After that, experiments can be performed in excess of
A to determine and k fo , and then runs starting with only C and D can be performed to determine
the reverse rate's parameters , , kbo , Eb .
The previous discussion indicates that if we want to determine a rate expression for a single
reaction, and we monitor the concentration of one component (one species) as a function of time (say of
limiting reactant A ), we can almost always, by proper selection of experimental conditions, reduce the
problem to one of the type:
dC A
k CA ; t 0 C A C Ao (32)
dt
where from a set of C A vs. t , i.e. C Ai vs. ti , data parameters and k have to be determined.
For the analysis of this type we can employ two basic methods:
- differential method or
- integral method.
This can be done either directly from the table of differences of the data i.e.
dC A C Ai C C Ai
ti Ai 1 ; ti t ti 1 (33)
dt ti ti 1 t i
or by numerical differentiation of the data, or by plotting a smooth curve through the C A vs. t data and
evaluating its slope graphically, or by matching the C A vs. t data by some function and evaluating its
derivative. Using the ratio of the differences from the table of data leads to very poor estimates of the
derivative when ti are uneven and large and C Ai are large. At the same time the question arises to
what value of t t t t , i.e. to what value of C t C C
i i 1 A i A Ai 1
does the derivative
C Ai
approximation correspond. The quality (i.e., poor quality) of error prone kinetic data usually does
t
not warrant the use of more sophisticated numerical differentiation methods.
C Ai
The best practical procedure is to plot verses t as a stepwise curve shown below. Since
ti
C Ai
was evaluated by using C Ai 1 , C Ai and ti 1 , ti it is valid for the interval ti 1 ti
t i
CAi
ti
tN+1 t
C Ai
The area under every step is the value of the ordinate times the value of the corresponding time
t i
C Ai
interval ti 1 ti , i.e., ti C Ai C Ai 1 which measures the variation in concentration
t i
during one time interval. The area under the whole stepwise curve is:
N
C Ai
ti
ti C AO C AN 1 (34)
i 0
and measures the overall concentration change of A during the experiment. Yet we know that the
dC A
overall concentration change we can obtain by integrating the derivative with respect to time:
dt
tn + 1
dC Ai
C AO C AN 1
to
dt
dt (35)
This tells us to plot a smooth curve on our step-wise plot in such a way that the area under the smooth
curve is equal to the area under the stepwise curves. Since both areas are C AO C AN 1 the smooth
dC A
curve gives us our best estimate of . This was done on the figure in the previous page. Now
dt
dC A
for every ti we can read the value of the derivative ti by going to the smooth curve at that
dt
value of ti and reading of the corresponding value of the ordinate. Since for every ti we have a
dC A
measured C Ai value, we can now pair the values ti and C Ai .
dt
dC A
means
dt
evaluated at ti
dC A
3. Now plot ti versus C Ai on log-log paper. By doing this
dt
dC A
log log
k log CA (36)
dt
a x
y
we should be able to pass a straight line through the data points provided that the rate indeed was of n-th
order form and not of a more complex form.
The slope from the log-log plot gives our . From any point then we can determine log k and k . It is
advisable to determine k from a number of points and to take an average value.
The strength of the differential method is that all that we have to assume is that the rate is of a
particular form (n-th order form). Both the reaction rate order and the kinetic constant k are then
extracted by the method. The main weakness of the method is poor accuracy.
Procedure:
1. Tabulate the C A vs. ti data.
1
2. Assume a particular order , i.e. 1, or or 2, etc.
2
3. Integrate the batch mass balance.
CA t
dC A
k dt (37)
C Ao C A o
F C A , C Ao kt
C Ao
for 1 , F C A , C Ao ln (37a)
CA
1 1
for 2 , F C A , C Ao etc. (37b)
CA C Ao
4. Form the pairs of values F C Ai , C Ao vs ti , i.e. evaluate the known function F C A , C Ao at all
measured values C Ai and pair them with corresponding ti by extending the previous table of data.
F C A , C Ao k t (38)
x
y
If the assumed order was correct one will be able to pass a straight line through the above plotted
points, the slope of the lines gives k and the assumed order has been confirmed.
If the assumed order was incorrect, a straight line plot cannot be obtained, and one has to go back to
step 2 and assume another and repeat all other steps.
The strength of the integral method is that the integration procedure smoothes the error of
experimental data and allows more accurate evaluation of k 's. The main weakness is that the method is
very tedious to use unless one has a good indication of what may be. Clearly the knowledge of the
mechanism would help.
In general, if one does not have any idea what may be, it is advisable to get an estimate of
and k using the differential method, and then retest the found and upgrade the accuracy of the k
value by the integral method.
In the above discussion we have assumed that the concentration of a reactant was monitored in
time. Naturally, all the above procedures hold if instead a product concentration is monitored, or a
physical quantity proportional to the composition of the system, or total pressure of the system (in case
of gas phase reactions which proceed with the change in the total number of moles). Let us consider this
last case:
S
aA bB dD gG or
j 1
j Aj 0 (1a)
S
n d g a b n
j 1
j (39)
rA k C A CB (40)
PAo PBo
C Ao CBo (41)
RT RT
The total initial pressure, if there were no inerts and products present, is given by:
PAo PBo P
C Ao CBo CTo To (42)
RT RT
Concentrations are related to extent per unit volume , as shown below:
PA NA N Ao aX
CA C Ao a
RT V v
P
CB B CBo b
RT
(43)
P
CD D d
RT
P
CG G g
RT
PA PB PD PG PT
CT C A C B C D CG CTo n
RT RT
P P P PTo
CT T To n T
RT RT n RT
or in general
S
P PTo
CT CTo j T (44)
j 1 j RT
Now let us get an equation in terms of total pressure:
dC A
k C A CB (45)
dt
1 dPA 1
k PA PB
RT
RT dt
dPA 1 +
k RT
PA PB (46)
dt
kp
a
PA PAo a PAo PT PTo (47)
n
PAo y Ao PTo
(48)
Mole fraction of A initially
dPA a dPT
(49)
dt n dt
PT PTo
PT PTo
a dPT a b
k p y Ao PTo yBo PTo n (50)
n dt n
k p n
PT PTo
PT PTo
dPT a b
y Ao PTo y Bo PTo n (51)
dt a n
This equation relates now the change of total pressure PT with time and from the PT ti vs. ti data ,
, k p can be determined by the previously outlined methods. Method of excess can be again used in
order to further simplify the apparent rate form.
Clearly, both the integral and differential method have one thing in common - they attempt to find
a relationship between properly arranged experimental data which is linear. The key is to find and plot a
straight line relationship.
dC A
log log k n log C (52)
dt A
x
y
F C A1 F C Ao kt f n (53)
1 1
n 1
n 1
k n 1 t (53a)
C C Ao x
A
y
For rate forms other than n-th order straight line relationships are also sought. For example for
Michaelis-Menten rate form:
dC A k CA
rA (54)
dt K CA
Differential Analysis:
1 1 K 1 CA K 1
or CA (55)
dC A k k CA dC A k k
x
dt x dt
y y
Integral Analysis:
C Ao
ln C k t 1
A
(56)
C Ao C A K C Ao C A K
x
EXAMPLE 1: Determine the reaction order and the rate constant for a single reaction of the type
A products based on the following experimental information obtained at isothermal conditions at
V const.
t min 0 5 10 15 20 25 30 35 40
mol
CA 1 0.58 0.41 0.32 0.25 0.22 0.19 0.16 0.14 0
lit
The last data point simply indicates that after a very long time (several hours as compared to minutes)
practically no A is found. Thus, at the experimental conditions used the reaction is practically
irreversible.
A. Differential Analysis:
C Ai
We can form ti and C Ai C Ai C Ai + 1 from the table, and
ti
t 5 05 5 10 10 15 15 20 20 25 25 30 30 35 35 40
C Ai
0.084 0.034 0.018 0.014 0.006 0.006 0.006 0.004
ti
C Ai
We can plot now vs. ti as a stepwise curve shown below. Now we have to pass a smooth curve
t
so that
0.08
CA
-
t
( litmol
min )
0.04
0.02
0
10 20 30 40 t (min)
the area under the stepwise curve and the smooth curve are approximately equal. From the smooth
dC
curve we now read of the corresponding values of A at desired measured concentrations C Ai .
dt
t 5 10 15 20 25 30 35
dC A
0.055 0.024 0.015 0.0105 0.0065 0.005 0.004
dt
dC dC
log A log k log C A i.e. we plot A vs. CA on a log-log plot.
dt dt
From the slope of the straight line that we managed to pass through the data points we find that
2.05 . From any point of the line now we could determine k . For example:
dC A
CA 0.58 0.41 0.32 0.25 0.22 0.19 0.16
k dt
2
CA k 0.163 0.143 0.146 0.168 0.134 0.139 0.156
1 2 3 4 5 6 7 8 9 1
0.1
9
8
7
3
dC
A
dt
2
0.1
log
0. 0014
slope 2. 05
0. 8
log
0.01 0.1
9
8
0.001
1 2 3 4 5 6 7 8 9 1.0
lit
The mean value of k turns out to be k 0.150 . Thus at the temperature of the
mol min
experiment we have determined
mol 2
rA 0.150 C A
lit min
Note: Although the values of the rate constants calculated from various data points vary
considerably, the variation is random and shows no trend with concentration level indicating
that the selected order is correct.
Integral Method:
Suppose that we have attempted to solve the same problem by the integral method. Since0 the
stoichiometry is A products , we will try first whether a first order rate form can fit the data (hoping
for an elementary reaction).
Assume 1 :
dC A
k CA
dt
CA t
dC A
C Ao
CA
k dt
o
C Ao
ln kt
CA
C Ao
2.3026
1
log kt
CA
2.3026
1
log C A log C Ao kt
We should then plot the experimental data on a semi log plot ( CA on the log scale and t on the linear
scale). If the assumed order of one is correct we should be able to obtain a straight line through the data
points. It is clear from the enclosed figure that a straight line cannot be obtained since the data show a
definite curve here - convex towards the bottom. If we connect the first and the last data point by a
straight line all the other data are below the line indicating that the concentration drop is faster than
predicted by first order behavior.
1-
9-
8-
CA 7-
6-
mol
lit 5 -
4-
3-
2-
1-
10 20 30 40 t (min)
dC A 2
k CA
dt
1 1
kt
CA C Ao
1
Plot vs. t on a linear plot.
CA
1
This time a straight line is obtained which properly intersects the ordinate at 1 1 . From the
C Ao
lit
slope of the line we get k 0.150 .
mol min
1 1
CA C Ao CA 0.58 0.41 0.32 0.25 0.22 0.19 0.16 0.14
k
t k 0.145 0.144 0.142 0.150 0.142 0.142 0.150 0.154
0
0 10 20 30 40 t (min)
lit
The mean value of k turns to be k 0.146 the difference between this mean value and
mol min
the one obtained by "eye fitting" the line through data points is 2.7% and is negligible as far as
engineering applications are concerned. Notice that the difference between the largest k 0.154 and
smallest k 0.142 is only 0.012 or less than 8.5% based on the smallest k -value. For the same data
the k 's evaluated by the differential method varied between a low of 0.134 and a high of 0.168, the
difference being 0.034 or 25% based on the smallest k .
Thus, for the same quality data the integral method tends to smooth out the errors and give more
consistent estimates for the rate constant.
mol 2
rA 0.150 C A
lit min
EXAMPLE 2:
Gas phase decomposition of di-t-butyl peroxide is monitored in a batch reactor of constant volume at
isothermal conditions of 170˚C. The run is started with pure di-t-butyl peroxide and the change of total
pressure of the system was recorded in time. From the data below find the rate expression and the rate
constant.
CH 3 3 CO OC CH 3 3 C2 H 6 2 CH 3 CO CH 3
A P 2Q
V = const, T = const
mol
rA k CA
lit min
We have seen before that we can also express the rate in terms of change of the partial pressure:
mm Hg
rA kp PA
min
where
k p k RT
1
dPA
rA k p PA
dt
PA
C A C Ao
RT
s P here
CT CTo j T
j 1 RT j 3 1 2
P PTo
PA PAo T
2
1 dPT 3 P PT
k p To
2 dt 2
dPT
dt
2
1
k p 3 PTo PT
DIFFERENTIAL ANALYSIS:
PT
Evaluate , plot vs. t a stepwise curve, from a smooth curve that has the same area underneath as
t
dPT dPT
the stepwise curve. Evaluate and calculate the corresponding 3 PTo PT . Plot vs.
dt dt
3 PTo PT on a log-log scale. The slope gives ; from the data find k p and k and their mean value.
The augmented table is shown below as well as the two figures (next page)
PT dPT
t PT 3 PTo PT kp
t dt
0 7.5 1.48 15.0 0.0987 k p 0.0812 min 1
2.5 10.5 1.20 0.95 12.0 0.0792
5.0 12.5 0.80 0.75 10.0 0.0750
10.0 15.8 0.66 0.52 6.7 0.0776
15.0 17.9 0.42 0.35 4.6 0.0761
20.0 19.4 0.30 0.25 3.1 0.0806
dPT
It seems that more than one straight line can be passed through the points on the log vs.
dt
( 3 PTo PT ) plot. The maximum slope seems to be 1.11 and the minimum 0.95. This indicates that
probably 1.0 .
dPT
kp dt
3PTo PT
These values are given in the last column of the above table.
PT
t
1.2 -
0.8 -
0.4 -
t (min)
0 5 10 15 20
18
10 - log
slope 2 1.11
20
log
28
dPT
dt log 15 2.0
slope 0.95
log 20 2.4
0.1 -
1.0 10.0 20.0
3PTo PT
mm Hg mol
rA 0.0812 PA 0.08 PA or rA 0.0812 C A 0.0022 PA
min lit min
INTEGRAL METHOD:
Suppose we assumed zero-th order
dPT
k" PT PTo k " t
dt
20
P
T
10
0 5 10 15 20
t (min)
Assume 2nd order:
dPT
k " 3 PTo PT
2
dt
1 1
k " t
3 PTo PT 2 PTo x
y
0.3
1
3PTo PT
0.2
0.1
0 5 10 15 20
t (min)
Again data show a definite curvature and the reaction is not 2nd order.
Assume 1st order
dPT
k " 3 PTo PT
dt
2 PTo log 15 3.1
ln k" t slope 0.0342
3 PTo PT 20
log 3 PTo PT log 2 PTo 2.3026
1
k" t
3PTo PT
(mm Hg)
log 15 3.1
slope 0.0342
20
10.0 -
9-
8-
7-
6-
5-
4-
3-
2.0 -
0 20
5 10 15
t (min)
Now we do get a straight line.
mol
rA 0.08 C A 0.0022 PA
lit min
mm Hg
rA 0.08 PA
min
Notice again that the variation in the rate constant based on integral method is much less than if
differential method is used.
Example 3: Consider the reaction between sulfuric acid and diethylsulfate in aqueous solution
At isothermal conditions T 22.9 C starting with equimolar mixture of the reactants and with no
product the data presented below were obtained. Initial reactant concentration was 5.5 mol lit for
each of them. Find the rate expression.
t min 180 194 212 267 318 368 379 410 several days
mol
Cp 4.11 4.31 4.45 4.86 5.15 5.32 5.35 5.42 5.80
lit
Since for each mole of A reacted one gets two moles of P if the reaction went to completion one would
find 11 mol lit of P . Since only 5.80 mol lit of P are found this indicates that the reaction is
reversible.
C Peq
CPeq 5.80 C Aeq C Beq C Ao 2.6
2
C Ao C Aeq CPeq
1 2
dC p 2
Rp k f C A C B kb C p
dt
2
dC p Cp 2
k f C Ao kb C p ; t 0 Cp 0
dt 2
Let us use integral analysis. From the above rate expression at equilibrium
Cp2 kf 4 x 2 Ae
K 4.98
C A CB
1 x
2
eq kb
Ae
Cp t
dC p
o Cp
2
k
kf dt
o
2
C Ao 2 b Cp
kf
The integral on the left hand side can be promptly evaluated by using a set of integration tables. For an
exercise we will integrate it here:
kb 1
kf K
1 1 2 2
C p C Ao C p C Ao 0
4 K
2 C Ao
C p1, 2
4
1
K
1 A B
2 C Ao 2 C Ao
Cp Cp
1 1 2C Ao 2 C Ao 4 4
Cp Cp 1 1
4 K 4 4 K K
1 1
K K
K 2 Cp 2 K C Ao 2 k f C Ao
ln t
K 2 C p 2 K C Ao
K
2
4 x Ae
K
1 x Ae
2
x 2 xAe 1 xA 1 x Ae
ln Ae k f C Ao t
x Ae xA x Ae
0.232 C p 24.548
ln 4.929 k f t
4.232 C p 24.548
0.232 C p 24.548
i.e. y vs. t on a semi-log plot.
4.232 C p 24.548
t 0 41 48 55 75 96 127 146 162
y 1 1.240 1.295 1.369 1.595 1.853 2.256 2.742 2.809
1
y 1 1.255 1.311 1.390 1.627 1.897 2.322 2.831 2.901
1
dC p 1 1 kf CPeq 2 x Ae
k f C A kb C p K1
dt k b
1
C Aeq 1 x Ae
C Ao 1 1
ln kf1 1 t
C 1 1 C 2 K
Ao 1 P
2 K
1 1
ln 0.948 k f t
1 0.172 CP
or in alternative form:
1
x Ae kf
ln t
x Ae xA 2 x Ae
Now if we expect this rate form to hold we should also get a straight line on a semi-log plot of
1
y1 vs. t . The values of y1 are also calculated in the table on the previous page.
1 0.172 C p
Both forms are plotted on semi-log paper on the next page and both yield a reasonable straight
line!? Apparently from the experimental data given we are unable to distinguish between a reversible
2nd order reaction in both directions and a reversible first order reaction in both directions. If we
consider again the two integrated forms written in terms of conversion we can readily see that when
x Ae 0.5 the two forms become identical and indistinguishable from each other. Since under the
conditions of the experiment equilibrium conversion was xAe 0.527 , which is close to 0.5 , due to
experimental scatter we cannot distinguish between the two forms.
If we performed the experiments at different T so that xAe is far from 0.5 ; or if we used
nonstoichiometric ratio of reactants, we would find that the rate indeed is 2nd order in each direction. It
is important to determine the proper order since when designing a larger reactor we may be operating at
conditions when xAe 0.5 and the predictions of the reactor size for a desired production rate will
differ vastly based on the rate form.
20.0 -
10.0 -
9-
8-
7-
6-
5-
4-
3-
2-
slope log 20.0
450
0.00289
1-
100 200 300 400 500 t (min)
y1
10.0 -
9-
8-
7-
6-
5-
4-
3-
2-
slope log 15.0
400
0.00284
1-
100 200 300 400 500 t (min)
From the slope for the 2nd order rate form we get the value of the rate constant k f .
4.929
k f slope 0.00289
2.3026
The rate in mol lit min is given by the above expression at T 22.9 C . The temperature
dependence of the constants would have to be found by performing experiments at different
temperatures.
Caution: A note of caution is here in order. The integrated expression that we used for a reversible
2nd order reaction of the type A B 2 P was:
x 2 xAe 1 xA 1 x Ae
ln Ae kf C Ao t
x Ae xA xAe
Note 1: This expression is only valid when the experiment is performed with stoichiometric ratio
of A and B . It is not valid when C Ao CBo .
Note 2: The expression for the same reaction type is reported by Levenspiel (page 63, equation
56) and it looks almost exactly the same as the one above except that it has an extra factor
of 2 on the right hand side.
Question: Is there a mistake? Note that k f and k1 in the book (which are rate constants for the
reaction forward) will differ by a factor of two!!?
Potential
Trouble: The choice of the subscript f or 1 etc. does not indicate on which component the rate
constant is based. Thus, if one has only an integrated form to work with, one has no way
of knowing whether k1 is based on reactant or product, etc. Clarify that whenever
possible.
Note 3: Since the above mentioned ambiguity about rate constants always exists try to:
a. use integrated forms only when you know what the k 's are based on;
b. develop your own integrated forms by the help of integral tables.
This last choice after all is not that difficult. In the problem that we just solved we had to integrate
Cp Cp
dC p dC p
C
2
1
1 1 2 2
kf t
C p C Ao C p C Ao
0 2 0
C Ao p Cp
2 K 4 K
A quick look in the CRC Mathematical Tables shows that we have a problem of the type
dx
y
where x C p
y a bx cx 2
2
a C Ao
b C Ao
1 1
c
4 K
dx 1 2cx b q
The answer is: y
q
ln
2cx b q
where q 4ac b 2
2
4 C Ao
Thus in our case q .
K
Caution: Do not forget to evaluate the above expression at the upper and lower limit of integration
(expression at upper limit - expression of lower limit) since you started with a definite integral and CRC
Tables gives you the answer for an indefinite one. You should get the expression used in the problem.
where f f xi , x2 , x3 ,....
ln k ln -rA ln C A ln CB (59)
ln k ln k ln k
1 1 1 (60)
ln rA ln CA ln CB
2
rA
2 2 2
k C A C B
(61)
k
r
A CA CB
If we measured each concentration with precision better than 2% and the precision of the rate
measurements is 10% we get
2
k
0.12 0.022 0.022 0.0108 (62)
k
k
0.1039 (63)
k
The error in k is 10.4%
From the Arrhenius relationship we see
dk E dT
(64)
k RT T
The larger the error in T and the larger the activation energy, E , the larger the error in k .
We usually estimate E from
E ln k1 k0
(65)
R 1 1
T0 T1
E 1 1
ln ln ln k1 k0 ln (66)
R T0 T1
E
2
1
2
k 2 k 2 T 2 T 2 T 2
0 1 1 0 1 (67)
E ln k1 k0 k0 k1 T0 T1 T0 T1 T1
The larger the interval T1 T0 the smaller the error in E provided that classical Arrhenius form with E
= const holds over such a temperature range.
Useful guidelines can be found in Benson, Foundations of Chemical Kinetics.
To obtain the rate constant with an error of it is necessary:
1. To measure the concentration with an accuracy
Change in concentration
%
Largest concentration 1.4
2. To measure time (or equivalent) with accuracy of
Time interval
%
Largest time 1.4
3. To measure temperature with an accuracy of
T
% T in K
35 3500
2 1 1 1
t1 C Ao (69)
2 k 1
2 1
1
log t 1 log 1 log C Ao (70)
2 k 1
x
y
Starting with different C Ao 's and measuring the time that it takes for C Ao to drop to C Ao /2 and by
plotting t 1 vs. C Ao on a log-log plot one can determine overall order and rate constant k .
2
When investigating multiple reactions in batch systems of constant volume we have to monitor as many
variables as there are independent reactions in the system. For example, if we are studying the rates in a
system of the type
A B D desired product
2A U unwanted product
we could monitor C A and CD in time. If the experiments are performed in a constant volume batch
system and if we assume n-th order rate forms we have to evaluate the parameters
k1 f , k1b , k 2 f , k 2b , , 2 , , , which appear in the following description of the systems component
mass balance:
dC A
k1 f C A1 CB k1b CD 2k 2 f C A 2 k2b CU (71)
dt
dCD
k1 f C A1 CB k1b CD (72)
dt
We could again perform the experiments at low conversions of A and B so as to measure the initial rates
in the region where reverse rates can be neglected. Then:
dC A
k1 f C A1 CB 2k2 f C A 2 (73)
dt
dCD
k1 f C A1 CB (74)
dt
D
By forming the expression for a point yield y we get:
A
D dCD 1
y (75)
A dC A 2k2 f 2 1
1 CA CB
k1 f
CB CBo 1 CU CUo 2
dC 2k2 f
log A 1 log 2 1 log C A
k C
(79)
dCD 1 f Bo
dC A
From C A vs. CD data we can get estimates of as shown before, and by plotting the quantities
dCD
indicated above on a semi-log plot we get a slope of 2 1 .
Working now at moderate levels of CBo so that CBo CD CBo we can plot for the new set of
experiments
dC A
dC 1 2k
log D
2 1
log 2 f
k log CBo CD (80)
CA
1f x
y
From the slope one can get and from all points an average for 2k2 f k1 f .
From the slope 1 is determined and from the data points an average value of k1 f . Now since 2 1 ,
k2 f
and are already known, also k2 f and 2 can be evaluated.
k1 f
The above served to demonstrate that the differential method of analysis, as illustrated for single
reactions, can also be used in more complex reaction schemes. In order to do so one needs to plan the
experiments in a proper way and manipulate the data properly.
Similarly for many practical reaction orders of interest the expression for the yield can be integrated and
integral analysis can be applied. For example if we guess
2 1 1 1 1
dCD 1 1
(83)
dC A 2k 2 f 2k 2 f
CBo CD
1 1
1 C B 1
k1 f k1 f
CD
2k 2 f CA
CBo CD
1
1
k1 f
dCD dC A (84)
o C Ao
2k 2 f CBo
CD ln C Ao C A (85)
k1 f CBo CD
Thus, if the hypothesized orders 2 1 1 are correct, then a plot of
CBo k1 f
ln C Ao C A C
D (86)
CBo CD
2k 2 f
x
y
k1 f
should yield a straight line with a slope .
2k 2 f
Now it is difficult to rewrite the result in a form which would yield a straight line plot in a properly
selected coordinate system.
Clearly when dealing with multiple reactions a point is reached very fast when conventional methods of
integral analysis become tedious and often impossible to use. In such a situation one must perform a
sufficient number of experiments so that some statistical parameter estimation methods can be used.
The initial estimates may frequently be obtained by differential analysis as presented above.
species 1 summed up over all species j . This properly appears with the negative sign since it is equal
S
to disappearance of species 1. We could arbitrarily define k11 k
j2
j1 or in general
S
kii k
j 1
ji .
j i
The above scheme considers the possibility that every species in the system can produce all the others or
be produced from them. For example:
k 21
A1 A2 k11 k21 k31 k41
k12
k 42
k13
k 24 k22 k12 k32 k42
k 41 k2
k14 k 31 k 32
k33 k13 k 23 k43
k 43
A4 A3 k44 k14 k 24 k34
k 34
Naturally the above form is the most general one and in specific situations many constants can be a
priori set to be zero.
The problem in matrix form is:
dC
K C; t 0 C Co (90)
dt
where
C1
C k11 k12 ..... k1s
2 k k22 ..... k2 s
C . K
21
j i
Most often in order to deal with unit vectors every row is divided by CToT , total molar concentration,
and the equations are written in terms of mole fractions:
Cj
yj (91)
CToT
dy
K y; t 0 y yo (92)
dt
The problem is as follows:
From a continuous record of y vs. t (or from a set of N discrete measurements of y at
t1 , l 1, 2....... N where N S ) determine the constants comprising the matrix K .
An elegant and useful theory of monomolecular reactions of this type was developed by Wei and Prater
[Advances in Catalysis 13, 203-392 (1962)] and is also presented in abbreviated form by Boudart, M.
"Kinetics of Chemical Processes", Chapter 10, p. 210, Prentice Hall, N.J., 1968.
Basically the theory states that since K is an s s positive definite matrix it must have S real
eigenvalues m , m 0, 1, 2 .....s 1 . To each eigenvalue corresponds a particular eigenvector x m so
that the following matrix equation is satisfied:
K X X or K xm xm (93)
1 db 1
X X X K X b (98)
dt
and remembering
1
X X I Identity matrix (99)
KX X (100)
we get
db 1
b t0 b b0 X y0 (101)
dt
This is a decoupled matrix equation
db0
0b0 b0 b00 e ot
dt
db1
1b1 b1 b10 e 1t (102)
dt
dbs 1
s 1 bs 1 bs 1 bs 10 e s1t
dt
The above matrix manipulation shows that in principle one can find a set of S fictitious components B
so that the equations for quantities of B , b , are uncoupled and can be solved one by one. Furthermore
at equilibrium:
d y eq
0 K y eq (103)
dt
Remember m x m K x m , m 0, 1, 2,..... s 1 (104)
Since ym , m 1, 2....s are measured as a function of time and bm , m 0, 1, 2...s 1 , are obtainable
from ym 's only if k ji 's are known, which they are not, this means that one has to use a trial and error
method and search for suitable combinations of yio 's giving the desired straight line reaction paths from
which n and bi s can be determined leading to evaluation of k ji 's.
One of the best examples is the classical analysis of the isomerization of butane on alumina which was
studied extensively by Haag & Pines and Lago & Haag and was used by Wei & Prater to illustrate the
power of their method.
A1
1 - Butene
A2 A3
4a
3 5
e
4b
2
A2 A3
We can now perform an experiment starting with pure A2 (path 2). Since this does not yield a straight
line we know that pure A2 is not a component of b vector. However, Wei and Prater method shows
how to use the data from path 2 close to equilibrium in order to extrapolate to the side of the triangle and
select a new starting composition for run 3. Run 3 (path 3) does not give a straight line but with one
more correction leads to starting point for run 4. Path 4a is a straight line, thus, the composition on the
b
side of the triangle determines b10 and 1 can be found from ln 1 1t . However, a run 4b from
b10
pure A3 to verify the straight line path is necessary. Then matrix algebra allows the prediction of the
last straight line, path 5, and calculation of all 's and bi 's and kij 's by matrix manipulation as shown in
Boudart's book. Straight reaction path 5 can naturally be verified by experiments.
Certainly a powerful method to cut down on experimental work but restricted to first order reactions.
Cheng, Fitzgerald and park [Ind. Eng. Chem. Process Design and Development Vol. 19, No. 1, 59,
1977] proposed to following method for determining kij 's in a system of unimolecular reactions.
Let C C1 , C2 , C3 ....C N be a row vector of concentration of all the species and let these
concentrations be determined at equal time increments t . Then
C t C 0 e
K t
C 2 t C t e C 0 e
K t 2 K t
(109)
C m t C m 1 t e 2 K t
or in matrix notation
C t C 0
C 2 t C t e
K t
(110)
C m t
C m 1 t
C 0 C t
e K t
C t C 2 t (111)
C m t
C m 1 t
C 0 1
C t
1
K ln C t C 2 t (112)
t C m t
C m 1 t
All of the above computations (matrix inversion, logarithm of a matrix etc.) are readily programmable or
available on modern computers. Thus the above computations yield the matrix of rate constants
directly.
n-th order reactions:
A more general procedure valid for a set of n-th order reactions of known order (or even for complex
know rate forms) was developed by Himmelbaum, Jones & Bischoff [Ind. Eng. Chem. Fundamentals
Vol. 6, No. 4, 539 (1967)] for evaluation of the rate constants. In a batch constant volume system the
equations for materials balance for all components can be cast in the following form:
dC j M
dt
i 1
ki rji C ; j 1, 2, .... S (113)
C j tn C j to
k
i 1
i r C t dt
to
ji (114)
directly from measured data
by integration (since concentration dependence
of the rates is known) of measured data
M
C j t n C j to kX i 1
i nji (115)
The standard least squares method can now be used to minimize the following expression
N S
y yˆ nj
2
S
n 1 j 1
wnj
nj (116)
n indicates n-th data point at tn
N total number of data points
S total number of species