Ind. Eng. Chem. Process Des. Dev., Vol. 18, No.

3, 1979 357

REVIEW

Review of Measuring Methods and Results in Nonviscous Gas-Liquid

Mass Transfer in Stirred Vessels

Klaas Vant Riet

Gist-Brocades N. V., Research and Development, 2600 MA Delft, The Netherlands

A search has been made for relations between published results about gas-liquid mass transfer in stirred vessels
filled with water. The dynamic gassing-out method, the sulfite method, the separate determinations of A and k,
as well as a number of other experimental methods were judged on their own merits. Most methods appear to
be of limited applicability, and the sulfite method is even more restricted in this respect. Correlation of the various
reports on mass transfer is possible with power per unit volume and gas superficial velocity. This requires a rigid
differentiation between water without and water with ions in solution. Among the equipment variables it is in particular
the position of the sparger which appears to be of influence.

Introduction The purpose of this article is to review the relations

Stirred vessels are frequently employed to achieve a between hLA and the chief parameters from the literature.
large contact area between a gas and a liquid. If, for In this connection, the various experimental methods are
example, oxygen is the reactant to be transferred in these examined for reliability, the results for h& are presented,
agitated gas-liquid dispersions, the oxygen transfer rate and finally the influence of a number of vessel geometry
( O T R ) is determined by the specific contact area A, the parameters is dealt with. Discussions are restricted to
mass transfer coefficient kL ( l / k L provides a measure of aqueous systems, and the aim is to comment on as many
resistance), and the driving force CG/H - C L , according to features of these systems as possible rather than sum-
marize all relevant publications.
Measuring Methods
Dynamic Gassing-Out Method. The underlying
principle is that after a given liquid has been deoxygenated
where CG = the O2 concentration in the gas phase, CL = by passing, for example, nitrogen through it, the dissolved
the O2 concentration in the liquid phase, and H = the oxygen concentration profile is monitored following the
Henry constant. start of the air inflow.
The resistance in the gas phase, hc-l, has been neglected, As a rule, a polarographic electrode (the Clark cell) is
which is nearly always permitted (Calderbank, 1959; used. It follows from the mass balance equation for oxygen
Schaftlein and Russell, 1968; Joshi and Sharma, 1976). between tl and t2 that
Equation 1 is valid for the local CG and CL values.
Normally C G and CL are assumed unvariable throughout
the vessel. This means that both the gas phase and the
liquid phase need to be well mixed. The last assumption
is unnecessary if kLA is constant throughout the vessel.
This is reasonably acceptable for low viscous systems, in
contrast to highly viscous systems (Steel and Maxon, 1966).
Notably the assumption of a well mixed gas phase can be
dangerous in vessels with a tank height Ht vs. diameter T As early as 1951, Wise indicated the underlying principles
ratio H , / T >> 1. For these vessels plug flow for the gas of this method.
phase is a more accurate model. Different models are given This method seems to be an ideal one but one should
by Schaftlein and Russell (1968). not overlook a few restrictions on its applicability. Thus,
At first sight, there is not much consistency in the problems occur where as a result of rapid changes in
literature about mass transfer. One can assign various dissolved oxygen concentration with time, a response lag
reasons for this. Thus, there is a large variety of methods arises and the probe output is not directly related to the
to determine kLA, some of which occasionally yield in- instantaneous value of that concentration. The response
correct values. In addition, both kL and A can be affected lag itself is mainly due to diffusion through the membrane.
in several ways. Among the most important variables are In principle, the probe response time, T~ (the time
power consumption, gas superficial velocity, and liquid needed to record 63% of a stepwise change), would have
phase properties such as ionic strength, surface tension, to be much smaller than the mass transfer response time
and viscosity. Inasmuch as these variables are not always of system: l/hLA. In practice, this is seldom the case, and
selected in a consistent manner or may be latently present, therefore many models have been designed to calculate
different values can be found for kLA under seemingly kLA from the probe response all the same. These range
identical conditions. from one for first order response (Van de Sande, 1974) to
0019-7882/79/1118-0357$01.00/0 0 1979 American Chemical Society
358 Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 3, 1979

complicated models for diffusion (Heineken, 1970, 1971; Corrieu et al. (1976) reported about it. This method
Linek, 1972, and Linek et al., 1973). particularly requires accurate modelling of gas phase
It is pertinent to ask whether one really has to perform mixing for correct interpretation of measured OTR and
all these extensive procedures. Van t Riet (1975) has CL values.
demonstrated that in Van de Sandes model the ultimate Separate Determination of A . The separate deter-
error in KLA < 6% for T~ I l,/kLA, while Lineks model mination of A is another measuring method, and may
(1972) implicates an error <3% for T~ 5 1/5kLA. By comprise light transmission techniques (Calderbank, 1958;
adhering to these limits, it is possible to measure accurately Lee and Meyrick, 1970). Lockett and Safekourdi (1977)
enough a usually acceptable maximum value by means of have inspected the reliability of the method more closely.
commercially available electrodes with r p = 2 or 3 s to kLA To obtain a representative average A value, determinations
= 0.1 sd. One need not employ models like those of Linek in various places in the vessel are necessary. The design
and Heineken until these limits are exceeded. of equipment with a view to determine only the kLA value
A response lag of the electrode also arises where the mass would be a rather complicated matter.
transfer boundary layer is of importance for the Clark cell. A can also be calculated by measuring holdup (t) and
In stirred vessels with the Clark cell in the impellar dis- bubble diameter (db).
charge flow the influence of this layer can nearly always
be neglected. For information on mass transfer boundary (5)
layers and probe response, see Robinson and Wilke (1973),
Dang et al. (1977), Linek and Benes (1977), and Linek and It is possible to determine db photographically by suction
Vacek (1976, 1977). and taking measurements (Kawecki et al., 1967). Direct
When the electrodes used respond very rapidly, mea- suction and measurements in a capillary have been per-
surements may give far too high values. Linek et al. (1973) formed by Reith and Beek (1970), Todtenhaupt (19711,
have attempted to explain this by pointing out that gas and Van de Sande (1974). The extent to which a given
bubbles are also recorded so that the electrodes have too sample is representative is a major, if not insurmountable
great a response. This claim has been worked out in more problem, particularly, because samples which vary in
detail by Votruba and Sobotka (1976) and Votruba et al. density are sucked up, resulting in a selection of bubbles
(1977). according to size by pressure distribution at the capillary
Quite a different problem is presented by the gas phase entrance. Besides, dissimilarity in velocity between
residence time. If this value is not much smaller than bubbles and liquid makes it impossible to suck up isoki-
l/kLA, measuring results can be affected a good deal. The netically. One can also determine d b in a direct manner
reason is that in such a case CGfrom eq 2 can no longer photographically, as done by Valentin and Preen (1962)
be regarded as constant. Dunn and Einsele (1975) have and Koetsier and Thoenes (1972).
reported an extensive analysis with correction graphs. Pros and cons are similar to the light transmission
A further elaborated model has been described by Dang technique. Moreover, this method is rather laborious
et al. (1977). It appears from these models that a t high (Landau et al., 1977).
kLA values corrections are often rather rigorously required Sulfite Method. This method is based on the reaction
and that, as a consequence, measurements will lose much of absorbed O2 with Na2S03. Copper and cobalt ions can
of their accuracy. Problems become more serious at larger be used as catalyst. Mostly, there is a concentration range
vessel diameter especially a t low superficial gas velocity in which the reaction proceeds so rapidly that the dissolved
values. T G is given by oxygen concentration in the liquid is nil. This implies that
the reaction largely takes place in the boundary layer and
tV - EHt that the driving force is larger than if there were only
TG=--- (3)
us.y4?rT2 us physically controlled absorption. The apparent kLA thus
determined should then be converted into the kLA for that
The / u s ratio is reasonably invariable at defined power absorption.
per unit volume and us values (Smith et al., 1977). For On application of the sulfite method this causes diffi-
this case, as eq 3 shows, T G becomes proportional to H , culties. The conversion makes it necessary to determine
which means that there is a maximum H, value up to which both the reaction rate constant kR and the order of the
kLA can be measured using this method. reaction, and, as appears from the literature, problems arise
Smith et al. (1977) have worked out an experimental here. As early as 1941 Fuller pointed out that kinetic
design in which after gassing out the oxygen all N2bubbles values respond quickly to any changes in the catalyst
are allowed to escape, and incorporated this in a model. concentrations. Also, Linek and Mayrhoferova (1970) and
However, the holdup has to be built up, which together Linek and Tvrdik (1971) point out that k R values cannot
with changes in concentration in the gas phase causes both be adopted from the literature because these are rather
A and C G to become functions of time. Therefore, this dependent on the purity of the product. This is related
method does not offer many advantages. to the fact that additions may well distort k R . Onken and
OTR and CL Measurements. Especially in fermen- Schalk (1977) have demonstrated that the addition of 5%
tation systems, OTR is determined by measuring the glycerin reduces hR by a factor of 25. (The great influence
oxygen concentration of the air inflow and air outflow. At which additions may have could account for the extremely
the same time, CL is measured with a Clark cell. It is then large role of viscosity as observed by Paca and Gregr
possible to establish kLA directly with eq 1. (1977).)
The order of the reaction in oxygen may vary from 1at
(4) a partial pressure of 1 atm to 2 at low partial pressures,
as demonstrated by Reith and Beek (1973). Articles by
Notably in systems where these variables are also mea- Reith (1970) and Reith and Beek (1973) show that due to
sured for other reasons, it is, therefore, easy to determine various problems results can be interpreted incorrectly.
kLA. Finns review (1967) mentions some advantages of I t is obvious that on application of the sulfite method,
this method, about which remarkably little has been reaction kinetics are difficult to determine and are much
published elsewhere. Recently, Mukhopadhyay (1976) and dependent on the concentrations and purity of sulfite and
 Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 3, 1979 359

catalyst. They should, therefore, be established accurately As a model fluid it is only representative for strong ionic
for each single set of experiments. If in a publication such solutions.
data are lacking or other articles lay under contribution, Among the methods that involve separate determination
one has to interpret results very cautiously. To a somewhat of A and kL the light transmission technique is the most
lesser extent, this also holds for a mere comparison of, for reliable. The degree to which measurement is repre-
example, vessel parameters in one set of experiments where sentative and determination of kL are factors that influence
results need not be converted into physically controlled accuracy. Determination of A from bubble size, measured
kLA values. In the latter case, one should take extremely with suction into a capillary, can lead to difficulties due
good care that the liquids to be prepared are all of the same to unrepresentative samples. Direct determination using
composition. bubble size distribution measured photographically is
An additional problem is that for the sulfite technique tedious. Determination of kLA from OTR and CL is not
to be successful, high ionic concentrations (1 to 2 M) are well represented in the literature. It is a simple, reliable,
required. This implies that as compared to distilled water and straightforward method when the correct model for
A is much greater and that kL, too, can be affected. gas phase mixing is applied.
Therefore, the results are only applicable to one particular kLAResults
liquid, and cannot be simply extended to any other Power per unit volume and gas superficial velocity are
nonviscous liquid. On evaluation of the hL values re- major correlation coefficients for hLA. Therefore, equa-
ported, the influence of ions will be discussed in more tions of the following type are frequently found in the
detail. literature.
Finally, attention is called to an interesting observation
by Prasher (1975), who claims that while the specific
contact area exerts the same influence throughout the
vessel in case of an extremely rapid reaction, physically
controlled absorption varies according to the locally widely where P = gassed power input, V = liquid volume, U S =
differing degrees of turbulence and local bubble size. In gas superficial velocity, K = constant, and CY and /3 are
this connection, the two contact areas could not be exponents. When P is omitted, an equation like the one
compared readily. following can usually be adopted.
Other Measuring Methods. In addition to the sulfite kLA = KNalDaz(~S) (7)
technique, a number of other reactions have been applied
to assess mass transfer. The most commonly employed where N = stirrer diameter.
involves the measurement of C02,as performed by Mehta It is nearly always possible to convert eq 7 into eq 6. The
and Sharma (1971) for various liquid compositions. values found for CY and ,8 show a fairly great variation: 0.4
Robinson and Wilke (1974) combine C 0 2 absorption with < CY < 1 and 0 < p < 0.7. It is not unusual for the K value
dynamic gassing-out in order to permit a determination to remain unmentioned. For that matter, it correlates a
which they claim to be more accurate than the sulfite great deal with ct because P / V = lo3 - lo4. If a t a given
method (Prasher (1975) has criticized this approach). kLA and P/ V, a relatively small change is introduced into
CY,it causes great change in K.
In principle, as a reaction C 0 2 absorption is better Interpretation of the literature is, furthermore, ham-
controllable than the sulfite method. Yet some of the pered by two factors. (1) Where liquids of different ionic
objections inherent in the sulfite method (reaction kinetics strength, as in the case of sulfite subjected to the gass-
should be known beforehand and there remains an in- ing-out method, are used, any attempt to correlation af-
fluence of ions on A and hL) continue to exist. fords dissimilar results obtained under seemingly the same
In fact, a whole range of techniques can be employed. circumstances. (2) Relatively much store is set by the
Hydrogenation reactions, used by Van Dierendonck (1970), values of CY and 6. To what extent these can be varied
are found very suitable for the design of a hydrogenation without affecting the outcome significantly will be dis-
reaction. Several other reactions have been applied by cussed in more detail in the section on hLA. For that
Mehta and Sharma (1971) and Bossier et al. (1973). Miller reason, much consideration will be given here to the value
(1974) applies continuous C 0 2 absorption as a gassing-out of kLA itself, as it occurs in liquids of different ionic
method. strength.
This review of experimental methods to determine hLA Measurements in Ion-Free Water. First measure-
certainly does not pretend to be complete. Many methods ments in distilled water and tap water will be compared.
not mentioned here have been designed but usually these The salt concentration in tap water is approximately 0.02
are modifications of the techniques described, which means M, which causes no more than a minor increase in hLA (by
that they also have several of the reported advantages and 510%) over distilled water. In the following, these media
disadvantages in common. will be designated as pure water.
Conclusions Regarding Experimental Methods. The kLA in pure water has been determined in several
Dynamic gassing-out is a simple and reliable way of manners. Calderbank (1958,1959) makes use of the light
measuring mass transfer on the condition that probe transmission techniques to find A . Kawecki et al. (1967)
response time and gas phase residence time meet certain and Valentin and Preen (1962) have calculated A from the
requirements. Commercially available electrodes, with measured bubble size and holdup. In that case, kL remains
probe response time of 2-3 s, allow measurement up to kLA to be determined separately. Values reported in the lit-
= 0.1 s-l even without response models. Using these re- erature for hL in pure water vary from 1 to 6 X (see,
sponse models one can measure up to values several times for example, Calderbank and Moo-Young, 1961; Reith and
higher. The influence of gas phase residence time can be Beek, 1968; Linek et al., 1970; Valentin and Preen, 1962;
estimated simply from eq 3. Li et al., 1965; Robinson and Wilke, 1974). The k L value
Severe difficulties will be encountered using the sulfite used in the calculations in this review is 4 X which,
method. Even with the present knowledge reaction ki- judging by those mentioned in the literature, would appear
netics are still evasive and in any case the numerical values to be the best estimate. The most recent publications
would have to be established for each single experiment. reporting the determination of the kLA with the aid of the
360 Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 3, 1979

k + YAN ' T RIET (1975) GASSING OUT 80 3.6

SMITH (1977) ,, 4400 3.6
VOLUME I11 CALDERBANK 119591 TRANSMISSION 5,100 4.0
d . 1 0 - ~ 1 (ms-l.lo-') 12) VALERTIN (19621 BUBBLE DIAMETER 40 4.0
IC) CORRELATION EQUATION (71 4 0
11) CALDERBANK 11958) TRANSBISSION 5.100 0.5
( 2 1 VALERTIN (1962) BUBBLE DIAMETER 40 0.5
(3) VAN 'T RIET 119751 GASSING OUT 90 0 5
0 ROSI.NSON (1973)
0 HASSAN
SMITH
11977)
(1977)
::,, 2.5
2.5
0.46
0.46
180 0.44
$1 CORRELATIOX EQUATION (71 0.5

10-11

2.10
- 2 11
4.10' io3 + E (~rn-3)
Figure 1. Literature data and correlation equation at low superficial
gas velocities in pure water. Figure 3. Literature data and correlation equation at large superficial
gas velocities in pure water.

lations of Calderbank and Valentin, general formulas had

to be used. Obviously, the only sets of measured values
that show a good deal of deviation are those of Hassan and
Kawecki. Kawecki's experimental method, which involves
d
VOLmE "S
( m s - l , 10-'1
suction in a capillary, may well cause large errors. Hassan
B UWECXI (1967) BUBBLE DIAMETER 5 1.4 has performed measurements in a vessel similar to that
A
X VAN "I RIET
SMITH
(1975)
(1971)
GASSING OUT
,,
90
180
1.4
1 7
of Robinson, while the latter's results are not deviating.
v SUITH (19771 4400 1.8
Valentin's results, which imply a rather strong us de-
11) CALDERBANK 11958) TRANS~~ISSLOS 5,100 1.5
( 2 ) VALENTIh (19621 BUBBLE DIAllETER
(Cl CORRELAIIOR EQUATION ( 7 )
40 1 5
1.5
pendence, are in excess of measured kLA values in case of
large us values. All remaining results agree remarkably well
B while the experimental methods are rather different and
the volume may range between 2 and 2600 L.
The data can be correlated according to

for pure water; volume measured up to 2600 L; 500 < P/V

< 10000 (W/m3). The correlation is shown in Figures 1-3
as broken lines and is accurate within approximately
20-40 % .
Measurements in Water with Ions. The addition of
ions to water is known to reduce average bubble size
markedly. As demonstrated by Lessard and Zieminski
(1971) and Marucci and Nicodemo (1967) in model ex-
periments and by Koetsier and Thoenes (1972) in a stirred
vessel, it is especially the frequency of bubble coalescence
that decreases rapidly as the ion concentration increases.
Reith and Beek (1970) have examined gas phase mixing,
which also appears to be complete for ionic solutions a t
a sufficiently high stirrer speed, indicating some coalescing
mechanism. This, however, is in particular due to the
impeller coalescence with the cavity behind the stirrer
blade, which, as shown by Van 't Riet et al. (1976), is also
very intensive in ionic solutions.
The increase in A as a result of the decrease in bubble
diameter is further enhanced by the tendency of the
holdup itself to become larger. Under differing conditions,
Machon et al. (1977) have measured an increase by a factor
 Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 3, 1979 361

in a comparable geometry. The results of Smith e t al.

(1977) are only shown for the P/ V range in which the gas
REITH
LEE
(19681 SLLPHlTI
(1970) TRAiSUISSIOI
5.70,1380
21
4 7 0.79
holdup cannot be of influence.
LEE (19701
In most cases, the sulfite method utilizes 10-100 g of
sulfite/L, which implies a high ion concentration.
Therefore, the results obtained by this method are in-
cluded in this section. It is a safe generalization to say that
pertinent reports are scarce. Most measurements for the
determination of the kLA for physically controlled mass
transfer must be excluded because reaction kinetics have
been very inadequately checked, if at all. Eventually, the
results of Reith and Beek (1968) prove the most reliable.
Converted on the basis of kL = 4 X they are shown
in Figure 4.
It should be noted that the measurements made by Lee
yield values that are larger than those produced by the
P O
dynamic gassing-out method. This discrepancy may quite
well have been caused by the experimental method. They
themselves report to have measured in a region where the
interfacial area was mainly larger than the average one.
If this appears to be true, the measuring results of the
various authors agree rather well. The superficial gas
velocity has little influence in this respect. Westerterp et
0 45 al. (1963) and also Reith and Beek (1968) claim that a t us
1 7
> 0.5 the superficial gas velocity exerts only a slight in-
4.10' io3 - (Wm.3)
I , . , .
11 fluence. Zlokarnik (1970) has found a slightly better
correlation with uso * than with uso. Robinson and Wilke
Figure 4. Literature data and correlation equation in ionic solutions. (1974) have found a uso,36dependence a t us < 0.005. Also
the results of Lee and Meyrick (1970) indicate a slight us
of 2. Lee and Meyrick (1970) have demonstrated by means dependence. It may be expected that there exists a very
of the light transmission method that a t higher ion con- limited influence which, on simple linear regression of a
centrations A may be increased up to 10 times. Yoshida large number of measurements, disappears as a result of
and Miura (1963) have reported measurements showing the dependence of P / V . No author has applied partial
a much smaller variation, but they used a stirrer tip ve- regression techniques. When the uso,2is taken into account,
locity of 0.4-1.7 m/s, which is by no means sufficient to one obtains from Figure 4
ensure proper dispersion (according to a rough-and-ready
rule, that stirrer tip velocity has to be > 2-2.5 m/s). The
value of hL is likewise affected by the addition of ions
k L A = 2.0 X (F ) ~ , ~ ~ ~(9) ~ . ~
-both direct by the presence of ions on the bubble surface where 500 < P / V < 10,000 W/m3; 2 < V < 4400 L; strong
and indirect by changes in the bubble diameter. The direct ionic solutions; accurate for 2G409'0. On comparison with
influence is probably small.
Lessard and Zieminski (1971) and Raymond and Zie- eq 8 for pure water, it will be found that CY and 0 differ
markedly, as can also be seen in Figure 5, where ionic and
minski (1976) report an increase and also a decrease by pure water results are given.
ca. 10-209'0 in the presence of ions and a greater increase
in aluminum salt concentrations. Conclusions for hLA. On the condition that an ac-
curacy of, say, 20-4070 is regarded as acceptable, it may
Various authors (Calderbank, 1959; Calderbank and be said that the results presented in the literature show
Moo-Young, 1961; Linek et al., 1970; Li et al., 1965; Lessard much agreement. A distinction between water with and
and Zieminski, 1971; Yoshida and Miura, 1963; Topiwala
water without ions is, however, essential. By raising the
and Hamer, 1974) have discussed the influence of the ion concentration in solution, k L A is increased consider-
bubble diameter on kL. Most of them make mention of ably. Once a limit of, for example, ca. 10 g of NaCl/L has
a decrease in kL,due to changes in hydrodynamics as the
bubble diameter becomes smaller. Yet there are a few been reached, the increase is much smaller. The dis-
workers who claim the opposite. A t any rate, kL = 1-4 X tinction depends on PIV and us. It increases a t higher
lo4, which implies that it will not become lower by a factor
P/ V values; hence, the kLA's for ionic solutions are more
dependent on PIV than those for pure water.
in excess of 1-4. Bearing in mind the influence of ions on The formulas presented predict a kLA with an accuracy
kL and A , one may expect the product of kL and A to
increase 2-10 times, which, indeed, has been found by, in of 2G40%. For a given determination to be more accurate,
the kLA in vessel and fluid concerned should be measured
particular, Zlokarnik (1975). The results mentioned in the as exactly as possible. The question is whether this can
literature (with the exception of those obtained by the
sulfite method) have been measured by Topiwala and be achieved with the limitations of the experimental
Hamer (1973), Robinson and Wilke (1973, 1974), and methods and it might be better to be content with this
Smith et al. (1977) using the dynamic gassing-out method, margin.
while Lee and Meyrick (1970) have applied the light Vessel Geometry Parameters
transmission technique to establish A values. Influence of Stirrer Type and Number. In the
The results in question are depicted in Figure 4. The discussion of kL values in pure water and ionic solutions,
conversion of A values obtained by Lee, into kLA is based no mention was made of the type of stirrer. It is, however,
on hL = 4 X which, considering the literature, is an immaterial what sort of stirrer is employed because nearly
arbitrary but also the best possible estimate. For that all authors agree that this does not affect any correlation
matter, Reith and Beek (1968) have determined this value of kLA with P / V . A large variety of stirrers such as
362 Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 3, 1979

bubbles enter the vessel by way of the cavity. Bruijn et

al. (1974) have demonstrated that this diminishes dis-
persion and leads to a relatively higher value of power
consumption. It may be expected that, as a consequence,
mass transfer conditions become less favorable. Indeed,
data of Oldshue and Conelly (1977) show that spargers
larger in diameter than stirrers are less suitable.
Karwat (1959) describes a situation in which the sparger
is placed on the bottom. Bubbles from an open pipe then
can miss the stirrer and cannot be dispersed. This causes
a difference in kLA between an open pipe and a sintered
sparger. At higher power values, the difference becomes
smaller under the influence of an increasing circulation
capacity. Results of Reith and Beek (19681, indeed, point
that way, and Karwat has observed virtually no difference
at ca. 5000 W/m3.
There are more authors who have found differences such
as those outlined above. The reported stirrer tip velocities
are, however, much lower and, therefore the results of, for
example, Bartholomew et al. (19501, Miller (1974), and
Paca and Gregr (1976) are not representative of stirred
vessels.
Influence of Fluid Height. No direct correlations with
P/ V have so far been reported. Although Westerterp et
al. (1963) have not worked with P l V , the finding that a t
a given stirrer speed (and, in this connection, a practically
constant P ) kLA + T / H Le., + 1/V confirms the P / V
Figure 5. Correlation equations at several superficial gas velocities relation. Mehta and Sharma (1971) have observed a
for pure water and ionic solutions. linearly decreasing relation with H / T. Within the limits
set, this can, however, be described with the same degree
turbines, paddles, propellers, lochscheiben, rods, and of accuracy as linearly increasing with T I H .
self-sucking agitators, has been used (Judat, 1976; To-
piwala and Hamer, 1973; Valentin and Preen, 1962; The influence of the fluid height can be described in
Karwat, 1959; Van 't Riet, 1975; Zlokarnik, 1975; Smith terms of P/ V , yet only within a defined region, as it as-
et al., 1977). Several authors have not correlated kLA with sumes a well mixed gas phase and invariable kLA over the
P/ V but with various N,D combinations. On evaluation content of the vessel. At extremely high HIT ratios and
of their results, it becomes clear that P/ V correlation is one stirrer near the bottom the contents of a vessel can
not affected by the choice of stirrer (Westerterp et al., 1963; be more properly described as a well-stirred area in the
Mehta and Sharma, 1971; Friedman and Lightfoot, 1957). lower part along with a bubble column in the upper part.
Where it is unimportant which type of stirrer is used and Eventually, the mass transfer in the bubble column will
only the power dissipated in the fluid is essential, it may be hardly dependent on P/ V dissipated in the well-stirred
be expected that the number of stirrers is also immaterial, area.
as, indeed, is confirmed in the articles by Rushton et al. Influence of Stirrer Position. As stated above,
(1956), Moritz et al. (19741, and Oldshue (1966). It must neither the fluid height nor the type and number of stirrers
be added here that on some occasions a fluctuating dif- have any influence on kLA vs. P/ V. It may thus be said
ference (up to 40%) in the amount of mass transfer occurs that the stirrer position is also of no importance. With
but this may be related to experimental accuracy. regard to the stirrer position, two complications may occur.
For calculation of kLA for vessels with many stirrers a t (1) If the distance between stirrer and bottom is too
large Ht/ T ratios the introductory reservations concerning small, Le., smaller than the stirrer diameter, dissipated
gas phase plug flow must be taken into account. power decreases. This, as demonstrated by Mehta and
Exceptions to the above have in particular been reported Sharma (1971) and Karwat (1959), diminishes mass
by Steel (1962) and Westerterp et ai. (1963), who found transfer but data on any correlation with PIV are, un-
a difference by a factor of 10 at variations of D I T (T = fortunately, lacking in their publications.
tank diameter) of 0.20-0.60. This does not seem plausible ( 2 ) If the stirrer is positioned close to the surface, air
in the light of the findings outlined above. The role of can be entrained. At the same time, power consumption
flooding, surface air entrainment, and differences in sulfite decreases. This probably explains the reduced kLA, as
kinetics (see Reith, 1973) may well prove responsible. found by Karwat under such circumstances. He reports
Influence of Type of Sparger. Frequently, a sparger to have observed no differences within the whole range,
is used under the stirrer. Examples are a ring sparger and i.e., between 0.2 T < hR < 0.75T, but the stirrer speed
an open pipe. Van 't Riet (1975) and Bruijn et al. (1974) proved sufficient to bring about recirculation. Where the
have demonstrated that bubbles from the sparger come stirrer speed is too low and, as a result, recirculation does
under the stirrer and are invariably dispersed via the not take place, difficulties may arise (Nienow and Wisdom,
cavity. The choice of sparger then has little effect on the 1976).
mass transfer. Data obtained by Smith et al. (1977) and Surface Air Entrainment. This subject has been dealt
Oldshue and Conelly (1977) point that way. with above. Basically, it will not affect mass transfer,
As the diameter of the sparger relative to that of the provided that the superficial velocity of both sparged gas
stirrer becomes greater, the mechanism of dispersion and entrained gas is adequate. Under extreme circum-
undergoes more changes; thus, in the case of ring spargers stances this is also the case with a self-sucking stirrer where
having a diameter larger than that of the stirrer, fewer no sparged gas is present. This is exemplified by Figure
 Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 3, 1979 363

4 showing the data of Topiwala et al. (1973). N = stirrer speed, l / s

Information on the extent of entrainment has been OTR = oxygen transfer rate, mol/m3 s
reported by Van Dierendonck (1970), Biesecker (1972), and P = stirrer power consumption, W
Mehta and Sharma (1971). In general, the Froude number T = tank diameter, m
is an important feature. The stirrer tip speed shows little t = time, s
V = fluid volume, m3
variation during scaling-up on gas dispersion, which, us = gas superficial velocity, m/s
roughly speaking, means that Fr + N + 1/D. This explains a = exponent
why in larger vessels -especially those >1 m3 -surface p = exponent
air entrainment seldom occurs nowadays. In addition, the t = fractional gas hold-up
effect of entrainment decreases rapidly with increasing us T~ = probe response time, s
because the fluid surface becomes quieter. TG = gas phase residence time, s
Impeller location near the liquid surface will give surface Literature Cited
air entrainment even in large scale equipment. The su- Bartholomew. W. H., Karow, E. O., Sfat, M. R., Wilhelm, R. H., Ind. Eng. Chem.,
perficial velocities of sparged and entrained gas together 42, 1810 (1950).
Biesecker, B. O., VDI Forschungsh., 554 (1972).
determine the kLA value again when the gas phase is well Bossier, J. A,, Farritor, R. E., Hughmark, G. A,, Kao, J. T. F., AIChE J., 19,
mixed. 1065 (1973).
Sparged Gas Energy. At high gas inflow rates, energy Bruijn, W., Van 't Riet, K., Smith, J. M., Trans. Inst. Chem. Eng., 52, 88 (1974).
Calderbank, P. H.,Trans. Inst. Chem. Eng., 38, 443 (1958).
supplied by gas could be of great importance. Only a few Calderbank, P. H., Trans. Inst. Chem. Eng., 37, 173 (1959).
authors (Karwat, 1959; Smith et al., 1977) have recognized Calderbank, P. H., Moo-Young, M. B.. Chem. Eng. Sci., 16, 39 (1961).
Corrieu. G.. Lalande. M.. Perinaer. P., Rev. Ferment. Ind. Aliment.. 30, 125
this. At low power values (<lo3W/m3) and high gas inflow (1976).
rates, gas energy is relatively large. Its influence is more Dang, N. D. P., Karrer, D. A,, Dunn, I. J., Biotechnol. Bioeng., 19, 853 (1977).
limited in case of higher consumptions (>lo3 W/m3). Dunn, I. J., Einsele, A,, J . Appl. Chem. Biotechnol., 25, 707 (1975).
Finn, R. K., "Biochemical and Biological Engineering Science", N. Blakebrough,
Flooding. By flooding is meant that the gas handling Ed., Academic Press, London, 1967.
capacity of the stirrer becomes smaller than the amount Friedman, A. M., Lightfoot, E. N., Jr., Ind. Eng. Chem., 49, 1227 (1957).
Fuller, E. C., Crist, R. H., J . Am. Chem. Soc., 63, 1644 (1941).
of gas introduced. In extreme cases, the gas will bubble Hassan, I. T. M., Robinson, C. W., Biotechnol. Bioeng., 19, 661 (1977).
up around the stirrer and along its shaft. Circulation then Heineken, F. G., Biotechnol. Bioeng., 12, 145 (1970).
shows just an opposite pattern, the dispersing power of the Heineken, F. G., Biofechnol. Bioeng.. 13, 599 (1971).
John, 0. G., Chem. Ing. Tech., 43, 342 (1971).
stirrer decreasing and kLA being lower than may be Joshi, J. B., Sharha, M. M., Trans. Inst. Chem. Eng., 54, 42 (1976).
predicted on the basis of correlations. Judat, H., Thesis, University of Dortmund, 1976.
Karwat, H., Chem. Ing. Tech., 31, 588 (1959).
The deviation of kLA curves a t lower power consumption Kawecki, W., Reith, T., Van Heuven, J. W., Beek. W. J., Chem. Eng. Sci., 22,
values largely involves flooding. In case of flooding, kLA 1519 (1967).
is little dependent, if at all, on power consumption values. Koetsier, W. T.. Thoenes, D., Proceedings, Symposium on Chemical Reaction
Engineering, B 3-15, Amsterdam, May 1972.
Flooding is a rather complicated problem, and an extensive Landau, J., Boyle, J., Gomaa, H. G., AI Taweel, A. M., Can. J . Chem. Eng.,
discussion would lie outside the scope of this review. For 55, 13 (1977).
Lee, J. C., Meyrick, D. L., Trans. Inst. Chem. Eng., 48. T 37 (1970).
more information, see Judat (19761, Biesecker (19621, Lessard, R. R., Zieminski, S. A., Ind. Eng. Chem. Fundam., 10, 260 (1971).
Rushton and Bimbinet (1968), Zwietering (1963), Nienow Li, P. S.,West, F. B., Vance, W. H., Moulton, R. W., AIChE J., 11, 581 (1965).
and Wisdom (1976), Van 't Riet (19751, and John (1971). Linek, V., Mayrhoferova, J., Mosnerova, J., Chem. Eng. Sci., 25, 1033 (1970).
Linek. V., Mayrhoferova, J., Chem. Eng. Sci., 25, 767 (1970).
Conclusions for Vessel Geometry Parameters. On Linek, V., Tvrdik, J., Biofechnol. Bioeng., 13, 353 (1971).
correlation with P/ V , there is no influence of stirrer ge- Linek, V., Biotechnol. Bioeng., 14, 285 (1972).
Linek, V., Sobotka, M., Prokop, A,. Biofechnol. Bioeng. Symp., 4, 429 (1973).
ometry and number of stirrers on mass transfer in non- Linek, V., Vacek. V., Biotechnol. Bioeng., 18, 1537 (1976).
viscous systems. Linek, V., Vacek. V., Biotechnol. Bioeng., 19, 983 (1977).
In the event of full recirculation, the stirrer position is Linek, V.. Benes, P., Biotechnol. Bioeng., 19, 741 (1977).
Lockett. M. J.. Safekourdi, A. A,, AIChE J . , 23, 395 (1977).
rather immaterial. By placing the stirrer between the Machon, V., Vleck, J., Kdma, V., 2nd European Conference on Mixing, Cambridge,
middle of the vessel and one stirrer diameter away from England, Paper F 2, 1977.
Marrucci, G., Nicodemo, L., Chem. Eng. Sci., 22, 1257 (1967).
the bottom, an optimal power consumption value and mass Mehta, V. D., Sharma, M. M., Chem. Eng. Sci., 26, 461 (1971).
transfer are ensured. Miller, D. N., AIChE J., 20, 445 (1974).
At a fluid height in the range of 0.5 < T / H < 1.5, P / V Moritz, V., Silveira, R. S. A., Meireles. D. F., J . Ferment. Technol., 52, 127
(1974).
appears to be a variable with which kLA can be correlated. Mukhopadhyay, S. N., Ghose, T. K., J . Ferment. Technol., 54, 406 (1976).
The sparger may not be fitted at a distance from the center Nienow, A. W., Wisdom, D. T., Proceedings, 3rd Annual Research Meeting, "Mixing
and Dispersing in Multiphase Systems", Salford, March 1976.
larger than the stirrer radius. It may be useful to correlate Oldshue, J. Y., Biofechnol. Bioeng., 8, 3 (1966).
with the sum of power introduced by the stirrer and gas. Oldshue, J. Y., Conelly F. L., Chem. Eng. Prog., 73, 85 (1977).
In general, it may be concluded that on condition that Onken, U., Schalk, W., Chem. Ing. Tech., 49, 573 (1977).
Paca, J., Gregr, V., J . Ferment. Technol., 55, 166 (1977).
the vessel is properly stirred (which means absence of Prasher, B. D., AIChE J . , 21, 407 (1975).
flooding or surface air entrainment and the presence of a Raymond. D. R., Zieminski, S. A., Can. J . Chem. Eng.. 54, 425 (1976).
Reith, T., Beek, W. J., Proceedings 4th Evopean Symposium on Chemical R e a c t h
suitable gas sparger) and allowance is made for the ions Engineering, Brussels, p 191, 1968.
in the solution, P/ V and us are sufficient to describe kLA. Reith, T., Beek, W. J., Trans. Inst. Chem. Eng., 48, T 63 (1970).
Reith, T., Brit. Chem. Eng., 15, 1559 (1970).
Nomenclature Reith, T.. Beek, W. J., Chem. Eng. Sci.. 28, 1331 (1973).
Robinson, C. W., Wilke, C. R., Biofechnol. Bioeng., 15, 755 (1973).
A = specific contact area, m2/m3 Robinson, C. W., Wilke, C. R., AIChE J.. 20, 285 (1974).
CG = O2 concentration in the gas phase, mol/m3 Rushton, J. H.,Galhgher, J. B., Okkhue, J. Y., Chem. Eng. PTcg.,52, 319 (1956).
Rushton, J. H., Bimbinet, J. J., Can. J . Chern. h g . , 46, 16 (1968).
CL = O2 concentration in the liquid phase, mol/m3 Schaftlein, R. W., Russell, T. W. F., Ind. Eng. Chem., 80, 12 (1968).
D = stirrer diameter, m Smith. J. M., Van 't Ret. K., Middleton, J. C., 2nd European Conference on Mixing,
d b = bubble diameter, m Cambridge, England, Paper F 4, 1977.
H = Henry coefficient Steel, R.. Maxon, W. D., Biotechnol. Bioeng., 4, 231 (1962).
Steel, R., Maxon, W. D., Biofechnol. Bioeng., 8, 108 (1966).
H , = tank height, m Todtenhaupt. E. K., Chem. Ing. Tech., 43, 336 (1971).
h R = distance between stirrer and bottom, m Topiwala, H. H., Hamer, G., Biotechnol. Bioeng. Symp., 4, 547 (1973).
K = constant Topiwala, H. H.. Hamer, G., Trans. Inst. Chem. Eng.. 52, 113 (1974).
k~ = mass transfer coefficient in the gas phase, m/s Valentin, F. H. H., Preen, B. V., Chem. Ing. Tech., 34, 194 (1962).
Van de Sande, E., Thesis, University of Technology, Delft, 1974.
kL = mass transfer coefficient in the liquid phase, m/s Van Dierendonck, L. L., Thesis, University of Technology, Twente, 1970.
k R = reaction rate constant Van 't Riet, K., Thesis, University of Technology, Delft, 1975.
364 Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 3, 1979

Van t Riet, K., Boom, J. M., Smith, J. M., Trans. Inst. Chem. Eng., 54, 124 Zlokarnik, M., Chem. Ing. Tech., 47, 281 (1975).
(1976). Zwietering, Th. N., Ingenieur, 75, 63 (1963).
Votruba, J., Sobotka, M., Eiotechnol. Bioeng., 18, 1815 (1976).
Votruba, J., Sobotka, M., Prokop, A,, Eiotechnol. Bioeng., 19, 435 (1977). Received for review J u n e 26, 1978
Westerterp. K. R., Van Dierendonck, L. L., De Kraa, J. A,, Chem. Eng. Sci., Accepted December 20, 1978
~.
18. 157 119631.
~ -~~
\

Wise, W. S., J . den. Microbial., 5, 167 (1951).

Yoshida, F., Miura, Y., Ind. Eng. Chem. Process Des. Dev., 2, 263 (1963). P a r t of this article was presented as a paper a t the Engineering
Zlokarnik, M., Chem. Ing. Tech., 42, 1310 (1970). Foundation Mixing Conference, August 1977, Rindge, N. H.

ARTICLES

Simulation of Ammonia Synthesis Reactors

Chandra P. P. Singh and Deokl N. Saraf

Department of Chemical Engineering, Indian Institute of Technology, Kanpur-2080 16, India

A workable method to calculate diffusion effects within pores of a catalyst pellet, for a complex reaction, has been
developed. Suitable rate equations have been selected to describe the rate of ammonia synthesis reaction over
catalysts of different make. This, in conjunction with the method developed to calculate the effectiveness factor,
has been used to obtain a mathematical model for ammonia synthesis reactors. Reactors having adiabatic catalyst
beds with interstage cooling as well as autothermal reactors have been considered. Plant data and simulation results
are, generally, in very good agreement.

Introduction A mathematical model considering all physical and

Synthesis of ammonia from hydrogen and nitrogen is chemical processes in the reactor has been described in this
one of the simplest kinetic reactions. The synthesis is paper. A method to solve transport equations to evaluate
straightforward, there is no side reaction, and the product the effectiveness factor, up to the desired accuracy, is
is stable. The physical and thermodynamic properties of developed and used in the model calculations.
the reactants and products are well known (Gillespie and Rate Expressions
Beattie, 1930). However, the mechanism of this reaction The literature contains innumerable rate expressions.
over the synthesis catalyst (iron catalyst) is not well Those reported until the early thirties have been sum-
understood. This has led to numerous rate equations, all marized by Frankenburg (1933) and Emmett (1932,1940).
of which are of complex order. Because of the complexity In 1940 Temkin and Pyzhev developed a rate equation
of these rate equations it is difficult to account for the which offered a satisfactory kinetic approach to the
diffusional resistances to the transport of reactants and synthesis and decomposition of ammonia over doubly
product in catalyst pores. This, in addition to limited promoted iron catalysts. Since then this rate equation as
reliability of rate equation, makes it difficult to have a such or in modified forms has been most extensively used,
mathematical description of the processes taking place although some doubts about the generality of the equation
inside an ammonia synthesis reactor. have been raised (Adams and Comings, 1953; Emmett and
Major changes have taken place in the design of am- Kummer, 1943; Hays et al., 1964). The modified form of
monia synthesis reactors since the first commercial pro- the Temkin equation (Dyson and Simon, 1968) used in this
duction started in 1925 (Johansen, 1970). Most of these work, is as follows
changes have been based on historical plant data rather
than an insight into the physical and chemical processes
taking place in the reactors. However, the use of com-
puters in design, optimization, and control made it nec-
essary to have a mathematical description of the process.
Simulation models for ammonia synthesis converters of where r N H 3 = reaction rate, kg-mol of NH,/(h m3 of cat-
different types have been developed for design, optimi- alyst), K 2 = velocity constant of the reverse reaction,
zation (Murase et al., 1970; Singh, 1975), and control (Shah, kg-mol/(h m3),fN2, fH1, fNHB = fugacities of nitrogen, hy-
1967; Shah and Weisenfelder, 1969) purposes. drogen, and ammonia, respectively, k , = equilibrium
T o describe the reactor operating conditions as accu- constant of the reaction: 1.5H2+ 0.5Nz * NH,, and a =
rately as possible, the simulation model should take into constant. According to some workers a = 0.5 for all iron
consideration all the physical and chemical processes catalysts (Morozov et al., 1965; Temkin and Pyzhev, 1940;
taking place in the reactor. In order to avoid the com- Temkin e t al., 1963) whereas others obtained values
plexity resulting from such a consideration, earlier workers ranging from 0.4 to 0.8 (Anderson, 1960; Bokhoven et al.,
have attempted only approximate simulations. 1955; Bokhoven and Van Raayen, 1954; Brill, 1951; Buzzi
0019-7882/79/1118-0364$01.00/0 0 1979 American Chemical Society