TRANSPORT PHENOMENA AND REACTION IN FLUIDIZED

CATALYSTBEDS
Terukatsu Miyauchi and Shintaro Furusaki

Department of Chemical Engineering

University of Tokyo
Tokyo, Japan

Shigeharu Morooka

Department of Applied Chemistry

Kyuahu University
Fukuoka, Japan

and

Yoneichi lkeda

Fluidization Engineering Laboratory

Tokyo, Japan

I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
A. Purpose and Outline of the Review . . . . . . . . . . . . . . . . 276
B. General Properties of Fluidized Beds . . . . . . . . . . . . . . . 277
C. Historical Development of FCB Studies . . . . . . . . . . . . . . 281
D. Quality of Fluidization in Relation to Particle Properties . . . . . . . 283
11. Flow Properties of Fluid Beds . . . . . . . . . . . . . . . . . . . . . 285
A. Particle Properties in Relation to Fluidity . . . . . . . . . . . . . 285
B. Formation, Splitting, and Coalescence of Bubbles . . . . . . . . . . 290
C. Operation of Fluid Beds . . . . . . . . . . . . . . . . . . . . . 297
D. Properties of Bulk Flow Inside Fluid Beds . . . . . . . . . . . . . 297
E. Axial Distribution to Bed Structure . . . . . . . . . . . . . . . . 305
F. Effect of Bed Internals . . . . . . . . . . . . . . . . . . . . . . 307
111. Turbulent-Flow Phenomena in Bubble Columns and Fluidized
Catalyst Beds . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 10
A. Two-Phase Bubble Flow in Recirculation . . . . . . . . . . . . . . 311
B. Comparison of Theory with Experiments on Bubble
Columns . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
275
Copyright @ 1981 by Academic Press, Inc.
ADVANCES IN CHEMICAL ENGINEERING, VOL. 11 All rights of reproduction in any form reserved.
ISBN 0-12-008511-9
276 M I Y A U C H I . F U R U S A K I . MOROOKA. A N D I K E D A

C . Phenomenological Background of Turbulent Kinematic

Viscosity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
D . Interpretation of Experiments for FCB . . . . . . . . . . . . . . . 327
IV. Longitudinal Dispersion Phenomena as Derived
from Flow Properties . . . . . . . . . . . . . . . . . . . . . . . . . 330
A . Theory of Longitudinal Dispersion for the Recirculation
Flow Regime . . . . . . . . . . . . . . . . . . . . . . . . . . 331
B . Longitudinal Dispersion in a Bubble Column . . . . . . . . . . . . 335
C . Longitudinal Dispersion in a Fluidized
Catalyst Bed . . . . . . . . . . . . . . . . . . . . . . . . . . 338
V. Bubble Phenomena in Relation to Bed Performance . . . . . . . . . . . 340
A . Mean Bubble Velocity without Bulk Recirculation . . . . . . . . . . 341
B . Mean Bubble Velocity with Bulk Recirculation . . . . . . . . . . . 344
C . Bubble Splitting in Turbulent Flow . . . . . . . . . . . . . . . . 350
VI . Heat and Mass Transfer in Fluidized Catalyst Beds . . . . . . . . . . . 360
A . Bubble Dynamics . . . . . . . . . . . . . . . . . . . . . . . . 360
B . Heat and Mass Transfer through the Bubble Wall . . . . . . . . . . 363
C . Mass Transfer between Bubble and Emulsion Phase . . . . . . . . . 365
D . Particle Capacitance Effect . . . . . . . . . . . . . . . . . . . . 371
E . Axial and Radial Mixing of Heat and Mass . . . . . . . . . . . . . 373
F. Review of Wall Heat Transfer in Fluid Beds . . . . . . . . . . . . 379
VII . The Successive Contact Mechanism for Catalytic Reaction . . . . . . . . 381
A . Experimental Facts and Reactor Models . . . . . . . . . . . . . . 381
B. The Successive Contact Mechanism . . . . . . . . . . . . . . . . 390
C . Overall Rate Constant KO, Based on L , . . . . . . . . . . . . . . 399
VIII . Further Properties of the Successive Contact Mechanism . . . . . . . . 402
.
A Axial Distribution of Reactivity in a Fluid Bed . . . . . . . . . . . 402
B. Axial Distribution of the Contact-Mechanism
Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . 407
IX . Nonisothermal Effect on the Bed Performance . . . . . . . . . . . . . 413
A . Effect on Steady Reaction . . . . . . . . . . . . . . . . . . . . 413
B . Stability of Fluid Bed Reactors . . . . . . . . . . . . . . . . . . 421
X. Discussion and Summary . . . . . . . . . . . . . . . . . . . . . . . 425
A . Applicability of Reactor Models . . . . . . . . . . . . . . . . . . 425
B . Development of Industrial Fluidized Catalyst
Beds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 426
C . Recent Trends in Fluidized Catalyst Beds . . . . . . . . . . . . . 429
D . Technical Problems in FCB Design . . . . . . . . . . . . . . . . 430
Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . 432
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 431
Note Added in Proof . . . . . . . . . . . . . . . . . . . . . . . . . 448

1. Introduction

A . PURPOSE
A N D OUTLINE
OF THE REVIEW
Fluidization today seems a very young and active field. even though
Charles E . Robinson originated the technique a century ago . It is being
applied today in the chemical industry to catalytic reactions. noncatalytic
 F L U I D I Z E D CAT AL YST BEDS 277

reactions, and many physical operations. Warren K. Lewis and Edwin R.

Gilliland can be said to have begun modern engineering on the fluidized
catalyst bed, which was started in an elaborate study on hindered settling
and developed in a great number of theses at MIT (Ml).
The purpose of this review is to consider the uses of fluidized beds for
catalytic reactions, using a viewpoint quite different from studies directed
toward the physical handling of solid particles or toward gas-solid non-
catalytic reaction. At this point it is useful to outline the scope of the
review.
First, it will be shown that flow properties of the fluidized catalyst bed
(FCB) are clearly different from those of other conventional fluidized
beds. The different treatment required is very significant for research and
development on fluidized catalytic beds. Next, factors affecting the flow
properties are discussed, especially particle size distribution, and also
heat and mass transfer, and mixing properties.
Reactions in the FCB are then discussed. The mechanism of successive
contact (M25) is presented. Contact efficiency is greater in the dilute phase,
i.e., freeboard region, than in the underlying dense phase. In the FCB, the
dilute phase plays an important role in advancing the catalytic reaction
when reaction rates are high. This factor provides a basis for identifying
the appropriate reaction model and clarifying the effect of the dilute phase
on selectivity and stability.
To understand particle circulation in the FCB, the simplified theory
(M31)for a bubble column is applied so as to analyze flow patterns of
bubbles and “emulsion.” Finally, applications of all these studies are
investigated. Apparatus problems and the development of improved
FCBs are given considerable attention.

B. GENERAL
PROPERTIES
OF FLUIDIZED
BEDS
In fluidization, a suspension of fine solid particles behaves like a liquid
during the upflow of a supportive gas or liquid phase. Thus the bed of
fluidized solid itself may be analyzed similarly to liquid systems. The
gas-lift effect produces internal recirculation, by providing a descending
flow of high particle concentration and an ascending flow of low particle
concentration. This effect resembles the circulation in bubble columns.
Whereas bubble columns contain dispersed gas and a continuous liquid
phase, the fluidized bed comprises the bubble phase and the emulsion
phase in which particles have gained fluid-like properties by the interstitial
gas flow.
The interrelation of pressure drop and superficial gas velocity appears
to determine the various states of fluidization, as shown in Fig. 1 by
278 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

UPANSION CONVEYING RATE

AT ~ATURATION'

FIG. 1 . Various states of fluidization (after Y l l , Y12).

Yerushalmi et al. (Y 11, Y 12). With increasing gas velocity, the pressure
drop increases while the bed is in a fixed state; then it stays almost
constant, equal to the fluidizing density of the bed after the minimum
fluidization velocity. Further increase of gas velocity makes the bed
fluidize vigorously; a change occurs from the bubbling bed to the turbu-
lent bed and finally to the fast fluidized bed, and the amount of particles
entrained by the gas flow likewise increases. The charge of particles in a
fluidized bed will not remain constant unless the particles are separated in
cyclones and returned to the bed.
Most industrial fluidized beds using fine powder are found to be in the
category of turbulent beds; typical of these is fluidized catalytic cracking.
On the other hand, most noncatalytic reactions and physical operations
use coarse particles and are operated as bubbling beds. Squires (S14) has
pointed out that the state of fluidization varies significantly with fluidized
particle diameter, and has identified the following two categories:
 FLUIDIZED CATALYST BEDS 279

(a) Fluid bed. The fluidization of fine particles, generally all through 20
mesh; usually with a substantial part smaller than 200 mesh, often smaller
than 325 mesh. Superficial velocities generally below 60 c d s e c .
(b) Teerer bed. The fluidization of coarse particles, generally all larger
than 60 mesh; with a substantial part larger than 10 mesh. Superficial
velocities generally above 150 c d s e c .
Considering Squires’ criteria and observed industrial performance
(G15, K24, M2,010, V12,Z5), Ikeda (14) modified the criteriaas shown in
Table I, stressing that the fluidized catalyst bed should be viewed as the
typical fluid bed (11-110). The general flow features of a fluid bed operated
under relatively high gas velocity are illustrated in Fig. 2 (M25). The
dense bed consists of the emulsion phase and bubble phase; the emul-
sion phase exhibits liquid-like properties, in the case of the fluid bed.
Intense circulation of dense phase results from the central rapidly ascend-
ing bubble-rich phase and the peripheral descending emulsion phase of
low bubble content. The boundary of the dense phase and the dilute phase
is rather ambiguous; a transition region exists in which bubbles collapse
and particles are blown in from the emulsion phase. In the transition

TABLE I
TYPICAL
FLUIDBEDAND TEETERBED

Characteristics Fluid bed Teeter bed

Mean particle size 50-70 wm 0.1-2 mm (mostly -0.4 mm)

Range of sizes 5- 100 pm 0.05-5 mm
U,[cdsec] 30-90 30- 150
UJUrn, 50-200 3- 10
Lrn, [cml 50-500 50- 150
1-2 0.1-0.5

Constitution Bed divides into dense phase + Density of bed is almost constant,
dilute phase, although surface of bed rather well
surface of dense phase is defined
ill-defined
Behavior of Most gas flows as uniformly small Large bubbles, which may
bubbles bubbles coalesce progressively
during rise
Behavior of Circulative flow is developed Generally relative movement of
emulsion (possibly central upward, particles; partially circula-
phase peripheral downward) tive flow
Stability Slugging unlikely, even for Frequent slugging for
large LJD, L,, > 1-1.5 m
280 M I Y A U C H I , F U R U S A K I , MOROOKA, A N D I K E D A

FIG. 2. General flow features of the fluid bed.

 FLUIDIZED CATALYST BEDS 28 1

region, inversion of the circulating flow can occur. In the dilute phase,
particles are suspended particulately in the stream of ascending gas. The
content of suspended particles decreases with height, except in the upper
part of the dilute phase where it stays constant.

C. HISTORICAL OF FCB STUDIES

DEVELOPMENT
As is well known, the fluid catalytic cracking (FCC) process was the
first application of the fluidized bed to catalytic reactions. Esso Research
and Development Company (now Exxon) developed the process in a very
short period (15 months) from the operation of the pilot plant to the
start-up of the commercial plant, under the supervision of Warren K.
Lewis of MIT. A major reason for the quick success of the development is
that Lewis farsightedly chose fluid beds for FCC, instead of teeter beds
which had been applied for the Winkler process, and backed up this
choice with sustained basic research. Fluid beds are preferable for cataly-
tic reactions, because of: (1) sufficient inventory of catalyst, ensuring
adequate contact time, heat-transfer area, and process inertia (fly-wheel
effect); (2) good fluidization, with only small perturbations of pressure
drop, and consequent stable operation; (3) good fluidity-particles easy to
handle, with low rates of erosion of bed apparatus and catalyst attrition;
(4) easily manufactured catalyst particles, produced in large quantity in
spray dryers. Moreover, the purpose of fluidized-bed operation was to
remove coke formed and still operate continuously, rather than to achieve
high conversion (conversion of 50-60% was acceptable). Frantz (F6)
stated, commenting on the good fortune of matching FCC and FCB, “If
some other reaction had been chosen for the commercial fluid unit, fluidi-
zation as a commercial operation would have been put back on the shelf
for several years.”
Carlsmith and Johnson (C2) have reviewed the research and develop-
ment of the FCC process. Drawing on their data, Ikeda (14) has recalcu-
lated scale-up ratio and operating results, as shown in Table 11. The feed
rate of oil is roughly proportional to the cross-sectional area of the reac-
tor. Thus a large superficial velocity was chosen for the small-scale reac-
tors, and scale-up was conducted under a constant superficial velocity.
The success of FCC encouraged applications of the fluidized catalyst
reactor to other catalytic reactions. Successful applications can be found
in fluid catalytic reforming, production of alkyl chloride by oxychlorina-
tion, production of phthalic anhydride, acrylonitrile synthesis by ammox-
idation, and production of maleic anhydride.
In applying the fluidized bed to catalytic reactions, the catalyst must
have the appropriate properties and activity. For production of phthalic
282 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

TABLE I1
SCALE-UP &TI0 AND REACTION
IN T H E DEVELOPMENT
OF FCC

Small-scale Large-scale Commercial

pilot plant pilot plant plant

Reactor
Bed diameter, DT 2 in. 15 in. 15 ft.
Relative diameter 0.133 1 12
Relative cross section 0.018 1 144
Bed height, L , [ftl 20 20 28
Relative bed height 1 1 1.4
LtI4 120 16 1.87
Regenerator
Bed diameter 4 in. 22 in. 19.5 ft.
Relative diameter 0.18 1 10.6
Relative cross section 0.03 1 112
Bed height [ft.] 20 31 37
Relative bed height 0.67 I 1.2
LtIDT 60 17 2.3
Capacity [bbVdayl 2 100 15,000
Relative capacity 0.02 1 150
Product
Gasoline [vol. %I 50.2 49.7 49.5
Gas oil [vol. %I 50.0 50.0 50.0
Dry gas [wt. %] 4.2 5.0 6.2
Carbon [wt. %] 3.2 2.8 2.8

anhydride, Betts (B11)stated that the ground catalyst easily suffers attri-
tion and produces dusty fines; these give rise to heat-exchanger fouling,
cyclone or filter blockage, and after-burning. Betts showed that the mi-
crospherical catalyst resisted attrition and eliminated the difficulties listed.
According to Graham and Way (G14), fluidized beds are ideally suited to a
moderately active catalyst because they involve contact times and large
masses of catalyst; the more highly active fixed-bed catalysts may give
reduced yield due to overoxidation and should be avoided in fluid beds.
The Fischer-Tropsch synthesis produces hydrocarbons from H2 and
CO. Fluid beds were applied to it in the Hydrocol process using ordinary
turbulent beds and in the Kellogg process using fast beds (Fig. 1). Devel-
opment of the Hydrocol process has been reported by Grekel et al. (G15)
and Hall and Taylor (Hl). The main difficulties in scale-up of these
fluidized beds were incomplete fluidization and low conversion.
On the basis of this experience with the Hydrocol process, Volk et al.
(V12) proposed providing internals in fluid beds in order to achieve good
 F L U I D I Z E D CATALYST BEDS 283

fluidization and raise the contact efficiency by performing the following

functions: (1) prevention of growth of bubbles; (2) promotion of lateral
movement of gas and solids; (3) prevention of formation of dead pockets
of solids, (4) prevention of elutriation of fine particles, so as to maintain
the original particle size distribution; ( 5 ) allowance for periodic removal of
the entire bed of solid particles from the reactor.
Volk et al. (V12) used vertical surfaces (tubes or half-rounds) with
success in a version of the Hydrocol synthesis known as the HRI process.
The conversion of CO gas for several reactor diameters was correlated by
the following equivalent diameter:

4(free cross-sectional area)

(1-1)
Deq = (wetted perimeter of all vertical surface)

The HRI process concept has greatly contributed in the application of

fluid beds to catalytic reactions.
Perhaps the most successful application since FCC is the Sohio process.
The reasons for the success are explained as follows. First, the pro-
cess achieved high conversion and selectivity in FCB, even in a reaction
system involving several parallel and consecutive reactions. Second, the
process was industrialized directly by the use of fluid beds, rather than by
passing through the stage of fixed beds. Callahan et al. (Cl) have reported
on development of the initial catalyst, bismuth molybdate, for the Sohio
process.

OF FLUIDIZATION
D. QUALITY I N RELATION
TO PARTICLE PROPERTIES

In industrial fluidized beds, it is always important to establish good

fluidization, and especially so for high-performance catalytic reactions
with conversions over 90% and selectivity over 70%. Volk et al. (V12)
summarized the variables defining good fluidization and good gas-solid
contact, as follows: (1) Gas entry-designed so that the gas entering the
bed is well distributed. (2) Gas velocity-high enough to keep the solid in
motion, but not so high that gas channeling occurs. (3) Bed height-
relatively low. With other factors constant, the greater the bed height, the
more difficult it is to obtain good fluidization. (4) Gas and solid
densities-relatively high-density gas and low-density solid. The closer
the ratio of densities of gas and solid, the easier it is to maintain good
fluidization. ( 5 ) Particle size-a wide range of sizes gives more stable
fluidization than particles of uniform size. (6) Reactor internals-serving
the functions described earlier.
284 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

I l l I l l 1
20 50 100 200 500 1000 2000
m i c l c size dp b m l
FIG. 3. Classification of particles by Geldart ((34, G5).

Geldart (B 1, G4, G5) has classified fluidized particles into four groups,
taking account of the fact that the behavior of fluidized beds is consid-
erably affected by properties of the fluidized particles:
Group A . Powders having a particle density less than about 1.4 g/cm3;
in particular, those with porous structure and a mean diameter in the
range of 20-100 pm.
Group B. Powders having a particle density in the range of 1.4 to about
4 g/cm3, and a mean size in the range of 40-500 pm.

TABLE 111
DESIRED PROPERTIES O F FLUIDIZED CATALYST PARTICLES

Characteristics Optimal properties Main reasons

Particle shape Spherical, smooth surface Fluidity; resistance against

attrition
Bulk density 0.4- 1.2 Good fluidization. If larger than
[g/cm31 this, the bed has a tendency
toward slugging. If smaller,
attrition and carry-over increase
Particle size Average 50-70 pm; Fluidity; good fluidization
distribution fraction with (d, 5 44 pm)
should be 10-20% for
LPCR, 20-40% for HPCR"
Attrition Attrition rate <O. l%/hr Minimize catalyst consumption,
resistance maintain particle size distribu-
tion and particle shape

" HPCR = high-performance catalytic reaction, such as ammonia oxidation;

LPCR = low-performance catalytic reaction, such as FCC.
 FLUIDIZED CATALYST BEDS 285

Group C . Powders in which surface forces become overwhelmingly

important, e.g., all powders having a mean size less than about 30 pm.
Group D . Powders of size usually over 600 pm.
This classification is illustrated in Fig. 3, in terms of size and density of
particles. Comparing Geldart’s classification with Table I, we find that
group A roughly corresponds to the typical fluid bed, and group B (as well
as part of group D) corresponds to the typical teeter bed. According to
Geldart, the behavior of gas bubbles and the effective viscosity of
fluidized beds can be interpreted by a “surface/volume diameter” regard-
less of the particle size distribution.
Ikeda (14) paid attention to the effect of the fines fraction ( d , < 44 pm),
and deduced the optimal size range to obtain good fluidization and good
gas-solid contact efficiency from the results of industrial operations (010,
V12, ZS; see Table 111). His criterion for good fluidization is indicated as
A’ in Fig. 3, a target in developing or selecting catalysts for FCB.

It. Flow Properties of Fluid Beds

A. PARTICLE PROPERTIES I N RELATION

TO FLUIDITY

1. Effect of Fines
Particle size distribution has a great effect on most aspects of fluidiza-
tion. Zenz (24) pointed out that fluid-solid systems could be classified in
terms of a bed viscosity, which is probably controlled by the rubbing of
coarse particles against one another. According to Trawinski (T25, T26),
the fines between coarse particles act as a lubricant, reducing bed viscos-
ity. The minimum concentration of fines is the quantity required to coat
each coarse particle with a monolayer of fines. Extending this aproach to
multidispense systems, Zenz (24) calculated the optimum size distribution
of fluidized particles, shown in Fig. 4 and expressed by a skewed normal
probability distribution, which agrees with a typical size distribution of
FCC particles.
Ikedaet al. (15, IlO), investigating the effect of fines on bed expansion,
pressure fluctuation, and discharge rate of particles from a 8-cm-i.d. fluid
bed, found that bed expansion and discharge rate increase and pressure
fluctuations decrease with increasing content of fines (<44 pm). The range
of particles smaller than 44 p m is called “fines fraction”; its role has
been summarized by Ikeda (12) as shown in Table IV. The desirable prop-
erties of fluidized particles utilized as catalyst were given by Ikeda (12) as
286 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

FIG.4. Typical size distribution of particles used in fluid bed.

shown in Table V. Morse and Ballow (M44), de Groot (D7) and Matsen
(M10) have also recognized that the quality of fluidization is improved by
using smaller particles with a wider size distribution. Geldart (G4, GS)
classified the fluidized beds into four categories characterized by density
difference (p, - p,) and mean size of particles, as presented in Section I.
The terms “fluid bed” and “teeter bed” (Squires, S14) seem to corre-
spond to group A and group B, respectively, although the criteria which
distinguish among the groups are not well established.

2. Expansion of the Emulsion Phase

The expansion of a gas-solid fluidized bed at gas rates between the
minimum-fluidization velocity and the initial bubbling velocity was stud-
ied in 1966 by Davies and Richardson (D6). This homogeneous expansion
of the bed is considerably larger when finer particles are fluidized. Several
studies have been carried out on bed expansion (D6, D15, M41); on
particle-particle forces and interactions (D14a, M6, R9), and on the flow
rate of gases through beds (D6) under the condition of particulate fluidiza-
tion.
Figure 5 shows typical expansion data measured by Morooka et al.
(M41) using unclassified FCC catalyst particles. As soon as the bed
reaches minimum fluidization, it starts to expand. After bubbles start to
form (i.e., under aggregative fluidization) the bed expands slowly and then
 TABLE IV
EFFECTSOF THE CHANGEOF WEIGHTFRACTION
OF PARTICLES
SMALLER
THAN
44 pm (“FINESFRACTION”)
(12)
~~ ~

Change of weight fraction and its causes Main effect

When ( I ) attrition resistance of particles is ( I ) Slugging prevails

large, (2) Erosion of bed increases due to poor
(2) catalyst recovery system is less fluidization
efficient, and (3) Load on catalyst recovery system
(3) fine particles are not supplied becomes unstable, and instantaneous
and coarse particles are not load on the system increases
discharged during the operation, (4) Pressure and temperature control
then particles smaller than 44 pm become becomes difficult
low in weight fraction. ( 5 ) Yield of products decreases because
side reactions come to prevail.
Occasionally, afterburning takes place
attrition resistance of particles is (1) Channelling prevails
small, (2) Large clusters of particles grow due to
catalyst recovery system is agglomeration
efficient, and (3) Operational troubles occur due to
coarse particles are not supplied particle adhesion (Decrease of heat
and fine particles are not transfer, excess reaction at walls in
removed during the operation, freeboard, and clogging in pipes and
then particles smaller than 44 pm become cyclones will take place.)
high in weight fraction. (4) Load on catalyst recovery system
increases
( 5 ) Afterburning takes place
(6) Temperature becomes unstable locally
and side reactions prevail

u, [cmlsrcl
FIG. 5. Expansion and pressure drop of a fluid bed (after M41).
 TABLE V
DESIRABLE PROPERTIES OF FLUIDIZED
PARTICLES
~~ ~~ ~ ~~~~

Property Main interests Troubles in commercial- Supplementary methods

of particles Optimum values of operation scale fluid bed of improvement

Chemical Apparent contact time Stability of operation High activity: excess reaction in Dilution with inert particles
activity 3 < T < 30 sec Homogeneous temperature in dilute phase Multistage fluid bed
bed LJDT = 1-2 Low activity: decrease of reaction
yield
Particle size Weight meAn diameter Improved fluidity Severe erosion if particles are coarse Installation of internals
distribution 50 < dp < 70 pm Good fluidization Difficulty in recovering particles Multistage fluid bed with
+80 pm, 5-20 wt. % Decreased catalyst loss and (10 pm horizontal baffles
-40 pm, 10-30 wt.% attrition
Attrition Rate of attrition Avoidance of catalyst loss Very fine particles produced by Reduction of gas velocity
resistance (Forsythe Maintenance of optimum size attrition (cf. Table IV) injected through
method) distribution distributor
<O. 1-0.5
wt. %/hr

Shape of Microspherical Low tendency toward attrition High erosion rate when irregular Installation of internals
particles Decreased erosion hard particles are used Blending with MS particles
Particle Below 2-3 g/cm3 Good fluidization Unsatisfactory fluidization when Blending with low-density
density high density particles are used particles
Installation of internals

* From 12.
 FLUIDIZED CATALYST BEDS 289

its volume remains constant or even decreases with increasing gas veloc-
ity; the volume of the emulsion phase is always smaller than at the bubble
point. Thus it is not correct to select the minimum-fluidization bed height
as the standard bed height for calculation of gas-bubble holdup. The
equivalent height of emulsion phase in the aggregative bed can be deter-
mined by sedimentation, extrapolating the sedimentation curve to time
zero (M41, R9). The results of Morooka et al. (M41) show that, at higher
gas velocities, the expansion ratio of the emulsion phase (L, - L,)/L, is
independent of gas velocity and bed diameter, as illustrated in Fig. 5; the
sedimentation velocity corrected by the deaeration velocity equals the
superficial gas velocity in the same manner as in a liquid-solid fluidized
bed (R7).
The expansion ratios in the emulsion phase measured by Oltrogge ( 0 5 )
and Morooka et al. (M41) at Uf >> Umfare correlated in Fig. 6 with the
parameter N,,. According to Miyauchi and Yamada (M32), the mean par-
ticle diameter in the parameter N,, is
4 = (EAn d i / x A n d j ) ” 2 (2-1)

The ratio ( L , - L,)/L, is about 0.1 when typical FCC particles are
fluidized.

loo .-
-
I I I I I 1 1 1 1 I 1 I I 1 1 1 1 J

--
Morooka rtrb(M41)
key particles
A broad PVC

-
m

4
P 8 - particles -
5 6-
-
0 -
4
* 4-
4
b
- -
Oltroggc ( 0 5 ) *\ -
-A polyprop lene
o broad FEC
o narrow F C C

FIG.6 . Expansion ratio of emulsion phase (after M41).

290 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

B. FORMATION,
SPLITTING,
A N D COALESCENCE
OF BUBBLES

The frequency of gas bubbles which are formed steadily through an

orifice in a fluidized bed has been studied by Harrison and Leung (H6).
Their results show that the mechanism of chainlike bubble formation is
the same as that in an inviscid liquid. If all of the excess gas above the
minimum fluidization velocity passes through in the form of gas bubbles,
the diameter of a sphere having the same volume as the originated bubble
is represented by two equations (in units of cdsec). Van Krevelen and
Hoftijzer (V6) found that

where AT is the bed’s cross-sectional area, and nd is the number of

orifices. Harrison and Leung (H6) have provided a similar relation:
dbi = 0.347[&( uf - umf)/nd]2’5 (2-3)
where nd is the number of holes in the perforated plate used as gas dis-
tributor, The initial bubble diameter measured by Miwa et al. (M17),
Hiraki and Kunii (H13), and Chiba et al. (C4) shows good agreement with
Eq. (2-2). The initial equivalent diameter of bubbles formed through a
porous plate is given by Miwa et al. (M17):

dhi = O.O0376(Uf - (2-4)

where d b i and (Vf - Umf)are in units of cm and c d s e c , respectively.
Comparisons between the experimental results and the equations are
shown in Fig. 7.
Intensive studies have been carried out on the ascending bubble diame-
ters in free fluidized beds (C5, K27, R14, R16, W9). Various correlations
for estimating bubble diameters have appeared (M36, R13, W9). How-
ever, the particles utilized in these experiments belong to group B of
Geldart’s classification. For this type of particle, bubble diameters are
expressed as a function of bed diameter, of distance of the bubble
above the distributor, of initial bubble diameter, and of physical proper-
ties of the fluidized particles. Mori and Wen (M36) emphasized the former
three factors and proposed the equation:

where dbM is a maximum bubble diameter, estimated as

 FLUIDIZED CATALYST BEDS 29 1

Hiraki (H13) 4.55

Miwa(M17)
Chiba(C4) 0.5

0.1
1 2 4 6 8 10 20 40

FIG. Comparison of the observed initial bubble diameter with calw-ted diameter
(after M36).

The ranges of data from which their correlation was obtained are
0.5 5 Urn,I20 cm/sec, Uf - Urn,I48 cm/sec
0.006 5 d,, 5 0.045 cm, DT 5 130 cm
The correlation of Werther (W9) gives nearly the same results as Eq. (2-5).
Both correlations predict growth of bubble diameter and decrease of bub-
ble frequency due to coalescence. However, experimental data in fluid
beds have been excluded from the above correlations.
In contrast to bubbles in teeter beds, naturally occurring bubbles in fluid
beds appear to split and recoalesce very frequently, and occasionally a
bubble frequency will increase at a height further from the distributor.
Rowe (in D5) and Toei et al. (T17) studied the mechanism of bubble
splitting. Toei et al. (T17) found that small downward cusps were gener-
ated at the two-dimensional bubble surface near the stagnation point,
which grew and thereby caused bubble deformation and, frequently, bub-
ble splitting. Figure 8 shows the effect of particle diameter on the average
total frequencies of bubble disturbance and splitting. The results of Toei et
al. (T17) indicate that the frequency of disturbances or splittings de-
creases with increasing particle diameter, and is independent of bubble
292 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

x,l
and splittings (T17).
-0.21 0.28 0.36
~ m m ~

FIG.8. Effect of the particle diameter on average total frequencies of disturbancesfdsl

diameter. Therefore bubbles which form in a fluid bed with fine particles
have more tendency to split than to coalesce, as discussed in Section V,C.
Morooka et al. (M43) measured properties of bubbles in fluid beds with
FCC particles. The fluid beds were 7.9 and 19.5 cm i.d., and L, was
100-200 cm. Bubble signals were detected by a hot-wire probe, originally
designed by Yamazaki and Miyauchi (Y3-3, which consists of a pair of
parallel tungsten wires 10 p m in diameter with a span of about 3 mm
and a separation of 12 mm. The electrical resistance of the hot wires
changes with the density difference between the bubble and emulsion
phase. The disturbance induced by this probe is presumably not greater
than that of the miniature capacitance probes of Werther (W6) and
Burgess and Calderbank (B17). Figure 9 shows the velocity of bubbles
ascending along the central axis of fluid beds. The effects of bubble diame-
ters, gas distributors, and axial position of the probe on the ascending
bubble velocity are rather small.
The measured bubble velocity is much greater than predicted from
equations derived by Davidson and Harrison (D3) and Taylor (TS),
respectively:
ub = uf - + 0.711(gdb)1’2 (2-7)
and
~b = 0.71 l(gd,)”’ (2-8)
 FLUIDIZED CATALYST BEDS 293

I I I I I I I I I 1 1

200- L,=2 rn

I I I I
3 4 6 8 10 20 40
Uf (cmlsec)

FIG.9. Velocity of bubbles ascending along the center of a fluid bed (M43).

The reason, pointed out by Turner (T28), is that the bubble velocity in a
fluidized bed without circulation will obey Eq. (2-8), but that bubbles in
fluid beds ascend in an intense circulatory flow of the emulsion phase.
The arithmetic mean of bubble signals measured by Morooka et al.
(M43) is shown in Figs. 10 and 11. Fig. 10 illustrates the effect of superfi-
cial gas velocity on the mean length of bubble signals. Some slugs appear
in the range of U,< 30 cm/sec at the upper part of a 7.9-cm-i.d. bed.
Figure 11 shows the relationship between the mean length of bubble sig-
nals and axial probe positions for operations at higher gas velocity. If an
initial bubble diameter is larger than the equilibrium diameter, the bubble
has a tendency to split.
Actual bubble diameters can be calculated from the mean bubble sig-
nals according to the relationship given by Anderson (A6) and Ueyama
and Miyauchi (Ul). Assuming (1) the bubbles are spherical, (2) an ascend-
ing velocity is independent of a bubble diameter, and (3) the wall effect is
negligible, the following equation is obtained:

where M ( l ) is a distribution function of bubble signals measured with a

point-shaped probe, and F(dJ is a distribution function of bubbles actually
294 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

4 --e-
0

i0 FCC calalvsl
parliclrs
A

'0 10 20 30 40 50 60
Uf (cmlsecl

FIG. 10. Mean length of bubble signals in fluid beds (M43); perf. = perforated.

existing in a fluid bed. Integration of Eq. (2-9) yields:

= (m/n) lom
P 2 M ( l )dl
(2-10)
joa
dV'(dJ d(dJ jox
F M ( 1 ) dl
where rn and n are positive integers. The actual volume-surface mean of
bubbles is calculated with Eq. (2-10) as follows:

(2- 11)

Under the conditions of Morooka et al., d32 is estimated as 4.5-6.5 cm.

Matsen (MlO), in studies of bubble diameters in a 61-cm-i.d. fluid bed,
recognized that small difference in particle size distribution could produce
significant changes in the character of fluidization. With Coke-5 particles
(ap at 50 cumulative wt.% = 70 pm, pp = 2.1 g/cm3), his probe measure-
ments showed initial formation of 25-cm-diameter bubbles, which broke
into smaller stable bubbles of -6-13 cm. The maximum stable bubble size
in a 13.8-crn4.d. fluid bed with 26-pm particles was estimated as 2.5 cm.
Massimilla (M3) also reported that smooth fluidization was obtained with
fine particles.
Werther (W9) measured bubble diameters in a 100-cm-i.d. bed, using
quartz sand (Umf= 1.35 cm/sec) and spent FCC particles (Umf= 0.20
cm/sec). In the quartz sand bed a rapid increase was observed in average
bubble size with height above the distributor, whereas in the FCC bed
 FLUIDIZED CATALYST BEDS 295

.L V
m-

1
0 100 200
Lm (cm)
FIG. 11. Mean length of bubble signals in fluid beds operated under relatively high gas
velocity (M43); perf. = perforated.

bubble growth decreased with increasing height. When the FCC particles
were fluidized at Uf = 10 cm/sec, the volume-surface mean of bubble
diameter approached the equivalent value (3.5 cm) at 40-50 cm above the
perforated plate. In a similar study, Ikeda (15) calculated' the effective
bubble diameter to be about 8 cm in a 360-cm-i.d. baffled fluid bed (equiva-
lent bed diameter = 70 cm, Uf = 45 cm/sec, U,, = 0.6 cm/sec).
Tsusui (T27) studied the effect of particle characteristics in a fluid bed of
6-cm i.d., using a hot-wire probe. The difference of flow characteristics
was most evident at high velocity of bubbles and high diameters, as shown
in Fig. 12. Fresh FCC particles showed the smoothest fluidization, ex-
pected for the high ascending velocity. Removing fine particles (d, I44
pm) from original FCC particles gave a low ascending velocity and larger
bubbles. The heavier MS catalyst showed very low ascending velocity
and a typical slugging flow. Silica microspheres, which are very light ( p ,
= 0.3 g/cm3) and average around 70 p m in diameter with a very narrow
particle size distribution, showed a fluidity similar to that of the FCC
particles without fines. The actual bubble size distributions, F(db), are
shown in Fig. 13.
All these results indicate that bubble diameters in fluid beds are strongly
influenced by the phenomenon of bubble splitting. Kehoe and Davidson
(K8, K9) observed slugging in a fluid bed with 62-pm catalyst particles,
although their results are presumably due to the narrow size distribution
of particles. An experiment conducted by Lanneau (L2) showed the pres-
ence of large bubbles in a 7.6-cm4.d. fluid bed with typical fine micro-
spherical alumina catalyst; however, his capacitance tip measured 4.8 x
6.4 x 19 mm. Since many bubbles present were probably smaller than the
296 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

--
9 -
50-
---- __-------
Eq.(2-8)
&--
- 2 -
,
- ,, r3,uz";beT' --
-
I
/
(Stewart eta/.,S21) -
/
I 1 I I
0' I

FIG.12. Ascending velocity of bubbles under the conditions of U,= 20 c d s e c , D, = 6

cm, and L, = 40 cm (T27). ( I ) FCC particles, (2) MS catalyst, (3) FCC catalyst without fines,
(4) silica balloons, ( 5 ) heavy fluid bed catalyst.

1.4 I I I I I I I I I I

1.2

1.0
n

\
5
fi 0.8

--
8 0.6
LL

0.4

0.2

0 1 2 3 4 5 6
db [em]

FIG. 13. Frequency distribution functions of actual bubble size (T27). Experimental
conditions are the same as in Fig. 12.
 F L U I D I Z E D CATALYST BEDS 297

probe tip, as shown in Fig. 13, Lanneau’s measurement seems to show the
maximum size present under given conditions.

C. OPERATION
O F F L U I DBEDS

Operation of the bed is closely related to the properties of particles

discussed in Section II,A. Fluid beds are usually operated in the
turbulent-flow region to obtain good contact and sufficient throughput of
reactant gas. In general, the superficial gas velocity Uf in fluid beds is
considerably larger than the minimum fluidizing velocity Umf;the ratio of
Uf/Umfis usually more than 100. Also Uf/u, is more than unity, where u, is
the terminal velocity of an average-size particle. Bubble size does not in-
crease to an unlimited extent, but approaches a certain limiting value
determined by the turbulence intensity; when large bubbles are injected
into a turbulent bed by a single nozzle, the bubble size reduces to an
asymptotic value as bubbles ascend (M43).
In commercial-scale fluid beds, a superficial gas velocity usually more
than 30 cm/sec is preferred to provide sufficient turbulence, fluidity, and
throughput of reactant. Increase of the contact efficiency by increasing the
gas velocity has been observed by Lewis et al. (L12) and Gilliland and
Knudsen (G7).
However, high gas velocity, may lead to problems in bed operation. It
increases the entrainment loss of fluidizing catalyst particles. It also may
give rise to excessive reaction in the particle-disengaging space, which
sometimes will lead to reactor instability or to decreased selectivity (as
discussed in later sections). Attrition or erosion of the reactor is more
likely at higher gas velocity. Thus, there is an optimum gas velocity for
fluid beds, which for most catalysts is in the range of 20-80 cm/sec,
usually 40-60 cm/sec.
Operation of a fluid bed depends on both height and diameter. Slugging
properties have been studied extensively by Davidson and co-workers
(H17, K8, K9, S21). Generally the ratio Lf/DT is much larger than unity
for small-scale reactors (C2). For solids particles in a teeter bed, this ratio
will lie well into the slugging region. However, the fluidity in fluid bed is
quite different from that in a teeter bed, as explained in Section II,A.

D. PROPERTIES FLUID
OF BULKFLOWINSIDE BEDS

1. Average Holdup of Gas Bubbles

Leva (L7) proposed a correlating equation for expansion in aggregative
fluidized beds, but poor agreement between calculated and observed
298 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

voidage was found for fluidization of FCC particles with atmospheric air
(Z5). Bed expansion can be expressed in terms of average bubble holdup,
which also enters the calculation of contact efficiency (Section VII). Aver-
age bubble holdup is defined by
Zb = (LE - L€.)/L, (2-12)
where L, is the equivalent height of the emulsion phase measured by the
flow interruption method (R9). Experimental data by Morookaet al. (M40)
and van Swaay and Zuiderweg (V8) show (in Fig. 14) that the effect of bed
diameter on the bubble holdup is very small. At higher gas velocities,
experimental values of bubble holdup in fluid beds agree with those in
bubble columns. At U, = 50 cm/sec, z b does not exceed 0.5, and a
fluidized-bed height is at most twice that of a settled bed. In contrast with
fluid beds, the expansion of teeter beds is a function of both particle
properties and bed diameters (H17,K8, K9).
The data of Wilhelm and Kwauk (W13) plotted in Fig. 14 reveal higher
values of z b because of severe slugging in a small bed. Figure 15, given by
de Groot (D7), shows that bubble holdups in beds with narrow-range silica
are much smaller than with broad-range silica. Thus, in larger fluidized
beds of silica with a narrow size distribution, very large bubbles form with
high ascending velocities. The effect of temperature and pressure of the
I 1 I I [ I l l I I I 1

0.5 -
glass beads

c.

c. Bubble column
4
v Kato cta/.(K6)
I1
v Dr>20 cm

0.1
- 4
-
9-
0 6.6crn -
8- -
A 7.9 -
7-
0 12
6- v 19 -
0.05 - 0 /' -
1
do, I 1 1 1 1 1
I I I I &
2 4 6 8 10 20 40 60
Ur (crn/sec)
FIG.14. Holdup of gas bubbles in fluid bed (after M40).
 FLUIDIZED CATALYST BEDS 299

50.':
- 1
T
I
' ' '' " ' I '
I I

1 1

10-

1 ' 1'
-
O bll 012 ' Ol, '0.'6';.;;
D, (m)

FIG. 15. Bed expansion as a function of bed diameter at LJ, = 20 c d s e c . (D7).

fluidizing gas on bed expansion has been studied by de Vries et al. (D8).
As Fig. 16 shows, the influence of the gas pressure is not large for wide-
range microspherical silica. These authors mention that emulsion-phase
fluidity is quite good at high temperature.

2. Lateral Distribution of Bubble Holdup

El Halwagi and Gomezplata (E8), Larroux et al. (L3) and Nishinaka et
al. (N7) studied lateral distributions of bubbles in fluid beds. Their results
show that bubbles in a dense phase are likely to ascend in the central part
of a bed.
Nishinaka et al. (N7) have determined the lateral distribution of gas
bubbles in fluid beds with diameters of 6.6, 12, and 19 cm, as plotted in

8 T = 300'C

c
'I. T= 20.C
40 r

30

1 .o
I "

0 0.5 1.5
Pabsr bar

FIG.16. Fluid bed expansion for SCP catalyst as a function of pressure (D8). Fraction
t44 p n 20%, U,= 18 c d s e c , and DT= IOcm.
J
300 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

FIG. 17. Lateral holdup distributionof gas bubbles in fluid bed with 12-cm inner diameter
(NV.

Fig. 17. The distribution is nearly independent of axial position. The hold-
up of gas bubbles at radial distancer from the center axis of the bed is
expressed as
(EbO - Eb)/EbO = [(EbO - E b ~ ) / E b l l l ( T / R ) ~ (2-13)
where ebOis the holdup of gas bubbles at the center axis of the bed, and Ebw
is the holdup of gas bubbles extrapolated to the bed wall and may be taken
as approximately zero. Parameter n in Eq. (2-13) is variable with gas
velocity and bed diameter (Figs. 18 and 19). Their experiments were

Uf Ccmhccl

FIG.18. Effect of superficial gas velocity on n (Nn.

 F L U l D l Z E D CATALYST BEDS 301

FIG. 19. Effect of superficial gas velocity on ( E -~ cbw)/cb0(N7).

carried out with FCC particles fluidized in beds of Lf/DT1 . 2 . Different

distribution behavior has been reported for teeter beds with Uf/Umf5 LO
and Lf/DT5 1 (W7,8).
The lateral distributions of bubble holdup in fluid beds are very similar
to those of bubble columns for gas-liquid systems. Experimental results
by Akehata et al. (A2), Pozin et al. (P6), Ivanov and Bykov (I12),
Yamagoshi (Y2), Miyauchi and Shyu (M31), Hills (H9), Kato et al. (K7)
and Ueyama and Miyauchi (U3) are reasonably well expressed by Eq.
(2-13). When a gas phase is distributed with a perforated plate or a nozzle
sparger, the parameters in Eq. (2-13) for the bubble columns are in the
range o f n = 1.7-2.5.

3. Flow Patterns in Fluid Beds

The lateral distribution of gas bubbles induces bulk recirculation of the
emulsion phase. For teeter beds, the flow pattern of solid particles has
been studied by Werther (W7, WS), Burgess and Calderbank (B17), Oki
and Shirai (03) and Whiteheadet al. (W12), in experiments carried out with
alumina particles, quartz sand, and petroleum coke under conditions of d,
2 . 8 3 p m and (U,- Umf)/Um,I14. The mode of bulk circulation was
centrally descending and peripherally ascending for &/L& 5 1, while the
ascending flow moved toward the center with increasing Lf/D,. Werther
(W8) showed that this circulation flow is caused entirely by bubbles which
carry solid particles upward in their wakes.
In contrast to the teeter bed, less work has been carried out on the bulk
flow pattern in fluid beds (L1 1, M29). Measurements at the relatively high
gas velocities of practical interest ( U , > 10 cm/sec, U,/U,, >> 1) show
that the rate of circulation exceeds that of solid particles conveyed by the
bubble wake, and results from the buoyant force induced between the
centrally ascending bubble-rich phase and the peripherally descending
emulsion phase of low bubble content (see Fig. 2). The behavior greatly
resembles that in bubble columns, as will be discussed quantitatively in a
later section (H9, K15, M31, P3, M,T23, U3, Y2, Y22, and Kato and
Morooka, unpublished data). For a bubble column, Shyu and Miyauchi
302 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

(S13) have proposed a semiempirical equation for the circulating liquid

velocity at the center axis:
U ~ O= 6.8Uf'5D$2s (2-14)
where uzoand U fare in cm/sec and D T is in cm.
In fluid beds, the ascending velocity of bubbles at the center axis has
been measured by Morooka et al. (M43) and Tsutsui and Miyauchi (T27)
with the assumption that the slip velocity of bubbles relative to the adja-
cent emulsion equals the ascending velocity of a single bubble; the circula-
tion velocity of the emulsion phase along the axis is calculated by
UeO = UbO - 0.71 l(gdb)"2 (2-15)
When the fluid bed is operated under relatively high gas velocity, the
mean bubble size changes only slightly with axial bed height and depends
essentially on the flow intensity. Under these flow conditions, ueoof fluid
beds and uloof bubble columns are seen to be nearly equal, as shown in
Fig. 20. This indicates that the mechanism of circulation in the fluid bed
agrees well with that in the bubble column. Studies on solid circulation in
a fluidized bed with a draft tube have been demonstrated by Yang and
Keairns (Y7, Y8) and LaNauze (Ll); their models also are based on a
density-driving force between annulus and draft tube.

4. Apparent Emulsion Viscosity

Fluidized beds have many liquidlike properties. An apparent viscosity
is often assigned analogous to Newtonian fluids, measured by means of

FIG.20. Circulation velocity ascending at the center of fluid bed and bubble column.
Data of bubble columns are measured by: 8,Pavlov (P3), DT = 17.2 cm; €B, Pozin et al.
(P6), 30.0 cm; A,Yamakoshi (Y2), 25.0 cm; curve, Yoshitome and Shirai (Y22), 15.0 cm;
0 , Miyauchi and Shyu (M31). 10.0 cm; 0 , Hills (H9), 13.8 cm; V , Ueyama and Miyauchi
(U3), 60.0 cm; 4, Koide et al. (KIS), 504.0 cm; @, Kato and Morooka (unpublished data),
12.0 cm.
 FLUIDIZED CATALYST BEDS 303

Stormer viscometers, concentric cylinder viscometers, rotating spindles,

falling spheres, etc. Because these viscometers expend part of their
energy in accelerating the particles, this produces change in their orienta-
tion, and because voidage in the bed is affected by the immersed objects,
the data on apparent viscosity of fluidized beds have to be carefully exam-
ined.
Most previous work has been reviewed by Grace (G13) and Schiigerlet
al. (in D5).Botterill (B12) and Hetzler and Williams (H8) correlated the
apparent viscosity of liquid- and gas-fluidized systems, applying a free-
volume theory which may be used successfully for glass-forming
(polymeric) liquids. Saxton et al. (S3) proposed another approach to a
free-volume theory. They compared theoretical expectations with the ex-
perimental data obtained in liquid-fluidized systems. Their extension to
gas-fluidized systems (in cgs units) became, for sand,

(2-16)

and for polyvinyl acetate powder,

(2-17)

[ p b (g/cm sec), d , (cm), and Uf (cm/sec)]. The former expression coin-

cides roughly with the data of Schiigerl er al. (in D5),but the latter gives
values smaller than the data of Furukawa and Ohmae (F20) by a factor of
2. This implies an additional dependence of p b on d,.
Grace (G13) studied the relationship between the included angle of
spherical-cap bubbles and the apparent viscosity, and found that apparent
viscosity in fluidized beds was between 4 and 13 poise. Hagyard and
Sacerdote (H20) measured the bed viscosity by means of the decay of
oscillation of a vertical cylinder immersed in the bed. The kinematic vis-
cosity decreased from about 5 cm2/sec to 1-0.5 cm2/sec.
The general tendencies of the experimental results are as follows: The
apparent viscosity in an unfluidized bed is very high (perhaps infinite) but
drops rapidly with increasing gas velocity. When particles are fully
fluidized, the bed viscosity becomes rather independent of gas velocity
and increases with increasing particle size and particle density; i.e., it
shows a gradual change from a displacement effect to a collisional momen-
tum transfer.
Furthermore studies on the effect of particle size distribution, by
304 M I Y A U C H I , F U R U S A K I , MOROOKA, A N D IKEDA

Mathesonet al. (M7), show that relatively small amounts of fines added to
coarse particles sharply decrease the bed viscosity. Based on this evi-
dence, Zenz (24) has calculated an optimum size distribution of fluidized
particles (as mentioned in Section II,A,I), The importance of the apparent
viscosity in relation to flow characteristics of the bed has been stressed by
Matheson et al. (M7), Rice and Wilhelm (R5),Finnerty et al. (F3), Grace
(G13), and others. Matheson et al. (M7) have found that the slugging
tendencies of a bed can be expressed in terms of the apparent viscosity.
Their results for FCC particles containing 20 wt.% of 46-pm-diameter fine
particles gave p b equal to 6 X g/cm . sec; this small bed viscosity
agreed with the visible high fluidity of the bed.
Meanwhile, Rice and Wilhelm (R5)have investigated the stability of
interfacial surface between the particle-free gas phase and the dense-bed
phase. According to their analysis, the lower bed interface (or upper
interface of a bubble) is inherently unstable, whereas the upper bed inter-
face (or lower interface of a bubble) is inherently stable. Thus an instabil-
ity initiated at the upper bubble surface may reduce bubble size; this
tendency is presumably stronger in a bed of lower viscosity. The apparent
viscosity which applies to the surface dynamics has been studied by Fin-
nerty et al. (F3), Takamura (T2) and Morooka and Kato (unpublished
data). They have measured wavelengths, frequencies, and attenuation
rates for waves generated at the bed surfaces. In the experiments of
Takamura (T2) and Morooka and Kato (unpublished data), rectangular
fluidized beds of 6 x 60 cm and 8 x 80 cm were utilized, respectively, and
bed heights were kept at about 30 cm. The attenuation rates of surface
waves were calibrated with glycerol-water solutions. Their results, shown
in Fig. 21, indicate that the apparent viscosity of FCC particles is the
smallest among particles employed in the experiments.

FCC particles
4- dp=60~m
$=l.Og/cm

I
-
E
2- ‘a,
&
Y
D
Y a

30:- \ Porous alumina particli

02 0.4 0-6 0.8 1 2 4 6 810

U, Ccmlstcl

FIG.21. Apparent viscosity of fluidized bed (T2 and Morooka and Kato, unpublished).
 FLUIDIZED CATALYST BEDS 305

E. AXIALDISTRIBUTION
TO BED STRUCTURE

1. Bed Density
With increasing gas velocity, the bed surface diffuses gradually, and a
dilute phase is formed on the dense bubbling bed owing to entrainment of
fluidizing particles. This axial distribution of bed density has been mea-
sured by several workers (B4, F2, L13, M25, M33). Fig. 22 shows the
axial distributions of bed density measured by Miyauchi et al. (M33) in
1969. Their experiment was performed in a 7.9-cm-i.d. fluid bed of FCC
particles having a mean diameter of 58 pm.
The influence of the amount of particles in the bed on the axial porosity
distribution has been reported by Bakker and Heertjes (B4) for glass
beads of 175-210-prn-diameter, fluidized in a 9.0-cm teeter bed. Their
results show that the quantity of axially suspended particles in the dilute
phase is increased noticeably by increasing the initial bed mass from 200
to 2000 g. In contrast with this behavior, the bed density distribution given
by Kaji (Kl) for the dilute phase shows less dependence on the initial bed
height Lq under constant Uf. He has measured the distribution for a 7.9-
cm-diameter bed of FCC catalyst, having a mean size of 58 pm, by vary-
ing L , at 32, 51, and 71 cm, and Uf for 27-58 cm/sec. The mean bed
density in a dense fluid bed can be calculated by the following equation
knowing the mean bubble holdup in the dense fluid bed given in Section

2. Transition Zone and Return Flow of Emulsion

A transition zone exists between the dense and dilute phase. In this
region, particle density reduces gradually to the low level of the dilute

9 0.3
n
I
E
20.2
FIG.22. Longitudinal bulk B
density distribution (M33). 2
0.1

0
50 100 150 200
Longitudinal position [cml
306 M I Y A U C H I , F U R U S A K I , MOROOKA, A N D I K E D A

phase. A circulating flow is formed in the transition zone, changing the

direction of flow. When there happens to be an upward flow of the emul-
sion, the ascending velocity of bubbles is presumably expressed by u, +
0.71 l(gdb)”2,where u, is the circulating velocity of emulsion. When the
circulation returns, the rising velocity of bubbles becomes slowed in the
upper part of the dense phase, which gives rise to greater holdup of the
bubble phase. This is probably a reason for the large mass-transfer
rate in the transition zone (described in Section VIII). Other causes of
such good contact in the transition zone may be coalescence and rupture
of bubbles at the end of the dense phase, the latter giving great distur-
bance in the dense phase and pushing particles into the dilute phase. The
return flow of the emulsion also occurs at the vicinity of the distributor,
and here the holdup of bubbles is reported to be larger than in other parts
of the dense phase (Basov et al., B6). Although the flow pattern of bubbles
is different in the transition zone, good contact in this zone is experimen-
tally justified as written in Section VIII.

3 . Properties of the Dilute Phase

The gas leaving the top of the dense phase carries entrained particles
with it. Zenz and Weil(Z6) studied the mechanism of particle entrainment
from nonslugging fluid beds, observing a transport disengaging height
(TDH), above which the rate of decrease in entrainment approaches zero.
The TDH, in most cases the design optimum for location of cyclones, is
dependent on bed diameter and superficial gas velocity. An empirical
correlation for estimating the TDH given by Zenz and Weil (26) largely
agrees with the results of Tanaka and Shinohara (T6). According to Zenz
and Weil (Z6), the entrainment rate can be replaced by the saturation rate
for a particle whose diameter equals the geometric mean (50 wt.%) of all
particles with terminal velocities less than the superficial gas velocity.
Measurements of entrainment from fluid beds have been carried out by
Zenz and Weil (Z6), Overcashier et al. (01l ) , Lewis et al. (L13), Fournol
et al. (F5), Hanada (H3), and Kono (K28). Figure 23 shows that their
results are strongly affected by particle size and size distribution, particle
density, and other factors. Calculations by the procedure of Zenz and Weil
(26) predict smaller values than the experimental data for mixed-size
particles, and larger values for uniform-size particles.
The entrainment rate and the density distribution below the TDH were
intensively studied by Lewiset al. (L13)with closely sized particles. Their
results on the entrainment rate were correlated by the following equation.

e/Uf = B exp[-(b/UJ2 - aHl (2-19)

 F L U I D I Z E D CATALYST BEDS 307

FIG.23. Entrainment rate from the bed with cracking catalyst particles.

where H is freeboard height, and a, 6, and B are parameters dependent on

various factors: the effect of particle size distribution on the parameters is
not known.
Elutriation-the selective removal of fines by entrainment-has been
studied by several workers (K25, L6, M23, 08, T7, W2). Most of this
work has been reviewed by Kunii and Levenspiel (K24) and Leva and
Wen (in D5).However, Merrick and Highley (M13) point out that early
correlations for elutriation rate constant are inaccurate for very fine parti-
cles, since they assume u t proportional to d: (rather than to d p )and hence
predict that the elutriation rate constant reduces to zero as the particle
size reduces to zero. Thus, care must be taken when such empirical equa-
tions are applied to a fluid bed containing fine particles.

F. EFFECTO F BED INTERNALS

The uniformity and contact efficiency of a fluid bed can be improved by
immersion of a surface within the bed (G15,15, V12). Ikeda has defined a
generalized equivalent diameter, considering that the commercial reactors
have not only vertical internals but also horizontal baffles. Figure 24 indi-
308 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

0 Modified lnternals
I I I I
2 4 6 8 10 12 14
Equivalent Diameter, in.

FIG.24. CO conversion in Hydrocol reaction for several reactor diameters (V12).

cates that the equivalent diameter of a large reactor gives the same con-
version as a small reactor of the same actual diameter. The proposed
equivalent diameter is
(Deq1-I = (DeV1-l + (DeH1-l (2-20)
where Devand DeHare the component equivalent diameters based on the
vertical surfaces and the horizontal surfaces, respectively. Ikeda carried
out an acrylonitrile synthesis by the Sohio process, in fluid beds with
diameters from 8 cm to 3.8 m. Reaction yields were compared under the
same reaction conditions and catalyst, and the effective bubble diameter
was calculated by modifying the bubbling bed model originally proposed
by Kunii and Levenspiel (K22, K23). The effective bubble diameter is
given by
(db)eff = 1-9(De~)''~ (2-21)
where (db)eff andDeq are both in cm.
Several investigations have been made on the flow characteristics in
multistaged fluid beds. Nishinaka et al. (N6, N8, N9) have measured the
average bubble holdup, the lateral distribution of bubble holdup, and the
longitudinal dispersion of solid particles in four- and eight-stage fluid beds
installed with various horizontal baffles. As shown in Fig. 25 the average
bubble holdup (except for beds baffled with tube plates) is correlated by
the equation of Nishinaka et af., (N8):
Eb = 0.08 ufo.73A-O.3NO.O5
r (2-22)
where A, is the free area of baffle in %, and N is a number of horizontal
plates; Uf is in cm/sec. The longitudinal-dispersion coefficient of solid
particles has been found to decrease with decreasing free area of baffles
(N6). The intermixing mass velocity of solid particles between adjacent
 FLUIDIZED CATALYST BEDS 309

FIG.25. Correlation of average gas bubble holdup in free and baffled fluid beds (N8).

stages is related to the longitudinal-dispersion coefficient of solid particles:

Intermixing mass velocity = p,Ui = E,,p,/AL (2-23)
where Ui is solid intermixing velocity based on particle density, and AL is
the length of one stage. This particle-intermixing velocity is correlated
in the same manner by Overcashier et al. (011). In Fig. 26 the solid
intermixing velocity is plotted against g z '/Ufzand is expressed by the fol-
lowing equations.
For slots and perforated plates
u; = llUf/gz' (2-24)
and for tube plates
u,, = 2.5U,2/gzt (2-25)
where z is the height of the dilute zone under eac.. baffle. Other work has
been carried out with regard to holdup of solid particles and axial distribu-
tion of bed density (B2, G16, N3), residence-time distribution of gas and
solid particles (011, R1) and entrainment from the bed ( 0 1 1, H3). Most of
these investigations have been reviewed by Harrison and Grace (in DS),
Botterill (B 12), and Verma (V10). An experimental study on mass-transfer
3 10 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

FIG.26. Relationship between intermixing velocity of solid particles and gz'/V: (N8).

rate between the bubble and the emulsion phase in baffled fluid beds (M42)
is mentioned in Section V1,B.

111. Turbulent-Flow Phenomena in Bubble Columns

and Fluidized Catalyst Beds

In the preceding section, the flow properties of fluidized catalyst beds

have been clarified mainly on the basis of experimental observations. In
the case of FCC catalyst, the apparent viscosity of the emulsion is usually
very small, and the emulsion shows good fluidity. Catalyst particles
once charged into a fluidized bed reactor are usually in service for several
months. Hence it is justifiable to prepare the particles very carefully, so
that the fluidized bed shows the best fluidization possible. This kind of
careful preparation is usually impractical in the case of single-pass parti-
cles such as coal, mineral ores, or grain.
 FLUIDIZED CATALYST BEDS 311

It is interesting to see how far modern technology for fluidized catalyst

beds has served to achieve good fluidization. Our criterion of good fluidi-
zation is a gas-fluidized bed of a low viscosity liquid (such as water),
where the low-viscosity liquid would set the lower limit to the fluidity of
the emulsion. Such a gas-fluidized liquid bed is the well-known bubble col-
umn, which has been studied extensively. Our objective is to understand
the behavior in the recirculation flow regime, since the superficial gas
velocity of practical interest is usually more than 30 cm/sec for fluidized
catalyst beds and for these conditions intense recirculation of the emul-
sion has been observed (note Fig. 2 and Section II,D,3).
Currently available data for the flow properties of the fluidized catalyst
bed are fragmentary, since the local motion of the emulsion phase is
difficult to measure experimentally. Therefore, it is useful to clarify the
flow properties of the bed in terms of our knowledge of bubble columns.
First, the fluid-dynamic properties of the bubble columns will be ex-
plained; then, the available data will be adapted to apply to fluid catalyst
beds. The reader will be able to picture an emulsion phase of carefully
prepared catalyst particles operating in intense turbulence for fluidized
beds under conditions of practical interest. This turbulence distinguishes
the flow properties of fluid catalyst beds from those of widely studied
teeter beds.

A. TWO-PHASE FLOWI N RECIRCULATION

BUBBLE
In a bubble column, as is widely known, the bubble swarm rises uni-
formly when superficial gas velocity is low (less than about 2 cm/sec) and
uniform-size bubbles are released at a bottom gas-distributor. This type of
uniform flow, called the bubble flow regime, has been utilized widely for
various gas-liquid contacting operations (K2 1, 09, S 1). The bubble-flow
regime seems to correspond to the well-known bubbling fluidized bed,
where, however, the bubbles grow to larger size due to coalescence as
they rise. Upon increasing the gas velocity, the uniform bubble flow be-
comes unstable, and intense recirculation of mixed gas and liquid sets in
due to buoyant forces induced between the centrally ascending bubble-
rich phase and the peripherally descending bubble-lean phase. This condi-
tion, named the recirculation flow regime (P3, Y2, Y21), resembles the
flow in fluid catalyst beds operated at practical gas velocities.

1. Simplified Theory of Recirculation for a Bubble Column

Theories of recirculation have been presented for the bubble column
with largel/D,, without continuous liquid feeding (H9, M3 1). The studies
312 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

use the same equation of motion, but different assumptions for the turbu-
lent kinematic viscosity.
The time-averaged flow pattern of liquid in a bubble column may be
visualized as shown in Fig. 27, where the column is in recirculation flow,
and the ratio LID, is large compared to unity. Under steady flow, the
basic equation of motion (H9, M31) is:
-(l/t') d(m)/dr = dP/dz + (1 - Eb)Plg (3-1)
where the left-hand side of the equation gives the force acting on a fluid
element by radial shear-stress difference, and the right-hand side gives the
forces acting by axial static-pressure gradient and by gravity. A similar
equation of motion has been applied by Wijffels and Rietema (W16) to
determine the flow properties of liquid-liquid spray columns.
The shear stress T is related to the mean axial velocity of liquid through
the sum of the molecular and turbulent kinematic viscosities:
T = -(VM + Vt)p,(dU/dr) (3-2)
where vMis usually negligible in comparison with v, except in the vicinity
of the column wall. The turbulent shear stress for the bubble column with
L / D , >> 1 is expressed by the following equation (U2):
-vt dU/dr = (1 - Eb)ULU: (3-3)
where U: and u: are the radial and axial velocity fluctuations of liquid,
respectively. Equation (3-3) is obtained by time-averaging the fluctuation
velocity components for two-phase flow.
The boundary conditions to solve the basic equations, Eqs. (3-1) and
(3-2), were given earlier (M31). As shown in Fig. 27, the liquid flow under-
goes upflow between a and b and downflow between b and c, both with a
developed turbulence, and also downflow in the laminar sublayer between
c and d near the wall; at the wall, r = R (or y = R - r = 0) and u = 0.
Consistent with the concept of the universal velocity profile (S4, SS), a
buffer layer should be present between the turbulent core and the laminar
sublayer. However, it is hard to distinguish the laminar-turbulent buffer
layer from the turbulent core, since bubbles from the turbulent core enter
irregularly into the buffer layer and agitate it. Hence, partly to simplify the
mathematical treatment and partly to satisfy a physically sound interpreta-
tion of this bubbling buffer layer, the thickness 6 used for the laminar
sublayer is taken somewhat greater than in single-phase flow and the buffer
layer is neglected.
According to the concept of the universal velocity distribution, the
velocity profile in the laminar sublayer is given by
uIv, = v.y/vM for v , y / v , 5 5 (3-4)
 FLUIDIZED CATALYST BEDS 313

r -0 r =R
y =o
FIG.27. Recirculation flow pattern in a gas bubble column.

where u. = ( / ~ ~ l / pisJthe
~ ’ friction
~ velocity. The velocity distribution in
the turbulent core is
u/u, = 5.75 log(u,y/u,) + 5.5 for u,y/uM L 70 (3-5)
By equating Eqs. (3-4) and (3-9, one obtains a valuey defined as the
thickness 6 of the laminar sublayer, for which:
u,S/uM = 11.63 (3-6)
The velocity us at y = 6 is given by combining Eqs. (3-4) and (3-6):
ug = 11.63~.= 1 1 . 6 3 ( ~ ~ w ~ / ~ J 1 ’ 2 (3-7)
According to the analysis of many experimental data (M31, US), the
thickness 6 is usually much smaller than the column radius R , so that us
essentially equals the peripheral velocity luwl of the turbulent core. As a
314 M I Y A U C H I , F U R U S A K I , MOROOKA, A N D I K E D A

consequence, the boundary condition at r = R for the turbulent core is

given by
Iu,I = ug = 1 1 , 6 3 ( ~ ~ , ~ / p ~ ) ~at
'* r = R (3-8)
Another boundary condition is obviously
duldr = 0 at r = 0 (3-9)

2. Solution of the Simplijied Theory

The solution of Eqs. (3-1) and (3-2) with the boundary conditions of
Eqs. (3-8) and (3-9) has been given by Miyauchi and Shyu (M31) for the
case without liquid feeding, and by Ueyama and Miyauchi (US) for the
case with continuous liquid feeding. The procedure here follows the latter
reference, with extension to a simplified treatment (M27).
Multiplying by 2nr dr and integrating radially from r = 0 to r = R , Eq.
(3-1) is transformed to
-dF/dz = , (1 - E b ) p l g
( ~ / R ) T-I- (3-10)
where <b is the mean gas holdup, defined by
(3-11)

By introducing Eqs. (3-2) and (3-10) into Eq. (3-1) and neglecting vM in
comparison with vtrthe following basic equation is obtained for the turbu-
lent core:
-(I/r)d[v,r(du/dr)l/dr = (2/R)(TW/Pl) - (Eb - Eb)g (3-12)
This equation is solved with respect to u under the simplifying assump-
tions (M31, US) that vt is constant in the turbulent flow region, and that
bubbls gas holdup Eb is distributed radially according to the relation:
Eb/<b = 2(1 - 4') (3-13)
where 4 = r / R . The physical significance of the assumption regarding vt
will be explained later; trial calculations to find a radially variable vt giving
the best fit to the experimental velocity profile of the bubble-liquid mixed
phases have shown that a constant vt is quite workable if taken as the
radial-cross-sectionally averaged mean value of a distributed vt. This as-
sumption simplifies the mathematical treatment, and also is supported by
Figs. 17 and 18 (refer to Section II,D,2). Becausen = 2 the gas holdup QbO
at the center axis is twice Eb. The empirically adjustable viscosity term vt
has different values when n is changed, but the flow properties of the
 FLUIDIZED CATALYST BEDS 315

mixed phases are insensitive to n. Therefore, n = 2 is a satisfactory

average for all the experimental data.
With these assumptions, Eq. (3-12) is solved directly, with use of Eqs.
(3-8) and (3-9):

The numerical value of vt will be given in the next section. In Eq. (3-14),
the first term of the right-hand side has been shown by Ueyama and
Miyauchi (US) to be essentially negligible for the recirculation flow re-
gime. As a consequence, one has the following simplified quantitative
relation:

@Y2 (3-15)

The undetermined term Iu,I is given by applying a mass balance for the
liquid phase:

loR27rr(l - EJU dr & 7rR2UL= 0 (3-16)

where the column is continuously supplied by a liquid feed with superficial

velocity k U , (+ indicates countercurrent gas-liquid flow and - concur-
rent) ( U , = 0 if the liquid is stationary).
By introducing Eq. (3-13) for Eb, Iu,,, is obtained from Eqs. (3-15) and
(3-16):

(3- 17)

where the 2 sign has the same meaning as that in Eq. (3-16). With this
Iuwl,the radial distribution of linear liquid velocity u is given by Eq. (3-15).
Another simplified liquid velocity profile is also obtained directly from
Eq. (3-15):
(u + lu,l)/(uo + lu,l) = ( 1 - @Y (3-18)
where uo is the linear liquid velocity at the column axis, + = 0, given from
Eqs. (3-15) and (3-17) as:

(3-19)

+*
The radial position r , (or = r J R ) where the liquid velocity u relative to
the column wall is zero (point b in Fig. 27) is obtained from Eqs. (3-15) and
316 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

(3-17) by setting u = 0. For the simple case of UL = 0 which is usually

utilized (batch operation for liquid), the above procedure gives the rela-
tion:
C#)* = (1 - eb)"*, where eb ((2 - 3Eb)/[6(1 - Eb)]}"' (3-20)
The shear stress at the wall T , is given from Eq. (3-8):
TWIPI = -(luwl/ll .63)' (3-21)
where (u,l is calculated adequately from Eq. (3-17) (or still more accurately
from US).
Another useful relation results from the mass balance for the gas phase
of the bubble column:
loR27rr(u + us)€,,dr = nR2UG (3-22)

where us is the slip velocity of a bubble relative to the liquid surrounding

it. According to experimental measurements of us in the recirculation flow
regime (U3, Y2), us is approximately constant radially, and hence a mean
constant value 6, may be used. With Eqs. (3-13) for q,, Eq. (3-18) for
u , and Eq. (3-19) for uo, Eq. (3-22) is integrated to

UG/<b f u~/(1
- Eb) = us + ( U o + 1~,1)/6(1 - iib) (3-23a)
With uo + Iu,I given by setting C#) = 0 in Eq. (3-15), Eq. (3-23a) becomes
uc/<b k uL/(l - Eb) = 6 s + (gDT/192~t)Eb/(l- Eb) (3-23b)
where the f sign has the same meaning as in Eq. (3-16). The second term
on the right-hand side of the equation is the apparent increase of slip
velocity due to recirculation flow. [Calculation shows that this term equals
U*/Eb,where U* is the net linear velocity of the recirculating liquid aver-
aged over the column cross section. '] This term is a natural consequence of
recirculation flow, and does not appear in the bubble-flow regime where
the following well-known relationship applies (N4,Y20):
uL/(l - E b ) = 6 s
uG/Zb h (3-24)
The mean gas holdup Eb in the recirculation flow regime is obtained at
UL = 0 by solving Eq. (3-23b) with respect to gb:

) "'1 (3-25)

U* =
I,"
(~rR*)-l 2nru dr.
 FLUIDIZED CATALYST BEDS 317

where
(Y = U,/ii, and y = gDz,/(192 u,U,)
The reason & is approximately proportional to U$z and decreases only
slightly with column scale-up in recirculation flow (cf. Fig. 14) is ex-
plained from Eq. (3-25) by including the behavior of Us and vt, which are
discussed in the next section.
The mean flow rate o f the recirculating liquid is useful when known. For
upflow, we take the superficial value:

0 = (7rR*)-' I:* 2rr(l - eb)u dr = 2 lo6* (1 - eb)U+ d+ (3-26)

With integration limits in Eq. (3-26) from r . to R with U, = 0 (no continu-
ous liquid feed), the mean superficial downflow velocity results. By using
Eq. (3-13) for Eb and Eqs. (3-15) and (3-17) for u, with U, = 0, Eq. (3-26),
when integrated, is
= ( g o $ / 192~t)' &e;(4 - 3Cbeb) (3-27)
with eb as given in Eq. (3-20). Once 0 is known, one can easily obtain the
interstitial mean velocities of upflow and downflow.

B. COMPARISON WITH EXPERIMENTS

OF THEORY ON BUBBLE
COLUMNS
Experimental facts which support the recirculation flow and its turbu-
lence properties are given here, mainly in relation to the simplified theory
developed in the preceding section.

1 . Interstitial Velocity Profile of Recirculating Liquid

The available data were correlated by Ueyama and Miyauchi (US), who
found that the normalized velocity profile Eq. (3-18) represents the data
reasonably well. Figure 28 illustrates the experimentally measured veloc-
ity profile of water containing bubbles from a single-nozzle gas-distributor
(Y2). The bubble column was 25 cm in diameter operating in the recircula-
tion flow regime at a superficial gas velocity U Gof 5.2 cm/sec. Estimated
values o f u o= 45 cm/sec and Iu,I = 26 cm/sec, amved at by trial, give the
best fit of the normalized profile data, and Eq. (3-18) provides a reasonable
approximation.
The radial distance 4" at which the mean liquid velocity is zero, given
by Eq. (3-20), is illustrated in Fig. 29 as a function of Eb, where Eb is taken
from the following correlation:
Gb = O.O54U&" (3-28)
318 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

$ = r/R
0 0.1 02 0.3 0.4 0.5 Ok-I

profile

0.1

'0 0.l 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
fi r/R [-I
FIG. 28. Interstitial liquid velocity profile and normalized velocity profile [data by
Yamagoshi (Y2)],D, = 25 cm, I!/= , 5.2 cd s ec, air-water system at room conditions.
Adjusted u,, and I u, I are 45 and 26 c d s e c , respectively.

with V Gin centimeters per second. This equation is a good approximation

for Eb of a bubble column operated in the recirculation flow regime at V , =
0 and is equipped with a single-nozzle gas sparger. The columns tested are
in the range of& = 10-60 cm and aspect ratioL,/DT z 2 (H11, U3).
In Fig. 28, the radial position Cp* at u = 0 is 0.66 from the measured
profile and 0.63 from the normalized profile, where the term (u + lu,I)/(uo
+ lu,l) = 26/(26 + 45) = 0.366. These two values of Cp. are shown in Fig.
29 with some additional bubble-column data (U3) and some fluidized-bed
data (M29). Equation (3-20) seems applicable to the bubble-column data.
When a perforated plate is used as a gas sparger in a bubble column, the
mean gas holdup Eb is influenced by the size and arrangement of holes as
well as column diameter as a result of change in both of bubble size and
the flow pattern of the bubble swarm. These effects have been reviewed
and correlated by Kato and Nishiwaki (K6) for air-water systems where a
sparger is perforated uniformly. When the hole diameter 6 is smaller than
1.0-1.4 mm and the gas velocity U GiS low, Eb for a given diameter column
 FLUIDIZED CATALYST BEDS 3 19

FCC-Fluidized bed

0
Mean gas holdup 5 [-I
FIG.29. Radial position 4%as a function of ib.

changes with 6, U,,and the number of holes. However, &, tends to be

variable only with U G ,for U Gabove 10-15 cm/sec, where it approaches
the values observed for a column with 6 2 2 mm.
Essentially the same performance is observed for a qolumn equipped
with a porous sintered metal or glass plate as a gas distributor.
When the hole size 6 is equal to or greater than about 2 mm, Eb increases
in proportion to U, in the low-gas-flow range ( U , 5 5-6 cm/sec) for
different diameters of columns. As UG increases beyond this range, Eb
increases approximately in proportion to (UG)1/2, decreasing slightly at
constant gas velocity as the column diameter DT increases. Here, the
extent of decrease seems to depend on the type of sparger, which depends
slightly on D, for a perforated plate, and is essentially independent of DT
for a single-nozzle sparger.

2. Turbulent Kinematic Viscosity

This term may be determined in various ways from experimental data.
A frequently used method for single-phase flow (S4) is to differentiate an
experimental velocity profile graphically, and to obtain the kinematic vis-
cosity ut from the velocity gradient with basic relations [e.g., Eqs. (3-1)
and (3-2)]. The currently available data for velocity profile do not appear
accurate enough to yield ut as a function of radial position. As an alterna-
tive, Hills (H9) has assumed a relation between the local vt and the radial
position:
Vt a eb(1 - eb) cc (3-29)
320 M I Y A U C H I , F U R U S A K I , MOROOKA, A N D I K E D A

The last proportionality comes from his experimental test of the relation.
Along the column axis, u equals uo and is approximately proportional to
(U&)1’4, as is seen in Fig. 30. Hence, vt a Dq’4/V24along the column
axis.
The constant effective vt assumed in the present simplified treatment is
obtained by knowing uo and IU,~. Along the column axis ( 4 = 0), Eq.
(3-15) gives
u0 + IuwI = gPT$/(32vt) (3-30)

With uo and Iu,l obtained by trial and <b from Eq. (3-28), vt is given by Eq.
(3-30). In the case of Fig. 28, Eb is calculated to be 0.123 (the measured
value is 0.127) and vt = 33.2 cm2/sec.
In the same manner, vt has been calculated by Ueyama and Miyauchi
(US)from a group of available experimental velocity profiles. Fig. 31
shows the recalculated vt as a function of DT, matched empirically by the
power function:
vt = 0.1280$’0 (3-3 1)

where vt is in cm2/sec and DT in cm. Earlier work of Miyauchi and Shyu

(M31) also gives Eq. (3-31) with a different coefficient. Recent experience
with a plant-size bubble column (DT= 550 cm) suggests that the Y, de-
pendence onD, may be extrapolated to larger-size columns for estimating
the flow properties (K15).
The dependence of vt on UGis (U5)approximately proportional to U&/6,
but not always consistently. This dependence is confirmed when vt is
calculated theoretically from uo (in the next section). Figure 32 shows a

150
100
’ J 7
$5;

- 3
J
2

1°1 2 3 4 5 7 10 2 3 450 7 100

Superf iciol gas velocity 9, [crn/sec]
FIG. 30. Dependence of liquid velocity at the center axis, uo, on the superficial gas
velocity CJG and the column diameter DT;v,is taken from Fig. 33 and C,, from Eq. (3-28) for
the calculated curves. Data are taken from Hills (H9).
 FLUIDIZED CATALYST BEDS 32 1

FIG.3 1. Turbulent kinematic viscosity vt as a function of column diameter DTfor bubble

columns; ut increases rapidly with DT,but is little influenced by U,.Note that UG = 1.9-93
cdsec.

correlation based on this dependence. The line B-B is for the bubble
columns and is expressed (in cm-sec) by
v1 = 0.160Lp,'6D3~2 (3-32)
The correlation shown in Figs, 32 and 33 as curve A-A is the turbulent
kinematic viscosity of the emulsion phase of fluidized cracking catalyst
beds estimated by an indirect method explained in Section IV,C in rela-
tion to axial dispersion of the emulsion. The correlation is given in cm-sec
units by
vl = 0.079U&'6Dk70 (3-3 la)
The exponent 1.70 on DTis different from 1S O of Eq. 3-32, but Eq. 3-3 la
gives a better correlation for axial dispersion of the emulsion.
We have seen that vt can be obtained from Iu,I or uo. In the case of Fig.
28, ul = 28.1 cm*/sec from Eq. (3-17), and vl = 36.1 cm*/sec from Eq.
(3-19). Some data are available for uo (H9, P3, p6) from which vt are
calculated and shown in Fig. 33. In general, vt as calculated from uo tends
to give a larger value than from uo + IuwI,perhaps due to lack of suffi-
 0, Fml
FIG. 32. Turbulent kinematic viscosity u, as a function of DT and U,. Line A -A is for

D, [cml
FIG.33. Turbulent kinematic viscosity v, as determined by different methods: the lines
A-A and B-B are the same as those given in Fig. 32.
 FLUIDIZED CATALYST BEDS 323

ciently accurate measurements. Once a mean vl is obtained, Eq. (3-19)

accounts well for the performance of uo, as illustrated in Fig. 30 for the
recirculation-flow regime with UG 2 3.8 cm/sec. Figures 28-33 show
definitely that the,flow field in the bubble columns is intensely turbulent in
the recirculation-flow regime, since vl is much greater than vM.

3. Mean Slip Velocity of Bubbles Relative to Liquid

Equation (3-23a) gives Us from experimental knowledge of uo + Iu,I and
Gb. In the case of Fig. 28, uo + Iu,( = 71 cm/sec and Eb = 0.123, so that ii,
= 28.8 cm/sec at U, = 5.2 cm/sec withD, = 25 cm. Figure 34 (from US)
summarizes Us so obtained as a function of U,. New data are added for a
commercial-scale column (K15,DT = 550 cm). The mean slip velocity ii,
remains essentially unchanged at 50 cm/sec for U , 2 . 2 0 cm/sec, but
drops gradually as UGdecreases below this value. Also the column diame-
ter has-essentially no effect, for DT 2 10 cm. The intense turbulent field
induced in the column seems to keep a steady bubble size, by dynamic
balance between colaescence and redispersion of bubbles.
Experimental Us values determined as the difference between the local
mean velocities of bubbles and liquid are also shown in Fig, 34 as “open
keys.” For these bubbles, the volume-surface mean diameters d32 have
been measured experimentally. The free-rise velocity of the bubbles (Tl,
V6) is
Us = (gd32/2)1’2 (3-33)
The ii, thus obtained, shown in the figure as half-open circles (U3), is in

- 10’

<X ?
u E
u 5
4
3
I3
2

10

Superficial gas velocity 9,[cm/sec)

FIG. 34. Dependence of average slip velocity ic, on the superficial gas velocity UG;
column diameter DT has little effect on 4. A porous plate distributor seems to give
smaller &.
324 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

reasonable agreement with other data, suggesting that bubbles in highly

turbulent flow have little interaction with one another. Data from Nicklin
(N4) show smaller ii,, perhaps due to use of a porous plate gas-distributor
instead of a sieve plate or a single nozzle.

4. Bubble Volume Fraction

So far ut and ii, have been given numerically. Figure 36 (upper part)
shows g b as calculated by Eq. (3-25) for a bubble column with DT = 19 cm
using values of these two variables. (The fluidized-bed data in the figure
will be discussed in Section III,D,4.) Curve C-C expresses Eq. (3-28).
Curve BQB is from Eq. (3-25) with iis from Fig. 34 and ut from Eq. (3-32);
BQP, the same, except with ut from Eq. (3-31). Curve BQP matches better
than BQB to curve C-C.
When a bubble column is scaled-up to a larger DT,calculations similar
to the above show that g b decreases only slightly with increasing DT,and
retains the same functional dependence on U, because of the properties of
ut and U s .

5 . Solution for u, and U,

By reversing the procedure of obtaining g b from Eq. (3-25) after deter-
mining ut and ii,, these two parameters can be determined i f g b is known as
a function of U, for a given DT.This step involves the use of Eq. (3-23b)
or, equivalently, of Eq. (3-25).
For simplicity, we take U, = 0. Also, for a bubble column, fis is con-
stant for U, 1 . 2 0 cm/sec. We designate this value as is0. Eq. (3-23b) is
then rewritten:

(3-34)

By taking i b from Eq. (3-28) and plotting u G / g b as ordinate and the

co-factor of (l/ut) as abscissa, iiso and l/ut are obtained simultaneously as
shown in Fig. 35. Here ut is assumed as being variable only with DT(as for
curve BQP in Fig. 36), with DT = 19.0 cm and U, 2 . 2 0 cm/sec. Figure 35
gives ut = 14.1 cmz/sec and iiso= 46 cm/sec. On the other hand, ut = 19.1
cmz/sec from Eq. (3-31) and iiso = 50 cm/sec from Fig. 34.
As is seen from Fig. 36, Eq. (3-28) (curve C-C) is lower than the calcu-
lated curve BQP. Lower z b means that the liquid is more easily recircu-
lated, so that a lower ut is obtained. According to several test calculations,
the above method tends to give vt 25-35% smaller than by Eq. (3-31)
(cf. Fig. 33), but gives a reasonable value for k0.
Essentially the same procedure is applicable to a fluidized catalyst bed,
 FLUIDIZED CATALYST BEDS 325

l -
D, = 19.0 cm
& = 0.054U$z
UG2 20 cm/sec

100

9, = 141 cmysec
~ = 4 6crri/sec

00

FIG. 35. Simultaneous determination of Y, and Us0 from Pb based on Eq. (3-34).

U, tcrn/secl
2 3 4 5 7 10 2 3 450 7 100

U, [cm/sec]

FIG. 36. Comparison of the calculated &, by Eq. (3-25) with the data given in Fig. 14.
Unless otherwise mentioned, U, = 49.5 c d s e c . Data are for FCC-catalyst beds of different
bed diameters. Curve E-E is an average of the data. The calculated curves BQB and BQP
are for bubble columns, FQF and FQP are for FCC-catalyst beds.
326 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

where Us remains nearly unchanged for U, 2 . 7 - 8 cm/sec as will be

shown. This procedure, although approximate, provides a convenient tool
for data analysis, since the dependence of Eb on U, is easy to measure and
not many data are yet available on turbulence in fluidized catalyst beds.

c. PHENOMENOLOGICAL BACKGROUND
OF TURBULENT
KINEMATIC
VISCOSITY
The turbulent kinematic viscosity v, has been introduced in the basic
relation, Eq. (3-2), and is correlated with the operational variables through
the experimental data (cf. Figs. 31-33) with a simplifying assumption that
vt is constant radially as an experimentally adjustable parameter.
It is desirable to establish a phenomenologically sound basis by relating
v, to well-established fluid-dynamic concepts, since we need to know the
possibility of scale-up or of applying the concept of vt to a fluidized
catalyst bed, as discussed in what follows (M27).

1. Bubble Column of Low-Viscosity Liquid

The radial velocity distribution inside a bubble column has been given
by Eq. (3-15). Differentiating this with respect to 4, one finds the velocity
gradient at the column wall, ( d ~ / d & ) + to
= ~be
, zero, which means that
peripherally descending liquid flow slides freely along the column wall. In
other words, the column wall keeps the downflow in a cylindrical form
at the periphery, but exerts only a negligible frictional force.
As a consequence, the peripherally descending liquid flows freely past
the centrally ascending liquid. This counterflow is inherently unstable; a
time-varying vortex-like mixing zone should be induced between the
counterflowing two streams, leading to intense turbulence. Takahashi (T5)
observed such vortex-like motion in measuring a two-point velocity cor-
relation along the column axis in a bubble column of 25 cm in diameter.
The time-varying turbulent motion of the bubble column seems to re-
semble the free turbulence of open jets (S4) or in the mixing zone of two
parallel streams of different velocities (D13).

2. Turbulent Viscosity on the Basis of Prandtl's Hypothesis

Prandtl (P7, S4) established a simplified equation for the apparent
kinematic viscosity of free turbulent flow based on experiments with free-
turbulent flow by Reichardt (R2). He assumed that the dimensions of the
lumps of fluid moving traversally during turbulent mixing are comparable
in magnitude to the width of the mixing zone. The virtual kinematic vis-
cosity vt is now formed by multiplying the maximum difference in time-
 F L U I D I Z E D CATALYST BEDS 327

mean flow velocity by a length proportional to the width of the zone. In

the case of the bubble column, the maximum difference in the velocity is
u,, + lu,I (see Fig. 27) and the length is the column radius R . Correcting the
effect of the liquid-volume fraction CL, u, is approximately expressed as:
ut = kGLR(uo + lu,l) (3-35)
where k denotes a dimensionless number to be determined experimen-
tally. It follows from Eq. (3-35) that Y, may be considered constant over
the whole width at every cross section of the column.
Introducing Eq. (3-30) with 4 = 0 for uo + Iu,I in the above expression
and solving Eq. (3-35) with respect to u,, we obtain the relation
Ut = ( k ” 2 / 8 ) [ g p & , ( 1- Eb)]1/2 (3-36)
Introducing Eq. (3-28) for Cb, ut is approximated by
ut = [0.0318(kg)1/2]U&&6D3,/2 (3-37)
With k = 0.0259, Eq. (3-37)reduces to Eq. (3-32) (in cm-sec), a correlation
based on experiment. This k value is the same order of magnitude as for
free turbulence flow (S4).
For liquid-liquid spray columns Wijffels and Rietema (W17) have stud-
ied the combined effect of bulk turbulent flow and the drop motion sus-
pended in it on the intrinsic turbulent viscosity uf . For the contribution of
turbulent flow alone they assume that u t is given by k*R Iuwl, with k*
adjusted at 0.0167 from experiment. Numerically, k* is the same order of
magnitude ask, although smaller, with u,, + Iu,( taken here instead of Iu,I.
For bubble columns the contribution of bubble wake motion has been
included in the bulk turbulent motion, since the recirculation flow is in-
duced primarily by the buoyant force of bubbles. The bubble-wake mo-
tion in the intense turbulent flow seems quite unstable, and is difficult to
distinguish from the motion of the turbulent bulk liquid.
Based on the interpretation developed here, the assumption of constant
u, and its empirical expression by Eq. (3-32), seem to be sound physically
for bubble columns. As shown in later sections, the flow properties of a
fluid catalyst bed of good fluidity (e.g., cracking catalyst) are in many
respects similar; hence the same interpretation can be applied to the fluid
catalyst bed (FCB).

D. INTERPRETATION FOR FCB

OF EXPERIMENTS

Thus far the experiments considered have been confined to bubble col-
umns, which provide a fair number of data over a wide range of conditions
and are therefore useful for testing the simplified theory. Equivalent data
328 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

are often not available for fluidized beds of carefully prepared particles,
especially with solids other than FCC catalyst.
The main uncertainty in applying the simplified theory to a fluidized
catalyst bed probably lies in the question as to whether the shear-stress
term in Eq. (3-10) is negligible in comparison with the buoyant-force term.
In what follows, although we still lack direct experimental proof regarding
the shear-stress term, the simplified theory will be seen to account well for
the bed performance (M27).

1. Interstitial Velocity Profile of the Emulsion Phase

As stated in Section II,D,3, the pioneer work of Lewis et al. (L11)
showed intense recirculation of the emulsion phase in an air-fluidized
MS-catalyst bed. The time-averaged velocity profile in the recirculation-
flow regime is still lacking, although Yamazaki (Y3a) has revealed the
profile at a low gas velocity ingeniously by utilizing thermal response. The
radial position 4. where the mean velocity is zero relative to the bed wall,
measured by Morooka (M43) for a FCC bed of 7.9-cm diameter, is shown
in Fig. 29. The deviation from prediction by Eq. (3-20) is about 0.3 cm and
is comparable with the resolution power of the strain-gauge probe.

2. Turbulent Kinematic Viscosity

Several data for mean ascending velocity ubo of bubbles along the bed
axis, in FCC-catalyst beds, are shown in Fig. 37. Velocity Ubo equals
Us + uo, with U, from Eq. (3-33) for a given &, and uo from Eq. (3-19).
Withu, = 49.5 cm/sec ford32 = 5 cm (cf. Fig. 341, Eb from Fig. 36, and vt

t o ) 4 5 7 10 2 3 & S O 7 100

Superficial gas vel. U,&m/sec)

FIG.37. Mean bubble velocity ubo along the column axis; experimental data are for
FCC-catalyst beds. Full curves are calculated by UbO = + u owith 4 = 49.5 c d s e c anduo
from Eq. (3-19). Umr is neglected due to Urn,<< U,.
 FLUIDIZED CATALYST BEDS 329

from Eq. (3-31a), theu,, which results is shown in Fig. 37. The calculation
matches the data for columns with diameters of 5.2 and 19.5 cm, but not
for those with diameters of 7.9 cm, for reasons which are not obvious.
In Fig. 37 the bubble velocity in slugging beds [cf. Eq. (5-4b)l is also
shown. The FCC-catalyst bed of 5.2-cm diameter is fluidizing smoothly in
the recirculation-flow regime with no indication of slugging when L, is
smaller than 50-60 cm.
By reversing the above procedure of obtainingu,, by knowing iis,Eb and
vt, one can calculate vt by knowing fi,, c b and ubo. These vt are shown in
Fig. 33 as open circles. They show reasonable agreement with the correla-
tion A-A for fluidized catalyst beds, where the data for 7.9-cm diameter
with U, < 30 c d s e c are omitted.

3. Mean Slip Velocity of Bubbles

The mean bubble size in a fluidized bed ii, has been discussed in Section
I1,B. As discussed, d,, for a fluidized catalyst bed of good fluidity may be
taken as approximately 5.0 cm [cf. Figs. 10 and 11, and Eq. (2-ll)] for
U , z 10 c d s e c . With Eq. (3-33), this dS2gives ii, = 49.5 c d s e c , which is
shown in Fig. 34 as a dashed line. It is interesting that the mean slip
velocity is essentially the same as for a bubble column, when V , 2 20
c d s e c . As noted in Section II,B, d32and ii, are very sensitive to change in
particle size, size distribution, shape, and density.

4. Bubble Volume Fraction

The averaged volume fraction I&,calculated by Eq. (3-25), is shown in
Fig. 36, for bubble columns and also for FCC-catalyst beds (M40). The
mean slip velocity of bubbles is again taken as ii, = 49.5 c d s e c . Also, vt
is calculated by Eq. (3-31a) for curve FQF and by Eq. (3-31) for curve
FQP, while curve EE is an empirical fit of the data. As in the case of a
bubble column, curve FQP matches better with curve EE, although FQP
is consistently higher. Curve EE tends to decrease the slope for
UG5 7-8 c d s e c , perhaps due to the decrease in ii,; the scatter of data
makes the behavior unclear. This is explained by the difference in the
region of U, < 20 c d s e c . The bubble column shows higher Eb values than
those for the fluidized bed, which is due to the bubble column’s lower ii,
(cf. Fig. 34).

5. Solution for vt and ii,

As shown above, ii, is constant at about 49.5 cmlsec for a fluidized
catalyst bed of good fluidity when UG > 7-8 c d s e c . Hence vt and ii, are
3 30 M I Y A U C H I , F U R U S A K I , MOROOKA, A N D I K E D A

obtained in this range by plotting the variables of Eq. (3-34) as deter-

mined from the experimental data. Figure 38 shows the calculation, with
gas holdup data taken from curve EE for DT = 19.0 cm in Fig. 36. The
resulting turbulent kinematic viscosity, lower than would be obtained
from curve FQP in Fig. 36 or curve AA in Fig. 33, is shown in Fig. 33 as
open triangles.
In summary, the calculations show convincingly that modern
fluidized-catalyst-bed technology has attained an emulsion fluidity nearly
equivalent to that of low-viscosity liquids. With such fluidity, data ob-
tained for a bubble column shed light on the performance of a fluidized
catalyst bed, and vice versa.

IV. Longitudinal Dispersion Phenomena as Derived from Flow Properties

Longitudinal dispersion in the continuous phase (the liquid phase for a

bubble column, and the emulsion phase for a fluidized catalyst bed) is
closely related to flow properties of the equipment. Here, we wish to
describe the longitudinal dispersion phenomena in terms of the fluid-
dynamic properties of the equipment. The prime purpose is to test
whether the fluid-dynamic analysis developed earlier is sound, but lon-

1’1 D~ = 19.0 cm
Fluidized RI: bed
-
&b VS. UG Fig,36

u,2 7 cm/sec
P

FIG.38. Simultaneous determination of vt and &, from mean gas holdup data by using
Eq. (3-34).
 FLUIDIZED CATALYST BEDS 33 1

gitudinal dispersion itself is important for estimating the reactor perfor-

mance (D2, G17, K11, L9, M20, P4, S8, V11, Y l ) .

A. THEORY
O F LONGITUDINAL
DISPERSION
FOR THE RECIRCULATION
FLOWREGIME
A gas bubble column is taken here as a model equipment undergoing
longitudinal dispersion of the continuous phase. The theory obtained is
equally applicable to a fluidized catalyst bed of good fluidity exhibiting
similar flow properties. The following procedure is from Miyauchi (M27).

1. Underlying Concept
In the bubble column the velocity profile of recirculating liquid is shown
in Fig. 27, where the momentum of the mixed gas and liquid phases
diffuses radially, controlled by the turbulent kinematic viscosity vt. When
UL = 0 (essentially no liquid feed), there is still an intense recirculation
flow inside the column. If a tracer solution is introduced at a given cross
section of the column, the solution diffuses radially with the radial diffu-
sion coefficient E , and axially with the axial diffusion coefficient E,. At the
same time the tracer solution is transported axially b y the recirculating
liquid flow. Thus, the tracer material disperses axially by virtue of both
the axial diffusivity E, and the combined effect of radial diffusion and the
radial velocity profile.
The latter mechanism assumed is the well-known Taylor dispersion
(T9, TlO, T l l ) , which has been studied extensively ( A l l , G6, L9, T14,
S2). High-speed motion pictures taken by Towell et al. (T23) in a 40-cm
bubble column (R3) have shown the presence ofturbulent eddies, on a scale
roughly equal to the column diameter, with systematic large-scale circula-
tion patterns superimposed. Their pictures showed that liquid near the
wall flowed downward, while liquid near the center of the column flowed
upward, consistent with the flow theory developed earlier and with the
Taylor dispersion mechanism.
Axial dispersion in the bubble column is usually well expressed by the
following one-dimensional diffusion model with respect to liquid concen-
tration c:
ac/ae = E~ a2C/a~2- ii ac/az - aqc) (4- 1)

where ErT is the total axial dispersion coefficient (interstitial value), ii is

defined by ii = UL/EL,the mean interstitial liquid velocity, and @(c) is the
reaction or mass-transfer term. The total dispersion coefficient EzT is split
332 M I Y A U C H I , F U R U S A K I , MOROOKA, A N D IKEDA

into the contributions of two mechanisms:

EzT = E,, + Ez
where E,, is the contribution of Taylor dispersion, and Ez that of the usual
axial diffusion averaged over the entire column cross-sectional area.
The total dispersion coefficient is usually determined by measuring the
concentration distribution of tracer material steadily back-mixed up-
stream, or by knowing the impulse response of the column for tracer
liquid (see Section IV,A,3).

2. Basic Relations
The equation of continuity for the liquid phase of the bubble column is
obtained from a mass balance on tracer material for differential gas-liquid
mixed phases. With the same procedure as for a homogeneous flow, the
following equation is obtained when @ ( c ) = 0 and the aspect ratio
LID, >> 1:
-
r ar
[r( -EreL$)] - ueL ac = cL ac as (4-3)

Here, the axial diffusion term, cLEr d2c/dz2,is tentatively omitted for
simplicity, and the radial mass flux N, (with v, = ELE~for simplicity) is
defined by
N , = -eLEr ac/ar = -u, a c / a r (4-4)
When the reference plane moves with mean net liquid velocity of
kUJ(1 - G,,) (+ indicates cocurrent and - countercurrent), Eq. (4-3)
remains valid; u , the liquid velocity relative to the moving reference
plane, is given from Eqs. (3-15) and (3-17) by setting UL= 0. This is
justified because the net flow term, which defines the moving reference
plane, cancels with the net flow term included in the linear velocity term
u , [see Eqs. (3-15) and (3-17)]. In what follows in this section we takeu as
velocity relative to the moving plane.
We now consider the case of steady back-mixing experiment to mea-
sure the total axial diffusivity EzT. A small amount of fresh liquid is con-
tinuously introduced into the column cocurrently or countercurrently with
the gas flow. A still smaller amount of tracer liquid is steadily introduced
at a downstream section. Under such circumstances the concentration
distribution of the tracer at a given cross section is steady, and the axial
concentration gradient a c / d z is constant radially. As a consequence, Eq.
(4-3) simplifies to
-- -(
i a -rv,
r ar
5) - ucL ac- oz- (4-5)

with a c / a z almost independent of r at the steady state.

 FLUIDIZED CATALYST BEDS 333

The axial dispersion coefficient E,, given by the Taylor mechanism for
the above case is given by the procedure of Tichacek et al. (T14) for
turbulent pipe flow. Integrating Eq. (4-5) twice with respect to r, one has

The net transport Q past any reference plane moving with liquid at
the mean velocity +U,/(l - Eb) is

Q = 27~loR(c, - cR)ueLrdr (4-7)

Q is seen to be proportional to the axial concentration gradient ac/az
when Eqs. (4-6) and (4-7) are combined. Hence the mean axial disper-
sion coefficient Ex, is defined by
Q = - rrR2ELE,,(dc/az) (4-8)
Combining Eqs. (4-6)-(4-8), E,, is given by

where 4 = r / R , and u is obtained from Eqs. (3-15) and (3-17) by setting

u, = 0.
3 . Radial and Axial Diffusivity ;Simplification of Eq. (4-9)
For integrating Eq. (4-9), urn(= ELE,) should be known as a function of 4
and operating variables. However, the momentum diffusivity ut is the only
term we know, with essentially no systematic data for u,. In the case of
free turbulence of a homogeneous fluid, the diffusivity of a scalar quantity
like heat and mass is estimated to be about two times that of momentum
(S4) and the two diffusivities are not far apart for turbulent pipe flow (S8).
However, such a relation is not available yet for gas-liquid bubble flow in
bubble columns. Generally the local radial mass diffusivity v, may be
expressed by a u t , with a being a numerical coefficient of order unity.
In the recirculation flow regime Eissa et al. (E6) have reported radial
diffusivities E, of 1.21 cm2/sec at UG = 6.3 c d s e c and 1.26 cm2/sec at
7.8 cm/sec for a 5-cm-diam bubble column equipped with a multiple-
orifice-plate gas distributor. These diffusivities amount to 60% of the eddy
kinematic viscosity; ut = 2 cm2/sec for DT = 5 cm (Fig. 31). Reith et al.
(R3) have measured a radial diffusivity of 10 cm2/sec at UG = 9.3 c d s e c
for a 14-cm-diam bubble column equipped with a perforated plate orifice
having one 2-mm-diam hole per cmz; this is slightly lower than the average
kinematic viscosity for the same size column (ut = 11.4 cm2/sec). At UGof
6.0 and 10 cm/sec, Pozin et al. (P6) report a radial diffusivity of 35-40
334 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

cm2/sec in the core region of a 30-cm-diam bubble column with a gas

distributor plate having 1.5% free area with 2.5-mm hole diam. The turbu-
lent kinematic viscosity read from Fig. 31 is 42 cm2/sec for a 30-cm-diam
column.
As for axial diffusivity Ez included in Eq. (4-2), the situation is not much
different from that for radial diffusivity. Pozin et al. (P6) report that the
mean value of axial diffusivity EZis 2.5 times the value of E,., when radial
and axial diffusion are assumed to be homogeneous and nonisotropic
throughout the bubble column. Hence it is reasonable to assume for the
local axial diffusion coefficient E, = E,, with 5 of order unity. On the
other hand, one has the relations that E, = v,/eL [see Eq. (4-4)] and
v, = a v t , so that E , equals the product (a5/EL)vt.Accordingly, the cross-
sectionally averaged axial diffusivity E, is expressed by the relation
E, = ((YS/EL)v,,where the overbar shows an effective mean value.
Two approximations are introduced, for simplicity, to perform the in-
tegration of Eq. (4-9) for the Taylor mechanism dispersion coefficient Ear.
First, the term v, included in the second integral of Eq. (4-9) is eliminated
from the integral by taking an effective mean value i, (= G v J , with ~5
being a mean of a. Second, the local liquid holdup eL is eliminated from
the multiple integral by using the mean liquid holdup EL, with a correction
factorf of order unity (f 2 1).
With these simplifications, Eq. (4-2) is rewritten for subsequent use:

where qr and qz are defined by qr = f E L / & and qa = LYS/EI., respectively.

Properties of qr and q, are not yet evident from experiment, but they are
of order unity and should be weak functions of the superficial gas velocity
U, because they include the liquid holdup q,.Also, qz increases andq,just
as eL decreases, with increasing U G .
In regard to axial dispersion in unbaffled bubble-flow equipment like
liquid-liquid spray columns, gas bubble columns, or fluidized catalyst
beds, a close similarity has been supposed as a result of bubble flow and of
turbulence induced by bubbles (B3, M33). Baird and Rice (B3) have as-
sumed that the Kolmogoroff concept for eddy viscosity in isotropic turbu-
lence is applicable to evaluate EZT in the unbaffled bubble column under
turbulent conditions, concluding that EzT is 0.35 D$'3~$3 in cm-sec units,
with E , the rate of energy dissipation per unit mass of continuous phase.
Experimental observations indicate that axial dispersion follows the
Taylor dispersion mechanism.
The eddy viscosity vt as given earlier for bubble columns has also been
correlated (M27) according to Kolmogoroff's local-isotropy concept, to
 FLUIDIZED CATALYST BEDS 335

relate vt for the columns with data for atmospheric turbulence taken by
Richardson (R6, 01). This approach gives vt approximately equal to 0.12
14/3~$3in cm-sec units; for atmospheric turbulence E,,, is assumed to be 5
cm2/sec3(01) and 1 one-half the distance between pairs of observational
points; for the bubble column 1 is assumed to be one-half the column
radius. Physically the Prandtl hypothesis may be a sounder approach.

B. LONGITUDINAL
DISPERSION
I N A BUBBLE
COLUMN
Many data are available for the longitudinal dispersion coefficient of the
bubble column in the recirculation flow regime. Most are summarized in
Fig. 39. Of these, Kato and Nishiwaki (K6, 0 and 0),Ohki and Inoue

Hole size, 2 & 3mm

UC [cm sec]

FIG.39. Longitudinal dispersion coefficients of liquid in bubble columns (SN = single

nozzle; PP = perforated plate). The full circles are calculated with the use of Eq. (4-12).
336 M I Y A U C H I , FURUSAKI, MOROOKA, A N D IKEDA

(02), Towell (T24, @), and Hikita and Kikukawa (H11) have utilized a
pulse response technique or step response; and the rest of the data have
been obtained by the steady back-mixing method. These are correlated as
follows (M27).

1. Basic Relations for Longitudinal Dispersion Coeflcient

Equation (4-10) is solved for the bubble column by utilizing Eqs. (3-15)
and (3-17) for u ;the u included in Eq. (4-10) is obtained by setting U L= 0
in the two earlier equations. The final expression, in cm-sec units, is

For bubble columns of relatively small diameter the first right-hand-side

term of Eq. (4-11) is much larger than the second term, so that E,T is
nearly proportional to D$/v: at a given U,. As a consequence, the func-
tional dependence of vt on D , can be tested accurately by inspecting the
dependence of EZTon D,.
Introducing Eq. (3-32) for vt and Eq. (3-28) for Zb, Eq. (4-11) is trans-
formed in cm-sec units to

(4-12)

where empirically qr = 1.864/U&l4andq, = 0.813U&!2, with UGin c d s e c ,

to give Eq. (4-1 1) the best match with experimental data. The given de-
pendence of 4+ and q, on U, looks qualitatively sound, but note that a
slight change in the dependence of vt and g b on U , and DT results in a large
influence on q r and q,.

2. Comparison of Theory with Experiment

The total axial dispersion coefficient E,, calculated by Eq. (4-12) is
shown in Fig. 39 as a function of DT and U,. Agreement with experiment
seems satisfactory except for data from Argo and Cova (AlO), who used
an off-center single-pipe gas distributor tending to create tangential swirl-
ing motion, which may have modified the liquid recirculation flow pattern.
They also found a higher gas holdup Z b than is given by the correlation
utilized in Eq. (4-12), which could also lead to higher EzL.Their axial-
dispersion coefficients correspond to DT = 65 cm, about 1.5 times that of
their column. Data of Aoyama et al. (A9) for 5 - , lo-, and 20-cm-diam
columns equipped using a sieve plate as a gas distributor fall slightly
above the calculated curves.
The superficial liquid velocity varies over the range of 0-2.18 cdsec for
the data available; this seems to have very little effect except at very low
 FLUIDIZED CATALYST BEDS 337

gas velocity. Also, the gas distributor construction has little effect in the
recirculation flow regime, provided the distributor holes are placed sym-
metric around the column axis.
Eq. (4-12) is simplified further by approximating the term in parentheses
as a function of U G .In cm-sec units this leads to
ExL = 1.35UjjzPdz (4- 13)
for U G = 5-60 cm/sec. For DT = 10-30 cm Eq. (4-13) is modified to
UGDT/E,L = 2.0( UG/(gDT)1/2)112 (4-14)
which is nearly the same as the dimensionless correlation by Kato and
Nishiwaki (K6) for UG/(gDT)*’z2 0.03. Also, Aoyama et al. (A9) report
that E Z L is proportional to Pdzfor the columns with a sieve plate as a gas
distributor. An empirical correlation by Beduraet al. (B7) gives E z L equal
to 2.4uOG.33D$4.
Some correlations are available for the apparent axial diffusivity E L E z L .
Towell and Ackerman (T24) have given ELE,L = 1.23 Uj12PJ2. Deckwer et
al. (D14) obtain the same relation with 2.0 in place of 1.23. Towell and
Ackerman’s correlation is reduced to ErL = 1.5 L13j8P2zin cm-sec units.
Equation (4-12) explains the axial dispersion coefficient reasonably well,
as shown in Fig. 39, where the range of experimental variables is DT 2.I0
cm, the perforation hole size 2 2 mm and UL 5 2.18 c d s e c , and the liquid
is water or aqueous solutions with the liquid viscosity between 0.86 and
14.5 cp. and the surface tension = 33-75.5 dyne/cm.
At low gas rate, the size of gas bubbles arid the gas hold-up are fairly
sensitive to the gas-distributor structure (A9, F7,K17, 0 2 , T15). Mixing
and bulk recirculation of liquid are induced by the interaction of liquid and
bubble motion, so that axial dispersion is influenced by the structure of
gas distributor and may show strange behavior (A9, 02). For example,
Ohki and Inoue (02, Fig. 9) show that ErLincreases rapidly with U Gto a
maximum at UG = 5 c d s e c , then decreases gradually, and approaches
typical recirculation-flow behavior regime at around UG= 16 c d s e c . In
this example a sieve plate with 91 holes of 0.4-mm diam is utilized for a
ldcm-diam column. Aoyama et al. (A9) found maximum and minimum
Ex,, in the low-gas-flow rate region of their 5.0-cm-diam bubble column
using a sintered porous plate for gas distribution. These anomalies are
closely related to the properties of bubble holdup in the transition from
bubble flow to the ultimate circulation flow with increasing UG.The prop-
erties of axial dispersion in these flow regimes may be explained by the
Taylor dispersion mechanism (T9, A1 1); thus,

(4-15)
338 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

where ko is a constant characteristic of the liquid flow pattern and U is an

effective mean liquid velocity, and measures the axial spread of liquid due
to velocity distribution. When many small bubbles of uniform size are
generated and rise up through the liquid rather homogeneously, the liquid
flow pattern and velocity are modified and radial mixing is suppressed. E,,
changes as a result of relative magnitude Uz and k&,, sometimes increas-
ing or decreasing quite rapidly as UG increases.
For organic liquid aeration, Hikita and Kikukawa (H11) show that
EzLcc l/pt.lz with negligible effect of surface tension. Alexander and
Shah (A5) obtain essentially no effect of liquid viscosity p L or surface
tension on EZL.
For recirculation flow the Taylor dispersion mechanism was introduced
by Shyu and Miyauchi (S13). Equation (4-12) is a revised result for it. For
this flow regime, Ohki and Inoue (02) developed an expansion model with
parameters adjusted to the data available, and also introduced the Taylor
dispersion mechanism for the low-gas-velocity region of uniform bubble
flow.

C. LONGITUDINAL
DISPERSION BED
CATALYST
I N A FLUIDIZED

Most data available from past studies are summarized in Figs. 40 and 41
for the longitudinal dispersion coefficient of the emulsion phase in
fluidized catalyst beds of good fluidity. Such coefficients were first mea-
sured in _the pioneer work of Bart (1950, B20) for a wide range of gas
velocity ( V , = 7.5-90 c d s e c ) . His values with a 3.2-cm-diam column are
approximately one-half those by Morooka et al. (M40) in the higher flow
rate region. The basis for their correlation follows (M27).

1. Underlying Concept for Longitudinal Dispersion

When the velocity profile of the emulsion phase is similar to that of the
liquid phase in a bubble column, Eq. (4-11) will apply to the fluidized
catalyst bed. This similarity seems to be well justified as mentioned in
Sections III,A,4-5, although there is no direct calculation of the turbulent
kinematic viscosity ut from the measurement of velocity profile in the
fluidized catalyst bed.
For bubble columns Em is estimated from Eq. (4-11) by determining
u, and Eb. For the fluidized bed, reversing the process utilized for the
bubble column, ut can be calculated from z b and an experimental EZT.
As discussed in Section IV,B, 1, EzT is approximately proportional to
PT/u;at a given U G ,so that the dependence of ut on DT is fairly accurately
determined. Also, the ut dependence on UG should be weak as mentioned
in Section III,A,4 in relation to the Prandtl hypothesis. Systematic data
 F L U I D I Z E D CATALYST BEDS 339

for v, of the fluidized catalyst bed are not yet available from flow
properties.
For the FCB we adopt the notation E, instead of E,.,. Experimental data
for E, have been processed to know how they relate to DT under a given
UGand to UGunder a given DT. With these relations in hand, Eq. (4-1 1) is
utilized to find the turbulent kinematic viscosity vt of the fluid catalyst bed
as a function of DT and U,, applying the 9,. and qr utilized for the bubble
columns [Eq. (4-12)]. The vt thus obtained has been given in Fig. 32 as
curve A-A, matching Eq. (3-31a). As is evident from Fig. 32, the turbulent
kinematic viscosity for a fluidized catalyst bed with good fluidity is ap-
proximately the same as for the bubble column, and this suggests strongly
that the two systems behave similarly. In the recirculation flow, for ex-
ample, the emulsion phase should be in intense turbulence. The bubble
sizes in this flow regime should be similar for the two cases as is found in
Fig. 34.
Introducing Eq. (3-3 la) for vt into Eq. (4-1 I), again taking q 7 and q z
from Eq. (4-12), the longitudinal dispersion coefficient E , of the emulsion
phase is given in cm-sec units as follows:
E%s = (EzT)ernulsion

11

The equation applies for to U, 2.10 c d s e c , utilizing the approximation

taken for Eb. The equation simplifies further, with a slight change in coeffi-
cient to fit best the data available (in cm-sec units):
E,~= 1 2 . 0 ~ ~ 2 ~ : 9 (4- 17)
where the range of parameters is UG = 10-50 cm/sec and DT = 3.2-200
cm.

2. Comparison of Eq. (4-16) with Experimental Data

The totzl axial dispersion coefficient E,, of the emulsion phase calcu-
lated from Eq. (4-16) is shown in Figs. 40 and 41 as a function of DT and
U G , in reasonable agreement with experimental data. The E, shown
here are for particles of good fluidity. Also, note in Figs. 40 and 41 that the
second term in the right-hand side of Eq. (4-16) tends to be significant as
DT increases. At U , = 20 c d s e c the contribution of this term to E , is
10% for DT = 100 cm and 22% for DT = 300 cm. At U , = 50 c d s e c the
contribution decreases down to about 0.75 times the mentioned levels.
Figure 41 illustrates the influence ofDT and the properties of bed parti-
cles on E, at UG = 20 c d s e c . Although data from de Vries et al. (D8, V8,
340 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

5.10' """ ' ' " ' 1 1

'I
w

5
4

)1
1oz
7

U, [cm/sec]

FIG.40. Longitudinal dispersion coefficient E,, of emulsion phase in fluidized catalyst

beds as a function of U c and DT.

@) and those from de Groot (D7, V8, $) deviate from those for FCC
particles at smaller or larger bed diameters, respectively, the general trend
is nearly parallel.
The turbulent kinematic viscosity u, of the fluidized catalyst bed has
been determined, as Eq. (3-3la), from the use of axial dispersion coefficient
E,. This is a natural consequence of the analogy between the bubble
column and the fluidized catalyst bed of good fluidity (such as in fluidized
catalytic cracking). The mean gas holdup (Fig. 36) and the mean'bubble
velocity along the bed axis (Fig. 37) are reasonably well predicted by
applying Eq. (3-31a) for the fluidized cracking catalys! bed.

V. Bubble Phenomena in Relation to Bed Performance

The bubble holdup Eb for a bubble column or for a fluidized bed is easily
obtained as a function of superficial gas velocity U G .Useful information
 FLUIDIZED CATALYST BEDS 34 1

lo'

n 5

. de Vries et al. ( 8

lo2
Bed diameter D, [cm]

FIG. 41. Influence of particle properties on axial dispersion of the emulsion phase;
U G = 20 cm/sec with variable bed diameter D T ;keys are given in Fig. 40. Data by May
(U,= 25 cm/sec) and those by Stemerding (U,= 10 cm/sec) are extrapolated to UG = 20
cm/sec according to Fig. 40.

has been extracted from the Zb versus V , relation for the fluidized bed
(D3, N4, P8, T29). In recirculation flow an additional term appears in the
apparent mean slip velocity shown on the right-hand side of Eq. (3-23b).
This equation has been utilized to determine the turbulent kinematic vis-
cosity vt and the mean slip velocity ii, (cf. Sections 111,B,2, III,B,5, and
III,D,5). In this chapter, the equation is further examined in relation to
bed performance, since the turbulence properties of the bed result from
interaction between bubbles and the continuous phase. As shown in Fig.
34, the mean slip velocity of bubbles in a fluidized catalyst bed of good
fluidity is essentially the same as that for a bubble column when UG k 20
c d s e c . A criterion under which bubble size approaches a dynamic
equilibrium is obviously needed for predicting or evaluating the perfor-
mance of scaled-up beds.

A. MEANBUBBLE WITHOUT BULKRECIRCULATION

VELOCITY
A brief survey is needed here of bubble flow, with or without a bubble
wake. Bubbles of a uniform size are generated by a gas distributor and rise
uniformly through the continuous liquid phase in a bubble column, or
342 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

through the emulsion phase for a fluidized catalyst bed. This one-
dimensional bubble flow has been discussed extensively (B14, D4, N4,
27). A bubble column without continuous liquid feed ( U , = 0) is taken for
simplicity.

1. Velocity of the Bubbles

Case a of Fig. 42 shows the steady bubbling of gas through stagnant
liquid in a bubble column, according to Nicklin (N4), where U, = 0.
Superficial gas velocity UGis related to the ascending velocity iib of bub-
bles relative to the wall:
UG/Eh = kh (5- 1)
where Eb is the gas holdup under continuous bubbling.
Case b of Fig. 42 shows bubbles in exactly the same configuration, but
rising relative to the stagnant liquid above in a swarm of finite size. In this
case the bubble swarm has the same gas holdup Eh as case a, but rises at a
velocity i b o (relative to the wall) different from fib. In case a there is no net
flow of liquid across the section A-A'. In case b there must be a net
downward flow of liquid across the section A-A' to cancel the upward
flow of gas.
The net downward flow of liquid is the amount necessary to fill in the
bubble cavities left behind the swarm per unit time, i.e., EbiihO. Hence the
interstitial velocity of the downward flow of liquid through the bubble
swarm is GbfibO/(l - &), and the ascending velocity of the swarm is re-
duced from &, by this amount:
fib0 = fib - EhfibO/(l - Eh) (5-2)

- Free

9,

A --
gb
"b

$ Gas
1

flow
Gas flow
is stopped
(a) (b)
FIG.42. Example of the rise of swarm of bubbles: (a) for continuous bubbling; (b) for a
rising swarm of bubbles.
 F L U I D I Z E D CATALYST BEDS 343

Equations (5-1) and (5-2) give the following relation (N4):

uG/ Eb- + UG
-fi
= u-b - hO (5-3)
Taking UG - Urn,instead of U G , Eq. (5-3) is equally applicable to the
fluidized bed, where Umfis the superficial velocity at incipient fluidization
and is usually negligible in comparison with UGfor the fluidized catalyst
bed.
Nicklin et al. (N5) have shown that an equation similar to Eq. (5-3)
is applicable to slug flow. A single slug rises at a velocity equal to
0 . 3 5 m . If Eq. (5-3) is modified to allow for a nonuniform velocity
profile in the liquid regions between the slugs, an empirical equation
results:
UG/gh = 1.2u(3 + 0.35(gDT)’/2 (5-4a)
For fluidized beds Eq. (5-4a) is modified (D3, G19, 0 7 , P10, S19):
(UG - umf)/gh = (UG - umf) + 0.35(gDT)’/2 (5-4b)
The onset of slugging represents the terminal stage of bubble coalescence
when a bubble spans the entire cross section of the bubble column (B15,
S21).
If the bubbles in case a of Fig. 42 do not interfere with each other (D3),
then the velocity of the bubbles should equal the free-he velocity Us0 of a
single bubble in a cross-sectional area of liquid large enough for the liquid
phase to have no net vertical velocity. This has been indicated by Turner
(T28) in regard to the fluidized bed, for which he modified Eq. (5-1) to the
form
fib = (UG - Umf)/gh = fiso = 0.71g”2V26 (5-5)
where v h is the volume of a spherical-cap bubble. This relation has been
once applied to a bubbling bed (L5), but later abandoned (F12).

2. Influence of the Bubble Wake

So far, the influence of bubble wake on mean bubble velocity f i b relative
to the column wall has not been mentioned, since Eq. (5-3) has been
formulated on the basis of fib(,, which already includes the effect of the
wake (although it lacks a correction for wake fraction). In bubbling-bed
models (FlO, F12, K24, L5, S18) an upward flow of solid carried by the
bubbles and bubble wakes leads to a downflow of solid (that has been
assumed uniform) in the remainder of the bed. Then the bubble velocity f i b
relative to bed wall should be smaller than the slip velocity of the bubble
Us relative to emulsion, since the bubble phase is retarded by downflow of
344 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

solid. The rate of downflow of solid will be a maximum when the wake is
formed at the gas inlet and released at the free surface without being
renewed during the travel; also, the rate of downflow will be essentially
zero when renewal is very fast during the travel. The actual situation lies
between these two extremes (R17); and essentially the same situation
applies in the gas bubble column.
We now consider the extreme case of negligible renewal of bubble wake
for a continuously bubbling bed. The wake fraction in the bed isfwZh.The
bubble phase is composed of a bubble and a wake, and the fraction of this
phase is (1 + fw)& In the bubbling-bed model (F12, K24) the bubble
phase is assumed to ascend at velocity ii, relative to the emulsion phase.
The rate of release of wake solid at the free surface of the bed is then
fw&iih, where iib is the rising velocity of the bubble phase relative to the
bed wall. Released solid constitutes the downward return flow through the
emulsion, its velocity being given by f&uh/[ 1 - (1 + f,).h] This return
flow retards the ascending motion of the bubble phase, and leads to the
following relations:
fib = - fwEbuh/[ 1 - (1 + fw)eb] (5-6)
or
fib = (u, - umf)/g = [l - fwEh/(l - Eh)]& (5-7)
As a consequence i i h is always smaller than the slip velocity ii, for the
extreme case of negligible renewal of the wake and equals Us whenf, = 0
or when the rate of renewal is very rapid for finitef,. In the latter example
the wake loses its identity relative to the emulsion.
In conclusion, for a homogeneously bubbling bed, the velocity of bub-
ble rise i i b relative to the bed wall is equal to or lower than the bubble slip
velocity ii, relative to the continuous phase.

B. MEAN BUBBLEVELOCITY WITH BULKRECIRCULATION

This section is concerned with the behavior of Eq. (3-23b) when applied
to all bubble-flow equipment (M27).
In the preceding section the bubble velocity fib in a homogeneously
bubbling bed has been shown to be no greater than the slip velocity iiS.
However, numerous measurements do show iib exceeding ii,; they are
generally expressed by the following empirical formula, with p an empiri-
cal coefficient characteristic of the flow properties:
fib = (u, - umf)/cb = + p(uG - umf) (5-8)
where Urn,is the superficial gas velocity at incipient fluidization and is zero
 FLUIDIZED CATALYST BEDS 345

for the bubble column; Us is the mean slip velocity of bubbles, equal to the
free-rise velocity Us, when the interaction between bubbles is negligible.
Towel1 et al. (T23) give p = 2 for their 40-cm-diam. bubble column under
fairly high speed aeration. In Davidson and Harrison's well-known equa-
tion (D4), p equals unity with ii, = iiso, although their equation is derived
on the basis of homogeneous bubble flow.

1. Simplified Recirculation-Flow Model

The relation given by Eq. (5-8) is obtained easily for a bulk recirculation
flow of the continuous phase. We consider a bubble column of radius R in
which the column liquid is in upflow centrally with a constant interstitial
velocity iil,. Peripherally the liquid descends, forming a recirculation flow.
For simplicity, it is assumedZthat gas bubbles distribute uniformly in the
central upflowing liquid and ascend with it, and that no bubbles rise
through the peripheral downflow. Here, the mean gas holdup of the
upflow is &* and that averaged over the total column cross section is Eb.
The bubble wake released at the free surface, joined with the upflow,
constitutes the downflow; so that there is no downward return flow
through the bubble layer. Under these circumstances, the mean bubble
velocity i i b relative to the column wall is
iib = + fil, (5-9)
The mass balance for the bubble phase gives
rR2uG = r r : c b . i b (5-10)
where r . is the radius of the central upflow stream. In Eq. (5-10) rr!Cb, is
obviously equal to rR2ib. Hence
iib = UG/Eb = fis + iil, (5-11)
In this simple example til, corresponds to p(UG- Umf),with Urn,= 0, in
Eq. (5-8), and the bubble wake shows no influence on i i b . Here the simple
recirculation flow gives the additional term El, even if there is no turbulent
motion.

* When a gas bubble column of& = 20 cm is operated at UG = 20 cm/sec, the fraction of

gas throughput which rises through the central upflow zone amounts to 89.3% of the total
gas, where the fraction p, is given as follows:

lo''* +
pb = ( T R z ~ , ) - ~ 21rr(u U&b dr
=I-(I--) u;/2 .1- 1.5ib
uG/cb 3(1 - Zb)
Hence, the simplification of 100% is not unreasonable.
346 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

2. Properties of the Bubble Wake

For one-dimensional homogeneous bubble flow, the wake usually re-
duces the ascending velocity fib of the bubbles. The extent of reduction,
however, depends on the rate of wake renewal, as shown in Section
V,A,2. For the case of Section V,B,I, the wake has no influence on f i b .
Thus the way by which the wake influences f i b depends strongly on the
physical situation encountered. In what follows the typical recirculation
flow under intense turbulence will be treated, since this is the focus of our
interest. We now show that the influence of the wake can be neglected in
this flow regime.
Several investigators have observed the size of wakes in the turbulent
flow field. Ahlborn (Al) has observed that the wake formed behind a solid
cylinder gets much smaller as the cylinder approaches a turbulence grid.
Kojima et al. (K16) observe for a spherical-cap bubble that the size of
wake when the main stream is disturbed by a turbulence grid is not much
different from when the main turbulent stream is undisturbed. They indi-
cate that the volume of wake is approximately 4.7 times greater than the
bubble volume. However, their bubbles are located downstream from a
turbulence-promoting bar (a single glass cylinder of l-cm diameter) a dis-
tance 8.3-38 times the bubble diameter, e.g., these bubbles would be
located 15 cm downstream from the bar. Accordingly the bubbles do not
seem unstable; the shape of the bubbles changes irregularly, but the wake
is not clear whether washed out frequently or not.
When a spherical-cap bubble rises through a uniform swarm of smaller
bubbles in water, velocity enhancement of the cap bubble is observed, the
enhancement being associated with a change in its shape to an axially
elongated form (H10,Fig. 4). In this case the wake will probably be
reduced considerably in volume.
We now take a frequently encountered case of a bubble column of
20-cm diameter, operated at UG = 20 c d s e c , with water. The bubble
holdup is about 24% [Eq. (3-28)], and many bubbles of fair sizes ascend
irregularly through the turbulent water phase. I f a wake with 4.7 times the
bubble volume accompanied each bubble, the continuous water phase
would be essentially composed of wakes. Hence we have no way of dis-
tinguishing wakes from turbulent bulk liquid. Also, each bubble should be
relatively spherical although of irregular shape, as suggested from Hills’
observations (H 10) and also observed experimentally. Washing out of
wakes seems to occur frequently, due to the irregular motion of bubbles in
the intense turbulence. At one moment a bubble may have wake-like flow
behind it, and at the next moment the flow may be washed out. As a
consequence it is hard to distinguish the wakes from the turbulent bulk
liquid.
 FLUIDIZED CATALYST BEDS 347

When the wakes are renewed rapidly, they have little influence on bub-
ble motion (Section V,A,2).A relatively small influence of wakes is par-
ticularly the case when bubble motion is in the recirculation-flow regime
(Section V,B, 1).
The influence of wake motion on bulk turbulence induced in the liquid is
understood more clearly by inspecting oscillograms which show the fluc-
tuation of local liquid velocity. Figure 43 shows such oscillograms taken
by Kikuchi (K30) with a hot-wire probe. The bubble column is 8.0 cm in
diameter, water-filled to a 170-cm height, with the probe 115 cm above the
bottom gas distributor. In the column bubbles of constant volume (100 cc)
are injected successively at constant time intervals, either at 2.1 sec (case
a) or 0.50 sec (case b). Case c is an example of continuous bubbling at
UG = 6.45 c d s e c .
In case a, wake motion is obviously identified for each bubble passing
over the probe. The local velocity of liquid changes periodically in reso-
nance with the bubble frequency. The resonance, however, is consid-
erably disturbed as the bubble frequency increases. The oscillogram
shows the onset of irregular liquid motions, which have different frequen-
cies from those of bubbles. The relation between wake motion and local
liquid motion becomes quite dubious. This tendency is increased in case
b, where liquid motion exhibits fairly low frequencies. These flow proper-
ties are succeeded by continuous bubbling (case c). Here the local liquid
motion is governed by the properties of bulk liquid motion, rather than by
the bubble wakes.
Based on these discussions, the influence of bubble wakes has been
omitted in what follows, since the wake motion is difficult to distinguish
from bulk turbulence. However, the question of wake effects has not yet
been settled experimentally.

3. Mean Bubble Velocity in Recirculation Flow

Under negligible influence of bubble wakes, the gas balance for the
bubble phase is:
(5-12)

In the same manner as for Eq. (3-23), a radially constant mean Us is

taken for the slip velocity u s . Also, with an obvious relation that
Eh = 1 - ( 1 - Eh), Eq. (5-12) is modified to

nR2UG = loR2Tru dr loR

2TrU(1
- - Eh) dr -k TR2g& (5-13)

The second integral is the net liquid flow through the column and is zero
 FIG.43. Hot wire oscillogram showing the fluctuation of local liquid velocity; air-water
system, DT = 8.0 cm,S = start of bubbling; data by T. Kikuchi. (a) Time interval is 2.1 sec
for bubble injection. (b) Time interval is 0.5 sec. (c) Continuous bubbling at UG = 6.45
cdsec.
 FLUIDIZED CATALYST BEDS 349

when U, = 0. Hence the equation simplifies to

fib uG/& = & + u*/$ (5-14)

where U* equals the first integral of Eq. (5-13), given in a footnote to Eq.
(3-23b). Equation (5-14) is general for conditions of constant 4 and no
liquid feed, since it assumes no particular radial-velocity distribution. For
the fluidized catalyst bed, ( UG - Umf)is taken for UGin Eqs. (5- 12145-14).
From Eqs. (5-8) and (5-14) (with Umf= O ) , one has for the bubble col-
umn

In the case of recirculation flow, as developed in Section III,A, u * / g b is

given by the second term on the right-hand side of Eq. (3-23b). Hence
p = (uO/uG)gb/(l - gb) (5-16)
where uo = gpT/192v,. When the quadratic equation with respect to Gb,
which is obtained from Eq. (3-23b), is modified into the form of Eq. (5-8),
p is given as follows:
p = 1 + ( 0 0 - fis)gb/uG. (5-17)

The data necessary for calculating p using the above equations have al-
ready been given in Section 111.
Figure 44 shows p as calculated by Eq. (5-16) as a function of UGand
DT. Generally, p changes with these two parameters. ForD, = 40 cm, p is

UG km/sec]
FIG. 44. p as a function o f UGand DT for a bubble column, where UJZb = U s + PUG.
350 MIYAUCHI, F U R U S A K I , MOROOKA, A N D IKEDA

approximately 2 in conformity with Towel1 (7'23). Also, p 2 1.1 for the

usual bubble columns, operated in the recirculation-flow regime.
Figure 45 illustrates the plot of ( U , - Umf)/Ebversus ( U, - Umf)accord-
ing to Eq. (5-8) for fluidized beds. In the figure, Eq. (5-4)is also shown for
slugging beds. The mean gas holdup for the FCC-catalyst bed is taken
from Fig. 36. The plot for the FCC bed shows clearly that the bed is
fluidized smoothly, without slugging. The plot also shows data for a bed of
fluidized glass beads of mean diameter of 287 pm. (Data are taken from
W13; cf. Fig. 14). The bed behavior is seen to approach that of the slug-
ging bed as ( U G - Umf)increases beyond 20 cm/sec. Also, the average /3 is
approximately 0.8 for U, < 20 cm/sec, showing the possibility of bulk
recirculation of the emulsion.
The plot shown in Fig. 45 is useful in understanding the flow properties
of bubble-flow equipment. Also, it is quite probable that the Davidson and
Harrison relation (D4, P11) with p = 1 is for a fluidized bed with bulk
recirculation, although the recirculation pattern may be different.

c. BUBBLE
SPLITTING IN FLOW
TURBULENT
The mean slip velocity Us has been shown in Fig. 34 for the bubbles
observed in bubble columns or in fluidized FCC beds. Here the mean

D, = 7.55cm, 287 p m glass beads

air-fluidized (W-13 )

0' 10 20 30 40 50 0
( 0- &I 1 rcm / =I
FIG.45. (UG - Umf)/Zbas a function of (VG - Urn,)for fluidized beds. FCC-catalyst
beds show no indication of slugging.
 FLUIDIZED CATALYST BEDS 35 1

bubble size remains essentially unchanged as U, increases beyond 20

cm/sec, suggesting a dynamic balance of coalescence and splitting of
bubbles in the turbulent flow. Recently Yamazaki et al. (Y6) have ob-
served the bubble-size distribution in a FCC-catalyst bed (8.1 cm in
diameter with a quiescent bed height of 83 cm). They used a multichannel
hot-wire system with a resolving power of 0.5 cm in bubble size. Figure
46a shows how the bubble size distribution at a given bed section changes
with the height of the sectionz above the gas distributor (a canvas-coated
sieve plate with I-mm hole size). At the lower bed section, the distribution
is unimodal, but with increasing z it shifts gradually to bimodal distribu-
tion as a result of bubble coalescence. Figure 46b shows the volume-

&l I 11.0(cmlsac1
Z (cm)

----
- 38.0
57.0
............ 87.5

1 b ' 20 ' io 60 ' 80 ' 100 ' 120

Z (cm)
'
FIG.46. (a) Bubble size distribution at various heights z from the gas distributor. (b)
Volume-surface mean bubble diameter ds, as a function o f z. DT = 8.1 cm, U G = 1 1 cm/sec,
L , = 83 cm; air-fluidized FCC-catalyst bed (Y6).
352 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

surface mean bubble size calculated from the distributions; the mean size
becomes constant withz z 3OT.
Clift et al. (C6), in their study of bubbles in fluidized beds, indicate that
instead of having a discrete “maximum stable bubble size” we can expect
bubble splitting to occur over a relatively broad and continuous range of
bubble sizes. Whether or not a particular bubble splits will depend not
only on size but also on angular position, wavelength, and amplitude of
disturbances of the bubble interface. It seems likely that measured maxi-
mum stable bubble diameters correspond to mean diameters for systems
in which dynamic equilibrium has been achieved between coalescence
and splitting.
There are three main mechanisms which determine the bubble sizes:
coalescence of bubbles, splitting of a bubble under a given disturbance,
and the occurrence of disturbances in the bubble-flow equipment. The
latter two will be discussed in what follows. (Coaledcence is discussed in
C7, M10, T16, and W10).

1. Splitting of a Bubble
In 1950 G. I. Taylor (T8) investigated whether a small disturbance
superimposed on the interface between two immiscible fluids of different
densities was amplified or damped away when the fluids were accelerated
at a constant rate. Four years later Bellman and Pennington (B 10) devel-
oped this concept further to include the influence of liquid viscosity and
surface tension. Rice and Wilhelm (RS) applied Taylor instability to the
mechanism of bubble formation in a fluidized bed. Recently, Clift and
Grace (C7) have shown that the first stage of bubble splitting in a fluidized
bed is to develop an indentation in the bubble roof, which then moves
around the bubble periphery while growing to form a curtain of particles.
Splitting occurs if the lower edge of the curtain reaches the base of the
bubble before the top passes the equator. It was suggested that this phe-
nomenon results from instability of the type discussed by Taylor, where a
heavy fluid overlies a lighter one.
For a gas bubble in liquid, Henriksen and Ostergard (H7) have observed
that the bubble is broken up by disturbances created by the downward jet
of liquid. According to their observation, a finger of liquid projects down
from the roof and eventually divides the bubble in two. The reason that
bubble breakup in methanol is easier than in water is due to lower surface
tension. Based on these observations, they support the hypothesis of Clift
and Grace that the bubble is broken up as a result of the Taylor instability.
Linearized stability analysis has been performed (B 10) by superposing a
sinusoidal disturbance of wavelength A (which equals 27r/k, with k the
 FLUIDIZED CATALYST BEDS 353

wave number) on the interface. The disturbance is assumed to grow with

time 0 according to exp (no),where n is the growth factor. Clift et al. (C6)
extend the analysis to the fluidized bed, where the particles and fluid
comprising the dense phase are considered separately by treating the
dense phase as two superimposed, continuous fluids. If the viscosity and
density of the fluidizing fluid are very much smaller than the correspond-
ing quantities for the particulate phase (i.e., for gas fluidized beds) the
stability analysis is found to be identical with that by Bellman and Pen-
nington (B10) for real fluids with negligible interfacial tension and with the
upper fluid much more dense and viscous than the lower.
Predictions from these theories for a two-dimensional interface are pre-
sented in Fig. 47 for vM = 0.01-10 cm2/sec, (T = 0 and 60 dyn/cm for the
upper dense phase, where the lower light phase is air at room condition.
The curves for zero surface tension are for a fluidized bed, and the curves
for (T = 60 dyn/cm are for a bubble column of an aqueous solution.
In the case of zero surface tension, the analysis shows that the growth
factor n is always positive for all wavelength A, hence the disturbance is
always amplified. Increase in dense-phase kinematic viscosity results in a

WbMe phase :
density = 1.26 Xl@g/cm’

80
Continuous phase:
:U=O dyne/cm
: U=GOdyne/crr

: kinematic viscosit

0
wvelength : 1 = 2 it/ k [cm]

FIG.47. Growth factorn as a function of wavelength of disturbance, A , and the kinematic

viscosity of the continuous phase for zero surface tension (full curves) and for
u = 60 dynekm (dotted curves) (B10, C6).
354 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

reduction in the growth factor and an increase in the most sensitive wave-
length, at which n is a maximum. For the gas-fluidized bed, uM is the
apparent kinematic viscosity of the particulate phase defined ((26) by uM=
pp/ppese,where pup is the effective Newtonian shear viscosity of particu-
late fluid, pp the density of single particle, and eSethe volume fraction of
particles in the emulsion phase. Clift et al. (C6) have shown that the
interstitial gas velocity has virtually no effect on the stability of a bubble
roof. In their analysis the velocity is varied in the range of 1-100 cm/sec.
When the surface tension is finite, the disturbance tends to be sup-
pressed by the surface tension, the influence of which is noticeable due to
smaller curvature of the interface as the wavelength decreases. Hence
there is a lower limit for wavelength, below which the growth factor takes
negative values. The wavelength Amin where the instability is neutral is
(H7):
Amin = 2da/ptg)”2 (5-18)
For an air-water system at room temperature Amin equals 1.70 cm, and for
air-methanol Amin is 1.04 cm. The influence of surface tension on growth
factor is shown in Fig. 47, where the behavior discussed previously is
observed clearly.
As noted previously, disturbances initiated on the roof of a bubble are
swept around the periphery, so that in practice a bubble does not split
unless the disturbance has grown sufficiently before the tip of the growing
spike reaches the side of the bubble. Clift et al. (C6) have estimated the
likelihood of splitting by comparing the time required for a disturbance to
grow by a given factor with the time available for growth.
The required growth time may be provided by T, = l / n , where T, is the
time required for a small-amplitude disturbance to grow by a factor e.
Although the T, approach is no longer accurate for disturbances which
have grown beyond the scale described by the linearizeed analysis, the
estimate is retained in the absence of a better alternative.
An estimate of the time available for growth, T,, may be obtained from
the dense-phase tangential velocity r/$ (equal to r b &/do, with $ the
angle measured from the vertical) at the bubble-emulsion interface:

(5-19)

where the disturbance originates at $ = Thus T~ becomes large if the

disturbance originates very close to the nose of the bubble. Observations
of splitting bubbles suggest that disturbances usually develop on either
side of the nose.
Two cases have been studied. Case A has the bubble nose a node when
 FLUIDIZED CATALYST BEDS 355

the disturbance originates, and case B has a node located A/4 from the
bubble nose (so that the nose is an antinode in the initial disturbance).
With these the maximum time available for growth, T ~ , , is calculated from
Eq. (5-19). A bubble is liable to be split by a disturbance for which T,, >
T,; here, only disturbances with wavelength less than the arc length from
the nose to the equator, A IA,,, = m& are considered, since a distur-
bance with wavelength greater than,,,A represents a gross deformation of
the bubble rather than a perturbation on the interface.
Figure 48 shows the calculation for a bubble with a radius of 2 or 5 cm in
a gas-fluidized bed. In the bed with uM = 4 cmz/sec, the bubble withr, =
2 cm shows T ~ > , T, for A in the range 0.35-3. I cm, so that a disturbance in
this range may cause splitting. However, the same bubble in the bed with
uM = 10 cm2/sec always shows T,, < T,, and therefore should not split
whatever the wavelength of the disturbance. By contrast, the bubble of
5-cm radius in either bed is liable to be split by a broad range of A. The
most sensitive wavelength (minimum T,; point m in Fig. 48) does not
correspond to the A most likely to cause splitting; in some cases,,A is
shorter than the most sensitive wavelength. When the most sensitive
wavelength is within the range of possible disturbance, the ratio T,,/T~ is a
maximum for a wavelength less than the most sensitive.
An important conclusion of the above approximate Analysis is that the
effective kinematic viscosity of the emulsion is the dominant factor de-
termining both the initial growth of instabilities and the most sensitive
wave number. Thus, prediction of the effect of system properties on bub-
ble stability depends on prediction of the effect on the effective kinematic

FIG.48. Comparison of maximum time available for growth,ram,with time for a distur-
bance to grow by a factor e , re, for a gas-fluidized bed; disturbance is case B (after C6).
356 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

viscosity. Based on this concept, CliR et al. (C6) have suggested that
bubbles split more readly in a small-particle system and that fluidization is
more aggregative as the ratio pp/pg is increased. As for particle size dis-
tribution, Tsutsui (T27) has shown that the wide size distribution of a
standard FCC catalyst gives better fluidization than the FCC catalyst
having a very narrow size cut with the same equivalent mean diameter.
Geldart (G4, G5) indicates that fine particles of Group A (see Fig. 3) show
good fluidity. Better fluidity of given particles is primarily determined by
the volume-surface mean diameter of the particles and is not much influ-
enced by their size distribution, according to Geldart.
Basic understanding of the effective kinematic viscosity of the emulsion
is potentially important from the viewpoint of planning and operating
industrial fluidized beds. There are mechanisms which lead to the forma-
tion of very small bubbles; these are omitted here since the performance
of larger bubbles usually dominates the rate processes under high rates of
aeration.

2. Onset of Disturbances Effective in Bubble Splitting

As explained previously Henricksen and Ostergaard (H7) have ob-
served that a bubble in water or methanol is broken up by the disturbances
created by the downward jet of the liquid. Plate A in Fig. 49 shows a
bubble that is about to break up because of the disturbances created by
two liquid jets.
For a bubble to break up, disturbances should take place in the continu-
ous phase, induced by the interaction between bubbles and the continuous
phase. If the disturbances induced by an ascending bubble are almost
completely damped away before the amval of the next bubble, the suc-
ceeding bubble will hardly be broken up by the Taylor instability. If not,
the bubble is liable to be split by the residual disturbances. As the extent
of disturbance increases, bubbles split to smaller sizes and finally attain a
steady mean size and size distribution as a result of the dynamic balance
between splitting and coalescence (cf. Fig. 49). The decay of distur-
bances, however, depends on factors such as the strength of disturbances,
viscosity of the continuous phase, frequency of bubble passage, and col-
umn dimensions.
When bubbles are successively fed to the bubble column filled with
quiescent liquid, the first bubble will rarely be broken up due to the
absence of disturbances on the bubble roof. The bubble rises as a slug or a
spherical-cap bubble, depending on the gas flow rate and the column
diameter. However, the first bubble leaves disturbances behind so that
the succeeding bubbles will be broken up when the disturbances persist at
 FLUIDIZED CATALYST BEDS 357

FIG. 49. Splitting of bubble by fluid-dynamic disturbances: (a) from Henricksen and
Dstergaard (H7); (b) and (c) from T. Kikuchi. (b) Air-water system, DT = 8 cm, U , = 2.15
c d s e c , continuous bubbling. (c) Air-aqueous glycerol system, D, = 8 cm, CJ, = 4 cdsec,
continuous bubbling.
358 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

a certain high level; this is clearly observed in plate C of Fig. 49. Plate B in
the figure shows a large bubble about to be split by the turbulence induced
by continuous bubbling.
Figure 50 shows the influence of molecular kinematic viscosity v M of
liquid on the mean bubble size, when a 15-cm-diam. bubble column is
operated at the superficial air velocity of U, = 6.7-12.7 cm/sec (see data
in K10). The mean bubble size is seen to increase only slightly as uM
increases from 0.04 to 0.5-0.7 cm2/sec (aqueous glycerol solution); the
flow properties here are well into the recirculation-flow regime. As vM
increases beyond 0.7 cm2/sec the mean bubble size starts to increase
rapidly, transferring finally to the high-viscosity branch PQ of insufficient
breakup due to disturbances of low level. A similar phenomenon has also
been observed for a larger-diameter bubble column by Ueyama and
Miyauchi (U4),where a different viscosity is obtained for the flow transi-
tion effective to bubble splitting.
The mean bubble size that concerns us here is on the order of 5 cm, so
that the Eotvos number Eo (equal to dggpl/c+) is well over 40 for usual
bubble-column liquids. The bubbles are of spherical-cap type under this
condition, which is essentially equivalent to a Weber-number criterion We
(equal to dbpliii/(T) > 20, since U, = (H4, H5). The bubbles in a
fluidized catalyst bed satisfy the above criterion, since CT -
0. Conse-
quently, surface tension has relatively little effect, and instead the splitting
is closely related to disturbances induced by the bulk turbulence, the
intensity and the scale of which are mainly governed by the fluidity of the
continuous phase and the operating gas velocity.
Figure 51 shows the observed bubble diameter db in fluidized beds as a

30
20 -
___--_-
Column ----
diama

T
a ?
Id 5
0 UG: 6.7 - 12.7 cm/sec
A UG= 8.9 cm/ sec

dD1 2 34570.1 2 3 4 5 7 1 2 3 4 5 7 10 2 3
viscosity [poise]

FIG. 50. Change of mean bubble size with liquid viscosity under constant aeration.
Aqueous glycerol solution, D , = 15 cm, U c = 6.7-12.7 c d s e c . Data by Kimura (K10).
 FLUIDIZED CATALYST BEDS 359

2 8 S I r ' ' ' " " " 1

8 1 '

Observe; bubble diameter

lo

2-

U, , Terminal velocity [cm/sec]

FIG.51. Observed bubble diameter d b versus particle terminal velocity u,; data from
Horio and Wen (H21). Mean bubble size by Morooka et al. (M43) and that by Yamazaki
et al. (Y6)are added.

function of the particle terminal velocity ut. The data are mostly taken
from Horio and Wen (H21). Values of mean steady bubble sizes under
turbulent aeration are from Morooka et al. (M43) and Yamazaki et al.
(Y6). As the size of particles decreases while good fluidity is retained,
db tends to approach the steady bubble diameter d b s observed for bubble
columns of low-viscosity liquids (Fig. 34). In this domain wheredb = d b ,
dynamic equilibrium exists between bubbles of various sizes as a result
of turbulent motion prevailing in the emulsion; there is a steady mean
bubble size here rather than the maximum stable bubble size.
The maximum stable bubble diameter dbms defined by Davidson and
Harrison (D3) seems to apply for an emulsion of large particles, where the
turbulent motion is weak. Bubbles tend to coalesce with each other to
grow ultimately to dbms. The observed bubble diameter db is closely re-
lated to bubble splitting and coalescence as a result of turbulence. The
data in Fig. 51 are expressed by an approximation that db is the sum of dbs
and dbmsor
db = dbs + dbms

= dbs + (~,/0.711)'/g (5-20)

360 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

Scientific approaches to improve bed fluidity are potentially important

for fluidized bed technology. Also, further quantitative relations between
bubble splitting and bed properties would be very helpful in planning and
scaling-up fluidized catalyst beds.

VI. Heat and Mass Transfer in Fluidized Catalyst Beds

Heat and mass transfer constitute fundamentally important transport

properties for design of a fluidized catalyst bed. Intense mixing of emul-
sion phase with a large heat capacity results in uniform temperature at a
level determined by the balance between the rates of heat generation from
reaction and heat removal through wall heat transfer, and by the heat
capacity of feed gas. However, thermal stability of the dilute phase de-
pends also on the heat-diffusive power of the phase (Section IX). The
mechanism by which a reactant gas is transferred from the bubble phase
to the emulsion phase is part of the basic information needed to formulate
the design equation for the bed (Sections VII-IX). These properties are
closely related to the flow behavior of the bed (Sections 11-V) and to the
bubble dynamics.

A. BUBBLEDYNAMICS
Although gas bubbles ascending through the emulsion of fine catalyst
particles are constantly splitting and coalescing (Sections 11, 111, and V),
they are largely free of particles (H14, K13,T19). Such a bubble, may be
pictured as essentially spherical, with the lower of its volume occupied
by particles (Fig. 52). As the bubble rises, it displaces solid particles
around it and cames some particles upward in its wake. Its rise velocity ii,
is proportional to the square root of its frontal diameter. If ii, is lower than
the velocity of the interstitial gas umf= Umf/cmf, the gas from the emulsion
enters the lower part of the bubble and leaves through the roof; however,
for small particles, 1, is usually larger than umf,and the circulation pattern
as shown in Fig. 52 is encountered. Gas circulates up the central core of
the bubble and, upon emerging from the roof, encounters drag from the
particles streaming past. Consequently this gas is swept downward rela-
tive to the bubble and is re-entrained at the bottom because of the prevail-
ing pressure gradient there.
It follows that as the bubble rises more quickly, the gas which emerges
from its roof will penetrate outward a smaller distance before being swept
downward; i.e., the limit of penetration will be nearer to the bubble wall.
The total volume enclosed by the limit of penetration is called the cloud,
 FLUIDIZED CATALYST BEDS 36 1

yAxis 01 symmetry

FIG.52. (a) Gas and particle flow pattern near a bubble, a = 2.5 and (b) a photograph
showing cloud size re and bubble size rb by using NO,-containing-bubble technique,
a = 2.5, 230 pm ballotini (after R17).

and the region between the bubble-cavity wall and the limit of penetration
is called the cloud overlap. This region and part of the wake are the only
regions where the gas in the bubble can contact particles.
These physical pictures of bubble dynamics have been developed by
Davidson and Harrison (D3), Murray (M46, M47), Pyle and Rose (B),
Rowe et al. (R17) and others. Jackson (Jl) considered that a mantle
of bed with increased voidage exists near the roof of the bubble. Most of
the work on fluid-mechanical theories of aggregative fluidization have
been reviewed by Jackson (in DS).
The radius of the cloudr, has been given by Davidson and Harrison (D3)
for a = &/Urn, > I :
rc/rb = [(a + 2)/((r - i ) y 3 (6-1)
Alternatively, Murray (M47) proposes for (Y > 1:
(a - l)(rc/rb)4 - a(rc/rb)= cos 8 (6-2)
The experiments of Stewart (S20) show (Fig. 53) that the radius of an
observed bubble cloud at 8 = 0 falls between Eqs. (6-1) and (6-2). Figure
54 shows the thickness of the cloud-overlap region estimated by Eqs. (6-1)
and (6-2) for a fine microspherical catalyst fluidized by ambient air, where
the minimum fluidization velocity Urn,is calculated by Wen and Yu's
equation (W4). The cloud-overlap region is seen as being limited to at
most a few layers of fine particles (M30). Rieke and Pigford (R8) observed
experimentally for a two-dimensional fluidized bed that the gas in the
 txprrin
El Jackror

FIG. 53. Ratio of cloud radius rc to bubble radius rb on vertical axis above three-
dimensional bubbles. Comparison of experimental values with various theoretical predic-
tions. The curve “Jackson (modified)” refers to resultsobtained by Jackson’s type of analy-
sis, but with the Davies-Taylor type rise velocity k v g r , , where k is chosen to give the best
fit to the experimental rise velocity (from S20 and D5).

2 3 4 5 678910
db (Em)

FIG.54. Thickness of gas cloud. pp = 1.0 g/cm3, emf= 0.5, fluidized by ambient air.
 F L U I D I Z E D CATALYST BEDS 363

bubble wake does not return to the bubble cavity while circulating. This is
in conformity with the observation by Rowe et al. (R17); see Fig. 52. As a
consequence, the observed gas-flow patterns suggest that there is no di-
rect gas communication between the wake and the cloud-overlap region.
Progress in understanding the bubble dynamics has made it possible to
formulate transport equations for heat and mass transfer through the bub-
ble wall. which we now consider.

B. HEATA N D MASSTRANSFER
THROUGH THE BUBBLE
WALL
Most of the experimental work so far has been concerned with the
performance of an isolated bubble; few publications are available on
fluidized catalyst beds under a high aeration rate. In 1961 van Deemter
(Vl) analyzed the transient and steady-state gas mixing data by Mason
(given in V1) taken in a 7.6-cm-diam. cracking catalyst bed. Van Deemter
based his calculation of the overall mass transfer coefficient kebab on the
two-phase theory of fluidization. Further, axial mixing of gas and solid
particles has been studied for fluidized catalyst beds by de Groot (D7),
Botton (B13), van Swaay and Zuiderweg (V8), and Morookaet al. (M42).
For crushed silica fluidized at ambient temperature and pressure in beds
with diameter 0.3-1.5 m, van Swaay and Zuiderweg (V8) have reported
the apparent HTU:

(Hob)app -
- a = (1.8 - *)(3.5
k uG
ob b OFZ5
- -)LPZ5
2.5 (6-3)

where (&,)app, DT, and Lf are in meters. This relation is expected to apply
for D, and Lf up to 10 m, with 15% fines content of the solid (d, < 44 pm).
The use of microspherical particles with a larger fraction of fines showed
somewhat better mass transfer; but the dependence of HTU or of k o b a b / &
on bed dimensions was the same, as shown in Fig. 55. The coefficients are
less dependent on gas velocity and bed height for particles with good
fluidity. The data by de Groot (D7) indicate a doubtful effect of bed height,
in contrast to Eq. (6-3).
Morookaet al. (M42) studied the effect of gas adsorption on mass trans-
fer between bubbles and emulsion phase, while measuring the residence
time distributions of tracer gases for helium, carbon dioxide, Freon 12,
and Freon 22 in 12- and 19-cm-diam. free and baffled beds of cracking
catalyst. The k O & values are calculated from the distributions for differ-
ent gases according to the two-phase model, providing for the adsorption
equilibrium.
The quantity kebab increases with increasing superficial gas velocity VG;
it shows a higher value when gas is more strongly adsorbed by the parti-
364 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

6 --' II '' I I 1 I I IIll I

key f f [mi tracer
I
investigator
1 I I 1 1 1 1 -
-
n 4- 0 0.9 He vanDeemtcr(V1) -

-
V
Q, - @ 2-2-4.9 H2 de Groot (D7) -
A 1-3-3 H2 Botton(B13)
-
v)
\
c
2-.
0 v 1.6-3.4 H2 Botton(B13)

I
I I 1 I I l 1 1 l I I I 1 I l l 1 1 I I I l l
2 4 6 8 2 4 6 8 2 4 6 8
1 10 1 o2
m, C - 3

FIG.56. Effect of adsorption equilibrium constant on k o f l b / & G in free and multistep

fluid beds (after M42).
 FLUIDIZED CATALYST BEDS 365

I 1 I
20'C'\
$5
4
\
I
i8.I
10 15 20
Fraction of fines ( < 4 4 p m ) '10

FIG.57. Effect of fines on kebab in fluid bed (data taken from D8).

cles (rn larger). The effect of increasing gas velocity or adding horizontal
baffles is mainly to increase the mean bubble holdup in the bed.
The group k & h / E h m is plotted in Fig. 56 as a function of the gas
partition ratio m,. Obviously k,,h increases with m,,owing to decreased
particle-side mass-transfer resistance. Data of Yamazaki and Miyauchi
(Y4)are plotted in Fig. 56 and match well as an extension of the Morooka
data. Using a quick-response thermoelement, Yamazaki measured the
temperature distributions of the bubble and emulsion phases, from which
the overall heat-transfer coefficient is calculated for 7.9- and 19-cm-diam.
cracking catalyst beds. Since the heat capacity of the emulsion phase is
much higher than that of the bubble phase, the overall coefficient is nearly
equal to the film coefficient h h a h for the bubble side. The heat-transfer
coefficient reduces to the mass-transfer coefficient according to the equal-
ity h h a h / C p g & = k b a h .
The effect of added fines content in increasing k&h, measured by de
Vrieset al. (D8), is shown in Fig. 57. Other work on heat and mass transfer
for teeter-type beds is reviewed by Kunii and Levenspiel (K24) and in
Davidson and Harrison's book (D5).

C. MASSTRANSFER
BETWEEN BUBBLE
A N D EMULSION
PHASE
Gas exchange during catalytic reaction between bubbles and the emul-
sion of catalyst particles is essentially a diffusional phenomenon, with
simultaneous adsorption and reaction in the emulsion. Based on this con-
cept, the overall mass transfer has been analyzed by Miyauchi and
366 M I Y A U C H I , F U R U S A K I , MOROOKA, A N D I K E D A

Morooka (M30), Chiba and Kobayashi (C3), and Drinkenburg and

Rietema (D17). According to this approach, the bubble-side film coeffi-
cient khahtends to provide the main resistance to mass transfer when gas
adsorption onto the particles is strong and the reaction rate constant is
high. The following theory is by Miyauchi and Morooka (M30).

1. The Underlying Concept and Its Formulation

As stated in Section VI,A, bubbles ascending through the usual fluid
catalyst bed show a sufficiently large a (equal to &/Umf)to make the
thickness of the cloud-overlap region very thin, as illustrated in Fig. 54.
Consequently the streamlines for gas and particles are essentially parallel
to the bubble surface. Convective bulk flow of “cloud” gas across the
bubble interface is negligible. Hence it is assumed that only the portion of
cloud gas which ascends along the bubble axis region flows into the roof of
the bubble at a volumetric flow rate qf, then descends through the cloud-
overlap region and flows out of the bottom of the region to recirculate
again into the bubble void (see Fig. 58). The recirculating gas flowing
through the cloud-overlap region exchanges the reactant gas by diffusion
from the bubble void to the bulk emulsion phase.
With a simplifying assumption that local adsorption equilibrium is in-
stantaneously attained between gas and particles, the mass-transfer pro-
cess is expressed by Fig. 58 for the cloud-overlap region. Here the influ-
ence of bubble wall curvature is neglected, since the region is very thin.
When no catalyst is suspended in the bubble void, the equations of con-
tinuity for the reactant gas are as follows (M30):
ac,lae = D~ a 2 C b l a $ (6-4)

fqt m&c,
=bob cb
CC
4 + C.0
c3 diffusion

bubble porticle emulsion

void -cloud phase
overlap

FIG.58. Schematic diagram of mass transfer near a bubble in fluid bed (M30).
 F L U I D I Z E D CATALYST BEDS 367

for the bubble void;

m ac,/ae = Deffa2c,/ax2 - ~TC, (6-5)
for the cloud-overlap region;
m dc,/de = D,, a2celdx2- krce (6-6)
for the emulsion phase, where m = Efe + m s E s e , with m, = (Cs/Cf)equilibrium.
c, is moles of adsorbed gas per unit volume of particles including intra-
particle pore volume. Also, a first-order irreversible reaction is assumed
to take place in the particulate phase.
The initial and boundary conditions are as follows:
for e = 0,
cb = CbO, Cc = Cco, Ce = Ceo

for e > 0,
x= -ma . ch = ChO

x = 0: -DG ach/dx = -Deff ac,/ax, cb = C,

= 6: -Deff ac,/ax = - D , ~acelax,

~ C, = C, (6-7)

x = +m: c, = ceo
The initial concentration of the cloud-overlap region c,, is obtained from
the reactant gas balance taken at the roof of the bubble, where the catalyst
particles are fed to the cloud-overlap region from the emulsion above at
volumetric flow rate q s :
CbOqf + 4qsceO = (qf +q d c C O (6-8)
When a >> 1 , the slip velocity of particles relative to the gas is negligible
in comparison with the particle velocity, so that qf/qe is nearly equal to
q s / E s , . With this equality Eq. (6-8) is modified to the form:
CcO = (EfecbO + %%eceO)/m (6-9)
The concentration distributions in each phase and the mass flux across
the bubble surface have been solved to satisfy Eqs. (6-4)-(6-6) under the
restrictions of Eqs. (6-7) and (6-9) (M26, M30). The overall mass-transfer
coefficient between the bubble cavity and the emulsion phase kob has been
given as follows:
(6- 10)
368 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

The film mass-transfer coefficients for the bubble and the emulsion phase
( k b and k,, respectively) are given by

kb = (2/7T"2)(D~/Pb)112 (6-12)
ke = ( 2 / T '/')( mDeff/Tb)' (6-13)
where Tb is the time necessary for local surface renewal; p is the Hatta
number (or reaction factor) for unsteady gas absorption with a first-order
irreversible reaction, as given by Danckwerts (Dl):

p = (mH + k)erf( 3)+ exp( - '9) (6-14)

with m H = (krDeff)'12/ke.J in Eq. (6- 1 l), associated with the cloud-overlap

region, is plotted in Fig. 59 and conforms to the relation:

J =zerf(
4mH
$) - l*& exp[ - (4mrr:, x +E)]dx
X
(6-15)

where Pe = ma2&/4dbD* J is approximated accurately enough by

the following equation for 0 5 Pe 5 0.1 and 0 5 mH 5 2, the usually
encountered domain for fluidized catalyst beds of fine particles:
J = (rr Pe)'j2 - Pe - Pe0.84
0.454mH1.40 (6-15a)
The physical data needed to compute kob are DG, m, k,, Umf,db,
and ere (efe can be approximated by emf). With the Higbie penetration
theory for bubbles, t b included in Eqs. (6-12) and (6-13) is approximated
by db/fis, so that Eq. (6-10) is modified to

(6-16)

083
- 1.0

-
+ 1.3

-
1.5

q
IS
L 20

2 40
35

Pe
FIG.59. Integral J as a function of Pe and m H (M26).
 FLUIDIZED CATALYST BEDS 369

Here, the tortuosity factor x is approximately 1.5, as given by Hoogscha-

gen (M23). When Pe 5 0.01 and mH 2 3, Eq. (6-11) reduces to
PI.= P = m H . Consequently, Eq. (6-10) is modified to
(kob)-' = (kb)-' + (krDeff)-'" (6- 17)
kob is independent of ke under these circumstances.
The gas flow velocity through the emulsion phase is close to the mini-
mum fluidization velocity Urn,.When the particles are spherical and have
diameters of several tens of microns, this flow condition gives a quite
small particle Peclet number, d,Umf/DG.For example, the Peclet number
is estimated as 0.1-0.01 when 122-pm-diam. cracking catalyst is fluidized
by gas, with U,, = 0.73 cm/sec and D, = 0.09 cm2/sec;and it is estimated
as 0.001-0.01 for 58-pm-diam. particles, with Urn,= 0.16 cm/sec. The
mechanism of mass transfer between fluid and particles in packed beds is
controlled by molecular diffusion under such low Peclet numbers, and the
particle Sherwood number kfd,/DGis well over 10 (M24). Consequently
with intraparticle diffusion shown to be negligible (M21), instantaneous
equilibrium is established to be a good approximation [see Eq. (6-24)].

2. Comparison with Other Theories

Davidson and Harrison (D3) have assumed that both convective and
diffusional flows contribute to gas exchange between bubble and emulsion
phases. Partridge and Rowe (P2) have developed a boundary-layer equa-
tion for the transfer from cloud to emulsion phase. Kunii and Levenspiel
(K22), in their model for estimating the gas-exchange coefficient, assume
two transfer steps-between bubble void and cloud-particle overlap re-
gion, and between cloud-particle overlap region and emulsion phase. The
mass-transfer coefficient for the former step is assumed of the form given
Davidson and Harrison, and for the latter step of the form of the Higbie
penetration model. In terms of mass-transfer coefficients, the equation of
Kunii and Levenspiel is:
(kob)-' = (kbc)-' + (kcej-' (6-18)
where
kbc = +Urn,+ 0.975D&'2(g/db)''4 (6-19)
kce = ( 2 / r '/')( E&fDGfib/db) (6-20)
where iib is given by Eq. (5-8), with p = 1 and Us = Us,,.
Chiba and Kobayashi (C3) have presented the following equation,
based on the Murray stream function (M47):
370 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

They further introduce the equilibrium influence of physical adsorption on

to the particles. In terms of the present notation and with replacement of
emfby the nearly equal term Efe, their relation is written:
(6-22)
where
r) = [I + $[I - (~,/m)](a - 1)-1]1/2 (6-22a)

The relation between proposed equations (6-10) and (6-18) merits

discussion. When pp= 1 (no reaction) and m = Emf (no partition to parti-
cles), k o b as calculated by Eq. (6-10) or (6-16) is smaller than that by Eq.
(6-18) for fine particles, since the former takes Us for bubble velocity and
the latter uses iib, which is larger than Us. Equations (6-20) and (6-13) are
generally different because m and Emf are not necessarily equal. The
bubble-void resistance to mass transfer has been assumed negligible in
Eq. (6-21). This equation is rendered applicable to the case of arbitrary m
by utilizing Eq. (6-22). Equation (6-13) can be rewritten in a form similar
to Eq. (6-22), with the observations that m = eye + ntsese, and Ere = Emf:

The correction factor r ) to Eq. (6-22) is nearly equal to unity for (Y > 10, so
that Eq. (6-22) is almost equivalent to Eq. (6-23) for smaller particles.
Equation (6-21) is independent of mH or rate of chemical reaction.
Drinkenburg and Rietema (D 17) have presented a numerical computa-
tion of kob based on the stream functions given by Davidson and Harrison
(D3) and by Murray (M47). The bubble-void resistance to mass transfer
has been neglected. Enhancement of gas transfer rate by diffusion with
simultaneous chemical reaction (Fig. 5 of D17) is reasonably well ex-
pressed by Eq. (6-11), the enhancement being expressed as the Hatta
number. Enhancement by physical adsorption (Fig. 2 of D17) is also ap-
proximated by Eqs. (6-22) or (6-23) for smaller particles.
When the reaction data by Lewis et al. (L12) are analyzed by utilizing
Eq. (6-lo), it turns out that PI.amounts to 1.45 (free-flow bubbles) for the
highest reaction rate constant. Hence it seems reasonable to take into
consideration the influence of pr on kob. Also, it has been shown expen-
mentally that k b is not necessarily negligible for fluid beds (Y4).
If the adsorption equilibrium is not attained instantaneously, a different
analysis is needed. Toei et al. (T18) studied the mechanism of heat and
mass transfer between bubbles and emulsion phase under such circum-
stances. The dependence of diffusion rate on bulk flow across the bubble
interface also becomes important when coarse particles are fluidized
(H16). For two-dimensional bubbles Chavarie and Grace (C7a) compared
various interphase mass-transfer models.
 F L U I D I Z E D CATALYST BEDS 37 I

3. Applicability of the Theories

Many experimental tests have been made on the theories previously
given, although most workers have fluidized particles coarser than the
typical cracking catalyst under relatively low gas velocities. When the gas
component is not adsorbed, the observed mass-transfer coefficients [see
Kunii and Levenspiel (K22, K24), Chiba and Kobayashi (C3), and Drink-
enburg and Rietema (D IS)] agree with those obtained by the theories.
Miyauchi (M26, M30) has shown that the influence of chemical reaction on
k o b is expressed adequately by Eq. (6-16). Under relatively high aeration
rates Morooka et al. (M42) measured the bubble diameter (cf. Section
II,B) and calculated kob and k o b a b / & d & from Eq. (6-16). AS shown in
Fig. 56, the prediction is in good agreement with the experimental results,
when helium is used as the tracer gas. However, the prediction form, > 1
is much smaller than the observed data, though the adsorption equilibrium
is taken into account in Eq. (6-16). Nguyen and Potter (N2) mention the
same effect, but details are not reported.
The reason that kbabis higher than calculated from Eq. (6-12) may be
explained qualitatively by three effects: ( 1) splitting, coalescence, and
rupture of bubbles (T18, T20); (2) direct contact of gas and particles in the
transition zone from dense phase to dilute phase (F18); (3) the influence of
the particle capacitance effect (M21, M22) as a result of a small steady
interchange of particles between the bubble void and the emulsion. An
example of this is the case where particles are raining through the bubble
(D18, R8, W1) and (4) asphericity of the bubbles (D18). If the particle
capacitance effect (discussed in the next section) is responsible for high
experimental values for kbab,such values should not be applied to the
usual catalytic reactions, where m is on the order of unity and particle
capacitance has little effect on kba,,. For design purposes it is normally
better to use experimental mass-transfer coefficients obtained by a prop-
erly sized fluid bed for the reaction system of interest.

D. PARTICLE
CAPACITANCE
EFFECT
As is well known, the temperature in a fluidized catalyst bed is nearly
uniform even when considerable heat is generated in the bed (G8, L11).
This comes from the bed’s large content of solid particles, which disperse
the heat axially as a result of particle heat capacity. Similarly, the gas
concentration in the bed becomes uniform if the partition of transferring
gas component is favorable to the particle phase. The circulating particles,
holding a large amount of transferring component, iocally exchange the
component with the surrounding gas phase; the larger the partition ratio is
of the component, the higher the rate of concentration equalization in the
gas phase of the bed. Enhancement of heat and mass transport in the bed
372 M I Y A U C H I , F U R U S A K I , MOROOKA, A N D I K E D A

as a result of particle movement is named the particle capacitance effect; a

general formulation has been presented by Miyauchi (M21, M22).
When each catalyst particle is in adsorption equilibrium locally with the
gas phase, the axial-dispersion term of the emulsion phase gas in the equa-
tions of continuity is (Eef + GE,,) a2cfe/az2,where E,, and E,, are the
dispersion coefficients in the z direction for gas phase and particles, re-
spectively. The emulsion of fine particles entrains interstitial gas without
appreciable relative motion, giving Eef = EfeE, and E,, = E,E,, where
E , is the dispersion coefficient of the emulsion in the z direction. As a
consequence, the axial-dispersion term is expressed by ( E f , + msEse)E,
d2cfe/az*or m E , azcfe/a2z.In other words, as a result of the capacitance
effect, the axial dispersion coefficient of the adsorptive gas is now mE,,
which is much larger than Eef when m >> 1.
Essentially the same enhancement takes place for the transport term by
convection. That term is ultimately given by -mu, acfe/dzrwhere ue is the
mean velocity of the emulsion in the z direction.
When the particles move relative to the gas phase, as when they rain or
shower through a gas bubble, they can transport the adsorptive gas rap-
idly to or from the emulsion phase. The same is true for heat transport
(see Section VI,C,3).
The time constant T, for a particle of radius rp placed in the emulsion
phase approaching the adsorption equilibrium with the surrounding gas
has been estimated (M21) to be (in cm . sec):
r, = r2,m/20De (6-24)
where D, is the effective diffusivity of adsorptive gas in the particle. For
r p = 30 pm, when m = 10, T~ is on the order of 10-3-10-4 sec, so that local
partition equilibrium is well established in the usual fluidized catalyst
beds. This has been shown experimentally by measuring the mean resi-
dence times of adsorptive tracer gases in a fluidized cracking catalyst bed
(M39). In the work of Nguyen and Potter (Nl) with coarser particles,
adsorption equilibrium was not attained. Gilliland and Mason (G9) men-
tion the influence of adsorption on the bed transient response early in the
1950s.
Few studies have been performed on the physical adsorption of reactant
gases at higher temperatures. Eberly (El, E2), Eberly and Kimberlin
(E3), and Tam and Miyauchi (T4) measured the adsorption equilibria of
hydrocarbons on catalysts. The last-named authors have obtained high-
temperature equilibria of benzene and nine normal paraffins (CSHI4to
CI4H3,,)for a silica-alumina cracking catalyst at 150-450°C. Their results
suggest that a fair amount of hydrocarbon is physically adsorbed on the
cracking catalyst in high-temperature operations.
 FLUIDIZED C A T A L Y S T BEDS 373

The particle capacitance effect in bubble mass transfer was shown ear-
lier (see Section V1,C) for the streaming emulsion outside a bubble, so
that rn is included in the emulsion-side film coefficient k,. Chiba and
Kobayashi (C3) and Drinkenburg and Rietema (D17) introduce the same
effect. Yokota et al. (Y13), in studies of mass transfer from submerged
surfaces in a gas fludized bed, find that the rate is strongly enhanced by
adsorption of the transferring gas component on fluidized particles. The
role of the particle capacitance effect in heat transfer has been discussed
by Mickley and Fairbanks (M14), Mickley et al. (M16), Drinkenburg
(D16), Yoshida et al. (Y18), and Kunii and Levenspiel (K24).

E. AXIALA N D RADIAL
MIXING
OF HEATA N D MASS

1. Overall Mixing of Nonadsorprive Gas

The extent of longitudinal mixing of nonadsorptive gas has been mea-
sured by Gilliland and Mason (G8,G9), Mason (in R4), Stemerding (in
R4), May (M12), Overcashieret al. (Oll), De Maria and Longfield (DlO),
Miyauchiet al. (M33), Muchiet al. (M49), and Ogasawaraet al. (012). All
these experiments were carried out under relatively high gas velocities of
practical interest. The mixing process in the beds is agproximated by the
one-dimensional diffusion model:
a c p e = E, aZc/azz - u, ac/az (6-25)
where c is the mean concentration of gas, E , is the superficial longitudinal
dispersion coefficient of gas, and UGis the superficial gas velocity.
Experimental values for E , available so far for fluidized catalyst beds
are summarized in Fig. 60, where the full curve has been calculated by
combining Eqs. (6-31) and (6-34) (see Section VI,E,3). In obtaining E,
experimentally there have been two kinds of methods, one of which
utilizes a step response and the other a steady backmixing of a nonadsorp-
tive tracer gas. In Fig. 60,the data obtained by the step-response method
(solid keys) are seen generally to agree with the calculated curve, where
the response curves by Gilliland and Mason (G9) and that by May (M12)
are treated by the step-response theory of Yagi and Miyauchi (Yl) to
obtain EJU,. The same quantity obtained by the steady back-mixing
method (open keys) is seen to scatter considerably, giving lower E,/ U, for
smaller DTand higher values for larger DTwhen compared with the calcu-
lated curve. Gilliland and Mason (G9) have indicated the inadequacy of
the steady back-mixing method for measuring axial-mixing of gas, al-
though the method suggests information as to flow properties of the beds.
374 M I Y A U C H I , F U R U S A K I , MOROOKA, A N D I K E D A

FIG. 60. Longitudinal dispersion of gas in fluid beds (after Miyauchi). The full curve
is calculated from Eqs. (6-3 I ) and (6-34) for He-air system.

Other works on overall axial and radial mixing of nonadsorptive gases in

fluidized beds have been reviewed by Kunii and Levenspiel (K24) and
Potter (in D5).

2. Mixing of Solid Particles

Axial mixing in fluidized catalyst beds has been explained in Section IV,
where bed particles of good fluidity are shown to give a fair amount of
dispersion as a result of the Taylor dispersion mechanism. Porous parti-
cles with diameter larger than about 150 p m or heavy particles such as
glass beds show considerably different flow properties from the fluidized
beds of good fluidity. Morooka et al. (M40) find that particle-caused
dispersion approaches total dispersion, as size and density decrease and
as the size distribution broadens. They explain the deviations of data
from Shrikhande (S12) and de Groot (D7) from their correlation, which
 FLUIDIZED CATALYST BEDS 375

applies mainly to broad size distributions of porous particles smaller than

100 p m in mean diameter and to equal-sized glass beds smaller than about
20-30 pm.
De Vries et al. (D8) have studied the effect of fines content on the axial
dispersion coefficient of the particles. As shown in Fig. 61, good fluidiza-
tion is obtained when at least 10% of fines (-44 pm) are present. Axial
dispersion of particles has also been studied by other workers (L8, M5,
P5, S6, T3, W14). Lateral dispersion of particles is covered by studies
performed under relatively limited conditions (B16, G1, H15, L l l ,
M37).

3. Axial Dispersion of Fluid with Particle Capacitance Effect

The overall dispersion coefficient of gas has been observed to increase
when a more adsorptive gas is utilized as a tracer for transient response
under constant aeration (Stemerding in R4; M33, M42). Figure 62 shows
such examples taken for a fluidized cracking catalyst bed (M33), where
the partition ratios of adsorptive gas m (equal to efe + m,e,,) are measured
as 0.6, 4.5 and 10 for hydrogen, carbon dioxide, and Freon 12, respec-
tively. Enhancement of gas dispersion by adsorption has also been
studied by Yoshida and Kunii (Y15),Yoshida et al. (Y17),Yates and
Constants (Y9), Zalewski and Hanesian (Z3), and Nguyen and Potter
(W.
Overall longitudinal dispersion of an adsorptive gas is obtained by solv-
ing the interaction between several rate processes and the particle capaci-
tance effect. Morooka et a f . (M39) have analyzed the behavior of the
fluidized catalyst bed with recirculation flow as given in Fig. 2. The
simplifying assumptions for modeling the bed behavior are essentially the
same as in Section VII,B,2, where the influence of directly contacting

I- / /
0
.0l
V
- 8
0 5 10 15 20 25 30
% FINES

FIG.61. Longitudinal dispersion coefficient of solid particles as a function of size fraction

<44 pm. D T = 30 c d s e c , L , = 2-2.5 m, and superficial gas velocities are: (0) 10 c d s e c ,
( 0 )20 c d s e c (DW.
376 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

Uf Ccmlsec3

FIG.62. Influence of particle capacitance effect on overall longitudinal dispersion coeffi-

cient of gas (M33).

particles with bed gas has been neglected, since the influence is usually
small for transient response of tracer gas at a moderate partition ratio.
The simplifications taken here are, first, that the catalyst particles are
usually several tens of pm in mean diameter, so that the interstitial gas
velocity through the emulsion phase is about umf(equal to Umf/Emf) and is
neglected in comparison with UGand emulsion-phase recirculation veloc-
ity. Also, the particles and the gas in the emulsion are in local adsorption
equilibrium the partition ratio m (equal to efe + mseSe)applying for unit
volume of the emulsion phase. Second, the central upflow and the periph-
eral downflow recirculate in the bed as ideal plug flows at approximately
the same linear velocities. Furthermore, the bubble phase ascends only
through the central upflow emulsion.
With these simplifying assumptions, the equations of continuity are
given for transient response of a tracer gas as follows (M38):
Eb &,/a6 = -uc ach/az - kohah(Cb - cfeu) (6-26a)
for the bubble phase;
mieu acfeu/a8= -mu, acfe,/dz + kohab(Ch - cfeu)
-mkexaex(cfeu - CfecJ (6-26b)
 FLUIDIZED CATALYST BEDS 377

for the centrally ascending emulsion phase;

mZed dcfed/ae = mu, dcted/az + mk,,a,,(cfe, - cfed) (6-26c)

for the peripherally descending emulsion phase, where k & a b is the overall
coefficient of mass transfer between bubbles and emulsion, and kexaexis
the exchange rate coefficient between upflow and downflow portions of
emulsion phase as a result of radial mixing (Fig. 2). The ie,and i e d are the
volume fractions of ascending and descending emulsion phases, related by
the equation ieu +
Zed = E,. U, = i,,~,,is the superficial circulation veloc-
ity of emulsion phase, whereu,, is the interstitial velocity of the ascending
emulsion. Also Ue = & ued, as a result of continuity of the recirculating
emulsion [see Eq. (3-26)l.
The initial and boundary conditions to solve Eqs. (6-26) for an impulse
response of the adsorptive gas are:
8 < 0: c b = cfe, = Cfed =0
e 2.0: x = 0; cb = S(e), cfeu = cfed (6-27)
x = Lf; cfeu = Cfed

The variance of the residence times of the tracer gas with respect to
mean residence time is obtained from Eqs. (6-26) and (6-27), according
to van der Laan (V3), as follows:
0 2 - 1 mie 1
2- (ib+ mi,) + N,,
-(Nv 2 ( i b + mi,)
(6-28)
+

N2, (A, - A,)(l - e-”)(l - e+)

(e-hi - e-hz)

where Ne, = mkexaexLf/UG,Not,= k,i,abLf/U,, N, = mUe/UG,and

Xi = Nob(1 -I-~ / N V ) ( I p)/2, N o b ( 1 -k 1/Nv)(1 - p ) / 2
A2 =

p = [I 4NeX/(I N~~)N,,b]’’*
Here, we note that the influence of particle motion on an impulse response
has been treated by van Deemter (V2), Yoshida and Kunii (Y15), Fryer
and Potter (F10, F1 I), and Nguyen and Fryer (N2), but their models are
all different from the one treated above.
Morookaet al. (M42) further solved the impulse response of a tracer gas
for a fluidized catalyst bed according to the one-dimensional two-phase
diffusion model (Vl), where the influence of the particle capacitance effect
was considered under the assumption of local adsorption equilibrium. The
equations of continuity for the tracer gas are:
Eb acb/ae = - u, acb/az - kobab(Cb - ce) (6-29a)
378 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

for the bubble phase;

mge dC,/de = E, d2Ce/dZZ kobab(Cb - C,) (6-29b)
for the emulsion phase, where E, is the apparent longitudinal dispersion
coefficient of emulsion phase. Since the slip velocity of gas through the
particles is negligible for the emulsion phase of small particles, E, is con-
nected with E, [cf. Eqs. (4-16) and (4-17) and Section VI,D] with reason-
able accuracy:
E, = mGeEZs (6-30)
Under the same boundary conditions as given by van Deemter (Vl) the
variance of residence times is given by

(6-3 1)
where Nob = kobabLf/UG, (PeB), = U,L,/mZ,E,, and
= [l + 4(PeB),/Nob]1’2, A = 4No,/2
For the present simplified recirculation model, the axial dispersion
coefficient is given by the following approximate equation for arbitrary m :
( m Ue)2/mkexaex
= Ee (6-32)
This relation is derived by considering a steady-state tracer exchange
between the ascending and descending emulsions and the bubbles, and
comparing the resulting equation with that of the diffusion model. Since
E, = mEeEZs,the above equation is equivalent to the following relation
from van Deemter (V2):
Glkexaex = EeEs (6-33)
Once the variance of the residence time distribution has been obtained,
the axial dispersion coefficient of the tracer gas E , is calculated by combin-
ing the following equation for the one-dimensional diffusion model (V3)
with Eq. (6-28) or Eq. (6-31):

where (PeB), = U G L f / E , the

, column Peclet number for the tracer gas [cf.
Eq. (6-25)]. The full curve given in Fig. 60 is calculated by combining Eqs.
(6-31) and (6-34) for a helium-air system, where the numerical values
necessary to the calculation are given in the figure and the dispersion
coefficient of the emulsion E, has been taken from Fig. 41. Calculation
gives good mean values for the data obtained by step response (solid
 FLUIDIZED CATALYST BEDS 379

keys), but deviates considerably from those by steady back-mixing (open

keys). The quantity EdUG is reported to be little influenced by changing
UGwidely (G9, R4).
The apparent axial dispersion coefficient of the emulsion phase,
mCeE,/UG, is also shown in Fig. 60 as a dotted curve. Axial dispersion of
gas is seen to be larger than that of emulsion for smaller DT,but is not
necessarily larger for larger-diameter beds. With the operating variables
chosen here, E,/UG seems to approach mCeE,/UG for 4. 2 200 cm. De
Maria and Longfield (D10) suppose thatE,/UG can be considerably larger
than mCeE,/UG at increasing D,.
Figure 60 indicates that the overall mass-transfer coefficient k&ab is
obtained from the use of transient-response methods, and this approach
has been utilized extensively. However, the approach seems to be more
appropriate to small-diameter beds. In fact, E,/ U, for large-diameter beds
is mostly governed by the term mEeE,/UG, but only little influenced by
kbab.
The influence of longitudinal dispersion on the extent of a first-order
catalytic reaction has been studied by Kobayashi and Arai (K14),'
Furusaki (F13), van Swaay and Zuiderweg (V8), and others. They use the
one-dimensional two-phase diffusion model, and show that longitudinal
dispersion of the emulsion has little effect when the reaction rate is low.
Based on the circulation flow model (Fig. 2) Miyauchi and Morooka (M29)
have shown that the mechanism of longitudinal dispersion in a fluidized
catalyst bed is a kind of Taylor dispersion (G6, T9). The influence of the
emulsion-phase recirculation on the extent of reaction disappears when
the term cp defined by Eq. (7-18) (see Section VII) is greater than about
10. For large-diameter beds, where cp does not satisfy this restriction,
their general treatment includes the contribution of Taylor dispersion for
both the reactant gas and the emulsion (M29).
Mixing and dispersion of gas and particles in the transition zone and in
the freeboard region are important to know, in relation to the mechanisms
of the reactions taking place (see Sections VII-IX). However, little work
has thus far been done in these areas.

F. REVIEWO F WALLHEATTRANSFER
I N FLUID
BEDS
Numerous measurements of heat transfer between the fluid bed and
immersed surface have been carried out; these have been reviewed by
Kunii and Levenspiel (K24), Davidson and Harrison (D5), Botterill (B12),
and Zabrodsky ef al. (Zl, 22). Wen and Leva (W3) and Wender and
Cooper (W5) correlated heat-transfer data at the walls of fluidized beds.
However, the data utilized for their correlation were obtained with a
380 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

fluidized bed less than 10.2 cm in diameter, and measurements under fluid
bed conditions were not included. Heat transfer between fluidized beds
and immersed vertical tubes was correlated by Wender and Cooper (WS).
This correlation includes data with fine porous catalyst particles by Olin
and Dean (04) and Mickley and Fairbanks (M14):

is in sec/cmz. The quantity C,

for d p p g U G / p g= 10-2-101, where cPgpg/kg
represents the correlation factor for nonaxial tube location, given in Fig.
63.
The time-averaged heat-transfer coefficient h , between the vertical wall
and fluidized beds, using fine particles, is controlled by the unsteady heat
transfer due to packet renewal at the heat-transfer surface (M14, M16,
Y18). According to this mechanism, h , is expressed as:
hw = (kerpeCpe/r$l’Z (6-36)
where I is the proper characteristic contact time of packets. Therefore,
h w / s is a function of i or of U G , as shown in Fig. 64; data on
wall-to-bed heat transfer in bubble columns are given in the figure. In the
case of bubble columns, corresponding values for the water-air system
are used in calculating (keffplc,r)1/2.
The correlation by Kato (K5)includes
data by Fairer al. (Fl) and Kolbelet al. (K18, K19). The rate of the packet

Center of vessel Wall of vessel

I

FIG.63. Correction factor CRfor nonaxial location of immersed tubes (V14).

 FLUIDIZED CATALYST BEDS 38 1

q’0L I I I I I I I I I I I 1
-
I l l l L
u
- 8-
-
yc-=---
p-
E6- ?bubble column -
- bubble column --------- -
-
- 0 dp;9pm -
/-<:- -
.I.--

--
\ -

c
f2- fluid bed
a
104 -
dp = 3 8 , 4 8 p m @ 140
I I 1 I I 1 1 1 1 1 I I I 1 1 1 1
1

renewal on the vertical heat-transfer surface in fluid beds agrees with the
data in bubble columns. The larger the particle size that is used in
fluidized beds, the smaller the rate of the packet renewal becomes com-
pared with that in bubble columns. Vreedenburg (V13) correlated data on
heat transfer between horizontal tubes and the fluidized bed, including
data for fluid beds; the data are widely scattered, and further study ap-
pears to be needed.
The bed-to-wall heat-transfer coefficient in a fluidized bed at high tem-
peratures is larger than at room temperature (Y19). Questions have been
raised about the effect of radiative heat transfer at high temperatures
(B12, Y19), and more studies are necessary on this problem.
In the transition zone and the freeboard region, heat transfer between
bed and wall is a function of bed density. Shirai et al. (S9, S11) studied
heat transfer from a sphere immersed in the fluidized bed and showed the
trend of decreasing heat-transfer coefficient with decreasing bed density.

VII. The Successive Contact Mechanism for Catalytic Reaction

A. EXPERIMENTAL
FACTSA N D REACTORMODELS
Early attempts to approximate gas-solid contacting in fluid catalyst
beds were based on the assumption either of isothermal plug flow of the
fluidizing gas through the bed with the catalyst uniformly distributed or of
isothermal complete mixing of the gas within the bed. The simple disper-
sion model, falling between the above two cases, was also used (G8, R4).
Evidence from both large-scale and laboratory observations (G9a, L12),
382 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

however, indicated conversions substantially below those predicted by

the completely mixed case. Due to this, a number of models to describe
fluidized reactor behavior based on the two-phase theory of fluidization
(G9b, 52, S7) have been proposed over the last 20 years. Most assume
with this theory that all gas in excess of that necessary to produce mini-
mum fluidization passes through the bed as the bubble phase. For the fluid
catalyst beds, the net flow of gas through the emulsion phase may be
negligible because of its relatively small magnitude in comparison with the
bubble-phase gas. In what follows this simplification has been made con-
sistently.

1. Overall Rate Constant of Reaction

When a fixed bed of catalyst is operated isothermally as an ideal piston-
flow reactor without radial concentration gradients and without change
in the gas density, the differential mass balance for reactant along the
bed is given by
-UG dc/dZ - l i =~ 0 (7- 1)
where UGis the superficial gas velocity, c the reactant concentration, z the
axial distance from the reactor inlet, and a first-order irreversible reaction
with the rate constantk, is assumed to take place in the bed. This equation
is integrated to give the bed outlet concentrationco at z = L,, the fixed bed
height:

When the bed is not fixed, but is fluidized freely at the same gas velocity
as above, with UG much greater than the minimum fluidization velocity
(Imf,the bed becomes aggregative and deviates from the piston-flow reac-
tor. As a consequence, the apparent overall rate constant of reaction KO,
for the fluidized bed is smaller than k,, and the extent of reaction is
expressed by
(CO/Co)fluid bed = exp(-KoRLq/ uG) (7-3)
K O , is the same physical quantity as the specific converting power Q of
Lewis et al. (L12).
Usually, the fluidized catalyst bed constitutes an aggregative dense
phase forz between 0 and Lf, and a dilute phase forz > Lf. Hence, another
apparent overall rate constant k,, is defined on the basis of Lf. Taking Lf
instead of L,, Eq. (7-3) is rewritten:
(CdCo)fl"id bed = exp(-koRLf/ UG) (7-4)
 FLUIDIZED CATALYST BEDS 383

where obviously the relation

KoRLq = koRLf (7-5)
holds.
The object of the following treatment is to establish a physically sound
reactor model to obtain koR,based on the flow and transport properties of
fluidized catalyst beds. Bed performance for chemical kinetics other than
the first-order reaction may be computed after a sound bed performance
has been established.

2. General Properties of Reactor Models

Many investigations have been performed to make clear the nature of
gas flow through fluidized particles by studying gas mixing, residence time
distribution, and gas bypassing. Various kinetic reactor models have been
proposed to describe gas-solid contacting in fluidized reactors (C9, D3,
F10, F12, G7-9, G9a, H16,113, K4, K14, K24, L5, L12, M8, M12, M29,
M48, 0 6 , P1, P2,P9,S7, V1, V2). Essentially these models are variations
of the two-phase concept (52, S7) of a gas bubble phase and an emulsion
phase, constituting together a dense phase, with gas being interchanged
between the bubble and emulsion phases. Some models take into consid-
eration the gas-cloud region associated with each bubble (K4, K12, K24,
P2, F9).
The main differences between the models lies in whether or not some
fraction of the catalyst is in direct contact with the bubble gas, and in the
extent of axial mixing in each phase. Properties of various models have
been discussed extensively by Gilliland and Knudsen (G7)in relation to
the extent of reaction in experimental fluidized bed reactors, considering
that allowance for direct contact between bubble gas and a certain amount
of catalyst in it is the sole way to account for the contact efficiency. Unless
a fraction of the catalyst particles is assumed to be entrained in the bubble
gas, the bubble size calculated to fit the reaction data is found to decrease
with increasing catalyst activity at otherwise identical fluidization condi-
tions, in which the bubble size should remain constant. Essentially the
same decrease in bubble size was observed by Miyauchi and Morooka
(M29) in their analysis of the data by Lewis et al. (L12), and by Furusaki
(F14) in his fluidized bed data for the Deacon reaction.
The general two-phase model of fluidization contains such parameters
(see Fig. 65) as the volume fraction of bubbles E b , the volume fraction of
particulate phase E, (equal to 1 - z b ) , the overall mass transfer coefficient
k,,a, (sometimes also the bubble-side and emulsion-side film coefficients
kbaband keahrrespectively), the fraction of catalyst directly contacting
384 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

I
- - Phase boundary

Return flow-- -- --- Bubble phase

(cb,% ,G)

I
Err ---
I, --- Dense phase , L,

fw KJ ---
&ob ab,____ - ---Emulsion phase
k b ab. (C.,E 1
k e ab,

- -- Distributor

T Gas, UG(=EG)

FIG.65. Schematic expression of general two-phase dense bed model and the parameters
included (homogeneous return flow is by F10-12, K24, L.5).

with bubble gas v, the cloud-wake fraction fwzb, and the axial dispersion
coefficient of the emulsion Ezs.
In the case of fluid catalyst beds, the mean bubble size stays approxi-
mately constant axially, so that the above parameters are treated as re-
maining unchanged axially. Also, the interstitial gas flow through the
emulsion is neglected in comparison with the feed gas flow U G ,so as to
considerably simplify the treatment. Usually the bubble phase is assumed
to be unmixed axially.
Under the above simplifications, the two-phase consecutive model by
Shen and Johnston (S7)with vertically unmixed emulsion is written for a
differential bed height:
-uG dcb/dz - kobab(Cb - c,) =0 (7-6a)
 FLUIDIZED CATALYST BEDS 385

with
kobab(Cb - ce) - gekrc, = 0 (7-6b)
where the original capacity coefficient is expressed as the product of the
overall coefficient k o b and the specific surface area of the bubbles a b in-
cluded in the bubble-emulsion mixed phases. A first-order irreversible
reaction takes place in the emulsion. Solving the equations, the overall
rate constant as defined by Eq. (7-4) is given by
koR = [(kebab)-' + (Gek)-'l-' (7-7)
When a part of the emulsion volume v is included in bubbles and con-
tacts freely with the bubble gas, the effective fraction of the emulsion
phase is Ze - Y , and Eq. (7-6) is rewritten:
-UG d C b / d Z - kobab(Cb - Ce) - VkrCb = 0 (7-8a)
kobab(Cb - ce) - (ge - v)krce = 0 (7-8b)
This case is the direct-contact model corresponding to the vertically un-
mixed emulsion (VUE) by Lewis el al. (L12), a special case of the model
by Mathis and Watson (M8). Solving the equations, the overall rate con-
stant is
{(16bab)-' + [(G - v)k]-'}-' (7-9)
The bubble flow model by Orcutt et al. (C9, 06) uses the film coefficient
kbUb between the bubble gas and the emulsion. In their vertically unmixed
emulsion k,, is given, with negligible gas flow through the emulsion, by
Eq. (7-7) with k b a b in place of kebab. When the emulsion is mixed per-
fectly, the extent of reaction is given by
c,/co = e-Nb + (1 - e+")/[l + EeNr/(l- e-"")] (7-10)
where N b = k b a b L f / u G and Nr = kLf/UG. Hence, k o R is obtained by
equating Eqs. (7-10) and (7-4).
In the bubbling bed model of Kunii and Levenspiel (K24), there are two
transfer steps for the bubble mass transfer, namely, the transfer between
bubble void and cloud-particle overlap region k b a b and that between the
cloud-particle overlap region and the emulsion phase k e a b . They further
assume that the cloud-particle overlap region and the bubble wake are
mixed perfectly, and contact freely with the cloud gas. Their basic equa-
tions in the present notation are (for their case 2):
-uG dcb/dz - kbab(Cb - c,) = 0 (7-lla)
kbab(Cb - c,) - fwgbkrc, - keab(C, - ce) = 0 (7-llb)
keab(cc - ce) - (ge - f w g b ) k r C e = (7- 1Ic)
386 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

wherefwEbis the fraction of catalyst located in the cloud-wake region, and

the catalyst directly contacts gas of concentration c c . Equation (7-1 lb)
applies for the cloud-wake region, and Eq. (7-1 lc) for the emulsion not
including fwib. These equations lead to

with
kor = {(keab)-' + [(ze - fwEb)krI-'I-' + fwihkr (7-13)
Here k b is given by Davidson and Harrison (D3), and ke by the Higbie
penetration mechanism (see Section VI where k h c and k,, are utilized in
place of k b and k,, respectively).
It is interesting to observe the limiting cases where one particular
mechanism dominates the whole reactor performance. One extreme is the
case where k, << k o b U b , leading to (for all models):
koR = E,k, for k, << k o b U b (7-14)
Obviously, model identification is not possible under this condition. An-
other extreme is k , much larger than the bubble mass-transfer coefficient.
The extremes differ from one another, as follows:
koR = kebab (7-15)
for the two-phase consecutive model, Eq. (7-7);
koR = kh'b (7-16)
for the bubble flow model, Eq. (7-lo), and bubbling bed model, Eq. (7-13);
koR = vk, (7- 17)
for the direct contact model (VUE), Eq. (7-9). Here we note that koR for
the models not including directly contacting catalyst are upper-end limited
by the mass-transfer coefficient kobUb or k b U b . Only the direct-contact
model shows koRincreasing linearly with k, under a given flow condition.
These tendencies are more clearly visualized from Fig. 66, where koR is
shown as a function of k , at constant gas velocity. Numerical values given
to various parameters are those for an ethylene hydrogenation run by
Lewis et ul. (L12).
The models show considerably different k o R when k, 2 . 1 sec-'. The
bubble flow model (BFM) is for perfectly mixed emulsion, but a k,, only
slightly larger is obtained for BFM when the emulsion is vertically un-
mixed. For the bubbling bed model (BBM) koR is fairly sensitive to differ-
ent assumptions forf,; anfw of about unity is recommended ( F 1 1 , K24) to
fit the reaction data available. The value fw = 0.35 is taken from Rowe
(D5) for 75-prn-diam. spherical particles.
 FLUIDIZED CATALYST BEDS 387

FIG.66. Overall reaction rate constant koR as a function of reaction rate constant k , for
various reactor models; = 25 c d s e c , ih
= 0.27, k&ab = 0.37 sec-I, k @ b = 0.94 sec-I,
k e a b = 0.61 sec-I, Y = O.lO,f, = 1.0 and 0.35; data are for a fluidized catalyst bed.

The direct-contact model with vertically unmixed emulsion (VUE)

shows considerably different behavior from other models for k , 2 5
sec-'; k,, increases approximately linearly with k,, but the rest of the
models are upper-end limited by kbab or k,&. As a consequence, the
models are best identified with one another by comparing their perfor-
mance with the data obtained fork, 2 5 sec-I.

3. Comparison of Models with Experiments

Catalytic hydrogenation of ethylene by nickel- or copper-impregnated
cracking catalyst is taken here for comparison. Figure 67 shows typical
experimental koRvalues taken under a constant superficial gas velocity U G
by varyingk,. Curve LGG is based on the data by Lewis et a f . (L12), GK
by Gilliland and Knudsen (G7), and FKM those by Furusaki et ul. (F18).
The calculation ofk,, will be explained later. The dotted curves are calcu-
lated by the two-phase consecutive model (TCM) and by the bubbling bed
model (BBM)for UG = 20 and 25 cm/sec, where the mean bubble size is
4.5 cm and the wake fractionf, = 1.0.
Comparing Fig. 67 with Fig. 66 shows that only a direct contact model
such as VUE accounts for the experimental data. For other models to
388 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

R Jsec-'J
FIG.67. Overall rate constant k,, for experimental ethylene hydrogenation runs by Ni-
or Cu-impregnated cracking catalyst beds. Dotted curves were calculated by the two-phase
consecutive model (TCM) and by the bubbling bed model (BBM). Full curves were calcu-
lated by the successive contact model.

account for the data, kbab or k&ab should be increased with increasing k;
this would imply that the bubble size decreases with increasing catalyst
activity at otherwise identical fluidization conditions, whereas the bubble
size should remain constant. A similar tendency is also observed for
catalytic ozone decomposition by a fluidized catalyst bed (see Fig. 68),
although not so decisively.
The direct contact model has some difficulties, however. In fluidized
beds, gas bubbles of very low solid content are usually considered to exist
in the dense phase (H14, K13, T19). Also, the cloud layer is negligibly
thin, due to small Umffor the usual fluid catalyst beds, according to equa-
tions of Davidson and Harrison (D3) and Murray (M47). The streamlines
of gas phase through a bubble have been observed to pass through the
cloud, but not through the bubble wake (R17). Thus there seems little
possibility of believing that the bubble gas is in direct contact with a sub-
stantial amount of catalyst in the bubble phase (see also Section V1,A).
Furthermore, the direct contact model is applied to the data by Gilliland
and Knudsen, and u in Eq. (7-9) is calculated to fit the data. Calculation
(M26) shows that the volume of catalyst, with an apparent density the
same as for the emulsion, which contacts the bubble gas freely exceeds
the volume of bubble gas itself ( u / c b = 3.3, 2.0, and 1.5, respectively, for
U G = 10, 20, and 30 cm/sec). This seems to be unsound physically.
 FLUIDIZED CATALYST BEDS 389

rQ r [s~c.']
FIG.68. Summary of catalytic reaction data by fluid beds for ethylene hydrogenation and
ozone decomposition: curves were calculated by the successive contact model.

Similarly the bubbling bed model may be visualized as a kind of direct

contact model when kb& is much greater thank,, in Eq. (7-12), since the
equation is now reduced to Eq. (7-9) with an apparent equality OffWEb = v
under this extreme. Actually, however, the transfer of bubble gas to the
cloud-wake region is limited by k b a b , so that the amount of catalyst f W &
should be greater than u to account for the above data; such a large f W E b
again seems unreasonable.
So far, the models explained here view the fluidized catalyst bed as
being composed only of dense phase, so that there is difficulty in locating
directly contacting catalyst. This difliculty may be avoided by placing the
390 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

catalyst somewhere outside the dense phase. Based on the axial distribu-
tion of bed density, Miyauchi (M25) located the catalyst in the dilute
phase and developed the concept of the successive-contact mechanism. A
part of the catalyst may be placed in the jetting zone above the gas dis-
tributor, as explained in Section VIII. Catalyst erosion could be excessive
with too much jetting, however. Also, such a large amount of directly
contacting catalyst as observed for the Gilliland and Knudsen data is
difficult to locate solely in the jetting zone.
Perhaps the influence of dilute phase on the progress of reaction must
have appeared in the minds of many investigators, but it was never formu-
lated. It was stated (M33) that “Visually the particles in the dilute phase
are dispersed rather uniformly, so that under these flow conditions this
portion of the particles may take part in increasing the catalytic conver-
sion.” Interesting discussions by Riley (R12) are another example.

4. Other Reactions
Fluidized catalytic reactions have been industrially operated in the
“fluid bed” conditions, but most of the research has been carried out for
the “teeter bed.” Several studies of fluidized catalytic reaction are listed
in Table VI, which are of interest in considering transport phenomena in
fluidized catalyst beds.

B. THESUCCESSIVE
CONTACT
MECHANISM
A reactor model is developed to include reaction taking place in the
dilute phase, and to be reasonably consistent with the known flow proper-
ties of fluidized catalyst beds operated under relatively high gas velocity.
According to this model, reaction proceeds successively in the dense
phase and in the dilute phase.

1. Flow Properties of Fluidized Catalyst Beds

As stated previously, fluid beds are operated under high gas velocities
with small particles (V,/U,,,, 2 100, Uc/u, = 1-3). The apparent density
of particles is rather small, i.e., usually less than 1.5 g/cm3 (and preferably
less than 1.0 g/cm3).
The bed consists of the dense phase and the dilute phase, and the
former constitutes the emulsion and bubble phases (Fig. 2). In many re-
spects the flow behavior resembles that in gas bubble columns operated at
about equivalent gas velocity (Sections 11-V). Intense circulation of the
dense phase takes place by the buoyant force induced between the cen-
trally ascending bubble-rich phase and the peripherally descending
 TABLE VI. CATALYTIC
REACTIONS IN FLUIDIZED BEDS"

4 4 UG u
rn1 Reaction rate Refer-
Author Reaction Luml [cml [cdsec] [cdsec] constant [sec-'1 ences

Pansing Regeneration of cracking catalyst 44.5-79 25 3- 18 - 28.5 Cr at UG = 18' PI

(4 types)
Askins et al. Regeneration of cracking catalyst - 1220 47 - - A 12
Mathis and Dealkylation of cumene by 74- 149 5-10 1.5-2.5 0.6 -1 M8
Watson SiOz-Al,03 catalyst
Lewis et al. Hydrogenation of ethylene Ni on 80- 188 5 6-45 0.13 1.1- 14.3 112
MS catalyst (av. 122)
May Catalytic cracking 80- 150 5- 150 3-60 - - m12
Gometzplata Catalytic cracking of cumene, 74- 149 7.6 1.8-7.3 - -0.11 g18
and SiOI-AIz03 catalyst
Shuster
Massirnilla and Oxidation of ammonia by 44- 149 11.4 2-16 0.27 0.086 M4
Johnstone alumina catalyst (av. 105)
Orcutt et al. Ozone cracking by Fe03on FCC 20-60 10.2-15.2 3.6-15 0.3 0.05-7 06
catalyst
Ishii and Isomerization of cyclopropane 74- 105 2.3-15 0.9-20 0.3 - I1 1
Osberg
Echigoya et al. Dealkylation of cumene 105-149 5.3 1.2- 14.2 0.25 - E4, E5
Hydrogenation of benzene
Gilliland and Hydrogenation of ethylene Cu on 44- 149 5.2 3-49 - 0.06-0.95 G7
Knudsen FCC catalyst (av. 52, 120)
de Vries et al. Oxidation of HCI 30-150 -20 - -0.5 D8
van Swaaij and Ozone cracking by FQO, 5-60 -20 1 <I v7
Zuidenveg
Furusaki HCl oxidation by CuC1,-KCI- Av. 84 5.5 10-60 -0.6 - F14
SnCl, on MS SO,
Tone et al, Cracking of methyl cyclohexane 88- 105 4.5-5.2 1-3 0.25-0.3 - T2 1
by FCC catalyst
Furusaki et al. Hydrogenation of ethylene 62 5.3 10 - 30 0.2 4.3 - 21 F18

Selected for fluid beds. ' Cr = gm-coke/gm-catalyst .

392 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

bubble-lean phase. Mean bubble size stays relatively unchanged axially;

the size seems to be characteristic of flow intensity in the bed. The upflow
and downflow emulsions exchange with each other by radial eddy mixing.
Thus the general flow features may be visualized as in Fig. 2.
In the dilute phase the catalyst population decreases with increasing
axial height from the dense-bed surface (Section 11). It has been shown
(M26) that the amount of directly contacting catalyst is approximately
equal to “a” of Lewis et al. (L12), if one takes the amount as the volume
of catalyst existing above Lf. The data are compared in Fig. 69. Here, Lf is
taken as the height above the distributor where the volume fraction of the
emulsion equals that of the bubbles, i.e., Ee = Eb; E, becomes meaningless
in the dilute phase or in the transition zone, since Eb becomes dubious.
However, Eed (instead of Z,J is used as the hypothetical volume fraction if
the dispersed particles in the dilute phase are concentrated into the same
density as the emulsion. This treatment is convenient in calculating con-
version in the dilute phase.
One important point is that the phase boundary which separates the
dilute phase from the dense phase is somewhat arbitrary to define and is
located with difficulty. Instead of defining the phase boundary, it might be
more reasonable to place a transition zone between the above two phases.
In what follows, however, the concept of phase boundary has been

FIG.69. Comparison of experimental“a” with calculated values based on the successive

contact model (M26).
 FLUIDIZED CATALYST BEDS 393

utilized, owing partly to the simplicity of developing a reactor model and

partly to the lack of sufficient information to define a transition zone.
When the dense phase height Lf is taken at E, = 0.50, the bed density
distributions show that Lfremains approximately unchanged for different
gas velocities (M25).

2. Formulation of the Model

The previously mentioned flow properties of the beds lead to the follow-
ing formulation for the successive-contact model (M25). Simplifying as-
sumptions additional to those given in Section VII,A,2 are, for the dense
phase (M29):
(a) Ascending or descending velocity constant for bubble phase or for
circulating emulsion phase, respectively.
(b) Bubbles ascend mostly with the upflowing dense phase, so that they
are included only in this phase (see footnote on p. 345).
(c) The upflow and downflow emulsions exchange a part of the emul-
sion with each other at the exchange rate coefficient kexaex(Fig. 2).
(d) A small quantity of freely suspended particles has been observed in
the bubble phase (H14, K13,T19). These particles, with the volume frac-
tion u , are allowed to directly contact the bubble gas.
(e) Physical adsorption equilibrium is almost instantaneously attained
between particles and fluid in the emulsion phase with the partition ratio
m = Efe + rn,ese, with m, = cSe/cfe.
(0 An isothermal and irreversible first-order reaction takes place stead-
ily in the bed.
With these premises, analytical solutions have been presented for the
dense phase (M25, M29). The solutions show essential features of Taylor
dispersion with chemical reaction, similar to the case treated by Cleland
and Wilhelm and others (C8, S2, W15) for cylindrical-pipe flow (see M29).
Here, the axial dispersion coefficient of gas affecting the steady reaction
diminishes as the rate of chemical reaction increases, compared with the
dispersion coefficient affecting axial mixing without accompanying
chemical reaction. These solutions are general, but inconvenient for
computation. Several simplifications have been discussed; one simplifica-
tion potentially applicable to the usual fluid-bed operations is to adopt
the restriction:
(7-18)

With this behavior of cp the influence of the emulsion-phase circulation

apparently disappears ( U, is the superficial circulation velocity of the
emulsion phase) and the computation is greatly simplified.
394 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

With (c z 10, the following equations of continuity are obtained for the
dense phase:
- UGd C b / d Z - kobab(Cb - Cf,,) - VkrCb = 0 (7-19)
kobab(Cb - Cfeu) - m k e x a e x ( c f e u - c f e d ) - ( i c e - v)krcfeu =O (7-20)
mkexaex(cfeu - cfed) - kgekrcfe, =0 (7-21)
where half of the emulsion phase is in ascending flow and the other half is
in descending flow. Equation (7-19) is for the bubble phase; Eqs. (7-20)
and (7-21) are for the ascending and descending emulsion phase, respec-
tively. The term (iGe - v) is the fraction of the ascending emulsion and 4Ge is
that of the descending emulsion.
When the exchange rate of emulsion between the upflow and the
downflow is high enough, the emulsion gas concentrations ce, and c e d
become nearly equal, so that Eqs. (7-20) and (7-21) are combined and
reduced to Lewis' VUE model, Eq. (7-8.b).
Equations (7-19)-(7-21) are combined by eliminating ce, and Ced, yield-
ing
-UG dCb/dZ - (16, + V4)Cb = 0 (7-22)
where l/kor = l/k,b& + 1/tGekr, and 6 is given in Eq. (7-28). For large-
diameter beds, cp could be smaller than 10 (see M29).
The bubble gas, with initial concentration co, leaves the dense phase at
z = 4 with concentration cLp and enters the dilute phase. For the dilute
phase the diffusion-model approach is more general, since a fair amount of
gas mixing is observed in a large-diameter bed (H19). Here, however, the
usual piston-flow assumption is made for simplicity, since in small-scale
beds the particles in the dilute phase are suspended rather uniformly.
Actually the observed flow properties of the phase, including the transi-
tion zone, seem to be more complicated.
The equation of continuity with the piston-flow assumption is
-UG dc/dZ - gedkrC = 0 (7-23)
where the physical meaning of Zed has been explained in Section VII,B, 1.
The reactant concentration coof the gas leaving the dilute phase at z =
& is obtained by integrating Eqs. (7-22) and (7-23) as follows:
co/c0 = exP{-[kor + (v + e)krlL,/UG) (7-24)
where

(7-25)
 FLUIDIZED CATALYST BEDS 395

Accordingly, eL, is the total fraction of catalyst (with the same density as
the emulsion) suspended in the dilute phase. The total fraction of catalyst
in the dense phase is ZeL,, so that the following equality exists for the
successive contact model:
L, = ( e + Ze)Lf (7-26)
Comparing Eqs. (7-24) and (7-4), koR can be given by
koR = [(kebab)-' + (4Eekr)-']-' + (v + e)k, (7-27)
where 6, defined by Eq. (7-28) is the fraction of catalyst in the dense phase
effectively utilized for reaction:
4 =1 - (v/E,) + [2 + ( C A / m k e ~ d I - ' (7-28)
Equation (7-27) is reduced to the VUE model by Lewis et al., Eq. (7-9),
when e = 0 and Zek,/mkexaex<< 2; v is usually small in comparison with e
(Section VIII). In this case, with 4 = 1, Eq. (7-27) is reduced to the
simplest expression of the model:
koR = [(k,,a,)-' + (Zekr)-']-' + ek, (7-29)

3 . Experimental Evidence
Ethylene hydrogenation runs are summarized in Fig. 68, where the
data by Lewiset al. are taken from their smoothed curves (Figs. 7 and 8 of
L12). For these, approximate numerical values for cp and 4 are cp = 25-38
and4 = 0.95-0.73,for UG = lOcm/sec,andcp = 52-63and4 = 0.98-0.83
for UG = 40 cm/sec. Also, the simplest expression [Eq. (7-29)] is utilized
to calculate koR.The Lewis-Gilliland-Glass plot (see Section VII,C,2) has
been applied to obtain kebab and e from the experimental data; the other
data are similarly processed.
The data by Orcutt et al. (06) for catalytic ozone decomposition have
been processed in the same manner. Those by van Swaay and Zuiderweg
(V7, V8) for ozone decomposition are treated as follows. Their data are
presented in terms of an apparent HTU, (Hob)app, as defined by
(Hoblapp = u G / k ~ R - UG/Eekr (7-30)
where UG/ko,equals HoR,the overall height of a reaction unit by Hurt
(H18). Because (&,b)app equals UG/(kobab>app, they presume Eq. (7-7)
holds. Now koR is calculated using Eq. (7-30) by knowing the expenmen-
tally given (Hob)app; see Fig. 70.
In processing the above data, v is consistently assumed to be negligible
in comparison withe, and Eq. (7-29) is utilized in calculatingk,,, including
the full curves given in Fig. 68.
396 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

D~cm ut 1 db e 1 6b
200 - - 0 23 15 142 0.036 1 015
A 10 10 I 2.8 0.031 I 0.12
1 1 ' 'I I ' "'I

2o t
k, [sdl
FIG.70. Comparison of the calculated (Hob)app
with the data of van S w a y and Zuider-
weg (VS).

In the figure the data obtained by Gilliland and Knudsen from the use of
copper-impregnated cracking catalyst (quiescent bed density pmf = 0.33
g/cm3) show considerably higher koR than those by Lewis el al., who used
nickel-impregnated cracking catalyst (pmf = 0.52 g/cm3). Since the ex-
perimental runs have been performed under nearly isothermal conditions,
the higher koR by Gilliland seems mostly to come from their using less
dense catalyst than Lewis et al. used. In other words, at least qualita-
tively, a fair amount of catalyst is suspended in the dilute phase in con-
formity with higher e (or a ) .
Furusaki et al. use a standard cracking catalyst of wide size distribution
with smaller mean particle size, higher e, and higher UG than those of
Lewis et a f . (cf. Fig. 68).
The successive contact model seems to be sound in accounting for the
fluid-bed performance given in Fig. 68, since the model satisfies the flow
and transport properties of the beds as well. The model is also applied
successfully to catalytic oxidation of HCl (the Deacon reaction) in a fluid
bed (F14).
Recently, Yates and Rowe (YIO) have observed, on the basis of their
model for catalyst distribution in the freeboard region, that this region can
usually exert a considerable influence on the course of the reaction. Their
observation is essentially parallel with the concept of the successive con-
tact mechanism. However, they use the bubbling bed model in calculating
the reaction in the dense phase, so that the effect of directly contacting
catalyst seems to be corrected two times, first partially in the dense phase
and then in the freeboard region (see Section VII,A,3).
 FLUIDIZED CATALYST BEDS 397

4. Relative Contribution to the Overall Extent of Reaction

As explained in Section VI, the bubble mass-transfer coefficient k,&, is
given by
(kebab)-' = (khUb)-' + (Prkeab)-' (7-3 1)
where Pr is the modified Hatta number. Also, the variables affecting the
reaction in a fluid bed have been formulated as Eq. (7-27). It is important
to know their relative contributions, i.e., P,, the amount of catalyst in the
dilute phase e, the fraction of catalyst directly contacting with the bubble
gas v , and the jetting zone. Of these, v is difficult to measure directly by a
physical method, but is usually small in comparison withe (Section VIII).
For a typical example we take the data from Lewis et al. for the system
C2H4-Hz, where U , = 16 cm/sec, d , = 122 p m and DG = 0.891 cm2/sec.
Then, the mean bubble size is approximately 5.0 cm (Fig. 511, so that k,&b
is obtained from Eq. (7-31) (Section VI). With these and 8 = 1, k,, is
calculated from Eq. (7-29) by taking k, and e as variables.
The results (M28) are given in Fig. 71, where HoR= U,/k,,. From this it

-
'I
T 5-
lo'

8 4-
I
3-
3= 16 cm/iec i--
Cb = 0.16 E. = 0.04
2- e = 0.074 , QOZO
d,= 5cm
10 r
7-
5-
4-
3-

FIG.71. Relative contribution of parameters to HoR = LIG/koR);P I ,P,, and P3 are the
respective locations where k,$,, e , and 0,start to be significant (M28).
398 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

is obvious that e is most significant; p, is important only for very large k ,

(220 sec-', point P3). In the region of very small k, (10.1 sec-') the
reaction is the rate-determining step. Points where kebab and e begin to
affectH,, are shown as points PI and P2, respectively, for an approximate
estimate. Point P2 for an e which is starting to be significant changes
considerably as e changes.
The properties of point P2are more clearly observed from Fig. 70, for
the experimental apparent HTU by van Swaay and Zwiderweg (V7, V8).
The HTU decreases as k , increases beyond about 1.5 sec-', showing an
apparent enhancement of bubble mass transfer due to ek, (M28). This k,
corresponds to point P2. The full curves are calculated by Eqs. (7-29) and
(7-30) from the use of numerical values given in Fig. 70. The equations
give two extremes:
(Hob)app = U G / k o b a b (7-32a)
for negligible ek,; and
(Hoblap, = UG/ekr (7-32b)
for k, >> kobUb.
Axial distribution O f kobUb has been shown to have only a minor effect on
the performance of fluid catalyst reactors (K14, M28). It has been shown
in Section I1 that (a) bubbles from a single nozzle break up in rising a
certain distance to attain a final size; (b) bubbles from a perforated plate
associate together when rising; and (c) db stays fairly constant axially
thereafter.
Basov et a/. (B6) measure the longitudinal-bed density distribution for
crushed cracking catalyst (average d, = 120-130 pm). They observe that
near the distributor (0-20 cm above it), there is a region of varying density
with height, then stays unchanged. The density is low near the distributor,
since dense-phase circulation is low there and particles may not be easy to
supply from the emulsion above. This lower region is called the jetting
zone. Here d b may be smaller or larger than the average d b of the bed,
depending on the type of distributor and on energy input to the jet.
To see the influence of the jetting zone on the extent of reaction, H,,
(equal to U,/k,,) has been calculated under the assumptions that (a) koba,
is five times greater in the jetting zone than in the dense phase above, (b)
the height of the jetting zone is one-fifth of the bed height Lf,and (c) the
other four-fifths has kebab such that the mean k,bUb averaged over L ,
equals that in the example given in Fig. 71. Calculation gives the dotted
curves given in Fig. 71. The effect of the jetting zone appears insignificant,
as far as kebab is concerned. However, a fraction of directly contacting
catalyst seems to be located in the jetting zone, although it is not very
significant (see Section VIII).
 FLUIDIZED CATALYST BEDS 399

Behie and Kehoe (B9) state that the effect of the jetting zone is salient,
especially for the case of very fast reaction. They use a very high mass-
transfer coefficient in their calculation (koJ = 57 cm/sec, uJ = 30 m/sec)
(B8). On the other hand, Eb could be very small in the jetting zone if the
dense phase holds only the jets of high speed and not bubbles, since Eb =
UJu,, = WJu, << 1 due to u, >> 0. Actually, the fraction of bubble phase
is high in this zone (B6, E8). In case of fluid beds, the spout above the
nozzle of the distributor is quite unsteady and splits into bubbles within a
very short distance. As for this, the observation by van Krevelen and
Hofiizer (V6) for a gas jet submerged in liquid is quite suggestive for
practical values of UG.

c. OVERALL RATECONSTANT K O , BASEDO N L ,

So far, the overall rate constant has been based on the fluidized bed
height Lf. This definition of koR has some advantages, since the parameters
which appear in formulating the reactor theory are directly related to flow
properties of the bed and are thus easier to understand physically. How-
ever, K o R , based on quiescent bed height L,, is more convenient for design
purposes, since L, is known once the catalyst inventory is known. In
contrast, Lf changes with U G ,and may be difficult to define when the bed
is aerated with large gas velocity. In such a case Lf may even be zero when
the catalyst inventory is small. Also, K O , has the definite advantage that,
as stated in what follows, the Lewis-Gilliland-Glass plot is applicable to
the successive contact model without any trial calculations.

1. Reactor Models Based on K O ,

Equation (7-5) gives the relation that KoR equals koRLf/Lq.
For the suc-
cessive contact model the ratio L,/L, equals (Ee + e) from Eq. (7-26), so
that we have:
KoR = koR/(Ee + e) (7-33)
For the rest of the models there is no dilute phase. Hence e = 0 and L, =
E,Lf, or
KoR = koR/Ee (7-34)
Equation (7-33) is combined with Eq. (7-27), so that a KoRis obtained for
the successive contact model:
KoR = (1 - u)[F'+ kF'1-I + ak, (7-35)
with F = kobUb/(& - v) and a = (v + e)/(Ee+ e ) .
400 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

When e = 0 in Eq. (7-39, the above model is reduced to the direct

contact model (VUE) of Lewis et al. This relation is also obtained by
combining Eqs. (7-34) and (7-9). In an essentially similar manner the other
models are reduced to the form of Eq. (7-35) and these are summarized in
Table VII. Various extreme cases of chemical reaction, bubble mass trans-
fer, or directly contacting catalyst being controlling are easily obtained
from this table.
Figure 72 shows a numerical example taken from the catalytic hydro-
genation of ethylene by Gilliland and Knudsen (G7). Reaction data are
shown at U G= 30 cm/sec. The rest of the parameters are given in Fig. 72,
and these make it possible to calculate KoRfor different reactor models.
The experimental data are first processed by the Lewis-Gilliland-Glass
plot to obtain F and a, from which e and k&ab are calculated for the
successive contact model (SCM, u = 0). This example perhaps constitutes
one of the most severe tests of different reactor models. In Fig. 72
DCM(VUE) approximates the data equally well.
Except for the mechanistic concepts, there is little difference between
the two models as far as the capability of correlating data is concerned.
The differences are that (a) LJL, is different for the two models due to
existence of the dilute phase, and (b) as calculated from F is 0.166

TABLE VII
REACTORMODELS"
KoR FOR VARIOUS

KoR= ( I - a ) / ( l / F + I / k J + ak,
koR/&R
Reactor model F U (=La/&)
~~

Two-phase consecutive kobab/Ze 0 4

model (S7)
Bubble-flow model (VUE)b Ce
(06)
Direct contact model (VUE)b 4
(L12)
Successive contact model
(usually Y << e ) 2. +e
(M25, M26)
Bubbling bed model
(K22-24) Qe
 FLUIDIZED CATALYST BEDS 40 1

Direct contact

0 Data by Gilliland-Knudsen .
a1 UG= 30 cm/sec (G- 7 1

FIG.72. Overall rate constant KO, based on the quiescent bed height L , calculated by
various reactor models; U G= 30 cmlsec, Zb = 0.31, kebab = 0.17 sec-I, kbab= 0.42
sec-',k,ab = 0.27 sec-',a = 0 . 6 6 , ~= ~l,f, = 1.0 and 0.35.

sec-' for SCM and 0.057 sec-' for DCM; the latter value seems too small.
From Table VII one has
= [kobab/(&- Y)IDCM(VUE) = (kobab/ge)SCMb=O)
so that
(kobub)DCM(VUE)/(kobab)SCM(u=O) a (7-36)
= (ge - v)/ge = 1 -
In the present example a = 0.657 and 1 - a = & (see also discussion given
in Section VII,A,3).

2. Evaluation of e and kObab+he Lewis-Gilliland-Glass Plot

In evaluating the transport properties in fluid catalyst beds during reac-
tion, it is necessary to utilize reaction data obtained at relatively high
reaction rate. The reactor models of different mechanisms have been re-
duced to the form of Eq. (7-33, as shown in Table VII, including the
bubbling bed model when k b a b >> Kor. Eq. (7-35) is equivalent to one
developed by Lewiser al. (L12) for their direct contact model of vertically
unmixed emulsion (VUE). As a consequence, Eq. (7-35) is transformed
(L12) to:
k&R/(kr - K o ~=) &/(k - koR) -k F (7-37)
402 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

CI '-100+140 MESH

Y
5 -
AV dp=122MICRONS
I
sr'2
Y

Y
8
1
INTERCEPT=F

1 I I I
Ob 5 10 15 20
kf (kr- K0R)

FIG.73. Typical Lewis-Gilliland-Glass plot (after L12).

The plot of k;KoJ(kr - KoR)versus e / ( k- KoR)for the data taken at

constant U, should give a straight line of slope a and intercept F, as shown
in Fig. 73 (L12). In this way, one can estimate a andF by reaction data for
several catalysts of different activity. This evaluation method, named the
Lewis-Gilliland-Glass plot (F18), gives reliable F and a during reaction,
since the original data are obtained from the reaction itself. For the suc-
cessive contact model, k&&, and e are calculated from F and a, since v is
usually small in comparison with e (Section VIII).

VIII. Further Properties of the Successive Contact Mechanism

The concept of the successive contact mechanism has been given its
simplest form by dividing the fluidized catalyst bed into two parts-dense
phase and dilute phase. The concept has been found to apply to bed
performance, as shown in the preceding section. The reactor model has
been developed on the basis of several simplifying assumptions, partly to
retain mathematical simplicity as a workable design equation accounting
for the relative effects of the variables, and partly due to a relative lack of
information about bed performance. Further properties of the mechanism
are examined here, particularly as to axial distribution of reactivity inside
the bed.

A. AXIALDISTRIBUTION
OF REACTIVITY
IN A FLUIDBED

The dilute phase as defined in Section VII,B,l partly includes the transi-
tion zone. Gas-solid contacting there is much more complicated than is
 FLUIDIZED CATALYST BEDS 403

indicated by the simplifying assumptions for the successive contact

model.
Tsutsui (T29) measured the axial bed density distribution inside a
2-inch-diam. fluid bed with the purpose of reproducing the runs by Gilli-
land and Knudsen (G7). The fluid-flow experiment was intended to deter-
mine whether or not so large an amount of particles was suspended in the
dilute phase. Since the catalyst utilized by Gilliland was not available,
Tsutsui used microspherical carbon and, glass “baloons” (ballotini) and
mixtures of them with cracking catalyst as the fluidizing particles to adjust
the quiescent bed density. He found that the quantity of particles sus-
pended in the dilute phase was qualitatively parallel with those observed
for the runs by Gilliland, but not always quantitatively.
Later, Furusaki et al. (F17) studied the hydrogenation of ethylene by
fluidized Ni catalyst to obtain the axial reactivity distribution. Here the
samples of bed gas were removed by a traveling sampler placed at the
center of the bed during steady reaction, so that the sample taken in the
dense phase shows an average of the concentration in the bubble and
emulsion phase. Figure 74 shows an example of the axial concentration
profile.
Apparently the reaction seems to have almost ended near the dis-
tributor; this is because the sample has been mostly taken from the emul-
sion phase. The calculated concentration profile, assuming ( -
-

0.2, is close to the observed profile. However, the axial reactivity dis-
tribution inside the bed is not always clear, although this kind of experi-
ment does give useful information. A similar experimental approach has
been utilized by other investigators (C7a, F12).

FIG.74. Axial concentration profile of bed gas, observed by taking out the gas by an
axially traveling sampler (F17).
404 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

An alternative to the above method is to introduce one of the reactants

at a given vertical bed height and to measure steady-state conversion at
the bed outlet by varying the height. This method, named the differential
reactivity test, was suggested by T. Kikuchi and applied experimentally to
catalytic hydrogenation of ethylene by Furusaki et al. (F18).

1. The Differential Reactivity Test

When a fluid bed is operated as an ideal piston-flow reactor as in the
simplest case of the successive contact model [Eq. (7-29)], the equation of
continuity is given by
-UG dc/dz - r(c) = 0 (8-1)
If the reaction rate r(c) is the product of a function of vertical position
koR(z)and that of concentration f(c), Eq. (8-1) is integrated to

where one of the reactants at initial concentration co, is introduced stead-

ily into the bed at height z measured from the distributor, and leaves the
bed at z = L , with the concentration co. Differentiating Eq. (8-2) by z , we
have:
koR(z)lUG :f
= ( d / d z )J C dc/f(c) (8-3)

This is the general relation to obtain the local overall rate constant koR(z)
from experimental runs. When the reaction is of first order,f(c) = c, then
koR(d/ UG = d[ln(co/cO)l/dz (8-4)
Thus, the local slope of the plot of In(co/cO)versus z for the data taken
under a constant UGwill give koR(z)/UG, and from this koR(z).

2. Application of the Test to a Fluidized Catalyst Bed

Distribution of the local reactivity koR(z)was tested by hydrogenation of
ethylene (F18). Reaction proceeded in a fluid bed 5.3 cm in diameter and
130 cm in height. Nickel-impregnated cracking catalyst was fluidized by
hydrogen, introduced through the gas distributor at a chosen superficial
velocity. Ethylene was injected into the fluidized bed through the injection
nozzle at several vertical positions. Conversions thus obtained are plotted
against vertical position z in Figs. 75 and 76. In these figures, the slope is
large near the distributor and at the transition zone: pointf corresponds to
the dense bed height Lfat Ge = 0.5 (M25). Thus, local reactivity is large at
these regions. The dense bed may be divided into three sections: the
 FLUIDIZED CATALYST BEDS 405

z(cm)

FIG. 75. Longitudinal distribution of reactivity for ethylene hydrogenation by Ni-

impregnated cracking catalyst, k , = 10 sec-l, catalyst lo00 cm3 (F18).

Level of C2H, injection Z [cd

FIG.76. Effect of the reaction rate constant k , on the longitudinal distributionof reactiv-
ity, ethylene hydrogenation, catalyst 600 cm3, U o = 30.5 c d s e c . Nozzle designs u and b
refer to Fig. 3 of ref. F18.
406 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

jetting zone near the distributor, the main dense phase, and the transition
zone between the dense and dilute phases. Besides these three zones, the
dilute phase must also be considered.
From the differential reactivity test, k,, is obtained as a measure of
reactivity. Another measure is to define the local contact efficiency 7,by
the following relation:
7, = koR(Z)/+ (8-5)
where Eed is taken instead of i,in applying 7,to the dilute phase. In the
case of the dense phase, the fraction of catalyst v is directly contacting
with the bubble gas, so that koR(z)equals k,, + vk, from Eq. (7-22); then 7,
for this case is given by
7,= t$'([k,(k,,t,Ut,)-* -k (5ce)-1]-1 + V ) (8-6)
Values of qCare given in Fig. 77 for the data of Fig. 76. The line of the
dense phase in the figure is calculated by Eq. (8-6) assuming $, = 1 and v =
0. The term ko&, is 0.4 sec-' at U, = 30.5cm/sec and is obtained from the
use of the Lewis-Gilliland-Glass plot (Section VI1,C ,2).
The contact in the dilute phase and the transition zone is different from
that in the dense phase. The former is related to mass transfer between the
gas phase and an agglomerate of solid particles, whereas the latter is
mainly related to mass transfer between bubbles and emulsion including a
certain amount of directly contacting catalyst. Upon consideration of hin-
dered settling of the swarm of particles (S16, Z7), we also find contact
efficiency to be a function of the population density of particles. Normaliz-
ing by the E, of the main dense phase, &,, E,/E,,de is chosen as a variable

' A ' O m

$0.5-

OO Ib 2b 2 ( c m30) Lb o;

FIG. 77. Dependence of T~ on the height from the distributor, catalyst 600 cm3,
U G = 30.5 cm/sec (F18). Point a indicates the upper limit of the dense phase.
 FLUIDlZED CATALYST BEDS 407

FIG.78. Correlation of the contact efficiency with Ee/& for the dilute phase including
the transition zone (F18).

(Zed/Ze,de for the dilute phase) to affect vC because these are primary data
which can be obtained by measuring the solid density in the bed.
Figure 78 shows this relation. The data are rather scattered, but most of
the efficiencies are inside the shaded area of the figure. The most probable
value is given by the solid line, expressed by:
v, = 1 - 0.75(E,/E,,de)0.4 (8-7)
The equation shows that the strong interaction of particles affects the
contact efficiency. The interaction seems to increase rapidly with increas-
ing particle population. Figure 78 can be used to estimate local contact
efficiency when the density distribution of solid catalyst is known.

OF CONTACT-MECHANISM
B. AXIALDISTRIBUTION CONTRIBUTIONS
So far, the local reactivity distribution has been explained in terms of
overall rate constant koR and overall contact efficiency vc, following
Furusaki et al. (F18). The overall rate constant is further split into two
parts, the local mass-transfer term and the local fraction of directly con-
tacting catalyst, by developing a concept of the local Lewis-Gilliland-
Glass plot (M27a). The relative contribution of the variables will be ex-
plained here according to (M27a).

1. Axial Distribution of Local Overall Rate Constant

Figure 76 has shown the distribution of local reactivity. To see the
distribution more clearly, the smoothed curves which connect the data in
408 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

the figure are drawn and differentiated graphically with respect to the axial
bed height z at which one of the reactants, ethylene, has been injected.
The differentiation gives, by Eq. (8-4), the local overall rate constant koR:
k R = 2.303 UG d [ l ~ g , o ( ~ o / ~ ~ ) ] / d ~ (8-8)
Figure 79 shows the axial distribution of k,, obtained in this manner.
The local reactivity is high in the distributor region and in the transition
region (z = L,) as well. Also, the reaction proceeds in the freeboard
region, indicating the progress of reaction by freely suspended catalyst.
The reactivity is rather low in the main dense phase, in contrast to that
anticipated from the direct-contact model (L12) or the bubbling-bed
model (K24). More reaction takes place in the transition zone than in the
distributor region.
The local reactivity distribution, Fig. 79, is further processed in the
following sections to determine the relative contribution of directly and
indirectly contacting catalyst to the progress of the reaction.

2. Local Lewis-Gilliland-Glass Plot

As stated already in Section VII,B,2, Eqs. (7-19)-(7-21) hold for a given
cross section of the dense phase (see Fig. 2). These equations have been

FIG.79. Axial distribution of the overall rate constant koRfor different catalyst activities,
catalyst 600 cms, U G = 30.5 c d s e c (M27a).
 FLUIDIZED CATALYST BEDS 409

combined, leading ultimately to Eq. (7-22) or

-U, d c b / d Z - koRCb = O 03-91
where
koR = k,, + ~ k =, { ( k , , b ~ b ) - ' + [(Ce
~)k]-'}-' + vk, (8-10)
-

For the usual experimental beds, Z,kr/mke,ae, << 2, so that 6 as given by

Eq. (7-28) is assumed here to be
5= 1 - (v/Ze) (8-11)
For a given cross section of the dilute phase (this phase includes a part
of the transition zone) the particles contact the ascending gas in a compli-
cated way. Swarms of suspended particles may show a mass-transfer
resistance for gas-particle contact, whereas a very dilute suspension of
catalyst will not involve any appreciable mass-transfer resistance. Ac-
cordingly, one can assume that the fraction v d directly contacts the reac-
tant gas; the rest of the catalyst Ged - v d , in dense clusters, indirectly
contacts the gas with the overall mass-transfer coefficient k,a,. Under the
simplifying assumption of ideal plug flow of gas through the dilute phase,
the equations of continuity are given as follows:
- UGd c / d Z - kmU,(C - C,) - VdkC = O (8- 12a)
for free-flowing gas at concentration c, and
kmac(c - cc) - (Zed - vd)krcc = 0 (8-12b)
for dense-cluster gas at concentration c,.
Eliminating the concentration c, for Eqs. (8-12) we obtain
-UG dcldz - koRcC =0 (8-13)
where
koRc = {(kocac)-' + [(Ged - ud)krl-l}-l + udkr (8-14)
Equations (8-10) and (8-14) are essentially the same type of equations.
Also, they are each equivalent to Eq. (7-9), the direct contact model of
VUE by Lewis et al. (L12). As a consequence, the local mass transfer
term (kebab or kocuC)and the local fraction of directly contacting catalyst ( v
or vd) should be obtained from the use of the Lewis-Gilliland-Glass plot.
At a given bed section, koRor koRcis obtained from Fig. 79 for each run
of different catalyst activity under constant aeration. Consequently, the
local parameters are determined by this plot. In contrast to Eq. (7-9),
which is for the overall bed performance, Eqs. (8-10) and (8-14) are for the
local performance. In this sense the plot as applied to the latter equations
is called the local Lewis-Gilliland-Glass plot (M27a).
410 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

Combining Eq. (7-37) with a and F as given in Table VII for the direct
contact model (VUE) or directly modifying Eq. (8-lo), one has the follow-
ing local relation for the dense phase section (with &Z, = Z, - u ) :
OR = vkr 4- p:(1 - kR/Z~kP)(k,bab)P,=l/(l - Y/ce) (8-15)
where is defined [see also Eq. (7-31)] by:

(8-16)

Calculation shows for the data of Fig. 79 that p: = 1.21 and 1.35 fork, =
10 and 20 sec-*, respectively.
In processing data by Eq. (8-15), first assume p: = 1 and determine
(k&,)&=l, from WhiChkb andk, are obtained (Section VI). With thek, and
k,, P;* is calculated as a function of k,. Usually, further iteration is not
needed, because pp* = 1 is a good approximation.
The plot kR/( 1 - kR/ZeQ)p: versus Q/(1 - kR/Ze,%)p: for data such as
those in Fig. 79 at a given cross section should give a straight line of slope
u and intercept (k,,ba&,=l/( 1 - v/Z,) as shown in Fig. 80a. In this way, one
can estimate the local u and (kbab)&=lfrom the reaction data for the dense
phase.
Similarly, the relation for the dilute phase is obtained by modifying Eq.
(8-14). Equation (8-15) again applies by taking kOm, ko,a,, Zed, and vd,
respectively, in place of koR,kebab, Z,, and v. Also, simply assume p: = 1
for the dilute phase, since the mass transfer term is usually negligible
here.

3. Axial Distribution of Local Mass-Transfer Coeficient and Directly

Contacting Catalyst
Figure 80a is a plot of Eq. (8-15) for the mean koR of the main dense
phase (z = 10-20 cm) given in Fig. 79. Figure 80b is the same plot for the
dilute phase data (z = 50 cm). For the dilute phase the latter plot shows
that the straight line passes through the origin, indicating that the mass
transfer term is negligible; for the dense phase the intercept remains al-
ways finite. Similar plots give the axial distribution of the mass-transfer
term (k&ab)4=1 and the fraction of directly contacting catalyst v (Fig. 81).
Having in mind the possibility of inaccuracy inherent to data smoothing
and graphical differentiation, it is still quite interesting to see the charac-
teristic distributions of (k&&)&,1 and u. Here the dense phase height L , is
approximately 35 cm. The term (k&b)&=l remains approximately con-
stant in the dense phase, and drops rapidly to zero abovel,. Mass transfer
seems somewhat better in the transition zone and in the jetting zone than
 FLUIDIZED CATALYST BEDS 41 1

FIG.80. Local Lewis-Gilliland-Glass plot for the data of Fig. 7 9 (a) at z = 10-20 cm;
= 50 cm (M27a).
(b) at z

in the main dense phase, but not very different. As a consequence, mass
transfer in the dense phase can be effectively expressed by a mean value
for ( k h u h ) & = l *
In contrast to this, the fraction v distributes in a complicated manner.
Of the total directly contacting catalyst (the area enveloped by the v
412 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

Axial bed height 2 [cml

FIG.81. Axial distribution of local mass transfer termk,Ga, and fraction of direct contact
catalyst v for the data of Fig. 79 (M27a).

versus z curve, Fig. 81), about 15% is located in the jetting zone (z = 0-7.5
cm), another 15% in the main dense phase (z = 7.5-30 cm), about 40% in
the transition zone (z = 30-40 cm), and the remaining 30% is in the dilute
phase (z 2.40 cm). The v distribution curve implies that direct contacting
of catalyst with the reactant gases is most noticeable in the transition zone
and in the dilute phase, less so in the jetting zone, and rather lacking in the
main dense phase. We note that the relative contribution of variables given
here is based on only one experimental example that would allow separa-
tion of v and k,,,a,. More data are needed for further progress.
The kinetic approach so far presented as the successive contact mecha-
nism, in its simplest form, has been applied to very complicated fluid-bed
behavior. Nevertheless the model has consistently paralleled the observa-
tions. A fair amount of directly contacting catalyst is shown to exist in the
transition zone and in the freeboard region, with less in the jetting zone,
under aeration with relatively high gas velocity. The relative location of
the directly contacting catalyst will certainly influence the selectivity of
complex reactions. Directly contacting catalyst in the freeboard region
may sometimes produce a drastic effect on thermal stability of the bed
(R12; also Section IX).
The successive contact mechanism seems to provide a reasonable ap-
 FLUIDIZED CATALYST BEDS 413

proach to fluid catalyst bed design, by making it feasible to take into

consideration the combined influence of bed design and reaction kinetics
of a given chemical system on both optimum selectivity and bed stability.
Fluid beds with internals are frequently utilized for exothermic reac-
tions. Qualitative and quantitative effects of the internals design have
been discussed extensively (see Section 11; 12, 15, L12). The internals are
believed to increase /?,,a,, and perhaps v (note), and also to decrease the
catalyst content in the dilute phase by reducing both the circulation rate of
the dense phase and the ascending velocity of bubbles (M26). These ef-
fects are still under investigation.

IX. Nonisothermal Effect on the Bed Performance

In the dilute phase the contact efficiency is rather high but the rate of
heat transfer is low, and hence the temperature may not remain uniform in
this region. Thus, the temperature effect on conversion and selectivity
becomes important when a substantial part of the entire reaction occurs in
the dilute phase. This chapter discusses salient features of the tempera-
ture effect on fluid bed reactions.

A. EFFECT REACTION
ON STEADY

1. Bed Design and Local Reactivity

Several types of internals have been proposed so far. Some studies are
presented by Volket al. (V12)and Grekelet al. (G15), who studied several
arrangements of tubular internals and baffle trays. Volk suggested vertical
tubes to increase the contact efficiency. For fluidized beds with diameter
of 5-198 inches, conversions for beds of the same equivalent diameter (cf.
Section 1,C) were found equal. A large heat-exchange surface in the fluid
bed is necessary for non-isothermal reactions, so the use of vertical tubes
is quite practical.
Grekel et al. (G15) pursued the idea of vertical tube internals, but also
recognized the effect of the horizontal grid structure. They increased con-
tacting efficiency by the use of spaced and staggered horizontal tubes,
which could have the effect of increased direct contact below and just
above the grid. However, care must be taken in applying such a design,
because an unstable hot spot may form in a region of good gas-solid
contact and poor thermal conduction. This instability in the dilute phase,
which is discussed later, is more extreme in the vicinity of a grid. Horizon-
tal internals tend to decrease the flow circulation rate and to produce
414 MIYAUCHI, FURUSAKI, MOROOKA, A N D lKEDA

underlying freeboard space, thus sometimes permitting occurrence of a

temperature increase.
Grekel et al. also discussed the effect of the region of vicinity of the
distributor. This region is important, since there is some catalyst directly
contacting the entering gas phase (cf. Section VIII). Designs for the dis-
tributor region are proposed in many patents. However, the effect of the
distributor disappears within a few decimeters because of the violent
turbulence within the dense phase.
Furusakiet al. (F18) presented an increase of temperature in the bottom
part of the dilute phase for the reaction by activating fine particles. Re-
sults for homogeneous catalysts are shown in Fig. 82. Here, the tempera-
ture rise is not so significant as the case of activated fine particles (shown
in F18) because of the effective cooling through the wall of the reactor.
However, if the cooling surface area is small compared with the reactor
volume, the temperature rise will be more significant, especially in the
upper part of the dilute phase. In the lower part, the intensive longitudinal
mixing of solid particles prevents temperature from ascending signifi-
cantly. The necessity of cooling devices in the dilute phase will be shown
later. In some cases the internal cooler or an emergency shot of coolant is
provided. The effects of internals are also discussed by Overcashier et al.
(Oll), Morooka et al. (M42), and Nishinaka et al. (N6, N8, N9).
Heat transfer between the bed and wall is important in considering
temperature rise in the dilute phase (F16). This is discussed briefly in
Section VI,G.

2. Overall Conversion
The effect of the dilute phase on overall conversion of the reactant is
straightforward. Conversion in the bed increases if the contact efficiency
increases due to suspended catalyst in the dilute phase. If temperature

34 - 3 1 5 0.07
5
315-400.04 3
40 - 45 0.01 1.5
45 - 50 0.001 0.1
50 - 0.005 0.5 I
50 90
z (cm)

FIG.82. Temperature profile in a fluid bed. Standard catalyst 600 cm3. Bed diameter
5.5 cm. Solid line shows calculated value using Eqs. (9-7) and (9-8). T, = temperature at
the axis.
 FLUIDIZED CATALYST BEDS 415

increases in the dilute phase, conversion will further increase. Thus, from
the viewpoint of conversion of the reactant, it is recommended to use the
direct contact effect in the dilute phase and transition zone. Stability and
selectivity must now be discussed more thoroughly.

3. Effect on Selectivity
Industrially, selectivity is often as important as conversion in consider-
ing the efficiency of the reactor. In isothermal reactions, the dilute phase
and transition zone may cause better selectivity due to better contact in
that region. But in nonisothermal reactions, the effect will be different
because of the temperature effect. Mixing of gas and solids in the dilute
phase is not sufficient, and this may cause a temperature distribution for
exo- or endothermic reactions.
As an illustration, a simplified model will be considered here to show
the importance of the nonisothermal effect in the dilute phase. We assume
that cp 2 10, reaction in the bubble phase is negligible, and temperature in
the dense phase is uniform. Then the material balance equation for the
dense phase is:
uf(dcb/dz) + kbab(cb - ce) = 0 (9-1)
kbah(Ch - ce) - EehCe = (9-2)
and for the dilute phase,
U,(dc/dz) + qccdAre-E/RTc
=0 (9-3)
where the temperature is a radially averaged value. The above equations
are rearranged into dimensionless forms:

(dCb/dZ) + Nob(Cb - ce) = 0 (9-4)

Nob(Ch - ce) - EeNrCe = 0 (9-5)
(dC/dZ) - q,EedNAre-r’/BC
=0 (9-6)
In calculating temperature distribution in the dilute phase, solid motion
must be accounted for. Solid motion in the dilute phase is shown in
Sections I1 and VI. Laboratory-scale fluid beds exhibit a circulating
flow of solid particles with ascending central core and descending
peripheral region. The enthalpy balances for both ascending and de-
scending zones for the steady state are
416 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

Xmc,,p, Uf( dTd / d z ) + !tVcEed ( -A H)Ar C E I R T d C

+ hex&.x(Tu - Td) - hwaw(Td - Tw) = 0 (9-8)

where XU, is the superficial flow of solid particles; hxuexis the
volumetric heat exchange coefficient between the ascending and de-
scending emulsions; hwawmeasures heat transfer between the descending
emulsion and the wall. In deriving the above equations, the radial
concentration distribution and the flow of the descending gas are
neglected. Converting to dimensionless forms,
-(I -k hm)(d8,/dZ) + $7)cEedflNAre-r’eUC - NHex(eu- 8,) = 0 (9-9)
Am(d e d / d Z ) + $ q c E e d a N A r e-r’edc
+ NHex(8u - 8,) - N H w ( 8 d - Ow) = 0 (9-10)
where [equal to ( -AH)co/c,,p,Tde] denotes the dimensionless heat of
reaction and r (equal to E/RTde) is the dimensionless activation energy.
For an adiabatic condition, NHw = 0. Also, for NHex>> 1, 8, becomes
equal to 8,. Thus,
-(dO/dZ) + qcE&aNAre-r’eC= 0 (9-1 1)
For NHex = 0, the temperature of the ascending zone is given by
-(I + Am)(d&/dZ) + ir]&daNAre-r’eUC= 0 (9-12)
In the actual bed, solid exchange between the ascending and descending
zone is significant, and the circulation is seen as a series combination of
localized recirculation. If this scheme of solid mixing is more realistic, the
temperature profile may be suitably expressed in terms of an effective
thermal diffusivity, which will be used later for discussion of instability in
the dilute phase.
In this section, calculations were carried out for the case of q, = 1 ,
NHex >> 1 , and Eed = (Z - 4)&e/(l - ZJ. The last relation means
that catalyst density decreases linearly with bed height from the top of
the dense phase to some point (Z,) in the dilute phase. This 2, is not the
total bed height, but a hypothetical intermediate height where the amount
of catalyst particles becomes negligible.

a. Consecutive Reactions. The problem of selectivity for B in con-

secutive reactions (A I ,B f C) is referred to as selectivity of type I11 by
Wheeler (Wll). Wheeler’s analysis concerned the effect of pore diffusion
on selectivity of catalysis. According to him, selectivity decreases along
with the increase of diffusion resistance. In case of fluid bed reactors, the
mass-transfer resistance between bubble and emulsion phases causes
 FLUIDIZED CATALYST BEDS 417
similar effects; the selectivity of B is greater for the case of larger N & .
Also, in case of isothermal reactions it will be greater if e is chosen larger,
because direct contact is assured in the dilute phase. For nonisothermal
reactions the aforementioned expectation is not true because of the tem-
perature distribution existing in the dilute phase.
For this reaction system, rates of increase of A, B, C are given by R A =
-krlcA, RB = krlCA - kr2cB,and R , = kr2cB,respectively. The k,’s may
be expressed by the Arrhenius equation k , = A, exp( - E / R T ) . Calculations
were made according to the equations given above. Details of the equa-
tions for this special case are shown elsewhere (M28). Results for the
isothermal reactions are given in Fig. 83. Here, the rate of reaction 2 is
assumed to be smaller than that of reaction 1 by a factor of 10. For Noh <
3, the yield of B is higher for 2, = 2 than that at the exit of the dense
phase, i.e., Z, = 1. If Z , is larger, the amount of catalyst in the dilute
phase is large, and the yield of B becomes large, as is obvious from Fig.
83. Thus the use of the dilute phase improves selectivity for isothermal
consecutive reactions.
For nonisothermal reactions, the previous discussion does not hold.
Figure 84 shows the yield of B with respect to the dimensionless heat of
reaction 0.If R is positive, the reaction is exothermic; if negative-
endothermic; and if zero-isothermal. Values of R for real systems may
sometimes be even larger, but because of solid mixing the effective R
would be reduced. From the figure, it is seen that yield of B decreases
significantly with increasing R, and especially so for large activation
energy of reaction 2 (B + C) which proceeds more rapidly at higher tem-
peratures in the dilute phase. Many oxidation reactions operated in fluid

1.01
I n=o

-
1 2 3 L
Nob
1 2
FIG.83. Selectivity of isothermal consecutive reactions (A B +C).
418 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

I 2
FIG. 84. Selectivity of nonisothermal consecutive reactions (A + B +C).N,, = I,
Z, = 2, El = 83.6 kJ/mole, NArl = 2 X 108, NAa
= NArl/Io.

beds are exothermic with E2 > El and E2 >> 10 kcal/mole. In these reac-
tions it is profitable to suppress the role of the catalyst in the dilute phase.
In case of endothermic reactions, the yield of B does not vary signifi-
cantly from isothermal reactions. For E2 > 30 kcal/mole, the use of the
dilute phase even becomes profitable. Therefore, it may be desirable to
use the dilute phase in case of endothermic reactions.
The case of solids mixing with consecutive reactions is shown in Fig.
85. The magnitude of solid mixing is considerable (Npe= O.l), but the
selectivity is affected strongly by the heat of reaction, especially for the
case of highly exothermic reactions. The qualitative effect of dilute phase
on selectivity is not affected by solid mixing; therefore the effect will be
discussed for the case of negligible solid mixing, at smaller values of
effective R.
b. Denbigh's System with Optimal Temperature Distribution. Denbigh
(D11)proposed the following reaction system as a general case for organic
reactions;
3
AAX+Y
 FLUIDIZED CATALYST BEDS 419

1 2
FIG. 85. Variation of selectivity of consecutive reaction (A 4 B + C ) . N , , = 0,
NM1 = 2 x lo8, NAr2= 2 x lo', n2= a,, NP. = 0.1, El = E2 = 83.6 kJ/mole.

This is a system with an optimal temperature distribution, while the sys-

tem in the previous section is one with an optimal residence time. In this
system, the cases of (1) El C E2 and E3 > E4or (2) E, > E2and E3 < E4are
especially interesting. Several calculations were made for these cases;
details are not shown here (cf. M28). Generally speaking, the use of the
dilute phase is preferable for the case of the Denbigh system, except in the
case of exothermic reactions for El > E2 and E3 < E d . In this case, high
conversion in the dense phase is preferable, but increasing No,, does not
give significant improvement because the yield of P increases while that of
Q decreases. Therefore, operation with a large Uf may possibly be the
economically optimal step, even for the cases mentioned above. Care is
needed if Y decomposes further to form by-products; in this case, for
exothermic reaction of Y it is preferable to suppress the dilute phase for
the same reason as stated above for consecutive reactions.
c. Parallel Reactions wirh an Equilibrium. The parallel reaction
[A + X, A + Y] is another important reaction system. If the orders of
the two reactions are equal, selectivity does not depend upon the type of
reactor. An example of industrially important parallel reactions is
[A X, A 1 Y] with X the required product (D12). Examples of this
reaction system are hydrocarbon and methanol synthesis reactions, where
thermodynamically favorable side reactions are suppressed by the use of
420 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

appropriate catalyst. Packed-bed reactors with an internal cooling system

have been used for this type of reaction. As expected, X should be re-
moved from the reactor as soon as it is produced. In fluid beds, if reaction
1 is exothermic (AHl < 0), the equilibrium is unfavorable for high tem-
peratures. Thus, use of the dilute phase for exothermic reaction is un-
favorable and a large N o b is necessary in order for the reaction to proceed
in the dense phase. On the other hand, if reaction 1 is endothermic
(AHl > 0) the equilibrium is more favorable at higher temperature
(SZ > 0). Assuming SZ, = SZ3, the yield of the unwanted Y is plotted
against AH1 or Nob in Figs. 86 and 87. Here, input values of
NArl= 1.387 x 10l2 and El = 126 kJlmole give krlLf/Uf= 5.0 at
T = 573°K. If SZ = 0.1 then AH = - 17 kJ/mole at c o = 2.1 x lo-' mole/
cm3 (-lo%), and Tde = 573°K. It is shown that yield of X will decrease for
exothermic reactions with small values of Noh. Thus, in this reaction sys-
tem, a large N o b is preferred in order to obtain more desired product (X).

4. Discussion of Selectivity
In discussing the yield from a reactor, the temperature distribution
inside the reactor must be investigated. In the dense phase of fluid beds,
the heat-transfer coefficient between the bed and wall has been, widely
studied (L10, M14, M15, M16, T22, V5, W3, W5). Botterill (B12) has
reviewed the recent literature; studies of heat transfer in the dilute phase
are quite limited in number. Shirai (S9, S l l ) , Furusaki (F15), and
Morooka et al. (M5O) studied the heat-transfer coefficient in the dilute
phase as well as in the dense phase. They found that the heat-transfer
coefficient between the bed and the wall decreases as the bed density
decreases, which will cause an axial distribution of temperature in bed.

0-Lo' " ' 0

" " "50
A H [kcal /mole]

FIG.86. Effect of heat of reaction on selectivity for X in competitive reactions with an

I 3
equilibrium: A S X, A + Y. Parameters are same as in Fig. 87, except Nob = 1 .
2
 FLUIDIZED CATALYST BEDS 42 1

0.4,

os'l
OO
, , , , . ,
5
Nob
, , , , ,

FIG. 87. Effect of Nob on yield of byproduct Y in the competitive-reaction system

3
A X, A + Y. NMl = 1.387 x NA, = NArJIO, El = Es = 128 kJ/mole, Z , = 3,
2
KR = 10.

The influence of the temperature distribution on selectivity vanes ac-

cording to the reaction scheme. Among such schemes, the consecutive
reaction (A + B + C) qualitatively represents many organic reactions
with by-products. As shown in the previous section, the use of dilute
phase is recommended for endothermic reactions, but prohibited for
exothermic reactions. This conclusion agrees with the development of
fluid bed reactors for partial oxidations (exothermic) and cracking (en-
dothermic). This knowledge may help one to design or develop new fluid
bed contactors.

B. STABILITY BEDREACTORS
OF FLUID

In discussing the stability inside reactors, we must deal with several

stability problems, such as multiplicity of steady states, local stability
against perturbation, sensitivity to external conditions, oscillation, etc.
(F19, S S ) . Froment (F8) and Endoh et a!. ( E l l ) discussed sensitivity in
packed-bed reactors. Froment (F8) showed that multiple steady states and
oscillation are rare for catalytic packed-bed reactors, when (PeB), >> 1 and
Nre >> 1. In case of fluid bed reactors longitudinal mixing of solid particles
is sometimes much larger in magnitude, so NPe becomes much smaller
than (PeB), . This will introduce the possibility of multiple steady states
and of instability. On the other hand, stability against perturbation is in-
creased by solid mixing. Conversely, with less solid mixing, the steady
state is essentially unstable and perturbations lead to random fluctuations
(F15). Stability in the dense phase is discussed in other articles (B18, B19,
E10).
The most important stability problem in fluid beds is that of runaway
422 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

phenomena in the freeboard and also in the spaces under buffle plates
inside the dense bed-the so called dilute-phase regions. Good contact
and low heat-transfer rates are the major causes of runaway. The most
significant temperature rise is considered to occur by a coupling effect of a
medium degree of temperature rise due to good contact in the dilute phase
(including the transition zone) and of noncatalytic reactions which are
initiated by the temperature rise. Thus a model reaction to consider the
possibility of runaway is:

Here, A is the raw material, B is the required product, C and D are

by-products, but D is produced by noncatalytic reactions (3-3, such as
combustion or polymerization. The production of D is highly exothermic,
and also the contributing reactions have large activation energies. Hence
the production of D tends to give rise to violent reaction.
The material balances and enthalpy balance in the dilute phase are given
by the following equations in dimensionless form:

The boundary conditions for these equations are (at Z = 1) CA = CAo,

CH= CBo,C, = Cco, 6 = 1 and (at 2 = Z , ) d6/dZ = 0. For simplicity,
N A r 3 = NAr4 = NArS,r3= r4 = Ts,and rl = Tz are assumed. The Peclet
number Npeaccounts for the temperature diffusion due to solid mixing. If
N,, is zero, the temperature in the dilute phase is completely uniform;
NHw is the dimensionless heat-transfer coefficient; and sometimes the ef-
fect of radiative transfer must also be accounted for.
An example of concentration and temperature distribution is given in
Fig. 88. Reaction 1 is an ordinary reaction with the activation energy of
83.6 kJ/mole (20 kcdmole); the rate of reaction 2 is chosen as one-tenth of
that for reaction 1. The noncatalytic reactions 3-5 are rather violent with
the activation energy of 167.2 kJ/mole and heats of reaction (A + D) of
3 x lo3 kJ/mole. This reaction does not occur at the temperature of
the dense phase, but is quite active at higher temperatures because of the
 FLUIDIZED CATALYST BEDS 423

z (-1
FIG.88. Temperature rise by noncatalytic reaction with axial mixing of solid particles.
= 2 X lo', NAr2= NArl/lO,NAr3 = 1.08 x IOI3, El = E2 = 83.6 kJlmole, E3 = 167
kJ/mole, Npe = 0.1, N,, = SO, 8, = 0.9, a, = 0, = I , O3 = SO.

large activation energy. The value NHw= 50 is that obtained for the
dilute phase with an effective diameter of about 5 cm. The values
NArl= 2 x 1O'and El = 83.6 kJ/mole give k,,L,/Uf = 5 ; 113 = 50 means
-AH3 = 2800 kJ/mole at c,! = 6.4 x lo-' mole/cm3 (-30%) at 573°K
and 1 atm. From Fig. 88 it is obvious that there is considerable temperature
rise in the dilute phase, especially due to the noncatalytic reactions.
The reaction system is very sensitive to NPe and NHw. The stabilizing
effect of the dense phase, due to good thermal conductivity, is completely
lost in the dilute phase. If Npe is larger, the temperature rise at the bottom
of the dilute phase becomes so large as to be uncontrollable.
The case of large Nre may be studied by calculation for the case of
unmixed solid particles. Calculation of this case shows that cooling of the
dilute phase is essential for steady-state operation of the reactor for such
unstable reactions. It is hard to recommend using the dilute phase, be-
cause side reactions should be avoided (Fig. 89).
The extreme case of small N P eis the case of the thermally completely
mixed dilute phase. As shown above, multiple steady states are possible.
The analyses by heat generation and rejection curves (V4)are shown in
Figs. 90 and 91. Q is the dimensionless heat being generated or rejected,
and 8 is the dimensionless temperature in the dilute phase. The value of
NHw(Lt - L f ) / L ,is 100-200 for the transition zone and 30-50 for the dilute
phase. There are two stable steady states and one unstable point for E3 of
167 and 293 kJ/mole. If E3 is reduced to 40 kJ/mole, only one stable point
424 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

z [-I
FIG.89. Runaway in the dilute phase. NArl = 2 X 108, NA, = NArl/lO,NAr3= 2.92 X
l P 4 , N , = 31, R, = sl, = 1, sl, = 50, T, = IWC, N H w= 1.17 El = E2 = 83.6 kJ/mole,
E3 = 293 kJlmole.

0 [-I
FIG. 90. Thermal effect in the dilute phase, with thermally complete mixing.
El = E2 = 83.6 kJ/mole Es = 167 kJ/mole, Tde = 3WC, R, = $2, = I , sl, = 50,
NAr1 = 2 x lon, NArz = NArl/lO, NArS = I x L,/Lf= 3.42.8, = TIT,.

e 1-1

FIG.91. Thermal effect in the dilute phase, thermally complete mixing. Parameters
are same as those in Fig. 90 except Es = 293 kJ/mole NArs = 2.9 x
 FLUIDIZED CATALYST BEDS 425

is possible. Thus, spontaneous ignition is possible for large values of El,

and instability is probable for these unstable reaction systems even in the
case of thermal homogeneity.
In conclusion, the possibility of highly exothermic noncatalytic reac-
tions must be carefully eliminated, for all values of solid mixing. This is
done by the control of composition of the feed mixture so as, for example,
to suppress an excess of oxygen. Also, appropriate cooling is necessary in
the dilute phase.

X. Discussion and Summary

A. APPLICABILITY
O F REACTOR
MODELS
The main advantages of the reactor models are:
i. Reaction performance can be expressed by the minimum number
of measurable parameters.
u. The relationship between the parameters and variables, e.g.,
catalyst properties, reactor design, and operational variables, be-
... comes clear and easy to apply.
111. The equations obtained are simple and easy to use.

Most of the models for fluidized catalytic reactions have not devoted
enough attention to flow properties of FCBs. The important features of the
FCB are as follows: (a) Bubbles grow much more slowly than in teeter
beds; (b) A circulating flow exists, centrally upward and peripherally
downward; (c) Above the dense phase are the transition region and the
dilute phase, where particle density decreases gradually with height.
These properties introduce another uncertainty in developing a large-
scale FCB, which must be operated at high superficial velocity (about 50
c d s e c ) and at a large bed height (5-10 m), with more than 30 percent of
fines fraction. The successive contact mechanism (M25, M26) is the
simplest theory dealing with FCB flow properties. The necessary pararn-
eters to calculate reactor performance are as follows: (a) bubble diameter;
(b) particle distribution through the bed height, which may be obtained
from static pressure distribution; (c) contact efficiency distribution in the
transition region and the dilute phase. If relationships between these three
parameters and the operational variables are obtained from FCB perfor-
mance in more than two laboratory steps, it is possible to calculate the
same reaction schemes in large-scale fluidized catalyst beds.
The direct contact model of Lewis et al. (L12) is a special case of the
successive contact mechanism. If the effect of circulation of the emulsion
426 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

on reaction can be ignored, and contact efficiency in the dilute phase is

considered to be loo%, then the overall contact efficiencies by the two
models approximately coincide.
The successive contact mechanism and simplified theory of recircula-
tion provide much useful information concerning the FCB. For example,
Furusaki et al. (F18) showed experimentally, in hydrogenation of
ethylene, that the dilute phase plays an important role in the progress of
the reactions, and affects conversion and selectivity significantly. They
also showed that the fines fraction promoted the reaction in the dilute
phase, as well as improved the fluidity. Variables such as fraction of
catalyst in direct contact, improvement of contact efficiency, and temper-
ature control in the dilute phase were found to affect significantly the
extent of reaction in FCB. This result gives important perspective for
reactor design, as to size distribution of catalyst particles, control of activ-
ity distribution with respect to particle sizes, selection of superficial veloc-
ity of gaseous flow, and improvement of the internals arrangement.
The effect of dilute phase on selectivity depends on the type of reaction.
Generally, reaction systems with exothermic side reactions should avoid
use of the dilute phase. On the other hand, it is advantageous to enhance
the use of dilute phase for isothermal or endothermic reactions. Discus-
sions on this subject by Miyauchi and Furusaki (M28) are pertinent in the
design of FCBs.
According to the simplified theory of recirculation, the turbulent vis-
cosity of the fluid bed determines its flow pattern. Turbulent viscosity for
FCC particles may be given by a function of the bed diameter, as for
bubble columns of low-viscosity liquids. This relation will provide useful
information for practical design of FCBs, such as control of backmixing of
particles and gas, gas distribution from the distributor, design of the
sparger, control of aging of the catalyst, and prevention of erosion.

B. DEVELOPMENT
OF INDUSTRIAL
FLUIDIZED
CATALYST
BEDS

1. Fluid Catalytic Cracking of Petroleum

The initial application of fluidized beds in the petroleum industry was
the upflow dilute-phase reactor (M45). The obvious disadvantage of this
design is that all of the flowing catalyst passes overhead and must be
removed in dust-removal equipment. Later, the basic design which finds
widest application is the downflow dense-bed reactor (M2), which has the
following major advantages (K26):
i. Catalyst load to the cyclone decreases, and the cyclone may be
made smaller. Loss of catalyst and rate of corrosion decrease.
 FLUIDIZED CATALYST BEDS 427

ii. Catalyst inventory in the reactor is large enough to stabilize the

operation, so that liquid oil may be introduced.
iii. Plant size becomes smaller, and the cost of the construction de-
creases.
Catalyst also serves as a heat-transfer medium between the reactor and
the regenerator, circulating in a high flow rate between the two units.
Development of the microspherical catalyst in 1946, from AI2O3-SiO2sol
by means of spray drying, contributed to major advances in fluidized
catalyst-bed technology, especially for improvement of catalyst fluidity,
decrease of attrition loss, and decrease of erosion in transfer lines.
Recently, new upflow operation (riser cracking) has become popular,
through the development of highly active zeolite catalyst (V15).

2. Fluid Catalytic Reforming

Fluid catalytic reforming is a process of catalytic isomerization of a
naphtha fraction, obtained by direct crude distillation, to a high-octane-
number gasoline. The catalyst used is molybdena-alumina with a mean
particle diameter -60 pm. Superficially this process appears similar to
FCC, but there are some important differences. This process is operated
at a much higher pressure (about 14 atm); because the pressure drop in
catalyst recirculation is small compared with total pressure, control of
solid handling is sometimes difficult. Compared with catalytic cracking,
reaction rates are slow. Consequently the reactor is considerably larger
than the catalyst regenerator, and the resulting low catalyst recirculation
rate gives poor thermal efficiency (G3).

3. Production of Phthalic Anhydride

Catalytic oxidation of naphthalene to phthalic anhydride is the second
application of FCB which was initiated in 1945. Riley (R12) reported
problems on afterburning and on the control of bed temperature, and
introduced many improvements in the fluidized bed. However, the most
important change to make scale-up easier and to achieve good perfor-
mance was the use of microspherical catalyst, or moderately active
German-type catalyst (B 1 1, G 14).
This was the first application of a fluidized bed to a process that necessi-
tates removal of large heat of reaction (about 450 kcaygm-mole
naphthalene) and high yield. The fluidized process surpassed fixed-bed
processes in safe operation at high concentration, in yield, in reduced
pollution, and in plant cost-probably because the reaction was rather
simple and the products were stable. Fluid bed catalysts currently in use
428 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

consist principally of vanadium oxide on a silica gel base, and have a

particle-size range of less than 300 p m (G14).

4. Acrilonitrile by the Sohio Process

The Sohio process is considered one of the most successful applications
of FCB. Problems in industrial application of the reaction arose from the
strong exothermicity of propylene ammoxidation and from the intermedi-
ate production of acrylonitrile in the consecutive reactions (V9). It is par-
ticularly noticeable that the catalyst gives high selectivity, and the reactor
design aims at better fluidization and higher contact efficiency than in the
FCC process.
The first plant was put on stream in 1960. Since then, catalysts have
been modified, i.e., CAT-A (P * Mo Bi), CAT-21 (U Sb), and CAT-41
(P . Mo . Bi * Fe X) have been used successively; yield of acrylonitrile
based on propylene has also improved from about 57%, to 63%, and to
67% due to the modifications of catalyst (A8), all microspherical particles
with the properties of typical fluid bed catalysts (S15).

5 . Fischer-Tropsc h Synthesis
The synthesis of hydrocarbons from H2and CO is strongly exothermic.
The Hydrocol process mentioned in Section I is a modified Fischer-
Tropsch synthesis, using powdered iron catalyst in a dense-phase
fluidized bed. In spite of elaborate scale-up based on intensive research,
many problems arose in commercial plants, and the operation was termi-
nated in 1957 for economic reasons (Z5).
The major difficulty of this process arose from the physically and chemi-
cally unstable catalyst. The iron particles broke down in size very rapidly
under the synthesis conditions. The fluffy nature of the product made
fluidization difficult, if not impossible. Also, catalyst particles were
“waxed up,” i.e., coated with carbonaceous materials, and activity de-
creased significantly (S22).
Another difficulty of this process was low conversion in the commercial
plant, which could not be predicted from the results of the pilot plants.
Grekel et al. (G15) reported the effects of particle size distribution, gas
inlet devices, and internals, on contact efficiency. Volk et al. (V12) also
emphasized the effect of bed internals. These developmental studies be-
came a very useful guide for applications of fluidized catalyst beds.
The Kellogg process has been successful for its dilute-phase transfer-
line reactor in Sasol’s South African plant. The void fraction in the reac-
tion zone is more than 95%, and the superficial gas velocity ranges from 3
to 12 d s e c (G2). Careful consideration is thus necessary in applying FCB
 FLUIDIZED CATALYST BEDS 429

to processes where the catalyst can suffer physical and chemical attrition,
and is so heavy that good fluidization is hard to attain.

6. The Two Newest Applications in Japan

Oxychlorination of ethylene to produce ethylene dichloride was put on
stream in 1969 by Mitsui Toatsu Chemicals. Formerly, chlorine and/or
alkyl chlorides had been produced in fluidized beds by the Shell process
(A7, E12, F4). The previous catalyst had been made by dipping and
calcining CuC1, onto various carriers. The present catalyst is made by
spray-drying a gel of mixed CuClz and AI2O3,followed by calcining, and
has a particle size distribution suitable for fluid beds. It is very active,
and has a long life and high attrition resistance (M19).
The reasons that fluid beds are applied to this process are: (a) The high
heat-transfer capacity is appropriate for this highly exothermic system.
(b) The reaction product is stable, and consecutive reactions do not occur.
(c) Operational safety can be expected. Scale-up was successful with the
use of baffles to prevent channeling and of a gas distributor to obtain high
contact efficiency. Several steps of pilot-plant test were examined in de-
veloping the process. Performance of the commercial unit revealed the
overall yield of 97%, which had been anticipated (M18).
Another example is the production of maleic anhydride by Mitubishi
Chemical Industries, Ltd. in 1970, which is oxidation of BB-fraction.
Synthesis of maleic anhydride from butylene is a typical consecutive reac-
tion, and it was considered difficult to get high yield by the use of fluid
beds. However, it is highly exothermic and application of fluid beds is
preferable.
In scaling-up this process, appropriate reactor design was conducted
elaborately by the results of several tests. The following two points are
characteristic of the scale-up. One is the development of catalyst suitable
for fluid beds. Selectivity should not decrease by the increase of hy-
drocarbon concentration. Also, reaction rate of the secondary reaction
should be small. These problems were solved by the development of
rather simple P-V catalyst (K2, K3). Another device is the instantaneous
mixing inlet mechanism for hydrocarbon and air. By this, activity decline
of the catalyst can be prevented (T12). Thus a process has been developed
which is equally as profitable as the benzene process.

c. RECENT
TRENDS
IN F L U I D I Z ECATALYST
D BEDS
Recently many inventions about fluidized catalyst beds, which are use-
ful in industrial applications, have been reported (16-8), initiating trends
430 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

counter to traditional technology since FCC. Some of the important im-

provements are listed with patent references:
(a) Improvements in use of catalyst:
i. Effective execution of consecutive reactions by mixed fluidization
of catalysts of different activities (German Offen 2603770, Japan
75-30825).
ii. Accomplishment of good fluidization by means of mixed fluidiza-
tion of inactive particles of different size and of different specific
gravity (Japan 73-3 1827, 74-27263).
(b) Improvements in feed systems of reactant gases:
i. Prevention or decrease of decline of catalyst activity by instan-
taneous mixing of reactant gases (Japan 74-29166, 74-29167, 74-
29 168).
ii. Simultaneous manufacture of plural main products by simulta-
neous feed of different main reactants (Britain 1238347, Japan 76-
166 15).
iii. Establishment of uniform fluidization by introducing a reactant
into the middle stage of fluidized beds (South Africa 7308033, Bri-
tain 1238347).
iv. Increase of selectivity by means of split feed of reactants (U.S.A.
3546268; Britain 1208191).
(c) Improvements in internals:
i. Establishment of good fluidization by a modified arrangement of
vertical internals (Japan 75-15772, 77-7905).
ii. Increase of contact efficiency by means of improved arrangement
of sieve trays (Japan 74-47725, 74-47726).
iii. Decrease of backmixing by providing horizontal ridges on the ver-
tical surfaces (Britain 1359377).
(d). Prolonged catalyst activity:
Providing an autoregeneration zone in fluid beds and improving its
effect (Britain 1126617, Japan 76-291 14).
(e) Modification to decrease the carry-over of catalyst particles:
Curtailing carry-over by arranging the internals in the upper part of
the dense phase (i.e., the transition region) and in the dilute phase
(U.S.A. 3859405).

D. TECHNICAL
PROBLEMS
I N FCB DESIGN

Marshall (M2) presented the following six factors in the design of

downflow fluidized solid reactors.
1 . Required reaction volume: This is decided by experimental determi-
nation of the necessary space velocity. It will be influenced by the extent
 FLUIDIZED CATALYST BEDS 43 1

of gas mixing and gas backmixing and by the ratio of length to diameter of
the bed.
2. Desired diameter: This is determined by gas velocity, which will be
in the range between the minimum fluidization velocity and the velocity
determined by excessive entrainment. The most economic velocity will be
set by balancing gas circulation cost with other costs.
3. Type of gas distributor: There are several types such as conical
bottom, packed bed, and grid.
4. Freeboard above the reactor bed: Determined by the need for dust-
removal equipment, and for free settling prior to entry to that equipment.
5 . Disengaging-space diameter: The enlarged diameter will lower the
gas velocity, aid settling of particles, and reduce carry-out.
6. Equipment required for heat exchange, if any.

Even now, the above considerations are the basis for FCB design.
Elaborate devices are installed on the gas distributor and internals for
high-performance reactions (Section I).
A panel discussion on fluid particle technology (Philadelphia, 1973) was
significant in correctly recognizing the central problems in industrial a p
plication of fluidized beds to catalytic reactions (H2). Some of the com-
ments are communicated in the following paragraphs.
Optimum size distribution is important for a fluid bed reactor (Bergoug-
nou). Models based on bubbles are not yet capable of predicting the wall
effect (Wen). Vertical baffles are most effective in breaking up large bub-
bles (Volk). The height of the bottom ends of vertical tube bundles above
the grid will set the attainable bubble size at the bottom of the bundle. The
bundles then essentially maintain the bubble size (Zenz). Horizontal per-
forated baffle plates reduce the mean residence time of elutriable fine
particles in a fluidized bed (Buckham). Observations on attrition in cy-
clones indicate that it is an exponential function of velocity (Tenney).
In general, industrial application of FCBs is possible from pilot-plant
tests (including cold runs) of appropriate scale, with the 35-year experi-
ence of industrial plants and research since the initiation of FCC. How-
ever, the following conditions are necessary. First, the catalyst should
have such properties and activity as to be suitable for FCB. Second, the
heat of reaction must be appropriate for use in FCB, considering heat
transfer and the heat capacity of fluidized beds. Third, the reaction prod-
ucts must be stable, and high selectivity must not be required if consecu-
tive reactions are involved. Generally, the catalyst is in an equilibrium
state in industrial reactors due to physical and chemical impacts, and
artificial controls from outside corresponding to those impacts.
Problems on the optimization of industrial fluidized catalyst beds are
still left unsolved. Certainly optimization in a rigorous sense is impossible.
432 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

However, approximate treatment is possible. Ikeda and Tashiro (19) re-

port an optimization of catalytic reactions in fluid beds. They find that the
maximum yield of the intermediate product decreases, and that the op-
timum contact time increases for first-order consecutive and parallel reac-
tion systems if contact efficiency in the reactor decreases. They also
showed the most economical “equilibrium activity” and the optimal size
distribution of catalyst.
Modern technology on fluidized catalyst beds started late in the 1930s
with the pioneer work by W.K. Lewis and E. R. Gilliland, and has found
a place in some large-scale industrial applications. It will find further
applications, as its phenomenological and theoretical background become
clearer. Specific flow properties which distinguish the fluidized catalyst
beds from the usual teeter beds have been highlighted. The study of bed
properties has been advanced by the concept of successive contact mech-
anisms. However, the science of fluid catalyst beds is still young, and
much is still unknown about their physical and chemical behavior. The
technology, following the trends reported in this review, has benefited
from constantly improving technical knowledge. In particular, better fluid-
ity of the beds has been found to lead to higher reactor efficiency.
ACKNOWLEDGEMENTS
The authors wish to express their thanks to Professor Y. Kato, Kyushu University, for
stimulating discussions on this article, and to Professor T. Vermeulen for his valuable
comments and help in editing the manuscript. Thanks are also to Mr. T. Kikuchi of the
University of Tokyo for drawing figures, and to Misses K. Sekiguchi and T. Hongo,
University of Tokyo, for typing manuscripts.

Nomenclature

A, Free area of baffles; also, volume of bed

frequency factor in the Ar- B Width of bed
rhenius plot C,C,,CB
AT Cross-sectional area of fluid C,,C, Concentration in dilute phase
bed (dimensionless)
a Fraction of catalyst in direct Cb Concentration of free gas in
contact with gas phase, bubble phase (dimension-
Eq. (7-35) less)
ab Contact area of bubbles per C, Concentration of free gas in
unit volume of bed emulsion phase (dimen-
u, Specific surface area of clus- sionless)
ter phase c Molar concentration; also
aex Area for the exchange of ma- concentration in freely
terial between ascending flowing gas in dilute phase
and descending emulsion c b Concentration in bubble
a, Area of heat transfer per unit phase
 FLUIDIZED CATALYST BEDS 433

Concentration in cloud- E, Apparent longitudinal dis-

overlap phase; also con- persion coefficient of gas in
centration in cluster phase emulsion phase based on
(Section VIII) empty vessel
Concentration in emulsion Eef r,E, in case of fine particles
phase E,, E,,E,, in case of fine particles
Concentration in descending ED Eotvos number, d:gp,/a
emulsion (dimensionless)
Concentration in ascending g, Radial diffusivity
emulsion E, Cross-sectionally averaged
Concentration in fluid phase radial diffusivity
in emulsion E,, Lateral dispersion coefficient
Concentration at Lf of solids based on empty
Concentration in inlet gas vessel
Concentration in outlet gas E, Axial diffusivity defined by
Apparent heat capacity of Eq. (6-25)
emulsion E,, Axial diffusivity by the
Heat capacity of fluid Taylor dispersion
Heat capacity of gas E, Total axial diffusivity Eq.
Heat capacity of liquid (4-2)
Heat capacity of solid E,, ErT of liquid in bubble col-
Concentration in solid phase umn
in emulsion E,, E, of emulsion in fluidized
Effective gas diffusivity in bed
particle 8, Cross-sectionally averaged
Equivalent bed diameter axial dithsivity
Effective gas diffusivity in e Fraction of catalyst in dilute
emulsion phase with respect to vol-
Equivalent bed diameter on ume of catalyst in emul-
horizontal surfaces sion phase, Eq. (7-25)
Equivalent bed diameter on eb Refer to Eq. (3-20) (dimen-
vertical surfaces sionless)
Gas diffusivity F Gas exchange rate between
Diameter of nozzle the bubble and emulsion
Diameter of bed phase based on the volume
Diameter of bubble of the emulsion after
Effective bubble diameter Lewis et al. (L12)
Initial bubble diameter f(c) kor(z)/r(c)
Maximum bubble diameter F(d,) Distribution function of
Maximum stable bubble di- bubble sizes existing in
ameter given by Davidson bed
and Harrison (D3) f, Ratio of wake volume to the
Steady bubble diameter in total bubble volume
turbulent bubble flow Gr, Grashof number, zb&/&
Diameter of particle (dimensionless)
Number-averaged mean g Gravitational acceleration
bubble diameter H Height of freeboard
Volume-surface mean bubble (&JaPP Apparent overall height of a
diameter transfer unit refer to Eq.
Activation energy (6-3)
434 MIYAUCHI, FURUSAKI, MOROOKA, AND IKEDA

HoR Overall height of a transfer [(kobabl-' + (&k)-']-'

unit for whole bed refer to Eq. (7-4)
AH Heat of reaction refer to Eq. (8-14)
hb Bubble-side heat-transfer Rate constant for first-order
coefficient irreversible reaction based
hex Heat-exchange coefficient on volume of emulsion
between ascending and phase
descending regions (c.g.s. Reaction rate constant for
units) reactions 1, 2 . . .
hob Overall heat-transfer coeffi- Effective thermal diffusivity,
cient from bubble to emul- CIS PSEZS%e
sion phase Height of mixed phases
Wall-to-bed heat-transfer Equivalent height of emul-
coefficient sion phase
Parameter defined in Eq. Height of dense-phase fluid
(6-15) bed
Overall reaction rate con- Probe position measured
stant based on settled bed from the distributor
Equilibrium constant (dimen- Height of fluid bed at Urn[
sionless) Height of settled bed
Numerical coefficient (di- Total bed height
mensionless); also wave Distance between stages
number Distance between observa-
Wave number tion points; also length of
Mass-transfer coefficient in bubble signals
bubble phase Arithmetical mean length of
Mass-transfer coefficient de- bubble signals
fined by Eq. (6-19) Distribution function of b u b
Mass-transfer coefficient de- ble signals measured with
fined by Eq. (6-20) a point-shaped probe
Mass-transfer coefficient in Adsorption equilibrium con-
emulsion phase stant for emulsion, equal
Effective thermal con- to ere + m e w ; also, cusps/
ductivity in emulsion
phase *lk
Mass-exchange coefficient Adsorption equilibrium con-
between ascending and stant for catalyst particles
descending emulsion (cse I c i e )
Mass transfer coefficient be- Number of horizontal baffle
tween gas and particle plates
Thermal conductivity of gas Arb/ uf
Numerical constant, Eq. k,a,4/ VG
(4- IS) (dimensionless) mkxaexbluf
Overall mass-transfer coeffi- h w a w b l UrcAsPs
cient between bubble and Lthexaex 1Ufiwps
emulsion phase kDbabLflUf
Mass-transfer coefficient to L iuccuapglks
or from cluster phase Radial mass flux in Eq. (4-4)
Mass-transfer coefficient be-
tween jet and emulsion
 FLUIDIZED CATALYST BEDS 435

m U,l CJG ity of gas in emulsion

Growth factor of distur- phase
bances; also constant de- Superficialnet liquid velocity
fined by Eq. (2-13) Minimum bubbling velocity
Number of holes of perfo- Minimum fluidization veloc-
rated plate ity
Static pressure Superficial intermixing ve-
Peclet number, rn 6*ri,/4dbDerr locity of solids through
UG&/E, baffle plate
UG4.r/mEZ,s, Superficial circulation veloc-
Pressure drop of bed ity of solids in emulsion
Specific converting power; phase
also net transport defined Tangential velocity along the
by Eq. (4-7);also dimen- bubble-emulsion interface
sionless heat generated or Time-averaged interstitial
rejected in Figs. 90-91 velocity of liquid
Flow rate of gas in gas cloud Interstitial mean liquid ve-
phase locity, Eq. (4-15)
Flow rate of solids in gas Ascending velocity of bub-
cloud phase bles
Radius of bed; also gas con- UG/Eb,mean bubble velocity
stant; also column radius Bubble velocity along the
Reaction rates column axis
Coordinate of radial position Rising velocity of finite-size
r atu = 0 bubble swarm
Radius of bubble Circulation velocity of emul-
reaction rate sion
Radius of gas cloud Circulation velocity of emul-
Radius of particle sion along column axis
Temperature; also radially Velocity of jet
averaged temperature in Circulation velocity of liquid
dilute phase Circulation velocity of liq-
Temperature of coolant uid along column axis of
Temperature in descending bed
zone of dilute phase Area-averaged circulation
Temperature in dense phase velocity of liquid
Temperature in ascending Mean interstitial velocity of
zone of dilute phase upflow, Eq. (5-9)
Temperature at heat-transfer Interstitial velocity of gas in
well emulsion phase
Characteristic contact time 11 along the column axis
of pockets Slip velocity of bubble rela-
Superficial gas velocity as tive to liquid
defined by Eq. (3-26) Mean slip velocity of bubble
Velocity as defined by Eq. Free-rising velocity of single
(5-14) bubble
Superficial circulation veloc- Terminal velocity of a parti-
ity of emulsion Ze,ce, cle
Superficial gas velocity Absolute liquid velocity at
Superficial circulation veloc- column wall
436 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

Radial velocity fluctuation of 4 u Volume fraction of ascending

liquid emulsion, 6"+ ced = e,
Axial velocity fluctuation of Cr Volume fraction of gas,
liquid €f + cs = 1
u6 Velocity at y = 6, Eq. (3-7) Volume fraction of gas in
vb Volume of a bubble emulsion, ere + cEC= I
Vo g G / 1 9 2 v , , characteristic ve- Local liquid holdup (dimen-
locity sionless), cb + cL = 1
V* Frictional velocity Averaged value of el,
We Weber number, dbpld/u Rate of energy dissipation
(dimensionless) per unit mass
Coordinate of position Volume fraction cf at Umr
R-r Volume fraction of solids,
ZlL, Pf + cs = 1
ZJLf ese Volume fraction of solids in
Coordinate of bed height emulsion, ere + c,, = 1
Length of space existing di- r) Correction factor defined by
rectly under horizontal Eq. (622a)
bame plate r)C Contact efficiency
Total bed height 0 Polar coordinate with origin
at bubble center, also di-
GREEKLETTERS mensionless temperature
(T/Td,), also time
a kJumf;also numerical coeffi- h Wavelength, 2a/&;also ratio
cient of order unity (di- of superficial flow of solids
mensionless) to that of gas in dilute
p Hatta number; also numeri- phase
cal coefficient of order Minimum wavelength, Eq.
unity (dimensionless) (5-18)
pr Modified Hatta number for Ao Minimum eddy size
the emulsion Lc Volume ratio of catalyst in
p$ refer to Eq. (8-16) bubble phase
r E/RTde Apparent viscosity of fluid
6 Thickness of gas cloud; also bed
thickness of laminar sub CLB Viscosity of gas
layer PL Viscosity of liquid
c Entrainment rate P P Effective Newtonian shear
cb Local gas-bubble holdup viscosity of particulate
ib Averaged value of cb fluid
c, Volume fraction of emulsion, PI Turbulent viscosity
ce +
cb = 1 v Fraction of catalyst particles
i, Averaged value of e, contacting directly with
Volume fraction of descend-
ced gas
ing emulsion; also ce in di- Fraction of catalyst in direct
lute phase contact with the bubble
idAveraged ce in dilute gas
phase; also averaged value v in the dilute phase
of ced Turbulent diffusivity of mass,
i& Averaged E, in dense phase v, = av,
 FLUIDIZED CATALYST BEDS 437

uM Molecular kinematic viscos- cp Parameter defined by Eq.

ity (7-18)
u( Turbulent kinematic viscos- x Tortuosity
ity d~ Angle measured away from
6 Defined by Eq. (7-28) the vertical (radians)
p,, Bed density (-AH)CPlCpgPgT~,
pp Density of emulsion phase in w Frequency of wave
fluid bed
pr Density of fluid SUBSCRIPTS
pg Density of gas
p, Density of liquid b bubble
pp Density of particles including d descending flow
inner pore volume de dense phase
pQ Density of settled bed di dilute phase
ps Density of solids e emulsion
u Interfacial tension ex exchange
u2 Variance f fluid phase
T Shearing stress g gas phase
Time available for growth H horizontal, also heat
Maximum value for T, L, I liquid
Local surface renewal time M molecular
Growth time of disturbances o overall
Time constant for particle p particle
Shear stress at column wall r radial
Rate of reaction or mass s solid, or slip
transfer t turbulent
Radial coordinates r / R (di- u ascending flow
mensionless) V vertical
Radial coordinates r*/R (di- w wall
mensionless) z axial

References
Al. Ahlborn, F., 2. Tech. Phys. 12, 482 (1931).
A2. Akehata, T., Shirai, T., Sugita, M., Yoshino, K., and Shirai, K., Hokkaido Meet.
Soc. Chem. Eng. Jpn. Preprinr No. B8 (1968).
A3. Akita, K., and Yoshida, F., Ind. Eng. Chem. Process Des. Dev. 12, 76 (1973).
A4. Akita, K., and Yoshida, F., Ind. Eng. Chem. Process Des. Dev. 13, 84 (1974).
AS. Alexander, B. F., and Shah, Y. T., Chem. Eng. J . 11, 153 (1976).
A6. Anderson, T. T., AlChE J . 10, 776 (1964).
A7. Anonymous, Chem. Eng., Oct., 392 (1953).
A8. Anonymous, Hydrocarbon Process. 46, 141 (1967); 48, 146 (1969); 52, 99 (1973).
A9. Aoyama, Y., Ogushi, K., Koide, K., and Kubota, H., J . Chem. Eng. Jpn. 1, 158
(1968).
AIO. Argo, W. B., and Cova, D. R . , Ind. Eng. Chem. Process Des. Dev. 4, 352 (1965).
A1 I. Ark, R., Proc. R . SOC.A235, 67 (1956).
A12. Askins, J. W., Hinds, G. P., Jr., and Kunreuther, F., Chem. Eng. Progr. 47, 401
( 1951).
438 MIYAUCHI, FURUSAKI, MOROOKA, AND IKEDA

B1. Baeyens, J., and Geldart, D., “Fluidization and Its Applications,” p. 263. Toulouse,
France, 1973.
B2. Bailie, R. C., Chung, D. S., and Fan, L. T., Ind. Eng. Chem. Fundam. 2,245 (1963).
B3. Baird, M. H. I., and Rice, R. G., Chem. Eng. J. 9, 171 (1975).
B4. Bakker, P. J., and Heertjes, P. M., Chem. Eng. Sci. 12, 260 (1%0).
B5. Bartholomew, R. N., and Casagrande, R. M., Ind. Eng. Chem. 49, 428 (1957).
B6. Basov, V. A., Makhevka, V. I., Malik-Akhnozarov, T. Kh., andOrochko, D. I., Int.
Chem. Eng. 2, 263 (1%2).
B7. Bedura, R., Deckwer, W. D., Warnecke, H. J., and Langemann, H.,Chem. Ing.
Tech. 46, 399 (1974).
B8. Behie, L. A., Doctoral dissertation, Western Ontario University, 1972.
B9. Behie, L. A., and Kehoe, P., AIChE J. 19, 1070 (1973).
B10. Bellman, R.,and Pennington, R. H., Q.Appl. Math. 12, 151 (1954).
B11. Betts, W. D., Ind. Chem. July, 370 (1963).
B12. Botterill, J. S. M., “Fluid-Bed Heat Transfer.” Academic Press, New York, 1975.
B13. Botton, R. J., Chem. Eng. Prog. Symp. Ser. 66(101), 8(1970).
B14. Bridge, A. G., Lapidus, L., and Elgin, J. C., AIChE J. 10, 819(1%4).
B15. Broadhurst, T. E., and Becker, H. A., AIChE J . 21, 238 (1975).
B16. Brotz, W., Chem. Ing. Tech. 28, 165 (1956).
B17. Burgess, J. M., and Calderbank, P. H., Chem. Eng. Sci. 30, 1511 (1975).
B18. Burkur, D., and Amundson, N. R., Chem. Eng. Sci. 30, 847, 1159 (1975).
B19. Burkur, D., Wittmann, C. V., and Amundson, N. R., Chem. Eng. Sci. 29, 173(1974).
B20. Bart, R., Sc.D. Thesis, Mass. Inst. of Tech, Cambridge, Mass., 1950, cited by
G. H. Reman (R4).
B21. Bohle, W., and Swaay. W. P. M., “Fluidization,” Proc. 2nd. Eng. Foundation
Conf. 1978, p. 167, Cambridge Univ. Press, London and New York (1978).
B22. Bukur, D. B., Ind. Eng. Chem. Fundam. 17, 120 (1978).
C1. Callahan, J. L., Grasselli, R. K., Milberger, E. C., and Strecker, H. A., Ind. Eng.
Chem. Prod. Res. Dev. 9, 134 (1970).
C2. Carlsmith, L. E., and Johnson, F. B., fnd. Eng. Chem. 37, 451 (1945).
C3. Chiba, T., and Kobayashi, H., Chem. Eng. Sci. 25, 1375 (1970).
C4. Chiba, T., Terashima, K., and Kobayashi, H., Chem. Eng. Sci. 27, 965 (1972).
C5. Chiba, T., Terashima, K., and Kobayashi, H., J . Chem. Eng. Jpn. 6, 78 (1973).
C6. Clift, R., Grace, J. R., and Weber, M. E., Ind. Eng. Chem. Fundam. 13, 45 (1974).
C7. Clift, R., and Grace, J. R., Chem. Eng. Prog. Symp. Ser. 116 67, 23 (1971).
C7a. Chavarie, C., and Grace, J. R., Ind. Eng. Chem. Fundam. 14, 75, 79, 86 (1976).
C7b. Chavarie, C., and Grace, J. R., Chem. Eng. Sci. 31, 741 (1976).
C8. Cleland, F. A., and Wilhelm, R. H., AlChE J. 2, 489 (1956).
C9. Calderbank, P. H., and Toor, F. D., in “Fluidization” (J. F. Davidson and D. Hani-
son, eds.), p. 383. Academic Press, New York, 1971.
DI. Danckwerts, P. V., Trans. Faraday SOC.46, 300 (1952).
D2. Danckwerts, P. V., Chem. Eng. Sci. 2, 1 (1953).
D3. Davidson, J. F., and Harrison, D., “Fluidized Particles.” Cambridge Univ. Press,
London and New York, 1963.
D4. Davidson, J. F., and Harrison, D., Chem. Eng. Sci. 21, 731 (1966).
D5. Davidson, J. F., and Harrison, D. (eds.), “Fluidization.” Academic Press, New
York, 1971.
D6. Davies, L., and Richardson, J. F., Trans. Inst. Chem. Eng. 44, T293 (1966).
D7. de Groot, J. H., Proc. Int. Symp. Fluidization, Eindhoven, p. 348.
 FLUIDIZED CATALYST BEDS 439

D8. de Vries, R. J., van Swaaij, W. P. M., Mantovani, C., and Heijkoop, A., Proc. 5th
Eur., 2nd I n t . Symp. Chem. React. Eng. B9-59 (1972).
D9. Deckwer, W. D., Burckhart, R., and Soll, G., Chem. Eng. Sci. 29, 2177 (1974).
D10. DeMaria, F.,and Longfield, J. E., Chem. Eng. Prog. Symp. Ser. 38 58, 16 (1962).
DII. Denbigh, K. G., Chem. Eng. Sci. 8, 125 (1958).
D12. Denbigh, K. G., “Chemical Reactor Theory.” Cambridge Univ. Press, London and
New York, 1965.
D13. Dimotakis. P. E., and Brown, G . L., J . Fluid Mech. 78, 535 (1976).
D14. Deckwer, W., Graeser, U., Langemann. H., and Serpemen, Y., Chem. Eng. Sci. 28,
1223 (1973).
D14a. Donsi, G., and Massirnilla, L., “Fluidization and Its Applications,” p. 41. Capadues,
Toulouse, 1973.
D15. Donsi, G., and Massimilla, L., AIChE J. 19, 1104 (1973).
D16. Drinkenburg, A. A. H., Proc. Int. Symp. Fluidization, Eindhoven, p. 468 (1967).
D17. Drinkenburg, A. A. H., and Rietema, K., Chem. Eng. Sci. 27, 1765 (1972).
D18. Drinkenburg, A. A. H., and Rietema, K., Chem. Eng. Sci. 28, 259 (1973).
D19. Darton, R. C., Trans. Inst. Chem. Eng. 57, 134(1979).
D20. De Lasa, H. I. and Grace, J. R., AIChE J. 25, 984 (1979).
El. Eberly, P. E., J. Phys. Chem. 65, 68 (1961).
E2. Eberly, P. E., J . Phys. Chem. 66, 812 (1962).
E3. Eberly, P. E., and Kimberlin, C. N., Trans. Faraday Soc. 57, 1169 (1961).
E4. Echigoya, E.,Toyoda, K., and Morikawa, K., Kagaku Kogaku 32, 364 (1968).
E5. Echigoya, E., Iwasaki, M., Kanemoto, T., and Niiyama, H., Kagaku Kogaku 32,
571 (1%8).
E6. Eissa, S. H., El-Halwagi, M. M., and Saleh, M. A., Ind. Eng. Chem. Process Des.
Dev. 10, 31 (1971).
E7. Eissa, S. H., and Schiigerl, K . , Chem. Eng. Sci. 30, 1251 (1975).
E8. El Halwagi, M. M., and Gomezplata, A,, AIChE J. 13, 503 (1967).
E9. El Nashaie, S., Chem. Eng. Sci. 32, 295 (1977).
E10. El Nashaie, S., and Yates, J. G., Chem. Eng. Sci. 28, 515 (1973).
E l l . Endoh, I., Furusawa, T., and Matsuyama, H., Kagaku Kogaku 41, 232 (1977).
E12. Engel, W. F., and Waale, M. J., Chem. Ind. Jan. 76 (1962).
Fl. Fair, J. R., Lambright, A. J., and Anderson, J. W., Ind. Eng. Chem. Process Des.
Dev. 1, 33 (1962).
F2. Fan, L. T., Lee, C. J., and Bailie, R. C., AIChE J. 8, 239 (1962).
F3. Finnerty, R. G., Maa, J. R., Vossler, A. M., Yeh, H. S., Crouse, W. W., and Rice,
W. J., Ind. Eng. Chem. Fundam. 8, 271 (1%9).
F4. Fleurke, K. H., Chem. Eng. March, CE41 (1968).
F5. Fournol, A. B., Bergougnou, M. A., and Baker, C. G. J., Can. J. Chem. Eng. 51,401
(1973).
F6. Frantz, J. F., Chem. Eng. 69, Sept., 161 (1962).
F7. Freedman, W., and J. F. Davidson, Trans. Inst. Chem. Eng. 47, T251 (1969).
F8. Froment, G. F., Adv. Chem. Ser. 109 p. 1 (1972).
F9. Fryer, C., Ph.D. thesis, Monash University, Australia, 1974.
F10. Fryer, C., and Potter, 0. E., Ind. Eng. Chem. Fundam. I t , 338 (1972).
F11. Fryer, C., and Potter, 0. E., “Fluidization and Its Applications,” p. 440. Cepadues,
Toulouse, 1973.
F12. Fryer, C., and 0. E. Potter, AIChE J. 22, 38 (1976).
F13. Furusaki, S., Kagaku Kogakic 32, 1033 (1968).
440 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

F14. Furusaki, S . , AIChE J. 19, 1009 (1973).

F1S. Furusaki, S., Dr. Eng. dissertation, University of Tokyo, 1976.
F16. Furusaki, S., and Miyauchi, T., Annu. Meet. SOC. Chem. Eng. Jpn. 42nd, Reprint
(1977).
F17. Furusaki, S., Kikuchi, T., and Miyauchi, T., Autumn Meet., 9th. Soc. Chem. Eng.
Jpn. Reprint (1975).
F18. Furusaki, S.,Kikuchi, T., and Miyauchi, T., AIChE J. 22, 354 (1976).
F19. Furusawa, T., Endoh, I., and Matsuyama, H., Kagaku Kogaku 33, 949 (1969).
F20. Furukawa, J., and Ohmae, T., Ind. Eng. Chem. 50, 821 (1958).
GI. Gabor, J. D., AIChE J. 10, 345 (1964).
G2. Garrett, L. W., Chem. Eng. Prog. 56, (4), 39 (1960).
G3. Geldart, D., Chem. Ind. 1474 (1967).
G4. Geldart, D., Powder Technol. 6, 201 (1972).
G5. Geldart, D.,Powder Technol. 7, 285 (1973).
G6. Gill, W. N., and Sankarasubramanian, R.,Proc. R. Soc. (London) Ser. A 316, 314
( 1970).
G7. Gilliland, E. R., and Knudsen, C. W., Chem. Eng. Prog. Symp. Ser. 116 67, 168
(1971).
G8. Gilliland, E. R., and Mason, E. A., Ind. Eng. Chem. 41, 1191 (1949).
G9. Gilliland, E. R.,and Mason, E. A., Ind. Eng. Chem. 44, 218 (1952).
G9a. Gilliland, E. R., Mason, E. A., and Oliver, R. C., Ind. Eng. Chem. 45, 1177 (1953).
G9b. Gilliland, E. R.,N . J . Section Meet. AIChE May (1953); cited by W. F. Pansing (Pl).
GIO. Gondo, S., Tanaka, S., Kazikuri, K., and Kusunoki, K., Chem. Eng. Sci. 28, 1437
(1973).
(311. Gordon, A. L., and Amundson, N. R.,Chem. Eng. Sci. 31, 1163 (1976).
G12. Goto, Y.,Okamoto, T., and Terahata, T.,Annu. Meet. Soc. Chem. Eng. Jpn. 34th,
April (1%9).
G13. Grace, J. R.,Can. J . Chem. Eng. 48, 30 (1970).
G14. Graham, J. J., and Way, P. F., Chem. Eng. Prog. 58, (l), 96 (1962).
G15. Grekel, H., Hujsak, K. L., and Mungen, R., Chem. Eng. Prog. 60, (I), 56 (1964).
G16. Guigon, P., Bergougnou, M.A., and Baker, C. G. J., AIChE Symp. Ser. 141 70, 63
( I 974).
(317. Gunn, D.J., Chem. Eng. (London) CE 153 (I%@.
G18. Gometzplata, A., and Shuster, W. W., AIChE J. 6 , 454 (1960).
(319. Griffith P., and Wallis, G. B.,Trans. Am. Soc. Mech. Eng. 83C, 307 (1961).
(320. Grace, J. R., and De Lasa, H. I., AIChE J. 24, 364 (1978).
(321. George, S. E., and Grace, J. R., AIChE Symp. Ser. No. 176, 74, 67 (1978).
H1. Hall, C. C., and Taylor, A. H.,J. Inst. Petrol. 41 (376), 102 (1955).
H2. Halow, J. S.,AIChE Symp. Ser. 141 70 (141), l(1974).
H3. Hanada, K., B. Eng. thesis, Dept. of Applied Chem., Kyushu University, 1975.
H4. Harberman, W. L., and R. K. Morton, Trans. Am. SOC. Civil Eng. 121, 227 (1956).
H5. Harmathy, T. Z., AIChE J. 6, 281 (1960).
H6. Hamson, D.,and Leung, L. S . , Trans. Inst. Chem. Eng. 39, 409 (1961).
H7. Henriksen, H. K., and Qstergaard, K., Chem. Eng. Sci. 29, 626 (1974).
H8. Hetzler, R.,and Williams, M. C., Ind. Eng. Chem. Fundam. 8, 668 (1969).
H9. Hills, J. H., Trans. Inst. Chem. Eng. 52, 1 (1974).
H10. Hills, J. H., and Darton, R. C., Trans. Inst. Chem. Eng. 54, 258 (1976).
H11. Hikita, H., and Kikukawa, H., Chem. Eng. J. 8, 191 (1974).
H12. Hinze, J. O., AIChE J. 1, 289 (1955).
H13. Hiraki, I., and Kunii, D.,Kagaku Kogaku 33, 680 (1969).
 FLUIDIZED CATALYST BEDS 44 1

H14. Hiraki, I., Yoshida, K., and Kunii, D., Kagaku Kogaku 29, 846 (1965).
HIS. Hirama, T., Ishida, M., and Shirai, T., Kagaku Kogaku Ronbunshu 1, 272 (1975).
H16. Hovmand, S., and Davidson, J. F., Trans. Inst. Chem. Eng. 46, T190 (1968).
H17. Hovmand, S.,and Davidson, J. F., “Fluidization” (J. F. Davidson and D. Harrison,
eds.), p. 193. Academic Press, 1971.
H18. Hurt, D. M., Znd. Eng. Chem. 35, 522 (1943).
H19. Handlos, A. E., Kunstman, R. S., and Schissler, D. O., Ind. Eng. Chem. 49, 25
(1957).
H20. Hagyard, T., and Sacerdote, A. M., Ind. Eng. Chem. Fundam. 5, 500 (1966).
H21. Horio, M., and Wen, C. Y., AIChE Symp. Ser. No. 161, 73, 9 (1977).
H22. Hoebink, J. H. B. J., and Rietema, K., “Fluidization,” Proc. 2nd. Eng. Foundation
Conf. 1978, p. 327, Cambridge Univ. Press, London and New York (1978).
11. Ikeda, Y., Kagaku Kogaku 27, 667 (1963).
12. Ikeda, Y.,Kagaku Kogaku 29, 57 (1965).
13. Ikeda, Y., Kagaku Kojo 9, (12), 63 (1965).
14. Ikeda, Y., Kagaku Kikai Gijutsu 18, 191 (1%5).
15. Ikeda, Y., Kagaku Kogaku 34, 1013 (1970).
16. Ikeda, Y., Kemikaru Enjiniyaringu 973 (1977).
17. Ikeda, Y., J. Jpn Petrol. Inst. 20, 619 (1977).
18. Ikeda, Y., and Tashiro, S . , Kagaku Kogaku 29, 956 (1965).
19. Ikeda, Y., and Tashiro, S., “Saikin no Kagaku Kogaku,” p. 195. Maruzen, Tokyo,
1%7.
110. Ikeda, Y.,Tashiro, S., Yamaguchi, T., and Kawai, M., Annu. Meet. SOC. Chem.
Eng. Jpn. 35th, Preprint D202 (1970).
I l l . Ishii, T., and Osberg, G. L., AIChE J . 11, 279 (1965).
112. Ivanov, M. E., and Bykov, V. P., Theor. Found. Chem. Eng. 4, 119 (1970).
113. Iwasaki, M.,Furuoya, I., Sueyoshi, H.,Shirasaki, H., and Echigoya, E., Kagaku
Kogaku 29, 892 (1965).
J1. Jackson, R., Trans. Inst. Chem. Eng. 41, 21 (1963).
J2. Johnstone, H. F., Batchelor, J. D., and Shen, C. Y., AIChE J . I, 318 (1955).
K1. Kaji, H., M.S. thesis, Dept. of Chem. Eng., University of Tokyo, 1%6.
K2. Kamimura, S . , J . Jpn. Petrol. Inst. 20, 627 (1977).
K3. Kamimura, S . , Kagaku Kogaku 33, 1082 (1973).
K4. Kato, K., and Wen, C. Y., Chem. Eng. Sci. 24, 1351 (1969).
K5. Kato, Y.,Kagaku Kogaku 27, 366 (1963).
K6. Kato, Y., and Nishiwaki, A., Kagaku Kogaku 35, 912 (19711, I n t . Chem. Eng. 5,
182 (1972).
K7. Kato, Y., Nishinaka, M., and Morooka, S . , Kagaku Kogaku Ronbunshu 1, 530
( 1975).
K8. Kehoe, P. W.K., and Davidson, J. F., AIChE Symp. Ser. 128 69, 34 (1973).
K9. Kehoe, P. W.K., and Davidson, J. F., AIChE Symp. Ser. 128 69, 41 (1973).
K10. Kimura, T., M.S. thesis, Dept. of Chem. Eng., University of Tokyo, March, 1972.
K11. Klinkenberg, A., Jr., Trans. I n s t . Chem. Eng. 43, TI41 (1965).
K12. Kobayashi, H., “Saikin no Kagaku Kogaku,” p. 17. SOC.Chem. Eng., Japan, 1966.
K13. Kobayashi, H., Arai, F., and Chiba, T., Kagaku Kogaku 29, 858 (1965).
K14. Kobayashi, H., and Arai, F., Kagaku Kogaku 29, 885 (1965).
K15. Koide, K.,Morooka, S., Ueyama, K., Matsuura, A., Yamashita, F., Iwamoto, S.,
Kato, Y., Inoue, H., Suzuki, S.,and Akehata, T., J. Chem. Eng. Jpn. 12, 98 (1979).
K16. Kojima, E., Akehata, T., and Shirai, T., J . Chem. Eng. Jpn. 8, 108 (1975).
K17. Kolbel, H., Beinhauer, R.,and Langemann, H., Chem. Ing. Tech. 44,697 (1972).
 442 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

K18. Kolbel, H., Borchers, E., and Muller, K., Chem. Ing. Tech. 30, 729 (1958).
K19. Kolbel, H., Borcher, E., and Martins, J., Chem. Ing. Tech. 32, 84 (1960).
K20. Kumar, A., Degaleesan, T. E., Laddha, G.S.,and Hoelscher, H. E., Can. J. Chem.
Eng. 54, 503 (1976).
K21. Kumar, R.,and Kuloor, N. R., “Advances in Chemical Engineering,” Vol. 8, p. 256.
Academic Press, New York, 1970.
K22. Kunii, D., and Levenspiel, O., Ind. Eng. Chem. Fundam. 7, 446 (1968).
K23. Kunii, D., and Levenspiel, O., Ind. Eng. Chem. Process Des. Dev.7, 481 (1968).
K24. Kunii, D., and Levenspiel, O., “Fluidization Engineering.” Wiley, New York, 1969.
K25. Kunii, D., and Levenspiel, O., J. Chem. Eng. J p n . 2, 84 (1969).
K26. Kunii, D., Akiyama, T., and Takagi, K., “Shin Kagaku Kogaku Koza,” IV-5, Fluidi-
zation. Nikkan-Kogyo-Shimbun, Tokyo, 1957.
K27. Kunii, D., Yoshida, Y., and Hiraki, I., Proc. Int. Symp. Fluidization, Eindhoven,
p. 243 (1967).
K28. Kono, T., B. Eng. thesis, Dept. of Appl. Chem., Kyushu University, 1976.
K29. Khan, A. R., Richardson, J. F., and Shakiri, K. J., “Fluidization,” Proc. 2nd.
Eng. Foundation Conf. 1978, p. 351, Cambridge Univ. Press, London and New
York (1978).
, K30. Kikuchi, T., private communication (1977).
L1. LaNauze, R. D., Powder Technol. 15, 117 (1976).
L2. Lanneau, K. P., Trans. Inst. Chem. Eng. 38, 125 (1960).
L3. Larroux, G . J., Kim, Y. G.,and Bankoff, S . G.,Chem. Eng. Sci. 27, 447 (1972).
L4. Latham, R. L., and Potter, 0. E., Chem. Eng. J. 1, 152 (1970).
L5. Latham, R. L., Hamilton, C., and Potter, 0. E.,Br. Chem. Eng. 13, 666 (1%8).
L6. Leva, M., Chem. Eng. Prog. 47, 39 (1951).
L7. Leva, M., “Fluidization.” McGraw-Hill, New York, 1959.
L8. Leva, M., and Grummer, M., Chem. Eng. Prog. 48, (6), 307 (1952).
L9. Levenspiel, O., and Bischoff, K. B., “Advances in Chemical Engineering,” Vol. 4,
p. 95. Academic Press, New York, 1963.
L10. Levenspiel, O., and Walton, J. S . , Chem. Eng. Prog. Symp. Ser. 50, l(1954).
L11. Lewis, W. K., Gilliland, E. R., and Girouard, H., Chem. Eng. Prog. Symp. Ser. 38
58, 87 (1962).
L12. Lewis, W. K., Gilliland, E. R., and Glass, W., AIChE J . 5, 419 (1959).
L13. Lewis, W. K., Gilliland, E. R.,and Lang, P. M., Chem. Eng. Prog. Symp. Ser. 38 58,
65 (1962).
MI. MIT, A Bibliography of Fluidization Research at the Massachusetts Institute of
Technology, 1939- 1955.
M2. Marshall, S . , Chem. Eng. May, 219 (1953).
M3. Massimilla, L., AIChE Symp. Ser. 128 69, 11 (1973).
M4. Massimilla, L., and Johnstone, H. F., Chem. Eng. Sci. 16, 105 (1961).
M5. Massirnilla, L., and Westwater, J. W., AIChEJ. 6, 137 (1960).
M6. Massimilla, L., Donsi, G.,and Zucchini, C., Chem. Eng. Sci. 27, 2005 (1972).
M7. Matheson, G. L., Herbst, W. A., and Holt, P. H., Ind. Eng. Chem. 41, 1099 (1949).
M8. Mathis, J. F., and Watson, C. C., AIChE J. 2, 518 (1956).
M9. Matsen, J. M., Chem. Eng. Prog. Symp. Ser. I01 66, 47 (1970).
M10. Matsen, J. M., AIChE Symp. Ser. 128 69, 30 (1973).
M11. Matsen, J. M., and Tarmy, B. L., Chem. Eng. Prog. Symp. Ser. 101 66, 1 (1970).
M12. May, W. G., Chem. Eng. Prog. 55, (12), 49 (1959).
M13. Merrick, D., and Highley, J., AIChE Symp. Ser. 137 70, 336 (1974).
M14. Mickley, H. S., and Fairbanks, D. F., AlChE J. 1, 374 (1955).
 FLUIDIZED CATALYST BEDS 443

M15. Mickley, H. S., and Trilling, C. A., Ind. Eng. Chem. 41, 1135 (1949).
M16. Mickley, H. S . , Fairbanks, D. F., and Hawthorn, R. D., Chem. Eng. Prog. S y m p .
Ser. 32 57, 51 (1961).
M17. Miwa, K., Mon, S., Kato, T.,and Muchi, I., Kagaku Kogaku 35, 770 (1971); I n t .
Chem. Eng. 12, 181 (1972).
M18. Miyauchi, Takeshi, and Oyamada, N., Jpn. Petrol. Inst. 20, 624 (1977).
M19. Miyauchi, T., Sato, Y.,Michiki, H., and Fujimoto, K., Japan Patent 45-39616, 1970.
M20. Miyauchi, Terukatsu, “Ryukeisosa-to-Kongotokusei.” Nikkan-Kogyo-Sinbun,
Tokyo, 1960.
M21. Miyauchi, T., Annu. Meet. SOC. Chem. Eng. Jpn. 30th. Preprint No. 6110 (1965).
M22. Miyauchi, T., Annu. Meet. SOC.Chem. Eng. Jpn. 3Ist. Preprint No. 244 (1966).
M23. Miyauchi, T., J. Chem. Eng. Jpn 4, 238 (1971).
M24. Miyauchi, T., CHISA Congr., 4th. Prague Sept. (1972).
M25. Miyauchi, T., J. Chem. Eng. Jpn. 7 , 201 (1974).
M26. Miyauchi, T., J . Chem. Eng. Jpn. 7 , 207 (1974).
M27. Miyauchi, T., Fall Meet. SOC. Chem. Eng. Jpn. 11th. Invited Lecture, Oct. (1977).
M27a. Miyauchi, T., and Kikuchi, T., Annu. Meet. SOC. Chem. Eng. Jpn., 42nd,
Hiroshima April (1977).
M28. Miyauchi, T., and Furusaki, S., AIChE J . 20, 1087 (1974).
M29. Miyauchi, T., and Morooka, S., Kagaku Kogaku 33, 369 (1969); Int. Chem. Eng. 9,
713 (1969).
M30. Miyauchi, T., and Morooka, S., Kagaku Kogaku 33, 880 (1969).
M31. Miyauchi, T., and Shyu, C. N., Kagaku Kogaku 34, 958 (1970).
M32. Miyauchi, T., and Yamada, K.,Kagaku Kogaku 35, 547 (1971).
M33. Miyauchi, T., Kaji, H., and Saito, K., J . Chem. Eng. Jpn. 1, 72 (1968).
M34. Mori, S., and Muchi, I., Kagaku Kogaku 34, 510 (1970).
M35. Mori, S., and Muchi, I., J . Chem. Eng. Jpn. 5, 251 (1972).
M36. Mori, S., and Wen, C. Y., A K h E J. 21, 109 (1975).
M37. Mori, Y.,and Nakamura, K., Kagaku Kogaku 29, 868 (1965).
M38. Morooka, S., Dr. Eng. dissertation, University of Tokyo, 1969.
M39. Morooka, S., and Miyauchi, T., Kagaku Kogaku 33, 569 (1969).
M40. Morooka, S., Kato, Y., and Miyauchi, T., J. Chem. Eng. Jpn. 5 , 161 (1972).
M41. Morooka, S., Nishinaka, M., and Kato, Y., Kagaku Kogaku 37, 485 (1973).
M42. Morooka, S., Nishinaka, M., and Kato, Y.,Kagaku Kogaku Ronbunshu 2,71(1976);
Int. Chem. Eng. 17, 254 (1977).
M43. Morooka. S., Tajima, K., and Miyauchi, T., Kagaku Kogaku 35, 680 (1971); Int.
Chem. Eng. 12, 168 (1972).
M44. Morse, R. D., and Ballow, C. O., Chem. Eng. Prog. 47, (9), 199 (1951).
M45. Murphree, E.V., Brown, C. L.,Fischer, H. G. M., Gohr, E. J., and Sweeney, W. J.,
Ind. Eng. Chem. 35, 768 (1943); Chem. Tech. 6, 523 (1976).
M46. Murray, J. D., J . Fluid Mech. 21, 465 (1965).
M47. Murray, J. D., J. Fluid Mech. 22, 57 (1965).
M48. Muchi, I., Mem. Fac. Eng. N a g o y a Univ. 17, (I), 79 (1965).
M49. Muchi, I., Mamuro, T., and Sasaki, K., Kagaku Kogaku 25, 757 (1961).
M5O. Morooka, S., Maruyama, Y., Kawazuishi, K.,Higashi, S., and Kato, Y., Kagaku
Kogaku Ronbunshu 5, 275 (1979).
M51. Morooka, S., Kawazuishi, K. and Kato, Y., Powder Technol. 26, 75 (1980).
NI. Nguyen, H. V., and Potter, 0. E.,Adv. Chem. Ser. 133, 290 (1974).
N2. Nguyen, H. V.,and Potter, 0. E., “Fluidization Technology,” Vol. 2, p. 193. Hemi-
sphere, Washington, D.C., 1976.
444 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

N3. Nguyen, X. T., Bergougnou, M. A., and Baker, C. G. J., Can. J. Chem. E n g . 51,573
(1973).
N4. Nicklin, D. J., Chem. Eng. Sci. 17, 693 (1962).
N5. Nicklin, D. J., Wilkes, J. O., and Davidson, J. F., Trans. Inst. Chem. Eng. 40, 61
( 1962).
N6. Nishinaka, M., Morooka, S., and Kato, Y., Powder Technol. 9, I (1974).
N7. Nishinaka, M., Morooka, S.,and Kato, Y., Kagaku Kogaku Ronbunshu 1,81(1975).
N8. Nishinaka, M., Morooka, S., and Kato, Y., Kagaku Kogaku Ronbunshu 1, 605
(1975).
N9. Nishinaka, M., Morooka, S., and Kato, Y., Kagaku Kogaku Ronbunshu 2, 96(1976).
N10. Nguyen, H. V., Whitehead, A. B., and Potter, 0. E., “Fluidization,” Proc. 2nd.
Eng. Foundation Conf. 1978, p. 140, Cambridge Univ. Press, London and New
York (1978).
01. Ogura, Y., “Taiki-Ranryu-Ron” (Atmospheric Turbulence), 5th. ed., pp. 2-7, 68.
Chijin-Shokan, Tokyo, 1959.
02. Ohki, Y., and Inoue, H., Chem. Eng. Sci. 25. 1 (1970).
03. Oki, K., and Shirai, T., “Fluidization Technology,” Vol. 1, p. 95. Hemisphere,
Washington, D.C., 1976.
04. O h , H. L., and Dean, 0. C., Petrol. Eng. March, C-23 (1953).
05. Oltrogge, R. D., Preliminary outline of thesis, Technical University of Eindhoven and
University of Michigan, 1971.
0 6 . Orcutt, J. C., Davidson, J. F., and Pigford, R. L., Chem. Eng. Prog. Symp. Ser. 38
58, 1 (1962).
07. Ormiston, R. M., Mitchell, F. R. G., and Davidson, J. F., Trans. Inst. Chem. Eng. 43,
T209 (1965).
08. Osberg, G. L., and Charlesworth, D. H., Chem. Eng. Prog. 47, 566 (1951).
09. Ostergaard, K., “Advances in Chemical Engineering,” Vol. 7, p. 71. Academic
Press, New York, 1968.
010. Othmer, D. F., “Fluidization,” Reinhold, New York, 1956.
011. Overcashier, R. H., Todd, D. B., and Olney, R. B., AIChE J. 5, 54 (1959).
012. Ogasawara, S., Kihara, M., Nishiyama, M., Shirai, T., and Morikawa, K., Kagaku
Kogaku 28, 59 (1964).
PI. Pansing, W. F., AIChE J . 2, 71 (1956).
P2. Partridge, B. A., and Rowe, P. N., Trans. Inst. Chem. Eng. 44, T335 (1966).
P3. Pavlov, V. P., Khim. Prom. (9), 698 (1965).
P4. Perkins, T. K., and Johnston, 0. C., SOC. Petrol. Eng. J . 70 (March 1963).
P5. Potter, 0. E., and Thiel, W. “Fluidization Technology,” Vol. 2, p. 185. Hemisphere,
Washington, D.C., 1976.
P6. Pozin, L. S.,Aerov, M. E., and Bystrova, T. A., Theor. Found. Chem. Eng. 3,714
(1969), (English trans.).
W.Prandtl, L., ZAMM 22,241 (1942); see H. Schlichting, “Boundary Layer Theory,”
Chap. 19, 4th English ed. McGraw-Hill, New York, 1960.
P8. Pyle, D. L., and Harrison, D., Chem. Eng. Sci. 22, 1199 (1967).
P9. Pyle, D. L., and Rose, P. L., Chem. Eng. Sci. 20, 25 (1965).
PIO. Pyle, D. L., and Stewart, P. S . B., Chem. Eng. Sci. 19, 842 (1964).
P11. Pyle, D. L., and Harrison, D., Chem. Eng. Sci. 22, 531 (1967).
R1. Raghuraman, J., and Verma, Y.B. G., Chem. Eng. Sci. 30, 145 (1975).
R2. Reichart, H., VDI Forschungsh. (Berlin) 414 (1942).
R3. Reith, T., Renken, S.,and Israel, B. A., Chem. Eng. Sci. 23, 619 (1968).
R4. Reman, G. H., Chem. Ind. Jan., 46 (1955).
 FLUIDIZED CATALYST BEDS 445

R5. Rice, W. J., and Wilhelm, R. H., AIChE J. 4, 423 (1958).

R6. Richardson, L. F., Proc. R. Soc. (London) Ser. A 110, 709 (1926).
R7. Richardson, J. F., and Zaki, W. N., Trans. Inst. Chem. Eng. ’32,35 (1954).
R8. Rieke, R. D., and Pigford, R. L., AIChE J. 17, 10% (1971).
R9. Rietema, K., Proc. Inr. Symp. Fluidization Eindhoven p. 154 (1967).
RIO. Rietema, K., Proc. Int. Symp. Fluidization Eindhoven p. 176 (1967).
RII. Rietema, K., and Hoebink, J., “Fluidization Technology,” Vol. 1, p. 279. Hemi-
sphere, Washington, D.C., 1976.
R12. Riley, H. L., Trans. Inst. Chem. Eng. 37, 305 (1957).
R13. Rowe, P. N., Chem. Eng. Sci. 31, 285 (1976).
R14. Rowe, P. N., and Everett, D. J., Trans. Inst. Chem. Eng. 50, 55 (1972).
R15. Rowe, P. N., and Partridge, B. A., Chem. Eng. Sci. 18, 511 (1%3).
R16. Rowe, P. N., and Partridge, B. A,, Trans. Inst. Chem. Eng. 43, T157 (1965).
R17. Rowe, P. N., Partridge, B. A., and Lyall, E., Chem. Eng. Sci. 19, 973 (1964).
R18. Rietema, K., Chem. Eng. Sci. 34, 571 (1979).
R19. Rowe, P. N., MacGillivray, H. J., and Cheesman, D. J., Trans. Inst. Chem. Eng.
57, 194 (1979).
R20. Rowe, P. N., Santoro, L., and Yates, J. G., Chem. Eng. Sci. 33, 133 (1978).
SI. Sada, E., “Kiho-Ekiteki-Kagaku.” SOC.Chem. Eng. Japan, 1969.
S2. Sankarasubramarian, R., and Gill, W. N., Proc. R . SOC. (London) Ser. A 333, 115
( 1973).
S3. Saxton, J. A., Fitton, J. B., and Vermeulen, T., AIChE J . 16, 120 (1970).
S4. Schlichting, H., “Boundary Layer Theory,” 4th ed. McGraw-Hill, New York, 1960.
S5. Schmitz, R. A., “Chem. Reaction Eng. Reviews” (H. H. Hulburt, ed.), Adv.
Chem. Series 148, 156 (1975).
S6. Schiigerl, K., Proc. In?. Symp. Fluidization, Eindhoven, p. 782. (1%7).
S7. Shen, C. Y., and Johnstone, H. F., AIChE J . 1, 349 (1955).
S8. Shenvood, T. K., Pigford, R. L., and Wilke, C. R., “Mass Transfer.” McGraw-Hill,
New York, 1975.
S9. Shirai, T., Kagaku Kogaku 26, 637 (1962).
S10. Shirai, T., “Ryudoso,” 2nd ed. Kagaku-gijutsusha, Kanazawa, 1973.
SI 1. Shirai, T., Yoshitome, H., Shoji, Y., Tanaka, S., Hojo, K., and Yoshida, S., Kagaku
Kogaku 29, 880 (1965).
S12. Shrikhande, K. Y., J. Sci. Ind. Res. 14B, 457 (1955).
S13. Shyu, C. N., and Miyauchi, T., Kagaku Kogaku 35, 663 (1971).
S14. Squires, A. M., Chem. Eng. Prog. 58, (4), 66 (1962).
S15. Standard Oil Co. (Br.) 1, 126, 617 (1968).
S16. Steiner, H. H., Ind. Eng. Chem. 36, 618,840 (1944).
S17. Stemerding, S., see Reman, G. H., Chem. Ind. Jan., 46 (1955).
S18. Stephens, G. K., Sinclair, R. J., and Potter, 0. E., Powder Technol. 1, 157 (1967).
S19. Stewart, P. S. B., Ph.D. dissertation, Cambridge University, 1965.
S20. Stewart, P. S. B., Trans. Inst. Chem. Eng. 46, T60 (1968).
S21. Stewart, P. S. B., and Davidson, J. F., Powder Technol. 1, 61 (1967).
S22. Storch, H. H., Glumbic, N., and Anderson, R. B., “The Fischer-Tropsch and
Related Syntheses.” Wiley, New York, 1951.
S23. Sit, S. P. and Grace, J. R., Chem. Eng. Sci. 33, 1115 (1978).
TI. Tadaki, T., and S. Maeda, Kagaku Kogaku 25, 254 (1961).
T2. Takamura, T., B. Eng. thesis, Dept. Chem. Eng., University of Tokyo, 1970.
T3. Talmor, E., and Benerati, R. F., AIChE J. 9, 536 (1963).
T4. Tam, Le Van, and Miyauchi, T., Kagaku Kogaku 35, 650(1971).
446 MIYAUCHI, FURUSAKI, MOROOKA, AND IKEDA

T5. Takahashi, M., M.S. thesis, Dept. Chern. Eng., University of Tokyo, 1977.
T6. Tanaka, I., and Shinohara, H., J. Chem. Eng. Jpn. 5, 57 (1972).
“7. Tanaka, I., Shinohara, H., Hirose, H., and Tanaka, Y.,J. Chem. Eng. Jpn. 5, 5 1
( 1972).
T8. Taylor, G. I., Proc. R . SOC. (London) Ser. A201, 192 (1950).
T9. Taylor, G . I., Proc. R . SOC. (London) Ser. A219, 186 (1953).
T10. Taylor, G. I., Proc. R . SOC.(London) Ser. A223, 446 (1954).
TI1. Taylor, G. I., Proc. R . SOC.(London) Ser. A225, 473 (1954).
T12. Terahata, T., Tazawa, S., Kakizaki, T., Minoda, N., Miyazima, S., Ito, M.,Imai, H.,
Kawazu, T., Japan patent No. 49-29166, 1974.
T13. Tett, H. C., “Fluidization,” p. 1’. SOC.Chem. Ind., London, 1%4.
T14. Tichacek, L. J., Barkelew, C. H., and Baron, T., AIChE J. 3, 439 (1957).
T14. Todt, J., Liicke, J., Schiigerl, K., and Renken, A., Chem. Eng. Sci. 32, 369 (1977).
T16. Toei, R., and R. Matsuno, Proc. I n t . Symp. Fluidization, Eindhoven, p. 271 (1967).
T17. Toei, R., Matsuno, R., Oichi, M., and Yamamoto, K., J . Chem. Eng. Jpn. 7 , 447
( 1974).
T18. Toei, R., Matsuno, R., Hotta, H., Oichi, M., and Fujine, Y.,J. Chem. Eng. Jpn. 5,
273 (1972).
T19. Toei, R., Matsuno, R., Kojima, H., Nagai, Y.,Nakagawa, K., and Yu,S . , Kagaku
Kogaku 29, 851 (1965).
T20. Twi, R., Matsuno, R., Nishitani, K., Hayashi, H., and Imamoto, T., Kagaku
Kogaku 33, 668 (1969).
T21. Tone, S . , Seko, H., and Otake, T., J. Chem. Eng. Jpn. 7, 44 (1974).
T22. Toomy, R. D., and Johnstone, H. F., AIChE Symp. Ser. 5 49, 51 (1953).
T23. Towell, G. D.,Strand, C. P., and Ackerman, G. H., Paper presented at Am. Inst.
Chem. Eng. and Inst. Chem. Eng. Joint Meeting, p. 1272, London (1965).
T24. Towell, G. D., and Ackerman, G. H., Proc. 5th Eur., 2nd Int. Symp. Chem. Reaction
Eng. Amsterdam B3-1 (1972).
T25. Trawinski, H., Chem. Ing. Tech. 23, 416(1951).
T26. Trawinski, H., Chem. Ing. Tech. 25, 201 (1953).
T27. Tsutsui, T., and Miyauchi, T., Kagaksu Kogaku Ronbunshu 5, 40 (1979).
T28. Turner, J. C. R., Chem. Eng. Sci. 21, 971 (1%6).
T29. Tsutsui, T., M. S. thesis, Dept. of Chem. Eng., University of Tokyo, 1974.
T30. Tellis, C. B., and Hulburt, H. M., Proc. Pacgc Chem. Eng. Con$ 1st SOC. Chem.
Eng. Jpn. AIChE Kyoto Session 8, p. 178 (1972).
UI. Ueyama, K., and Miyauchi, T., Kagaku Kogaku Ronbunshu 2,430 (1976).
U2. Ueyama, K., and Miyauchi, T., Kagaku Kogaku Ronbunshu 2, 595 (1976).
U3. Ueyama, K., and Miyauchi, T., Kagaku Kogaku Ronbunshu 3, 19 (1977).
U4. Ueyama, K., and Miyauchi, T., Kagaku Kogaku Ronbunshu 3, I15 (1977).
U5. Ueyama, K., and Miyauchi, T., AIChE J . 25, 258 (1979).
V1. Van Deemter, J. J., Chem. Eng. Sci. 13, 143 (1961).
V2. Van Deemter, J. J., Proc. Int. Symp. Fluidization Eindhoven p. 334 (1967).
V3. Van der Laan, Chem. Eng. Sci. 7 , 187 (1958).
V4. Van Heerden, C., Ind. Eng. Chem. 45, 1242 (1953).
V5. Van Heerden, C., Nobel, P., and van Krevelen, D. W., Chem. Eng. Sci. 1,51(1951).
V6. Van Krevelen, D. W., and Hoftijzer, P. J., Chem. Eng. Prog. 46, (1). 29 (1950).
V7. Van Swaaij, W. P. M., and Zuiderweg, E J., Proc. 5th Eur. 2nd Int. Symp. Chem.
Reaction Eng. B9-25 (1972).
V8. Van Swaay, W. P. M., and Zuiderweg, F. J., “Fluidization and Its Applications,”
p. 454. Cepadues, Toulouse, 1973.
 FLUIDIZED CATALYST BEDS 447

V8a. Van Swaaij, W. P. M., “Chem. Reaction Eng. Reviews,” (D. Luss and V. W.
Weekman Jr., eds.), A m . Chem. SOC. Symp, Ser., 72, 193 (1978).
V9. Veatch, F., Callaham, J. L., Idol, J. D., and Milberger, E. C., Perol. ReJner 41, ( l l ) ,
187 (1962).
VIO. Verma, Y. B. G., Powder Technol. 12, 167 (1975).
V l l . Vermeulen, T., Moon, J. S., Hennico, A., and Miyauchi T., Chem. Eng. Prog. 62,
(9), 95 (1966).
V12. Volk, W., Johnson, C. A., and Stotler, H. H., Chem. Eng. Prog. 58, (9,44 (1962).
V13. Vreedenberg, H. A., Chem. Eng. Sci. 9, 52 (1958).
V14. Vreedenberg, H. A., Chem. Eng. Sci. 11, 274 (1960).
V15. Venuto, P. B. and Habib, E. T. Jr., “Fluid Catalytic Cracking with Zeolite
Catalysts.” Marcel Dekker, 1979.
W1. Wakabayashi, T., and Kunii, D., J . Chem. Eng. Jpn. 4, 226 (1971).
W2. Wen, C. Y., and Hashinger, R. F., AIChE J . 6, 220 (1960).
W3. Wen, C. Y., and Leva, M., AIChE J. 2, 482 (1956).
W4. Wen, C. Y., and Yu, Y. H., Chem. Eng. Prog. Symp. Ser. 62 62, 100 (1966).
W5. Wender, L., and Cooper, G . T., AIChE J . 4, 15 (1958).
W6. Werther, J., Trans. Inst. Chem. Eng. 52, 149, 160 (1974).
W7. Werther, J., AIChE Symp. Ser. 141 70, 53 (1974).
W8. Werther, J., Powder Technol. 15, 155 (1976).
W9. Werther, J . , “Fluidization Technology,” Vol. 1, p. 2 IS. Hemisphere, Washington,
D.C., 1976.
W10. Werther, J., Int. J. Multiphase Flow 3, 367 (1977).
WlOa. Werther, J., Chem. Ing. Tech. 50, 850 (1978).
W I I . Wheeler, A., “Catalysis 11” (P. H. Emmett, ed.), p. 105. Reinhold, New York, 1958.
W12. Whitehead, A. B., Gartside, G., and Dent, D. C., Powder Technol. 14, 61 (1976).
W13. Wilhelm, R. H., and Kwauk, M., Chem. Eng. Prog. 44, 201 (1948).
W14. Wollard, J . M., and Potter, 0. E., AIChE J . 14, 388 (1968).
W15. Wan, C. G., and Ziegler, E. N., Ind. Eng. Chem. Fundam. 12, 55 (1973).
W16. Wijffels, Ir. J-B., and Rietema, K., Trans. Inst. Chem. Eng. 50, 224 (1972).
W17. Wijffels, Ir. J-B., and Rietema, K., Trans. Inst. Chem. Eng. 50, 233 (1972).
XI. Xavier, A. M., Lewis, D. A. and Davidson, J. F., Trans. Inst. Chem. Eng. 56,
274 (1978).
Yl. Yagi, S., and Miyauchi, T., Kagaku Kogaku 17, 382 (1953).
Y2. Yamagoshi, T., B.S. thesis, Dept. Chem. Eng., University of Tokyo, 1969.
Y3. Yamazaki, M., J. Chem. Eng. Jpn. 8, 420 (1975).
Y3a. Yamazaki, M., Annu. Meet. SOC.Chem. Eng. Jpn., 40th. Nagoya (1975).
Y4. Yamazaki, M., and Miyauchi, T., J. Chem. Eng. Jpn. 4, 324 (1971).
Y5. Yamazaki, M., and Miyauchi, T., Kagaku Kogaku Ronbunshu 3, 261 (1977).
Y6. Yamazaki, M., Ito, N., and Jimbo, G., Kagaku Kogaku Ronbunshu 3, 272 (1977).
Y7. Yang, W. C., and Keairns, D. L., AIChE Symp. Ser. 141 70,27 (1974).
Y8. Yang, W. C., and Keairns, D. L., “Fluidization Technology,” Vol. 2, p. 51. Hemi-
sphere, 1976.
Y9. Yates, J. G., and Constans, J. A. P., Chem. Eng. Sci. 28, 1341 (1973).
Y10. Yates, J. G., and Rowe, P. N., Trans. Inst. Chem. Eng. 55, 137 (1977).
Y11. Yerushalmi, J., Turner, D. H., and Squires, A. M., Ing. Eng. Chem. Process Des.
Dev. 15, 47 (1976).
Y12. Yerushalmi, J., Gluckman, M. J., Graff, R. A., Dobner, S., and Squires, A. M.,
“Fluidization Technology,” Vol. 2, p. 437. Hemisphere, 1976.
Y13. Yokota, T., Hidaka, Y., and Yasutomi, T., Kagaku Kogaku Ronbunshu 1,399( 1975).
448 MIYAUCHI, FURUSAKI, MOROOKA, A N D IKEDA

Y14. Yoshida, A., B.S. thesis, Dept. Chem. Eng., University of Tokyo, 1971.
Y 15. Yoshida, K., and Kunii, D., J . Chem. Eng. Jpn. 1, 11 (1%8).
Y16. Yoshida, K., and Kunii, D., J . Chem. Eng. J p n . 7, 34 (1974).
Y 17. Yoshida, K., Kunii, D., and Levenspiel, O., Ind. Eng. Chem. Fundam. 8,402 (1969).
Y18. Yoshida, K., Kunii, D., and Levenspiel, O., Znt. J . Heat Mass Transfer 12, 529
(1%9).
Y19. Yoshida, K., Ueno, T., and Kunii, D., Chem. Eng. Sci. 29, 77 (1974).
Y20. Yoshitome, H., Kagaku Kogaku 27, 27 (1963).
Y21. Yoshitome, H.. Dr. Eng. dissertation, Tokyo Inst. Tech., 1967.
Y22. Yoshitome, H., and Shirai, T., J . Chem. Eng. Jpn. 3, 29 (1970).
Y23. Yoshitome, H., Mannami, Y., Mukai, K., Yoshikoshi, N., and Kanazawa, T.,
Kagaku Kogaku 29, 19 (1965).
Y24. Yacono, C., Rowe, P. N. and Angelino, H., Chem. Eng. Sci. 34, 789 (1979).
Y25. Yerushalmi, J. and Cankurt, N. T., Chem. Tech. 8, 564 (1978).
Y26. Yoshida, K., Nakajima, K.,Hamatani, N. and Shimizu, F., “Fluidization,” Proc.
2nd. Eng. Foundation Conf. 1978, p. 13, Cambridge Univ. Press, London and
New York (1978).
Z1. Zabrodsky, S. S., “Hydrodynamics and Heat Transfer in Fluidized Bed.” Gosener-
goizdat, Moscow, 1963; English trans., MIT Press, Cambridge, Massachusetts, 1966.
22. Zabrodsky, S. S., Antonishin, N. V., and Parnas, A. L., Can. J . Chem. Eng. 54, 52
(1976).
23. Zalewski, W. c., and Hanesian, D., AIChE Symp. Ser. I28 69, 58 (1973).
24. Zenz, F. A., Petrol. Rejiner 36 (9,261 (1957).
Z5. Zenz, F. A., and Othmer, D. F., “Fluidization and Fluid-Particle Systems.”
Reinhold, New York, 1960.
26. Zenz, F. A., and Weil, N. A., AIChE J . 4, 472 (1958).
Z7. Zuber, N., Chem. Eng. Sci. 19, 897 (1964).

Note Added in Proof

Some related literature that appeared subsequent to the preparation of this article may
be of interest to the reader and is therefore listed here:
Topic Reference

Flow in fluid beds B22, R20, XI, Y24

Bubble phenomena D19, R18, R19, Y26
Heat and mass transfer B21, H22, K29, S23
Axial dispersion N10
Elutriation or dilute phase phenomena G21, M50, M51, Y25
Reactions in freeboard or grid region D20, G20
Models V8a, WlOa