Reaction Engineering for
Environmentally Benign Processes
Reactor Selection Strategy
M.P. Dudukovic
Module 6
 Homogeneous systems
 Heterogeneous systems
 Systems (multi-scale) approach
S1
Approach to Reactor Selection
1. Identify number of phases present at reaction
conditions (thermodynamics)
–
–
Single – Homogeneous system
Multiple – Heterogeneous systems
2. Identify stoichiometry, number of reactions,
energy requirements (e.g. adiabiatic temperature
rise/fall)
3. Identify mechanism (if possible) and plausible
reaction pathways and active intermediates
4. Decide on the purpose of reactor selection
 Evaluation of kinetic data
 Data for scale-up
 Commercial design
S2
Chemical Reaction Engineering Basics
 Molecular Level
– Mechanisms and kinetic rates
 Eddy (Particle) Level
– Micromixing & kinetics
– Intra phase diffusional effects (Thiele modulus,
effectiveness factor)
– Inter phase transport effects
 Reactor Level
– Ideal flow patterns (CSTR, PFR)
– Non-ideal flow patterns between phases
– Contacting patterns
– Mixing
S3
For Homogeneous Systems:
 Identify the magnitude of heat transfer requirement
 Assess the effect of ideal flow patterns on volumetric
productivity and selectivity
 Select the best ideal flow pattern (batch, semi-batch,
continuous flow stirred tank reactor – CSTR, plug flow
reactor – PFR)
 Optimize your objective function (related to profit) using
as manipulative variables:
- Feed reactant concentrations and their ratio
- Feed temperature
- Reactor temperature or temperature profile
 Approach the ideal by practical design as much as
possible.
Keep things simple whenever possible!
S4
HOMOGENEOUS SYSTEMS
(Optimizing Volumetric Productivity)
Batch Reactor
t  0, C A  C A0
t  t , C A  C A0 1  X A 
t s  shut down time
Molesof A
C VX



  A0 A
t  ts
 reactedper unit time
t
C A0

CA
dCA
 reaction ime
t
 R A 
Continuous Flow Stirred Tank Reactor (CSTR)
FA0
FA = FA0(1-XA)
CA0
T = const.
Molesof A



  FA0 X A   RA  V
 reactedper unit time
Plug Flow Reactor (PFR)

FA0
CA0
FA = FA0(1-XA)

Molesof A



  FA0 X A   RA V
reacted
per
unit
time


X
1 A dX A
 V /[
]
X A 0  RA
S5
Volumetric Productivity for Product P then is
Fp
V
molesof P producedper unit time
reactor volume
For CSTR
 Fp

V

 p

    R A 
 CSTR  a 
For PFR
 Fp

V

 p a
 p

    RA 
X
1 A dX A
 PFR  a 
X A 0  R A


where  p a is the ratio of stoichiometric coefficients
The ratio of volumetric productivities in the two systems
F V 
F V 
p
PFR
p
CSTR

 R 
A
 RA 
Is the ratio of average reaction rate in a PFR and the reaction rate at exit
conditions of the CSTR
S6
Homogeneous Systems (optimizing selectivity)
A+B
2nd Order
P
desired product
A+A
2nd Order
S
undesired product
• Which is the optimal flow pattern ?
• What is the optimal selectivity ?
(at fixed feed concentrations, feed ratio of FA0/FB0 and conversion of B)
A
PFR
B
P+S
A
P+S
B
A
PFR
P+S
B
PFR
P+S
A
B
B
A
initially only B
initially only A
S7
In multiple reactions it is useful to consider the point yield behavior

Then in CSTR
While in PFR
RR
rateof formationof desired product

 R A rateof disappearance of key reactant

CR  exit CAo  CAexit
CR 

C Ao
  dC
A
C Aexit



CA
CA
CA
Production rate of R is maximized:
 In a CSTR for systems with d/dCA<0 (undesired reactions of higher
order than the desired one)
 In a PFR for systems with df/dCA>0 (undesired reactions of lower
order than the desired one)
 In a reactor combination for nonmonotonic yield curve
S8
Other Simple Rules Worth Remembering
 In consecutive reactions production of intermediate is
always more favored in a PFR than in a CSTR
 For exothermic reactions maximum volumetric productivity is
reached at an optimal temperature which is a function of
conversion
 When desired reaction has the highest activation energy
select the highest temperature for best selectivity
 When desired reaction has the lowest activation energy
lowest practical temperature optimizes selectivity
 For intermediate activation energy of desired reaction an
optimal temperature or temperature profile can be found
For lumping complex reaction schemes into patterns to
analyze see Levenspiel, O., Chem. React. Eng.
S9
Recognize that selected ideal flow patterns may only be approached in
practice.
Determine the deviation from ideal flow patterns by examining the residence
time distribution (RTD) of the system either derived from the solution of the flow
field or experimentally determined on a reactor prototype (cold flow model), pilot
plant or on the actual unit.
Et  dt  fractionof outflowof residence timearoundt 
PFR
E
Between
PFR & CSTR
CSTR
Et  dt  fract ionof out flowof residence t imeabout t 
E
t
t
t

t   t E t  dt 
o
1
  2
t
t
t
t
V
Lesser values indicatestagnancy
Q

2
E
exponential
decay

2




t

t
E
t
dt



0 PFR
1 CSTR
o
0   2  1 Model by dispersion or tanksin series model
 2 1
Bad! Avoid!
S10


In scale-up of systems with broad RTD  2  0.2 we need to assess whether
transport limitations can develop on a micro-scale (i.e. in bringing reactants in
contact or in supplying them to the soluble catalyst, enzyme or cell). This is
particularly important for non-premixed feeds.
We need to assess the scale of the smallest turbulent eddies in the system which is
determined by the amount of energy dissipated per unit mass of the system. For
example
14
  3    kinematicviscosity
K   
  energydisspated per unit mass
  
Characteristic diffusion time  D 
K 2
D
D  molecular diffusivity
C
Ao
Characteristic reaction time  R   R 
A o
D
 0.3 No microscaletransportlimitations
R

 D 5
Strongmicroscaleeffect- mixingdominatedsystem
R

In between micromixing models needed!
S11
Example:
WaterSolution:
k  O10 m
D 
k
2
D

10 
3 2
105
 101 s 
All reactions with R > O(1 second ) will not cause transport limitations.
Reactors with large  2 can be considered in maximum mixedness condition
k  O100 m
Highly Viscous Solution :
D 
K 2
D
10 

2 2
108
 104 s 
Only reactions with R > 105(s) will not cause transport limitations.
For most systems mixing and reaction occur simultaneously and proper micromixing
model is needed.
Proper treatment of this topic is not available in most standard reaction engineering
text. References and related notes can be obtained upon request.
S12
Two extreme micromixing models are:
1. Segregated Flow – All fluid elements remain segregated by
age on their sojourn through the system and elements of
different ages mix only at the exit.

C AS . F .   C Abatch t  E t  dt
o
dCAbatch
where
dt
   R A 
t0
C A  C Ao
2. Maximum Mixedness – All fluid elements of same life
expectancy are together at all times.
C AM . M .  C A   0  obtained by solving
dC A
d
  R A   
E
 E t  dt
C
Ao
 CA

t

dC A
0
d
S13
Micromixing Effect
Reactions:
A
2A
System:
k1
k2
P
1st order
S
2nd order

CSTR, t = 48 min;
k2CA0/k1 =
0.5
Exponential RTD
Laboratory System:
1 L vessel, 1500 rpm
Large System:
5000 gallon vessel; 300 rpm
Selectivity in the Lab.:
CP/CS = 98 at XA= 0.98
Selectivity in the Large Unit:
CP/CS = 15 at XA= 0.98+
Model Predictions:
Maximum Mixedness Flow:
CP/CS = 100
Segregated Flow:
CP/Cs = 4.5
S14
Key issues associated with selection and scale-up of reactors for
homogeneous reactions
 Developing sufficient knowledge of molecular level events to propose
mechanism and establish reaction pathways, key reactions and their parameters.
 Determining optimal ideal flow pattern and maintaining the same flow pattern
(same t and  2 with scale-up).
 Avoiding bypassing and stagnancy with scale-up.
 Maintaining same level of micromixing with scale-up is needed but hard as power
dissipated per unit volume decays with scale and affects micromixing adversely.
 Maintaining adequate heat transfer rate with scale-up is difficult as heat evolved
by reaction  volume and heat removed  surface. With scale-up in general S/V
is reduced which may lead to problems unless corrective steps are taken.
 Control of temperature, pressure, pH etc. becomes more difficult with increased
scale.
 Homogenous catalyst or soluble enzyme recovery, a cinch in the lab, becomes a
major chore in large units.
 Solvent separation is a problem.

Heterogenize the system whenever possible, do not use solvents unless
absolutely necessary!
S15
The objective in multiphase reactor selection and design is to minimize
the manufacturing costs in producing the desired marketable product.
For conversion cost-intensive processes one must achieve both high
volumetric productivity and high product concentration.
 v  Fp  p V
m

m v kg m 3 h

- volumetric productivity
Fp km ol h  - molarproductionrate
 p kg km ol - molecularweight
 
V m3
- reactor volume
For recovery cost intensive processes (e.g. often encountered in
biotechnology) one must achieve high product concentration cp(kg/m3).
cp   p Cp


c p km ol m 3  molarconcent rat
ion
In either case proper reactor selection is required since reactor type and
performance affects significantly the whole process.
S16
Reactor

Performanc
e
-Conversion
-Selectivity
-Production Rate
Input and
Mixing


f 
; Rates ;
Pattern
 OperatingVariables
- Flow Rates
- Inlet Conc. & Temp.
- Kinetics
- Transport
- Macro
- Micro
- Heat Removal
LHS
RHS
feed, Q
L( Cb )  R( Cb ,Tb )
Lh ( C b ) 
T ,C0 , P0
 ( HR ) j R j ( C b ,Tb )
j
T ,C , P
j
  f kinetics; transport 
product, Q
Reactor performance determines the number of separation units and their load and
hence profoundly affects process economics and profitability.
 Production  Volume Averaged  Reactor

  
  

 Rate   ReactionRate   Volume
S17
In heterogeneous systems the volume averaged reaction
rate (volumetric productivity) is a function of:
 Molecular scale – kinetics and rate forms
 Single particle (single eddy) scale effects on diffusion and
reaction in the particle, specific phase interfacial area
effect on inter-phase mass and heat transfer
 Reactor scale effect via contacting pattern and phase RTD
influence on the average rate and via flow regime effect on
phase holdups and inter-phase transport coefficients.
S18
As a reminder consider the diffusional effects on the rate in a porous particle with
uniformly deposited active catalyst in pores
 AverageRate 

  P articleEffectiveness 

 P er UnitVolume  
Factor


 of P article 


 RA  particle   p  RA s
Where typically  p 
 Rate Evaluatedat 


 P articleOutside 
 Surface Conditions


tan h  p
p
1-
With Thiele modulus  p
p 
2
p 
p  0
0.1 -
p
D
R
0.01 0.001 -
Vp
10-4 -
Sp
|
0.01 0.1
 RA  particle   RA S
|
1.0
|
|
10
100
1000
p
True kinetics, activation energy is observed. Doubling catalyst activity doubles the rate.
Rate independent of Sp/Vp
 S p  n1
1
 p    R A  particle 
C AS  k De   C AS
V 
p
 p
2
Approximately ½ E observed. Reduced order
 R A  part .
 Vp

S
 p
1

 ; catalystactivity1 2 , catalstloading1 2


S19
Now one must also consider inter-phase transport
k s S p C Ab  C As    RA s  p V p
And for first order reaction one gets
CA
 R A obs part   0  R A bulk 
b
1
1

k s S p kv V p  p
The denominator contains the sum of external resistance and internal +
kinetic resistance.
Of course we need the rate per unit reactor volume so
 RA obs react  1   B o  RA bulk
Clearly how much catalyst we packed in (bed voidage B) affect also
volumetric productivity.
Finally flow pattern will affect how (-RA)bulk is averaged and flow pattern
affects transport coefficient ks.
p  
 R A  particle

1
C AS 
p
 Sp
k De 
V
 p

 CA
S


n 1
2
Approximately ½ E observed. Reduced order
 RA  part .
 Vp

S
 p




1

; cat alystact ivit y
1 2
, cat alstloading
1 2
S20
REACT ION: Ag   Bl   P1
Gas Limiting Reactant (Completely Wetted Catalyst)


KINET ICRAT E : k v A m ol m 3 cat.s   A
per unit catalyst volume

RAT E IN CAT ALYST: k v  p 1   B  As m ol m 3 react. s
per unit reactor volume

T RANSP ORTRAT E m ol m 3 react. s
per unit reactor volume
 Ag

: K 1 a B 
 A1 
 Ha

- Liquid - solid : k s a p  Al  As 


- Gas - liquid


OVERALL (AP P ARENT RAT
)
E m ol m 3 react.s :
Ag
Ag
HA
R A   o k v 1   B  H A 
1
1
1


K l a B k s a p 1   B  k v p
S21
A System Approach to Multiphase
Reactor Selection
Economics
Reactants
Reactor Type
&
Contacting Pattern?
Process Requirements
• Maximum selectivity
• Maximum conversion
• Maximum productivity
• Stable
• Easy scale-up
• Operability
Products
Environmental Constraints
• Minimum pollution
Why System Approach?
• Number of configurations extremely large
• Limits to intuitive decision making
• Innovations are possible
S22
Multiphase Reactor Selection Methodology
I. Volume / Interfacial Area for the Phases
~ dp for gas-solid systems
~ b for gas-liquid systems
~dp and b for G-L-S systems
II. Contacting & Flow Pattern
a) RTD for each phase (PF, backmixed)
b) Co – Counter – Cross current?
c)  Split addition
 Product removal in situ, etc.
III. Flow Regime
 Homogeneous
 Churn turbulent
 Dense phase riser (air lift)
 Dilute phase riser (spray)
S23
Example: Recovery of Oil From Oil Shale
Process Requirements (Wish List)
• Maximize “oil” recovery (99%+)
• Scale-up to mega-size units ( 500 kg/s feed)
• Minimize reactor volume
• Handle fines well
>200 G-S Reactor Configurations possible!
After Krishna (1989)
S24
S25
Shell’s SPHER 3
Bed Concept
Chevron’s STB
(staged turbulent bed)
S26
Decisions to be made:
I. Particle Size
II. Contacting Pattern
a. Overall contacting flow pattern of gas and solid phases:
b. RTD of each phase:
III. Gas-Solid Fluidization Regime
Krishna (1992), Adv. Chem. Eng.
S27
Kinetics & Transport Phenomena
Affecting Process Performance
Oil Shale Pyrolysis
Wallman et al (1980), AIChE Meeting, San Francisco
S28
Residence time required for heating up of particle to
95% of Tg = 482°C
Residence time required for isothermal backmixed reactor
(174 min)
Conversion of kerogen 99%
Residence time required for isothermal plug flow reactor
(8 min)
Large throughputs  minimize reactor
Volume  need small residence times  need particles in
range I  need plug flow of solids
S29
Desired product (heavy oil) yield improved with
small particle size (dp < 2mm).
In grinding shale to make 2mm particles fines
may be formed too.
S30
I. Particle
Size
Selection
Tree
II. Contacting
Flow Pattern
Selection
Tree
III. Flow
Regime
Selection
Tree
S31
To reduce oil degradation, must remove oil as soon as
formed  in situ product removal
Wilkins et al (1981), 2nd World Congress, Montreal
S32
A. Counter-Current Contacting
B. Co-Current Contacting
C. Cross-Current Contacting
Reactor volume requirement  need plug flow of solids
Rapid oil removal  cross flow for gas
S33
Proper fluidization regime should now be chosen to
accommodate:
• Desired particle size (small)
• Desired solids holdup (large)
• Desired contacting pattern (solids-plug flow, gas short
contact time)
• Excellent heat transfer
S34
The “Ideal” Reactor:
Multi-Stage Cross-Current Fluidized Bed
Meets the criteria:
• Small particles
• Plug flow of solids
• Short vapor residence time (cross-flow)
• Good mixing and heat transfer
• Scale-up possible – study one train
Shell Shale Retorting Process
(Shell Research)
Krishna (1992)
S35
For Shale Example
Possible Reactor Combinations
3


particle
size
range
13
 5  195


gas - solid
fluidizati on
contacting
regimes
pattern
• Sequential design making leads to success without brute force
evaluation of all options.
Choice of wish list effects final result. Add:
Choice should be based on known technology

Moving bed reactor
S36
This example illustrates how consideration of all
scales leads to successful reactor selection
It also teaches that in situ separation when
possible is of high value and can sometimes be
achieved by:
 Catalytic distillation
Selective adsorption or absorption
Membrane separation
Other means (e.g. dynamic reactor operation, etc.)
Think Out of the Box!
S37
Our task in catalytic reactor selection, scale-up and design is to
either maximize volumetric productivity, selectivity or product
concentration or an objective function of all of the above. The key
to our success is the catalyst. For each reactor type considered
we can plot feasible operating points on a plot of volumetric
productivity versus catalyst concentration.
m vmax
m v
 kg P 
  specific activity
S a 
 kg cat h 
 kg cat 
x 3
  catalyst concentrat ion
 m reactor 
Sa
xmax
x
 vmax is determined by transport limitations and xmax by
Clearly m
reactor type and flow regime.
Improving
limited.
S a only improves
m v if we are not already transport
S38
Chemists or biochemists need to improve Sa and together with engineers work on
increasing xmax .
Engineers by manipulation of flow patterns affect
In Kinetically Controlled Regime
m v

x,
m vmax .
Sa
xmax limited by catalyst and support or matrix loading capacity for cells or
enzymes
In Transport Limited Regime
m v

p
Sa , x p
0  p  1/ 2
Mass transfer between gas-liquid, liquid-solid etc. entirely limit m v and set m vmax .
Changes in S a , do not help; alternating flow regime or contact pattern may help!

Important to know the regime of operation
S39
Schematic of Bubble Column Type of
Photo Reactors (Commercially Used)
A train of bubble columns (sparged reactors) through which liquid
toluene and chlorinated products flow in series while chlorine is added
into each column and hydrogen is removed from the column.
Typical selectivity to benzyl chloride: 90% But Toluene conversion is
less than 30%. Can one do better?
S40
Schematic of Photo Reactive Distillation System
Configured into a Semi-Batch Model
Allows in situ product removal
and toluene recycle.
Selectivity to benzyl chloride: 96% + up to toluene conversion of 98%.
Z. Xu (1998)
S41
PROBLEM
Reaction System:
Current Reactor:
Production of herbicide intermediate, aryl
amino-alcohol (AA) via hydrogenation of
aryl nitro-alcohol (NA)
complex
semi-batch dribbling liquid reactor with
suspended catalyst slurry
DISADVANTAGES:
 Low volumetric productivity
 Poor selectivity
 Catalyst filtration and separation
problems
 Pressure limitations (due to shaft)
Khadilkar et al., AIChE J., 44(4), 912 (1998)
Khadilkar et al., AIChE J., 44(4), 921 (1998)
S42
Reaction Network
S43
Conclusions
 Liquid trickling flow pattern is preferable to a suspended
catalyst mixed slurry to obtain the desired yield and productivity
of Amino Alcohol.
 Yield improvement is observed with decreasing feed
concentration, liquid flow rate and temperature due to
suppression of NA decomposition and subsequent side
reactions.
 Productivity of AA is a complex function of flow, feed
concentration and temperature with optimal productivity being
determined by the level of acceptable by-product
concentrations.
 Laboratory trickle bed performance data is shown to be an
effective means to obtain the network kinetic parameters by
proposing a plausible mechanism and optimizing the reactor
model generated data. This is particularly effective in cases
where conventional slurry and basked methods are rendered
ineffective by the dominance of side reactions.
S44
References
1.
Dudukovic, M.P., Larachi, F., Mills, P.L., “Multiphase
Reactors – Revisited”, Chem. Eng. Science, 541, 19751995 (1999).
2.
Dudukovic, M.P., Larachi, F., Mills, P.L., “Multiphase
Catalytic Reactors: A Perspective on Current Knowledge
and Future Trends”, Catalysis Reviews, 44(11), 123-246
(2002).
3.
Levenspiel, Octave, Chemical Reaction Engineering, 3rd
Edition, Wiley, 1999.
4.
Tranbouze, P., Euzen, J.P., “Chemical Reactors – From
Design to Operation”, IFP Publications, Editions TECHNIP,
Paris, France (2002).
S45