Chemical Reactor Analysis and Design Fundamentals

James B. Rawlings Department of Chemical Engineering University of Wisconsin Madison, Wisconsin
John G. Ekerdt Department of Chemical Engineering The University of Texas Austin, Texas



Madison, Wisconsin

This book was set in Lucida using J:l.TEX, and printed and bound by Worzalla. Cover design by Cheryl M. and James B. Rawlings, and Jobo W. Eaton. Copyright © 2002 Nob Hill Publishing. All rights reserved. Nob Hill Publishing Cberyl M. Rawlings, publisher Madison, WI 53705 No part of this book may be reproduced, in any form or by any means, without permission in writing from the publisher. library of Congress Control Number: 2001135832
Rawlings, James B. Chemical Reactor Analysis and Design Fundamentals / by James B. Rawlings and John G. Ekerdt. p. cm. Includes bibliographical references (p.) and index. ISBN 0-615-11884-4 (cloth) 1. Chemical reactors. I. Ekerdt, John G. II. Title.

To Cheryl, Melanie and Blake, and Carol, Alice and Barbara, for their love and encouragement.

Printed in the United States of America.












Chemical reactors are at the core of the chemical engineering discipline, and chemical reactor analysis and design is one of the distinguishing -' courses that clearly separates the chemical engineers "from the other engineering professionals. Given that chemical reactor analysis and design is a mature and stable topic in the curriculum of chemical engi· neering, however, it is natural to ask what is the motivation for a new text on this topic. We offer our motivations here. This book grew out of the combined experience of the two authors teaching this subject to undergraduates for more than 30 years, Given the rapidly changing landscape of scale and type of reactors of interest to practicing chemical engineers (chemical vapor deposition reactors, pharmaceutical fermentors, micro-reactors, as well as traditional catalytic crackers, bulk polymerization reactors, etc.), it seems unwise to emphasize one industrial sector and treat its reactor types in detail. Practicing chemical engineers work in a broad array of industrial sectors, and many will change sectors during their careers, If chemical engineering has any important distinguishing characteristic, it is a set of fundamentals that apply to all scales and all types of reaction and transport processes involving chemical change, This book is all about reactor fundamentals, Rather than presenting many facts about reactors, we focus on the framework for how to think about reactors - a framework for thinking that enables one, with some experience, to establish any of these facts for oneself, and discover new facts given new situations. All engineering and science textbooks do this to some extent; in this text, we will do it to a rather large extent, Computations matter in this subject. Reactor fundamentals, like the fundamentals in any subject, are few in number. But the diversity of the consequences of these fundamentals is enormous, Computational approaches provide a powerful and general approach to systematically investigating these consequences Without making unrealistic simplifying assumptions, We attempt to exploit the significant advances in computing algorithms, software, and hardware in order to revise and streamline the presentation of reactor fundamentals. We focus on two high-level languages intended for numerical computation, Octave and


MATLAB. Octave is freely available for a variety of hardware platforms and can be down·loaded from MATLAB is commer. cially available from The MathWorks, Inc., and is becoming a commonly available tool of industrial engineering practice. These languages allow us to focus on essentials and ignore programming details, which is the goai of any "high·level" language. Students should not feel compelled to recall the detailed information in the figures, but recall only the concepts, principles and main results. Students will have the computational tools to re-create the figures in this or any other textbook on this subject. For example, all calculations required for the figures in this text were performed with Octave. Students should not feel compelled to memorize design equations for reactor types. The goal is to develop suffiCient expertise so that students can set up appropriate mOdels from the basic principles for each new problem they encounter. That is the time-tested way to instill confidence that one can analyze a new Situation, which We fully expect to be the experience of practicing engineers. Newly practicing engineers certainly will need to learn the economics of manufacturing their main products, main reactor configurations, detailed energy-recovery schemes, contacting patterns, new catalysts, and reactor mOnitoting and control systems. Rather than make inadequate efforts to include all of this information as a survey during the reactor deSign course, we feel this material deserves separate coverage, or can be learned during the early years of engineering practice. We do not regard it as poor preparation that some of this information is not included in University curricula; rather, we regard learning this material on the job as part of the maturation process of the practicing professional chemical engineer.

Acknowledg ments

W benefited greatly from the help and advice of many colleagues durine the preparation of this text. Taking Warren Stewart's graduate reacmodeling course greatly inlluenced the first author's way of thInking about this subject. More recently, Warren also proVlded1J'i.any helpful comments on computational issues involving reaction-dIffuslOn, and ameter estimation. Harmon Ray contributed many mSIghtful and comments on a variety of topics, especially on Issues of reactor mixing and industrial polymerization reactors. Bob Brrd provided many helpful comments on presenting the matenal related to transport phenomena. He also provided detailed feedback on a reVlew copy of



Finally, we hope the text conveys some of the excitement that chemical engineers feel for chemical reactions and chemical reactors. Under. standing the fundamentals prepares you to investigate new situations. Grounding in the fundamentals enables you to explore With confidence and be creative. Enjoy. JBR Madison, Wisconsin JGE Austin, Texas

the text. d 'd d de Marty Feinberg read an early version of the notes, an proVl e . tailed feedback on many issues including reaction rates a~d presentmg computational material. Yannls Kevrekidis used a prelirmnary verSlOn of the notes as part of a graduate reactor modeling course at Prmceton. His feedback and encouragement are very much appreCIated. John Falconer also provided many helpful comments. Tom Edgar provided much-needed departmental support and encouragement when the authors began writing. Roger Bormecaze, Bud: die Mullins and C. D. Rao used early verSlOns of the notes and Oc tave software in teaching the undergraduate course at The UmversIty of Texas. Their feedback was very helpful. . . Several graduate students contributed to the software, whIch IS an essential part of this text. Ken Muske developed some examples for the chapter on mixing. John Campbell worked on programs for managing thermochemical data, Sankash Venkatash contributed programs on parameter estimation and reaCtion-dIffusIOn, and DomImk Nagel contributed programs to calculate potential-energy surfaces. Chns Rao provided helpful feedback on mixing issues, and Matt Tenny and Ene Haseltine provided helpful input on the cover design. Eric also helped prepare the solution manual. We have benefited from the feedback of many classes of undergraduates both at the Umverslly of Wlsconsm and The University of Texas. We appreciate the enthUSIasm WIth WhICh they received some very early and incomplete drafts of the notes. Our publisher deserves speCial recognition for provIdmg enough

flexibility so that we could realize our original Vision, while still insisting that we actually finish the project. Finally, we would like to acknowledge the essential contributions of John Eaton. When the two authors first outlined the goals of this writing project, we enVisioned creating in parallel a convenient computational environment. The original goal of the computational enVironment was to enable students to quickly and conveniently define and solve new reactor deSign problems as a means to gain deeper understanding and appreciation of chemical reactor fundamentals. Starting from this simple idea, John Eaton created the Octave software project. We could not have written this text in the way we planned without Octave. Every figure in the text requiring calculation was prepared with Octave. John contributed in a myriad of other ways. He designed the typesetting macros and the text version control system. We maintained more than 10 versions of the source files on more than 7 different computing systems at various stages of production. With John's expert guidance, we easily managed this complexity; in fact, the computing tools worked so well, that the computational aspects of the book production were enjoyable.


1 Setting the Slage Introduction . . . . . . . . . . . . . . . . . . . . ......... 1.1 Some Classifications and Terminology ............ : ... . 1.2 ........... 1.3 Scale . . . . . . . . . . . . . . . . . . . . . . . Some Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4 1.4.1 Chemical Vapor Deposition . . . . . . . . . . . . . . .
1.4.2 Hydrodesulfurization . . . . . . . . · · · · ..... .

1 1 2

7 10 10


1.4.3 Olefin polymerization . . . . . . . . . . . . . . . . . . . 1.4.4 Hepatitis B Virus Modeling . . . . . . . . . . . An Overview of the Text . . . . . . . . . . . . . . . . . . . . .

12 IS
18 20

2 The Stoichiometry of Reactions 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . Chemical Reactions and Stoichiometry . . . . . . . . . . . . 2.2 2.3 Independent Reactions . . . . . . . . . . . . . . . . . . . . 2.4 Reaction Rates and Production Rates . . . . . . . . . . . Computational Aspects of Stoichiometry .... : ..... . 2.5 2.5.1 Computing Production Rates from Reaction Rates 2.5.2 Computing Reaction Rates from Productlon Rates 2.5.3 Measurement Errors and Least·Squares Estunatlon
2.6 Summary and Notation . . . . . . . . . . . . . . . .







2.7 Exercises . . . . . . . . . . . . . . . . . . . . . .




3 Review of Chemical Equilibrium 63 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 3.2 Condition for Reaction Equilibrium .............. . 75 3.2.1 Evaluation of the Gibbs Energy Change of ReactIOn 3.2.2 Temperature Dependence of the Gibbs Energy .. . 76 Condition for Phase Equilibrium . . . . . . . . . . . . . . . . 79 3.3.1 Ideal Mixtures . . . . . . . . . . . . . . . . . . . . . . . . 80 82 3.3.2 Nonideal Mixtures , . . . . . . . . , . . . . . . . . . . ,
3.4 Heterogeneous Reactions . . . . . . . . . . . . . . . . . . . . .

3.5 Multiple Reactions . . . . . . . . . . . . . . . . . . . . . . . . .

3.5.1 Algebraic Approach . . . . . . . . . . . . . . . . 3.5.2 Optimization Approach . . . . . . . . . . . . . .

82 88 88






3.6 Summary and Notation ...................... 3.7 Exercises .. " ........................... . 4 The Material Balance for Chemical Reactors 4.1 General Mole Balance ....................... 4.2 The Batch Reactor 4.2.1 Analytical Solutions for Simple Rate Laws . . . . . . 4.3 The Continuous-Stirred-Tank Reactor (CSTR) . . . . . . . . 4.4 The Semi-Batch Reactor .... ..... " ........... . 4.5 Volume Change Upon Reaction . . . . . . . . . . . . . . . . . 4.6 The Plug-Flow Reactor (PFR) . . . . . . . . . . . . . . . . . . . 4.6.1 Thio Disk Volume Element 4.6.2 Steady-State Operation . . . . . . . . . . . . . . . . . . 4.6.3 Volumetric Flowrate for Gas-Phase Reactions ... . 4.6.4 VolumetriC Flowrate for Uquid-Phase Reactions .. 4.6.5 Single Reaction Systems . . . . . . . . . . . . . . . . . 4.6.6 Muitiple-Reaction Systems . . . . . . . . . . . . . . . . 4.7 Some PFR-CSTR Comparisons . . . . . . . . . . . . . . . . . . 4.8 Stochastic Simulation of Chemical Reactions . . . . . . . . 4.9 Summary and Notation ...................... 4.10 Exercises
•• " < ••••••••••••••••••••• < " ••••••••••••• ••••••••••••••••••••• < •••••••••


6.4 6.5


133 133 146 146 147 147 148

6.6 6.7

6.3.2 Stability of the Steady State . . . . . . . . . . . . . . 6.3.3 Sustalned Oscillations, Umit Cycles . . . . . . . . . The Semi-Batch Reactor . . . . . . . . . . . . . . . . . . . . . The Plug-Flow Reactor . . . . . . . . . . . . . . . . . . . . . . 6.5.1 Plug-Flow Reactor Hot Spot and Runaway . . . . . 6.5.2 The Autothermal Plug-Flow Reactor . . . . . . . . . Summary and Notation . . . . . . . . . . . . . . . . . . . . . Exercises . . . . . . . . . . . . . . . . . . . . . . . . .~/ ...

. . . . . . . .

295 305
316 317 322 326



164 173 176 189

5 Chemical Kinetics 5.1 Introduction 5.2 Elementary R~a~;i~~ : : : : : : : : : : : : : : : : : : : : : : : : 5.3 Elementary Reaction Kinetics . . .. . . . . . . . . . . . . . 5.4 Fast and Slow Time Scales 5.4.1 The Reaction Equilibrium Assumption . . . . . . . . 5.4.2 The Quasi-Steady-State Assumption . . . . . . . . . . 5.5 Rate Expressions . . . . . . . . . . . . . . . . . . . . . . . . . . 5.6 Reactions at Surfaces . . . . . . . . . . . . . . . . . . . . . . . 5.7 Summary and Notation . . . . . . . . . . . . . . . . . . . . . . 5.8 Exercises . . . . . . . . . . . . . . .
• • « • « •••••••••••••••
•••••••••••••• >

190 195

207 209
218 226 237 255

7 Fixed-Bed Catalytic Reactors 7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.2 Catalyst Properties . . . . . . . . . . . . . . . . . . . . . . . . . 7.2.1 Effective Diffusivity . . . . . . . . . . . . . . . . . . . . 7.3 The General Balances in the Catalyst Particle . . . . . . . . 7.4 Single Reaction io an Isothermal Particle . . . . . . . . . . . 7.4.1 Fitst-Order Reaction io a Spherical Particle .... . 7.4.2 Otber Catalyst Shapes: Cylioders and Slabs .... . 7.4.3 Other Reaction Orders . . . . . . . . . . . . . . . . . . 7.4.4 Hougen-Watson Kinetics . . . . . . . . . . . . . . . . . 7.4.5 External Mass Transfer . . . . . . . . . . . . . . . . . . 7.4.6 Observed versus Intrinsic Kinetic Parameters .. . 7.5 Nonisothermal Particle ConSiderations . . . . . . . . . . . . 7.6 Multiple Reactions . . . . . . . . . . . . . . . . . . . . . . . . . 7.7 Fixed-Bed Reactor Design . . . . . . . . . . . . . . . . . . . . . 7.7.1 Coupling the Catalyst and Fluid . . . . . . . . . . . . 7.S Summary and Notation . . . . . . . . . . . . . . . . . . . . . . 7.9 Exercises . . . . . . . . . . . . . . . . . " . . . . . . . . . . . . . 8 Mixing in Chemical Reactors S.l Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.2 ReSidence-Time Distribution . . . . . . . . . . . . . . . . . . 8.2.1 Definition . . . . . . . . . . . . . . . . . . . . 8.2.2 Measuring the RTD . . . . . . . . . . . . . . . . . . . 8.2.3 Continuous-Stirred-Tank Reactor (CSTR) . . . . . . 8.2.4 Plug-Flow Reactor (PFR) and Batch Reactor .... 8.2.5 CSTRs in Series . . . . . . . . . . . . . . . . . . . . . . 8.2.6 Dispersed Plug Flow . . . . . . . . . . . . . . . . . . . 8.3 Limits of Reactor Mixing . . . . . . . . . . . . . . . . . . . 8.3.1 Complete Segregation . . . . . . . . . . . . . . . . 8.3.2 Maximum Mixedness . . . . . . . . . . . . . . . . . . 8.3.3 Mass Transfer and Umits of Reactor Mixing ...

355 355
357 360 363

365 366
374 377 380

385 390 392 395 400 400
416 419

275 275 276 277 277 286 290

6 The Energy Balance for Chemical Reactors 6.1 General Energy Balance . . . . . . . . . . . . . . . . . . . . . . 6.1.1 Work Term .... ...... , ........... . 6.1.2 Energy Terms . . . . . . . . . . . . . . . . . . 6.2 The Batch Reactor . . . . . . . . . . . . . . . . . . . . . . . . . 6.3 The CSTR 6.3.1 Stead~-s;a;e' M~I~i~ii~i~Y' : : : : : : : : : : : : . . . . .

433 . 433 . 434 434 435 . . 439 . 442 . 444 . 446 . 454 455 458 . . 464

. . . . .. 1. . . .2. • . 8. A.2. . . .. . . . . 8. . . .1.. . . . . . . . r. . .10 Top view of a single-wafer CVD reactor . . . .5 Examples in Which Mixing is Critical . . . . . .2 Automatic Stopping Times (ddas rt) . . . . . . . . . . . . . . . . .. . . . . . . .. . . . . . 9. . .14 Expanded view of a multi-bed HDS reactor . . . . . . . . . .1 599 . . 9. . 3. .. .. . . . . . . . . 9. . . . .. . .. . 1. . 8. . .. .. . . .2 The General Case . . . . . .2 Gibbs energy versus reaction extent at constant T and P. . . . ... . . .. .I linear Algebra and Least Squares. . . .. . . . . . . . . . . . .2...16 The chemical events comprising the reproduction cycle of the hepatitis B virus. .. . . . . . . . .4 Summary and Notation . . .2 Eigenvalues and Eigenvectors .xiv Contents 8.1 Review of the Normal Distribution .9 polyethylene reactor . ..4 Least Squares with Unknown Variance . . . .15 Simplified polymerization process . .2 Expanded view of the internals of a batch reactor . . . . . .. . .3 An industrial Case Study . 1. . . . . . .1 Differential-Algebraic Equations (dassl) . 1. . .6 Design of Experiments . . . . . . . . . . . . . . . .. . . . . . . . . . ... . . .11 Representative organosuifur compounds .3 52 63 65 75 3. .. . . . . . . . . . Index Schematic diagram of a batch reactor. . 9. .4. . .. .3. .3 Scbematic diagram of a CSTR. . . . . . . . . . . . . . . . . .. . .2 Experimental Reactors for Kinetics Studies . . . . . .6 Exercises . . . . . .. . . .1 A Single Convex (Concave) Reaction Rate . . . . . . . . . 3.2. . .. . Estimated reaction rates from six production-rate measurements subject to measurement noise. .3 Gibbs energy Gversus reaction extent £'. . . . . . . . . . .. . . . 1. .. . . . . . 3.6 Summary and Notation . . . .. . . . . . . . . . . .. . . . . . . . . .. . . . . . .2 Data Modeling and Analysis . . . . . . on) . . . . . 9 Parameter Estimation for Reactor Models 9. . . . . . .. . 9. . . . . . .. . . . . . . . 9. . .1.6 Cross-sectional view of a 1 III combinatorial screenlng reactor 1. . 3. . . ..1 Analytical Probes for Concentration . . . . 8. .2. . .5 Schematic diagram of a plug-flow reactor. . . . . . . . . . . . . . .. . . . . . . . . . . . . . . 1. .2 2. . . . . 8. .8 Top view of a combinatorial reactor gas-distribution network 1. . .1. . . .7 Assembled view of 256 I-ilL reactors built into a 3-in by 3-in support wafer .6 Gibbs energy contours for the pentane reactions as a function of the two reaction extents.2. . 1. . 9. . . .2.. . .3 Least-Squares Estimation . 84 89 95 . .. . . 2. . . . . . . ..12 Hydrodesuifurization process. .. . . . . . . . . . . . . . . . . . .. . . 1. . 9. . 9. . . I l I ThermodynamiC system. . . . 1.5 Gibbs energy versus two reaction extents at constant T and P. . . . . . . . . . . . . . 1.. . . . .. . . . .3 Characterizing Catalysts and Surfaces .S Boundary-Value Problems and Collocation (colloc) . . . . . . . . . . . .7 Parameter Estimation witb Differential Equation Models . .. . . . . . . . . .4 limits of Reactor Performance . . . . .. . . . . . . .. . . . . . . . . . . for the 5iH2 reaction . . . . .7 Exercises . . .. .. A. . . . . . ... . . . 9.4 Schematic diagram of five CSTRs in series . .1 Defining the reaction rate. 4 4 5 6 6 7 8 9 9 11 13 14 15 16 17 19 42 S2 2.. . . Estimated reaction rates from 500 production-rate measurements subject to measurement noise. . . .2. . . 1. . . . . . . . . . .. .1 Experimental Methods . . . . . . . . . . A. . . . . .4 Partial pressures of components A and C versus liquidphase composition in a nonideal solution . . . . . ...1 Defining Functions (funct. .. . . . . .4 Sensitivities of Differential Equations (ddasac) .. . . . .2 Solving Nonlinear Algebraic Equations (fsolve) . . .. 9. . . . . . . .. . .. . . . . . 3. . . . .13 Cross-sectional view of a commercial HDS reactor . .3 Differential Equations (l sode).. . 1. . A. . . . .2. . . 9. 9. . . .. A. . 9. ... . . . . . . . .. 469 469 474 479 486 490 501 501 501 506 510 513 514 515 517 523 527 532 538 551 562 565 575 575 578 578 579 580 582 584 584 585 589 595 list of Figures 1.3. A. . .. . . . .. .. .3 Solving Nonlinear Optimization Problems (npso l) A. . . . .. .. . . . . . A. .2.1 A Computational Metbods A.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. .5 Nonlinear Least Squares . . . . . . . . . . . . . . . . . 1. A. . A. . . . . . .. . . .2 Nonlinear Algebraic Equations and Optimization . . .. . . .4. .5 Exercises ..

. . . deterministic and average stochastic models. 169 4.. . . . .. CEO = O. . ... . . .22 Molar f10wrate of ethane. . .. . • . . n :$ O . 183 C com4. .2 4. . . . . . . 5. . . • . . . . .9 Batch reactor with nth-order kinetics. .. . . . . . . . . . . . . irreversible kinetics in a batch reactor. L! = 0. . . . .5. . . .. . reversible kinetics in a batch reactor. . . . . .C . . . . . . . 163 4.!. . . . . . . . . .c• • . . . . . . 129 4.15 Semi·batch reactor volume for different monomer addition poliCies . . .. .10 Reaction rate versus concentration for nth-order kinetics r = kc1..6 Fnll model solution for k! = 1. . . . . . . .4 First-order. . . . . . . . . . . . . . . . 171 4. • . . . . . . . 132 4. • .. . . . . . . . . . .3 7 Deterministic simulation of reaction A + B pared to stochastic simulation. . . . . . . Potential-energy surface for the F. .17 Semi-batch reactor monomer content for different monomer addition policies .. . . .B . .. . 145 4. . ..k=1. . Hz reaction. 154 4. . . . . . . . .9 Comparison of eqUilibrium assumption to full model for kJ = 1. . . . . . . . . . . . . . .7 Full model solution for kJ = 1. . . . . . . .. . . . . . .. . ./ = 1. . . . . . . . . . . . . . . . nthorder kinetiCS. . . . . . irreversible kinetics in a batch reactor. . . . . . . .32 Species cccDNA versus time for hepatitis B virus model.. .8 Two systems with identical intensive properties and different extensive properties. . 185 = 5. . lx. . . . .31 Stochastic simulation of the first-order reactions A. . . . . : . . . . . .. . . r = kc. . . . . . . . . . . .7 3. . . . . . . r = kc. . . 158 4. . . . . . n < 0 . . . . . . . .36 An alternative CSTR-separator design. . . . . . . . .12 Two first-order reactions in parallel in a batch reactor . . . ..5. . . . . . . . 145 4. . . . ..14 Reaching steady state in a CSTR. . . . . . " . . . . .. 144 4.6 Second-order and first-order kinetics in a batch reactor. . . . . . . . . . . . . . . . . n > 0 .. . . . 127 4. . . . " 115 4. . . . . . . . . . . . . . . two representative stochastic trajectories. • • . . . . . . . : . . . . . deterministic and average stochastic models . . . . . . . . . . . 5. negative order. .5 Morse potentialfor Hz and HF.. 111 Batch reactor volnme element. . 5.26 PFR versus CSTR with recycle and separation. . . . . k-J = 0. k2 = Lz = 1. . .28 Choosing a reaction with appropriate probability ••".3 First-order.3 S. . . . . . . . . . .xvi List of Figures List of Figures xvii 3. . . .5 First-order. . 100 Vessel with semipermeable membrane and cbemical reaction 2A = B. .2 = 10. .8 Concentrations of Band C versus time for full model with increasing kz with Kz = kzl Lz ~ 1. n < 1 . . . . . .19 Plug-flow reactor volume element . . . . . .33 Species rcDNA versus time for hepatitis B virus model. . . . . . . . . . . .. 144 4. . . . . . . .B-C starting with 4000 A molecnles. . . . . . .10 Normalized concentration of C versus dimensionless time for the series reaction A .20 Benzene conversion versus reactor volume. . . . . . . . . . . . 113 4. . . 124 4. 118 4.. . .23 Molar flowrate of ethane versus reactor volume for inlet temperatures of 1000. . . . ko = kC. .34 Envelope versus time for hepatitis B virus model. . kJ = 1. . . . . . . n ~ 1. L! = 0. . . . . . . . . log scale.. . . .. . .B-C starting with 100 A molecules. . . . . . ethylene and NO versus reactor volnme for ethane pyrolysis example. kz = k. .. 122 4. . 1050 and 1100 K. . . 158 4. .35 Species cccDNA versus time for hepatitis B virus model.········· . .1 5. . . . . . . . . • 167 4. . . .25 The CSTR is smaller than the PFR for irreversible. . deterministic and average stochastic models.!i. .. . 162 4. . . . . . . . Comparison of measured and calculated rate constant versus temperature for oxlrane decomposition. . . . . . . .. . . . . . . . 5. . .. . llS 4. . . . . . . cAo=l. . . . . . . .!. .2 5. .. . . . . .. . . . . 171 4. . .. . .5. . . . 161 4. . . . . . .21 Component mole fractions versus reactor volume. .29 Stochastic simulation of the first-order reactions A. 171 4. 120 4. nthorder kinetics. . . 125 . . • • • . n :2: O. . . . . . . . . . . .27 Overall and per-pass conversion of A as a function of fractional recycle. . . . . . . . . . . . . . . LJ ~ 0. . . 169 4. . . . . . . . . . . . 5.. . .. .11 Two first-order reactions in series in a batch reactor . . CAD = 1. . . . . . . . . . . ..1 4. . . . . . .!. . . . . . . . . . . .. .7 Reaction rate versus concentration for nth-order kinetics r = kcA. . . .4 5. 4. .8 Batcbreactorwith nth-order kinetics. . ... . . . . . . . 128 4. . . . . . . 147 4. . . . . . . . . . .. . . . . . . . . . . . 154 4. .. . Contour representation of the potential-energy surface. . . kz = Lz ~ 10. . 160 4. . 102 Reactor volnme element. . .I = 1.. . . . . . . .13 CSTR volnme element. . . . .. . . . . . . . . . . . . . .. .. . . Reaction-coordinate diagram. 172 4. . . . . . . . . . . . . . . ko = kc. .30 Stochastic simulation of the first-orderreactions A-B-C starting with 1000 A molecules. positive order. . . . 168 4. . . .18 Semi-batch reactor polymer content for different monomer addition policies . • . . . . . .5. 124 4. . . . . . . . . . . . .24 The PFR is smafler than the CSTR for irreversible. . .16 Semi-batch reactor feed flowrate for different monomer addition policies . . . . 196 197 198 199 208 210 210 211 215 221 . . . . . . . . _ . . . .

. . . . . . . . .. . . .. . . . . . • . . 224 5. K. . . . . . . ' Eigenvalues of Jacobian matrix vs. . . . . . . .. . 252 5. . . . . .log scale . . . . .. . ... . . . .. . . . . .33 6_34 Rates of heat generation and removal for e = 1. . ... . .4 6.. .. 245 5. .19 Langmuir isotherm for CO uptake on Ru.13 Molar flowrates of C2H6. . . . . Tubular reactors with interstage cooling . . . . . ... Phase portrait of conversion versus temperature for several initial conditions. . . .29 6.. . . . . . . . . . . A B . . . . ..3 x 10' kJ/kmol. . . .22 Five possible arrangements around a single site. . .10 6. 243 5. . .log scale . . . .. . Phase portrait of conversion versus temperature at showing stable and unstable limit cycles. . . . . . . Temperatures and molar flows for tubular reactors with interstage cooling. . . . ..32 6. .. . .. reactor temperatm'e . and point B is unstable . . . . . . .8 6. . . . . residence time . . 6. . . . . . .20 Linear form of Langmuir isotherm for CO uptake on Ru. Steady-state conversion versus residence time. . . .21 6. . . .23 Dimensionless C02 Production rate versus dimensioniess gas-phase CO and 02 concentrations. time for e = 35 min. . . . . . . . . . . .18 6. points A and C are stable. . . .9 min. . .16 Surface and second-layer (dashed) atom arrangements for several low-index surfaces. .. reactor conversion .. . . . .. .24 6.3 6. . . . . . ignition and extinction points . . CSTR and heat exchangers. . .. . . .. .14 6. . . . . . . . . .. . . .. . . .2 6.30 6.. . . . . . .. . . .. Steady-state temperature versus residence time. . . . . . . . . . .17 Schematic representation of the adsorption/desorption process. . 234 5.. . _ . .B . . . . . . . . .9 6. . .. . . .12 Fractional error in the QSSA concentration of C versus dimensioniess time for the series-parallel reaction. . . .f~r !1~~ ~. . . . . .. . . . . . .20 6. . . .79 min. . .. 247 5. . . . resid~nce t~e . ignition and extinction points. 225 5. . . Steady-state temperature vs. Marble on a track with three steady states. . . C2H4 and CH4 corresponding to the exact Solution. .1th an ignition point (A) and a stable steady state (C). . . 6. . .17 6. ..1 6. . .18 Fractional coverage versus adsorbate concentration for different values of the adsorption constam. . . . .. .. . .. . .. ..22 Steady-state temperature versus residence time for !1HR = .. .. . . . . ... point C is unstable and is surrounded by limit cycles .15 6. . . .. . . 94 2 . .11 Fractional error in the QSSA concentration of C for the series reaction A . . . . . . . . 241 5. . 248 5. .26 6. . . e = 35 min. . . . .7 Steady-state converSion versUS residence time for !1HR ~ -·3 x lOs kJ/kmol. . . . point A is the unique. . . .25 6.. . .. . .. . . . . . .11 6.. . . . . 296 = :~:di~~t~~~~ra~r~ ~ers~s. .. .. Steady-state conversion vs. . . . . . . . . . . . . . Plug-flow reactor volume element. . . .16 6. . 236 5_15 Comparison of the molar flowrates of C2HS and C2 H4 for the exact solution (solid lines) and the simplified kinetic scheme (dashed lines). _ . . . . . . .. . . . . . . . Rates of heat generation and removal for e = 30. . .12 6. 6.. . .. .. . . .. " .. .. . . . . . . .. .14 Fractional error in the QSSA molar flowrate of CZH4 versus reactor volume. . .. . .C . Marble on a track with three steady states. . . . . _ Marble on a track v. . . . . . . '" . . .27 6. .3 min. . . . . Conversion of A versus reactor temperature. . . . . . . Phase portralt of conversion versus temperature for feed initial condition. . . . Rates of heat generation and removal for e = 15 min. . . . 6. . . . . . 238 5. Steady-state conversion versus residence time for different values of the heat of reaction . . . Conversion and temperature vs. e = 35 min. . . . . .6 Steady-state temperature versus residence time for different values of the heat of reaction . . . . . .xviii List of Figures List of Figures xix 5. .5 275 277 288 290 293 293 294 6. . . . 6. . 298 299 299 303 303 304 304 304 306 306 307 307 309 309 310 311 312 312 313 313 314 315 317 320 322 . .19 6.. . . .. ... _ Eigenvalues of the Jacobian matrix versus reactor conversion in the region of steady-state multipliCity . . . 237 5.. . . time for e = 72.13 6. . . . . . _ . . . . . residence time . . . . . . Marble on a track in a gravitational field. 245 5. .. . . . 252 5. .25 Adsorbed oxygen concentration versus gas-phase oxygen concentration. . . . . . . ..21 Two packing arrangements for a coverage of e = 1/2. . . . _ .. . ConverSion and temperature vs. . . .. . . . Real and imaginary parts of the eigenvalues of the Jacobian matrix near points C-P. .28 6. . .31 264 Reactor volume element . . . . . . . . .. . . . . . . . .. time for e ~ 30 min.. Eigenvalues of Jacobian matrix vs. .. . . . . . . . .23 6.. . . . .. .. ConverSion and temperature vs. .24 Dimensioniess C02 production rate versus a single dimensioniess gas-phase concentration .. Eigenvalues of the Jacobian matrix versus reactor converSion in the region of steady-state multiplicity. Flow streams entering and leaving the volume element. . . stable steady state. . . .C . .. . . . . . . . . . . . . . . .

.. .. .. . . . c z . . . . ' . 7. 384 6. .. .27 Molar concentrations versus reactor volume .. . . . . . . " 350 7. . ' . . . . . ' .. .13 Effectiveness fact . .. . with energy flux e ss uxes N.. oni~ ~oi~ fr~~t. . .. . . . . . . . .p' conSider th f ' c . . fluld concentration of 02 (xl. . . . CO (+). . . . . .. .. .. . .. . . . . .. . .22 Concentration profiles of reactants. . . .29 Fluid molar concentrations versus reactor volume. . .. . .. .. . .15 Effect of external mass transfer on pellet surface concentration. . . 7.11 Dimensionless con . . . . first-order reaction in a spherical pellet with.19 Effectiveness factor versus nonnalized Thiele modulus for a first-order reaction in a nOnlsothermal spherical pellet.39 Re actor and co' I hree steady-state solutions . .. . . 326 at reactor inlet. . . 7. ' . .. I 2 DimenSionless . . .. " 7. . .. . . . . . . . . .18 Asymptotic behavIor of the effectiveness factor versus Thiele modulus.. . at reactor inletrature at reactor outlet versus temperature 6. . . . augen-Watson kinetics . .og sca e. . ... . : ' . . 372 lC and semi-I' fini. ...38 Coolant temp::CtlOn IS used to preheat the feed. .ill """""'" 378 pellet' rea r r versus Thiele modulus in a spherical . isothermal reaction in a fixed-bed reactor. . .. . . . . . ' . .. . 369 order react. ...24 Concentration profiles of products ' .. . . 325 exothermic r P ug-fl~w reactor. C3H6 (*). . . " """""'''' concent ra rIOn . . ... . . . . . . . 7. ... . . . ' C IOn orders less than unit 7. . the heat released by the 6.. .37 Autother~ai . C3H6 (*). . ' . ' . .. . .. . .. . ... . . . . .:. . . . . .30 Fluid temperature and pressure versus reactor volume. . .1 Expanded vIews 0 f fix 7. " ' " 7. . . ' .. .... 7.26 Molar flow of A versus reactor volume for second-order. . . . 7.. . . . .. . UIlstable middle (8). . " 7. . . . . . .o~ 'v~. . . . . " .. ' _ . : ' . . 0 ant temperature profiles versus reactor 4 33 : .. .20 Dimensionless concentration versus radius for the nonisothermal spherical pellet: lower (A). . . " 376 pellet' r ' r versus ThIele mOdulus in a spherical " eaction orders greater than u 'ty 7. .. 7. . . . . " 7./ . . 374 cylinder and slab or versus Thiele modulus for the sphere. . .s~~ ~~a~~o~ i~n'g. .] 0 EffectIveness facto . . . . .. . : ' . . . 7. y.. . . . .. . . 7.. . .. . ' . . .] 6 Dimensionless concentration versus radiUS for different values of the Biot number. . . . . ' . semi-infinite cylinder n te slab.35 Molar flow of 0- I i . .. CO (+l. ength a for sphere. . . 3 30° 6'4 °1 cAmom 6. .. . .17 Effectiveness factor versus Thiele modulus for different values of the Biot number. . . . and upper (Cl steady states.. . ..3 Hyperbolic t ' J' . . . .. . 7. . ' .. '..4 DI SlOn ngonometnc position for dif~oncentratlOn versus dimensionless radial ' 7' 5 Effect lveness f erent values of the Thiele modulus ' .. . unstable middle (B). • . . . . . '. .. . . first-order reaction in a spherical pellet. the equations show the coupling between the catalyst particle and the fluld ' . . . 378 reaction' centratlOn versus radius for zero-order . .. . . . . . 7.. ' . 7. .acto r versus Thiele modulus for a firstIOn m a sphere 7. . .. . . . . I) 7. : . . . .versus z for large . . . .. . . . 0 l ant tempe . ... .. . . . . 384 385 387 3 88 389 394 396 396 398 398 401 404 408 411 411 414 414 415 415 419 .p = L . . .. . . . . . .. .. . . . . .7 Characterise I .. ' . . ' . . aud upper (Cl steady states. .. . . ' . . .. 350 ovmg wIth the PFR flUId veloCIty. . 329 length . . ' . . 7.. ..42 Volume element'~' " . . 372 order rea t' ... .6 Effectiveness fac ' . . . .. . . . • • . . .functions sinh... .25 Fixed-bed reactor volume element containlng fluid and catalyst particles. . .. . . . . . . Hougen-Watson kinetics. . " ..h 7. . .. .14 Effectiveness factor versus appropriate Thiele modulus in a slab.. 7. .28 Dimensionless equilibrium constant and Thiele modulus versus reactor volume " " " " " " " " " " " " 7. . . 364 'men' Ie s s . .. . . . cosh. . . . . . : .tor versus Thiele modulus for a firstc IOn m a sphere (log I . .. 7. . 382 ulus in a sl b' H or versus an mappropriate Thiele moda. . . m a spherical pellet 379 7. . . . . ...8 Effectiveness fact ' . . . .. . .2 Volume element a ed-bed reactor. fluid concentration of 02 (x). . ' . .. 325 perature p ature versus length for different feed tem6. . .. ...31 Reactor pOSitions for pellet profiles. .21 Dimensionless temperature versus radius for the nonisothermal spherical pellet: lower (Al. tanh ..xx List of Figures List of Figures xxi feed Ie xy ene versus reactor length for different mperatures 636 ReactOrtem er . 7. . 7.. ' . 369 7. first-order reaction in spherical pellet .o ' . . ~ture at reactor outlet versus temperature 6.9 Effectiveness fac. . . ' .. . . 357 density E d In a statIOnary solid containing energy and rna aufl molar Concentrations Cj.. . . ' . . e unctIOn z (e) in place of -( ) 7.. . .23 Concentration profiles of reactants (log scale). . . .32 Pellet CO profiles at several reactor positions. . . .... . .. .33 Catalyst pellet with slab geometry. . .

. . . . . . . . 8. . .13 PFR followed by C5TR (A). . .11 Residence-time distribution p(8) versus for plug flow with dispersion number D. .. . . . . . . . . .25. and C5TR followed by PFR (B). . .3 9. . . . . .. Multivariate normal for np = 2. . .10 pte) versus e for plug flow with disperSion number D.. 8. . . . . ideal CSTR. . . . . . . 8. 8. 472 8. . . .. . . . . . Mass spectrometer schematic. . . outlet flows adjusted to achieve given RTD. .43 Value of fooo j(e)p(O)de as the function pW) varies . . rate of the mean (rm ) is less than the mean of the rate (rs ). .. .. . . . . . . . . . . . 471 8. . . . . .. . . . . . . 495 8. . . 487 8.2 Family of pulses and the impulse or delta function. . . . . . . . . . . . . . " . . . . . .17 Maximum mixed flow as a plug-flow reactor with side entrances. . 490 8. . . . . .34 Effectiveness factor versus Thiele modulus for different values of the Biot number. . . . .16 Alternate representation of completely segregated flow (A). . . . . . . . . . . . . . . . . ideal CSTR. . .6 Plug-flow reactor with moving front of tracer. differentiability is not required. . T = 2 . . and 2-CSTR mixlng 484 model with p = 0. . . . . . . . . . . 0 . . . .32 Imperfect mixing (top reactor) leads to formation of an A-rich zone. . 8. . . . 449 8. . . .12 Start-up of the tubular reactor. . . . . . . . . . . .1.'. . optimal sequence to achieve 95% conversion is PFR-CSTR-PFR. . . 8."verse of reaction rate versus concentration. . . . . which is modeled as a small CSTR feeding a second CSTR (bottom two reactors). .xxii List of Figures List of Figures xxiii 7. ..24 Differentiable convex and concave functions. .19 Dimensionless effluent concentration versus dimensionless rate constant for second-order reaction . . . . . . . . . . . . . . . . . . .31 RID for the optimal reactor configuration . .1 9.8 RID p(8) versus e for n CSTRs in series. . . . and as a function of internal flowrate.. . . . . . . . . . .. . . 487 8. . . . .29 Reaction rate versus concentration of limiting reactant. . M = 0. . . . . . . . . . . . . . . . ..2 9.28 Mean segregated and mixed concentrations versus 8. . . . . . . . . 476 8. . .5 Gas chromatograph schematic. . . . . . . . . . . .42 Some convex functions. . . . . and as a function of internal flowrate in a 2-CSTR mixlng model . . . . . . . . . 479 8. 473 8. . . . . . . 8.25 Two volume elements before and after mixing. . " . . . . . . . . . . . . . . . 8. . .18 Volume element in the state of maximum mixedness. . . . . . . . . . . . . . . .15 Completely segregated flow as a plug-flow reactor with side exits. . . . . . . . . . . " . . . . . . . . . . . . .3 CSTR and tracer experiment. .. 2. . 8. t . 504 505 511 515 516 . CA (z. . . . . . . . . . . 475 8. . .. . . . in a 2-CSTR mixing model. . . 8. • . . . 8. . . . . .. . . T = 435 437 439 440 441 443 444 445 446 2. . .. 498 9.39 Tracer concentrations in the feed and effluent streams versus time. . . . . . . . . . . . . . . . . . . 497 8. . . . . 8. . . . . . . . second-order reaction in a cylindrical pellet. .. . 8. . . . e 449 452 452 454 455 456 458 459 463 464 465 468 468 469 8. . . 8. . . . . .27 Two tubes before and after mixing the entering feed . . . . . . . . . .. . . .. . ..30 I. .5. . . . .14 Comparison of the effluent concentrations for the two cases shown in Figure 8.26 Convex rate expression and the effect of mixing. . . and an intermediate mixing pattern (C).. . .4 C5TR and volume elements in a game of chance. . . . . . t) versus z for various times. . . . . . . inlet flows adjusted to achieve a given RID. . . . 8. . . ideal CSTR. . . . . . . . . . . 8. . 474 8. . . .. . . . . . . . . . .33 Conversion of reactant for single. . .4 9. .. . . . . . . . . . . . . . 8. . . . . .20 CSTR followed by PFR (A) and PFR followed by CSTR (B) as examples of complete and partial mixing .36 Conversion of reactant A versus reactor length for different dispersion numbers. . 8. . . . 8. . . . . maximum mixed flow (B). . . . 429 8. . . . 8. . . . . . . . .7 Series of n equal-sized C5TRs with resideuce time Tin. 494 804 1 Effluent concentration versus time after unit step change in the first reactor. . . . . . . . . is a contradiction.22 Total concentration of A in the reactor effluent versus particle size . . . . . . . . . . . . " . rate expression is neither convex nor concave.. . .5 C5TR residence-time distribution. . .13 .34 Yield of desired product C for single. . T = 2. . 491 8040 Reactor before and after stirrer is started . .. . . . . . Univariate normal with zero mean and unit variance. . . . . . . .38 Reactor configurations subjected to a step test in tracer concentration. . . 8. . . . . . . . . . 483 8. . . . . .23 Particle concentrations of A and B versus particle age for three different-sized particles. . . 8. . . . . . p = QrIQ2. . T = 2. . . . . . . . . . . . . . . . . . curves crossing at e. . . . . . . . . . . . . . . 483 835 Step response for single. . . . 8. . . . . . . . . .. . . . . . . . . . Volumetric chemisorption apparatus. .21 Adding two liquid-phase feed streams to a stirred tank. . . . .37 Yield of deSired product B versus reactor length for different dispersion numbers. . . . . 470 8. . ...9 P(8) versus e for n CSTRs in series. . . . .1 Arbitrary reactor with steady flow profile. . . . . . . . . . . .. . . .. . " .

. . . . . . . . . isothermal reaction in a fixed-bed reactor. . . . Dimensionless concentration versus radius for the nonisothermal spherical pellet: lower (A). . . . . . . . . .16 Model fit to all adsorption experiments . .. . . . .. . . . .. .3 7 Parameter estimates and confidence intervals for reduced model with redesigned experiment. . Transformed data set. . . .7 A. . Relative error in the effectiveness factor versus number of collocation points. . . _ . . . .. . . . . . . . . . . unstable middle (Bl. . .13 Parameter estimates with only 10 data points. . . . . 9. A.12 9. . . . " 522 523 Reducing parameter correlation by centering the data. . . . . . . . . . . . . . . . .9. . . . . . . .. . . . . and upper (Cl steady states . . . . . . . . . . . . .22 Monte Carlo evaluation of confidence intervals .S A. . . . . . . . . . . .. 541 543 9. . . Solution to first·order differential equation dCAldt ~ -kCA' and sensitivities S.lnk versus liT . . . . . .xxiv List of Figures Ust of Figures xxv The geometry of quadratic form x T Ax ~ b. . 526 9.19 Uncertainty in activation energy E and rate constant In k m versus next measurement temperature. .. . . two approximations and exact solution. . . . . 536 9. . " 521 Distribution of estimated parameters . . . . 594 594 596 . . . . . . .. _ . . . . . . . . 9. . . . 9.. . . .. .. . .18 Effect of next measurement temperature on parameter confidence intervals. . . . . C and D versus time. . . . . 9. . . .. . . 554 9. .9 9. . . . . . .. . . . ~ oCAlok and S2 ~ oCAlocAo . . .39 !latch-reactor data for Exercise 9. 9.6 9. .2 A. . . . . . . . . . . . . . . .7. . . . . . . . . . . :. . . 537 9. .21 Experimental measurement and best parameter fit for nthorder kinetic model. . . .25 Envelope versus time for hepatitis B virus model. . . . .42 Batch·reactor data for Exercise 9. . " 9. . . . . . . . . 9. . . .41 A second experiment for Exercise 9. . . . .. 530 9.14 Confidence intervals with known (solid line) and unknown (dashed line) error variance..24 Species reDNA versus time for hepatitis B virus model' initial guess and estimated parameters fit to data. . 9. . . . Ink versus liT. .. . . . Gibbs energy contours for the pentane reactions as a function of the two reaction extents. . Magnified view of figure A. . . . 517 520 Measured rate constant at several temperatures.. . .36 Fit of LC measurement versus time for reduced model.. . . . . . . . . . .. 3 mns with different measurement error variance. . . 9. .35 Predictions of LC measurement for reduced model. .10. : . . . . ..30 Depiction of an LC curve for determining the concentration of intermediate C and product D. . .9 Estimated reaction rates from 2000 production-rate measurements subject to measurement noise. . .40 Batch·reactor data for Exercise 9.23 Species ceeDNA versus time for hepatitis B virus model' initial guess and estimated parameters fit to data. . . . .. . .. . . 9. . . . . . Values of X2 and Fversus the number of data points when estimating 2 and 5 parameters. .17 Drawing a line through two pOints under measurement error: points far apart (A). 556 9. ..27 Species reDNA versus time for hepatitis B virus model. ..43 A second experiment for Exercise 9. . . Molar flow of A versus reactor volume for second-order. r ~ kc. '.15 Model fit to a single adsorption experiment. . . .. .29 Semi-batch reactor addition of component B to starting 553 material A. .20 Uncertainty in activation energy E and rate constant In k m versus number of replicated experiments. intial guess and estimated parameters fit to data. . . ... .33 Concentrations of species A. . . .31 Base addition rate and LC measurement versus time. .. .S A... . . . 531 9. . . . . ..4 A.8 9. . . 546 549 9. .10. Dimensionless concentration versus dimensionless radial position for different numbers of collocation points. . . . . 533 9. . . 526 9. .. . . .34 Concentration of species B versus time. . . . . . .2. . . . .{ . . .6 A. . . . 546 9.7 9. . . . . 534 9.. . . .11 9.3. . . 520 Several replicate data sets.10 9. . .l 557 560 561 561 566 567 571 571 572 572 577 581 588 589 591 592 A. . . . . 525 9. 557 9.26 Species eeeDNA versus time for hepatitis B virus model. 546 9. . . 555 9.. . . . . . . . . . . . . . early time measurements have been added. . . . .38 Batch reactor data for Exercise 9. . points close together (Bl. . . . . ..3 A. . . . . . .9. . function C (r) and its values at five collocation points . .28 Envelope versus time for hepatitis B virus model. .. . . 549 549 9. 9. .32 Comparison of data to model with optimal parameters . .. 9. . . . . . . . " 9.

. . .9 Energy balances for the semi-batch reactor. . . . . . .4 Steady state and eigenvalue with largest real part at selected points in Figures 6. . . . . .. . . . . . . ..6 5. .5 PFR operating conditions and parameters for a-xylene example. . . . . . . .11 Thermodynamic data for allyl chloride example. .. . . . . . and semi-infinite slab.. . . . .2 4. . . . . . . .3 Parameter values for limit cycles. .2 7.3 Industrial reactions over heterogeneous catalysts. . . . . . . . 6. . . 359 Porosity and tortuosity factors for diffusion in catalysts. . 6. .1 5. Ethane pyrolysis kinetics. . . . . . . . . . . . . . . . .6 Parameter values for Example 6. . . . . . .8 6. . . " 6. . 173 Molecular partition function terms . . . . . . . . . . . . . .. ._ ! . .. . . . 6. . . . . . . . . . . . . . . . . Gas-phase oxygen concentration and adsorbed oxygen concentration. . . . . . . . 6. . 376 . . . . . . Fast aud slow time-scale models in extents of reactions . . . . . . . . . . . _ . .1 7. Chemisorption and physisorption properties . . . . " 6. . . . .. 363 Closed-form solution for the effectiveness factor versus Thiele modulUS for the sphere.21. . .. . .1 6. . .2 5. . . . .7 Energy balances for the batch reactor.. 7. .8 Energy balances for the CSTR. .20 and 6. . . .. . . . . . . . . . . . .2 Reactor balances for constant-density and ideal-mixture assumptions. . . . . . . . Exact and QSSA solutions for kinetic Schemes I and II. . Fast and slow time-scale models in species concentrations.7 5. . .. . . .6 . . . . . . . . . . 201 206 213 214 222 233 239 264 291 296 305 308 324 329 332 333 334 335 340 Parameter values for multiple steady states.1 4.4 5. . . . . . . Selected values of steady-state temperatures and residence times. . . . . . . . . . . . 6. 6. . . . . .list of Tables 4. .5 5. . . . 6. . 140 Summary of mole balances for several ideal reactors. . .3 5. . . semi-infinite cylinder. . . . . . Parameters for the trioxane reactiou. . 139 Reactor balances for general equatiou of state.10 Energy balances for the plug-flow reactor.3 5. _ . . . . . . . . . . . . . . .

. . . . . . . . . . . . .. . .1 5. . .3 4. . . . . . . . . . . . . Minimum in G for an ideal gas . . . . . . . . . . . . . . . . Ethane pyrolysis in the presence of NO . . . . . . . . . The PFR versus CSTR 'vith separation . . . . . . . . . . .. . . . . . . . . . . . . . . . . . .. . . . . . . deterministic simulation of a virus model . . . . .3 5. . .2 2. .4 7. . . . Stochastic VS. . . . . . . . .8 Stoichiometric matrix for a single reaction . . .. . .9 253 . . . . . . . . .4 2. . . . . . .4 2. . . . . . . . . . . . . . . 30 30 31 33 34 37 39 45 66 68 73 82 90 93 130 131 141 149 152 155 156 161 170 205 226 228 230 232 244 249 250 3. . . . . .4 5. . . . . . . Equilibrium composition for multiple reactions . . . . . . . Reaching steady state in a CSTR . . .1 2. . . . . . . . . . . . . . . .5 4. . . . . .. . . Columns of v . . . . . . . . . . . . . . .3 2. . . . . . Production rate of acetone . . . . . . . . . More species than reactions . . . . . Free·radical polymerization kinetics . .7 5. . . . . . Chemical and phase equilibrium for a nonideal mixture . . . . . . . . . . . Maximal set of reactions for methane oxidation . . . . " Equilibrium CO and 0 surface concentrations . . . . . . Phenol production in a CSTR . . . . . . . . .4 3.5 2. .9 Ideal-gas equilibrium . . . . . Semi-batch polymerization . . .. . . . . . . .3 8.. .. . . . . . .1 8. . . . .5 7.. . . . . . . . . . . . . _ . . . . . . . . . . . . . . . . . . Multiple reactions with optimization. . Ideal-gas equilibrium. . . . . .6 Controlling mechanisms for pellet reaction rate given finite rates of internal diffusion and external mass transfer 389 Kinetic and mass-transfer parameters for the catalytic converter example.' Computing a rate constant . . . . . . . . . . . . .8 5. . _ _ _fl" . . . . . . . . . . . . . . . . Benzene pyrolysis in a PFR . . . . . Rowsofv Stoichiometric matrix for CVD chemiStry . . . . . . .6 4. . . . . . . . . . . . . . . . . . revisited. .2 5. . . .2 8. . . . . . . .1. . . Benzene pyrolYSis. . . . . . . . . . . . . .5 5. .3 3. . .' . . . . . '. . . . . . .2 4. . . . . .xxviii List of Tables 7. . . .6 2. . . .5 3. . . . . Production rate of oxygen . . . . .8 4. . . . . . . . . . . . . . .6 4. . 412 Mass-transfer and kinetic parameters for Example 8. . . . . . . . . . . ..2 3. . revisited . . . . . . . . . . . . Conservation of mass . . . . ...4 4. . .1 4. . . . . . Ethane pyrolysis . . . . . . . . .6 5. . . . . . . . . . . . . . . Production rate of methane 5. . . . . . .7 4. Production rate of C02 . . . . . . Fitting Langmuir adsorption constants to CO data .. . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . Reactor and kinetic parameters for feed-mixing example. . . . . . Parameters for the dispersed PFR example.. 451 466 480 485 list of Examples 8. . .7 2. . . . . . 399 Feed f10wrate and heat-transfer parameters for the fixedbed catalytic converter. . Production rate for SiH2 . . . . .1 3. . . . . . . . . . . . . . . . . . . .. Changing flowTate in a PFR . . Mass-transfer and kinetiC parameters for micromixing problem.

. . . . . 8. . .5 Fitting rate constants in hepatitis B virus model . .2 Liquid-phase batch reactor . 9. . . . . . . . . . Multiple-reaction. sizing the reactor. .. 9. . .. . . .. . ... . . . In addition to meeting production and purity goals. .. _ . . 590 A4 A simple fixed-bed reactor problem .. .. . . . .. . As we will see.. .1 Introduction Transient start-up of a PFR ... or fabricating microelectronic devices. ... . . ..1 7. .. ... nOnisothermal fixed-bed reactor . . . .6 Ammonia synthesis . . 592 > ••• > • > • • • • The chemical reactor lies at the heart of most chemical processes. and minimizing the amount of feed or product that must be kept in storage. . . . . . .4 Optimal is neither segregated nor maximally mixed . . . . . . . . . First-order.. but the consequences of the prinCiples arc rich. 6. . .1 Constant-pressure versus constant-volume reactors . .. In the overall process. . . .. .5 7. .. . . The chemical reactor is that essential component in which the feed is converted into the desired products. 1. . .. . . . . . is a highiy integrated operation. . .. . .. Estimating the rate constant and activation energy .2 Order matters . . .. : : : : : : : : : : : : : 8. .3 Temperature control in a CSTR. . ... . " Catalytic converter .. . . specifying initial and feed conditions. .. and concentrations of species. 6. . . .. . . .... . 6. . isothermal fixed-bed reactor. . . . .. .2 281 283 287 319 323 327 371 397 403 405 406 409 412 1 Setting the Stage 7. 8. the process design and operation also are influenced by the sometimes conflicting goals of minimizing energy consumption. sometimes the reactor is a living organism..3 7. . . .. . . . . . These principles can be applied at many size scales to many different types of chemically reacting systems. " Fitting single and multiple adsorption experiments .6 Maximizing yield in dispersed plug flow . . . . . 6. . . . . .. .2 9. . 8. . " 586 A3 Single-pellet profile . .3 8.7 Using the Thiele modulus and effectiveness factor . . . . . . . While the reactor of interest is often a man-made vessel.. . . . . . . specifying operating conditions. . . refining petroleum. pressure. 9.. . . . .xxx List of Examples 6. . .. . . isothermal fixed-bed reactor. . we construct and illustrate a set of reactor analysis and design principles. . using . . . .. 7. . . 8. . . .4 Fitting reaction-rate constant and order ...5 Mixing two liquid-phase streams in a stirred tank . We will illustrate throughout the text... including predicting performance. . . . . .4 7. . The reactor is normally followed by separation processes that separate the products from the unreacted feed and the reactor byproducts. The chemical reactor is the place in the process where the most value is added: lower-value feeds are converted into higher-value products... .6 7. . . . . .. . . . . . . . . ..1 9. . Second-order. manufacturing pharmaceuticals. diverse and complex. 6. . . . . and selecting the reactor type. . the prinCiples are small in number. . .. whether processing fine or commodity chemicals.. . . . Unknown measurement variance and few data points . . ... " 577 A2 Two simple sensitivity calculations . ... ....1 450 451 461 474 479 484 519 525 529 541 545 Al Estimating reaction rates . Mass-transfer limitations in a fixed-bed reactor . . Many aspects of reactor analysis and design are treated in this text.. . . . .5 Oxidation of a-xylene to phthalic anhydride .. .... . . .. . The modern chemical processing plant. . . . . . . .. . . . Hougen-Watson kinetics in a fixed-bed reactor. . . . .3 Two CSTRs in series . .. . . . . .4 PFR and interstage cooling . . The design and operation of the reactor often determine the success or failure of the entire process.. To accomplish these goals.. feedstocks are delivered to the chemical reactor at the appropriate temperature.. . . .

or in some situations. semi-batch and continuous. We make extensive use of numerical methods and high-level programming languages in this text to solve complex models. but the reaction occurs in a different phase. and other models are needed to describe the reactor. and coupled sets of ODEs or algebraic equations are needed to describe the temperature. often a solid phase. and a single ODE or algebraic equation describes a single species concentration or extent of a single reaction. and. Concentration. and the photochemical reactions of chlorocarbons in the troposphere that lead to ozone destruction for gas-phase reactions. pressure and species concentrations. differential algebraic equations (DAEs). such as specialty fioe cliemicals. The batch and continuous-stirred-tank reactors are assumed to be ideally well mixed. continuous-stirred-tank. If the flow in a tubular reactor is not turbulent (Reynolds Number less than lOL 104 ). We model many chemical reactors using three maln reactor archetypes: batch.e. The material and energy balances of these three reactors are sets of first-order. Batch reactors also are used in 1. many complex chemical reactors can be well approximated by these three simple reactor types. the reactor is charged with reactants. The heterogeneously catalyzed reactions such as zeolite-catalyzed cracking of high-hailing crude oil and iron-catalyzed ammonia synthesiS from hydrogen and nitrogen involve two phases. temperature and pressure are therefore the usual dependent variables that are solved as functions of time or distance along the reactor as the independent variahle. Sometimes the reactants and products are transported through the reactor in one phase. and after some processing time the contents of the reactor are removed as product. continuous-stirred-tank. Another iIDP. I.2 Some Classifications and Terminology Ideal mixing and plug flow. Multiphase reactions need not involve heterogeneous catalysts. Examples include: steam cracking of ethane to ethylene. such as synthesis gas (CO + H2) reactions over iron-based catalysts that are suspended in a high molecular weight alkane to help moderate the reaction temperature and dissolve the reaction product. flow in the limit of large Reynolds number. and plug-flow reactors. the fusing of limestone and charcoal to produce calcium carbide. The operation of the reactor can be classified as batch. high-level programming languages are readily availahle and easily can he used to solve the complex models. occurs via free-radical intermediates in the liquid phase of the reactor. the diversity of systems and issues addressed by chemical reactor analysis and design. only in the axial or tube length direction. Batch reactors. high value-added products. such as describing a single reaction in an isothermal reactor. the reaction takes place. pressure and species concentrations are independent of spatial position within the reactor. The reactants and products of homogeneous chemical reactions are in a single phase. continuous-stirred-tank. the design problem is approached in the same manner using the same set of prinCiples. Plug flow often describes well the limit of fully developed turbulent flow. and the manufacture of low-volume. and plug-flow reactors. numerical methods are required to solve the models. Batch. Fortunately. semi-batch and continuous operation. and esterification of an acid and alcohol to produce an ester for liquid-phase reactions. enzymatic isomerization of glucose to fructose. The plug-flow reactor describes a special type of flow in a tuhe in which the fluid is well mixed in the radial direction and varies . In all but the simplest cases. Homogeneous and heterogeneous reactions. The great simplification that has hecome standard practice in introductory reactor design texts is to neglect the momentum balance and a careful treatment of the flUid flow pattern within the reactor. More often the design involves many reactions and nonlsothermal operation.ortant reactor classific'ation pertains to the phase in which the reaction occurs. nonlinear ordinary differential equations (ODEs) or nonlinear algebraic equations. Batch. noncatalytic reaction is the low-pressure epitaxial growth of Si(100) films from gas-phase disilane (Si2HG). Regardless of the complexity. The reaction of liquid p-xylene and gaseous 02 to produce liquid terephthalic acid. Sometimes the reactor model is quite simple. and the interfacial reaction between strontium oxide and silicon dioxide to form strontium silicate for solid-phase reactions. By virtue of their design and the typical operating conditions.2 Setting the Stage 1. which means that the temperature. and plug-flow reactors are defined by certain idealized assumptions on the fluid flow. In hatch operation. are often used for liquid-phase reactions. depicted in Figure L l. Another two-phase. pharmaceuticals and fermentation products. Sometimes three phases are present simultaneously during the reaction.2 Some Classifications and Terminology 3 many different examples. The batch. then the flow may not be well modeled as ideal plug flow..

The CSTR is used extensively in situations where intense agitation is required. tubes for adding (sparge tube) and removing (sample tube) fluid. It is common to employ a cascade or series of CSTRs in which the effluent from the first reactor is used as feed to the second and so forth down the cascade (Figure 1. The cascade permits one to realize high conversion of reactant. The wellmixed assumption can be realized more easily for liquids than gases. composilton and temperature . while mjnimizing total reactor vohnnc. and often heating the reactants to the reaction temperature. The semi-batch process is similar to the batch process except feed addition occurs during the batch cycle. Cour~ tesy of Autoclave Engineers. The great flexibility of batch processing allows the reactor to be used for the manufacture of many different products. The continuous-stirred-tank reactor (CSTR) is shown in Figure 1. and cleaning the vessel before the next charge of reactants. emptying the vessel. the velocity proflle becomes plug-like.4 Setting the Stage 1. and the suspension of a solid or second liquid within a continuous liquid phase. so manufacturing costs can be considerably higher than a corresponding continuous process. and the contents of the reactor are assumed to be well mixed. The addition/removal policy allows one to control the reaction rate or heat release during reaction.3. tubular reactor as depicted in Figure L5.2: Expanded view of the internals of a batch reactor. Figure 1.2 provides an expanded view of the internals used in a batch reactor showing the cooling coils. The plug-flow reactor (PFR) is a constant cross-section.1: Schematic diagram of a batch reactor. Product mayor may not form during these preparation steps and they often involve manual labor. Reactants and products flow into and out of the reactor continuously. Figure 1. so CSTRs are often used for liquid-phase reactions.3: Schematic diagram of a CSTR. Products may also be removed during the semi-batch process. The cycle often ends by bringing the contents to a discharge temperature.2 Some Classifications and Terminology S T Figure 1. Under turbulent flow conditions.4). and baffles to ensure complete mixing. The velocity. Inc. The semi-batch reactor also may provide more complete Qj Cj Q T -Ej cjf Ij VR T Figure 1. the composition and temperature of the effluent stream are identical to those of the reactor. such as the addition of a gaseous reactant to a liquid by transfer between the buhbles and the continuous liquid. the effluent composition of the CSTR is identical to the conditions that exist in the reactor. Division of Snap-Ute. Polymerization reactions are sometimes conducted in CSTRs. which greatly Simplifies the material and energy balances. The batch cycle begins by charging the reactants to the vessel. situations where it is not practical to implement a continuous process. use of the reactor volume in reactions such as polymerizations that convert lower-density reactants to higher-density products during the course of the reaction. This unit is placed in a cylindrical vessel that comprises the exterior of the batch reactor. The fluid composition and temperature undergo a step change when passing from the feed stream into the interior of tlie reactor.

Copyright 2001. the reactor is referred to as a fixed-bed reactor. As many as 256 different catalyst formulations can be deposited into shallow rights reserved. If the PFR is filled with a porous catalyst and the fluid flowing in the void space is turbulent.6] 7. -I ~~ Cj2 Thermal Isolation and product transport ej2 Thermal insulation Q.ign patents pending. used .3 Scale 7 Qf ejf T Heating plate I f r:--. U. Additional US and for~ e.004..S. PFRs are employed in situations that require high capacity and high conversion. Plug-flow reactors are used for gas-phase and liquid-phase reactions. version of reactant per unit volume than the CSTR. lab-on-a-chip reactor systems.e.6 presents a schematic view of 1 pi test reactors that are used for combinatorial screening of heterogeneous catalysts. S. PFRs also are used in situations involving highiy exothermic or endothermic reactions. there is more efficient use of volume in a PFR.t. Inc. Q Cj T P 1.6: Cross~sectional ! restnctor Bon~ed . or surrounded by a high-temperature boiling fluid for exothermic reactions. Patent No. 6. are functions of only the axial position.3 Scale The size or volume of chemical reactors varies widely. Figure 1. PFRs are the reactor of choice for gas-phase reactions due to the problems of mixing gases in a CSTR. i. Figure 1. lo CjJ Catalytic reactor QJ Catalyst support wafer t Bonded wafer distributor! i ejl restrictor T1 VRI T.7 presents an assembled view of these reactors. one is interested in determining if a reaction proceeds and in minimizing the scale of the experiment so many combinations or conditions can be screened rapidly.6 Setting the Stage 1. In the combinatorial reactors. the effluent of each reactor becomes the feed to the next. '2 3 I T3 Cj3 VR3 T] view of a 1 Jil combinatorial screening reactor. a bundle of small-diameter tubes can be placed in a furnace for endothermic reactions. to several hundred thousand liters for certain petroleum refining operations. 1 T3 CJ4 1 I~ ~ 14 V R4 Qs Cj5 Cj5 1'5 VRS Ts Figure 1. we are usually interested in only the steady-state profile in the tube and neglect the dynamics. Figure 1. In this text. Symyx Technologies. We will see that the isothermal PFR usually leads to higher con- .9RS.3S(). 1'2 VR2 T. For this reason..ier distributor/ Sorbent support wafer Cooling plate ! I--Q.5: Schematic diagram of a plug-flow reactor.. r~ Q3 Cj3 l Ie Adsorption detection reactor Figure 1. Reactor volumes can range from hundreds of nanoliters for combinatorial.--' with permission.4: Schematic diagram of five CSTRs in series.

7: Assembled view of 256 ! -pL reactors built into a 3-in by 3-in support wafer. In d Vl'ftl(.1 "iU!! !ill'.f(>I! 1..\l (I!sp!.ti (nld1lIUI.iqllhl !Ii UPl'fd\(' ell )'). \dlldi L'-> !ll\' ITdC!(l! . Ilw residence time can Lw be \-aricd from j () Ins 10 j S. dicLJ!t'd h) (l cksirc to miL\:imi/c productioll gi\l'!l d J]'\('d n\lc!io!l [d!e.l . till' df".\C!()l'.)\ jS dis( lIssnl flll'llln )11 \n-li(l1i I_~L -J.' 1 itljiJ1\'t'l Dlvl'.lj(!j !C. l'lw B-llTl' 111m clis!rihuti()ll nc!\\(}rk 10 dtTompihl1 Ihis is slHm'n in J j.l!f(lfl (OUll(". Symyx Technoio 9125. nJd:. Copyright 200!.s]\{'ly p(lraJJl'l clnicc.ll1.8 Setting the 5tage Figure 1. Symyx Technologies. (hi) i(. hgll!'l' 1. i \\vlJs in.8: Top view of a combinatorial reactor gas-distribution network. ! l1l'sl' n'dC]()rS drc llH'rv!()fi' Llrgc. For \'XdJ)lP!c. In rrH)('f dtdrnetl'i f(.(). t ..9: Pnlyethyil'!h' j('. Copyright 2001. Inc.. Inc used with permission. !!'d(l(1( I'.K .l\ rt'(\['!()L "> <l q-. till' S(It' of p('lroh'ull"l rdllJ ing fTdctOfs ic. is but olle oj' tlw dvsigll challenges lor Ihis small-sc(1/v.nif()rtn H(m di:-lrihutlon 10 cdeh oj' I Ill' n'actor:-..~ing l11c gas 1l()\\Tdll's.\1 !IF' oppnsilc enei of t!ll' spcclrllm..lV.:\ '. H! dl.1(tul tlU!I.V li..UO! )l10PCLl tfw.'> \\t' til'jlll\' ri'\id"jw(' (1Il]('.l-lll h) . jll( !Im\ s\ .. tIll' hdlcll !!llll' dl'll'I'!llI!W\ Ill\' jlrodlll-llof) r(lll' 101 J'f'.!i !nL !-OJ' (!f i\U[(H i.1(t01 1\ ! his PI'OCC. Figure 1. used with permlsslOn.(JUij P\I drHi (.1 I'il I. !()\\ dl'llSi! \ p()l~ vI ll~ Il'llV pul) nwri/dl io]] Figure 1.j-in \uPJlort \\'afcr.

I I I ! I 1. If tbe rate is high. Devices such as heteroJunction bipolar transistors and field-effect transistors are fabricated by a series of steps that involve masking the semiconductor wafer with a coating. if the rate is low. IS. Typically the reactants are hydride gases. The automobiles we drive and the fuels we use to power them depend on products made iu chemical reactors. 1. as we would anticipate for a high-temperature oxidation or pyrolysis reaction. such as disilane (5i2H6) and germane (GeH4). The rate of chemical reaction has units of moles/time-volume and is a function of temperature and composition. Silicon-germanium alloys have higher mobilities than silicon and a lower bandgap enabling faster speeds and lower power. Chemical reaction engineering is the core discipline that designs. a thin film (tens of nanometers to microns) is grown onto 5i(OO 1) wafers. uniform and independent of position. analyzes and creates the processes to convert natural resources into other intermediate chemicals and final products. Batch times and residence times can be as short as milliseconds for high-temperature. genera"Y employs cluster tools that Single-wafer reactors are used permit a wafer to be shuttled between in the manufacturing processthe different reaction/process chaming steps (Figure 1. De\1ce speed and performance are related to the amount of Ge in the alloy. the time needed for a desired production rate is small.10). Courtesy of Applied Materials. reduced-order kinetic model that accounts for surface and gas-phase components.10 Setting the Stage 1. and control of the fjlm compOSition is critical. including the ink on this page and the paper itself. This means the gas phase can he modeled as a CSTR in which gas enters and leaves the reactor in the flow streams and undergoes reaction at the surface of the wafer. Electronic devices and computer chips require a myriad of chemical reaction steps in their manufacture. When the pressure is low enough. This information can be compactly summarized in a fairly small. the top is supported in the horizontal view of one reactor shows the pins on position. tem is used to ensure uniform flow of reactants over the wafer surface. Much is known about the chemistry that takes place during film growth IS.10: Top viewofa single-wafer wafers (currently 300-mm dieVD reactor. As microelectronics manufacturing moves to ever larger Figure 1. and lower cost is driving the current silicon-based devices close to their physical limits. 13J. In a bers without exposfng it to contami~ single-wafer reactor. at pressures on the order of 0. 19.4 Some Examples We live in a chemical world and most of the items you use in daily life result from chemical processes.4 Some Examples 11 volume divided by the feed volumetric flowrate. or as long as days for liquid-phase fermentation reactions in batch reactors. lower power requirements. a gas distribution syswhich a wafer rests during the CVO reaction. 7.4. such as ammonia oxidation and HCN synthesis in flow reactors. and adding layers of metal and dielectrics to build up the . Ill. this quantity has the units of time. and can be thought of as the time required to displace the equivalent of one reactor volume. Similarly. gas·phase reactions.1 to several torr in a process knov"n as chemical vapor deposition (CVD). We are surrounded by and regularly use items that are made from plastics and polymers.1 Chemical Vapor Deposition of Silicon-Germanium Alloy Films device from the wafer surface 118. a long time is needed. implanting dopant ions such as boron or phosphorous. The reactor volume is determined by the reaction rate and amount of product to be manufactured. For silicongermanium alloy-based devices. One solution that permits the traditional silicon-based circuit design to be retained is to use silicon-germanium alloys in place of silicon in the active region of the device. 2. 12. the diffusion lengths of the gas-phase components are long enough that one can assume the gas phase to be well mixed. selectively removing regions of the mask with lithography. and the wafer is heated radiative!y to maintain the entire cross section at a constant temperature. Single-wafer processing ameter wafers are routine). the wafer nation or ambient conditions. Chances are one or more of the articles of clothing you are wearing are made from synthetic fibers. and can be used to predict the The demand for faster microelectronic devices with higher reliability.

the surface. it is difficult to where the symbols (s). undesired side reactions occur.2r6 .s 2Ge(s) + H2(g) SiCs) + GeH(s) Ge(s) = the aromatic character of the molecules. r3 rs Ys + 2r7 .H6(g) + Si2H6 (g) + 6Si(s) ~ 6SiH(s) + 2Sj(b) 6SiH(s) + 2Ge(b) 6SiH(s) + Sj(b) + Ge(b) Alkyl-substituted thiophene 4Si(s) + 2Ge(s) ~ 5Si(s) + Ge(s) Alkyl·substituted dibenzothiophene GeH4(g) + 4Ge(s) GeH4(g) + 3Si(s) + Ge(s) k· . respectively. The production rates of the ulne speCies follow from the stoichiometry of the eight reactions and the reaction rates as we will see in Chapter 2. the film growth rate.4r4 RGefs ) RCeH(s} RSiH(S) ~ RSi(b) RCe(b) RHz 6rj + 6r2 + 6r3 + 2rs . sulfides. The fixed·bed reactors typically operate in the temperature range 525750 K. HDS is an exothermic reaction and if the reactor temperature becomes too high.3rs + 2r6 + rs -2r2 .Sr3 .Y3 . During HDS the carbon-sulfur bonds are broken and the hydrogen replaces the sulfur and H2S is formed. and the Si/Ge film alloy composition. The total sulfur content of crude oil can easily be several percent and it must be reduced to several parts per million (ppm)_ This sulfur is removed by treating the petroleum feed with hydrogen at high pressures in a fixed·bed or trickle·bed cat· alytic reactor. Multi· pIe beds operating in series and at different temperatures.12 presents a simplihed process diagram.11: Representative organosulfur compounds illustrating Benzonaphthothiophene 25iH(s) 2':<: 2GeH(s) SiB(s) + .4r2 . such as the hydrogenation of the unsaturated bonds and hydrogenolysis of the C-C bonds. when the alkyl substitution position is near the S atom.2 4GeH(s) + Ge(b) 2GeH(s) + 25iH(s) + Si(b) 25j(s) + H2(g) R Figure 1. The feed is heated and mixed with hydrogen in the inlet to a fixed·bed catalytic reactor..2 Hydrodesulfurizat.I<: ks k.12 Setting the Stage 1.4. The production rates enable us to follow the changing composition of the gas phase. The reactions are Si2 H6(g) + Si.4 Some Examples 13 film growth rate [9 j. (b) and (g) refer to the wafer surface. the wafer bulk. One such kinetic model consists of eight reactions among nine species.on Crude oil contains a number of organosulfur compounds (RS). and the gas phase.11. with inter- .rs = 4r4 + 2rs 2r 7 + Ys = 2rl + r] + Ys = 2rz + r3 + r4 = r6 + r7 some of which arC shown in Figure 1. 1. thiophenes and alkyl· substituted thiophenes [8].Y2 RGeH4 = -Y4 - desulfurize the molecule. indud· ing Rsils ) ~ = -6Yj . The sulfur must be removed from the petroleum feeds to prevent it from pOisoning precious metal catalysts that are used in many refinery processes and to meet the re· strictions for sulfur content in fuels. and are given hy RSi2Ho = -YI . and hydrogen pressure range 35-100 atm. Figure 1. the process is referred to as hydrodesulfurization (HDS) [3].

LLC.14 can hold up to 60. Optimal reactor operation and design are two methods to maximize sulfur removal. considerable research effort is focused on finding new catalysts to improve activity. bed cooling. Polyethylene encompasses a family of scmicrystallinc polymers with ethylene as the major building block 161. with more than 33 billion pounds of the resin produced in the Unlted States in 2000 11]. Within each section containing catalyst.13 and 1. and linear-low-density polyethylene (LLDPE).11. and in addition to hydrogenation of the C-5 bonds. Separate reactors in series permit the reactor sequence to be changed as catalyst activity degrades with time. trays to support the catalyst particles and distribute the flows uniformly across the reactor cross section are shown. and to adapt to the ever mOre stringent emiSSion standards 110]. Olefin Polymerization Polyethylene is a major commodity plastic. In addition. high-density polyethylene (HDPE). After cooling the outflow of one of the beds.000 kg of catalyst. Describing the rate of these kinds of complex reactions on catalyst surfaces is covered in Chapter 5. I.12: Simplified hydrodesulfurization process diagram. selectivity and durability. where we show the rate of the HDS reaction can be expressed by gas Quench gas Quench gas Quench Outlet Figure 1. the reactor temperature can be controlled by the heat removal. the organosulfur molecules are highly unsaturated. three separate fixed-bed catalytic reactors in series. The organosulfur molecule must adsorb and orient itself so that the suifur atom is over an active site. The reaction rate is influenced by the relative amounts of reactants (adsorbed hydrogen and organosulfur molecule) and the reactivity of the catalyst. Figure 1. Adapted from McCulloch [8].14 shows an expanded view of the internal construction of a fixed-bed reactor. In this way. reactor 1 is used as a guard bed in which the metals are removed to not poison the HDS catalyst. such as an exposed molybdenum atom on a molybdenum disulfide catalyst. The catalysts have different types of sites. Multi-bed reactors are generally employed with exothermic reactions.14 present two views of a multi-bed fixed-bed reactor in which the beds are built into one vessel. As can be seen from Figure 1. Hydrogen must adsorb dissociative!y. the reaction is allowed to proceed in the next section. one wants to avoid saluration of all the C=C double bonds.4 Some Examples /5 Heat exchangers and furnaces Figure 1. The resins are loosely grouped into three classes: low-density polyethylene (LDPE). which complicates the kinetic model 110]. The catalytic reaction chemistry is quite complex. Figures 1.3 reaction. quenching is necessary to limit the temperature this increase for exothermic Adapted from UOP 1. reactors 2 and 3 permit control over temperature and sulfur removal. is one way to control the temperature and to ensure the complete removal of the sulfur. These concepts are discussed in Chapter 6.14 Setting the Stage /. and the fixed-bed sections are operated adiabatically.13: Cross-sectional view of a commercial HDS reactor containing four fixed- bed reactors with interstage cooling/quenching.4.DPE is a homopolymer of ethylene with side-chain branching at a frequency . the reactor is modeled as a fixed bed. Commercial HDS reactors such as shown in Figure 1.

000 psi and temperature range of Gas recycle 200-300°C. By changing the density. metal and electrical wires. from 10-30 atm. Ethylene. VOCh) on a support (silica.5-2. Courtesy of UOP LLC of 2-50 per 1000 carbons in the chain. Polyethylene is used in films for hags and packaging.. and. and are removed after they settle in a side leg of the loop. The polymerization process begins with an initiation reaction between ethylene and an initiator to form a primary radicaL The primary radical undergoes step-wise chain growth (propagation) until it is terminated by a variety of mechanisms that stop chain growth. Its density is in the range 0. Inc. Gas-phase polymerization also is used for HDPE in the process represented schematically in Figure 1.14: Expanded view of a multi-bed HDS reactor showing the gas distribution tray at the top and the catalyst support tray in the middle. titanium or hafnium containing cyc!opentadienyl rings) and an organoaluminum cocatalyst.940 g/cm 3. the chemical resistance and mechanical properties can be tallored to a variety of uses.915-0. molecular weight.15: Simplified polymerization control is accomplished by process. catalyst and diluent are fed continuously. The polymerization reaction itself is exothermic and the process is operated to prevent excessive temperature Increases. and operate in the pressure range of 30. carhon black) and an organoaluminum cocatalyst (A1(C2HSl3). and a much lower density of side branches than LDPE. and branching-group frequency and type. 6J. ranging from garbage and trash bags to whole blood storage. Copyright ene converted within the re©1996 [6].9 displays a commercial LDPE tubular reactor. containers. This material is used by permiSlimitIng the amount of ethylsion of John Wiley & Sons.DPE reactors are either stirred autoclaves or tubular reactors. HDPE requires a catalyst [22. which acts as the polymerization medium. a copolymer with butene or hexene. these particles reside in the reactor loop for 0. graphite. 17.4 Some Examples 17 Figure 1.940 g/cm3 HDPE can be a homopolymer or.510 branches per 1000 carbons in the chain. and bas a density greater than 0. High-density polyethylene is produced at lower pressures. Low-density polyethyleue is produced in a free-radical initiated. and piping and tubing.5 hr. The l.0 m in diameter and 200 m in total length. more commonly. highpressure process. One such slurry process involves circulating the reaction mixture at 5~ 12 m/s around a loop that can range from 0. MgCIz. Figure 1. Polymer particles form and are removed contin· uously.16 Setting the Stage 1. 15. HDPE processes can employ a slurry of catalyst. Ethylene conversion can reach 95-98%.5-1. a metallocene (a complex of zirconium.000-45. these reactors can range in diameter from 1. Typical tubular reactors have diameters of 15-30 in and lengths of 17-25 ft. The process temperatures are high enough that the polyethylene is in the melt state. Free-radical polymerization mechanisms and kinetics are discussed in Chapter S.5-5 m. TiCh. and the high flowrate ensures turbulent flow within the loop [6J. actor and/or removing heat through the reactor walls. Various catalysts have been used including: chromium oxide supported on an inert porous substrate such as silica. 0. The cylindrical reactor can be . coatings for paper. LLDPE is a copolymer of ethylene and small amounts of (X-olefins. The heat of reaction is removed by a cooling jacket that surrounds the reactor. a Ziegler catalyst consisting of titanium or vanadium compounds (TiCI4. The temperature Figure 1. polymer and diluent. and the presCatalyst sures are high enough to form a supercritical mixture of ethylene and polyethylene.

The understanding and modeling of these biochemical events is an area of current research activity [14.3) amino acids cc~"JA (1. Reaction 1.16: 1. When we change to this context.6) Figure 1.16: The chemical events comprising the reproduction cycle of the hepatitis B virus.4.16. down to less than a few hundred molecules. p.2) (1. rcDNA+ These reactions correspond to the following steps in Figure 1. 4.4 Some Examples 19 25-30 ft in diameter and have a five-to-one length-!O-diameter ratio. Reaction 1. In Chapter 9 we explore estimating the rate constants that appear in this virus model given the kinds of laboratory measurements that are available. 5.4 and 1. 16].4) (1. Reaction 1. but may prove useful to explain potential loss of active cccDNA.1) (1.2 accounts for Step 12.5) (1.4 Hepatitis B Virus Modeling As a final example. High linear velocities serve to fluidize the catalyst particles that form.6 accounts for Steps 13-15. 1. Reactions 1. Reaction 1. which shows some of the biochemical reaction events that occur in a single cell during the reproduction cycle of the hepatitis B virus [4. 9-11. Consider Figure 1. Catalyst is added continuously and polymer particles are removed once they reach a particular size (about 500-1000 /Jm). The following is a simplified but useful model of part of this reproductive system nucleotides nucleotides + rcDNA cccDNA envelope envelope cceDNA ~ rcDNA cccDNA envelope degraded secreted or degraded secreted virus (1.3 accounts for Steps 5-7. The reactor typically operates at 220-370 psig and 160-205°F. In Chapter 4 we use this simple model to make quantitative predictions about the evolution of the viral speCies concentrations. fluidization facilitates removing the heat of reaction. 21. These same reactors are used for LLDPE. 3. if we also model the cell population. 20. The choice of catalyst and the presence of inert gases and comonomers are used to regulate the polymer resin properties and molecular weight. .16.5 are not present in Figure 1.1 accounts for Steps 5. The chemical reaction modeling principles remain valid. The reactor diameter is larger at the top. We also show how to model systems that have small concentrations of species. the human liver.7671. Courtesy of ASM Press [4J. we wish to display the "ide scope of chemical reaction modeling principles. which lowers the linear velocity and acts to disengage the polymer particles from the unreacted gases.18 Setting the Stage 1. 2. the "chemical reactor" of interest becomes the living cell or.

We should not underestimate the complexity of some of the systems of interest to chemical engineers. isothermal reactions at this point. We illustrate the use of these assumptions in developing kinetic expressions. In the catalyst particle. The reactor behavior can become complex and interesting.5 An Overview of the Text Chapter 2. Chapter 6. and species are present in extremely different concentrations. The main goal of this text is to build the skill set with which chemical engineers apply reaction modeling tools to understand chemically reacting systems. if we wish to increase our understanding of a chemically reacting system. we must consider the combined reaction-diffusion problem to be able to evaluate the temperature and the species con- 1. we make free use of matrices and linear algebra to summarize compactly the reaction stoichiometry. We derive and illustrate the use of tbe basiC material balances for single and multiple reactions using molar concentration or molar flow as the only dependent variable. so we delay a comprehensive study of reaction rates and reaction rate expressions until Chapter 5. We feel it works best to introduce the chemical reactor as soon as possible in Chapter 4. and we explore some of the complex behavior such as multiple steady states and Sustained oscillations. The use of numerical optimization methods is illustrated for solving complex reaction equilibrium problems involving many reactions. This caution is perhaps especially true for biological systems. We treat both homogeneous and heterogeneous reactions. continuous-stirred-tank. Because we cannot include all the details.20 Setting the Stage 1. The chapter concludes with a discussion of mechanisms for reactions occurring at the surfaces of solid catalysts. but knowl· edge of the equilibrium state of the system allows one to define limits of reactor performance and identify operating conditions to realize desired production rates. ~ Chapter 5. In Chapter 6 we develop the energy balance for the chemical reactor. The coupling of the material and energy balances also provides some surprises. We explore such issues as how to remove or add heat to the system to overcome equilibrium limitations on conversion. We consider homogeneous. single-phase. skillful model building is often an indispensable part of the overall investigation. In Chapter 5 we provide a simple. Chapter 4. We also introduce the use of stochastic simulation to simulate reaction kinet~in well-mixed reactors. Since most processes involve multiple chemical reactions. In Chapter 4 we develop the material balances for the three reactor types: batch (and semi-batch). and develop the accounting system for tracking the change in the reaction extent and the species concentration. and how to integrate a process to use the heat of reaction to preheat the feed entering the reactor. and it is possible to make naive use of modeling approaches. We define chemical reaction rate and species production rate. Chapter 7. The conditions for equilibrium are developed using the Gibbs energy and the chemical potential or species activity. Next. . and plug-flow reactors. On the other hand.5 An Overview of the Text 21 Tailoring a model to make successful quantitative predictions of a system of interest is still more of an art than a science. we briefly review the important facts concerning the equilibrium state of a system undergoing chemical reaction. Chapter 3. in Chapter 3. Because chemical reactions occur at extremely different rates. our models are always incomplete. We begin in Chapter 2 by discussing stoichiometry or the quantitative rela· tionship between the different chemical species undergoing chemical reaction. theoretical framework for predicting the rate of a chemical reaction and for relating detailed statements of the reaction chemistry to reaction rates. Chapter 7 considers the industrially important case of heterogeneous reactions taking place in solid catalyst particles. we can often reduce the complexity of the reaction mechanism without changing the main features of the modeL We develop the two main procedures for reducing the reaction mechanism complexity: the reaction equilibrium assumption. The remaining text is divided into eight chapters. especially when feedback mechanisms or autocatalytic steps are involved. Simple models often can explain complex system behavior. We also briefly review phase equilibrium so that we are prepared for multi-phase reactions. Most chemical processes do not reach equilibrium. such as free-radical polymerization kinetics. We also consider the volume change upon reaction and show how to employ the equation of state to complement the species material balances in situations in which the fluid density is not constant. and the quasi-steady-state assumption. and produce models with little connection to reality and little predictive value. The combined material and energy balances provide us with a rich description of many chemically reacting systems. The condition for chemical and phase equilibrium is generally stated as the minimization of an appropriate energy function or maximization of entropy.

computational fluid dynamics software is evolving to the point that this approach is becoming tractable for many problems of interest. . The fundamental reactor modeling principles covered in Chapters 2-8 provide the framework in which we think about chemical reactors. The material and energy balances for reaction-diffusion in catalyst particles result in fairly challenging boundary-value problems. We introduce the reactor residence time distribution and describe the general issues of mixing in chemical reactors. Methods for solving ordinary differential equations and differential-algebraic equations are the real workhorses of reaction engineering. they are required in solving the material and energy balances. estimating model parameters from data. which is not usually addressed in a systematic manner in introductory treatments of reactor analysis and design. we used optimization for three main purposes: finding the equilibrium state for complex situations involving multiple reactions and multiple phases. we require values for the model parameters. Second. therefore. We describe the limitations of these approximate mixing models. Append. accurate modeling requires one to solve for the complete velocity proflle in the reactor. The tradition in introductory reactor design courses is to neglect a careful treatment of the fluid flow and use the simplified. We understand whicb phenomena cause which observed reactor behaviors. In this text. First. In some cases.x A: Computational Methods. they help in the problem of estimating model parameters from data. To make these models useful in stan· dard industrial practice. and we discuss numerical methods for solving them. Chapter 8. We also make extensive use of methods to compute the sensitivities of solutions of differential equations to the parameters appearing in the modeL These sensitivities are needed for two purposes. But when we want to make quantitative predictions of reactor performance. Altbough that topic is beyond the scope of this text.22 Setting the Stage /. ideal reactors for modeling. in Appendix A we summarize the numerical methods that have been necessary to solve the reactor models presented in the text. and solving reactor design problems.5 An Overview of the Text 23 centrations. We show how to couple the fluid balances to the catalyst particle balances in order to predict the overall fixed-bed reactor behavior. Packing the solid catalyst particles in a tube and passing a fluid stream over them produces the fixed-bed reactor. Optimization methods are used extensively in science and engineering. Linear algebra and matrices are introduced to handle reaction stOichiometry involving multiple reactions. we must be able to conveniently determine or estimate these parameters from experimental data collected on the system of interest. In Chapter 8 we explore what to do when the reactor flow pattern is not well represented by these ideal mixing assumptions. Finally. Chapter 9 covers tbis important topic of parameter estimation. and which design variables should be changed if we wish to alter the reactor performance. they provide one means for solving the challenging boundaryvalue problems that arise with simultaneous reaction and diffusion in catalyst pellets. It is a simple fact that most of the parameters needed for the chemistries and reactor configurations of interest are not available in the literature. Chapter 9.

Summers. Lett. [6] Y. Bramblett. [8] D. T. I. Ekerdt. fifth edition. Smith. Matoaka. Bioi. TechnolA. Neutral. 119J Y. Washington. Viral. Russell and W. New Jersey. ]. 1999. andJ. T. [2] T. New York. Yu. Reno. New Jersey. McGraw-Hill Publishing Company. 1983. Hydrotreating Technology for Pollution Control Catalysts.. Upper Saddle River. Smith. Morphogenetic and regulatory effects of mutations in the envelope proteins of an avian hepadnavirus. Streettnan and S. Hepadnavirus envelope proteins regnlate covalently closed circular DNA amplification. 8(1):61-67. American Society for Microbiology. J. Microb. Vac. Sci. Sinnlah. App/. 1993. Oxford University Ptess. D. Kissin and L. Karasawa. 1. and Processes. Rev. Holbert. L. November 200l. Leach. G. London. Shennan. Viral. 2000. fourth edition. [211 J. 2000. Lee. Greene. Rev. Phys. 1171 M. Fundamental Of Microelectronics Processing. A surface kinetics model for the growth of ShxGe x films from hydrides. 1996. S. an Introduction. Breiland. 115J K. E..1990. Horwich. McCnlloch. Kinetics of hydrogen desorption from germanium-covered Si(100). editor.r. S.]. 2001. Friedrich. M. Polyethylene. Suda. Madison. Srivastava. and M. Academic Press. Chianelli. J. University of Wisconsin. Upper Saddle River. Lubben. P. 1991. S. R. W.. Bansleben. D.1990.. Deal. R. Yin. L. Blackie Academic and Professional. 64(1):51-68. Loeb. N.. [71 H. Connor. Griffin. L.C. 79(26):4351. P. M. M. Adsorption and thermal dissociation of disilane on Si(100)2xl. single-component nlckel (Il) polyolefin catalysts that tolerate heteroatoms. 1996. H. ASM Press. . Eng. Summers. R. [S] S. Inc. and Control. 369:51-68. Kroschwitz. Enquist. Weinberg. Silicon VISI Technology Fundamentals. Occelli and R. Applied IndusMal CatalysIs. 1996. Wisconsin. Si(OOI)2x 1 gas-source molecular-beam epitaxy from Si2H6: Growth kinetics and boron doping. 1989. Lu.. Mason. - 19] [10) 1111 (12) 113) [141 c. Catalysis. 13] R. New mechanisms for hydrogen desorption from covalent surfaces: The monohydride phase on Si(100). Henderson. Q. Phys. J. March 2000. J. M. M.1994. Pathogenesis.o/. Hepatitis B virus biolo". New York. John Wiley & Sons. Janda. c. Mol. Solid State Electronics_ Ptentice Hall. Catalytic hydrotreating in petroleum refining. Yates. [4] S. Modelling and Control of Silicon and Germanium Thin Film Chemical Vapor Deposition. Lewis. New York. G. 1997. M.. Deposition of silane on SiO 1l)-(7x7) and Si(100)-(2xl) surfaces below 500 'c. and K. pages 702-784. D. M. App!. Stevens. Farrauto and C. M. and D. Phys.}. Pebsworth. [18J B. Flint. L. V. 76(3):1884-1888. Practice and Modeling. 287:460-462. Marcell Dekker. M. Hasan.. [22] T. Annual MehE Meeting. New York. Grubbs. W. Bartholomew. W. 2000. D. PhD Dissertation. J. Chern. Gates. and J. [20J J. Prentice Hall. P. Kirk-Othmer Encyclopedia of Chemical Technology. and P.200l.. H. Science. 73(7):3525-3530. K. In J. Seeger and W. [16J R. Racaniello. and A. Skalka. Beach. volume one. 2000.]. A. V. pages 69-121.Jo. B. and P. Polymer Chemistry. J.. M. D. UK. and]. Greenlief. In B. ]. 1990. 1990. A. News. Surface Science. S. F. M. Principles of Virology: Molecular Biology. Chern. R. New York. C. Huang. 92(5):3144-3153. H. Krug. volume 17. Younkin. E. Russell and J. Greene. Computational and experimental analysis of dnck hepatitis B virus intracellular kinetics. 64(6):2819-2824. editor. Nevada. 62(5):567-570.Bibliography 25 Bibliography [1] Facts and figures for the chemical industry. Banerjee. and A. Middlebrooks. 65(3):1310-1317. Plummer. Fundamentals of Industrial Catalytic Processes. N. Phys.

lWebster's New World College Dictionary.1 Introduction Stoichiometry is defined as the determination of the proportions in which chemical elements combine or are produced and the weight relations in any chemical reaction. ExerCises are provided in Section 2. Section 2. The reactor analYSis book by Arls fl] influenced several sections of this chapter.6 provides a summary of the important concepts and relationships that may prove useful as a study guide or quick reference. The last several exercises introduce new material and show how the stoichiometry fundamentals presented in this chapter lead into other interesting topiCS. fourth edition. 2000.3 introduces the concept of linearly independent reactions and discusses the implications of mass conservation on the stoichiometric matrix_ Section 2.2 The Stoichiometry of Reactions 2. .7 for reinforcing the concepts and to further develop one's understanding. 1 In this chapter we explore and develop these quantitative relations between the different chemical species undergoing chemical reaction. References for further study are provided at the end of the chapter.4 defines the rates of reactions and rates of production of chemical species due to the reactions. Section 2. The next section establishes the accounting procedure for tracking chemical change and introduces the stoichiometric matrix_ Section 2.5 explores the issues involved in calculating production rates given reaction rates and vice versa_ We also formulate and solve the least-squares problem of extracting best estimates of reaction rates given production rate measurements containing errors.

and A6 = C02. . H.5) Reactions 2. The overall stoichiometry of this reaction is (2. I The matrix appearing in Equation 2. we can choose AI = H. and N02. and A3 = N02.1) The convention that we follow is reactants appear on the left-hand side of the chemical reaction symbol. and products appear on the rightband side.6) Again notice the sign convention that products have positive coefficients and reactants have negative coefficients in Equations 2. we can choose AI = NO. OH. H2. 02.7) A6 Reactions 2. A3 = OH. and writing A to denote the column vector of the Aj. our final summary of tbe water gas shift reaction appears as vA =0 2Boldfac{' letters provide a mnemonic device. It consists of a single reaction among three species. In this example. A3 + As ~ ~ . CO. These reactions have been proposed to describe a silicon chemical vapor deposition process. represent the jth species taking part in a reaction. For the first example. The stoichiometry of this reaction is 2NO + 02 = 2NOz I I I I In order to organize the way we discuss chemical reactions. In the water gas shift example. and CO 2 .. NO. A4 = H20. the following notation is convenient. and in determining the H2 content in the feed for fuel cells. .2 Chemical Reactions and Stoichiometry 29 2. Using the Aj notation.2) -A4 . consisting of 20 reactions among 14 species.6 2 Equations 2. In the first example. =. and is appropriately known as the stoichiometric matrix. A2 = H2. (2.2 Chemical Reactions and Stoichiometry We shall consider three primary examples as a means of illustrating the main concepts in this chapter.6. As = CO. (2. It also is known that the follOwing two reactions are needed to describe what is happening at the molecular level. it consists of three reactions among six chemical species.4) I i I I I l 0 1 I 1 1 0 0 I -1 -1 1 0 I 0 -1 1 0 1 1 A2 A3 A4 As =l~l (2. which is an important step in the production of microelectronic materials.As + A6 + A2 = 0 -A4 -AI +A2 +A3 = 0 -AJ As +A6 +AI = 0 (2.4 comprise a simple reaction network. Using the rules of matrix multiplication. The second example is known as the water gas shift reaction. consider two molecules of nitric oxide and one molecule of oxygen reacting to form two molecules of nitrogen dioxide.. .28 The Stoichiometry of Reactions 2. Giving the stoichiometric matrix the symbol v. The third example illustrates the complexity of common industrial reactions of interest.6 as AI Tbe rate of this reaction is important in determining the CO/C02 ratio in exhaust gases from internal comhustion engines. H20 + H OH + CO = Hz + OH C02 + H (2. there is a single chemical reaction and three different chemical species taking part in the reaction.3) = (2.7 provides an efficient description of the stoicbiometry for the reaction network. and it illustrates the case of mUltiple reactions. A j.- A6 + A2 A2 + A3 A6 + A. There are three chemical reactions and six different chemical species taking part in tbe three reactions.5 suppress the identities of the species for compactness. we can express the water gas shift reactions as I A4 + As A4 + A. The second example is the water gas shift reaction.2-2. The first example is one of the reactions responsihle for smog formation in the atmosphere. one can express Equations 2.j = 1. We can further compress the description by moving all of the variables to the right-hand side of the chemical reaction symbol and replacing it with an equality sign. Let the symbol. A2 = 02.6 now resemble a set of three linear algehraic equations and motivates the nse of matrices. H20.

A3 = NOz.12) = C02 + H2 Example 2. give Reactions 2. in this casc. After piling up this much abstraction to describe what started out as a simple set of three reactions. Example 2.8 gives i ~ =i -~ -~J o -1 0 -1 1 I (2. More preCisely.11) (2. Notice we have written the original third reaction first and the original first reaction third.\4 ~ A2 A6 A3 .2: Columns of v Since we are free to assign chemical species to the Aj in any order we choose. and the index j runs from 1 to n" the total number of species in the network We say that v is an 11y x 11s matrix. We can assign the species to the A as follows: Al = NO.30 The Stoichiometry of Reactions i I 2. Example 2. Instead of using Al = Hand A6 = C02.\2 As ~ + . and A6 = H. A2 = 02. A:J + + + As Al ~ A6 + + Al A4 . we can ascribe the following elements of a third stoichiometric matrix. The reaction can then be written as It is clear from examining Equations 2. we express it as v = [-2 -1 2] o = C02 + H H20 + H = H2 + OH OH+CO H20 + CO (2. ~ ~ A'I + A2 A'2 + A3 A'1 + A'6 (2. let us work a few examples to reinforce the concept of the stoichiometric matrix. We now explore what happens if instead of expressing the water gas shift reaction as it appears in Reactions 2. consider what happens if we change the order of the species in the water gas shift example. what is the stoichiometric matrix if Al is chosen to be CO. A:1 = (lB. 2NO + 02 A4 = H20.7 that switching the identities of species one and six in the A vector has necessitated switching the first and sixth columns in the stoichiometric matrix.8) Extracting the modified v matrix from Equations 2. What is the impact of this change on the v matrix? I Solution Using the original ordering of the A vector.5 are then modified to ~ A'4 + A'5 A'4 + A'6 A'3 + A:" .2 Chemical Reactions and Stoichiometry 31 The element vi} in the stoichiometric matrix is the stoichiometric coefficient for the jth species in the ith reaction. A2 ~ I-Iz. Therefore one can make the connection between the columns of v and the species taking part in the reactions.9 and 2. The index i runs from 1 to 11" the total number of reactions in the network.2-2. As = CO.1: Stoichiometric matrix for a single reaction Find the stoichiometric matrix for the nitric oxide example. 0 The stoichiometric matrix for a single reaction is a row vector. A'I ~ C02.3: Rows of v Just as we are free to assign the species to the A vector in any order we choose.9) = 2N02 Solution The nitric oxide example consists of one reaction and three species. we are also free to express the reactions in any order we choose.4.10) (2. the jth column of the v matrix supplies the stoichiometric numbers of the jth species in all of the reactions. and A6 is chosen to be H'l Solution Switching the identities of the first and sixth species in the A vector gives us the following modified vector.

ns ~ 14. 5i2H2.1 S.:= simpler nitric oxide and water gas shift chemistries presented previously. SiH4 SiH4 SiH4 + SHh :. We can therefore make the connection between the rows of v and the reactions 3 The ith row of the stoichiometric matrix contains the stoichiometric numbers of all species in the ith reaction.H2 5 iHz + Hz S13::= SizH2 + Sb SizH. Polymerizations and long-chain-producing reactions consist of thousands of species and associated reactions.SiH z SiBj + Hz + SbH2 Hz + SbH4 SlzH4 Si2H6 SiHj S.: : : : . The following chemistry has been proposed to describe a silicon chemical vapor deposition (CYD) reaction.. We now introduce the third example. one may permute the columns and rows into any order and maintain a valid stoichiometric matrix. + SiH = c.13) As one might expect.: : : : .::::: 5bHz H SbH3 2Si. We then look through all of the reactions and identify the different species taking part. Example 2. Developing and understanding that procedure is the topic of the next several sections. H2 + SieHj Sh -:. 5i2. which is an important process in the production of microelectronic materials. A matrix "ith many zero entries is called sparse. for example. Obviously the stoichiometry of these complex problems is intractable if we do not develop a systematic. Since there is no reason to prefer one ordering of species and reactions over another. SiH4. + SiH Sill4 + Si + Si H2 Si2Hs 2SiH2 SiB2 rb Hz + Hz +Si2Hz + Si + SiB = = ~ (2.Ii 5 = = = = = SiH 4 + SiH] -:.14. The only difference is that Reactions 2. Notice that for this example v is a 20 x 14 matrix. After wtiting this out we notice that there are 14 different species. SiH. Indeed the principles for analYZing a CVD reactor with this chemistry are exactly the same as the principles for analyzing the I :t Another mnemonic. and it contains many zero entries.:.7. I . SiH3. Reactions 2. Practice filling out a few rows of the v matrix and check it with the values given in Equation 2.. 5i. Si2H4.14) Sillj + SiB2 Sh:. In fact. Solution The first thing we notice is that there are 20 reactions or nr ~ 20. All of the reactions in the CVD chemistry.14 and find the stoichiometric coefficients of each species in each reaction. exchanging the orders of the first and third reactions causes us to exchange the first and third rows in the v matrix as evidenced by comparing the matrices in Equations 2. More will be said about this issue in the discussion of mechanisms in Chapter 5. Si2H3. Si2HS + SiH4 + SiH SiH. SiJ. SiH2.32 The Stoichiometry of Reactions 2. which is a more complicated reaction network.= = Si = ::= :.14 are a Simplified version of 120 reactions that were originally postulated for this reaction network [3J. are unimolecular or bimolecular. Si2Hs.2 Chemical Reactions and Stoichiometry 33 ~ -~ -~ -~ ~J 1 0 -1 -I I (2. H2. Do not forget the convention that species appearing as products in a given reaction have positive coefficients and those appearing as reactants have negative coefficients. H3 Si. With an A chosen. automated procedure.14 are quite simple compared to many mechanisms that have been proposed for combustion problems in which it is not uucommon to have severalliundred reactions.14 are complex enough that we use a computer to keep track of the algebra for us. Chemical engineers should also bear in mind that Reactions 2. 5jHz + SiH] + SiZH6 Hz H Sill. The large number of zeros simply reflects the physical fact that very few molecules can take part in a particular reaction.=:: Si. 0 From the previous two examples it is clear that one could develop a large number of stoichiometric matrices to describe the same set of chemical reactions.:j + SiB4 + Hz:::::= H SiH4 +. The student should not be dismayed by the complexity of this reaction network..13 and 2. Si 2 H6.4: Stoichiometric matrix for eVD chemistry Determine the stoichiometric matrix corresponding to Reactions 2. it is a simple matter to look through Reactions 2. A possible assigmnent to the A vector is: H.

Mo.16) = OH + CO = . we know or MH2 MOH MH. and A6 = C0 2. M. Another way to state conservation of mass involves molecular weights of the species. For the first reaction.15) For the water gas shift reaction. using the ordering of the species. For example in the nitric oxide reaction.3 Independent Reactions 3S Mo. In the nitric oxide reaction. H20 + CO = C02 + H2. not nuclear) reactions. + 2MNO. for multiple reactions vM~O in which M.I is the molecular weight of species j.. the number of molecules is not conserved in general. let us again consider the water gas shift reaction H20 + CO H20 + H = C02 + Hz Hz + OH C02 + H (2.. A4 = H20.7 multiplied by M. AI = H. The mass. is conserved in chemical (Le. 2NO + 02 = 2N02. M = [MH MH2 MOH MH20 Mea Mco. Solution In a chemical reaction. so the mass is conserved. ~ [-2 -1 2lr~~ J=o l MNOz If we put the molecular weights of tbe species in a vector. however. As = CO. Tbe second and third reactions simply fill out the second and third rows of v so that again. = 2MNO. three reactant molecules react to form two product molecules. Equivalently 0 1 I 0 0 0 0 -1 -1 J v~ 0 J 0 0 0 0 0 0 0 -1 0 0 0 0 0 0 a 0 0 0 0 -J -1 -1 -J -J -1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 -1 0 -1 -I 0 -J 0 0 -1 -1 0 0 0 0 0 0 0 0 -1 0 0 0 0 0 0 -J 0 1 0 0 -J -J 1 0 0 1 0 2 J -I 0 -I 0 I J -I -J -1 1 -1 0 -1 0 -J 0 0 0 0 0 0 0 0 0 0 0 0 I 0 0 0 0 0 a 0 0 0 J -1 -1 -1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 -I 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 I 0 0 1 0 0 1 1 0 0 0 -1 0 1 0 0 0 -1 0 0 0 0 -1 1 0 0 0 0 -1 0 0 -1 1 0 2 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 J 0 0 0 0 0 0 0 0 0 0 0 0 0 0 -2MNO .5: Conservation of mass Show that conservation of mass in a chemical reaction can be stated as vM~O Mco M co. which is the first row of v in Equation 2. Az = H2. 0 0 0 0 0 0 (2.O -J 0 [ o 1 0 -1 -1 1 ] = 0 Example 2. A3 ~ OH. the molecular weights of reactants and products are related by 2MNO + o 2. It is clear in the above example that the atoms (N and 0) are conserved.34 The Stoichiometry of Reactions 2.3 Independent Reactions To motivate the discussion of independence of chemical reactions. The transpose of the matrix means to exchange the rows for colunms and vice versa. conservation of mass for this single reaction can be written as. r I 0 0 0 -J 0 0 I 0 0 0 0 0 0 0 0 0 0 in which the superscript T means the transpose of the matrix.

16 are independent. is not required to follow the subsequent development. We now consider the stoichiometric matrix for the water gas shift reaction presented in Equation 2. we .17 can equal Reaction 2. It is not necessary to eliminate extra reactions and work with the smallest set. so the first and second reactions could be chosen as an independent set. the concept is important and is examined further. The rank of a matrix is defined as the number of linearly independent rows (or equivalently. is equal to the rank of v. For this example then. columns) in the matrix. The situation is not always this simple as we will see from the chemical vapor deposition chemistry.19) 1 -1 If we deleted the first reaction from the network. because no multiple of Reaction 2. Therefore. then the set of reactions is not independent. the third reaction in Reaction 2. If you are familiar with linear algebra. likewise there is no way to produce H2 or H20 from only Reaction 2.17. Finally. Example 2. any two of the reactions comprise an independent set. If we can.16. H20 + III CO So tbe first and third reactions could be chosen as the independent set of two reactions. although helpful. but on how we can exploit the results of the algorithm to analyze sets of chemical reactions. This discussion is not meant to imply that there is something wrong with the first reaction in Reactions 2. Before making the problem more complicated. Can we express the first reaction as a linear combination of the second and third reactions? By linear combination we mean multiplying a reaction by a number and adding it to the other reactions. the number of linearly independent reactions in a network. we can again ask the question whether or not it can be written as a linear combination of the first and third reactions.36 The Stoichiometry of Reactions 2.17) (2. we explore the question of whether or not the three reactions listed in Reactions 2. In any case. Gaussian elimination with partial pivoting. The answer is yes because the second reaction is the first reaction minus the third reaction. H2 0 --- We can make an important mathematical connection to the preceding physical argnments.16 is equal to the first reaction minus the second reaction. ni. so the set of three reactions is not independent. would the remaiulng two reactions be independent? H20 + H OH+CO = H2 + OH H = C02 + The answer is now yes. There is no way to produce CO or C02 from only Reaction 2.18. The linear independence of the reactions in a reaction network is equivalent to the linear independence of the rows in the corresponding stoichiometric matrix. The focus of our attention is not on the algorithm. + + ! OH II II H _ + H2 _ + HI H20 + .6: More species than reactions Show that mass conservation implles that any independent set of reactions has more speCies than reactions.7 H2 + • + + • H20 + CO + C02 III vJl -~ (2.3 Independent Reactions 37 The issue of independence centers on the question of whether or not we can express any reaction in the network as a linear combination of the other reactions. Familiarity with these concepts. Indeed if we focus attention on the second reaction. It is clear from inspection that the first reaction is the sum of the second and third reactions. the issue of independence of reactions is obviously related to the rank of the stoichiometric matrix.18. You should consult Strang [7] or another linear algebra text for a lucid explanation of the algorithm. The question of whether or not the ith reaction can be written as a linear combination of the other reactions is the same as the question of whether or not the ith row of the v matrix can be written as a linear combination of the other rows.18) ~ ~ =~ 0 -1 0 -1 o 0 1 1 J (2. There are efficient numerical algorithms available for finding tbe rank of a matrix and a set of linearly independent rows. Before making any of these statements precise.~Iore how to automate the preceding analysis by exploiting the stoichiometric matrix. but it is sometimes preferable.

20 is nonzero.. 4 Because the rows are linearly independent.38 The Stoichiometry of Reactions 2. We start with a set of species. We may use molecules of pure elements (02) rather than atoms (0) to save work as long as tbe atoms themselves do not appear in the species list. Notice that one must consider linearly independent reactions for the statement in the example to be true. If we considered an arbitrary network for example. list formation reactions for every speCies in the list from its elements. First. we conclude there are more columns (species) than rows (independent reactions). This matrixvector mUltiplication can be expressed as a linear combination of the columns of v with the elements of the M vector as the coefficients in the linear combination [ =! -6 -8 -3 -4 =~ ! 1 6 8 Since we have not used independent reactions. Recall v = [-2 -1 2 for the single independent reaction.3 Independent Reactions 39 Solution From Example 2. 2NO + 02 = 2NO. a given set of reactions may be linearly independent. Up to this point. ns > ni and v is a wide matrix (Le. The remaining set is a maximal linearly independent set of reactions for the original species list. Example 2. 40ne could solve for column j of v by movIng the remaining terms to the other side of the equality and dividing by Mj.a/l and not wide. all of the multipHers are CH4 CO Hz 02 1 T nonzero. This set is guaranteed by construction to be a maximal linearly independent set for an enlarged species list. In other words. In this case v= Consider the columns of the v matrix as column vectors.7: Maximal set of reactions for methane oxidation ConSider the oxidation of methane in which the observed species are: A = [C02 H20 6NO + 302 = 6N02 8NO + 402 = 8NO. 4NO + 202 = 4N02 Maximal sets of linearly independent reactions. It is necessary to write these reactions in the order in which the species . In our case. which Is: nonzero. but we want to be sure we have not left out some valid reactions. 0 1 The last equation implies the columns of v are linearly dependent because the molecular weights are nonzero. we know that mass conservation is equivalent to vM=O These reactions arc obviously not independent because they are the same reaction written four times. and we would like to know the largest number of linearly independent valid chemical reactions among these species. v is J. This procedure is perhaps best illustrated by example. not a square or tall matrix). which is a sufficient condition implying the weaker condition of mass conserving. we have started with a set of reactions and investigated constructing subsets of these reactions that are linearly independent. Then eliminate through linear combinations of reactions any new elements that were introduced in the formation reactions and that do not appear in the original species list. Now consider the reverse problem.5. which is in agreement with the example. We first write formation reactions for all species from the elements. which includes the original species plus any new elements introduced in the formation reactions. By valid chemical reactions we mean element conserving. That is possible as long as one of the M j multipliers in Equation 2. The following describes a systematic approach to this problem.

25 with the sum of row 1 and negative of row 3.26) . one has to know the reaction rates. See Exercise 2. which is not in the orig· inal species list. s So the rank of v is four. The stoichiometric matrix above corresponds to the following choice of reactions CH4 + 02 2H2 + 02 C02 + 2Hz Notice we have zeros below the diagonal of the first four rows and columns. As an example.22) (2.4 Reaction Rates and Production Rates in order to describe the change in composition in a reactor. That reduces v to 2C + 0. When we reach the last row. therefore. = v' = 1 0 0 2 [ o 0 ~l 0 2 0 0 ~2 2 ~2 ~4 ~l ~l 1 Notice we have introduced the element carbon. either a zero already exists in the last column or we remove this last row because we have no rows below tbe last with which to zero that element. We can add C to the list and the enlarged species list is • A = [ CO.24) = = because the V2.25) Therefore the maximal linearly independent set for the original species contains three reactions. Proceeding down the rows we replace row 3 by twice row 3 minus row 4. C ]T inspection of the zeros in the first three columns of v' tells us that the rows are independent. so are the entire rows. C does not take place in the second reaction. we would repeat this procedure starting with v.4 Reaction Rates and Production Rates 41 appear in the species list. We can use 02 and H. 2H20 CH4 2CO (2.14 (2. we consider the third reaction in the CYD chemistry. 2. = CO. the last column of zeros removes the C from the species list and we have The stoichiometry of the formation reactions are then summarized by vA in which = 0 vA in which =0 . H. which is why we wrote them in the first place.9 for a proof that there are no other valid linearly independent reactions among these species and elements.21) (2. and. The approach is to replace formation reactions involving C with independent linear combinations of the four reactions that eliminate C from the set.23. Because these portions of the first four rows are independent.23) (2.40 The Stoichiometry of Reactions 2. For example we could replace Reaction 2. + 0. See also Exercise 2.21 and the negative of Reaction 2.7 is zero already. If we now multiply out these equations. C +2H. We leave tbe second row unchanged 5Examjne the locations of the zeros in the first four rows and columns of V. therefore. Equivalently we replace row 1 in Equation 2. to express the formation reactions. Reactions 2. i. because 0 and Hare not in the species list C+ 02 2H. We always add linear combinations of rows below the row on which we are making the zero in order not to disturb the pattern of zeros below the diagonal in the first columns of the matrix..~[! 0 2 0 0 0 0 1 0 0 0 0 2 0 ~2 ~2 ~l v=[g ~l 0 2 0 ~l 0 2 ~2 ~4 0 2 0 ~2 ~l ~l 1 1 ~l 0 ~l 0 ~2 :1 (2.6. the matrix has full rank. and.O CH4 CO H20.21 with the sum of Reaction 2. We now wish to elhnlnate carbon from the species list. = 2H20 2CO + 4H2 = 2CH4 + 02 o If we were going to remove other elements besides C. the reactions. We know by inspection that these reactions are linearly inde· pendent.e.

Fi' nally. If the system is at equilibrium. however. an SizH6 molecule is consumed. But that is a separate issue. which are the usual units for extent and reaction rate in this text. If the reverse event occurs more often than the forward event. RS iH 4 RSiH2' = = ~ y r Notice that we have three production rates. We delay a more complete discussion of elementary reactions and reaction mechanisms nntil Chapter 5. RSlzHp is therefore directly related to the reaction rate.27 that each time the forward reaction event occurs. r. l! is difficult to measure reaction rates directly. We define the reaction extent. SizH6' rate.. then the change in E is zero and the forward and reverse events occur in equal numbers. then the change in E and reaction rate are negative. because there is only a single reaction. which we take up later. it may be difficult to know whether or not a reaction is an elementary reaction.4 Reaction Rates and Production Rotes 43 H H H-Si~~Si-H I I I I H H Figure 2. to keep track of the number of times this reaction event occurs. Notice that if r is positive RSi2Hfi is positive as we expect because Si2H6 is being produced~ Similar arguments lead to relating the other pro· duction rates to the reaction rate. E.27) If the forward event (an SiH4 molecule and an 5iH2 molecule lurn~ ing into an Si2H6 molecule) occurs more often than the reverse event (an ShH6 molecule decomposing into an SiH4 molecule and an 5iH2 molecule). If we . is the net number of reaction events that occur in the time interval 6t. Each time the reverse reaction occurs. one for each species.27 are #!(time·volume). we are laking the reac· lion stoichiometry literally. One can imagine turning SiH4. If one divides by Avogadro's number. It is important to connect the reaction rate to the rate of change of the concentrations of the various species in the reactor. Rj. as the rate at which the jth species is produced (moles!(time·volume» due to the chemical reo actions taking place. Imagine that we could somehow count up the net number of times an SiH4 molecule hit an SiH2 molecule and turned into an Si2H6 molecule during a short period of time. is defined as the number of times this reaction event takes place per time per volume. but small enough that there is negligible spatial variation in the average concentrations of the components or the reaction rate within V. The production rate of Si2H6. then the change in E is positive and the reaction rate is pos· itive.42 The Stoichiometry of Reactions 2. and that issue does not prevent us from defining the reactinn rate. for the reaction The reaction rate. Under this continuum assumption. but oniy one reaction rate.1: Defining the reaction SiH2 + SiH 4 . because we do not directly sense molecular transformation events.E. This means we can take the volume V large enough to average the random fluctuations of the molecules. tJ. we often deal with physical situations in which we assnme the material behaves as a continuum and we can ignore the discrete na· ture of the molecules. r.. an Si2H6 molecule is produced. We can measure con· centrations. We define production rate. We will also see that for complex reactiug systems. The extent E is a number of molec' ular change events and therefore the units of r in Equation 2. we can speak of the reactiou rate as defined at a point in space within some larger reacting system or physical reactor equlpment_ _ Notice in the defiultion of reaction rate. We are postulating that these collision and transformation events are taking place at the molecular leveL These literal reactions are known as elementary reactions.:::. It is clear looking at the stoichiometry in Equa· tion 2.1. the units of ex· tent are moles and the units of reaction rate are moles!(time·volume). which are the quantities we nsually care about in a commer· cial reactor. SiH2 and Si2H6 molecules loose in a box of some fixed volume V as depicted in Figure 2. The change in the reaction extent. The reaction rate is then y=-- 6E MV (2.

29. and is produced in the third reaction.7 is the same as the corresponding column of the matrix in Equation 2. H20+CO H20 + H OH + CO = C02 + = Hz The fundamental relationship between the reaction rates and the production rates now emerges.8: Production rate for 5iH2 + (-1) Yz + (1) Y3 ~ -Yz +YJ Consider the second species.4 Reaction Rates and Production Rates 45 now introduce the production rate vector.vill see in the next section under what conditions that solution can be found. We therefore write RH ~ (0) Yj tion 2.26 Consider what happens to the relationship hetween the production and reaction rates if there is more than one reaction. Equa- H2 + OH C02 + H = Three reaction rates are required to track all three reactions. results from these three reactions? We notice that H does not take part in the first reaction.44 The Stoichiometry of Reactions 2. Hz. Let Yj denote the reaction rate for the ith reaction. each row of the matrix in Equation 2. in terms of the reaction rates for the CVD example? . the two matrices are transposes of each other.28 is just the transpose of the row vector that comprises y ~ [-1 . H. We can therefore summarize Equation 2. That computation is a simple matter of matrix multiplication. Moreover. rate can therefore be expressed as You should examine the remaining four species and produce the following matrix equation.7 and 2.29) Notice that the column vector in Equation 2. We .30 implies that one can always compute the production rates from the reaction rates. Notice that the first row of the matrix in Equation 2. is consumed in the second reaction. Its production What is the production rate of SiR. is not so simple as it involves solving a set of equations. Example 2.29. we can sununarize the connection between the three production rates and the single reaction rate by RH RH2 ROH 0 1 0 -1 -1 1 (2. The reverse problem. Compare the matrices in Equations 2. In other words. Recall the water gas shift reaction.29 as (2. It is produced in the first and second reactions and does not take part in the third reaction.Ill. Yz Y3 J (2. deducing the reaction rates from the prodUction rates.7 is the same as the first column of the matrix in Equation 2.28) RH20 -1 1 1 -1 1 0 -1 -- 0 -1 1 Rco RC02 0 0 l y. R.29. which follows from Reaction 2.30) in which yT denotes the transpose of the stoichiometric matrix. What production rate of atomic hydrogen.

46 The Stoichiometry of Reactions 2. Extracting the stoichiometric numbers of SiHz for each of these reactions gives 6 [ 0 -1 1 1 1 0 1 -1 -1 -1 0 -1 0 -1 0 ~1 = [ ~J (2. H H2 OH H2 O CO CO2 Si] = (2.33. Consider again the water gas shift reaction chernistry.45) (2.H3 = ~ (2. in some chosen units of moles/(time·volume).49) H (2. and to consume OH. we briefly summarize Octave and MATLAS as high-level programming languages for numerical solution of reactor analysis and design problems.50) Looking at Reactions 2. MATiAS is commercially available from The MathWorks. 1 = l 0 1 I I 0 -1 -] -] I 0 -1 0 0 -J 0 fl. application of the fundamentals requires computational tools.4 so the row vector above is indeed the transpose of the fifth columo of the v matrix in Equation 2.44) (2. we note that SiH2 takes part in Reactions 2. 2.15. 0 The production rates of the six species due to these reactions are then computed as RH RH2 ROH RH20 Rco RC02 2.38) H. + SiH3 + SizH6 Hz (2.42) (2.43. Octave is freely available for a variety of hardware platforms and can be down-loaded from www. + Si.37) Si2HZ SiH4 + 5iHz Sh SizHz + 5i2 SizH-4 + SiH4 + H.37. and is becoming a commonly available tool of industrial engineering practice. 2.lh Si = 2SiH. + SiBz Hz H. SiH4 + SiH3 ::::: SizHs + SiH4 + SiH SiH. + Si. computing R from r is a simple matter of matrix mnltiplication. 2. but for now just assume we know the three reaction rates are.1 'Computing Production Rates from Reaction Rates H= SiH3 + Hz + SizH4 ::::: SiZH6 H2+ Hz + SiH = Siz H SiH3 Please perform this calculation for yourself.34.5 Computational Aspects of Stoichiometry As we have seen in this chapter.33) (2. and 2. + SiB SiH4 + Si + - = SizHs SiH 3 + 5iHz = (2. Sill2 was chosen as the fifth species in Exercise 2.octave. Hz and C02. H.40) Si2 Hz + Si + = SizH3 2Si. problems of realistic and complex reaction stoichiometry involve matrices and linear algebra.5 Computational Aspects of Stoichiometry 47 Solution SiH4 SiH4 = = ~ SiH.H2 = SizH. . prepares one to apply the fundamentals in realistic and complex industrial situations.46) (2.31) 5iHz + SiB.36) (2. 2. H2 = SiHz (2. H20 and CO at the given rates.32) (2. Inc.39-2.39) (2.5.34) (2.47) (2.31.. Moreover. After the fundamental concepts are in place. You may wish to choose another component such as Sill4 and check another column of the v matrix. if the computing environment is organized properly. = = SiB.31-2.41) (2.51) or RSiH2 =[ 1 0 -1 1 0 0 1 0 2 1 -1 -1 -1 0 0 0 0 0 0 0 ] r In Chapter 5 we discuss means for predicting reaction rates given species concentrations. the experience of solving nontrivial problems reinforces the understanding of the fundamental concepts and further tiWe are so sure you will not forget that products have positive stoichiometric numbers and reactants have negative ones that we will not repeat it again. + SiHz + SiH 5i = 5i.48) (2. 3 -1 J -3 -4 4 The effect of the three reactions is to produce H.35) (2.43) (2. In Appendix A. = SizHs As discussed previously.

It is clear from inspection of Reactions 2.2 Computing Reaction Rates from Production Rates By inspection. to help infer the corresponding reaction rates. As we discuss in Chapter 5.3 Measurement Errors and least-Squares Estimation ~.56 gives RE ~ -(RA + Re) as before.56) and r3 is arbitrary_ The second equation places the restriction RE ~ r2.55 and 2. which is characteristic of using sets of reactions that are not linearly independent.54) ~ [-~ -i ~ J 1 0-1 The second equation tells us RE ~ h . We explore subsequently what happens when this restriction is violated by the production-rate measurements.5.54 that the third reaction is the sum of the first two. we explore what happens if we use an independent set.(R A + Re ). which is the usual case When working with reaction mechanisms. not the reaction rates themselves. We require more information. 2. we explore what happens when the extra conditions on the production rates are violated by the data. B and C.55) and the production rates and reaction rates are related by and from the third equation (2. Computing the rank of v confirms that oniy two reactions are independent. Finally. we can deduce that r3 is not determined by the production rates. If we omit the third reaction. using the linearly independent reactions. Before we continue with the full set of three reactions. we can compHl'e'both reaction rates and we find a restriction on the possible production rates.52-2.53) (2. [ Re J Jl -i -~ -1~ J r~~ 0 1 In It might appear at first glance that we can compute j r given R because we have three equations and three unknowns.5. such is not the case. we cannot deduce the reaction rates from production rates. r. In that case we seek to determine the rate constanrs from the production rate measurements. If we now return to the original set of three isomerization reactions. Because r3 is arbitrary.48 The Stoichiometry of Reactions 2. we emphasize that using linearly dependent sets of reactions presents no problem at all if one is interested oniy in computing the production rates from given reaction rates. for example. To make the concept clear. So we conclude that measuring production rates is not enough to tell us reaction rates. In other words. but because the reactions are not linearly independent. ~ Another common task arising in the analysis of reactors and reaction mechanisms is to measure production rates by monitoring changes in species concentrations. A=B B=C C=A The stoichiometric matrix for these reactions is v -RA and the third equation gives Y2 ~ Re (2. the first equation gives Y. A=B B=C the production rates and new reaction rates are related by I I o As a final topic in this chapter. Imagine . wblch upon substitution of Equations 2. consider the simple isomerization reactions between species A.52) (2. We take up that important problem in Chapter 9.5 Computational Aspects of Stoichiometry 49 2.Y2 = . and that from the first equation (2. the extra information usually is provided by postulating a reaction mechanism and applying the laws of mass action to the elementary reactions.

Notice that we obtain reasonable estimates of the true reaction rates.05 ] 1.07 r. Figure 2.50 The Stoichiometry of Reactions 2. which is not too surprising because we have six equations and only two unknowns in Equation 2. For example.02 1.98 -1. that is each production-rate measurement.06 1. we find r = [0.97 2.92 -2.03 1.1 3 2 -3 -1 1 The effect of this error is to make the equations inconsistent. A system with more equations than unknowns is called over-determined.98 -1.03 F instead of the correct value r = [1 2 F. If we add small amounts of random noise to the data given in Equation 2.95 1.96 -1.09 -2.96 Figure 2. If one is given R and trying to solve for r from R ~ vT r (2.06 3. all of which are subject to small random errors.05 -1.06 -1.04 3.3 displays the estimated reaction rates given 500 production-rate measurements.99 -1.57) We can compute the production rates when the two reaction rates are in whicb the superscript -1 indicates a matrix inverse. one is usually interested in knowing the values of the two reaction rates that come closest to satisfying the six equations simultaneously. -2.98 1.2 shows the estimated reaction rates for the six productionrate measurements.97 -2.03 1. is in error by a small amount so the measured R is R = [-2. If the reactions are not linearly independent this inverse does not exist and the least-squares solution is not unlque as b e f o r e . As a final example.04 and we obtain R = [-2 3 2 -3 -I 1 r (2.93 1. H. we produce the follOwing sLx measurements.59) = 2.04 2.03 0.07 1.03 1. and compute the least-squares estimate of r for that measurement.01 1.99 l.5 Computational Aspects Of ~tolchlOmetry ~ I the following two linearly independent reactions of the water gas shift are taking place H20 + CO = C02 + Hz H20 + H then the least-squares solution is given by r = (y yT )-lYR I I = 0 1 0 -1 -1 I Hz + OH so the production rates are given by RH RH2 ROH RHzO Rco Rco.04 -2. then we are using the classic least-squares approach.00 1. Computing least-squares solutions to over-determined equations is a useful computation in linear algebra.05 2.01 -2. unmodeled effects undoubtedly cause discrepancy between predicted and measured values of R.58. we obtain six estimates of the reaction rates. let's conSider the case in which we have repeated measurements of the production rates. If we measure how close the equations are to zero by squaring the error in each equation and summIng over all equations. Notice that the estimated rates are again scattered about the true value.96 -0.01 -2. If we take each colunm of Rmeas.08 1.94 -2.92 0.06 2. one for each measured production rate. If the reactions ate linearly independent. -1 1 I -1 0 0 [ ] r] r2 (2.92 3.57.98 2.98 l. let's assume that the production rate of the first species. Normally we would not expect to find a solution for arbitrary R.01 -2. the . r = r I 2]T Next consider what happens when we have many measurements available.94 2. When the equations are inconsistent.58) Rmeas Now if R is determined by measuring species concentrations.03 1. then the matrix product v yT has an inverse and the least -squares solution is unlque.93 3.983 2. _ If we compute the least-squares solution to the inconsistent data given above.97 2. only those R that are generated by mUltiplying an r by yT . so there no longer is an exact solution for the reaction rates.98 -0. Notice a small error in the H production rate has translated into small errors in both inferred reaction rates. We can perform these operations in one matrix equation via and the result is rest ~ [ 0.

. We define the rate of reaction i.2 2. so the rank of v is the number of linearly independent reactions in the network. ri..85 0.15 1.2 A set of reactions is linearly Independent if no reaction in the set can be written as a linear combination of the other reactions In the set. to be the net number of times a reaction event occurs per time per volume.1 LIS 1. ns and ns is the number of species in the reaction network. but engineers benefit from thinking about extracting information from data and models at an early stage_ 1..2: Estimated reaction rates from six production~rate mea- In which A j represents chemical species j.05 r2 - 2 - 0 0 8 0 0 mean of these values is close to the true value. we can calculate directly the production rates of all species. i'"'l n.8 0. and the shape of the frequency distribution for the reaction rate estimates begins to emerge.9 1.2 :z:: VijA} ~ 0.2 .6 Summary and Notation 2. Vijri. The stoichiometric coefficients are contained iu the stoichiometric matrix v.95 1. The rank of a matrix is the number of linearly independent rows (equivalently columns) of the matrix. .2 2..85 0.. Given the rates of all reactions.9 1. .. .95 1 Tj 1. 0.15 !2.95 1. j~l. 2. .85 1. .1 !- I .6 Summary In this chapter we have introduced the compact notation for keeping track of the stoichiometry of chemical reactions. You might want to generate the same plot with 5000 random measurements and see how much more detail is apparent in the probability distribution of estimates. L05 f f 2.8 I I f f - - . Further discussion of parameter estimation from data is delayed until Chapter 9. Rj = i"d figure 2. ..9 0.95 I LOS 1. nr and nr is the number of reactions in the network We can summarize the reaction stoichiometry with one vector equation vA~O 2. In which Vi} is the stoichiometric coefficient for species j In reaction i.1 2..9 0. . - 2. The reaction rate is a fundamental concept that allows quantitative prediction of rates of conversions of reactants to products. i=l. ns . j surements subject to measurement noise.3: Estimated reaction rates from 500 production-rate measurements subject to measurement noise..8 0.05 r2 r:go o 0 0 0 0 2 o o 1..52 The 5toichiometry of Reactions 2.85 1. Unear Independence of reactions is equivalent to linear independence of the rows of v.8 0.1 LIS 1. . ~ j l1 r Figure 2. i ~ 1.

54 The Stoichiometry of Reactions 2.7 Exercises or as an equivalent vector equation. How many of the reactions are linearly independent? (c) Now that you have found the number of independent reactions.+ 3NO. :). there may not exist an exact solution of reaction rates.e.66) (a) Determine the stoichiometric matrix. .7 Exercises Exercise 2. it is a simple matter to compute the species production rates with Equation 2.63 in the final set. or your favorite software package to determine the ajl Aj E' j M] ni nr ns Yi Rj Ei vi) formula number for element l in species j jth species :in the reaction network lth element comprising the species reaction index. +NO (2. If the set of chemical reactions is linearly independent. and the species list.Os N0 2 + NO. Notation = = = = ~ 2N.60.04 0. :)] rank(stoi2) Do you see how the indices in forming stoi 2 work out? Notice we do not have to enter any more matrices after we build the original stoichiometric matrix to test the ranks of various reaction networks.stoi(6. ni.60) 2. 2N20s N. MATLAB. that satisfy Equation 2. If the observed production rates contain experimental errors.1: Finding independent sets of reactions Consider the following set of chemical reactions. reactions) from the original stoichiometric matrix..2. which ni of the original set of 6 reactions can be chosen as an independent set? Try guessing some set of ni reactions and determine the rank of the new stoichiometric matrix. computing the corresponding reaction rates requires additional information.64) (2..62) (2.. in generaL Given observed production rates.. one is normally interested in finding the reaction rates that most closely satisfy Equation 2. for this reaction system so the reaction network is summarized by vA~O (h) Use Octave. nr species index. stoi. Hint: you want to examine the rank of sub-matrices obtained by deleting rows (I. if you assign the original stoichiometric matrix to a name. Can this be correct'! Why or why not? . Stop when you have determined successfully one or more sets of ni independent reactions. v. N.62 and 2.. j "'" 1. In Octave.60.61) NO. (2. (d) What do you think of a coJ!eague's answer that contains Reactions 2. for example.63) (2. . by stoi2 ~ [stoi(1:4. A. ns molecular weight of the jth species number of independent reactions in reaction network total number of reactions in reaction network total number of species in reaction network reaction rate for ith reaction production fate for jth species extent of reaction i stoichiometric number for the jth species in the ith reaction rank of the stoichiometric matrix. then you can obtajn the rank of the stoichiometric matrix associated with deleting the fifth reaction.60. Oue cannot solve uniquely the reverse problem. In trus situation. .2. Y.65) NO+ NzOs N02+ N02 = (2. . i = 1. (2.04 + 0. The closest solution in a least-squares sense is easily computed with standard linear algebra software. then one can uniquely solve the reverse problem.+ N03 N03 Given the rates of reactions. such as rate expressions for the elementary reactions in a reaction mechanism. + N03 NzOs NO.

what are the two reaction rates that are consistent with these data.2: The stoichiometric matrix (a) "'lIal is the stoichiometric matrix for the following reaction network [2]7 By inspection. + 202 = C02 + 2HzO CH4 + ~Oz = CO + 2H20 (a) Are these reactions linearly independent? How many reactions comprise a linearly independent set? (b) List all sets of linearly independent reactions. in terms of the reaction rates for the reactions. . As· sume the production rates have been measured and are.51. for this reaction system so the reaction network is summarized by vA~O 2: I=l n. RJ . COz.2. Hz + Br. Which reaction is included in all of the linearly independent sets of reactions? Why'! . 5J. and CH4 that is fed into a furnace with O2 and produces CO. Is t. . n" which ni of the original set of six reactions can be chosen as an independent set? Try guessing some set of ni reactions and determine the rank of the new stoichiometric matrix.56 The 5toichiometry of Reactions 2. when the production rates are the same? Can we determine \-vhfch set of reactions is really causing this measured production rate? 1 Hz + 202 = H20 CH. in some units of moles/(time·volume). RH RHz ROB -I 3 1 Exercise 2.4: independent reactions for bromine hydrogenation Consider the follOWing set of chemical reactions [4... how many of the reactions are linearly independent'? How would you check your answer if you had access to a computer? H. v. The follOwing chemical reactions have been suggested to account for the products that fotm.7 Exercises 57 Exercise 2. A.ny (b) Use Octave or MATLAB to detennine the rank of the matrix using the rank What is the production rate of the jth species. VijA} = 0. and H2 0. Hz. Stop when you have detennined successfully one or more sets of ni independent reactions.3: Finding reaction rates from production rates Consider again the water gas shift reaction presented in Equation 2. and the species list. i = 1.5: independent reactions for methane oxidation Consider a mixture of CO.his answer unique? (c) How can these reaction rates differ. Br+ H2 = H + Br2 = H+HBr= 2Br = = 21 HI + H 1+ H2 = H+lz=HI+I (b) Given a stoichiometric matrix for a reaction network with ns species and - nr reactions ia) Detennine the stoichiometric matrix. 1 Ru2 0 Reo RC02 -3 -2 2 CO+ 202 = C02 (a) If you choose the first two reactions as a linearly independent set. Exercise 2.+ Iz = 2H! I. Yj? function. Is this answer unique? (b) Repeat the calculation if you choose the second and third reactions as the linearly independent set of reactions. Br2 = = 2HBr 2Br HBr+ H HBr + Br Hz+Br Br. How many reactions are linearly independent? (c) Now that you have found the number of independent reactions. Exercise 2.

= v. k. Exercise 2.j vjA j ~ 0 satisfies the following e equa· tions to balance tbe elements a12 2N20S = N20S = NO. Consider an arbitrary m x n matrix.butene k_J k'l isobutene cis . L2 = cis .58 The Stoichiometry of Reactions 2. (b) Determine A.9: Null space and fundamental theorem of iinear algebra Equation 2.z to hold the formula numhers. L. (a) Determine A. Show that any chemical reaction.7 Exercises 59 Exercise 2. A J' j ~ I.67 and establish that tbe number of linearly independent reactions that satisfy the element balances is butene I .2 butene = ks isobutene s ..2 ..2 . ] [ "u a21 . k.. B. E'.butene trans .7.67) Determine the rates of production of each component in terms of the rates of each reaction. Butene isomerization reactions are shown below.butene L. e.rank(j\ ) The numerical support for computing null spaces is excellent.. we can express (2. for the methane oxidation in Example 2.butene 1 .67 in Exercise 2. in which a)l is the formula number for species j corresponding to element l." cis . .8 is begging to be analyzed by the fundamental theorem of linear algebra [7J. E and j\ Determine the rates of production of each component in terms of the rates of each reaction.7: Production rates from reaction rates (a) Consider the following set of chemical reactions." 1~ [ a" 0 0 oJ N02+ 3N02 N204 02 +NO a" If we define the s x e matrix the element balance by j\ a. 02 and C are not observed as species. . oxygen (molec· ular) and water. . . (b) Repeat if Hz.7.. For example. so we explore that concept here. The null space of matrix B.2 .. +N0 3 NzOs NOz+ N02+ NO+ NO.butene = trans . I ~ I. consider Equation 2. 2Nz04 + 02 [ Vj V2 NO. Exercise 2.• s be comprised of the e elements. . + = N03 = NzOs = NO. written N(B ).. an a. Eand j\ for the species hydrogen (molecular). Writing the cbemical formulas for tbe species in the usual way Exercise 2. is defined to be the set of all vectors x such that B x ~ O. The dimension of N(B ) is the number of linearly independent vectors x satisfying B x ~ O.2 .8: Restrictions from element balancing Let the s species.butene k. the Octave command null (8) returns a matrix with columns conSisting of linearly independent vectors in the null space of matrix B . One of the remarkable results of the fundamental theorem of linear algebra is the relation rank(B ) + dim(N(B )) ~ n Armed with this result. .6: Methane oxidation and maximal independent sets (a) List a maximal set of linearly independent reactions if 02 as well as C are not observed as species in the methane oxidation reactions in Example 2.

If other quantities are conserved. the surface site.Titten among the 5 species? Give an example with more than one species in which you cannot vvrite any reactions.113-166J.13 presume that conserving elements in the chemical reactions is the only restriction. n max . Exercise 2.10: Umits on numbers of independent reactions (al Species A. n max ? (c) If you enforce the stronger condition of element balancing. Exercise 2. (b) If 5 > e. the usual case.e and c. the method of analysis remains valid. Let i be the number of reactions in the maximal independent set.13: Generalizing independent reactions TImfu 7 (h) What is the smallest possible number. how many reactions can be written? Prove you can write at least one valid reaction. Imagine a more flexible chemical world in which we demand conservation of only mass for a valid chemical reaction..60 The Stoichiometry of Reactions 2.9 and determine upper and lower bounds on the rank of. when we study reactions with catalytic surfaces in Chapter 5. I (a) Write a few mass-conserving reactions among these species that are not valid chemical reactions.I.>l based on s. how many linearly Independent valid chemical reactions are possible? Exercise _2. Before these developments.rank(. in whiclfi' of the species are chemical compounds.>I land i for this example. We will see one such conserved quantity. pp. cbemical transfonnation of the elements themselves. but the answers change. what are the largest and smallest nnmbers of linearly independent reactions that can be .>I ) from Exercise 2.77 the null space introduced in Exercise 2. B and C are observed in a reacting system. Le. was on the table as an early goal of research.12: An alchemist's view of stoichiometry In tbe early 1700s. Do you obtain tbe same size maximal set as in Example 2.e) '" i '" mints . n :5 Consider 5 species comprised of e elements. Assume we observe Hz.9 to be sure you have accounted for all linearly independent reactions.s .7 Exercises 61 (bl Determine rank(. not pure elements. (c) List a chemical example that has a maximal set with n reactions for each n in nmin::$.11: Generating possible independent reaction sets (a) If one has 5 species comprised of e elements. such as lead to gold. What is the largest possible number of linearly independent reactions among these species.e) Hint: use the result i ~ 5 .. then.. the alchemists were unaware of the defining role that electrons played in bond formation and transformation of chemical species. Show max(O. Exercise 2. We should point out that these lintits on numbers of reactions in Exercises 2.5 and the idea of I I I .8-2. The structure of the nucleus. and H20 as chemkal species. elucidation of bond formation with electron sharing and the construction of the periodic table were major triumphs of the chem~ leal and physical sciences in the nineteenth and early twentieth centuries [6. by inspection. 0. any mass-conserving rearrangements of the protons and neutrons jn the nuclei afC also considered valid chemical reactions. or find a counterexample to this statement. (b) How many linearly independent mass-conserving reactions can be con~ structed? You may want to use the result of Exampic 2.

131(2):425-434. A mathematical model of the cou- 3. It is by design a review. Kee. 1920. the consideration of the thermodynamic restrictions on reactor performance may motivate one to make operations or design changes that allow these restrictions to be changed and reactor performance improved. pled fluid mechanics and chemical kinetics in a chemical vapor deposition reactor. New York. ]. Specifying the temperature. 1980. For these reasons we begin with a review of chemical equilibrium. and assumes the reader is familiar with the fundamental concepts of classical thermodynamics as discussed in the many introductory textbooks on the subject [14.1. nj. Zur Theorie der Reaktionsgeschwlndigkeiten in Gasen. Elementary Chemical Reactor Analysis. MA. 16. [4J K. pressure. 26:49-54. ConSider the single-phase. Benson and R. rium. J. F. However. Electrochem. Chern. . the calculation of the equilibrium state can allow one to make concluSions about the limits of reactor performance that would be achievable if this state were reached. Coltrin. Butterworths. it is still important to understand what equilibrium is and how to draw quantitative conclusions from considering the equilibrium state of a system. (5) M. Srinivasan. i2l S. 23(1):200-202. 1955. 59(15):635-667. and the number of moles of each component. R. pressure. W. 9. New York. Phys.1989. T. Although essentially all reactors in Figure 3. Oxford University Press. E.1: Thermodynamic operation are not at chemical equilibsystem.Bibliography 3 Review of Chemical Equilibrium [lJ KAris. In addition. Linear Algebra and its Applications. Strang. Annalen der Physik.4]. Chemical reactors are not at eqUilibrium because the net rates of chemical reactions are seldom allowed to become zero. It is convenient for our purposes to describe the state of this system by the variables temperature. [7J G.1 Introduction This chapter is intended to remind the reader of some important facts concerning the equilibrium state of a system that can undergo chemical reaction. Academic Press. 1984. [3J M. Polimyt Reaktionsisochore und Reaktionsgeschwindigkeit vom Standpunkte der Statistik Zeitschrift fUr Electrochemie und angewandte physikalische Chemie.. 1919. and l A Miller. [6J R L Silver. Some complexities in the reaction system:Hz + h = 2Ht ]. multi component system depicted in Figure 3. 1998. Stoneham. The Ascent of Science. Soc. P. 5. Herzfeld.

is the energy function of these state variables that conveniently describes the condition of equilibrium. In particular the equilibrium state has a minimum Gibbs energy with respect to any changes in the number of moles at fixed T and P.6) -SdT + VdP + 'f. is the fugacity of component j and Jj is the fugacity of component.5. P (3.1 gives an expression for the change in the Gibbs energy due to a small change in the reaction extent.) T OE . fJj For the closed system. and number of moles of each component is developed in standard textbooks to be dG = -SdT + VdP + 'f.c. (3. P and E. for a small change in the extent of this single reaction.E (3.l1 j dTIj j G (3. The Gibbs energy of the system. The expansion of a total differential in terms of the partial derivatives allows one to deduce from Equation 3. . The activity and fugacity of component j arc related by 0]= T" Ij r J "VfJi = L j J.2 Condition for Reaction Equilibrium Consider a closed system in which the TIj can change oniy by the single chemical reaction.4) in which OJ is the activity of component j in the mixture refereuced to some standard state and Gj is the Gibbs energy of component j in the same standard state.1) in which S is the system entropy.2 Condition for Reaction Equilibrium 65 and number of moles of eacb component then completely specifies the equilibrium state of the system.7) S.2: Gibbs energy versus reaction extent at constant T and P.2 = Gj + RTlnaj (3. 'f.3) (3. the numbers of moles of each component change by dTIj = vjd£ Substituting this relation into Equation 3. (Vjl1j) dE j (3. dG = Consider a plot of G at some fixed T and P versus the reaction extent. V is the system volume and I1j is the chemical potential of component j. E 3. sketched in Figure 3.6 to compute equilibrium composition is inconvenient because cbemical potential is often expressed in terms of other quantities such as activity or fugacity.(3G) __ oT V (DC) oP = P.i in the standard state.5) in which f. The difference in Gibbs energy between two states at slightly different temperatures. G is only a function of T. A necessary condition for the Gibbs energy to be a minimum is that the derivative be zero.2.64 Review of Chemical Equilibrium 3. G.vjAj = 0 j Figure 3. (?.2) The direct use of Equation 3. pressures. Given the definition of reaction extent in Chapter 2. or in light of Equation 3. One can express 11 j in terms of the activity of component j.

Solution We recall that the fugacity of a component in an ideal-gas mixture is equal to its partial pressure. The standard Gibbs energy change for this rearubn at 400 K is -3.L (3. Example 3. na? = j 0 (3. and solid mixtures: pure component j at LO atm pressure and the system temperature.8) The term 2. but this pressure is close to 1.9) I.72 kcal/mol [I 71. K.vjGj +RTL. YJ' as the unknown variables rather than the partial pressures. we can simply calculate K at 400 K from Equation 3.66 Review of Chemical Equilibrium 3. The activity of component j is then simply L.0 atm pressure and pure component j. Equation 3. an ideal-gas mixture. Determine the equilibrium composition for this system at a pressure of 2. Since I1G' is given at the system temperature. 400 K.'.'. ~ l.5 atm and temperature of 400 K. strong functions of the system temperature.6.0 atm for most gases. Substituting the activity of the components in Equation 3.Vjllj j = L.10) and the standard state. From Equation 3. and B represents the butene.9 and exponentiation leads to the following simple condition for chemical equilibrium (3.12) The quantity is the fugacity of component j in the standard state.0 atm pressure and the system temperature. Sometimes the gaseous mixture standard state is defined to be the pressure corresponding to unit fugacity. I1G'.15) .j VjG. It is common convention in reaction equilibrium calculations to use the following standard state for gaseous.0 atm. K = 108. The following example illustrates these concepts and the calculation of the equilibrium composition in the simplest possible setting.0 atm because the partial pressure of a pure component j at 1. and simple logarithm addition and exponentiation identities yields I1G' + RTln I.10. (3. the units of Pj have to be atm or a j is not dimensionless as it shouid be. Multiplying both sides of Equation 3. aj = 1 atm (3.13 in subsequent discussion.2. however. Equation 3. we have (3.14) alaB From the definitions of K (Equation 3. One such in which P represents the trimethylpentane. K = .14 tben gives K ~ Pp PIPS If we choose to use the mole fractions of the components. It is obvious from the definition of the standard state that and are not functions of the system pressure or composition. I. 3-trimethylpentane I + B - P G. You may assume the gases are ideal at this temperature and pressure and this reaction is the only one that occurs.vjlnaj j j (3. because the standard state is l. it is clear that K also is a function of the system temperature. I.2 Condition for Reaction Equilibrium 67 Standard state. I represents isobutane. We now attempt to calculate the equilibrium composition from Equation 3. is known as the standard Gibbs energy change for the reaction. which is pure component j at 1.9 motivates the definition of the equilibrium constant.11.0 atm total pressure is 1.7 by Vj and summing yields reaction is isobutane + I-butene = 2.12. Using the equilibrium condition. liquid. but not a function of the system pressure or composition..0 atm in Equation 3. Gj and are.13) (3.11) Notice that when we suppress the 1.10) Rearrangement of Equation 3.1: Ideal-gas equilibrium The reaction of isobutane and linear butenes to branched (8 hydrocarbons is used to synthesize high octane fuel additives.

E TIED .E TIro . 2:Yj = 1 j consumed than produced by the reaction.E') (YEO .18 giving Yf Summing Equations 3. Notice that the mole fractions are now expressed in terms of the initial mole fractions.73 X lO2 yp = 0.YEo = YlO = 1/2 gives £.19) = nBO ~E.E')KP . which are known.2: Ideal-gas equilibrium. Ypo = a.18) YP = 1 .E TIro . The negative sign is therefore chosen.15 (and sununing to one) could be an equilibrium composition.885 in which TIT is the total number of moles in the vessel. All we know at this point is that any set of mole fractions satisfying Equation 3. We proceed to describe the mole fractions of all components in terms of a single variable.18 sum to one for all E'.2 Condition for Reaction Equilibrium 69 in which P is 2. nB Dividing top and bottom of the rigbt-hand side of the previous equations by TIro yields.18 into Equation 3.l. and only one equation. Let TIjO represent the unknown number of moles of each component inltially contained in the vessel. which is the single unknown. and the solution is E' = 0..E') = a The equilibrium equation is quadratic in E'. Since the vessel is closed.16 by TIr produces equations for the mole fractions in terms of the reaction extent.15 then provides the single equation to determine E'.17) TIro . Equation 3. Example 3. the gas-phase mole fractions at equilibrium. In other words.E' )(YlO .19. TIp = TIpo + E (3. The equilibrium mole fractions are then computed from Equation 3. K = (Ypo + E')(I . and a reaction extent.E') P Multiplying through by the denominator and collecting terms gives (YEO .16) The correct solution is chosen by considering the physical constraints that mole fractions must be positive.E' in which E' = E / TI TO is a dimensionless reaction extent that is scaled by the initial total number of moles.469.E' Ypo + £" (3. We can of course add the equation that mole fractions sum to one.E' YEO . the equilibrium composition cannot be determined with the information given. The total oumber of moles decreases with reaction extent because more moles are The equllibrium at 400 K favors the product trimethylpentane. The gas is contained in a closed vessel that is inltially charged "ith an equimolar mixture of isobutane and butene.1 with the following additional information.jO + VjE.15. the extent of reaction.73 X 10 2 YB = 5.2(1 + KP) E'(l + KP) + {l/4)KP = a The two solutions are . Notice also that the mole fractions in Equations 3. The problem is that we have three unknowns. 0 .16 produces = 5. Solution We show how the initial composition of the gas in a closed vessel allows us to determine the equilibrium composition. Substituting in the given initial composition. Notice in particular that we do not know the volume of the vessel or the inltial number of moles. the number of moles at some other time can be related to the reaction extent from the stoichiometry using n} = n. YlO .5 atm.E' YJ = I-E' YE = 1 . If the positive sign is chosen in Equation 3. 2 (3. It is important to realize that the mere statement that a system is at equilibrium at a certain temperature and pressure does not necessarily tell one everything about the system.(ypo + E')(1 .E but we still have three unknowns and only two equations. n10 . Substituting Equations 3. Dividing both sides of Equations 3. revisited Compute the equilibrium composition of Example 3. E = 1± {. 0 The next example illustrates what kind of additional information is required to be able to find the equilibrium composition. the reaction extent is greater than 1/2.E npo + E Yl = YB = yp = (3. just that the vessel inltially contains an equimolar mixture of the two reactants.E')(ylO . which is physically impossible because It would result in negative numbers of moles of the reactants. The reaction has moved from the initial coudition of pure reactants to an equilibrium state of relatively pure product.68 Review of Chemical Equilibrium 3.

23) Performing the sununation on the two terms yields Equation 3. p'~ • . Dividing Equation 3.P" " J Y.23.. (3.P . we have the inequality in Equation 3. we would like to show a 2 G IOE 2 > 0 at this value of reaction extent as depicted in Figure 3.28) (il2C) .30) We now proceed to calculate a 2G/OE 2.2 j Vi [Vj v YjO + ViE' . Summing over all components yields nT Substituting Equation 3. Before leaving the equilibrium of ideal gases.30 may not appear very useful.27) Differentiation with respect to 1 ~ nTO E' yields (3.21) Differentiating this equation with respect to gives and P are constant) in which v = Lj Vj.29) Since the reaction extent only has physical significance when it corre· sponds to a nonnegative number of moles for all components.21 gives Y = J (-~~-.26) . J (3.Iv (3 11J ) .O + ViE' ('I + VE "J - . v- (3.2 j The chemical potential of a component in an ideal gas is Ilj = Gj + RTln(Pj/l atm) = Gj + RT [lny} + InP 1 1 (3.29 yields Ilj = Gj + RT [In(Yjo + VjE') -In(l + vE'j + InP E' (T = nTO + VE > 0 (3.T. oniy for an arbitrary number of compo· nents. We stan with Equation 3. we can express the derivative with respect to the dimensionless reaction extent.20) From Equation 3.25) --- We remark in passing for later use that ~ = nT nro 1 +VE' > 0 V)l J J nToRT l+VE'(JZC) -aE'2 . Substituting into Equation 3. Recall that for the closed system undergoing a single reaction. We begin by performing the same steps we used to determine the mole fractions in Example 3. ) T. an uninteresting case.24 into Equation 3. (OC) T OE ~- Although Equation 3. it is instructive to show that the equilibrium state is in fact a minimum in the Gibbs energy and that this minimum is unique. then the minimum is unique.JOE' T. To show that this value corresponds to a minimum (and not a maxhnum or inflection).2 Condition for Reaction Equilibrium 71 Second derivative of G.L T.5.70 Review of Chemical Equilibrium 3.20. The total number of moles is strictly positive to preclude the case of no material in the system. if we define two Variables. Notice v = 0 if moles are conserved in the reaction. repeated here. E' (aZC) T.24) Multiplying through by 1 + V E' gives (3.P _ OE.P = RT [YjO .P = OE. we can describe the number of moles by ( gives 1 nToRT VjE' - cc1-+~VV7~E-' ] ] n '0 + VE J J nTO + VI: (3.22) Defining the dimensionless extent as before.20 by Equation 3. If we can show that j)2GloE 2 > 0 for every E.22 can be written as Yj = 1 + VE' (3.1 + VE' = EinTO YjO + VjE' (3. The procedure in the preceding example determined the value of the reaction extent for which BG /dE = O.

Since 1 + VE' > 0 from Equation 3. J 1+ (iJ G) _" a2" b 2 _ (" ab)2 nyoRT 2 VE' iJE'2 T. also known as the Schwarz inequality. As the last example on the equilibrium of an ideal gas.31) G~ j njG.3 7 gives (338) Due to the definition of E'. Evaluation of G.3: Minimum in G for an ideal gas ~ C.35 and 3.38 yields G~ -E'inK + I(Yjo j + VjE'} [In((YjO+_VjE')) +lnp] 1 + vE' which can be rearranged to ~ G ~ -E' InK + (l + VE') InP + I.2 is indeed the correct picture for an ideal gas. R and T are positive. The chemical potential of a component in an ideal gas is given by Equation 3.2. As we shall see. (3. E') ~ G . ~ I j njoG. P. Aris provides an alternative treatment of this same question Ill.35) The Cauchy inequality [2]. with k 2c 0).P..30 can be written as Substituting the above into Equation 3. the shape of the function G is the same as G.29. and the equilibrium state is unique.20.L.25 and nyO. +RTInj [lny. J J . In particular. Equation 3.34 gives L. and strict inequality holds in Equation 3.2 and show that E' corresponding to the minimum of G agrees with the result of Example 3. Since the G function has positive curvature at every E.F . . the definition of K and Equation 3.33) We now define a modified Gibbs function to measure changes from the Initial standard Gibbs energy and to scale by the temperature and the total initial moles as follows.6 for a derivation) (3. we can conclude (3.3 7) Substituting Equations 3.p > 0 which is what we set out to show. the ratio njlnTo is easily seen to be equal to YjO + VjE'.31.36 into Equation 3. +lnP] j (3. Notice that this inequality is true for every E corresponding to physically meaningful compositions.>G' The equality is achieved if and only if the two vectors are nonnegatively proportional (aj ~ kb j for every j.32) Note that for this single reaction case. Substituting this relation. repeated here.24 into Equation 3. states that for any two vectors a and b (3. J j j L.SnoG' J J j - (3. we calculate the actual function G (T.33 is equivalent to (~:nT. Gis simply a shift of the G function up or down by a constant and then rescaling by the positive constant II(nToRy).36) nToRT (3. Figure 3. E'). + InPj Plot the Gfunction for Example 3.. nj ~ njO + VjE. and the minimum with respect to E' is at the same value of E' for the two functions.72 Review of Chemical Equilibrium 3. the equilibrium extent then can be determined by plotting G versus E'.34) G ~ Illjnj j From Equation 3. The Gibhs energy of a single phase can be computed from the chemical potentials and mole numbers of the components from (see Exercise 3. (YjO + VjE') In (YO ++ VE') 1 J J _ J VEt (3. + Et. which gives I j njG. + RT [Iny. Ilj Example 3. Since aj ~ y. G(T.39) Notice that because T and P are known values. )0 ) it is clear that the vectors a and b are not proportional. :~E' bJ .2 Condition for Reaction Equilibrium 73 Equation 3.

thus pushing the minimum in the Gfunction to smaller values of E.5 atm. which means there are more product than reactant molecules in the stoichiometry of the reaction.5.P.45 0. Lj Vj ~ v ~ -1. the opposite situation arises.1 Evaluation of the Gibbs Energy Change of Reaction We Usually calculate the standard Gibbs energy change for the reaction.25 E' 0.5 -1.E') ~ -E'lnK(T) + (1. so Ypo ~ O.41 aod plotting the results gives Figure 3.39 that for an ideal gas.1 0. which is in good agreement "'ith the value 0.2 Condition for Reaction Equilibrium 75 Solution We apply Equation 3. the graphical representation of Gis more informative on issues of whether or not the solution is indeed a minimum and whether or not the minimum is unique.'lith an equimolar mixture of reactants. This term often gives rise to a large effect of temperature on equilibrium. 3.n In(I .35 0. Because the minimum is close to E' ~ 0.5 -1 E'ln(E') + 2(0.E') (3.3 0. because the term (I + E'V) InP is then constant with E' and does not affect the location of the minimum.YIO ~ YEO ~ 0. For single liquid-phase or solid-phase systems.39 to the chemistry of Example 3.3.469 calculated in Example 3. Hence for v > 0 an Figure 3. Substituting these values into Equation 3.E')InP+ o G -0. Although we prefer the simple calculation in Example 3.47 at equilibrium. then an increase in the pressure causes a decrease in the equilibrium reaction extent.2. I+B 1 -LSS -L1\9 = P (3. 0 Notice from Equation 3.0. The temperature effect on the Gibbs energy is contained in the htK(T) term. The standard Gibbs energy offonnation of species j is defined as the difference in the Gibbs energy . Recall we must use atm for the pressure units because of the way we expressed chemical potential in Equation 3. If v is positive.9 For this stoichiometry. dG".5 we see again that the equllibrium favors the product. because the chemical potential of a component in a liquid-phase or solid-phase solution is usually a weak function of pressure. We turn our attention to the evaluation of this imPOrtaot temperature effect in the next section. the pressure does not affect the equilibrium.(l .2. If positive.. If v is negative.5 .29.2 to obtain a numerical value for E'. hy using the Gibbs energy of formation of the species.74 Review of Chemical Equilibrium 3.2 0.15 0. Substituting these values into Equation 3. For v ~ 0. It also gives us confidence that the analysis of the previous section is correct because Ghas positive curvature for all E' of physical interest.4 0. The pressure is 2.3: Gibbs energy Gversus reaction extent c'.5 The equilibrium constant at 400 K has been computed to be K ~ lOS. Magnifying this region of the plot in Figure 3. This can be verified by examining the term (l + E'V) InP in Equation 3. which are COmmonly available thermochemical data. increasing the pressure increases this term more at larger values of E' than at smaller values.39 gives G(T. The effect of pressure on the equilibrium extent depends on the sign of v.5 -2 Recall that the raoge of physically significaot E' values is 0"E'. The reaction started . in agreement with Le Chatelier's principle.41) ·1.3 shows that E' ~ 0.05 0. the pressure enters directly in the Gibbs energy "'ith the In P term.E') In(O_S - n . the effect of pressure on equllibrium is usually small.2.40) 0. equilibrium shift favoring reactants is the effect of a pressure increase.39.5 o 0.

44. Vendors offer a variety of commercial thermochemical databases to address this need. The list of Induded compounds is available at: http: Writing Equation 3..asp. .43) We then take the appropriate linear combinations of the species Gibbs energy of formation to compute the Gibbs energy change for the reaction at 25'C [.).76 Review of Chemical Equilibrium 3.nk in which X is any extensive mixture property (U.42) 1". industrial problems. the Gibbs energy of the elements cancel.G" I::. Finding appropriate thermochemical data remains a significant challenge for solving realistic.47 is valid also for the corresponding partial molar quantities Gj = H j TSJ (3.0 atm.11.G' is discussed in the next section. =-S aT 1 For educational purposes and to gain some experience with thermochemical databases. other students/chemsearch. In Recall from Equation 3.49 for the entropy change and substituting into Equation 3.2 Temperature Dependence of the Standard Gibbs Energy (3. the partial molar property of component j measures the change in an extensive property as a small amount of component j is added to the mixture at constant T. Equation 3. To accomplish this conversion.2 Condition for Reaction Equilibrium 77 of the species and its constituent elements in some standard state.0 atm. so we need to convert from 25'C to the system temperature.". i'iT = al::. L aT L VJS j j 0 J I j VijG ff (3.44) Defining the term on the right-hand side to be the standard entropy change of reaction. S.' The standard state temperature 25'C is often not the temperature of the system ofinterest.T5 As shown in Exercise 3.Gi = Notice in this context that the subscripts P and i l j denoting constant pressure and'composition are redundant due to the definition of the standard state and are dropped to avoid confusion. Gif = Gj - I n.1"I::. the Gibbs energy of formation of H20 is (3.W .byu. pure component j at a pressure of 1.G" I::.S" (3. etc. A web·based student version of the database provides students with access to data for 100 (ommon compounds at no charge: http://dippr. V. 3. For example.3 that the change of the Gibbs energy with temperature is the negative of the entropy. G =H.S' gives (3.W . leading to Equation 3. I::.45 with the stoichiometric coefficients of a single reaction gives '\' o(vjGjJ ~ '\' _ . 1 The standard state for the elements are usually the pure elements in their common form at 25'C and l. This expression can be used to evaluate the change in the Gibbs energy in the standard state.46 gives l/dippr.H. you may wish to try the Design Institute for Physical Property Data (DlPPR) database.49) Solving Equation 3.1"5.48 for a component in its standard state gives G" H. :ffilnming Equation 3. the temperature dependence of I::.45) ajlG EI (3.48) = (aG) P. G. and summing with the stoichiometric coefficient yields J D.2. Recall the definition of a partial molar prop- erty is Xj = (::) J T. Many companies also maintain their own private thermochemical databases for compounds of special commercial interest to them.47) Partial molar (3. P and other mole numbers.46) Let H denote the enthalpy and recall its connection to the Gibbs energy.P.A.G" = I::.byu.

but if "C. The partial molar heat capacity for component j is defined by Cp 2 Consider a multicomponent. T1 RT2 To evaluate the integral one needs to have the standard enthalpy change as a function of temperature.' = 0. multiphase system that is at equilibrium and denote two of the phases as <X and {3. Similarly if the pressures are not equal.RT2 "W Using differentiation formulas. known as the van't Hoff equation.52 is correct.. one expects a mass transfer of the component from the phase Wifh higher chemical potential to the phase with lower chemical potential until the chemical potentials become equaL In reality.. Similarly for the pressure.nk Pj =Pj' {3 j ~ 1.52 also may be a useful app=imation. 3.. Let Tk and pk represent the temperature and pressure of phase k and represent the number of moles of component j in phase k.3 Condition for Phase Equilibrium Equation 3.RT2 which finally can be expressed in terms of the equilibrium constant (3.50. multiplying by the stoichiometric numbers and summing produces in which ilCf.50 can be integrated directly to give In ~ ilH' (3.50 can be integrated between two temperatures to give T2 I aIn K _ T1 aT dT - IT2 I'>H' d1. then ilHO is constant and Equation 3. R T2 Tr (K2) _ (2. The chemical potential of component j in phase k is denoted by The hat is to emphasize that this is the chemical potential of component J in the multicomponent mixture.52) K. ~ . but the error in assuming constant standard change in enthalpy should be examined if heat capacity data are available. the . expresses how the equilibrium constant varies with temperature. one expects to see a transfer of heat from the hotter to the colder phase until the equilibrium condition is satisfied. Most of our attention then is focused on the condition that the chemica! potential of component j must be equal in all phases. If the temperatures of the phases of a two-phase mixture are not equal. T a ~ T~ pa "IX ~p~ h ~ __ ( OH) J aT (3.51) P. Finally.3 Condition for Phase Equilibrium 79 Rearranging this equation and division by R T gives 1 a"G' "G' RTaT . _2. then ilH' is still independent of temperature and Equation 3.) a(~) aT ilW = .RT2 ~ . This information can be computed from the partial molar heat capacities. if the chemical potential of a component is higher in one phase than another phase.z:: vjC.ns (3.78 Review of Chemical Equilibrium 3. is the heat capacity change upon reaction ilCf. the temperatures of all phases are equal and the distinction of temperatures for each phase is unnecessary and one can speak of tbe system temperature T. the left-hand side can be rewritten as If the standard heat capacity change of reaction is zero over the temperature range. -Equation 3.'} } The standard enthalpy change between a reference temperature To and an arbitrary temperature T can be evaluated from Ml" (Tl = ilW (To) + I T To "CpdT Since these equations apply to any two phases. The conditions for equilibrium when more than one phase are present are developed in standard texts to be nJ ilJ. A loose but useful physical intuition is connected to the equilibrium conditions. one expects a momentum transfer from the system at higher pressure to the one at lower pressure.. of course.2.53) Evaluating this expression at the standard state. Equation 3.50) Notice the Hj themselves may be strong functions of temperature.

54 for two phases c< and f3 and equate their chemical potentials we deduce liquid (and Solid) Solutions Again the simplest mixing rule for liquid (and solid) mixtures is that the fugacity of component j in the mixture is the fugacity of pure j at the temperature and pressure of the mixture times the mole fraction of j in the mixture. at t h e system temperature and 1.60) I? (ideal gas) The exponential term in Equations 3. The simplest rule for evaluating the fugacity of a component in a mixture is the ideal mixture assumption.60 is called the Poynting correction factor. then the second component follows Heury's law: (3. We let Xj denote the mole fraction to distinguish liquid-phase and solid-pbase from gas-phase compositions. If we express Equation 3. which is defined by (3. pressures of interest to obtain (3. pp. one can establish from the Gibhs-Duhem relations that if the first component obeys the ideal mixture.1ng linear mixing rule I""c = I jGYj (ideal mixture) (3. The simplest mixing rule for calculatiug the fugacity of component j in the mixture containing component j at some mole fraction Yj is the follov.56) j An ideal gas.3.59 and 3. obeys this mixing rule and furthermore the fugacity of pure j at the mixture T and P is simply the system pres= P. . 3.57) This apprOximation is usually valid when the mole fraction of a component is near one. the chemical potential is often expressed in terms of the fugacity. We integrate between the two Consider a gas-phase mixture of interest at some temperature and pressure.3 Condition for Phase Equilibrium 81 gradients and fluxes do not separate so ideally and we observe the simultaneous transfer of heat. Again.1The standard state is again pure component J. for liquids and solids. momentum and mass of all components while the system approaches equilibrium from some arbitrary initial condition [3. We start with the standard relation Vj In this and the next section we briefly review the methods for expressing the fugacity of components in a mixture and calculating the phase equilibrium. for example. Note that it is not generally equal to Ii (unless the mixture is ideal over the entire composition range). In a two-component mixture. Let IG denote the fugacity of pure component j in the gas phase at the temberature and pressure of the mixture. the partial molar volume is generally small. Sometimes we know the fugacity of a component at one pressure and we need to evaluate it at another pressure. 767-768J.58) jiX =jfi J J j= IJ 2 " ' J n s (3. The fugacity of component j in an ideal gas is therefore sure.1 Ideal Mixtures in which k2 is the Henry's law constant for the second component.80 Review of Chemical Equilibrium 3. Fugacity pressure dependence. Gaseous Solutions RT in which Vj is the partial molar volume. - (3.55) One can therefore use either the equality of chemical potentials or fugacities as the condition for equilibrium. the partial pressure of component j. the partial molar volume is usually constant over reasonable pressure ranges giving (3.54) in which /1~ denotes the chemical potential of component j in a standard state.0 atm pressure. and the Poynting correction may be neglected if the pressure does not vary by a large amount.59) For condensed phases. We review this case first.

These coefficients may be available in several forms. Use Henry's law to calculate the composition of all three phases for K ~ 4.4 Heterogeneous Reactions 83 3. However. P~" If the system pressure is not greatly different . Correlations may exist for systems of interest or phase equilibrium data may be available from which the coefficients can be calculated. correcting factors from the ideal mixture assumption are used. Assume K ~ 1.13. 4>j.82 Review of Chemical Equilibrium 3.4: Chemical and phase equilibrium for a nonideal mixture Consider the liquid-phase reaction A(I) + B(l) Solution Part 1. the fugacity of A is computed from the activity coefficient A f'L -~ - InYA ~ lnyc ~ xz' [AAC + 2(AcA xl AAclxAJ [ACA + 2(AAC . The calculation of the partial pressures of A and C in a vapor phase in equilibrium with a liquid-phase mixture of A and C follows from equating the chemical potentials of the components in the different phases. The fugacity of pure liquid A at the system T and P is not known.Yj.Ac. In the liquid phase.18.23. (3.62) = e(l) (3. it is simply the vapor pressure. The following example illustrates the important issues. What is the equilibrium pressure? 3.50 atm The analogous correcting factor for the liquid phase is the activity coefficient.4 Equilibrium Composition for Heterogeneous Reactions 4. Phase Ill: ideal-gas mixture of A. Plot the partial pressures of A and C versus The introduction of these coefficients obviously has shifted the burden from evaluating the fugacities to evaluating the activity and fugacity coefficients. as follows All three phases are in intimate contact and we have the following data: AAC ~ 1. Repeat for K ~ 0. 5.3. one defines the fugacity coefficient. assume the activity coefficients are given by the simple Margules equation. 'L Jj in which perature.l Xc 1 JL AXA}'A (3.0 PB ~ 0.63) Phase II: pure liquid B.. XA for a vapor phase that is in equilibrium with oniy the A-C liquid phase.65 atm ACA ~ 2. the fugacity of pure liquid A at the system temperature and the vapor pressure of A at the system temperature is known. Use the Margules equation to calculate the composition of all three phases. Phase I: nonidealliquid mixture of A and e only.2 Nonideal Mixtures For mixtures that display significant deviations from ideality. Sketch the meaning of Henry's law on the plot and verify your calculation from the plot.7. For illustration purposes. Example 3.12. For gaseous mixtures. The following list is a representative sample of the numerous texts devoted to this important topic [11. Band C. 3.5]. 2. Repeat 4 with an ideal mixture assumption and compare the results.4 PA~ 0. Pi is the vapor pressure of component j - at the system tem- ~ fjx/Yf L 1. Since the gas phase is assumed an ideal-gas mixture. In this section we illustrate the calculation of chemical equilibrium when there are mUltiple phases as well as a chemical reaction taking place.50 atm Pc ~ 0. Compute the Henry's law constants for A and C from the Margules equation. .61) that occurs in the following three-phase system.

4 plots PA.64 gives ~=~~ 0.7 0. XA.4: Partial pressures of components A and C versus liquidphase composition in a nonidea! solution. C. Notice that the system exhibits positive deviations from an ideal mixture. using the Margules equation to compute the activity coefficient in Equation 3.65) The analogous expression is valid for Pc. dashed line: ideal mixture. Figure 3. The unknowns in the problem are XA and Xc in the A-C mixture.3. and use the fact that the mole fractions sum to one in the gas and A-C liquid phases. Substituting in the values gives kA = 2.62 gives. (3.4 Heterogeneous Reactions 85 1 ~ which is valid for XA smalL Comparing to Equation 3. In general we would set up six equations in six unknowns and solve the problem numerically.63 and using Equation 3. Henry's law for component A is in which K is evaluated for a liquid-phase standard state. Finally. Figure 3.84 Review of Chemical Equilibrium 3.8 1 These values can be checked by examining Figure 3. from the vapor pressure of A at the system temperature. bUlthe structure of this problem allows an analytical solUtion.64 into 3. The validity of this assumption can be checked after the eqUilibrium composition is computed.PC and PI = PA + Pc versus the composition in the liqUid phase. one expects a large value of the equilibrium constant to favor the formation of the product. Evaluating the activities first of component A.4.6. fugacities of liquids are only weak functions of pressure L fA - - PA (3.7. Part 2. We require six equations for a well-posed problem. YA.63 and equating to 3. as discussed in Section 3.2 PA / > 0 0 > The analogous expression holds for component C. 0. and the system pressure.6 Pc 0.1.:. There are therefore six unknowns.4 So the Henry's law constant for component A is - 0. . the component for which we assume Henry's law. For K = 4. then the following approximation is valid because. solid line: Margules equation. which is also plotted in the figure.2 OA XA 0. YB and Yc in the gas phase.6 0. The slope of the tangent line to the PA curve at XA = 0 is equal to k A • The negative of the slope of the tangent line to the Pc curve at XA = 1 is equal to kc.8 which is also valid for small equation for XA = 0 gives XA· Computing YA from the Margules 11 os c.65. kc = 3.64) Substituting Equation 3. the chemical equilibrium provides the sixth equation and the problem can be solved. We can equate the fugacities of each component in the gas and liquid phases. We therefore assume that XA is small and Henry's law is valid for component A. Consider first the condition for reaction equilibrium.

Xc is assumed near one. The fugacity is then evaluated from Substituting in the provided data.0 atm. Ye ~ OA3 Part 3. but we do know that PAis the fugacity of pure liquid A at the system temperature and the vapor pressure of A at this temperature.86 Review of Chemical Equilibrium 3. the equilihrium composition may not allow Henry's Using Xc =1- XA and solving Equation 3. Pc ~ OA8 atm Ie Ie Xc a c = Ie = Ie ~L ~L L Now Ih and Ie are the fugacities of pure liquid C at the system temperature and the system pressure and 1.i(ACA + 2(AAC = = 0. IA is the fugacity of pure liquid A at the system temperature and 1.l Again the assumption of Henry's law is justified and the system pressure is low. this value is unknown. Henry's law constant and known liquid-phase compositions gives PA = 0. ACA) (I .24 atm Substituting these activities into the reaction equilibrium condition gives K= _ Xc xAkA/P'A.51. From this value. This low pressure justifies the assumptions we have made about fugacities of pure liquids at the system pressure. The fugacities of all components in the ideal-gas mixture are equal to their partial pressures. YB = OA5.4 Heterogeneous Reactions 87 From the standard state for reactions.50 atm. The answers are XA = 0. the vapor-phase concentrations can be computed from the ratios of partial pressures to total prl'1'>imre.XA))l .65 .C The solution is Xc ~ {).12.95 K = XA exp [0:' xA)2(AAC + 2(AcA (3. Therefore.4. The system pressure is therefore P = L 11 atm.0 atm is not large. then it is a good assumption to assume these fugacities are equal giving. The difference between 0. the assumption of Henry's law for component A is reasonable. ForK = 0. Part 4. YB = Xc = OJ)3 Yc Since component B is in a pure liquid phase.67 can be solved numerically to give XA = 035.23 one expects the reactants to be favored so Henry's law is assumed for component C You are encouraged to work through the preceding development again for this situation.05. so we assume IA = PA' Since XA is assumed small.65 and 1.AAC )xA) 1 (l-xA)exp [x. Again.66) XA L P = 1. Xc 0. PB = 0. respectively. The vapor compositions now are computed from the phase equilibrium conditions. which gives YA ~ 0.97. the rest of the solution can be computed.09 IB Iii ~L = IE = I IB (3. the same reasoning leads to tiL = B YA = 0. from Figure 3. equating these to the fugacities of the liquid phase gives Equation 3. If the system pressure turns out to be reasonably small. Finally. we have one equation in one unknown. = 0.0 atm. By expressing Substituting in the provided data gives XA Xc ~ 1 XA.13 atm. OAO.66 for yields law to be assumed for either A or C In this case we must solve the reac· tion eqUilibrium condition using the Margules equation for the activity coefficients. Recall that the partial pressures of A and C now must be computed from the fugacity of the liquid phase using the activity coefficients j= A.67) Therefore. For K = 1.

5: Gibbs energy versus two reaction extents at constant T and P. the solution is XA ~ 0. G is a minimum as a function of the ny reaction extents. Necessary conditions are therefore ~ 0.96 for K ~ 0.5 Yc ~ YA ~ 0./l} L. VijdEi j = -SdT + VdP + ~ (~Vij/lj) dEi (3. 1.. 0.69 gives IL.l0) .ny Letting E.054 for K ~ 4.69) At constant T and P. be the reaction extent for the itb reaction. We again consider a single· phase system but allow nr reactions 2:>ijA j J (~~) Thin ~ 0.5 Multiple Reactions 89 YA ~ 0. dnj ~ Li VijdE. P YB ~ ~ 0.37 atm G(Ei) Using the full Margules equation instead of Henry's law for the other K values gives XA ~ 0. the ideal mixture assumption leads to significant error as might be expected from the strong deviations from ideality shown in Figure 3.5. the equi· Iibrium condition becomes K ~ Xc XA which can be solved to give XA ~ I/O + KJ.88 Review of Chemical Equilibrium 3.2.7 and XA ~ 0. 0.30. For K Xc ~ ~ ~ 1. the change in the number of moles of component j for a closed system due to the ny chemical reactions is CHiS) as illustrated in Figure 3..28 1. gives dG ~ -SdT + VdP + L. YB 0. The phase and reaction equilihrium conditions generally lead to mathematical prob· lems of two types: solving nonlinear algehraic equations and minimiz· ing a nonlinear function subject to constraints. .4. in good agreement with the Henry's law solution presented above. I J i ~ l'2'''~j (3. 3./ljdnj j Determining equilihria for reacting systems with multiple phases and reactions can require significant computational effort.23. Finally.5 for the case of two reactions.. Part 5. 0 We can compute the change in Gibbs energy as before 3. In this section it is assumed that the required thermochemical data arc available.37. VijfJj ~ 0.5. Evaluating the partial derivatives in Equation 3.5 Multiple Reactions dG ~ -SdT + VdP + L. but find· ing or measuring these data is often another significant challenge in computing equilibria for systems of industrial interest. P ~ LOB atm As can he seen from comparison with the solution to Part 4. = 1.23 Figure 3.47. Yc ~ 0.1 Algebraic Approach Substituting in the change in the number of moles with reaction extents. .36. if one assumes that the A-C mixture is ideal.

£.49 kcal/mol for this reaction at 400 K.5 Multiple Reactions 91 In other words.351 Other initial guesses result in the same solution. 3The mole fractions are real and positive. = YP2 = 0 2 Using the initial guess from the solution of Example 3.3 trimethylpentane denoted as PI and reaction 2 be the formation of 2.£2 YP. YP. we have K2 = = 0. Substituting Equation 3.E~ 20 + E~ - (3.3 and 2.£. tl. Forming the mole I j Vijllj = j VijG.) - (YPw + £.2.£.£.469.7 for the chemical potential in terms of the activity into Equation 3. The news oniy gets worse when we consider tbat a host of other products are also possible. = 0.90 Review of Chemical Equilibrium 3.£.2.4-trimethylpentane may also form with the same stoichiometry.3.£.'Y of formation of many of these possible 2It is best to use a numerical tool to solve these equations when there is more than one reaction.3.3. defining the equilibrium constant for reaction i as (3.j VijC.2.72) £.£2.4-trimethylpentane (3.4-trimethylpentane denoted as P2. £2 = 0.2.£.£2 YB ~ 1 £2 £. From the Gibbs energy changes.£.2. n1 Notice that the existence of the second reaction causes us to produce considerably less 2. + RT I j YJO .4 isomer. physically meaningful solution 3 The mole fractions are computed from Equations 3. Yl = YB = 0.G.) 0 (YP20 + £.3-trimethylpentane in Example 3. YP. YP. = 0. leading us to believe that this solution is the unique. The Gibbs enerl.£.) £.£j .71) Substituting in the mole fractions gives two equations for the two unknown reaction extents.5: Equilibrium composition for multiple reactions In addition to the formation of 2.= -4.2. £2=0.) =0 Example 3.3-trimethylpentane in favor of the 2.2.£1 . YB ~ 0. - £2 Y P + £2 1 . ..2. = n10 ~ £1 E1 ~ £'2 £2 np2 n8 = 11BO = np:!O + £2 .Gi = L.031.72 to the two reactions gives Kl ~ = I I j j .70 gives The total number of moles is then ny fractions yields YI = = nyO . the necessary condition for a single reaction applies for each reaction in the network for the multiple reaction case. such as the 2.E~ . gives Vijllj = tl.£.4 is preferable due to its high octane content.73) These two equations are solved with the initial condition of an equimolar mixture of reactants.680 284 We now set up a table to calculate the four compositions from the extents of the two reactions.2. + RT Vi} Inaj Applying Equation 3. it is knOW<1 that 2.£2 1 . 1 .74 giving YI Recalculate the equilibrium composition for this example given that tl. This is bad news in this example because the production of 2. ~ YPlO Vij In aj YEO .3-tetramethylbutane and still others. [1 = 0. £2)(YBO . isobutane + I-butene = 2.)(1 .74) Defining the standard Gibbs energy change for reaction i. .258.HYBo .2:'I'L YIYEP Finally.133. . . gives the solution Cl ~ 0.2..5. PKdyIO PK2(YIO - allows one to express the reaction equilibrium condition as (3.031.E~ .E.2. Solution Let reaction] be the formation of 2.)(1. . .C.4 isomers as well as 2.

78) J The other main approach to finding the reaction equilibrium is to minimize the appropriate energy function. nonnegative mole fractions.5 Multiple Reactions 93 products show that they are presenl in significant amounts at equilibrium_ It is clear that one cannot allow the system to reach equilibrium and still hope to obtain a high yield of the desired product.92 Review of Chemical Equilibrium 3. . the activity of component j.79) i Onr final statement. We recall for an ideal-gas mixture.5.] lnaj 1 (3.76) Solution We start with the modified energy function. The min notation means function G(E.6: Multiple reactions with optimization to We again use Equation 3. This use of catalysts to promote the rates of desirable reactions is discussed further in Chapter 5. and find the equilibrium composition by minimizing the Gibbs energy.IT. In Ki + L 1 [YjO + L VijE. If we sum on j in Equation 3. If such a catalyst is available. Example 3.78. as shown in Example 3. + Gj ~ viiEi + [njO + ~ VUE.3.75) and express the chemical potential in terms of acti\1ty To find the equilibrium compOSition. or.76 and introduce this expression into Equations 3. we minimize this modified Gibbs energy over the physically meaningful values of the ITy extents. if we wish to use intensive variables. I (3. p It is convenient to define the same modified Gibbs energy function that we used in Equation 337 G(T. we can write the energy [unction as G= =:2:>.) with respect to the argument E. we obtain Using the definition of the equilibrium constant. This optimization-based formulation of the problem. therefore. = EdIT TO. again. is simply its partial pressure diVided by 1.77) a) = 1 atmYj (ideal-gas mixture) .P.68 to track the change in mole numbers due to multiple reactions. in this case the Gibbs energy [19j.0 atm. Equation 3.79.2 Optimization Approach in which E. (3. EI ) = ~ rtyoRT (3. that they produce nonnegative mole numbers. The solulion to problems like these is often to develop a catalyst that selectively increases the reaction rate for the desired reaction in comparison to the undesired side reactions.75 and 3. Equation 3. minimize tbe Using the two previous equations we have /ljITj = ITjoG.. 0 3. We first extend our evaluation of the Gibbs energy for the single reaction case to allow multiple reactions. for finding the equilibrium composition for multiple reactions is to solve the optimization problem (3... . We can express these constraints as l=y·O= "V'E' <0 L J U t - j j = 1. the reaction is halted after a high yield of the desired product is formed and before the equilibration to the other products can reduce the yield.77.80) /lj = Gj +RTlnal subject to Equation 3.71.] RTlnaj GnjO~l Revisit the two-reaction trimethylpentane example. We start with (3. The main restriction on these extents is. can be more informative than the algebraic approach discussed above.

+ £2 s . + 0 0 -YPIO-E.4 0.YIO + E. V. The constraints on the extents are found from Equation 3..45 0.56 -2. and the Gibbs energy contours depicted in figure 3. for a single reaction or 2.81) in wbich Vi 0. then determines the two equilibrium extents. .569 The solution is in good agreement with the extents computed using the algebraic approach. Vijll J j = 0.3 £2 = j Vi)" 0. the Gibbs energy was chosen as the convenient function for solving reaction equilibrium problems when the temperature and pressure are specified.81 as a function of the two reaction extents.6 Summary na J j ._.:.6 Summary and Notation 9S Substituting tbis relation into Equation 3.45 0.o - £2 s o 0.1 0._.6. -: .79.55 -2. G= -2.25 0.6: Gibbs energy contours for the pentane reactions as a oS E~. in which the equilibrium constant is defined to be . j ••• .15 0. We also rely heavily on optimization tools in Chapter 9 on parameter estimation. We see immediately that the minimum is unique.35 0.4 0. The numerical solution of the optimization problem is equilibrium condition is that the Gibbs energy is minimized.6 shows the lines of constaut Gibbs energy determined by Equation 3. 0 Optimization is a powerful tool for solving many types of engineering modeling and design problems. .J i.25 E'I 0.. -2 -I -.6 that G is defined oniy in the region specified by the constraints.35 -2.YEO + E.5 . = 0. 1.05 0 - o --- The minimization of tbis function of E.78 and rearranging gives 0.94 Review of Chemical Equilibrium 3.. Notice in Figore 3.5 0..- (3. £2 = 0. The fundamental for multiple reactious.2 0.5 0 Substituting in the initial conditions gives the constraints Figure 3.3 0.05 0. For this problem they are £2 s .133.1 0.2 0.15 0. sO -YP. 1'lr i = I Ki = 3..53 -2. nr In this review of reaction equilibrium. Figure 3. function of the two reaction extents.351. This fundamental condition leads to several conditions for equilibrium such as Vjllj=O j E.

k = 1. l = 1. and the following methods and approximations were presented: ideal mixture. and simple correlations for activity coefficients. . . partial molar heat capacity of component j Ith element constituting the species in the reaction netvvork fugacity ofComponent j h Gibhs energy G Gj Partial molar Gibbs energy of component j !lGi standard Gibbs energy change for reaction i enthalpy H Hj partial molar enthalpy of component j !lH. np nr ns P equilibrium constant equilibrium constant for reaction i element index. . If optimization software is available. Henry's law. because it depends on temperature via Notation aj activity of component j ajl formula number for element I in species j jth species in the reaction network iJT= iJ InK 6. In (K2) KI = _ 6. ny species index. The equilibrium "constant" is not so constant. .96 Review of Chemical Equilibrium 3.2..2. ne total number of reactions in reaction network 1 total number of species in reaction network pressure partial pressure of component j entropy temperature volume reaction extent reaction extent for reaction i cheuilcal potential of component j stoichiometric number for the jth species in the ith reaction stoichiometric number for the jth species in a single reaction Vj vi} . i = 1. Two approaches to this problem were presented. .. pressure and chemical potential of each species in all phases. standard enthalpy change for reaction i EI - i j k K Ki reaction index. we require numerical methods to find the equilibrium composition. The evaluation of chemical potentials of mixtures was discussed. which is the usual situation faced in applications.... When more than one reaction is considered..W RT2 Aj CPj or. . j ~ 1.2.6 Summary and Notation 97 You should feel free to use whichever formulation is most convenient for the problem.H- R (~ _ ~) T2 TI The conditions for phase equilibrium were presented: equalities of temperature.... the optimization approach is more powerful and provides more insight. if the enthalpy change does not vary vvith temperature.2. ns phase index. We either solve a set of nonlinear algebraic equations or solve a nonlinear optimization problem subject to constraints.

5.:0:'.2 trans ~ 2 butene butene (c) What is the equilibrium gas·phase composition at 2.0 atm and temperature of 500 K 0 23.1 Exercises 99 3.3: More tban one reaction Consider the following two gas-phase reactions A+ B = ~ C 2C D (a) What are the two equilibrium equations relating the two unknown ex· tents of reaction and the initial gas~phase compositions? You can as~ sume the pressure is low enough that the gases are ideaL (b) What is tbe equilibrium gas-phase composition at 2.33 20. ~ 2.11 kcal/mol.2 . isobutene hexane:.94 21. k. -:..2: Equilibrium and linearly independent reactions Butene isomerization reactions are shown below. What assumptions did you make? of the reactions.7 Exercises Components I-butene cis-2·burene trans-Z-burene isobutene 6Gf (keal/mol) Exercise 3.10 kcal/mol.1: Calculating an equilibrium composition Consider the follovving reaction in a dosed vessel at a pressure of 1.butene k 4 cis .62)? k. Exercise 3.10 21.:0:'. 1 .0 atm total pressure given initial gas-phase composition YA ~ YB ~ 0. Yc ~ YD ~ 0. isobutene k.2. At what pressure must you run this reaction at a temperature of 500 K to achieve this conversion? (c) At what temperature mnst yon run this reaction at a pressure of LO atm to achieve 97% conversion'! (d) What do you know about this system that prevents you from taking these answers too seriously? Exercise 3. k. (a) Determine the equilibrium composition for this system for an initia] equimolar mixture of isobutane and butene.5? Solve the two nonlinear equations numerically. 6W ~ -20. 3-trimethylpentane p (a) Determine the minimum number of reactions that are needed to calcu- late the equili15rium composition of the butenes.YB ~ YD ~ 0 (partial answer: YA ~ 0..4: Temperature effect on reaction equilibria Hexane can equilibrate to methyl-substituted pentanes and methylcylopen· tane according to k.1a7 Assume for economic reasons that you must increase the equilibrium conversion to 97%.23 isobutane + I-butene + = 2.0.0 atm total pressure given initial gas-phase composition YA ~ Yc ~ O. -:.5. (dl Why is isobutene the predominate butene formed? (b) What is the equilibrium conversion of the reactants in part 3. methylcydopentane + H2 k __ l The Gibbs energy of formation of the components at 400 K are [Ill hexane k<:' 2 methylpentane . What assumptions have you made? (e) Compute the equilibrium composition at 400 K and LO atm pressure starting from pure 1 ~butene.:::=. K.butene k .butene 1 _. If a computer is not available. cis . = k. and Kr ~ l. you can use trial and error (partial answer: YA ~03n Exercise 3. B (b) Set up the linear algebra problem to determine the equilibrium extents -' The standard Gibbs energy and enthalpy changes for this reaction at 500 K are 6Go ~ -4. cis ~ 2 ~ butene trans 2 . butene 1 ~ butene k :.98 Review of Chemical Equilibrium 3.

At this condition the hydrocarbons are gases. To be fair to the other energy fllilctions.100 Review of Chemical Equilibrium 3. the reaction is exothermic with Mi" ~ -10 kcal!mol. and integrate Equation 3. P is fixed. fln"j 1. G 1°O K (kcal/mol) 43. so we are concerned about the formation of a liqUid phase in the reactor. The Clausius-Clapeyron equation well represents the vapor pressure: of component C as a function of telnperature 2 T. what system function would be at an ex' tremum (maximum or minimum) if tbe following were specified.10 -31.83) ent extensive properties. it remains constant during the process. P. In addition to P what other thermodynamic property is fixed? The reactor is initially filled with an equimolar mixture of A and B. n.02 45.0 atm. molar property x ~ IXJnj j in which X is any extensive thermodynamic property. dStot 2: O.5: Other equilibrium conditions In the text. (a) Over what temperature range does the reactor contain a liqUid phase'! Exercise 3.34 in the text.6: Partial molar Gibbs energy and chemical potential (a) Derive Equation 3. what other thermodynamiC property is fixed? (f) The system is insulated and In PZ~ = c _ !1Hvap RT (3..76 methylcylopentane 2-methylpentane hydrogen -47.63 42.7 Exercises 101 Component n~hexane "H foOO K (kcal/mol) -46. fdG f[-sdT+vdP+ j)JjdnJ ] _ .7 at given T. Perform the integration and express G2 and nj2 in terms of Gj and njl and fl to complete the argument.5 kcal/mol. .10 kcal!mol. fJnj Figure 3.69 t. Consider first a system at equilibrium depicted in Figure 3. Provide a short derivation vertfying your result starting mth the second law. Components A and B are not very soluble in liquid C The heat of vaporization of component Cis "H"p .7: Two systems with identical intensive properties and differ- (e) The system is insulated and V is fixed.0 atm total pressure. . System 2 is speCified by the variables T. Consider a reversible change from system 1 to system 2..1 0 in which liHJoO K and "GJoo K are the heats of formation and the Gibbs free energy of formation of the compounds at 600 K from the elements. Also find the composition of the mixture at this temperature. for example. imagine slowly removing the partition depicted in Figure 3. In addition to V. Hj. along "'ith T and P. • (d) 5 and V are fixed. (h) T and V are fixed. (h) Show this same argument produces the follOWing result for any partial Exercise 3. The equilibrium state (h) If the reaction is endothermic with "11' . p. over what temperature range does the reactor contain a liqUid phase? . Because fJ j is an intensive variabJe.. (a) T and P are fixed. The equilibrium constant at T ~ 298 K is K ~ 8. Exerdse 3. respectively [17l. and value of the Clausius-Clapeyron constant is c _ 7. Find the temperature where equal amounts of rnethylcylopentane and 2-methylpentane fonn if pure hexane is allowed to equilibrate at 1. we focused our attention on minimizing Gibbs energy' to deter~ mine the chemical equilibrium at fixed temperature and pressure.. j = 1.82) (c) 5 and P are fixed . Hs< Now consider system 2 to he simply fl identical copies of system 1. .. and the system pressure is P ~ 1.S3. P.7.39 0 of system 2 therefore has the same temperature and pressure (intensive properties) as system 1 but the number of moles of each component (extensive property) changes by constant factor fl.7: Creating a second phase with reaction Consider the gas-phase reaction A+B = C (3. • Product C has a fairly low vapor pressure.

05.44 in the text for the Gibbs energy of reaction.8.0 atm. in species j.8: Vessel with semipermeable membrane and chemical re~ action 2A = S.krc~ Exercise 3. What is the equilibrium composition and pressure in the vessel if Pe ~ 5. The equilibrium constant at temperature To is K = 0. At 400 K. We expect these rate constants to depend on temperature because the frequency and energy of collisions increases with the mean velocity of the molecules which increases with temperature. IS 0. which is proportional to d given the typically large numbers of molecules. Using the result of Exercise 2. quantify the fraction of collisions that result in successful reactions. An isothermal rigid vessel Hutlally contains pure component B at pressure Po and temperature To_ l! is separated by a semipermeable membrane from a Source of pure component A at pressure Pe . l=l in which GEl is the Gibbs energy of element I and a jl is the formula number for element l. B mixture 2A = B pure A Figure 3. - =B I n. verify Equation 3. The reverse rates expression follows from assuming the bimolecular reverse reaction is proportional to the probability of collisions of two E molecules.10: Reconciling kinetics and thermodynamics Consider the gas-phase reaction A kr 2:: VijG]! = D. The net reaction rate is given by the difference of forward and reverse rates r = kfcA . Po = 1. (a) In order for this kinetic view to be in agreement with the results of ther· modynamics. kr = K4>A kJ 4>1 What can you conclude about the dependences of rate constants kJ and kr on the composition? (b) If A and B form an ideal mixture.11: Gibbs energy of formation Consider the Gibbs energy of formation of species j GJf = C.102 Review of Chemical Equilibrium 3. ajlCEI inside the vessel due to the presence of a catalyst.84) J.6 and CO is 0.kJ.7 Exercises 103 To P A -Pe A.8: The van't Hoff relation Methanol can be manufactured by the gas-phase reaction The forward and reverse rate constants.580 cal/moL Determine tbe temperature at which the equilibrium mole fraction of H2 . LiC' = -333 cal/mol and LiW = -22. what can you conclude about the de- Exercise 3. we suppose everyone is familiar with the classical rate expressions for the forward and reverse reactions 7j = kfcA. Anticipating the discussion of kinetics in Chapter 5.d = 28 k.4.5. . Tbe membrane is permeable to component A but not component E.8. No reaction occurs outside the vessel.9: Semipermeable membrane and chemical reaction Consider the membrane separation sho'An in Figure 3.Gi I J"d nr rts Vii 2:: ajtGi:t 1=1 Tte (3. Multiply this equation by the stoichiometric coefficient Vij and sum on j to produce Exercise 3. A kinetic view of the equilibrium condition is obtained by setting the net reaClion rate to zero. show that the rate constants also satisfy the restriction A batch reactor is charged with a Hz and CO mixture at LO atm and the pressure IS mamtamed at LO atm.0 atm? An ideal·gas mixture may be assumed. The system undergoes chemical reaction 2A pendence of kf' kr on the composition? Exercise 3. Y r = krc~ The forward rate expression follows from assuming that the unimolecular forvvard reaction is proportional to the concentration of A molecules. The initial mole fraction of Hz is 0.k r .

Gibhs himself used the follOwing style of argument [8. the following style of argument is used to derive the phase rule. . and that j?n+2-rr (3.rr (3. Gibbs's phase rule states (3. If yes.. .85. So. following the discussion by Feinberg [6J. Can reactions go to completion in a multipJe~phase system? If no. and (ill) infinitely many solutions. I'j.90) See Noll [lOJ and Feinberg [6J for further discussion of this issue.1) equations..75. how many linearly independent solutions does Equation 3.7 Exercises /05 Exercise 3.9 to show j = n + 2 . What is missing from this argument? Provide sets of two equations with two unknowns that have (iJ no solution.104 Review of Chemical Equilibrium 3. Slart with the usual set of variables dG = -SdT + VdP + IIl/Hj J Notice we cannot state a phase rule vvith chemical reaction if we have not defined Hnearly independent reactions. Because the rank can exceed neither the number of rows (rr) nor the number of columns (n + 2). Apply the set of equilibnum relations given in Equation 3.87.rank(M) (3.I mole fractions in each phase Tk . Call j the difference between the number of equations and the number of unknowns...87. P' 'x' J I:·: ~V2 n2 i I n' j n2 n nIT n dT dP dl'I SIT ~vrr nIT I dl'n [:1 (3. which applies to every phase. .. we know that j is nonnegative. .72. prove it.. P.. . show the equilibrium condition i=l. (b) Rather than reject the phase rule.71 and 3.87-96].87 admit? Use the result of Exercise 2. differentiate Equation 3..87) The phase equilibrium relations tell us the variables T.53. j = 1.. (til exactly one solution. . Therefore. Count these equations and show there are (n + 2)(rr . (d) Consider r linearly independent chemical reactions and add the reaction equilibrium conditions to Equation 3. What are the dimensions oflbe modified M matrix? Show the modified phase rule with reaction is imental evidence. we have the same number of equations and unknownsrand are supposed to conclude that the intensive variables of the equilibrium state are therefore uniquely detennined. . and show tbat this difference produces Equation 3. let's work a little harder to derive it. for which we have ahundant exper- in whichM is the rr x (n + 2) matrix appearing in Equation 3.88) j=l. the number of independently vari- able intensive properties is simply the number of linearly independent solutions to the set of linear equations listed in Equation 3. Specify the following intensive vanables to define tbe state of the system: temperature. and n .I) unknowns.86) Count these unknov-ns and show there are rr (n . and subtract to obtain the Gibbs-Duhem reIations 0= -SdT + VdP njdllj I j cannot be satisfied by any reaction going to completion and completely eliminating one of the components. . Let j be the number of independently variable intensive properties of the composite system not allowing chemical reactions. pp. G I')nj. n-l k=l. In particular. which is another nlotivation for defjrung linearly independent reactions in Chapter 2.13: Phase rule Consider a system at equilibrium with n components and IT pbases.85) (a) In many textbooks. if j = 0. Hint: consider Equations 3. so we can write the set of equations _VI ni Exercise 3.12: Reactions going to completion In a single~phase system.ny To obtain a relation among only intensive variables. provide an example. n are equal in all phases.89) (c) What assumption about equilibrium states is required before we can conclude rank(M) = IT and tberefore (3.87. pressure.

When is equality achieved in the Cauchy ioequality? We have therefore shmVIl H is a positive definite matrix. corresponding to Figure 3.14: Multiple reactions and ideal-gas equilibrium We would like to show that. Notice the same argument applies if we make the weaker assumption of an ideal mixture rather than an ideal gas.106 Review of Chemical Equilibrium 3.31. for every x '* 0 . fij) = I. Equation 3.1 Exercises 107 Exercise 3. Given the relation nj = njO + VijEi use the chain rule to show corresponding to Figure 3.92) (c) To show H is positive definite. To reduce notational complexity we consider T and P constant in all partial derivatives of this problem.2. The following argument extends the single-reaction discussion starting on page 70 to multiple reactions !I9. and its first derivative aG I anJ /1j' Taking the second derivative gives a matrix H with elements a2G a/1' Hjl"'" _ _ _ = _ J anion.91) (b) Using tbe chaio rule show (3. j . Assume we bave fir linearly independent reactions. H Hk t - _ a2G z: z: j VijR). to show the right-hand side is nonnegative. = (a) Use the ideal-gas chemical potential /1) Gj + RTln(PYj) and show Using the definition of mole fraction Yj = nj/nT. /1m. the equilibrium state of an ideal gas is unique. we must show that the following quadratic form is positive xTHx = 2.92 into the above equation and show alG GE2 >0 Next use the Cauchy inequality. 15. (d) We now find the second derivatives of Gwith respect to reaction extent. fir = 2. 7J.5 for H is positive definite if xTHx> 0. sbow aYj = 6 j . Consider the function G(T. sioce H is positive definite and the reactions are linearly independent. _ Yj anI ny nT (3. for multiple reactions.Vk' 1 OEiOEk is a positive definite matrix. E) has a positive second derivative in the: single reaction extent. P. we sbowed that the Gibbs energy G P.XjHjIXt > 0. show that fl is positive definite also. In the single (T. an. In matrix notation we have shown for every x '* 0 = fl = VHVT (e) Finally. reaction case. or Substitute Equation 3. A matrix Let H be the matrix of second derivatives in extents.2. Then we must show that the nr x nr mafrix of second derivatives.

]. namics. Anal.c Chemical eqUlltbflum ill [191 W. voL Ill. Cambridge. M. Butterworths. Ughtfoot. Gibbs. Prentice 118] J. 1969. [14] J. Conn. Dantzlg. and G. Boston. ModelL Thermodynamics and its Applications. [7] M. 1876. The Chemical Thermodynamics of Organic Compounds. 1984. Elementary Chemical Reactor Analysis. Soc. In The SCientific Papers of J. N. G.1958. Stoneham. 1999. .1989. [3] R.. Stull. D. Woodbridge. 2002. Upper Saddle River. [5] J. W. Bird. Prentice-Hall. Stewart. Uchtenthaler.. 1999. j . Prentice Hall. fourth edition. F. Rational Mech. M. B. Upper Saddle River. S. Introduction to Chemical Engineering Thermodynamics. [13J S. f2J E.J.. 200L tI71 D. New York. Shapiro and L S. E. 13(2):353-375. Shapley. The Mathematical Association of America 1961. 28(5). McGraw Hill. Phys. An Introduction to Inequalities. fIll B. Hall. N. third edition. editor. Z. Denbigb. Prau. Thermodynamics.. Models for Thermodynamic and Phase Equilibria Calculations. On the equilihrium of heterogeneous substances [Trans. Schieber and J.. Feinberg. [l6J J. CT. 1878. and M.. [6J M. J. Chemical and Process Thermodynamics. 70:219234. M. Prentice Hall. Math. Bellman. pp. 343-524. New York. Elliott and C T. Sinke. Arch. R. Wiley & Sons. second edition. McGraw·Hill. and E. MA. [8] J. and E. Ox Bow Press. 1997 .Bibliography 109 Bi bl iog raphy W Tester and M. The Principles of Chemical Equilibrium. volume one.751-705. pp. New Jersey. Transport Phenomena. L Sandier. pages 55-353. t h'lrd e d"mon. B. New Jersey. . C. Smith. B. New Jersey. 200L [12] J. C. ' Johnson . O'ConnelL Properties of Gases and LiqUids. Acad. G. Chern. R. Kyle. sixth edition. Jolm Wiley & Sons. On certain convex sets of measures and on phases of reacting mixtures. Mass action laws and the Gibbs free energy.mitz. Willard Gibbs. Introductory Chemical Engineering Thermody.. M. Molecular Thermodynamics of Fluid·Phase Equilibria.1979. Marcel Dekker. 1970. 38:1-12. Upper Saddle River. 1998. Beckenbach and R. de Azevedo. [9J B. Feinberg. R. P. '[ M Wh1 e. 1981. Abbott. New York. Rational Mecl!. Prausnitz. Van Ness. Aris. On Gibbs's phase rule. On the convexity of the ideal gas Gibbs free energy functiolL Personal communication. . 1965. de Pablo. Lira. Noll. j. complex mixtures. [1] R. and]. Eugle· wood Cliffs.. Anal. 1993. Indust Appl. and G. 1994. .]. Poling. E.. [4] K. E. [15J N. John. 108-248. New Jersey. W. 2002. G. If. Cambridge University Press.. [10] W. New York. Westrum Jr. Arch.

The statement of conservation of mass for this system takes the form.1 General Mole Balance Consider an arbitrary reactor volume element depicted in Figure 4. The molar concentrations of component j in the two streams are given by CjO and ejl and the production rate of component j due to chemical reactions is Rj . which has inlet and outlet streams with volumetric flowrates Qo and Q" respectively.4 The Material Balance for Chemical Reactors 4. rate of accumulation { of component j } rate of inflow } { rate of outflow } { of component j of component j rate of generation } of component j by { chemical reactions (4.1: Reactor volume element.1.1) ~ + \ I / v Qo CjO /" / Rj "- \ Figure 4. .

volume is constant or changes during the course of the reaction. . if the reactor is iso· thermal. One can. and the rate laws are assumed to be quite simple.5 applies whether the reactor volume element..2 The Batch Reactor The batch reactor is assumed to be well stirred. of course.6) Be sure to use Equation 4. the material balance for the batch reactor must be solved numerically.2 gives the general batch reactor design equation. which is sometimes a good approximation for liquid·phase reactions. we simply use some common reaction-rate expressions \Vithout derivation. then an· alytical solutions of the material balance are possible. even if one must compute a numer· ical solution for a problem of interest.1 as. Figure 4. The effect of parameter values on the solution is usually morc transparent. include volume elements with more than two flow streams by summing with the appropriate sign over all streams entering and leaving the reactor. which do not take place in any reactions.2. However. n s . and nothing is added or removed from the reactor until the stopping time. rio Therefore we require an expression for the reaction rates in terms of the concentrations of the species. and the careful study of analytical solutions can often provide insight that is hard to extract from numerical computations. the inflow and outflow stream flowrates are zero. analytical solutions prmide insight that is difficult to achieve with numerical solutions.112 The Material Balance for Chemical Reactors 4. the production rate can be computed directly from the stoichiometry and the reaction rates for all reactions.2 are simple to evaluate. :t Iv cjdV = QOCjO = QjCjl + Iv RjdV (4. These rate expressions may be regarded as empirical facts until the next cbapter when the theoretical development of the rate expressions is provided.5 to give _J de =R dt J = (4. If the reactor volume is constant. we can write Equation 4. Comparing a numerical solution procedure to an analytical solution for a simplified problem provides some assurance that the numerical procedure has been constructed correctly. VR call be divided out of both sides of Equation 4. In particular.2) Equation 4. Rate expressions.5 rather than Equation 4. Then the verified numerical procedure can . This topiC occupies the majority of Chapter 5.1 Analytical Solutions for Simple Rate laws 4.4) VR RjdV ~ RjVR In complex and realistic Situations. Notice also that we are assuming that component j enters and leaves the reactor volume element only by convection with the inflow and outflow streams.6 if the reactor vol· ume changes significantly during the course of the reaction. there will be two or fewer flow streams. Because the reactor is charged with reactants at t ~ 0.2 The Batch Reactor 113 In terms of the defined variables. In this case it is natural to consider the entire reactor contents to be the reactor volume element as in Figure 4. the solution procedure should be checked for errors by comparison to known solutions.2: Batch reactor Equation 4. .. we require an expression for the production rate. so there are no con· centration gradients anywhere in the reactor volume.3) (4.2 applies to every chemical component in the system. First. j = 1. For the balances in this chapter. Qo = Q 1 = 0.2. due to the closed form of the solution. Analytical solu· tions are valuable for at least two reasons. including inerts. Secondly. we are neglecting diffusional flux through the boundary of the volume element due to a concentration gradient. To solve the reactor material balance. Substituting these results into Equa· tion 4. 4. As shown in Chapter 2. and V = VR • Because the reactor is well stirred. Ri' for each component. the integrals in Equation 4. The diffusional flux will be considered during the development of the material balance for the packed·bed reactor.2. - f f VR c-dV = J CVR ) (4. For the purposes of illustrating the material balances in this chapter.

3 for several values of the rate constant k. log scale.1 k = 0.. Because the reaction is irreversible.. . 1 aa L~k:~~S~~__~~=±~=-~:L======~====~ 1 2 3 4 5 t Figure 43: First·order. Consider the first-order.2 k=2 ~ ~ 0.4.JL---o 1 234 S t Figure 4..4: Fjrst~order. As will be discussed in more detail in Chapter 9.8 requires an initial condition to have a unique solution.... irreversible kinetics in a batch reactor. the A concentration decreases exponentially from its initial value to zero with increasing time. irreversible reaction A 1<.. The rate constant determines the shape of this exponential decrease. t ~ a 1~----'------'-----'-----'r-----' The solution to the differential equation with this boundary condition is (4. The next several sections derive analytical solutions for some simple rate laws. Of course.114 The Material Balance for Chemical Reactors 4. Rearranging Equaflon 4.2 The Batch Reactor 1/5 be used with more assurance on the full problem for which no other solution is available. the batch reactor is assumed to be operating at constant temperature in this discuSSion. 0. Equation 4. first-order.9 and taking logarithms gives 0.01 k~S which is plotted by using a log scale in Figure 4.7) 0. B (4.L---L---J. irreversible kinetics in a batch reactor..---"'-. Application of the material balance for a constant-volume reactor gives the following differential equation (4. This procedure is sometimes recommended as a way to determine rate constants for first-order reactions by plotting experimental concentration data and determining this slope.8) in which the negative sign arises because the production rate of A is RA ~ -y due to the stoichiometry of the reaction..8 0. We denote the initial concentration of A in the reactor as CAD. this procedure is a poor way to determine a rate constant and should be viewed oniy as a rough approximation.001 L-----1-...5 k~l k~2 0..5 in which the reaction rate is given by y ~ kc A. Notice one can get an approximate value of the rate constant by calculating the slope of the straight line given by InC CA I CAO) versus t.9) which is plotted in Figure 4..6 k k 0. irreversible.

Consider now the same first·order reaction. but assume it is reversible (4. that the sum of concentrations A and B is a constant.+k. The value of this constant is known at t = 0. CAh = aexp( -(k.. Adding Equations 4. and Ll in Reaction 4. It is shupier to note. One particular solution to the equation is the constant solution CAp and solve this differential equation with the initial condition for B.14 for t = 0 gives CAO = a + k 1 Ll ~l (CAO + CEO ) k + and the reaction rate is r and B are now dCA = k.e-(k. One could write down the material balance for component E. + k_l (CAO + CBO) [1 .)t 1 (4. + k_tlt).. dCAh ---. in which a is an arbitrary constant to be determined from the initial condition. Notice that CA + CB remalns constant.8 and 4. ~- dCB dt = R B = k CA (4. c = ae-1k. (4. or more simply. by switching the roles of A and Band k.It + {kl + k ) CAh = 0 -I dt Therefore.13) The B concentration can be determined by substituting Equation 4. Writing Equation 4.11 into the material balance for A gives dCA Substituting in this value of a into Equation 4. so CB can be computed from Equation 4.L. The I tion can be written as the sum of what IS called the parncular salu~~~ and the solution to the homogeneous equation [2J. Substituting Equation 4. yielding (4. +k.12. which can be rearranged for the B concentration.11 into 4.cs. L. however.11) The solution to this equation already has appeared in the previous section.cB =r = ~- deB dt k. The material balances for A Solving this equation for a yields dt = -r = -k.116 The Material Balance for Chemical Reactors + A kl + L. CE(O) = CBO. reversible. after substituting the knmvn solution for CA(t).12) The constant a is nOW determined from the initial condition.10 gives d(CA+CB) =RA+RB=O = k k_~ 1 + (CAO -1 + CEO) You should substitute this back into Equation 4.10) Equation 4. +k.k_JCB with the same initial condition CA(O) = CAO. CA + cn is a constant and independent of time. The full solution to Equation 4.16) . (CAO + CBO) (4.. It + k-l k.14 and rearranging terms gives the final solution CA = cAOe-1k. linear differential equation.13 is then CA = CAh + CAp.13 to check that it is indeed a solution. The homogeneous equation refers to the differential equation with a zero forcing term on the right-hand side.CA .2 The Batch Reactor Tl7 The B concentration is easily determined from the A concentration.15 and rearranging.14) first-order.15) dt = -kICA + LJ (CAO + CBO - cAl which can be rearranged into (4. CB(O) = CBO.13 is a nonhomogeneous.cA + L.cA .

(4. Notice that because KI is larger than 1. values.5 for Kl = 110.----.118 The Material Balance for Chemical Reactors 4. Defining KI = kif k_l allows us to rewrite this as CAs - .+kt CAO 1 )-1 (4. Equation 4. Integrating both sides between the inItial condition and a final condition of interest gives Performing the integrals gives in which CAs is the steady-state concentration of A.4 appears on the left -hand side and only a function of time appears on the right-hand side.16 are plotted in Figure 4.. but certain ones do. Taking the limit t .----~----_. We next calculate the values of the steady-state concentrations. Consider the irreversible reaction A Jz.=-kt CAO 1 1 CA Finally solving for CA gives CA = 1 + KI 1 (CAD CB + CEO) = ( . DivIding hoth sides by c~ and putting the time differential on the right -hand side gives 1. CHO = O.18) 0. For small K. CAs = 1/3. the steady state would favor the reactant A..5. the secondorder reaction decays more slowly to zero than the first-order reaction.-----.2 The Batch Reactor 119 lr----.15 gives in which only a function of C. O.4 in which the rate expression is second order. .6 displays this solution and the first-order solution with a rate constant chosen such that the inItial rates are equal.5: First~order.15 and 4. the forward reaction is You should check that this solution does indeed satisfy the differential equation and inItial condition. CAD = 1. Equations 4.19) Performing the same calculation for CBs gives = 1 + KI KI (CAD + CBO) CSO = These results are shown in Figure 4. O. CBs = 213. Figure 4. B = kc~. kl = This is our first nonlinear differential equation. reversible kinetics in a batch reactor. the concentration of A does not go to zero as in the irreversible case.B favored and the steady state favors the product B. but goes to a nonzero steady-state value.6 C CBs 0. irreversible. k_l = 0. Nonlinear differential equations do not have analytical solutions in general. Notice that although hoth solutions have the same qualitative features.5 = 2 and CAO = 1. r halance and initial condition are dCA dt = -kc" 2 (4.17) The material 0.18 can be solved analytically because the equation is separable. Second-order. Notice that with the reversible reaction.5.00 in Equation 4.2 CB(t) OL-----L-____~____~_____ L_ _ _ _~ o 1 2 3 4 5 t Figure 4.

then from Equation 4. the steady-state solution is = -y = -kcAcB CBs dCB = 0. irreversible reaction. 4.a. Component A remains at steady state even though the reaction is irreversible. ' j Notice that if CAO = CEO. one can solve for the concentration in this case.CBOe(CBO~CAD)kt]~l CAO (4. B starts out in excess and all of the A is depleted.8 SUbstituting this expression into the material balance for A yields . OA 0.23. A+B The concentration of component C also can be computed from noticing that the material balance for C is dec by J:. but again the special structure of this problem allows an analytical solution.CA and therefore. A starts out in excess and all of the B is depleted.6 CA dt This equation also is separable and can be integrated to give (you should work through these steps). (4.24) first order 0 0 1 2 3 4 t Figure 4. CCs = CBO + Ceo = -y = -kcAcB In general we need to solve sets of nonlinear equations numerically. C dt = kCACB O.6: Second-order and first-order kinetics in a batch reactor CAO = I k = 1. Component B is present at steady state and the final amount of C is equal to its starting value plus all the A that is present inltially. Notice that we can subtract the material balance for B from the material balance for A to obtain d(CA -cel = 0 dt which implies that in this reaction.120 The Material Balance for Chemical Re.19.11.25) CA Consider another second·order. Component B can be computed from Equation 4.. The final amount of C is equal to its starting value plus all the B that is present inltially.CAO + C80) dCA 0. the steady-state solution is CCs = CAO + Ceo In this case.~ct. d(CA + cell dt Cc = The material balance for components = CAD + Ceo dCA dt dt Notice that if CAO > CBO. For CBO > CAO.CE is independent of time.2 second order CA = (CAO-CBO) [1.23 CA(t) = CB(t) for all t and this case reduces to the solution previously presented in Equa~ tion 4.20.2 The Batch Reactor 121 of B in terms of A. CA . Be' cause this value is known inltially.= -kCA(CA . or by switching the roles of A and B in Reaction 4. .24 as discussed in Exercise 4. The concentration of C is given by .23) 0.CAO) [1 _Cno e(CAO~CBO)kt]~l CAO (4. in which the rate law isy A and Bare = kCAC8. One also can obtaln that result by taking the Iinlit in Equation 4. giving CB = (CEO .

but care must be exercised in using this kinetic model when the concentrations are small.5 4 n ~ 3 3. should he examined carefully if the concentration reaches zero during model solution. models with order less than one.= [I CAD . .5 2 1 1 has units of inverse lime. Let the reaction rates be given hy simple first· order rate expressions in the corresponding reactants. n*1 in wbich Reactant A decomposes to form an intermediate B that can further react to form a final product C.5 3 Figure 4. Consider the following two irreverSible reactions l Performing the integration and sohing for CA CA gives ~ [cAg+! + (n-Ilkt]cb.8 sbows the be· havior of CA versus t for values of n greater than one. Notice the rate becomes unbounded as CA approaches zero.7 for a variety of positive n values. irreversible.2 The Batch Reactor 123 5 4. it is often necessary to modify the right·hand sides of the material balance as follows dCA nth-order.29) n *1 (4.9. t zero - n<1 n<:1 (4. the more slowly the A concentration approaches zero at large timc. This equation also is separable and can be rearranged to deA = -kdt Note that this amendment of the material balance is equivalent to pUlting a discontinuity in r(cA) at CA = in Figure 4. 0. and especially order less than zero.28) dt ~ -r ~ -kc. 0. The material balance for Reaction 4.26) B C (4...5 :3 r 2 2.\ .. Figure 4. Inhibition reactions are not uncommon.7: Reaction rate versus concentration fornth~order kinetics r = kc~. + (n -l)kotl~ j A~ B (4.17 "'fth nth-order reaction rate expression is dCA = { -kc. dt (4.5 1 2 2. so. ' Finally notice in Figure 4.26 yielding 1.10 that for n < 0.11.26 for n < L First notice in Figure 4.26 as discussed in Exercise 4. the rate decreases with increasing reactant concentration.5 1 _{l/koO-n). When using an ODE solver to compute solutions that can reach zero in finite time. The nth-order rate expression r ~ kc. n ~ O. Notice the larger the value of n.' .30) We can divide both sides by CAO to obtain c.5 a 0 0.27) 00. ° c. again. The n ~ 1 case already is given in Equa· tion 4.9 that CA reaches zero in finite time for n < L This time can be obtained by setting the term in brackets to zero in Equation 4.122 The Material Balance for Chemical Reactors 4.10.. Two reactions in series. Some care should be exercised when using Equation 4. is displayed in Figure 4.. the reactant inhibits the reaction. which is not physi· cally realistic. which also can be recovered by taking the appropriate limit in Equation 4.

The particular solution can be guessed to be of the form eBP = a2e-k.------------------------------~ The material balances for the three components are 0.5 1 2 t Figure 4.. r = kc~.t as before. l..2 The Batch Reactor I "J 1 5 0.: L kc~. n. --' = RA = -rJ dt n=2 1 dec . r kC~Ol = 1.k.9: Batch reactor with nth-order kinetics.2 1 2 o 0.6 CA eAO n =5 0.5 0 4 ----0. e'k.-2 3 0. r -= h ~.t o 0.10: Reaction rate versus concentration for nth-order kinet- ics..t due to the form of the forcing term ou the right-hand side.5 1 1.5 3 t Figure 4.=Rc dt = rz = k2CB The material balance for component A can be solved immediately to give CA = CAoe-k.8 0.8 0.5 4 3.5 1 -1/2 -J ~.124 The Material Balance for Chemical Reactors 4.(l n)l for n < L I t I The homogeneous solution is again eBh = a.2 1/2 I 1. n:s O.5 deB k C dt + 2 B = k ]cAOe .t.4 4 I I I 4.. ko = kc j = 1 note the concentration reaches zero In fimte 10 time t = I/(k. The material balance for B becomes 0. Substituting the particular solution into the differential equation and solving for a2 gives k.5 3 I I I r 2.6 dc. .5 2 0 0 1 2 3 5 0 2.8: Batch reactor with nth-order kinetics. ko Figure 4._ n=..5 2 1.

k2 ~ L Two reactions in parallel. and kl and k2 in Equation 4. Exercise 4. +k11t) (4.11 for an initial condition of pure reactant A Notice that A decays exponentially as before.32) Substituting cArt) into the material balance for B gives dCB _ dt - k c 1 AO e-(k. B forms at intermediate times and then disappears leaving only C as the final product. 0 1 2 3 4 5 t Figure 4. Therefore Cc can be computed from c OA 0. Consider next two parallel reactions of A to two different products.35) Finally. the rate constants determine which product is favored.31) Notice that the case of k1 ~ k2 can be handled by taking limits of Equation 4. +k.126 The Material Balance for Chemical Reactors 4.34) = ceo + CAD k k2 k 1 + 2 (1 - e -(k. +k2It) (4. (4.36) These results are plotted in Figure 4. Notice that because the two parallel reactions compete for the same reactant. component C can be determined from the condition that CA + CB + Cc is constant or by Switching the roles of Band C.3 L To determine the C concentration. Band C are displayed in Figure 4. A. notice from the material balances that d(cA + CB + celld! ~ 0. which implies CA + cn + ce is constant. Substituting a1 and 0. Large values of k1 / k2 favor the formation of component B compared to C and vice versa. ce The material balance for A can be solved directly to give (4. CBO ~ Cco ~ reactions in series in a batch (4.1 1: Two ftr5t~order CAD ~ 1.2 The Batch Reactor 127 so the particular solUtion is known. 0. . k] ~ 2.8 CA(t) edt) 0.35.33) This equation is now separable and can be integrated directly to give CB ~ CRO Assume the Fates of the two irreversible reactions are given by Yl k t CA and Yz ~ k2CA· The material balances for the components are + CAO k k1 k 1 + 2 (1 - e -(k.2 0 The concentrations of A.6 (4. Band C.12. The initial condition then determines all CB(O) ~ CBO ~ at + a2 Solving for a1 yields a1 ~ CEO a2 into the CB solution gives - 1 k 1cAO/(kz k t ).10 asks for the optimal time to quench the reaction to achieve the maximum amount of intermediate B.

3 The Continuous-Stirred· Tank Reactor (CSTR) 129 1 0. Pain! selectivity: The point (or instantaneous) selectivity is the ratio of the production rate of one component to the production rate of another component. However. We refer to this .on. This practice often confuses the issue and should be avoided. Yield: The yield of component . it is convenient to define selectivity. ns (4. k2 = 2... ~--~- - d (CjVR) ~··t·. Writing Equation 4. It is best to state the definition in the context of the problem being solved.i is the fraction of a reactant that is converted into component I The continuous-stirred-tank reactor (CSTR) is also a well-stirred reactor so there are no concentration gradients anywhere in the reactor volume..tents 10 be the reactor volume Q Qf element as in Figure 4. yield and conversion.. . j = 1.12: Two first-order reactions in parallel in a batch reactor.QCj + RjVR..128 The Material Balance for Chemical Reactors 4.. the relations in Equa/1 \ tions 43 and 4.6 c 4.5 3 t Figure 4.13.= QfCjf .37) If the reactor volume is constant and the volumetric f10wTates of the inflow and outflow streams are the same. Conversion has several definitions and conventions. . and \ I / V = VR. 0.13: CSTR volume element. Equation 4.3 The Continuous-Stirred-Tank Reactor (CSTR) o 0.4 apply to the CS1R also. When two or more reactions are in this kind of competiti.-. stirred.2 for this reactor gives. CBO = Ceo = 0. We again consider the entire reactor con. and conversion to quantify the efficiency of the reaction in forming a desired final product.38) in which e ~ VRIQf is called the mean residence time of the CSTR.R J . CAO = 1.5 1 1.13. tor is the flow streams shown in Figure 4. these terms are often invoked without careful definition in the hope that the meaning is clear from context. kl = 1. Since the reactor is well Cj Cjf . yield. Its flowrate is denoted Q.37 reduces to (4. There are several ways to define selectivity. The difference between the CSTR and batch reacFigure 4.. We denote the feed stream with flowrate Qf and component j concentration Cjf' The outflow stream is flowing out of a well-mixed reactor and is therefore assumed to be at the same concentration as the reactor. The reSidence time distribution will be discussed further in Chapter 8. Overall selectivity: The overall selectivity is the ratio of the amount of one component produced to the amount of another component produced.8 Conversion: Conversion is normally defined to be the fraction of a component that has been converted to products by the reaction network.5 2 2.

2B the feed concentration of A is C Af ~ 2 mOl/L. Starting with Equation 4. Consider the reaction of mmene hydroperoxide (CHP) to phenol and acetone (4.3 The Continuous-Stirred-Tank Reactor (CSTR) 131 balance as the constant-density case. Plot the concentration of A versus time for constant feed concentration cAf ~ 2 mol/L if the reactor is initially filled with feed so CAD ~ 2 mol/L CAD ~ Figure 4. 0.40) 1 + (0. which was solved earlier. We will see in Chapter 6 that the nonisothermal reactor behavior can be much more complex than that shown in Figure 4.kCA (4.2: Phenol production in a CSTR Next we consider a simple reactor-sizing problem.182 mOm One can divide Equation 4.41) 1 dt ~ B ( cAf (0) ~ CAO . liquid-phase reaction in an isothermal CSTR shown in Figure 4. The steady state of tbe CSTR is described by setting the time derivative in Equation 4.QCj J Qjcjf 2 mol/L (steady state) (4.1 mln 1 )(lOOmln) = .14 shows the transient solution for the two initial conditions 0 and CAD = 2 moVL Notice botb solutions converge to the same steady-state solution as time increases even though the starting conditions are quite different. Because the reaction is isothermal and takes place in the liquid phase.39 through by VR to obtain for the constantdensity case (steady state.14. constant density) Example 4. and the rate constant is k ~ 0.1 mln -1. I.13 A Parts 2 and 3. Exercise 4. (439) Substituting the production rate the steady-state concentration CAs RA ~ -kCA and solving for CA gives ~ 1 + kB CAf Substituting in the numerical values gives CAs ~ Conversion of reactant j is defined for a steady-state CSTR as follows x. 0 Solution Par! J.1: Reaching steady slate in a CSTR Consider a first-order.6 provides another convenient method for solving linear differential equations analytically. we find the analytical solution (4. It is often a good approximation for llquid-phase reactions.130 The Material Balance for Chemical Reactors 4. Plot the concentration of A versus lime for constant feed concentration cAf = 2 mol/L jf the reactor is initially filled with an inert so CAO ~ 0 mol/L 3. the residence time of the reactor is 0 ~ 100 mln. 2. _ QfCjf . The constant-density mass balance for component A is given by Equation 438 dCA CA (4.37 to zero.44) .42) Equation 4.CA ) .42 can be arranged in the form J<.43) given feed.4 I for component A Example 4. Find the steady-state concentration of A in the effluent for the a linear differential equation of the same form as Equation 4_13. we as some the fluid density is constant. Using either of these methods.

We then set Q = 0 in Equation 4. and the reaction rate is first order in the concentration of CHP and acid 15]. like the CSTR.12 ill 1)(0.9 m 3 jhr and k = 4.45 into the above equation gives XA= kO 1+kO 1.182 = =2'-. Normally there is no outflow stream. but allows a feed addition policy while the reaction takes place. Therefore we assume Q = QJ and Equation 4.45) The conversion of CHP (for Q = QJ) is 4. like the batch reactor. Compared to the batch reactor. the reaction rate is given by r = kcCHP.4 The Semi-Batch Reactor Figure 4.: k(l . J 4: Reaching steady state in a CSTR.41 applies.85) = 37 m 3 (4.37 to obtain for the semi-batch reactor (4.5 Volume Change Upon Reaction The preceding examples in the batch ~eactor and CSTR sections were solved under the simplifying assumption that the composition of the . Solution The reactants and products are liquids at 85'C and the volume change is neglected. RA = -kCA. which means this is a pseudo-first-order reaction.12 hr'l.132 The Material Balance for Chemical Reactors 4.5 :::l 1 CAO ~ Solving for 0 in terms of conversion gives :a§ <.:.-. The flowrate into the reactor is QJ ~ 26.:. the semi-batch reactor provides more complete use of the reactor volume in reactions such as polymerizations that convert from lower density to higher density during the course of the reaction. cA~cAI+RAO The semi-batch reactor is a cross between the batch reactor and CSTR..46) where A designates CHP.XA) QfxA Substituting in the known values gives the required CSTR volume o 10 20 30 40 50 60 70 80 VR = (26.XA) Substituting tbe relation V R = QJO and solving for VR gives 0. and solving for the CHP concentration gives CAl CA = One may choose to operate a semi-batch reactor to control the reaction rate or heat release during reaction by slowly adding one of the reactants in the feed stream.15) t (min) o 4. This liquid-phase reaction is conducted using a small amount of acid as a catalyst [12].4 The Semi-Batch Reactor /33 Substituting Equation 4. 1 + kO (4. The semi-batch reactor is initially charged with reactant. Find the reactor volume to achieve 85% conversion of CHP at steady state. For the reactor temperature (85·C) and feed conditions. Substituting in for the production rate of CHP.9 m 3 jhr)(O.J 0= 2 XA k(l .5 CAs ~ "' VR 0.

If we have a single-phase system at equilibrium. Substitution into Equation 4. This assumption is useful for ideal-liquid mixtures in which the pure components have similar densities. and is a correct statement for ideal-gas mixtures.50 is clearly a total mass balance. Pi' We know The term that conservation of mass in chemical reactions implies 2: j R jM) = 0 (see Example 2. Notice that because the total mass balance is a linear combination of the mole balances. the"molar concentrations of the components. If we assume an ideal mixture.47 by the molecular weight of species j and sum over all species we obtain.dt . We briefly review the situation when the density does not vary with composition and then treat the general case.50) in which Vj is the partial molar volume of component j in the mixture. k.. (4.47) Equation 4. is known if the Cj are known. CSTR and semi-batch reactors. This relationship is one form of the equation of state for the mixture (4. P and c}.47 covers the batch. Therefore. The total mass balance is the starting point for the derivation of governing equations for numerous special situations in which the density and the volume of the reactor change.49. ns (4. which we denote P P Nonce that one should always replace a high-concentration species with the mass balance so that catastrophic loss of precision does not occur in the subtraction to obtaln Ck' . In this section we establisb the appropriate modeling equatious for the general case. in which the total mass in the reactor changes in time due to the inflow and outflow of mass. (4. Notice that chemical reactions play no role in the total mass balance. this reduces to ideal mixture in which Vi is the specific volnme of pure component j.37./34 The Material Balance for Chemical Reactors 4.52 shows that the equation of state is a single constraining relationship on the set of possible mixture concentrations. In this chapter we consider the temperature and the pressure to be known. We may some· times prefer to include the mass balance in place of one of the mole balances. depending on the physical properties of the reaction mixture and the manner in which the reactor is operated. which is an intensive variable.= QjPt" .48 leads to Ear example. If we multiply Equation 4. . it is not independent information. depending on how we specify Qi and Q. j ~ k.5). Equation 4. and the temperature and the pressure completely specify all intensive variables of the system. The partial molar volumes are functions of T.52) = L cjMj j=l n.49) Substituting the definition of density. it is convenient to express the equation of state as (4.48) The term 2:) etMj is the mass density of the reactor contents. The concentration of the kth species is . d(pVR) (4. the density of the reaction mixture. We start with the material balance for a well-mixed reactor.5 Volume Change Upon Reaction 135 reaction mixture has no effect on the density of the mixture.Qp . Equation 4. j then obtained by subtraction (4. .51) = 1. Equation 4.. ".53) CjiMj is the mass density of the feedstream. Because T and P are fixed constants. into Equation 4. We shall assume that a thermodynamic equation of state is valid even when the reactor is not at equilibrium. which is a function of oniy T and P. to ensure that numerical roundoff does not cause small violations of the mass balance.2. fixed quantities.. we could express the equation of state in terms of the partial molar volumes as } .

55 then determines the change in volume of the reactor during the filling operation. The constant-mass react0J:. For the batch reactor. In the general case. are specified. Q. dVR dt - dt <) . Qj." This terminology should be avoided because one cannot write a conservation statement for volume.53. Q = Qf = O. The difference in pressure between the top and bottom of the reactor is determined by ph in which h is the height difference. Because T. • Batch reactor. not molar concentrations. In the nonconstant density case.P are regarded as known. we begin by taking the differential of the equation of state f(cj) = 0 to obtain df dt = " of dc} = " fdej = L oc) dt J L ) dt J 0 (4. P. then we can conclude from Equation 4. Equation 4. p. We assume throughout that the feed conditions. VR. All three cases enumerated above can be handled by modifying the residual equations provided to the DAE solver.56) in which f . Howrate out of the reactor is specified.55) which is sometimes referred to as a "volume balance.55 that Q = Qj. In this case Q(t) is specified. Because the mixture density. If the outflow of the CSTR is regulated so that the CSTR has constant volume. (4. In this Situation.47. 3. Constant ·volume reactor. the set of unknowns is nj. The differential equation for the reactor volume is a consequence of the mass balance and the simplifying assumption of constant-mixture denSity. is independent of composition. In this situation pVR is speCified as the additional equation.of Now differentiating the product term Ii ) . YR. The calculations for the nonconstant-density case may be greatly simplified by using a differential-algebraic equation (DAE) solver. We can therefore conclude from Equation 4.54) p =Pj The total mass balance then reduces to (4. Equation 4.136 The Material Balance for Chemical Reactors 4. nj. If the cross-sectional area is independent of height. Equations 4. as required in the reaction rate expressions and the equation of state. fixed constants in this chapter. 1.>:an be achieved if a differential pressure measurement is used to control the flowrate out of the reactor and the reactor has constant cross-sectional area. In the semi-batch reactor. then the pressure difference is also proportional to pVR. VR is speCified as the additional equation. 2.55 that a batch reactor with constant density has constant volume. So if the flow controller maintains constant pressure difference. we switch to nj from Cj because the material balance is fundamentally a balance on total moles in the reactor. Constant-mass reactor. This type of operation may be achieved if the flowrate out of the reactor is controlled by a flow controller. If we know the n j and VR. The constant -volume reactor can be achieved by allowing overflow of the reactor to determine flowrate out of the reactor. The three reactor types we have considered thus far are summarized below. We do not have to differentiate the equation of state or perform other algebraic manipulations that are required if one uses an ordinary differential equation (ODE) solver. to produce a differential equation for the reactor volume. the total mass in the reactor. the reactor is filled during operation so Qj is specified and positive for some time and Q = O. • Semi-batch reactor. The final equation is provided by a statement of reactor operation. The equation of state. We have the Us equations from the component mole balances. To specify the right-hand sides of the material balances. Nonconstant density. we also require the value of Q. Cjf. then the values of the Cj are known.oc) (CiVR) gives the relation dt V dei _ d(CjVR) _ '. • CSTR (dynamic and steady state). In the general case. we can consider the following variables to fully determine the state of the reactor: T.5 Volume Change Upon Reaction 137 Constant density. it does not vary with time either and we can set it to the feed value. We consider here three important cases of reactor operations. The solution to Equation 4.47. We therefore have ns + 2 unknowns to determine and require ns + 2 equations. A semi-batch reactor is operated in this way "ith Q = 0 until the reactor is filled with the reactants. provides one additional equation. it also maintains constant mass.

" "0 ~~ "....0 '" V ~ 00 U3 >. ..:" r:::J r:::J - "~ "''" " . Tbe remaining two equations for the constant-mass.2. ~""> ~~ :.." " "0 N + £" '" ''"" ... the change in specific volume upon reaction i...: . .i? -~ > <l W . Ii = alloc. Depending on the form of the equation of state.JjCj s:-. For example.~ fA! i:j " ""' " r:::J r:::J " Po '" CI".. We have TIs + 2 unknowns and the component balances at the top of Table 4. ~ "-. and "'Ii = L'. We can either specify algebraic equations to obtain DAEs for the model.:: 0 fA! + ~ .:: :..~ <tl ""' 2 which is a change in a derivative property upon reaction. ~~ " . these two extra equations are either differential equations or merely specify the values of constants.1..VR dt f 2.:: . in which the products are denser than the reactants.. The constant-density and ideal-mixture cases for the three reactor operations are summarized in Table 4.138 The Material Balance for Chemical Reactors 4.'" ..:~ .~ fA! II + ""' "" u " . r:::J " ".j ]jc)1 _ Q + 2.:: " ..:: S N ~~ r:::J M " Q I I "0 '" '" OJ .:: " . "I fA! + r:::J I '-..JjCj Expressing the production rate in terms of the reaction rate yields dVR = Q 2. The general equation of state case for the three reactor operations is summarized in Table 4... " ..~ d(CjVR) " dVR Llr--dt .. ..:" (4.JjCj 2...§ '" :g ~ I ::<1 I 0. constant-volume and specified-outflow operations are provided lower in the table.. But for the remaining two equations. r:::J Ij Vij]j = Ij Vij OCj 01 . r:::J v " ..j]j (QICjf-QCj + RjVR) dt = 2.:: ::.~ '" "0 ~ " "- -...... The ns component balances are the same as in Table 4. .... Vt... . as shown in Equation 4..5 Volume Change Upon Reaction 139 Multiplying by ]j and summing gives ~" dCj VR Llj dt j = .1. . ]j = V....:" ::._ ' r:::J .1 provide TIs equations.~ .:: :.: r:::J .~ ~ " . we have a choice.r.." <l ....Lljcjdt = 0 j j . " > <1 ::....:: .57) "" + " I -~ . r:::J ~ r:::J '0 "" Q.) OJ ''"" ro cO " r:::J ~ 0 " r:::J '" ::'" r:::J r:::J I '" Po '" ~ '" S :S1 ~ 8 '"0 ~ ::0 '-- 0 '" 0 0 OJ 'cr " Po U '" "" CJ '" - '" '" 0 -.. the pure component speCific volumes..: - ... In these cases.i "'1.. > " .. if an ideal mixture is assumed for the equation of state.. or differentiate the algebraic relations to obtain ODEs.... DAEs may be preferred because the ODEs require the partial derivatives of the equation of state.. we set up and solve a nonconstantdensity case. these derivatives may be complex expressions. In the following example.~ in which "'Ii is defined to be "'Ii = r:::J fA! B ..: ~ in which we have used Equation 4.57..56~ Substituting the material balance into this equation and solving for the rate of change of reactor volume gives dVR 2. " ..~ '" r:::J " ~.

and the percentage increase in polymer production achieved in the second operation. N ~ C! II ~ .c + C! I ~~ <l ..1 min-I. C! Q '" 0 <to .kM = (4. OJ dt = QjOCMj . -".-". . II 0 0 II dt = QjOCMj - kCMVR II N + E ". "' C! C! M M(O) a (4. You may assume an ideal mixture. and the rate constant is k = 0.~ -i£ <J <J N tor to half its total volume. 4-..5 Volume Change Upon Reaction 141 Example 4."'" ' .:.~ N if? ~ ~N Calculate the total polymer mass production..58) 1j 0 0- '0 C) C ! i1 0 OJ 0 '0 .J Consider a solution polymerization reaction.~ C! H II Cl.. irreversible reaction MJ<.~ + ..= QfO + LlVkM VR(O) = 10 m' in winch .. <1- . The monomer feed is adjusted to keep the reactor filled while the reaction goes to completion.3: Semi-batch polymerization ~ <l '""' ~~ N " .'...2: '0 u OJ in a.: II '0 OJ A 20 m 3 semi-batch reactor is initially charged with solvent and initia- p. ~ ". Solution While the reactor is filling..~ - ~ .N ~2 + C! I 2.. A pure monomer feed is ~wly added at f10wrate QjO = 1 m 3 /min to fill the reactor in semi-batch operation to control the heat release. and can wTite the monomer balance as dM <to . The monomer feed is shut off and the reaction goes to completion. 0 Q ~ '" ~ II 0 -'0 4-. '. ::t: v .~ .: v I . We denote the total number of moles of monomer by M = CMVR. -.:: M - . the densities of monomer and polymer are PM = 800 kg/m3 . C! "' C! . '" '" 0J '" S dt ..:l ~ ~ Cl.140 The Material Balance for Chemical Reactors 4.. the monomer mole balance is d(CMVR) C! II '- 'C! II ~ ~ 0 Ui 0 II t. " "iil cO -'" ~ I~ I o ~ <to . " ~ 0 Cl. the volume is given by dVR 0 :.."'- 4-. in which CMj = PM I MM is given.:l ~ "" 'u OJ 0.59) '" '" 0- For an ideal mixture.:: ~ - ? H II ~ ~.'if.. II ~ <l ~ " C! II 'w C! " "" OJ . which can be modeled as a first-order.. ~ ~ . and Qj = QjO is constant duriog the filling operation.::.• . pp = 1100 kg/m 3 The monomer molecular weight is MM = 100 kg!kmol.P Cl.'" g 4-.:: C! "~ II ~ c. ~ -: + "~ ~-: NH 0 II " ... " ~ u II ~ II 0 . 1... Consider two cases for the subsequent reactor operation.

66) (4.60 and integration gives VR(t) ~ VR(ttl +L'. After t I we consider the two operations..67 to obtain t . we note from the stoichiometry that the mass production rate of polymer Rp is and substitution into Equations 4.13 and can be solved in the same manner. the volumes are additive.60) Notice the reactor volume decreases after II because L'.VM(tl) Rp ~ -RMMM so the mass balance for total polymer P is given by dP dt ~ RpVR ~ kCMMMVR ~ (kMM)M (l_e.58 and 4.60 and integrating F(t) =P(ttl+ MMM(t!l [ ( )] (l+L'. In the first operation. We can perform an independent.60 and integrating gives Substitution of this flowrate into Equation 4. P(t) ~ F(ttl + MMM(ttl (1 .VCMf)kM which can be solved to give (4.VcMf) l~exp -(l+L'. VR(ttl ~ 20 m 3 • Note we have a single nonlinear equation that must be solved numerically.e-kU-trl) . Operation 1.18.59 and 4.18. (4.18.e. Notice the final polymer production is larger in Operation 2 because of the extra monomer addition. II .5 Volume Change Upon Reaction 143 To compute the polymer mass. so we take the limit as t -. These results also are plotted in Figures 4. t (4.58 is of the same form as Equation 4. Substituting in the numerical values and solving for tl yields II = 11. Operation 2.V is negative..67) I (4.61.VkM Substituting this result into Equations 4.2 min Note the reactor would have filled in 10 min if the density were constant.VcMf)k(t-tl) .58 gives dM dt = -(1 + L'. Therefore in which initial condition MUll is given by substituting II in Equation 4. 10 m 3 of polymer are produced because in an ideal mixture. Substituting this value of Qf into Equations 4. Qf ~ 0 after tl.62) M(t) ~ M(ttl exp (-(1 + L'.kt ) . we have no simple calculation.61) = -6.64) (4. simple calculation of the total polymer in Operation 2. Equation 4.65) The polymer mass can be calculated by substituting this result into Equation 4.00 in Equation 4. Because the reactor volume is constanr. tl (4.k Ji )."we can solve Equation 4.V CMf )k(t ~ tIl) .142 The Material Balance for Chemical Reactors 4.65 and 4. In Operation 2.15-4. and VR(tI) ~ 20 m 3 We solve Equation 4.15-4..64 to obtain t1 S in good agreement with the loug-time solution for Operation 2 in Figure 4. yielding M(t) = Qf~Mf (1 .62 for the time II when the reactor fills.59 gives (4.63) We can solve Equation 4.59 for the feed flowrate during the secondary monomer addition Qf Let tl be the time that the reactor fills. The extra time reflects the available volume created by converting some of the monomer to polymer during filling. For Operation I. t . t tl . These results are plotted in Figures 4. (t-t tl'.

. time (min) Figure 4.8 - 10000 2 --- '2 M g " cJ 0.2 0 . 20 . .. 35 I I 40 45 50 0 5 10 15 time (min) Figure 4. . . 20 25 30 time (min) 35 40 45 50 Figure 4. <:l . 0. 15 1 .18: Semi-batch reactor polymer content for primary monomer addition (Operation 1) and primary plus secondary monomer additions (Operation 2). .15: Semi-batch reactor volume for primary monomer addition (Operation 1) and primary plus secondary monomer additions (Operation 2). . 5 . 1 ... . .5 Volume Change Upon Reaction 145 20 2 1 6000 5000 15 e ~ 0 0 be 4000 3000 2000 1000 ..- 6000 4000 2000 0 / / /- '" 0 -0.4 0. 12000 11------.. 10 .16: Semi-batch reactor feed fiowrate for primary monomer addition (Operation 1) and primary plus secondary monomer additions (Operation 2). . .17: Semi-batch reactor monomer content for primary monomer addition (Operation 1) and primary plus secondary monomer additions (Operation 2). . 2 be <lJ 8000 / / / / ~ 0 ---.2 5 S .6 0. o o 0 5 --0 5 10 15 20 25 30 35 40 45 50 10 15 20 25 30 35 40 45 50 time (min) Figure 4... 25 30 . . .".2 § - e .. .144 The Material Balance for Chemical Reactors 4.


The Material Balance for Chemical Reactors

4.6 The Plug-Flow Reactor (PFR)


Both P(tll and M(tJ) can be evaluated by substituting tl ~ 11.2 min in Equations 4.61 and 4.63, respectively. The results are
p(tj} ~ 3582 kg,

M(tll ~ 53.95 kInol,

PI ~ 8977 kg


Note the value j\ ~ 8977 kg is in good agreement with the long-time solution for Operation 1 in Figure 4.18. Finally, the increase in production rate is

h _ PI

x 100%

~ 22.5%

By using the volume of the reactor more efficiently, the total polymer production increases 22.5%. 0


\ I /
l'.V -

/ I \

Rj '-



The Plug-Flow Reactor (PFR)


z +!::..z

Plug flow in a tube is an ideal-flow assumption in which the fluid is well mixed in the radial and angular directions. The fluid velocity is assumed to be a fnnction of only the axial position in the tube. Plug flow is often used to approximate fluid flow in tubes at high Reynolds number. The turbulent flow mixes the fluid in the radial and angular directions. Also in turbulent flow, the velocity profile is expected to be reasonably flat in the radial direction except near the tube wall. 4.6.1 Thin Disk Volume Element

Figure 4.19: Plug-flow reactor volume element.

If the tube has constant cross section, A c , then velocity, v, is related to volumetriC flowrate by v ~ Q/ A c , and axial length is related to tube
volume by z ~ V / A c , so Equation 4.69 can he rearranged to the familiar form [1, p.584J

Because of the plug-flow assumption, it is natural to take a thin disk for the reactor volume element as shown in Figure 4.19. The concentration does not change over the volume element because there is complete mixing in the radial and angular directions and because the axial distance l'.z is small. The element has volume l'. V ~ Acl'.z in wbich Ac is the tube cross-sectional area. Writing Equation 4.2 in this situation gives

4.6.2 Steady-State Operation

One is often interested in the steady-state operation of a plug-flow reactor. Setting the time derivative in Equation 4.69 to zero gives,

a (e,l'.v)


. I ' ~c,QI z -cJQ. z + n z ' +R,l'.V .

The product CjQ ~ Nj is the total molar flow of component j. One also can express the PFR mole balance in terms of the molar flow,


Dividing the above equation by l'. V and taking the limit as l'. V goes to zero yields,



~ Rj



4.6.3 Volumetric Flowrale for Gas-Phase Reactions To use Equation 4.72 for designing a gas-phase reactor, one has to be able to relate the volumetric Ilowrate, Q, to the molar flows, N j, j ~





The Material Balance for Chemical Reactors

4.6 The Plug·Flow Reactor (PFR)


1,2, .. , ns. The important piece of information tying these quantities together is, again, the equation of state for the reaction mixture. The equation of state is a function relating the system temperature, pres· sure, and all molar concentrations, f(T,P,c}) ~ O. Because the molar flow and concentration are directly related,

in which Mf is the feed mass !lowTate. The total mass balance indicates that the total mass flow is constant with respect to axial position in the PER because chemical reaction cannot alter the total mass flow. We can solve for the volumetric /lowrate by rearranging Equation 4.76

Q~ Mf


the equation of state is also a relation between temperature, pressure, molar flows, and volumetric f1owrate. For example, the ideal-gas equa· tion of state, C = P jRT, can he stated in terms of molar concentrations, ej, as

If the liquid density is considered constant, p

Pr, then_ .

constant density


P = RT

In terms of molar /lows, the equation of state is
LjN! P

and the volumetric f10wrate is constant and equal to the feed value. Equation 4.78 is used often for liquid-phase reactions. If the density variation is significant, however, Equation 4.77 should be used instead . Finally, if we denote the time spent in the tube by e ~ VI Q, if Q is constant, we can rewrite Equation 4.71 as



de d~ ~ Rj ,

constant flowrate


One can solve the previous equation for the volumetric flowrate,

Equation 4.74 is used with the mole balance, Equation 4.72, to solve gas-phase PFR problems under the ideal·gas assumption. To evaluate the concentrations for use with the reaction rate expressions, one then rearranges Equation 4.73 to obtain

which is identical to the constant-volume batch reactor, Equation 4.6. in other words, for the constant·flowrate case, the steady-state profile in a PFR starting from a given feed condition is also the transient profile in a hatch reactor starting from the equivalent initial condition.
4.6.5 Single Reaction Systems

Example 4.4: Changing flowrate in a PFR A pure vapor stream of A is decomposed in a PFR to form Band C

Nj P Nj ~ Q- ~ RT LjN j



Volumetric Flowrate for liquid·Phase Reactions

AJ<. B+C


Consider the equation of state for a liquid·phase system to be arranged in the form

= f(T,P,c})

The mass density is related to the volumetric f]owrate and total mass flow, M ~ NjM}, via
M =pQ

Determine the length of 2.5 em inner-diameter tube required to achieve 35% conversion of A. The reactor temperature is 518°C and the pressure is 2.0 atm. Assume the pressure drop is negligihle. The reaction rate is first order in A, k = 0.05 s-] at the reactor temperature, and the feed f10wrate is 35 L/min. Solution Writing the mole balance for component A gives
dN'l = RA dV

Multiplying Equation 4.72 by M, and summing on.f produces




M(O) =



The Material Balance for Chemical Reactors

4.6 The Plug-Flow Reactor (PFR)


The production rate of A is RA = -r = -keA. Substituting the production rate into the above equation gives, (4.81) The volumetric flowrate is not constant, so we use Equation 4.74, whicb assumes an ideal-gas equation of state,

The previous differential equation can be separated and integrated,

Performing the integration gives,
= -



(NA + NB + NcJ



kP V

The ideal-gas assumption is reasonable at this reactor temperature and pressure. One can relate the molar flows of B and C to A using the reaction stoichiometry. Tbe mole balances for Band Care
dNB dV

The conversion of component j for a plug-flow reactor operating at steady state is defined as
x' - Nf N J J J Nf

= RB = r,

- - =Rc=r

dNc dV

Adding the mole balance for A to those of Band C gives
d(NA+NB) =0 d(NA+Ncl =0




The stoichiometry does not allow the molar flow NA + NB or NA + Nc to change with position in the tube. Because both of these quantities are known at the tube entrance, one can relate NB and Nc to N4,

Because we are interested in the V corresponding to 35% conversion of A, we substitute NA = (] - xA)NAf into the previous equation and solve for V, RT V = - kpNAf[2In(1-xA) +xA1 Because Qf = NAfRT IP is given in the problem statement and the tube length is desired, it is convenient to rearrange the previous equation to obtain

NA +Nc =NAf +NcJ



Qf [21nO

x,d + xAl

Rearranging the previous equations gives,

Substituting in the knmvn values gives

= NAf + NEf

- NA

NAf + NCf - NA

= - ( 0.05

4) " 35 x ]0 cm /miu) ( S-l 60 s/min 17(2.5 cm)2 [2In(l - .35) + .301



Substituting the relations for NB and Nc into Equation 4.82 gives

z = 1216 cm = 12.2 m o
4.6.6 Multiple-Reaction Systems



(21VAf + NBf + NCf- NA)

Because the feed stream is pure A, NEf = Nef = 0, yielding

= RT

(2NAf - NA)

Substituting this expression in Equation 4.81 gives the final mole balanee,

The modeler has some freedom in setting up the material balances for a plug-flow reactor with several reactions. The most straightforward method is to write the material balance relation for every component,


The Material Balance for Chemical Reactors

4.6 The Plug-Flow Reactor (PFR)


Recall the production rate of each species is a simple linear combination of the reaction rates times the appropriate stoichiometric coefficients,

Solution The mole balances for the four components follow from the stoichiometry,




j ; ; ; ; 1,2, .. " ns

The reaction rates are expressed in terms of the species concentrations. The Cj are calculated from the molar flows with Equation 4.73, and Q is calculated from Equation 4.74, if an ideal-gas mlxture is assumed. Example 4.5: Benzene pyrolysis in a PER Hougen and Watson [9J analyzed the rate data for the pyrolysis of benzene by the following two reactions. Diphenyl is produced by the dehydrogenation of benzene,

The initial condition forthe ODEs are NB(O) ~ NBf andND(O) = NN(O) ~ NT(O) = O. Because tbe number of moles are conserved in both reactions, the total molar flux does not change "1th reactor volume. The volumetric flowrate can be evaluated from Equation 4.74, which in this case reduces to,

Triphenyl is formed by the secondary reaction,

The rate expressions are substituted into the four ODEs and they are solved numerically. The total conversion of benzene, XB = (N Bf NBl/NBf' is plotted versus reactor volume in Figure 4.20. A reactor volume of 403 L is required to reach 50% conversion. The composition of the reactor versus reactor volume is plotted in Figure 4.2 L 0 In the previous approach, a differential equation is required for each component. Therefore n, differential equations are solved simultaneously in this approach. One can instead solve the problem by using One differential equation for each independent reaction for a lOtal of n, differential equations. Because n, < ns, this second approach produces fewer differential equations. Normally n, is not much less than n s , and the computational expenses of the two approaches are simllar. Whether we wTite differential equations for the speCies or the reaction extents is largely a matter of taste. In any case, the following procedure is guaranteed to give the mlnimal set of differential equations. Instead ofwTiting the material balance for eacb component, define extents for each reaction that satisfy

The reactions are assumed to be elementary so that the rate expressions are

in which the subscripts, B, D, T and H represent benzene, diphenyl, triphenyl and hydrogen, respectively. Calculate the tube volume required to reach 50% total conversion of the benzene for a 60 kmol/hr feed stream of pure benzene. The reactor operates at 1033 K and LO atm. Plot the mole fractions of the four components versus reactor volume. The rate and eqUilibrium constants at T ~ t033 K and P = LO atm are given in Hougen and Watson,
k, = 7

10 5 Llmol . hr

K, = (1.31

k, = 4

X 105

Llmol . hr


The initial conditions for these extents can be defined to be zero,



The Material Balance for Chemical Reactors

4.6 The Plug-Flow Reactor (PFR)



The molar flow for each species can be computed by summing the stoichIometric number tImes each extent,


~ Nif +






As with the previous method, the Cj are calculated from the molar flows with Equation 4.73, and Q is calculated from Equationil.74. Example 4.6: Benzene pyrolysis, revisited Rework the previous example using the minimum number of differential equations. Solution






V (L)





Defining the extents to track the rates of the two reactions in Equations 4.83 and 4.84 gives

Figure 4.20: Benzene conversion versus reactor volume.


(4.87) (4.88)


0.8 -

The stoichIometry allows all molar flows to be calculated from these two extents,

0.6 YJ
0.1 -



+ 2EI +

E2 ~ NBf
~ NDf

NH - £1 - £2 ~ NHf


+ E2


E2 ~ NT!

Substituting in the feed conditions gives,



NB ~ NBf - 2EI - E2

NH = Cl
NT ~ £2

+ £2


= [1 ~ £2









V (L)





These relations can be used along "ith Equation 4.86 and Equations 4.85 to evaluate the rates Yl and Y2 in Equations 4_87 and 4.88 in terms of £1 and £2. In this approach one only has 10 solve two differential equations rather than the four required in the previoos approach. The solutions are the same in the two approaches. 0

Figure 4.21. Component mole fractions versus reactor volume.


The Material Balance for Chemical Reactors

4.6 The Plug-Flow Reactor (PFR)


Example 4.7: Ethane pyrolysis in the presence of NO The thermal decomposition of ethane is inhibited by nitric oxide. Laidler and Wojciecbowski (l0] have suggested the following sequence of elementary reactions to describe pyrolysis.

change in temperature creates such a large effect on the amount of ethane reacting?

The reaction rates are functions of molar concentrations of the components, and the concentrations can be found from the molar flowrates of each component _

~ C2HS + HNO


d + CZH6 -

C2H, +






We need to know all the molar flows, and they can be found from the PFR mass balance dN _ _ = Rj = 1 VijYi dV ,.


The kinetic data for lhis reaction are, where k i Reaction

~ Ai exp(-E;/RT)

which produces for this example






2 3 4 5 6

Yz ~ k2CC,Hs

= k3CHCC2H6

1.0 X 10!4 cm 3 /mol s 3.0 x 10 14 S-1 3,4 x 10 12 cm 3 /mol s

r4 = k4CHCNO

LO x 10 12 cm 3 /mol s
LO x 1013 5- 1 LO X 10 12 cm 3 /mol s

rs = kSCHNO Y6 = ~6CC2HsCHNO

(kJ/mo!) 217.6 165.3 28.5 0 200.8 0



-Yl - Y3
-Y! - r4

+ Y6 + rs + r6


dNczHs dV dNHNO dV


Yl - r2

+ Y3

+ Y4

rs -

Assume the reaction takes place in an isothermal, ISOO-cm 3 PFR operating at constant pressure (LO atm). The feed to the reactor consists of a mixture of ethane and NO with a molar ratio of 95% ethane and 5% NO. The inlet volumetric flowrate is 600 cm 3 Is. Consider a base-case reactor temperature of 1050 K. 1. Develop the necessary design equations to calculate the molar flow of ethane, NO and ethylene, and determine the effluent composition from the reactor. Be sure to specify the initial conditions. 2. Solve the design equations and calculate the molar flowrates of etbane, ethylene and NO as a function of reactor volume for a reactor temperature of 1050 K. 3. Examine the effect of increaSing and decreasiog the temperature 50 K on the molar f10wrate of ethane. Why do you think a small

The rates are calculated from the rate constants and concentrations; for example, the first rate is

in which

The feed contains only CZH6 and NO at molar flows given by
p Nif ~ Yjf RTQr

leading to NC2 H6J ~ 6.62 x ] 0- 3 moJjs and NNOJ ~ 3,48 x 10- 4 mol/s. The numerical solution of Fquations 4.89 at 1050 K is shown in figure 4.22. We see the consumption of ethane and the production of


The Material Salance for Chemical Reactors

4.7 Some PFR·CSTR Comparisons


0'()07 0.006

'" ::::, "'


0.005 0.004 0.003 0.002 0.001 0 0 200



ethylene. The ethane molar flows at the three different reactor temper, atnres,are shown in Figure 4.23. The solution shows a small change in ethane molar flow initially, followed by a more rapid change of molar flow with increasing distance in the reactor. The rate is low initially until a suffiCient amount of H is formed so the third reaction can begin to contribute to the ethane consumption. The rates are very sensitive to temperature because the activation energies are quite high for the reactions that lead to the loss of ethane. For example, theJ'irst reaction that initiates the pyrolysis and the second reaction that produces H, which contributes to the consumption of ethane, have high activation D energies.

400 600 800 1000 1200 1400 1600 V (cm 3 )

Some PFR,CSTR Comparisons

figure 4.22: Molar flowrate of ethane. ethylene and NO versus reac· tor volume for ethane pyrolysis example.

Given the two contInuous reactors in this chapter, the CSTR and the PFR, it is natnral to compare their steady,state efficiencies in convert, ing reactants to products. For simplicity, consider a constant, density, liquid·phase reaction with nth'order, irreversible reaction rate
A.Jz. B


= kc~


0.007 . 0.006

For this sitnation, the steady,state PFR material balance is given by Equation 4.79





~ -riCA)


'" "' ::::, 0.004 0
0.002 0.001 0 0 200 400 600


We rearrange and solve for the time required 10 change from the feed condition CAj to some exit concentration CA


:i' 0.003





--,-dc 4

I r(cA)

(cm 3 )

1000 1200 1400 l600

in which the prime is to remind us that this variable is a dummy variable of integration. Figure 4.24 shows the graphical interpretation of this ) result. The area under the curve l/r(c A is the IOtal time required 10 achieve the desired concentration change. To achieve this same concentration change in the CSTR, we start with Equation 4.41, and solve for e gi\1ng

Figure 4.23: Molar flowrate of ethane versus reactor volume for inlet temperatures of 1000, 1050 and 1100 K.


CAl - C,4


This result also can be interpreted as an area in Figure 4.24. Notice that this area is the height, l/r(cA), times the Width, CAj - CA, of the


The Material Balance for Chemical Reactors

4.7 Some PFR·CSTR Comparisons



c' A
Figure 4.24: Inverse of reaction rate versus concentration; area un-

der curve is the PFR residence time 8; area of the


gle is the CSTR residence time; PFR is more efficient; irreversible, nth-order kinetics with positive order, n > o.

Figure 4.25: To achieve the same conversion, the CSTR is smaller than the PFR for irreversible, nth·order kinetiCS, nega· tive order, n < O.

Example 4.8: The PFR versus CSTR with separation rectangle sho'tm in Figure 4.24. Obviously if l/r(cA) is a decreasing function of CA, or, equivalently, r(cA) is an increasing function of CA, to achieve the same conversion, the PFR time (or volume, VR = QjO) is less than the CSTR time (volume). We make a final comment to reinforce this point. The reaction rate for the PFR varies with length. The rate is high at the entrance to the tube where the concentration of A is equal to the feed value, and decreases with length as the concentration drops. At the exit of the PFR, the rate is the lowest of any location in the tube. Now considering that the entire volume of the CSTR is reacting at this lowest rate of the PFR, it seems intuitively clear that more volume is required for the CSTR to achieve the same conversion as the PFR. So for this simple, single-reactant case, we only need to know if r(cA) is an increasing or decreasing function of CA to know which reo actor is more efficient, i.e., requires less volume. If the reaction order is positive (the usual case), the I'FR is more efficient, as shown in Fig· ure 4.24. If the reaction order is negative, the CSTR is more efficient, as shown in Figure 4.25. If the reaction rate is neither strictly increasing nor strictly decreasing, the reactor configuration with the smallest volume becomes a series of CSTRs and PFRs. Such an example is shown in Chapter 8, Example 8.4. Levenspiel provides further discussion of these interesting cases Ill, pp. 128-156J. We have noticed that a PFR achieves higher conversion than an equivalent volume CSTR for the irreversible reaction with first-order kinetics
A- B

ConSider the case in which we add separation. Find a single CSTR and separator combination that achieves the same conversion as the PFR. You may assume a perfect separation of A and B, the feed is a pure A stream, and ke = 1 for the PFR. Solution The PFR achieves a fractional conversion of A



= 1 - exp(-ke) = 0.632

For an equivalent volume CSTR without separation, the conversion of Ais XCSTR = 1 - N,,/ NAo = ke / (l + ke) = 0.5 The goal is to increase the achievable conversion in the CSTR using separation. Education in chemical engineering principles leads one im· mediately to consider recycle of the unreacted A as a means to increase this conversion. Consider the flowsheet depicted in Figure 4.26. A fraction of the outflow from the CSTR is recycled, the product B is removed and the unreacted A is combined with the feed as the inflow of the CSTR.


The Material Balance for Chemical Reactors

4.7 Some PFR-CSTR Comparisons




0.4 0.2

t" "-"'-=:-o: -: :-: -: -: -=-~-=-=-~-=-~-=-~-~-~-~-~-~-~-~-~-~- --+ ----~- ---;c~~;


per pass

pure B
Figure 4.26: PFR versus CSTR with recycle and separation.









Figure 4.27: Overall and per"pass conversion of A as a function of fractional recycle, DC

Given the assumption of perfect separation, we can achieve essentially complete conversion of A for kB ~ I with complete recycle, so our goal here is to calculate the fractional recycle, lX, that achieves exactly the PFR conversion. For kB < 1 the achievable conversion is less than one as discussed in Exercise 4.14. Notice first that the mean residence time of the CSTR, B', is less than that for the PFR, B, because the flowrate has increased due to the recycle. Notice that with perfect separation, pure A streams are combined at the mixer, and QO/QI ~ NAO! NAI so
B' ~ VR!QI ~ (VR/Qo)(QoIQI) ~ BN,\O/NAI

stream. Because the inlet flow of A is not specified, it is convenlent to divide the preceding equations by NAO, define dimensionless molar flows, and rearrange to obtain reactor: splitter: mixer:
NA,(l + kB/NAt1- NA! - 0 NA - (l lX)NA2 - 0 1 + lXNA2 -NA!

We may consider four variables to specify the state of the system: c<, NAI, N A2, NA; and we can write three component A material balances for the reactor, splitter at reactor exit and mixer at reactor inlet
reactor: NA2 ~ NAI/(l + kBN40/NAI ) NA
~ (l =





N AO + aN/!,2

The separator balance is trivial because the separation of A is perfect, and, therefore, the molar flow of A is conserved across the recycle

We can specify a single variable as known and solve for the remaining three with the three equations. For example, if we specify the recycle ,N AI , N A2, and compute fraction, c<, we can solve Equations 4.91 for the conversion from XA - 1 - N A. Figure 4.27 shows the reSUlting conversion of A plotted as a function of ex. We can see from Figure 4.27 that the PFR conversion is achieved at about ex ~ 0.65. If we want a more accurate answer, we can set NA ~ exp( -kB) = 0.3678 and solve


The Material Balance for Chemical Reactors

4.8 Stochastic Simulation of Chemical Reactions


nnmerically for a, N AI, N A2, with Equations 4.911 , and the result is




CSTR equivalence principle. Example 4.8 is motivated by a recent result of Feinberg and Ellison called the CSTR equivalence principle of reactor-separator systems [41. This surprising principle states: For a given reaction network "ith ni linearly independent reactions, any steady state tbat is achievable by any reactorseparator design with total reactor volume V is achievable by a design with not more than ni + 1 CSTRs, also of total reactor volume V. Moreover the concentrations, temperatures and pressures in the CSTRs are arbitrarily close to those occurring in the reactors of the original design. Applying this principle to the last example, we know that any achievable concentration of the PFR for a single reaction is achievable with a CSTR and separation. Note the number of CSTRs can be reduced from 11i + 1 to 11i in certain Situations, such as the one considered in Example 4.8. And we know the concentration in the CSTR will be achieved somewhere in the PFR.

also highly developed. The stochastic modeling approach is appropriate if the random nature of the system is one of the important features to be captured in the model. These situations are becoming increasingly important to chemical engineers as we explore reactions at smaller and smaller length scales. For example, if we are modeling the chemical transformation by reaction of only a few hundreds or thousands of molecules at an interface, we may want to examine explicitly the random fluctuations taking place. In biological probiems, we often conSider the interactions of only several hundred or several thousand protein molecules and cells. In sterilization problems, we may wish to model the transient hehavior until every last organism is eliminated. It is perhaps best to illustrate features of the stochastic approach with a simple example. Instead of the common case in which we have on the order of Avogadro's number of reacting molecules, assume we have oniya hundred molecules moving randomiy in the gas phase and we wish to follow the reaction

A~ B
Bh C

(4.92j (4.93j

in a constant-volume batch reactor. In this section we take reaction

statements quite literally. We assume these reactions are not merely observed stoichiometries, but actual molecular events. The probability of reaction is assumed proportional to the


Stochastic Simulation of Chemical Reactions
in which Xj is the numberof component j molecules in the reactor volume. Note Xl is an integer, unlike Cj of the deterministic model, which is real. The reaction probabilities play the role of the rate expressions in the deterministic models. Given the stoichiometry and the reaction probabilities, we would like to simulate the expected behaVior of the reaction network. One way to accomplish this task is the Gillespie algoritJ:un, which we describe next. The basic idea of the Gillespie algorithm is to: (1) choose randomly the time at which the next reaction occurs, and (ti) choose randomly which reactions occurs at that time. Of course we do not choose completely randomly. If the total reaction probabtli~ ties are large, it is intuitively clear that the time interval until the next reaction should be small, and, if reaction probability '"1 is much larger than r2, the first reaction is more likely to occur at the next reaction

We wish to introduce next a topic of increaSing importance to chemical engineers, stochastic (random) simulation. In stochastic models we simulate quite directly the random nature of the molecules. We will see that the deterministic rate laws and material balances presented in the previous sections can be captured in the stochastic approach by allowing the numbers of molecules in the simulation to become large. From this viewpoint, deterministic and stochastic approaches are complementary. Deterministic models and solution methods are quite efficient when the numbers of molecules are large and the random behavior is not important. The numerical methods for solution of the nonlinear differential equations of the deterministic models are
I Note one can solve this simple problem analytically as wel1. Eliminate aN ,L~ from the second and third equations in Equation 4.91. Substitute the result into the first equation and solve the resulting quadratic equation.



The Material Balance for Chemical Reactors

4.8 Stochastic Simulation of Chemical Reactions


time. The beauty of the Gillespie algorithm is the simple and statistically correct manner in which these two random choices are made. In a series of papers, Gillespie makes an elegant argument for the use of stochastic simulation in chemical kinetic modeling [7, 8] and provides the following simulation algorithm [7, p.23451.
L Initialize. Set integer counter n to zero. Set the initial species numbers, Xj(O),j = I, ... ns. Determine stoichiometric matrix v


and reaction probability laws (rate expressions)



Figure 4.28: Randomly choosing a reaction with appropriate proba~ bility; the interval is partitioned according to the rela-

for all reactions.
2. Compute reaction probabilities, r, action probability rtot = r,.

k,h(xj). Compute total re-

tive sizes of the reaction rates; a random number P2 between zero and one is generated to determine the reaction; in this case, m = 2 and the second reaction is


3. Select two random numbers, PI. P2, from a uniform distribution on the interval (0, O. Let the time interval until the next reaction be (4.94)

Determine reaction m to take place in this time interval. The idea here is to partition the interval (0,1) by the relative sizes of each reaction probability and then "thtow a dart" at the interval to pick the reaction that occurs. In this manner, all reactions are possible, but the reaction is selected in accord with its probability. See Figure 4.2 8. 4. Update the simulation time t(n+ 1) = t(n) +T. Update the species numbers for the single occurrence of the mth reaction via

... ns

Let n = n + L Return to Step 2. If riot is the total probability for reaction, e~~Yw,T is the probability that a reaction has not occurred during time interval T, which leads directly to Equation 4.94 for choosing the time of the next reaction. We will derive this fact in Chapter 8 when we develop the reSidence-time distribution for a CSTR. Shab, Ramkrishna and Borwanker call this time the "interval of quiescence," and use it to develop a stochastic simulation algorithm for particulate system dynamics rather than chemical kinetics !I4J.

Figure 4.29 shows the results of this algorithm when starting with XA = 100 molecules. Notice the random aspect of the simulation gives a rough appearance to the number of molecules versus time, which is quite unlike any of the determinlstic simulations presented in Section 4.2. In fact, because the number of molecules is an integer, the shnulation is actually discontinuous with jumps between simulation times. But in spite of the roughness, we already can make out the claSSic behavior of the series reaction: loss of starting material A, appearance and then disappearance of the intermediate species B, and slow increase in final product C. Note also that Figure 4.29 is only one shnulation of the stochastic modeL Unlike the deterministic models, if we repeat this simulation, we obtain a different sequence of random numbers and a different simulation. To talk about expected or average behavior of the system, we must perform many of these random shnulations and then compute the averages of quantities we wish to report. Next we explore the effect of increaSing the initial number of A molecules on a single simulation. The results for 1000 and 4000 initial A molecules are shown in Figures 4.30 and 4.31, respectively. We see the random fluctuations become less pronounced. Notice that even with only 4000 starting molecules, Figure 4.31 compares very favorably v..ith the deterministic simulation shown in Figure 4.11 of Section 4.2. Another striking feature of the stochastic approach is the trivial

31: Stochastic simulation of the first-order reactions A-~B-C starting with 4000 A molecules. simulation of the first-order reactions A-B-C starting with 1000 A molecules. Xj o 2 3 4 5 Figure 4. The microscopic level is characterized by discontinuous. The solution time depends on the number of Simulation steps. however.8 StochastiC Simulation of Chemical Reactions 169 100 80 XA 1000 800 Xc Xc 60 Xj Xj 600 400 200 0 0 40 20 0 1 2 3 4 5 0 1 2 3 4 5 t Figure 4. 4000ll------------------------::===========. . If reliable deterministic rate laws are available.30: Stochastic first-order reactions A-B-C starting with 100 A molecules. and also on whether or not we must repeat the simulations to calculate averages. random molecular motion and the probability of collision as the basis for chemical reaction rate. In fact. level of programming effort required to make the simulations. The macroscopic level is characterized by is more accurate to use the term pseudo-random number here to distinguish something we compute from a truly random number.168 The Material Balance for Chemical Reactors 4. Usually large uumbers of simulation steps are chosen when one has large numbers of initial molecules. 2 and many well-developed algorithms are available for that task. It builds a clear intuitive connection between the microscopic and the macrOscopic. at some paint it becomes marc efficient to use the deterministic models as the number of molecules increases. be large. The computational time required for performing the stochastic simulation may. But the stochastic approach is invaluable in several ways.29: Stochastic simulation of the t Figure 4. the biggest numerical challenge is producing the random numbers.

101) e ~ [1 f '" 8000 6000 4000 2000 0 0 50 deterministic avg. the random fluctuations may be an important physical behavior to include in the model.33: Species reDNA versus time for hepatitis B virus model.5 x 10. Example 4. stochastic - amino acids cc~"JA 50 100 t (days) 150 200 (4.6 (4. 100 t (days) 150 200 Figure 4. . explain why not. which is known as the thermodynamic limit. 250 200 reDNA cecDNA envelope degraded secreted or degraded secreted virus (4.025 1000 0.31 and then finally to the deterministic Fignre 4.170 The Material Balance for Chemical Reactors 4.102) Compare the results of a deterministic simulation to the average of 500 stochastic simulations.25 2 7. As stressed earlier.95) (4. We illustrate with the hepatitis B virus model introduced in Chapter 1.8 Stochastic Simulation of Chemical Reactions /71 smoothly varying concentrations.98) (4. Watchlng the transition in Fignres 4. nucleotides nucleotides + reDNA cceDNA envelope rcDNA+ envelope cccDNA 25 -< Z 20 15 10 tl u u deterministic avg. and use the following rate constants [XA Xe Xc 10000 r ~ [1 0 0 r 1 (4.29-4.32: Species eeeDNA versus time for hepatitis B virus model. stochastic 0.97) (4. and deterministic rate laws and material balances.9: Stochastic versus deterministic simulation of a virus model Consider the hepatitis B virus model given in Chapter 1. ~ 150 100 50 0 0 deterministic avg. and no rcDNA and no envelope protein.96) (4.100) Figure 4. the stochastic approach is essential and a deterministic approach cannot be substituted.11 is a nice illustration of this connection and provides logical support for the construction of the deterministic rate laws. deterministic and average stochastic models. deterministic and average stochastic models_ Assume the system starts with a single cccDNA molecule. It is possible to prove that the average of stochastic simulations converges to the deterministic simulation as the number of molecules becomes large. deter- ministic and average stochastic models. In this situation.99) tl . If these results are not the same. stochastic 5 0 0 50 100 t (days) 150 200 Figure 4..34: Envelope versus time for hepatitis B virus model..

111) I ~ Q u u u Z. two representative stochastic trajectories. We should bear this feature in mind when using deterministic models with small numbers of molecules. Figure 4. and the plug·flow reactor (PFR).95-4. Rc Y3 Ys Y6 Simulations. Notice the first stochastic simulation does fluctuate around the deterministic simulation as expected.32-4.105) (4. shows complete extinction of the virus. l R 4 RB J ~ [Y2 .172 The Material Balance for Chemical Reactors 4_9 Summary and Notation 173 35 30 25 BATCH constant volume CSTR deterministic stoch 2 constant density steady slale SEMI-BATCH d(CjVR) dt dej dt d(CjVR) dt dCj dt ~ RjVR Rj ~ (4. and C is envelope.110) (4. Figures 4. The second stochastic simulation.3: Summary of mole balances for several ideal reactors.Y4 . 20 15 10 ~ QfCjf .107) (4. irreversible I I 'I l I . it occurs for 125 of the 500 4.32-4. B is reDNA.3 summarizes the mole balances for these four reactors. the semi-batch reactor.9 Summary We have introduced four main reactor types in this chapter: the batch reactor.104) (4. We also have introdoced some of the basic reaction-rate expressions: • first order. Notice these results are not the same.112) Table 4.109) (4.35: Species cccDNA versus time for hepatitis B virus model.Y2 J (4.100 are given by Yl Y2 Y3 Y4 Ys Y6 k1xA k2XB k3 X A k4 X A ksxc k6XE X C I d(cjQ) = Rj dV dej Rjl de ~ e ~ V/Qf (4. So the average stochastic simulation in Figures 4.103) . Solution The reaction rates and production rates for Reactions 4.35 shows two representative stochastic simulations for only the cccDNA species.34 show the deterministic model simulation and an average of 500 stochastic simulations. The two types of stochastic trajectories therefore explain why the average stochastic model is not equal to the deterministic model. Table 4.34 consist of 75% trajectories that fluctuate about the deterministic trajectory and 25% trajectories that go to zero.106) (4. 0 in which A is cccDNA. In fact. the continuous-stirred-tank reactor (CSTR).108) (4. That is another possible steady state for the stochastic + Rj Cjf + Rje QfCjf + RjVR o(CjQ) 5 0 c} d(cjVR) dt OCj at ~ I ~ 0 50 100 t (days) 150 200 PFR steady state constant flo"Tate ~ av + Rj Figure 4. __ (j(cjf .Y6 YI . however. and we should investigate why not.QCj + RjVR 1 ~ .

With small numbers of molecules. vijV:. K. Xj Xj Yj . The probability of occurrence of a given reaction is assumed proportional to the number of possible combinations of reactants for the given stoichiometry. k. the CSTR achieves higher conversion than the same'volume PFR if the rate is a decreasing function of a component composition (n < 0).density mixture. In the case of mUltiple reactions with nonlin' ear rate expressions (Le" not first'order reaction rates). an ideal mixture. a differentialalgebraic equation (DAE) solver may be convenient. Ytot total polymer mass volumetric flo\\Tate feed volumetric flmvTate reaction rate for ith reaction total reaction rate. NjM j molecular weight of species j reaction order moles of species j.9 Summary and Notation 175 • • • • • • first order.Li Yj gas constant production rate for jth species time temperature reactor volume variable partial molar volume of species j speCific molar volume of species j reactor volume change in volume upon reaction i. tion (ODE) solver is indispensable for solving these problems. M M MJ Mj reactor tube cross-sectional area preexponential factor for rate constant i concentration of species j feed concentration of species J steady-state concentration of species j iutial concentration of speCies j activation energy for rate constant i reaction rate constant for reaction i equilibrium constant for reaction i total mass flow. Cj cjf Cjs CjO We developed the equations required to compute the volume of the reactor if there is a significant volume change upon reaction. number of molecules of species j in a stochastic simulation molar conversion of species j mole fraction of gas-phase species j R Rj t T V Vj V. reverSible second order.1) pressure P P Q QJ T. We require an equation of state for this purpose. we introduced stochastic simulation to model chemical re' actions occurring with small numbers of molecules. nonlinear hepatitis B virus modeL E.2 describe the appropriate balances for a constant. QCj uniformly distrihuted random number on (0. Two pseudo'random numhers are chosen to determine: (i) the time of the next reaction and (ii) the reaction that oc' curs at that time. NJMj total monomer mass total feed mass flow. however. The stochastic model uses hasic prohability to compute reac' tion rate. We demonstrated this fact with the simple. the halances must be solved numerically. Finally. A. . and a mixture with a general equation of state. Conversely. irreversible nth order. A high'quality ordinary differential equa. QCJ feed molar flow of species j. For a complex equation of state and nonconstant'volume case.174 The Material Balance for Chemical Reactors 4. The smooth behavior of the macroscopic ODE models is recovered by the random simulations in the limit of large numbers of reacting molecules. . Tables 4. Several of these simple mass balances with basic rate expressions were solved analytically.1 and 4. VRCj number of reactions in the reaction network number of species in the reaction network n nj nr ns Nj NjJ P molar flow of species j.I VR tN. reverSible reactions Notation A. Each of these ran' dom simulation trajectories has a rough appearance and the average of many of these simulations is required to show the expected system behavior. We showed that the PFR achieves higher conversion than the CSTR of the same volume if the reaction rate is an increasing function of a component composition (n > 0 for an nth'order rate expression). the average of the stochastic simulation does not have to be equal to the deterministic simulation. irreversible two first'order reactions in series two first'order reactions in parallel two second'order.

llquid-phase isomerization reaction carried out in a solvent containing dissolved catalyst at 25°C in a batch reactor A~B The apparent first-order reaction-rate constant. What is the order of the reaction if it takes 10 minutes to reach 75% conversion'! What is the order of the reaction if it takes 20 minutes to reach 75% conversion? What is the order of the reaction if jt takes 30 minutes to reach 75% conversion? Exercise 4. A chemist colleague of yours has studied the catalyst deactivation process and has proposed that it can be modeled by ka = -~~ Exercise 4.B. decreases "h1th time becanse of catalyst deterioration. of volnme VR !3 in series? (c) What is the conversion if the original reactor is replaced by a plug-flow reactor of volume VR? r = kc1 The conversion of A reaches 50% in a five¥minute run. isothermal gas-phase batch reactors for the elementary and irreversible decomposition of A to B. what is kd? (d} How long does it take to reach 75% conversion? . VR I Qf stoichiometric coefficient for species j in reaction i mass density e Vij r = kCBC~ p T simulation time in a stochastic simulation takes place in a CSTR of volume VR in the presence of a large excess of reactant B..10 Exercises Exercise 4.10 Exercises 177 z Ei reactor length variable Exercise 4.".0 atm and k 0.35 min"l.6 hr I. reactor 2: The reactor pressure is held constant {reactor volume therefore changes)..3: CSTR performance A liquid-phase reaction A+B. Sketch your solution. te) If it takes two hours to reach 50% conversion and the fresh catalyst has (a) What is the fractional decrease in the concentration of A in reactors 1 and 2 after five minutes? (b) What is the total molar conversion of A in reactors 1 and 2 after five minutes? a rate constant of 0. A-I<. C.176 The Material Balance for Chemical Reactors 4.. Both reactors are charged Vvith pure A at 1. in which k is the fresh catalyst rate constant and kd is the deactivation-rate constant.. (h) Solve the mole balance for CA (t). (a) What is the steady-state conversion if the original rea". extent of reaction i reactor residence time.r is replaced by two CSTRs of volnme VR /2 in series? (b) What is the conversion if the original reactor is replaced by three CSTRs 4. ka. 2B k 1 + kdt reactor 1: The reactor volume is held constant (reactor pressure therefore changes). Assume the reactor achieves 50% conversion of A at steady state.4: Catalyst deactivation in a balch reactor Consider the irreversible.2: Constant volume versus constant pressure batch reactor Consider the following two well-mixed. (a) Write down the mole balance for this reactor.1: Order versus rate of reaction Uquid A decomposes with nth-order kinetics in a batch reactor A .

Recall yield is the fraction of reactant converted into a specified product.8: Nonconstant density with a liquid-phase reaction ~Ax+b ~ dt x(O) xo (4. CA(t) versus t.71 x 10'J S--1 Ea ~ 18. (a) Take the Laplace transform of Equation 4.113 and show that it satIsfies the differential equation and initial condition. for the tank initially filled "ith solvent.D Y1 Write down and solve the dynamic material balance for component A.02 min-'.113 applies to tbe reversible. Sketch the solution.CH. Clearly label on your plot the initial and klC~ "" r2 k 2 cB steady-state concentrations for both curves.115 into Equation 4. A--". The tank volume is VR. first-order reaction case given in Equation 4.-CH-. B+C The reaction is first order in A and irreversihle. CAD ~ cAf.113 appJies to the reversible.113 and show _() + H2 0 -.113) Propylene glycol is produced by the bydrolysis of propylene oxide according to tbe following reaction H.115 and verify Equation 4.178 The Material Balance for Chemical Reactors 4. (c) For a 50 m' tank with flowrate of 7 Lis and rate constant k ~ 0. and the reaction is performed in a 1000 L CSTR operating at bO'~C The feed conditions and physical properties arc as follows [l5J: . Methanol is added as a solvent. the reaction has been found to be first-order in propylene oxide (4. which is also constant. The feed to tbe first Exercise 4. so tbat Equation 4. first-order reaction case given in Equation 4.05 L/mol·min and k2 ~ 0. first·order. constant-coefficient differential equation dx reactor is pure A.0 kcal/mol K (c) Substitute Equation 4. Uquid density change due to reaction may be neglected. h. At time zero. The tank initially is charged with species A at concentration CAO.15. Exercise 4. CAi = 1. Substitute these values into Equation 4. and the residence time in each reactor is 15 min. 10 Exercises 179 Exercise 4. sketch the solution for the tank initially filled with feed. (a) (h) _. A. what is the steady-state conversion of A? in whicb k.13. On the same plot.S mol/L. The Laplace transform is a bandy metbod for solving linear differential equations. (e) Identify A. Exercise 4.7: Multiple reactions in CSTRs in series The folloWing liqUid-phase reactions take place in two identical CSTRs in series. Substitute these values into Equation 4.42. Determine tbe overall conversion of A and the yield of B for the series of reactors. The feed concentration of A is CA!. the feed pump is turned on and delivers constant fiovvrate J Qj.022 min-'. ~ 0.114) propylene oxide I I x 5 ~-A s- Xo + 5 (5~ A·) b OH OH propylene glycol (b) Invert the transform and sbow In the presence of excess water.C~-CH-CH. b.5: Dynamic CSTR A CSTR is used to convert A to products Band C via the following liquid-phase reaction (d) Identify A. so tbat Equation 4.115) r "" kepo and tbe rate constant is [61 ko = You may want to use the following partial fraction formula s(S~Ai ~ *D~ s~Al 4.115 and verify Equation 4.CH3 "0/ (4.B+C B.43. CAD ~ O.6: Linear differential equations and Laplace transforms Consider the linear.

116) 6600 0 1300 <a) Show that even though the density is not assumed constant in Equa· tion 4. constant volume or constant mass? Look at the conversion of propylene oxide and production rate of propy' lene glycol for the two cases.31 for the intermediate B concentration (a) What is the optimal stopping time in order to maximlze the concentra· tion of intermediate B? (b) Check your result against Fignre 4. What are you wasting in constant masS operation? (b) Resolve the constant reactor volume operation under the assumption that all densities are equal to water. (c. and VR for the following two situations. ~ M)! Pj are the pure component specific molar volumes. upon reaction even though the solvent does not take part in any reactions.9: linear density versus concentration assumption Consider the important case of liquid-phase reaction in an excess of solvent Let component j = 5 be the solvent and consider the solution density at some reference concentrations.29 reactor mass.11: Some limits (a) For the second·order.24 CAO cJ ()} *' CRO .02 76.116.859 1. A float in the top of the tank is used to adjust Q to maintain reactor volume constant at 1000 L.c'o) in which C s is the actual solvent concentration in the mixture. wt.116 is in which V. and a differential pressure measurement is used to adjust Q to maintain constant M.C Mscso + 2: Mjcjo ) *s In this situation. irreverslble reactions in Reaction 4.10: Maximizing an intermediate We considered the following two first-order. which also changes. Neglect any change in the pure component densities "'lith temperature in the temperature range 25-60'C. Q. upon substitution of the previous equation p 58. The reactor is initially charged with pure solvent. Exercise 4. Cj = CjO. in genera!. a useful approximation for the density-concentration relationship is the linear form IU] fJ=fJo+ L.08 18. (g/mol) Inlet feedrate (L/hr) 1300 Of. 2.0361 0. How much error in the conversion and production rate do you commit under this assumption? Ah B B -"" C and derived Equation 4. Exercise 4.11.000 1. (a) Compute the steady· state concentrations of all components.180 The Material Balance for Chemical Reactors 4. .10 Exercises 181 Component propylene oxide water propylene glycol methanol Density (g/cm 3 ) 0. equivalently. ~ kit Exercise 4. irreversible reaction A+B-". we showed solving the material balance produces Equation 4.04 = Mscso + J*' Mjcj (4.7914 Mol. one still ohtains for the reactor volume in a CSTR dVR dt Assume the mixture is ideal so that ~ Qf Q (b) Show that the approximation error in the density in Equation 4. Po = (c) How does this result simplify for the case k.11 32. Which operation do you recommend. I.Mj(cj J'~ s in the batch reactor.

7.1 in this equation and show that it produces the first-order reaction result ~ =e CAO (d) Plot e versus XA for b very hypotheticaL ~ 1. Consider the following hypothetical reaction taking place in the gas phase in an isothermal. Derive an expression for the tube volume required to achieve molar conversion of A.26.xample 4. hence the idea gas law. change the feed concentration of NO from 5 ppm to 5%.12: Constant gas Ilowra!e.19 -<- CA = I ) ( ~.36: An alternative CSTR'separator design.13: Ethane pyrolysis. (a) Ideal gas equation of state.8 shmvn in Figure 4.100. for the follOWing two cases.. Assume the feed consists only of ethane and NO. we showed solving the material balance produces Equation 4. XA.15: An alternative reactor-separator design Feinberg and Ellison considered an alternative design for F. and after the recycle split in Figure 4.T . rx? What happens if one attempts to recycle more material than this limiting value? (b) Constant gas flowrate.36.~ + kt CAD I (b) For the n-order." it doesn't matter if they are heavy or light. Notice the separator is placed ahead of the recycle spIlt in this Your result should be a function of only XA and h.14: Recycle effects Consider again Example 4. Explain the effect of the feed NO concentration on the ethane conversion. constant pressure PFR with a pure A feed stream A . and that the pressure and volumetric nowTate are the same as in Example 4. VRc.5. large or small error? One of the interesting features of (ideal) gas molecules is that they demand a certain amount of "territory..2. 1. VRIl. irreversible chemical reaction A.:. in which ko=kcAol Take the limit n '. deSign. and compute the limiting conversion for a CSTR With recycle for kO OS What is the corresponding limiting fractional recycle.vhich we call the dubious assumption.10 Exercises 183 Take the limit CAD ~ Ceo 0 in this equation and show it produces the other second-order react jon result given in Equation 4.182 The Material Balance for Chemical Reactors 4.117) Exercise 4. .7. Figure 4. and examine the effect on the concentration profile for ethane in the PFR at 1050 K. Exercise 4. Note for b ~ 100.B r = kCA pureB -- in the batch reactor. which we call the correct answer.bE (4.26 ~ ~ [1 CAD + (n -l)ken'}. revisited Using the reaction chemistry and kinetics presented in Example 4.117 is kl (e) If you are unwilling to commit errors larger than 10%. {c} Compute the fractional error committed when using the dubious assumption Exercise 4. bjg or small ~ all gas molecules occupy the same amount of space at a given temperature and pressure. for what band XA values must you object to the dubious assumption? Bottom line: are you generally making large or small errors when you neglect the change in the nnmber of moles when modeling a gas-phase reaction in a PFR? Exercise 4. Reaction 4.8.

3600 B. and a simulation that leads to extinction.102. I integrate it Over the CSTR reactor volume element (as depicted in Figure 4.954. Perform enough simulatious so that some of your trajectories lead to extinction of all three species: cceDNA.119) .C (c) Repeat the stochastic simulation for an initial condition of 4000 A. A student in your class asks you to explain the following difficulty he is having. "Since this equation is Hue at all points in the volume element.184 The Material Balance for Chemical Reactors 4.37: Deterministic simulation of reaction A + B :.8. (CjVj) - R] dO j 0 (4.8 Exercise 4.4 (a) Compare and contrast a simulation that leads to a steady-state infection. Consider the foHovving situation from an instructor's pOint of view. Figure 437b a representative comparison for one sequence of pseudorandom numbers. "I start mth the continuity equation for species j . 360 B and zero C molecules. C com+ pared to stochastic simulation starting with 400 A molecules. '0 ~ 0.' + '7.16: Stochastic simulation of a virus model Perform several random simulations of the virus model given in Equations 4.1l8) (b) Compare the result to a stochastic simulation using an initial condition of 400 A. Exercise 4.:: 0 u E u -- 0." 5 min.. Notice from the units of the rate constants that kJ should be divided by 400 to compare simulations. k = .18: From microscopic to macroscopic (a) Solve the deterministic material balance for a constant-volume batch reactor with kJ ~ 1 moJjL·min k-J J min"'! cdO) ] mol/L CB(O) ~ O. 0. Remember to scale kJ appropriately.17: Stochastic simulation for nonlinear kinetics Consider the reverSible.13). Band C concentrations out to t = ~ 0 rnoljL It is often easier to provide a correct explanation than it is to find the error in someone else's incorrect explanation." Iv [Gil.::::. what is the longest time required to reach extinction? vVhat fraction of your trajectories lead to extinction? (dl Which parameters in the model have the largest affect on the behavior? Figure 4. (b) 1 Compare the two reactor-separator designs. What are the advantages and disadvantages of each design. Are the fluctuations noticeable with this many starting molecules? Exercise 4. o (4.6 .2 2 time (min) 3 4 5 (c) Of the trajectories that lead to extinction. 10 Exercises 185 (a) Calculate the lIo"Tate of the recycle stream that achieves the PFR conversion given in Example 4.g mol/L Plot the A. second-order reaction A+B k. zero C molecules. is it possible to tell by looking at the trajectory whether it wiJI lead to infection or extinction? About how many days are required before you have a reliable idea of where a trajectory "ill finish? 0. reDNA and envelope protein. (b) During the first 25-50 days.

. Consider the algebraic equations appearjng in Table 4. Some care should be exerCised.12S) The solution to Equations 4. n" are the x variahles and VR and Q are the y variables. then iJB (iJyT = 0 and the DAE has index two or higher.106 is correct. for example.123) Exercise 4.125 with respect to time produces (b) Write out the series expansion for e KI and verify the solution derived in Section 4.2 are of the following form in which the time derivatives of only the x variables appear explicitly in the model. The DAEs in Table 4. cases L and 2. and K is an ns x ns matrix of Tate constants. is index one.120) DB.. 2. Show that.126) dt dx =f(x. The batch reactor material Exercise 4.2.y) (4.1 (c) Repeat for two ftrst~order parallel reactions in a batch reactor . with no algebraic relations.120 gives" f[ OC) ot." f v CJ(V]· n)d<T = S ~QfcJf + QCj (4.186 The Material Balance for Chemical Reactors 4.12!) "Substituting this into Equation 4. iJYJ "It is clear to me that the surface integral picks up the flux of species j into and out of the reactor with the flow streams. the complexity of the ODE and DAE models appearing in Tahle 4. reactor volume constant. to be index one? (4. for example." fv [ act in which og/iJyT is a Jacobian matrix with elements OB ) ( iJyT i) = at ~RjJdQ+f5 c(vJ ·n)d<T~O J (4.. Without differentiating the equation of state. in which R is an n~-vector of production rates.2.y) g(x. reactor mass constant. how can case 2. however.20: Matrix exponential Consider the general linear kinetic expression written in vector-matrix form R(c) = "But this doesn't agree with Equation 4. (a) Write K for two first-order series reactions jn a batch reactor The total moles of the species. he does not really want to hear a restatement of why Equation 4.124) (4. has index zero.122) "Finally since Cj and Hi don't vary over the well-mixed reactor. I simplify to" de VR d( ~ QfCjf ~ Qc) + RJVR If og {oyT is a non singular matrix. J (4.2.l ~R]dQ~QfCI'+QC>~O J. he really wants to know what is wrong with his Ke derivation. are high-index DAEs and case 3. Differentiating Equation 4. 3! o This series converges for all A.106!" How do you respond to this student? Remember. The matrix exponential can he defined using the power series 1 eA=l+A+l + ~A·3 + .5 that it was convenient to specify reactor models for nonconstant -density cases by using Df"ills instead of ODEs. balance can be written as de ~Ke e(O) = Co (4.19: Index and DAEs VyTe saw in Section 4. A=B B=C The index of a DAE is the number of times that the algebraic equations must be differentiated with respect to t to determine dyldt 131. Q specified. and 3. Notice if the y variables do not appear explicitly in the algebraic equations. c is an ns~vector of concentrations. Dllis \vith index zero and one are generally much easier to solve than thnse with index two or higher [3]. be modified to be index one? Can you find a simple way to mollify case 1. Compare. then we can solve this equation for dy /dt and the DAE has index one.126 can be written as c = eK1c(J. to avoid creating a DAf that is diffjcult to solve. 10 Exercises 187 "} then use the divergence theorem that we learned in the Transport Phenomena course to convert the divergence term to a surface integral and obtain.2 for the noncenstantdensity cases: I. An ODE. in which e Kt is the matrix exponential. as expressed in Table 4.

Florida. General kinetic bounds on productivity and selectivity in reactor-separator systems of arbitrary design: L Principles. Eng. II3J T. lSI J. Chern. Gillespie. Gillespie. volume 35. New York. 1990. 188:404-425. The reaction inhibited by nj~ trie oxide.. and E. second edition. Part Three: Kinetics and Catalysis. E. 23(6):897-904. 2. W. second edition. 16J T. C \\least. 12] W. Chern. R. 1972. A rigorous derivation of the chemical master equation. D. New York.1 Introduction The purpose of this chapter is to provide a framework for determining the reaction rate given a detailed statement of the reaction chemistry. 1997. Pujado and S. Frank K. 17] D. Feinberg and P. and L. Proceedings of the Royal Society of London. New York. 18J D. Seubold and W.Bibliography 5 Chemical Kinetics 11 J R. Ramkrishna. Chemical Reaction Engineering. editor. Marcel Dekker. Vaughan. Introduction to Chemical Engineering Analysis. New York. Boca Raton.. J. Levenspiel. sixth edition. 1989. 5. j. and T. F. Bird. 1977. 1979. W. Brenan.. M. Numerical Solution or lnitial+alue Problems in Differential-Algebraic Equations. The goals here are to develop a chemical kinetics basis for the empirical expression. 1121 P. Math and Physics. New York. The quasi-steady-state assumption (QSSAI and the reaction equilibrium assumption allow us to generate reaction-rate expressions that capture the details of the reaction chemistry with a minimum number of rate constants_ The concepts include: • the elementary reaction • Tolman's principle of microscopic reversibility • elementary reaction kinetics • the quasi-steady-state assumption • the reaction equilibrium assumption This chapter complements Chapter 9 on data analysis and empirical reaction-rate expressions in which the goal is to find a mathematical relation that describes the reaction rate over the range of experimental conditions. Borwanker. Stewart. and to show that kinetic analysis can be used to take mechanistic insight and describe reaction rates from first prinCiples. Eng. T. eRe Press. John Wiley & Sons. 60th edition. B. Exact stochastic simulation of coupled chemical reactions. Laidler and B. Russell and M. Furusawa. Physica A. 75:3790-3792. AfChE}. C. 2001. 1961. 81:2340-2361. Soc.]. New York.1969. In J. 19S3. 1947. T. Nishimura. Kinetics and mecbanisms of the thermal decomposition of ethane II. Ellison. W. Simulation or particu" late systems using the concept of the interval of quiescence. Amer. pages 372-391. Hougen and K. Encyclopedia or Chemical Processing and lJesign. John Wiley & Sons. John Wiley & Sons. New York. Campbell. 40(14):3181-3194. E. DiPrima. editor.. 1972. and J. 114J B. Phys. H. 14J M. Denn. Chern. L. Sifniades. John Wiley & Sons. Phenol. Shah. 260A(] 300):103-1 13. M. E. 2(1):95-100. The stoichiometry of an elementary reaction defines the concentration dependence of the rate expression. 1977. !15] R. 1992. . H. Chemical Process Principles. Transport Phenomena. Watson. 13J K. Japan. Petzold. John Wiley & Sons. N. E. S. Elsevier Science Publishers. Acid-catalyzed decomposition of cumene hydroperoxide. We use several concepts from the subject of chemical kinetics to illustrate two key points: 1. McKena. Wojciechowski. Chern. lightfoot. J. R. j. Boyce and R. Miyaucru. A. Res. [9J O. III J O. [1 OJ K.. Experimental study of a bistable continuous reactor. Elementary Differential Equations and Boundary Value Problems. Inc. 2002. D. eRe Handbook or Chemistry and Physic'>'. Ind.

the reaction rates for the forward and reverse paths are proportional to the concentration of species taking part in the reaction raised to the absolute value of their stoichiometric coefficients.5 to ketene and a methyl radical. Catalysis has a signlficant impact on the Unlted States economy' and many important reactions employ catalysts.190 Chemical Kinetics 5.4. The third radical also undergoes unimolecular decomposition in elementary Reaction 5. is explained as a linear combination of the five reactions in the proposed mechanism. Or the reaction statement may represent the actual molecular events that are presumed to oCCur as the reaction proceeds. More than 60% of the products and 90% of these processes were based on catalysis!27j. By balancing the species in the overall stoichiometry. The reaction order in all species is determined directly from the stoichiometry. Reaction 5.1 million.6. for the thermal decomposition of acetone at 900 K to ketene and methylethyl ketone (2-butanone) [33J. and the methyl radical combines with the third radical to produce methyl ethyl ketone in elementary Reaction 5. any correspondence hetween the stoichiometric coefficients and the reaction order is purely coincidentaL We use three examples to illustrate overall reaction stoichiometry and elementary reactions. 63 major products: and 34 major process innovations "vere introduced by the chemical industry. Elementary reactions are usually unimolecular or bimolecular because the probability of collision between several species is low and is not observed at appreciable rates. The reaction statement may represent the change in the relative amounts of species that is ohserved when the reaction proceeds. CH2CO and CH3COC2HS. CH]CO and CH3COCHz. The acetyl radical undergoes unlmolecular decompOSition to a methyl radical and carbon monoxide in the second elementary reaction. (5. on the other hand. If we assign the species to the A vector as follows . If the elementary reactions cannot he comhined to form the overall stoichiometry. The first involves the mechanism proposed {Catalysis is at the heart of the chemical industry. The thermal decomposition of acetone generates four stable molecules that can be removed from the reaction vessel: CO. The physical characteristics of catalysts are discussed in Chapter 7. The former is referred to as an overall stoichiometry or an overall reaction statement. Three radicals are formed and consumed during the thermal decomposition of acetone: CH]. CH4.1.2 Elementary Reactions and Microscopic Reversibility StOichiometric statements such as A+B = C are used to represent the changes that occur during a chemical reaction. For an overall stoichiometry. and the petroleum refining industry. then the mechanism is not a valid description of the observed stoichiometry. The overall reaction stoichiometry.75 million in 1990 (27). one can determine that the overall stoichiometry is produced by adding twice the third reaction to the remaining reactions. We show subsequently that for an elementary reaction.6) in the first elementary reaction.2 Elementary Reaction 191 In this chapter we also discuss heterogeneous catalytlc adsorption and reaction kinetics. acetone undergoes unJrnolecular decomposition to a methyl radical and to an acetyl radicaL These radicals continue to react the methyl radical reacts to methane and a third radical (CH3COCHz) in elementary Reaction 5.2) (5.3) (5A) 5. The latter is referred to as an elementary reaction. These statements can be interpreted in two ways.19805.5) (5. 2Between 1930 and the . The elementary reaction is characterized by a change from reactants to products that proceeds without identifiable intermediate species forming. These three radicals are reaction intermediates and cannot be isolated outside of the reaction vessel. which had sales of $140 billion and employment of 0. The overall reaction can be represented by and is proposed to proceed by the following elementaIj1"reactions CH3COCH3 CH3CO CH3 + CH3COCH3 CH3COCH2 CH3 + CH3COCHZ + CH3CO CH3 CH3 CO + CH4 +CH3COCH2 CHzCO CH3 + CH}COCzH s (5.2 We describe the kinetic prinCiples that are needed for rate studies and demonstrate how the concepts for homogeneous reactions apply to heterogeneousy catalyzed reactions with the added constraint of surfacesite conservation. with sales in 1990 of $292 billion and employment of 1.

the elementary reactions describe the detailed pathway between reactants and products. The student in which subscripts g and s refer to gas phase and adsorbed species. all of which are adsorbed on the surface.9) (5.2 Elementary Reaction 193 Species Al A2 A3 A4 As A6 A7 /18 Formula CH3COCH3 CH3 CH3 CO CO Ol)COCH 2 CH2CO CH4 CH3COC2H. methylene. CO and Hz.21) which is the overall stoichiometry given in Reaction 5. developing and verifying a mechanism and a complete discussion of these factors is beyond the scope of this text.19) (5. that is to say the reaction contributes to tbe appearance of intermediates or products. over a ruthenium catalyst [8J.20) If we multiply v by [ 1 1 0 2 1 1we obtain 1 2 1 [ -3 0 1 0 1 Hs CH3s + S CH4(g) + 25 Hs- (5. are generated and consumed during the photochemical reaction.12) In this reaction mechanism.21). - - Name acetone methyl radical acetyl radical I carbon monoxide acetone radical ketene methane methyl ethyl ketone should work through the steps to determine the linear combination of the mechanistiC steps that produces the overall reaction. Numerous factors must be considered in proposing. Hydrogenation of methyl leads to methane in elementary Reaction 5.7) and one possible mechanism is [12. and S refers to a vacant ruthenium surface site.192 Chemical Kinetics 5. during . stable molecules arC formed. and methyl. In this example. H20(g) + S 2Hs CHs+ S CHz s + S (5.8) (5. Adsorbed CO (COs) either occupies surface sites or diSSOCiates to adsorbed carbon and oxygen in elementary Reaction 5. respectively. In each of these examples.16) (5.1.8).16 and 5. that is to say the reactants could form the products (and the products could form tbe reactants). nitrogen dioxide is activated by absorbing photons and decomposes to nitric oxide and oxygen radicals (elementary Reaction 5. the overall reaction is twice the fourth reaction added to the remaining reactions. Keep in mind that the elementary reactions must be possible chemically. and the elementary reactions should be kinetically Significant. 10] N02+ h v O+NOz NO} + NOz NO} + NO NO+ 0 N03 NO + 02 + NOz 2N02 NO+02 = (5.21. For example. and the products form at the surface and desorb (elementary Reactions 5.15 ) (5. The second example involves one of the major reactions responsible for the production of photochemical smog.15. The overall reaction is (5. The adsorbed carbon undergoes a sequential hydrogenation to methyne.17).10) (5. During the overall reaction. the reagents adsorb (elementary Reactions 5.17) - Hs Hs - (5. The overall reaction is 2N02 + hv 2NO + 02 (5. 0 and NO].14) (5.11) (5.14 and 5.18) (5. The third example involves the synthesis of methane from synthesis gas. in this case NO and 02.13) and one possible mechanism is CO(g) + then the stoichiometric matrix is -1 a v~ -1 0 0 1 1 -1 1 1 1 -1 a 0 0 1 0 1 0 0 0 0 0 0 I a a 1 -J 1 0 0 a 0 1 0 0 0 0 1 0 COs + Os + H2(g) lh(g) + Cs + CHs+ CHz s + CH3s + 2S S S COs Cs + 0. and radicals. As in the previons example.

22 forms the basis for predicting reaction rates and is applied to homogeneous and heterogeneous systems. This reversible nature of elementary reactions is the essence of Tolman's prinCiple of microscopic reversibility [37. One criterion for a reaction to be elementary is that as the reactants transform into the products they do so without forming intermediate species that are chemically identifiable. Laidler (21) and Hill 114) provide a good discussion of some of these kinetic arguments and present additional examples of complex reactions. 699).22.3 Elementary Reaction Kinetics In this section we outline the transition-state theory (T5T). 40J and the material outlined here is far from comprehensive it is intended to provide a framework for developing a physical picture and mathematical model of a chemical reaction. the priociple requires not only that the fOtal number of molecules leaving a given quantum state in unit time is equal to the number arriving in that state in unit time.22) in which j E 1Z. the most probable path in one direction must be the most probable path in the opposite direction. Equation 5. Before describing TST it is necessary to develop the concept of the reaction coordinate. the Gibbs energy difference in the two directions must be equal to the overall Gibbs energy difference . symmetry and steric bases.CH3COCzHs CH3 + CH3 C2H6 2CH3COCH2 . In the case of a system in thermodynamic equilihrium. you need only an understanding of molecular partition functions at the level presented in undergraduate physical chemistry texts. Our purpose is to show how the rate of an elementary reaction is related to the concentration of reactants. We use a two-body reaction example to calculate and illustrate a potential energy surface from which a reaction coordinate diagram is constructed_ Elementary reaction kinetics is a rich field [26.194 Chemical Kinetics 5. therefore the probability of methyl recombining is much lower than the methyl being consumed in the other steps. The potential energy V (r) . we may use three coordinates to locate the center of mass of the molecule. the recombination of CH3COCH2 can be discounted because the chemical instability of the bond formed in the recombination product (CH3COCH212 would cause this product to decompose rapidly back into the radical CH3COCH2. and j E 'Pi represent the reactant species and the product species in the lth reaction. but also that the numher leaving by anyone particular path. using only the products of the elementary reaction. Because of its wide use. The remaining 3N ~ 6 coordinates are used to specify the relative positions of the atoms. Furthermore. The result is (5. 5. Transition state theory erST) is reViewed briefly in the next section to describe the flow of systems (reactants products) over potential-energy surfaces. For convenience.the ratio of rate constants for an elementary reaction must be equal to the equilibrium constant for the elementary reaction.3 Elementarv Reaction Kinetics 195 acetone pyrolysis several radicals are produced and these could recombine in a variety of processes that are chemically possible such as CH3 + CH3COCHZ ~. Molecules containing N atoms require 3N Cartesian coordinates to specify the locations of all the nuclei. respectively.21. which can be used to predict the rate of an elementary reaction.6 is listed in the mechanism presented above because the other two radical recombination reactions occur at rates too small to be significant. for the purposes here. A second aspect of an elementary reaction is that the reverse reaction also must be possible on energy. Only two are needed to describe the orientation if the molecule is diatomic. Furthermore. Using Tolman's principle. Transition state theoty is based on the principles of statistical mechanics and. only Reaction 5. the remainder of this section describes the concepts and assumptions that underlie Equation 5. and three more coordinates to orient a chosen axis passing through the molecule. Various models or theories have been postulated to describe the rate of an elementary reaction. is equal to the number arriving by the reverse of that particular path. Methyl recombination was ruled out because the methyl concentration is anticipated to be much lower than the concentration of CH3COCHZ.(CH3COCH2)2 However. This assumption (at equilibrium the uumber of molecules going in unit time from state 1 to state 2 is equal to the number going in the reverse direction) should be recognized as a distinct postulate and might be called the principle of microscopic reverSibility. p.

5 -1 SO 0 3 4 5 1 2 2.5 Figure 5. A slice through the energy surface of Figure 5.2: Potentja!~energy 2 surface for the F.2.l 'a "- s 50 0 -50 -100 1 e <a u -120 > 1 1 1 1 1 1 o 1 1 .3 gives a contour representation of Figure 5.1 presents the Morse functions for HF and H2 molecules. requiring four dimensions to plot the surface. Figures 5. . the potential-energy surface consists of two valleys connected by a pass.reproduces the curve for H2 in Figure 5. The contour lines are isoenergetic. Figure 5. and for a diatomic molecule the potential energy can be represented with a Morse function V(r) = D [e.2 at large values of rHF in which F atoms exist along with H2 molecules . F + H-H = {F-H-H} = F-H + H Several sources provided the data used to calculate this potential-energy surface [30. a surface of energies . H2 reaction. for a collinear colliSion between a diatomic molecule and an atom. 1 2 HH-HF ----- o 1.1: Morse potential for Hz and HF. we require three coordinates to describe the internuclear distances.196 Chemical Kinetics 5. I.2fir . and f3 is related to the vibrational frequency of the which H atoms exist along with HI' molecules . A slice at large rfjfj -. We might choose the two distances from the center atom and the bond angle as the three coordinates. A plot of the energy would be a threedimensional surface. The potential energy is a strong function of the distance of separation between the atoms. 32.2 and 53 illustrate. If we increase the number of atoms in the molecule to three. The potential energy is a function of all three coordinate values. Figure S. Moore and Pearson provide a more detailed discussion of computing potential-energy surfaces [26].reproduces the curve for HF in Figure 5. only the two distances relative to a central atom arc required to describe the molecule. Figure 5.2 shows a representative view of such a surface for the collinear collision between F and H2. If we fix the bond angle by choosing a collinear molecule. As the F atom is brought into contact with H2 in a collinear fashion. say..3 Elementary Reaction Kinetics 197 200 150 100 v (kcal/mol) o -40 -80 . The depth of the well is governed by the magnitude of D and the curvature at the bottom is governed by the vibrational frequency. 31].2r8r ] in which D is the dissociation energy. and we can view this case as a three-dimensional plot. Figure 5. r is the displacement from the equilihrium bond length.1. The potential energy can be expressed as a function of these two distances. of the molecule is related to the relative position of the atoms.

'" ~ -105 -110 -l1S -- . The reaction coordinate represents travel along the minimum energy path from reactants to products.4: Reaction-coordinate diagram. a reaction-coordinate diagram can always be constructed from a minimum energy path along a potential-energy surface. For this example. it is the arrangement of atoms at the point in energy space between reactants and products.~\ ~ ----=-<--. The saddlepoint location defines the relative positions of all nuclei in the system.5 3 Figure 5. there is a one-to-one correspondence between the energy associated with a rHF-rHH coordinate pair in Figure 5. the maximum value of the energy along the minimum energy path is a saddle point. Figure 5. However. Statistical mechanics uses the properties of individual molecules to describe the thermodynamic properties of the bulk system such as the energy.3. Because of the repnisive and attractive forces present during the collision.1 is a reaction-coordinate diagram. This transition-state complex is not a chemically identifiable reaction intermediate. and movement along this pathway is the reaction coordinate. The dashed line is referred to as the reaction-coordinate pathway. which is the average energy of all molecules in the system. in this case it is exothermic by 34 kcal!moL The barrier height of 4 kcal/mol between reactants and products is used to calculate the activation energy for the reaction. and two H's) at the saddle point is called the transition-stale complex. I:.3 Elementary Reaction Kinetics 199 3 I I I I I -90 2. The molecular structure (the relative positions of F. proceeds along the minimum energy path and ends in the HF valley at large rHH. The difference in energies between reactants and products is the heat of reaction.5 o \\ o I I I 0 1 2 3 4 5 6 7 0.3: Contour representation of the potential-energy surface.5 1 III\1~V - . The reaction starts in the H2 valley at large rHF.-- ~ I ::a ~ .:. is possible that depends on the values of rHF and rHH.198 Chemical Kinetics 5. Any change of relative positions of f and the two H's for the transitionstate complex (as long as the motions are collinear) results in the complex reverting to reactants or proceeding 10 products.5 1 2 2. which displays the energy change during this reaction..- -120 -125 -130 0. Potential-energy surfaces such as Figures 5. just as in Figure 5.5 11111\ -95 -100 If I 2 "$ f5 '" 1.3. There is a minimum energy path along the valley of decreaSing rHF and constant rHH' This path is shown by the dashed line in Figures 5.3 cannot be depicted for polyatomic systems with N > :3 because more variables wonid be needed to describe all the relative positions of the nuclei.2 and 5. The principle of microscopic reversibility for this elementary reaction implies that the same structure at the saddle point must be realized if the reaction started at HF + H and went in reverse. . Figure 5. This energy is found by summing over the product of all possible en- .2 and 5.3 and the energy presented in Figure 5. --.4.

3 Elementary Reaction Kinetics 201 ergies and the probability of finding each energy in the system. vibrational and electronic energy states in the molecule ( Component translation rotation (linear) rotation (noniinear) vibration electronic Units length~3 I I -- Expression (tLn = (2rrmkBT)3i2 h3 qrot= - 8rr2IkB T (Th 2 - qrot= 8rr'8rr 3 (fA IB Ic ) 1/2 (kB T)3i2 lfh 3 3N~6 - - I j qvib = n i exp (-hv. Helmholtz energy.26) . the degeneracies of the electronic states of HF.1 lists relationships needed to calculate the various terms of the molecular partition function. and rate constants using T5T. including the internal energy. the activity for any species j is aj = (t)j kJT The molecular partition functions can be used to calculate the equilibrium constant for the reaction between F and Hz. The molecular partition function per unit volume for an ideal gas is the product of the translational. enthalpy and Gibbs energy. and the electronic energy levels of HF. at some particular orientation of atoms.)H Hf. This same microscopic description can equally well be used to calculate the enthalpy or any other thermodynamic property of the system.!.. For this gas-phase reaction K Using the principles of statistical mechanics we can write for this gasphase reaction K = = amaH aFaH2 = (. In Our development of rate constants we make use of the molecular partition functions.. ('i) F ('i) Hz flidli V V (5./2k. We use X to designate the transition-state complex.1: Molecular partition function terms. A to denote one reactant. The partition function of the system Q is related to the molecular partition function of the individual molecules in the system. pressure..B 1 . H" F and H . It is also possihle to compute the chemical potential of a component in a mixture from the partition function. The individual probabilities are normalized with a quantity known as the partition function.24) kBT The concentration of the transition-state complex is required to express the rate as a function of concentration. If activity is defined as (5.the bond distances in HF and H2. entropy. Therefore. and B to denote the other reactant A+B--X Table 5. Further. F and H. )HF (.1_ fx (5./ kB T) Bei TL - I states qelec = 2. aAaB ~ This example serves to illustrate how a microscopic description .23) -":!. All thermodynamic properties can be expressed in terms of the partition function.can be used to determine a thermodynamic property.: i exp (-cd kB T) I Table 5. Q.. rotational. the vibrational frequencies of HF and Hz.exp (-hv. the reacting molecules (atoms) are in (~t (~)B (~ L fAj~ _.. H2. The reaction-coordinate diagram illustrates the idea that. This ability to take a microscopic description of the molecules and determine macroscopic properties is exploited to calculate equilihrium constants for a reaction.200 Chemical Kinetics 5. the partition function is a normalization function and We use a microscopic description of the molecules in the system to calculate the partition function. (5. ~) j = ( ~ ) tran QrotQ0bQelec a configuration that either goes forward to products or reverts back to reactants.25) and the equation of state is. TST is based on the assumption that the transition-state complex is in equilibrium with the reactants forming it.

6 linear) vibrational modes remalning because one was used in the argument containing v*.8 em-.33) which leads to qvib f = r _ 3 I exp(-hv.35) in which k' in which 3N . O. and it equals 0.24 and 5.8 cm./kBTl r = kCACB (5.202 Chemical Kinetics 5. the rate has units of molecules/time· volume.L This limit is sometimes referred to as the claSSical high-temperature limit.5 modes describe a linear molecule).28) TST invokes the idea tbat the rate in the forward direction (H2 + F HF + H) is equal to the number of transition-state complexes moving to the right of on the reaction coordinate diagram.30) Equation 5.27 and solving for ex gives V 'ran qco.32 except for the concentration of reactants. For the example worked here.1 for a temperature of 300 K. nonlinear molecule comprised of N atoms (3N . represents all the terms that are composition independent. The variable k is the rate coostant and many texts consider this to be a compositionindependent term.qelecqvib 1 * exp(-hv*/2kBT) exp ( hv* / kEn (5.vith Equation 5.31 becomes * kBT V x = V tran qrotqelecqvib hv* (q)* (q) kBT Combining this result .29 gives r=h(~t(~)B (~): ¢A¢BZ ¢x CACB (5./2kBn l-exp(-hv.27) I I 1 At this point. TST assumes that the complex either moves to the right or to the left within the frequency of the single vibration (molecular motion) that best represents the bond to be made (or broken). The composition dependence is absent for ideal gases and it is reasonable to neglect . We refer to this special frequency as v*.1 at a temperature of 28.30) h (~L (~)B (~): (5.800 K for a stretching frequency of 2000 cm. (5. the quantity exp(-hv. We then rewrite the molecular partition function for the transition-state complex as q) (V x (q) = Combining Equations 5. Most rate expressions use the form of Equation 5.36) This approximation is valid for hvi/kBT s. A frequency of 20. and we can write (5.29) The vibrational partition function is in which q.l . In the limit of a very weak vibration (small Vi). The argument hv / kB T equals 0. I (Cj) = l.31) Cx = (5.3 Elementary Reaction Kinetics 203 then aAaB = ¢x fAls 1 ¢A'i>B fx ZkBT CACB (5.34) Equation 5.6 vibrational modes are needed to describe a polyatomic. we assume a weak bond is made (or broken) as the the transition-state complex transforms to products.34./2ksT) 1. Since movement along the reaction coordinate describes molecules (or atoms) coming together and bonds rearranging.ib now has 3N 7 (or 3N .1 at a stretching frequency of 20. For a reaction involving ideal gases.1 is reasooable for a rocking motion or a hindered rotation about a bond. The rate "constant" is not strictly constant. At the classical high-temperature limit (shown in Equation 5.32) The rate constant contains all the terms on the right-hand side of Equation 5.exp(-hvdkBn ~ kBT (5. It changes with temperature and even composition.33 can be rewritteo as (5.

. TST teaches us that elementary reactions depend on the amount (concentration) of each reactant ralsed to the absolute value of its stoichiometric coefficient because of the equilibrium assumed between the transition-state complex and the reactants.. Because the masses of the transition-state complex and the trioxane ate equal. The Arrhenius rate expression is one very common way of representing the temperature dependence. (5.38) rj "'" k i Cj c jV" . ko ~ lO'5. The temperature dependence is rigorously found by accounting for the effect of temperature on each term in Equation 5. The ring expands during the reaction. The composition at equilibrium is found by expressing the activities in terms of concentrations.D"-. in principle. .O. and the experimentally observed rate constant is [16].33. / 0 CHz jEl'.lThere is an exception for unimolecular reactions in the lovv~pn.38 cannot be set to zero to define the equilibrium concentration.3. to calculate the rate of any elementary reaction. Recall also that the production rate of any species j is Rj = (5.5-trioxane at 750 K.4 kcal/mol Solution It is reasonable to assume the reacting mixture is ideal and k is given by Equation 5.5 ± 2. Applying Eqnations 5. although true in general. The reversibility condition.32 and a potential-energy surface. (~) is the same for the complex and trioxane molecule .1: Computing a rate constant for trioxane decomposition Using TST.'ssure k = kBT h (~).39) where ko is temperature independent. The equilibrium constant is equal to a ratio of activities.204 Chemical Kinetics 5.2.ODlPieX ('i) oxirane V (SAO) regime.3 Elementary Reaction Kinetics 205 compositional dependence of k in many situations. The vibrational frequencies and moments of inertia of the molecule and the transition complex are listed in Table 5.2S. Often ko varies weakly with temperature ko = ATE with . it is possible. nor is the ratio of the forward and reverse rate constants equal to the equilibrium constant for the elementary reaction. as computational chemistry develops better empirical methods for determining the potential-energy surface. may not be important in a particular reaction.2 :5 B :5 L The Arrhenius expression is accurate provided Ea » BkBT. k = koexp (-EalkBT) Note that Li may be small compared to ki' in which case the reverse reaction is insignificant.06 S·l Equation 5. This is a useful result because.38 and 5.37) L VijYi i". it becomes possible to predict reaction rates from first principles and compare them against experimental information or possibly avoid experimental determination of the reaction rate. This allows us to generalize the results illustrated above to all elementary reactions 3 and adopt the convention ~Vlj k -i (5. . HzC CH2 I I n n 0"-./. The equilibrium constant K is given by Ea = 47. With Equation 5.} n. predict the value of the rate constant for the unimolecular decomposition of 1.36.39 to N03 in the photochemical smog example gives Example 5. The decompOSition proceeds by the concerted rupture of three C. trao . JER.0 bonds in the ring to form CH 2 0 .

200. 752. 1400.5 compares the experimental and calculated values of the rate constant over the temperature range 700-800 K. . . A factor of two is not a large disagreement if you consider the temperature dependence of the rate constant.381 x 10 23J/K)(750 K) One of the characteristic features of many chemically reacting systems is the widely disparate time scales at which reactions occur.3025 100.1200.23 J/K)(750 K) 6. 978. kcal/mol = (l)exp ( (1.700.3 cm)(vjcm. 2850.626 x 10 34 1 8)(2.'UJ'e 5.100. 120.1 )) 1. 1400. 3025. k = (1. oxirane _n n 29 [=1 exp(-lj2(1.998 x 10lOcm/s) = 1. 1242.92 x 1O. the following is used in place of hj kB T 5.15.56x 1013 s-I)(1.92 x 1O. 1242. 2850.l ). 1178.I )) = 1 94 x 1O~13 .4 S-1 at tbis temperature. 1410.3.987 cal/mol K)(750 K) = -51400 cal/mol ) 96.oxirane 1. 1305. 1200.3 cm)(vjcm. Substituting all the individual terms into Equation 5. em trioxane Value I Using the vibrational frequencies listed in Table 5./(96. 249.4 Fast and Slow Time Scales he keT = (6.92xlO 3cm)(vcm I)) ____ ' t = 59 6 transition complex The reaction frequency is kET h ~ (1.3)(l20.5)(249. 1178. 2850.945.4) Tbis value compares reasonably well to the experimentally determined value of 27.05 x 10. 1122. 1100.b.92 x 10 3 em (1. 173. 2850. 30 exp(-lj2(1.524.200. It is not unusual for complex reaction mechanisms to contain rate constants that differ from each other by several orders of magnitude. 1481.JTAJiJC) complex (~IAIBlc) oxirane = v'(T25.complex . 1178. the vibrational partition functions are 296. 3025.0 125.1495.2.4 )(96.945.l )) = 3. 1200.15 =1 and the ratio of electroniC terms is 1.2: Parameters for the trioxane reaction.381 x 10. 1178. 2850. 466.25 x 10-16 l-exp(-(1.53)(596)(1. 1481.2850.296.9 S-I A ratio of the qrot reduces to (qrot) complex (qrot)oxirane (qrotlcomplex (qrotloxirane ( .53 .626 x 10 34J S Assuming degeneracies of unity the electronic terms are qelec. 1070.05 x 10.5. Fil.4)(96.40 yields Table 5.524. 1410.05 x 10.4 Fast and Slow Time Scales 207 Parameter frequencies.2 51. amu-A2 trioxane transition complex potential-energy barrier.4 qelec.oxlrane i=l tlvib.206 Chemical Kinetics 5. 1200.200.92 x 1O. 13 The vibrational terms use the frequency v in reciprocal lime and we are given the frequency in wave numbers (cm.4.945. 1495. 1235. Moreover..3025 qvib. complex qvib. complex moments of inertia. 1070. = q".4.15 ) = 14. 1383. 1200. 1000. 1481.2) = 1. 1481. 3025. 3025.945.exp(-(1. 96. the . To use wave number units.3 cm)(vjcm. 1305.

kz ~ L2 ~ 1 and initial conditions CA (0) ~ l.38 1000lT (1/K) 1. the physical insight provided by these approximations remains valuable. even stiff differential equation models with fairly disparate time scales.8 shows the rapid equilibration of CB and Cc at small times for increasing k2 holding K2 ~ k21 L2 ~ L We now analyze the kinetic model. measurement of low-concentration species. 5. that is not their primary purpose today. we reduce the full mechanism on the basis of fast and slow reactions. A1thougb these approx' imations were often used in earlier days to allow easier model solution. this remaining slow reaction is called the rate-limiting step.1 The Reaction Equilibrium Assumption 100 experimental k (1/s) In the reaction equilibrium assumption. The remaining. cdO) ~ 0 shown in Figure 5. Figure 5. so the second reaction equilibrates quickly. the first reaction is slow and can remain far from equilibrium for a significant period of time. In a given mechanism consist- 10 calculated using TST 1 L-~ __ ~ __-L__-L__ ~ __ ~ __ ~ __L -__ L-~ 1. and make the equilibrium assumption on the set of fast reactions.24 1. The next two sections describe two of the most widely Assume that the rate constants k2.42 1. some reactions may be so much faster than others.. Moreover the reductiou of complex mechanisms removes from consideration many parameters that would be difficult to estimate from available data. Because the only other reaction in the network is at eqUilibrium. that they equilibrate after any displacement from their equilibrium condition. Comparison of measured and calculated rate constant versus temperature for oxirane decomposition. first reaction is called the rate-limiting step. We can express the mole balances in terms of either the extents of the two reactions or the can· . On the other hand.4 Fast and Slow Time Scales 209 used methods of model simplification: the equilibrium assumption and the quasi-steady-state assumption.41) concentrations of highiy reactive intermediates may differ by orders of magnitude from the concentrations of relatively stable reactants and products. Most models. The time scale for this equilibration is about 0. CB(O) ~ 0. The remaining slow part of the system takes about 4 min to reach equilibrium.1 min. L2 » kl' L 1. k~2 are much larger than the rate constants k" k~l. Notice the characteristic feature is that eqUilibrium is now quickly established between species Band C.7 shows the solution for k2 ~ k2 ~ 10.3 1. can be solved efficiently with modern ODE solvers.44 ing of multiple reactions.26 1. these disparate scales often allow us to approximate the complete mechanistic description with simpler rate expressions that retain the essential features of the full problem on the time scale or in the concentration range of interest. On the other hand.32 1. By contrast. the slow.36 1.4.208 Chemical Kinetics 5.34 1. LI ~ 0. These widely different time and concentration scales present challenges for accurate estimation of rate constants. Now conSider what happens as kz and Lz become larger. It is perhaps clearest to start with the full model and demonstrate what happens if k2. and even numerical solution of complex models.6. We show that the rate of this slow reaction does indeed determine or limit the rate at which the concentrations change. But the equilibrium approximation is more general and flexible than this one case. If we take the extreme case in which all reactions except one are assumed at equilibrium. C (5. We may decompose a full set of reactions into two sets consisting of any number of slow aod fast reactions. slower reactions then govern the rate at which the amounts of reactants and products change. We illustrate the main features with the simple series reaction B k. Figure 5. Consider the solution to the full model with rate constants kl ~ 1.28 1. Figure 5.4 1.

for the batch reactor.4 0..6 ~ .CB. let denote a fast time variable via T t Figure 5.2 0. k2 = k 2 = 10.8 C '0 OJ CA .15 k2 k2 k2 k2 0.2 0 0 2 CB dt = Yl = klCA dt ="'2 = . k2 = L2 = 1. . but we start with the two reaction extents because there are only two independent reactions..35 S . Cc = Ceo + £2 (5.45 0. k 1 0.3 0. Cc in which.. Both approaches are instructive so we examine both..5.05 t Figure 5..:: " Cc A Cc ...25 3 4 5 0 0. The full model can be expressed as dEl 0.2 0. EdO) =0 " u " dEZ k2 C B - L2CC.4 0. 1. = t Figure 5. so a small change in t is equivalent In a large change in If we divide .7: Full model solution for k. j 0.6 0.. = 0. I.6: Full model solution for k.s . 1 0. 0.4 0. " c " u centrations of the three species.42) T 3 4 5 To describe the fast time-scale behavior displayed in Figure 5. '" " u " " .8: Concentrations of Band C versus time for full model with increasing k2 with K2 = k2! L2 = 1.1 '" ~ u 0 u 0 u " 0.8 CA 0.1 = 10 = 20 = 50 = 100 0..15 0..5.210 Chemical Kinetics 5A Fast and Slow Time Scales 211 1 0.:: 0.L.05 0 0. = = L2t T..25 0. L... CB .8.2 0 0 1 2 .

212 Chemical Kinetics 5.Cc = O. we can compute the relaxation on the slow time scale.5 and Exercise 4. Equation 5. E] (T) ~ 0. Now if we take the limit as L z .. and the product does not go to zero. The big advantage of the reduced model is that values of kz and k~2. we wish to impose the equilibrium condition K2CB . We can express the model either as a set of two ODEs or a DAE consisting of one ODE and one algebraic equation. On the slow time scale. Notice the time derivative of <2 should not be set to zero on the long time scale because dE2 dt = L2 (KZCB - ce) Although the term in pareutheses approaches zero for a fast second reaction. and setting the equation to zero. In the reduced model.. overthe small time during which the fast reaction equilihrates. so we reexamine the equilibrium assumption in terms of the concentration variables. this equation is equivalent to £2 = = Ceo + £2 Table 5. K2 K2 . shown in the left.and right-hand columns of Table 5. We require species balances for reactors with flow terms. respectively. but are inadequate to describe systems "'lith flow terms such as the CSTR.9. these equations reduce to dE1 ODEs dE2 dT = K2CB - ce EdO) =0 ' E](O) ~ 0 =a dT which shows the extent of the slow reaction does not change. to obtain 4 4Nolice we again are choosing between a DAE and an ODE description of the model as in Section 4. Extents of reaction are convenient variables for a batch reactor. if one is deSired. In either case. K2. are not required. Notice the concentrations of all components are in excellent agreement after a short time. The steady state of the fast model becomes the initial condition for the slow mode!. The mole balances for the full model are .E ] + ---CBO 1 + K2 I + Kz - 1 ---ceo 1 + K2 (5.---Cco E2(0) 1 + K2 1 + K2 Cc Given this steady state achieved on the fast time scale. If we solve the slow time-scale model. In terms of extents. To determine a differential equation for E2.3 summarizes the results so far.42 into the right-hand side of the differential equation for £z.z approaches infinity. appears. shown in the top and bottom portions of the table. =0 =0 = K2 CB =0 Cc dE] -=Yt - dt DAEs dEz dT K2 I = ---CEO .00.3: Fast and slow time-scale models in extents of reactions. we produce the results in Figure 5. k. We com· pute the steady state achieved hy E2 on this fast time scale hy suhstituting Equations 5. A network with fast and slow reactions leads to fast and slow time-scale reduced models.4 Fast and Slow Time Scales 213 both full model equations by L2 we produce Equations Fast Time Scale(T = L2tJ Slow Time Scale(t) •• in which Kz ~ kz/L2 is the equilibrium constant of the fast reaction. we instead differentiate the equilibrium condition.43) This equilibrium relationship takes the place of the differential equa· tion for E2 on the slow time scale.43. the dynamics of the slow time scale are determined only by Y" verifying that the slow first reaction is the rate-limiting step. which are large and hard to estimate from data.19 Table 5. which yields E2s E2(0) £. only their ratio.3.

: co U '" + U -'" v U 0 u " § 0.equilibrium assumption . right portion of Table 5. " + v + ''" " '" C/O u V I u u " I 0 "' 0 '" l...Ll ~ O.4 Fast and Slow Time Scales 215 1 full model .Yz..~ u '"" Vl '" u Q) "' " " S u u " " .. ~ L2 ~ 10.J. we can add the ...0 ttl [.k.) u I 0 O. - .4.S .S...5 3 t Figure 5. which could serve as a replacement for anyone of the ODEs above d(cA + eB + if 0 dt eel = 0. f 0 " u 0 " v " Co 1. II 0 '" " " I 0. --=y? dec dt - '" " + J I V - s u" " ~ 0 I u C" '" v co '" N ~'13 "'II- " ~) t".6 ~ .s " on ..5 1 "'" u " "'I~ ~'" "" u 0 "' " 0" I '-'.44 and the mole balance for component A.. u " 0 0. ~ II I (/j '" N "' We know that only two of these differential equations are independent.. ~ u ~ 0" '" 0" " u " u u 0 <. w '" " """ <YII ". Deriving the correct slow time-scale model is less obvious and sometimes leads to confusion..9: Comparison of equilibrium assumption to full model for ~ "' " 0 0 " "' "" " 0 u '" " 0 '" v '" c c U v I """ II i= I '" co ~ u '" " C I " u '" u v I v I.214 Chemical Kinetics 5... we could add the three expressions to deduce the implied algebraic constraint.2 0 0 i kl ~ '" 0.: -'" u " 0 "' " + 0 Co u '" I 0 I I 'ti" .Il " 0' 0 0 <: '" 0 J I The short time-scale model is similar to the case with reaction extents. + CB + Cc = CAD + CEO + Ceo (5. in which the second reaction does not appear. CA.5 2 2.z II 1- ~I+ deB dt = Yj I . If we prefer differential equations.. shown in tbe lower.J c: 0 ':.4 + v -Ii II II JI~ "'I~ '" ~"" u"" ~i~ " vl~ " II 0 .44) '. then constitute a complete set of equations for the slow time scale. The equilibrium assumption is made by adding the algebraic constraint that Y2 = 0 or KlCB Cc ~ O.. Table 5.4 shows the short time-scale result in the left hand column.: I .- u "" " " 0'11"' ".: I - I v c. This algebraic equation coupled with Equation 5.

we see that the two series reactions in Equation 5.4 are zero-order terms in a full perturbation series solution. We start with the slow time-scale model given in Table 5.46) We see in the next section that this case is well described also by making the qnasi-steady-state assumption on species B. 35].4. Finally. (5. if the fast reactions are irreversible. one can also knit the slow and fast time-scale models together and obtain a more accurate reduced model valid at all times. It is usually simpler to compute numerically the solution to the full model if these rate constants are known. fast reaction. Kumar. we obtain a second relation. K2 I + K2 CB(O) =0 This result also is listed in Table 5.45 involves no approximation. The inner (fast) solution is valid in a boundary layer (small time) that decreases in size as the perturbation parameter goes to zero (L2 goes to infinity) as shown in Figure 5. and also discuss the resulting control problems arising with chemical reactors having both fast and slow time-scale kinetics [20]. ~~ ° fast reactions. we obtain from Table 5. further Simplification is possible.4 7) Under the equilibrium assumption. If we differentiate the equilibrium condition.3 and 5. which describes the irreversible reaction of A directly to C with rate constant k. Equation 5. In other words. Ramkrishna and coworkers [II] provide a more complete discussion of this approach. The slow and fast time-scale models presented in Tables 5.216 Chemical Kinetics 5.41. We see immediately from the slow time-scale differential equations that all concentrations are driven by y" again showing the first reaction is the rate-limiting step. However. K2 = 0. The inverse of the large rate constant plays the role of the perturbation parameter in this theory. = kJ CA. This knitting procedure is known as matching the inner (fast) and outer (slow) solutions in the mathematical theory of singular perturbations [29. Although this series solution may be of interest to gain analytical insight into nonlinear kinetic models.45) Notice we have not set Y2 = to obtain this result. A~ C (5. .46. which allows us to solve for both dCBldt and dcc Idt. or A going reversibly to B. let the second reaction be fast and irreversible in Equation 5.4 Fast and Slow Time Scales 217 mole balances for components Band C to eliminate Y2 (5. In the simple series reactions. the higher-order corrections require as much information as the solution of the full model. cd t) = 0. K2dcBldt ~ dcc/dt = 0. Equation 5. Notice that this result is not derivable by setting Y2 = in the full giving dec dt = --~Y.47. If we take the second reaction as irreversible in the backward direction. 23. the higher-order terms depend on the values of k2 and L2. Another interesting extension is the case of larger sets of reactions. Normally one is interested in the solution to only the slow timescale model and ignores the small errors at early times. depending on the magnitude of the equilibrium constant of the second. Christofides and Daoutidis provide further discussion of the application of the singular perturbation method.41 may reduce to A going directly and irreversibly to C.8. eB (t) = 0. Higher-order corrections also can be computed as shown by O'Malley 129]' for example. The equilibrium assumption for larger sets of reactions is elegantly handled by finding the null space of the stoichiometric matrix of the ° cdO) = CBO + Cco In this limit the B disappears.4 cs(O) = CEO + Cco cdO) = 0 which describes the reversible reaction between A and B with no second reaction.4 and take the limit as Kz . Equation 5. and y.

we reduce the full mechanism on the basis of rapidly equilihrating species rather tban reactions as in the reaction equilibrium assumption. These observations are simple consequences of the physical pictures presented ahove and the errol' formulas derived from them. that the concentrations. hut as pointed out by Turimyi et a!. but must then test to determine if these reaction intermediates are QSSA species [38. short lifetimes.2 that reaction networks can involve the formation and consumption of intermediate species.cj) i=l nr (5. it is advantageous to reduce the number of estimated kinetic parameters. such as a constantvolume batch reactor. we still use these characteristics to identify prospective QSSA species. where k < n s . We show in the uext section how the solution of Equations 5. their net rate of formation can he set equal to zero. ions and molecules in excited states.c i= 1 nr 5 ) (5. The key issue is how well the approximate solution that invokes the QSSA describes the exact solution. B . The examples in Section 5. not much has been gained by reducing the set of ODEs required for the model solution. that the induction period is usually short. however. Given today's computing capability. 26]. This leads to the following set of algebraic equations for the QSSA species dt dCjS ~0~ 2.) do not necessarily make a good QSSA species. radicals. The idea to set the production rate of a reaction intermediate equal to zero has its origins in the early 1900s [2. After solving for these small concentration species. A detailed discussion of the error analysis is given by Turimyi et al.48. we write the following differential equation for each component dc- d( ~ Rj ~ 2.48) A system composed of ns species leads to ns differential equations. ' For a spatially homogeneous reaction system.49 leads to reaction-rate expressions in terms of only the non-QSSA species concentrations.4. Having stated that certain reactive characteristics (high rates of con- sumption. etc. [38. and that most QSSA species are radicals. Here we illustrate the features of the QSSA and its application to a QSSA species. highly reactive species that are chemically identifiable but unlikely to exist outside the reaction mixture.4] and has he en termed the Bodenstein-steady-state.278 Chemical Kinetics 5. CH3CO.: Vijri ~ 9(T.5. The conditions are: the error in the concentration of the important species (reactants and products) calculated with the QSSA and vvithout the QSSA (the exact solution) be small. A valuable feature of the QSSA is that it eliminates hard-to-estimate rate constants from the model as demonstrated in Example 5. If one knows all the k i values.: vijri ~ f(T. low concentrations. If experimental data are heing fit to a kinetic model.49) The system of differential-algebraic equations is solved subject to the same initial conditions. it is just as easy to simulate the entire set of ODEs as it is to solve the reduced set of ODEs. radicals. There have been several empirical observations or conelusions hased on the investigation of small model reaction systems that showed that the rates of' consuming reactions of QSSA species are unusually high.2 contained radicals (CH3. To illustrate the QSSA we consider two simple elementary reaction schemes. [38]. In some cases the intermediates are transitory. We use the term quasi-steady-state assumption (QSSA). and the net rates of reaction of QSSA species arc unusually low. p. one is then able to construct reaction-rate expressions for the stable reactants and stable products in terms of reactant and product concentrations only. We have seen in Section 5. A number of other qualifying conditions have been applied.4 Fast and Slow Time Scales 219 5.2 The Quasi-Steady-State Assumption In the quasi-steady-state assumption. and the exact solution is found by solving Equations 5. Examples of these transitory species include atoms. or quasi-steady-state assumption. 172] are not universally applicahle. which is the goal. and the time domain over which the QSSA is applied he selected to ensure the calculated error is minimized. The intermediate species are referred to as QSSA species [38). which may he valid in certain circumstances. When the transitory species (reaction intermediates) have certain kinetic properties that we discuss next. The QSSA is applied to k QSSA species. which enables their concentration to be determined in terms of reactants and products using algebraic equations. Scheme I is given hy A -'l B and Scheme II is given by A k . pseudo-steady-state. short induction times. CH3COCHZ and N03) and an atom (0).

6. Note CBs is not constant.10 illustrates that as k2 » kl.48 for Scheme I is CA = cAOe~klt k2 ~ kl k2 ./k l • The curve cCleAO = (I-e. Also see ExerCise 4. it changes linearly With the concentration of A. Figure 5.78 into Equation 5. Similarly. the rate of decomposition of B is much greater than its rate of formation. We next examine the effect of increasing k2 on the concentration of C. Substitution of Equation 5. kl (5.48 for Scheme II isS (5.52) CB = CAO Cc = CAO I (k2(l-e. figure 5. but whether or not the B equilibrates quickly to its quasi-steady value. CB = Cc = O.5 for the initial condition CAS = CAD. The solutions to Equations 5.48 for components A and C results in dCAs -.220 Chemical Kinetics 5. we write dCBs dt = 0 = klcAs ~ for Scheme I. for Scheme l. CBs = ees = O. batch reactor be CA = CAO. when component B is defined to be a QSSA species. for Scheme n.B .) (5.e.78 ) k'"~1 + k 2 CAs .59) "'The Laplace transform is a good approach for solvlng these differential equations.e-.61) k 2eBs which leads to (5.60 and 5.k .53) (5.55) (5.57) 1 2 3 4 5 figure 5..48 for components A and eB s - . which is a necessary condition for B to be a QSSA species..) -kl(J -' ---.52 versus time for different values of k..k .10 shows the normalized concentration of C given in Equation 5.60) = k (5.kl (rk" k . Substitution of Equation 5.61 are presented in Table 5.54) o (5. The exact solution to the set of Equations 5.10: Normalized concentration of C versus dimensionless time for the series reaction A .59 into Equation 5.51) (5.= - dt dt dccs k leAs I CAs (5. If component B is defined to be a QSSA species..4 Fast and Slow Time Scales 221 Let the initial concentration in a constant-volume.) _e-k2'») (5.C for different val· ues of k2/kl.56) (5.k . the exact solution is equivalent to Cc = CAO (1..50) (5.) is shown also (the condition where k2/kl = 00). The issue is not if CB is a constant.58) For k2 » k.k21kl = 1 k21kl = 10 k2/ k l=50 k2/kJ=oo The exact solution to the set of Equations 5. isothermal.

the reaction product C.l:ll '" "" " I I - c + .e. As kz/ kJ increases. the error decreases and the induction time (dimensionless because we plotted kIt) to reach an acceptable error..5 for the initial condition CAs = CAO.I N .'1 ... ~ c ~ ~ ~I ~I i Sll 'J1' '" " I I 10. .. s CI . <c.80 are presented in Table 5. '" + '" I ~ " " " u u u CJ ~ " II u "" " II 0 ~ u "' 6 c " 0 h U'> < "- " "' " " N h U'> 21 PO en . ...l:l U .~ '" " 0 ~ c Ii c I"" I"" J1 .~ i2 '£.:.222 Chemical Kinetics 5A Fast and Slow Time Scales 223 '" . I '<c. Turimyi et al.. Equation 5. '\' c i~ ~ C .111 -< . An error of 1% is reached at kIt = 0.80) ~ I I "5 -< Vl '" I '0 "'" 0 . ~ '" I ~ ~ 10. 2 . [38] have shown that the induction period to reach the QSSA is likely to be several times the lifetime of the longest-lived QSSA species. If Component B were a radical or atom. The fractional error between the QSSA and exact values for Cc is shown in Figure 5.:: t.79) (5. LO Ii 1 Ie "0 Vl ~ 10 :.. 10. it could be treated as a QSSA species provided k2 i k I were large enough because large kz / k 1 leads to small errors in the predicted concentration of product.. Figure 5. N " I I~ + I"': 1'< . m '.. I I - '£. -"'.11 illustrates several points..e: ..1 < v i I 0) '-' U ::. '" -. -? N + . 'e '< i3 ~ l "" ~ NI'" -" "".ith decreasing ~ c '" "" -" ~ ~i I _I"" I ! es + ~I"" "" "" ~i3 ~ " '" -..1 for k2/kJ = 104. and the induction time to reach the quasi-steady state.10..11. ~ CO R '£. tbe error in the predicted concentrations of the reactants and products must be acceptable and the induction time should be small. ~ ..~ "I -< "" "' "' t. v ~ u ~ S := s Q) " "" I ~ U Il! Vl The solutions to Equations 5. .:. ~ U il " v . CBs = eCs = O. The induction time should be such that the quasi-steady state is established rapidly (i.68). ~rIf. which corresponds to a conversion of about 10% for A. the amount of A that reacts before an error of 1% is reached decreases as indicated in Figure 5. C . As the lifetime of the intermediate product decreases (the reciprocal of b). ~ " u u II ~ 'i u .. such as 10. c N ""I 'i' '" ~ e. -I.S ~ C ~ ~ "" I El 0 Ul '" . Figure 5.. In this simple example the net rate of production of the intermediate never reached zero over the values of kIt examined.. <'_J --- '" " I 0 "" l~ N --::-- + .:i1..5 permit us to examine the error in the predicted concentration of an important species. At sufficiently large values of k21 kl the intermediate B can be considered a QSSA species because the error in the actual and approximate concentration of the QSSA reaches an acceptable level. .:. The results in Table 5. t.11 illustrates this change in induction time I".79 and 5.. h '" .11 for much larger ratios of k21kJ and for shorter dimensionless limes than are presented in Figure 5.e. - . demonstrating it is not necessary for the net rate of production of QSSA species to be zero. rf '~ i " I II c 0 0 c:. before a significant amount of reactant A has disappeared).10 to show CclCAO versus dimensionless time (kl t) for different values of k21 k J • The dotted curve shows ecsl CAO (Equation 5.65 is used in Figure 5.:l 1- ~ Co (} " " " " II ~ ~ '" " c VJ <: C/o Vl "" '" ~ if:: '"J1 .. h ~ h CO ~ R h C results in ~ (5. For the QSSA to be accurate.2 '. decreases.:: h ~ ~ ~ ~ ~ " .

81 to the components that are QSSA species. Figure 5. here T ~ (II LJ + 11 k2).5 1 1.B .6 0.2 more rapidly. Now both k.--- 0. but can be applied equally well to a PFR. As the lifetime decreases.72 and 5.9 1 0. In addition. Material balances for batch and plug-flow reactors are ordinary differential equations.0001 10& Ie-OS le-06 0 0.001 -' --.C Figure 5. This effect can be seen by comparing the two curves for LJ = kz = 100k l and k"l = kz = 1000kl.5 kl t 0. k-dkJ = 1. the error in Cc reaches an acceptahle level such as 10. J 2: Fractional error in the QSSA concentration of C versus dimensionless time for the series"parallel reaction.LI/kJ = 100. These algebraic equations can be used to simplify the reaction expressions and reduce the number of equations that must be solved simultaneously.12 was generated using Equations 5.1 . appropriate use of the QSSA can eliminate the need to know several clifficult-to-measure rate constants.8 0. The QSSA is a useful tool in reaction analysis. Lllkl = 1000. The required information can be reduced to ratios of certain rate constants. For this reason.5 kIt 2 2.2 0.4 0.224 Chemical Kinetics 5. we write for a QSSA species R.11: Fractional error in the QSSA concentration of C versus dimensionless time for the series reaction A ---.. In the next section we show how the QSSA is used to develop a rate expression for tbe production of a component from a statement of the elementary reactions. it is best to define the QSSA based on the production rate of species j. = 0. A=B-C.J u u u 0. Ldkl = 1000.01 <. Similar concentration plots can be developed for tbe components in Scheme II to examine the effect of increasing the rate of decomposition of intermediate B relative to its rate of formation. we would be justified in letting B be a QSSA species.1 0.0001 Ie-OS ° 0. j = QSSA species (5.5 3 Figure S. their material balances become algebraic equations.l and k2 need to varied.3 0.01 1 0. which can be more easily estimated from data.1 0. By applying Equation 5. and illustrate the kinetic model Simplification that results from the QSSA model reduction.81) The QSSA was developed for a constant·volume batch reactor. In summary. Note also that the error at long times does depend on the relative values of LJ and k2.4 Fast and Slow Time Scales 225 1 0. for different values of k2/kj. and there are subtle differences in the induction time for comparable T. . rather than the time derivative of the concentration of species j. Then provided T is sufficiently small and k.7 0. vi ~Id k2!k 1=104 0.77 to illustrate the effect of decreasing the lifetime T of B.J and/or k2 are sufficiently large.001 10 5 <.J u I 00 0.

84) (5.- .88.k4Cs - kSC2CS From Equation 5. ks C2 (5.1'l CH}COCHz . Apply the QSSA to each of these species Rz ~ () ~ k.82 and 5. an elementary reaction is treated as being at equilibrium.5 Rate Expressions 221 5. These algebraic statements come from applying the QSSA to reaction intennediates. acetyl and acetone radicals are ~ -k c. A.6.2: Production rate of acetone The thermal decomposition of acetone is represented by the following stoichiometry (5. + k2C3 . 2.!<:! CH3CO CH3 + CH3COCHz ~ Solution Let the species be designated as: Species + + CH3COCHZ (5. Perform the necessary algebraic manipulations and substitute the resulting intermediate concentrations into the rate of production statement from step 1. Heterogeneous problems are considered in Section 5. In some cases.k SC 2 C S R3 ~ 0 ~ k. k.85) h CH3 + CH3 COCH3 .90 gives [5 = . You may assume the methyl. The goal is to determine the rate in terms of measurable quantities. either by assumption or because of kinetic insight.82) (5.87) 3. their concentrations can be found in terms of reactant and product concentrations by writing a necessary number of algebraic statements.83) (5. and the temperature.9 ]) Adding Eqnations 5.92) . Identify the species that do not appear in equilibrium reactions CH3CO CO (5. . but additional constraints are placed on heterogeneous problems so the synthesis reaction example is presented after discussing these constraints.k3C.5 Developing Rate Expressions from Complex Mechanisms QSSA species. k.C. . such as the concentrations of the reactants and products.84 (5.226 Chemical Kinetics 5. 5. If the rate of production statement developed in step 1 contains the concentration of reaction intennediates that are QSSA species.C3 Rs ~ (5. The approach taken for homogeneous and heterogeneous reactions is similar. CH3COC2HS . (5. c. The following procedure enables one to develop a reaction-rate expression from a set of elementary reactions.1.89 Cs Use the follov.39. 1. 3 and 5 are QSSA species. C.86) and write a statement for the rate of production of these species using Equations 5.38 and 5.k.89 aod 5. CH3COCH3 CH3 CH3 CH4 CH2CO + In this section we develop rate expressions for the mechanistic schemes presented in Section mechanism to determine the production rate of Dcetone 1331. Name acetone methyl radical acetyl radical carbon monoxide acetone radical ketene methane methyl ethyl ketone Write the production rate of acetone using Reactions 5. Species 2..C2 + k 4cs .88) Example 5.90) 0 ~ ksc.89) ( . Az A3 A4 As A6 A7 As Formula CH3 COCH3 CH3 Cfl)CO CO CH3COCHz CH2CO CIi.

105) o Example 5. This result can be substituted into Equation 5..kSCIC3 (5. Alternately.104 (5. Assume atomic 0 and N03 radicals are QSSA species.102 and 5. =0 ~ k2CIC3 .91 and 5.106) 2NO + 02 (5.93) N03 + N02 N03 + . and products.L2C4 . we can use Equation 5.95 as the basis for designing experiments to detertnine if the rate of production is first order in acetone and to detertnine the magnitude of the first-order rate constant. Adding Equations 5.96) The intermediate N03 concentration is found by solving Equation 5.104 and rearranging gives derive an expression for the production rate of Oz.103 and 5.87 to give the rate in terms of measurable species.' (5.k3CIC4 . we could calculate the rate of acetone production.k2CIC3 + k~2C4 . SUbstituting Equation 5.101) (5.. kJ k-2 C4 C3= +--~~ (CIO?) k2+ks k2+ksc] .95) (5. NOz + hV. N02..). Solution (5.99) (5.228 Chemical Kinetics 5. kerf.]03 for e3.108) .98) (5.102) by defining an effective rate constant keff = 3 -k! + 2 Species 3 and 4 are QSSA species. NO 2<! in which the positive sign is chosen to obtain a positive conceutration.3: Production rate of oxygen Given tbe following mechanism for the overall stoichiometry 2NOz + hv ~ This result can be used to simplify the rate expression. k_2 NO + 0 N03 NO + 02 + N0 2 2NOz NO+02 -.103) This Simplified form illustrates the rate is first order in the acetone concentration.88 gives terms of the reactant. Applying the QSSA to these reaction intermediates gives R3 = R4 0~ klcl . based on kinetic theories.100) (5.92 into Equation 5. If.5 Rate Expressions 229 Inserting the concentrations ShO\\11 in Equations 5. The production rate should he in (5.94) 0+N02 Let the species he designated as: Equation 5.. we knew all the individual rate constants.104) Adding Equations 5.97) O+N02 (5. NO and 02.107 into Equation 5.0.94 can he Simplified to The production rate of molecular oxygen is RI = -keffel (5.105 (5.k4C2C4 (5.0.

4: Free-radical polymerization kinetics Polymers are economically important and many chemical engineers are involved with some aspect of polymer manufacturing during their careers.106 and 5.111 because the polymer concentration for long chains goes to zero.k. A simple polymerization mechanism is represented by the following set of elementary reactions. we set their production rates to zero =0 Propagation: R.110) j~l in which the rate of the jth propagation reaction is given by Tpj = kpCMCR j R 1 Making the QSSA for all polymer radicals. An example is the thermal dissociation The first term is the formation of growing chain i by propagation (initiation for i = 1).109) Equation 5. Consider free-radical polymerization reaction kinetics as an illustrative example. It is valid to perform the infinite sum in EqUation 5. and ktmn = k t is independent of the chain lengths of the two terminating chains. M. 0 Example 5. and the third term is loss of chain i by termination reaction with chains of all other lengths. Notice if we add these equations. Initiation: k. and RJ is a growing polymer chain of length j.108 Rs = k. is a dead polymer chain of length j. Develop an expression for the rate of monomer consumption using the assumptions that k pj = kp for all j. limj_oo Cl'j = O. Rj+l Rj +M kpj Termination: Rm +Rn ~Mm+n in which M is monomer.111) . We solve Equation 5. C2 (k2k3 + k3kS)CI + (k2k4 + k4ks)c2 + L2kS (5.CI _ k . of benzoyl peroxide as initiator to form a benzyl radical that subsequently reacts with styrene monomer to form the first polymer chaln. the initiation reaction generates the free radicals. propagation terms on each line cancel corresponding entries on the lines directly above and below. k2k4 C. The termination reaction presented here is a termination of two growing polymer chains by a combination reaction.109 is rather complex. You may make the QSSA for all growing polymer radicals. +M R2 +M - kpl R R =0 =0 2 k p2 k J 3 R3 +M 2 R4 + (5.230 Chemical Kinetics 5. under reaction conditions several terms in the denominator are expected to be kinetically unimportant leading to a simple power-law reaction-rate expression. which initiate the polymerization reactions.l1! for the total growing polymer concentration and obtain . Solution From the propagation reactions we can see the rate of monomer consumption is given by RM =- I jed Ypj = -kpCM I w CRj (5. _ . Polymerization reactions raise interesting kinetic issues because of the long chains that are produced. In free· radical polymerizations.5 Rate Expressions 231 The rate expression now can be found using Equations 5. the second term is loss of chain i by propagation with monomer.

6 times the ethylene concentration. The exact solution of this problem uses the methods (5. C2H4 and Hz. Assuming the radicals CH3.H"C" - (S. In this numerical solution.116) kSCflCC. We are interested in determining the effluent concentration from the reac· tor and in demonstrating the use of the QSSA to simplify the design calculations. C2H4 and CH4 are plotted in Figure 5. -' (5. The pyrolysis is performed in a 100 cm 3 isothermal PFR.k]CCsHs + k4CC. Solution The rate of ethylene formation is The rate constants are listed in Table 5.CCzl'" .H6 . From the initiation reaction.13. The feed consists of ethane in a steam diluent. The concentrations of the radicals. are on the order of 10.110 yields the monomer con· sumption rate RM Reaction 1 2 ~ -kpCM)YJ / k. the initiation rate is given by y} ~ k}c}.6: Ethane pyrolysis kinetics.lh ~ [) ~ k2CCH.4 1.2% of the molar flowrate of the other products.117) .6 for the elementary reactions. The preexponential factor Ao has units of S-l or cm 3/mol s for first.112) 3 4 5 LO 2. C2HS and Hare QSSA species. with the concentrations of each component given by H+C2H6 - 14 The results for CZH6. in which k ~ Aoexp(-E/RT). CH3.Tate of CH4 is only 0. C2Hs and H. develop an expression for the rate of ethylene formation.113) Notice that this result also provides a mechanistic justification for the production rate used in Example 4.kSCflCC. RH ~ RC2HS 0~ k3CCz!ls k4cHCC.and second-order reactions.115) (5. operating at a constant pressure of LO atm and at 925 K. 0 Example 5. The overall stoichiometry for the process is not a simple balance of ethane and the products. The following mechanism and their kinetics have been proposed for ethane pyrolysis 122] developed in Chapter 4 and involves solving a set of eight initial-value ODEs of the form subject to the inltial conditions.5: Ethane pyrolysis This example illustrates how to apply the QSSA to a flow reactor. Verify that this apprOximation is valid. ~ 0 ~ 2klCCzH6 .3 in which monomer consumption rate was assumed linear in monomer concentration.6 Ao x 10 17 x lOll x 10 14 x 10 12 x 1013 E(kJ/mol) 356 44 165 28 0 i Table 5. Note the molar fiOv. respectively.H6 (5.CC.114) Next use the QSSA to relate C2Hs to stable reactants and products ReH. The Hz concentration is not shown because it is almost equal to the C2H4 concentration.0 3. The inlet partial pressure of ethane is 50 torr and the partial pressure of steam is 710 torr.0 3.232 Chemical Kinetics 5. each of the molar flows are computed as the volume increases from 0 to 100.l1.k2CCfl. Ethane pyrolysis to produce ethylene and hydrogen also generates methane as an unwanted reaction product. The feed enters at a flowrate of 35 cm 3 /s. which upon substitution gives the final monomer consumption rate (5.5 Rate Expressions 233 Substituting this result into Equation 5.

117 gives (S.13: Molar f10wrates of C. such as 100 cm 3 If the reactor volume were very small (a very short reSidence time).115.118 into Equation 5.119..H6.123.124) Adding Equations 5..5e-05 3e-05 methane . If one had experimental data. - Y2 .. the error is less than 5% after 5..l1S) Inserting Equation 5.124) because these components are present at levels on the order of 10..5e-OS / / ---- -- --- _ (ll + \jI(ll)2 + klk4) 2k3 2k3 k3ks The student should prove that 5e-08 At 925 K. The error is less than 10% after 3...5e-05 / / / / (5. such as effluent ethane. 5. During the numerical solution the concentrations needed in the rate equations are computed using (5. hydrogen and methane concentrations for different residence times (VR I Qf). dNc/H6 (5.6 cm 3 Figure 5.116 yields and Equations 5.120) which can be rewritten as (5.122 and 5.14 provides evidence that the QSSA is valid for sufficiently large reactors. oj _. such as 20 em3 . ethylene.123) ethane 0 S '" / / Ie-OS / / / / / 5e-06 / / I / le-08 0 100 dV ~ -y.::. k ~ 0.]21) . The reader might question what has been gained in this QSSA example.5 Rate Expressions 235 3.797 4e-OS 3e-08 2e-08 The validity of the QSSA is established by solving the set of ODEs S-I. i..119) Finally (5. the full model solution requires the simultaneous solution of eight ODEs and the reduced model requires solving five ODEs.1 em 3 of reactor volume. and CHI corresponding to the exact solution.234 Chemical Kinetics 5. Shnilarly. Figure 5. one Would need to fit these data to find the five rate constants for either the (5.14 shows the error in the molar flowrate of ethylene that results from the QSSA. Note we can neglect H. subject to the initial molar flowrates. Ul § ~ 0 2. After all.H 4 ~ Y3 Figure 5. the QSSA would not be appropriate.116 and 5.9 cm 3 and less than 1% after 18..e. 5.6 less than C2H6.::. it takes about 3% of the volume for the error to be less than 10%.. ethylene OJ 2e-05 / / / / / / '" ~ :.Y4 + Ys 0 0 20 40 60 volume (cm 3 ) 80 --=y? dNcH4 dV dV • dNc. CH3 and C2HS in the molar flowrate balance (Equation 5. C2H. / / / 6e-OS in which k ...122) c 1.k3 .

This suggests we can neglect CH4 in a species balance when computing mole fractions.125 using Equation 5.14: Fractional error in the QSSA molar flowrate of C. full or the reduced model. the results in Figure 5.5e-05 ethane Ze-05 1.001 '" '0 '"" S 0 o 10 20 30 40 50 60 70 80 90 100 0 20 40 60 80 100 Volume (cm3 ) volume (cm 3 ) Figure 5. The reverse situation of knowing a mechanism but not the rate constants could pose a difficult optimization problem when fitting all the rate constants. Our intention in this chapter is to provide the student with a simple picture of surfaces and to provide a basis for to relate RC.797 s". However. Using the mass action statement to describe the reaction stoichiometry and the definition of the rate of productiou of a component in a single reaction nearly equivalent predictions are made for the Simplified scheme based on the mass action statement.15 shows that . Here one would expect a first-order expression to describe the appearance of ethylene adequately.121 for the rate expression. Reaction 5. they should be used because the QSSA buys very little in this case. to the result for CZH4 found in Equation 5. Figure 5.5e-05 3e-05 0.:i ~ -' 0.236 Chemical Kinetics S. 3. it is possible to solve the problem analytically. This Simple example demonstrates that when one has all the kinetic parameters..1 '" :a g ~ 2.---------------------------. is a minor product at this temperature and this also would be found in the experimental data.6 Reactions at Surfaces 237 1 . QSSA analysis predicts the ethane pyrolysis rate should be first order in ethane concentration and the rate constant is k = 0. Figure 5.l21.15: Comparison of the molar f10wrates OfC2H6 and C2H4 for the exact solution (solid lines) and the simplified kinetic scheme (dashed lines).ll" and RH. ver· sus reactor volume. Therefore the mass action statement (5. 0 5.13 demonstrate that CH.125) does a reasonable job of accounting for the changes that occur in the PFR during ethane pyrolysis.6 Reactions at Surfaces Many books have been written on this topic and it continues to be an area of intense research.Ii. and the QSSA analysis can provide a framework for simplifying the kinetic expressions against which the data are tested.5e-05 le-05 Se-06 ethylene '" co ~ <.

The manner by which the adsorbed speCies is held and the properties it exhibits once adsorbed determine the type of adsorption . A number of criteria have been applied to distinguish between phys· ical and chemical adsorption. This simple example illustrates the steps in a cat· alytic reaction . Table 5. the Langmuir isotherm.126) (5. fcc (Ill) ~~ bee (110) 00 fce (100) Property amount specificity heat of adsorption activated I Chemisorption limited to a monolayer high typically> 10 kcal/mol possibly Physisorption multilayer possible none low (2-5 kcal/mol) generally not Table 5. The dynamics of the process by which the incident adsorbate finds the adsorption site is used to construct a rate expression and rate constant for the adsorption step from first principles.7: Chemisorption and physisorption properties. Adsorption occurs when an incident atom or molecule sticks to the surface.16: Surface and second·layer (dashed) atom arrangements for several low-index surfaces. ment and different combinations of nearest neighbors that can exist at a surface. In some cases the distinction is not clear.7 lists several properties that can be used to distiuguish between physisorption and chemisorption. For example. bee (100) fcc (110) Figure 5.127) (5. The adsorbing species can be bound weakly to the surface or it can be held tightly to the surface. Catalysts.126 and 5. The density of atoms at the low·index planes of transition metals is on the order of 1015 cm~2 Figure 5. their properties. an fcc(ill) surface atom has six nearest-surface neighbors. This figure illustrates the packing arrange' where the subscript ads refers to adsorbed species and 5 refers to vacant surface sites.adsorption. Keep in mind the goal is to develop a rate expression from a reaction mechanism.128. and desorption of products also shown as part of Reaction 5.adsorption of reactants in Reactions 5. The interested student can find a more complete and thorough discussion of adsorption. For this discussion. surface reaction and desorption. and the nature of catalytic sur· faces are discussed in Chapter 7. A heterogeneously catalyzed reaction takes place at the surface of a catalyst. we approximate the surface as a single crystal with a known surface order.6 Reactions at 5urfaces 239 }§{.36. To facilitate our discussion of heterogeneous reaction kinetics let us conSider the oxidation of CO on Pd.1. it is necessary to discuss the elementary steps that comprise a surface reaction . an fcc metal. To streamline the pre· sentation. To understand how we construct sur· face reactions and develop rate expressions. We also need to have a model of a surface at which the elementary steps occur. and an fcc(llO) surface atom has two nearest·surface neighbors. This rate expression can be used to direct the regression of rate data when determining the parameters for a het· erogeneously catalyzed reaction.238 Chemical Kinetics 5. . The most distinguishing characteristics are the degree of coverage and the specificity. We find that the interaction between an adsorbing molecule and the surface of a single crystal depends on the surface structure.16 presents the atomic arrangement of low-index surfaces for various metals. the discussion is limited to gas-solid systems and a single chemical isotherm.128. reaction of adsorbed components in Reaction 5. an fcc(l 00) surface atom has four nearest-surface neighbors.127. The foUowing mech· anism has been proposed for the oxidation reaction over Pd(ll1) [9] 02 + 2S ~ 20ads CO+ S ~ COads COads + Oads C02 + 25 (5.3].128) evaluating heterogeneously catalyzed reaction kinetics and reaction· rate expressions.24.13. surface reaction and desorption in numerous textbooks [5.physical or chemical.

6 torr. Physisorption is similar to a condensation process and has practically no dependence on the solid surface and the interaction between the adsorbate and adsorbent. the surface sites can be divided into two groups. T) S is known as the sticking coeffiCient. Small heats of adsorption (2-5 kcal/moll do not always indicate physisorption. accommodation with the surface. This limit of a single layer or monolayer is exploited later to derive a statement of conservation of sites. The upper limit need not have a one-to-one adsorbate to surface atom stoichiometry.240 Chemical Kinetics S. Since all sites are equivalent. and is a function of the surface coverage and temperature. sites +. and binding to the surface are complex. but not always. The sorface-site balance is therefore which. The flux of CO molecules to the surface (molecules/area. and adsorbate-adsorbate interactions are neglected. Since most surfaces contain approximately 10" atoms/cm 2 . and entropy considerations are satisfied. Ad· sorption occurs when an incident CO molecule finds a vacant site_ The dynamiCS of the collision process.6 Reactions at Surfaces 241 Chemisorption: Chemical adsorption occurs when a chemical hand or a partial chemical hand is formed between the surface (adsorbent) and the adsorbate. The discussion that follows treats only chemisorption_ Figure 5. time) is proportional to the number density of incident molecules and the mean kinetic velocity. Therefore. equals 3. 300 K. the number of atoms in a monolayer strikes the surface per second. the exact functional dependence can be determined once the dynamics of the adsorption process are known [24]_ We use the Langmuir adsorption to model the adsorption process. however. for example.83 x I020PCO molccule/s'cm' at 1 Total number of surface sites per unit area Number of -1 vacant. 290 miJIion times the number of atoms in a monolayer strike the surface per second.17: Schematic representation of the adsorption/desorption process. per unit area i OCCUPi~d Number of sites per urHt area J .6 x 10. at 2. At one atmosphere. In general. adsorption proceeds if a vacant site is available_ The Langmuir process also assumes adsorption is a completely random surface process with no adsorbate-adsorbate interactions. Heats of adsorption greater than 10 kcal/mol are definitely associated with chemisorbed species. and is given by CO(g) CO(g) kads kdes 0 0 0 0 c c 0 c 0 0 0 0 c 0 c 0 0 c 0 c c 0 0 c c c c c c c Pd (111) surface -- Figure 5. The rate of adsorption can be written as the product of the flux times the probability of sticking rads ~ FS(8. it is best to look at more than one property when trying to distinguish between chemisorption and physisorption. leading to the specificity of the process. a gas such as N2 condenses on solids maintained at 77 K_ Adsorption is an exothermic process and the magnitude of the heat of adsorption is used to distinguish chemisorption from physisorption.17 depicts a schematic of a simple adsorption process. Not every collision leads to adsorption. vacant and occupied. The Langmuir adsorption model assumes that all surface sites are the same_ As long as the incident molecules possess the necessary energy. Just as water vapor condenses on any cold surface placed in ambient. for pressure in torr. the formation of a bond limits the coverage to at most one chemisorbed adsorbate for every surface atom. and the overall rate of this process can change for the same adsorbate on different metals or even different low-index surfaces of the same metal. humid air. Physical adsorption is used to measure the area of high-area oxide catalysts and oxide-supported metal catalysts_ Physisorption isotherms and their use are discussed in Chapter 7. saturation can occur after one-third of all sites are occupied_ Physisorption: Physical adsorption occurs once the partial pressure of the adsorbate is above its saturation vapor pressure.

: Cj j"d n. Since experimental results are often consistent wirh the precursor state in equilibrium with the gas phase.17 shows the model of a surface consisting of vacant sites and sites covered by chemisorbed co.Cea Solving for the surface concentration gives k.242 Chemical Kinetics 5. When the adsorption-desorption reactions are in equilibrium. (5. an equilibrium constant. the Langmuir description of the adsorption process represented by (5. associative Langmuir adsorption isotherm. and the forward rate of adsorption and tbe reverse rate of desorption. Since active catalysts generally are dispersed an high-surface-area inert carriers.18: Fractional coverage versus adsorbate concentration for (5. Figure 5. the amount of adsorbed CO is related to the gas-phase partial pressure of CO. Equations 5. the units for cm often are number of sites per gram of catalyst.130 represent the conservation of sites for noninteracting adsorbates on uniform surfaces. (5. S. The surface reaction. Figure 5.0 1.cco 6This is an oversimplification of the adsorption process. k. The co in the gas phase is adsorbing on the surface and desorbing from the surface.4 0. As long as the precursor state is jn equilibrium with the gas phase. The rate expressions follow directly from the reaction mechanism because the reactions are treated as elementary steps.129 and 5. Figure 5.2 0 0 50 - 100 150 Cco (arbitrary units) 200 250 = COads k_. are given b y6 CO + S 0. () -co = K]cco 1 + K. In the limit K 1Cco » I. Since adsorption is exothermic.132 and rearranging leads to the Langmuir isotherm for single-component.135) Reaction 5.8 0.129) 1 10. A more acceptable view is to have CO adsorb into a precursor state and move around on the surface until it finds a sultable adsorption site. K. Equating the adsorption and desorption rates at equilibrium gives k.131) different values of the adsorption constant.133) Substituting Equation 5. and the total number of surface sites. associative adsorption - Cco = ~~--- CmKICCO 1 +K]cco (5.134 and 5. K. we prescot lhe Langmuir model.132) Dividing both sides of Equation 5. Ceo = ~k CCOCv = K lCCOCv ~l The conservation of sites gives (5.130) Equations 5. The units of the rate are mol/time·area.O.] Bco 0.ccoCv = L. decreases with increasing temperature and higher CO partial pressures are required to reach .131 gives the same result as is obtained by a more accurate description of the adsorption and desorption dynamics.18 presents the general shape of the fractional coverage as a function of pressure for different values of K.134) (5.135 asymptotically approach and J.6 0.0 Dividing by the total number of surface sites leads to the fractional form of the site balance. respectively.133 into Equation 5.134 by the total concentration of surface sites cm leads to the fractional form of the siugle component.6 Reactions at Surfaces 243 or Cm = Cv + 2.

From Figure 5..136. ~ ~ 1<. .model o data 2.4 .2 2 "be . Carbon monoxide is known to adsorb associatively on ruthenium at 100'C. the concentration at which this saturation occurs is a strong function of K.94 300 2. ~ .. In all cases the surface saturates with the adsorbate...20 we can estimate using least squares the slope.8 "'" 0 u 1. 1. are assumed to be in adsorption-desorption equilibrium along with CO.89 Jimol/g o We now consider multiple· component adsorption..134 and 5.. em. The linear form is ohtained by taking the inverse of Equation 5... however.6 1.20 plot the data for Equations 5. and the zero concentration intercept.19 and 5.. we add the following two reactions B+ S D+S k_ 2 kl k_. Bads Dads Figure 5.model o data 7 Figures 5. Single component adsorption is used to determine the number of active Sites.2 o -' for the slope. Sr---------------------------------~.. Pco (torr) 1 4 6 8 10 12 14 16 18 Figure 5.28 150 1. respectively. . The adsorption data can be fit to either the Langmuir isotherm (Equation 5..6: Fitting Langmuir adsorption constants to CO data The active area of supported transition metals can be determined by adsorbing carbon monoxide..> 1..19: Langmuir isotherm for CO uptake on Ru. Use the following uptake data for the adsorption of CO on 10 wt % Ru supported on Ah03 at 100'C to determine the equilibrium adsorption constant and the number of adsorption sites.... Example 5.134 and multiplying by Ceo Ceo = _1_ + Ceo (5.244 Chemical Kinetics 5.6 Reactions at Surfaces 245 saturation.. and then calculate KJ = 0. If other compo· nents.35 g/Jimol.. 0..1S illustrates the effect of K on the coverage.77 250 1. Figure 5.06 400 2.20: Linear form of Langmuir isotherm for CO uptake on Ru... em) for the intercept and l/e rn 2.136) Ceo K1c m em A plot of the linear form provides 1/ (K.134) or to the "linear" form of the isotherm. such as Band D.4 1. 100 1.21 co adsorbed (Jimol/g cat) Solution .822 cm 3 /g...l 3 4 6 8 10 12 14 16 18 = k.63 200 1.190 cm 3 jJimoi em = 2.

6 Reactions at Surfaces 247 Following the development that led to Equation 5.138) A The site balance becomes (5. The probability of finding pairs of vacant sites can be seen by considering adsorption onto a checkerboard such as Figure 5. cmKcoeco 1 + Kleco + K2CB + K3C]) (5. 4 can happen four ways. Therefore. Hill [15J and Kisliuk [18.0 and the probability a gas-phase molecule can dissociate on two adjacent vacant sites is p ~ O. with the center site always vacant. Figure 5. the rate in the forward direction depends on the probability of finding pairs of reaction centers. the probability of two adjacent . arrangement 3 can happen six different ways. the probability of dissociative adsorption is p ~ 10/24. such as dissociative adsorption. when multiplied by the respective concentration. The probability of any configuration is (l .141) Adsorption requires the 02 molecule to find a pair of adsorption sites and desorption requires two adsorbed 0 atoms to be adjacent.o)nom. Also.139 into Equation 5. in turn. Arrangement 1 can happen only one way. and 5 can happen only one way.137) (5. From Chapter 3 we know the equilibrium constant is related to the exponentiai of the heat of adsorption. in which n is the number of vacant sites and m is the number of occupied sites. For Figure S.21 presents two possibilities where half the sites (the shaded squares) are covered. This probability.22. Therefore.0)2. if components have very different heats of adsorption. describe how the probability of finding pairs of sites is used to develop rate expressions on surfaces. determines the component or components that occupy most of the sites. associative desorption.21: Two packing arrangements for a coverage of e ~ 1/2. we can show the equilibrium surface concentrations for Band D are given by = K2CBCl! CD = K3CDCv CB (5. which is exothermic. the more strongly a component is chemisorbed to the surface. Let two sites be adjacent if they share a common line segment (Le. When a bimolecular surface reaction occurs..16.21B has 10 vacant adjacent site pairs.140 have special significance and the magnltude of the equilibrium adsorption constant. and whether or not the adsorbates are mobile on the surface.and Hayward and Trapnell [13].246 Chemical Kinetics 5. depends on whether or not the sites are randomly located. or a bimolecular surface reaction. _ Ceo ~ B Figure 5. The probability of finding vacant adsorption sites is 0 s P s 1/2 for these two examples because the probability depends on the packing arrangement. Each of the terms Kicj accounts for sites occupied by species j. The probability for random packing can be developed by considering five different arrangements for a site and its four nearest neighbors. the denominator may be well approximated using only one Kicj term.139) Substituting Equation 5. as shown in Figure 5. Similarly. Figure 5. Boudan and Djega-Mariadassou [3]. 19] discuss lattice statistics and the probability of finding pairs of sites in two-dimensional arrays presented by the regular arrangement of surface atoms illustrated in Figure 5.132. For random or independent packing of the surface the probability for finding pairs of vacant adjacent sites is p ~ (l . However. the larger the equilibrium constant. which can be found by counting the line segments on adjoining squares.140) The terms in the denominator of Equation 5.132 and rearranging leads to a different isotherm expression for CO.21A the probability an incident gas atom striking this surface in a random location hits a vacant site is 1 .21. do not count sharing a vertex as being adjacent) and let 0 equal the fraction of surface covered_ The checkerboard has a total 24 adjacent site pairs. arrangement 2 can happen four ways because the covered site can be in four different locations. The CO oxidation example also contains a dissociative adsorption step (5. Therefore.

+ 25 = 20ads k-l k.e)3 + 3(1. however.134 and 5. Example 5.146 are different because each represents a different gas-surface reaction process.8) 2 e 3 P = (1. (3/4)4P2. Both share the asymptotic approach to saturation at suffi· ciently large gas concentration. At equilibrium. (5. + 3(1. which can be found using the site bal(5. the rate of adsorption equals the rate of desorption leading to 5 2 Co = . and rearranging leads to the Langmuir form for single-component dissociative adsorption Co = 1+~ sites being vacant is (5.fKlCO.146) with one atom in the center and the other pointing either north.8)48 + 3(1 8) 3 8 2 + (1.8) 2e The Langmuir isotherms represented in Equations 5.129. Returning to Reaction 5. The probability of finding other pair comhinations is the product of the fractional coverage of the two types of sites.6 Reactions at Surfaces 249 L __ • r-~ 3 4 (5.145) Figure 5.149) Cm=Cv+Cco+Co . Combining Equation ~- (5.jKICO.148) The remaining unknown is ance cv.142) An incoming gas molecule can hit the surface in one of four positions.e)e2 + e l 3 Using a binomial expansion. Cv ~ (5. Therefore. we use v • v to represent the probability of finding pairs of vacant sites for dissociative adsorption. cm.e)5 + 3(1 .e)2[(1. east or west. (x + y)3 = term in brackets can be written as X3 + 3x 2y + 3xy2 + y3. For random molecule orientations. (l/2)6P3 and (l/4)4p4. 02 + 2S k.143) Co = VKzco. the probabilities that a molecule hits the arrangement with a correct orientation are: PI. south. the probability of a suc· cessful collision is P = PI + 3P2 + 3P3 + P4 P = (1.7: Equilibrium CO and 0 surface concentrations Determine the equilibrium CO and 0 surface concentrations when 02 and CO adsorb according to CO+ S = COads L. e e Solving for the surface coverages in terms of the concentration of vacant sites gives Ceo = K 1 ceoc'v - (5. Equation 5.147) and using concentrations of vacant sites instead of fractional coverage.22: Five possIble arrangements around a single site.8) + eJl Solution At eqnilibrium the net rates of the two reactions are zero (adsorption rate equals desorption rate) or p=(1-e)2 Therefore.248 Chemical Kinetics S.8)'[(1. the = k_2 k. we write the rates of dissociative adsorption and associative desorption of oxygen as (5.144) The units on kl are such that rads has the units of mol/time' area. 20 ads P = (1 .145 'Nith the site balance.141 0.

dissociative adsorption of 02.CCO + . If we hold one of the gas-phase concentrations constant and vary the other. Cco CmKICCO = c:---ccc:-'"=-="'i.. A large reaction rate in Equation 5.=== -- _ v em C = 1 + K.) and plot this production rate versus dimensioniess concentrations of CO and 02 in Figure 5.JKzC02 We next substitute the vacant site concentration into Equations 5.250 Chemical Kinetics 5. If this reaction step is essentially irreversible.23. Figure 5.8: Production rate of C02 Find the rate of C02 production using the reaction mechanism listed above for CO oxidation..24 shows the result when each of the dimensioniess gas-phase concentrations is held fixed at 1.147 and 5.l (hc. = Rco. Notice from the plot shown in Figure 5. increases.151) Note this rate expression is based on the probability of finding two dissimilar paired adsorbates.cco) changes in slope from 1. Assume the 02 and CO adsorption steps are at equilibrium.23. 1 + K. High gas-phase concentration of either reactant inhibits the reaction by crowding the other speCies off the surface. and then decreases the rate upon further increase in concentration. and the bimolecular surface reaction CO+ 02 + S Co = cmJKZCO.153) For ease of visualization. the rate expression is (5.jKzco. we define a dimensioniess production rate k_l = COads k.. As the concentration of 0.152 requires large surface CO and surface 0 concentrations. But if one gasphase concentration is large relative to the other. What are the slopes of the production rate at high and low CO concentrations? . then that species saturates the surface and the other species surface concentration is small.150) The rate of C02 production is given by (5.148 to give the surface concentrations in terms of gas-phase concentrations and physical constants 1 + K 1cco + .6 Reactions at Surfaces 251 Combining Equations 5.152) The surface concentrations for CO and 0 were determined in the preceding example.. Notice that the production rate goes to zero if either CO or 02 gas-phase concentration becomes large compared to the other. which causes a small reaction rate.CCO + JKZCO.149 yields Solution (5.0. This feature is typical of competitive adsorption processes and second-order reactions. Notice again that increasing the gas-phase concentration of either reactant increases the rate until a maximum is achieved. we are taking a slice through the surface in figure 5.147-5.24 that logR versus log(K.152 gives the rate of C02 production (5. Make a log-log plot of the production rate of C02 versus gas-phase CO concentration at constant gas· phase 02 concentration. 2S k2 20ads L2 k3 ~ fI. Substituting these concentrations onto Equation 5.0 as CO concentration increases.0 to -1. Example 5. the slope COads + Oads C02 + 2S The reaction requires COs and Os to occupy adjacent sites. _ o The CO oxidation reaction involves associative adsorption of CO.

.01 ---- Oads + H2(g) H2(g) + 2S Hads ~ H20(g) + S k.2.23. k~.158) 0.. The reaction is proposed to proceed as follows over a ruthenium catalyst [8]. Example 5.15 0.0] 0.160) (5.25 0.2.162) k. Hads k-7 ks - S 1e·05 0. But now it is necessary to add a site· balance expression (Equation 5.2 0.153 are known as Hougen· Watson or Langmuir·Hinshelwood kinetics [17].156) (5.9: Production rate of methane 10 1 0.129) in Step 3. slices through Figure 5.5.1 Figure 5.153. 0.1 0.001 0.001 2Hads CHads + Ol2ads + CH 3s + S S Cads + k~5 (5.157) (5. (LOOOI CHads + Hads CH2ads ~ + k_6 k.5 to -0.252 Chemical Kinetics 5. R 0.24: Dimensionless C02 production fate versus a single di· mensionless gas~phase concentration while holding the other fixed at 1.05 a 100 100 Rate expressions of the form of Equation 5.1 1 10 lOO lOOO 10000 CH3ads + Hads CH4 (g) + 25 dimensionless concentration Figure 5. The development proceeds ex· actly as outlined in Section 5.159) (5. We complete the section on the kinetics of elementary surface reactions by returning to Fhe methane synthesis reaction listed in Section 5. These values also can be deduced by taking the logarithm of Equation 5.0. You may assume the reactions shown to be reversible are at equilibrium and that tbe surface is dominated by adsorbed CO.23: Dimensionless C02 production rate versus dimension" Develop a rate expression for the synthesis of methane. 0 0. This form of kinetic expression is often used to describe the species production rates for heterogeneously catalyzed reactions.161) (5. k. 1 and one possible mechanism is i{ versus K 1Ceo R versus K2Co.2 + Oads (5.154) less gas·phase CO and 02 concentrations.6 Reactions at Surfaces 253 changes from 0. The overall reaction is (5.155) S k.1 CO(g) + COads + S h k~l COads Cads (5.

3/2K4 (C6)3/2 -K ="-~ 1 C1 The rate of methane formation is given by Reaction 5.167) Substituting Equation 5..-:-. but this statement does not describe how the reactants convert into the products.JC3 R12 = yk3ks KzK5K6K7vC6~ (5.166) We are told the surface is saturated with CO.163) Rl2 = yk 3k8K.167 into the expressions for the hydrogen and carbon monoxide surface concentrations. • Most reaction processes consist of more than one elementary reo action.e.164) From the reaction stoichiometry. Oeo " 1.164 and rearranging gives 7See Exercises 5.254 Chemical Kinetics 5.165 into Equation 5. If only CO adsorbs Co 03 = =2. which simplifies to From statements of equilibrium for the reversible reactions C4 Cg = o 5. = em Therefore KIC1 » 1 and -' . i.162. CHZads CHJads CH4(g) Name carbon monoxide vacant site adsorbed carbon monoxide adsorbed carbon adsorbed atomic oxygen hydrogen water adsorbed atomic hydrogen adsorbed methyne adsorbed methylene adsorbed methyl methane r. A3 A4 As COads Cads Oads A6 A7 A8 Ag AlO All A12 H2(g) H.O(g) Hads CHad. 9 and 10 into the expression for com· ponent II gives - We introduced several concepts in this chapter that are the building blocks for reaction kinetics.165) - = kg CSCll k3 C6 --~- Substituting Equation 5.7 5ummarv and Notation 255 Solution Let the species be designated as: Species A1 Substitution of this expression into Equation 5.7 Summary K 2-=C3 C2 C5 = Ks Cz _ ClO = K6--- C9C8 Substituting for components 4.j =C~. The reaction process can be represented by a mass action statement. • The reaction order for elementary reactions is determined by the stoichiometry of the reaction -3- Cll = Kz K sK6 K 7 -2-C2 c S R12 CgC3 (5. The atomistic description of chemical events is found in the elementary reactions.5. and combining the resulting expressions with Equation 5.9 and 5.168) mat ely make this assumption.KsK6 K 7C m ... First principles calculations can be used to predict the order of these elementary reactions and the values of their rate constants.163 gives Formula CO(g) S A. '" KICI cm (5.= 1 +K1c1 cm c. Cs = R7 and? (5. such as CzH6 . .C2H4 + Hz in Example 5.10 for B more complete description of when you can legiti" (5.166 leads to I (5.

FJ fugacity of species j standard-state fugacity of pure species j flux of species j from the gas phase to the catalyst surface degeneracy of the ith electronic energy level Planck's constant moment of inertia for a rigid rotor moment of inertia about an axis Boltzmann constant reaction equilibrium constant mass of a molecule (atom) molecular weight of species j moles of species j molar flowrate of species j probability of finding adjacent pairs of sites lifetime of a component c/>j S. total pressure partial pressure of species j product species in reaction i molecular partition function of species j qelec electronic partition function qrot qvib rotational partition function vibrational partition function Q Qj partition function volumetric flowrate at the reactor inlet reaction rate for ith reaction gas constant - production rate for jth species reactant species in reaction i time V(r) absolute temperature intermolecular potential energy reactor volume Notation activity of species j Ei compressibility factor energy of the ith electronlc level extent of the ith reaction concentration of species j steady-state concentration of j concentration of species j on the catalyst surface total active surface concentration (monolayer coverage concentra- Ei 8j Bu Vij fractional surface coverage of species j fractional surface coverage of vacant sites stoichiometric number for the jth species in the ith reaction symmetry number. and (ii) the rate of formation of highiy reactive intermediate species can be set to zero. . For heterogeneous reactions.8 Exercises 257 • Two assumptions are generally invoked when developing simplified kinetic rate expressions: (i) some of the elementary reactions are slow relative to the others. and is called the rate-limiting step. k1 k1 kB K The difunctional starting material tA) combines with the primary reactant (H) to form the monofunctional intermediate (C). surface reaction) are generally treated as elementary reactions. 1 for a heteronuclear molecule.1: Equilibrium assumption and nonlinear kinetics It is common to add groups to difunctional molecules in two consecutive steps: A+B = C. we add a site balance to account for the vacant and occupied sites that take part in the reaction steps. the fast reactions can be assumed to be at equilibrium. 2 for homonu" tion) concentration of vacant surface sites fJ if clear molecule T f. In many cases. • The chemical steps involved in beterogeneous reactions (adsorption. If only one step is slow. Exercise 5. which subsequently combines With B to produce the desired final product (D).8 fugacity coefficient for species j Exercises h I I.256 Chemical Kinetics 5. this single reaction determines all production rates. the concentration of these intermediates is found from the resulting algebraic relations rather than differential equations. desorption. one reaction is the slow step and the remaining steps are at equilibrium.

SFsOOOSFs (a) Write down the production rate of each species in the mechanism. We call this set of differential equations the full mode~ .2: Nitrogen dioxide reaction mechanism The following overall stoichiometry has been observed for the decomposition of gaseous dinitrogen pent oxide to gaseous nitrogen dioxide and oxygen [28]. The first reaction is not at equilibrium and SFsO and SFs(h are radicals. (b) After making the quasi-steady·state assumption (QSSA) for the intermediates. and 0. Then we can eliminate the ODE for component D in the simplified model and replace it with this algebraic equation. N20s NO + NO" = ~ NO.. Develop an expression for the rate of bis-pentafluorosulfurtrioxide decomposition in terms of stable molecules. (b) Let's assume the first step is very slow compared to the second. NO. the second reaction is at equilibrium compared to the first). ~ I.3: Using the QSSA Consider the simple series reaction mechanism taking place in a volume batch reactor B=C Lz k. (c) What are the differential equations for concentrations of components A. (e) Now solve the Simplified model. write down the production rate of the reactants and products. Exercise 5.4: Applying the QSSA to develop a rate expression The thermal decomposition of bis-pentafluorosulfurtrioxide (SFs0 3 SFs) SFS03S FS - Consider the follovdng proposed mechanism to explain this as easily as the set of three ODEs in part 5. showing cf(t) for the full model and the simplified model.1e Why might you prefer the simplified model anyway? Be specific. for species B? (e) Make the QSSA on species B and write down the resulting differential equations for species A and C. either analytically or numerically. Exercise 5.Os. Make three plots.8 Exercises 259 (a) Write down the material balances for components A. + SFsO. (d) Plot CA. + NO] NO + O 2 + N02 SFsOzSF s + I 20. 2SFsO + 0. C and D if this reaction takes place in a batch reactor. Band C in the simplified model? Hint: don't just set rz ~ 0 in the full model but eliminate instead.e. You can solve this model analytically or numerically. constant~ (a) Write down the mole halance differential equations for species A. and rate constants kl ~ k. in the reactor? How many rate constants would you need to determine experimentally to use the QSSA model to predict these same concentrations? kul = k} SFsO + SFsO. is postulated to proceed by the follOwing mechanism for oxygen partial pressures greater than 100 torr [7]. Exercise 5. (c) How many rate constants would you need to determine experimentally to use the full model to predict the concentrations of N. B. 2N20s = 4N02 + 0. (b) For what values of rate constants would you expect the QSSA to be valid r. NOz + NO) - 2NO. one for each species. . k2 ~ Ll ~ 20. Solve for the equilibrium concentration of component D by setting the rate of the second reaction to zero (i. SFsO. Are you satisfied using the QSSA for this situation? What advantage does the simplified model provide? (d) A numerical ODE solver like Octave can solve the full set of four ODEs in part s. Cs and Cc versus time for the solution to the full model for initial conditions CA (0) ~ I. 2SFsO '} SFsOOSF.258 Chemical Kinetics 5. Band C. CB (0) ~ Cc (0) ~ 0.

CHO CH3+ CHO = 0. Ce. (c) Compare the answer for this simplified model that uses k. determine the fate of methane formation io terms of only stable molecules. has been proposed to involve the following reactions [33].= 0 in (d) in the second one. Feel free to solve it analytically or numerically. In this mechanism CH}. Predict the reactor volume required for 98% of the ethane to react. Then set Y2 . set Yz 0 and solve for the remaining ODEs for CA and CB. (b) Detailed analysis of the products in a PFR reveals that the reactions can be represented with the following mass action statement Exercise 5.5: Using QSSA to develop a simplified rale expression Example 4. The feed to the reactor consists of a mixture of ethane and NO with a molar ratio of 95% ethane and 5% NO. (a) Apply the quasi·steady state analysis to these reactions. In the first one.7. but for simplicity they .5 Ceo = 0 Which Simplified model correctly describes the full model in the limit of a fast second reaction? What went wrong in the approach that didn't work? Exercise 5. Hz and NO.8 Exercises 261 Exercise 5. where HNO. C. are reaction intermediates to find the rate of ethylene production in terms of stable molecules.H4 . and write down differential equations for CA and CR. (e) We "rite down two candidate simplified models to try to describe this situation. but now eliminate from the differential equations. H. Assuming the reaction intermediates (CH" CH3CO and CHOl are quasi-steady-state intermediates. and C2H. solve for Cc as above. ~ 100 Exercise 5.7 to answer the following are ignored here. questions. Plot the concentrations versus time for each modeL k.rr with the full solution to this problem in which all components are followed as illustrated io Example 4.7 presented elementary reaclions for ethane pyrolysis in the presence of NO. CH. CH3CHO. We call this set of differential equations the full modeL where tb) For what range of rate constants would you think it is reasonable to assume the second reaction is at equilibrium compared to the first re~ action? and it is necessary to follow only the concentrations of C2H6.260 Chemical Kinetics 5.8: Mechanisms and elementary reactions CH3CO CH3+ CO (a) What 1S the important difference between a reaction with an observed stoichiometry and an elementary reaction? (b) Ust two simple tests a mechanism must pass to be considered a valid description of an overall stoichiometry. rather than setting it to zero. and determine the activation energy for keff. Band C.7: Two candidate reduced models Consider the simple series reaction mechanism taking place in a constantvolume batch reactor (a) Write down the mole balance differential equations for species A. CH3CO and CHO afe reaction intermediates. (e) Solve the three models for the follOWing values (choose any time and concentration units you like). The radical CHO undergoes further reactions than are shown. eSO T. Use the data provided in Example 4. Assume the reaction takes place in an isothermal PFR operating at constant pressure (1 atm) and constant temperature of 1100 K.6: Acetaldehyde decomposition mechanism and kinetics The pyrolysis of acetaldehyde. The iulet volumetric flow rate is 600 cm 3/s.

C . B. Develop the production-rate expression for propylene.262 Chemical Kinetics 5.8 x 104 crn 3 jmol KIHzi 4.9. Exercise 5. Develop the production-rate expression for cumene. (d) Describe the transition-state concept and give a cbemical example. Develop production-rate expressions for the follo"ing two situations. Determine the fraction of surface sites that are covered by CO and atmnic hydrogen in the absence of any subsequent reactions. If this reactor achieves steady state. PUz = 6 atm and T "'" 398 K. (c) After glancing at the overall stoichiometry.9 x 103 cm3 /mol The gas-phase conditions are Pco 3 atm. and the adsorption and desorption of benzene and the surface reaction are at equilibrium. X . and the efflnent gas containing propylene. (b) The adsorption of cumene is rate limiting and irreversible. Carbon monoxide and hydrogen compete for surface sites in the adsorption phase of the synthesis process. What simple experiments can you do to determine which of the previous two sets of assumptions actually describe tbe chemistry of cumene disproportionation? Explain how you would interpret the results of your proposed experiments to make this determination. B+P a colleague claims that the production rate of curnene must also be the negative of the production rates of propylene and benzene. Propylene does not adsorb on the surface. How many unknown parameters does your model contain? . from the CSTR material balances what can you say about tbe relationships betvveen the steady~state production rates of gas~phase cumene. benzene. were collected for adsorption of oxygen on a Pd catalyst. (a) Derive an expression for the Langmuir isotherm to model the concentration of adsorbed oxygen in terms of the concentration of gas-phase oxygen.9: Disproportionation of cumene The disproportionation of cumene (C6HsCH(CH3)2) to benzene and propylene (a) Assume you are running this reaction in a well-mixed gas-phase CSTR containing a small.10: Overall stoichiometry and production rates Consider again the disproportionation of cumene described in Exercise 5. and the adsorption and desorption of benzene and curnene are at equilibrium. Exercise 5. Do you agree? Why or why not? Exercise 5.8.8 Exercises 263 (c) Name and describe the tV-lO major assumptions that are made to simplify the rate expressions for complex reaction networks..: B . propylene and vacant sites. The adsorption is dissociative 02 + 2X ~ 20· X k 1 k.25. The A· X notation means A adsorbed on the catalyst surface. B· X L3 in which C. and any unreacted cumene is withdrawn..11: Fractional coverage for multicomponent adsorption The conversion of carbon monoxide and hydrogen into synthetic fuels occurs over metals. [6] used initial rates to support the mechanism given in Exercise 5. P and X represent cumene. respectively.9. which are plotted in Figure 5.12: Oxygen adsorption The data in Table 5. Gas-phase cumene is fed to tbe reactor. What can you conclude about the production rates of gasphase cumene. benzene and propylene under these conditions? (c) Corrigan et a1. (a) The surface reaction is rate limiting and irreversible. B(g) + X = k. The adsorption reae!ions are: CO(g) + X H2 (g) + 2X = = CO· X 2H· X = Keo ~ 3. benzene. solid-catalyst bed. benzene and propylene? (b) Assume that you run the experiment while maintainingtdw concentra- proceeds by the following catalytic mechanism [6]: C(g) + X tions of adsorbed cumene and benzene compared to the gas-phase concentrations. C~ Exercise 5. X + Pig) k_1 b = k1 C· X '-.

Don't just plug these numbers into a least~squares calculation without looking at the re-plotted data first.000 K. We are interested in selecting an adsorbent material that maximizes the amount of propane that desorbs while changing from P2 to Pl.21 5.25: Adsorbed oxygen concentration versus gas-phase oxygen concentration. What is LlC for for K .000 cm 3 /mol The heats of adsorption are Mil ~ -20.0 76.59 Exercise 5. "0 0 "- .14: Competitive adsorption and reaction rate Consider the following heterogeneous reaction in which A and B associatively adsorb on a catalyst surface and undergo reaction to product P. .88 4.§ I" 00 2e-06 f0 . (a) Write down an expression for LlC ~ C(P2) . which we desorb while driving. - '" v blJ 4e-06 I3e-06 f0 Assume the propane "filling station" can charge the hed with a propane supply pressure of Pz ~ 10 atm and the hed is exhausted at a propane partial pressure of PI = 0.25.oo? Sketch a plot of LlC versus K .00 1..8 Exercises 265 CO 2 x 10 3 (mol/L) Co x 106 3. 0.000 Mi4 c -12.30 2. B+ A X k __ 2 B-X P-X+X P-X X+B-X p+ X k 4 (b) Use the experimental data to obtain preliminary estimates of tbese parameters.15 12. 1995 [25].06 cO 2 0. At 373 K the equilibrium constants for adsorption are Kl~190.000 Profe~sor cal/mol Eduardo Gland! problem came from a seminar given at UT -Austln by on April 18. 0 0.c(PJl. We revisit these data in Exercise 9.51 4. K is the adsorption equilibrium constant and P is the gas-phase cmKP 6e-06 5e-06 f~ I I I I I I I I 0 0 0 0 0 0 0 propane pressure.14 0. ~ 580.16 Exercise 5.2 15.5 107 122 137 152 l. Table 5.30 5.62 2. to use as a fuel replacement for liquid gasoline. You may wish to replot the data as l/eo versus 1/ JCQ.1 Figure 5.8: Gas·phase oxygen concentration and adsorbed oxygen concentration.8 provides the numerical values of the data plotted in Figure 5.10 9. Let's assume that propane adsorbs on the adsorbent material as a Langmuir isotherm. which subsequently desorbs.59 4.1 atm.2 6l.02 0.056. c ~ 1 + KP in which c is the molar concentration of adsorbed propane-.08 (mol/L) 0.04 0.2 91.13: Adsorption of propane 8 (mol!g·cat) Table 5. 0.95 5."'C~ is the monolayer coverage.88 2.12 . Assume we are going to place a large adsorption bed underneath our automobile to store adsorbed propane. Does the Langmuir model look reasonable for these data? Explain why or why not. K ~ 0.264 Chemical Kinetics 5.7 after we have developed more general parameter estimation methods in Chapter 9. I I I (b) What is the optimal adsorbent material (K value) and what fraction of le-06 I0 I 0 I I the adsorbed propane can be taken off as fuel when using this optimal adsorbent material? .000 30.35 5.

~ -20.77 1. Why do the curves for the two components have the shapes they do? Was studied over a wide range of partial pressures of reactants A and B.94 2. Why are the y-axis intercepts different for Y A ~ 0 and YA ~ 17 the gas contains a 50:50 mixture of A and B at a pressure of 1.15: Competitive adsorption Consider two gas~phase 2.06 maximum in the rate versus composition curve at higher pressures? 150 200 250 300 400 Exercise 5.000 cal/mol The irreversible heterogeneous catalytic reaction A+B-. Plot Y3 for the case where the temperature is 400 K.000 K. ~ . (b) Determine the magrtitude of rate of the surface reaction (Y.16: Associative versus dissociative adsorption The active area of supported group-VIII metals can be determined hy adsorbing carbon monoxide. In the experimental observations reported below the partial .Ru + 0 COg + Ru At 373 K the equilibrium constants for adsorption are KI 190. the mole fraction of A is 0. 000 6H. the total pressure is 1. Comment on the shape of the curves.C+D (a) Examine the effect of changing the temperature from 300 to SOD K on the fractional coverage of A and B.. Comment where T is in Kelvin.9.) at 373 Kif on the shape of the curves.0 atm. (An inert component that does not adsorb makes up the balance of the gas phase. (c) Examine the effect of changing the composition of the gas in contact with the catalyst on the rate.8 Exercises 267 The rate constant for r3 is k3 ~ 7. Why does one of the components appear to reach an asymptotic saturation level at a lower pressure than the other one? (c) Examine the effect of the gas-phase composition on the coverage of A and B. What is the value of the rate if a 50: 10:40 mixture of A:B:P is present instead? em ~ 0.Ru COg + 2Ru = C . Plot B.4 and BB versus T when the pressure is 1.1 and the mole fraction of B is varied between 0 and 0. the mole fraction of P is zero.28 1.0008 mol/g cat. (a) Develop a Langmuir-Hinshelwood rate expression (Le .17: Hougen-Watson kinetics The heats of adsorption are Mil ~ -30.Ru (b) VVhat is the concentration of total surface ruthenium atoms in terms of 580. Plot SA and eB versus pressure between 0 and 3 atm when the temperature is 373 K and the gas contains an equimolar mixture of A and B.21 components that adsorb associatively on a heteroge· (a) Test the data to determine if the adsorption is associative or dissociative.0 atm.63 1. The type of adsorption is a function of the metal type and the adsorption temperature. neous catalyst = CO .266 Chemical Kinetics 5. The follOwing data describe the adsorption of CO on 10% Ru on AI. Why is there a Exercise 5.000 cm'lmol mol of sites/gm of catalyst? Exercise 5.03 at 100·C Peo (torr) 100 CO ads (Ilmoljg cat) 1. Problems arise with the use of CO because it can adsorb associatively or dissociatively. (d) Now repeat part (c) for the 5 and 10 atm total pressure.6x IO'8 exp (-ISOOO/Tl (gcat' ) (8) (em') (mol) (b) Examine the effect of pressure on the fractional coverage of A and B.0 atm and the gas contains an equimolar mixture of A and B. the rate is expressed in terms of gas-phase components) for the surface reaction when you assume the surface reaction is the slow step and the adsorption and desorption processes are at equilibrium.) Discuss the shape of the curve. and products C and D. Plot 8A and 8B versus the mole fraction of A as it-is varied from o to I when the temperature is 373 K and the pressure is La atm.

I.18: Power-law approximation of Hongen-Watson kinetics The following catalytic reaction is conducted in a 0. At high Pc the curve has slope -2. . X + D(g) What do you expect the values of f3 and y to be and why? Exercise 5. The bimolecular surface reaction is . 2.D Hz + 25"::' 2B. You then fit the rate data to the following power~law rate expression - II ·X r = k CACS2CC C\' f3 y iX. Item Kl K2 K4 CA eBl C A.C . At low Pc the curve has zero slope.e. (b) A ·X+B·X-. At high PE the curve has a slope of -1. Explain your reasoning and justify your answer. Exercise 5. B2 and C are in adsorption· desorption equilibrium with the surface. Specifically." irreversible and rate limiting.D C+ Additional data for this problem follow.440 S. Use the observations to establish the reaeHon mechanism. Characteristics of the plot of log rate versus log P varied 1.D I Mechanism 1: A(g) + X B(g) + X C(g) + X D(g) + X Cc - A·X B·X C·X D ·X ratc-limiting step k3 em Value 130.A ·X Assume you can perform experiments that would enable you to determine the rate of the surface reaction as the concentrations of A.C ·X +D ·X Mechanism 2: A(g) + X B(g) + X C(g) + X D(g) + X .268 Chemical Kinetics 5. which candidate reaction mechanism is consis~ tent with all of the data. X -. At intermediate PE the curve has a maximum. A+S~Aads k.500 6.8 Exercises 269 pressure of one component was varied while holding the partial pressures of the other three components fixed.40xlO.d. Be and eel at the tabulated conditions.C ·X rate-limiting step D ·X A(g) + B . Observations: Partial Partial The mechanism for this reaction is shown below.19: Ethylene hydrogenation The hydrogenation of ethylene to hydrogen is found experimentally to follow the reaction orders given in the following table 134J. At low PA the curve has a slope of I.l b pressure varied A pressures fixed B.41xlO 8 1. Band C are varied plus and minus 50% from the values in the table. .. 2.C.25 em radius spherical pellet.C. '-.83xlO-s 1.5 Units cm'/mol cm3 /mol cm3 /mol mol/em' mol/cm3 mol/cm 3 g cat2 /(mol em3 s) mol/g cat - (a) Determine the fractional surface coverages of A. k.8 xlO. Cads + S B A. L At low PE the curve has a positive slope slightly less than L 2.4 L17xlO. At intermediate PA the curve has a positive slope less than I. Band C (i.. A.100 6.5 7. GA . 3. k] Aads + Bads ----.B.

Czarnowsld and H. Hayward and B. k. O..d. Photolysis of nitrogen dioxide to produce transient 0. Butterworths. 1984. Kinetics of Heterogenous Catalytic Reactions. [lOJ H. Neglect the reverse of Reaction 4. Eng. second edition. C2H4 + H2+ C2 H 4. Princeton University Press. HilL . Res. . S. Chern.. The kinetics and the mechanism of the thermal decomposition of bis-pentafluorosulfurtrioxide (SF. page 73. 5 25 Had" kl k -I k. Ertl. Neglect the reverse of Reaction 3. L Develop a rate expression for the rate of ethane formation for Scheme L Let Reaction 3 be the rate·limiting step and assume all other reactions are at equilibrium Assume that adsorption from the gas phase follows a Langmuir adsorption isotherm. Soc. Assume that adsorption from the gas phase follows a Langmuir adsorption isotherm. 1960. k_2 c. Le. F. MA. Theory of Adsorption and Catalysis. 1970. Catal" 58:170. H. [6] T.19 Bibliography The following steps have been suggested for the catalytic reaction mecha· nism [39. Ekerdt and A. The rate of the formation of ethane Is that of the rate· limiting surface hydrogenation step. W. New York.. G. 11:613-619. Kirk. S.. C2H4 ads ' 2Hads C2HS. Berichte der Bunsen-Gesellschaft fUr Physikalische Chemie.. [l2] A. New York. W. New York. 1979.Os. Djega·Mariadassou. Le.. Amsterdam. S. Develop a rate expression for the rate of ethane formation for Scheme II. the coverage of ethyl. For Scheme I. the rate·limiting step is Reaction 3.. Trapnell. R. K. There are two choices. the rate·limiting step is Reaction 4. assume the surface is either vacant or covered with adsorbed ethylene.3 S [4] D. Int. Reading. introduction roStatistical Mechanics. Woodruff. 1990. 1982. Some basic aspects of reaction engineering of precipitation processes. B.).:::: Eine Theorie der photochemischl'f!' Reaktions· geschwindigkeiten. Princeton. is Scheme I or II the more probable mechanism and why'! J. D. Chern. Addison·Wesley Publlshing Company. editor.]. Eng.. Harker and H. 34(10):32233230. L Hill. BeU. 1977. 85:329-397. Gandhi. Ind. Chern. For Scheme II. Kinetics of catalytic cracking of cumene. The influence of mass. Gaedtke and 2.. 3. In D. 103:496-508. Schumacher.C.:.85 0. Synthesis of hydrocarbons from CO and H2 over silica~supported Ru: Reaction rate measurements and infrared spectra of adsorbed species. N]. 1913. Engel and G.d" and then adsorbed ethane. the coverage of atomic hydrogen and adsorbed ethane are negligibly small. assume the surface is either vacant or covered with adsorbed ethylene and ethyl. fifth edition. [! IJ K. Phys. Chapman and L.53]. Boudart and G. Z. When performing a site balance. atomic hydrogen and adsorbed ethane are negligibly small. UnderhllL The interaction of chlorine and hy· drogen. OOOSF. [IS] T. Ramkrisbna. 1979. [9J T..74 -0. Chemisorption. Troe. volume four. Trans. where 5 is a vacant surface site. C2HS.. T. 1975. Kumar. Chem. There is a stepwise addition of adsorbed hydrogen to form an adsorbed ethyl group. Let Reaction 4 be the rate·limiting step and assume all other reactions are at + [IJ A. John Wiley & Sons. John Wiley & Sons. An Introduction to Chemical Engineering Kinetics and Reactor DeSign. G..1953. Clark. Primary processes in the photolYSis of N02. Johnston. Adamson. and R. The Physics of Solid Surfaces and Heterogeneous Catalysis. . + [3] M. Wash· ington. C2H6.]. 79(2):184-191. Bodenstein. Garver. In this mechanism etbylene adsorbs associatively and H2 adsorbs d!ssociatively. Chern.59 -0. 77(9):1153-1156. Physical Chemistry of Surfaces. M. [2J M. NO] and N. [8] J. 49(11):603-610. Elsevier. J. and D. [5J A. E. P. L. Academic Press. Phys. [7J J. [l4J c. Chern. Rase. 1973.210 Chemical Kinetics Catalyst Rh Ru Co Order in H2 0.]. Based on the experimental data listed in the table. Corrigan. Kinet..1995. 1913. [13] D. Frog. When performing a site balance.95 0. C. p..55 Order in C2H4 -0.

Quasi-unimolecular and quaSi-bimolecular steps lu complex reactious. Surfaces. SL4M Rev. 97:163. National Academy Press. pages 243-3J 7. Phys. Macmillan. ]. N. Inc. New York. and E. Hochgreh and F. Pearson. Phys. Tomlin. Solid catalysts and reaction rates. Investigation of an ab initio energy-surface for F + H2 ~ [33] F. Matranga. Daoutidis. Watson. F. Chern. New York. Chern. Chern. 47(7):1569-1579. Eng. 96:295-297.. Engle· wood Cliffs. SegeL Mathematics Applied to Deterministic Problems in the Natural Sciences. 1950. 1958. J. Soc. Chern. Slemrod. Hougen and K. [39J R. Phys. Chern. 52(7):3449-3457. S. PUys. D. 1967. and M. Modified semiempirica! approach to the H3 potential-energy surface. [38] T.. 23:693. L The uninhibited reaction. 53(8):1491-1504. Advances in Catalysis and Related Subjects. ]. Chern. Solids. Proceedings afthe Royal Society of London. W. 1981. Cornell University Press. Math and Physics. Ind. Chemistry in Two DimenSions. [34J G. The thermal decomposition of organic compounds from the standpoint of free radicals. Wojciechowski. 1957. R. Kumar. Somorjai. Niemantsverdriet.. C. The structure and activity of metal-ansilica catalysts. The mechanism of some chain reactions. W. [26] ]. Kinetics and Mechanism. In D. Phys. Schult and L L van Reijen. [20] A. Segel and M. J Laidler. L Schreiber. [40] R. ]943. A. Porter. 1991.3.. 1998. A. and V. J. New York. J. HF + H. The sticking probabilities of gases chemisorbed on the surfaces of solids. L Dryer. C. [21] K. Turimyi. 1974. E. ] 992. Christofides. Komarewsky. 35(5):529-541. O. J. On the error of the quaSi-steady· state approximation. G. [18] P. 1981.. The quasj-steady-state assumption: A case study in perturbation. Kisliuk.. Chemical Kinetics.. 1995. M. W. Solids. 1989:[36J G. 1958. Herzfeld. Phys.5-trioxane at 700-800 K. O'Malley Jr. Weston and H.212 Bibliography Bibliography 273 [16] S. [24J S. A. desorption.. Sci. New York. 13:1. Moore and R.1992. 1987. Eng. Lett. Washiugton. 29(3):319-322. ]. Chern. Surf. 1. diffusion. Eng. D. A. Semiempirical VB calculation of the (H2Iz) interactlon potentiaL J Chern. D. Chern. third edition. Rep. N. Slugular perturbation modellug of nonlinear processes with nonexplicit time-scale multiplicity. editors. Laidler and B. New Jersey. D. U 9J P. [28] R. New York. Kisliuk.. [17] O. A. National Research Council. and P. Phys. [37J R. Chemical Kinetics and CatalysiS. c. Sims. Sci. Chemical Kinetics. A. Pilling. New York. Chern. tenth edition. 1974. Amer. P. Eley. Chern.. Raff. 31(3):446-476. 5:78. Tolman. 1991. Rice and K. L Thompson. J. 1992. 1972. 1970. Ogg Jr.. 1924. 3:95. Phys. Lombardo and A. G. L Myers. W. Lin and L A. Schwarz. Duration of molecules in upper quantum states. 56:284-289. Ithaca. ]. 1934. Harper Row. Rev. Phys. E. Springer Verlag. Distribution of reaction products (theory). third edition. Stivers. 47(11):4751-4757. [22J K. [25J K. Storage of natural gas by adsorption on activated carbon. Sci. and L. [29J R. G. L. [30J L Pedersen and R. Porter. Fraukenberg. c.C. ]. 260A(1300):91-102. [27] Catalysis looks to the future.. R. New York. T. [23] c. Gland!. 1993. and reaction of gases on metal surfaces. Kinetics and mechanisms of the thermal decomposition of ethane. Chern. A. ].. [35] L A. VI. [31] ].. 18(4):572-573. Singular Perturbation Methods far Ordinary Differential Equations. van Santen and]. B. Academic Press. 1961. . John Wiley & Sons. Prentice Hall. [32] L M. Plenum Press. Decomposition of 1. The sticking probabilities of gases chemisorbed on the surfaces of solids -II.. Bell A review of theoretical models of adsorption. Palanyi and]. D.

we are again neglecting flux of mass through the surface of the volume element except at stream locations 0 and 1. We derive the energy balance by conSidering an arbitrary reactor volume element. we reqnire a model to determine the temperature of the reactor. . In this chapter. shown in Figure 6.1. and the total energies of . The molar concentrations of component j in the two streams are given by CjO and Cjl.1: Reactor volume element. ma Eo CjO v Figure 6. J General Energy Balance To specify the rates of reactions in a nonisothermal reactor. respectively. The volume element has inlet and outlet streams with mass flowrates rno and rnl. The temperature is determined by the energy balance for the reactor. as we did in deriving the material balance in Chapter 4.6 The Energy Balance for Chemical Reactors 6.

some cases one might need to consider also electrical and magnetic energies" For example. and W is the rate of work being done on the system.5 reduces to (6.m]E] ' '" Note that the units in the product mP/p are (m/t)(m/lt 2 )/(m/P).d jonic species between the plates in a battery cell. (il. the work done when moving the system boundary. allows us to rewrite Equation 6. W . 6.2 as d (_ . rate of heat } { rate of work } { added to system + done on system In terms of the defined variables. The rate of heat added to the system is denoted by Q. dE dt ~ moEo . Obvious contributions to the total energy arise from the internal. Recalling the definition of enthalpy. etc. we consider only these forms of energy. or expressed on a per'unit mass basis.2.2: Flow streams entering and leaving the volumO""element. In this case Wf is 1 ml (fI +K +1»1 + Q + Ws + lirb (6.5) Wf ~ vaAoPo .2 The Batch Reactor Since the batch reactor has no flow streams Equation 6. '-v-' The total energy may be regarded as composed of many forms. Consult Bird for a more detailed development from microscopic conSiderations IS.3) To calculate the work done by the flow streams.216 The Energy Balance for Chemical Reactors 6.2 The Batch Reactor 217 the streams are denoted by Eo and El.1 Work Term 6. we might conSider the motion of charg<.2) in which the hat indicates an energy per unit mass. fI ~ (j + P / p. 2 For our purposes in this chapter. rate a f energy { accumuI ' atlOn } + ~ Po Ao v Vo rate of energy rate of energy } . Q]PI 6. s and !irb . _) dt (U +K +$) ~ ma H +K +$ 0 W total work & flow streams + ~ shaft work + & boundary work (6. which are usually much smaller than the pressure force. kinetic and potential energies. Be aware that the opposite convention in which W points out of the system in Figure 6. and the exiting stream has pressure p] and velocity V]. and let AD and A] be the areas on the bounding surface where the streams intersect the boundary as shown in Figure 6.vIAIP] ~ QoPa -.1.2 Energy Terms It is convenient to split the work term into three parts: WI' the work done by the flow streams while moving material into and out of the reactor. The overall rate of work can then be expressed as +Q +W (6. leaving system entermg system { } { by outflow by inflow (6. the shaft work being done by stirrers. we can write Equation 6. H ~ U + PV.1. we assume the entering fluid has pressure Po and a uniform velocity Vo normal to the bounding surface.6) We also can express the volumetric flowrate as a mass flowrate divided by the density. Q ~ m/ p I We are neglecting any normal viscous forces. a unit of energy per time as we expect. The statement of conservation of energy for this system takes the form.1 as.1 is also in use. which gives (mI 2 /t 2 )/t. .1) Figure 6. compressors.

v. Taking the differential of the definition of enthalpy gives for V ~ VR dH ~ --' (6.nj (OP) av T.14) The first partial derivative is the definition of the heat capacity.OfTVR dt + L.nk (6.n} ap We now conSider several special cases.13) For single-phase systems. Notice thIs heat capacity is the extensive heat capacity of the reactor contents. Normally we neglect the work done by the stirrer. (6.VR dt ~ Q whIch upon substitution into Equation 6.7 gives dH dP· dt .OfT)VRdP + L Hjdn j ) (6.2 The Batch Reactor 279 In chemical reactors. If the reactor operates at constant pressure (dPJdt ~ 0) or the fluid is incompressible (Of ~ 0).9 can be written compactly as dH ~ VRPCpdT + (J .12) dU + PdVR + VRdP Forming the time derivatives and sobstitution into Equation 6.n. nt. we normally assume the internal energy is the dominant contribution and neglect the kinetic and potential energies.8 gives .".9 are the partial molar enthalpies. dT dP" dn} . Neglecting kinetic and potential energies and shaft work yields dU +pdVR ~ dt dt so Equation 6.11 yields (6. aT dT + V. is the coefficient of expansion of the mixture. Cp.7) in which Wb ~ -PdVRldt. VRPCr/It .10) Q Forming the time derivatives from this expression and substituting into Equation 6. We can express the intensive heat capacity on either a molar or mass basis. we consider the pressure as a function of T. and fornling time derivatives and substituting into Equation 6. then Equation 6.11j L [HI j ".nk dnt For reactor operation at constant volume. pressure and number of moles. ~V(l-Ofn To derive the constant-volume case.n) T. and express the pressure differential as dP ~ in which Of ~ (l/V)(av lonp.218 The Energy Balance for Chemical Reactors 6.9) (6.TVR ( 3P ) anj ] dnj TY.11 gives [ VRPC P . dV ~ 0. We choose to use the heat capacity on a mass basis. Hi OP) (. It is convenient to use enthalpy rather than internal energy in the subsequent development.TVR (a3P " T) ] ddTt + \ . unless the mixture is highIy viscous and the stirring operation draws signlficant power [13J.8) (6. so the total heat capacity can be expressed as Cp ~ VRPCP The second partial derivative can be expressed as ( OH) T. we consider the enthalpy as a function of temperature. The partial derivatives appearing in the sum in Equation 6.11) We note that the material balance for the batch reactor is (6.-H} dt ~ Q J (6. V(V ~ VR).13 reduces to Incompressible-fluid or constant-pressure reactor. Normally we express thIs quantity as an intensive heat capacity times the amount of material in the reactor.n j dP+L(:H) dn} j n J T..nk dt- 12 .15) ~V_T(av) aT P.n J dV + j (OP) anj T. and express its differential as in whIch tlHRI is the heat of reaction tlHRi ~ L vijH} } dH~(~~) dT+(~~) P.n.P.

Cv = Cp). (6.).T) (6. . and using the number of moles of A. p. Note that Equation 6_15 must be solved simultaneously. The material balance for component A in the batch reactor is d(cAVR) _ R V dt . Equation 6. Write the material and energy balances for these two reactors. Equation 6. nA = CA VR yields dnA dt = - k (T)nA (6.12. lfwe consider the constant-volume reactor with incompressible fluid (a ~ 0. ~ RT/V. Example 6. gas-phase batch reactors for the elementary and irreversible decomposition of A to B. = UOA(7'a .A R Substituting in the reaction-rate expression. in the constant-pressure case. which provides the information for how the temperature changes. ilHR.17) where Vi ~ Ij Vi). and that this balance is the same as the balance for an incompressible fluid in a constant-volume reactor. may be assumed constant over the composition and temperature range expected. Two cases where Equation 6. k(Tl = koexp(-E/Tl balance gives Constant-volume reactor.15 cau be Simplified include the case of adiabatic operation (0. and usually numerically. is the same for both reactors. r = k( TlCA. with the batch-reactor material balance. If we consider an ideal gas. the physical situations they describe are completely different.0 atm and k has the usual Arrhenius activation energy dependence on temperature. Although the same final balances are obtained for these two cases.18) The heat of reaction. 2B (6.1: Constant-pressure versus constant-volume reactors Consider the follOwing two well-mixed.19) Reactor 1: The reactor volume is held constant (reactor pressure therefore changes).16 reduces to Equation 6. A J<.16) Reactor 2: The reactor pressure is held constant (reactor volume therefore changes). Cp. We must solve it simultaneously with the energy balance. Liquid-phase reactions often are carried out in batch reactors. (6.20) Notice the temperature dependence of k(T) prevents us from solving this differential equation immediately like we did for the isothermal reactor. t 5 is therefore a common batch-reactor energy balance.V.280 The Energy Balance for Chemical Reactors 6. the same energy balance applies for any fluid mixture (ideal gas. Which reactor converts the reactant more quickly? Solution Material balance. Equation 6. The material balance.n. Substitution into the constant-volume energy Both reactors are charged with pure A at 1. We also notice that. Equation 6.20. adiabatic. Substitution of Cv and the material balance yields the energy balance for the constant-volume batch reactor Constant-volume reactor. it is straightforward to calculate aT ~ 1 and (op/onj)Y. etc.15 is valid for any reactor operation with an incompressible fluid. ~ OJ and the case of heat transfer through a jack"t or cooling coil 0. _.2 The Batch Reactor 281 We note that the first term in brackets is Cv ~ VRPCV [15. 43J in which UO is the overall heat-transfer coefficient. ideal gas. and heat capacity of the mixture. incompressible fluid.15 as it should because Equation 6.

!:.0 mol/L. VR = 1200 L Solution 1. The is deA = -kCACB dt - The stoichiometry of the reaction. What is the maximum rate at which heat must be removed by Comparing the production rates of A and B produces the cooling coil (kcal/min) and at what time does this maximum occur? Because there is no B in the reactor initially. Ceo = O. We consider first the constant-volume reactor. and this work results in a lower temperature and slower reaction rate compared to the constant-volume case.23) dt Cp Rate constant. we know from thermodynamics that the two total heat capacities are simply related. CPA = CPB = 20 cal/moj·K. The energy balances for the two reactors are not the same.HR is negative and the positive RT is subtracted.21 into the constant-volume case yields dT dt (/'. For the A 2B stoicbiometry. CAO = CBO = 2.15 dT CPTt = A + B. we substitute the rate expression and v = 1 into Equation 6. We also see the denominator is smaller because Cp is positive and the positive nR is subtracted.2 The Batch Reactor 283 Energy balance.. 0 Example 6. C.HRknA constant pressure (6.282 The Energy Balance for Chemical Reactors 6.01725 L/mol·min. subtracting nA from both sides yields for the total number of moles 4. and the material balance for B gives . The reaction proceeds more quickly in the constant-volume case.22 and 6. l!. Assuming constant density. k = 0. 'What is the total amount of heat (kca!) that must be removed by Cp -nR (6.22) and the temperature differential equation for the constant-pressure case is dT /'.nA)R constant volume (6.(2nAO . The energy balance for the constant-pressure case follows from Equation 6.23 that the numerator in the constant-volume case is larger because /'. at Z7"C Partial molar heat capacities. -- -llHRknA in which Cp = VRPCp is the total constant-pressure heat capacity. What is the adiabatic temperature rise for this reactor and what is its Significance? Additional data: Substitution of the above and Equation 6.HR .RT) knA Cp . How long does it take to reach 95% conversion? 2.17 to obtain constant -pressure reactor is expending work to increase the reactor size.2: Uqnid-phase batch reactor The exotbermic elementary liquid-phase reaction in which Cv = VRPCV is the total constant-volume heat capacity. Cv = is carried out in a batch reactor with a cooling coil to keep the reactor isothermal at 2rc.21) the cooling coil when this conversion is reached? 3.HR = -10 kcal/mol A. The reactor is initially charged with equal concentrations of A and B and no C. For an ideal gas. the material balance for component A We see by comparing Equations 6. at 27"C Heat of reaction. Therefore the time derivative of the temperature is larger for the constant-volume case. Cpc = 40 cal/mol K Reactor volume. 1.

it is convenient to evaluate the total heat capacity as Substituting CA = 0. If the beat removal is manipulated to maintain constant reactor temperature. giving t1Tmax = -10 x 103 cal/mol 2(2 mol/L)(20 cal/mol K) (0 .15 vanishes leaving (6.1 n. We assume the incompressible-fluid energy balance is accurate for this liquid-phase reactor.25 "1th constant heat capacity gives 3.c:J Because we are given the partial molar heat capacities (see Equation 3. Substituting r = kd into Equation 6. so we can express the righthand side of the previous equation as :z:: Cpjnj = :z:: CpjnjO + Et1Cp j. o . which occurs for CA = CA(I. Integrating both sides gives Integrating Equation 6.25) Separation of variables and integration gives t= i [C~.284 The Energy Balance for Chemical Reactors 6_2 The Batch Reactor 285 or CA gives = Cn.05cAo and the values for k and CAO gives t=55lmin 2.j n. The adiabatic temperature rise is calculated from the energy balance without the heat-transfer term The adiabatic temperature rise indicates the potential danger of a coolant system failure. the time derivative in Equation 6. Substitution into the material balance for species A dCA _ -k 2 dt cA Substituting the material balance dnA/dt = -rVR gives (6.2 mol/L) t1Tmax " 250 K 4. If we assume the partial molar heat capacities are independent of temperature and composition we have t1Cp = and ° VRPCp = :z:: CpjnjO j.51 for the definition of partial molar heat capacity).24) For a batch reactor. In this case the reactants contain enough internal energy to raise the reactor temperature by 250 K.". . the number of moles can be related to the reaction extent by nj = njO + VjE.24 yields The maximum temperature rise corresponds to complete conversion of the reactants and can be computed from the given data The right-hand side is a maximum in absolute value (note it is a negative quantity) when CA is a maximum. j Substituting dCA/dt = -r and multiplying through by dt gives in which t1Cp = Ij VjCpj .

Writing Equation 6. which is normally small for a liquid.QpH + Q + Ws + Wb (6. If the heat capacity of the liquid phase does not change significantly with composition or temperature. (6. there is a final approximation that is often useful.30 can be set to zero yielding.29 dt dU dU -.32 and 6. a variety of important special cases may be considered. for a single-pbase system we consider the change in enthalpy due to changes in temperature.3.33) j And if we neglect the shaft work dt dVR _. Its flowTate is denoted Q. A common case is the liquid-phase reaction.. possibly because of the presence of a large excess of a nonreacting solvent.34) d/ - Qfcif ..Qc.8 in the chapter summary. (6. (6. enthalpy Hf' and component j concentration cff. and we neglect the pressure effect on enthalpy. the time derivatives in Equations 6. Again.J+().30) . Substitution into Equation 6.L ri 6HRi VR + QfPjCp(Tf + T) + Q ~ 0 (6. Again we denote the feed stream with flov.i A=R k. pressure and the number of moles of component j.Hj ~ Cpj(Tr .T) Q (6. + Plit ~ QfPfHf .32) (6.31) in the next section we consider further simplifying assurriptions that require less thermodynamic data and yield useful approximations.3: Temperature control in a CSTR An aqueous solution of species A undergoes the following elementary reaction in a 2000 L CSTR as depicted in Figure 6.33 gives The material balance for the CSTR is dn . which usually is well approximated by the incompressible-fluid equation.28) Substitution into Equation 6.5 for this reactor gives. we obtain HJf .28 and rearrangement yields 6HRi Yi VII+ LCJfQr(Hjf-H. we again assume that the internal energy is the dominant contribution to the total energy and take tbe entire reactor contents as the volume element.3 The CSTR Dynamic Operation For the continuous-stirred-tank reactor (CSTR).QpH + Q or if we use the enthalpy rather than internal energy (6.Tate Qf' density Pf. k.3 The CSTR 287 6.26) and 6. Note that the heats of reaction 6HRi are evaluated at the reactor temperature and composition.33 provide ns + 1 algebraic equations that can be solved simultaneously to obtain the steady-state concentrations and temperature in the CSTR.29) Example 6. The outflow stream is flowing out of a well-mixed reactor and its intensive properties are therefore assumed the same as the reactor contents. dH ~ VRPCpdT + (l . These are listed in Table 6. 6H" ~ .286 The Energy Balance for Chemical Reactors 6. Steady-State Operation If the CSTR is at steady state. For a liquid-phase reaction.27 gives QpH + Equations 6. ~ QfPfHf . L VijYiVR i (6.27) As in the batch reactor.exT)VRdP + L Hjdnj j Substitution into Equation 6.18 kcal/mol .

2.34 to this reactor gives .-K (cAI . 1 Substitution into the material balance for A gives Q(cAI .cAl . is 4 mol/L and feed 1I0wrate.cA»)tlHRV" . because kl and Kl are given functions of temperature. Cp ~ 1 call g K.kdcA . What are the heat duties of the two heat exchangers if the feed enters at 25°C and the product is to be withdrawn at this temperature? The heat capacity of reed and product streams can be approximated by the heat capacity of water.2CAf to achieve 80% conversion. Solving this equation numerically gives T Figure 6.35) lfwe set CA = O.33 x 10 3 kcal/min o .35 for CA.9 x 10-lI e 9059fT or for XA ~ 1 .8(l + l/KIl Figure 6.cAl .QrPCp(Tf .3: CSTR and heat exchangers. The steady· state material balances for components A and R in a Q2 ~ QjPCptlT ~ = (250 L/min)(1000 g/L) (1 cal/g K)(298 -9 x lO3 kcal/min 334 K) constant-density CSTR are Q(CAj . T. It may be valuable to first make a plot of the conversion as a function of reactor temperature.4 displays XA versus T and we see that the conversion 80% is just reachable at 334 K. At what temperature must the reactor be operated to achieve 80% conversion? 1 + k. ~ ~ A word of caution is in order here.CA)) 1 . ~ 334 K The feed concentration. If we solve Equation 6. (CA .Kl (cAI . we hrr\7e one equa· tion and one unknmvn. Solution 1.288 The Energy 8alance for Chemical Reactors 6. there are two solutions. Because the reaction is reo versible.cAl)VR 1 =0 (6. is 250 L/min. CR) + rVR ~ 0 Adding these equations and noting cRf • 0 gives 5.-K (C4{ . 1 ~ QI = kj(cA . Qj.rVR Q(cRf ~ 0 Applying Equation 6.35. A simple calculation for the heat·removal rate required to bring the reactor outllow stream from 334 K to 298 K gives 2. we do not know if 80% conversion is achievable for any temperature when we attempt to solve Equation 6. we have CA 3 X 10 7 e-5838fT min-' = QiVR + kl/Kl cAf QiVR + kItl + llKIl Kl 1. CAf. and that for any conversion lower than this value.T) 1 .3 The CSTR 289 Substituting this result into the rale expression gives r ~ k.cAlcAj 1. The reaction·rate constants have been determined experi· mentally kJ.

T) I (6.4: Conversion of A versus reactor temperature. The parameters appearing in the problem are: CAj.37) d(CAVR) + RA VR in which Cps ~ PjCp.4 UO Value 298 298 4.~ QfcAJ . The material halance for component A is ~.1: Parameter values for multiple steady states. In this section we explore the fact that the steady state of the CSTR is not necessarily unique.l K kg/m 3 kJ/kmol 0.8 0. We introduce this topic with a simple example [24]. CPs. This phenomenon is known as steady-state multiplicity. kJ/kg K kmol/m3 min. The temperature appears in the rate-constant function.0 0. liquid-phase streams. With these assumptions.HR.1. the material halance simplifies to Figure 6. As many as three steady-state solutions to the material and energy balances may exist for even the simplest kinetic mechanisms. a heat capacity per volume. kme'ElllT -lITm) A second equation is provided by the energy balance. Consider an adiabatic. We wish to study this solution as a function of one of these . Tr. The production rate is given hy 260 280 300 320 T(K) 340 360 380 400 For the steady-state reactor with constant-density. and. 0. tl. in this example.36) The coupling of the material and energy balances for the CSTR can give rise to some surprisingly complex and interesting behavior. km.0 x 105 0 Units K K . Tm . constant-volume CSTR with the following elementary reaction taking place in the liquid phase A Equation 6.3.290 The Energy Balance for Chemical Reactors 6. The solution of Equations 6.3 The CSTR 291 Parameter Tf Tm 0. 6. B Dividing through by VR and noting U O gives ~ 0 for the adiabatic reactor We wish to compute the steady-state reactor conversion and temperature. E.36 is one nonlinear algebraic equation in two unknowns: CA and T. we assume the heat capacity of the mixture is constant and independent of composition and temperature.001 8. The data and parameters are listed in Table 6.QCA G ~ -kCAtlHR + f[f(Tj .1 Steady-State Multiplicity I0 ~ CAi k(T) ~ (1 + kO)CA I (6. Even the steady-state solution of the material and energy balances holds some surprises.6 Cp cAf km E P tlHR 0. the steady-state energy halance reduces to J'.2 Table 6.0 x 103 103 -3.37 for CA and T provide the steady-state CSTR solution.36 and 6.0 2.

HR ~ 0 lines In Figures 6.6: Steady-state temperature versus residence time for dif- ferent values of the heat of reaction (!!.. This case is much as we ex· pect.S: Steady-state conversion versus residence time for differ- ent values of the heat of reaction (6.. three solutions in this case. the conversion of A Increases monotonically with residence time because the reactant molecules have more time to react at longer resi' dence times.292 The Energy Balance for Chemical Reactors 6.8 0..HR ~ +5 X 104 kJ/kmol. there would be a small upset and the reactor would increase in conversion and temperature as it approached the new steady state at the new residence time. We also should note that we have assumed the reaction mixture remains in the liquid phase at these temperatures and the reactor pressure. plotted in Figures 6. the reactor residence time. Note that if the heat of reaction is more exothermic than -10 kJ/kmol. at low conversion of A and low temperature. one can see the forma· tion of the s·shaped steady·state temperature curves. Consider the situation at the ignition pOint.~1O.3 0.7 and 6.6 / / / I x 0. In these cases... Consider a reactor at steady state for a small value of residence time.5 and 6. For intermediate values of the heat of reaction. there is no 1 0. 6. however. Now consider the solution for exothermic reaction.2 .HR x 10 4 kJ/kmol).9 min at x ~ 0. there is a range of reSidence times in which there is not one but several steady·state solutions.~20.. Let the reactor he at steady state for () ~ 30. We first examine the iso· thermal case. -5 +5 0. -10 . In this case.5 and 6.4 0. Because the heat of reaction is positive.1 0 1 0--100000 10 100 1000 10000 e (min) Figure 6. Equation 6.. the steady·state temperature decreases with increasing conversion and residence time. tempera. \ \ -30 -- -20 --- -10 _.9 0.6. Consider the endothermic case.HR ~ -5. The reactor is said to exhibit steady·state multiplicity for these values of residence time.3 The CSTR 293 parameters. If the feed tlowrate were decreased slightly ((J increased).-20 ----. The rate con· stant increases significantly with temperature.09 and T ~ 311 K. we already have shown that ke x ~ 1 + ke The x and T values for this case are plotted versus residence time as the 6. - -5 o --- +5 - T(K) 360 340 320 300 280 260 1 10 Ion 1000 () (min) 10000 10()()O() Figure 6. The lower temperature decreases the rate constant and one has to operate the reactor at significantly longer residence times to achieve the same conversion as the isothermal case as showu in Figure 6.~30 X 104 kJ/kmol. The points at which the steady· state curves turn are known as ig· nition and extInction points. 460 440 420 400 380 -30 . \ \ \ \ . If there is a small decrease io feed flowrate so the residence lime is 32 min. . Figures 6. for t!. e.IlR x j () 4 kJ/kmol). J 0 min. Assume the heat of reaction is zero. without a log scale on the residence·time axis. ture increases 'Nith conversion and hence residence time.7 0.37 then implies T ~ Tf for all residence times. .5.6.HR ~ -3 X 10' kJ/kmol case again. and higher conversions are achieved at smaller reSidence times as the exothermicity increases.5 0.8 show the 6..

8: Steady-state temperature versus residence time for 6. A smalllncrease In feed flowrate causes the residence time to decrease enough so that no steady-state solution exists near the current temperature and concentration.2 Stability of the Steady State 20 25 e (min) 30 35 40 45 Figure 6.1 0 0 5 10 15 20 e (min) 25 30 35 40 45 Figure 6. We can compute this maximal temperature correspondlng to complete conversion as follows. The upper and lower branches. ignition and extinction pOints. The curves of low temperature and low conversion connecting to the ignition point are known as the lower branch of solutions. r .-cAO"'h .294 The Energy Balance for Chemical Reactors 63 The C5TR 295 1 0. The choice of which branch the reactor approaches depends on the exact nature of the perturbation the reactor experiences. ignition and extfnction points. are stable steady states. We next discuss why some steady states are stable and others are unstable.3 0.9 0. which has wide applicability in anaIyzlng the stability of any system described by differential equations.imax- TJ . The points on the middle branch. An unstable steady-state solution has the property that small perturbations to a reactor operatlng at this steady state causes the reactor to leave this solution and approach either the steady state on the upper or lower branch.3.HR == -3 X 10 5 kj/kmol. The reactor returns to these steady states after small perturbations.6 steady state near the temperature and concentration of the reactor.8 extinction polnt OJ 0.5 OA 0. A large release of heat occurs and the reactor ignites and moves to the steady state near x = 1 and T = 448 K. This situation is much the same as a pencil initially balanced on end on a desk. flHR = 460 440 420 400 T(K) extinction polnt 380 360 340 320 300 280 0 5 10 15 A reactor operatlng near the extinction point can exhibit the opposite phenomenon. Which direction the pencil falls is not easHy predicted and depends in a sensitive wayan the actual perturbation the pencil experiences. The curves connecting the ignition and extinction points together are called the middle branch. are unstable solutions. First we present a plausibility argument and develop some phYSical intuition by constructing and exatnining van Heerden diagrams 1251_ In the next section we present a rigorous mathematical argument.2 0.150 K --. . A rapid drop in temperature and increase in concentration of A occurs as the reactor approaches the new steady state. This discussion comes in two parts. The curves connecting the extinction polnt to the high temperature and conversion at large residence time are known as the upper branch. ignition point 6. while satisfylng the steady-state equations.7: Steady-state conversion versus residence time for -3 X 105 kJ/kmo!. on the other hand.Cps x 0.

another indicator that paint E is an extinction paint. Consider plotting these two functions as T varies.9 311 D 15 338 E 1. we call the heat-generation rate. and E.10_ To the right of A we notice that the heat-removal rate is larger tban the heat-generation rate. If we first substitute the solution of the mass balance.9: Steady-state temperature versus residence time for MIR ~ -3 X 105 kJ/kmoL Point e(min) T(K) A B 1. the heat-generation rate is larger than the heatremoval rate causing the reactor to heat up and move back to the right.9.79 min in Figure 6. consider a decrease to the reactor temperature. e .36. Equation 6. the system responds by resisting our applied perturbation. the slope of the heatremoval line increases and the intersection corresponding to point A shifts slightly.2: Selected values of steady-state temperatures and residence times.term is the enthalpy difference between the inflow and outflow streams.2 15 300. Equation 6.2. In other words. Band C on the lower branch. D on the middle branch.5 G 30.38) 360 340 320 300 280 260 0 5 10 A D The first term. If we were to increase tbe reactor temperature slightly. we wonld be to the right of point A in Figure 6.296 The Energy Balance for Chemical Reactors 6. which moves the temperature back to the left. Point A is a stable steady state because small perturbations are rejected by the system. That causes the reactor to cool. whJch we call the heat-removal rate. The second .9 447 Table 6.37. Using this argument we see the situation at point E is e. the solution at point E disappears. we obtain Obviously we have a steady-state solution when these two quantities are equal.Plausibility Argument Consider again the steady-state temperature versus the residence-time curve shown in Figure 6. however. Qg.79 421 F 15 446. We have labeled seven points on this curve: A. C in whJch we emphasize the temperature dependence of the rate constant. These two functions are plotted for = 1.3 The CSTR 297 460 440 420 400 380 T(K) a single equation for the single unknown T F G (6. Notice the two intersections of the heatgeneration and heat-removal functions corresponding to steady states A and E.5 C 30. To the left of point A. the reaction rate times the negative of the heat of reaction. If we decrease the residence time slightly.10. 15 20 25 30 35 40 k(T) ~ kme~E(l/T~lITm) e (min) Figure 6. F and G on the upper branch. We also can make a plausibility argument on the stability of steadystate A. Because the two curves are just tangent at point E. Reactor Stability . for c" into the energy balance. Qr. The heat-removal rate is simply a straight line with slope Cps / The heatgeneration rate is a nonlinear function that is asymptotically constant at low temperatures (k(T) much less than one) and high temperatures (k(T) much greater than one).79 298. The numerical values at these points are listed in Table 6. Similarly.

.. perturha- 25000 -~ M ~ 20000 . . Increasing the temperature causes the system to cool.c: '" ! / 5000 0 -5000 250 300 / / / C (ignition) removal generation 350 T(K) 400 4S0 SOO Figure 6.298 The Energy Balance for Chemical Reactors 6_3 The CSTR 299 500000 400000 ~ M 60000 SOOOO S S 300000 / E (extinction) ~ M 40000 / . ----. Tbat is characteristic of an unstable solution.12 displays the heat-generation and heat-removal rates for e 30. increasing temperature causes heat generation to he larger than heat removal. Figure 6. and decreasing temperature causes heat generation to be smaller than heat removal. e / 15 min. All points on the middle hranch are similar to point D. . the slope of the heat-generation rate is larger than the heat-removal rate_ For point D..c: ~ "- S OJ 30000 / .. Point F on the upper branch also is stable because the slope of the heat -generation rate is smaller than the heat-removal rate at point F.c: '" C) 100000 / / / / A . Consider next the points on the middle branch. ------F " :. but decreasing the temperature also causes the system to cool.-------- ~ 10000 C) 15000 / / / / / G / / ! / . Figure 6. The system tends to increase this initial perturbation and does not return to the steady state. Figure 6.11 displays the heat-generation and heat-removal rates for points E. Both of these perturbations are amplified by the system at point D.? ~ 200000 / / / / / / . D and F.9 min. J 2: Rates of heat generation and removal for 0 = 30. just as at point A. and this solution is unstahle./ / g '" . is similar to extinction point E. Point E on the lower branch is stable as was point A. / / / / D (unstable) 0 -100000 250 removal generation 0 500 -10000 250 B 300 350 T(K) 400 450 300 3S0 T(K) 400 450 500 Figure 6_10: Rates of heat generation and removal for e /1. / / / / removal generation 20000 / / / / 10000 . however.11: Rates of heat generation and removal for e ~ 15 min. the ignition paint.79 min..9 min. At point D. 30000 quite different from the stability perspective. Notice that point G on the upper branch is stable and pOint C.

CAs £2 ~ T .. we call the steady state unstable.43) For our case.) + ~: (T ~ 15) + higher-order terms Iz(CA. 1~) + ~{~ (CA - C. taking partial derivatives shows the Jacobian is -kCAs ~lE_ . T (6. the evolution of the perturbations. Consider again the dynamic model of the reactor dCA ai. (Ji' 1 (6. If so. which are valid when the perturbations are small.39) ~ -(Ta .) + OCA (CA ~ CAs) The stability discussion of the previous section accompanying the van Heerden diagrams is useful and provides insight. we can gain further insight without resorting to full numerical solution. If not. Obviously we can solve numerically the nonlinear differential equations to answer this question.. rather than approximately. Consider the Taylor series expansion of the noniinear functions II.T) + .4. We also develop analysis tools that tell us more than just whether a steady state is stable or unstable. . but in the other direction they are amplified.44) Cps 0 If Equation 6. h we can compactly summarize the linear deviation variable equations in the convenient notation dE dt ~]£ (6.43 governed exactly. particularly in simple situations.Rigorous Argument around the steady state idcA. we also gain insight on the dynamic response of the system to perturbations.T 6HR Cps e (6. Ts) Imagine the reactor is initially at this steady state and at t ~ 0 we perturb the temperature and concentration by small amounts. we would be essentially finished. TJ ~ ideM. Linear . but then we answer the question on a case-by-case basis. as satisfying CA. 1) ~ i2(CAs. We would like to know whether or not the system returns to the steady state after this initial conditiou perturbation. In this section we develop proper arguments that determine stability rigorously in complex situations where physical intuition may not be as readily forthcoming. Ts .42) o ~ 11 (CAs. fz with respect to Given some values of parameters. Substituting the Taylor series expansion for 11. we call the steady-state solution (asymptotically) stable. Reactor Stability . CAS.41. we define steady-state solutions to these equations. 1.41) dt "'-£2 oh ilCA aT If we define the Jacobian matrix of 11.300 The Energy Balance for Chemical Reactors 6.Ts ~ cAl . + oj.--kCA Cps UO If . h and neglecting the higher-order terms. produce the follOwing approximate linear equations for the deviation variables dEl dt ~ all DCA EI + aT .3 The CSTR 301 lions in one direction are rejected. from the steady state that measure how far the system is £1 = CA . £2. (1 ~ Ts) aT + higher-order terms £1.40) For convenience we summarize these two equations as the following noniinear differential equations We can differentiate these deviation variables to see that they also satisfy Equation 6. I~) o ~ 12 (CAS. afI E2 -~-El dE2 312 (6.CA _ kCA dt dT dt e (6. By linearizing the nonlinear differential equations. Define deviation variables.

1 0.2 -0. which may be real or complex. they come as complex conjugate pairs. Both eigenvalues are real for all values of residence time.2 -0. This connection forms the basis for all linear stability analysis.2 ~ ~ D The solution to the linear model.2 and shown in Figure 6.8 -1 280 A B C FG (cos(bt) + isin(bt)) and the period of these oscillatory terms is (6AS) 300 320 340 360 380 400 420 440 460 T(K) Figure 6.6 -0.43.4 -0.302 The Energy Balance for Chemical Reactors 6. Equation 6. The important feature for us to note is that the solution contains linear combinations of the terms eAt in which A are the eigenvalues of J.43 contains oscillatory terms due to the relation e1a. We also know that if the eigenvalues are complex. In fact.bilt ~ eat '" ~ 0 -0. this linear problem is only important to us because it can be connected to the solution of the ful! nonlinear problem. the ignition point.1 0.14: Eigenvalues of the Jacobian matrix versus reactor conversion. We next use this result to analyze the multiple steady states of the previous section. In our case the 2 x 2 J matrix has two eigenvalues.41. Equation 6.8 0. The interested reader should see the first chapter in Guckenbeimer and Holmes [l1J for an introduction into the mathematical details and important restrictions on this general statement.4 0.6 0.3 The C5TR 303 differential equations can be solved analytically. Notice that the steady state loses stability at point C.7 0. Although precise statements of this connection require more mathematics than we can assume. Stability is regained on the upper branch at 0. this general statement is valid.8 0.9 x Figure 6.8 -1 .6 0.4 0.4 -0.43.13: Eigenvalues of the Jacobian matrix versus reactor temperature. In Figure 6.2 0 -0. .6 0. is well described by the corresponding behavior of the linear model.4 0.3 0. For the models analyzed in this text.6 -0. as one of the eigenvalues changes sign from negative to positive. ~ ignition extinction ~ '" 0 0.5 0. Stability of the linear equations is determined by the real parts of the eigenvalues. the 10eal behavior of the nonlinear model.2 0. the solution to Equation 6. is asymptot· ically stable if and only if the eigenvalues of J have strictly negative real parts.8 0. If complex. tbe general result may be summed up for "nlce" systems by saying [11 J 1 For small initial perturbations from the steady state.13 we plot the eigenvalues of J versus the steady-state reactor temperature and label the paints A-G as indicated in Tahle 6. Finally. Equation 6.9.

If we replot these results using a log scaling to stretch out the x-axis. we would expect to see eigenvalues with negative real part and nonzero imaginary part because the system exbibits a decaying oscillation back to the steadystate position after a perturbation.':. and linearized the model at point A. Consider the same simple kinetic scheme as in the previous section. Units K K ! kJ/kg K I kmol/m3 min~l A B c Tf Tm Cp E Figure 6. in wbich the eigenvalues are plotted versus the conversion instead of temperature. point A is the unique. stable steady state.304 The Energy Balance for Chemical Reactors 6.7 Paramo Value 298 298 4.3 Sustained Oscillations. . extinction and hysteresis.15: Marble on a track in a gravitational field.004 1. Here we have three steady states.7 and 6.37. The steady states A and Care ob· viously stable and B is unstable. A. at a given operating condition. now marked point A.3.1.0 2. limit Cycles We next explore the fact that the dynamic behavior of the CSTR can be more complicated Iban multiple steady states with ignition. Now consider the track depicted in Figure 6. the ignition and extinction points are labeled.15. points A and C are stable. we obtain the results in Figures 6. If we corn· pute the solutions to the steady-state mass and energy balances. "1th its zero imaginary eigenvalues. Perturbations from point B to the right are attracted to steady-state C and perturbations to the left are attracted to steady-state A. the extinction point. The oscillation decays because of the friction between the marble and the track. figure 6. Notice that the activation energy in Table 6. Based on our physical experience with such systems we conclude immediately that the system has a single steady state.2 L Notice the steady-state solution curve has become a bit deformed and the simple s-shaped multiplicities in Figures 6.19.0 0. Equations 6. the three positions where the tangent curve to I I I . If we expressed Newton's laws of motion for tbis system. I I ! the track has zero slope.14 provides another view of the same phenomenon.8.16.16: Marble on a track with three steady states. 6. Students may find it helpful to draw an analogy between the chemical reactor with multiple steady states and simple mechanical systems that exbibit the same behavior. as the eigenvalne changes sign back from positive to negative.2 x 10 5 cycles.5 x 104 103 -2. B but "ith the following parameter values. 18 and 6. wbich is asymptotically stable. wbich remains stable.3: Parameter values for limit A C Figure 6. point C.16 so there is just a single point of zero slope. and point B is unstable.3 is significantly larger than in Table 6. Tbis analysis validates the plausibility argument and shows the entire middle branch is unstable. Consider a marble on a track in a gravitational field as depicted in Figure 6. position A. Small perturbations push the marble over to the only remaining steady state. with these new values of parameters.HR K kg/m3 kJ/kmol Table 6. Now consider the track depicted in Figure 6.. We have flattened the track between points A and B in Figure 6.17. In fact. The significant difference between the reactor and marble systems is that the marble decays to steady state in an oscillatory fasbion. c4f km(Tml P 6. Tbis situation is analogous to the chemical reactor with multiple steady states.36 and 6. we obtain the results displayed in Figures ().20 and 6. all steady states may be unstable and the reactor may exbibit sus· tained oscillations or limit cy' c1es. point E. Point A now corresponds to a reac' tor ignition point as shown in Figures 6. and the reactor.3 The CSTR 305 A Figure 6. returns to the steady state without overshoot or osclHation.17: Marble on a track with an ignition point (A) and a stable steady state (C).

01 0.3 The CSTR 301 1 1 0.9 0. Figure 6.19: Steady-state temperature versus residence time" figure 6.21: Steady-state temperature versus residence time scale.9 0.5 0. log .3 0.7 0.1 0 0.6 C B D E x 0.1 1 F A 10 100 1000 B (min) Figure 6.7 0.6 0.20: Steady-state conversion versus residence time log scale.2 0.18: Steady-state conversion versus residence time.4 0.8 0.306 The Energy Balance for Chemical Reactors 6. B (min) Figure 6.001 0.1 0 0 20 40 60 80 100 x 0.2 0.4 0.3 0.5 0.8 0.

970 I T(K) 305 396 339 336 327 306 E F 0. Now we take a walk along the steady·state solution curve and examine the dynamic behavior of the system.7 0. The middle branch between A and B is unstable and the upper branch after point B is stable again as the positive eigenvalue passes through zero and becomes negative.0330 400 300 200 ::' Table 6.8 0. From the beginning of the curve through points A and B.4 0. they have the same real part and opposite imaginary parts so the eigenvalue curves from D to F are mirror images.519 Re(A)(min I) 0 0 -0. and the eigenvalue of the Jacobian with largest real part.8 0. The two eigenvalues remain real and negative as we pass through point C and then merge into the same value at point D.20 and 6. .1 0.7 0. the pair of eigenvalues cross back into the right half of the complex plane and 0 0.159 0. 0. At point E.4: Steady state and eigenvalue with largest real part at se· lected points in Figures 6. Figure 6.6 0.22 versus the conversion.218 -0.24 shows the real and imaginary parts of the eigenvalues from points CE Notice the two eigenvalues approach each other along the real axis until they become identical at point D.893 0.5 _1L--L__~~L--L__~~__.308 The Energy Balance for Chemical Reactors 6. As in the s-shaped multiplicity of the previous section.3 71.962 0.3 The CSTR 309 Point A B C D O(min) ILl x 0. After point D the imaginary parts of the eigenvalues are no longer zero.~ ~ I A B 0 -0.2 0. But the real parts of the eigenval· ues remain negative from paint D to E and the steady states along this portion of the upper branch are stable. and we have a complex pair of eigenvalues.3 0.2 L Table 6.008 19.8 have taken on a mushroom shape with the new parameters. The two eigenvalues are complex conjugates.3 0.L__~~__~ o 0.2 20.5 0. the two eigenvalues are real.905 I 0.5 ~ M .4 0.4 summarizes the locations of these paints in terms of the residence times. As we continue along the upper branch.21.7 29.22: Eigenvalues of the Jacobian matrix versus reactor con· version in the region of steady-state multiplicity. the steadY'state conversions and temperatures. however.373 0 0 Im(A)(min~l ) 0 0 0 0 0.1 0.9 1 x Figure 6. j « ~ 100 ~ a -100 -200 A B and 6.20 and 6.5 0. we see that an eigenvalue becomes positive at point A. Figure 6. We have labeled paints A-F on the steady·state curves in Figures 6.23: Eigenva!ues of the Jacobian matrix versus reactor conversion in the region of steady-state multiplicity -~ magnified y-axis. Their values are plotted in Figure 6.125 0.2 0.2 I 0. some surprising things hap· pen.6 0. which is again an ignition point.9 1 x Figure 6.23 shows a magnified view.

A simulation at thJs residence time is shown in Figure 6.28. the amplitude of the emerging oscillation is small and thc period of the oscillation is well predicted by Equation 6.24. and the shape is more like a pure sine wave.2 x 0.' -_ _. \ \ - 340 T(K) 320 300 280 600 \ \ . Near point E in Figure 6.' -I -0. If we decrease the residence time to e ~ 30 min. We see that the solution does not approach the steady state but oscillates continuously. T(t) I \ \ / . that there is no lower-branch steady state for these values of residence time.45 (T ~ 2IT I b).. 0 .2 o 0.1 -0.3 The CSTR 311 03. Notice also that the amplitude of the oscillation is large. Several other initial conditions inside the limit cycle are ShO'>l1 also. Figure 6. Let us see what simulations can show us. the reactor ignites. they do not damp out at large times.8 -0.-----.2 0 100 200 400 300 time (min) 500 Figure 6.27 explores the effect of initial conditions.159 .25: Conversion and temperature vs. ignites again.26. more than 80 K in temperature and 50% in conversion. Figure 6. have positive real parts. All of these initial conditions wind onto the same final limit cycle. however. The trajectory starting in the upper left of the figure has the feed temperature and zero A concentration as its initial condition.19. the eigenvalue analysis indicates the upper branch is unstable. Notice the amplitude of these oscillations is much smaller.6 -0. then slowly decays.1-_ _. including starting the reactor at the unstable steady state.. So a pressing issue now is what the reactor does after it is perturbed from these unique and unstable steady states. We say that the limit cycle is a global attractor because all initial conditions wind onto thJs same solution. Notice from Figures 6_18 and 6. \ \ --. we are close to point E.----~ 1 400 380 0.24: Real and imaginary parts of the eigenvalues of the Jacobian matrix near pOints C-F.------r----~r_----. time for 8 35 min.26 is shown again..2 E 0. where the stability of the upper steady state changes.1 o -0.310 The Energy Balance for Chemical Reactors 6.3 '--_ _-'--_ _. It appears we have unique steady states and they are unstable. This kind of plot is known as a phase plot or phase portrait and is shown in Figure 6.26. These oscillations are sustained.4 shows the imaginary part of the eigenvalue at point E is 0.5 I I I I I I I I I I 360 I I I I I \ 1\ I ' I I I . and eventually winds onto the steady limit cycle shown in Figure 6.' -_ _--'-_ _. Time increases a5 we walk along the phase plot. Solving the dynamic mass and energy balances '''ith thJs value of residence time produces Figure 6._----_... A residence time of e ~ 35 min is between points E and F as shown in Table 6.25. We can obtain another nice view of this result if we plot the conversion versus the temperature rather than both of them versus time. Table 6. where the transition to instability occurs. -0.4. The trajectory starting with the feed temperature and concentration as in Figure 6. From pOint E to F.

\ / \ 0. x 0.9 0. Figure 6.L-_. 400 380 l~------~~==~~~~~ 0. '- . .. \ .---.27: Phase portrait of conversion versus temperature for sev' eral initial conditions. II ~ 35 min.3 .J 280 a 100 200 300 400 500 600 700 800 time (min) Figure 6. .8 0.. . -- j .8 0.! OL-~ 300 T(t) __ ~ __ L-~ __ ~ __ L-~ __ ~ __ L-~~ 290 :000 310 320 330 340 350 3GO 370 380 390 400 T(K) a L.". . . .7 .9 0. .' .1 OL--d__. . . .28: Conversion and temperature vs.7 I 0..29: Conversion and temperature vs. Figure 6. j ." . \ \ 380 x(!) 360 \ .---. .5 OA 0..---.5 0.3 min. . . .L._. - 340 T(K) 0.4 0.=:0 3S min...L__ L-~ __ ~ 290 300 310 320 330 340 350 360 370 380 390 400 T(K) Figure 6. .5 0. 360 340 T(K) I' j j l j' " 320 ." .3 x 0. time for 0 = 72.6 x 0.---. j ..3 The CSTR 313 I~------~======~~~~ 0. .L__ L-~ T(t) __. \ .26: Phase portrait of conversion versus temperature for 100 200 150 time (min) 250 280 300 feed initial condition. time for 1 e = 30 min. " .._L-_L-_L-_.6 x 0.-r<.312 The Energy Balance for Chemical Reactors 6. . .' ' ....L.--. . .-. .2 O.---..5 ... . . j .L__ L-~ -- j 320 300 __.2 0.--._-'-_.. . . 0." . j . .

The analogy is not perfect because a single limit cycle does not surround the unstable point C as in the chemical reactor shown in Figures 6. dashed line indicates frictionless track....! min. and a stable steady state. To see how limit cycles can remain after the steady state regains its stability. We have thtee steady states.8 x 0.. 73.. consider Figure 6. Notice. Imagine the track consists of a frictionless material to the right of point B..16.1 min showing stable and unstable limit cycles..31.. finally.' . We see that the unstable limit cycle is the boundary between the regions of attraction of the stable steady state and stable limit cycle. Although in violation of the second law. Consider the marble and track system depicted in Figure 6.6 0. But the analogy may prove helpful in demystifying these kinds of reactor behaviors. that does not mean that limit-cycle solutions are not possible in this part of the parameter space. The amplitude of the oscillation in the marble system depends on the initial perturbation and a COntinuous family of limit cycles surrounds the unstable steady-state C. As we pass point F. A B c Figure 6.31: Marble on a track with three steady states. indicated by a solid Circle. Note that all initial conditions outside of the stable limit cycle would converge to the stable limit cycle from the outSide.12 explores the regions of attraction of the steady state and the stable limit cycle in further detail. which stabilizes the unstable limit cycle. Consider what is the essential difference between these thermodynamic systems. We also have a quantitative measure of a perturbation capable of knocking the system from the steady state onto a periodiC solution. the steady state is not asymptotically stable..9 0. the amplitude of the oscillations is not small near point F. constructed for B = 73. Phase portrait of conversion versus temperature at (} = stable limit cycle. The figure depicts the stable steady state.5 0..29. point C is not asymptot- ically stable and is surrounded by limit cycles.. Perturbations from point C do not return to the steady state but continually oscillate..3 min.' .) The unstable limit cycle is in turn surrounded by a ]This curve was fon. surrounded by an unstable limit cycle.24 indicates the steady state is stable for B :... A simulation near point F is shown in Figure 6..' . in contrast to point E. the eigenvalues in Figure 6. .30. Exercise 6.. Indeed.24 cross back into the left-half plane and the steady state is again stable.4 ' . We may modify our simple mechanical system to illustrate somewhat analogous limit -cycle behavior.~~~ted by reversing time in the integration of the mass and energy balances.314 The Energy Balance for Chemical Reactors 6.' .27 and 6. all initial conditions inside the unstable limit cycle are attracted to the stable steady state..30.7 0.. Without friction in the vicinity of pOint C. we can either use our imagination or consider extremely low friction material. indicated by the dashed line. even though Figure 6. All initial conditions in the region between the unstable and stable limit cycles would converge to the stable limit cycle from the inside. The reader may also want to consider why we were compelled to violate the second law to achieve sustained oscillations in the simple mechanical system but the reactor can continually oscillate without such violation.159 min = 40 min which is in good agreement with the simulation shown in Figure 6. steadystate A is again stable and steady-state B is unstable as in Figure 6...30..' 325 315 320 300 305 310 T(K) figure 6. so the predicted period is T = ZIT 10. At this point we cheat thermodynamics a bit to achieve the desired behavior.3 The CSTR 315 1 0.: 72.28..

especially if confronted with apparently complex operating data.5 of Chapter 4. L'lV \ I / Q(z+L'lz) Cj(z / IR \ " z + L'lz j + L'lz) fI(z + L'lz) Figure 6. Q(z) (. The main results are summarized in Table 6. at (pUAcL'lz) = mHz . then we have ns + 3 unknowns. even under isothermal operation. which then appears in the material and energy balances_ The development of the semi-batch reactor energy balance follows directly from the CSTR energy balance derivation by setting Q = O. then an additional statement of reactor operation.5 for thls element and neglect kinetic and potential energies and shaft work..74 in the CSTR Table 6. hydrolysis of acetyl chloride [3]. and Q is the heat transferred through the wall. chemical reactors are a rich source of other types of behavior including deterministic chaos [18. Just because these behaviors are possible does not mean that they are normally observed in reactor operation for something as simple as A goes to B. Note in particular that Equations 6. we have n. if the reactor outflow term is nonzero as well.e.5 The Plug-Flow Reactor To derive an energy balance for the plug-flow reactor (PFR).316 The Energy Balance for Chemical Reactors 6. R is the tube outer radius. we obtain a· _.81-6. lead to complex types of behavior. The energy balance of this chapter provides an additional equation. +q . normally expressed using an overall heat-transfer coefficient Q = U02rrRL'lz(Ta .0. such as constant reactor volrnne or mass. equations from the material balances and one equation of state. we must specify the reactor pressure or some other system constraint that determines the reactor pressure. for a single-phase system. such as polymerization kinetics [19. If we write Equation 6.32.17. Bailey [4] summarizes some cases of experimentally observed oscillations in CSTRs including chlorination of methyl chloride 17].4 The Semi-Batch Reactor As in Section 4. As discussed in Chapter 4. It is good practice to enumerate the unknowns and applicable equations when setting up a model for the nonisothermal semi-batch reactor. Complex kinetics.141.23.5 The Plug-Flow Reactor 317 further Reading on CSTR Dynamics and Stability Consideration of the coupled mass and energy balances for the CSTR have led to possible behaviors that may seem surprisingly complex for even the simplest kinetic mecbanism. the reaction between methanol and hydrogen chloride [12]. 6..32: Plug-flow reactor volume element.9 at the end of thls chapter.I iJ at Ac iJz • . If we use the reactor volume as the single extensive variable. and the temperature and the pressure specify all intensive variables of the reactor.T) Dividing by AcL'lz and taking the limit L'lz . As in Chapter 4. gives ~-(pU) = -~~(QpH) a . Finally. The chemical engineer should be aware of the complex behavior that is possible 'hith simple nonlinear CSTR models. is required to determine Q.mi{z+"z + Q in whlch Ac is the cross-sectional area of the tube.83 in the semi-batch reactor Table 6. the molar concentrations of the components. In addition to the multiple steady states and sustained oscillations. I.72-6. an irreversible first-order reaction.9 are identical to the corresponding Equations 6. 22]. conSider the volume element in Figure 6. 6. and the reaction between sodium thiosulfate and hydrogen peroxide 18].8. without the nonlinearity introduced by the rate constant's exponential dependence on temperature.

) The feed consists of component A diluted in an inert N2 stream. the pressure drop is usually negligible.46) In unpacked rubes.4: PFR and interstage cooling Consider the reversible.I A=B kJ k. adiabatic.51) 'I Hjdnj j yields a Gibbs-Duhem equation for enthalpy ° Equation 6.5 The Plug-Flow Reactor 319 in which q ~ (2/R)UO(Ta - Tl and we express the mass f10wrate as m .Nsf ~ 0. Since the inert stream is present in such excess. gas-phase reaction ~ V pCpdT + (l . we have Ideal gas. we assume that the heat .Qp.47 with reactor volume to obtain (6. ex ~ 0.6 in Chapter 3). (6. We have H ~ Lj njHj because H j is a partial molar enthalpy (see also Exercise 3. V .4.48) . In the steady state. we have The second term in Equation 6. or neglect pressure drop.47) We can form the derivative of Equation 6.50 is the usual energy balance for PFRs in this chapter. .49) pH - 'IcjHj j The enthalpy term can be broken apart into two terms (6. (6.I. Substitution of this result and Equation 6.33. Its differential is therefore dH ~ Hjdnj + njdHj If the fluid is incompressible. and Qf ~ IO.HR. and Equation 6.49 reduces to 'I j Incompressible fluid. (Refer to the discussion in Section 1. The next chapter considers packed-bed reactors in which the pressure drop may be significant.50) We treat the Lj cjdHj term by deriving the Gibbs-Duhem eqnation for enthalpy as follows. .49 is more appropriate in that situation unless one can safely make the ideal-gas assumption. exT ~ 1.OO() ft 3 /hr at Pf ~ ISO psi a and Tf ~ 83WR. plug-flow reactors With an interstage cooler between them as shown in Figure 6.2 for more on interstage cooling.exTlVdP - 'I njdHj j We express an intensive version of this result by dividing by V and rearranging to obtain Gibbs-Duhem equation (6. Example 6. + q.exTlVdP + 'I f'>.10.Acz and partial molar enthalpy to express the enthalpy per volnme. N Af / Nif ~ O. and for an ideal gas. dH ~ V pCpdT + (1 . For both of these cases.318 The Energy Balance for Chemical Reactors 6. The reaction is carried out in two long.46 then gives (6. which is valid for singlephase systems.r.48 into Equation 6.46 is expressed using the material balance for the PFR in which we nse the volnme instead of length of the reactor. Equation 6.dT dP QpCp dV + Q dV ~ - Subtracting this expression from Equation 6.

L. What other assumptions did you make? 2.NA L. If the reactors are long. Using the mole balance. Make sure all variables in your equations are expressed in terms of T and NA. What is the conversion at the exit of the first reactor? 5. The heat of reaction is tlHR = -5850 BTU/lbmol and can be assumed constant.320 The Energy Balance for Chemical Reactors 6. (NAI .j~l Nj = NAf + Nlf and the volumetric flowrate is and the reaction rate is r = ~ (k'NA -k_I(NAf.CA . 3. The value of the equilibrium constant is K = k. Solve these two equations numerically and determine tbe temperature and conversion at the exit of the first reactor.NA reactors. express T at the exit of the first reactor in terms of the feed conditions and N A.NA) = 0 Dhiding by k-J and solving for N4 gives NAI NA=---1 +K. Assume that economics dictate that we must run this reaction to 70% conversion to make a profit How much heat must be removed in the interstage cooler to be able to achieve this conversion at the exit of the second reactor" What are the temperatures at the inlet and outlet of the second reactor? 6.33: Tubular reactors with interstage cooling. you can substitute the material bal· ance into the energy balance to obtain one equation for T. How would you calculate the actual conversion achieved for twO If we assume the nilxture behaves as an ideal gas at these condi· tions. Using the energy balance. K.NA )) RT NAf + Nlf which is in terms of T and NA. Solve this equation to determine the temperature at the exit of the first reactor. Alternatively. . express N A at the exit of the first reactor in terms of the feed conditions and the equilibrium constant. Notice we have two equations and two unknowns because K is a strong function of T. The adiabatic PFR energy balance for an ideal gas is given by (653) 2.5 The Plug-Flow Reactor 321 PFRs of speCified sizes (rather than "long" ones) with this value of Q? Solution L The steady-state molar flow of A is given by the PFR material bal· ance Figure 6. so we have the total molar flow is 2. 4. 1. C = P/RTor The material balance for inert gives dN} / dV = 0. for long reactors. we may assume that the nilxture is close to equilibrium at the exit.NA = NAf .52) capacity of the nilxture is equal to the heat capacity of nitrogen and is independent of temperature for the temperature range we expect. Write down the mole and energy balances that would apply in the = k.1 k-I = 1.k-ICB = (kINA . so we conclude = NAf + NSf .5 at the feed temperature. -' and the rate expression for the reverSible reaction is given by r (6. r = 0 or k.NBl/Q = The molar flow of B is given by dNBldV NB r.

-'3NAf 738 oR ~ 2 x 10 5 BTU/hr Figure 6. The mole fraction of o-xylene is less than 2%. reactant feed concentration. A sudden. Substituting r = -dNAldV into the energy balance and mUltiplying through by dV gives dT = t._4.HR Tl .NAjl mep 4. If we assume the heat of reaction and the heat capacity are weak functions of temperature and composition. an ideal gas mixture. for example.322 The Energv Balance for Chemical Reactors 6.53. x = is highly exothermic The reaction is carried out in PFR tube bundles with molten salt circulating as the heat transfer fluid [10].34: Temperatures and molar flows for tubular reactors with interstage cooling. Under these conditions. 6. If the reaction is highly exothermic. Reactor runaway is highly dangerous. The a-xylene is mixed with air before entering the PFR.5. a point along the reactor length at which the temperature profile achieves a maximum. 6.1 (1 + 0. Q = 200.1 T2 = 738'R [J Plug-Flow Reactor Hot Spot and Runaway in which Cx is the o-xylene concentration.J4N~ 874 'R n h ~4'-4LN_A_f 1 ~ Q 726 oR _ _ _ _0_. and a small increase in the inlet temperature or . Integrate Equations 6. The kinetic parameters are adapted from Van Welsenaere and Froment and given in Table 6.5 [26]. The following example.000 BTU/hr. T .34. The results are summarized in Figure 6. we can perform the integral yielding t. The operating pressure is atmospheric. illustrates the PFR hotspot and reactor nmaway.5 The Plug·Flow Reactor 323 N'-'ALf _ _ _ _ _ 830 'R ~ o_. gas-phase reactions in a PFR.56 1) = 0. The reaction rate is limited by maintaining a low concentration of hydrocarbon in the feed.52 and 6. Tl = 874'R. the temperature profile can be very sensitive to parameters. the heat release generally leads to the formation of a reactor hot spot. Solution Ifwe assume constant thermochemical properties. large increase in the reactor temperature due to a small change in feed conditions is known as reactor runaway. and express the mole and energy balances in terms of reactor length.Tlf = ---(NA . oxidation of o-xylene to phthalic anhydride.830 + 80.04~2e2944IT 0. Calculate the temperature and o-xylene composition profiles. For exothermic.H~ dN QpCp A The term Qp = m in the denominator is the mass flowrate. can lead to large changes in the temperature profile. the large excess of oxygen leads to a pseudofirst-order rate expression 5. which is constant and equal to the feed mass flowrate. Example 6.5: Oxidation of o-xylene to phthalic anhydride The gas-phase oxidation of o-xylene to phthalic anhydride 3. and operating conditions are normally chosen to keep reactors far from the runaway condition.

0002 0 0 0. the temperature spikes quickly to a large value and all of the o-xylene is converted by z = 0.5 0. Tighter control allows us to operate safely at higher T(K) 660 640 620 600 0 0.001 0.4 we obtain dN x = -A r dz C Figure 6. This safe value obviously also depends on how well we can control the composition and temperature in the feed stream.35 shows the molar flow of 0xylene versus reactor length for several values of the feed temperature. which is a classic example of reactor runaway.992 0.36.8 z (m) 1 1.35: Molar flow of Q·xylene versus reactor length for differ- ent feed temperatures.0016 0.6 0.2 625 kg/s 6311------------~ Table 6.36: Reactor temperature versus length for different feed temperatures.019 1.361 X 103 2.0 1. We see a hotspot in the reactor for each feed temperature.373 0.8 z (m) 1 1. .0006 0.0012 J Z 'iii' 0.6371 x 10.0014 0.6 625 625 1.2 OA 0. The corresponding temperature profile is shown in Figure 6.0125 0. Nx RT N y = 2rrRUO • QpCp 740 720 700 680 and the total molar flow is constant and equal to the feed molar flow because of the stoichiometry.0008 615 0.f.T) r in which = k. notice if we increase the feed temperature above ahout 631 K.HR Qp Value 1922.3 Units s 1 K K atm m m kJ/kg K kJ/m 2 sK K kJjkmol 0.2 IA Figure 6. To avoid this reactor runaway..3636 x 104 -1. Finally.2 ]. dT dz = -f3r + y(Ta . Notice the hotspot temperature increases and moves down the tube as we increase the feed temperature.5: PFR operating conditions and parameters for o·xylene ex· ample.6 0. OA 0. we must maintain the feed temperature below a safe value.0004 0. Figure 6.5 The Plug-Flow Reactor 325 Parameter km Ta Tm Pf tp 1 R UO Yxj EIR t:.324 The Energy Balance for Chemical Reactors 6.6 m.

PH. we have the possibility to lower the reactor operating cost by heat integration_ The essential idea is to use the heat released by the reaction to heat the feed stream. which increases the production rate. The reactor system is an annular tube. This reactor configuration is known as an autothermal plog-flow reactor. More accurate thermochemical properties and a more accurate equation of state do not affect the fundamental behavior predicted by the reactor model.. Ammonia synthesis is also interestingbecause of its large impact on the early development of the chemical engineering discipline. it was undoubtedly the synthesis of ammonia by BASF. catalysis.37 125]. another discipline was emerging. conSider the heat integration scheme depicted in Figure 6. As a simple example of this concept. that of chemical engineering .326 The Energy Balance for Chemical Reactors 6. during the reactor start up.:Annulus in which feed is further heated °oogy ~..55) in which PN. and K is the equilibrium constant for the reaction forming one mole of ammonia.6: Anunonia synthesis feed temperatures and feed a-xylene mole fractions. p. Tbe feed passes through the outer region and is heated through contact with the hot reactor wall. resistance of materials. at the beginning of the twentieth century.37 [IJ. high pressures. 00 0 9g_~?__ Feed Preheater Products Reactants Figure 6. PA are the partial pressures of nitrogen. and flows countercurrently to the feed Caiculate the steady-state conversion for the synthesis of ammonia using the autothermal process shown in Figure 6. equilibrium reactions. The heat released due to reaction in the inner region is used to heat the feed in the outer region./N [ p3!2 /I - FA PH ~~2 ] (6. 0 6. and design of large-scale apparatus. 101] While physicists and chemists were linking up to understand the structure of matter and giving hirth to physical chemistry.- 8°00 0000 0 a 00 Catalyst bed where reaction occurs 0000 T -t-m-B~g A so +_--.5. hydrogen. The feed then enters the inner reaction region. no external heat is required to preheat the feed. Of course. Example 6. we assume the thermochemical properties are constant and the gases form an idealgas mixture. particularly in the United States. respectively. A rate expression for the reaction (6. and ammonia. For illustration.2 The Autothermal Plug-Flow Reactor In many applications. successfully achieved in 1913 in Oppau. Although recycle of energy can offer greatly lower operating costs. . Quoting Aftalion 12. . the dynamics and control of these reactors may be complex.54) over an iron catalyst at 300 atm pressure is suggested by Temkin [21] r = LdRT K 2 (1. which is filled with the catalyst. external heat must he supplied to ignite the reactor. it is necessary to heat a feed stream to achieve a reactor iillet temperature having a high reaction rate. When the reactor is operating at steady state.. 00 ooo"-b °ooQ-o 0°°00 000 00 0 0°0 0 0 °SOOOB °0° 0 0'b080 00 stream. If the reaction also is exothermic.5 The Plug-Flow Reactor 327 Ta (0) = T(O) . We next examine an ammonia synthesis example to show that muitiple steady states are possible.37: Autothermal plug-flow reactor. which forged the linking of chemistry with physics and engineering as it required knowledge in areas of analysis. the heat released by the exothermic reaction is used to preheat the feed.

5 -2. 600 500 400 T(K) in which q is the heat transfer taking place between the reacting fluid and the cold feed q = ~UO(Ta R T) The material balances for the feed-heating section are simple because reaction does not take place without the catalyst.----. Notice the heat terms are of opposite signs in Equations 6. To.B.59) Figure 6. a ~ (6.38: Coolant temperature at reactor outlet versus temperature at reactor inlet. Without reaction.5 The Plug-Flow Reactor 329 The steady-state material balance for the ammonia is dNA dV NA(O) ~Pararneter p = RA = 2r = NAJ Qo Ac l Tar Value :lOO 0. (6. the volumetric flowTate is given by E.HRr + q Table 6.----. q (6. (l) versus Ta (0)..16 1 12 323 Units atm rn"/s m2 m K and the other molar flows are calculated from y~---- 2rrRU" QpCp 0.342 4250 -1.NAJ) NH = NHJ .6.2 x 104 7.57 using dVa = -dV.dTa QpCp dV ~ 300 400 500 700 600 Ta(O)(K) 800 900 1000 .794 x 1011 2 x 104 11m m 2 s K/mol cal/mol cal/mol K = NNJ . heat of reaction and mixture heat capacity assumed constant.56.----r----r---.58) (6.---.3/2(NA .---. the molar flow of all species are constant and equal to their feed values and the energy balance for tbe feed-heating section is A B c 300 200 100 QaPa C Pa dV = -q dTa .1/2(NA .56) 700r----.C). we obtain .1' QpCp LlG' LlW k. and switch the direction of integration in Equation 6.57 and 6.l/R The energy balance for the reactor is the usual ~ dT QpCp dV = -6.NAJ) NN f3 ~ LlHR .328 The Energy Balance for Chemical Reactors 6. If we assume the fluid properties do not change significantly over the temperature range of interest. .6: Parameter values for Example 6.lO -' If we assume an ideal gas in this temperature and pressure range. intersection with coolant feed temperature T/-lf indicates three steady-state solutions (A.57) Ta(O) = TaJ in which the subscript a represents the fluid in the feed-heating section.

which are based on those used by van Heerden [25]. Solution Figure 6. because Ta is specified at the exit of the reactor.39 and 6.40: Ammonia mole fraction versus reactor length. which we have from the fact that the heating fluid enters the reactor at z = 0. A simple solution strategy is to guess the reactor iniet temperature. 0. Notice three steady-state solutions are indicated in Figure 6. and then iterate on the guessed values until we match the specified boundary conditions.38 shows the results for the parameter values listed in Table 6. 0. unstable middle (B).39: Reactor and coolant temperature profiles versus reac- tor length. solve the model to the exit of tbe reactor.------.6 Summary and Notation jj/ 1000 900 T 800 700 T(K) C Finally we require a boundary condition for the reactor energy balance. unstable middle (B).15 0.40.60) 600 -.------. we solve the initial-value problem. and plot tbe resulting Ta (VR) as the solid line in Figure 6. and upper (Cl steady states. We guess the missing values required to produce an initial-value problem.1 ---0.38.6.-----.330 The Energy Balance for Chemical Reactors 6. It is important to operate at the upper steady state so that a reasonably 0 large production of ammonia is achieved.60 is a boundary-value problem.T) y" A T.2 C Equation 6.------. lower (A).05 A 2 B 4 6 z (m) 8 10 12 Figure 6.25 r-----. T(O) = Ta(O). We will see more about boundary-value problems and shooting methods when we treat diffusion in Chapter 7. The intersection of that line with the feed temperature TaJ = 323 K indicates a steady-state solution. 0.38 for these values of parameters.Ta 300L-----L-----J-____. Combining these balances and boundary conditions and converting to reactor length in place of volume gives the model dNA dz dT dz = Ta 2A r C NA(O) = NAJ I (6. rather than an initial-value problem.__ B Ta sao 400 -in which ~ ~ = -fJr = + y(Ta .. The profiles in the reactor for these three steady states are shown in Figures 6..-----. This strategy is known as a shooting method. and then compare the computed feed preheat temperature to the specified value TaJ.L____-L____ ____ a 2 4 6 8 10 12 z (m) Figure 6.T) T(O) = Ta(O) dT" dz (T . . We solve the initial-value problem.60. and upper eCl steady states. lower (Al. Equation 6. For given values of Ta(O).

VR. [6. Constant-volume reactor 2: 6.QpH + Q + - -. b.6 Summary and Notation jjj Neglect kinetic and potential energies dU Tt ~ Neglect kinetic and potential energies (6. . (6.62) (6..72) .T) + Q ~ 0 .70) (6.VR + . continuous-stirred-tank.F.HR' .66) Slngie phase . . .64) (6.orTVR + dt dt dnj ~ _.H j ) + Q .RTvtl I'.VR + Qf{lfCp(Tr .HR.71) Single phase (6.QpH + Q dt dt . and plug-flow reactors.6 Summary Tables 6. (C].+. Q+W. H j dt ~ QjPfHf .dT VRPCP-.69) Neglect shaft work (6.7-6. semi-batch.68) + 2:c]fQf(H]f'. Incompressihle-fluid or constant-pressure reactor (6.7: Energy balances for the batch reactor. .10 summarize the important energy balances for the batch.8: Energy balances for the CSTR.dT dP VRpCPTt . ideal gas . 2: ] ) + Q (6.HR . which is reasonably straightforward. Constant-volume reactor h.. .VR n j J (6. 6. . ideal gas -dT VRPCVTt ~ - 2: (6. constant Cp.I Constant-volume reactor."TV.FNR + 2: c]fQf(Hjj . + Wb (6.VR + 2: CjjQf(Hjf .V.73) a. Q . choosing the proper energy balance requires some care.dT VRPCV-dt ~- .HJl.dT dP VRPCp. ] (6. J (6.orVR 2: v'] (.11k ] r.2:6. dt ~ . It is unwise to select an energy balance from a book without carefully considering the assumptions that have been made in the derivation of that particular c. + P .75) J Table 6. .. Steady state.+ Wb dt Negiect shaft work dU ~ QfpfHf .74) b.~ QjpfHj-QpH +Q dt dt dH dP _.H J ) + Q (6.76) 6.63) dU dVR -.HR' . P = Pj .. Incompressible-fluid or constant-pressure reactor . In contrast to the material balance.~ QjPfHf .HRiF.65) a.61) . cjjQj(Hif -Hj 2: .U ~ - h.QpH + Q ] (6. .l Constant-volume reactor.P) T.J Table 6.332 The Energy Balance for Chemical Reactors 6. W.

". and steady. Wb Neglect kinetic and potential energies and shaft work (6.flow reactors. these balances arc sets of initial.batch. The nonlinear nature of the energy and material balances can lead to multiple steady'siate solutions.90) ll. Table 6. a.HR. lncompressible~fluid or constant¥pressure reactor • dT VRPCp- dt ~ .9: Energy balances for the semi. and the reaclor can exhibit sustained oscillations.".HRiYiVR + cjfQf(Hjf . Nonisothermal reactor design reqnires the simultaneous solution of the appropriate energy balance and the species malerial balances. (6. and some books liSI energy balances that are simply incor' reel. ' di ~ QjpjHf + Q + W.82) j 2: llHRi r .86) dH dt dP dt . or ideal gas a.87) Single phase[tJ . In very limited situations (constant thermodynamic properties.85) Neglect shaft work R dt +P T t ~ QfPjHf + Q --VR-~QfPfHf+Q Heat transfer with an overall heat-transfer coefficient dU dV " (6. iJz (QpH) + q.. Adding to the confusion. dT dP . b. single reaction... For the batch.10: Energy balances for the plug..334 The Energv Balance for Chemical Reactors 6.83) . dT dP dn j " VRPCPdi .1 Constant C p .6 Summary ancl Notation Neglect kinetic and potential energies dU '-'.A.84) . These reactor behaViors were illustrated with exothermiC CSTRs and autothermal tubuiar reactors. HjTt ~ QfPfHf + Q J 2: (6. many books do not state clearly what assumptions have been made in deriving various energy balances. See Denn [9J for a thorough discussion of common errors in energy balances and a list of books containing incorrect energy balances.batch reactor. ~ ..TiVR + QfPfCp(Tf .H) + Q (6.value ODEs that must be solved numerically.state solutions may be unstable. + q_ . Steady state dV (QpH) ~ q d ' . (6.2: ll.T) + i Q (6.HRinVR + 2: CjfQf(Hjj i j ' Hj) + Q (6.aT) dV ~ - . Table 6. incompressible fluid a. semi. Neglect pressure drop.89) (6.80) (6. Steady. adiabatic)._..78) i 1 iJ ' otJ .TVRdt dt ~- .state plug.88) . (pU) (6.. dT dP VRPCp. dT VRPCp dt ~- (6. _.79) (6.81) Single phase QpCp dV + Q(l. h. energy balance. one can solve the energy balance to get an algebraic relation between temperature and concentration or molar flowrate. .f1ow reactor.TVR dt + .

Li Yi gas constant production rate for jth species time temperature temperature of heat transfer medium mean temperature at which k is evaluated overall heat transfer coefficient internal energy per mass velocity of stream k reactor volume variable partial molar volume of species j specific molar volume of species j reactor volume change in volume upon reaction i.n) extent of reaction i reactor residence time.jvij P Pk .ummary ana lVotaClon :J:J/ Notation A Ac Ai Cj cjf Cjs C jO Cp Cp j CPs Cp Cv !lCp Ei Ek Ek !lG- Hj H .336 Tile tnergy Balance Tor Lhem/wl Keacrors 6.Tj reaction rate for ith reaction total reaction rate. km K. 2. VRCj number of reactions in the reaction network number of species in the reaction network molar flow of species j. if = kUO(Ta .6.T) volumetric flowrate feed volumetric flowrate Ytot R Rj t T Ta Tm o Vk UO V Vj Vi VR Xj Xj Yj z IX E.j vij Vj rate work is done on the system number of molecules of species j in a stochastic simulation molar conversion of species j mole fraction of gas-phase species j reactor length variable coefficient of expansion of the mixture.b ::. (J = Vn I Qf stoichiometric coefficient for species j in reaction i -' L.j vijHj enthalpy change on reaction at standard conditions reaction rate constant for reaction i reaction rate constant evaluated at mean temperature Tm equilibrium constant for reaction i kinetic energy per unit mass tubular reactor length total mass flow of stream k reaction order moles of species j. IX ~ (lIV) (aV IOTjp.HRi !lW k.. usually modeled as Q ~ un A(7~ . !lHR . Qc] feed molar flow of species j. !lCp ~ 2. ~ 2.j v)CPj activation energy for rate constant i total energy of stream k total energy per mass of stream k Gibbs energy change on reaction at standard conditions partial molar enthalpy enthalpy per unit mass enthalpy change on reaction. mass density mass density of stream k potential energy per mass . B heat transfer rate to reactor. K ns Ni NJf P P'j q Q Qf heat transfer area reactor tube cross-sectional area preexponential factor for rate constant i concentration of species j feed concentration of species j steady-state concentration of species j initial concentration of species j constant-pressure heat capacity partial molar heat capacity heat capacity per volume constant-pressure heat capacity per mass constant-volume heat capacity per mass heat capacity change on reaction.j. Qc) pressure partial pressure of component j heat transfer rate per volume for tubular reactor.

2.92) points so we can see a smooth curve.1: Batch energy balance with heat transfer Consider Example 6.m) ) (6.3 X ]04 ~ -41 .~ (~ exp .2: Batch energy balance with constrained heat-transfer Consider Example 6. 7 k caIImm 551 which is much less in absolute value than the required -828K'c~l/min caleulated in Example 6. The rate expressions are Exercise 6. What value of heat-transfer coefficient (U°AjVR) should your design achieve so that this temperature is not exceeded.01725 L/mol·min Tm ~ 300K EalR ~ 2660 K ~ k m ( .7 Exercises so the temperature dependence of the rate constant is given by k Exercise 6. .338 The Energy Balance tor Lhemlcat Keactors 6. Put in enough Exercise 6.1 txerCises 339 6.01725 L/mol·min ~ n = kIPc]H(jPC12 Y'2 = k2PCjHijPCl2 The thermodynamic data for this reaction are listed in Table 6.0 atm') and are -27200 kJ ~ 206000 x exp~k2~11.2.7xexp and the temperature of the heat transfer Iluid is Ta ~ 27"C. Q ~ . (d) Plot CA and T versus time for the nonadiabatic case with heat exchange: UO AjVR ~ 0. (e) Assume the batch is ruined if the temperature exceeds 350 K during the run.3: CSJR energy balance with multiple reactions Allyl chloride is to be produced in a 0. Assume the control system can only remove heat at this so the temperature dependence of the rate constant is given by constant rate.83 ft3 CSTR [20] (6. The partial molar heat capacities can be calculated using cal/mol K 300 K Ea1R~2660K . and 39Z"F.2 with the following additional data: km Tm 0.2 "'ith the foUaffing additional data: k m ~ 0. The reactor pressure may be assumed constant.2. The rate constants have units of Ibmol/(hr ft3.11116J.85 lbmol/hr. (a) What is the maximum temperature achieved in this reactor and when does this maximum occur? (a) What is the adiabatic temperatnre rise for the reactor? (b) How long does it take to reach 95% conversion if the reactor operates (b) How long does it take for the control system to return the temperature to Z7"C? (c) When does the reactor achieve 95% conversion? isothermally at Z7"C? (c) How long does it take to reach 95% conversion if the reactor operates adiabatically? Plot CA and T versus time for this case. How long does it take to reach 95% conversion with your design? How should yOll operate the reactor if you want to speed things up but cannot violate the 350 K limit? -6860 RT~ where T is in degrees Rankine and R is in Btu/ObmolOR).93) The feed is a 4:1 molar ratio of propylene to chiorine and it enters at a feed rate of 0.0 atm of pressure.91) The heat-transfer equipment accidentally has been under-deSigned hy using tbe average rather than the maximum required rate of heat transfer .01 kcal/(min L K) (6.

E. liquid·phase reaction A .l m] S--l. the rate at which B forms is greater than the rate at whlch ( forms. .219 18. Tin K has reaction-rate expression Find the steady+state temperatures and conversions.] 3.3a with the value at the inlet temperature. C3HsC] and C3H6C/z.6: Finding the CSTR steady state The first-order reaction A.442 0.277 ~O.95) (6. Determine the effluent temperature and concentration of A for a reactor operating at the conditions listed in the following table. How many steady states can you find? Which steady states are stable and which are unstable? in which the units for the rate are kmol/m 3 5. 500/T) 2.496 8. 2~C3H6C1z 0.771 ~ 5.94) (c) Explain why the answer to Part A or Part B is greater.Hr l t:. Compare the heat of reaction in Part 6. (6.15 ~22.42 ~O.48 x 106 exp(~7550/T) kmol/m" mJ/s J/kmol J/kgK kg/m.4: CSTR stability An adiabatic CSTR with a first·order.19xlO. and comment on the differences in the molar f10wrates and the temperatures in Parts 6.Hr2 V Qf CpI kl k2 1000 100 0.9241 D (x109) Exercise 6. CA is in kmol/m 3 .49 Table 6.06 ~39.266 17.16 X 10 14 exp( ~ 12.6.7 exerCises Component H29S K J A B C (x 105) ~O. I.96) 5. are each first order in the concentration of A.604 7.2976 72 2. determine the rate at which heat is transferred to the surroundings to achieve a selectivity of five.235 -2.3b.60 (xl0z) 6..S: CSTR energy balance The liquid-phase reactions A. The feed to a nonisothermal (STR contains pure A at 45 o( and 5 mol/L Additional informiltion for the reactions and the reactor are provided below.602 7.866 0.11.B kcal/mol Clz C3H6 C3 H5CI HCI 1.00 4.OOO Mt .729 ~6.6 (:p p lOB 4.e.3a and 6.432 0. (b) Now consider an approximate solution along the lines presented by Smith 120J in which tbe heat of reaction is computed at the inlet temperature and assumed constant.HR Value Units Exercise 6. operates at the conditions shown below r:'" If cAf QJ /).( (6.l 10 3 18x 10 3 VR k 4. Thermodynamic data for allyl chloride example.931 A. and the reactor temperature for adiabatic operation.88 ~O.200 ~12. achieve a selectivity of four. 5001T) Units g/min cal/mol cal/mol liter liter/min cal/g K min-l.22 3.09x K 10.8082 5. Parameter Value 93. You need to use an algebraic equation solver for this problem. T in K (a) Defining the selectivity as S ~ RB / Rc.34U The Energy Balance tor Chemica! Reactors 6.52 X 109 exp( ~8. a reasonable first guess on the temperature is 1200'R.B 298 3 60 x ~2. T in K min-I.695 0. (a) Compute the molar f1mnates of Clz. We want to operate this reactor such that the selectivity to B is greater than the selectivity to C. (h) Determine the rate at which heat is transferred to the surroundings to Exercise 6.OOO ~15. and T is in K.

225 -8. Check that the dynamic simulation converges to a steady state consis~ tent with Figures 6. (a) Write down the dynamic CSTR equations for the adiabatic reactoL How many parameters are there and what are their values? Choose consistent units.500 -3700 1000 932 100 0. T versus residence time plots in Figures 6.6 A-B r kCA Exercise 6.H R VI< (b) Choose a residence time of 0 ~ 10 min.7 and 6.16 X 10 14 1. Choose as initial conditions for CA and T. (a) Determine if a CSTR operating at 347. how might you modify the reactor design to stabilize the steady state? Sketch your control scheme.22 3. Start the reactor from the feed conditions and simulate for 1000 min. According to Figures 6.09 x 108 4. Does your simulation point out any safety issues? Can you think of a less violent metbod to start up the reactor? What values of (CA.9: Removing heat in a CSTR The liquid-phase reaction kglm' m3 m'ls VR Qf 18 x 10-3 60 x 10.8 (th!s is your program debugging check).HR ko P krnol/m3 J/cm2 sec K J/kmol l/s kg/m3 J/kgK m' Cp V R QJ Next simulate the dynamic reactor. Does anything unusual happen? Describe what the reactor is doing. T) steady-state values and what are the unstable ones'? You can just read these off the plots.7: Testing a CSTR operating condition for stability Consider the reaction of Exercise 6. How many steady-state solutions are there for this residence time. PJ Qf ! I U" Cp ko A Value 12.7 and 6. t----t Parameter tJ.342 The Energy Balance for Chemical Reactors 6.19x]OJ 18 X 10..48 x 106 4. (d) Change the residence time to e ~ 32 min.3 60 x 10. x.7 Exercises 545 Parameter E Ah Value 7550 3600 Units K cm' K K To Tr CAJ 312 298 3 0. Once you are sure the program is working correctly simulate the reactor starting from CA ~ cAf.6 Ua tJ.T Tf' Increase oniy the initial temperature and simulate again.7 and 6. (c) If it were advantageous to operate at an unstable steady state.0 298 3 -2. Hand in plots of CA(t) and Tit) starting one degree helow and above this critical temperature. nonadiabatic CSIR operating at 95 "C The feed contains pure A at 45 "C and 5 mol/L Using the additional information provided below determine the temperature of the cooling fluid medium. Be sure to show and label dearly your sensor and actuator (if you are interested you can simulate the controlled reactor and see what controller gain is required to achieve stability). How warm can the reactor be started before it ignites and reaches the upper branch steady state instead of th~wer branch. values near the lower-branch and upper-branch steady states. Parameter E Tf cAf tiHR ko Value 75.09 x lOs 4.48 x 10 6 ]OJ 4.6 K is stable or unstable for the following conditions.8.8.19 x 10 J ]03 Units K K Cp P krnol/m3 J/kmol l/s J/kg-K Exercise 6.8: CSTR ignJlion Consider the CSTR parameter values listed in Table 6. T) are you trying to avoid during start up? m 3 /s Exercise 6. how many steady~state solutions are there at this residence time? What are the stable (CA.6. SimoJate the reactor for a time long enough for it to reach steady state.1 and tbe steady-state is to be carried out in a nonisothermal.8 X 104 Units K cal/mol L giL L/min cal! min em:! K cal/g K 1 em E min· 2 I .54 0.

batch reactor Equation 6. in a liquid'phase. (b) Draw a separate sketch of conversion of A and reactor temperature ver~ sus residence time for an exothermic reaction in an adiabatic CSTR that exhibits steady. are known at the iniet to the PFR. Pressure drop in the reactor can be neglected. Be sure that all variables in both the mole and energy balances can be evaluated in terms of your dependent variables.72 in Table 6. What is special about these points? What does hysteresis mean? What are the practical operational problems associated "'th trying to operate a (STR near one of these points.19 3.731 psi·ft 3 /lbmo)'R (e) If the equilibrium conversion of component A determined in 6.dT 2 0 QpCp dV ~ -6. at any point in the reactor? (b) Write down the mole balance for component A. concentrations. calculate the outlet temperature using . Also write down an expression to evaluate the volumetric flowrate.HR T + RU (Ta T) (d) Compute the equilibrium concentration and conversion of component A that could be achieved if the reactor operated isothermally at the iniet temperature given below. Gniy A and Bare present in the reactor feed stream. Is it possible to operate near one of these points? If not. gas~phase isomerization reaction shawn below: B(g) k. TI.09 X 10 5 103 6. I.state multiplicity. what is the sign of the heat duty? Exercise 6. and the molar flux of A. HE. perature and composition. Remove any reference (a) At what temperature must the reactor be operated to achieve 80% con* version? (b) What is the heat duty on the heating/cooling system when operating at this steady state and are you heating or cooling the reactor. and pressure. (el Describe an ignition and extinction point and locate them on the sketch of the (SIR with steady. Exercise 6.l kmol/m 3 A(g) m3 s-j kJ/kmol kg/m 3 m'/s This isomerization is to be carried out in an adiabatic PFR.14: Adiabatic plug-flow reactor Consider the elementary. Assume ideal gas behavior. The following data and parameters are known. Qj. (a) What does the stoichiometry tell you about the relationship between the molar llux of B.14a. determine the appropriate energy balance equation for the PFR in this problem. K ~ kl / L j ~ 1.e.1 Exercises Exercise 6.11: CSTR stability (a) For the reaction of A to products.5 Units K K kj/kg K Exercise 6. how do you overcome the operational problems described above? to CB in this equation with the result from 6.48 x 10" -2. constant.0 x 10. temperature. Tf ~ 830'R. HA.state multiplicity. The heat capacity and density of the mixture can be assumed independent of tern. Now derive the semi.12: Atttactors in the CSTR Show which initial conditions are attracted to which solutions in Fignre 630. The feed rate.volume (STR with heating/cooling system.8 by mak. B (6. Pf ~ 180 psia. PI. QJ because it is not con+ stant (c) Starting from the general energy balance equation for a PFR given below.14d is achieved in the adiabatic reactor. Why are these two energy balances identical even though they apply to different reactor ty])es? _ .018 4. Parameter E Cps cAf Tf V R ko 6. R 10. why not? If so.97) Exercise 6. ing the assnmptions listed in the table.HR P Qf Value 7554 298 4.9. C AJ and cBf. The molar ratio of A to B in the feed is 4:1.344 The Energy Balance for Chemical Reactors 6.81 in Table 6. T and H A .10: CSTR and energy balance Consider the reaction A -".5.0 0.13: Semi-batch reactor balance Derive the CSTR energy balance given by Equation 6. sketch the conversion of A versus mean reactor residence time for an isothermal CSTR..

What do you notice when you compare the extent of product formation and reactor temperature for both cases? Why do you think this happens? (c) Now compare the exact solutions developed above to the approximate solution presented in Smith 120J where tbe component heat capacities and the heat of reaction are evaluated at the inlet temperature and as~ sumed constant. Plot these dependent variables versus reactor length. Tbe feed is a 4:1 molar ratio of A and B and enters at 500 'F. (b) Compute the molar flowrates of C12.700/TJ The nnits for rare lbmol/hr ft 3 . 000 cal/mol) and the compo· nent heat capacities (CpA ~ CPB ~ 50 cal/mol K) are constant.H6C!" and the reactor temperature as a function of reactor length for adiabatic operation. for the exact and and approximate solutions on different graphs. Plot these dependent variables versus reactor length. The heat of reaction is tJ.17: PFR energy balance with multiple reactions Allyl cbloride is to be produced in a 25-ft long 2-in ID tube operating as a PFR 1201.11 1161.85 lbrnol/hr and 2 atm total pressure.99) Exercise 6.1des 93% conversion of B.:. A constant wall temperature of 392 OF can be reallzed by balling ethylene glycol on the outer surface of the reactor wall. The reaction~rate -6860 expression is in which T is in OR and R is in Btu/(lbmoloR).85 lbmol/hr. adiabatic and constant wall temperature.Hr ~ . (a) Compute the molar flowrates of Cr" C3H 5 c! and C. Assumiug constant thermodynamic properties.06 x 105 exp (-13. gas~phase reaction is to be conducted in a PFR A~ B r =k (CA ~ ~) You may assume the heat of reaction (tJ. determine the length of a 2'in inner-diameter (ID) tubniar reactor that operates adiabatically and pro· .7xexp~ k2 = is carried out in a PFR.H6Cl. P is in atm. Cp ~ 20.14d. .000 ft3/hr (6. Is it sensible that the error is greater for adiabatic operation? Why? Exercise 6. tJ. k2PC3H6PC12 k ~ 2.98) (6.20.7 Exercises 34/ the inlet conditions specified in 6. and T is in -R. and 392 'f . Plot the molar flowrates of C3HsCI and C.85 lbmol/hr and 2 atm pressure. at a feed rate of 0.16: Temperature and the equilibrium limit The following exothermic.. The thermodynamic data for this reaction are listed in Table 6. The equilibrium constant at 298 Kis K lOs Determine the maximum possible conversion of A if the feed contains pure A at a temperature of 300 K. reversible. one for adiabatic and one for nonadiabatic operation.Hy -48. Assume the heat capacities of A and B are the same and the heat capacity and heat of reaction are constant at the values helow. 000 Btu/lbmol and the mixture heat capacity is C p ~ 110 Btu/lbmol'R. temperature as a function of reactor length for constant wall temperature operation. COI1Ullent on the errors this approximation introduces. The inside heat-transfer coefficient is 5 Btu/hr ft' of for a feed rate of 0. The rate expressions are in which Y2 .The partial molar heat capacities are expressed in units of cal/mol K and can be calculated using Cpj ~ Aj + Bj T + C j T2 + Dj T3 Two modes of operation have been considered for tbis PFR. The reactor pressure may be assumed constant. C"HsC! and C3H6Clz.346 The Energy Balance for Chemical Reactors 6.15: PFR energy balance with a single reaction Tbe gas-pbase reaction The rate constants have units of lbrnol/(hr ft' atm 2 ) and are kJ ~ 2. and the reactor Exercise 6.0 B11J/lbmolo R . The feed is a 4:1 molar ratio of propylene to chlorine and it enters at a feed rate of 0.HR ~ 5850 BTU/lbmol . Qf ~ 10. 2 atm of pressure. Prepare similar plots for the temperature.06 X lOS exp (-27200/ (RT)) 11.

100) iJ(vP. Con· tinue to decrease y until you produce Figure 6."ish to convert the CO in a waste stream to C02 by reaction in an adiabatic bundle of empty tubes.z + The rate constant has units of 1bmol/(hr ft3 atm 2) and is given by kJ 2. The heat of reaction is "fiR ~ co + 40. y. Use the parameter values given in Table 6. and plot the resulting Ta(l) as in Figure 6. determine the minimum inert flowrate (in Ibmoljhr) that must be added to the feed to keep the reactor temperature below 880 R when the conversion of chiorine is 95%. -"H Nz cal/mol 'C 6. We wish to find the value of y such that tbe middle and upper steady states coalesce.value problem.60.348 The Energv Balance for Chemical Reactors 6. Consider an adiabatic PFR Q ~ 0.5 we derived the energy balance for the PFR by considering a fixed volume element.21: Moving volume element in PFR Exercise 6. Ij fbi Decrease the heat·transfer rate by 20% and repeat this calculation. = co.38.9 8. What is the critical value of y that corresponds to Figure 6.20: The effect of heat transfer in an autothennal reactor Consider wbat happens in the anunonia synthesis system described in Exam· pie 6. The feed is a 4: I molar ratio of propylene to chlorine and it enters at a feed rate of 0. Component CO CO2 02 Cpj Btu/lbmol'R If you assume "H R and Cp are constant and do not change as an inert is added to the feed. Dividing by iJ. The mean heat capacities and heats of formation at 25 'C are as follows.42 94.fI) ~ () DZ ilt (tUOl) . constant~pressure PFR in Section 6. You want to calculate the maximum temperature that might be reached in the tube to select appropriate material of construction for the tube bundle.85 Ibrnol/hr. (a) What is the maximum outlet temperature that could be achieved? Wbat assumptions do you make to compute this temperature? (b) If this temperature is too high for the materials on hand.05 0 0 Exercise 6. and is mixed \\lith a pure air stream to achjeve 100% excess air to CO. (a) For a range on:. what simple process design changes can lower the outlet temperature? (c) What would you do ooxt jf you wanted a more accurate estimate of the temperature of the eflluent from the reactor? Additional information.41? (c) What happens if you operate the autothermal reactor with a value of y less than the critical value found in part 6.06x IOsexp(-27200/RT) where T is in "R and R is in Btu/Obmo!"R).20h? kcal/mol 26. This stream enters the reactor tube bundle at 100·C. and 392 'F (852"R). solve the initiaJ.35 7.6 if we decrease the heat·transfer parameter. The rate expression is For the fixed volume element we have convection terms because material streams through the boundary. The reactor tern· perature must be maintained below S80°R.92 . 2.41.98 12. 0 Exercise 6.z and taking the limit leads to tbe energy balance J(pO) + ¥Vb z ero (6. for which we sbowed mHz+t. -48.6.18: The adiabatic limit in a PFR We .0 atm pressure. Equation 6.19: Using a simplified PFR energy balance Consider the following reaction in an adiabatic.000 Btu/lbmol and the heat capacity of the reactor feed miXture is The waste stream is a mixture of CO and air. (0) values. but no work term for moving the boundary because the element is fixed in space.7 Exercises Exercise 6.

100 to this moving element produces -' (6. t) plane. Hints: Figure 6. It is not difficult to add the heat teTIll to this development as well. n at ~at+azat a(pO) e(pO) oz (6. Applving Equation 6. [O(PU) "Z: + pU i)i.. starting with Equation 6. so this volnme element may expand or contract as it moves down the reactor.a"z] at 3. These lines are known as the characteristic lines for Equation 6.. In the transport literature this derivative also is called the material or substantial derivative. The right-hand side contains the usual partial derivatives in which \'y'e hold either z or t constant.350 The Energy Balance for Chemical Reactors 6.101. upper and middle steady-state solutions have coalesced. 2.102) B.103 is the rate of change of internal energy as we move Vvith the fluid along a characteristic line in Figure 6. in which ozjJt = v is the fluid velocity because we consider the moving element. hut the work term for moving the boundary is nonzero.z by considering the work done while moving the two ends of the element. Notice also the two circular eods of the element move with different velocities if the fluid density is not constant.41: Coolant temperature at reactor outlet versus temper~ 1. that choice does not affect the final energy balance. Show that we still arrive at the same energy halance. at the criti~ cal value of heat-transfer coefficient.C 300 200~--~--~----~--~----~--~--~--~ 200 300 400 500 600 700 Ta (O){K) 800 900 1000 in which the underline represents the quantities in the moving element.102 for the moving element.102 to obtain o lJi(pUA. Compute the change in the volume element size due to the change in the flUid velOcity with position. Equation 6.42.42: Volume element moving with the PFR fluid velocity. The left-hand side of Equation are zero.7 Exercises 351 800 700 600 T(K) 500 400 A in which v is the PFR fluid velocity. Combine these results to obtain Equation {UOL .101.z) ~ A.42. The location of the moving element traces out one of the lines shown in the (z. 4. First show 1::tb"'" -A c - " o(Pv) ature at reactor inlet. Consider a volume element that moves with the fluid velocity as depicted in Figure 6. az-6. and show 3Llz "" au &z at = ez- "Z(t2) Use the chain rule to show for pU e(pUI pU(z. TaU) versus Ta{O). Differentiate the left-hand side of Equation 6. The moral of the story is that the choice of volume element is based on convenience.103) Z o Figure 6. Notice for this choice of volume element the convection ter.

V. Eng. Englewood Cliffs.. 1974. Instabilities in two-phase reactors. Chern. Inc. R. 1161 B. Prog. Eng. John Wiley & Sons. Stability analysis of polymerization in continuous stirred-tank reactors. New jersey. Smith. A. R. [25) c. O'ConnelL Properties Gases and Liquids. Acta Physicochimica u'R. Chemical Reactor Analysis and Design. pages B3-47-83-58. Prausnitz. Properties and reactor design. H. 67:3040-3044. 1969. Adomaitis and A. Douglas. 45(6):1242-1247.1953.. Eng. and J. Roy.. [20) J. Korean]. R.1957. Pitzer. 1991. 1987. S. Sci. j. Sci. Aftalion. Poore. M. 47(15/16):4121-4132. Schork and W. M. Bush. G. Vau Welsenaere and G. New York. Chaotic Behavior of Deterministic Dissipative Systems. 1992. New York. J. M. van Heerden. Cambridge University Press. Pyzhev. New York. Bischoff. AIChE]. Cbemical Heritage Press. B. [26J R. 198G. Experimental evidence of chaotiC states in the Belousov-Zhabotinskii reaction. 6:123-131. third edition. W. Baccaro. Chaos during the oxidation of carbon monoxide over platinum ~ experiments and analysis. A. Experimental investigation of limit~cycle behavior for vinyl acetate polymerization. Proceedings of the fifth European Symposium on Chemical Reaction Engineering. 1990.Bibliography Bibliography [18] R. E. Chemical Engineering Kinetics. 25:1503-1516. 1977. H. 1992. 15(2):105-123. In (24) A Uppal. pages 758-813. Chem. L Hudson. Sci. F. Schmitz.S. 11 51 K. An experimental study of oscillating reactors. Polym. Sci. 1995. S. Ray. F. Marek and l. J.. R. 29:967-985. In L Lapidus and N.. New Yark. The equations of change and the macroscopic mass. An experimental study of oscillatory states in a stirred reactor. Prentice· Hall. A History of the International Chemical Industry. Autothermic processes. E. F. [141 M.. second edition. Ray. Amundson. Kinetics of ammonia synthesiS'1'hi promoted iron catalysts. Else. Sci. Sci" 47(15/16):4133-4140. McGraW-Hill. 12(3):327-356. Kenney. Hancock and C. Cbang and R. 1970. 1981. The dynamic behavior of continuous polymerization reactors . M. Ray. Chern. A. K. 1986. 1983. (21) M. 1988. Henderson. and J. Bailey. Sci. Eng. and R. Razon. Eng. second edition. M. Chem. Chern. Chang. Chern. j. [l3J L S. GraZiani.ier Publishing Company. Froment. Complex dynamics in full-scale reactors.. Ind. momentum. Eng. Bird. Ray. A.. and energy balances. Froment and K. D. S. and Bifurcations of Vector Fields. 30:21-34. B. Pitman Publishing. [19J F. Teymour and W. editors. Chern. Chem. n] 12] (3J [4] [5) [6) 17) 18] 19J (10) [Ill 1121 R. Marcb 1970. Process Modeling. Denn. Chem. Chern. The dynamic behavior of continuous polymerization reactors . 43(4):887-898.VI. Viewpoints on transport pbenomena.. Dynamical Systems. Eng. B. B. Parametric sensitivity and runaway in fixed bed catalytic reactors. j. M. Eng. Philadelphia. Sci.. and A. pages 42-50. Schmitz. On the dynamic behavior of continuous stirred tank reactors. The dynamics of the continuous emulsion polymerization of methylmethacrylate. Chemical Reactor Theory. 16(2):249-254. Translated by Otto Tbeodor Benfey. Chern. Y. ].. H. and J. Eng. Teymour and W.. Amsterdam. 309:1-26. Appl. McGraw Hill. New York. P. A. Eng. Soc. N.1998. Eng.. Springer Verlag. Thermodynamics. Schmitz. Bird. J. [22J F. or . Guckenheimer and P. Chem. Nonlinear Oscillations. Sci. Proc.. McGraW-Hill. Holmes. The bifurcation behavior of an autothermal packed bed reactor. 1977. 34:1259-1276.. Poling. M.-M. 1940. March 1987. Phys. A Review. P. [23J F. G. The measurement and prediction of sustained temperature oscillations in a chemical reactor. Cinar.S. F. 1972.. Periodic phenomena.1975. Galtonde. 41:1561-1576. third edition. Schreiber. N. 200L 1171 L F. Temkin and V. 2001. H.

and the catalyst particles. t n S (7_1) (7. as a function of concentration and temperature. Equations 7. For heterogeneously catalyzed reactions. in which mass and energy transfer are very rapid.. ri. The concentrations . and intraparticle transport control. which occupies the interstitial region between catalyst particles. reactor as long as the plug-flow conditions are satisfied. we showed in Chapter 6 that the molar flows and temperature are governed by the follOwing differential equations j = 1.2 .1 and 7.2) A rate expression. plug-flow reactor. These balances are applicable to reactions that are homogeneous or heterogeneous. For a steady-state.. Catalytic systems operate somewhere between two extremes: kinetic control. the heterogeneous case is referred to as a fixed-bed reactor. tubular. for each of the nr reactions is required to complete the problem description.7 Fixed-Bed Catalytic Reactors 7.1 Introduction The analysis and rational design of reactors requires the simultaneous solution of material and energy halances. in which the reaction is very rapid. Separate material and energy balances are needed to describe the concentration and temperature profile inside the catalyst pellet. the e!"fects of IntraparticJe transport on the rate of reaction must be considered. In a fixed-bed reactor the catalyst pellets are held in place and do not move with respect to a fixed reference frame. in whicb the reactions occur. Material and energy balances are required for both the fluid..2 can be used for any reaction within the .

Table 7. The other steps are inherently faster than the slow step(s) and can accommodate any change in the rate of the slow step.7. For kinetic or reaction control. Figure 7. 2. This chapter's goal is to descrihe the reaction at any location within the reactor. transport of reactants and energy from the external surface into the porous pellet. the pellet.1: Expanded views of a fixed·bed reactor. are covered in Chapter 5. During any catalytic reaction the follOWing steps occur: L transport of reactants and energy from the bulk fluid up to the A D B Figure 7. catalyst pellet exterior surface. 3. That is the reason we are using a catalyst. deriving ana· lytical solutions develops intuition and clearly illustrates the coupling between the pellet and the fluid. hetween the surface and the bulk. (Om. if step 1 is the slowest process.1 lists some of th~ important commercial . surface and bulk temperatures are assumed to be equaL The five steps Just listed describe the simultaneous processes that occur in the fixed·bed reactor. except in some special cases. desorption and surface reaction. step 3 is the slowest process. Essentially all reaction occurs within the cata· lyst particles. These processes occur at a flUid·solid interface and the rates seal" direct Iv with the tolal nllmber of surface sites.1 presents several views of the fixed·bed reactor. The system is intraparticle transport controlled if step 2 is the slow process (sometimes referred to as diffusion limited). The fluid in contact with the external surface of the cat· alyst pellet is denoted with subscript 5. we model the system on the scale of Figure 7. adsorption. 7. D i .1 C The problem is solved for one pellet by averaging the microscopiC processes that occur on the scale of Figure 7.2 Catalyst Properties The chemical steps for catalytic reactions: adsorption. In this chapter. Finally. chemical reaction. the reaction is said to be externally transport controlled. In certain cases.1). Even though the applicability of these relationships is limited. we use a tilde on the variables within the catalyst pellet. Numerical methods are used to solve the resulting coupled differential equations. lD over the volume of the pellet or over a solid surface volume element.2 Catalyst Properties 357 of each component and the temperature in the pellet are related to the concentrations and temperature in the bulk fluid through the boundary conditions that complete tbe problem description. and S. analytical solutions to the transport of mass and energy within the catalyst pellet can be used to develop al· gebraic relationships that couple the rate of reaction vvithin the pellet to the temperature and concentration that exist in the bulk flUid.356 Fixed·Bed Catalytic Reactors 7. This procedure requires an effec· tive diffUSion coeffiCient. The coupling of transport processes with chemical reaction can lead to concentration and temperature gradients within the pellet. or both. When we need to discuss both fluid and pellet concentrations and temperatures (see Section 7. to be identified that contains information about the physical diffusion process and pore structure. transport of products into the bulk fluid. transport of products from the catalyst interior to the external surface of the pellet. For the remainder of this section. The species production rates in the bulk fluid are essentially zero. Equa· tion 7. and desorption of products at the catalytic sites. Usually one or at most two of these steps are rate limiting and act to influence the overall rate of reaction in the pellet. I applies to the bulk fluid in the interstitial volume. 4.

2 propylene . and up to 2.2 Catalyst Properties 359 catalysts and their uses [12].Hz + C02 ~ 7(p""(-. e. Po is the normal saturation vapor . 14. Pd. 0. The solid density is denoted Ps. + 0.3-1. . . ~ 2 503 ~ V20S (on metal oxide support) Bismuth molybdate :Mixed oXlctes of Fe and Mo Naphthalene + 9/20. Total catalyst area is generally determined using a physically adsorbed species. In the case of metal catalysts. promoted by alkali metals) VzOs or Pt Paraffin hydrocracking Olefin polymerization. 0. SiOz-Al203. Pt.ethylene oxide (with C02 + H20) 2 S0..paraffin Metals (e. The pore structure is a strong function of the preparation method. Ni.3) Table 7. . .. acetone + Hz ~ isopropanol Complete oxidation of hydrocarbons. In the physisorption process multi-layers are allowed to form. or it can be made into a high-area oxide. . the metal is generally dispersed onto a high-area oxide such as alumina. e. Copyright © 1992 [12]. These carrier oxides can have surface areas ranging from 0.. The procedure was developed in tbe 1930s by Brunauer. Emmett and and Teller [7].. 15] provide an excellent introduction to catalysis and catalytic processes.g.05 m 2 /g to greater than 100 m 2 /g. e. in which c is a parameter related to the heats of adsorption in the first layer and subsequent layers. The BET isotherm model treats the first layer differently from all subsequent layers and leads to the following expression V Catalyst Reaction C C bond hydrogenation. + 2(Oz +2HzO CH30H + 02 phthalic anhydride ProPYlene + 1/202 . e. The pellet volume consists of both void and solid.5 em diameter hollow cylinders or rings.. the nmnher of sites per unit reactor volume should be such that the rate of product formation is on the order of 1 mol/L·hour [21].1: Industria! reactions over heterogeneous catalysts. olefin + Hz . 13.polyethylene Olefin metathesis.g._-=pCC-)"C(~l~+~( C VmCP - 1) P I Po) (7. Inc. Ti or Zr Metal-oxide-supported oxides of WorRe Ag(on inert support. and the isotherm they developed is referred to as the BET isotherm. 'A-ith other components.g. . The carrier oxides generally are pressed into shapes or extruded into pellets.3 em diameter cylinders with a length-to-diameter ratio of 3-4. such as N2.acrolein formaldehyde (with CO.7 em diameter spheres for fixed-bed reactors. zeolites) Y-AlZ 0 3 CIl30H Paraffin dehydrogenation. Cu.g. AI203) Re + Pt (supported on fJ-Al203 or Y-A1203 promoted with chloride) Solid acids (e. References [12.paraffins + olefins + H20 + C02 (+ other oxygen-containing organic compounds) (Fischer-Tropsch reaction) CO + 2H z ~ Fe or Co (supported and promoted with alkali metals) eu (supported on ZnO.3-0.358 Fixed-8ed Catalytic Reactors 7. The pellet void fraction (or porosity) is denoted by € and c ~ PpVg in which Pp is the effective particle or pellet density and Vg is the pore volume.1D shows a schematic representation of the cross section of a single pellet.O) H20 + CO .g" Cr.ethylene + butene Ethylene + 1/2 02 .g. + H. The following shapes are frequently used in applications: 20-100 /lm diameter spheres for fluidized-bed reactors. Pore sizes can be as small as molecular dimensions (several Angstroms) or as large as several millimeters. This material is used by permission of John Wiley & Sons.g. PO Fe {supported and promoted with alkali metals) Ni (=0 bond hydrogenation. PO Metal (e. Figure 7. .. Metal oxides also can be dispersed on a second carrier oxide such as vanadia supported on titania.03) Metals (e. To make a catalytic process commercially viahle.g" Pd.1I cm 3 /g pellet.oxidation of CO 3Hz + Nz ~ 2NH3 CO + 3Hz .. as powders or on supports) or metal oxides (e. CH4 + HzO (methanatjon) CO + Hz .. isomerization and dehydrocyclization Paraffin cracking and isomerization Alcohol olefin + H20 Pd supported on acidic zeolite Metal-oxide-supported complexes of Cr..g. and catalysts can have pore volumes (Vg) ranging from 0.. ethylene . The pores can be the same size or there can he a bimodal distribution with pores of two different sizes. a large size to facilitate transport and a small size to contain the active catalyst sites.g" Ni. e.

of the pores is then Ya = . For example.0018583 V fp (2.162 nm 2 • The total area per gram of catalyst is denoted by S9' 7. One then assumes each adsorbed molecule in the volume Vm covers a particular area. P in.1 A number of correlations have been developed for bulk flow that describe binary interactions. Finally.. pressure and composition.AB Effective Diffusivity We define a diffusion coefficient that takes into account the random distribution of solid and void as one moves from the exterior to the interior of the pellet.2 discussed pore sizes and pore size distributions. any void space transport can be described by the diffusion of a substance under a concentration gradient. all the binary interactions must be considered. i. in which Vm is the volume of gas that corresponds to a monolayer of the entire solid. r a . Starting at any pore opening in Fignre 7. Knudsen flow dominates and diffusion is described by collisions with the pore walls. In multicomponent systems (lis species). The correct and complete treatment requires the use of Stefan-Maxwell relations.. 1\.5 em at 1. respectively.\ < Y a by an order of magnitUde.\ ~ J2rra zp and it has a value of approximately 10. (Yk-YJ·~~)+Ib (7. Uptake data are fit to Equation 7. This one limiting case illustrates tliat the diffusion coefficient can change with temperature. For Knudsen flow the diffusivlty of species j is DK ~ 8 r a RT J 3 Y21TMJ I (7. r----~· DAB ~ 0. the diffusivity of species j in the mixture. D Jm . For N2 at 77 K the area of each molecule is 0. + 2. We ooly consider the first two in this text.4) The type of diffusion in the pore.5) Substitution of the gas constant gives . Assume here that all pores are straight cylinders and that the pellet has a single pore size distribution.0 atm and room temperature for simple gases. bulk flow dominates and diffusion is described by collisions in the gas phase. For a gas-phase molecule the mean free path. Again for the single reaction n.) /"A MB~ P(TAB[lD. This idea leads to the concept of an effective diffusivity. Liquid-phase diffusion is described by liquid-liquid intermolecular collisions. The average radius.1D and moving inward along a straight line to the center.2. transport is described by surface or configurational diffusion. when oniy one reaction is present for a multicomponent system. and v is the volume of gas adsorbed at a particular pressure P. Inside.e.atm and (TAB in Angstrom units. as well as the contributions relative molar fluxes make to the transport of each component.360 Fixed-Bed Catalytic Reactors 7. can be found using in which J - 8-~ '\en. hi 2Vg Sg (7. The diffusion rate is controlled by the microscopic details of the transport in the voids. the footprint of the molecule at the adsorption temperatnre. Section 7.. kET D/K ~ 9. When. v ] which predicts the diffusivity in cm' /s for T in K and MJ in g/moL Bulk or Knudsen flow dominate when Y" and . A combined diffusivity is needed when ra and .\ > ra by an order of magnitude.\ are of the same order of magnitude.7 x 103 ra .G) . so that the balance can be wTitten over the entire pellet.\ vary by at least an order of magnitude.- which predicts the diffusivity in cm 2 /s for Tin K.3 to determine c and Vm. When . when Y a is approximately equal to the dimensions of the diffusing reactants. depends on whether the diffUSing species collide more often "itli each other or with the pore wall surface. One such correlation is the ChapmanEnskog relation. is in which Vk and Yk represent the stOichiometric coefficient and mole fraction of component k. bulk or Knudsen.2 Catalyst Properties 361 pressure of the adsorbate at the adsorption temperature./ . one quickly encounters solid.

7 can be used to describe diffusion down a straight cylindrical pore. 1. Consider the volume element depicted in Figure 7. Now if one considers the solid to be an equivalent homogeneous medium. one finds T = 3. however.7) Equations 7. As more anisotropy in the pore distribution is introduced by the manufacturing process.7 Girdler G·58 Pd on alumina 0. 0.56 pelletized Cr203 supported on Aha. T increases. (7. Because of this variability in T.1C.3 6. If the pores are modeled as straight cylindrical tubes of various lengths.2.0 I I Table 7.2: Porosity and tortuosity factors for diffusion in catalysts.22 2. . the rate of mass transfer is solved rigorously in a pellet at the level depicted in Figure 7. The texts by Aris [2J and Petersen [lS] illustrate the approach. In short. Similarly. De. As the microscopic description changes..] 0.9 0.39 2.416 1.5 pelletized boehmite alumina 0.8) in which D j refers to either Djm or D JK and 1. J= The integral!. The conservation of mass and energy for the volume element .6 and 7.1C.3 The General Balances in the Catalyst Particle In this section we consider the mass and energy balances that arise with diffusion in the solid catalyst particle when considered at the scale of Figure 7. cross-linked. We also can use these microscopiC models to describe diffusion radially inward. It is doue to remind liS that as a diffusing molecule travels radially inward it encounters solid and void but it diffuses mainly in the void regions. Equation 7. A porous solid does not consist of straight cylindrical pores.3 The General Balances in the Catalyst Particle 363 in which Nk is the molar flux of component k in the pore.34 2. pore model can be used to predict the valoe of T.0 pelletized Ag 0. would apply to other diffusing species as well. all of which can be measured experimentally.59 3.5% Pd on alumina 0.5% Ca alloy 0. rather it must follow a tortuous path that effectively adds to the distance of travel. the molecule cannot travel in a straight line from the exterior to the center point. For binary systems (components A and B) with equimolar counter diffusion. which is the level depicted in Figure 7. Models for pore structure have been proposed that describe the pore size distrihution and orientation as a function of location within the pellet [2J. The values of T generally range from about 2 to 7. Table 7. is an integral over the bounding surface of a dimensionless concentration gradient for j that depends on the microscopic description of the pore surfaces. ¢) within the pellet.2 lists experimental values for T [8].S and 7. changes. 7. (7. each having the same length and radius.43 3.10 to give an equation of the form 0.9 leads to in which T is the tortuosity factor.0% Pd on alumina pelletized Ag/S. e. is generally represented as 1. then the same rate can described in terms of an effective diffusion coefficient.9) Comparison of Equations 7. it should be determined experimentally for the catalyst of interest. Rigorous models such as the parallel.ns Let E be the total energy within the volume element and e be the flux of total energy through the bounding surface due to all mechanisms of transport.3 10. pore volume and pore area.. this same 1. The use of the ratio E/T has no real physical Significance.5 _.362 Fixed-Bed Catalytic Reactors 7. accounting for the changing pore shape and the presence of void or solid as a function of any position (Y.2.6 reduces to D Ac Catalyst 100-11 Opm powder packed into a tube E T 1 I 1 --=--+-DAB DAK (7.5 1. . .8 Haldor-Tops0e MeOH synthesis catalyst I 0.3 . 7. pore size distribution. These microscopic descriptions can be used to predict the porosity. Assume a fixed laboratory coordinate system in which the velocities are defined and let v 1 be the velocity of species j giving rise to molar flux N.

ns (7. and neglect kinetic.12) In this chapter. c1uding the flux of species j due to gradients in the other species con" centrations. in which k is the effective thermal conductivity of the solid...ns (7. 767768] provide further discussion of these contributions to the mass !lux. using the definition of tbe heat of reaction.N+Rot J J' oE = -'V. In multiple"reaction.10 produces (7.364 Fixed"Bed Catalytic Reactors 7.4 Single Reaction In an Isothermal Particle 36S e We approximate the total energy flux by e = -k'VT + j I j NjH j This expression accounts for the transfer of heat by conduction. we consider the internal energy to be the dominant contribution to total energy. Set" ting the time derivative to zero in Equation 7. 7. and the flux of energy due to concenl'fation gradients (Dufour energy) and radiation. in. pressure gradients. solid particle.. k'VT . we must solve these equations numerically.HRiri .14) 0= kV2T .2" . and setting the time derivative to zero. As in Chapter 6. .10) 0= V'. . l. thermal conductivity and partial molar enthalpies are independent of temperature and compo· sition to produce the follOwing coupled mass and energy balances for the steady· state problem o = DjV 2 cj +R j . therefore. Next we consider the fluxes.. . e ilt Substituting Equation 7. and the flux of species j due to temperature gradients (thermal diffusion).2. so the assumption of constant transport and thermodynamic properties is driven by the lack of data.4 Single Reaction in an Isothermal Particle We start with the simplest cases and steadily remove restrictions and increase the generality.HRiri (7.k'VT-I[('V·Nj)Hj+Nj. noniosthermal problems.13) implies _J oC =-'V. so E = U.11) j . Stewart aod Lightfoot [5.IN j (7.2: Volume element in a stationary solid containing energy density E and molar concentrations Cjt with energy flux e and mass fluxes Nj. pp. and gravitational or other ex· ternal fields (forced diffusion). and not ana" lytical convenience.I6. we consider a single . Bird. ot ='V. We have neglected any viscous heating effects.15) in which D j is an effective diffusivity for species j.12. In this section.I6. we assume the mass flux is well approximated by Finally.'VHd j . So the energy balance becomes iJ(pO) Figure 7. Since we are considering the diffusion of mass in a stationary. Notice we are ne· glecting a variety of other possible mechanisms of mass transport. j = 1. and transport of energy due to the mass diffusion. (7. 'VHj in which R j accounts for the production of species j due to chemical reaction. potential and other forms of energy. yields j:::.13 into Equation 7. we often assume that the diffusivity. we are concerned mostly with the steady state.

In this section we assume the concentration atthe outer boundary of the pellet. Because the reaction is first order. 1 Some We now wish to solve Equation 7.4 Single Reaction in an Isothermal Particle 367 reaction taking place in an isothermal particle. first-order reaction.20) = CAs.17 has two dimensional quantities. the term modulus was retaiu<:d. Thiele number would seem a better choice. Later when Thiele's name was assigned to this dimensionless group. first-order reaction occurring in an isothermal spherical pellet A.!£ (r2 dCA) dr dr (7. For the sphere. catalyst density or catalyst area.!£ (-2 dr de) y2 dY r _". and assuming pellet symmetry in e and <p coordinates gives D Ay2 l.366 Fixed-Bed Catalytic Reactors 7. but we will find that a better choice is to use the pellet's volume-to-surface ratio. Equation 7. but the tenn Thiele modulus has become entrenched. Be aware that the unlts for a heterogeneous reaction rate constant are sometimes expressed per mass or per area of catalyst.' Therefore the two boundary conditions for Equation 7. We might naturally choose the sphere may prefer to use other boundary conditions at r co.= 0 '¥ C (7. 0. Next we consider other catalyst shapes. and neglect the external mass-transfer resistance.2 Tbe Thiele modulus quantifies the ratio of the reaction rate to the diffusion rate in the pellet. (7..16) Dividing through by the various dimensional quantities produces Substituting the production rate into Equation 7. Thiele used the term modulus to emphasize that this then unnamed dimensionless group was positive. is known.!<.19) l. and then other kinetic expressions such as the Hougen-Watson kinetics of Chapter 5.17 are CA in which <I> is given by reaction rate diffusion rate Thiele modulus (7. the first-order rate constant k has unlts of inverse time. These boundary conditions are equivalent for the problems conSidered here. as illustrated in Example 7. We start with the spherical particle. r=R The single dimensionless group appearing in the model is referred to as the Thiele number or Thiele modulus in recognition of Thiele's pioneering contribUtion in this area [20]. 7. In these cases. expressing the equation in spherical coordinates. then other reaction orders. We end this section by considering the effects of finite external mass transfer. CAs.2. so we use that quantity to nondimensionalize the concentration In this section we consider a single. E. the model is linear and we his original paper.17) r=3 in which DAis the effective diffusivity in the pellet for species A As written here.19 with the given boundary conditions.1. We require two boundaty conditions for Equation 7. irreversible.21) dCA = 0 dr At this point we can obtain better insight by converting the problem into dimensionless form. such as boundedness of CA or dCAldr. or a zero source/sink condition.14. length and concentration.4. the reaction rate expression includes the conversion factors.17.18) Sp 4rrR2 3 The only concentration appearing in the problem is the surface concentration in the boundary condition. this characteristic length is Vp §rrR3 R a=-=--=(7. and the symmetry of the spherical pellet implies the vanlshing of the derivative at the center of the pellet. . Equation 7.1 First-Order Reaction in a Spherical Particle radius R as the length scale.20 can be considered as a special case of a more general definltion (7.

4: Dimensionless concentration versus dimensionless radial position for different values of the Thiele modulus.25) Figure 7.0 The constants C1 and C2 are determined by the boundary conditions. dZu _ <l>zu d1' 2 =0 (7.r smhr = 2 tanhr = ~~ dcoshr .19 provides a simpler differ· ential equation for u en.-------~------~--~~~--------~~_.ll--. It is often convenient in spherical coordinates to consider the variable transformation -(-I' 4.22 gives the solution to the model _(=) _ C <I> = 2. Some of the properties of the hyper· bolic functions are 1 0.24) This solution is analogous to the sine and cosine solutions if one reo places the negative sign with a positive sign in Equation 7.8 0.3.22) 2 1 cosh r Substituting this relation into Equation 7.23.6 0. u W) = C1 cosh <1>1' + C2 sinh <1>1' (7.. = 0. cosh and tanh."::.5 l' 2 2. Substituting Equation 7.24 and 7. Note for small values of Thiele modulus.4 displays this solution for various values of the Thiele modu' Ius. and applying the boundary condition at l' = 3 gives C2 = 3/ sinh 3<1>.23) o ·2 -3 -----------~~--------- .2 <I> = 0.4 0.1 .3: Hyperbolic trigonometric functions sinh.l' sinh 3<1> i (7.368 Fixed·Bed Catalytic Reactors 7. These functions are shown in Figure 7. h dr =sm r dsinhr dr =coshr <P = 0. the reaction rate is small .S I 1..4 Single Reaction in an Isothermal Particle 369 can derive an analytical solution. Figure 7. e r e.5 Sinhr coshr c <I> ~ 1.:~-?~ tanh r with the transformed boundary conditions u=3 u=o 1'=3 1'=0 sinh r The boundary condition u = 0 at l' = 0 ensures that c is finite at the center of the pellet The student may recall from the differential equa· tions course that the solution to Equation 7. cr =-=- uW) r 3 (7.':.24 into the boundary condition at l' = 0 gives C.0 0 0 O. Substituting these results into Equations 7.23 is r Figure 7.5 3 r sinh <1>1" .

85 g/cm3 . the rate of consumption of reactant within the pellet is equal to the rate at which material fluxes through the pellet's exterior surface.3 cm radius spherical catalyst pellet at T = 450 K.007 cm 2 /s.pelleL For IJ near zero. and we see the two asymptotic limits of Equation 7. The effective diffusivity of A in the pellet is D A ~ 0.26) (7.5 and 7.6 shows that the asymptote I) = 1/<1> is an excellent approximation for the spherical pellet for <I> 2: 10. Figure 7. the rate of reaction is much greater than the rate of diffusion.3 cm pellet to find the value for the rate constant k. and we say the pellet is diffusion limited. the reaction rate is large compared to the diffusion rate. the concentration everYlvhere in the pellet would be equal to the surface concentration.31 into the definition of effec· tiveness factor to obtain for the first-order reaction in the spherical pellet 1) ~ iii tanh 3<1> 1[1 . and at large <1>.15 cm radiUS spherical pellet. B) takes place in a 0. Solution We can use the production rate and pellet parameters for the 0.32) effectiveness factor The effectiveness factor is a dimensiouless pellet production rate that measures how effectively the catalyst is being used. Determine the production rale at the same temperature in a 0.1: Using the Thiele modulus and effectiveness factor The first-order. at steady state. irreversible reaction (A -.4 Single Reaction in an Isothermal Particle j// compared to the diffusion rate.25. the effectiveness factor is near unity.tanh 3<1> . We use the direct method here and leave the other method as an exercise.5 x 10.6 is particularly useful. (7.5 mol/(g s). The first and more direct method is to integrate the local production rate over the pellet volume. We can perform this calculation in two ways. and the reactant is converted to product before it can penetrate very far into the pellet.. the pellet's production rate is -2.7 atm partial pressure of A. and the pellet concentration becomes nearly uniform. the effectiveness factor is much less than unity. At small <1>. and then compute the Thiele It is instructive to compare this actual pellet production rate to the rate in the absence of diffusional resistance. and changing the vanable of mtegration to x = <l>r gives R Ap = - 3<1>2 sinh 3<1> k CAs 3<1> x smh x dx (7.27) We define the effectiveness factor. the pellet reacts at a low rate. the entire volume of the pellet is reacting at the same high rate hecause the reactant is able to diffuse quickly through the.310 Fixed·Bed Catalytic Reactors 7.29) Figures 7. For I) near unity. The integral can be found in a table or derived by integration by parts to yield finally (730) RAp ~ -kc As . Conversely. and we say the pellet is reaction limited. Equation 7. We now calculate the pellet's overall production rate given this con· centration profile.3<1> <I> 1[1 1] I Example 7. The reactant is unable to penetrate significantly into the interior of the pellet and the reaction rate is small in a large portion of the pellet volume.31) . We can substitute Equations 730 and 7. The log·log scale in Fig· ure 7. when the diffusion rate is much larger than the reaction rate. The pellet's diffusional resistance is large and this resistance lowers the overall reaction rate. The pellet density is Pv = 0. For large values of the Thiele modulus. to be the ratio of these two rates (7.26 and converting the integral to dimensionless radius gives kCAs R Ap = --9- r3 crYJo -(-) °dY r J0 .3<1> 1] (7.33.33) in which the local production rate is given by RA (r) = -kCA (r).28) Substituting the concentration profile. The two expressions are volume integral surface flux (assumes steady state) (7. corresponding to the limit <I> = O. The second method is to use the fact that.6 display the effectiveness factor versus Thiele modulus relationship given in Equation 733. I) ~ 1/<1>. If the diffusion were arbitrarily fast. For large values of Thiele modulus. The pellet rate for this limiting case is simply (7. I). IJ ~ 1. At 0. Substituting the local production rate into Equation 7.

3 0. We have three unknowns.60 <I> = 1.1 Using this result.36 into 7.) = -2.35 for k.3~] = 1.3 7 gives <I> h'~ 3<1> .90 x 10 mol/em -5 3 (-2. The production rate is therefore RAp = -0.4 0.1 0 -' <I> [ (7. 0 . RA P (450 K) = 1.smol/cm3 ) = -3.36) The production rate is given in the problem statement.93 This equation can be solved numerically yielding the Thiele modulus 0.35 gives the rate constant om k = 2.6: Effectiveness factor versus Thiele modulus for a ftrstorder reaction in a sphere (log-log scale).35) 0.964. k.8 0.5 x 1O_ SmOI ) (0.685 (2.D ACAs 2 (7.9 modulus.685.125 mol cm3 S Substimting these values into Equation 7. so the Thiele modulus is half as large or <I> = 0.34) 0.61 S·l The smaller pellet is half the radius of the larger pellet.7 O.JL.90 x 10. and the following three equations (7.<I>.5: Effectiveness factor versus Thiele modulus for a firstorder reaction in a sphere.38 x 10. and substimting that result and Equation 7.2 0.1 ) (1. 85 gS cm3 .6S.7 atm 82.5 0. give one equation in the unknown <I> tanh 3<1> 1 _ ~] 3<1> __ RA p a .312 Fixed-Bed Catalytic Reactors 1. Solving Equation 7.B I) (7. Equation 7. I). effectiveness factor and production rate for the smaller pellet. Notice that this type of increase is possible only when the pellet is in the diffusion-limited regime. which gives I) ~ 0. We see that decreasing the pellet size increases the production rate by ahnost 60%.34.06c:ot~m 3) Figure 7.3 7) 0 2 4 6 8 10 oJ> 12 14 16 18 20 The surface concentration and pellet production rates are given by CAs =( = 0.4 Single Reaction in an Isothermal Particle 313 1 0. 5 mOls Figure 7.

The maximum difference hetween the effectiveness factors of the sphere and the slab l] is about 16%.38) The effectiveness factor for the different geometries can be evaluated using the integral and flux approaches.7. d"C ~ 0 Y~O in which the boundary conditions for all three geometries can be compactly expressed in terms of q. and the sphere is the smallest at all values of Thiele modulus. .7: Characteristic length a for sphere. The results are summarized in Table 7. a~R. depicted in Figure 7.39 and 7. which lead to the two expreSSions _ I) I)~ 1 (q + 1)q rq + 1 _ q _ Jo cr dr P"-.4 Single Reaction in an Isothermal Particle 375 R.41 for the sphere./3 The characteristic length a is again best defined as the volume-tosurface ratio. To have a simple analytical solution.8. are modified Bessel functions of the first kind of orders zero and one. See Exercise 7. We can summarize the reaction-diffusion mass balance for these three geometries by (7.27. semi-infinite cylinder and semi-infinite slab. Equations 7. the cylinder is intermediate. Note that Jo and 1.q+l (7. and occurs at <I> ~ 1. Although the functional forms listed in Table 7..~ 0 C Y~q+l Figure 7. we must neglect the end effects.1 for the slab geometry.40) (7.5 and <I> > 7. we see in Figure 7. Analytical solutions for the slab and cylinder geometries also can be derived. we therefore consider in addition to the sphere of radius R s .3 appear quite different.41) 1 d"C -4>2 dr I in which sphere cylinder slab The associated boundary conditions are r ~ Rs r = Rc r~L CA = CAs ~0 { sphere cylinder slab ali geometries dCA dr r ~ 0 We have already solved Equations 7. the semi-infinite cylinder of radius R e .3.6. respectively. the difference between all three effectiveness factors is less thao 5%. The effectiveness factor for the slab is largest. For <I> < 0.39) ~~ yq dY r dY "c~1 (-q de) _<1>2. and the semi-infinite slab of thickness 2L.4.374 Fixed-Bed Catalytic Reactors 7. which gives for these geometries sphere cylinder slab The dimensionless form of Equation 7. The effectiveness factors versus Thiele modulus for the three geometries are plotted in Figure 7. q ~ 2.26-7.38 is -' (7. 7. The three curves have identical small <l> and large <I> asymptotes.8 that these solutions are quite similar.2 Other Catalyst Shapes: Cylinders and Slabs Here we consider the cylinder and slab geometries in addition to the sphere covered in the previous section.

43) For reactions other than first order.4 Single Reaction in an Isothermal Particle 311 7.3 Other Reaction Orders Sphere I) = 1[1 1] 1> tanh 3<1> . Figure 7. Figure 7.3: Closed-form solution for the effectiveness factor versus Thiele modulus for the sphere. cylinder and slab. Notice the discontinuity in slope of the effectiveness factor versus Thiele modulus that occurs when the order is less than unity. Notice that the definition of Thiele modulus in Equation 7. We will see.44) -' (7.. Re~all from . The reaction-diffusion equation and boundary conditions for this case are (7. We consider nth-order.9 shows the effect of reaction order for n . irreversible reaction kinetics A Jc. As the reaction order increases." 1 in a spherical pellet. r = kc~ Slab I) (7.. and semi-infinite slab.46) We might naturally define the Thiele modulus as In Equation 7.21.-- c=1 I) dY dc = a I qT1 am I) _ I) - r (q + 1)" Jo n+ <I> -n-qd- err r~q.47 has achieved the desired goal of giving all reaction orders a common asymptote at high values of <1>. the effectiveness factor decreases.1 1 <I> !O 100 Figure 7. the reaction-diffusion equation is nonlInear and numerical solution is required.3<1> tanh <l> <I> (7.44) .001 am 0..47) 1 sphere(7A2) .10 shows the effectiveness factor versus Thiele modulus for reaction orders less than unity. semi-infinite cylinder.316 Fixed-Bed Catalytic Reactors 1.45) Table 7.43) ----slab(7. that many of the conclusions from the analysis of the first-order reaction case still apply for other reaction orders.cylInder(7. B.42) Cylinder I) 1 Id2<1» = ---<I> 10(2<1» = (7.4. bowever. ~ 1-\ dr d~1 [ 0. but the results for various reaction orders have a common asymptote if we Instead define ThIele modulus nth-order reaction (7.8: Effectiveness factor versus Thiele modulus for the sphere.

an inner region of the pellet has identically zero A concentration. for large ¢ the inner region of the pellet has zero A concentration.5 3 r Figure 7.577 <P = 0. For large values of Thiele modulus.1 1 <P = 0.5 1 1.8 <P = 0. This radius increases . the concentration of A reaches zero in finite time.8 displays the discontinuity in the slope of concentration versus time. Notice that the discontinuity in slope of IJ versus if> in Figure 7. For larger values of ¢.577. the A concentration reaches zero at finite radius.4 Single Reaction in an Isothermal Particle 379 n = 2 ----n = 5 -----n = 10- n~1 -- 1 0. Equation 4.9: Effectiveness factor versus Thiele modulus in a spherical pellet. the A concentration reaches zero just at the center of the pellet.11 shows the reactant concentration versus radius for the zero-order reaction case in a sphere at various values of Thiele modulus_ For ¢ = 0.2 0_1 0. the same kinetic effect causes the discontinuity in 1) versus <P.378 Fixed-Bed Catalytic Reactors 7.6 I) C 0.8 <P = 10 2 ]0 0 0 0.28 shows the discontinuity that is introduced into the mass balance when the concentration reaches zero. reaction orders greater than unity. For orders less than unity.4 0. reaction orders less than unity_ the discussion in Chapter 4 that if the reaction order is less than unity in a batch reactor.11: Dimensionless concentration versus radius for zeroorder reaction (n = 0) in a spherical pellet (q = 2).1 0. figure 7. and Figure 4.1 1 10 figure 7. n=O n = 1/4 ----= n=1/2------ n 3/4 - n=1 1) 0. Figure 7. the diffusion is slow compared to reaction.4 0.5 2. and the A concentration reaches zero at some nonzero radius inside the pellet.10: Effectiveness factor versus Thiele modulus in a spherical pellet. In tbe reaction-diffusion problem in the pellet. 10 occurs at this same <P value.

In this section we study an example reaction mechanism of this form. without resorting to numerical integration.4 Single Reaction in an Isothermal Particle 381 with increase in Thiele modulus so that for <!> ~ la. This goal was accomplished for the nth-order reaction as sho"n in Figures 7.380 Fixed-Bed Catalytic Reactors 7.52.497. and taking the limit produces (7_54) e~l z=O Z=oo de ~ 0 e~l r~1 dz dr de ~ 0 (7.47. The new dimensionless group <p represents a dimenSionless adsorption constant The effectiveness factor is calculated from RAp -(Sp/Vp)DA dCA/drl rea lJ = RAs ~ -kcmKAcAs/(l + KACAs) Given the discussion in Section 5. for all values of <p.9 and 7.52) y=l products r ~ kCrn KACA 1 + KACA (7. it is reasonable 10 expect our best catalyst rate expressions may be of the Hougen-Watson form. to evaluate the effectiveness factoL . If we consider the slab catalyst geometry.51) The effectiveness factor is obtained by changing variables in Equation 7.48) This expression arises when gas-phase A adsorbs onto the catalyst surface and the reaction is first order in the adsorbed A concentration. Consider the follo"ing reaction and rate expression A - which becomes upon definltion of the dimensioniess qua. producing l+<pdCI (7.4. the mass balance is d2CA D A -2 dr - Now we wish to define a Thiele modulus so that lJ has a common asymptote at large q.6 of adsorption and reactions on catalyst surfaces.10 by induding the factor (n + 1) /2 in the definition of <!> given in Equation 7. 7. only a small outer shell of the pellet has nonzero A concentration. Asymptotic behavior for large 4>.54. Then we take the limit 4> ~ 00 to obtain the model describing the asymptotic lJ ~ 1 12<P dd~ 1_ '¥ r (7.50) in which we now have two dimensionless groups (7.4 Hougen-Watson Kinetics We use the tilde to indicate 4> is a good first guess for a Thiele modulus for this problem. We now turn our attention to finding the right Thiele modulus for the HougenWatson kinetics problem. but we "ill find a better candidate subsequently. Equations 7. As a first step we change the independent variable from r to (7. First we define dimensionless concentration and length as before to arrive at the dimensionless reaction-diffusion model (7.55) 17 ~ -~ dz zoo We need to perform one integral of Equation 7. The same qualitative picture holds for all values of reaction order less than one. First we calculate lJ for large 4>. Changing variables in the reaction-diffusion model.50.49) in order to remove 4> from the differential equation [18J.53) kem KACA 1 +KACA a and the boundary conditions are CA = CAs r~L dCA ~ 0 dr r~O We would like to study the effectiveness factor for these kinetiCS.

1 + </>e Obviously it does not pay to try to guess an appropriate Thiele modulus for these more complex rate expressions. we obtain Figure 7_13 in which the various values of </> have different asymptotes.57 tells us what we need to know. First.14 displays things more clearly. The variable transformation is p(z) = p(z(e» = pic) c~ </» . May and June of 1965.13 and 7. which corresponds to z = 0 (see Figure 7.12).56 at e = 1. consider the function z(c) in place ofc(z).14.14. we obtain the results in Figure 7. 113] This is the essential idea in three papers published independently in March.12: Dimensionless concentration C versus z for large ~. To quote from Aris [2.p (1+</» I '12 (</> -In(l + </») (7. A more limited form was given as early as 1958 by Stewart in Bird. We wish to redefine the Thiele modulus so that this equation reads simply -' Figure 7. that the dimensionless reaction rate is I </> InO + </>C)) ] 1 l/2 (7.382 Fixed-Bed Catalytic Reactors 7. to obtain dp __ ~ de P . Stewart and Lightfoot [4. Using the Thiele modulus defined in Equation 7.57) o z 00 o 1 Equation 7. and substituting the result into Equation 7. Figure 7. define a transformation that makes it a function of e.54. Figure 7. We can accomplish this with a simple rescaling of the Thiele modulus via <!> = Define p(z) = dc/dz to be the desired first derivative.56) e + </>c dimensionless reaction rate Hougen-Watson kinetics . see Bischoff [6]. Equation 7. 338]. p. An asymptotic analysis as presented here is required to find the appropriate scaling. instead of considering P as a function of z.12 illustrates what we have in mind. We make the following variable transformation. p.58) We now use the familiar chain rule of differentiation dp = dp de = dp (z) = dp _ dz de dz de P de P and substitute the mass balance. This equation can be separated f: f: pdp = C [ to yield 1 +e</>c de Performing the integral on the left-hand side and solving for P p gives -=-zf Jo 1 + </>c e de J 1 2 / in which we choose the correct sign to correspond to Figure 7.58. The only nonobvious step in this section is the next one.4 Single Reaction in an Isothermal Particle 383 1 00 Evaluating Equation 7. we have the following two dimensionless groups for this problem (7.55 produces I) z 1 ~ =.12 (</> l~(l + </») <l> So.12. Equation 7. This idea appears to have been discovered independently by three chemical engineers in 1965. Aris [1] and Petersen [19). notice from the mass balance.51.54. Equation 7. If we use our first guess for the Thiele modulus. in summary. Performing the integral on the right-hand side yields The payoff for this analysis is shown in Figures 7.

1 <p = 10 <p = 100 <p = 1000 0.10 corresponds to spherical geometry. 10 100 Figure 7. we replace the boundary condition . is not satisfied. then the boundary condition 0.14: Effectiveness factor versus appropriate Thiele modulus in a slab. Comparing Figure 7. Figure 7.1 0. If the external mass-transfer rate is low. Notice that we approximate the zero-order reaction discontinuity in the slope of IJ versus <P for large <p. Hougen-Watson kinetics.10 we see this trend.1 0.5 External Mass Transfer IJ 2De (<p -lnD + <p)) If the mass-transferrate from the bulk fluid to the exterior of the pellet is not high. which causes the small differences in the effectiveness factor plots. 1 <p = 0.1 1 10 Figure 7. Again we see that as long as we choose an appropriate Thiele modulus.13: Effectiveness factor versus an inappropriate Thiele modulus in a slab. Recall that Figure 7.15 provides a depiction of the concentration profile under low and high external mass-transfer rates. the apparent order of the reaction changes depending on the size of the dimensionless adsorption constant <p.1 1 Figure 7.1 <p = 10 <p = 100 <p = 1000 and. and we are using the slab geometry in this section. For small <p. Hougen~Watson kinetics. 7. the concentrations in the bulk fluid and external catalyst surface are Significantly different. we expect zero-order reaction behavior. and if <P > 2 or <P < 0.4 Single Reaction in an Isothermal Particle 385 1 CAj CA ~ -R I 0 a r I CAt CAs CA <1'= IJ R I ~ I I -R 0 a r - .14 to Figure 7. The largest approximation error occurs near <P = 1.2. we can approximate the effectiveness factor for all values of <p with the first-order reaction. and for large <p. the approximation error is negligible. To obtain a simple model of the external mass transfer. as we have seen in Chapter 5.384 Fixed-Bed Catalvtic Reactors 7.15: Effect of external mass transfer on pellet surface concentration.R I <p = 0. we expect first-order reaction behavior.4.

the concentration profile in the pellet becomes more uniform and tbe dimensionless surface concentration decreases.64) The effectiveness factor can again he derived by integrating the local reaction rate or computing the surface flux.17 shows the effect of the Biot number on the effectiveness factor or total pellet reaction rate. When B = 00. and the curve for <P = LO in Figure 7.4 is the same as the curve for B = 00 in figure 7.c).64 for different values of B. = L =0 Notice we are comparing the pellet's reaction rate to the rate that would be achieved if the pellet reacted at the bulk fluid concentration rather than the pellet exterior concentration as before. If B is small. and the result is l) =.59) 0. The lower concentration leads to lower reaction rates.c) Figure 7.4 Single Reaction in an Isothermal Particle 387 above with a flux boundary condition DA dCA = k m ( cAJ . = B (1. to enhance external mass transfer and make the Biot number large.62) B=oo 2.7 0.5 0.5 Y 2 2. Summarizing.9 0.3 0. Evaluating this constant using the external boundary condition gives -n _ ~ c r - r sinh<Pr sinh3<P + (<P cosh 3<P (sinh3<P)13) IB (7. the log-log The solution to the differential equation satisfying the center boundary condition can be derived as in Section 7 A to produce c('P) = ~ sinh<Pr r (7.5 004 0. for finlte external mass transfer. corresponding to slower external mass transfer.fl(3<P» IB (7. With decreasing B. Therefore one normally designs the reactor and chooses operating conditions.65) in whicb .5 0.CA ) . Figure 7.5 1 1. Notice that the slope of the log-log plot of l) versus <P has a slope of negative two rather than negative one as in the case without external mass-transfer limitations (B ~ (0). we obtain the dimensionless boundary con· dition (7. Figure 7. such as large gas velocities.p J [ Iltanh3<P-l/(3<P) ] 1 + <P (1/ tanh3<P .63) in which C2 is the remainlng unknown constant.64 reduces to Equation 7.2 0.59 by al cAJD A.18 shows this effect in more detail.16 plots Equation 7.25. Equation 7.16: Dimensionless concentration versus radius for djffer~ ent values of the Biot number. dr 1 r=R (7. first-order reaction in a spherical pellet with q.8 0.0 c 0. the dimensionless model and boundary conditions are (7. in which (7.16.1 0 0 0.386 Fixed-8ed Catalytic Reactors 7.61) is the Biot number or dimensionless mass-transfer coefficient.6 in which k m is the external mass-transfer coefficient. If we multiply Equation 7. Figure 7.1 3 d'C dr d'C dr = B (1 .60) ~.

01 0. one can easily sketch the straight line asymptotes shown in Figure 7.1 1 10 100 <I> Figure 7. 0.1 2.. The possible cases are summarized in Table 7.18. . given the value of the Thiele modulus.. Both mechanisms of diffusional resistance. Thiele modulus. first-order reaction in le-06 0.17: Effectiveness factor versus Thiele modulus for different values of the Biot number.01 0.0001 Ie-OS Ie-OS 0.01 1/ 0.4.0 . and whether internal diffusion or external mass transfer or both limit the pellet reaction rate. 3Students who are taking or already have taken the process control course might want to think about what happens to the amplitude of a signal as it passes through a series of first-order processes.001 0. Then. then it comers again and decreases the slope to negative two at <I> = JR. the log-log plot first comers with a slope of negative one at <I> = 1. Biot number B <1 Thiele modulus <!><B B<<I><l 1 < <I> 1<B <!> < 1 l<<p<B B<<I> --- Mechanism controlling pellet reaction rate reaction external mass transfer both external mass transfer and internal diffusion reaction internal diffnsion hoth internal diffusion and external mass transfer I I I Table 7.0001 0.1 0. one can determine the approximate concentration profile.4: The controlling mechanisms for pellet reaction rate given finite rates of internal diffusion and external mass trans- fer.01 IJ 0.388 Fixed-Sed Catalytic Reactors 7.01 0.4 Single Reaction in an Isothermal Particle 389 1 1 0. plot comers with a slope of negative two at <I> = JR. first-order reaction in a spherical pellet. the diffusion within the pellet and the mass transfer from the fluid to the pellet. 3 Given the value of the Biot number.1 1 10 <I> 100 1000 10000 Figure 7. If B is large. show their effect ou pellet reaction rate by changing the slope of the effectiveness factor by negative one.18: Asymptotic behavior of the effectiveness factor versus a spherical pellet. .001 0.1 BJ = 0.5 0.

e. and a first-order reaction instead of the true reaction order kob = kmA = nob = n nob I Everything goes according to plan when we are reaction limited. We express the reaction rate as (7.66) We also know that at steady state. mass-transfer coefficients also have fairly smalllemperamre dependence compared to reaction rates. but the internal diffusion is slow.4 Single Reaction in an Isothermal Particle 391 7. and we would estimate the intrinsic parameters correctly in Equation 7. measure the rate rl as a function of CA. In this case we have Normally.66 r 1 -. Ediff « E rxn . assume the reaction and diffusion are fast compared to the external mass transfer. First assume that both the external mass transfer and internal pellet diffusion are fast compared to the reaction. The observed activation energy from Equation 7. Finally. !k aVn+ . varies as the square root of the intrinsic rate constant. Let Erxn be the intrinsic activation energy of the reaction. Then 1) = 1.70) We call the values of k and n the intrinsic rate constant and reaction order to distinguish them from what we may estimate from data.B. Assume the following nth-order reaction takes place in a catalyst particle A . = (n + 0/2 (7. Normally the temperature dependence of the diffusivity is much smaller than the temperature dependence of the reaction. We even get an incorrect reaction order: a first-order reaction appears half-order. Also consider what happens if we vary the temperature and try to determine the reaction's activation energy. diffusion and external mass transfer may all manifest themselves in the measured rate. and then find the values of the parameters k and n that best fit the measurements.67) So we see two problems.k ob C(n+l)J2 (7. k. I) = 1/<1>. be represented also in Arrhenius form. independent of the true reaction's activation energy. Diffusion limited.390 Fixed-Bed Catalytic Reactors 7. Next assume that the external mass transfer and reaction are fast.66 = kmACAb (7. so we would estimate an activation energy that is one-half the intrinsic value.67 gives rl We now study what happens to our experiment under different ratelimiting steps. .6 Observed versus Intrinsic Kinetic Parameters We often need to determine a reaction order and rate constant for some catalytic reaction of interest. Here we show only that one should exercise caution with this estimation if we are measuring the rates with a solid catalyst. kGb. The effects of reaction. The typical experiment is to change the value of CA in the bulk fluid.69 is Eob = Ediff + Erxn 2 so both activation energies show up in our estimated activation energy.69) (7.71) If we vary CAb and measure rl. Mass transfer limited. with Ediff the activation energy of the diffusion coefficient. so the observed activation energy would be almost zero.68) As kob nob =! /z 1 D. The diffusion has affected the measured rate of the reaction and disguised the rate constant. Let the temperature dependence of the diffusivity. the rate is equal to the flux of A into the catalyst particle (7. We explain this procedure in much more detail in Chapter 9. a second-order reaction appears first-order. The rate constant we estimat. and using the definition of Thiele modulus and Equation 7. D A. and so on. Then we have CAb» CAs and Equation 7. we would find the mass transfer coefficient instead of the rate constant. Reaction limited.4.

)] Ts + tlHRDA (eA . Substituting Equation 7. we can integrate this equation twice to give the relationship between temperature and A concentration T .c) ) dC) 1 + /l(l c) ~~=o c=1 1'=3 1'=0 (7. and estimate the parameters D A and kmA simultaneously with k and n_ We develop techniques in Chapter 9 to handle this more complex estimation problem.5 Nonisothermal Particle Considerations 393 The moral to this story is that mass transfer and diffusion resistances disguise the reaction kinetics_ We can solve this problem in two ways_ First. (1 ( }'/3(l-C») ] l(y.75) = -tl.2 ~~ If"3 1/2 (7.73) We then substitute the rate constant into the mass balance.cAl (7. and we assume fast external mass transfer.HRDA n2 CA ~ v k which relates the conversion of the reactant to the rise (or fall) in temperature.:t.4A on the Weisz-Hicks problem. in which we have RA = -r and Equations 7.74. we can express the rate constant as k(T) }'/l(l . and <I> is the usual Thiele modulus. we can arrange the experiment so that mass transfer and diffusion are fast and do not affect the estimates of the kinetic parameters_ We can accomplish this by making the catalyst particles small (<I> small) and using a flow reactor with high fluid flowrates to ensure a high mass-transfer coefficient. If this experimental design is impracticalor too expensive.Ts Equation 7.CAs) / k It now simplifies matters to define three dimensionless variables }'= E = -tlH!!D ACAs kTs ' - 7. Consider the singlereaction case.392 Fixed-Bed Catalytic Reactors 7. Because we have assumed constant properties.74) kV'T = LlHRr We can eliminate the reaction term between the mass and energy balances to produce n2T _ v - tl.72 into the rate constant expression gives k(T) = ks exp [!i. we can compute the effectiveness factor for the nonisothermal pellet using the usual relationship ry = . in which T. We wish to explore the effects of this temperature rise on the catalyst performance.14 and 7. Given an exotherntic reaction. that the appropriate Thiele modulus for this problem is kCf) = ksexp[ -E(} . if the particle's thermal conductivity is not large compared to the rate of heat release due to chentical reaction.74 is sometimes called the Weisz-Hicks problem in honor of Weisz and Hicks's ontstanding paper in which they computed accurate numerical solutions to this problem [22J.72) We now consider a first-order reaction and assume the rate constant has an ArrheniUS form. however. the temperature rises inside the particle. for example. Given the solution to Equation 7. /l is a dimensionless heat of reaction. If we perform the same asymptotic analysis of Section 7. and assume a spherical particle to obtain the final dimensionless model 1'2 d1' Je. (7. (1'2 d1' = <I>'c exp (}'/l(l. we can alternatively model the effects of the mass transfer and diffusion.HRDA k • (CAs .5 Nonisothermal Particle Considerations We now consider situations in which the catalyst particle is not isothermal. We have already written the general mass and energy balances for the catalyst particle in Section 7.76) . we find.15 reduce to DAV 2 CA =r in which}. With these variables.c) ] = ksexp [ 1 +/l(l-c) (7../l)= 2Jo cexp l+/l(l-c) dc .3.JJ [. (1 Ts a P Ts . is a dimenSionless activation energy. Again we use the tilde to indicate we will find a better Thiele modulus SUbsequently... is the pellet exterior temperature.

if there is a large external heat transfer resistance. . the pellet's interior temperature is significantly higher than the exterior temperature Ts. . and we wish to compute these profiles accurately.76 because the large <P asymptotes are the same for all values of y and f3.19 shows the effectiveness factor versus Thiele modulus for activation energy y = 30 and a variety of heats of reaction. but the scope of these is limited. The next example involves five species. The first interesting feature of Figure 7. The three temperature and concentration profiles correspond to an ignited steady state (C). The normalizing integrall(y.6 Multiple Reactions 395 1000~----~~------------------------------ .. For realistic values of the catalyst thermal condUctivity. The effectiveness factor. We show in Figures 7. As we showed in Chapter 6.74 for this case. however.001 0. it is possible for the effectiveness factor to exceed unity in a nOnisothermal pellet. the pellet can often be considered isothermal and the energy halance can be neglected [171.. Consider the case 4> = 0.1L-~~~~~~~L-~~~L-~~-u~~~~ 0. If we consider what is happening within the pellet. We use the collocation method. f3) can be expressed as a sum of exponential integrals [2] or evaluated by quadrature. A second striking feature of the nonisothermal pellet is that multiple steady states are possible. and an unstable intermediate steady state (E). 511. Figure 7. ATis provides further discussion of these cases and shows that many steadystate solutions are possible in some cases [3.19: Effectiveness factor versus normalized Thiele modulus for a first~order reaction in a non Isothermal spherical peliet. 7. we consider the case of multiple reactions.19. which we see in Figure 7. Figure 7. Boundaryvalue problems (BVPs) are generally much more difficult to solve than initial-value problems (lVPs). we find that even numerical solution of some of these problems is challenging for two reasons.01. The rate constant inside the pellet is therefore much larger than the value at the exterior. Indeed.. We focus on numerical computation as a general method for these problems. p. Note that <P is well chosen in Equation 7. f3. whicb is described in more detail in Appendix A. however. f3 = 0. whether we achieve the ignited or extinguished steady state in the pellet depends on how the reactor is started. and ATis provides a comprehensive list [2].0001 0. It is not unusual for species concentrations to change by 10 orders of magnitude within the pellet for realistic reaction and diffusion rates. Multiple steady-state solutions in the particle may still occur in practice.19 is that the effectiveness factor is greater than unity for some values of the parameters. k s • Even though the concentration is lower inside the pellet because A is con- .20 and 7. Because the effectiveness factor compares the actual rate in the pellet to the rate at the surface conditions.6 Multiple Reactions As the next step up in complexity. an extinguished steady state (A). this effect may not be too surprising.19. For the highiy exothermic case.has three possible values for this case. First.394 Fixed-Bed Catalytic Reactors 7. Second. Some analytical solutions are available for simple cases with multiple reactions. the increase in the rate constant is more significant and the product r = keA is larger inside the pellet.4 and y = 30 shown in Figure 7. 100 y = 30 '7 10 1 0. Notice that feature is more pronounced as we increase the exothermic heat of reaction.01 0. A detailed description of numerical methods for this problem is out of place here.1 1 10 surned. steep concentration profiles often occur for realistic parameter values.21 the solution to Equation 7. we are solving boundary-value problems because the boundary conditions are provided at the center and exterior surface of the pellet.

2 0.25 0.5 [16].ining 2% CO. O.02.4 0.4 q.1 0.79) (7.35 0.5 1 with the boundary conditions 1. = 0. unstable middle (8).45 0.77) (7.15 0.t<l.4 0. and upper (C) steady states.5 3 de _ 1 =0 dr (7. Solution We solve the steady-state mass balances for the three reactant species. Example 7.83) Figure 7.80) isothermal spherical pellet: lower (A). unstable middle (8).396 Fixed-Bed Catalytic Reactors 7.20: Dimensionless concentration versus radius far the non- (7. = 0.01 Consider the oxidation of CO and a representative volatile organic such as propylene in a automobile catalytic converter containing spherical catalyst pellets with particie radius 0.5 2 CO + 202 9 C3H6 + 202 - 1 C02 3C02 + 3H20 (7.'i) = ~R.2 0 0 0. dr (r2tt.01 y c The rate constants and the adsorption constants are assumed to have Arrhenius form. Calculate the steady-state pellet concentration profiles of all reactants and products.4 q.22 and 7. The mass-transfer coefficients are taken from DeAcetis and Thodos [9J. The reactions of interest are c 1.175 cm.82) (7.05 A = 30 f3 = 0.C)HGl. The reactant concentration profiles are shown in Figures 7. and both diffusion and external mass-transfer limitations.78) 1 2.5 y = 30 f3 = 0. B D J 1 !!.5 2 2. The pellet may be assumed to be isothermal. The parameter values are given in Table 7.0 atm pressure and 550 K con. [l6J Figure 7.2 . and upper (C) steady states. The model is solved using the collocation method. Dimensionless temperature versus radius for the non~ D j de = k mj ( ejf ~ Cj ) d: j isothermal spherical pellet: lower (A). dr } y=R (7.2: Catalytic converter 0. = {CO.6 B 0.S 0.05% (500 ppm) C3H6.3 1'0. .81) a 0 0. 1 A two reactions with Hougen-Watson kinetics. 3% 02 and 0.23.6 Multiple Reactions 397 1.21.5 3 with rate expressions given by Oh et al.8 C 0. The particle is surrounded by a fluid at 1.

CO (+).0% 0.12 0.).05 % em K K K K Kcoo KC3H{iO Dca D02 DCj H6 - kmco k mo2 kmc3H6 i Units i mol/em 3 .0469.::. however.02 I E CCOj CO?! co C3 H6 I I cC:.099 x 10 6 2.22: Concentration profiles of reactants! fluid concentration of 02 (x).16 r (em) Figure 7.0487 0.14 0. we can integrate these .16 0.04 0.14 0.398 Fixed·Bed Catalytic Reactors 7.579 x 108 0. Obvionsly the consumption rate is large compared to the diffusion rate for these species.175 13. \7 2 CC02 DH2ov2CH20 E w le-11 1e-12 le-13 Ie·14 Ie-IS 0 0. The product concentrations could simply be calculated by solving their mass balances along with those of the reactants. C3H6 (. Notice that 02 is in excess and both CO and C3H6 reach very low values within the pellet The log scale in Figure 7.08 0.109 -409 19l 2.04 0. Because we have only two reactions.90 02 + T R El E2 Eeo EC 3 H6 '(5 '-J S 4e·07 u 3e·07 l2e-07 iIe-07 f0 0 0.0487 3.5: KinetIc and 0. s mOl/em 3 • s em3 /mol em3/mol cm 2 /s em 2/s em2!s em!s cm/s ! em!s I Table: 7.06 0. 0.18 mass~transfer parameters for the catalytic con- I verter example.23 shows that the concentrations of these reactants change by seven orders of magnitude.108 15.07 3. but not dramatic.02 0. Ie-OS Ie-06 Ie-O? I e-08 M 02 Je-09 0 co S Ie-lO v .0% 3. con~ Because the diffusivHies are assumed constant.12 0.6 Multiple Reactions 399 7e·07 6e·07 5e·07 M Parameter x P Value 1.18 2 -Reo = -RC3 H6 = -RC{h = _· RH 2() _. If we take the following mass balances Dco\7 2 cco DC3H£ \7 CC3 Hti Dco.H&f 550 0.1 0.08 0.1 r (em) 0. 0. = 'Yl = r2 = -Yj 3r z = ·-3 r 2 and form linear combinations to eliminate the reaction-rate terms for the two products.90 4.07 x 1019 IA7 x 10 21 8. The external mass-transfer effect is noticeable. the concentrations of the products are computable from the stoichiometry and the mass balances.06 0.013 X Units N/m 2 K 10 5 Parameter kJO k 20 Value 7. we obtain Figure 7.23: Concentration profiles of reactants (log scale): fluid centration of 02 (X)j CO (tL C3H6 (*).

In the fixed-bed reactor.IS 0. __ D deeo _ 3D co dr CJH.22 and 7. we track the molar flows of all species.1 Coupling the Catalyst and Fluid 3.24.1 r (em) 0.400 Fixed-Bed Catalytic Reactors 7. .!l. the temperature and the pressure.-k-.[ 3Dc3H6 .14 0.86) the length of the reactor. We have constructed a detailed understanding of all these events.7 Fixed-Bed Reactor Design 2.[deeo + 3DclH6-d. Steady state.1 1 I I I I I 0 0. r) to obtain for the products D D 5e-07 co. we now are prepared to pack a tube with a bed of these particles and solve the fixed-bed reactor design problem.02 0. 7. dr dr c. Particles are small compared to (7. no radial profiles.OS The right-hand sides are available from the solution of the material balances of the reactants.04 " O. dr dcco. We can no longer neglect the pressure drop in the tube because of the catalyst bed. we keep track of two phases..7 Fixed-Sed Reactor Design 401 once on (0. the well-known Ergun equation [10]. We see that C02 is the main product. Note the products flow out of the pellet.S4) We make the following assumptions: 1. 13 ~ § " 2e-07 Ie-O? !- H2 O I Substituting in the relationships for the products gives Cco. Neglect axial diffusion in the bed. Plug flow in the bed.7.:£CC 3Ho 4e-07 !3e-07 CO2 H2 0 dCH20 __ 3D The exterior boundary condition can be rearranged to give Cj -Cjf = - ~J ~. Plotting these results for the products gives Figure 7.24: Concentration profiles of the products.16 Figure 7.] Dcodr meoz r 1 dCC3Ho] = cH20f . The reaction-diffusion processes take place in the solid-phase catalyst particles. 7.06 O. 0 0.12 0.d mHzO r 0 .-k-. CH20 = 1 dCC3H6 ccod . dr dCC3H6 . Given our detailed understanding of the behavior of a single catalyst particle.8S) (7. 4. fluid concentration of CO2 (xl. which are flowing into the pellet.23. and now we assemble them together. In the flUid phase. unlike the reactants shown in Figures 7. Fluid. Therefore. The two phases communicate to each other by exchanging mass and energy at the catalyst particle exterior surfaces. Uniform catalyst pellet exterior. H20 (+l. The flUid-phase streams through the bed and transports the reactants and products through the reactor. We use an empirical correlation to descrihe the pressure drop in a packed tube. we have the follOwing differential equatIOns for the fluid phase (7.

:. Equations 7.. We next examine several packed-bed reactor problems that can be solved without solving this full set of equations.3: First-order. z=O z=O z=O (7. we equate the enthalpy change on reaction experienced by the fluid phase to the enthalpy change on reaction taking place inside the particles. ' Vp ] dr r..:.Vp k dT I dr (7...7. we equate the production rate Rj experienced by the fluid phase to the production rate inside the particles. The feed to the reactor is pure A at 1. This infonnation can be presented in a number of equivalent ways. we track the concentrations of all species and the temperature. and summarizes the coupling relations between the two phases.EE.Cj r=O r=R r=O r~R (7.3 cm pellets are to he used. Pp. Example 7. We can easily measure the density of the pellet. and the zero slope conditions at the pellet center de dr de) Dj dr 1EB = PBipp _J =0 = kjm (e} _ ) . We neglect any pressure effect inside the catalyst particle.90) (7.94) (7.97) roR Catalyst particle.-' rate j / vol =- Njf.91) Notice we require the bed porosity to convert from the rate per volume of particle to the rate per volume of reactor.R vol cat I vol .402 Fixed-Bed Catalvtic Reactors 7. Assume the reactor operates isothermally at 450 K and that external mass-transfer limitations are negligible. and the density of the bed. These ex- .88) (7.92) (7. which requires numerical solution of the full set of equations.5 atm at a rate of 12 mol/so The 0.87) (7. The bed porosity or void fraction. Finally. is defined as the volume of voids per volume of reactor.95) dT = 0 dr -dT k-=kT(T-T) dr Figure 7. From the definition of bed porOSity.~_ _~.o rate j / vol cat Sp D (l .93) (7. = Tf' P = Pf. Finally. Inside the catalyst particle.. Analogously. we present Example 7. which leads to a bed density PH = 0. The volume of catalyst per volume of reactor is therefore 1 .89) ' . PH._". We have the following differential equations for the catalyst particle rate heat / vol rate heat! vol cat (7.1 to find the fixed-bed reactor volume and the catalyst mass needed to convert 97% of A.7 Fixed-Bed Reactor Design 403 The fluid-phase boundary conditions are provided by the known feed conditions at the tube entrance Nj T ~ pressions are given below Rj '-.84-7.25 shows the particles and lluid.96) Sp vol cat I vol .97 provide the full packed-bed reactor model given our assumptions.EE) - dej - I (7. which is where the reaction takes place. isofhennal fixed-bed reactor Use the rate data presented in Example 7. we have the relation or if we solve for the volume fraction of catalyst The boundary conditions are provided by the mass-transfer and heattransfer rates at the pellet exterior surface.6 g!cm3 . EH. Coupling equations..

the equations show the coupling between the catalyst particle balances and the overall re' actor balances. 0 Example 704: Mass-transfer limitations in a fixed-bed reactor Solution We solve the fixed·bed design equation dNA dV Reconsider Example 7. in which CA is the A concentration in the fluid.kdr V P I T""R VB ~ . For the first-order. 6 3 (0.85 0. the Thiele modulus is independent of A concentration.3 given the following two values of the mass· transfer coefficient EBll)kcA kml ~ km2 ~ ~ RA = -(1- 0. P NA RT NAJ SubStituting these values into the material balance. finding the size of a fixed·bed reactor is no more difficult than finding the size of a plug·flow reactor.17 indicates that we expect a Significant reduc· tion in the effectiveness factor due to mass·transfer resistance in the first case.03NAf NAJ dNA NA -' ~ 789 kg ~ tlHR t r lp i ~ Sp . we express the concentration of A in terms of molar flows for an ideal·gas mixture Inspection of Figure 7.3~] ~ L~3 [1 5.165 1)2 = 0.6 ) (82.03NAj. isothermal reaction.. Equation 4.1) ~ 1 1 (0. and little effect in the second case.~8] = 0.404 Fixed·Bed Catalytic Reactors 7.( 0.007~~2/S = 1.75.7 Fixed·Bed Reactor Design 405 Mass The total molar flow is constant due to the reaction stoichiometry so NA + NB ~ NAJ and we have C.129 In Chapter 4.397 . and is therefore independent of axial position in the bed Solution First we calculate the Biot numbers from Equation 7.6)(1. We see from this example that if the Thiele modulus and effectiveness factors are constant.03) ~ 132 x 10 cm and We Figure 7.65 indeed shows 171 = 0.5) In(0.25: Fixed-bed reactor volume element containing fluid and ~ PBVR ~ l~gO (132 X 106 ) catalyst particles.07)(0.429)(2.06)(450)(12) .007) ~ [ta~3<l> .dT .61 and obtain B ~ (0. Evaluating the effectiveness factors with Equation 7.07 cmls L4 cmls between the limits NAf and 0.001) B 2 <l> = ~~ ~A = . rearranging and integrating over the volume gives Energy VR ~ -(1-eBl ( RTNAJ) I7kP JO.33 ~ 0.- .4 = .l) ~ 20 (0.93 cm The effectiveness factor is also therefore a constant I) = ~ 0 f~261'- = (lA){O.

Equation 7.25NAf· We again express the concentration of A in terms of the molar flows As a final exercise.101 shows that <l> varies with the molar flow.397 Therefore.101) J<.34> 1[1 1] (7.24. The desired conversion of A is 75%.45 em. The pellet density is Pp = 0.42. each with molar flowrate of 10 mol/s.429) (789) = 2051 kg 0.101 and 7.98) Summarizing so far. the first mass-transfer coefficient is so small that more than twice as much catalyst is required to achieve the desired conversion compared to the case without mass· transfer limitations.98. we can make the large <l> approximation 1)=- 1 <I> (7. we solve one differential equation. and Equations 7. isothermal fixed-bed reactor Estimate the mass of catalyst required in an isothermal fixed-bed reactor for the second-order.429) (789) = 852 kg As we can see. to compute NA versus VR. Equation 7. You may assume the fluid and pellet surface concentrations are equal. in which we use Equation 7. The second mass-transfer coefficient is large enough that only 8% more catalyst is required. We can then easily calculate the required catalyst mass from the solution of the previous example without masstransfer limitations.5: Second-order.429 from the previous example with no mass-transfer resistance. We march in VR until NA = 0.99) V' = (0.165 0. tanh 3<1> . which means <I> and 11 vary along the length of the reactor as NA decreases.47 <I> 1 = !Ie [(n + l)kC~-] 1/2 =!Ie [ <I> for nth-order rea@ons given in (n.49 to 3. A r Next we use the definition of Equation 7. As shown in Figure 7. the total pressure is 4.99 for CA.102) d:N. The catalyst is a spherical pellet with a radius of 0.103) . between the limits NAJ and O.68 g/cm3 and the bed density is PB = 0. we can approximate the effectiveness factor for the second-order reaction using the analytical result for the first-order reaction.25 x 1O S cm 3 /mol s The gas feed consists of A and an inert. (7. The required reactor volume and mass of catalyst are: VR = 361 t. Equation 7.+ l)k 2De 3 2De 3 (~ NA ) RT 2NAJ n-l] 1/2 (7.25N A f' The solution to the differential equation is shown in Figure 7.A = RA = -(1 . 0 Example 7. the total molar flow is constant and we know its value at the entrance to the reactor = ( 0. given that <l> ranges from 6. So for this particular example. and the new values of the effectiveness factors VRl R2 As in the previous example.17 ( 2~:f ) 1/2 (7. We are asked to estimate the catalyst mass needed to achieve a conversion of A equal to 75%.102 for <I> and I). heterogeneous reaction. B = kd k = 2. <I> decreases from 6. Solution We solve the fixed-bed design equation dV NA(O) = NAJ For the second-order reaction.49 to 3.60 g/cm3 The effective diffusivity of A is em' /s and may be assumed constant.£B)l)kc~ (7. l] = ij.0 atm and the temperature is 550 K.100) Substituting in the parameter values gives <I> = 9.406 Fixed-Bed Catalytic Reactors 7.7 Fixed-Bed Reactor Design 407 which we can compare to 11 = 0.

One advantage is purely practical.98 and rearrange we obtain dNIl. zero C02 and the remainder inerts.099 x 106 exp(409/T) cm3 /mol VR = (l-cBlJK(P/RT)3/2 (7.1 cm. ma1/s cm 3 pe 11 et The constants are provided below k = 8.6: Hougen-Watson kinetics in a fixed-bed reactor The following reaction converting CO to C02 takes place in a catalytic.103.408 Fixed-Bed Catalytic Reactors 7. one might question the value of this closedin which T is in Kelvin. on the other hand. We may not have a computer available. The feed to the reactor consists of 2 mol% CO. . If we substitute this approximation for I). or halve the rate constant. As sho"m in Figure 7. The reactor size and the percent change in reactor size are 1'. provides the solution's dependence on all parameters. The other important advantage is insight. isothermal reaction in a fixed-bed reactor.7% Given that we have a result valid for all <j> that reqoires solviug only a single differential equation.26 the approximation error is small.104) Dco = 0. for example. Find the reactor volume required to achieve 95% conversion of the CO. before using Equation 7. Instructors are usually thinking about in-class examination problems at this juncture.0 atm (7. and Equation 7.(RI3)~3/DA(2NAf)3/2 A r = 1 + Kcco kccoco. Solution Given the reaction stoichiometry and the excess of 02. and approximate the reaction as pseudo-first order .73 K = Separating and integrating this differential equation gives 4 [(1XA)-112 X 10 12 exp( -13.0487 cm 2 /s Large <1> approximation The results for the large <1> approximation also are shown in Figure 7.7 Fixed-Bed Reactor Design 409 10 9 8 Vi "- 7 6 C g :2: -< 5 4 3 2 1) I) =¢ 1 [1 1 ] tanh 3<1> .104_ 0 Example 7. we can neglect the change in co.9 that we are slightly overestimating the value of I] using Equation 7. Rememb~to check that the Thiele modulus is large for the entire tube length. [16J. Notice from Figure 7. so we underestimate the required reactor volume. Equation 7_104. It is not readily apparent from the differential equation what would happen to the reactor size if we double the pellet size. The catalyst pellet radius is 0.105) The following rate expression and parameters are adapted from a different model given by Oh et a1.26: Molar flow of A versus reactor volume for second~order.--7.26. The rate expression is assumed to be of the Hougen-Watson form =- 1 <1> 0 50 100 150 200 VR (Ll 250 300 350 400 Figure 7. to obtain a closed-form solution. however. 500/T) cm 3 /mol s -ljN A{(R/3lJ3/DA 8.100 into Equation 7. fixed-bed reactor operating isothermally at 838 K and l. 10 mol% 02. _ -O-CBlJK(P/RT)3!2N312 dV .3<1> form solution.

Values indicate i} = 1 lIP is a good approximation for entire reactOr.28: Dimensionless equilibrium constant and Thiele modulus versus reactor volume.8. entrance exit = q. Recall that the volumetric flowrate varies in this reactor so conversion is based on molar flow. S 8e-06 ~ -0. dNco dV = M u .. . I I I C0 2 = Neal - Nco 01-" 0 CO I N = NOd + 112(Nco + Ncof) 50 100 150 200 2S0 The concentrations follow from the molar flows assuming an ideal-gas mixture 350 Figure 7. From the reaction stoichiometry.. Reactor volume VR = 233 l 150 a is required to achieve 95% conversion of the CO.= 79.27: Molar concentrations versus reactor volume. a 50 100 VR (l) 30 25 20 <p = 32. q.:---. we evaluate <p = KcoccQ.14 that IJ llq.-... we compute the Thi.4. Figure 7_28 shows how q. Figure 7..27.. which justifies the pseudo-first-order rate expression.74 q.Nco!) Ncoz . is an excellent approximation for this reactor. 35 To decide how to approximate the effectiveness factor shmvn in figure 7. Substituting this equation for r) into the mass balance and solving the differential equation produces the results shown in Figure 7. = 326.2e-OS Ie-OS ~ which is of the form analyzed in Section 7.~ in which Ceo is the fluid CO concentration.410 Fixed-Bed Catalvtic Reactors 7. 150 15 10 5 200 250 100 50 0 <p It is clear from these values and Figure 7.4e-05 molls cm 3 pellet 1.ele modulus given in Equation 7. at the entrance and exit of the fixedbed reactor. and <p vary o with position in the reactor.4. We can write the mass balance for the molar flow of CO.tB)lJr(cco) E " u 6e-06 4e-06 2e-06 . we can express the remaining molar flows in terms of Nco No z = NOd + 112 (Nco . With <p evaluated..0 <p = 1.14..7 Fixed-Bed Reactor Design 411 inCO k1cco r= 1 + Kcco k' = kCod 1. not molar concentration.58 and obtain 300 250 200 /. The concentration of 02 is nearly constant.

nonconstant thermochemical properties. the pellet exterior CO concentration has dropped by two orders of magnitude.412 Fixed-Bed Catalytic Reactors 7.0 atm to 1. to ensure convergence in the fluid profiles. nonisothermal fixed-hed reactor Evaluate the performance of the catalytic converter in converting CO and propylene.2 (the differences are caused by the different feed pressures). We see that as the reactor heats up. the reaction rates drop.31 and 7. we compute the numerical solution to a problem of this type. The pellet concentration profiles are solved with the collocation approach. 0 Table 7. . hoth diffusion and external mass-transfer limitations.4 x 10 0. The good news. essentially zero reaction is occurring.3 cal/(cm2 Ks) cal/(mol CO K) -67. It can be numerically challenging to calculate rapid changes and steep profiles inside the pellet.S. In the previons examples.7 Fixed-Bed Reactor Design 413 Parameter Pf Tj R. At 890 cm3. The reaction chemistry and pellet masstransfer parameters are given in Table 7.HR2 C p Ilf Ph Pp Units N/m2 K em cmls K 325 5. The feed conditions and heat-transfer parameters are given in Tahle 7. The fluid concentrations are shown in Figure 7. As the reactions go to completion and the heat losses cool the reactor. Example 7. and we can calculate an accurate overall pellet reaction rate.6% of the CO and propylene.HRl t.55 atm must be large enough to account for the remaining pressure drops in the tall pipe and muffler. we have exploited the idea of an effectiveness factor to rednce fixed-bed reactor models to the same form as ping-flow reactor models. The reason is that when steep profiles are present. temperature and pressure. For these cases.32.028 x 10. the heat losses then reduce the temperature to less than 500 K by the exit. the CO hegins to diffuse back into the pellet.02 x 10 5 550 5. The reactor temperature and pressure are shown in Figure 7. A bed volume of 1098 cm 3 is required to convert the CO and C3H6. essentially all of the reaction occurs in a thin shell near the pellet exterior. The feed enters at 550 K.2 g/(cm s) 0. The pellet and fluid concentrations are coupled through the mass-transfer boundary condition. marked by (j) in Figure 7. but this approach is also limited and can take us only so far. and nonisothermal behavior in the pellet and the fluid.0 75 Solution The fluid balances govern the change in the fluid concentrations.6.51 g/cm 3 0. By 490 cm 3 in the reactor. Finally. Notice the catalytic converter exit pressure of 1.30. however.5 X 10. the pellet CO concentration profile at several reactor positions is displayed. The feed profile. It is always a good idea to vary the numerical approximation in the pellet profile.23 of Example 7. is simllar to the one shown in Figure 7. The pressure drops from the feed valne of 2. and the reactor experiences about a 130 K temperature rise while the reaction essentially completes.7: Multiple-reaction. Determine the amount of catalyst required to convert 99. uf Ta UO t. by changing the number of collocation points.32. the profiles become much flatter near the exit of the reactor.68 g/cm 3 Value 2. the reaction rates become large and the CO is rapidly converted inside the pellet. We use the collocation method to solve the next example. we have no alternative but to solve numerically for the temperature and species concentrations profiles in both the pellet and the hed. pressure and volumetric f1owrate. Figure 7.25 cal/(g K) 0. We can calculate accurately down to concentrations on the order of 10"15 as shown in Figure 7.32. two reactions "ith Hougen-Watson kinetics. In the general case. we must contend "'lth multiple reactions that are not first order. is that accurate pellet profiles are generally not required for an accurate calculation of the overall pellet reaction rate. and the profile inside the pellet has become very steep.63 x 103 3 cal/(mol C3H6 K) -460.29.29 also wpws that oxygen is in slight excess.55 atm at the exit. and by that pOint. and nonconstant fluid temperature. As a final example. which involves five species. This approach is useful and solves several important cases.6: Feed flowrate and heat-transfer parameters for the fixedbed catalytic converter. In Figures 7.

30: Fluid temperature and pressure versus reactor volume.414 Fixed-Bed Catalvtic Reactors 7. S le-1O u 0 (!J § le-12 1e-14 1e-16 ® .16 r (em) Figure 7. . .32: Pellet CO profiles at several reactor positions.08 0.I:f 0 ® o 0. .7 Fixed-Bed Reactor Design 415 Ie-OS M le-06 le-07 Ie-OS (j)@@ '- S u ® ® § Ceo .06 0.1 0. Figure 7. .31: Reactor positions for pellet profiles.. .12 0.02 0. . . .29: Fluid molar concentrations versus reactor volume.14 0. Figure 7. 0 Je-09 le-lO 04090 490 VR (em3 ) 890 990 1098 I e-ll 0 200 400 800 1000 1200 le-06 1e-08 M Figure 7..04 0.:::. .

8 Summary This chapter treated the fixed-bed reactor.84-7. which measures the rate of production divided by the rate of diffusion of some component. and adsorption. These included: • Single reaction in an isothermal pellet. reaction and desorption on the catalyst surface. and determined the temperature and conceutration prol1les within the pellet. Hougen-Watson rate expressions. and described bulk and Knudsen diffusion phenomena.97. irreversible reaction in a spherical particle. We started with a general overview of the transport and reaction events that take place in the fixed-bed reactor: transport by convection in the fluid. Multiple steady-state solutions in the particle may occur in practice. but for realistic values of the constant for the BET isotherm concentration of species j concentration of species j at the catalyst surface dimensionless pellet concentration total number of active surface sites c em DAB OJ DjK binary diffusion coefficient effective diffusion coeffiCient for species j Knudsen diffusion coefficient for species j Djm Dr Ediff diffusion coefficient for species j in the mixture pellet diameter activation energy for diffusion experimental activation energy intrinsic activation energy for the reaction heat of reaction i rate of transport of species j into a pellet fobs Erxn MiRi Ii 10 modified Bessel function of the first kind. for tbe single-reaction cases.8 Summary and Notation 417 7. irreversible reaction. zero order . This motivated the definition of the pellet effectiveness factor. Vp I Sp reactor cross-sectional area Biot number for external mass transfer * Particle with significant external mass-transfer resistance. a tubular reactor packed with catalyst pellets. * First-order. For the single-reaCtion. catalyst tbermal conductivity. we solved numerically the full problem given in Equations 7. We solved the reaction-diffusion problem in the pellet coupled to the mass and energy balances for the fluid. We summarized the transport properties of the catalyst particles. We next presented tbe general mass and energy balances for the catalyst particle. however. These problems are challenging because of the Sleep pellet prol'illts that are possible. Multiple steady-state solutions also are possible for this problem. A complete analysis of the first-order reaction in a sphere suggested a general approach to calculate the production rate in a pellet in terms of the rate evaluated at the pellet exterior surface conditions. we cannot use the simple Thiele modulus and effectiveness factor approach. For complex mechartisms. for complex reactions involving many species. the Thiele modulus. and must solve numerically the complete reaction-diffusion problem. we performed dimensional analysis and found a dimensionless number. nonisothermal problem. For simple reaction mechartisms. We showed the effectiveness factor can be greater than unity for this case. • Multiple reactions.416 Fixed-Bed Catalytic Reactors 7. • Single reaction in a nonisothermal pellet. we were still able to use the effectiveness factor approach to solve the fixed-bed reactor problem. the pellet often can be conSidered isothermal and the energy balance can be neglected. This case was further divided into a number of special cases. Notation a A. if there is a large external heat-transfer resistance. Next we solved a series of reaction-diffusion problems in a single catalyst particle. which is a function of the Thiele modulus. * nth order. In order to simplify the model. and we used the Ergun equation to calculate the pressure in the flUid. we showed several ways to couple the mass and energy balances over the flUid flOv. Finally. B c Cj Cjs characteristic pellet length. diffusion inside the catalyst pores. * * Reaction in a semi-infinite slab and cylindrical particle. we solved the socalled Weisz-Hicks problem. we assumed an effective diffusivity could be used to describe diffusion in the catalyst particles.1ng through a fixed-bed reactor to the balances within the pellet.

Assume that the pellet height. Ic. DA. .. are much greater than the width.418 Fixed-Bed Catalytic Reactors 7. P PB Pp p.9 Exercises Exercise 7.AB fixed-bed porosity or void fraction effectiveness factor mean free path bulk fluid density stoichiometric number for the jtb species in the itb reaction integral of a diffusing species over a bounding surface bulk fluid density reactor bed density overall catalyst pellet density catalyst solid-phase density hard sphere collision radius tortuosity factor Thiele modulus dimensionless function of temperature and the intermolecular potential field for one molecule of A and one molecule of B 7.33. w. C4S_ Species A has an effective diffusivity. first order effective thennal conductivity of the pellet mass-transfer coefficient for species j nth-order rcaeUan rate constant pore length molecular weight of species j number of reactions in the reaction network total molar flow. irreversible reaction A J. uo Vrn V modified Bessel function of the first kind. and length. kjm kn L MJ fir N NJ p Q r rip y R R RJ RJj R jp Rjs 59 5p T Tr T. 2:J Nj molar flow of species j pressure volumetric flo\tYTate radial coordinate in catalyst particle average pore radius rate of reaction i per unit reactor volume observed (or experimental) rate of reaction in the pellet total rate of reaction i per unit catalyst volume dimensionless radial coordinate spherical pellet radius gas constant production fate of species j production rate of speCies j at bulk fluid conditions total production fate of species j per unit catalyst volume production rate of speCies j at the pellet surface conditions BET area per gram of catalyst external surface area of the catalyst pellet temperature bulk fluid temperature pellet surface temperature overall heat-transfer coefficient volume of gas adsorbed in the BET isotherm volume of gas corresponding to an adsorbed monolayer reactor volume coordinate pellet void volume per gram of catalyst volume of the catalyst pellet reactor volumetotal mass of catalyst in the reactor mole fraction of species J position coordinate in a slab porosity of the catalyst pellet Vi} 1. h w y Figure 733: Catalyst pellet with slab geometry.1: Isothennal slab with first-order kinetics A first-order. a T <I> OD. The catalyst exterior is assumed to be at constant species A concentration. h.9 Exercises 419 I. 2B takes place in a catalyst pellet with tbe slab geometry depicted in Figure 7. in the peUet and the catalyst is isothermal. I.

this becomes Exercise 7. (f) Plot 1]{<l». and species B has an effective diffusivity.10.2: Pellet profile of product and reactant Consider the same reaction and slab geometry of Exercise 7.2. ~ 1 and CB. CBS. I.106.1.] A 2B but include the mass-transfer resistance. plot this result and compare to Figure 7. Comment on the differences and similarities. What is the largest error that you would commit by using the sphere in place of the slab geometry and for what <l> value does it occur? .. same figure. RB that ~ . for <l> ~ 112. show tration? Nondimensionalize the model and boundary conditions. ~ O.CAs(l-CA) and the boundary conditions are x~1 (e) Let's compare the A and B concentrations. (b) Solve the model and find c~dx). (c) Compute the effectiveness factor versus Thiele modulus.5. Using Equation 7. Assume the dIffusivities are equal and that CA.3: Slab with finite external mass transfer Consider the same reaction and slab geometry of Exercise 7. (a) Write the dimensionless material balance and define the Thiele modulus and Biot number for this situation. What is the appropriate characteristic length. Plot this result for the same values of Biot number as in Figure 7.. CE.! . which may be different from What happens to the terms iJ2 CA li! y 2 and i! 2CAli!z2 if h and I are much greater than w? Why? Make this simplification and substitute in the production rate for A for the final dimensional model.17. DB. and show the final dimensionless model is (h) Solve this equation by performing two integrals and show CB = CBs + 2~.16 and compare the results. Plot cAlx) and CB(X) for <l> ~ 10. 10 on a log-log scale for the sphere and slab in the Exercise 7. What are the boundary conditions for the model? (b) Now consider writing the model in dimensionless form.1. Plot CA{r) for the sphere for the same <P values.106) In rectangular coordinates.420 Fixed-Bed Catalytic Reactors 7. Explain what happens to the product B concentration with this larger value of DB. Comment on the differences and similarities. Comment on the effect of pellet geometry on the overall reaction rate. - (a) The steady-state material balance for component B is DBV 2CB + RB ~ 0 For the given stOichiometry. <l> . Comment on the differences between CA and CB.. a? What is the characteristic concen~..O. x ~-I What is the definltion of the Thiele modulus for this problem? (c) What is the solution of this model? (d) Perform the integration to compute the effectiveness factor and show tanh<l> ~~-<l> (e) Plot CA(X).O. x . The catalyst exterior is assumed to be at constant species B concentration.2RA. A -'" 2B Now let's compute the product concentration profile in the pellet. (d) Now assume DB ~ lODA and replot CA and CB.9 Exercises 421 (a) The steady-state material balance for component A is (7.

83 x 1.015 cm'/s.400 Units cm3 /mol cm3 /mol cm 3/mol K 523 0. A series of experiments was performed using 0.0 atm. A+B.5: Thiele modulus and Hougen-Watson kinetics The following catalytic reaction is conducted in a O.96 g/cm3 cylinder (0.0 atm of pure A. At 150"C and 1.7: Thiele modulus and rate data The second-order catalytic reaction A. The reactant and product are gases at the reaction conditions.100 6. Using the following data.422 Fixed-Bed Catalytic Reactors 7.63 X 10~5 mol/cm 3 ·s when the concentration of A is 3.41 X 10 8 1.O (atin9 i -2.25 X 10.17x104 FA l. The following tahle lists additional information. The intrinsic activation energy is 20 kcal/mol.5 7.79 g/cm 3 0.70-cm diameter catalyst beads. A B k~4.6: Thiele modulus and first-order kinetics A catalytic reaction that is first order in the concentration of A A-B Exercise 7.79 em length) (h) Estimate the concentration of A at the center of the pellet. The catalyst vendor has changed fabrication procedures and has verified that the intrinsic rate constant listed above is unchanged. a plot of log rate versus log CA was a straight line of slope 3/2.25-cm radius spherical pellet. C is carried out in a spherical pellet with a O.500 64. Exercise 7. The rate of reaction at 398 K is 2.84 g/cm3 sphere (0.5 mol/cm 3 . Cc kJ Cm 1. determine if the mass of catalyst needs to be changed for the new catalyst to maintain the same production rate of B.045 5. The bimolecular surface reaction is irreversible and rate limiting.54x 107~ s gmol 6 Parameter Kl K2 K4 T D.48 em diam. 1_" CE. Property effective diffusivity particle density bed density shape Old catalyst 0.20-cm radius. what is the rate at 448 K for the same concentration of A? The mechanism for this reaction is shown below.0072 cm'/s 1. (a) Assuming DA is independent of temperature.]07) r ~ kc~ mol/s·g Cads + S C+ is carried out Over a spherical catalyst pellet. n'C) ISO 2OO L- RA (molls· g) oo6.4: Thiele modulus and a second-order reaction The following second-order catalytic reaction is currently carried out in a fixedbed reactor operating isothermally at 650 K.635 em diam) New catalyst 0. Additional data needed for this problem afe proVided in the following table. A.A Value 90. B2 and C are in adsorption~ desorption equilibrium with the surface.9 Exercises 423 Exercise 7.75 g/cm' 0.40 x 10~5 1O~4 Pure A is fed to the reactor and a conversion of 93% is achieved at a pressure of 2.8 X 1O~5 cm2 /s mol/em' mol/em' mol/em' g cat 2/mol·cm J ·s mol/g cat - (a) Detennine if the concemratiou of A changes appreciably within the pel- let from the value it has at the pellet surface.37 x!03 LO . The effective diffusivity of A in the pellet is 0.B (7. 0.0095 cm'/s 1. Exercise 7.0~3.17 x 10.

an inner diameter of 0.5 cm.'J g/cm 1 g/cm 3 kg/kmol 10'4 g/cm·s .B The intrinsic rate constant is 5.74 x 1013 exp (-38. A --".5 atm.3 cm. What is your final choice of catalyst size. The remainlng A in the product stream must be separated in a second unit for recycle. The catalyst pellet density Pp ~ 0. Can you meet these design constraints with aU three of the catalysts? Once you have met the constraints. What additional information would you require to perform a better design? Additional data for the catalyst and reaction are listed in the follOWing table.15 X 10' cm s·g·mol 6 I Value 0. and a length of 0.0 atm. 0.9 Exercises 425 Esthnate the value of We IN AI required to achieve 50% conversion of A at 200"C in a fixed-bed reactor using the same catalyst but with a bead diameter of LO cm. the minhnal final conversion.58 g/cm3 .5 mol/s. Your project is to make a design change with the new catalyst to hnprove the reactor efficiency. conversion. The reactor achieves 90% conversion of A. sure. Parameter Effective diffusivity Particle density Bed density Molecular weight of A Viscosity of gas ~ Exercise 7. The pellet density Pp ~ 0. The existing tubes are IOO-em long and 20-cm2 crosssectional area.84 kg/hr of B. The reactant and product are gases. In particular you are considering changing the catalyst size. A~ B kJ ~ 1. The desired production rate per tube is 18. You have been assigned to model the reactor and decide which is the best choice.35 cm and a length of 0. and 0. the maxima! tube length (it is fine to use reactor length less than 100 em. The catalyst is a cylindricalshaped pellet with a radius of 0. you can optimize the reactor operation over the remaining design decision variables. in which the units of R are cal/mol K. You may assume the bulk and pellet surface concentrations are equal. Pure A is fed to the reactor at a total pressure of 1.00375 1. an additional compressor has to be installed upstream of the reactor. A.0 atm and the temperature is 550 K.84 g/cm 3 • The reactor bed density PH ~ 0. the conversion in the reactor must be at least 93% and the outlet pressure must be at least 1. Assume the diffusion process is dominated hy Knudsen diffusion. one simply leaves the end of the tube empty of catalyst). bed length) affect both the capital and operating costs of this process. Another team member wants to use the largest diameter catalyst to minimize the pressure drop.5 em. the maximal inlet pressure.075 cm. and Thiele modulus Versus reactor length for your final design. The catalyst vendor has told you that they easily can make spherical pellets with diameters of 0. The catalyst is a hollow cylinder with an outer diameter of 1.B k ~ 5." The current catalyst has reached the end of its nseful lifetime and needs to be replaced. Exercise 7.9: A hollow-cylinder catalyst pellet in a fixed-bed reactor Esthnate the mass of catalyst required in an isothermal fixed-bed reactor for Pure A is red to the reactor with a nominal inlet pressure of 2.10: Catalyst size and production rate The following second-order catalytic reaction is carried out in a fixed-bed reactor operating isothermally at 650 K.008 cm2 Is and may be assumed constant.. Some of the design constraints that you must satisfy include the production rate. You also have been assured that the effective diffusivity. and reaction rate constant do not vary among the catalysts in this size range. O::~. The feed consists of pure A at 1. 630 K.3 cm2 Is.7 cm. and the unlts of T are K. Determlne the mass of catalyst required to achieve 90% conversion of A if the reactor operates isothermally. and at what nominal inlet pressure will you run the reactor? Include a plot of the pressure.424 Fixed-Bed Catalytic Reactors 7. the total pressure is 3. If the reactor effluent pressure is below 1.73 g/cm 3 and tbe bed density PH ~ 0. Please write a brief report discussing how the design variables (catalyst size. Exercise 7..30 em. One of your team members wants to use the smallest diameter catalyst to minimize the total mass of catalyst required.8: First-order reaction in a fixed-bed reactor The following first-order catalytic reaction is conducted in a fixed-bed reactor. and the diffusion coefficient is constant.52 g/cm3 At 630 K the effective diffusivity in the pellet is lAO x 10. The effective diffusivity of A is 0. and the minlma! outlet pres- the second-order reaction. For the separation unlt to run efficiently.75 Ll2 100 . at molar llowrate 0.5 atm. inlet pressure.0 atm. an additional compressor must be installed dovvnstream of the reactoT.15 em.21 x lQscm 3 /mol·s The feed is pure A at a molar llowrate of 15 mol/s.0 atm. If you choose to operate above the nominal inlet pressure. bed density.OOO/RT) S-I.

0 L/mol·s at 800 K. first-order reaction. (b) Choose a value of <l> ~ I and draw a sketch of cA(r) and cB(r) versus r for an irreversible. What happens to the required bed length to achieve 85% conversion if the feed pressure is doubled and the feed volumetric flowrate is held constant? You may assume that the flowrate is small enough that the pressure drop across the bed is unimportant.0 atm and 800 K. The exit conversion is 85%.14: Second-order reaction in a fixed-bed reactor The follOWing second-order gas-phase catalytic reaction is conducted in an isothermal fixed-bed reactor: A.13: Reaction-limited and diffusion-limited fixed-bed reactors A second-order catalytic reaction A. Consider the following two cases. the rate of reaction is zero. respectively. and the reactor bed densiry is 0. '-1 B (7.426 Fixed-Bed Catalytic Reactors 7. Make this sketcb as accurate as possible. and show the material halance for A in dimensionless variables is d 2c _ <l>2c _ 0 dx' What is the Thiele modulus for this problem? How does it differ from the irreversible case? What are the dimensionless boundary conditions? (e) Solve your model and show C = (1 CAe/CAJ ) cosh<l>x cosh~ . CAe.2 em are used to pack the reactor.108) The external mass-transfer rate to the outer surface of the catalyst is very large and the fluid has A and B concentrations cAf and cBf. How do you nondimensionalize your length for these other tvvo geometries.12: Mass transfer with reaction for reversible kinetics ConSider the reversible first-order elementary reaction in an isothermal solid catalyst slab of thickness w surrounded by a fluid A is carried out in an isothermal fixed-bed reactor containing spherical catalyst pellets. If you are using dimensionless variables. The reed is 15. in terms of D. CAJ' CSj. The effective diffusivity of A inside the catalyst pellet is 2.000 L/hr of pure A at 10. Write down the reaction rate expression. The catalyst has a pellet density of 0.- < Exercise 7. The reactant and product are gases. x = x w/2 .11: Mass transfer with reaction (a) Draw a sketch of the effectiveness factor versus the Thiele modulus for a spherical pellet. Find an expression for the equilibrium A concentration. Spherical catalyst pellets of radius 1.06 x 10. (c) At equilibrium. What are the boundary conditions for these two material balances.9 Exercises 427 Exercise 7. A .52 g/cm". (a) What is the new required bed length if the overall pellet reaction rate is controlled by the intrinsic reaction rate? (b) What is the new required hed length if the overall pellet reaction rate is controlled by the pellet diffUSion rate? You may assume the entire bed length is either in the intrinsic reaction rate or pellet diffusion rate-limited regimes and that the external mass-transfer reSistance between the fluid and the solid catalyst is negligible.4 cm 2 Is. Species A has effective diffusivity DA and species B has effective diffusivity DR.90 g/cm".B Exercise 7. (c) Comment on the differences and similarities between tbe effectiveness factor versns Thiele modulus plot for slabs and cylinders compared to the spherical case. (d) Provide a concise explanation of the reactor design and operation tradeoff involved in choosing hetween a catalyst pellet that gives a diffusioncontrolled reactor versus one that gives a kinetically controlled reactor. Be sure to label clearly your axes and put as many numbers on the plot as you can. Calculate the mass of catalyst necessary to achieve 85% conVersion of A. in an isothermal pelle!.B The intrinsic rate constant is 15.B. (a) Write down the steady-state material balance for speCies A and B conSidering the reaction and diffusion in the slab. and D = DA/DB. (b) Add the two material halances together and find an expression for CR in terms of CA. which may not be the same. cAfl CUi' and K = kl / L I(d) Define dimensionless concentration and distance via ~_ Exercise 7. be sure to indicate what quantities you have used to make the variables dimensionless.

110 using Figures 7. 338][18j I dc' z(e)~ c -r~======= (7.112.16: External mass-transfer effects on a second-order reaction Consider the irreversible. J(e")dc" I 1000.007 0.13 and 7.5 mol/so Name experimental rate catalyst radius effective dlffusivity of A catalyst pellet density catalyst bed density Symbol 0. 0. dz de ~ a z=oo Exercise 7.54.B. kc1 e~l taking place in a semi-infinite cyllndrical pellet.t)dx .4. with general kinetic expression (7.9 Exercises 429 Exercise 7. Use the data in the following table to determine the mass of catalyst required for 95% conversion of A in a fixed-bed reactor that can be modeled as a PFR. <P small <P large Verify the result in Equation 7. Using Equation 7.85 0. second~order Figure 7.34: Effectiveness factor versus Thiele modulus for different values of the Biot number.2S-cm diameter spherical catalyst pellet at LO atm of A and 510"C.5. (b) Verify Equation 7. Notice the Hougen~Watson and nth~order reaction kinetics are special cases of this more general form. 0. .14.18: General kine tics Consider the large <P reaction-diffusion equation. z(e) by taking two derivatives with Exercise 7. The fixed-bed reactor you are asked to size is isothermal (51 O·C).428 Fixed-Bed Caralytic Reactors 7. constant pressure (LO atm).112) .34.1 0. p. and the feed enters at NAI ~ 1. second~order reaction in a reaction r = cylindrical pellet. verify the following limits in which fCc) is the dimensionless reaction rate. and <P (7.110) ~ (a) Apply the change of variables of Section 7. Compare your calculation to Figure 7.125 0.17: Hougen-Watson kinetics Consider the Hougen·Watson kinetics of Section 7. a(ll }(x.4. Compute the effectiveness factor versus Thiele modulus for tbe follOwing values of the Biot number.58.15: Using experimental data from a fixed-bed reactor The irreversible catalytic reaction A-B is first order in the concentration of A.4 and show the solution is 14.01 R DA Pp PH Value L04xlO 4 0.11 j) Hint: recall the Leibniz rule for differentiating an integral in which the limits of integration are variable. Equation 7. The experhnental rate was measured in a O. 2.109) (7. B ~ 00. A .55 Units mol/is·cm') cm cm'/s glcm' g/cm' Exercise 7.1.4. d dt fb(l) .111 is satisfied by respect to e of Equation 7.


Fixed-Bed Catalytic Reactors

Exercise 7.19: Multiple·reaction, noulsothermal fixed·bed reactor
Consider again the catalytic converter problem of Exercises 7.2 and 7.7. Farrauto and Bartholomew provide a table of exhaust gas velocities for various driving conditions [11]. The gas velocities range from 0.54-25.2 mls as the engine operation varies from idling to fuli load. We used 0.75 mls in Exercise 7.7, which corresponds to a low engine load. Calculate the fluid profiles in the bed for uf ~ 5 mls feed gas velocity. What is the conversion of CO and CoHs if the bed volume found in Example 7.7 is used? How much more catalyst is required to reach 99.6% conversion of CO and C3H6 under the new engine operation?


Il] R. AIis. A normalization for the Thiele modulus. Ind. Eng. Chern. Fundam ..
4:227,1965. R. AIis. The Mathematical Theory of Diffusion and ReiUitivn in Permeable Catalysts. Volume I: The Theory of the Steady State. Clarendon Press, Oxford, 1975. R. Arts. The Mathematical Theory of Diffusion and Reaction in Permeable Catalysts. Volume Ii: Questions of Uniqueness, Stability, and Transient Behavior. Clarendoo Press, Oxford, 1975. R. B. Bird, W. E. Stewart, and E. N. Lightfoot. Notes on Transport Phenomena. John Wiley & Sons, New York, 1958. R. B. Bird, W. E Stewart, and E. N. Lightfoot. Transport Phenomena. John Wiley & Sons, New York, second edition, 2002. K. B. Bischoff. Effectiveness factors for general reaction rate forms. AIChE J, 11:351, 1965. S. Brunauer, P. H. Emmett, and E. Teller. Adsorption of gases in multimolecular layers. ]. Arney. Chern. Soc., 60:309-319, 1938. J. B. Butt. Reaction Kinetics and Reactor Design. Prentice Hall, Englewood Cliffs, New Jersey, 1980. J DeAcetis and G. Thodos. Flow of gases through spherical packings. Ind. Eng. Chern., 52(10):1003-1006, 1960. S. Ergun. Fluid flow through packed columns. Chern. Eng. Prog., 48(2):8994,1952. R. J. Farrauto and C H. Bartholomew. Fundamentals of Industrial Catalytic Processes. BIaclde Academic and ProfeSSional, London, UK, 1997. E. C Gates. Catalytic Chemistry. John Wiley & Sons, New York, 1992. B. E Leach, editor. Industrial Catalysis, volume one. Academic Press, New York, 1983. B. E. Leach, editor. Industrial Catalysis, volume two. Academic Press, New York, ]983. B. E. Leach, editor. Industrial CatalYSiS, volume three. Academic Press, New York, 1984. S. H. Oh, ]. C. Cavendish, and L. L. Hegedus. Mathematical modeling of catalytiC converter lightoff: Single-pellet studies. AIChE]" 26(6):935-943, 1980. c. J. Pereira, j. S. Wang, and A. Varma. A justification of the internal isothermal model for gas-solid catalytiC reactions. AIChE J., 25(6):10361043, 1979.





1I2] [l31






US] E. E. Petersen. Chemical Reaction Analysis. Prentice Hall, Englewood Cllffs, New Jersey, 1965. [l9] E. E. Petersen. A general criterion for diffusion influenced chemical reaclions in porous solids. Chern. Eng. Sci., 20:587-591, 1965. [20] E. W. Thlele. Relation between catalytic activity and size of particle. Ind.
Eng. Chem., 31(7):916-920, 1939.

Mixing in Chemical Reactors

[21] P. B. Weisz. Zeolites· new horizons in catalysts. Chern. Tech., 3:498, 1973. [22] P. B. Weisz and J. S. Hicks. The behaviour of porous catalyst particles in view of internal mass and heat diffusion effects. Chern. Eng. Sci., 17:265275, 1962.

8.1 Introduction
The three main reactor types developed thus far - batch, continuousstirred-tank, and plug-flow reactors - are useful for modeling many complex chemical reactors, and to this point we have neglected a careful treatment of the flUid flow pattern within the reactor. In this chapter We explore some of the limits of this approach and develop methods to address and overcome some of the more obvious limitations. Scope of problem. The general topic of mixing, even in the restricted context of chemical reactors, is an impossibly wide one to treat comprehensively. Obviously the phases of matter that are mixed playa large role and a comprehensive treatment would consider: liquid/liquid, liquid/solid, liquid/gas, and gas/solid mixing. Even in the contacting of two liquid phases, the miscibility of the two phases plays a large role in determining whether or not mixing occurs. We could also consider the mixing of different solid particles, in which particle size plays a large role. In this chapter, we will restrict ourselves to flUid-phase systems. One natural approacb to describing mixing is to solve the equations of motion of the fluid. In flUid systems, the type of fluid flow is obviously important, and we should consider both laminar and turbulent flow, and various mechanisms of diffusion (molecular diffusion, eddy diffusion). Usiug fluid mechanics to describe all cases of interest is a difficult problem, both from the modeling and computational perspectives. Rapid developments in computational fluid dynamics (CFD), however, make this approach increasingly attractive [l]. A second, classical approach to describing mixing is to use simple tests to experimentally probe the system of interest. These empirical testing approaches do not use any of the structure of the equations of motion, but they can provide some rough features of the mixing taking



Mixing in Chemical Reactors

8.2 ReSidence-Time Distribution


place in the system under study. In this chapter we fjrst develop this classical approach, and find out what kinds of understanding it can provide. We also identify some of the limitations of this approach. Nauman and Buffham provide a more in-depth treatment of many of the claSSical topics covered in this chapter, and provide many further citations to the research literature [18J. Convection and diffusion. We should also mention one point to bear in mlnd when thinking about how convection and diffusion affect mixing. One mlght intuitively expect that to enhance mixing and reduce spatial variation in concentration, one should seek conditions that maximlze the rate of diffusion. Although this notion is correct for mixing on the finest length scales, it is generally much more important in macroscopic scale processes to decrease variations on the larger length scales. Mixing in this regime is enhanced primarily by improving the convection, and diffUSion plays only a small role. In simple terms, one does not expect to appreciably decrease the time requited to mix the cream in one's coffee by increasing the coffee temperature (enhanced diffusion); one instead turns a spoon a few times (enhanced convection). On the finest length scales, mixing is accomplished readily for small molecules by the random process of molecular diffusion; in fact, the random molecular motions are the only effective mixing processes taking place on the finest length scales.

Figure 8.1:

Arbitrary reactor with steady flow profile.

as convection "ith the established flow, which itself might be turbulent, we would start recording a distribUtion of residence times and we would create the residence-time probability density or reSidence-time distribution. If the reactor is at steady state, and after we had collected sufficient residence-time statistics, we expect the residence-time distribution to also settle down to a steady function. Let p{O) represent the probability density or residence-time distribution, and P(O) the integrated form so


Residence-Time Distribution


probability that a feed molecule spends time o to 0 + dO in the reactor (8.1) probability that a feed molecule spends time zero to 0 in the reactor (8.2)

Consider an arbitrary reactor with single feed and effluent streams depicted in Figure 8.1. Without solving for the entire flow field, which mlght be quite complex, we would like to characterize the flow pattern established in the reactor at steady stale. The residence-time distribution (RTOl of the reactor is one such characterization or measure of the flow pattern. Imagine we could slip some inert tracer molecules into the feed stream and could quety these molecules on their exit from the reactor as to how much time they had spent in the reactor. We assume that we can add a small enough amount of tracer in the feed so that we do not disturb the established flow pattern. Some of the tracer molecules might happen to move in a very direct path to the exit; some molecnles might spend a long time in a poorly mixed zone before finally finding their way to the exit. Due to their random motions as well

The two versions of the probability function obviously contain the same information and are related by




p(O) ~ dP(O) dO

Some prefer to use exclusively PW) because if PCB) is not differentiable, then even defining p(O) is somewbat problematic. We sidestep this issue by introducing the impulse or delta function to describe p( Ol m the case where PW) has jump discontinuities.
8.2,2 Measuring the RTD

As a thought experiment to define the IUD, querying tracer molecules on their exit from the reactor is a fine concept. But we plan to actually


Mixing in Chemical Reactors

8.2 Residence-Time Distribution


measure the RTD, so we require an implementable experiment with actual measurements. We cannot measure tbe time spent by a particular tracer molecule in the reactor; to us, all tracer molecules are identical. We can measure concentration of tracer molecules in the effluent, however, and that will prove sufficient to measure the RTD. Imagine an experiment in which we measure the concentration of tracer in the feed and effluent streams over some time period, while the reactor maintains a steady flow condition. From the definition of the RTD in the previous section, the effluent tracer concentration at some time t is established by the combined exit of many tracer molecules with many different residence times. The concentration of molecules that enter the reactor at time t' and spend time t - t' in the reactor before exiting is given by cf(t')P(t - t')dt' These molecules are the ones leaving the reactor at time t that establish effluent concentration ce (t), so we have

assumption using deviation variables. Because the feed concentration is constant at Cf after time zero, we can take it outside the integral in Equation 8.5 and obtain

= cf



= cfP(t)

So for a step-response experiment, the effluent concentration versus time provides immediately the integrated form of the IJ:!'idence-time distribution (8.6) step response Pulse and impulse responses. An impulse response is an idealized experiment, but is a useful concept. As we will see it provides the RTD directly rather than in the integrated form. To motivate the impulseresponse experiment, imagine we abruptly change the inlet tracer concentration from zero to a large value and retrtrn it to zero after a short time as sketched in Figure 8.2. Such a test is called a pulse test. The pulse test is no more difficult to implement than the step test; it is merely two step changes in feed concentration in rapid succession. In some ways it is a superior test to the step response, because ao(t) by returning the tracer concentration to zero, we use less tracer in the experiment and we cause less disruption of the normal operation of the reactor. The impulse response is an idealized limit of the pulse response. Consider a family of pulse tests as sketched in figure 8.2 of shorter and shorter duration 1\1. We maintain constant total tracer addition by spiking the feed with higher and higher concentrations so that the product Cfl\t = a is constant. The impulse response is the limit of this o experiment as I\t - O. We call Figure 8.2: Family of pulses and this limiting feed concentration the impulse or delta function. versus time function the delta funCtion, ao (1). It is also called

The inlet and outlet concentrations are connected through this convolution integral with the residence-time distribution. If we conduct the experiment so that the feed tracer concentration is zero before an initial time t = 0, then the integral reduces to

= J~ Cf(t')P(t-



Notice we can change the variable of integration in Equation 8,4 to establish an equivalent representation

J~ Cf(t -



which is sometimes a convenient form. This connection between the inlet and outlet concentrations, and the RTD, allows us to determine the RTD by measuring only tracer concentrations. We next describe some of the convenient experiments to determine the RTD. Step response. Although we can in prinCiple use any feed concentration time function to determine the RTD, some choices are convenient for ease of data analysis. One of these is the step response. In the stepresponse experiment, at time zero we abruptly change the feed tracer concentration from steady value Co to steady value Cf. For convenience we assume Co = O. Exercise 8.4 shows that we can easily remove this


Mixing in Chemical Reactors

8.2 Residence-Time Distribution

the Dirac delta function or an impulse, hence the name, impulse response_ The constant a is the amplitude of the delta function. The mathematical problem with this approach is that the limit is not a welldefined function. The values at all pOints other than t ~ 0 go to zero as we take narrow enough pulses, and the value at t ~ 0 is unbounded. But mathematicians have taken a bold step and defined this new kind of function by what it does rather than its numerical values_ The main property of the delta function is that, because it is so narrowly focused, it extracts the value of an integrand at a point in the interval of integration,

of experiment and assessment of uncertainty in the RTD is highiy recommended. Shinnar [22J provides a detalled discussion for further reading on the science and art of determining and using RTDs to solve industrial reactor problems_
8.2.3 Continuous-Stirred-Tank Reactor (CSTR)

roo g{t)o(tJdt ~ g(O),

all g{t) normalized




~ 1,

So if we can approximate ef(t)

~ ao(t),

then we have from Equation 8.3

~ a foo o(t')p(t -

~ ap(t)

So for an experiment approximating an impulse, the effluent concentration versus time provides the residence-time distribution directly

= c,(e)la,

impulse response


We do not require any sophisticated mathematics to use tbese gener-

alized functions such as the delta function. We introduce them here because they make our work simpler, not more complicated. You will also find these functions useful in the process control course when you analyze the dynamiCS and feedback control of processes. Moreover, empirical tests, such as the step- and impulse-response experiments, provide a general method for modeling the dynamiC behavior of many processes. Using these methods to determine a reactor RTD is just one small example of their uses in chemical process modeling. Arbitrary feed concentration experiment. It may be inconvenient or impossible to perform a step or impulse test on a reactor. In some situations it is possible to measure a low-concentration impurity in the feed and effluent streams, and construct the RTD from those measurements. In these cases the feed concentration cf(t) is an arbitrary, but measured and known, function of time. One must take care that the nOise in the measurements does not obscure the RTD, and replication

We next examine again the wellstirred reactor introduced in Chapter 4. Consider the following stepresponse experiment: a clear fluid with flowrate Qj enters a well-stirred reactor of volume VR as depicted in Figure 8.3. At time zero we start adding a small flow of a tracer to the feed stream and measure the tracer concentration in the effluent stream. We assume we keep the flow of the Figure 8.3: CSTR and tracer experiment. tracer small enough that we do not disturb the existing flow pattern in the reactor. We expect to see a continuous change in the concentration of the effluent stream until, after a long time, it matches the concentration of the feed stream. We are by now experts on solving this type of problem, especially in this simple situation without chemical reaction to complicate matters. Assuming constant density, the differential equation governing the concentration of dye in the reactor follows from Equation 4.38
Qf dt~VR{Cf-C),



in which c is the concentration of the dye in the reactor and effluent stream. In Chapter 4, we named the parameter VR I Qf the "mean residence time" and gave it the symbol 11_ For our purposes here, we have already reserved II for the distribution of reactor reSidence times, so we define T ~ VR1Qf. We show subsequently that T is indeed the mean of the RTD for a CSTR. We already derived the solution to Equation 8.10, where so we have immediately
(8.11 )


Mixing in Chemical Reactors

8.2 Residence-Time Distribution

which upon differentiation gives CSTR residence-time distribution



T ~

T ~

1/2 1


It is often helpful to have two different views of the same result, so we next consider the well-mixed reactor as a game of chance and re-derive the RTD for the CSTR. Imagine we divide the reactor into a number of volume elements, depicted in Figure 8A. At each time instant, some group of volume elements is wIthdrawn from the reactor, and new volume elements enter from the feed stream. We assign a probability law to model which volume elements are wIthdrawn from the reactor. The well-mixed assumpFigure 8.4: CSTR and voltion is equivalent to the statement ume elements in a game of that all volume elements are equally chance. likely to be wIthdrawn from the re· actor. One can view the mixing as quickly (instantaneously in the case of perfect mixing) randomizing the locations of the volume elements so that each one is equally likely to be nearest to the exit when the next element is wIthdrawn in the effluent stream. We also see why perfect mixing is a physical idealization; it is not possible to instantaneously randomize the locations of the volume elements wIth finite power to the stirrer. We have n volume elements, each with volume Vn ~ VRln. In time interval 61, a volume of QI':c.t leaves with the effluent stream. The number of elements leaving in the time interval is therefore QI':c.t I Vn out of n, and the fraction of elements ,,1thdrawn is 1.2

p(e) 1
0.8 0.6 0.4 0.2 0 0



Figure 8.5: CSTR residence-time distribution.

the probability that an element has residence (survival) time


it.t is



(1 -

t.t T )8 61


We now take the limit as 61 - O. Recall from calculus limO + ax)l/x




Taking the limit and using this result in Equation 8.13 gives


~ e~eiT

which. after normalization. is again the residence-time distribution of the CSTR. The normalization constant is computed by integration

so the residence-time distribution is Now consider the probability that a particular volume element is still in the reactor after i time intervals. The probability of removal is j, so the probability of survival is 1 f. After i trials, each of which is independent. the probability of survival is the product (J .~ fJ i. Therefore
p(e) ~

I -c o'"


which is plotted in Figure 8.5 for a variety of mean residence times. We


! ,

16) . We can therefore compactly wTite Equation 8.v t > 0.v t < 0.2 Residence. Jo ee~8!T de roo r~ T = o z ~ [-TerOIT _ (de~8IT]I~ =T and we have establisbed that the mean of the RTD for the CSTR is indeed T = VR/Qf. In fact. which explains the exponential decrease in p (0). The Heaviside function is defined as follows H(t)={~: t<O t>0 (8.6. in which v is the velocity of the axial flow and 1 is the reactor length. II is convenient to introduce the unit step or Heaviside function to summarize this result.z) = so the effluent tracer concentration is given by (8. Again. An integration by parts produces e= Jo e p(O) de = 1. Notice the exponential distribution tells us that it is unlikely for a volume element to remain in this reactor for long because at each instant there is a constant probability that the element is withdrawn in the effluent. Consider a step test in which the reactor is initially free of tracer and we increase the feed tracer concentration from zero to Cf at time zero. to choose the time of the next reaction in the stochastic simulation of chemical kinetics in Chapter 4. e For z . From these physical conSiderations.4 Plug-Flow Reactor (PFR) and Batch Reactor crH(t .z/v) The simple flow pattern in the PFR produces a simple residence-time distribution. a classic and slightly counter-intnitive case: the most likely value of e. = VR/Qf. if the mixing is not perfect.17) Because we are armed with the delta function. for example. For z . the maxe p (e) occurs at e = 0.O e (8. { cl' z-vt>O z-vt<O We can differentiate this result using Equation 8. the time until the next occurrence of the event is distributed as a decreasing exponential function.Time Distribution 443 next compute the mean of this distribution.442 Mixing in Chemical Reactors 8.14 as p(O) = Z le~e. we can even consider differentiating this discontinuous function to obtain the relationship dH(t) dt = 6(t) (8. the tracer front has not reached location z at time t so the tracer concentration is zero.15) Figure 8. As shown in Figure 8. zero is the most likely residence time. our derivation of the R1D of the C5TR shows the following general principle: given an event with constant probability of occurrence.2. the front has passed location z at time t and the tracer concentration is equal to tbe feed value cf. which is different from the mean of the distribution. we can write the reactor tracer concentration immediately c(t.6: Plug-flow reactor with moving front of trace.18 to obtain (8. this maximum does not occur at e = 0 because it requires finite time for a volume element entering with the feed to be convected to the exit.20) 0. the tracer travels in a front that first reaches the reactor exit at time t = l/v. This principle is used.19) and the integrated from of the residence-time distribution is therefore (8.18) Equation 8.Z) = 8.16 can then be summarized compactly by c(t.. 1t is difficult to win for long in this kind of game of chance. Moreover.

22) which is plotted in Figure 8. The mean of this distribution is then also VR /Qj.5. All material is charged to the reactor at time t = 0 and remains in the reactor until the final batch time. the RTD for a batch reactor is immediate..8 for T ~ 2. p(OJ = y(n.21 also apply to the batch reactor. which we may call Then Equations 8. n~l 2 1 2 3 n=2 n=3 11 = 11= 11~ n 5 10 25 100 -.24) Notice the gamma function is a generalization of the familiar factorial and for n an integer.. 1..5 CSTRs in Series Consider dividing the volume of a single CSTR into n equal-sized CSTRs in series as shown in Figure 8. e.444 Mixing in Chemical Reactors 8. We can compactly summarize the RTD for the plug-flow reactor hy o 1 2 3 4 5 e Figure 8. Figure 8. pee) = D(e e) PFR and batch reactors P{&) = H(e - e) (8.21) fact Eqnation 8.7 1 .I " " which shows that VR / Qf is the mean residence time for the PFR as well as the CSTR.2 Residence..Time Distribution 445 2. we can show p (0) = fooo ttn-Ile-'dt (8.8: Residence-time distribution p (e) versus e for n CSTRs in series. which is verified hy using Equation 8.. . If we solve for the effluent concentration after an impulse at time zero. T = 2.7: Series of n equal-sized CSTRs with residence time Tin. all tracer molecules spend exactly the same time VRIQj in the reactor.25) . Notice the residence-time distribution becomes more narrowly focused at T as n increases.23) 8.7.5 11 ~ pCB) In other words.22 for large 11 is another approximation for D(e we integrate Equation 8.-------------------------------------.. If the single CSTR has volume VR and residence time T ~ VR I Qf' each of the CSTRs in series has volume VR In and residence time T In. [(n) = {n -1)1 = (~)n ~~l~e~nO/T T (n-I)I ( nn) in which the gamma function is defined as n11) (8. TJ.22 we obtain If Likewise. even though the residence-time distrihutions of these two reactors are quite different. In The incomplete gamma function is defined by the same integrand on the finite interval (8.2.

27) n = 25 n=]OO--«-' 3 4 5 e Figure 8. We specify a zero slope condition at the tube exit.. in the PFR a change in the feed concentration travels as a perfect fronL But measurements in actual packed-bed reactors may show a fair amount of spreading or dispersion as this front travels through the reactor. Further discussion of the applicability aud some alternatives to the Danckwerts boundary conditions are given by Bischoff [3]..22 to obtain a series representation Pie) = z =1 (8.--' ac a2c -v.29) Ideal plug flow may not he realized in many packed-hed reactors.2 0 0 1 2 convection diffusion - (8.T (826) (n . 8. 0.6 These two boundary conditions have become knovm as Danckwerts boundary conditions [6].+2X.6431. dx d 2c . we write a material balance over a small region containing the entry point. D. p. It is now convenient to introduce a coordinate moving with a scaled velocity x=---= J4D/t J4Dt/T Dispersed Plug Flow in which z-vt z-t/T (8.2. and consider diffusion and convection terms.8 molecular diffusion caused by the random motion of the molecules.4 0.27. e for n CSTRs in series.446 Mixing in Chemical Reactors 8. Evaluating the incomplete gamma function is numerically better than summing this series when n is large. but they were derived at least 45 years prior to Danckwerts in a classic paper by Langmuir [16].9 shows Equation 8.23 for a range of n values. The diffusion term introduces a second-order derivative in Equation 8. The dispersed plug-flow model given in Equation 8. Molecular diffusion in the direction of the convected flow is always negligible in packed bed reactors operating at reasonable flowrates. As shown in Figure 8.=0 . T = 2. You can perform integration by parts on Equation 8.(ne/T)2 '" I! + 2! + + (ne/T)n<i) -nO.6. The coeffiCient.+ Dl DZ2 oz ~ 0.30) . c(O) = Cj. A diffusion-like term has been added to the mass balance.6 P(e) ac at '--v-' n = ] n=2 n=3 n=5 n= 10 accumulation '--.2 Residence-Time Distribution 441 1 0.28) 1_(1 + netT .272]. the inlet boundary condition is no longer justthe feed condition.27 allows for this spreading phenomenon. and Parulekar and Ramkrishna [21]. such as Taylor dispersion [2. is called the dispersion coefficient Note this term is not meant to represent true dimensioniess dispersion number Transforming variables from z and t in Equation 8. that we used in the PFR.9: P(B) versus Given diffusion in the tube. p. de (8. The spreading of concentration fronts is caused by other phenomena. To derive the boundary condition at the inlet. Levenspiel [17. so we now require two boundary conditions. Figure 8.1)1 e .27 to x gives I -x~ d . We develop next a model that allows for deviations from plug flow.

-.2 and we have calculated the integrated form of the RTD for the dispersed PFR with simplified boundary conditions (8.5 2 p(8) = 1/2 [1-erf( ~)J .5 1 Equation 8..5 1 1.34 to obtain the dispersed plug-flow RTD (8. T 2.30 with boundary conditions listed in Equations 8. I " 2 D=0.2 / -' . The solution to Equation 8.5 2 2.34) 1... We can differentiate Equation 8.t) 3 2. I I I I 0.34 is plotted in Figure 8.2 I ° The integral can be expressed in terms of the error function..33) D=0. for z.t 2 dt (8.5 3 3.5 Substituting in the original variables and setting 2' = 1 to obtain the response at the reactor outlet as a function of time gives C(2'= l.2 Residence-Time Distribution 449 Rather than use the Dane kwerts boundary conditions. See Exercise 8.35) 0.4Dt/T (8. .10: e for plug flow with dispersion number D.OI D=0.448 Mixing in Chemical Reactors 8.5 ° ° 0.31) These boundary conditions correspond to stating that c ~ for z. : .10 for comparing the dispersed PFR RTD with Danckwerts boundary conditions to the one we calculate here with the simplified boundary conditions.10 for T = 2 and various dispersion numbers D.8 X=oo (8. t values corresponding to arbitrarily long times after the step change has passed.002-D=O. which is defined as erf(x) = ° T = 1 3 4 5 e pee) versus 2. Figure 8.002-D=O. 2 0 rrr IX e. we can approximate the behavior with the following simplified boundary conditions 1 c = I. c = 0.4 .31 is ° 0.5 5 e e for plug flow This RTD is plotted in Figure 8.05 D=0.'/J . • I . and that c = I.11: Residence-time distribution p(O) versus with dispersion number D.32) 3. t values corresponding to arbitrarily long times before the step change arrives.1 L Figure 8.Ol D=0.6 p(8) 0. x = -00 0.05 D=0.5 4 4.

5 I Pe DA vI DA[ which measures the rate of diffusion compared to the rate of convection. Dispersed plug flow with reaction. Example 8.27 for dispersed plug flow to account for chemIcal reaction. which is called orthogonal collocation. the mass-transfer analog of the inverse of the Peele! number. we have solved exclusively steady-state profiles in tubular reactors. the inverse of the Peclet number is often called the Bodenstein number in European literature. rather than the usual ODE for the steady-state profile.36. second-order reaction 2A~· Figure 8.5 I Units L/mol'!nln moljL m2/min m/mIn m Table 8. Obviously tubular reactors experience a start-up transient like every other reactor. What are the reSidence-time distributions for the two reactor systems? Wbat are the overall conversions for the two systems? E. The kinetic and reactor parameters are given in Table 8.13. but was defined and used by Langmuir [161.01 0.ililmp/e 8. t=O = (8. liquid-phase. as given in Equations 8. iJ 2c +Djl +Rj az t (8.39) (8AO) oc' at = -v a~ iJc-. is related to another dimensionless group.5. Component A is transported by convection and diffusion down the reactor. D A. Equation 8.2kc A iJ2 CA 2 (8. and the dispersion number contains the effective axial dispersion coefficient. We modify Equation 8. D.4(Z.1: Mass·transfer and kinetic parameters for I'. Appendix A describes the method we use for this purpose.450 Mixing in Chemical Reactors 8. v 1 0. the concentration profile has reached its steady value. .t) =0.2: Order matters ConSider two arrangements of a PFR and CSTR of equal volume in series as shown in Figure 8. Levenspiel makes a compelling case that these two quantities have different origins and motivations and deserve different names..36) 1 Danckwerts boundary conditions. an initial condition is required C. Calculating the transient tubular-reactor behavior involves solving the partial differential equation (PDE)... Given the low value of dispersion in tbis problem. Weller [261 provides an interesting discussiou of the history of the literature on the Bodenstein number.1: Transient start-up of a PFR Compute the transient behavior of the dispersed plug-flow reactor for the isothermal. Up to this point in the text. The feed enters the reactor and the A con· centration at the iniet rises rapidly. After about t = 2. The key difference is the Peclet number contains the molecular diffusivity. Solution The mass balance for component A is oeA at = -v OCA iJz + DAl oz2 . We see the reactor initially has zero A concentration.1.2 Residence-Time Distribution 451 The dispersion number. can be applied without change.28.. I "". and the reaction consumes the A as it goes. D l .12 shows the transient profiles. To further complicate matters.38) The boundary conditions are z=o z Finally. and this time-dependent behavior is also important and interesting. (8. the steady-state profile is close to the steady-state PFR profile for 0 this problem. Parameter k C4j Value 0. Example 8.1. which does not appear to have been defined or used by Bodenstein.37) The reactor is initially filled with solvent.

exp(-(t T)/T» H(t . That feed concentration to the PFR is then simply delayed by T time units to give for tbe CSTR-PFR arrangement.25.8 1 With this feed into the CSTR. ~ 0. We see that tbe conversions are not the same and that the PFR-CSTR gives higher conversion (lower outlet concentration) than the CSTR-PFR for all values of kTCo for a second-order reaction. For this case the intermediate stream is given by a delayed step input C] (t) = H(t - T) 0 0 0.. "I so again for this case p(e)=(1-exp(-(t-T)/T»H(t-T) Co C2 and the two residence-time distributions are equal.T) Figure 8. The steady-state conversions for both arrangements are also simply calculated.2 t~O edt) ~ 1. .5 pee) ~ (1-exp(-(t-T)/T))H(t-T) "" " Next consider a unit step test into the PFR-CSTR arrangement. 0 " 1 "2.4 z (m) 0.2 Residence-Time Distribution 453 Solution 1 Consider a unit step test for the CSTR-PFR arrangement.6 0. Because we are not even told kTCO.6 0.) 1 (B) Figure 8. we check over a range of values. -_ -1 + -vl + 4kTCi -_ C( c t. So we wish to compare P(C(co» for the CSTR-PFR caSe and C(P(co» for PFR-CSTR case. = 1 + k Tei = P(c.8 0.:::.2 0. CA (z. For a single CSTR.13: PFR followed by CSTR (A). t) versus z for various times. the effluent is merely the CSTR response to a unit step change after we shift the starting time of the step forward T time units. The outlet concentration for the CSTR is given by 0. § c t ~ 2.exp(-t/T) :::l .) 2kTC. the steady-state inlet and outlet coneen\rations are related by Co / (A) ct. For a single PFR.14 displays tbe result.5. and CSTR followed by PFR (B).12: Start-up of the tubular reactor.452 Mixing in Chemical Reactors 8. Figure 8.4 0. C2(t) = (1 . the inlet and outlet concentrations are related by Co C2 co/c. 0 .

as in Figure S.454 Mixing in Chemical Reactors 8.16. Although we may fiod it instructive to separate these two phenomena in the simple reactor mixing models under discussion. We can determine the limits of achievable mixing consistent with a measured reactor residence-time distribution. We defioe next the two extreme limits of mixing consistent with the given RTD.15: Completely segregated flow as a plug-flow reactor with 0. in actual reactors this separation may be unrealistic. This last point is perhaps more clear if we redraw the reactor configuration as an equivalent bank of PFRs of different lengths without side exits.3 Umits of Reactor Mixing 455 1 0. Accurate prediction of reactor performance may require solUtion or apif someone were to characterize your learning in this course by measuring your hours spent in the classroom (RTD). however. Imagine we know a reactor RTD. jf the students were attentive Dr asleep.15. they would hopefully obtain a positive correlation between learning and residence time. These mixing limits do provide some insight into the limits of achievable reactor performance. and micromixing: the exchange of material between different volume elements during their residence times. inclnding material transport dne to diffusion.14: Comparison of the effluent concentrations for the two cases shown in Figure 8.----3 4 5 proximate solution of the equations of motion for the fluid.3 0.-0-Figure 8.2 0.. pee).3 limits of Reactor Mixing We have seen in the previous section that complete knowledge of the reactor residence-time distribution is insufficient to predict the reactor performance.. the residence time e of the material reaching that location increases.1 Complete Segregation -' Figure 8. Although we have characterized completely the time tracer molecules spend in the reactor. although this might be difficult to achieve in practice.6 004 side exits. We know from Example 8.8 0. it is the limit of no mixing between volume elements. and so on. We would want to know if the instructor was prepared for lecture and saying reasonable things.3.2 that this constraint does not tell us the complete state of mixing in the reactor.---.9 0. either through direct measnrement in a tracer experiment or solution of the equations of motion or some other means. A PFR with this removal rate then has the speCified RTD. But we would naturally want to evaluate the environment inside the classroom during these hours if we were going to make more accurate predictions of learning. outlet flows adjusted to achieve given RTD. 8. The first limit is called complete segregation. 8. In defense of the simple mixing models. This reactor . We feed the fraction pie) of the total flow into each tube of residence time e so as to achieve the given !tTD for the composite reactor system. No material in two volnme elements with different residence times is ever exchanged because the plug flow has zero backmixing. although this connection remains an active area of research as discussed in section 804.A \25].7 C2 CSTR-PFR-PFR-CSTR .13. We can realize this limit by considering the ideal reactor depicted in Figure 8. As we progress down the plug-flow reactor. These ideas were laid out in influential papers by Danckwerts and Z.vietering. Co 0.5 0.1 0 1 2 kTCO --. We can imagine withdrawing from the reactor at each location or e a fraction of the flow corresponding to the required RID value pee). we have not characterized their surrounding environment in the reactor during this time 1 In the literature these two effects are sometimes termed macromixing: the distribution of residence times of molecules moving with the flow. they do provide another important insight. Each tube has a single e valne according to its length.

3 Limits of Reactor Mixing 457 feed passing througb that tube Cs ~ r p(8)c(8)d8 (8. but experience suggests a strongly decreasing function like the exponential (8. sucb as z ~ exp( -e). we require tbe solution at 8 ~ 00 to determine the effluent concentration of the reactor. Equation 4. Differential equations on semi-infinite domains are termed singular.(B) stops changing. if the residence-time distribution is zero beyond some maximum residence time. we can map the semi-infinite domain onto a finite domain and let the ODE solver do the work for us. we would need to check the solution for a sequence of increasingly large e values and terminate the ODE solver when the value of c.41. Alternatively.112.456 Mixing in Chemical Reactors 8.(O) ~ 0 system is called completely segregated because there is no exchange of material between lhe various tubes. C c.42) in which cW) comes from the solution of Equations 8. We assume a single reaction and constant density throughout the follOwing discussion.. then it is straightforward to integrate the initial-value problem on 0 5 8 5 8 max . 8max .(O} ~ 0 Figure 8. dc d8 ~ R(c). maximum mixed flow (B). It is often convenient to remove the explicit integration required by Equation 8. Let e.(8} B ~ J: pW)c(e)d8 From this definition it is clear that Cs (8) satisfies the following differential equation and initial condition ~~ ~ p(8)c(8). but. and an intermediate mixing pattern (C). so . in general. and the reactor effluent concentration is the limit of c.16: Alternate representation of completely segregated fiow (A). and require some care in their numerical treatment as we discuss next. On the other hand.43) c. It is a simple matter to predict the behavior of this completely segregated reactor. Each tube acts as its own private reactor that processes material for a given amount of time and then discharges it to be mixed with the other reactors at the exit. Let c (8) represent the concentration of a reactant in a volume element that has spent time 8 in the reactor. c. (8) as 8 ~ 00.42. Numerical solution. Because of the semi-infinite domain. e(O) ~ cf Notice that this is an initial-value problem. Many transformations are possihle.43 as an initial-value problem as written with an ODE solver.41) The concentration of the effluent stream is then simply computed by multiplying the concentration of each tube by the fraction of the total . We can solve Equation 8. Each tube of specified length or volume V can be assigned a residence time according to 8 ~ V I Q. Because the reactor is segregated. We can combine the two differential equations for convenient numerical solution of the segregated case de ~ R(e) d8 dc de ~ p(8)c(8) e(O) ~ Cf (8. (e) represent the effect of combining streams with residence times less than or equal to 8. each tube satisfies the constant density PFR equation.

We face the equivalent problem if we mix two identical. A2. maximum mixedness.17: Maximum mixed flow as a plug-flow reactor with side entrances.17 128]. It is useful to plot c(z) to make sure the z = 1 end does not exhibit some unusual behavior.18: Volume element in the state of maximum mixedness. These are values of the variable of integration beyond which the ODE solver will not step. the e locations increase from zero at the exit of the tube to large values at the entrance to the tube. that usually indicates the transformation was successful. If we mix two groups of molecules with different times-to-go. If the ODE solver terminates successfully. causes the right·hand side to go to infinity at z = 1. Consider the equivalent representation of maximum mixedness in Figure 8. Again. We next derive the differential equation that governs the maximum mixedness reactor [28]. the mixture must later be separated again so that the molecules may exit the reactor at the specified times Al and A2. The shading means that the material at these locations is completely mixed with the material from the other tubes at these same locations [25].3. we have no process to separate them again based on their initial locations because the molecules are identicaL Such a separation process would violate the second law of thermodynamics.B. e = 00 so R(e) = 0. the mixing process at a given location is mixing material with different times spent in the reactor.43 de R(c) dz (1_Z)2 dcs p(z/(l-z»)c dz= (1_Z)2 to c(O) = cf (8. The feed stream is distributed along the length of the PFR and injected at the appropriate rate at various side entrances corresponding to different e "locations" to achieve the required RTD. Indeed.16. It is not possible to mix material with different times-to-go without changing their exit times for the following reason.16. As depicted in Figure 8. We would specify z = 1 as a critical value because the right-hand side is not defined past z = L At the value z = 1. we transform Equation 8. ConSider an expanded view of the reactor in a state of maximum mixedness shown in Figure 8.16. A C(A) Figure 8. Notice that we have aligned the exits of the tubes in figure 8. After mixing. we are mixing material not with common time spent in the reactor but rather with a common time-to-go .458 Mixing in Chemical Reactors 8. we would specify the right-hand sides are zero because the reaction will have equilibrated at z = 1. some care with must be taken because the denominators are also going to zero. Notice that because the flow has been reversed compared to the segregated case.B. pure-component gases initially on opposite sides of a partition. and pee) = 0 at e = 00. inlet flows adjusted to achieve a given RTD. we rewrite the derivative as de = dedz = O_Z)zde de dzde dz Using this result. Al. Reactors with these speCified sidestream addition policies are conceptually important in understanding recent research on achievable reactor performance as discussed in Section 804.18. 8.\. We allow an infinitely long tube if we wish to allow RTDs such as the CSTR defined on a semi-infinite domain.3 Limits of Reactor Mixing 459 G-Figure 8. and we simply exchange one singularity for another. Therefore when we mix material between tubes.44) es(O) =0 Most modern ODE solvers allow the user to specify critical stopping values. A more gentle transformation and its inverse are Z=-1+e' e e= z -z Using this change of variable. Such a separation is not possible because molecules are not distinguishable based on their times-to-go in the reactor. by reversing the flow in the segregated reactor as shown in Figure 8. feed with Ilowrate Qp(8)dO is added to each tube having .2 Maximum Mixedness We realize the opposite mixing limit.

50 was zero. a critical stopping value should be set at z = I to avoid an undefined right-hand side. care should be exerCised at z = 0 because the denominator goes to zero at z = 0(8 = (0). For c to remain bounded as A .\) e(A)-ef ptA) ( ) --::J 5 (8. The analogous transformation to the segregated reactor is Z= .\ = O.48) Provided we know the limit p(. Example S. A . irreversible reaction. the entrance to the longest tube.'\ p(A')dA' . Given the RTD of two equal-sized CSTRs in series for a single.47) (8.etA) LlA becomes fro Mb" p(A')dA' _ e(A)p(A) + efP(A)+ R(e) [ p(A')dA' l[ I-P«(l-zl/z) ( c-cf ) -R(c) ] p«l-z)/z) C(O) = Coo (S.0) interval into (0. Again. the material balance for this element consists of the following terms: entering at A + LlA: leaving at A: feed addition: production: Q boundary condition is required at A = 00.(A». We wish to transform the'\ E (00. A=oo (fOO A+C.1).A P(A')dA') e(A + AA) (8.49) We "'ish to calculate the reactor effluent concentration.50) =0 Taking the limit as AA . So if we consider a volume element to be a mixed section of the composite reactor between timesto-go A and A + LlA.45) Q ( [ p(A')dA') eUt} Q (p(A)M)cf Q ( [ p(A')dA') LlA R(e.PtA»~ as A . we stipulate the boundary condition dc dA =0. which is given by e(A) at A = O.00. we call this value Coo. we can solve Equation 8.49 directly for the boundary condition on e at A = 00. we wish to integrate from A = 00.460 Mixing in Chemical Reactors 8. Plotting c(z) and examining the z = 0 end for unusual behavior is recommended.\l/(l . compute the reactor effluent concentration for the following cases: segregated flow. to the combined lUbe exits at . this first-order differential equation is singular. maximum mixedness and two ideal CSTRs.erA) foo p(A')dA' + cfP(A)M+ A R(e)M r p(A')dA' =0 in which the minus sign arises hecause we are changing the direction when integrating in the transformed z variable.00.0 and rearranging gives dA = de peA) fA' p(A')dA' ( e(A) - Cf ) - R(e) Equivalently we can express the integral in terms of the integrated form of the R1D and "'Tile and we integrate from z = 0 to z = L Again. di= de I-P(.49 then We can combine the first two integral terms and divide by LlA to obtain e(A + M) . - zE Numerical solution.3: Two CSTRs in series We illustrate the results of these sections with an example taken from Zwietering [28]. Note that Coo " Cf. 'rate per volume volume of reactor element 1 +x X=-Z l-z Considering the reactor is at steady state. secondorder. As in the segregated reactor case. (S.46) (S.3 Limits of Reactor Mixing 461 residence time e to achieve the desired RTD. we set the accumulation to zero and obtain c(A + LlA) The derivative becomes de dc dz 2 dc -=--=-z d'\ dz d'\ dz foo lvr6. Equation S. We set the right-hand side to zero at z = 0 because we determined the value of Coo such that the bracketed term in Equation 8.

19 displays the solution to this differential 0 equation for a range of K values.1 1 0.16.. Notice that all the physical constants of the reactor combine into the single dimensioniess constant K..7 0.19: Dimensionless effluent concentration = cleo versus dimensionless rate constant K = kTCo for second-order reaction.8 0. C ~ C / Co and x ~ 0/ T. Figure 8.3 0. the reactor with the CSTR preceding the PFR is in the condition of maximum mixedness .3 Limits of Reactor Mixing 463 Solution The residence-time distribution for two CSTRs in series is given by Equations 8.462 Mixing in Chemical Reactors 8. The transformed derivative satisfies de dc dz 2de dx = dz dx = -z dz so the final differential equation is dc __ .16. The balance for the maximum mixedness case becomes -0.16. whIch we can transform via z~--.2. Again we have an equation on a semiinfinite interval.4 maximum mixedness .- ~ ~~ e. Weinstein and Adler [25] also proposed an interesting general conceptual mixing model by allowing a general mixing pattern between the various tubes as depicted in Figure 8.Z2 2(1-z)/z+1 C(O) 1) +Kc 2] = Coo The effluent of the maximum mixed reactor is given by the solution e(z) at z ~ 1.0) is transformed to z E (0.. no segregated streams remain to be combined at the reactor exit.9 0. If we apply the zero slope condition at . we obtain the quadratic equation 0 10 K 100 1000 Figure 8. c whIch can be solved for Coo.2 CSTRs ----segregated .2 0. The material is mixed at the earliest possible times as it travels down the tubes.20. The situation depicted by the maximum mixedness reactor of Figure 8. The segregated reactor depicted in Figure 8.51) -~---cc dx 2x+ 1 in whIch K ~ kCoT.15 and Figure 8.26 for n ~ 2. 1 1+x l-z X=-z in which x E (00. The material remains segregated until it reaches the common exit where the segregated streams are finally mixed in the reactor effluent.\ ~ 00.1) + K-2 (8.6 C 0.1).22 and 8.01 0. Comparing these two reactor configurations.z) / z (-c dz. Intermediate conditions of mixing.P(fI) ~ (1 + ZO/T)e-Ze!T in whIch T ~ VR / Qf and VR is the total volume of the CSTRs.B is sometimes called early mixing.1 [ 4(1 . the RTD for all cases is given by 2 CSTRs in series. Consider again the two reactors in Example 8.1 Defining dimensioniess variables.(. the equation becomes dc 4x (. The reactors have identical RTDs.A is sometimes referred to as late mixing or mixing as late as possible.17 and Figure 8.Z8 / T 1. p(fl) 1 0.5 0.. The conceptual mixing pattern is sketched in Figure 8.

the stirrer is modeled as shearing the liqUid A feed stream into small t uniformly sized particles of A and randomly distributing them in the continuous phase containing B. This physical picture.464 Mixing in Chemical Reactors 8. although idealized. 8. pp.20: CSTR followed by PFR (A) and PFR followed by CSTR (B) as examples of complete and partial mixing.1-171. it displays an intermediate state of mixing. In other words.3. uniformly sized "particles" of component A dispersed in the con- tinuous phase containing component B dissolved in a solvent. As shown in Figure 8. The PFR preceding the CSTR is not in the condition of maximum mixedness. such as suspension and emulsion polymerization reactors. the macromixing is excellent. We show in Section 8.3 Mass Transfer and Limits of Reactor Mixing Consider the follOwing physical picture to help in our understanding of segregated flow and maximum mixedness. Let the mass transfer between the particles and the continuous phase be described by a mass-transfer coeffiCient. We therefore have an ideal CSTR residence-time distribution. and determining the residence-time distribution would not indicate anything except an ideally mixed reactor. and could be considered well mixed or segregated.20.21 shows the classic situation in which we mix two liquid-phase feed streams in a stirred tank for the second-order reaction A+B. The size of the A "particles" is one measure of how well the stirrer is working. we would see the classic step response for the CSTR. complete mixing gives the lowest conversion possible consistent .16.C (8.52) Figure 8.4 that because the reaction rate is second order. is motivated by several types of real reactors. ~ :::::=:::: 00 []o A particles o because the CSTR is the condition of maximum mixedness and the feed to the PFR is therefore well mixed.1th the given RTD. so the mass balance for components A and B inside the . similar to case C in Figure 8. We model the action of the stirrer as shearing the fluid A stream into small. so the different portions of the RTD in the PFR section have identical compositions. RTDs for the two configurations are equal. if we inject tracer with the A stream or the B stream.21: Adding two liquid-phase feed streams to a stirred tank. Ottino provides a well-illustrated discussion of the detailed results of fluid shear [20.2 in which the CSTR-PFR arrangement had lower conversion than the PFR-CSTR. This conclusion is consistent with the calculation performed in Example 8. We assume these "particles" of component A move rapidly about the reactor with the fluid flow. nor is it segregated. Figure 8.. Now we model the micromixing.3 Limits of Reactor Mixing 465 A B A fluid Figure 8.

Residence-time distribution measures the reactor macro mixing. these particles remain at the iniet feed conditions for a large time. The initial conditions follow from the fact that the particles are initially formed from the pure A feed stream.1 ~R C< [(X 0= Q2CBf .Q2CB - ~ [ex Jo kCACBp(G)dtJ + kCACB] r' 1 + ex (8. respectively. -' CA)S .10 aod 100 11m.53 simultaneously with Equations 8. The particle size governs the tnicromixing. Small particles have large mass-transfer rates and eqUilibrate with the continuous phase and the particles in the reactor with different ages leadlng to the case of maximum mixedness. Orthogonal collocation is described briefly in Appendix A. which is excellent. First the residence-time distribution corresponds to a perfect CSTR regardless of particle size.2 We compute the total A and B concentration . For r = 111m.0 mm.54) We use orthogonal collocation on z = e/(l + 8) to solve Equations 8. We see that if the stirrer is able to produce A particles of 1. the A and B concentrations in the particles rapidly change from the feed values to the continuous phase values as they spend time in the reactor.2: Mass-transfer and kinetic parameters for micromixing problem. They are washed out of the reactor before they cao produce hardly aoy reaction rate.0 11m or less. operates as ao ideally mixed CSTR. Therefore. Particles of intermediate size then describe the reactors in intermediate states of mixing.23 provides a detailed look inside the particles for r ~ 1.kCACBV CB)S . then the reactor is essentially in the state of maximum mixedness. Large particles have small mass-transfer rates and do not exchange much material with the continuous phase nor therefore with particles of other ages. Figure 8. which leads to essentially zero rate of reaction. we wTite the overall.3 Limits of Reactor Mixing 467 Parameter k kmA kmB Value 1 1. and the total surface area aod rate of mass traosfer are small.4 ~ Units L/mo!·tnin cm/tnin cm/tnin mol/L mol/L min \ in the effluent by sumtning over both particle and continuous phases IX CAj CBj Ql/QZ e = VR/(Ql + Qz) 1 1 1 CAt ~ -l-IX.oo 1 em ~ -1cB(e)p(8)de + -1-CB 0 +a 10 Table 8. The variables CA and CB are the continuous-phase concentrations of A aod B. if the A particles are larger than about 1. the particles are 100 times larger. This case corresponds to essentially complete segregation.Ql J" o f: kCACBP(e)de + kCACB] cB(e)p(e)de . For r = 100 11m.(8)p(e)de + c:---1-1-c + ex Jo + ex +lX IX J. This equilibration with the continuous phase is rapid because the particles are small. This case corresponds to segregated flow. Figure 8. This case is close to maximum mixedness. equivalently.54124]. The kinetic and mass-transfer parameters are given in Table 8. the total surface area aod rate of mass transfer are therefore large. this example is instructive for two reasons. To determine the A aod B concentrations in the continuous phase. particles is given by V~.22 shows CA (e) for particle sizes ranging from 0.466 Mixing in Chemical Reactors 8.67 x 10-4 1. Summarizing.1 11m to 1.67 X 10. Segregated flow essentially reduces the reaction rate to zero because the A and B species cannot come into contact. aod kmA and kmB are the A and B mass-transfer coefficients. Gniy as 8 increases do they have time to communicate with the continuous phase.kCACBV CB(O) =0 (8.0 cm. V and S are the particle volume and area.roo c. then the reactor operates essentially as a segregated-flow reactor. ~ kmACCA V~C: ~ KmBCCB - We next study the effect of particle size. or. steady-state mass balances for both components () ~ Q1CAf - Ql fa" cA(e)p(e)de - Q2 CA .53) in which 8 is the time the particle has been in the reactor. At the other extreme.

2 t.. and Exercise 8.23: Particle concentrations of A and B versus particle age for three different-sized particles. In that case. o .-. r = 1 11m 1.6 r--------------------..3.24: Differentiab!e convex and concave functions.2 r = 10 11m CA r= IOO llm 1 .~~~~ 1 10 100 r (11m) 1000 10000 8.. d 2 j(x) <0 dx .4....4 limits of Reactor Performance O~~~~~·L-'~~~~·~'~~~UU_·. the simplest version pertains to functions having at least two derivatives..1 " 0.. . as shown In Figure 8. ~-- ! I I I o 2 4 e (min) 6 8 10 Figure 8..... and requires justification. segregated flow I--------------------------:::.22: Total concentration of A in the reactor effluent versus particle size........ 1.. . .4 ..1 - 8. ----.. maximum mixedness 0.5 j(X) convex j(X) concave ~ o 0. Nauman and Buffham [18] also provide a detailed discussion. .. The general result for the single reaction is Given a single reaction ..::===~--___j 0.. As presented In the Introductory calculus course...3... the highest (lowest) conversion for a given RTD is achieved by the segregated reactor and the lowest (highest) conversion is achieved by the maximally mixed reactor.4 - J x 0.-- To generalize the results of Examples 8. This nonobvious result is a significant generalization of the numerical Examples 8_2 and 8.... The argument presented next first appeared in Chauhan et al....4 Limits of Reactor Performance 4b9 0. the nth-order reaction-rate expression r = concave if n :s L Note that first-order rate functions are both convex and concave......2 and 8.8.~-----..i'~-" -.--. ....1 A Single Convex (Concave) Reaction Rate Figure 8....4 ------ ------------------------ 0...:-~=_=_ x __ ______________________________________ _ :::_:::_: Figure 8... is convex if n "....24 d 2 j(x) > 0 dx -..-------------------------------------. a function is convex (concave upward) if its second derivative is everywhere greater than or equal to zero. A function is concave (concave downward) if its second derivative is everywhere less than or equal to zero.ith convex (concave) reaction rate expression.--.468 Mixing in Chemical Reactors 8.. [51.2 - For example.6. j convex j concave 0..6 0...--.. we define convex and concave functions.--.--.3§ 0.

4 Limits of Reactor Performance 471 mix r(CB) . .26: Convex rate expression and the effect of mixing. CB and a...!lVA +r(cB)VB Cm Figure 8.---------.' Figure 8. . in which a is now the flmvrate fraction in tube A {X = -~.470 Mixing in Chemical Reactors 8.27.25. The rate of the mean rm is less than the mean of the rate Ys for convex reactions. We can track the mean concentration for the segregated case Cs by simply summing the molar flows for tubes A and B divided by the total flow Cs = aCA + (1.alce (8. which we may choose to mix or maintain segregated as material proceeds down their lengths.. . ... Osasl in which a is the volume fraction of element A = ~=ccc- VA VA + VB On the other hand. = ar(cA) + (1 a . . r(cA) Step 1.26.. the concentration is Cm ~ CAVA +CBVB V V ~ aCA + (1 ..25: Two volume elements before and after mixing.. . . . consider the two volume elements shown in Figure 8. we lower the overall reaction rate... Without loss of generality assume the stoichiometric coeffiCient for the lim· iting species is negative one.a)CB A + Step 2.1 -' . . so that the segregated rate per volume is Y.a)r(CB)...13). if'we mix the two volume elements..55) This result is the key to understanding what happens in the general reactor. the material balances for the segregated tubes are dCA B The total reaction rate per volume after mixing is therefore rm = r(cm) de = -r(cA). For constant density.. . . The opposite conclusion applies if we have a concave rate expression. Again assume a single reaction takes place and the reactionrate expression is a convex function of a single reactant species. CB(O) dCB = CBO As shown in Figure 8. Now consider two tubes as shown in Figure 8.. or in which e ~ V /Qf. Note that in this discussion CA and CB represent concentration of the same reactant species in volume elements or tubes A and B. .--- QA QA +QB . To start. rate of the mean (rm ) is less than the mean of the rate (rs ). de ~ -r(cB). . When the volume elements are segregated the total reaction rate Y s is simply Ys ~ r(c. if we mix the contents. In fact.. for all CA. this statement can be lakeo as the definition of convexity (Exercise 8.

57) all 0 CAD. 0. We continue in this fashion. and all feed concentrations and flowrates. we conclude dC m > dc s 0 = 0. Figure 8.. How do we know.28: Mean segregated and mixed concentrations versus 0. Note this pairing and mixing procedure does not affect the RTD.28.472 Mixing in Chemical Reactors 8.57.a)cBO (8. Let 01 be the first such crossing time... Equation 8.a)r(CB)].B. = _ ~~ = - [r(acA Cs 0= 0. cm 2 C s as shown in Figure 8. Our goal now is to show cm 2 c. Consider now the mixed case. and the lowest possible conversion.. we achieve the reactor of maximum mixedness in Figure 8. differential equation for We have assumed that em Cm. and replace the segregated tubes with mixed tubes that achieve lower conversion than the original system. Finally. gives + (1alcB)]. o e _.a)CRO)] 2 This argument and result apply equally well for all cso and a. CAD.C.dO' But this relationship contradicts the assumption that the C5 and em curves cross each other. [r(acAO + (1. for at leas! some small distance down the tube. If the tubes are mixed at some point. at 0 1 so the d:.4 Limits af Reactor Performance 473 Cm C ".. Therefore there can be no time 0 1 at which the curves cross and we conclude cm(O) 2 C5(O).a)cE.[ar(CA) + (1. We know at 0 = 0 dCm dO = _. mix them.16. Comparing the right-hand sides of these two differential equations and using the conveXIty of r(c). that the curves do not cross each other at some later time? Assume this crossing can happen as shown in Figure 8.a)r(CB)] = dO dcs If the initial derivatives have this relationship we know. make the same argument that we made in Step 2. consider a segregated reactor with arbitrary residence· time distribution as depicted in Figure 8. however. dO . and we establish a contradiction. for all reactor positions. C4 and CB have some well·defined values and Step 3. We select any pair of tubes.. Cs Figure 8.. -[ar(CA) + (l ."'". Cs (0) = aCAO + (I .16. CBO and a. . At 0 1..28.56 gives 01 [ar(cA) + (1 . then the material halance for the concentration after that point is (8.56) The differential equation for still applies and Equation 8.55 dc s dO = . 0 = We also can write a differential equation for c. by simply differentiating Equation 8.. averaging the two segregated tubes produces C s . Cs = aCA + (l . curves crossing at 8 1 is a contradiction. and after we pairwise mix all the segregated tubes with mixed tubes.a)r(cB)].27: Two tubes before and after mixing the entering feed. the mixed feed tube produces Cm. which we may call 0 = 0.

Then we find any minima in the inverse rate function and construct CSTRs from those values until we intersect the inverse rate curve.83. Examining the plotted I /r function in Figure 8. We present next an example that dispels this notion. optimal sequence to achieve 95% conversion is PFR-CSTR-PFR.3 0. Notice these are also the places where dO /r(c)) / de ~ O.30.30. Levenspiel [17] shows how to find the optimal reactor configuration for this type of example.2 The General Case Figure 8.35 0.30.30: Inverse of reaction rate versus concentration. We can calculate the sizes of the reactors as follows.25. we nse PFRs. We next find the pOint where dr(c) / de ~ O. This example is based on one presented by Glasser. For a feed concentration of 5. which gives ~ 1 /5c 2 + 0.4 Limits of Reactor Performance 475 0.4. we see the optimal configuration is a PFR-CSTR-PFR. In the remaining sections of the curve where the inverse rate is a decreasing function of concentration. and then discuss what is known about the limits of reactor performance.58) which is plotted in Figure 8.29: Reaction rate versus concentration of limiting reactant. find the comhination of CSTRs and PFRs that achieve 95% conversion with the smallest total reactor volume.29.4: Optimal is neither segregated nor maximally mixed Consider the rate expression ric) Solution As we mentioned in Chapter 4. Setling the derivative of Equation 8. One might expect that the limits of reactor mixing determine directly the limits of reactor performance for more general kinetic schemes as well as the single convex or concave rate expression of the last section. this configuration is sketched in Figure 8.15 0. 0 1 2 3 4 5 6 c 8.25 r(e) 10 8 6 1 ric) '-" r\l~--~)~01 0.05 0''-(")-~J2 8} 4 2 a 0 1 2 3 c 4 5 6 0 Figure 8.58 to zero gives a quadratic equation with two roots: 0. Determine the RTD for this reactor configuration.05e (8.501 and 1. Unfortunately nature is more subtle. What conversion is achieved in a segregated reactor with this RTD? What conversion is achieved in a maximally mixed reactor with this RTD'! . We know from the problem statement that Co ~ 5.474 Mixing in Chemical Reactors 8.2 0. rate expression is neither convex nor concave. the smallest volume can be achieved with a series combination of CSTRs and PFRs. First we plot the inverse of the rate as shown in Figure 8. We choose the one corresponding to the minimum in 1 /r.1 0. Example 8. Hildebrandt and Godorr [111. C3 ~ 0.

We should also note that by considering all reactor combinations. however. that the extreme pOints of the attainable region boundary are made up entirely of plug-flow reactor trajectories. we can write the RTD immediately p(O)= ~2exp[ 0-(~12+03)]H(0-(81+03» which is plotted in Figure 8. Given a feed stream of known composition and a set of chemical reactions with known rate expressions.416 Mixing in Chemical Reactors 8. They showed. yields. it seems conceptually difficult to formulate a procedure by which we can calculate this set. recent research progress has been made.C2) which gives Cl = 3. for example. determine the set of all possible steady-state species concentrations that can be achieved by any combination of chemical reactors. The general problem can be defined in this way.07 0.95 segregated 0.-. To size the PFRs we simply use the PFR design equation and obtain 0.02 We see that these two mixing limits do not bound the performance of the actual reactor sequence with the given RTD.44. They also showed that combinations om ° ° 10 20 30 40 50 60 70 80 0 figuration. This set was proposed by Horn almost 40 years ago and named the attainable region [13]. = - I C1 Co r .03 0. Then the residence time is given by 02 = l/r(cZ)(cl .C -(-)dc = 3. is to build insight into the effects of mixing on reactor behavior under the constraint of a fixed. they are off by more than 20%. Figure 8.4.3 L With the RTD in hand. Feinberg and Hildebrandt [9] characterized the boundary of the attainable region. 1 03 = - I C3 Cl .30. these simple mixing models are insufficient.-75=- j t . we can compute both tbe maximally mixed.68 maximally mixed_~__0. reactor cases. In spite of the seeming difficulty in finding the attainable region. In this section we would like to provide a brief overview of what is known about finding sharp bounds on reactor performance. In fact. which is of importance because it bounds the posSible steady. 0 The attainable region. which is a considerable generalization from the single RTD that was considered in the mixing discussion in previous sections.d c = L07 r(el 1 These results are displayed in Figure 8. The results of those two calculations are summarized in the following table ". we cannot bound the performance of an actual reactor between the segregated and maximally mixed mixing limits. Glasser and Hildebrandt revived recent interest in this problem [12. such as the segregated reactor and the maximally mixed reactor. if we are mainly interested in determining bounds on achievable reactor states (conversions.'" optimal I ("'"m'" 0..04 0. Equation 8.4 Limits of Reactor Performance 411 0. we are also considering all possible residence-time distributions. and presumably measurable.08 0.state concentrations. As we have seen in Example 8. which we solve numerically. The primary use of the classical mixing models.94.31: Residence-time distribution for the optima! reactor con~ Next we find the concentration Cl such that l/r(c)) = 1/r(c2). This results in a cubic equation. if the rate expression is neither convex nor concave.95. Feinberg provides a nice summary overview of the history and many recent developments [7]. Equation 8. 11].06 0. Because we have a series of CSTRs and PFRs. and segregated.05 p(8) 0. etc).50. Even with a single reaction.. excellent. RTD. . Because the set is defined for all possible reactor combinations.

Grossmann and Westerberg provlde an overvie\-v in Chapter 1:1 of their text. Forecasting is always risky business.478 Mixing in Chemical Reactors 8.32. and certainly fan within the scope of the course in reactor analysis and design. :!Bieg\ef. Hybrid methods involving superstructures and geometric considerations have also been proposed [15].59) in the presence of the undesirable side reaction A--"'. Upper Saddle River.5 Examples in Which Mixing is Critical 479 of PFRs. Because the superstructure does not enumerate all possibilities. and what are called differential side-stream reactors (PFRs with addition of feed along the side of the tube). and almost all industrial designs would fall into this category. 8. D (8. CSTRS and mixing.2 Manousiouthakis has recently proposed an infinite dimensional state-space approach (IDEAS) that requires only PFRs. CSTRs. Prentice Hall PTR. which is modeled as a small CSTR feeding a second CSTR (bottom two reactors). we would like to close the chapter by presenting a few more chemical mechanisms for which reactor mixing can playa critical role. If we wish to allow separation as well as chemical reaction. and optimizing an objective function among the possible reactors [14]. researchers have proposed computational procedures to approximate the attainable region and solve reactor synthesis problems.5 Examples in Which Mixing is Critical Returning to the topiC of mixing. Figure 8. S'ystt'ma1ic tvtclflods of Chemica! Pmass Dl!slgn. linear optimization problems. In addition to properties and conceptual characterization of the attainable region. it seems likely that new and highly useful results on pure reactor and reactor-separator synthesis problems will be forthcoming. A full analysiS of the convergence properties of the finite dimensional approximation is not yet available. we can anticipate problems. Example 8. To understand what happens in this situation. These ideas and results may have immediate industrial impact.60) If the rate of the second degradation reaction is fast compared to the rate of mixing of the two feed streams. C (8. 1997. Component A is assnmed to be the limiting .5: Mixing two liquid-phase streams in a stirred tank A classic mixing problem arises when we must bring two liquid-phase feed streams together to perform the second-order reaction A+ B --"'. The advantage of this approach is that one solves only convex.32: Imperfect mixing (top reactor) leads to formation of an A-rich zone. provide the means to find all of the attainable region extreme pOints. The CSTR equivalence principle of Chapter 4 is an example of the strikingly simple and general results that recently have been achieved for the reactor-separator synthesis problem 18). Some computational methods are based on finding the boundary of the attainable region using the reactor types that characterize the attainable region extreme paints. New Jersey. consider the mixing model depicted in Figure 8. Some ofthese are based on proposing a superstructure of reactor types and numbers. but given the rapid pace of recenl progress. the solution may not be close to the true attainable region. A person skilled in reactor design may be able to choose reactor numbers and types well and overcome this general difficulty on specific reaction networks of interest. The disadvantage is the problems are infinite dimensional and require a finite dimensional approximation for calculation. but the approach shows promise on numerical examples [4). then the problem switches from a pure reactor synthesis problem to a reactorseparator synthesis problem.

For large Qr. CA2.klCACBe We can solve numerically the two equations for the two unknowns CA.(QI (QI + Qr)CAI + QrCA2 . ideally mixed CSTR. to vary the degree of mixing.59 and 8.(Ql + Q2)CA2 (k 1 cA2CB2 + k2C~2)VR2 Q.Q. o 0= QICA[ . We assume the density of this liquid-phase system is constant. we expect the two-reactor mixing model to approach the single. and the sizes of the two reactors. respectively.-CA2 . the conversion and yield are the usual quantities of interest in competing parallel reactions of the type given in Reactions 8.min min min min Solution The steady-state mass balance for the Single.60.3. The concentration of C in the outflow is determined from the change in the concentration of E.3: Reactor and kinetic parameters for feed-mixing example. CBl. well-mixed CSTR is ° ° ~ Q.32.5 Examples in Which Mixing is Critical 481 Parameter kl k2 Value 1 n e.1 Units min 1 L/mol. well-mixed reactor.(Ql + (k I cA2 CB1)VR2 . we model the formation of an A-rich zone near the feed entry point. + Qr)CBl + QrCB2 ~ kicA1CBl VRJ Reactor 2: I 0= (QI + Qy)CAl . we write mass balances for each reactor. Using this relationship and the defined parameters gives for conversion aod yield.klCACBVR e.(Q.(klCACB + k2c. Let CAl. CB.(1 + a)cE . reactant It is added at a low flowrate to a CSTR that contains an excess of reactant R In the top figure we depict the ideal-mixing case in which the rate of mixing is arbitrarily fast compared to the rate of eitber reaction.(Ql + Q2)cB . As discussed in Chapter 4. calculate XA and Yc versus Qr for the two-reactor mixing model shown in Figure 8. + Q2)CA . Discuss whether or not the residencetime distribution is a reliable indicator for problems with yield in the imperfectly mixed reactor.)CB2 o ~ Q2CBj + (QI + Qr)cBI- Q. Qr.480 Mixing in Chemical Reactors 8.(l + a)cA .(le ~ cBf . CB2 be the unknown A and B concentrations in the tworeactors. VRI and VR2.(kIcAICBI + k2C~I)VR' and compare the result to the single.(kICACB + k2C~)VR ~ QZCBj .CB2 . The mass balances are Reactor 1: Given the parameters and rate constants in Table 8. So in the bottom figure. and define the overall conversion of reactant A and yield of desired product C as follows: ° ° ~ aCAf . But this ideal mixing may be impossihle to achieve if the reaction rates are reasonably large. Then calculate the residence-time distribution P(()) for tracer injected with the A feed stream for the two models. + e2 QrlQ2 Defining the following parameters a~Q1/Q2 I p ~ varies allows us to write these as Table 8.CAf . For the two-reactor system. ~ VR1/Q2 e2~VR2IQ2 e~VRIQ2 ~ 2 2 1 2 3 0. This small CSTR exchanges mass with a larger reactor that contains the excess of reactant R We can vary the recycle flowrate between the two CSTRs.

that at low values of Qr. The result is shown in Figure 8. the A has been irreversibly converted to an undesired product D. Next we compute the outcome of injecting a unit step change in a tracer in the A feed stream. which corresponds to the poorest mixing and lowest yield.6 Yc 0.8 1 Figure 8. ideal CSTR. and fix the problem at the reactor. the conversion in the two-reactor system is higher than the single.4 0.4 P 0. It deserves careful study because it builds intuition and leads us to ask good questions when confronted with more complex cases. 0.PCA2 . we can design a separation system to remove the unreacted A and recycle it. and the single CSTR and two-reactor cases have similar yields and step responses. With low yield. well-mixed reactor. It is important to diagnose the low yield as a reactor mixing problem.33: Conversion of reactant A for single. 1 1 CSTR 0. If the conversion is low.6 XA r r r I 1 CSTR Figures 8. in a 2'CSTR mixing model.33 and 8. Qr = Q2. when P = 0.091 ~ al (1 + al because we inject tracer in only one of the two feed streams.482 Mixing in Chemical Reactors 8.4 2-CSTR mixing model 0.{k 1cA2CB2 + k2c:b)82 0= CBI + (a + P)CBl . Low yield is a qualitatively different problem than low conversion. o This example is one of the classic sets of reactions in which mixing has a significant impact on the reactor performance and the product yield. and as a function of internal flowrate. When p is reasonably large. We solve the transient CSTR balances and calculate the tracer concentration at the outlet. however. the tracer concentration reaches only Cis = 0. At the end of this chapter and also in Chapter 9 we show how to use this step response to determine the best value of p to model the mixing. which corresponds to poor nllxing at the feed location.(a + p)CAl + PCA2 . Notice.2 P Figure 8.(klCA2CB282 0. in a 2'CSTR mixing modeL .5 Examples in Which Mixing is Critical 483 We can summarize this case using the previously defined variables as four equations in four unknowns 2 0= aCAI .34: Yield of desired product C for single. the step test does reliably indicate the poor mixing.8 0.(l + aleB2 . this can be done analytically or numerically. Because the tracer does not take part in any reactions. however. or use it as feed in a second reactor.8 i- 0= -(a + P)CBl + PCB2 .35. A yield loss cannot be recovered by downstream processing. and as a function of internal flowrate.34 show the yield and conversion for the two cases as a function of Qy. We see the familiar single CSTR step response. ideal CSTR. Notice in all three step responses.PCB2 . The raw material is lost and cannot be recovered. p ~ QrIQ2. the yield changes from more than 90% to less than 15%.klCAlCB18 1 0= {a + P)CAl . In fact.6 0.(klCAlCBl + k2CAl)8l 1 t- 2-CSTR mixing model 0.{l + a)CA2 . For the two-reactor mixing model. p ~ QrIQ2. The conversion is not adversely affected by the poor mixing.2 I - ° i I a 0.2 0.

the B is largely converted to C and the yield is again low.484 Mixing in Chemical Reactors 8.5 varies Table 8. Approximately how large can the dispersion number be before the advantages of the PFR over the CSTR are lost? Solution The steady-state mass balances for components A and Bare dCA d 2cA v dz .61) (8. it is desirable to keep the average B concentration in the reactor low.1 0.08 0. which is undesirable in this case. and 2-CSTR mixing model with p = 0.07 0. dz 2 = RA V For example.63) (8. The B concentration is low near the tube entrance. ideal CSTR. = k. dz 2 = RB h B.66) .5 Examples in Which Mixing is Critical 485 0. If we make the tube too long. Because the flUid is a liquid. = k2Ci The second reaction can represent the first step in a polymerization process of species B.65) (8.0. however. and increases to its maximum value at the tube exit if we choose the right length or residence time. Calculate the steady-state conversion of A and yield of B versus PFR length for the kinetic and reactor parameters in Table 8. D A' = DBI = D/. yield is adversely affected by mixing. L should offer higher yield.4.D.03 0.04 0. r.05 0. because it maintains the same B concentration everywhere in the reactor. A PFR and we have assumed the dispersion numbers of both species are the same.01 5 10 Parameter kl k2 Value 1 1 1 0 1 1 CSTR P CAj Cnj Units min 1 L/mol·min mol/L m/min m m 2 /min =1 p=O v I D/ 0. we assume the velocity is constant.CA r2 (8. [23J discuss similar issues that arise in more complex polymerization reaction engineering problems. Intuitively the CSTR is a bad choice. Because the second reaction is second order in B.64) dCB d 2 cB dz .02 0. C. but achieve high B concentration near the reactor exit to maximize the production rate. Villa et al. to avoid yield losses.D. In this case.06 c. What is an appropriate reactor length to maximize yield of B? Study the effect of dispersion.62) in wbich 2B -"'.6: Maximizing yield in dispersed plug flow Consider the following two liquid-phase reactions in which B is the desired product A (8. Example 8.4: Parameters for the dispersed PFR example.35: Step response for single. 15 t (min) 20 25 30 Figure 8. We use Danckwerts boundary conditions for botb species z=O (8.09 0.

1 0.6 Summary and Notation 487 j ~ (A.2 0. For the kinetic parameters chosen. the conversion and yield are 0.25 XA 0. The derivation of the RTD of the CSTR also illustrated the following general principle: given an event with constant probability of occurrence. The RTD gives a rough measure of the flow pattern in the reactor. and therefore different performances.L. l r .001 0. We added a dispersion term to the PFR equations to model the spread of the RTD ohserved in actual tubular reactors.35 0..05 0.36..35 0.25 0. for example.1 0..5 8. and the plug-flow assumption of the PFR. the CSTR yield can be hnproved by about 8% with a PFR.79..3 7 show the conversion of A and yield of Bversus tube length for a tube designed to maximize the yield of B.4 0.15 x (m) Figure 8.3 0. ...05 D 0.001 ~ o~-~~~~~~~~~~--~--~--~-~ o 0. D ~ Figures 8._. and because the flowrate is con- stant.. We see that the dispersion number must be kept less than about 0. We computed the full.35 0. . the conversion of A increases. Given the concentrations. 0._-'-_-'-_.0 0. pulse test and (ideaiized) impulse test.36 and 8. We defined the residence-time distribution (RID) of a reactor..4 .05 0.l.2 0.....:D~~~1~00~0~========::::::::::::~7~~==..2 D ~ 0.1 0.. As the length increases above this value..1 0.36: Conversion of reactant A versus reactor length for different disperSion numbers.87....3 t~_. and showed how to measure the RTD "With simple tracer experiments such as the step test.L.486 Mixing in Chemical Reactors 8. but it does not determine completely the reactor performance.. and displayed the evolution of the concentration profile after a step change in the feed concentration. Indeed.5 m is appropriate. We introduced the delta function to describe this arbitrarily narrow !lTD.-JL---. .6 Summary x (m) In this cbapter we generalized the two flow assumptions of the idealized reactor models: the perfect mixing assumption of the batch reactor and CSTR.9 D ~ YB 0.85 D D ~ 1.15 0. may have identical RTDs.-__ D ~ 0.L_-'-_-L.45 0. We showed the CSTR has an exponential RTD... The PFR witb D ~ 0. 0 1.15 0. reactors with different flow patterns.001 achieves YB ~ 0.. but the yield of B drops rapidly..... to choose the time of the next reaction in the stochastic simulations of Chapter 4. the time until the next occurrence of the event is distributed as a decreasing exponential function..75 L--. defeating tbe main purpose of using a PFR. This principle was used.1 to maintain this advantage in yield...95 r--~ 0. A tube length of about 0.25 0.5 0.45 0. the high-dispersion PFR is essentially a CSTR.0 ~ 1000 0.1 0. B).. Figure 8.37: Yield of desired product B versus reactor length for different dispersion numbers.. transient behavior of the dispersed plug-flow model. As shown in Figure 8. .---l o 0.. and achieves YB ~ 0... The residence-time distribution of the PFR was shown to be arbitrarily sharp because all molecules spend identical times in the PFR.3 0.4 0._..

24 delta or impulse function. By contrast. RTD probability that a molecule spends time zero to e in the reactor.25 8(x) e A T e gamma funcHon of n. If we consider separation as well as reaction. J~ 0' p (0') d 0' time-to·go before molecule exits reactor e VRIQJ . rather general results such as the CSTR equivalence principle of Chapter 4 have recently been discovered. For more general reaction networks. convex reaction-rate expression. we showed that formation of a poorly mixed zone can lead to significant yield losses. Coo D effluent concentration in RTD measurement feed concentration concentration of species j concentration in a maximally mixed flow model concentration in a segregated flow model concentration boundary condition in maximum mixedness lllOdel dimensionless dispersion number. Next we discussed two contrasting cases in which mixing plays a critical role. Finally. Equation 8. D = DIT liz DA DJ{ molecular diffusivity dispersion coeffiCient for species j dispersion coefficient error function.18. Equa- tion 8. These topics seem ripe for further research progress. CFD methods may be used to validate simpler mixing models. We also showed that the mixing limits bound the possible reactor behavior for the case of a single. and Figure 8.2 residence time of tracer molecule in reactor mean residence time. Notation Cf Cj Cm C. simple mixing models may continue to play important roles in reactor analysis. the mixing limits do not bound the reactor performance. These validated. recent research on the attainable region has started to shed light on the possible reactor performance.7 and 8. the recent progress in the area of computational fluid dynamics (CFD) gives us reason to believe that direct solution of the equations of motion for the fluid will be a tractable approach for designing reactors and evaluating their performance !ll.488 Mixing in Chemical Reactors 8. We showed how a physical process such as mass transfer between a continuous phase and a particle phase can approach segregated flow for large particles (small mass-transfer rates) and can approach maximum mixedness for small particles (high mass-transfer rates).x) f(n) yield of species j reactor length variable incomplete gamma function of order n and argument x. Equations 8. in this example the reactor should be designed to approacb segregated flow. Equation 8. The two limits are segregated flow and maximum mixedness. In the mixing of two liquid reactants.6 Summary and Notation 489 We then examined the limits of reactor mixing consistent with a given RTD. Pe vl/DA volumetric flmvrate feed volumetric flowrate Qf r r particle radius in mixing model reaction rate of (Single) reaction production fate of species j RJ v VR xJ YJ fluid axial velocity reactor volume molar conversion of component j z y(n.17 mass~transfer coefficient tubular reactor length number of CSTRs in a mixing model probability that a molecule spends time e to e + de in reactor. good mixing leads to yield losses. design and optimization. For the general reaction network. for the kinetics of the second example. however. Equation 8. It seems reasonable to expect the classical RTD methods and simple flow models to complement the computationally intensive CFD methods.32 Dl erf(x) H(x) kmj I n pie) p(e) Pe Q Heaviside or unit step function. integrated form of the RTD Peelet number.

2: RTD to configuration Consider the feed and effluent tracer concentrations from a pulse test sho'NIl in Figure 8. Draw the simplest reactor configuration consisting of PFRs and CSTRs similar to those shown in Figure 8. Be sure to mark the mean of the RTD on your sketches. PFR-CSTR? or Exercise 8. r ~ ke/(l + ke) CSTR~PFR Which configuration achieves higher conversion for these cases. Determine T = VRIQJ for each reactor in your configuration.60 i 10 20 30 40 50 70 80 90 time (min) Figure 8. The total flow into each configuration is the same and each individual reactor has the same volume.7 Exerci ses Exercise 8..6 0. Is this configuration unique? Why or why not'? ~ fro Cf(t')P(t - t')dt' Assume the feed tracer concentration is not initially zero but some steady concentration Co.1 'b 0° 0 °0 --~ CSTR PFR E ~ concentration. Is each of the answers in A-f nnique? r = kc.38. at time zero the feed is abruptly changed to a new steady value cf '* Co.(I) Exercise 8. (t) ~ e..7 Exercises 491 --1 CSTR A ~ -1 -1 PFR PFR r CSTR 1. Equation 8.38: Reactor configurations subjected to a step test in tracer o .14 for the following reaction rates Exercise 8.4: Deviation variables Consider again the RTD relationship.3 e.2 1 B - H D CSTR CSTR t--- . 8.1: Configuration to RTD Make a qualitative sketch of the integrated form of the residence-time distribution P(G) by considering a step change in tracer concentration in each of the reactor configurations shown in Figure 8.2 600'0 COo ~00 0: 00 'b o 0 oP0 co 0 i °° o°c¢" d' '" I f F 0 ~ o i 0 0 oQ:1:b~O& i i i Figure 8.~ 0 <1 ~ 'il u 0 u <1) . '" <1 <1) 0.38 that you would use to model the reactor based on these test data. Consider a dimensionless deviation variable for effluent tracer concentration. (t )=(:<J cf Co . As in the step-response experiment. c.3: More equivalent RTDs with different conversions Prepare a plot like that in Figure 8. -{j ~ P '" " u k? 00' -10 o 0 0 °000 0.8 0 00:0 ° o 00 0° 00 feed effluent -° 0° 0 C 0.39.490 Mixing in Chemical Reactors 8.39: Tracer concentrations in the feed and effluent streams versus time.

6 33.7: RTD for laminar-flow reactor Calculate the residence-time distribution (RTD) for a tubular reactor undergoing steady.0133 0.40). A . he continues. in which v is the axial velocity and R is the tube radius.3. this mixing stuff doesn't make any sense. if the rate expression has the form (b) Establish that rm "r.4 48. e (min) 0 3. But you also told us in Chapter 4 that for these same kinetics. Displaying impeccable logic. Exercise 8.keoT and show the maximum mixedness case for the three-CSTR RTD is described by dx de = K-2 C 27x2 (.492 Mixing in Chemical Reactors 8. K . dcldx ~ 0 for x ~ 00 (8.0026 0.34Ibmol/ft 3 ) the reaction Exercise 8. Under the feed conditions (c Ai ~ 0. r kc 2 . irreversible reaction with a concave rate expression.5: Predicting performance The residence-time distribution is tabulated below for a flow reactor that has a feed rate of 16.67 for a range of K values and plol ((0) versus K.elT.0010 0. If we turn on a stirrer and suddenly mix the contenls of the PER (shov.V IQ and express the residence-time distribution for three equal-sized C5TRs in series.2 ft 3 /min.0061 II (min) 37. if the rate expression has the form Exercise 8.0627 0.1 at the temperature of the reactor. Exercise 8. "Assume we have a PFR operating at steady state at some overall conversion. Tables 1 and 2 in Zwietering [28].0 40.67) What "initial" value C (00) replaces the derivative boundary condition in Equation 8. 1. you claimed in class that the reactor's overall reaction rate wiH suddenly increase as well.0005 0.0604 0. Both of these statements cannot be true!" . x .8 0 0.0437 0. Plot this RID and compare to the RID for the plug-flow reactor. the RID is easily determined from the effluent concentration via P(e) ~ c.3 p(e) (min 1) 14.8: Other forms of the rate expression (a) Establish that 0.8 p(e) (min-I) Exercise 8. (t) satisfy? (c) Compare your results above with Zwietering's results for this problem. Define dimensionless variables (""'C/Co. laminar flow (Hagen-Poiseuille flow). Deterrn:ine the effluent concentration of reactant A when the total reed to the reactor is 16.2. with feed concentration Co.677 Solve Equation 8.1 51. Namely.9: Mixing it up Another perplexed student shows up at your office hours and complains.2 25.B. what differential equation and initial condition does the deviation variable c. let T .0001 0.6: Three CSTRs and maximum mixedness (a) Given feed flowrate Q and total reactor volume V.0000 rm " r. 51] 2Qf vir) ~ rrR' [ 1- (r)'] R is essentially first order in the concentration of A and the rate constant is 0. The velocity profile for Hagen-Poiseuille flow is 12.7 44.0004 0.'I] in Figure 8.4 ILl pie) (min-I) e (min) 18.lJ + 9x2 + fix + 2 c ~ . the PER has a higher steady-state conversion than a CSTR of the same volume." You and he discuss the simple case of a Single.23 min.2 still apply in this situation. Why do you suppose ZWietering does not report values for the maximum mixedness case for K ~ 3 and K ~ 50? Did you experience any difficulties computing these cases? What typographical error do you find in the published tables? See also the results in Example 8.0120 0.(e) (b) For a constant-density CSTR. With n ::.9 29.0415 0.5 22.02 Ibmol/ft 3 and cBI ~ 2. p. "Look.7 Exercises 493 (a) Show that the results of Section 8.7 7. (b) Consider second-order kinetiCS.2 ft 3 /min.0259 0.

Solve Equation 8.073 0.L (b) Compare your calculated values to Figures 8.1.187]).085 0. and compare the numerical results to your sketches. respectively.40: Reactor before and after stirrer is started: overall reaction rates before and just after mixing are Yl and Yz. the RTD of the dispersed PFR can also be solved analytically.68. ° t (min) C2 / elf 5 10 15 20 -" .092 0. p.086 0. (h) Draw the same sketch if the reaction rate is convex. Compare your analytical solution to your previous calculation. Van Swaaij and Beenackers [27.10 and 8.27 subject to Equations 8. The other parameters are given in Table 8. e in Westerterp.093 0. ~ I. Make the sketches before you perform the computation if you want to check your understanding of the issues.41: Effluent concentration versus time after unit step change in the first reactor. For what values of the disperSion number afe the RIDs similar.27 subject to Equatious 8.2.3. A unit-step change in inert concentration in the feed is made to the first reactor.28.00 0.10: Rm for dispersed PFR with Danckwerts Bes (a) Calculate the RTD for the dispersed PFR using the Danckwerts boundary conditions by solving numerically Equation 8. Compare your result with that given by Otake and Kunigita I J9] (see also the diSCUSSion Consider the 2-CSTR reactor model of Example 8. What is the best value of p'l .085 OD87 0. 0 3 6 9 12 15 18 21 24 27 30 0. Which approach do you prefer? Exercise 8.085 0. You may wish to try the same approach used in solving Example 8. n ~ 1/2. (a) Write the mass balances for an inert component that is added to the Q! feed stream depicted in tbe bottom of Figure 832 and show (8.7 Exercises 495 0.5. solve Equations 8.1 0 0 0 0 0 0 0 0 0 !:L elf 0 before: rl after: rz o. Iterate until you have a pretty good fit to tbe data.11: Fitting a reactor model to a step response Exercise 8. Co ~ 1.31. and the outlet concentration is shown in Figure 8.494 Mixing in Chemical Reactors 8. These fig~ ures were calculated for the dispersed PFR with the Simplified boundary conditions given in Equations 8.088 (c) Calculate the exact results for the cases k ~ 1. and compare to tbe data.-I 0 I I I I Figure 8. (a) Resolve this seeming contradiction by drawing a sketch of the overall reaction rate and the effluent concentration of the reactor versus time from the time just before the stirrer is started until the reactor comes to its new steady state \\lith the stirrer running.68) in which CI and Cz are the inert component concentrations in the first and second reactors. respectively. (b) Adjust p by hand.042 0.41.11 for the same values of the dispersion number and mean residence time. 30 25 t (min) Figure 8. and for what values of dispersion number are the RTDs different? (c) Because the model is linear.28 using either Laplace transforms or an eigenfunction expansion.

Can you suggest a general result for ordering PFRs and CSTRs given a single. for every y (8. Notice m depeuds on the value of x. We say fix) is a convex function if In terms of the chemical reactor.7 Exercises 49/ (c) Predict the conversion of A and yield of C for this reactor. miXing the reactor contents lowers or does not change the overall production rate. the PFR should precede the CSTR to max:irn1ze conversion. Convexity can be defined to apply to functions that are not differentiable such as cases Band C in Figure 8. given a convex production rate expression. CPCP.42.~(x») (y - x) We now choose m to maximize the function (f(z) . Use the definition of convexity of f to show fry) . the results stated for convex rate expressIons apply to functions like r = en in which n 2:: I. (a) Consider now a series of four reactors: two CSTRs and two PFRs.2 that for a convex reaction·rate expression.R(a)" m(cW) .13: Convex and concave functions Convexity is a fundamental notion in the theory of optimization. PCCP. PCPC. then there exists a constant m such that ~ r c(8)p(IJ)de and use Equation 8. which has the lowest conversion for a convex rate expression. the mean production rate is greater than or equal to the production rate evaluated at the mean concentration. The reason is shown in Figure 8. Hint: Let IX flax + (J - a)y) 5 af(x) + (l - a)f(y).42: Some convex functions. arbitrary function c and probability density pW). For n ~ 2.12: Ordering PFRs and CSTRs Consider a series of CSTRs and PFRs with a single nth·order reaction taking place in which r(c) ~ f(x) U x A f(x) J B C ken We showed in Example 8..69) . we should be able to safely reason that the lowest con· version is achieved with series: CCPP.a). We know in this case the series CCPP should achieve the highest conversion because the rate expression is concave.69 to show R(cW)) . Given f is convex and some value of x. convex or concave rate expression? The authors are not aware of a result of this type in the research llterature. which may be only piecewise dif· ferentiable and not monotone. but also apply to rate expressions like that shown in Figure 8. all x. Rank order all other configurations from lowest to high· est conversion. x x Figure 8.70) Exercise 8.42. fh) What analogous inequality holds when R is concave? fry) .42.f(x»/(z -x) over z to establish the result. Is your order in the two parts completely reversed or not? If not.20.f(x) " m(y . and CCpp. Be careful with the ordering of PCCP and CPPC. (a) Establish the follOWing fact about convex functions. (a) Derive Jensen's inequality. fh) Find m graphically for each of the functions sketched in Figure 8. The CSTR-PFR configuration is a maximally mixed reactor. ex:$l In chemical reactor problems.o:.496 Mixing in Chemical Reactors 8. can you explain why not? (c) Can you guess the solution for a series of six reactors: three CSTRs and three PFRs? Note there are IS possible reactor configurations in this case.. differentiability is not required. Le. Hint: -without loss of generality assume x < y and consider an arbitrary z such that x < z s y.x). which has six possible configurations: PPCC.y. which will prove useful.j(x) " (f(Z~ ..14: Jensen's ineqUality Jensen's inequality [101 states that given a convex function R. Exercise 8. CPPC. Exercise 8. J: R (c(e») p(lJ)d8" R (fr~ c(e)p(IJ)dIJ) (8. because that is also a maximally mixed reactor. for every c( 0) Multiply by pro) and integrate to obtain Jensen's inequality. Which of these achieves higher conversion? (b) Repeat for n ~ 1/2.

ondon. 1961.498 Mixing in Chemical Reactors fCe) Bibliography [1] A. Chern. Chern. J. Chern. Ind. Chern. M.72) (a) Show the maximum is achieved for Le. Figure 8. Lakshmanan and L T.. Toward a theory of process synthesis. Burri. [11] D. Eng. 1964... 40(14):3181-3]94. 200!.. editor.43 is merely a representation of the true situation because the ab· seissa is meant to depict a set of (unctions. Accepted for publication in Ind. Set. N. second edition. F. Feinberg and P. V. 1971. Hildebrandt and D. [31 K. pages 293-302.1621. Bird. 1953. Sci. P. !IS] A.l. B. 2002. Eng.15: Optimization of a residence-time distribution function Let f be a concave function and p( e) a residence· time distribution with mean o~ max pte) r ep(e)dO (8. Adler. D. Universal properties of the attainable region. Eng. Sci. 2:1-13. pages 30-39. Chern.. Eng. l141 A. John Wiley & Sons. tb) What analogous optbnizatian problem can you solve when f is a convex function? 3Notice Figure 8. Ellison. ]991. 52(10):1637-1665. and E. S. UK. f(e}p(elde as the function pte) varies. Synthesis of optimal chemical reactor networks. New York. Eng. IDEAS approach to reactor network synthesis: Application to attainable region construction.. [6] P. General kinetic bounds on productivity and selectivity in reactor-separator systems of arbitrary design: l. [7] M. Chern. the maximum is achieved when the residence~time distribution is arbitrarily narrow at the specified mean value. Glasser. ]13j F. Eng. D. C. On optimum mixing in continuous homogeneous reactors.1994. 1996. [2] R. Res. Chern. l. Eng. Principles. Chauhan. J. Bakker. Eng. p. [5] S. 45:2161-2168. nO] W. Ind. and S. Eng. P. and R. Glasser. Eng.43 3 By examining Equation 8. Chern. A. second edition.. Transport Pl'!enomena. Pergamon Press. not a set of scalar values as in an ordinary graph.73. Res. 46(5/6):1361-1383. 1997.. Hint: Use Jensen's inequality and choose cW) ~ G to show for all pee) (8. 33:1136-1144. Bell. Design reactors via CFD. Attainable and non-attainable regions in chemical reaction tecbnique. Haidari. and E. Exercise 8. Sci. [4] 1.71) Consider maximizing f over all possible RTDs having the specified meao 0 J" 0 f(e)p(8)de (8. Sci. E. New York. In Proceedings of the third European Symposium on Chemical Reaction Engineering.. A note on boundary conditions far flow reactors. [8] M. Eng. Chern. J. H. Marshall. for what p(e) is this upper bound achieved? For what f is this maximum unique? See also Nauman and Buffham [18. [9] M. Horn. Houdas. The attainable region for segregated. 2002. W. Wilson.. maximum mixed and other reactor models. Res.. Danckwerts.. John Wiley & Sons. ]6(1-2):131-133. Kakossis and C. Feinberg and D. 2002. Stewart. Manousiouthakis. Optimal reactor design from a geometric viewpoint . Chern. ]12\ D.73) which provides an upper bound for the maximum that is independent of pte) as sketched in Figure 8. Hildebrandt. Biegler. An Introduction to PrObability Theory and lis Applications: Volume II.. Feinberg. A. . Set. 27:585-591. Contiuuous flow systems: Distribution of residence times. Chern. Prog. Hildebrandt. December 200l.1972. 35:1344-1353. Godorr. Feller. Bischoff. Eng. B. Synthesis of isothermal reactor· separator-recycle systems. Accepted for publication in Compu1. The attainable region and optimal reactor structures. and V.43: Value of f. Chern.1990. Ind. Ughtfoot. M.

1908. Eng. Sci. Chemical Reaction Engineering. Drawing quantitative conclusions from these prinCiples requires not only the models. Ray. or. R.and contact-time distributions. Reactor Design in Chemical Reaction and Reactor Engineering.. E. 1959. A. 1967... Parulekar and D. M. 9.. Sci.. In J. Ottino. Effects of imperfect mixing on low-density polyethylene reactor dynantics.. 9 Parameter Estimation for Reactor Models 30(11):1742-1754. [19[ T. solution of the material balauce requires knowledge of the fluid-phase concentrations. The velocity of reactions in gases moving through heated vessels and the effect of convection and diffusion. M. Use of residence. Varma. Levenspiel. A.]. [23J 124J [25) {26) [271 128) c. Adler. Weller. from Danckwerts to Bodenstein and Damkbhler. Analysis of axially dispersed systems "ith general boundary candilons -I. 1998. editors. S. Evaluating the model parameters therefore usually requires something outSide the scope of the theory: experimental data. 1958. New York. Beenackers. New York. Chern. Westerterp. Kagaku Kogaku.1 9. [17J O. John Wiley & Sons. van Swaaij. W. Langmuir. J. J. C. and W. temperature. Villa. AIChE]. John Wiley & Sons. we develop convenient and reliable tools that aid us in extracting as much information as possible from the expensive and time-consuming process of building experimental facilities. W. Eng. 1967. Whether the reaction is homogeneous or heterogeneous. J. O. Stewart. Chemical Reactor Design and Operation. these prinCiples are efficiently expressed as mathematical models. Shinnar.1. Included in the list of variables are reactor volume. [21] S. Cambridge. Kunigita. Amer.. and making careful measurements. Chern. however. and effluent (final) concentrations of one or more components. [22J R. H.500 Bibliography [16J J. j.. 22:1483-1501. H. 22:144. K. M. Sci. The kinematics of mixing: stretching. [181 E. N. Otake and E. third edition. Much of this book is devoted to constructing and explaining the fundamental principles governing chemical reactor behavior. Sci. V. 1999. chaos and transport. Chern. Chern. P. Diliora. 39(11):1571-1579. 1983. and reaction rates are generally speCified in terms of component concentrations. B. 1989. and A. 28:262-264. Chern. Langmuir as chemical engineer . Solution of boundary. Boffham. Formulation. Mixing in Continuous Flow Systems. but the values of the parameters in the models as well. M. Ramkrishna. Ed. many of the parameters in these models are not known or available beforehand. John Wiley & Sons. iniet flowrates. T. Inc. New York. Marcel Dekker. [20J J. designing experiments.. Soc. 1986. Cambridge University Press. pages 63-]50. Villadsen and W. 22:65-75.1984. Chern. In this chapter we consider the issues involved in gathering data and estimating parameters from data with the expressed purpose of identifying a chemical reactor modeL In the process. Zwietering. Carberry and A. New York. J. Eng. Eng. The degree of mixing in continuous flow systems. 11:1-15. As we bave seen. In most practical applications. iniet (initial) concentrations of one or more components. so . 1984. Weinstein and R. Concentration or molar flowrate are the dependent composition variables in the design equations. Nauman and B.. 1994.1 Experimental Methods Analytical Probes for Concentration Experimental design includes specifying what variables to measure and how best to measure them. Eng. Mlcrontixing effects in continuous chentical reactors.value problems by orthogonal collocation. 44(7):1646-1656.

Infrared spectrometers vary by the means used to discriminate wavelengths. Dispersive instruments use a monochrometer and a grating to select the particular wavelength of interest or range of wavelengths that are scanned. in particular stretching. the sample must be transparent in the spectral region of interest. the C. Often only a few components out of an entire mixture can be monitored. The signal is recorded as a function of the mirror displacement. For example. or the mass/charge ratio for an ionized molecule can be probed by mass spectroscopy. b is the path length and c is the concentration of the absorbing species. 500-4000 cm. 17J.1. the thermal conductivity can be probed by gas chromatography. such as tbe components in gasoline. the C-C stretching mode of ethylene at 1623 em 1. The first example can be applied to the mixture while the second example requires the components to be separated before measuring the thermal conductivity. and liquids that vaporize at temperatures less than about 3S(YC. the O-H stretching mode of methanol at 3682 cm-'. 23J. where it is separated into components. is the molar absorptivity. the spectrometer reports absorbance versus wavelength. column and detector. such as separating the constituents in air. The energies of these vibrational motions are in the infrared region of the electromagnetic spectrum (2. special probing techniques such as attenuated total internal reflectance can be use to monitor the concentration. 3. ti' The general problem is to collect enough of the proper information so that the experimental data can be tested against appropriate kinetic and process models. Figure 9. liqUid and solid samples. Fourier transform spectrometers use an interferometer with a movable mirror to modulate the light source before it is incident on the sample.5-20 pm. and then !lows to the detector.1 Experimental Methods 503 we focus here on the important methods for measuring component concentrations in liquid-phase and gas-phase mixtures. Infrared spectroscopy can be used as a continuous monitoring tecbnIque by passing the gas or liquid through the cell as it exits a flow reactor. Most gases. 18. The intensity of a particular wavelength of light passing through the sample is recorded and compared to the intensity recorded when the cell is empty or filled with an inert substance_ The infrared spectrometer reports the intensity difference as an absorbance. The gas/liquid sample to be analyzed is injected as a pulse input in the injector and it flows along with the carrier gas through the column. With proper calibration using stand«rd mixtures. the Vibrational frequency can be probed by infrared spectroscopy. Infrared spectroscopy can be used for heterogeneous samples and adsorbates on catalyst surfaces [11. can be analyzed by gas chromatography. Gas chromatography is the workhorse of routine analysis [14. the composition is measured onIy at discrete sample times. as well as complex hydrocarbon mixtures. a unique chemical or physical attribute of the molecule under study is measured. and all wavelengths are sampled simultaneously.pc- . A ~ cbe € in which A is the measured absorbance. the absorbance intensity can be converted into a concentration using Beer's law because the path length and molar absorptivity are fixed variables. If the sample cannot be made tbin enough to be transparent.1 presents a Simplified schematic of a gas chromatograph. Because infrared spectroscopy is an optical technique. You are likely to encounter these methods in research and analytical services laboratories. and it is necessary to select the ones that provide the most information about the progress of the reactions. The quantitative aspects are based on Beer's law. The liquid-phase or gas-phase sample is placed in a fixed-path-!ength celL For gases this cell may be simply a tube \'ith infrared transparent windows affixed to each end. More . 16J because it can be used for simple gases.502 Parameter Estimation {or Reactor Models 9. 21. In most situations. It consists of an injector. To determine concentration. Fxamples include the C-H asymmetric stretching mode of a methyl group (CH3) at 2962 cm 1.0 asymmetric stretching mode of carbon dioxide at 2350 cm-'. Infrared spectroscopy can be applied to gas. Infrared spectroscopy exploits the absorption of electromagnetic radiation associated with molecular motion of chemical bonds. each section has a separate temperature zone. the R branch of the C-O stretching mode of carbon monoxide at 2170 em. Gas chromatography can resolve concentrations in the range of parts per billion to tens of percent. The final example samples the mixture and performs the mass discrimination as part of the analytical technique. and after averaging mUltiple measurements and transforming the interferograms. bending and rocking motions. and it is both a qualitative technique to identify the presence of molecules and a quantitative technique that can establish the concentration of the molecules. and the c-o stretching mode of acetone at] 718 cm-'. Analytical methods require calibration and proper selection of conditions to ensure the components of interest can be detected and their amounts quantified.1) II 9.

1 Experimental Methods 505 1 / I pro~ess gas stream ho.~:'". A carrier gas is selected. The thermal conductivity detector then operates by converting the voltage drop change to a temperature change as the gas mixture flowing past the wire changes from carrier gas only to carrier gas plus sample. To vent Injector -~~I ~ gas stream \-'-. the components interact with the adsorbent to differing degrees and are eventually separated into pulses of almost pure components. typically He or N2. The wire cools as a carrier gas flows Over it and a steady wire temperature is established as a base-line response.504 Parameter Estimation for Reactor Models 9. that has a thermal conductivity significantly different from any component in the mixture. to the wire resistance. not the composition of the gas mixture. The flame ionization detector operates by blending the column effluent with th and burning it with air in an ionizing Harne. The gas to be analyzed flows past a special valve.2: Mass spectrometer schematic. the wire temperature changes. Separation of the injected sample is accomplished by column selection. A flame ionization detector works best with hydrocarbons. which is temperature dependent. The thermal conductivity can be related to the composition by calibrating with carrier gas and samples of known composition. such as high-pressure liquid chromatography. that permits a small fraction of the flowing gas stream to enter the vacuum chamber. V ~ JR. extend the basic chromatographic technique to very high molecular weight compounds or to thermally unstable compounds. The detector identifies the thermal conductivity of the gas mixture. Mass spectroscopy can be used to determine the partial pressure of gases in a mixture. The voltage drop across the wire and the current through the wire are related via Ohm's law. the greater the signaL The flame ionization detector does not identify the composition of the gas in the flame. If the gas flowing past the wire changes composition and therefore thermal conductivity. Figure 9. The greater the concentration of a particular component in the flame. narrow tube that is either filled with a packing such as a porous adsorbent. The detector is typically a thermal conductivity detector or a flame ionization detector. The mass spectrometer is maintained in a high-vacuum chamber at pressures less than 10. Figure 9.2 presents a simple schematic of a masS spectrometer for gas monitoring. The column is a long. The electrometer merely records the presence of ions. perhaps sorting them by boiling point or polarity. called a leak valve. The ions arc counted with an electrometer.. The thermal conductivity detector is just a heated wire that has a constant current passed through it. Leak valve u Quadrupole rods Detector --::- Column Carrier gas Detector To vacuum pump ~~--------~----------------~ Figure 9. or contains a thin coating of a high-molecular-weight liquid on the inner tube surface that acts as the adsorbent.. As the injected pulse travels along the column length.. The science of gas chromatography lies in designing and selecting a column that separates the constituents. Standard mixtures are used to establish the elution order and time of elution for gases in a particular column. .1: Gas chromatograph schematic.6 torr.~. cialized chromatography techniques.

In the simplest case. Electrons (typically 70 eV) emitted from a very hot filament collide mth molecules in the ionizer region. I Experimental Methods 507 A steady state is established in which the amount of gas leaked into the chamber is pumped away. The current ~enerated by M+ and all M.6 and L8. Accurate and large-volume positive displacement pumps are costly. collimate and impart a constant kinetic energy to the ions. and small centrifugal pumps many not be available to handle the reactants. the reactants must be preheated and injected into the reactor Mthout any reaction occurring prior to the reactor iniet In an isothermal batch reactor. Depending on the instantaneous dc Voltage. of mass less than M. the reactor contents must be heated rapidly to the deSired temperature and held at constant temperature. steady and nonpulsating /low may be difficult to realize for liquids in a laboratory setting.. As illustrated in Figures 1. volumetric flowrate and temperature are the dependent variables.506 Parameter Estimation for Reactor Models 9. A batch reactor. If the reaction has a low activation energy. The effect of temperature is most easily explored by operating the isothermal reactor at different temperatures. Unless steps are taken.+ neutral fragments J which the data were collected. focusing lens system. a batch reactor may be nothing more than a beaker with a magnet stir bar for agitation. fragments at the detector is used to establish the partial pressure of each component in the vacuum chamber. the experimental data are fit against a model for the reactor in . the mass spectrometer generates a continuous measure of the gas-phase composition. Most experimental reactors are small to ensure large surface-tovolume ratio for good temperature regulation. The reactor models presented in Chapters 4-7 were based on specific flow assumptions.2. For example. The ionizer and lens system produce. it is necessary to operate the reaction at elevated temperatures to achieve a measurable rate. lowering the temperature does not quench the reaction. one often tries to design experimental reactors and select reaction conditions that permit isothermal operation. If using a flow reactor. The spectrometer has four major components: ionizer. efficient mixing. Pumping speeds of different molecules may vary. a predictable relationship between the relative mole fractions in the flowing gas stream and the relative mole fractions in the vacuum chamber can be established [20J.1. and concentration. M. It may prove necessary to dilute the sample or add components to cause side reactions to quench a reaction. the choice of a batch reactor is made by pumping considerations. liquid-phase reactions. The ions received at the detector induce a current to flow in the detector. M + e~ M' + 2e. Experimental reactors must retain these flow al'tributes if we msh to use the simple material balances for estimating model parameters. 9J. quadrupole section and detector. and isolation of the contents from ambient gases are important factors. such as the well-mixed reactor and plug flow reactors. The spectrometer operates on the principal that ions can be steered into the detector or diverted away from the detector. Starting and quenching the reaction can present formidable challenges. This process becomes more challenging when the reactor is nOnisothermaL Therefore. The electron impact ionization process ionizes the molecule or atom M + e~ M' + 2e- and ionizes and fragments the molecule into a number of ions. flow and batch reactors with volumes of several hundred nanoliters have been fabricated out of silicon wafers and used for organic synthesis screening studies. Many times. 9. if a reaction has a large activation energy. As configured in Figure 9. Various ionlzation techniques can be used [14. the most common is electron impact ionization. 13. only ions "ith a single m/z follow a stable path to the detector. The mass spectrometer measures the partial pressures of the molecules inside the vacuum chamber. Ultimately.2 Experimental Reactors for Kinetics Studies Time and position are the independent variables. the reactions may continue in the samples mthdrawn from the reactor. The quadrupole rods have a varying direct current (dc) potential and oscillating radio frequency (rf) field applied across pairs of rods that causes all ions except those "ith a particular mass/charge (m/z) ratio to follow an unstable path and collide with the rods. The experiment should be designed to ensure that the reaction takes place only in the reactor and under the conditions assumed in the reactor model. The situation becomes more complex if temperature regulation. which is at a large negative potential (-2 to -4 kiloelectronvolt (keV)). CSTR or PFR could be used for liquid-phase studies. Continuous. and to minimize the chemical inventory needed to conduct the experiments. The relative ratios of the Me fragments can be used to identify the structure of M.

1) tion 9. Gas flows are routinely and reproducibly regulated with mass flow controllers or manually with metering valves and flow meters. we have Q ~ NRTIP. It is quite common in experimental studies to operate a PFR as a socalled differential reactor. The isothermal PFR is usually nothing more complicated than an open tube maintained at a constant temperature. Commercial experimental reactors that achieve ideal mixing are available. The batch reactor and CSTR are rarely used for gas-phase studies because it is difficult to achieve ideal mixing of gases.Yjv. and we measure both CA and CB. .4.3 produces Q ~ Qf + RJ I i Viri/VR Substituting the Q relation into Equation 9. this enables a wide range of volumetric flowrates to be examined as an experimental variable. if we have the Single reaction A . Special considerations must be given to the selection of experimental temperature and catalyst particle size to minimize (and hopefully eliminate) internal transport limitations on the catalytic reaction rate. The design achieves a suffiCiently high linear velocity across the catalyst bed and volumetric flowrate in the reactor vessel so that the contents of the fluid phase can be assumed to be well mixed. The high internal circulation rate and linear velocities ensure uniform temperatures in the reactor. Sampling is best accomplished by directing the gas effluent through a gas sampling valve so that precise samples can be injected automatically into a gas chromatograph. For example.3 and rearranging produces I i (Vi] . The differential reactor is modeled using a limiting form of the PFR material balance design equation and generates data of the form rate versus concentration.3) in which R jf and rif are the production rates and reaction rates evaluated at the feed conditions. These reactors either hold the catalyst in a basket arrangement and circulate the reacting gases through the bed or they spin the bed at a high revolution rate. Figure 1. we obtain two estimates of the single reaction rate and the least squares solution is equivalent to taking the average.4.2B.508 Parameter Estimation for Reactor Models 9. The next requirement is that the flow pattern in the reactor is accurately represented by the well-mixed or plug-flow assumption. The subsequent discussion applies to gas-phase reactants.)rif ~ { / (Cj . depending on how they are operated. and summing Equa- Heterogeneously catalyzed reactions.2) (9. and special seals permit operation at pressures exceeding 100 atm. Macroscopic fluid models are combined with microscopic transport models in the catalyst particles to describe how concentration changes with time and position in a catalytic reactor. which determines the right-hand side of Equation 9. These well-mixed reactors can be treated as batch or CSTR. Gas-phase reactions. One then solves a leastsquares problem to determine all of the rif' By varying cJf' it is possible to measure efficiently the reaction rates over a range of concentrations.2 illustrates the batch reactor internals required to control the temperature and withdraw samples. The PFR design equation is (9.1 Experimental Methods 509 Samples are typically withdrawn as aliquots at preset time intervals via syringe and saved for subsequent analysis. Rate versus concentration data may prove useful in certain kinetic parameter studies.S explores what happens if we neglect the volumetric fJowrate change_) If we assume an ideal gas. we can approximate the derivative with a finite difference formula (9. Different tube diameters and lengths are used to change the reactor volume as an experimental variable.4) Experimentally. The reactor of choice for gas-phase reactions is tbe plug-flow reactor. Now we account for the change in volnmetric flowrate (Exercise 9. using Equation 9. one measures c] in the effluent stream.cJf) R (9. If we assume the molar !lows change by only a small amount. A differential reactor is a PFR in which the reactants experience only a small (differential) change in the extents of the reactions. in which N is the total molar flow.

I Experimental Methods 511 A second type of reactor is the plug-flow. Figure 9. fixed-bed reactor. however. Much more detail and many more techniques can be found in specialized books (33. An experiment proceeds by adding the catalyst and evacuating the entire apparatus until nothing is adsorbed on tbe catalyst and the pressure is below the detection level of the pressure transducer. distribution of active sites. characterize and monitor catalysts and catalyst surfaces. Our intent here is to discuss some of the more common and routine techniques. The axial aspect ratio ensures that axial dispersion is minimaL The radial aspect ratio ensures that channeling does not occur.3 Characterizing Catalysts and Surfaces Figure 9.10 The Reynolds number ensures turbulent flow and with it effective radial mixing. The experiment continues by closing Valve 1. 9. When these three dimensionless conditions are satisfied. The apparatus includes a pressure transducer for accurately measuring the pressure.1. The catalyst might be a metal dispersed on an inert carrier. or a zeolite (a crystalline and highly porous oxide).parameters that influence the catalytic activity.3 presents a schematic view of the essential components of a chemisorption apparatus. and then closing Valve 2. The catalyst is placed in Vessel 2 and this section of the apparatus is maiutained at constant temperature.27. and the pressure before and after admitting gas to Vessel 2. The velocity profiles are complex. A catalyst functions through the highiy specific interactions the active sites have with the reactants. Using an equation of state. The experimentalist is typically interested in the catalyst composition. Valve 1 is closed. one can determine the moles of adsorbed gas. a vacuum pump for establishing a low base pressure and removing all adsorbates from the catalyst. Dt > A wide variety of analytical probes are used to study. The equipment is straightforward. The pressure in Vessel 1 is noted before and after opening Valve L This sequence of filling and expanding is repeated until there is no change in the pressure upon expanding the gas into Vessel 2. presence of poisons/impurities after the catalyst has been used. The pressure is noted after reaching a steady value. with a flat ve[ocity prufile. Previous studies of transport in porous media have led to characteristic groups that gauge how well the Iluid flow can be characterized as plug-like [8]. This procedure generates the number of moles adsorbed as a function . 12J. a polycrystalline or amorphous mixture of metal oxides. The reacting gas must move through the reactor as a plug. such as the ideal gas law.5/0 Parameter Estimation for Reactor Models 9. Channeling refers to the situation in which the fluid close to the reactor walls travels faster than the fluid at the center of the tuhe. Valve 2 is opened. one can usually model the reactor as a PFR. the temperatures and volumes of Vessels 1 and 2. structure of the catalyst. Active site densities can be determined using adsorption isotherms as discussed in Chapter 5. and number of active sites . and gas is admitted to Vessel L Valve 2 is closed. These groups and their values include the particle Reynolds number Pressure transducer '\ To pump Valve 2 Vessell / Valve 1 the axial aspect ratio - L >30 Dp - From gas manifold Vessel 2 / Catalyst sample and the radial aspect ratio Dp . Valve 1 is then opened and gas both expands into Vessel 2 and adsorbs on the catalyst. and the pressure and temperature of Vessel I are recorded.32. a gas mixing manifold for introducing the adsorbate gas. and broad generalizations should be used with caution [28J.3: Volumetric chemisorption apparatus. opening Valve 2 and refilling Vessell. and two vessels of knowu volumes (V] and Vz) separated by a valve.

The heavier element produces a different intensity image. The two large questions we must address are: what model structure is appropriate to describe the reacting system of interest. provided the material is crystalline. and is in sufficiently large particles (~ 10 nm) and has a diffraction pattern that is not obscured by the support. 9. and. and analyzing the ions generated in the plasma with a mass spectrometer. X-ray photoelectron spectroscopy can be used to follow oxidation or reduction of the catalyst as well as identify the presence of elements. injecting the liquid solution into an inductively coupled plasma. Since the binding energy of innershell electrons typically changes with the valence of the element. Secondary or primary electrons are ejected from the sample at energies that are characteristic of the elements.2 Data Modeling and Analysis Having provided the context in which we perform experiments and make measurements. This problem is also important. and how certain are we about these parameter values. As particle size decreases. and this measurement has very higb elemental sensitivity and is quantitative. It is routine to resolve less than 0. we focus finally on the second question. The binding energy is used to identify the elements present [22]. Generally. what model parameters best represent the data we have collected. A sufficient number of particles must he measured to give a statistically meaningful average particle size. This technique can be applied before and after reaction to determine if the particles have changed size during the reaction. As the electrons travel through the sample. Composition also can be determined in a nondestructive manner using electrons or soft X-ray sources to excite the core levels of the atoms.6 eV and AI Kc< at 1486. A given crystal has a unique diffraction pattern. Crystallinity and particle size can be determined by X-ray diffraction. Electrons from other energy states in the atom rapidly fill the vacancy. The particle size is determined by measuring the size of the image and working back from the magnification used in the experiment. it has been the central focus of Chapters 1-8. In this technique. and . which can be used to determine the number of active sites and the adsorption equilibrium constant. Supported metals and metal oxides generally adopt their native packing wben they form crystallites sufficiently large to define a struc· ture. This scattering process leads to contrast images being projected on a detector. so the energy of the ejected Auger electrons provide a means of identifying the elements that are present [10]. Indeed. a mono-energetic beam of high-energy electrons is directed through a thin. Auger electron spectroscopy uses a focused electron beam (3-5 keY) to eject an inner atomic level electron and create a singly ionized excited atom. X-ray diffraction is used to assess the presence of amorphous versus crystalline phases and changes in the relative amounts of different crystalline phases with thermal treatment.2 nm and image individual atoms. understanding the fundamental principles and models that explain the many kinds of chemical reaction and reactor behaviors is one of the essential goals of the chemical engineering education. such as Pt dispersed on Al203. The digestion technique is destructive.5-10 nm of the sample surface. how to estimate model parameters from data. The kinetic energy of the ejected electrons is mOnitored and related back to the binding energy. A diffraction pattern identifies the atomic packing structure.6 eV) to eject core-level electrons from the unknown sample. Particle size also can be established using transmission electron microscopy. The energy of this ejected electron is monitored with an Auger electron spectrometer. solid specimen ($ 200 nm). it is easy to discern a heavy element from lighter elements. X-ray photoelectron spectroscopy uses soft X-rays (Mg Kc< at 1253. Elements have discrete electron energy levels with distinguishable binding energies. Composition can be determined by dissolving the catalyst in a suitable liquid. If the binding energy of this third electron is less than the energy transferred to it. it is difficult to make these techniques quantitative. it is ejected as an Auger electron. we now turn to the issue of how we extract the information contained in the data.2 Data Modeling and Analysis 513 of pressure. The ejected electron tecbniques identify what is present. This phenomenon is termed line broadening and it can he used to determine particle size 133]. they interact with the atomic constituents of the sample and scatter. The first question has occupied us up to this pOint. The inductively coupled plasma source is highly efficient at ionizing the injected sample. having selected a structure. In this section.512 Parameter Estimation for Reactor Models 9. the diffraction pattern becomes diffuse and the peaks broaden. The energy released in this de-excitation electronic transition can be transferred to a third electron. These ejected electrons have a mean free path that is energy dependent and can only exlt the sample if they are formed in the top 0.

and variance.2 0. Figure 9.5 displays a multivariate normal for p. ellipsoids or hyperellipsoids in three or more dimensions) we examine the eigenvalues and eigenvectors of the P matrix.1 Probability and statistics provide one useful set of tools to model the uncertainty in experimental data. we let x be an up-vector and the generalization of the normal is To understand the geometry of lines of constant probability (ellipses in two dimensions. the resulting vector points in the same direction as v.514 Parameter Estimation for Reactor Models 9. The rescaling factor is known as the corresponding eigenvalue A of A. = . the probability density p(x) for random variable x is given by Equation 95. lines of constant probability in the multivariate normal are lines of constant Figure 9.4: Univariate normal with zero mean and unit variance.5. which plays a central role in analyzing data. For distributions in more than one variahle.5 4. We adopt the following notation to "Tile Equation 9.e.2 Eigenvalues and Eigenvectors pi x ) = I I ex [ --(x (2rr)nr/2IPI1/2 p 2 m)Tp-l(x-ml] An eigenvector of a matrix A is a nonzero vector v such that when multiplied by A. Therefore the eigenvalues and eigenvectors satisfy the relation Av = Av." Equivalently. 9.1 = [3. The normal or Gaussian distribution is ubiquitous in applications. The notation !PI denotes determinant of P.N(m. I..5 2. (Y2. Solving tbis problem led Gauss to invent the least-squares method in the late 1700s. 1821 [15J.P) v*O . It is appropriate to start with a brief review of the normal distribution. OA 0.3 0. and is given by p(x) x Figure 9. Theory of the Combination of Observations Least Subject 10 Errors.2 1 ( 1 (x-m)") (Y2 (9. We also can write for the random variable x vector and only its magnitude is rescaled. Gauss.4 shows the univariate normal with mean zero and unit variance. x . Gauss's summary of this effort more than 175 years ago remains valid today: One of the most important problems in the application of mathematics to the natural sciences is to choose the best of these many combinations.2.5 ] 2. It is characterized by its mean.5 more compactly which is read "the random variable x is distributed as a normal with mean m and variance (Y2.1 Review of the Normal Distribution pix) 0.5.0 As displayed in Figure 9. symmetric matrix.2 Data Modeling and Analysis 515 has a distinguished place in the history of science and engineering. the combination that yields values of the unknowns that are least subject to errors. Accurate prediction of the motions of the planets based on astronomical measurements was one of the early motivating problems of parameter estimation. in which the np·vector m is the mean and the np x np·matrix P is called the covariance matrix.5) The matrix P is a real.2. C F. m.j2rr(y2 exp . 9.-------------------.

If we want to put simple bounds on the ellipse.6 displays these results: the eigenvectors are aligned with the ellipse axes and the eigenvalues scale the lengths. Substituting ()(v for x in this expression yields «()(vT)A«()(v) Aii = (i. i) element of A ~ 1 Figure 9.6. Figure 9.75 0.3 least-Squares Estimation =b Using the fact that Av = AV for the eigenvector gives ()(2AV T V =b because the eigenvectors are of unit length. This loss of information means we can put different tangent ellipses of quite different shapes inside the same box. The lengths of the sides of the box that is tangent to the ellipse are proportional to the square root of the diagonal elements of A ~ 1 .5xf + 2(2. The eigenvalues show us how stretched the ellipse is in each eigenvector direction. then we draw a box around it as shown in Figure 9. 9.2 Data Modeling and Analysis 517 p(x) = exp (-l/Z (3.6. We normalize the eigenvectors so v T v = 2: vI = 1 The eigenvectors show us the orientation of the ellipse given by the normal distribution. we solve for ()( and obtain ()(=~A fb Consider again the problem of fitting a straight line to data Yi "'" 'mxi + b . Each eigenvector of A points along one of the axes of the ellipse. we calculate how far we can go in this direction until we hit the ellipse x T Ax = b.516 Parameter Estimation for Reactor Models 9. Consider the ellipse in the two-dimensional x coordinates given by the quadratic xTAx which is shown in Figure 9. Figure 9.0x~)) xTAx =b p(x) 1 0. Notice the box contains much more area than the corresponding ellipse and we have lost the correlation between the elements of x. The size of the bounding box is given by length of ith side = ~bAii in which =b If we march along a vector x pointing in the eigenvector v direction.2.5 0.6: The geometry of quadratic form x T Ax = b.5: Multivariate normal for np = 2.5)X1X2 + 4.Z5 o Xl ----2 X2 ~-b-A:1- ---.

7. for models linear in the parameters. and so on.8 and produce a parameter estimate. but structural error is not accounted for correctly in thIs way. n p . The distribution of measurement errors creates a 1Notice the model slructure is usually in error also. . That can be shown to be the chi-square probability function [5].7) The best estimate of IJ in a least-squares sense is given by (9. nd and nd is the number of data points. ex. variables other than x may be required to predict y.8) a formula that you have used often. and the confidence level. a likely outcome if we differentiate the conceutration data to obtain the rate.9) We do not expect the best fit line to pass throngh all the data points. Example 9.518 Parameter Estimation for Reactor Models 9. we also can examine the distribution of parameter values given the observed measurements corrupted by the measurement errors. then (9. In fact. one may pose instead model discrimination tests to choose between competing models with different structures [30. so we modify the model to account for measurement error ' (9.. To make the estimation problem linear. Given the number of estimated parameters. we transform the data by taking the logarithm of Equation 9." We merely compute the size of the ellipse containing a given probability of the multivariate normal.6) The X 2 distribution is tabulated in many statistics handbooks [4] and is available in many computing environments...als.p) y =xe in which the parameters to be estimated are placed in the e vector in which tbe mean is the least-squares estbnate provided above and the covariance is P = (T2(X T X)-1 and the Y vector and X matrix are given by We also can calculate the parameter "confidence inte.1: Estimating the rate constant and activation energy Assume a reaction rate has been measured at several different temperatures in the range 300 K . We model the measurement error as a normal distribution with mean 0 and variance (T2 (9. Estbnate the pre-exponential factor and activation energy of the rate constant. Figure 9.g.7 shows a typical experiment. we can show the parameters are also distributed as a normal distribution IJ . e. 311· in which ko is the pre-exponential factor and E is the activation energy. the true relationship between y and x may be nonlinear. To illustrate the ideas we examine a classic problem: how to estimate the pre-exponential factor and activation energy of a rate constant. Solution We first model the rate (rate constant) as k = koexp(-E/T) (9. If the structure is in serIous doubt.10) in which e is a random variable. Imagine we create replicate datasets by drawing measurement errors e from the distribution given in Equation 9. 500 K. However.. i = 1. For each dataset we apply Equation 9.2 Data Modeling and Analysis 519 in which Yi is the measurement at Xi.10 Ink = lnko E/T . Notice that tbe measurement of the rate constant is somewhat noisy. Usiog matrix vector notation. we can "'Tite the equation for all the data as distribution of parameter estimates. The procedure outlined here lumps structural error into e as well.. and quantify your uncertainty in the estimated parameters.N(e. T .. scaled by the gas constant.

JL-_-L_ _. . .11) . Figure 9.8.05 I- • • ..0024 0.. (9.---".1 Ik 2. We construct the 95% confidence interval from Equation 9.----.9: Several replicate data sets. Ink versus liT.. OJ X X (I}..9 I r I~ 2.95) ~ 5. . In this problem np ~ 2 and ()( ~ 0.._--' 0.0022 0.15 f0 0 ~~~~~~~-1.55 L-..L. To generate these data we assume the model is correct and suppress the units of the parameter values Inko ~ 1. 0.I l o • • I o 2. 0.8 to estimate the parameter using the transformed model.85 f.. ..0032 0._.. .. Fignre 9. Ink versus liT.85 I1.. so Xi ~ liT.-...--. and we obtain from a statistics table X 2 (2.00220.6 f- ~ ~ 0 • • ~ + o x o .--"....0034 I/T(K-! ) 300 450 500 Figure 9.9 shows several more experiments. so we plot (I} ~ Figure 9. I ..r~ ~.'----.0026 0.61- The transformed data are shown in Fignre 9._-..520 Parameter Estimation for Reactor Models 92 Data Modelin9 and Analysis 521 o.. Figure 9.0026 0.55 L-..----.25 I2.8: Transformed data set..65 I0.L.N(O..8 L---1~ o x - : • + • o 9 ..001.95 I1..·~~~~~~~~~ -~-~r '~~~----'II~~~~~~··~~······~·T . .002 0.L..8 I0.L_ _L. o 0 o o o o o - e .85 '--'.0028 0.75 f0..003 0.J~_-L_ _. ..L.---"... To quantify the uncertainty.002 0. Then we can plot the distrihution of parameters. and Yi ~ In k i .. _ __L_ _ _L -_ ___L_ _ __'____' 350 400 T(K) .99 ~._. .-~ ..OOl) o o 0- - I .--'. imagine we replicate the experiment.--.. D 2 f1.65 I0. T...3 .0028 0... Notice the points are clustering in an elliptical shape.0) 0.L_---l 0...003 0._-L_-1_ _. • • o v • • ~....7: Measured rate constant at several temperatures.--~.O.2 I2..----r. Ink 0.L_.95.~ . E ~ 100 The measurements of In k are corrupted with normally distributed errors baving variance 0.99.9 I1. 2..0024 0..75 fInk 0. 0..--'. 0.9.0032 0... I ..7 I0.--'...~~-~.0034 I/T(K-!) We apply Eqnation 9... I .I-----.001 " T T /'- < 5. Each experiment that we perform allows us to estimate the slope and intercept.8 I0. 0.• A o o o o 0.10 shows the parameter estimates for 500 replicated experiments...7 I0.

522 Parameter Estimation for Reactor Models 9. reportiog confidence limits [ In k m ] E ~ [ 0. The ellipse is much more informative in this case.10.15 ] 60 But notice these limits are misleading.7 0.11 is fairly accurate with this many ran· dom experimental trials. we assumed the variance in the error was known to us.S5 0.10 and 9.75 Inkm 0.1 1. which iodicates Equation 9.-----r-----r----.10 does nol indicate the strong correlation between the parameters. Both Equations 9.2 Data Modeling and Analysis 523 IS0r----.11.15 1. Notice that the correlation between k m and E is much reduced compared to ko and E. (9. 24 out of the 500 poiots.95 1.10: Distribution of estimated parameters. In fact. Consider the mean of the temperatures at which data were collected.8% of the estimated parameters. Next we show how a simple re-parameterization oflhe rate constant can reduce the parameter correlation.05 1.2.9 0.2 lnko Figure 9.r----. The rectangle in Figure 9. For this problem. o 9.-----r----.----~0 180 0 00 o 0 0 160 140 120 160 140 120 00 co 0 E 100 E 100 80 60 __ ____ ___ L_ _ _ _L-___ L_ _ 1 so 60 40 20L-__ 0.95 1 1.4 least Squares with Unknown Variance In the previous problem. Note one often sees parameters reported with plus/minus limits.05 ] 60 is an accurate representation of the true 95% confidence interval ellipse. But io Equation 9.9 0.8 0. or 4. ature for Tm. If we estimate E and k m in place of ko we obtain the results shown io Figure 9. and re·parameterize the rate constant as in Chapter 6. It is often the case that we do not know the measurement .15 1.2 lnko Figure 9.1 1.12) We may wish to consider the mean of temperature or inverse temper.05 1.12 are two parameter models and we can convert between them using ko ~ k m exp(EITm). one might report Tm ~ 300 + 500 ~ 400 K 2 [ lnko ] E ~ [ 1 ] + 100 - [ 0. 20 0. lie outside this ellipse.10.85 0.11: Reducing parameter correlation by centering the data.75 ] + 100 - [ 0. Here we let Tm be the mean temperature This ellipse is also shown in Figure 9.8 40 ~ 0 0 ~ ~ ~ _ _~ 0.12 we estimate the rate constant at the mean temperature in contrast to infinite temperature (lIT ~ 0) in Equation 9. In fact.

12 shows the values of X 2 and npF as a function of the number of data points for 95% confidence limits. n p .524 Parameter Estimation for Reactor Models 9. Notice that npF is much larger than X2 at small nd. and the cases of two and five estimated parameters. Although we should expect larger confidence intervals when nd is smail. 25 nd . Given the number of estimated parameters. Notice the number of data points. that may not be what happens with a particular dataset as we show with the next example. shows up in the confidence interval when the error variance is unknown.nd np. this effect is rather small. .95 14 - For the confidence intervals. The measurement errors are drawn from a normal distributed with zero mean and variance a 2 = lO<l. in engineering problems we often must contend with rew data points. the larger values of F show information loss (larger parameter uncertainty) when estimating the measurement variance is the estimate of the unknovm error variance.12 quantifies this effect. nd. nd. i! can be shown that the distribution of parameters is a multivariate t-distribution (instead of a normal distribution) with its maximum at the least-squares estimate as before (9. That can be shown to be an F probability function [5]. we can see that if we have about 10 times as many data points as parameters. In the limit of large nd. nd > lOnp. we merely compute the size of the ellipse containing a given probability of the multivariate t -distribution.13 with two unknown parameters and only 10 data points. The F distribution is also tabulated in statistics bandbooks [4J and available in computing environments. With a small number of data pOints. from the data. because they are difficult or expensive to obtain.12: Values of x 2 and F versus the number of data points when estimating 2 and 5 parameters.lX) x 2 (nl-"cx) "p 2 .2 Data Modeling and Analysis 525 error variance. o 5 . Compute the best estimates of activation energy and mean rate constant and the 95% confidence intervals for the cases of known and unknown measurement variance. 0<.15) 15 30 35 40 45 50 Figure 9.2: Unknown measurement variance and few data points Consider the data shown in Figure 9. The sample variance also converges to the error variance in the limit of large number of data points. the F distribution converges to the X2 and the confidence intervals given in Equations 9. (9. Figure 9.14) ---------------------------------------- -' npF{np.14 are the same. The sample variance (9. 10 I . estimating the variance as well as the parameters from the data inflates significantly the confidence intervals.13) 0< = 0. 20 . the confidence level. On the other hand. Example 9. but must estimate it also from the data. I I I defines the confidence interval ellipse.9 and 9. and the number of data points. Figure 9. In this case.

127 0 9..._ _..75 0. the parameter estimation becomes a nonlinear optimization that must be solved numerically instead of a linear matrix inversion that can be solved analytically as in Equation 9.46).65 0.65 0. or such transformations simply do not exist.. o solution We calculate 0 from Equation 9. our 95% confidence interval for the unknown measurement variance case is smaller than the known measurement variance case. arithms to obtain a linear estimation problem.5 Nonlinear least Squares 180 160 E In the previous two examples we transformed the model by taking log- 140 120 100 80 0.0026 0.ih :s (0.85 0..---..13: Parameter estimates with only 10 data points.._---'_ _.. however.. For example.14: Confidence intervals with known (solid line) and unknown (dashed Ilne) error Variance.747 ] 153 ' S2 ~ 0.---r---. the problem becomes more challenging.0022 0.. 220 200 The two 95% confidence ellipses are shmvn in Figure 9. known variance o L-.l-_.9 and 9.6 0.L.001) (5.000454 Ink 0. the "approximate" confidence intervals remain very useful in nonlinear problems.L_ _.--.I.002 0..7 0..000454)(2)(4.95 1 Figure 9. the measurement error may be distributed normally in the original variables and the transformation may distort this distribution and subsequent parameter estimates and confidence intervals. As we will see..9 0...14.S. 0.8 0.003 0.8 lnkm 0... If we decide to treat the estimation problem using the nonlinear model..15 and obtain o o o o o o o O~ [0.99)....L... 8.-----..0028 I/nK-!) 0..l-_.O)TX T X(6 unknown variance 0. the confidence intervals become more difficult to compute.-. The numerical challenges for nonlinear models .0032 0. Can you resolve this apparent contradiction? What experiment can you propose in which we would find the confidence interval with esti· mated measurement variance to be about 50% larger than known measurement variance in agreement with the statistics in Figure 9.._---' (8 . As we will see. We obtain from a statistics table F(2.7 0.2..55 We construct the 95% confidence interval from Equations 9.75 ...461lO the two confidence intervals are given by (8 - 0) TX TX(6 .85 0.95) ~ 4.526 Parameter Estimation for Reactor Models 9.----. and they lose their strict probabilistic interpretation as "... Notice that although npF is 50% larger than X2 .0024 0.13 and the sample variance from Equation 9.6 0.14.2 Data Modeling and Analysis 527 0...-level confidence regions.--.0034 Figure 9. Moreover. O):s (0. In many situations we do not want to make such a transformation.

The system studied was the adsorption of H2 on a Pd catalyst with 5iOz support [25.nd-np.20) <1>(8) = 2:: (5'. . in which case H = 2X T X. <I> = 2:: (i'Hi .h(x" 8»2 min <I> (0) B These confidence intervals are correct only if the model is linear.528 Parameter Estimation for Reactor Models 9. 24J.JKsfCH. 1 + . It is known that the adsorption constant on the Pd is large. The intervals should be checked occasionally with Monte Carlo Simulations when the model is nonlinear.l We obtain the parameter estimates by solving the optimization problem subject to Equation 9.16. Consider the nonlinear model (9. eH = _ OT = [ . Both the catalyst and the support adsorb Hz so the adsorption isotherm model is Hz + 2Xp = 2H . Expanding in a secondorder multivariable Taylor series gives 1>(0) .4. we study an adsorption experiment.8 cmPJKiojCH2 cms-JI(SjCH.5 of Chapter 3).1>(8)+ ('\7<1»TIB.(j(8-8)+~ = 2H· Xs (0-B)THlo. We now consider the function <1>(8) near the optimum.CH') 2 (9.3: Fitting single and mnltiple adsorption experiments To illustrate the use of nonlinear models. The adsorption is assumed dissociative.l nd ~ (9.2 Uata Moaelmg ana Analysis can be addressed successfully in many reaction engineering problems if we employ high-quality numerical software. Call the solution to this problem B.o(O-8) (9. We can apply the methods of Section 5. Xp Hz + 2Xs i". which allows us to rearrange Equation 9.jCH. _ .16) and the least-squares objective nd in which S2 is again the sample variance S2 = 1 nd - 2:: (5'i _h(x"B»)2 = nd .17 to give (9.22 to determine the parameter estimates. (9.23) As the measure of fit to the data we choose a least-squares objective.18) Therefore.25) subject 10 Equation 9.19) .21) 1 + JKP 1>(8) np i".fKs cmS Cmp] (9...24) and propose the optimization problem min <1>(8) 8 (9.17) in which the gradient of the objective function is the vector of first derivatives of 1> with respect to the model parameters ('\7<1»} in which Xp represents a Pd vacant site and Xs represents a Si02 vacant site. so we may reduce the model to _ _ cms)KsfCH.(8-Bls2s2npF(np.6 and we use H as the A matrix.22) CH = emF + 1 + .fKs.e<) (9. lines of constant objective function are approximately quadratic functions as shown in Figure 9. and we have three parameters to estimate from data. We also can obtain order-of-magnitude confidence intervals using the relation (O-B)THI 0. We illustrate this check in Example 9. This quadratic approximation using the Hessian matrix evaluated at the optimum is accurate if we are in the neighborhood of the optimal parameter values.6 to derive the surface coverage of H atoms = oOj 31> and the Hessian of the objective function is the matrix of second derivalives 02 <1> Hkj = "BkOBj The gradient is zero at the optimum (see also Figure 3. . Example 9. + (9.

4 J (9.048] 1. .. • • "". [24. 60.1 32. but they have different meanings.3 40 Ie' s bi) 0 '-. ~ ~-------------------------- 50 bi) + ~ 45 ~ 13 . After the first adsorption experiment. because the adsorption experiments are per· formed after different reaction srudies have been performed..7 5.30..530 Parameter Estimation for Reactor Models 9. "li ~ •+ 0 A A + + + + • .081 4.> '" • • • 6'. and the catalyst regeneration step does not necessarily rerum the catalyst to exactly the same condition. compares favorably to the apparent saturation limit or total uptake value of 55 Ilmol/g reported by Natal·Santiago et a1.157].. For this catalyst sample.5 2.127 26. ... one might use the estimated values from a single adsorption experiment on a freshly prepared sample... The ad· ditional adsorption data are shown in Figure 9. Then the catalyst was regenerated with heat treatment. Notice because the reaction studies have introduced significant vari· ability in the data.16.2 Data Modeling and Analysis 531 5S 60 55 . Figure 9. Notice these are not replicate experiments. reaction srudies were per· formed with the catalyst sample. If the model is intended to represent the "average behavior" of the catalyst during tbe period it is used and regenerated. • + • 200 400 600 800 1000 1200 1400 CH2 (Pa) 35 30 o 100 200 300 400 500 600 700 800 900 0 CH2 (Pa) Figure 9..16: Model fit to all adsorption experiments.15 shows the best fit to a single adsorption experiment. and so on.~ ~ . 0 . The parameters and the 95% confidence intervals are given by the following parameter values and confidence intervals e~ A [JKS J emS emP ~ [0.. and a second adsorption experiment was per· formed. . followed by an additional reaction srudy. one would naturally use the estimated values from the many adsorption experiments.5 Ilmol!g.1 5: Model fit to a single adsorption experiment. Solution Figure 9.<. the confidence intervals for the parameter values are Significantly larger than those for the single adsorption experiment.. p.3 .. If the model is intended to represent the catalyst in its na· tive state.. ~ -.26) r jKs J l emS ~ 0895 0. Estimating the parameters by fitting all of these data simultaneously produces the solid line in Figure 9.4 (9.27) We can see that the model fit to the data is excellent. and the parameters are determined with fairly tight confidence intervals. I 6 and Tbe total coverage estimated here.1 1 ± emp [0. additional adsorption experiments are avail· able.. 50 45 40 35 • • '" '" <if. .4 J 8= A ± [0. Both sets of estimated parameters are valid.4 30. • *0 .

As shown in Figure 9. .2 Data MotJelmg ana AnalysIs 9. Imagine we wish to use a straightedge to draw a line on a piece of paper . We also must decide what small means when it comes to parameter uncertainty. and draw the line. let's build up some physical intuition about the problem we are addressing. a young child may reqoire rather large circles to bound the ex. compare data and model predictions. and require only the remaining small task of finding best parameter values and quantifying the parameter uncertainty.. we must assume the reader has had exposure to a course in basic statistics [41. To keep this discussion focused on parameter estimation for reactor models.17: Drawing a line through two points under measurement error: points far apart (A): points close together (B).. lay the straightedge next to the two points. Optimal experimental design generalizes this intuitive notion of how to construct the experiment to minimize the effect of the experimental errors on the estimated parameters. . It also can provide exper· . . Experimental design provides another option. and knows the differeuce between correlation and causation. <' A " . If we wish to ntlnimize the error in the constructed line. . for example.. . modify hypothesis. Optimal experimental design is an experimental design with the express purpose of making the parameter uncertainty as small as possible. Figure 9. We are not concerned at this point with embedding this step in the larger task described previously: deciding if our model structure is well chosen and the model assumptions are appropriate given the data. The main goal of our experimental design then is to choose experiments that make the uncertainty in the parameter estimates small. and repeat. B . the spread in slopes is large. . with the center or the circle representing the ideal location where we are supposed to mark the point. if the points are far apart. . iterative scientific method: hypothesize. ' . . .. our placement of the points is subject to small errors.17. But regardless of the current state of tecimology.. As shown in Figure 9. . ' . . . the measurement noise persists at some nonzero level. the spread in the slopes of the lines is small.6 Design of Experiments Experimental design is a large topic and we can only mention several of the important issues here. such as the one used to print this page.2. . The size of these circles tells us how much uncontrolled error exists in our experiment. A reasonably high-resolution printer. . . Before developing preCise. Now we ask the question: where on the line should we locate the two points to give us the smallest error in the line placement? Our experience and intuition probably lead us to guess rather qulckly that we want to spread the two points apart as far as possible. . In ali cases we face error in point placement and that error produces error in the final line. ' . We assume the reader understands the source of experimental error or nOise. An experienced draftsman or carpenter with steady hands may be able to measure carefully and place points within fairly small circles. If the pOints are close together.. Our goal at this point can be quite specific because we have already built the causal models of interest to us in Chapters 2-8. the small circle represents the regions in which we might actually mark a point. one option is to make the measurement error circles small. The goal of experimental design is to make this iterative learnlng process efficient. collect experimental data. We mark two points on the paper. may be able to locate a point to \Vithin 1/1200 inch. then a line with any slope is possible. . conSider that no matter how carefully we measure. . .532 Parameter tStimatlOn for KeaclOr MoaelS 9.. If the points are close enough that the circles overlap...17. Optimal experimental design. .. peeted errors.'' . The process of estimating parameters in reactor models is part of the classic. Improvements in measurement technology are always an attractive option. quantitative methods of experimental design. To make the problem more interesting and realistic.a task most of us face early in childhood.

If we add points at the right end of the line shown in Figure 9. The 500 K ellipse is therefgre stretched in the direction with negative correlation between In k m and E.18 shows the confidence intervals for three different choices of the next experiment.. 40 1000 K :' / . . This result is consistent perature. we must decrease the slope. we ask the natural question: where should we place the next experiment to obtain as much information about the parameters as possible? To answer this question we have to decide what information we seek.. T = 300 K. We explore one such nonobvious design issue in an industrial case study in Section 9. This result is exactly what we observed in Figure 9. we can guess immediately to put the next experiment at the upper or lower temperature limit if we want to know the activation energy. but with different orientations. T = SOO K. then that anchors somewhat the right end of the obtain less uncertainty in the slope. 1 Tm= 1/300+1/500 =37SK (9. we see that the next experiment should be placed at either T = 300 K or T = 500 K. Both values are equally good. the mean of inverse temperature.18: Effect of next measurement temperature on parameter confidence intervals. In this problem. So the 300 K ellipse is stretched in the direction with positive correlation between in k m and E. In this problem we chose Tm = 375 K.19 displays the size of the Sides of the box tangent to the 95% confidence ellipse as we vary the temperature of the next experiment.Q2 -0. / / / \ I I I I I ±E 0 -20 -40 • I -60 -80L-~ ____~____~____~____L __ _~~_ _~~ -0. If we are interested in the activation energy. Figure 9. spread the points as far apart as is practical. \ \ / / 20 . If we were able to violate our temperature limits and perform the next experiment at 1000 K. and to increase the value of In k m we must also increase the slope. but one is somewhat more important than the other.04 0. The choices of 30() K and 500 K produce two 95% confidence regions of the same size.534 Parameter Estimation for Reactor Models 9.2 Data Modeling ana AnalysIs 80 500 K'~_- _ __ --------"". to build some intuition about what to ask and expect of optimal experimental design. 300 K 60 ___ - ~/ -~'7- - ~ _ ~_ . imental designs in situations in which our intuition does not provide ready answers. motivated by Figure 9. figure 9.06 Figure 9. Figure 9. If experiments are either time consuming or expensive. We next explore what happens to our confidence intervals as we make different choices for the next experiment.02 0. One big advantage of the approach is that it is quantitative. Alternatively. we may want to know both of them.28) T :5 500 K If we are interested in the value of the rate constant at the mean tem- We cannot estimate two parameters unless we have at least two measurements. at the cost of slightly more uncertainty in the intercept.17 . and to increase the value of lnkm.3. and can only operate the experiment and collect rate data in the temperature range 300 K :5 with our physical intuition of drawing straight lines. With this background. Assume we have temperature constraints in our experimental system.13 (T = 300 K). but that is the main reason we studied this problem.17. we choose the first two measurements at the extreme limits of the temperatures. we now ask very specific questions and design experiments with precise goals. so.06 o ±lnkm 0. we may want to know only the intercept. if we add points at 500 K. we may want to know only the activation energy.19 shows we should place the next experiment at the mean temperature T = Tm· The results in this case are not too surprising. but would we know that the uncertainty would change from 82 to 72 when we did S07 If . For example. But for illustrative purposes.13. consider again the least-squares estimation of the mean rate constant and activation energy from measuring the rate constant at several temperatures. Le. then we have anchored somewhat the left end of the line in Figure 9. we see that we can achieve more certainty in the activation energy.

..01 o ±lnkm 0. Be aware. But replicating an experiment has another interesting and expected benefit...536 Parameter Estimation for Reactor Models 9.. To see this benefit. So replicating experiments provides another general avenue for increasing the confidence in the parameter estimates.. 0.. The second big advantage of this approach is that it is general and can be applied routinely in even complex experimental situations in which our intuition is not highly developed..' . ..045 -60 \ / 80 \ \ \ \ / / / / / 1 C/O \ / 78 \ / .. Figure 9. By replication we are confronted with the simple experimental fact that no matter how much care we take and how much effort we exert.E. What happens if we simply repeat the expertment and perform two sets of 10 measurements> The confidence region becomes smaller.. The confidence intervals continue to shrink as we replicate the experiments.046 0... Finally.. It increases our confidence in our conclusions.20: Uncertainty in activation energy E and rate constant lnk-m versus number of replicated experiments. and the interested reader can find a wealth of literature discussing its many aspects Ill.. We can answer the question of how many further experiments are required to meet a given bound on uncertainty.047 40 0.. we examine the impact of replicating experiments. 10 equally spaced temperatures per experiment. Assume now the experiments are inexpensive and we are not concerned about performing optimal experiments.0465 0. . that thls "brute force" .. however..i!J § ...... The irreproducibility of the experiment is one of the key motivations for the random variable descrtption of e in Equation 9.02 -0.8.03 Figure 9.... as shown in Figure 9... The differences among replicated experiments give us a direct quantitative measure of the ex· perimental error or noise.0475 / \ / 82 \ \ \ / / / / / 60 0... We also show the result of using 5 and 100 replicates of the experiment.6.0455 0.8 +1 20 ±E 0 -20 -40 \ / / o .14.2 Data Modeling and Analysis 84 ..a "" 76 +1 74 "'"" 0100 5 2 nsets = 1 72 300 350 400 T(K) 70 ' . that improvement is not large enough for our purposes. Replicate experiments. as discussed further in Exercise 9.a .. experimental design is a large topic.. but they are also due to our inability to control completely the experimental environment.' ....0445 450 500 -0.. These errors are partly due to our inability to measure accurately the outputs of interest.. . we would know further experiments are required. The main reason one replicates an experiment is to obtain a direct measure of the experimental reproducibility.' 0.. We simply make measurements at 10 evenly spaced temperatures as shown in Figure 9.20.01 0.13. the results of two experiments are never exactly the same. we examine again the least-squares problem for estimating km. . As we mentioned previously..03 -0.' .. If we perform one set of 10 measurements we achleve the confidence intervals we have shown previously in Figure 9.19: Uncertainty in activation energy E and rate constant In km versus next measurement temperature.02 0..

0 are the unknown model parameters. To get started. '.33) subject to Equations 9.36) in which x is the single material balance of interest. The Gauss-Newton approximation is not valid if the model solution is a highiy nonlinear function of the parameters. Note Xi is the only part of the objective function that depends on the model parameters. The differential equations make it much more expensive to evaluate the constraints while solving the optimization problem. or if the residuais are large and not randomiy distributed about zero at the optimal value of parameters. one may try numerical finite difference formulas to compute the Hessian. in the latter case one should question the model structure because the model does not well represent the data. e) = h(x) y Gauss-Newton approximation and sensitivities. Of course. OJ. 0) x(O) dx (9.2 Data Modeling and Analysis approach may be time consuming and expensive. Differentiating Equation 9. tbe second derivative of the model may be small compared to the first derivative.538 Parameter Enimation for Reactor Models 9. Xi = X(ti. we minimize this objective function to obtain our parameter estimates min¢(O) B (9.I (9.32 gives o¢ ae = - J 2" L (Xi i - xiJ '0 OXi (9. and y is the experimentally measurable quantity.35) lit = j(x.3 7) .ii is the experimental measurement at time Li. The major cbange is that the model constraint consists of nonlinear differential equations rather than linear or nonlinear algebraic equations as in the previous sections. for example. it may be necessary to include some of the initial conditions also as unknown parameters. ~ ax. reactor volume or length can take the place of time without changing the structure of the parameter estimation problem. We devote the rest of the chapter to developing methods for this problem.30) (9.31.7 Parameter Estimation with Differential Equation Models Now we turn to the single most important parameter estimation problem in chemical reactor modeling: determining reaction-rate constants given dynamic concentration measurements.34) v J Differentiating a second time gives the Hessian of the objective function. If the model fits the data well at the optimal value of parameters. As before. Alternatively. The first derivatives of the model solution Vvith respect to the model parameters are known as the model sensitivities. We can increase the efficiency of the optimizer if Two arguments support the Gauss-Newton approximation. 9.32) in which. we neglect the second term in Equation 9. in which case h(x) = x. however.35 to yield. (9. Computing a reliable finite difference approximation for a second derivative is not a trivial matter either. If the model is linear in the parameters. For the highiy noniinear case. and the step size should be carefUlly chosen to avoid amplifying the errors introduced by the finite numerical precision.29) (9. Xo is the initial condition. and Xi is the solution to the model at time ti. which we again use to construct approximate conffilence intervals (9. Recall the gradient of the objective function is the vector of first derivatives of ¢ with respect to the model parameters. Again.29-9. we consider a simple reactor model consisting of a single differential equation with a single experimentally measured quantity we provide an accurate gradient of the objective function. we define a least-squares objective to measure our fit to the data (9. s. the second derivatives are identically zero and the Gauss-Newton approximation is exact.31) = g(xo. As we see in several of the examples. In tbe GaussNewton approximation of the Hessian. Often t is time.I ~ vO. the residuals are small in magnitude and of different signs.2. The sum in tbe second term is then small. For simplicity let us assume here that x itself is measured. but in steady-state tubular PFRs.

21: Experimental measurement and best parameter fit for nth-order kinetic model. Compute Xi and Sij. These programs are generally more efficient if we provide the gradient in addition to the objective function.36 as Hkj = 22.5 00 0 - 0 0 o 0> o o 0 o o 0 oLL____ ~ ____ ~ ______L -____L __ _ _ _ 3 4 ~ o 1 2 5 t (min) We can express Equation 9. Caracotsios.I<P using Equations 9.38. If we define e to be the residual vector Cj\ 1 o o o o 00 o ~ 0 0 0 0 00 0 ~ ~q. rather than use finite difference formulas to obtain the sensitivities. If not converged. Calculate H using Equation 9.39. we can construct a fairly efficient parameter estimation method for differential equation models using standard software for solving nonlinear optimization and solving differential equations with sensitivities. kc~ . 0. we can summarize this relationship in matrix notation (9. This fact allows us to solve the model and sensitivity equations simultaneously with an ODE solver. go to ®- Given these expressions for the gradient and Hessian. these equations are derived in Appendix A. we can express Equation 9.40) @ Check convergence criteria. From the sensitivities and model solution. Stewart and S0renson developed this approach and produced an influential software code (GREG) for parameter estimation problems in chemical reaction engineering [6. (9. are usually controlled by an optimization package. solve the model and sensitiv- ity equations simultaneously given the current parameter values. 29. ® On convergence. The Hessian is normally needed only to calculate the confidence intervals after the optimal parameters are determined.4: Fitting reaction-rate constant and order We illustrate these methods on a classic reactor modeling problem: finding the rate constant and reaction order from isothermal concentration measurements in a batch reactor. nth order reaction A+ Bproducts. we can then calculate the gradient of the objective function and the Gauss-Newton approximation of the Hessian matrix.f'ikSij = 2 L S[iSij @ Evaluate <P and <.32 and 9.39) ® Update parameter values to minimize <P. set to current parameter values. Example 9. This step and the next If we write the last sum as a matrix multiplication. Calculate confidence intervals using Equa- e tion 9. Bard provides a readable account for further study on these issues [2J. 7].540 Parameter Estimation for Reactor Models 9.I<P = _2ST e (9. In terms of the sensitivities. ConSider an irreversible. Parameter estimation algorithm: CD Guess initial parameter values.38) Figure 9.2 Data Modeling and Analysis ~41 The sensitivities also can be described as the solution to a set of differential equations. r = ® Using an appropriate ODE solver. r "'" kc1.34 in matrix notation as <.19. Reliable and robust numerical optimization programs are available to find the optimal values of the parameters.

To treat this case.3 0.19 as we vary a.i CAO (9. usually the concentrations of several species are measured as well. determine the best values of the model parameters. Be aware that this conclusion may not be true for other nonlinear problems.542 Parameter Estimation for Reactor Models 9.41) (9. and then solve the nonlinear optimization problem in Equation 9.8< OJ 350 300 200 150 100 50 0 0 0. We also produce a value for the Hessiao for each dataset.22: Monte Carlo evaluation of confidence intervals. Notice the estimates are close to the correct values. Most reactor models of interest contain balances for several species and therefore consist of several differential equations. note that we had to solve 500 optimization problems to produce Figure 9. we consider the general dif- . We can generate a reasonable initial parameter set by guessing values and solving the model until the model simulation is at least on the same scale as the measurements.2 Data Modeling and Analysis taking place in a liquid-phase batch reactor containing a large excess of reactant B.1 0. Notice for all a values.471 = = ~:~~ ± r ~:!~ l 0. for each of these 500 datasets. The solution to the optimization problem and the approximate confidence intervals are given in Equation 9. Figure 9.2 0. which indicates the approximate confidence intervals given in Equation 9.01 to the model solution with the correct parameters. and we use the mean of these for H. approximately the correct number of parameter estimates fall within the corresponding ellipse.43. and we have fairly tight approximate confidence intervals. OJ '" . the points indicate the actual number of estimates that fall within each a-level ellipse for 500 replicated datasets.22. so we include it as a parameter to be estimated.19 are fairly reliable for this problem. a few Monte Carlo simulations in the final stages of a modeling stndy are well justified..43.22 shows the results of a Monte Carlo simulation study. But given the dramatiC increase in computational speed.6 0. it does not seem reasonable to assume we know CAO any more accurately than the other measurements. We provide this as the starting paint. Given the measured concentration of component A shown in Figure 9.43) The parameters that we used to generate the data in Figure 9. To determine the parameters in such models.S <<:I> 'Ul Given the data shown in Figure 9. We then compute the approximate confidence intervals using Equation 9.42) "d is 250 OJ c.19 with Equation 9.8 0.21.5 0. Solution The material balance for species A is dCA 500 450 400 . 0 Differential equation models with multiple measurements.7 0.33 using the least-squares objective as shown in Equation 9.39 for H.22.21 also are given in Equation 9.::.5 [0. and should be checked. for example. we solve the optimization prohlem to obtain the parameter estimates. In this study we generate 500 datasets by adding zero-mean measurement noise with variance 0 2 = 0. 00 = r c~o J r ~:~ J e l l k 0. The model therefore contains thtee unknown parameters OT = expected . Next we examine the quality of these approximate confidence intervals for this problem. This result is plotted as the points in Figure 9.052] (9.(O) = . This computational check is fairly expensive.32.4 0.9 dt = -kc.21.actnal 0 1 [k CAO n 1 Figure 9. Finally we calculate what fraction of these 500 estimates fall within each a-confidence region as defined by Equation 9.

x(O) ~ g(xo. The measured variables may differ in size from each other by orders of magnitude. Example 9.5: Fitting rate constants in hepatitis B virus model We illustrate using multiple measurements by revisiting the hepatitis B model introduced in Chapter 1 nucleotides cc!:!. Solution The reaction rales and production rates for Reactions 9.46.55) amino acids cc.r6 1 (9. To estimate the parameters we now solve rz n r4 rs r6 k. Xo are the initial conditions.)). ~ x(t. In weighted least squares.544 Parameter Estimation for Reactor Models 9.52) (9.r4 rl . and h. in which W is a symmetric. (9.'JA (9.).49) k 4cA suhject to Equations 9.48) Find rate constants k..54. however.25. As before we assume the nucleotides and amino acids are in large excess. - h(x(t.58) in which A is (ccDNA.::!?. 8)) (9.45) (9. . rcDNA and envelope protein measurements given in Figures 9.23-9. -k6 from the cccDNA. The elements on the diagonal are the weights aSSigned to each measurement type. The sensitivities are now a time-varying matrix. 8 are the unknown model parameters. B is rcDNA.Y2 .Ys .8) (9. We write mass balances for these components to produce three differential equations. ~ h(x(t. 30.r6 Y3 . he the residual vector at the ith sample time e. The reader should be aware that more general procedures are available for the multiple measurement case.53) (9. In this case. ~ (9.57) rcDNA+ y. it often does not make sense to sum the squares of the residuals. The simplest way to address this issue is to employ weighted least squares. we focus here on weighted least squares.Sf~~f We. positive-definite weighting matrix. and C is envelope. including the maximum likelihood method [29.56) (9. The influence a measurement has on the objective also would be influenced by the arbitrary choice of the units of the measurement.).46) V<l> ~ -2 I.!"A nucleotides + reDNA cccDNA envelope envelope - rcDNA eccDNA envelope degraded secreted or degraded secreted virus (9. kscc k6CB CC l~n ~ l r2 .54) (9.44) (9.2 Data Modeling and Analvsis 545 ferential equation model dx dt y ~ We can compute the gradient as before j(x. 8) ~ . we combine the different measurements by forming the weighted sum of the residuals. we again define a scalar objective function that measures our fit to the data. which is obviously undesirable.52 and 9.52-9. and y are the experimentally measurable quantities. For simpliCity of presentation. Let e.50) ' h(x) and the Gauss-Newton approximation of the Hessian is in which x is the vector of states that defines the reactor model.44-9.4 7) The objective function is then (9. When we have different measured quantities.51) Set. ~ H~2'STahf woh. and cccDNA catalyzes Reactions 9. x.s Li 'ax· DX T ' r t in which S. 31J.CA k2 CB k3 CA min<l>(O) B (9.57 are given by r. Usually W is chosen to be a diagonal matrix.

32 -1.08 18.025 1000 0.12 initial guess and estimated parameters fit to data.46 0.25: Envelope versus time for hepatitis B virus mode!. = e~ O.60 3.5 x lO~6 0.85 + 60 80 100 Binlt t (days) figure 9.91 18.30 -5. We choose 60 80 100 20 40 t (days) Figure 9.00 -0. rcDNA and envelope species.77 -0. initial guess and estimated parameters fit to data. of the measured value.23: Species cceDNA versus time for hepatitis B virus model.25 2.2 Data Modeling and Analysis 547 Z ~ Q U U U 60 50 40 30 20 10 0 0 I I I I I I I I I I I I I I The initial condition for the experiment is 0 0 0 0 0 0 °0 00 0 0 0 0 [c A CB Cc 0 ° 0 0° 0 r = [1 0 0 r and is assumed to be known precisely.59) 0 0 20 40 60 80 100 t (days) Figure 9. The measurement noise is assumed proportional to the size _ . When estimating several rate constants that may range in values hy several orders of magnitude.15 -0. This transformation often makes it easier for optimization software to find accurate solutions.32 (9. it often is useful to define the model parameters to be the logarithms of the rate constants. The data in Figures 9.17 1.30 1.239.60 0. The data are generated by assuming values for the rate constants and adding noise to the model solution.50 -(jAg 0. ko ~ 80 ~ (9.0 7.0 0.3 1.50 3. and automatically enforces the natural constraints that the rate constants are positive. intial guess and estimated parameters fit to data.3 0.60) CJ > c: 0" c '" 30000 25000 20000 15000 10000 5000 0 0 20 40 o o o o o ° o o o o o We generated a random initial guess for the parameters and produced the following optimal set of parameters and approximate confidence interval 0.25 were generated using the following rate constants 2.24: Species reDNA versus time for hepatitis B virus model. This weight matrix gives each measurement roughly the same relative weight in the objective function.8S 5.61) . Z Q u ~ ~ 600 500 400 300 200 100 0 0 I I I I I I 0 °0 0 0 0 0 ° 0 0 0 00 0 to reflect the relative sizes of the cccDNA.546 Parameter Estimation for Reactor Models 9.80 -Ll3 3.16 -5. (9.

'16.03 Vs ~ lOs V6 ~ -0. mated and perturbed parameter values. This outcome is not unusual when estimating many rate constants. 60 80 100 . We cannot graphically represent the six-dimensional confidence hyperelUpsoid as we did in Figure 9. V6 directions without greatly changing the value of the least-squares objective. consider the model solutions shown in Figures 9. the optimizer is able to find a good solUtion.03 -0. We may be satisfied with fitting the data in this way. We pursue model reduction next. which means we have little confidence in the estimated parameters in spite of the good fit to the data. Vs and V6. The eigenvalues are given in Equation 9. the other elements of this eigenvector are near zero. -0. esti~ . which is to examine the eigenvalues and eigenvectors. In spite of the poor initial guess. 0.23-9. But notice also in Equation 9. As you can see.03 (9. estimated and perturbed parameter values.00 The two smallest eigenvalues. are two orders of magnitude smaller than the largest four. also are given in Equation 9.. and the dashed lines correspond to the follOWing OJ p.62. Vs is aligned almost entirely in the (02. estimated and perturbed parameter values.62) 600 500 400 -< Z CO 300 u .2 Data Modeling and Analysis 549 The dashed lines in Figures 9. The Hessian is a 6 x 6 matrix for this problem because we have six parameters.1 x 106 lA x 106 2. These eigenvectors tell us that we can move the parameters a significant distance in the vs.25 show the solution of the model with the initial parameter values.27: Specles reDNA versus time for hepatitis B virus model.. As and .4 X 0.62 60 50 40 -< Z CO 30 u u 20 u 10 0 0 20 40 t (days) Figure 9.0 4 ) coordinate directions.03 0. The solid lines correspond to the estimated parameters in Equation 9. 0 '" 6 30000 25000 20000 15000 10000 5000 0 0 I I I 20 40 t 60 (days) 80 100 Figure 9.26: Species cceDNA versus time for hepatitis B virus model. which indicates we should remove about two degrees of freedom from the parameter estimation problem. The model solution using the optimal parameter estimates is shown by the solid lines in the figures.11 - 0. Os) coordinate directions. Similarly.28: Envelope versus time for hepatitis B virus model.61 that the confidence intervals are quite large. To demonstrate this fact.6!. The eigenvectors associated with these two eigenvalues.548 Parameter Estimation for Reactor Models 9. but we can do the next best thing. Notice we have an excellent fit to the data.6 in two dimensions. Notice that V6 is aligned almost entirely in the (83. we are not giving the optimizer a very good initial guess.28.26-9.00 0.\~ 1.0 X 108 4. but we may wish to reduce the model so that the estimated parameters have tighter confidence intervals. 200 100 0 0 20 40 60 80 100 t (days) Figure 9.

-:k. however. Notice that even though we have changed the parameters by a large amount. Similarly. we have a systematic approach to develop such understanding fairly qulckly when presented with a new problem of interest. ks We see again that the ratios of the rate constants k3/ k5 and k2/ k4 appear.06 0. the method of analysis can be used again but the model reduction may change. the least we should expect is that the computational and numerical procedures being employed are efficient and reliable. This study was conducted in collaboration with colleagues at Kodak and more details are available [26]. an excess of one reactant at the final time does not present a problem and the opportunity for improvement with better control is proportional to <I> ~ o 0. the payoff can be large. Please note that these particular conclusions depend on the particular vaiues of the rate constants we used to generate the data. 0 The complexity of this model probably precludes us from guessing quickly and intuitively that parameter pair k3 and k4 (or k3 and kl) cannot be determlued from the experimental data.00 0. Also the difference kl . We guess immediately that we should find the ratios of the rate constants k3/ I<s and k21 k4. if we multiply parameters k2 and k4 by some other constant.550 Parameter Estimation for Reactor Models 9. even if we use values of k3 and k4 that are off by an order of magnitude from the values used to generate the data.k::04~~ e We do not need to use the correct values of k3 and k4 to estimate <1>. To conclude this chapter we examine a small prototypical case study that illustrates some of the challenges we can expect when we model data from industrial facilities. because of the logarithmic transformation. k4 for this analysis to be valid.32 -1.64) and set k3 and k4 (or k3 and kl) to arbitrary values.66.66.46 1.14] -1. k6 (r.24] -1 00 .) ( k.!erent parameter values are chosen. the estimate of <I> is the same as in Equation 9. resulting in more efficient reactor operation. we must design a new experimen t to determine the values k3 and k4 if they are of interest.63. If very dif. A small quantity of one of the reactants often remains at the end of the reaction due to batch-to-batch variability in the purity and reactivity of the starting materials.k4 is the important parameter. if we multiply parameters k3 and ks by the same constant.65) CAs = 9.21 -0.12 <I>~ A 0. because they cannot be found using data shown in Figures 9. We also can gain insight by examining the steady-state solution to the model. we have not changed the simulations much at all. Note they cannot be completely arbitrary because we require k3 » k 1. and improved reactor designs. +k~4+k3) Notice in Equation'o---.26] [0.63) We added V6 to to obtain Equation 9. . more consistent and higher quality product. By estimating all parameters.58 to zero and solve for the steady-state concentrations we obtain 2 CAs _ = ~.70 [ -5.66) 2Note the zero solution is also a steady state. therefore. and we should estimate certain ratios and differences of rate constants instead of the rate constants themselves. When successful models are constructed. and are not compounding the difficulty of extracting models from the available data. In many chemical reactions. 2.. two main reactants are combined to yield a primary desired product.25.16 2. we do not change the fit to the data. not the individual values. In fact. to an excellent approximation k6 kJ.07 (9. k4. If we then perform parameter estimation with the reduced set of four parameters we obtain the tight confidence intervals shown iu Equation 9.47 4.56 -5. and.65 ± 0.08] (9. we do not change the fit to the data.12 [ -5. In other words. In some cases.23-9. and variability in the rate of side reactions.60 that k3 » k 1..3 An Industrial Case 5tudy parameters e= T [0..3 An industrial Case Study (I£<) (Is) (9. and analyzing the eigenvaiues and eigenvectors of the Hessian.. Reactor modeling in the industrial enviromnent is a challenging task. End-point problems. lfwe set the production rates given in Equation 9. Given the difficulty of the modeling challenge in the best industrial Circumstances. We therefore choose the model parameters to be (9.

perfect mixing.k2 CCCBlVR d(CDVR) _ k V dt . In these cases. The primary assumptions in this model are: isothermal operation. and reaction with impurities and inhibitors in the starting materiaL The initial conditions for the ODEs are: VR(O) The dihalogenated starting material (Al loses HX to the base (B) to form the mono-halogenated intermediate (e). negligible side reactions such as polymerization of the e and D. These materials may be difficult (impossible for practical purposes) to separate from the products. and therefore the unknown model parameters are given by = [kj k2 nAoJT. Modeling. a small amount of neither starting material is tolerable. an excess of one reactant is commonly added to ensure complete consumption of the other reactant. In some cases.552 Parameter Estimation {or Reactor Models 9. the parameter nAO is e chosen to be adjustahle to account for the unknown level of impurities . terial balances for the semi-batch reactor are dt dVR = Qf(t) d(CA VR) dt =-k1cACBVR d(CBVR) dt A+Bh C+B·HX c+Bh D+B·HX = QfCBf - (klCACB + k 2cCCBlVR d(CCVR) dt = (k1cACB . however.ZeB R (9. Although the initial A charged to the reactor is weighed. The reactor is initially charged with a weighed amount of component A and component B is added. and their presence even in small quantities may prevent further processing steps from occurring. The reaction of interest is the dehalogenation of a dihalogenated starting material to form the divinyl product.29: Semi·batch reactor addition of component B to starting material A. The excess addition of one reactant could also result in undesired reactions.29.3 An Industrial Case 5tudy 553 the quantity remaining. It is assumed that the halide groups are removed from the starting material in two consecutive reactions: B Figure 9. which is used in photographic film production. This general problem in which neither reagent can exist at the end of the reaction is called the "end-pOint control" problem. The reaction is carried out in a well·stirred semi-batch reo actor as shown in Figure 9.67) A XH2CeH2RCH2eH2X B e D Organic base H2e=eHRCH2eHzX H2e=eHReH=eH2 in which Qf is the volumetriC flowrate of base and cBf is the feed concentration of B. Over·addition of base causes polymerization and loss of the batch. In these cases. which is usually small. which subsequently loses HX to the base to produce the desired final product (D). The rna· = VRO = nA(O) nB(O) = nAO = ne(O) = nD(O) 0 It is assumed that cEf and VRO are known precisely. the cost of having the unreacted materials present in some amount greater than a small threshold is the cost of the entire batcb of chemicals. usually after the limiting reagent is exhausted. negligible volume change upon reaction or liquid chromatograph (LC) sampling.

30.--"r--.31.71) Conventional manufacturing procedure.6 0.0. The unknown parameters. The objective is to add enough B to consume all but 3% of the A.16 . Therefore only the C and D species show peaks in the LC output as depicted in Figure 9. An example of a typical B addition profile is given in Figure 9.68) (9.3 An Industrial Case Study 555 1.1 . Normal operating procedure is to charge the reactor with some (only approximately knmvn) amount of A.34.554 Parameter Estimation for Reactor Models 9.06 '" and the neglected side reactions. reports the relative amounts of C and D. This cycle repeats until the operator is satisfied that he is close enough (between 2-4%) to the target.----------------- 1'\ ~ "C5 ~ 1 - ~ o o 0.- w o +---.30: Depiction of an LC curve for determining the concentra~ tion of intermediate C and product D. in which klc is the proportionality constant for the LC. after a 7~minute delay.0. and add an inltial amount of B that is sure not to overshoot the end point.14 .3 L After the initial B is added. and reflect more the art rather than the science of reactor modeling.8 0.adds more B.'-'-'. Parameter estimation.33 and 9. and the operator again checks to see how close he is to the target before making another addition. After making tbe addition.32 The model predictions of all the concentrations are shown in Figures 9.70) This normalized C peak area.0.0. As can be seen in Figure 9. ~ '" '?i " 0 r. Ideally the operator would like to consume all the A.LC meas 0 r.. and . e.08 .based on experience . but the target is set at 3% to allow a margin for error.-.-'-"l-"-T'-.--t.4 0.0.02 o 100 200 300 400 500 600 700 800 900 time (min) Figure 9.-''-''T'--'-'-'r.'"'""-.12 CD 0.18 flowrate .69) ~ ~~ r. the operator required 7 additions and about 500 minutes after first turnlng on the LC before he bad determined how much B was required to reach end point. For composition analysis there is an on~line LC that draws a sample ev~ ery 8 to 10 minutes and.0.4 1. The penalty for overshoot is so high that only conservative addition steps are generally ever taken. Notice that making good decisions of this type depend on experience and judgment. The parameter estimation is performed by minlmizing the following relative least-squares objective function. the LC is switcbed on and the operator waits until the readings stabilize.2 - Figure 9. y. .31: Base addition rate and LC measurement versus time. By calculating normalized areas this constant can be removed.0. The areas of the two peaks are related to the molar concentrations by (9. is used for the parameter estimation. If we perform parameter estimation with these data we achieve the model fit shown in Figure 9. (9. The operator then sees how much C is remainlng.0. the LC readings are again allowed to stabilize.2 - . The reed flowrate of base is measured by a mass flow meter. The LC~detector wavelength is set to detect the R-vinyl bond. are estimated with the dynamic LC data. (9.04 At.. Note in which Yi is the measured normalized C peak and Yi is the model prediction at the ith sampling time.

but the rate constants have 200% uncertainty.~ ~ 2 1.1 0. That uncertaInty is a direct result of the problem structure...::. .3 An Industrial Case Study 557 0. If we could measure the B concentration.025 0. ( and D versus time..34: Concentration of species B versus time.2 o S OJ <l.I U P u 0.35 0.32: (omparison of data to model with optimal parameters.3 model-measurement 0 2.\ 600 I \~ .. 200 .556 Parameter Estimation for Reactor Models 9.I . we could achieve narrow confidence intervals for the rate constants as well.6%. For the purposes of end-point control. Unfortunately this concentration is not measured.5 a 0 100 200 300 400 500 time (min) 600 700 800 900 a a 100 200 300 400 500 600 700 800 900 time (min) Figure 9. After each base additiou.5 CA CD Cc 1 - p 0 <l.03 0. we have only rough information about the reaction rates hy observing the relative amounts of C and D. . . that the concentration of B provides the most information on the rate constants.::! .15 0.35 ± 0.~ kl k2 b u 0 '" 0 - 'p" OJ CB 0. '" u .035 M - .5 p 0.01 0. 300 I I '. The optimal values of the parameters and their approximate 95% confidence intervals are given by nAO 0. 800 900 o 100 400 500 time (min) 700 Figure 9. "0 0 S 0.015 0. however.013 2500 ± 4800 1250 ± 2300 ~ ..33: (oncentrations of species A.04 ~ 0. The relative amounts of C and D pin down accurately the amount of starting materiaL The rate constants determine the speed at which we arrive at these values.25 0.05 o ! '" "p 0 k .02 2. Fi9ure 9.J u '" '" S 0.005 0 Notice that the initial amount of A is determined to within 0.

82) in which (9.76 for the moles of A and C.(£1 nc = £1 nD = £2 . These results may be adequate. Due 10 the reactor-addition policy.75) (9. and conclude that under large kI. If we perform parameter estimation with this model and these data we obtain the results.76 then provides one algebraic equation for the two extents (9. We wish to make the equilibrium assumption as described in Chapter 5 to reduce the model. the difficulty we face is caused by the presence of large rate constants.00 ± 0.78 and substitute Equation 9. This analysis is clearest if we first rewrite the material balances in terms of extents of the two reactions as in Chapter 5 concentration of B is zero or the concentrations of A and C are zero. lfwe wish to cast the reduced model in differential equation form.80) Substituting Equation 9.72) The initial conditions for this model are VR(O) = = VRO fdO) £2(0) 0 =0 dt = Q fCB! (1 + knAO - nBadd + nBadd . Model reduction. .76) 2. we can differentiate Equation 9. nAO nB = nBadd . (9. Although it may not be inunediately apparent. and using Equation 9.78 again to eliminate £1 produces a differential equation for the second extent d£2 (9. The large parameter uncertainty tells us that the experimental data do not contain sufficient information to determine the rate constants.77) From the two reaction-rate expressions in Equations 9. we know B is the limiting reagent.35 ± 0. In this application. We must estimate them to determine the parameter we care about.£2 (9. The material balance for B in Equation 9. we regard ki and k2 as nuisance parameters. but if we wish 10 further improve the confidence in k = kIf kz and nAO. We seek to know how much total base to add. k2.558 Parameter Estimation for Reactor Models 9. not how quickly it "ill be consumed.81) We can easily translate from the two reaction extents to total moles of species via nA = nAO . if we assume equilibrium is established with these irreversible reactions.74) Solving this reduced modelforvarious values of k produces the results sho"TI in Figure 9.73) + E2) (9.72.79 to eliminate EI to obtain (9.3 An Industrial Case Study 559 the rate constants are irrelevant. but their values are not useful taus. That diagnosis is essential because it motivates our next step: model reduction.2£2 £Z)-I (9. But we still have more than 10% uncertainty in the ratio of rate constants k.78) The additional equation comes from examining the ential equations and noticing dEl tWO"'extents' differ.79) dt = (k nA) dEz nc dt dt= d£I dE2 dVR Q f dt = r.£1 (9. either the We have improved the situation compared to the uncertainty in the full model.35. we proceed by reexamining the experimental design.0083 2. = klcACBVR = k InAnB/VR dt = rz = k2 CCCBVR = ZnCnB/VR k We see that the concentration of B disappears from this slow time'scale model and the ratio k = kI/k2 appears instead of the individual rate constants. the B concentration is zero.29 (9.

We have removed the rate constants from the estimation problem by assuming their values are arbitrarily large and making the equilibrium assumption. . We do not have industrial operating data with early LC measurements.35 2. model-measurement 0 . we can easily remedy this final.36: Fit of lC measurement versus time for reduced model.36 in which we have simulated early LC measurements by solving the model with k Figure 9. \ \ S <l) ~ O..9 0 ~ . remaining problem.6 2..36 = kJik2 ~ 2.2 0. \. early time measurements have been added.2 0 1 . Although under the original reactor-addition policy.4 -' '-.0.34 2. Experimental design.355 2.1 0.2 2.. we see that these early measurements are actually the informative ones.'J~~1iIl 600 700 a 100 200 300 400 500 time (min) BOO 900 Figure 9.37: Parameter estimates and confidence intervals for reduced model with redesigned experiment. the data are much more informative.8 0. Given what we know at this point.8 1.7 0. But we still have a reasonably large uncertainty in the ratio of rate constants. .. Notice this noise is greater than the Figure 9. . we have the relatively large uncertainty in k shown in Equation 9.~\ "':-. tbere is little information left.~.:s ____t. .7 1. 2. Con· sider again the results shown in Figure 9.3 2. __ m """-"-~-~-- _ _ _ _ _ ~_~~ _ _ _ _ --.'~ -. By the end of the semi·batch addition. " en oJ <l) ~ '. But at early times..----- "\ \ k ~ 1/2 k=l k=2 k=B k = 32 LC meas 1.35.4 2.\ . ." .560 Parameter Estimation for Reactor Models 9..3 0."---~-~-="~ u -' Iii::_-f!Ef'l S 0. Consider Figure 9. We have the ability to query the model instead of performing expensive experiments to evaluate the impact of a proposed change.B 0. . We see that the ratio of the rate constants has its primary effect at early times. For this reason.345 2.4 0. nAO ~ 2.35 and adding measurement noise.1 k 2 1..6 LC 0.1 0 0 100 200 300 400 500 time (min) 600 700 BOO 900 .. .35: Predictions of LC measurement for reduced model.3 An Industrial Case Study 561 1 ~ """ . \ \ 0. \ '\\ -':' .5 0.82. when significant amounts of A and C remain in the reac· tor.9 1.6 0. but we can simulate the effect with our modeL Notice the advantage of modeling. . the operator had no reason to !Urn on the LC until later times.--. .


Parameter Estimation for Reactor Models

9.4 Summary and Notation


measurement noise shown in the actual operating data. We next per· form parameter estimation on the data shown in Figure 9.36 and obtain the following parameters and approximate confidence intervals 2.35 ± 0.0083 2.00


Figure 9.3 7 shows the confidence·interval ellipses and boxes when using the original dataset and using the dataset augmented with early· time data. As summarized in Figure 9.37 the primary benefit in adding the early time data is a more precise estimate of k. The nncertainty in k is reduced by more than a factor of six when adding the early lime LC measurements. The initial number of moles of A is relatively accurately determined by both datasets. Improved reactor operation. Given these modeling results, we can shorten the batch time significantly. First we switch the LC on at time t ~ 0. Then as the LC measurements become available, we estimate the initial number of moles nAO and monitor its confidence interval. As soon as the uncertainty in nAO reaches a suffiCiently low threshold, we are confident how much B is required and can add the remainder in one shot. Testing this approach with many datasets at Kodak allowed us to conclude that by the time the first large addition was completed, we obtained suffiCient confidence on nAO that the rest of the B could be added immediately. Such a procedure reduces the batch time from about 900 minutes with the conventional approach to about half that time with the model·based operation. The new operation essentially doubles the production rate without constructing new reactor facilities, which is significant for this capacity·limited chemical.

We also discussed how to model differential reactors, which are con· venient reactors for measuring reaction rates as a function of species concentrations. Next we covered analysis of data. We used probability and random variables to model the irreproducible part of the experiment. For mod· els that are linear in the parameters, we can perform parameter estima· tion and construct exact confidence intervals analytically. For models that are nonlinear in the parameters, we compute para~ter estimates and construct approximate confidence intervals using nonlinear opti' mization methods. We then covered the important topic of estimating parameters for differential· equation models. We employed computational methods for solving differential equations and sensitivities, and solving nonlin' ear optimization problems in order to tackle this challenging problem. The chapter concluded with an industrial case study. In this case study we illustrated the following steps: • Proposing an initial model given a rough idea of the reaction chemistry. • Estimating the parameters for the full model given composition measurements. • RedUCing the model based on the parameter confidence intervals. • Designing new experiments based on the reduced model's confidence intervals. • Using the final model to find new operating policies to double the production rate. We hope the examples and methods in this chapter serve to inspire students and practicing engineers to build models as part of nnder· standing new processes and chemistries of interest. The modeling experience often leads to deeper process nnderstanding and produces a compact summaty of current knowledge that is easily and efficiently communicated to other colleagues and team members. Process under· standing coupled with creativity often leads to process improvement and new discovery.



In this chapter we first summarized some of the analytical methods and experimental reactors used to collect reactor data, focusing the discussion on: • infrared spectroscopy • gas chromatography
• mass spectrometry


Parameter Estimation for Reactor Models

9,5 Exercises





b c cJ



nd np




absorbance path length for the absorbance measurement concentration of absorbing species concentration of component j measured concentration of component j at sampling time ti catalyst particle diameter fixed·bed reactor tube diameter Hessian matrix, Hkj ~ a'ip/O!JkO!Jj fixed·bed reactor tube length mass velocity number of data points number of model parameters molar flow of component j probability density function of random variable x volumetric flowrate feed volumetric flowrate reaction rate of (single) reaction reaction rate for ith reaction

Exercise 9.1: Bringing it all back home Given the new tools in this chapter, return to the data displayed in Chapter 2, Figures 2.2 and 2,3, Reaction rates and production rates are relaled by

If we measure production rates and wish to estimate reaction rates, we can model the measUrement process ""ith


vTr + e

in which e is assumed to have a normal distribution, (a) What is the formula for the 95% confidence interval on estimated reac· tion rates r?
(b) Place your 95% confidence interval on the set of points displayed in






production rate for jth species production~rate vector open tube area for flow sensitivity of state Xj vvith respect to parameter 8k sampling time reactor volume state vector vector of measured responses molar absorptivity bed porosity parameter vector fluid viscosity stoichiometric number for the jth species in the ith reaction transformed parameter vector objective function

Figure 23. These points are available at www.che, chemreacfun. Given the Octave code displayed in Chapter 2 lhat gener· ated Rmeas , what is the value of the measurement error variance. Of the 500 estimated reaction rales, how many of the values fall outside your 95% confidence interval ellipse? Notice that if you know (or have some means of estimating) the mea· surement error variance beforehand, you do not need the experimental measurements to know the confidence intervals, They are computable from the model before you do any experiments, (c) Now assume the measured production rates are only the first column in R m ,., in Chapter 2. Compute the least-squares estimate and 95% can· fidence interval in which you assume you do not know the variance in the measurement error and must estimate it from these data also. How do your two ellipses compare? Do you obtain a reasonable idea of tbe parameter uncertainty when you only have one measurement of the six species production rates? Exercise 9.2: Batch·reactor parameter estimation
A weU~mixed, isothermal, constant"volume batch reactor is initially charged with reactant A and tbe follOWing reaction takes place



Parameter Estimation for Reactor Models

9.5 Exercises













t (min)












0,83 0.68 0.59 0.51 0.43 0.38 0.33 0.30 0.26 (J,24 0.21

0.830 0.696 0.591 0.507 0.440 0384 0.339 0301 0.268 0.241 0,217 0.833 0.683 0.593 0.515 0.446 0.380 0.324 0.303 0.256 0.225 0.205 0.B62 0.699 0.561 0.511 Q.412 OA07 0.346 0.276 0.322 0.254 0.215

0.9 , - - - - - - - - - - - - - - - - - - - - - - - - - - - , o 0.8 0.7 0.6 0.5 -

o o o

0.9 0.8

_ _ (Y2





0.4 0.3 -


= le~4 2 = Se ~ 4

--.-- - ~ _.




0.2 0.1 --'-,-



~ §



10 15 20
t (min)

0.3 0.2 0.1






Figure 9.38: Batch reactor data for Exercise 9.2. The reactor volume is 10 L and the initial charge of A is 8.3 mol. The concentration of A versus time is measured and shown in Figure 9.38. You wish to fit these data using the following kinetic rate+law expression.
r = kc~









t (min)

Figure 9.39: Batch-reactor data for Exercise 9.3; 3 runs with different

measurement error variance.

(a) What is the solution to the model, CA(t), for this rate law? (b) Calculate from the data rough estimates of the parameters k and What are their units?

taking place in an isothermal, liqUid-phase batch reactor. Measurements of CA versus time for different measurement accuracies are displayed in Figure 9.39. The following method is sometimes recommended for finding the rate constant and order of this reaction.
1. Compute an approximation to the reaction rate hy "differentiating" the data CA 1+1 -, CAL rz= ' ,l=l,.,.,nd-l
tl+l - tl

Hint: consider approximating the time derivative deAl dt by t:.cAI t:.t and compare those approximate values of the rate using the data to kc~, A nonlinear transformation of the rate expression can make the estimation problem lineaL

Estimate the parameters by solVing a least-squares problem. Plot the fit of both models to the data and compare. Are the model fits comparable?

in which CAl is the measured concentration at sample time tl and yt is the "measured" rate at this sample time. Notice the rate is not actually measured; the concentration is measured. 2. Take the logarithm of the rate expression to obtain a model linear in the parameters log(r) ~ log(k) + nlog(cA)

Exercise 9.3: The perils of differentiating data
Consider the nth-order, irreversible reaction A ~ producls,


3, Use linear least squares with this model and the "measured" rate sus CAl to find k and n.




Parameter Estimation for Reactor Models

9.5 Exercises


Let's examine how well this method performs on the three datasets shown in Figure 9.39.

Exercise 9.4: Filting the RTD step response
Consider the RTD test data given in Exercise 8.1 L Use the least squares method of this chapter and find the best value of parameter p, given in Equations 8.68, that fits these data. Solve the model with this value of p and show the fit to the data. Resolve Exercise 8.11 and predict the reactor yield and conversion.

Radius (em) 0.016 0.071 0.121 0.191 0.246


(mol/cm 3 ) 2.39 x 10 .. 5 4.25 x 10-6 5.94 x 10- 6 7.84 x 10- 6 9.04 x 10- 6

Determine the value of the rate constant at 573 K.

Exercise 9.7: Oxygen adsorption, revisited


Exercise 9.5: The differential reactor and volume change
Your colleague has an idea that, since not much conversion occurs in the differ-

Consider again the adsorption data in Exercise 5.12. The data are listed in Table 5.8 and plotted in Figure 5.25. The adsorption is dissociative
02 + 2X

ential reactor, we should be able to assume the volumetric Ilowrate is constant to simplify the modeL Consider the gas-phase reaction A - 2B using a pure A stream and compare analyzing the data from a differential reactor in two ways. First, use Equation 9.4, which accounts for the changing volumetric flo-wrate, Second, neglect the change in volumetric Ilowrate, assume Q ~ Qf' and show that





(a) Write out again the expression for the Langmuir isotherm to model the concentration of adsorbed oxygen in terms of the concentration of gasphase oxygen. How many unknown parameters does your model cantam?

(b) Estimate these parameters using the nonlinear least-squares technique described in this chapter. (e) Consider again the transformation of the model into the linear form, i.e., consider llco as a function of 1/ ,jc02. Estimate the parameters again
using linear least squares.

Calculate the relative error e ~ (r - r)IT committed when neglecting the volumetric flowrate change in a differential reactor assuming you are measuring component A. Is this a large of small error? If it is a larger error, can you explain what went wrong "ith your colleague's logic?

(d) Compare the parameter estimates obtained with these two approaches by plotting the fit to the original data. Why are the fits so different? Which technique do you recommend and why?

Exercise 9.6: Analyzing experimental data
The heterogeneously catalyzed reaction

Exercise 9.8: Replication and confidence
Imagine you have performed an experiment and set up the least-squares problem


was studied in a small, research-scale CSTR. The CSTR design eliminates exlimitations (Le., you may assume the bulk fluid concentration equals the surface concentration). The feed consisted of pure A. The reactor pressure was 1.0 atm and the temperature was 573 K. The residence ternal


xe + e

time for the CSTR is 2.0 s. The following table presents effluent-concentration data for different catalyst sizes. The catalyst particles can be assumed sphericaL The effective diffusivity of A in the pellet is DA ~ 0.0008 cm2 /s. The pellet density equals the bed density, Pp ~ PB.

in which y contains the vector of measurements, X contains the matrix of independent variables, 0 is the parameter vector, and e is the measurement error, assumed to have known variance u'- We developed the following formula for the parameter estimates and t<~level confidence intervals
9~ (XTX)-lXTy

(e -

9) rxTx (e - 9) ,; u'x 2(n p , a)



Parameter Estimation for Reactor Models

9.5 Exercises


If you replicate this experiment

ns times, the new data matrix is

4.5 4 3.5 3


(a) Calculate xix,. From Equation 9.83, how much smaller is the confidence interval ellipse after 2, 5 and 100 replicate experiments? (b) Does the confidence ellipse change shape as well as size upon replication? Does your result agree with Figure 9.20? (c) If you want to reduce uncertainty by a factor of fl, how many replicate ex· periments are required? Explain why straight replication is a potentially expensive way to reduce uncertainty.

2 1.5 1 0.5




t (min)




Figure 9.40: Batch-reactor data for Exercise 9.9.


Exercise 9.9: Estimating a rate expression involving two reactants
Consider the irreversible reaction


taking place in an isothermal, liquid·phase batch reactor. Measurements of CA and CB versns time are displayed in Figure 9.40. We "ish to determine from the data the rate constant and the order of the reaction with respect to hotb A and B. (a) Estimate rate parameters, k, nand m, and initial concentrations, CAO and CBO, from the data in Figure 9040 without differentiating the concentration data. The data are available at chemreacfun. Notice your confidence intervals for k, nand m are rather large. Wby is this model poorly identified from these data? (b) Consider a second experiment shown in Figure 9.41. Estimate the parameters using only the second experiment. Are your confidence intervals for k, nand m any better using the second experiment? (c) Estimate the parameters using both experiments simultaneously. Are your confidence intervals any better using both experiments compared to using either experiment alone? Explain why or why not.



Figure 9.41: A second experiment for Exercise 9.9.

Exercise 9.10: living and dying with linear least squares
Consider differentiating the data of Exercise 9.9 as described in Exercise 9.3. Notice differentiating the CA data and the CB data provide two independent estimates of the same reaction rate. You may wish to average these to obtain a good single estimate of the rate. Taking the logarithm of the rate expression again produces a model linear in the parameters log(r) log(k) + nlog(cA) + mlog(cB)

Based on these results, to design experiments to detennine the order of the rate expression with respect to two different species, how shOUld you choose the initial conditions of the two species?

(a) To ensure that your method is coded correctly, consider first the "error free" data depicted in Figures 9.42 and 9.43. Estimate k, nand m using linear least squares and three different datasets: onJy the data in Figure 9.42, only the data in Figure 9.43, and the data in both figures.


Parameter Estimation for Reactor Models


!2J [3J Figure 9.42: Batch-reactor data for Exercise 9_10. [4J

A. C. Atkinson and A. N. Donev. Optimum Experimental Designs. Oxford






4 -

Po en


3 2 1 -



0 0









t (min)
Figure 9.43: A second experiment for Exercise 9.10. Do you get similar estimates with all three datasets? Are the confidence intervals similar? Explain why or why not. (bl Now apply the linear least-squares approach to the data in Figures 9.40 and 9-41. Are your estimates and their uncertainties comparable to the result of Exercise 9.97 Which method of estimating parameters do you recommend and why?



[12J [13J [14]


University Press, New York, 1992. Y. Bard. Nonlinear Parameter Estimation. Academic..llress, New York, 1974. L J. Bellamy. The Infra-red Spectra of Complex Molecules, volume one. Chapman and Hall, London, third edition, 1975. G. E. P. Box, W. G. Hunter, and J. S. Hunter. Statistics for Experimenters. John Wiley & Sons, New York, 1978. G. E. P. Box and G. C. Tiao. Bayesian Inference in Statistical Analysis. John Wiley & Sons, New York, 1973. M. Caracotsios and W. E. Stewart. Sensitivity analysis of initial value problems with mlxed ODEs and algebraic equations. Comput. Chern. Eng., 9(4):359-365,1985. M. Caracotsios and W. E. Stewart. Sensitivity analysis of initial-boundaryvalue problems with mlxed PDEs and algebraiC equations. Comput. Chern. Eng., 19(9):1019-1030, 1995. j. J. Carberry. DeSigning laboratory catalytic reactors. Ind. Eng. Chern., 56:39-46, 1964_ L G. Christophorou, editor. Electron-Molecule Interactions and Their Applications, volume one. Academic Press, Orlando, FL, 1984. L E. Davis, N. C. MacDonald, P. W. Palmberg, G. E. Riach, and R_ E. Weber, editors. Handbook of Auger Electron Spectroscopy. Physical Electronics Industries, Eden Prairie, Minnesota, second edition, 1976. A A Davydov. Infrared Spectroscopy of Adsorbed Species on the Surface of Transition Metal Oxides. John Wiley & Sons, Chichester, England, 1990. c. B. Duke, editor. SURFACE SCIENCE The First Thirty Years. Elsevier Science Publishers, Amsterdam, 1994. F. H. Field and J. L Franklin. Electron Impact Phenomena and the Properties of Gaseous Ions. Academic, New York, 1957. F. W. Fifield and D. Kealey. Principles and Practice of Analytical Chemistry. BIaclde Academic and Professional, Glasgow, UK, third edition, 1990. c. F. Gauss. Theory of the combination of observations least subject to errors: part one. part two, supplement. SIMI, Philadelphia, 1995. Translated by G. W. Stewart. R. L Grab, editor. Modern Practice of Gas Chromatography. John Wiley & Sons, New York, 1977.

Chromatography. P. editors.I linear Algebra and least Squares In this section we briefly introduce Octave and MATLAB.. Podkolzin and J. Bell and M. Natal·Santiago. L Hair. S. G. Infrared and Raman Spectra of Polyatomic Molecules. One is naluraily free to use any numerical solution procedure that is conveuient. [22J J. Optical measurements of porosity and fluid motion in packed beds. AIChE j. Adsorption data for Hz on Pd. and T. W. New York. Rawlings. editor. John Wiley & Sons. Infrared and Raman Spectra of Inorganic and Coordination Compounds.. E. John Wiley & Sons.che. pages 323-328. Durnesic. D. Van Nostrand Reinhold. [23] K. Open Tubular Column Gas Chromatography: Theory and Practice. Washington. Personal communication. Minnesota. 1989. 42(11):3055-3062. Bruemmer. Wachs. Molecular Spectra and Molecular Structure II. Microcalorimetric studies of interactions of ethene. HeHmann. J. and j. ButterworthHeinemann. Lee. Catal.. New York. M. P..w. Eden Prairie. Distillation Columns. Discrimination and goodness of fit of multiresponse mechanistic models.Perken-Elmer Corporation. fifth edition. Thomas and W. pages 1-1L American Chemical Society. 5J or. type help -i introduction at the Octave prompt or demo at the MATLAB prompt for information on how to enter matrices and perform matrix multiplication.globaldialog. Box. Consider . A. Nakamoto. E.!-jbraw!chemreacfun A. E. Foundations of Vacuum Science and Technology. Surfaces. and K. Inc. 1997. In this appendix we summarize many of the fundamental computational procedures required in reactor analysis and!-nobhill www. 1992. and Batch Processes. and G. 1992.. Chern. Part A Fundamentals and Techniques. VCH. isobutene. 1997. Dumesic.1945. G. Jerome. 1996. 38(5):641-650. Endpoint control in semi-hatcb chemical reactors. The MATLAB package is commerCially available from The MathWorks. 41(8):2161 .2170. L Stephenson and W. MATLAB.. We illustrate these procedures using the numerical computing language Octave [4.l D. W.c. 1998. 1986. [19J G. Eng. Sobol. 124J M. and show how to enter reaction networks quickly and perform linear algebra operations of interest. F. Stewart. Y. T. New York. Coroell University Press. Vibrational Spectroscopies for Adsorbed Species. P. D. E. M. S0rensen. ACS Symposium Series 131. Pt. Stewart. In A. Stoneham. and G. New York. [21) M.514 Bibliography [17J M. Octave is freely available for a variety of hardware platforms and can be down-loaded from www. Transmission infrared spectroscopy for high surface area oxides. 44(6): 1404-1412. A. Moulder. but we have found that these two environments provide reliable and high-quailty numerical solutions. E. 45:155-163. and is becoming a commouiy available 1001 of industrial engineering practice. Hamer. R. Somorja!. W. The Octave files required to produce these figures and tables are freely available at these two websites: www. 118J E. and J. Principles and Practice of Heterogeneous Catalysis. and PtSn. third edition. 198 L J. All computations required to produce every figure and table in this text were performed with Octave. 1980. M. AlChE j. Characterization of Catalytic Materials. AIChE J. Shon. 1992. Handbook otx ray Photoelectron Spectroscopy.octave. Weinheim. In Proceedings of the IFAC Symposium on Dynamics and Control of Chemical Reactors. G. and convenient user interfaces. 1984. 1992. Cortright. W. Elsevier. Sci. L Henson. Podkolzin. New York. S. and K. Model discrimination and criticism with Single-response data. Parameter estimation from rnultiresponse data. A. 1978. Herzherg. Box. 200 L J. After starting an Octave session. 1998. N. J M. editor. Caracotsios. L Hair. D. Many of the figures and tables in tbis text required numerical solutions of various types of chemJcal reactor problems. L E. Stewart. and isobu- A Computational Methods [25J [26) [27) [28) [29] {30) [31 J 1321 [33) tane with silica-supported Pd. Lett. Bomben. Amsterdam. Chemistry in Two Dimensions. ]. Bartle. John Wiley & Sons. T. Ithaca. W. equivalently. Stickle.. Lafferty. F. P. B. [20J J. Thomas. F. P. MA. A. editor. E. L.

3]. We know from Chapter 9 that the distrihution of estimates is a mnltivariate normaL We also know how o to calculate the <x-level confidence intervals.I: Estimating reaction rates Consider the fitst and second water gas shift reaction as an independent set and let [rj rz] = [1 2]. \ 1 0 -1 0 -1 1].95 1. endfor r_est=sto. and then compnte the species production rates from given reaction rates._ _ _. for example.576 Computational Methods A I Linear Algebra and Least Squares 577 again the water gas shift reaction chemistry.i)=0. A typical set of Octave or MATLAB commands to perform these operations wonld appear as follows octave:1> stoi=[O 1 0 -1 -1 1. 1 3 -1 -3 -4 4 Example A.9 0. octave:2> rank(stoi) ans = 2 octave:3> r=[1.95 1 LOS 1. Plot the distribntion of estimated reaction rates.' \ R-roeas.j by v T _ Fignre A. for i=l:npoints R-meas(: .2.L_ _ _ _L -_ _----l 0.1 + + ++ + + [ 0 1 1 0 -1 1 0 1 -1 -1 -1 0 -1 0 -1 1 0 1 2.1 shows the estimates. The command in Octave and MATLAB for the least-squares solution is r = stoi' \ R which can be recalled with the mnemonic of a "left division" of Equation A.1)+R. -111 -1 0 0]. we can solve for r as a least-sqnares problem. [rj r2 r31 = [1 2 3].9 L-_ _ _---L.1 Figure A. H Hz OH H2O CO CO2 2. npoints = 2000. r=[1.2J: R=sto. As we explored in Chapters 2 and 9.05 J + ++ + + + + + ++ + + + + + + r2 2 Say we would like to find the number of linearly independent reactions.l: Estimated reaction rates from 2000 production-rate measurements subject to measurement noise. Given the prodnction rates and reaction rates are related by R= vI'r (A-I) then the least-squares solution is given by r= (vyI' )-lYR in which the snperscript -1 indicates a matrix inverse. -1 1 1 -1 0 0. octave:4> R=stoi'*r R + + + + + + + + + 1. Create 2000 production rate measurements by adding noise to R = vI' r and estimate the reaction rates from these data.05*randn(6. Solution The Octave commands to generate these results are stoi=[O 1 0 -1 -1 1.'*r. from measured R. .

We explore computing solutions to these types of problems in Octave or MATLAB. To solve algehraic equations or optimization problems. Reactor equilibrium prohlems can be numerically challenging. The equilibrium condition requires the solution of two equations in two unkno~$. but finding or measuring these data is often another significant challenge in computing equilibria for systems of industrial interest. f(x) endfunction The first line states that the value of the variable y is returned by invoking the fnnction named f with the input value x.YP2 are defined in terms of the two reaction extents in Equations 3. yBO 0.2 Nonlinear Algebraic Equations and Optimization Determining equilibria for reacting systems "ith multiple phases and reactions can require significant computational effort.YPI in whichYI. inside quotes. yIO 0. provided in the variable xO.2 Nonlinear Algebraic Equations and Optimization 579 A. the following is a typical session to compute the solution given in Example 3. for example. and res i dua 1 is a vector containing the values of the two equilibrium conditions to be returned by the function. P = 2. But often we need to define our own functions.YP. on) is fil ename. Storing. yPIO = 0. m. = 0 PKIYIYB . we can use the built-in nonlinear equation solver to find its solntion. xO) The fso 1 ve function requires the name of the function. called m·files. With this function defined.5.5 . expm. yPl (yPIO + xCI» / d.Ys. As the computational complexity of the function increases. endfunction 0.2 50lv.2. gamma. incomplete gamma function. and even the best software can run into problems. Some that we have used in this text include the matrix exponential. pseudo-random numher generators. we would like to store our work so that it does not need to be re-entered at the beginning of each subsequent Octave session.YP2. Even for our purposes. info] fsolve("dgdx".2. yB CyBO . A. the naming convention Notice that x is a vector containing the two reaction extents provided to the function named dgdx when it is invoked. then. Octave commands and function definitions can be stored in text files.5. j(x) = x 2 for example. See also Section 3. After the function dgdx is defined.2. As we saw in Chapter 3. and several others. and an initial guess for the unknown extents.YP l .x(2» / d. We seek values of x such that resi dual is zero. = function residual = dgdx(x) K1 = 108. rand and randm. yP2 = (yP20 + x(2») / d.yPl.xCI) x(2» / d. An Octave function defining these two equations is Octave and MATLAB provide many convenient built-in functions.x(l) . Editing these files with a text editor during an interactive Octave session provides a convenient means for debugging new calculations.x (2) . is the calculation of the value of y to be returned. Type help -i and select the menu item Functi on Index for a complete list of Octave's built-in functions.74.. and returns the solution in x and a !Jag info indicating if the calculation was successful. It is highly recommended to examine these information !Jags after all calculations.5. K2 = 284. The body of the function. A. =0 PK2YIYB . docnmenting and maintaining the calculations performed in solving complex industrial problems is one of the significant challenges facing practicing chemical engineers.578 Computational Methods Nonlinear Algebraic Equations (fso 1 vel We consider again the problem in Example 3. yP20 d 1 x (1) . In this section it is assumed that the required thermochemical data are available. residualCl) P*Kl*yI*yB .1 Defining Functions (funet. residual (2) P*K2*yI*yB .1 for a brief discussion of thermochemical datahases. the command structure is function y y "" x*x. the user needs to provide a function that can be called by the solver to evaluate the nonlinear equations or the objective function. by entering Ex. we should he able to store and edit our computational problems. To define a function in Octave. the phase and reaction equilibrium conditions generally lead to mathematical problems of two types: solving nonlinear algebraic equations and minimizing a nonlinear function subject to constraints.5. yI (yIO .

05 0.ub.3 E'2 0.2 shows the lines of constant Gibbs energy determined by Equation A.13317 0.O.x(l) .25 E'I The other main approach to finding the reaction equilibrium is to minimize the appropriate energy function.(x(1)*log(Kl)+x(2)*log(K2)) + \ (1-x(1)-x(2»*log(P) + yI*d*log(yI) + \ yB*d*log(yB) + yPl*d*log(yPl) + yP2*d*log(yP2). yP20 d 1 .2.15 0. in this case the Gibbs energy.1 0.x(l) . lb.2) The minimization of this function of Ei then determines the two equilibrium extents. in which xO is the initial guess as before and the new variables are used to define the constraints (if any).2].35087 info = 1 0. First we define a function that evaluates G given the two reaction extents.aub) We proceed to translate this version of the equilibrium condition into a computational procedure. yIO = 0. infoJ=npsol (xO. [x. as shown in Example 3. These are required to prevent the optimizer from "guessing" reaction extent values for which the Gibbs energy function is not defined.5. we derived the following Gibbs energy function for an ideal-gas mixture 0. consider what the gi bbs function returns given negative reaction extents.2 0. yI (yIO . For example. and the following constraints are required to ensure nonnegative concentrations yPI = (yPIO + x(I» I d. we specify the sum of extents is less than 0. K2 = 284. P = 2.a.55 -2.5. In Chapter 3.2: Gibbs energy contours for the pentane reactions as a function of the two reaction extents.x(l) .} = fsolve("dgdx". This optimization-based formulation of the problem. can be more informative than the algebraic approach discussed above.2. function retval=gibbs(x) KI 108.35 = The value of info A.obj.5. Finally. (A. which define linear inequality constraints other .x(2» I d.4 0. The variable 1b provides lower bounds for the extents and ub provides upper bounds for the extents.5 Figure A.alb.1 0.580 Computational Methods A. We see immediately that the miuimum is unique.35 0.2 0.15 0. yBO 0.--2 -1 ----0--- Solving Nonlinear Optimization Problems (npso 1) 0.3.45 0.5 .xO) 0.3 1 indicates that the solution has converged... .5 with the last three arguments. retval ~ .25 -2. "gibbs". Notice that G is not defined for all reaction extents.x(2).4 0.5 x = 0. endfunction MATLAB To solve the optimization we use the Octave function npso 1 [6] or fuoction constr.45 0.2 as a function of the two reaction extents.05 0 -' o 0. yPIO = 0. 0.2 Nonlinear Algebraic Equations and Optimization 581 octave:l> xO=[O.56 -2.53 -2. yB (yBO .3 0. octave:2> [x.x(2) I d. Figure A. yP2 = (yP20 + x(2» I d.

A bigh-quality ODE solver is an indispensable tool of the reactor designer because chemical reactions can take place at widely different rates and time scales. and call the ODE solver. x lsodeC"benzene_pyrol_rhs".499: lb=[O.08205. Therefore. If the differential equations are nonlinear (j is not a linear function) and one must keep track of several species simultaneously. leading to stiff ODEs. alb=O. We can set alb = 0 because both extents are already specified to be positive. ub=[0.0. and the results are in good agreement with those computed using the algehraic approach. Q NBf*R*T/P. k1 = 7e5. the value of info = 0 indicates that the solution has converged.582 Computational Methods A. T = 1033. xdot(2) = k2*(cB*cO cT*cH/K2).35087 obj = -2. yB = ( NSf . cD = NO/Q. lsode 171. conv = ( 2*x(:. cT = NT/Q: cH NH/Q.ext(2). .48. then analytical solution is not possible and a numerical procedure is required.1) . xO.50)'. vaut). but volume or len~'in the steadystate PFR. volumes in tbis case. xO=[O. feed condition).2.499.5. If we examine the benzene pyrolysis example.13317 The ODEs required to describe transient-batch. k2 = 4e5. <. Type help -i differential equations in wbich the meanings of the symbol names should be apparent from the example. CSTR.2) ) / a suitable function is function xdot = benzene_pyrol_rhs(ext. 0. Type help ode23s for an overview of MATLAS'S ODE solvers.2*x(:. A.ub.a. and j is the function defining the material balances for the species.3 Differential Equations (lsode) xdot(l) = k1*(cB*cB .1) + x(:. The time scales for different reactions may be widely separated.cD*cH/K1).2.5. We recommend always using the so-called "stiff" solvers for chemical reactor modeling.OOl.OOl: aUb=O. cB = NB/Q.volume) global kl K1 k2 K2 R T P NBf NB NBf]=npsol(xO. a typical session to solve for equilibrium composition by minimizing Gibbs energy is octave:1> octave:2> octave:3> octave:4> x = 0.O."gibbs". xo. We then need to specify at wbich t values.499]. so A = [1 1 1 and aub = 0. In tbis case.2]. [x. + <2 s. and the Gihbs energy contours depicted in Figure A. R = 0.O].OOl]. P = 1. and steadystate PFR reactors are of the form ~ x(O) =j(x) = xo a=[1 1].2) ) /NBf. vout = linspace(O. NO = ext(l) . t is time in the batch reactor and CSTR.2*ext(1) ext(2).aub) 0. K2 = 0. Example 4. Optimization is a powerful tool for solving many types of engineering modeling and design problems. xO=[O.3 Differential Equations (l sode) 583 than simple bouuds. NT = ext(2). We also rely heavily on optimization tools in Chapter 9 on parameter estimation. K1 = 0. presenting difficult numerical challenges.O. endfunction The real workhorse for simulating chemical reactors is the ordinary differential equation (ODE) solver.5. For our problem. A typical Octave sessiou wouid be NBf 60e3. for an overview of Octave's ODE solver. NH ext(l) + ext(2).obj.5593 info = 0 in wbich x (t) is a vector of species concentrations to be found. the initial condition (in this case. the solution shouid be reported. The first task is to formulate a function wbich defines f.x(:.

dx/dt = f(x. ~ aoJdx. Brenan et al. and when we have created a new phase and need to change the ODEs govern· ing the system. are defined as the change in the solution with respect to the model parameters 5ij = (~~i) j (AS) (h:c<j S _ ox . which the ODE solver checks while marching forward until the function reaches the value zero.] ~ afT dx. octave:2> vout = linspace(0. in good agreement with Fig· ure 4. to see the details of how to use dass 1.1 Differential·Algebraic Equations (dassl) the PFR volume at which 50% conversion of benzene is achieved. t) Both DAEs and ODEs can be considered special cases of this structute. The program ddasrt enahles us to stop when specified conditions are met exactly.ext(2). The user provides one new function. [1] provide further reading on existence and uniqueness of solutions to these models. The nonconstant density. in which the function 1 i nspace computes SO linearly spaced points between 0 and 1600 L. are calculated for plotting. Finally.25 L.che.xB. tj. The results of this calculation are shown in Figures 4. vstopJ = ddasrt("benzene_pyrolJhs".2) ) /NBf. consider the more general form of implicit ODEs 0= f(dx/dt. which are considerably more complex issues than in the case of simple ODEs.wisc. function val = stop(ext.x.20 and 4. ( x(:.OJ (A3) (AA) x{O) = g(xo. and differential-algebraic equations. xdotO vout).l) + x(:. but how the solution changes with the model parameters. it is often valuable to know not just the solution to a set of ODEs. The reader can consult the software for Example 8. Initial conditions are required for dx / dt as well as x in this model. xO. To address these models.1 at www.20.COT (A.2) ) /NBf. stop. endfunction -' Some reactor models require a more general structure than the ODE. \ "stOP". octave:4> vstop(length(vstop)) ans = 403.4 Parametric Sensitivities of Differential Equations (ddasacl As discussed in Chapter 9. xdotO = [O. Consider the following vector of ODEs with model parameters O. A. to compute solutions to implicit differential. 0) The sensitivities.25 f A. The main difference between using dass 1 and 1 sode is the form of the usersupplied function defining the modeL A second difference is that the user must supply Xo as well as Xo. octave:l> xO = [O.SO)'. ( x(:. volume) global NBf xBstop NB = NSf 2*ext(1) . nonconstant volume semibatch and CSTR reactors in Chapter 4 are more conveniently expressed as differential-algebraic equations (DAEs). any quantities such as conversions or yields.584 Computational Methods A.O) d 00 a dt a. at t = 0 Petzold has provided a numerical package.1) . The PfR volume is found to he VR = 403.OJ.2) ) /NBf.1600. xdot. octave:3> [x.3. which are computable from x.3.2 L A.2 Automatic Stopping Times (ddas rt) We often need to stop ODE solvers when certain conditions are met.xC: .6) We can differentiate the Sij to obtain dt dS'j = dt [ax.NB/NBf: val = xBstop .5. xB 1 .4 5ensitivities of Differential Equations (ddasac) 585 yD yT yH ( xC: .OJ. ~~(t) =xo x{t) = Xo. ) J J (A7) . dass 1. xBstop = 0. Two examples are when we have reached a certain conversion. The following code illustrates how to find It = dx f(x.

3. S = iJcAlak. The result is (AI2) which agrees with Equation All. Equation AlO is our standard.8) solution 1. first-order linear equation whose solution is . Next consider the rate constant to be the parameter in tbe problem. Repeat steps 2 and 3 using the initial concentration as the parameter in the problem.6.2: Two simple sensitivity calculations Consider the differential equation describing a first-order.l3) = -kCA CA (0) = CAD (AlO) 1. 4. = -tcAor kt Evaluating the various partial derivatives gives e = k pro(A. Notice the sensitivity differential equation is linear in the unknown S.586 Computational Methods A. first-order linear equation. whose solution is Substitntlng this back into Equation A. The functions CA(t).SI(t) are shown in Figure A. e = k.S and writing the sum as a matrix multiply yields dS dt =jxS+ Ie (A g) 2. 3. The initial conditions for this matrix differential equation is easily derived from the definition of 5 S(O) = aXi(O) = OBi tJ Ix =-k so the sensitivity differential equation is dSl go = 0 aeJ aeJ dt = -kSl . Taking the partial derivative of CA (t) with respect to duces directly S. e = CAO. then S(O) = o. first write out the solution to the differential equation.CA dS dt =jx 5 +je = SdO) = 0 5(0) Be Notice that if none of the initial conditions are parameters for which one is computing sensitivities. isothermal.)2(0) = I This equation is our standard.4 Sensitivities of Differential Equations (ddasac) 587 Using the chain rule to perform the final differentiation gives (A. Next write the differential equation describing dS I dt.S(t) versus t. What is S (O)? Solve this differential equation and compare to your previous result. Plot CA(t).II) in which the subscript denotes differentiation with respect to that variable. Evaluating the various partial derivatives for this parameter choice gives Ix =-k Bo = I so the sensitivity differential equation is dSz dt = -kSz . or with Laplace transforms as in Exercise 4. 2. irreversible reaction in a constant-volume batch reactor dCA This equation is linear and can be solved with the method used in the series reactions in Chapter 4. Take the partial derivative of the solution directly to produce the sensitivity to this parameter. Taking the partial derivative of CA(t) with respect to e duces directly = dt CAO pro(A. 3. Example A.

In the collocation method. 2). Derivatives and integrals of the polynomial interpolant S. 21. The derivatives aod integrals of the polynomial interpolant can be . function global ca k xdot firstorder(t.4: Function c(r) and its values at five collocation points. we require the differential equation to be satisfied at more locations. 1). nth(sx. which seems a natural first chOice. turns out to have mediocre properties.3: Solution to first-order differential equation dCA/dt ~ -kCA. t Figure A.3. theta(l). 1J.S Boundary-Value Problems and Col/ocation (co I I DC) C 0.2. ~ oCA/ok and 5. 5XO. n c . we first compute the required derivatives of the polynomial approximation.l3. and sensitivities S. This choice is referred to as orthogonal collocation. The function is approximated by passing a polynomial through these points. tout). xO. Evenly spaced points. ~ are linear combinations of the function values at the pOints. theta. x. The locations of the collocation points have a large impact on how well the method works.S which agrees with Equation A. sxdot] = \ ddasacC'fi rstorder". xdot. Figure AA shows values of a function c (r) at five collocation points. To solve differential equations and boundary-value problems (BVP).5 Figure A. eaO]. and we ohtain a more accurate solution. [x. Then we then find the values of the function such that the differential equation is satisfied at the collocation points. we apprOximate a function by passing a polynomial through values of the function at selected points. Notice that only when the initial condition is the parameter is the sensitivity initially nonzero. ca = x. sxO = [0. A. The selected points are known as collocation points. Choosing the points as zeros of a member of a family of orthogonal polynomials turns out to have much better properties. ~ OCA/OCAD. The following Octave commands show how to use ddasac to solve Example A. Boundary-Value Problems and Collocation (colloc) o Stewart and Caracotsios produced the code ddasac to perform sensitivity calculations conveniently [3. If we increase the numher of collocation points. endfunction xO = caO. This sensitivity also is plotted in Figure A.L____ ~L- ____ ~ ____ ~ o 1 234 5 dcadk dcadcaO nthesx. _1L-____ ____. xdot -k*ca. SX.5 " S2 '5 -' o -0. theta = [k. theta) caO k x(I).588 Computational Methods A.

"left".14 to obtain r=O d 2 c 2 de 2 -+---<1> c=o dr 2 r dr The effectiveness factor is given by T) _ . Bijej n.2 -' 0 0. we satisfy the boundary conditions de I dr = at r = 0 and c = 1 at r = 3. The function is therefore ° function retval = pellet(c) global Phi A Aint Bint Rint n npts length(c): cint = c(2:npts-l): ret val (1) = A(l.3: Single-pellet profile Consider the isothermal.2 ° 1. At the two collocation endpoints.S Boundary-Value Problems and C. We solve for the concentration profile for the reaction-diffusion problem in a catalyst pellet to illustrate the collocation method.S: Dimensionless concentration versus dimensionless ra~ dial position for different numbers of collocation points.ollocatlon (co I lac) computed as linear combinations of the values at the collocation paints 1 0. Solve the problem for </> = 10.4 2 The strings "1 eft" and" ri ght" specify that we would like to have collocation points at the endpoints of the interval in addition to the zeros of the orthogonal polynomial. colloc(npts-2. (d) Plotthe concentration profile for nc = 5. "right"). based on tbe methods described by Villadsen and Michelsen [8J. first-order reaction-diffusion problem in a spherical pellet (A.14) 3 Figure A.4 nc nc nc nc = 30 = 10 =5 = 50 i""l f [R A B QJ = 1 f(c)dc = o 2.5 1 To obtain the locations of the collocation points and derivatives and integral weighting matrices and vectors. How many collocation points are required to reach accuracy in the concentration profile for this value of <1>. (e) How many collocation points are required to achieve a relative error in the effectiveness factor of less than 10~5? Solution First we perform the differentiation in Equation A.6 0. 10.30 and 50.<1>2 dr r-3 _ !de I (c) Compute the concentration profile for a first-order reaction in a spherical pellet. Yj 0.5 r -0. :)*c: . c 0.8 d2 d: I = 2. Qj/(ej) j""l n. Example A. we use the function co11 OC. We define the following Octave function to evaluate this equation at the interior collocation points.590 Computational Methods A. -0.

It is 0 not exact because the reaction is second order.c(npts).. 0 .8.01 0.. endfunction .+ 2*Aint*c. Notice the effectiveness factor approach gives a good approximation for the bed performance. If one is only interested in the pellet reaction rate.592 Computational Methods A.. Figure A. .\ The collocation method produces algebraic equations for the pellet profile as for the previous example.6 shows the relative error in the effectiveness factor versus number of collocation points.. about 16 collocation paints are 0 required to achieve a relative error of less than 10.B (A. Figure A.6: Relative error in the effectiveness factor versus number of collocation points.0001 Solve the model exactly and compare the numerical solution to the solution obtained with the approximate Thiele modulus and effectiveness factor approaches shown in Figure 7.S shows the concentration profiles for different numbers of collocation points. Also shown are the simplified effectiveness factor calculations for this problem from Example 7.5.6 Example A. second-order reaction A.97 .4: A simple fixed-bed reactor problem Here we set up the collocation approach of this section for the single.J Hounaary~vaJue PYOlJlems ana LOIJOcarwn (CO I IOC) 1 0. A magnified view is shown in Figure A.7.. We use a DAE solver to integrate the differential equation for NA and these coupled algebraic equations.. Solution The model for the fluld and particle phases can be written from Equations 7. We require about nc = 30 to obtain a converged concentration profile.26. The results for N A versus reactor length using 25 collocation points for the pellet are shown in Figure A.84-7. dCA = 0 dr retval(2:npts-l) Bint*c.15) . retval(npts) 1 . OJ le-06 le-08 "" le-lO le-12 le-14 le-16 CA = -' CA r=R r=O 5 10 15 20 25 30 nc Figure A.jRint Phi A2*ci nt.

S: Magnified view of Figure A.594 Computational Methods A. "' 7 § ~ 0 6 5 4 3 1) = 1) = H20 + H = OH+CO = C02 + Hz Hz + OH C02 + H 1 [1 1 ] =.j. Compare to Figure A.6 Exercises Exercise A. Exercise A. (a) Solve the problem with ddasrt and compute tbe time tl at which the reactor fills.3<P -' 1 i!> Take the second and third reactions as an independent set and estimate the reaction rates from the same 2000 production-rate measurements used in Example A.18. in which monomer is converted to polymer in a semi-hatch reactor.7.l: Water gas shift reaction and production rates Consider the water gas shift reactions HzO + CO 8 :::. show only that the sensitivity computed by taking the partial derivative of the solution to the .6 ~ 2.I. 2.7: Molar flow of A versus reactor volume for second-order. (b) Calculate the solution for both addition policies and compare your numerical result to the analytical result in Figures 4. Note: instead of solving the sensitivity differential equation. Figure A.3: Sensitivity for second-order kinetics Repeat the analysis in Example A.15-4. isothermal reaction in a fixed-bed reactor.2 for the differential equation describing a secand¥order1 isothermal. Explain the differences and similarities in the two figures. Compare your answer with the result tl = 11.2 min given iu Chapter 4.I.8 "(3 "' § 2.4 2. (a) What value of [Y2 Y31 produces the error-free production rate corresponding to [rl rz] = [1 2] 2 0 50 100 150 200 VR (L) 250 300 350 400 (b) Create 2000 production-rate data points with measurement noise as in Example A.2: Semi-batch polymerization reactor 3 Consider again Example 4. dt CA(O) =C.40 Figure A.2 Exercise A.D txerCises 10 9 A. two approximations and exact solution.3. tanh 3<P .I. (c) Estimate [rz r31 from these data and plot the result. irreversible reaction in a constant-volume batch reactor 300 320 340 360 380 400 deA = -kc..

<!> ~ 0. 1 0.4 0. Solve the problem for a range of dispersiou numbers. c 0. first-order reaetion-dlifusion problem with an ODE solver. 50.5 r 2 2.1.74-7]6. y ~ 30.g. Can you get a CSTR-like solution for large values of the dispersion number? Can you get a PFR-like solution for small values of the dispersion number? Do you have numerical accuracy problems at either of these limits? differential equation.2 0 0 0.6: Comparing shooting and collocation A Solve the isothermal. unstable middle (B).1. 10. Equations 7.g. and upper (C) steady states.596 Computational Methods A.9. (b) Compare and contrast your c(r) solutions to those in Figure A.9: Dimensionless concentration versus radius for the non~ prefer and why? isothermal spherical pellet: lower (A). increasing f3 Using orthogonal collocation. 1.00 L Compare your result to the calculation presented in Figure A. Explain and justify your anSWer.6 = 0.. forthe folloloVing parameter values: f3 ~ 0.4: Multiple solutions to the Weisz-Hicks prohlem Using orthogonal collocation. re-solve the Weisz-Hicks problem for the following parameter values: f3 ~ 0.!. how many solutions do you expect.6 0. satisfies the sensitivity differential equation and initial condition. Example 8. CA(t). and the shooting method. using <I> ~ 0.5 1 y f3 <P ~ 30 de = 0 = 0.. Ii ~ 2.2 Exercise A. Exercise A. Which method do you Figure A. Exercise A. <I> ~ 0. compute the solution c(r) to the Weisz-Hicks problem.8 B .6. . (a) How many solutions are you able to find? Given the results in Figure 7.7. (c) Solve the problem using the collocation method. an algebraic equation solver. Y ~ 30. 1.5 3 (b) Compare e(r) from the shooting method with the analytical solution.19. n = 1. Exercise AS Weisz-Hicks problem. 6 Exercises 1. Use initial guesses for the elr) profile based on the results in Figure A.~ (rqdC) _ _ 2_<I> 2 c" ~ 0 rq dr dr n+ 1 c = I.001 dr ' - C (a) Solve the prohlem with the shooting method for the first-order reaction in a spherical geometry.7: Transient solution of the dispersed PFR Revisit the transient PFR example of Chapter 8.001.

A. Numerical Solution of Initial· Value Problems in Differential·Algebraic Equations. Eng. 42. Spring 1995. 323. Eaton and J. L Michelsen.B. S. 11 Aris. 56 Benzene pyrolYSiS. Stewart. 327.. J. 1985. W. 529. 17 Bard. 11 Branch lower. Y. 82. [7J A C Hindmarsh. (40):11-18. 383. J. 246. 66 . S. 218 Bomben. 11 Brenan. AAC. 478 Bimolecnlar. March 15-17. 383. 358 Biegler. 112 Beach.. 226 Activity. 434. E. 276. Eaton_ Octave: Past.. H. 298 upper. 240 physical. S. Prentice-HaD. G. 72 Beenackers. http://www. K. 358 Ammonia synthesiS. 536 Attainable region.W.. 503 Batch reactor. 131 thermal decomposition. 80. 241.D.M. 2001.. Technische Universitiit Wien. R. CACHE News. 358 rate of. K. 239 dissociative. Eng. 1995. Scientific Computing. 191. 200 Adamson. 358 Buffham. 193.493 Bischoff. 502 _ . 383.J. 27. F.. 463. RA. Brenan. Leisch. G.B. 316 Bakker. LJ. 298 middle. A Saunders.. M. 238. 518. Proceedings of the 2nd International Workshop on Distributed Statistical Computing. Wrigbt. 459.W. WE. 551 Brunauer.. Amsterdam. 429. Caracotsios and W. 193. ODEPACK. M.. 513 Avogadro's number. 524. Abbott.. a systematized collection of ODE solvers.. Rawlings..R. User's guide for SOL/NPSOL (Version 4.. S. 238 Adiabatic operation.M.0): A Fortran package for nonlinear programming. Gil!.D. E. Camput Chem. 2001. J. 450 Bodenstein. 519. 239. 33 Biot number. J. M. C. 63 Acetone. 13 Bartle.. M. 316 Butene. [8J J ViIladsen and M.. 395 Arrhenius rate expressionl 204 Atkinson. 246 Boundary-value problem. 469. 238. E. [6J P. 298 Breiland.. 147.. 1989. D. 253. editor. editors.. 165 Baccaro.. 1983. L Campbell. Murray.E. 186. 569 chemical.A. Technical report.. North-Holland. 241 Aftalion.F. 488 Banerjee. 584 Bruemmer. 327 Alchemists.P. 9(4):359-365.E. 14J J.F. 280 Adiabatic temperature rise. Sensitivity analysis of initial-houndaryvalue problems with mixed PDEs and algebraic equations. 65 Activity coefficient.E. 532 Boyce. 478 Bush. 455. pages 55-64. Elsevier Science Pnb· lishers. 477 Auger electron spectroscopy.. LT.. R. 239. Stewart. 1986. 375 BET isotherm. K..P_. A. D. S. Bell. T. LD. 152. In R. Petzold. W. Englewood Cliffs New Jersey. [5J j. 540 Bartholomew. 326 Adsorption. 19(9):1019-1030.. 11 Bansleben. BA. 12] M. 513 Borwanker. 502 Belhnan. 364. 242. E. New York. 166 Boudar!.T. 498 Burri.cLtuWien. Camput Chem. Sensitivity analysis of initial value problems With mixed ODEs and algebraic equations. 447.. AC. Stepleman... Solution of Differential Equation Models by Polynomial Approximation. W.Bibliography Index IIJ K. E. 60 Alumina. B. Octave-a high level interactive language for numerical computations. 349 Applied Materials. 13] M. R. Systems Optimization Labo· ratory.. In L Hornik and F.P. 469 Bellamy. technical report SOL 86·2. Stanford Unlversity. 283 Adler. Austria. S.. 394.· 2001/Proceediogs/. and L R. 249 heat of. ISSN 1609-395X..M. 362. T. 469 Adomaitis.. 405 Bird. 495 Beer's law. 1978. see BVP Box.. present and future. see Reactor. 386. Department of Operations Research. 155 Bessel function. Caracotsios and W. 117 Bramblett... 447 Bodenstein number. and M. W. A. 503 Bell. II 316 Bailey. K. 72 Benson.

28. 13 7 Flow streams. 32 Czarnowski. 117 nonhomogeneous.. A. 358 Catalyst. ~ assumption. 147 . 377 reaction limited. x. 469 Chemica1 kinetics. 503. ]. 532 Experimental reproducibility.. 506 First-order irreversible reaction. see CVD Chemisorption. 355 batch reactor. R.E. 218 Chapman-Enskog relation.395. G. T. P.G. 316 Chapman.478 Error function. 358. 95 ideal-gas.. P. 205 Duke. 523. 131 CYD. P.286.B. J. 11.. 357 Extinction point.6. 116 Fixed-bed reactor. 156. 11 Dispersed flow. 363 BVP. 461 Cumene hydroperoxide. 66. 360 coefficient. 191 unimolecular. 588 Carberry.M. D. 515. 412 Catalytic process. 292 Farrauto. 397 Deal. 104-106. 190. 397. 357. 444. see Delta function Disilane. 326 Clark. 17 Conservation of mass.. M. c. 531 CSIR. RD. 63 Denn. 409 Chang. 136 Continuous~stirred~tank reactor. 511 ChianeHi. see DAE Diffusion bulk.L.006 Cinar. 63. 575 Effective dlffusivlty. J. 395 Catalytic converter. 114 reversible reaction. 435. I. 446 Djega-Mariadassou. 242 Constant density. 515. 240. L.. AA. 355 Flame ionization detector. 355 Catalyst particle characteristic length.£" 33 Complete segregation. 250. 540.. C.. 275.. p" 216 Davis. P. 113 linear.. 63 DeAcetis. 390 external mass transfer.22. 478 Flow controller. 506 Fifield. 360 Connor. 358 Cauchy inequality. K. 59 Elementary reaction kinetics.S.]. S. 238. E. 455 Dantzig.286 PFR.477. S. 128.J. G. 246 Donev.478 Feller. 375 cylinder. 209 chemical. 455 Complex mechanisms. 380 nonisothermal. 15 Cbristofides. 82 Ellison. 548 Eigenvector. 226 Concentration gradient. 580 Equivalence principle CSTR. 11. F. 316 Energy states. 366 general balances..]. G. 448 Eftl. 262. 64. J. 361 Knudsen. J.. 164. 85 composition.V. 117 Dirac delta function.164. see CSTR Conversion. 189 equilibrium. effective.E. F. 18 Enthalpy. 232 Experimental deSign. 101 Collinear collision..]'. 374 diffusion limited. S. 361 Chauhan. 164. 397. 510 Carbon monoxide adsorption. 282. 390 slab. 519. 238. 189. 151 Convolution integral. F.W. 244 oxidation. 216 Christophorou. 360 Diffusional flux.R. 186.. le. 316 Dryer. 497 Field.. 484 DlPtima.L.B. 11 Decomposition thermal. 119 Dlffereutial-algebralc equation.. 277.. 180. 187 External transport control. 536 Exponential of a matrix. 513 Davydov. . 191 Delta function. 164. 239.. 92 reaction. 548 Ekerdt.334 Density constant.0.P. 437 Denbigh. 200 Chemical vapor depOSition. A. 529. 18 Floudas.478 CSIRs in series. M. 112. 276 Flowrate volumetric. JA.317 semi-batch reactor.589 Campbell.W.584 Danckwerts boundary conditions...G. R. 39. J. ]74. 363 Hougen-Watson kinetics. 504 Flint. 436 Corrigan. J. E..439 CSTR equivalence principle. 478 Emmett. LE. 13 Fast time scale. 239. R. 79. 447 Danckwerts.... 404 Eigenvalue. 501. 107 Cavendish. 385 first·order... 397 Catalysis.. see Mass Conservation of sites. 106 Daoutidis. 63.. 129. M. 193 Deviation variables. 195 Elliott. 503 de Azevedo.374 spherical.. 137. 263 Cortright. 360 Effectiveness factor.G. 447.L. 529. 238 Clausius-Clapeyron equation. 537 Configurational diffusion.. 82 de Pablo.C. 392 n th-order reaction. 85 potential.. 362 combined. 88.. 213. 92. 371. 193... 366 steady-state multiplicity... 239 Ethane pyrolysiS. LG. 394. 360 Confidence interval. J..510 pellet.. 259 DAE. 278 Equation of state.F. 90. lM. M. 112 Dihora. 197 Coltrin. 10. 136 Desorb. 68 constant. 491 Differential equation analytical solution. 66 multiple reaction.2.N.. 5. 358 End-point control. 159. 531 Eaton. 239 Enquist. 117 nonlinear. M. Ill. 584 Caracotsios. A.O. 536 Douglas. 212 Feinberg. JR. W_. 136 nonconstant.. 277 CSTR.. M.W... 253 Element conservation. 510 Dumesic.600 Index Index OV I Butt. 200 Engel. 552 Energy balance. 13 5 Equilibrium approximation.R.

. 429 Levenspiel. 514 Gauss·Newton approximation. K. 316 Gamma function.5. see Reaction Heat·transfer coefficient.. 477 Godorr. 11 Grab. 450 Langmuir·Hloshelwood kinetics.L.. 192 Hayward. R. 217 Garver. 92.519.. A H" 488 l'lair. 437 Incompressible fluJd. AB. 316 Greenlief. 27.. lC. 355 intrapartide transport control..E.T. 355 equation. S. 216 Line broadening.aplace transform. B..Y. 553 Ura. 503. T.147. OA. 81. RH" II Lee. 17 Guckenheimer.. 279. l. 15. 82 LLDPE. 66 Isoenergetic. 545 Herzberg. 539 Germane. K. D.. 397. 355. 220 LDPE. 447.. A. 360 Kokossis. 226. 503 Gas·phase reactor.J. 195 of formation. RJ.5. 260 Hessian. D.. 447.. 474.L. 81 Henson. 65 coefficient. M. 477 Hill. A. 191. 66 mixture. 238.c. 323 Fuel cells.D.50. 265 Glasser..5.M. M. AC. 493 Haidari. 11 Lubben. 17 Henderson. 506 Laidler.. 409.. 551 Hancock. 63. I. 241 Langmuir isother!l:l. S. 503 Hamer. I.. 263 Gas chromatography. 357 Kirk. D..W. 86. 216 Kumar.... ]. 316 Kinetic control.160. see IVP Intermediate species. L.15 Loeb. see Hougen·Watson kinetics I. Y. 74 Leach. K.. G...447. CN.iside function..429.R. 502... 355. S.383..C. 15 Le Chatelier's principle. 18 Hougen.. 513 428 Huang. 497 Jerome. 80.L. 519. 117 Horn. 305 Un. Q.F.450. 59. 478 Laminar flow. 301 Jensen's inequality. 218 internal combustion engine.602 Index Index bU5 Franklin.I. J" 302 Hagen·Poiseuille flow. J.. 66. 192 Kealey. 134 Ideal·liquid mixture. 518 Hepatitis B virus. 17 Froment. 262. 33 Kenney. 316 Hunter. 503 Leibniz rule. 223 Industrial case study. 506 Kee. 28 Internal energy. 262 Kisliuk. 11 Gauss. N.V.M. 80 Ideal·gas equation of state. 105. 253. 380. 478 Krug.532 Hunter. 302 Homogeneous. P.. CF.E. 477 Graziani. 148 equilibrium.W. 292 ImpuJse function.. 493 Langmuir adsorption. 0.. 528. 580 difference.L 238. 532 Hydrodesulfurization.. 409 Helmholtz energy. 524.519. P. 319 induction tIme. 11. N..J. ]. 246 Kissen. 12. 393 Hildebrandt.. 82 Lightfoot. 281 Hea. 17 Knudsen flow. 246 Hindmarsh. 355 catalyzed reactions. 192 Gaitande. see Energy Interstitial region. B. 166 Glandt. CG. 35-39 Liquid chromatography. P. 205 Holmes. 443 Heftmarm. R. D.. 495 Kyle. S. CC. R. S. 27. R.. S. H. 13 Heat capacity.B. B.M" 18 Kumar. 278 partial molar. 551 Infrared spectroscopy. 477 Horwich. 187. 357 Intrinsic parameters. M.S 12 Unear algebra fundamental theorem. 358 Gates. 200. 238. 197 IVP.5. 241 Low·index crystal planes.. 474. H. Ideal mixing. 503 Grossmann. 106 Johnston. 105 Linear independence.D" 316 Harker. 435 ImpuJse response.571 nonllnear.. 11 Gibbs energy. 82 Hicks. 390 Isobntane.5. 11 Griffin. D. 582 Hochgreb.L. 478 Grubbs.M. E. 147 Gas·solid systems. 551 Johnson. J. F. 179 Gaedtke. 246 HDPE.P'. 503 lnitial·value problem. IS HDS.. E. 544 Least-squares estimation.L. D. E.. 474. 156.N. R. 194. At.195. 104 Gibbs·Duhem relations. 445 Gandhi. 217 Kunlgita. 2 Ideal mixture. 57. 238 Gates. 200 Henderson. K.. 517 unknown variance.474 Uchlenhaler. see HDS .F.K.. LL. M.232 Lakshmanan. W. 18 Lombardo. 18. E. 28 Fugacity.493 Limit cydes.. 527 weighted..... 82 Furusawa. C.. 194. 238 Lu..395 Jacobian matrix.G" 63 Lafferty. 503 Hegedus... 75 temperature dependence.. 539 Heterogeneous. AC. 581 Gillespie algorithm. 152 Hougen·Watson kinetics. J. G..G. P. 78 Heat of reaction.£. D" 11 MacDonald.H.L.. 18 Hudson. CT.524. 243 Langmuir. 64. 76 Gibbs. 506 Friedrich... Kodak. T.N. J. 318 Gill.. 355 reaction. N. 524 Lee..364. 134 Ignition paint. 278 Henry's law. 165 Gillespie. 358 Least squares.O. J. 502 Herzfeld..

P.347 Polyethylene high-density. see PFR Plummer. R.. 193 PDE. 33. 397 Pujado.R. 37. 128 Oxidation carbon monoxide.G. 358 Partition function. 339. 57 Poling.. 358 Pilling.J. 466. see HDPE linear-law-density. T.575-597 Matranga. M. 316 Margules equation. 122 Null space. 542 Moore. M. 79.589 Oscillations. 397. V..' free-radical. 33 Miyauchi. see l. 159. 82 Nonisothermal pellet. 195.47. 581 Myers. 575-597 ODE. 270 Nishimura. 82 Marshall. 179 Modell.. 47. 434. 512 Mass-transfer coefficieut external. 217.. 27. 278 conservation of. 216 O-xylene. 200 Parulekar. 5. S. 53 transpose. J.P.. lSI. 35 Maximum mixedness.M. 542 Ordinary differential equation. 186.450 Pearson. H. 543 Partial differential equation. T. 195.c. 386 Material balance...253 Michelsen. 280 Plausibility argument. R. 82.218. 80.. 529. 262 Rate . 192 Phthalic anhydride.. 166.G. T. 403 Periodic table. 502 Natal-Santiago. V.43. 219 Pebsworth. 104-105 Phenol. 197 Polimy.. 174. 15 Octave. 584 Ogg. 458. 250 Palmberg. 327 QSSA. E. 12 Miller. 465 Manousiouthakis. V. W. J.. M.131 Photochemical smog.R. 132 Pulse response. W. 514 normal. 197. 383. D. 33 stoichiometric. 216 Petersen. 258 Oh. 489 Pedersen. 465 Overall selectivity. 197 Poynting correction factor.R. 218.. 225..DPE Polymerization .. 347 O'Malley.F... 137. S. R. 151. 362 Pore volume. 27. 454. M.C.. 113 O'Connell.. 106 rank.S. 454. 493 McCulloch. 29.E. 514 density. 128 Polany. 317.53. 463. 60 Perturbation parameter.457. 358. E. 339. 82 Molar floWTate.D. 290 Pore structure.. 194.226 QSSA species.M.524 Gaussian.G. 397 total.339. 469. II Podkolzin.. 219 Morse function. 358. 179 Noll.....s. 478 Marek. lM.514 Production rate. 193 Metal oxides. 79 Mass balance steady·state.W. 136 Nonhomogeneous equation.M. 265 Matrix eigenvalue. 134. 355 MATLAB. 200 Monolayer. 196 Motooka..213. 77 Particle denSity. J.. 193.E. 505..1.189. 223 Pitzer. 347 Probability.439 Mechanism. 240 Monte Carlo shnulatlon. K. 226 Quasi-steady·state assumption. 575-597 Numerical solution..... 409 Ohm's law. W. 513 Multiple-reaction system. see Matrix Rase.W. 498 Niemantsverdriet. L. D. 429 Petzold. 515 exponential. 360 Mean residence time. 504 Olefin polymerization. see llDPE lOW-density. 358 Porosity.W. 12 Mean free path. RE. 11 Moulder... 584 PFR.. 197 Pellet density. 197. M. 81 Prausnltz. 37.B.442 PFR-CSTR arrangement. 82.. 358 Methane synthesis. J. 465 MicroscopiC reversibility. 323 Occelli.. 159. 18 Raff. 450. 119 nth-order irreversible reaction... M. K. see PDE Partial molar properties. 233 Molecular partition function.453 Phase equilibrium. 297 Plug flow. L.H. 17 kinetiCS. 18 Mass basis..F. 358.. 590 Microelectronic materials. 151 Murray. 531 Point selectivity. P. R. L.. F. 323 Physisorption. 435 Probability distribution chi-square. 82. 450. sustained..225... 27. 529. 281. 555 differential equation models. 112 Propylene oxidation.A. 403 Porter. 453 PFR-CSTR comparisons. 519 F. 437 Pyzhev.A.. 2. x. 195. K. 295.37 sparse.. see PFR Plug-flow reactor. see QSSA Racanlello. 531 Nauman. 582 implicit. 32 Micromixing. 112. S. 538 multiple measurements. 96 Phase rule. 393 Nonlinear differential equation. 447 Rank. 488 Mason. 126 Parameter esthnation. J.435 function.. 513 Parallel reactions. 199 Middlebrooks. 515 eigenvector. J. 136 Mass spectroscopy. 240. 105 Numerical methods. 22. 265 Nakamoto. AL.L. 15 Optlmlzation. 146. 34. H. IS. JW..604 Index Index bU::> Macromlxing. 230 gas-phase..R. 105 Nonconstant density. 45. 447 Pathway. 538. M. 111 transfer. 218. 47. 495 Ottino. see ODE Orthogonal collocation.L.N.A. 129. 381. 362.. 117 NOnldeal mixtures. 197 Potential-energy surface. 113. 305 Otake. 197 Ramkrlshna. L. j. 532. 17 Peclet number.. 59. AB. 187 positive definite. 230 Poore.

.H. L. 581 Schieber. L.277.. 149. K. 226 equilibrinm. 295 Somorjai.. 447..D.O. 278 Sinke. 67. 123 independent. 153 first-order. 339. 141. 190 surface. 279 heterogeneously catalyzed.M. 193. 128 point. 238 water gas shIft.112 Rate constant effective. 216 Selectivity overall.. 316.H.. 126 photochemical. 226 laws. 4. 439 Shon. 47.383. 63. 290.442..403 gas-phase. 463 semi-hatch. 469 convex. 27 . 228 first-order. 295 upper-branch. 44. M. A. 390 expression. 113. 141 reactor. II Russell. W. 374. 197 j 11 tubular. 518 Sifniades.. 392 linear combination. 29 Reaction bimolecular surface. 209 Rawlings.14. 406 second-order.F. 287. 469. 281 isothermal. L.. J. 326 batch.P. 191. 588 Sticking coefficient. 28. 226. 280 continuous-stirred-tank.. 164 Stoichiometric coefficient. F.CA. 290 autothermal. 316 Reactants. 225 kinetics.5. 98. 361 Step response. 132 Shah. 36 inhibition..Z. batch. W. J. 506 fixed-bed. 540 Spherical pellet. R. 41. 116 gas-phase. 43. 18 Sobol. 166 Shapirn. 56 Srivastava. 42... 28. 50. 66 Steady-state concentration. 106 Sherman. 380 Slemrod. 366.E.. 190. 5J. 192 polymerization. 316 Schreiber. 14 7 solution.. 510. 43.219. 281 imperfectly mIxed.L. fixed-bed. C. 577 rate expression.316. 80. J.112 reaction. 595 Reaction-rate constant. 72 Schwarz. 290 Stefan-Maxwell relations. 148. 11 Shinnar. 541. 189. irreversible. 106 Shapley. 434. 259 Schwarz inequality.. 484 Razon. 30 Stoichiometric matrix. B. 113 production. 35.. J.F.. 218. 42.W. heterogeneous... 114. J. 189. 128 Semi-batch polymerization.566 parallel. M. 507 network. G. 347 Smith. 3. 568 independent. 37 nitric oxide. see Matrix Stoichiometry. W. 41. 371 first-order. 28 chemical vapor deposition(CVDj. P. 11 Site balance. 513 Rice.M. 295 middle-branch. 493. 510 S0renson. T. Y_. 269 Schumacher. G.355. 11 combinatorial. 197 Scbuit. AM. 39.. 199.147. 292. 27. 203..A. 141. 541 Reaction-rate expression. 18 Slab. 21?Smith. 518.J" 316 Schreiber. R. 494 Rate-limiting step.E. 510 Stevens. 238... 450 Single ctystal. 195 Second-order irreversible reactioll 119.429. 260 Rigorous argument. 100 Slnniah. 111. 53 Reactor adiabatiC. 60 Sims.C. irreversible.. 300 Standard state. 64. 230 rate. see CSTR differential. 36 liquid-phase.F. J. B. 147. 4. G. 576 lower-branch. 133 Seubold.133. 137. 371 Srmivasan. 513 Stivers.. J.S. 2. 349 autothermal plug-flow. M. 132 Silicon-germanium alloys. 129 Residence-time distribution.224. 551.606 Index Index tJUI constant... 406 nonisothermal. 137 constant~pressure. 82 Saunders. 120 Seeger.. N. 112. irreverSible. S.364. 436 Stephenson. 558 Reactions in series. 340. 377. 195.A. R. 53. 122. 112 well-stIrred. 316 Schork. 63 Schmitz. 441 Russell. 32 coordinate. 395 operation. 279 constant-volume. reversible. 123 intermediates.. 290. 295 unstable steady-state. 85 extent. N.A. 575 Ray. see PFR segregated. 197 Stochastic simulation. 18 Stability analysis. 509. H. 238. 249 Skalka. 508 experimental. 10 Silver. 510 Riach. t. LA.M. 371 first-order. 238 Single-phase system. 119 statements.. 480 isothermal. 513 Solution least-squares.P. see RID Reynolds number.552 single~wafer.. 28. 118 second-order. 200 elementary. F.E. volume element. 507 chemical vapor deposition (eVD).B.E...K.. 241 Stickle.J. 466. 114. 250 cbemical.. 48 linearly independent. 300 RID. M. 283. 226 reversible. 508 heat of. 118 mnitiplicity. L. 201. 17 Stewart.L. 366. 7 constant-mass. 46. 280. 239.B. 18 Segel. 34 nth~order. 540. 216 Slow time scale. 228 Rate expression concave. 366. 346. 198 coordinate diagram. 199 Sandler. G.112. 275 plug-flow. 46. P. R.. 35-39 isomerization.L. 29. 180 Saddle paint.

67. J. R. M. ]. 195.. G. 360 Surface order. J. 28. T.M" 495 Van Welsenaere.. 461.M. R.. 512 Transpose. 327. 100 Suda.W" 238.J. 194 Tomiin. T.E. 223 Tortuosity factor.276 Wright. 18 Younkin.. 524 Titania. TK. 218.. 246 Troe.W.. 394. D... 7 Taylor dispersion. EW. 358. 11 Stull. I. G. 240 Vacant surface sltes.204 Turbulent flow. 33 UOP LLC. LK. 18 Surface diffusion. 367. D. 404 Thiele. 393 Weisz-Hicks problem. S.D.. 37. WE. 100 White. 270 Van Swaaij.j..R.R. 495 Weston.M" 152 Weast. 510 Thompson.C.. 403 Volume balance.G. 197 Tiao. 458. 98.E. CM.608 Index Index 609 overall. 435. 590 Virus mOdel. 238 Surface-site balance. 367 Thodos. 192 1'5T. 447 Teller. 358 Tolman. RW. Y. 393 Weller... F. W. 484 Villadsen. 179 Weber.E. 300 van Heerden. 98. 513 Yang. see Matrix Trapnell. 194... 201 Transition-state theory. 202 Villa. J-W.S" 92. 59 Streetman. Inc. RA. 202 Vibrational partition function. B.. LF" Jr. 241 Surfaces. 545 Void fraction. P. 269 van Santen. 82 Teymour..M. 466. 132 Vibration. 156. 327 Tester. 510. 512 X-ray photoelectron spectroscopy. 331 Van Ness. 136 Volumetric !lmvrate. 106 Wilson.P. 581 X-ray diffraction. 512 Water gas shift reaction. 504 Thermodynamics. A. 455. l... 478 Westerterp.232 Work... R.W. M. K. A. 67. 17 Yu. J. 63 van Reijen.. 218 Unimolecular. 316 Thermal conductivity. 146 Turimyi. S. 478 Wojciechowski. G. W. 11 Summers... 510 Thomas. 595 Watson. 459. 17 Zwietering. see TST Transmission electron microscopy. K. 363 Total mass balance.N" 455. R. 371. 195 Westrum. 323 van't Hoff equation. 463 Weisz. R. 358 Vaughan..S.. LL. 78 Vanadia. 3. 63 Thiele. F. 199. 239 van Heerden diagrams.c. 18 Zeolite. MR. C. 290 Vacant site. 223 Underhill. 15 Uppal. 367 Thiele modulus. 450 Westerberg.l. 218... 493 - . 295.L. 503 Yield.C. R. 510 Ziegler catalyst. Re. E. A. 147 Wachs. 128 Yin.. E. 513 Weinstein. 190 Strang.B" 358. 519.. 444 Transition-state complex. 237 Symyx Tecbnologies. 136 Tracer molecule. 397 Thomas.. 358 Temkin. W.