Combustion Notes PDF

LECTURE NOTES ON
FUNDAMENTALS OF COMBUSTION
Joseph M. Powers
Department of Aerospace and Mechanical Engineering

University of Notre Dame
Notre Dame, Indiana 46556-5637
USA
updated
July 23, 2010
2
Contents
Preface 11
1 Introduction to kinetics 13
1.1 Isothermal, isochoric kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.1.1 O O2 dissociation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.1.1.1 Pair of irreversible reactions . . . . . . . . . . . . . . . . . . 16
1.1.1.1.1 Mathematical model . . . . . . . . . . . . . . . . . 16
1.1.1.1.2 Example calculation . . . . . . . . . . . . . . . . . 22
1.1.1.1.2.1 Species concentration versus time . . . . . . 23
1.1.1.1.2.2 Pressure versus time . . . . . . . . . . . . . 24
1.1.1.1.2.3 Dynamical system form . . . . . . . . . . . . 26
1.1.1.1.3 Effect of temperature . . . . . . . . . . . . . . . . . 30
1.1.1.2 Single reversible reaction . . . . . . . . . . . . . . . . . . . . 31
1.1.1.2.1 Mathematical model . . . . . . . . . . . . . . . . . 31
1.1.1.2.1.1 Kinetics . . . . . . . . . . . . . . . . . . . . 31
1.1.1.2.1.2 Thermodynamics . . . . . . . . . . . . . . . 33
1.1.2 Zeldovich mechanism of NO production . . . . . . . . . . . . . . . . 37
1.1.2.1 Mathematical model . . . . . . . . . . . . . . . . . . . . . . 37
1.1.2.1.1 Standard model form . . . . . . . . . . . . . . . . . 37
1.1.2.1.2 Reduced form . . . . . . . . . . . . . . . . . . . . . 38
1.1.2.2 Stiffness, time scales, and numerics . . . . . . . . . . . . . . 49
1.1.2.2.1 Effect of temperature . . . . . . . . . . . . . . . . . 49
1.1.2.2.2 Effect of initial pressure . . . . . . . . . . . . . . . 50
1.1.2.2.3 Stiffness and numerics . . . . . . . . . . . . . . . . 52
1.2 Adiabatic, isochoric kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
1.2.1 Thermal explosion theory . . . . . . . . . . . . . . . . . . . . . . . . 54
1.2.1.1 One-step reversible kinetics . . . . . . . . . . . . . . . . . . 55
1.2.1.2 First law of thermodynamics . . . . . . . . . . . . . . . . . 56
1.2.1.3 Dimensionless form . . . . . . . . . . . . . . . . . . . . . . . 59
1.2.1.4 Example calculation . . . . . . . . . . . . . . . . . . . . . . 60
3
4 CONTENTS
1.2.1.5 High activation energy asymptotics . . . . . . . . . . . . . . 62

1.2.2 Detailed H2 O2 N2 kinetics . . . . . . . . . . . . . . . . . . . . . 66
2 Gas mixtures 71
2.1 Some general issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.2 Ideal and non-ideal mixtures . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
2.3 Ideal mixtures of ideal gases . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.3.1 Dalton model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.3.1.1 Binary mixtures . . . . . . . . . . . . . . . . . . . . . . . . 79
2.3.1.2 Entropy of mixing . . . . . . . . . . . . . . . . . . . . . . . 82
2.3.1.3 Mixtures of constant mass fraction . . . . . . . . . . . . . . 84
2.3.2 Summary of properties for the Dalton mixture model . . . . . . . . . 85
2.3.3 Amagat model* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
3 Mathematical foundations of thermodynamics* 93

3.1 Exact differentials and state functions . . . . . . . . . . . . . . . . . . . . . . 93
3.2 Two independent variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
3.3 Legendre transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
3.4 Heat capacity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
3.5 Van der Waals gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
3.6 Redlich-Kwong gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
3.7 Compressibility and generalized charts . . . . . . . . . . . . . . . . . . . . . 116
3.8 Mixtures with variable composition . . . . . . . . . . . . . . . . . . . . . . . 116
3.9 Partial molar properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
3.9.1 Homogeneous functions . . . . . . . . . . . . . . . . . . . . . . . . . . 119
3.9.2 Gibbs free energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
3.9.3 Other properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
3.9.4 Relation between mixture and partial molar properties . . . . . . . . 122
3.10 Frozen sound speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
3.11 Irreversibility in a closed multicomponent system . . . . . . . . . . . . . . . 125
3.12 Equilibrium in a two-component system . . . . . . . . . . . . . . . . . . . . 128
3.12.1 Phase equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
3.12.2 Chemical equilibrium: introduction . . . . . . . . . . . . . . . . . . . 130
3.12.2.1 Isothermal, isochoric system . . . . . . . . . . . . . . . . . . 130
3.12.2.2 Isothermal, isobaric system . . . . . . . . . . . . . . . . . . 135
3.12.3 Equilibrium condition . . . . . . . . . . . . . . . . . . . . . . . . . . 137
4 Thermochemistry of a single reaction 139

4.1 Molecular mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
4.2 Stoichiometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
4.2.1 General development . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
4.2.2 Fuel-air mixtures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
4.3 First law analysis of reacting systems . . . . . . . . . . . . . . . . . . . . . . 150
CONTENTS 5
4.3.1 Enthalpy of formation . . . . . . . . . . . . . . . . . . . . . . . . . . 150

4.3.2 Enthalpy and internal energy of combustion . . . . . . . . . . . . . . 153
4.3.3 Adiabatic flame temperature in isochoric stoichiometric systems . . . 153
4.3.3.1 Undiluted, cold mixture . . . . . . . . . . . . . . . . . . . . 154
4.3.3.2 Dilute, cold mixture . . . . . . . . . . . . . . . . . . . . . . 155
4.3.3.3 Dilute, preheated mixture . . . . . . . . . . . . . . . . . . . 156
4.3.3.4 Dilute, preheated mixture with minor species . . . . . . . . 158
4.4 Chemical equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
4.5 Chemical kinetics of a single isothermal reaction . . . . . . . . . . . . . . . . 164
4.5.1 Isochoric systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
4.5.2 Isobaric systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
4.6 Some conservation and evolution equations . . . . . . . . . . . . . . . . . . . 178
4.6.1 Total mass conservation: isochoric reaction . . . . . . . . . . . . . . . 178
4.6.2 Element mass conservation: isochoric reaction . . . . . . . . . . . . . 179
4.6.3 Energy conservation: adiabatic, isochoric reaction . . . . . . . . . . . 180
4.6.4 Energy conservation: adiabatic, isobaric reaction . . . . . . . . . . . . 181
4.6.5 Non-adiabatic isochroic combustion . . . . . . . . . . . . . . . . . . . 184
4.6.6 Entropy evolution: Clausius-Duhem relation . . . . . . . . . . . . . . 185
4.7 Simple one-step kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
5 Thermochemistry of multiple reactions 193

5.1 Summary of multiple reaction extensions . . . . . . . . . . . . . . . . . . . . 193
5.2 Equilibrium conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
5.2.1 Minimization of G via Lagrange multipliers . . . . . . . . . . . . . . 198
5.2.2 Equilibration of all reactions . . . . . . . . . . . . . . . . . . . . . . . 204
5.2.3 Zeldovichs uniqueness proof* . . . . . . . . . . . . . . . . . . . . . . 206
5.2.3.1 Isothermal, isochoric case . . . . . . . . . . . . . . . . . . . 206
5.2.3.2 Isothermal, isobaric case . . . . . . . . . . . . . . . . . . . . 213
5.2.3.3 Adiabatic, isochoric case . . . . . . . . . . . . . . . . . . . . 216
5.2.3.4 Adiabatic, isobaric case . . . . . . . . . . . . . . . . . . . . 220
5.3 Concise reaction rate law formulations . . . . . . . . . . . . . . . . . . . . . 221
5.3.1 Reaction dominant: J (N L) . . . . . . . . . . . . . . . . . . . . 221
5.3.2 Species dominant: J < (N L) . . . . . . . . . . . . . . . . . . . . . 222
5.4 Onsager reciprocity* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
5.5 Irreversibility production rate* . . . . . . . . . . . . . . . . . . . . . . . . . . 231
6 Reactive Navier-Stokes equations 237

6.1 Evolution axioms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
6.1.1 Conservative form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
6.1.2 Non-conservative form . . . . . . . . . . . . . . . . . . . . . . . . . . 239
6.1.2.1 Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
6.1.2.2 Linear momentum . . . . . . . . . . . . . . . . . . . . . . . 240
6 CONTENTS
6.1.2.3 Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240

6.1.2.4 Second law . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
6.1.2.5 Species . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
6.2 Mixture rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
6.3 Constitutive models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
6.4 Temperature evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
7 Simple solid combustion: Reaction-diffusion 247

7.1 Simple planar model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
7.1.1 Model equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
7.1.2 Simple planar derivation . . . . . . . . . . . . . . . . . . . . . . . . . 249
7.1.3 Ad hoc approximation . . . . . . . . . . . . . . . . . . . . . . . . . . 250
7.1.3.1 Planar formulation . . . . . . . . . . . . . . . . . . . . . . . 250
7.1.3.2 More general coordinate systems . . . . . . . . . . . . . . . 251
7.2 Non-dimensionalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
7.2.1 Diffusion time discussion . . . . . . . . . . . . . . . . . . . . . . . . . 252
7.2.2 Final form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
7.2.3 Integral form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
7.2.4 Infinite Damkoler limit . . . . . . . . . . . . . . . . . . . . . . . . . . 255
7.3 Steady solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
7.3.1 High activation energy asymptotics . . . . . . . . . . . . . . . . . . . 255
7.3.2 Method of weighted residuals . . . . . . . . . . . . . . . . . . . . . . 260
7.3.2.1 One-term collocation solution . . . . . . . . . . . . . . . . . 261
7.3.2.2 Two-term collocation solution . . . . . . . . . . . . . . . . . 264
7.3.3 Steady solution with depletion . . . . . . . . . . . . . . . . . . . . . . 266
7.4 Unsteady solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
7.4.1 Linear stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
7.4.1.1 Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
7.4.1.2 Separation of variables . . . . . . . . . . . . . . . . . . . . . 271
7.4.1.3 Numerical eigenvalue solution . . . . . . . . . . . . . . . . . 273
7.4.1.3.1 Low temperature transients . . . . . . . . . . . . . 274
7.4.1.3.2 Intermediate temperature transients . . . . . . . . 274
7.4.1.3.3 High temperature transients . . . . . . . . . . . . . 277
7.4.2 Full transient solution . . . . . . . . . . . . . . . . . . . . . . . . . . 277
7.4.2.1 Low temperature solution . . . . . . . . . . . . . . . . . . . 277
7.4.2.2 High temperature solution . . . . . . . . . . . . . . . . . . . 279
8 Simple laminar flames: Reaction-advection-diffusion 281

8.1 Governing Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
8.1.1 Evolution equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
8.1.1.1 Conservative form . . . . . . . . . . . . . . . . . . . . . . . 282
8.1.1.2 Non-conservative form . . . . . . . . . . . . . . . . . . . . . 282
CONTENTS 7
8.1.1.3 Formulation using enthalpy . . . . . . . . . . . . . . . . . . 282

8.1.1.4 Low Mach number limit . . . . . . . . . . . . . . . . . . . . 283
8.1.2 Constitutive models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
8.1.3 Alternate forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
8.1.3.1 Species equation . . . . . . . . . . . . . . . . . . . . . . . . 286
8.1.3.2 Energy equation . . . . . . . . . . . . . . . . . . . . . . . . 286
8.1.3.3 Schwab-Zeldovich form . . . . . . . . . . . . . . . . . . . . 287
8.1.4 Equilibrium conditions . . . . . . . . . . . . . . . . . . . . . . . . . . 289
8.2 Steady burner-stabilized flames . . . . . . . . . . . . . . . . . . . . . . . . . 291
8.2.1 Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
8.2.2 Solution procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
8.2.2.1 Model linear system . . . . . . . . . . . . . . . . . . . . . . 295
8.2.2.2 System of first order equations . . . . . . . . . . . . . . . . 296
8.2.2.3 Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
8.2.2.4 Linear stability of equilibrium . . . . . . . . . . . . . . . . . 297
8.2.2.5 Laminar flame structure . . . . . . . . . . . . . . . . . . . . 299
8.2.2.5.1 TIG = 0.2. . . . . . . . . . . . . . . . . . . . . . . . 299
8.2.2.5.2 TIG = 0.076. . . . . . . . . . . . . . . . . . . . . . . 303
8.2.3 Detailed H2 -O2-N2 kinetics . . . . . . . . . . . . . . . . . . . . . . . . 304
9 Simple detonations: Reaction-advection 309

9.1 Reactive Euler equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
9.1.1 One-step irreversible kinetics . . . . . . . . . . . . . . . . . . . . . . . 309
9.1.2 Thermicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
9.1.3 Parameters for H2 -Air . . . . . . . . . . . . . . . . . . . . . . . . . . 311
9.1.4 Conservative form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
9.1.5 Non-conservative form . . . . . . . . . . . . . . . . . . . . . . . . . . 313
9.1.5.1 Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
9.1.5.2 Linear momenta . . . . . . . . . . . . . . . . . . . . . . . . 313
9.1.5.3 Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
9.1.5.4 Reaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
9.1.5.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
9.1.6 One-dimensional form . . . . . . . . . . . . . . . . . . . . . . . . . . 315
9.1.6.2 Non-conservative form . . . . . . . . . . . . . . . . . . . . . 316
9.1.6.3 Reduction of energy equation . . . . . . . . . . . . . . . . . 317
9.1.7 Characteristic form . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
9.1.8 Rankine-Hugoniot jump conditions . . . . . . . . . . . . . . . . . . . 322
9.1.9 Galilean transformation . . . . . . . . . . . . . . . . . . . . . . . . . 325
9.2 One-dimensional, steady solutions . . . . . . . . . . . . . . . . . . . . . . . . 326
9.2.1 Steady shock jumps . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
9.2.2 Ordinary differential equations of motion . . . . . . . . . . . . . . . . 327
8 CONTENTS

9.2.2.2 Unreduced non-conservative form . . . . . . . . . . . . . . . 328
9.2.2.3 Reduced non-conservative form . . . . . . . . . . . . . . . . 330
9.2.3 Rankine-Hugoniot analysis . . . . . . . . . . . . . . . . . . . . . . . . 331
9.2.3.1 Rayleigh line . . . . . . . . . . . . . . . . . . . . . . . . . . 331
9.2.3.2 Hugoniot curve . . . . . . . . . . . . . . . . . . . . . . . . . 332
9.2.4 Shock solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338
9.2.5 Equilibrium solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
9.2.5.1 Chapman-Jouguet solutions . . . . . . . . . . . . . . . . . . 340
9.2.5.2 Weak and strong solutions . . . . . . . . . . . . . . . . . . . 342
9.2.5.3 Summary of solution properties . . . . . . . . . . . . . . . . 343
9.2.6 ZND solutions: One-step irreversible kinetics . . . . . . . . . . . . . . 344
9.2.6.1 CJ ZND structures . . . . . . . . . . . . . . . . . . . . . . . 346
9.2.6.2 Strong ZND structures . . . . . . . . . . . . . . . . . . . . . 350
9.2.6.3 Weak ZND structures . . . . . . . . . . . . . . . . . . . . . 351
9.2.6.4 Piston problem . . . . . . . . . . . . . . . . . . . . . . . . . 354
9.2.7 Detonation structure: Two-step irreversible kinetics . . . . . . . . . . 355
9.2.7.1 Strong structures . . . . . . . . . . . . . . . . . . . . . . . . 362
9.2.7.1.1 D > D . . . . . . . . . . . . . . . . . . . . . . . . . 362
9.2.7.1.2 D = D . . . . . . . . . . . . . . . . . . . . . . . . . 362
9.2.7.2 Weak, eigenvalue structures . . . . . . . . . . . . . . . . . . 366
9.2.7.3 Piston problem . . . . . . . . . . . . . . . . . . . . . . . . . 366
9.2.8 Detonation structure: Detailed H2 O2 N2 kinetics . . . . . . . . . 368
10 Blast waves 375

10.1 Governing equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
10.2 Similarity transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
10.2.1 Independent variables . . . . . . . . . . . . . . . . . . . . . . . . . . 377
10.2.2 Dependent variables . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
10.2.3 Derivative transformations . . . . . . . . . . . . . . . . . . . . . . . . 378
10.3 Transformed equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
10.3.1 Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
10.3.2 Linear momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
10.3.3 Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
10.4 Dimensionless equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
10.4.1 Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
10.4.2 Linear momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
10.4.3 Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382
10.5 Reduction to non-autonomous form . . . . . . . . . . . . . . . . . . . . . . . 382
10.6 Numerical solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384
10.6.1 Calculation of total energy . . . . . . . . . . . . . . . . . . . . . . . . 386
10.6.2 Comparison with experimental data . . . . . . . . . . . . . . . . . . . 389
CONTENTS 9
10.7 Contrast with acoustic limit . . . . . . . . . . . . . . . . . . . . . . . . . . . 389
Bibliography 393
10 CONTENTS
Preface
These are lecture notes for AME 60636, Fundamentals of Combustion, a course taught since
1994 in the Department of Aerospace and Mechanical Engineering of the University of Notre
Dame. Most of the students in this course are graduate students; the course is also suitable
for interested undergraduates. The objective of the course is to provide background in
theoretical combustion science. Most of the material in the notes is covered in one semester;
some extra material is also included. Sections which need not be emphasized are marked
with a .
The goal of the notes is to provide a solid mathematical foundation in the physical
chemistry, thermodynamics, and fluid mechanics of combustion. The notes attempt to fill
gaps in the existing literature in providing enhanced discussion of detailed kinetics models
which are in common use for real physical systems. In addition, some model problems
for paradigm systems are addressed for pedagogical purposes. Many of the unique aspects
of these notes, which focus on some fundamental issues involving the thermodynamics of
reactive gases with detailed finite rate kinetics, have arisen due to the authors involvement
in a research project supported by the National Science Foundation, under Grant No. CBET-
0650843. The author is grateful for the support. Thanks are also extended to my former
student Dr. Ashraf al-Khateeb, who generated the steady laminar flame plots for detailed
hydrogen-air combustion. General thanks are also due to students in the course over the
years whose interest has motivated me to find ways to improve the material.
The notes, along with information on the course itself, can be found on the world wide
web at http://www.nd.edu/powers/ame.60636. At this stage, anyone is free to make
copies for their own use. I would be happy to hear from you about errors or suggestions for
improvement.
Joseph M. Powers
powers@nd.edu
http://www.nd.edu/powers
Notre Dame, Indiana; USA

July 23, 2010
Copyright c 2010 by Joseph M. Powers.

All rights reserved.
11
12 CONTENTS
Chapter 1
Introduction to kinetics
Let us consider the reaction of N molecular chemical species composed of L elements via J
chemical reactions. Let us assume the gas is an ideal mixture of ideal gases that satisfies
Daltons law of partial pressures. The reaction will be considered to be driven by molecular
collisions. We will not model individual collisions, but instead attempt to capture their
collective effect.
An example of a model of such a reaction is listed in Table 1.1. There we find a N = 9
species, J = 37 step irreversible reaction mechanism for an L = 3 hydrogen-oxygen-argon
mixture from Maas and Warnatz, 1 with corrected fH2 from Maas and Pope. 2 The model
has also been utilized by Fedkiw, et al. 3 We need not worry yet about fH2 , which is known
as a collision efficiency factor. The one-sided arrows indicate that each individual reaction is
considered to be irreversible. Note that for nearly each reaction, a separate reverse reaction is
listed; thus, pairs of irreversible reactions can in some sense be considered to model reversible
reactions.
In this model a set of elementary reactions are hypothesized. For the j th reaction we have
the collision frequency factor aj , the temperature-dependency exponent j and the activation
energy E j . These will be explained in short order.
Other common forms exist. Often reactions systems are described as being composed
of reversible reactions. Such reactions are usually notated by two sided arrows. One such
system is reported by Powers and Paolucci 4 reported here in Table 1.2.
Both overall models are complicated.
1
Maas, U., and Warnatz, J., 1988, Ignition Processes in Hydrogen-Oxygen Mixtures, Combustion and
Flame, 74(1): 53-69.
2
Maas, U., and Pope, S. B., 1992, Simplifying Chemical Kinetics: Intrinsic Low-Dimensional Manifolds
in Composition Space, Combustion and Flame, 88(3-4): 239-264.
3
Fedkiw, R. P., Merriman, B., and Osher, S., 1997, High Accuracy Numerical Methods for Thermally
Perfect Gas Flows with Chemistry, Journal of Computational Physics, 132(2): 175-190.
4
Powers, J. M., and Paolucci, S., 2005, Accurate Spatial Resolution Estimates for Reactive Supersonic
Flow with Detailed Chemistry, AIAA Journal, 43(5): 1088-1099.
13
14 CHAPTER 1. INTRODUCTION TO KINETICS
j Reaction aj j Ej
1 O2 + H OH + O 2.00 1014 0.00 70.30
2 OH + O O2 + H 1.46 1013 0.00 2.08
3 H2 + O OH + H 5.06 104 2.67 26.30
4 OH + H H2 + O 2.24 104 2.67 18.40
5 H2 + OH H2 O + H 1.00 108 1.60 13.80
6 H2 O + H H2 + OH 4.45 108 1.60 77.13
7 OH + OH H2 O + O 1.50 109 1.14 0.42
8 H2 O + O OH + OH 1.51 1010 1.14 71.64
9 H + H + M H2 + M 1.80 1018 1.00 0.00
10 H2 + M H + H + M 6.99 1018 1.00 436.08
11 H + OH + M H2 O + M 2.20 1022 2.00 0.00
12 H2 O + M H + OH + M 3.80 1023 2.00 499.41
13 O + O + M O2 + M 2.90 1017 1.00 0.00
14 O2 + M O + O + M 6.81 1018 1.00 496.41
15 H + O2 + M HO2 + M 2.30 1018 0.80 0.00
16 HO2 + M H + O2 + M 3.26 1018 0.80 195.88
17 HO2 + H OH + OH 1.50 1014 0.00 4.20
18 OH + OH HO2 + H 1.33 1013 0.00 168.30
19 HO2 + H H2 + O2 2.50 1013 0.00 2.90
20 H2 + O2 HO2 + H 6.84 1013 0.00 243.10
21 HO2 + H H2 O + O 3.00 1013 0.00 7.20
22 H2 O + O HO2 + H 2.67 1013 0.00 242.52
23 HO2 + O OH + O2 1.80 1013 0.00 1.70
24 OH + O2 HO2 + O 2.18 1013 0.00 230.61
25 HO2 + OH H2 O + O2 6.00 1013 0.00 0.00
26 H2 O + O2 HO2 + OH 7.31 1014 0.00 303.53
27 HO2 + HO2 H2 O2 + O2 2.50 1011 0.00 5.20
28 OH + OH + M H2 O2 + M 3.25 1022 2.00 0.00
29 H2 O2 + M OH + OH + M 2.10 1024 2.00 206.80
30 H2 O2 + H H2 + HO2 1.70 1012 0.00 15.70
31 H2 + HO2 H2 O2 + H 1.15 1012 0.00 80.88
32 H2 O2 + H H2 O + OH 1.00 1013 0.00 15.00
33 H2 O + OH H2 O2 + H 2.67 1012 0.00 307.51
34 H2 O2 + O OH + HO2 2.80 1013 0.00 26.80
35 OH + HO2 H2 O2 + O 8.40 1012 0.00 84.09
36 H2 O2 + OH H2 O + HO2 5.40 1012 0.00 4.20
37 H2 O + HO2 H2 O2 + OH 1.63 1013 0.00 132.71
Table 1.1: Units of aj are in appropriate combinations of cm, mol, s, and K so that i has
units of mole cm3 s1 ; units of E j are kJ mol1 . Third body collision efficiencies with M
are fH2 = 1.00, fO2 = 0.35, and fH2 O = 6.5.
15
j Reaction aj j Ej
1 H2 + O2 OH + OH 1.70 1013 0.00 47780
2 OH + H2 H2 O + H 1.17 109 1.30 3626
3 H + O2 OH + O 5.13 1016 0.82 16507
4 O + H2 OH + H 1.80 1010 1.00 8826
5 H + O2 + M HO2 + M 2.10 1018 1.00 0
6 H + O2 + O2 HO2 + O2 6.70 1019 1.42 0
7 H + O2 + N2 HO2 + N2 6.70 1019 1.42 0
8 OH + HO2 H2 O + O2 5.00 1013 0.00 1000
9 H + HO2 OH + OH 2.50 1014 0.00 1900
10 O + HO2 O2 + OH 4.80 1013 0.00 1000
11 OH + OH O + H2 O 6.00 108 1.30 0
12 H2 + M H + H + M 2.23 1012 0.50 92600
13 O2 + M O + O + M 1.85 1011 0.50 95560
14 H + OH + M H2 O + M 7.50 1023 2.60 0
15 H + HO2 H2 + O2 2.50 1013 0.00 700
16 HO2 + HO2 H2 O2 + O2 2.00 1012 0.00 0
17 H2 O2 + M OH + OH + M 1.30 1017 0.00 45500
18 H2 O2 + H HO2 + H2 1.60 1012 0.00 3800
19 H2 O2 + OH H2 O + HO2 1.00 1013 0.00 1800
Table 1.2: Nine species, nineteen step reversible reaction mechanism for a hydro-
gen/oxygen/nitrogen mixture. Units of aj are in appropriate combinations of cm, mole,
s, and K so that the reaction rate has units of mole/cm3 /s; units of E j are cal/mole.
Third body collision efficiencies with M are f5 (H2 O) = 21, f5 (H2 ) = 3.3, f12 (H2 O) = 6,
f12 (H) = 2, f12 (H2 ) = 3, f14 (H2 O) = 20.
1.1 Isothermal, isochoric kinetics

For simplicity, we will first focus attention on cases in which the temperature T and vol-
ume V are both constant. Such assumptions are known as isothermal and isochoric,
respectively. A nice fundamental treatment of elementary reactions of this type is given by
Vincenti and Kruger in their detailed monograph. 5
1.1.1 O O2 dissociation
One of the simplest physical examples is provided by the dissociation of O2 into its atomic
component O.
1.1.1.1 Pair of irreversible reactions

To get started, let us focus for now only on reactions 13 and 14 from Table 1.1 in the limiting
case in which temperature T and volume V are constant.
1.1.1.1.1 Mathematical model The reactions describe Oxygen dissociation and re-
combination in a pair of irreversible reactions:
13 : O + O + M O2 + M, (1.1)
14 : O2 + M O + O + M, (1.2)
with
2
17 mole K kJ
a13 = 2.90 10 3
, 13 = 1.00, E 13 = 0 (1.3)
cm s mole
1
18 mole K kJ
a14 = 6.81 10 3
, 14 = 1.00, E 14 = 496.41 (1.4)
cm s mole
The irreversibility is indicated by the one-sided arrow. Though they participate in the overall
hydrogen oxidation problem, these two reactions are in fact self-contained as well. So let us
just consider that we have only oxygen in our box with N = 2 species, O2 and O, J = 2
reactions (those being 13 and 14), and L = 1 element, that being O.
Recall that in the cgs system, common in thermochemistry, that 1 erg = 1 dyne cm =
107 J = 1010 kJ. Recall also that the cgs unit of force is the dyne and that 1 dyne =
1 g cm/s2 = 105 N. So for cgs we have
10
erg kJ 10 erg erg
E 13 = 0 , E 14 = 496.41 = 4.96 1012 . (1.5)
mole mole kJ mole
5
W. G. Vincenti and C. H. Kruger, 1965, Introduction to Physical Gas Dynamics, Wiley, New York.
1.1. ISOTHERMAL, ISOCHORIC KINETICS 17
The standard model for chemical reaction, which will be generalized and discussed in
more detail later, induces the following two ordinary differential equations for the evolution
of O and O2 molar concentrations:

dO 13 E 13 14 E 14
= 2 a13 T exp O O M +2 a14 T exp O2 M , (1.6)
dt RT RT
| {z } | {z }
=k13 (T ) =k14 (T )
| {z } | {z }
=r13 =r14

dO2 13 E 13 14 E 14
= a13 T exp O O M a14 T exp O2 M . (1.7)
dt RT RT
| {z } | {z }
=k13 (T ) =k14 (T )
| {z } | {z }
=r13 =r14
Here we use the notation i as the molar concentration of species i. Also a common usage for
molar concentration is given by square brackets, e.g. O2 = [O2 ]. The symbol M represents
an arbitrary third body and is an inert participant in the reaction. We also use the common
notation of a temperature-dependent portion of the reaction rate for reaction i, ki (T ), where

i Ei
ki (T ) = ai T exp (1.8)
RT
The reaction rates for reactions 13 and 14 are defined as
r13 = k13 O O M , (1.9)
r14 = k14 O2 M . (1.10)
We will give details of how to generalize this form later. The system Eq. (1.6-1.7) can be
written simply as
dO
= 2r13 + 2r14 , (1.11)
dt
dO2
= r13 r14 . (1.12)
dt
Even more simply, in vector form, Eqs. (1.11-1.12) can be written as
d
= r. (1.13)
dt
Here we have taken

O
= , (1.14)
O2

2 2
= , (1.15)
1 1

r13
r = . (1.16)
r14
In general, we will have be a column vector of dimension N 1, will be a rectangular

matrix of dimension N J of rank R, and r will be a column vector of length J 1. So
Eqs. (1.11-1.12) take the form

d O 2 2 r13
= (1.17)
dt O2 1 1 r14
Note here that the rank R of is R = L = 1. Let us also define a stoichiometric matrix of
dimension L N. The component of , li represents the number of element l in species i.
Generally will be full rank, which will vary since we can have L < N, L = N, or L > N.
Here we have L < N and is of dimension 1 2:
= (1 2). (1.18)
Element conservation is guaranteed by insisting that be constructed such that
= 0. (1.19)
So we can say that each of the column vectors of lies in the right null space of .
For our example, we see that Eq. (1.19) holds:

2 2
= (1 2) = (0 0). (1.20)
1 1
The symbol R is the universal gas constant, where

7
J 10 erg erg
R = 8.31441 = 8.31441 107 . (1.21)
mole K J mole K
Let us take as initial conditions
O (t = 0) = b
O , O2 (t = 0) = b
O2 . (1.22)
Now M represents an arbitrary third body, so here
M = O2 + O . (1.23)
Thus, the ordinary differential equations of the reaction dynamics reduce to

dO 13 E 13
= 2a13 T exp O O O2 + O
dt RT

14 E 14
+2a14 T exp O2 O2 + O , (1.24)
RT

dO2 13 E 13
= a13 T exp O O O2 + O
dt RT

14 E 14
a14 T exp O2 O2 + O . (1.25)
RT
Equations (1.24-1.25) with Eqs. (1.22) represent two non-linear ordinary differential equa-
tions with initial conditions in two unknowns O and O2 . We seek the behavior of these two
species concentrations as a function of time.
Systems of non-linear equations are generally difficult to integrate analytically and gen-
erally require numerical solution. Before embarking on a numerical solution, we simplify as
much as we can. Note that
dO d
+ 2 O2 = 0, (1.26)
dt dt
d
O + 2O2 = 0. (1.27)
dt
We can integrate and apply the initial conditions (1.22) to get
O + 2O2 = b
O + 2b
O2 = constant. (1.28)
The fact that this algebraic constraint exists for all time is a consequence of the conservation
of mass of each O element. It can also be thought of as the conservation of number of
O atoms. Such notions always hold for chemical reactions. They do not hold for nuclear
reactions.
Standard linear algebra provides a robust way to find the constraint of Eq. (1.28). We
can use elementary row operations to cast Eq. (1.16) into a row-echelon form. Here our goal
is to get a linear combination which on the right side has an upper triangular form. To
achieve this add twice the second equation with the first to form a new equation to replace
the second equation. This gives

d O 2 2 r13
= . (1.29)
dt O + 2O2 0 0 r14
Obviously the second equation is one we obtained earlier, d/dt(O + 2O2 ) = 0, and this
induces our algebraic constraint. We also note the system can be recast as

1 0 d O 2 2 r13
= (1.30)
1 2 dt O2 0 0 r14
This is of the matrix form
d
L1 P =Ur (1.31)
dt
Here L and L1 are N N lower triangular matrices of full rank N, and thus invertible.
The matrix U is upper triangular of dimension N J and with the same rank as , R L.
The matrix P is a permutation matrix of dimension N N. It is never singular and thus
always invertable. It is used to effect possible row exchanges to achieve the desired form;
often row exchanges are not necessary, in which case P = I, the N N identity matrix.
Equation (1.31) can be manipulated to form the original equation via
d 1
=P
| {z L U} r (1.32)
dt
=
What we have done is the standard linear algebra decomposition of = P1 L U.

We can also decompose the algebraic constraint, Eq. (1.28), in a non-obvious way that
is more readily useful for larger systems. We can write
1
O2 = bO2 O b
O . (1.33)
2
Defining now O = O b
O , we can say

O b
O 1
= b + ( O ) . (1.34)
O2 O2 12 | {z }
| {z } | {z } | {z }
= b =D =
=
This gives the dependent variables in terms of a smaller number of transformed dependent
variables in a way which satisfies the linear constraints. In vector form, the equation becomes
b + D .
= (1.35)
Here D is a full rank matrix which spans the same column space as does . Note that
may or may not be full rank. Since D spans the same column space as does , we must also
have in general
D = 0. (1.36)
We see here this is true:

1
(1 2) = 0. (1.37)
12
We also note that the term exp(E j /RT ) is a modulating factor to the dynamics. Let
us see how this behaves for high and low temperatures. First for low temperature, we have

E j
lim exp = 0. (1.38)
T 0 RT
At high temperature, we have

E j
lim exp = 1. (1.39)
T RT
And lastly, at intermediate temperature, we have

E j Ej
exp O(1) when T =O (1.40)
RT R
A sketch of this modulating factor is given in Figure 1.1.
Note
exp(- E j /(R T))

1
Ej/ R T
Figure 1.1: Plot of exp(E j /R/T ) versus T ; transition occurs at T E j /R.
for small T , the modulation is extreme, and the reaction rate is very small,
for T E j /R, the reaction rate is extremely sensitive to temperature,
for T , the modulation is unity, and the reaction rate is limited only by molecular
collision frequency.
Now O and O2 represent molar concentrations which have standard units of mole/cm3 .
So the reaction rates
dO dO2
and
dt dt
have units of mole/cm3 /s.
Note that after conversion of E j from kJ/mole to erg/mole we find the units of the
argument of the exponential to be unitless. That is

Ej erg mole K 1
= dimensionless. (1.41)
RT mole erg K
Here the brackets denote the units of a quantity, and not molar concentration. Let us get
units for the collision frequency factor of reaction 13, a13 . We know the units of the rate
(mole/cm3 /s). Reaction 13 involves three molar species. Since 13 = 1, it also has an
extra temperature dependency. The exponential of a unitless number is unitless, so we need
not worry about that. For units to match, we must have

mole mole mole mole
= [a13 ] K 1 . (1.42)
cm3 s cm3 cm3 cm3
So the units of a13 are

2
mole K
[a13 ] = 3
. (1.43)
cm s
For a14 we find a different set of units! Following the same procedure, we get

mole mole mole
= [a14 ] K 1 . (1.44)
cm3 s cm3 cm3
So the units of a14 are
1
mole K
[a14 ] = 3
. (1.45)
cm s
This discrepancy in the units of aj the molecular collision frequency factor is a burden of tra-
ditional chemical kinetics, and causes many difficulties when classical non-dimensionalization
is performed. With much effort, a cleaner theory could be formulated; however, this would
require significant work to re-cast the now-standard aj values for literally thousands of re-
actions which are well established in the literature.
1.1.1.1.2 Example calculation Let us consider an example problem. Let us take T =

5000 K, and initial conditions b O = 0.001 mole/cm3 and b O2 = 0.001 mole/cm3 . The
initial temperature is very hot, and is near the temperature of the surface of the sun. This
is also realizable in laboratory conditions, but uncommon in most combustion engineering
environments.
We can solve these in a variety of ways. I chose here to solve both Eqs. (1.24-1.25) without
the reduction provided by Eq. (1.28). However, we can check after numerical solution to see
if Eq. (1.28) is actually satisfied. Substituting numerical values for all the constants to get
2 !
E 13 mole K
2a13 T 13 exp = 2 2.9 107 3
(5000 K)1 exp(0)
RT cm s
2
mole 1
= 1.16 1014 , (1.46)
cm3 s
1 !
E 14 mole K
2a14 T 14 exp = 2 6.81 1018 3
(5000 K)1
RT cm s
erg
4.96 1012 mole
exp erg
8.31441 107 mole K
(5000 K)
1
mole 1
= 1.77548 1010 3
, (1.47)
cm s
2
13 E 13 13 mole 1
a13 T exp = 5.80 10 , (1.48)
RT cm3 s
1
14 E 14 9 mole 1
a14 T exp = 8.8774 10 3
. (1.49)
RT cm s
@OD,@O2D HmoleccL
0.00150
O2
0.00100
0.00070
0.00050
t HsL
10-11 10-10 10-9 10-8 10-7 10-6
Figure 1.2: Molar concentrations versus time for oxygen dissociation problem.
Then the differential equation system becomes

dO
= (1.16 1014 )2O (O + O2 ) + (1.77548 1010 )O2 (O + O2 ), (1.50)
dt
dO2
= (5.80 1013 )2O (O + O2 ) (8.8774 109 )O2 (O + O2 ), (1.51)
dt
mole
O (0) = 0.001 , (1.52)
cm3
mole
O2 (0) = 0.001 . (1.53)
cm3
These non-linear ordinary differential equations are in a standard form for a wide variety
of numerical software tools. Solution of such equations are not the topic of these notes.
1.1.1.1.2.1 Species concentration versus time A solution was obtained numeri-

cally, and a plot of O (t) and O2 (t) is given in Figure 1.2. Note that significant reaction does
not commence until t 1010 s. This can be shown to be very close to the time between
molecular collisions. For 109 s < t < 108 s, there is a vigorous reaction. For t > 107 s,
the reaction appears to be equilibrated. The calculation gives the equilibrium values eO and
eO2 , as
mole
lim O = eO = 0.0004424 , (1.54)
t cm3
mole
lim O2 = eO2 = 0.00127 , (1.55)
t cm3
Note that at this high temperature, O2 is preferred over O, but there are definitely O
molecules present at equilibrium.
1.5 10-16
1.0 10-16
7.0 10-17
5.0 10-17
3.0 10-17
t HsL
10-10 10-9 10-8 10-7 10-6
Figure 1.3: Dimensionless residual numerical error r in satisfying the element conservation
constraint in the oxygen dissociation example.
We can check how well the numerical solution satisfied the algebraic constraint of element
conservation by plotting the dimensionless residual error r

+ 2 b 2b
O2
r= O O2 O

b
O + 2b
O2
as a function of time. If the constraint is exactly satisfied, we will have r = 0. Any non-zero
r will be related to the numerical method we have chosen. It may contain roundoff error
and have a sporadic nature. A plot of r(t) is given in Figure 1.3. Clearly the error is small,
and has the character of a roundoff error. In fact it is possible to drive r to be smaller by
controlling the error tolerance in the numerical method.
1.1.1.1.2.2 Pressure versus time We can use the ideal gas law to calculate the
pressure. Recall that the ideal gas law for molecular species i is
Pi V = ni RT. (1.56)
Here Pi is the partial pressure of molecular species i, and ni is the number of moles of
molecular species i. Note that we also have
ni
Pi = RT. (1.57)
V
Note that by our definition of molecular species concentration that
ni
i = . (1.58)
V
So we also have the ideal gas law as
Pi = i RT. (1.59)
Now in the Dalton mixture model, all species share the same T and V . So the mixture
temperature and volume are the same for each species Vi = V , Ti = T . But the mixture
pressure is taken to be the sum of the partial pressures:
N
X
P = Pi . (1.60)
i=1
Substituting from Eq. (1.59) into Eq. (1.60), we get

N
X N
X
P = i RT = RT i . (1.61)
i=1 i=1
For our example, we only have two species, so
P = RT (O + O2 ). (1.62)
The pressure at the initial state t = 0 is
P (t = 0) = RT (b
O + b
O2 ), (1.63)

erg mole mole
= 8.31441 107 (5000 K) 0.001 3
+ 0.001 , (1.64)
mole K cm cm3
dyne
= 8.31441 108 , (1.65)
cm2
= 8.31441 102 bar. (1.66)
This pressure is over 800 atmospheres. It is actually a little too high for good experimental
correlation with the underlying data, but we will neglect that for this exercise.
At the equilibrium state we have more O2 and less O. And we have a different number
of molecules, so we expect the pressure to be different. At equilibrium, the pressure is
P (t ) = lim RT (O + O2 ), (1.67)
t

7 erg mole mole
= 8.31441 10 (5000 K) 0.0004424 + 0.00127 ,
mole K cm3 cm3
(1.68)
dyne
= 7.15 108 , (1.69)
cm2
= 7.15 102 bar. (1.70)
The pressure has dropped because much of the O has recombined to form O2 . Thus there
are fewer molecules at equilibrium. The temperature and volume have remained the same.
A plot of P (t) is given in Figure 1.4.
P HdynecmcmL
8.4 108
8.2 108
8. 108
7.8 108
7.6 108
7.4 108
7.2 108
t HsL
10-11 10-10 10-9 10-8 10-7 10-6
Figure 1.4: Pressure versus time for oxygen dissociation example.
1.1.1.1.2.3 Dynamical system form Now Eqs. (1.50-1.51) are of the standard form
for an autonomous dynamical system:
dy
= f(y). (1.71)
dt
Here y is the vector of state variables (O , O2 )T . And f is an algebraic function of the state
variables. For the isothermal system, the algebraic function is in fact a polynomial.
Equilibrium
The dynamical system is in equilibrium when
f(y) = 0. (1.72)
This non-linear set of algebraic equations can be difficult to solve for large systems. We will
later see that for common chemical kinetics systems, such as the one we are dealing with,
there is a guarantee of a unique equilibrium for which all state variables are physical. There
are certainly other equilibria for which at least one of the state variables is non-physical.
Such equilibria can be quite mathematically complicated.
Solving Eq. (1.72) for our oxygen dissociation problem gives us symbolically from Eq. (1.6-
1.7)

E 13 e e e 13 E 14
2a13 exp O O M T + 2a14 exp eO2 eM T 14 = 0, (1.73)
RT RT

13 E 13 e e e 14 E 14
a13 T exp O O M a14 T exp eO2 eM = 0. (1.74)
RT RT
We notice that eM cancels. This so-called third body will in fact never affect the equilib-
rium state. It will however influence the dynamics. Removing eM and slightly rearranging
Eqs. (1.73-1.74) gives

13E 13 e e 14 E 14
a13 T exp O O = a14 T exp eO2 , (1.75)
RT RT

13 E 13 e e 14 E 14
a13 T exp O O = a14 T exp eO2 . (1.76)
RT RT
These are the same equations! So we really have two unknowns for the equilibrium state eO
and eO2 but seemingly only one equation. Note that rearranging either Eq. (1.75) or (1.76)
gives the result

14 E 14
eO eO a14 T exp RT
= = K(T ). (1.77)
eO2 a T 13 exp E 13 13 RT
That is, for the net reaction (excluding the inert third body), O2 O+O, at equilibrium the
product of the concentrations of the products divided by the product of the concentrations
of the reactants is a function of temperature T . And for constant T , this is the so-called
equilibrium constant. This is a famous result from basic chemistry. It is actually not complete
yet, as we have not taken advantage of a connection with thermodynamics. But for now, it
will suffice.
We still have a problem: Eq. (1.77) is still one equation for two unknowns. We solve
this be recalling we have not yet taken advantage of our algebraic constraint of element
conservation, Eq. (1.28). Let us use the equation to eliminate eO2 in favor of eO :
1 b
eO2 = O + b
e
O2 (1.78)
2 O
So Eq (1.75) reduces to

E 13 E 14 1 b
a13 T 13
exp eO eO = a14 T 14
exp O + b
e
O2 , (1.79)
RT RT 2 O
| {z }
=eO
2
Eq. (1.79) is one algebraic equation in one unknown. Its solution gives the equilibrium value
eO . It is a quadratic equation for eO . Of its two roots, one will be physical. We note that the
equilibrium state will be a function of the initial conditions. Mathematically this is because
our system is really best posed as a system of differential-algebraic equations. Systems
which are purely differential equations will have equilibria which are independent of their
initial conditions. Most of the literature of mathematical physics focuses on such systems
of those. One of the foundational complications of chemical dynamics is the the equilibria
fH@ODL HmoleccsL
300 000
200 000
100 000
@OD HmoleccL
-0.004 -0.003 -0.002 -0.001 0.001
-100 000
-200 000
Figure 1.5: Equilibria for oxygen dissociation example.
is a function of the initial conditions, and this renders many common mathematical notions
from traditional dynamic system theory to be invalid Fortunately, after one accounts for the
linear constraints of element conservation, one can return to classical notions from traditional
dynamic systems theory.
Consider the dynamics of Eq. (1.24) for the evolution of O . Equilibrating the right hand
side of this equation, gives Eq. (1.73). Eliminating M and then O2 in Eq. (1.73) then
substituting in numerical parameters gives the cubic algebraic equation
33948.3 (1.78439 1011 )(O )2 (5.8 1013 )(O )3 = f (O ) = 0. (1.80)
This equation is cubic because we did not remove the effect of M . This will not affect
the equilibrium, but will affect the dynamics. We can get an idea of where the roots are
by plotting f (O ) as seen in Figure 1.5. Zero crossings of f (O ) in Figure 1.5 represent
equilibria of the system, eO , f (eO ) = 0. The cubic equation has three roots
mole
eO = 0.003 , non-physical (1.81)
cm3
mole
eO = 0.000518944 , non-physical (1.82)
cm3
mole
eO = 0.000442414 , physical. (1.83)
cm3
Note the physical root found by our algebraic analysis is identical to that which was identified
by our numerical integration of the ordinary differential equations of reaction kinetics.
Stability of equilibria
We can get a simple estimate of the stability of the equilibria by considering the slope of
f near f = 0. Our dynamic system is of the form
dO
= f (O ). (1.84)
dt
Near the first non-physical root at eO = 0.003, a positive perturbation from equi-
librium induces f < 0, which induces dO /dt < 0, so O returns to its equilibrium.
Similarly, a negative perturbation from equilibrium induces dO /dt > 0, so the system
returns to equilibrium. This non-physical equilibrium point is stable. Note stability
does not imply physicality!
Perform the same exercise for the non-physical root at eO = 0.000518944. We find
this root is unstable.
Perform the same exercise for the physical root at eO = 0.000442414. We find this
root is stable.
In general if f crosses zero with a positive slope, the equilibrium is unstable. Otherwise, it
is stable.
Consider a formal Taylor series expansion of Eq. (1.84) in the neighborhood of an equi-
librium point 3O :

d e e df
( O ) = f (O ) + ( eO ) + . . . (1.85)
dt O | {z } dO =e O
O O
=0
We find df /dO by differentiating Eq. (1.80) to get

df
= (3.56877 1011 )O (1.74 1014 )2O . (1.86)
dO
We evaluate df /dO near the physical equilibrium point at O = 0.004442414 to get
df
= (3.56877 1011 )(0.004442414) (1.74 1014 )(0.004442414)2,
dO
1
= 1.91945 108 . (1.87)
s
Thus the Taylor series expansion of Eq. (1.24) in the neighborhood of the physical equi-
librium gives the local kinetics to be driven by
d
(O 0.00442414) = (1.91945 108 ) (O 0.004442414) + . . . . (1.88)
dt
So in the neighborhood of the physical equilibrium we have

O = 0.0004442414 + A exp 1.91945 108 t (1.89)
Here A is an arbitrary constant of integration. The local time constant which governs the
times scales of local evolution is where
1
= = 5.20983 109 s. (1.90)
1.91945 108
This nano-second time scale is very fast. It can be shown to be correlated with the mean
time between collisions of molecules.
1.1.1.1.3 Effect of temperature Let us perform four case studies to see the effect of
T on the systems equilibria and it dynamics near equilibrium.
T = 3000 K. Here we have significantly reduced the temperature, but it is still higher
than typically found in ordinary combustion engineering environments. Here we find
mole
eO = 8.9371 106 , (1.91)
cm3
= 1.92059 107 s. (1.92)
The equilibrium concentration of O dropped by two orders of magnitude relative to

T = 5000 K, and the time scale of the dynamics near equilibrium slowed by two orders
of magnitude.
T = 1000 K. Here we reduce the temperature more. This temperature is common in

combustion engineering environments. We find
mole
eO = 2.0356 1014 , (1.93)
cm3
= 2.82331 101 s. (1.94)
The O concentration at equilibrium is greatly diminished to the point of being difficult

to detect by standard measurement techniques. And the time scale of combustion has
significantly slowed.
T = 300 K. This is obviously near room temperature. We find

mole
eO = 1.14199 1044 , (1.95)
cm3
= 1.50977 1031 s. (1.96)
The O concentration is effectively zero at room temperature, and the relaxation time
is effectively infinite. As the oldest star in our galaxy has an age of 4.4 1017 s, we see
that at this temperature, our mathematical model cannot be experimentally validated,
so it loses its meaning. At such a low temperature, the theory becomes qualitatively
correct, but not quantitatively predictive.
T = 10000 K. Such high temperature could be achieved in an atmospheric re-entry

environment.
mole
eO = 2.74807 103 , (1.97)
cm3
= 1.69119 1010 s. (1.98)
At this high temperature, O become preferred over O2 , and the time scales of reaction
become extremely small, under a nanosecond.
1.1.1.2 Single reversible reaction

The two irreversible reactions studied in the previous section are of a class that is common
in combustion modeling. However, the model suffers a defect in that its link to classical
equilibrium thermodynamics is missing. A better way to model essentially the same physics
and guarantee consistency with classical equilibrium thermodynamics is to model the process
as a single reversible reaction, with a suitably modified reaction rate term.
1.1.1.2.1 Mathematical model
1.1.1.2.1.1 Kinetics For the reversible O O2 reaction, let us only consider reaction
13 from Table 1.2 for which
13 : O2 + M O + O + M. (1.99)
For this system, we have N = 2 molecular species in L = 1 elements reacting in J = 1

reaction. Here
1
mole cal
a13 = 1.85 3
(K)0.5 , 13 = 0.5, E 13 = 95560 (1.100)
cm mole
Units of cal are common in chemistry, but we need to convert to erg, which is achieved via
7
cal 4.186 J 10 erg erg
E 13 = 95560 = 4.00014 1012 . (1.101)
mole cal J mole
For this reversible reaction, we slightly modify the kinetics equations to

dO 13 E 13 1
= 2 a13 T exp O2 M , (1.102)
dt RT Kc,13 O O M
| {z }
=k13 (T )
| {z }
=r13

dO2 13 E 13 1
= a13 T exp O2 M . (1.103)
dt RT Kc,13 O O M
| {z }
=k13 (T )
| {z }
=r13
Here we have used equivalent definitions for k13 (T ) and r13 , so that Eqs. (1.102-1.103) can
be written compactly as
dO
= 2r13 , (1.104)
dt
dO2
= r13 . (1.105)
dt
In matrix form, we can simplify to

d O 2
= (r13 ). (1.106)
dt O2 1
| {z }
=
Here the N J or 2 1 matrix is

2
= . (1.107)
1
Performing row operations, the matrix form reduces to

d O 2
= (r13 ). (1.108)
dt O + 2O2 0
or

1 0 d O 2
= (r13 ). (1.109)
1 2 dt O2 0
So here the N N or 2 2 matrix L1 is

1 1 0
L = . (1.110)
1 2
The N N or 2 2 permutation matrix P is the identity matrix. And the N J or 2 1
upper triangular matrix U is
2
U= . (1.111)
0
Note that = L U or equivalently L1 = U:

1 0 2 2
= . (1.112)
1 2 1 0
| {z } | {z } | {z }
=L1 = =U
Once again the stoichiometric matrix is

= (1 2). (1.113)
And we see that = 0 is satisfied:

2
(1 2) = (0). (1.114)
1
As for the irreversible reactions, the reversible reaction rates are constructed to conserve
O atoms. We have
d
O + 2O2 = 0. (1.115)
dt
Thus, we once again find
O + 2b
O + 2O2 = b O2 = constant. (1.116)
As before, we can say

O b
O 1
= b + ( O ) . (1.117)
O2 O2 12 | {z }
| {z } | {z } | {z }
= b =D =
=
This gives the dependent variables in terms of a smaller number of transformed dependent
variables in a way which satisfies the linear constraints. In vector form, the equation becomes
b + D .
= (1.118)
Once again D = 0.
1.1.1.2.1.2 Thermodynamics Equations (1.102-1.103) are supplemented by an ex-

pression for the thermodynamics-based equilibrium constant Kc,13 which is:

Po Go13
Kc,13 = exp . (1.119)
RT RT
Here Po = 1.01326 106 dyne/cm2 = 1 atm is the reference pressure. The net change of
Gibbs free energy at the reference pressure for reaction 13, Go13 is defined as
Go13 = 2g oO g oO2 . (1.120)
We further recall that the Gibbs free energy for species i at the reference pressure is defined
in terms of the enthalpy and entropy as
o
g oi = hi T soi . (1.121)
o
It is common to find hi and soi in thermodynamic tables tabulated as functions of T .
We further note that both Eqs. (1.102) and (1.103) are in equilibrium when
1 e e e
eO2 eM = . (1.122)
Kc,13 O O M
We rearrange Eq. (1.122) to find the familiar

Q
eO eO [products]
Kc,13 = e =Q . (1.123)
O2 [reactants]
If Kc,13 > 1, the products are preferred. If Kc,13 < 1, the reactants are preferred.
Now, Kc,13 is a function of T only, so it is known. But Eq. (1.123) once again is one
equation in two unknowns. We can use the element conservation constraint, Eq. (1.116) to
reduce to one equation and one unknown, valid at equilibrium:
eO eO
Kc,13 = (1.124)
O2 + 12 (b
b O eO )
Using the element constraint, Eq. (1.116), we can recast the dynamics of our system by
modifying Eq. (1.102) into one equation in one unknown:

dO E 13
= 2a13 T 13 exp
dt RT

1 b 1 1 1
b
(O2 + O2 + (b
(O O )) (b O O ) + O ) O O ) + O )
O2 + (b
O O (b ,
| 2 {z }| 2 {z } Kc,13 | 2 {z }
=O2 =M =M
(1.125)
1.1.1.2.2 Example calculation Let us consider the same example as the previous sec-
tion with T = 5000 K. We need numbers for all of the parameters of Eq. (1.125). For O,
we find at T = 5000 K that
o erg
hO = 3.48382 1012 , (1.126)
mole
erg
soO = 2.20458 109 . (1.127)
mole K
So
erg erg
g oO = 3.48382 10 12
(5000 K) 2.20458 10 9
mole mole K
12 erg
= 7.53908 10 . (1.128)
mole
For O2 , we find at T = 5000 K that
o erg
hO2 = 1.80749 1012 , (1.129)
mole
erg
soO2 = 3.05406 109 . (1.130)
mole K
So
erg erg
g oO2 = 1.80749 10 12
(5000 K) 3.05406 10 9
mole mole K
13 erg
= 1.34628 10 . (1.131)
mole
@OD,@O2D HmoleccL
0.00150
O2
0.00100
0.00070
0.00050
O
0.00030
t HsL
10-11 10-9 10-7 10-5 0.001
Figure 1.6: Plot of O (t) and O2 (t) for oxygen dissociation with reversible reaction.
Thus, by Eq. (1.120), we have

erg
Go13 = 2(7.53908 1012 ) (1.34628 1013 ) = 1.61536 1012 . (1.132)
mole
Thus, by Eq. (1.119) we get for our system
1.01326 106 dyne

2
Kc,13 = erg
cm
7
8.31441 10 mole K (5000 K)
!!
erg
1.61536 1012 mole
exp erg
, (1.133)
8.31441 107 mole K
(5000 K)
mole
= 1.187 104 . (1.134)
cm3
Substitution of all numerical parameters into Eq. (1.125) and expansion yields the fol-
lowing
dO
= 3899.47 (2.23342 1010 )2O (7.3003 1012 )3O = f (O ), O (0) = 0.001.
(1.135)
dt
A plot of the time-dependent behavior of O and O2 from solution of Eq. (1.135) is given
in Figure 1.6. The behavior is similar to the predictions given by the pair of irreversible
reactions in Fig. 1.1. Here direct calculation of the equilibrium from time integration reveals
mole
eO = 0.000393328 . (1.136)
cm3
fH@ODL HmoleccsL
40 000
30 000
20 000
10 000
@OD HmoleccL
-0.004 -0.003 -0.002 -0.001 0.001
-10 000
-20 000
Figure 1.7: Plot of f (O ) versus O for oxygen dissociation with reversible reaction.
Using Eq. (1.116) we find this corresponds to

mole
eO2 = 0.00130334 . (1.137)
cm3
We note the system begins significant reaction for t 109 s and is equilibrated for t
107 s.
The equilibrium is verified by solving the algebraic equation
f (O )3899.47 (2.23342 1010 )2O (7.3003 1012 )3O = 0. (1.138)
This yields three roots:
mole
eO = 0.003 , non-physical, (1.139)
cm3
mole
eO = 0.000452678 , non-physical, (1.140)
cm3
mole
eO = 0.000393328 , physical, (1.141)
cm3
(1.142)
is given in Figure 1.6.
Linearizing Eq. (1.135) in the neighborhood of the physical equilibrium yields the equa-
tion
d
( 0.000393328) = (2.09575 107 ) (O 0.000393328) + . . . (1.143)
dt O
This has solution

O = 0.000393328 + A exp 2.09575 107 t . (1.144)
Again, A is an arbitrary constant. Obviously the equilibrium is stable. Moreover the time
constant of relaxation to equilibrium is
1
= = 4.77156 108 s. (1.145)
2.09575 107
This is consistent with the time scale to equilibrium which comes from integrating the full
equation.
1.1.2 Zeldovich mechanism of N O production

Let us consider next a more complicated reaction system: that of NO production known
as the Zeldovich 6 mechanism. This is an important model for the production of a major
pollutant from combustion processes. It is most important for high temperature applications.
1.1.2.1 Mathematical model

The model has several versions. One is
1: N + NO N2 + O, (1.146)
2: N + O2 NO + O. (1.147)
similar to our results for O2 dissociation, N2 and O2 are preferred at low temperature. As
the temperature rises N and O begin to appear, and it is possible when they are mixed for
NO to appear as a product.
1.1.2.1.1 Standard model form Here we have the reaction of N = 5 molecular species
with

N O
N

=
2 .
N
(1.148)
O
O2
We have L = 2 with N and O as the 2 elements. The stoichiometric matrix of dimension

L N = 2 5 is

1 1 2 0 0
= . (1.149)
1 0 0 1 2
The first row of is for the N atom; the second row is for the O atom.
6
Yakov Borisovich Zeldovich, 1915-1987, prolific Soviet physicist and father of thermonuclear weapons.
And we have J = 2 reactions. The reaction vector of length J = 2 is

a T 1
exp Ta,1
1

r1 1
T
N NO Kc,1 N2 O
r = = , (1.150)
r2 2
a2 T exp T
Ta,2 1
N O2 Kc,2 N O O

k1 N N O K1c,1 N2 O
= (1.151)
k2 N O2 K1c,2 N O O
Here, we have

1 Ta,1
k1 = a1 T exp , (1.152)
T

2 Ta,2
k2 = a2 T exp (1.153)
T
In matrix form, the model can be written as

N O 1 1
N 1 1
d
N = 1
r1

0 (1.154)
dt 2
O 1
r2
1
O2 0 1
| {z }
=
Here the matrix has dimension N J which is 5 2. The model is of our general form
d
= r. (1.155)
dt
Note that our stoichiometric constraint on element conservation for each reaction = 0
holds here:

1 1
1 1
1 1 2 0 0 0 0
=
1
0 = . (1.156)
1 0 0 1 2 1 0 0
1
0 1
We get 4 zeros because there are 2 reactions each with 2 element constraints.
1.1.2.1.2 Reduced form Here we describe non-traditional, but useful reductions, using
standard techniques from linear algebra to bring the model equations into a reduced form in
which all of the linear constraints have been explicitly removed.
Let us perform a series of row operations to find all of the linear dependencies. Our aim
is to convert the matrix into an upper triangular form. The lower left corner of already
has a zero, so there is no need to worry about it. Let us add the first and fourth equations
to eliminate the 1 in the 4, 1 slot. This gives

N O 1 1
N 1 1
d
N

= 1
r1

0 (1.157)
dt 2
N O + O 0
r2
2
O2 0 1
Next, add the first and third equations to get

N O 1 1
N 1 1
d
N O + N = 0
r1
2 1
r2 (1.158)
dt
N O + O 0 2
O2 0 1
Now multiply the first equation by 1 and add it to the second to get

N O 1 1
N O + N 0 2
d
N O + N = 0
r1

1 (1.159)
dt 2
N O + O 0
r2
2
O2 0 1
Next multiply the fifth equation by 2 and add it to the second to get

N O 1 1
N O + N 0 2
d

r1

+ = 0 1 (1.160)
dt

NO N2
0
r2
N O + O 2
N O + N 2O2 0 0
Next add the second and fourth equations to get

N O 1 1
N O + N 0 2
d

= 0
r1
N O + N2 1
r2 (1.161)
dt 0
N + O 0
N O + N 2O2 0 0
Next multily the third equation by 2 and add it to the second to get

N O 1 1
N O + N 0 2
d
N O + N + 2N = 0
r1

0 (1.162)
dt

2
0
r2
N + O 0
N O + N 2O2 0 0
Rewritten, this becomes

1 0 0 0 0 N O 1 1
1 1 0 0
0 d N 0 2 r1
1 1 2 0 0 0
dt N2 = 0 r2 (1.163)
0 1 0 1 0 O 0 0
1 1 0 0 2 O2 0 0
| {z } | {z }
=L1 =U
A way to think of this type of row echelon form is that it defines two free variables, those
associated with the non-zero pivots of U: N O and N . The remain three variables N2 , O
and O2 are bound variables which can be expressed in terms of the free variables.
The last three of the ordinary differential equations are homogeneous and can be inte-
grated to form
N O + N + 2N2 = C1 , (1.164)
N + O = C2 , (1.165)
N O + N 2O2 = C3 . (1.166)
The constants C1 , C2 and C3 are determined from the initial conditions on all five state
variables. In matrix form, we can say

N O
1 1 2 0 0 N C1

0 1 0 1 0 N = C2 (1.167)
2
1 1 0 0 2 O C3
O2
Considering the free variables, N O and N to be known, we move them to the right side
to get

2 0 0 N2 C1 N O N
0 1 0 O = C2 N (1.168)
0 0 2 O2 C3 + N O N
Solving, for the bound variables, we find
1
N2 C 12 N O 21 N
2 1
O = C2 N . (1.169)
1 1 1
O2 2 C3 2 N O + 2 N
We can rewrite this as
1 1
N2 2
C 1 2 21
O = C2 + 0
1 N O
. (1.170)
N
O2 12 C3 12 1
2
We can get a more elegant form by defining N O = N O and N = N . Thus we can say our
state variables have the form

N O 0 1 0
N 0 0 1

N = 1 C1 + 1 1 N O . (1.171)
2 2 2 2
O C2 0 1 N
O2 21 C3 12 1
2
By translating via N O = N O + b
N O and N = N + bN and choosing the constants C1 , C2
and C3 appropriately, we can arrive at
b
N O N O 1 0
N b 1
N 01 N O
b
N = N2 + 1
2 2 2 N . (1.172)
O b O2 0 1 | {z }
O2 b
O2 21 1
2 =
| {z } | {z } | {z }
= =b =D
This takes the form of

b + D .
= (1.173)
Here the matrix D is of dimension N R, which here is 5 2. It spans the same column
space as does the N J matrix which is of rank R. Here in fact R = J = 2, so D has the
same dimension as . In general it will not. If c1 and c2 are the column vectors of D, we
see that c1 c2 forms the first column vector of and c1 c2 forms the second column
vector of . Note that D = 0:

1 0
0 1
1 1 2 0 0 1 0 0
D= 1
2 2 = 0 0 . (1.174)
1 0 0 1 2 0 1
12 1
2
Equations (1.164-1.166) can also be linearly combined in a way which has strong phys-
ical relevance. We rewrite the system as three equations in which the first is identical to
Eq. (1.164); the second is the difference of Eqs. (1.165) and (1.166); and the third is half of
Eq. (1.164) minus half of Eq. (1.166) plus Eq. (1.165):
N O + N + 2N2 = C1 , (1.175)
O + N O + 2O2 = C2 C3 , (1.176)
1
N O + N + N2 + O + O2 = (C1 C3 ) + C2 . (1.177)
2
Equation (1.175) insists that the number of nitrogen elements be constant; Eq. (1.176)
demands the number of oxygen elements be constant; and Eq. (1.177) requires the number
of moles of molecular species be constant. For general reactions, including the earlier studied
oxygen dissociation problem, the number of moles of molecular species will not be constant.
Here because each reaction considered has two molecules reacting to form two molecules,
we are guaranteed the number of moles will be constant. Hence, we get an additional
linear constraint beyond the two for element conservation. Note that since our reaction is
isothermal, isochoric and mole-preserving, it will also be isobaric.
1.1.2.1.3 Example calculation Let us consider an isothermal reaction at

T = 6000 K. (1.178)
The high temperature is useful in generating results which are easily visualized. It insures
that there will be significant concentrations of all molecular species. Let us also take as an
initial condition
b mole
N O = bN = bN2 = bO = bO2 = 1 106 . (1.179)
cm3
For this temperature and concentrations, the pressure, which will remain constant through
the reaction, is P = 2.4942 106 dyne/cm2 . This is a little greater than atmospheric.
Kinetic data for this reaction is adopted from Baulch, et al. 7 The data for reaction 1 is
1
13 mole 1
a1 = 2.107 10 3
, 1 = 0, Ta1 = 0 K. (1.180)
cm s
For reaction 2, we have
1
9 mole 1
a2 = 5.8394 10 , 2 = 1.01, Ta2 = 3120 K. (1.181)
cm3 K 1.01 s
Here the so-called activation temperature Ta,j for reaction j is really the activation energy
scaled by the universal gas constant:
Ej
.Ta,j = (1.182)
R
Substituting numbers we obtain for the reaction rates
1
13 0 0 13 mole 1
k1 = (2.107 10 )(6000) exp = 2.107 10 , (1.183)
6000 cm3 s
1
9 1.01 3120 13 mole 1
k2 = (5.8394 10 )(6000) exp = 2.27231 10 3
.
6000 cm s
(1.184)
7
Baulch, et al., 2005, Evaluated Kinetic Data for Combustion Modeling: Supplement II, Journal of
Physical and Chemical Reference Data, 34(3): 757-1397.
We will also need thermodynamic data. The data here will be taken from the Chemkin
database. 8 Thermodynamic data for common materials is also found in most thermody-
namic texts. For our system at 6000 K, we find
erg
g oN O = 1.58757 1013 , (1.185)
mole
erg
g oN = 7.04286 1012 , (1.186)
mole
erg
g oN2 = 1.55206 1013 , (1.187)
mole
erg
g oO = 9.77148 1012 , (1.188)
mole
erg
g oO2 = 1.65653 1013 . (1.189)
mole
Thus for each reaction, we find Goj :
Go1 = g oN2 + g oO g oN g oN O , (1.190)
13 12 12 13
= 1.55206 10 9.77148 10 + 7.04286 10 + 1.58757 10 ,(1.191)
erg
= 2.37351 1012 , (1.192)
mole
Go2 = g oN O + g oO g oN g oO2 , (1.193)
13 12 12 13
= 1.58757 10 9.77148 10 + 7.04286 10 + 1.65653 10 ,(1.194)
erg
= 2.03897 1012 . (1.195)
mole
At 6000 K, we find the equilibrium constants for the J = 2 reactions are

Go1
Kc,1 = exp , (1.196)
RT

2.37351 1012
= exp (1.197)
(8.314 107 )(6000)
= 116.52, (1.198)

Go2
Kc,2 = exp , (1.199)
RT

2.03897 1012
= exp , (1.200)
(8.314 107 )(6000)
= 59.5861. (1.201)
Again, omitting details, we find the two differential equations governing the evolution of
the free variables are
dN O
= 0.723 + 2.22 107 N + 1.15 1013 2N 9.44 105 N O 3.20 1013 N N O ,
dt
8
R. J. Kee, et al., 2000, The Chemkin Thermodynamic Data Base, part of the Chemkin Collection
Release 3.6, Reaction Design, San Diego, CA.
@ND,@NOD HmoleccL
1 10-5
5 10-6
1 10-6
5 10-7 @NOD
1 10-7
5 10-8
@ND
t HsL
10-10 10-9 10-8 10-7 10-6 10-5
Figure 1.8: NO and N concentrations versus time for T = 6000 K, P = 2.4942

106 dyne/cm2 Zeldovich mechanism.
(1.202)
dN
= 0.723 2.33 107 N 1.13 1013 2N + 5.82 105 N O 1.00 1013 N N O .
dt
(1.203)
Solving numerically, we obtain a solution shown in Fig. 1.8. The numerics show a relaxation
to final concentrations of
mole
lim N O = 7.336 107 , (1.204)
t cm3
mole
lim N = 3.708 108 , (1.205)
t cm3
(1.206)
Equations (1.202-1.203) are of the form

dN O
= fN O (N O , N ), (1.207)
dt
dN
= fN (N O , N ). (1.208)
dt
At equilibrium, we must have
fN O (N O , N ) = 0, (1.209)
fN (N O , N ) = 0. (1.210)
We find three finite roots to this problem:

mole
1 : (N O , N ) = (1.605 106 , 3.060 108 ) , non-physical, (1.211)
cm3
mole
2 : (N O , N ) = (5.173 108 , 2.048 106 ) , non-physical, (1.212)
cm3
mole
3 : (N O , N ) = (7.336 107 , 3.708 108 ) physical, . (1.213)
cm3
Obviously, because of negative concentrations, roots 1 and 2 are non-physical. Root 3
however is physical; moreover, it agrees with the equilibrium we found by direct numerical
integration of the full non-linear equations.
We can use local linear analysis in the neighborhood of each equilibria to rigorously
ascertain the stability of each root. Taylor series expansion of Eqs. (1.207-1.208) in the
neighborhood of an equilibrium point yields

d e fN O e fN O
( N O ) = fN O |e + ( N O ) + ( eN ) + . . . ,
dt N O | {z } N O e N O N e N
=0
(1.214)

d e fN e fN
( N ) = fN |e + ( N O ) + ( eN ) + . . . . (1.215)
dt N |{z} N O e N O N e N
=0
Evaluation of Eqs. (1.214-1.215) near the physical root, root 3, yields the system

d N O 7.336 107 2.129 106 4.155 105 N O 7.336 107
=
dt N 3.708 108 2.111 105 3.144 107 N 3.708 108
| {z }
f
=J= |e
(1.216)
This is of the form

d e f
( ) = ( e ) = J ( e ) . (1.217)
dt e
It is the eigenvalues of the Jacobian matrix J that give the time scales of evolution of the
concentrations as well as determine the stability of the local equilibrium point. Recall that
we can usually decompose square matrices via the diagonlization
J = P1 P. (1.218)
Here P is the matrix whose columns are composed of the right eigenvectors of J, and is
the diagonal matrix whose diagonal is populated by the eigenvalues of J. For some matrices
(typically not those encountered after our removal of linear dependencies), diagonalization
is not possible, and one must resort to the so-called near-diagonal Jordan form. This will
not be relevant to our discussion, but could be easily handled if necessary. We also recall the
eigenvector matrix and eigenvalue matrix are defined by the standard eigenvalue problem
P J = P. (1.219)
We also recall that the components of are found by solving the characteristic polynomial
which arises from the equation
det (J I) = 0, (1.220)
where I is the identity matrix. With the decomposition Eq. (1.218), Eq. (1.217) can be
rearranged to form
d
(P ( e )) = P ( e ). (1.221)
dt
Taking
z P ( e ), (1.222)
Eq. (1.221) reduces the diagonal form

dz
= z. (1.223)
dt
This has solution for each component of z of
z1 = C1 exp(1 t), (1.224)

z2 = C2 exp(2 t), (1.225)
..
. (1.226)
Here, our matrix J, see Eq. (1.216), has two real, negative eigenvalues in the neighborhood
of the physical root 3:
1
1 = 3.143 107 , (1.227)
s
1
2 = 2.132 106 . (1.228)
s
Thus we can conclude that the physical equilibrium is linearly stable. The local time con-
stants near equilibrium are given by the reciprocal of the magnitude of the eigenvalues.
These are
1 = 1/|1| = 3.181 108 s, (1.229)

2 = 1/|2| = 4.691 107 s. (1.230)
Evolution on these two time scales is predicted in Fig. 1.8. This in fact a multiscale problem.
One of the major difficulties in the numerical simulation of combustion problems comes in
the effort to capture the effects at all relevant scales. The problem is made more difficult as
the breadth of the scales expands. In this problem, the breadth of scales is not particularly
challenging. Near equilibrium the ratio of the slowest to the fastest time scale, the stiffness
ratio , is
2 4.691 107 s
= = = 14.75. (1.231)
1 3.181 108 s
Many combustion problems can have stiffness ratios over 106 . This is more prevalent at
lower temperatures.
We can do a similar linearization near the initial state, find the local eigenvalues, and
the the local time scales. At the initial state here, we find those local time scales are
1 = 2.403 108 s, (1.232)
2 = 2.123 108 s. (1.233)
So initially the stiffness, = (2.403 108 s)/(2.123 108 s) = 1.13 is much less, but the
time scale itself is small. It is seen from Fig. 1.8 that this initial time scale of 108 s well
predicts where significant evolution of species concentrations commences. For t < 108 s, the
model predicts essentially no activity. This can be correlated with the mean time between
molecular collisionsthe theory on which estimates of the collision frequency factors aj are
obtained.
We briefly consider the non-physical roots, 1 and 2. A similar eigenvalue analyis of root
1 reveals that the eigenvalues of its local Jacobian matrix are
1
1 = 1.193 107 , (1.234)
s
1
2 = 5.434 106 . (1.235)
s
Thus root 1 is a saddle and is unstable.
For root 2, we find
1
1 = 4.397 107 + i7.997 106 , (1.236)
s
1
2 = 4.397 107 i7.997 106 . (1.237)
s
The eigenvalues are complex with a positive real part. This indicates the root is an unstable
spiral source.
A detailed phase portrait is shown in Fig. 1.9. Here we see all three roots. Their local
character of sink, saddle, or spiral source is clearly displayed. We see that trajectories are
attracted to a curve labeled SIM for Slow Invariant Manifold. A part of the SIM is
constructed by the trajectory which originates at root 1 and travels to root 3. The other
part is constructed by connecting an equilibrium point at infinity into root 3. Details are
omitted here.
-7
x 10
5
saddle 3 sink
0 SIM
SIM 1
[N] (mole/cc)
-5
-10
-15
spiral
-20 source
2
-4 -3 -2 -1 0 1 2 -6
x 10
[NO] (mole/cc)
Figure 1.9: NO and N phase portaits for T = 6000 K, P = 2.4942 106 dyne/cm2
Zeldovich mechanism.
@ND,@NOD HmoleccL
10-6
@NOD
10-8
10-10
10-12
@ND
t HsL
10-11 10-9 10-7 10-5 0.001 0.1 10
Figure 1.10: N O and N versus time for Zeldovich mechanism at T = 1500 K, P =

6.23550 105 dyne/cm2 .
1.1.2.2 Stiffness, time scales, and numerics

One of the key challenges in computational chemistry is accurately predicting species concen-
tration evolution with time. The problem is made difficult because of the common presence
of physical phenomena which evolve on a widely disparate set of time scales. Systems which
evolve on a wide range of scales are known as stiff, recognizing a motivating example in
mass-spring-damper systems with stiff springs. Here we will examine the effect of tempera-
ture and pressure on time scales and stiffness. We shall also look simplistically how different
numerical approximation methods respond to stiffness.
1.1.2.2.1 Effect of temperature Let us see how the same Zeldovich mechanism be-
haves at lower temperature, T = 1500 K; all other parameters, including the initial species
concentrations are the same as the previous high temperature example. The pressure how-
ever, lowers, and here is P = 6.23550105 dyne/cm2 , which is close to atmospheric pressure.
For this case, a plot of species concentrations versus time is given in Figure 1.10.
At T = 1500 K, we notice some dramatic differences relative to the earlier studied
T = 6000 K. First, we see the reaction commences in around the same time, t 108 s. For
t 106 s, there is a temporary cessation of significant reaction. We notice a long plateau
in which species concentrations do not change over several decades of time. This is actually
a pseudo-equilibrium. Significant reaction recommences for t 0.1 s. Only around t 1 s
does the system approach final equilibrium. We can perform an eigenvalue analysis both
at the initial state and at the equilibrium state to estimate the time scales of reaction. For
this dynamical system which is two ordinary differential equations in two unknowns, we will
always find two eigenvalues, and thus two time scales. Let us call them 1 and 2 . Both
these scales will evolve with t.
At the initial state, we find
1 = 2.37 108 s, (1.238)

2 = 4.25 107 s. (1.239)
The onset of significant reaction is consistent with the prediction given by 1 at the initial
state. Moreover, initially, the reaction is not very stiff; the stiffness ratio is = 17.9.
At equilibrium, we find
mole
lim N O = 4.6 109 , (1.240)
t cm3
mole
lim N = 4.2 1014 , (1.241)
t cm3
and
1 = 7.86 107 s, (1.242)

2 = 3.02 101 s. (1.243)
The slowest time scale near equilibrium is an excellent indicator of how long the system
takes to relax to its final state. Note also that near equilibrium, the stiffness ratio is large,
= 2 /1 3.8 105 . This is known as the stiffness ratio. When it is large, the scales in
the problem are widely disparate and accurate numerical solution becomes challenging.
In summary, we find the effect of lowering temperature while leaving initial concentrations
constant
lowers the pressure somewhat, slightly slowing down the collision time, and slightly
slowing the fastest time scales,
slows the slowest time scales many orders of magnitude, stiffening the system signifi-
cantly, since collisions may not induce reaction with their lower collision speed.
1.1.2.2.2 Effect of initial pressure Let us maintain the initial temperature at T =

1500 K, but drop the initial concentration of each species to
b mole
N O = b
N = b
N2 = b
O2 = b
O = 108 . (1.244)
cm3
With this decrease in number of moles, the pressure now is
dyne
P = 6.23550 103 . (1.245)
cm2
This pressure is two orders of magnitude lower than atmospheric. We solve for the species
concentration profiles and show the results of numerical prediction in Figure 1.11 Relative to
@ND,@NOD HmoleccL
10-8
@NOD
10-10
10-12
10-14
@ND
10-16
t HsL
10-8 10-6 10-4 0.01 1 100
Figure 1.11: N O and N versus time for Zeldovich mechanism at T = 1500 K, P =

6.2355 103 dyne/cm2 .
the high pressure P = 6.2355 105 dyne/cm2 , T = 1500 K case, we notice some similarities
and dramatic differences. The overall shape of the time-profiles of concentration variation
is similar. But, we see the reaction commences at a much later time, t 106 s. For
t 104 s, there is a temporary cessation of significant reaction. We notice a long plateau
in which species concentrations do not change over several decades of time. This is again
actually a pseudo-equilibrium. Significant reaction recommences for t 10 s. Only around
t 100 s does the system approach final equilibrium. We can perform an eigenvalue analysis
both at the initial state and at the equilibrium state to estimate the time scales of reaction.
At the initial state, we find
1 = 2.37 106 s, (1.246)
2 = 4.25 105 s. (1.247)
The onset of significant reaction is consistent with the prediction given by 1 at the initial
state. Moreover, initially, the reaction is not very stiff; the stiffness ratio is = 17.9.
Interestingly, by decreasing the initial pressure by a factor of 102 , we increased the initial
time scales by a complementary factor of 102 ; moreover, we did not alter the stiffness.
At equilibrium, we find
mole
lim N O = 4.6 1011 , (1.248)
t cm3
mole
lim N = 4.2 1016 , (1.249)
t cm3
(1.250)
and
1 = 7.86 105 s, (1.251)
2 = 3.02 101 s. (1.252)
By decreasing the initial pressure by a factor of 102 , we decreased the equilibrium concentra-
tions by a factor of 102 and increased the time scales by a factor of 102 , leaving the stiffness
ratio unchanged.
In summary, we find the effect of lowering the initial concentrations significantly while
leaving temperature constant
lowers the pressure significantly, proportionally slowing down the collision time, as well
as the fastest and slowest time scales,
does not affect the stiffness of the system.
1.1.2.2.3 Stiffness and numerics The issue of how to simulate stiff systems of ordinary
differential equations, such as presented by our Zeldovich mechanism, is challenging. Here a
brief summary of some of the issues will be presented. The interested reader should consult
the numerical literature for a full discussion. See for example the excellent text of Iserles. 9
We have seen throughout this section that there are two time scales at work, and they are
often disparate. The species evolution is generally characterized by an initial fast transient,
followed by a long plateau, then a final relaxation to equilibrium. We noted from the phase
plane of Fig. 1.9 that the final relaxation to equilibrium (shown along the green line labeled
SIM) is an attracting manifold for a wide variety of initial conditions. The relaxation onto
the SIM is fast, and the motion on the SIM to equilibrium is relatively slow.
Use of common numerical techniques can often mask or obscure the actual dynamics.
Numerical methods to solve systems of ordinary differential equations can be broadly cat-
egorized as explicit or implicit. We give a brief synopsis of each class of method. We cast
each as a method to solve a system of the form
d
= f(). (1.253)
dt
Explicit: The simplest of these methods, the forward Euler method, discretizes Eq. (1.253)
as follows:
n+1 n
= f(n ), (1.254)
t
so that
n+1 = n + t f(n ) (1.255)
Explicit methods are summarized as
9
A. Iserles, 2008, A First Course in the Numerical Analysis of Differential Equations, Cambridge Uni-
versity Press, Cambridge, UK.
easy to program, since Eq. (1.255) can be solved explicitly to predict the new
value n+1 in terms of the old values at step n.
need to have t < f astest in order to remain numerically stable,
able to capture all physics and all time scales at great computational expense for
stiff problems,
requiring much computational effort for little payoff in the SIM region of the phase
plane, and thus
inefficient for some portions of stiff calculations.
Implicit: The simplest of these methods, the backward Euler method, discretizes
Eq. (1.253) as follows:
n+1 n
= f(n+1 ), (1.256)
t
so that
n+1 = n + t f(n+1 ) (1.257)
Implicit methods are summarized as
more difficult to program since a non-linear set of algebraic equations, Eq. (1.257),
must be solved at every time step with no guarantee of solution,
requiring potentially significant computational time to advance each time step,
capable of using very large time steps and remaining numerically stable,
suspect to missing physics that occur on small time scales < t,
in general better performers than explicit methods.
A wide variety of software tools exist to solve systems of ordinary differential equations.
Most of them use more sophisticated techniques than simple forward and backward Euler
methods. One of the most powerful techniques is the use of error control. Here the user
specifies how far in time to advance and the error that is able to be tolerated. The algorithm,
which is complicated, selects then internal time steps, for either explicit or implicit methods,
to achieve a solution within the error tolerance at the specified output time. A well known
public domain algorithm with error control is provided by lsode.f, which can be found in
the netlib repository. 10
Let us exercise the Zeldovich mechanism under the conditions simulated in Fig. 1.11,
T = 1500 K, P = 6.2355 103 dyne/cm2 . Recall in this case the fastest time scale near
equilibrium is 1 = 7.86 105 s 104 s at the initial state, and the slowest time scale is
10
Hindmarsh, A. C., 1983,ODEPACK, a Systematized Collection of ODE Solvers, Scien-
tific Computing, edited by R. S. Stepleman, et al., North-Holland, Amsterdam, pp. 55-64.
http://www.netlib.org/alliant/ode/prog/lsode.f
Explicit Explicit Implicit Implicit

t (s) Ninternal tef f (s) Ninternal tef f (s)
102 106 104 100 102
101 105 104 100 101
100 104 104 100 100
101 103 104 100 101
102 102 104 100 102
103 101 104 100 103
104 100 104 100 104
105 100 105 100 105
106 100 106 100 106
Table 1.3: Results from computing Zeldovich NO production using implicit and explicit
methods with error control in dlsode.f.
= 3.02 101 s at the final state. Let us solve for these conditions using dlsode.f, which
uses internal time stepping for error control, in both an explicit and implicit mode. We
specify a variety of values of t and report typical values of number of internal time steps
selected by dlsode.f, and the corresponding effective time step tef f used for the problem,
for both explicit and implicit methods, as reported in Table 1.3.
Obviously if output is requested using t > 104 s, the early time dynamics near t
104 s will be missed. For physically stable systems, codes such as dlsode.f will still provide
a correct solution at the later times. For physically unstable systems, such as might occur in
turbulent flames, it is not clear that one can use large time steps and expect to have fidelity
to the underlying equations. The reason is the physical instabilities may evolve on the same
time scale as the fine scales which are overlooked by large t.
1.2 Adiabatic, isochoric kinetics

It is more practical to allow for temperature variation within a combustor. The best model
for this is adiabatic kinetics. Here we will restrict our attention to isochoric problems.
1.2.1 Thermal explosion theory

There is a simple description known as thermal explosion theory which provides a good
explanation for how initially slow exothermic reaction induces a sudden temperature rise
accompanied by a final relaxation to equilibrium.
Let us consider a simple isomerizaion reaction in a closed volume
A B. (1.258)
Let us take A and B to both be calorically perfect ideal gases with identical molecular masses
MA = MB = M and identical specific heats, cvA = cvB = cv ; cP A = cP B = cP . We can
1.2. ADIABATIC, ISOCHORIC KINETICS 55
consider A and B to be isomers of an identical molecular species. So we have N = 2 species

reacting in J = 1 reactions. The number of elements L here is irrelevant.
1.2.1.1 One-step reversible kinetics

Let us insist our reaction process be isochoric and adiabatic, and commence with only A
present. The reaction kinetics with = 0 are

dA E 1
= a exp A , (1.259)
dt RT Kc B
| {z }
=k
| {z }
=r

dB E 1
= a exp A , (1.260)
dt RT Kc B
| {z }
=k
| {z }
=r
A (0) = b
A , (1.261)
B (0) = 0. (1.262)
For our alternate compact linear algebra based form, we note that

E 1
r = a exp A , (1.263)
RT Kc B
and that

d A 1
= (r) (1.264)
dt B 1
Performing the decomposition yields

d A 1
= (r). (1.265)
dt A + B 0
Expanded, this is

1 0 d A 1
= (r). (1.266)
1 1 dt B 0
Combining Eqs. (1.259-1.260) and integrating yields

d
( + B ) = 0, (1.267)
dt A
A + B = b
A , (1.268)
B = b
A A . (1.269)
Thus, Eq. (1.259) reduces to

dA E 1 b
= a exp A A . (1.270)
dt RT Kc A
Scaling, Eq. (1.270) can be rewritten as

d A E 1 A 1 A
= exp 1 . (1.271)
d(at) bA RTo T /To b
A Kc b
A
1.2.1.2 First law of thermodynamics

Recall the first law of thermodynamics and neglecting potential and kinetic energy changes:
dE
= Q W . (1.272)
dt
Here E is the total internal energy. Because we insist the problem is adiabatic Q = 0.
= 0. Thus we have
Because we insist the problem is isochoric, there is no work done, so W
dE
= 0. (1.273)
dt
Thus we find
E = Eo (1.274)
Recall the total internal energy for a mixture of two calorically perfect ideal gases is
E = nA eA + nB eB , (1.275)
n nB
A
= V eA + eB , (1.276)
V V
= V (A eA + B eB ) , (1.277)

PA PB
= V A hA + B (hB , (1.278)
A B

= V A hA RT + B (hB RT , (1.279)
o
o

= V A cP (T To ) + hTo ,A RT + B (cP (T To ) + hTo ,B RT ,(1.280)
o o

= V (A + B )(cP (T To ) RT ) + A hTo ,A + B hTo ,B , (1.281)
o o

= V (A + B )((cP R)T cP To ) + A hTo ,A + B hTo ,B , (1.282)
o o

= V (A + B )((cP R)T (cP R + R)To ) + A hTo ,A + B hTo ,B , (1.283)
o o

= V (A + B )(cv T (cv + R)To ) + A hTo ,A + B hTo ,B , (1.284)
o o

= V (A + B )cv (T To ) + A (hTo ,A RTo ) + B (hTo ,B RTo ) , (1.285)

= V (A + B )cv (T To ) + A eoTo ,A + B eoTo ,B . (1.286)
Now at the initial state, we have T = To , so

b o b o
Eo = V A eTo ,A + B eTo ,B (1.287)
So, we can say our caloric equation of state is

b o b o
E Eo = V (A + B )cv (T To ) + (A A )eTo ,A + (B B )eTo ,B , (1.288)

= V (b A + b
B )cv (T To ) + (A b
A )eoTo ,A + (B b
B )eoTo ,B . (1.289)
As an aside, on a molar basis, we scale Eq. (1.289) to get
e eo = cv (T To ) + (yA yAo )eoTo ,A + (yB yBo )eoTo ,B . (1.290)
And because we have assumed the molecular masses are the same, MA = MB , the mole
fractions are the mass fractions, and we can write on a mass basis
e eo = cv (T To ) + (YA YAo )eoTo ,A + (YB YBo )eoTo ,B . (1.291)
Returning to Eq. (1.289), our energy conservation relation, Eq. (1.274), becomes

0 = V (b A + bB )cv (T To ) + (A b
A )eoTo ,A + (B b
B )eoTo ,B . (1.292)
Now we solve for T
0 = (b
A + bB )cv (T To ) + (A b A )eoTo ,A + (B b
B )eoTo ,B , (1.293)
b A o b B o
0 = cv (T To ) + A eTo ,A + B eT ,B , (1.294)
b
A + b
B b
A + b B o
A eoTo ,A b
b B eoTo ,B
T = To + A + B . (1.295)
b
A + bB cv b
A + b B cv
Now we impose our assumption that b

B = 0, giving also B = b
A A ,
A A eoTo ,A B eoTo ,B
b
T = To + , (1.296)
b
A cv b
A cv
A A eoTo ,A eoTo ,B
b
= To + . (1.297)
b
A cv
o o
In summary, realizing that hTo ,A hTo ,B = eoTo ,A eoTo ,B we can write T as a function of
A :
(b
A A ) o o
T = To + (hTo ,A hTo ,B ). (1.298)
b
A cv
o o
We see then that if hTo ,A > hTo ,B , that as A decreases from its initial value of b
A that T
will increase. We can scale Eq. (1.298) to form
o o !
T A hTo ,A hTo ,B
=1+ 1 . (1.299)
To b
A cv To
We also note that our caloric state equation, Eq. (1.290) can, for yAo = 1, yBo = 0 as
e eo = cv (T To ) + (yA 1)eoTo ,A + yB eoTo ,B , (1.300)

= cv (T To ) + ((1 yB ) 1)eoTo ,A + yB eoTo ,B , (1.301)
= cv (T To ) yB (eoTo ,A eoTo ,B ). (1.302)
Similarly, on a mass basis, we can say,
e eo = cv (T To ) YB (eoTo ,A eoTo ,B ). (1.303)
For this problem, we also have

Go
Kc = exp , (1.304)
RT
with
Go = g oB g oA , (1.305)
o o
= hB T soB (hA T soA ), (1.306)
o o
= (hB hA ) T (soB soA ), (1.307)
o o
= (hTo ,B hTo ,A ) T (soTo ,B soTo ,A ). (1.308)
So
o o !
hTo ,A hTo ,B T (soTo ,A soTo ,B )
Kc = exp , (1.309)
RT
o o !!
cv To hTo ,A hTo ,B T (soTo ,A soTo ,B )
= exp , (1.310)
RT cv To
o o !!
1 1 hTo ,A hTo ,B T (soTo ,A soTo ,B )
= exp , (1.311)
1 TTo cv To To cv
Here we have used the definition of the ratio of specific heats, = cP /cv along with R =
cP cv . So we can solve Eq. (1.270) by first using Eq. (1.311) to eliminate Kc and then
Eq. (1.298) to eliminate T .
1.2.1.3 Dimensionless form

Let us try writing dimensionless variables so that our system can be written in a compact
dimensionless form. First lets take dimensionless time to be
= at. (1.312)
Let us take dimensionless species concentration to be z with

A
z= . (1.313)
b
A
Let us take dimensionless temperature to be with
T
= . (1.314)
To
Let us take dimensionless heat release to be q with
o o
hT ,A hTo ,B
q= o . (1.315)
cv To
Let us take dimensionless activation energy to be with
E
= . (1.316)
RTo
And let us take the dimensionless entropy change to be with
(soTo ,A soTo ,B )
= (1.317)
cv
So our equations become

dz 1
= exp z (1 z) , (1.318)
d Kc
= 1 + (1 z)q, (1.319)

1 1
Kc = exp (q ) (1.320)
1
It is more common to consider the products. Let us define for general problems
B B
= = . (1.321)
A + B b
A + b
B
Thus is the mass fraction of product. For our problem, b

B = 0 so
B c A
= = A . (1.322)
b
A b
A
Thus,
= 1 z, (1.323)
We can think of as a reaction progress variable as well. When = 0, we have = 0, and

the reaction has not begun. Thus, we get

d 1
= exp (1 ) , (1.324)
d Kc
= 1 + q, (1.325)

1 1
Kc = exp (q ) (1.326)
1
1.2.1.4 Example calculation

Let us choose some values for the dimensionless parameters:
7
= 20, = 0, q = 10, = . (1.327)
5
With these choices, our kinetics equations reduce to

d 20 25
= exp (1 ) exp , (0) = 0. (1.328)
d 1 + 10 1 + 10
The right side of Eq. (1.328) is at equilibrium for values of which drive it to zero.
Numerical root finding methods show this to occur at 0.920539. Near this root, Taylor
series expansion shows the dynamics are approximated by
d
( 0.920539) = 0.17993( 0.920539) + . . . (1.329)
d
Thus the local behavior near equilibrium is given by
= 0.920539 + C exp (0.17993 ) . (1.330)
Here C is some arbitrary constant. Clearly the equilibrium is stable, with a time constant
of 1/0.17993 = 5.55773.
Numerical solution shows the full behavior of the dimensionless species concentration
( ); see Figure 1.12. Clearly the product concentration is small for some long period of
time. At a critical time near = 2.7 106 , there is a so-called thermal explosion with a
rapid increase in . Note that the estimate of the time constant near equilibrium is orders
of magnitude less than the explosion time, 5.55773 << 2.7 106 . Thus, linear analysis
here is a poor tool to estimate an important physical quantity, the ignition time. Once the
ignition period is over, there is a rapid equilibration to the final state. The dimensionless
temperature plot is shown in Figure 1.13. The temperature plot is similar in behavior to

1.0
0.8
0.6
0.4
0.2

0 1. 106 2. 106 3. 106 4. 106 5. 106
Figure 1.12: Dimensionless plot of reaction product concentration versus time for adia-
batic isochoric combustion with simple reversible kinetics.

10

1. 106 2. 106 3. 106 4. 106 5. 106
Figure 1.13: Dimensionless plot of temperature versus time for adiabatic, isochoric
combustion with simple reversible kinetics.
the species concentration plot. At early time, the temperature is cool. At a critical time,
the thermal explosion time, the temperature rapidly rises. This rapid rise, coupled with the
exponential sensitivity of reaction rate to temperature, accelerates the formation of product.
This process continues until the reverse reaction is activated to the extent it prevents further
creation of product.
1.2.1.5 High activation energy asymptotics

Let us see if we can get an analytic prediction of the thermal explosion time, 2.7 106 .
Such a prediction would be valuable to see how long a slowing reacting material might take
to ignite. Our analysis is similar to that given by Buckmaster and Ludford in their Chapter
1. 11
For convenience let us restrict ourselves to = 0. In this limit, Eqs. (1.324-1.326) reduce
to

d q
= exp (1 ) exp , (1.331)
d 1 + q ( 1)(1 + q)
with (0) = 0. The key trouble in getting an analytic solution to Eq. (1.331) is the presence
of in the denominator of an exponential term. We need to find a way to move it to the
numerator. Asymptotic methods provide one such way.
Now we recall for early time << 1. Let us assume takes the form
= 1 + 2 2 + 3 3 + . . . (1.332)
Here we will assume 0 < << 1 and that 1 ( ) O(1), 2 ( ) O(1), . . ., and will define
in terms of physical parameters shortly. Now with this assumption, we have
1 1
= . (1.333)
1 + q 1 + q1 + 2 q2 + 3 q3 + . . .
Long division of the term on the right side yields the approximation
1
= = 1 q1 + 2 (q 2 21 q2 ) + . . . , (1.334)
1 + q
= 1 q1 + O(2 ). (1.335)
So

exp exp (1 q1 + O(2 )) , (1.336)
1 + q

exp() exp q1 + O(2 ) , (1.337)
We have moved from the denominator to the numerator of the most important exponential
term.
11
J. D. Buckmaster and G. S. S. Ludford, 1983, Lectures on Mathematical Combustion, SIAM, Philadelphia.
Now, let us take the limit of high activation energy by defining to be

1
. (1.338)

Let us let the assume the remaining parameters, q and are both O(1) constants. When
is large, will be small. With this definition, Eq. (1.337) becomes

1
exp exp exp q1 + O(2 ) , (1.339)
1 + q
With these assumptions and approximations, Eq. (1.331) can be written as

d 1
(1 + . . .) = exp exp q1 + O(2 )
d

q
(1 1 . . .) (1 + . . .) exp .
( 1)(1 + q1 + . . .)
(1.340)
Now let us rescale time via

1 1
= exp . (1.341)

With this transformation, the chain rule shows how derivatives transform:
d d d 1 d
= = 1
. (1.342)
d d d exp
d
With this transformation, Eq. (1.340) becomes

1 d 1
1
(1 + . . .) = exp exp q1 + O(2 )
exp d

q
(1 1 . . .) (1 + . . .) exp .
( 1)(1 + q1 + . . .)
(1.343)
This simplifies to
d
(1 + . . .) = exp q1 + O(2 )
d

q
(1 1 . . .) (1 + . . .) exp .
( 1)(1 + q1 + . . .)
(1.344)
Retaining only O(1) terms in Eq. (1.344), we get
d1
= exp (q1 ) . (1.345)
d
This is supplemented by the initial condition 1 (0) = 0. Separating variables and solving,
we get
exp(q1 )d1 = d , (1.346)

1
exp(q1 ) = + C. (1.347)
q
Applying the initial condition gives

1
exp(q(0)) = C, (1.348)
q
1
= C. (1.349)
q
So
1 1
exp(q1 ) = , (1.350)
q q
exp(q1 ) = q + 1, (1.351)

1
exp(q1 ) = q , (1.352)
q

1
q1 = ln q , (1.353)
q

1 1
1 = ln q . (1.354)
q q
For q = 10, a plot of 1 ( ) is shown in Fig. 1.14. We note at a finite that 1 begins to
exhibit unbounded growth. In fact, it is obvious from Eq. (1.345) that as
1
,
q
that
1 .
That is there exists a finite time for which 1 violates the assumptions of our asymptotic
theory which assumes 1 = O(1). We associate this time with the ignition time, i :
1
i = . (1.355)
q
y1
1.4
1.2
1.0
0.8
0.6
0.4
0.2
*
0.00 0.02 0.04 0.06 0.08 0.10 0.12
Figure 1.14: 1 versus for ignition problem.
Let us return this to more primitive variables:

1 1 1
exp i = , (1.356)
q

exp 1
i = , (1.357)
q
exp
i = . (1.358)
q
For our system with = 20 and q = 10, we estimate the dimensionless ignition time as
exp 20
i = = 2.42583 106 . (1.359)
(20)(10)
This is a surprisingly good estimate, given the complexity of the problem. Recall the nu-
merical solution showed ignition for 2.7 106 .
In terms of dimensional time, ignition time prediction becomes
exp
ti = , (1.360)
aq
!
1 RTo cv To E
= o o exp , (1.361)
a E hTo ,A hTo ,B RTo
Note the ignition is suppressed if the ignition time is lengthened, which happens when
he activation energy E is increased, since the exponential sensitivity is stronger than

the algebraic sensitivity,
o o
the energy of combustion (hTo ,A hTo ,B ) is decreased because it takes longer to react
to drive the temperature to a critical value to induce ignition,
the collision frequency factor a is decreased, which suppresses reaction.
1.2.2 Detailed H2 O2 N2 kinetics

Here is an example which uses multiple reactions for an adiabatic isothermal system is given.
Consider the full time-dependency of a problem similar to the thermal explosion problem
just considered. We choose a non-intuitive set of parameters for the problem. Our choices
will enable a direct comparison to a detonation of the same mixture via the same reaction
mechanism in a later chapter.
A closed, fixed, adiabatic volume, V = 0.3061251 cm3 , contains at t = 0 s a stoichiometric
hydrogen-air mixture of 2 105 mole of H2 , 1 105 mole of O2 , and 3.76 105 mole of
N2 at Po = 2.83230 106 P a and To = 1542.7 K. 12 Thus the initial molar concentrations
are
H2 = 6.533 105 mole/cm3 ,
O2 = 3.267 105 mole/cm3 ,
H2 = 1.228 104 mole/cm3 .
The initial mass fractions are calculated via Yi = Mi i /. They are
YH2 = 0.0285,
YO2 = 0.226,
YN2 = 0.745.
To avoid issues associated with numerical roundoff errors at very early time for species
with very small compositions, the minor species were initialized at a small non-zero value
near machine precision; each was assigned a value of 1015 mole. The minor species all have
i = 1.803 1016 mole/cm3 . They have correspondingly small initial mass fractions.
We seek the reaction dynamics as the system proceeds from its initial state to its final
state. We use the reversible detailed kinetics mechanism of Table 1.2. This problem requires
a detailed numerical solution. Such a solution was performed by solving the appropriate
equations for a mixture of nine interacting species: H2 , H, O, O2 , OH, H2 O, HO2 , H2 O2 ,
and N2 . The dynamics of the reaction process are reflected in Figs. 1.15-1.17
12
This temperature and pressure correspond to that of the same ambient mixture of H2 , O2 and N2 which
was shocked from 1.01325 105 P a, 298 K, to a value associated with a freely propagating detonation.
Relevant comparisons of reaction dynamics will be made in a later chapter
0 H
10
O
H2
O2
OH
H2 O
5 H2 O 2
10 HO2
N2
i
Y
10
10
15
10
10 5 0
10 10 10
t (s)
Figure 1.15: Plot of YH2 (t), YH (t), YO (t), YO2 (t), YOH (t), YH2 O (t), YHO2 (t), YH2 O2 (t), YN2 (t),
for adiabatic, isochoric combustion of a mixture of 2H2 + O2 + 3.76N2 initially at To =
1542.7 K, Po = 2.8323 106 P a.
4000
3500
3000
2500
T (K)
2000
1500
1000
500
0 10 5
10 10
t (s)
Figure 1.16: Plot of T (t), for adiabatic, isochoric combustion of a mixture of 2H2 +O2 +3.76N2
initially at To = 1542.7 K, Po = 2.8323 106 P a.
7
10
P (Pa)
6
10 10 5
10 10
t (s)
Figure 1.17: Plot of P (t), for adiabatic, isochoric combustion of a mixture of 2H2 + O2 +
3.76N2 initially at To = 1542.7 K, Po = 2.8323 106 P a.
At early time, t < 107 s, the pressure, temperature, and major reactant species con-
centrations (H2 , O2 , N2 ) are nearly constant. However, the minor species, e.g. OH, HO2 ,
and the major product, H2 O, are undergoing very rapid growth, albeit with math fractions
whose value remains small. In this period, the material is in what is known as the induction
period.
After a certain critical mass of minor species has accumulated, exothermic recombination
of these minor species to form the major product H2 O induces the temperature to rise, which
accelerates further the reaction rates. This is manifested in a thermal explosion. A common
definition of the end of the induction period is the induction time, t = tind , the time when
dT /dt goes through a maximum. Here one finds
tind = 6.6 107 s. (1.362)
A close-up view of the species concentration profiles is given in Fig. 1.18

At the end of the induction zone, there is a final relaxation to equilibrium. The equilib-
rium mass fractions of each species are
YO2 = 1.85 102 , (1.363)

YH = 5.41 104 , (1.364)
YOH = 2.45 102 , (1.365)
YO = 3.88 103 , (1.366)
Y H2 = 3.75 103 , (1.367)
Y H2 O = 2.04 101 , (1.368)
0
10
H
O
5 H2
10
O2
i
Y
OH
H2 O
H2 O 2
10
10 HO2
N2
2 4 6 8 10
t (s) 7
x 10
Figure 1.18: Plot near thermal explosion time of YH2 (t), YH (t), YO (t), YO2 (t), YOH (t),
YH2 O (t), YHO2 (t), YH2 O2 (t), YN2 (t), for adiabatic, isochoric combustion of a mixture of
2H2 + O2 + 3.76N2 initially at at To = 1542.7 K, Po = 2.8323 106 P a.
YHO2 = 6.84 105 , (1.369)

YH2 O2 = 1.04 105 , (1.370)
YN2 = 7.45 101 . (1.371)
We note that because our model takes N2 to be inert that its value remains unchanged.
Other than N2 , the final products are dominated by H2 O. The equilibrium temperature is
3382.3 K and 5.53 106 P a.
Chapter 2
Gas mixtures
One is often faced with mixtures of simple compressible substances, and it the thermody-
namics of such mixtures upon which attention is now fixed. Here a discussion of some of
the fundamentals of mixture theory will be given. In general, thermodynamics of mixtures
can be a challenging topic about which much remains to be learned. In particular, these
notes will focus on ideal mixtures of ideal gases, for which results are often consistent with
intuition. The chemical engineering literature contains a full discussion of the many nu-
ances associated with non-ideal mixtures of non-ideal materials. Relevant background for
this chapter is found in standard undergraduate texts. 1 2 3 4 Some of these notes are
adaptations of material found in these texts, especially Sonntag, et al..
2.1 Some general issues

Here the notation of Sonntag, et al. will be used. There is no effective consensus on notation
for mixtures. That of Sonntag, et al. is more unusual than most; however, the ideas are
correct, which is critical. Consider a mixture of N components, each a pure substance, so
that the total mass and total number of moles are
N
X
m = m1 + m2 + m3 + . . . mN = mi , mass, units= kg, (2.1)
i=1
N
X
n = n1 + n2 + n3 + . . . nN = ni , moles, units= kmole. (2.2)
i=1
1
R. E. Sonntag, C. Borgnakke, and G. J. von Wylen, 2003, Fundamentals of Thermodynamics, Sixth
Edition, John Wiley, New York.
2
S. I. Sandler, 1998, Chemical and Engineering Thermodynamics, Third Edition, John Wiley, New York.
3
J. M. Smith, H. C. van Ness, and M. Abbott, 2004, Introduction to Chemical Engineering Thermody-
namics, Seventh Edition, McGraw-Hill, New York.
4
J. W. Tester and M. Modell, 1997, Thermodynamics and Its Applications, Third Edition, Prentice Hall,
Upper Saddle River, New Jersey.
71
72 CHAPTER 2. GAS MIXTURES
Recall 1 mole = 6.02 1023 . The mass fraction of component i is defined as Yi :

mi
Yi , mass fraction, dimensionless. (2.3)
m
The mole fraction of component i is defined as yi :
ni
yi , mole fraction, dimensionless. (2.4)
n
Now the molecular mass of species i is the mass of a mole of species i. It units are typically
g/mole. This is an identical unit to kg/kmole. Molecular mass is sometimes called molec-
ular weight, but this is formally incorrect, as it is a mass measure, not a force measure.
Mathematically the definition of Mi corresponds to

mi kg g
Mi , = . (2.5)
ni kmole mole
Then one gets mass fraction in terms of mole fraction as
mi
Yi = , (2.6)
m
ni Mi
= , (2.7)
m
ni Mi
= PN , (2.8)
j=1 mj
ni Mi
= PN , (2.9)
j=1 nj M j
ni M i
= 1
PN n , (2.10)
n j=1 nj Mj
ni M i
= PN n nj Mj , (2.11)
j=1 n
yi Mi
= PN . (2.12)
j=1 y j Mj
Similarly, one finds mole fraction in terms of mass fraction by the following:
ni
yi = , (2.13)
n
mi
Mi
= PN mj , (2.14)
j=1 Mj
mi
Mi m
= PN mj , (2.15)
j=1 Mj m
Yi
Mi
= P N Yj . (2.16)
j=1 Mj
2.1. SOME GENERAL ISSUES 73
The mixture itself has a mean molecular mass:

m
M , (2.17)
n
PN
i=1 mi
= , (2.18)
n
X N
ni Mi
= , (2.19)
i=1
n
N
X
= yi Mi . (2.20)
i=1
Example 2.1
Air is often modelled as a mixture in the following molar ratios:
O2 + 3.76N2 (2.21)
Find the mole fractions, the mass fractions, and the mean molecular mass of the mixture.
Take O2 to be species 1 and N2 to be species 2. Consider the number of moles of O2 to be
n1 = 1 kmole,
and N2 to be
n2 = 3.76 kmole.
kg kg
The molecular mass of O2 is M1 = 32 kmole . The molecular mass of N2 is M2 = 28 kmole . The total
number of moles is
n = 1 kmole + 3.76 kmole = 4.76 kmole.
So the mole fractions are
1 kmole
y1 = = 0.2101.
4.76 kmole
3.76 kmole
y2 = = 0.7899.
4.76 kmole
Note that
N
X
yi = 1. (2.22)
i=1
That is to say y1 + y2 = 0.2101 + 0.7899 = 1. Now for the masses, one has

kg
m1 = n1 M1 = (1 kmole) 32 = 32 kg,
kmole

kg
m2 = n2 M2 = (3.76 kmole) 28 = 105.28 kg,
kmole
So one has
m = m1 + m2 = 32 kg + 105.28 kg = 137.28 kg.
The mass fractions then are
m1 32 kg
Y1 = = = 0.2331,
m 137.28 kg
m2 105.28 kg
Y2 = = = 0.7669.
m 137.28 kg
Note that
N
X
Yi = 1. (2.23)
i=1
That is Y1 + Y2 = 0.2331 + 0.7669 = 1. Now for the mixture molecular mass, one has
m 137.28 kg kg
M= = = 28.84 .
n 4.76 kmole kmole
Check against another formula.
N
X kg
M= yi Mi = y1 M1 + y2 M2 = (0.2101)(32) + (0.7899)(28) = 28.84 .
i=1
kmole
Now postulates for mixtures are not as well established as those for pure substances.
The literature has much controversial discussion of the subject. A strong advocate of the
axiomatic approach, C. A. Truesdell, proposed the following metaphysical principles for
mixtures, which are worth considering. 5
1. All properties of the mixture must be mathematical consequences of properties of the

constituents.
2. So as to describe the motion of a constituent, we may in imagination isolate it from the

rest of the mixture, provided we allow properly for the actions of the other constituents
upon it.
3. The motion of the mixture is governed by the same equations as is a single body.
Most important for the present discussion is the first principle. When coupled with fluid
mechanics, the second two take on additional importance. The approach of mixture theory
is to divide and conquer. One typically treats each of the constituents as a single material
and then devises appropriate average or mixture properties from those of the constituents.
The best example of this is air, which is not a single material, but is often treated as
such.
2.2 Ideal and non-ideal mixtures

A general extensive property, such as E, for an N-species mixture will be such that
E = E(T, P, n1 , n2 , . . . , nN ) (2.24)
5
C. A. Truesdell, 1984, Rational Thermodynamics, Springer-Verlag, New York.
2.3. IDEAL MIXTURES OF IDEAL GASES 75
A partial molar property is a generalization of an intensive property, and is defined such that
it is the partial derivative of an extensive property with respect to number of moles, with T
and P held constant. For internal energy, then the partial molar internal energy is

E
ei . (2.25)
ni T,P,nj ,i6=j
Pressure and temperature are held constant because those are convenient variables to control
in an experiment. One also has the partial molar volume

V
vi = . (2.26)
ni T,P,nj ,i6=j
It shall be soon seen that there are other natural ways to think of the volume per mole.
Now in general one would expect to find
ei = ei (T, P, n1 , n2 , . . . , nN ), (2.27)
vi = vi (T, P, n1 , n2 , . . . , nN ). (2.28)
This is the case for what is known as a non-ideal mixture. An ideal mixture is defined as a
mixture for which the partial molar properties ei and vi are not functions of the composition,
that is
ei = ei (T, P ), if ideal mixture, (2.29)

vi = vi (T, P ), if ideal mixture. (2.30)
An ideal mixture also has the property that hi = hi (T, P ), while for a non-ideal mixture
hi = hi (T, P, n1 , . . . , nN ). Though not obvious, it will turn out that some properties of an
ideal mixture will depend on composition. For example, the entropy of a constituent of an
ideal mixture will be such that
si = si (T, P, n1, n2 , . . . , nN ). (2.31)
2.3 Ideal mixtures of ideal gases

The most straightforward mixture to consider is an ideal mixture of ideal gases. Even here,
there are assumptions necessary that remain difficult to verify absolutely.
2.3.1 Dalton model

The most common model for a mixture of ideal gases is the Dalton model. Key assumptions
define this model
Each constituent shares a common temperature.
Each constituent occupies the entire volume.

Each constituent possesses a partial pressure which sums to form the total pressure of
the mixture.
The above characterize a Dalton model for any gas, ideal or non-ideal. One also takes for
convenience
Each constituent behaves as an ideal gas.
The mixture behaves as a single ideal gas.
It is more convenient to deal on a molar basis for such a theory. For the Dalton model,
additional useful quantities, the species mass concentration i , the mixture mass concentra-
tion , the species molar concentration i , and the mixture molar concentration , can be
defined. As will be seen, these definitions for concentrations are useful; however, they are
not in common usage. Following Sonntag, et al., the bar notation, , will be reserved for
properties which are mole-based rather than mass-based. As mentioned earlier, the notion
of a partial molal property is discussed extensively in the chemical engineering literature
and has implications beyond those considered here. For the Dalton model, in which each
component occupies the same volume, one has
Vi = V. (2.32)
The mixture mass concentration, also called the density is simply

m kg
= , . (2.33)
V m3
The mixture molar concentration is

n kmole
= , . (2.34)
V m3
For species i, the equivalents are

mi kg
i = , , (2.35)
V m3

ni kmole
i = , . (2.36)
V m3
One can find a convenient relation between species molar concentration and species mole
fraction by the following operations
ni n
i = , (2.37)
V n
ni n
= , (2.38)
nV
= yi . (2.39)
A similar relation exists between species molar concentration and species mass fraction via
ni m Mi
i = , (2.40)
V m Mi
=m
z }|i{
m ni Mi
= , (2.41)
V mMi
=Y
i
z}|{
mi 1
= , (2.42)
m Mi
Yi
= . (2.43)
Mi
The specific volumes, mass and molar, are similar. One takes
V V
v = , v= , (2.44)
m n
V V
vi = , vi = . (2.45)
mi ni
Note that this definition of molar specific volume is not the partial molar volume defined in
V
the chemical engineering literature, which takes the form vi = n i
.
T,P,nj ,i6=j
For the partial pressure of species i, one can say for the Dalton model
N
X
P = Pi . (2.46)
i=1
For species i, one has
Pi V = ni RT, (2.47)
ni RT
Pi = , (2.48)
V
N
X XN
ni RT
Pi = , (2.49)
V
|i=1{z } i=1
=P
N
RT X
P = ni . (2.50)
V i=1
| {z }
=n
So, for the mixture, one has

P V = nRT. (2.51)
One could also say

n
P =
RT = RT. (2.52)
V
Here n is the total number of moles in the system. Additionally R is the universal gas
constant with value
kJ J
R = 8.314472 = 8.314472 . (2.53)
kmole K mole K
Sometimes this is expressed in terms of kb the Boltzmann constant and NA , Avogadros
number:
R = kb NA , (2.54)
1
NA = 6.02214199 1023 , (2.55)
mole
J
kb = 1.3806505 1023 . (2.56)
K
Example 2.2
Compare the molar specific volume defined here with the partial molar volume from the chemical
engineering literature.
The partial molar volume vi , is given by

V
vi = . (2.57)
ni T,P,nj ,i6=j
For the ideal gas, one has

N
X
PV = RT nk , (2.58)
k=1
PN
RT k=1 nk
V = , (2.59)
P
PN nk
V RT k=1 ni
= , (2.60)
ni T,P,nj ,i6=j P
PN
RT k=1 ki
= , (2.61)
P
=0 =0 =1 =0

z}|{ z}|{ z}|{ z}|{
RT 1i + 2i + . . . + ii + . . . + N i
= , (2.62)
P
RT
vi = , (2.63)
P
V
= PN , (2.64)
k=1 nk
V
= . (2.65)
n
Here the so-called Kronecker delta function has been employed, which is much the same as the identity
matrix:
ki = 0, k 6= i, (2.66)
ki = 1, k = i. (2.67)
Contrast this with the earlier adopted definition of molar specific volume
V
vi = . (2.68)
ni
So, why is there a difference? The molar specific volume is a simple definition. One takes the
instantaneous volume V , which is shared by all species in the Dalton model, and scales it by the
instantaneous number of moles of species i, and acquires a natural definition of molar specific volume
consistent with the notion of a mass specific volume. On the other hand, the partial molar volume
specifies how the volume changes if the number of moles of species i changes, while holding T and P
and all other species mole numbers constant. One can imagine adding a mole of species i, which would
necessitate a change in V in order to guarantee the P remain fixed.
2.3.1.1 Binary mixtures

Consider now a binary mixture of two components A and B. This is easily extended to a
general mixture of N components. First the total number of moles is the sum of the parts:
n = nA + nB . (2.69)
Now, write the ideal gas law for each component:
PA VA = nA RTA , (2.70)
PB VB = nB RTB . (2.71)
But by the assumptions of the Dalton model, VA = VB = V , and TA = TB = T , so
PA V = nA RT, (2.72)
PB V = nB RT. (2.73)
One also has

P V = nRT. (2.74)
Solving for n, nA and nB , one finds
PV
n = , (2.75)
RT
PA V
nA = , (2.76)
RT
PB V
nB = . (2.77)
RT
Now n = nA + nB , so one has

PV PA V PB V
= + . (2.78)
RT RT RT
P = PA + PB . (2.79)
That is the total pressure is the sum of the partial pressures. This is a mixture rule for
pressure
One can also scale each constituent ideal gas law by the mixture ideal gas law to get
PA V nA RT
= , (2.80)
PV nRT
PA nA
= , (2.81)
P n
= yA , (2.82)
PA = yA P. (2.83)
Likewise
PB = yB P. (2.84)
Now, one also desires rational mixture rules for energy, enthalpy, and entropy. Invoke Trues-
dells principles on a mass basis for internal energy. Then the total internal energy E (with
units J) for the binary mixture must be
E = me = mA eA + mB eB , (2.85)
m mB
A
= m eA + eB , (2.86)
m m
= m (YA eA + YB eB ) , (2.87)
e = YA eA + YB eB . (2.88)
For the enthalpy, one has
H = mh = mA hA + mB hB , (2.89)
m mB
A
= m hA + hB , (2.90)
m m
= m (YA hA + YB hB ) , (2.91)
h = YA hA + YB hB . (2.92)
It is easy to extend this to a mole fraction basis rather than a mass fraction basis. One can
also obtain a gas constant for the mixture on a mass basis. For the mixture, one has
PV = nRT mRT, (2.93)

PV
mR = nR, (2.94)
T
= (nA + nB )R, (2.95)

mA mB
= + R, (2.96)
MA MB

R R
= mA + mB , (2.97)
MA MB
= (m R + mB RB ) , (2.98)
mA A mB
A
R = RA + RB , (2.99)
m m
R = (YA RA + YB RB ) . (2.100)
For the entropy, one has
S = ms = = mA sA + mB sB , (2.101)
m mB
A
= m sA + sB , (2.102)
m m
= m (YA sA + YB sB ) , (2.103)
s = Y A sA + Y B sB . (2.104)
Note that sA is evaluated at T and PA , while sB is evaluated at T and PB . For a CPIG, one
has

o T PA
sA = sTo ,A + cP A ln RA ln , (2.105)
To Po
| {z }
soT,A

T yA P
= soTo ,A + cP A ln RA ln . (2.106)
To Po
Likewise

T yB P
sB = soTo ,B + cP B ln RB ln . (2.107)
To Po
| {z }
soT,B
Here the o denotes some reference state. As a superscript, it typically means that the
property is evaluated at a reference pressure. For example, soT,A denotes the portion of the
entropy of component A that is evaluated at the reference pressure Po and is allowed to vary
with temperature T . Note also that sA = sA (T, P, yA) and sB = sB (T, P, yB ), so the entropy
of a single constituent depends on the composition of the mixture and not just on T and P .
This contrasts with energy and enthalpy for which eA = eA (T ), eB = eB (T ), hA = hA (T ),
hB = hB (T ) if the mixture is composed of ideal gases. Occasionally, one finds hoA and hoB
used as a notation. This denotes that the enthalpy is evaluated at the reference pressure.
However, if the gas is ideal, the enthalpy is not a function of pressure and hA = hoA , hB = hoB .
If one is employing a calorically imperfect ideal gas model, then one finds for species i
that
o yi P
si = sT,i Ri ln , i = A, B. (2.108)
Po
2.3.1.2 Entropy of mixing
Example 2.3
Initially calorically perfect ideal gases A and B are segregated within the same large volume by
a thin frictionless, thermally conducting diaphragm. Thus, both are at the same initial pressure and
temperature, P1 and T1 . The total volume is thermally insulated and fixed, so there are no global heat
or work exchanges with the environment. The diaphragm is removed, and A and B are allowed to mix.
Assume A has mass mA and B has mass mB . The gases are allowed to have distinct molecular masses,
MA and MB . Find the final temperature T2 , pressure P2 , and the change in entropy.
The ideal gas law holds that at the initial state
mA RA T1 mB RB T1
VA1 = , VB1 = . (2.109)
P1 P1
At the final state one has
T1
V2 = VA2 = VB2 = VA1 + VB1 = (mA RA + mB RB ) . (2.110)
P1
Mass conservation gives
m2 = m1 = mA + mB . (2.111)
One also has the first law
E2 E1 = Q W, (2.112)
E2 E1 = 0, (2.113)
E2 = E1 , (2.114)
m2 e 2 = mA eA1 + mB eB1 , (2.115)
(mA + mB )e2 = mA eA1 + mB eB1 , (2.116)
0 = mA (eA1 e2 ) + mB (eB1 e2 ), (2.117)
0 = mA cvA (T1 T2 ) + mB cvB (T1 T2 ), (2.118)
mA cvA T1 + mB cvB T1
T2 = , (2.119)
mA cvA + mB cvB
= T1 . (2.120)
The final pressure by Daltons law then is
P2 = PA2 + PB2 (2.121)

mA RA T2 mB RB T2
= + , (2.122)
V2 V2
mA RA T1 mB RB T1
= + , (2.123)
V2 V2
(mA RA + mB RB ) T1
= , (2.124)
V2
(mA RA + mB RB ) T1
= , (2.125)
(mA RA + mB RB ) PT11
= P1 . (2.126)
So the initial and final temperatures and pressures are identical.

Now the entropy change of gas A is

TA2 PA2
sA2 sA1 = cP A ln RA ln , (2.127)
TA1 PA1

T2 yA2 P2
= cP A ln RA ln , (2.128)
T1 yA1 P1

T1 yA2 P1
= cP A ln RA ln , (2.129)
T1 yA1 P1
| {z }
=0

yA2 P1
= RA ln , (2.130)
(1)P1
= RA ln yA2 . (2.131)
Likewise
sB2 sB1 = RB ln yB2 . (2.132)
So the change in entropy of the mixture is
S = mA (sA2 sA1 ) + mB (sB2 sB1 ) (2.133)

= mA RA ln yA2 mB RB ln yB2 , (2.134)

R R
= (nA MA ) ln yA2 (nB MB ) ln yB2 , (2.135)
| {z } MA | {z } MB
=mA | {z } =mB | {z }
=RA =RB
= R(nA ln yA2 + nB ln yB2 ), (2.136)

nA nB
= R nA ln +nB ln , (2.137)
nA + nB nA + nB
| {z } | {z }
0 0
0. (2.138)
For an N-component mixture, mixed in the same fashion such that P and T are constant,
this extends to
N
X
S = R nk ln yk , (2.139)
k=1
N
!
X nk
= R nk ln PN 0, (2.140)
k=1 i=1 ni
| {z }
0
N
X nk
= R n ln yk , (2.141)
k=1
n
N
X nk
= Rn ln yk , (2.142)
k=1
n
XN
m
= R yk ln yk , (2.143)
M
k=1
N
X
= Rm ln ykyk , (2.144)
k=1
= Rm (ln y1y1 + ln y2y2 + . . . + ln yN
yN
), (2.145)
y1 y2 yN
= Rm ln (y1 y2 . . . yN ) , (2.146)
N
!
Y yk
= Rm ln yk , (2.147)
k=1
N
!
s s Y
= = ln ykyk . (2.148)
R R k=1
Note that there is a fundamental dependency of the mixing entropy on the mole fractions.
Since 0 yk 1, the product is guaranteed to be between 0 and 1. The natural logarithm
of such a number is negative, and thus the entropy change for the mixture is guaranteed
positive semi-definite. Note also that for the entropy of mixing, Truesdells third principle
is not enforced.
Now if one mole of pure N2 is mixed with one mole of pure O2 , one certainly expects
the resulting homogeneous mixture to have a higher entropy than the two pure components.
But what if one mole of pure N2 is mixed with another mole of pure N2 . Then we would
expect no increase in entropy. However, if we had the unusual ability to distinguish N2
molecules whose origin was from each respective original chamber, then indeed there would
be an entropy of mixing. Increases in entropy thus do correspond to increases in disorder.
2.3.1.3 Mixtures of constant mass fraction

If the mass fractions, and thus the mole fractions, remain constant during a process, the
equations simplify. This is often the case for common non-reacting mixtures. Air at moderate
values of temperature and pressure behaves this way. In this case, all of Truesdells principles
can be enforced. For a CPIG, one would have
e2 e1 = YA cvA (T2 T1 ) + YB cvB (T2 T1 ), (2.149)
= cv (T2 T1 ). (2.150)
where
cv YA cvA + YB cvB . (2.151)
Similarly for enthalpy
h2 h1 = YA cP A (T2 T1 ) + YB cP B (T2 T1 ), (2.152)
= cP (T2 T1 ). (2.153)
where
cP Y A cP A + Y B cP B . (2.154)
For the entropy
s2 s1 = YA (sA2 sA1 ) + YB (sB2 sB1 ), (2.155)

T2 y A P2 T2 y B P2
= YA cP A ln RA ln + YB cP B ln RB ln ,
T1 y A P1 T1 y B P1

T2 P2 T2 P2
= YA cP A ln RA ln + YB cP B ln RB ln (, 2.156)
T1 P1 T1 P1

T2 P2
= cP ln R ln . (2.157)
T1 P1
The mixture behaves as a pure substance when the appropriate mixture properties are de-
fined. One can also take
cP
k= . (2.158)
cv
2.3.2 Summary of properties for the Dalton mixture model

Listed here is a summary of mixture properties for an N-component mixture of ideal gases
on a mass basis:
N
X
M = yiMi , (2.159)
i=1
N
X
= i , (2.160)
i=1
1 1
v = PN 1
= , (2.161)
i=1 vi

N
X
e = Yiei , (2.162)
i=1
XN
h = Yihi , (2.163)
i=1
=R
N N
z }| { N
R X X yi Mi Ri R X
R = = YiRi = = yi , (2.164)
M i=1 i=1
N
X M i=1
yj Mj | {z }
=1
j=1
| {z } =M
N
X
cv = Yicvi , (2.165)
i=1
cv = cP R, if ideal gas (2.166)
XN
cP = Y i cP i , (2.167)
i=1
cP
k = , (2.168)
cv
XN
s = Y i si , (2.169)
i=1
yi Mi
Yi = , (2.170)
M
Pi = yi P, (2.171)
i = Yi , (2.172)
v 1
vi = = , (2.173)
Yi i
V = Vi , (2.174)
T = Ti , (2.175)
hi = hoi , if ideal gas, (2.176)
Pi
hi = ei + = ei + Pi vi = ei + Ri T , (2.177)
i | {z }
if ideal gas
Z T
hi = hoTo ,i + cP i (T ) dT, (2.178)
To
| {z }
if ideal gas
Z
cP i (T)
T
Pi
si = soTo ,i + dT Ri ln , (2.179)
To T Po
| {z }
=soT,i
| {z }
if ideal gas

o yi P o Pi
si = sT,i Ri ln = sT,i Ri ln , (2.180)
Po Po
| {z } | {z }
if ideal gas if ideal gas
Ri T
Pi = i Ri T = Ri T Yi = , (2.181)
vi
| {z }
if ideal gas
X N
Yi RT
P = RT = RT = , (2.182)
i=1
Mi v
| {z }
if ideal gas
N
X Z T
h = YihoTo ,i + cP (T ) dT , (2.183)
To
|i=1 {z }
if ideal gas
P
h = e+ = e + P v = e| +{zRT} , (2.184)

if ideal gas
N Z N
!
X T
cP (T ) P Y yi
s = YisoTo ,i + dT R ln R ln yi
i=1 To T Po i=1
| {z }
if ideal gas
(2.185)
These relations are not obvious. A few are derived in examples here.
Example 2.4
Derive the expression h = e + P/.
Start from the equation for the constituent hi , multiply by mass fractions, sum over all species,
and use properties of mixtures:
Pi
hi = ei + , (2.186)
i
Pi
Yi hi = Yi ei + Yi , (2.187)
i
N
X XN X N
Pi
Yi hi = Yi ei + Yi , (2.188)
i=1 i=1 i=1
i
N
X N
X N
X i Ri T
Yi hi = Yi ei + Yi , (2.189)
i=1 i=1 i=1
i
| {z } | {z }
=h =e
N
X
h = e+T Yi Ri , (2.190)
i=1
| {z }
=R
= e + RT, (2.191)
P
= e+ . (2.192)

Example 2.5
Find the expression for mixture entropy of the ideal gas.
Z
cP i (T)
T
Pi
si = soTo ,i + dT Ri ln (2.193)
To
T P o
Z T
o cP i (T) Pi
Yi si = Yi sTo ,i + Yi dT Yi Ri ln , (2.194)
To T Po
N N N Z T N
X X
o
X cP i (T) X Pi
s= Yi si = Yi sTo ,i + Yi dT Yi Ri ln , (2.195)
To T Po
i=1 i=1 i=1 i=1
N Z TX N N
X
o Yi cP i (T) X Pi
= Yi sTo ,i + dT Yi Ri ln , (2.196)
To i=1
T Po
i=1 i=1
Z T N
cP (T) X Pi
= soTo + dT Yi Ri ln . (2.197)
To T P o
i=1
All except the last term are natural extensions of the property for a single material. Consider now the
last term involving pressure ratios.
N N !
X Pi X Pi P P
Yi Ri ln = Yi Ri ln + R ln R ln , (2.198)
i=1
Po i=1
Po Po Po
N
!
X Ri Pi P P
= R Yi ln + ln ln , (2.199)
i=1
R Po Po Po
N !
X R/Mi Pi P P
= R Yi PN ln + ln ln , (2.200)
i=1 j=1 Yj R/Mj
Po Po Po

N !
X Y i /M i Pi P P
= R
P ln + ln ln , (2.201)
i=1
N
j=1 Yj /Mj
Po Po Po

| {z }
=yi
N !
X Pi P P
= R yi ln + ln ln , (2.202)
i=1
Po Po Po
N yi !
X Pi P P
= R ln ln + ln , (2.203)
i=1
Po Po Po
YN y ! !
Pi i Po P
= R ln + ln + ln , (2.204)
i=1
Po P Po
N y ! !
Po Y Pi i P
= R ln + ln , (2.205)
P i=1 Po Po
N
! !
Po 1 Y yi P
= R ln PN PN (Pi ) + ln , (2.206)
P i=1 yi Po i=1 yi i=1 Po
N y ! !
Y Pi i P
= R ln + ln , (2.207)
i=1
P Po
YN y ! !
yi P i P
= R ln + ln , (2.208)
i=1
P Po
N
! !
Y yi P
= R ln yi + ln . (2.209)
i=1
Po
So the mixture entropy becomes

Z N
! !
T
cP (T) Y P
s = soTo + dT R ln yiyi + ln , (2.210)
To T i=1
Po
Z N
!
T
cP (T) P Y yi
= soTo + dT R ln R ln yi . (2.211)
To T Po i=1
| {z } | {z }
classical entropy of a single body nonTruesdellian
The extra entropy is not found in the theory for a single material, and in fact is not in the form suggested
by Truesdells postulates. While it is in fact possible to redefine the constituent entropy definition in
such a fashion that the mixture entropy in fact takes
on the classical form of a single material via the
RT
definition si = soTo ,i + To cP iT(T ) dT Ri ln PPoi + Ri ln yi , this has the disadvantage of predicting
no entropy change after mixing two pure substances. Such a theory would suggest that this obviously
irreversible process is in fact reversible.
On a molar basis, one has the equivalents

N
X n
= i = = , (2.212)
i=1
V M
1 V 1
v = PN 1
= = = vM, (2.213)
i=1 v i
n
N
X
e = yi ei = eM, (2.214)
i=1
XN
h = yi hi = hM, (2.215)
i=1
XN
cv = yi cvi = cv M, (2.216)
i=1
cv = cP R, (2.217)
XN
cP = y i cP i = cP M (2.218)
i=1
cP
k = , (2.219)
cv
XN
s = yi si = sM, (2.220)
i=1
i
i = yi = , (2.221)
Mi
V v 1
vi = = = = vi Mi , (2.222)
ni yi i

V V
vi = = = v = vM , (2.223)
ni P,T,nj |n {z }
if ideal gas
Pi = yi P, (2.224)
RT
P = RT = , (2.225)
| {z v }
if ideal gas
RT
Pi = i RT = , (2.226)
vi
| {z }
if ideal gas
P
h = e+ = e + P v = |e +{zRT} = hM, (2.227)

if ideal gas
o
hi = hi , if ideal gas, (2.228)
Pi
hi = ei + = ei + Pi v i = ei + P vi = ei + RT = hi Mi , (2.229)
i | {z }
if ideal gas
Z T
o
hi = hTo ,i + cP i (T) dT = hi Mi , (2.230)
To
| {z }
if ideal gas
Z
cP i(T )
T
yi P
si = soTo ,i + dT R ln , (2.231)
To T Po
| {z }
=soT,i
| {z }
if ideal gas

o yi P
si = sT,i R ln = si Mi , (2.232)
Po
| {z }
if ideal gas
N Z N
!
X T
cP (T) P Y
s = yi soTo ,i + dT R ln R ln yiyi = sM
To T Po
|i=1 {z i=1
}
if ideal gas
(2.233)
2.3.3 Amagat model*

The Amagat model is an entirely different paradigm than the Dalton model. It is not used
as often. In the Amagat model,
all components share a common temperature T ,
all components share a common pressure P ,
each component has a different volume.
Consider, for example, a binary mixture of calorically perfect ideal gases, A and B. For
the mixture, one has
P V = nRT, (2.234)
with
n = nA + nB . (2.235)
For the components one has
P VA = nA RT, (2.236)
P VB = nB RT. (2.237)
Then n = nA + nB reduces to
PV P VA P VB
= + . (2.238)
RT RT RT
Thus
V = VA + VB , (2.239)
VA VB
1 = + . (2.240)
V V
Chapter 3
Mathematical foundations of
thermodynamics*
This chapter focuses on mathematical formalism which can be applied to thermodynamics.

Understanding of calculus of many variables at an undergraduate level is sufficient mathe-
matical background for this chapter. Some details can be found in standard sources. 1 2
3
3.1 Exact differentials and state functions

In thermodynamics, one is faced with many systems of the form of the well-known Gibbs
equation:
de = T ds P dv. (3.1)
This is known to be an exact differential with the consequence that internal energy e is a
function of the state of the system and not the details of any process which led to the state.
As a counter-example, the work,
dw = P dv, (3.2)
can be shown to be an inexact differential so that the work is indeed a function of the process
involved.
Example 3.1
Show the work is not a state function. If work were a state function, one might expect it to have
the form
w = w(P, v), provisional assumption, to be tested. (3.3)
1
M. M. Abbott and H. C. van Ness, 1972, Thermodynamics, Schaums Outline Series in Engineering,
McGraw-Hill, New York. See Chapter 3.
2
Edition, John Wiley, New York. See Chapters 13 and 15.
3
W. G. Vincenti and C. H. Kruger, 1965, Introduction to Physical Gas Dynamics, John Wiley, New York.
See Chapter 3.
93
94 CHAPTER 3. MATHEMATICAL FOUNDATIONS OF THERMODYNAMICS*
In such a case, one would have the corresponding differential form

w w
dw = dv + dP. (3.4)
v P P v
Now since dw = P dv, one deduces that

w
= P, (3.5)
v P

w
= 0. (3.6)
P v
Integrating Eq. (3.5), one finds

w = P v + f (P ), (3.7)
where f (P ) is some function of P to be determined. Differentiating Eq. (3.7) with respect to P , one
gets
w df (P )
=v+ . (3.8)
P v dP

w
Now use Eq. (3.6) to eliminate P v
in Eq. (3.8) so as to obtain
df (P )
0 = v+ , (3.9)
dP
df (P )
= v. (3.10)
dP
Equation (3.10) cannot be: a function of P only cannot be a function of v. So, w cannot be a state
property:
w 6= w(P, v). (3.11)
Consider now the more general form

N
X
1 dx1 + 2 dx2 + . . . + N dxN = i dxi . (3.12)
i=1
Here i and xi , i = 1, . . . , N, may be thermodynamic variables. This form is known in

mathematics as a Pfaff differential form. As formulated, one takes at this stage
xi : independent thermodynamic variables,
i : thermodynamic variables which are functions of xi
Now, if the differential in Eq. (3.12), when set to a differential dy, can be integrated to form
the function
y = y(x1, x2 , . . . , xN ), (3.13)
3.1. EXACT DIFFERENTIALS AND STATE FUNCTIONS 95
the differential is said to be exact. In such a case, one has

N
X
dy = 1 dx1 + 2 dx2 + . . . + N dxN = i dxi . (3.14)
i=1
Now, if the algebraic definition of Eq. (3.13) holds, what amounts to the definition of the
partial derivative gives the parallel result that

y y y
dy = dx1 + dx2 + . . . + dxN . (3.15)
x1 xj ,j6=1 x2 xj ,j6=2 xN xj ,j6=N
Now, combining Eqs. (3.14) and (3.15) to eliminate dy, one gets

y y y
1 dx1 + 2 dx2 + . . . + N dxN = dx1 + dx2 + . . . + dxN .
(3.16)
Rearranging, one gets

! ! !
y y y
0= 1 dx1 + 2 dx2 + . . . + N dxN .
(3.17)
Since the variables xi , i = 1, . . . , N, are independent, dxi , i = 1, . . . , N, are all indepen-
dent in Eq. (3.17), and in general non-zero. For equality, one must require that each of the
coefficients be zero, so

y y y
1 = , 2 = , . . . , N = . (3.18)
So when dy is exact, one says that each of the i and xi are conjugate to each other.
From here on out, for notational ease, the j 6= 1, j 6= 2, . . . , j 6= N will be ignored in
the notation for the partial derivatives. It becomes especially confusing for higher order
derivatives, and is fairly obvious for all derivatives.
If y and all its derivatives are continuous and differentiable, then one has for all i =
1, . . . , N and k = 1, . . . , N that
2y 2y
= . (3.19)
xk xi xi xk
Now from Eq. (3.18), one has

y y
k = , l = . (3.20)
xk xj xl xj
Taking the partial of the first of Eq. (3.20) with respect to xl and the second with respect
to xk , one gets
k 2y l 2y
= , = . (3.21)
xl xj xl xk xk xj xk xl
Since by Eq. (3.19) the order of the mixed second partials does not matter, one deduces from
Eq. (3.21) that
k l
= (3.22)
xl xj xk xj
This is a necessary and sufficient condition for the exact-ness of Eq. (3.12). It is a gen-
eralization of what can be found in most introductory calculus texts for functions of two
variables.
For the Gibbs equation, (3.1), de = P dv + T ds, one has
y u, x1 v, x2 s, 1 P 2 T. (3.23)
and one expects the natural, or canonical form of
e = e(v, s). (3.24)
Here, P is conjugate to v, and T is conjugate to s. Application of the general form of
Eq. (3.22) to the Gibbs equation (3.1) gives then

T P
= . (3.25)
v s s v
Equation (3.25) is known as a Maxwell relation. Moreover, specialization of Eq. (3.20) to
the Gibbs equation (3.1) gives

e e
P = , T = . (3.26)
v s s v
P
If the general differential dy = Ni=1 i dxi is exact, one also can show
RBP
The path integral yB yA = A N i=1 i dxi is independent of the path of the integral.
The integral around a closed contour is zero:

I I X N
dy = i dxi = 0. (3.27)
i=1
The function y can only be determined to within an additive constant. That is, there
is no absolute value of y; physical significance is only ascribed to differences in y. In
fact now, other means, extraneous to this analysis, can be used to provide absolute
values of key thermodynamic variables. This will be important especially for flows
with reaction.
3.1. EXACT DIFFERENTIALS AND STATE FUNCTIONS 97
Example 3.2
Show the heat transfer q is not a state function. Assume all processes are fully reversible. The first
law gives
de = dq dw, (3.28)
dq = de + dw, (3.29)
= de + P dv. (3.30)
Take now the non-canonical, although acceptable, form e = e(T, v). Then one gets

e e
de = dv + dT. (3.31)
v T T v
So

e e
dq = dv + dT + P dv, (3.32)
v T v
T
e e
= + P dv + dT. (3.33)
v T T
| {z } | {z v}
M N
= M dv + N dT. (3.34)
Now by Eq. (3.22), for dq to be exact, one must have

M N
= , (3.35)
T v v T
(3.36)
This reduces to
2e P 2e
+ = . (3.37)
T v T v vT

P

P
This can only be true if T v = 0. But this is not the case; consider an ideal gas for which T v = R/v.
So dq is not exact.
Example 3.3
Show conditions for ds to be exact in the Gibbs equation.
de = T ds P dv, (3.38)
de P
ds = + dv, (3.39)
T T
1 e e P
= dv + dT + dv, (3.40)
T v T T v T

1 e P 1 e
= + dv + dT. (3.41)
T v T T T T
| {z } | {z v}
M N
Again, invoking Eq. (3.22), one gets then

1 e P 1 e
+ = , (3.42)
T T v T T v v T T v T

1 u2
1 e 1 P P 1 2u
+ = , (3.43)
T T v T 2 v T T T v T 2 T vT

1 e 1 P P
2 + = 0. (3.44)
T v T T T v T 2
This is the condition for an exact ds. Experiment can show if it is true. For example, for an ideal gas,
one finds from experiment that e = e(T ) and P v = RT , so one gets
1R 1 RT
0+ 2 = 0, (3.45)
T v T v
0 = 0. (3.46)
So ds is exact for an ideal gas. In fact, the relation is verified for so many gases, ideal and non-ideal,
that one simply asserts that ds is exact, rendering s to be path-independent and a state variable.
3.2 Two independent variables

Consider a general implicit function linking three variables, x, y, z:
f (x, y, z) = 0. (3.47)
In x y z space, this will represent a surface. If the function can be inverted, it will be
possible to write the explicit forms
x = x(y, z), y = y(x, z), z = z(x, y). (3.48)
Differentiating the first two of the Eqs. (3.48) gives

x x
dx = dy + dz, (3.49)
y z z y

y y
dy = dx + dz (3.50)
x z z x
Now use Eq. (3.50) to eliminate dy in Eq. (3.49):

x y y x
dx = dx + dz + dz, (3.51)
y z x z z x z y
| {z }
=dy
!
x y x y x
1 dx = + dz, (3.52)
y z x z y z z x z y
!
x y x y x
0dx + 0dz = 1 dx + + dz. (3.53)
y z x z y z z x z y
| {z } | {z }
=0 =0
3.2. TWO INDEPENDENT VARIABLES 99
Since x and y are independent, so are dx and dy, and the coefficients on each in Eq. (3.53)
must be zero. Therefore from the coefficient on dx in Eq. (3.53)

x y
1 = 0, (3.54)
y z x z

x y
= 1, (3.55)
y z x z

x 1
= y , (3.56)
y z
x z
and also from the coefficient on dz in Eq. (3.53)

x y x
+ = 0, (3.57)
y z z x z y

x x y
= , (3.58)
z y y z z x

x y z
= 1. (3.59)
z y x z y x
If one now divides Eq. (3.49) by a fourth differential, dw, one gets

dx x dy x dz
= + . (3.60)
dw y z dw z y dw
Demanding that z be held constant in Eq. (3.60) gives

x x y
= , (3.61)
w z y z w z

x
w z x
= , (3.62)
y y z
w z
x w x
= . (3.63)
w z y z y z
If x = x(y, w), one then gets

x x
dx = dy + dw. (3.64)
y w w y
Divide now by dy while holding z constant so

x x x w
= + . (3.65)
y z y w w y y z
These general operations can be applied to a wide variety of thermodynamic operations.
Example 3.4
Apply Eq. (3.65) to a standard P v T system and let

x T
= . (3.66)
y z v s
So T = x, v = y, and s = z. Let now e = w. So Eq. (3.65) becomes

T T T e
= + . (3.67)
v s v e e v v s
Now by definition

e
cv = , (3.68)
T v
so
T 1
= . (3.69)
e v cv

e
Now by Eq. (3.26), one has v s = P , so one gets

T T P
= . (3.70)
v s v e cv

T
For an ideal gas, e = e(T ). Inverting, one gets T = T (e), and so v e = 0, thus

T P
= . (3.71)
v s cv
For an isentropic process in an ideal gas, one gets
dT P RT
= = , (3.72)
dv cv cv v
dT R dv
= , (3.73)
T cv v
dv
= (k 1) , (3.74)
v
T vo
ln = (k 1) ln , (3.75)
To v
T v k1
o
= . (3.76)
To v
3.3. LEGENDRE TRANSFORMATIONS 101
3.3 Legendre transformations

The Gibbs equation (3.1), de = P dv + T ds, is the fundamental equation of classical
thermodynamics. It is a canonical form which suggests the most natural set of variables in
which to express internal energy e are s and v:
e = e(v, s). (3.77)
However, v and s may not be convenient for a particular problem. There may be other
combinations of variables whose canonical form gives a more convenient set of independent
variables for a particular problem. An example is the enthalpy:
h e + P v. (3.78)
Differentiating the enthalpy gives
dh = de + P dv + vdP. (3.79)
Use now Eq. (3.79) to eliminate de in the Gibbs equation to give
| P dv
dh {z vdP} = P dv + T ds, (3.80)
=de
dh = T ds + vdP. (3.81)
So the canonical variables for h are s and P . One then expects
h = h(s, P ). (3.82)
This exercise can be systematized with the Legendre transformation, 4 which defines a set
of second order polynomial combinations of variables. Consider again the exact differential
Eq. (3.14):
dy = 1 dx1 + 2 dx2 + . . . + N dxN . (3.83)
For N independent variables xi and N conjugate variables i , by definition there are 2N 1
Legendre transformed variables:
1 = 1 (1 , x2 , x3 , . . . , xN ) = y 1 x1 , (3.84)
2 = 2 (x1 , 2 , x3 , . . . , xN ) = y 2 x2 , (3.85)
..
.
N = N (x1 , x2 , x3 , . . . , N ) = y N xN , (3.86)
1,2 = 1,2 (1 , 2 , x3 , . . . , xN ) = y 1 x1 2 x2 , (3.87)
4
Two differentiable functions f and g are said to be Legendre transformations of each other if their first
derivatives are inverse functions of each other: Df = (Dg)1 . With some effort, not shown here, one can
prove that the Legendre transformations of this section satisfy this general condition.
1,3 = 1,3 (1 , x2 , 3 , . . . , xN ) = y 1 x1 3 x3 , (3.88)

..
. (3.89)
N
X
1,...,N = 1,...,N (1 , 2 , 3 , . . . , N ) = y i xi . (3.90)
i=1
Each is a new dependent variable. Each has the property that when it is known as a
function of its N canonical variables, the remaining N variables from the original expression
(the xi and the conjugate i ) can be recovered by differentiation of . In general this is not
true for arbitrary transformations.
Example 3.5
Let y = y(x1 , x2 , x3 ). This has the associated differential form
dy = 1 dx1 + 2 dx2 + 3 dx3 . (3.91)
Choose now a Legendre transformed variable 1 z(1 , x2 , x3 ):
z = y 1 x1 . (3.92)
Then
z z z
dz = d1 + dx2 + dx3 . (3.93)
1 x1 ,x2 x2 1 ,x3 x3 1 ,x2
Now differentiating Eq. (3.92), one also gets
dz = dy 1 dx1 x1 d1 . (3.94)
Elimination of dy in Eq. (3.94) by using Eq. (3.91) gives
dz = dx + 2 dx2 + 3 dx3 1 dx1 x1 d1 , (3.95)
|1 1 {z }
=dy
= x1 d1 + 2 dx2 + 3 dx3 . (3.96)
Thus from Eq. (3.93), one gets

z z z
x1 = , 2 = , 3 = . (3.97)
1 x1 ,x2 x2 1 ,x3 x3 1 ,x2
So the original expression had three independent variables x1 , x2 , x3 , and three conjugate variables
1 , 2 , 3 . Definition of the Legendre function z with canonical variables 1 , x2 , and x3 allowed
determination of the remaining variables x1 , 2 , and 3 in terms of the canonical variables.
For the Gibbs equation, (3.1), de = P dv + T ds, one has y = u, two canonical variables,
x1 = v and x2 = s, and two conjugates, 1 = P and 2 = T . Thus N = 2, and one can
expect 22 1 = 3 Legendre transformations. They are
1 = y 1 x1 = h = h(P, s) = e + P v, enthalpy, (3.98)
2 = y 2 x2 = a = a(v, T ) = e T s, Helmholtz free energy, (3.99)
1,2 = y 1 x1 2 x2 = g = g(P, T ) = e + P v T s, Gibbs free energy.
(3.100)
3.3. LEGENDRE TRANSFORMATIONS 103
It has already been shown for the enthalpy that dh = T ds + vdP , so that the canonical
variables are s and P . One then also has

h h
dh = ds + dP, (3.101)
s P P s
from which one deduces that

h h
T = , v= , (3.102)
s P P s
From Eq. (3.102), a second Maxwell relation can be deduced by differentiation of the first
with respect to P and the second with respect to s:

T v
= . (3.103)
P s s P
The relations for Helmholtz and Gibbs free energies each supply additional useful relations
including two new Maxwell relations. First consider the Helmholtz free energy
a = e T s, (3.104)
da = de T ds sdT, (3.105)
= (P dv + T ds) T ds sdT, (3.106)
= P dv sdT. (3.107)
So the canonical variables for a are v and T . The conjugate variables are P and s. Thus

a a
da = dv + dT. (3.108)
v T T v
So one gets
a a
P = , s = . (3.109)
v T T v
and the consequent Maxwell relation

P s
= . (3.110)
T v v T
For the Gibbs free energy
g = |e +{zP v} T s, (3.111)
=h
= h T s, (3.112)
dg = dh T ds sdT, (3.113)
= (T ds + vdP ) T ds sdT, (3.114)
| {z }
=dh
= vdP sdT. (3.115)
So for Gibbs free energy, the canonical variables are P and T while the conjugate variables
are v and s. One then has g = g(P, T ), which gives

g g
dg = dP + dT. (3.116)
P T T P
So one finds
g g
v= , s = . (3.117)
P T T P
The resulting Maxwell function is then

v s
= . (3.118)
T P P T
Example 3.6
Canonical Form
If R/cP
P s
h(s, P ) = cP To exp + (ho cP To ) , (3.119)
Po cP
and cP , To , R, Po , and ho are all constants, derive both thermal and caloric state equations P (v, T )
and e(v, T ).
Now for this material

R/cP
h P s
= To exp , (3.120)
s P Po cP
R/cP 1
h RTo P s
= exp . (3.121)
P s Po Po cP
Now since

h
= T, (3.122)
s P

h
= v, (3.123)
P s
one has
R/cP
P s
T = To exp , (3.124)
Po cP
R/cP 1
RTo P s
v = exp . (3.125)
Po Po cP
Dividing one by the other gives
T P
= , (3.126)
v R
Pv = RT , (3.127)
3.4. HEAT CAPACITY 105
which is the thermal equation of state. Substituting from Eq. (3.124) into the canonical equation for
h, Eq. (3.119), one also finds for the caloric equation of state
h = cP T + (ho cP To ) , (3.128)
h = cP (T To ) + ho . (3.129)
which is useful in itself. Substituting in for T and To ,

P v Po vo
h = cP + ho . (3.130)
R R
Using h e + P v we get

P v Po vo
e + P v = cP + eo + Po vo (3.131)
R R
so
c c
P P
e = 1 Pv 1 Po vo + uo , (3.132)
cR R
P
e = 1 (P v Po vo ) + eo , (3.133)
cR
P
e = 1 (RT RTo ) + eo , (3.134)
R
e = (cP R) (T To ) + eo , (3.135)
e = (cP (cP cv )) (T To ) + eo , (3.136)
e = cv (T To ) + eo . (3.137)
So one canonical equation gives us all the information one needs. Oftentimes, it is difficult to do a
single experiment to get the canonical form.
3.4 Heat capacity

Recall that

e
cv = , (3.138)
T v

h
cP = . (3.139)
T P
Then perform operations on the Gibbs equation
de = T ds P dv, (3.140)

e s
= T , (3.141)
T v T v

s
cv = T . (3.142)
T v
Likewise,
dh = T ds + vdP, (3.143)

h s
= T , (3.144)
T P T P

s
cP = T . (3.145)
T P
One finds further useful relations by operating on the Gibbs equation:
de = T ds P dv, (3.146)

e s
= T P, (3.147)
v T v T

P
= T P. (3.148)
T v
So one can then say
e = e(T, v), (3.149)

e e
de = dT + dv, (3.150)
T v v T

P
= cv dT + T P dv. (3.151)
T v
For an ideal gas, one has

e P R RT
= T P =T , (3.152)
v T T v v v
= 0. (3.153)
Consequently, e is not a function of v for an ideal gas, so e = e(T ) alone. Since h = e + P v

for an ideal gas reduces to h = e + RT
h = e(T ) + RT = h(T ). (3.154)
Now return to general equations of state. With s = s(T, v) or s = s(T, P ), one gets

s s
ds = dT + dv, (3.155)
T v v T

s s
ds = dT + dP. (3.156)
T P P T
3.4. HEAT CAPACITY 107
Now using Eqs. (3.103,3.118,3.142,3.145) one gets

cv P
ds = dT + dv, (3.157)
T T v

cP v
ds = dT dP. (3.158)
T T P
Subtracting one from the other, one finds

cv cP P v
0 = dT + dv + dP, (3.159)
T T v T P

P v
(cP cv )dT = T dv + T dP. (3.160)
T v T P
Now divide both sides by dT and hold either P or v constant. In either case, one gets

P v
cP cv = T . (3.161)
T v T P
Example 3.7
For an ideal gas find cP cv . For the ideal gas, P v = RT , one has

P R v R
= , = . (3.162)
T v v T P P
So
RR
cP cv = T , (3.163)
vP
2
R
= T , (3.164)
RT
= R. (3.165)
This holds even if the ideal gas is calorically imperfect. That is
cP (T ) cv (T ) = R. (3.166)
For the ratio of specific heats for a general material, one can use Eqs. (3.142) and (3.145)
to get

s
cP T T
k= = P , then apply Eq. (3.56) to get
s
(3.167)
cv T T
v
s T
= , then apply Eq. (3.58) to get (3.168)
T P s v

s P T v
= , (3.169)
P T T s v s s T

v s P T
= , (3.170)
s T P T T s v s

v P
= (3.171)
P v
T s
The first term can be obtained from P v T data. The second term is related to the
isentropic sound speed of the material, which is also a measurable quantity.
Example 3.8
For a calorically perfect ideal gas with gas constant R and specific heat at constant volume cv , find
expressions for the thermodynamic variable s and thermodynamic potentials e, h, a, and g, as functions
of T and P .
First get the entropy:
de = T ds P dv, (3.172)
T ds = de + P dv, (3.173)
T ds = cv dT + P dv, (3.174)
dT P
ds = cv + dv, (3.175)
T T
dT dv
= cv +R , (3.176)
Z Z T vZ
dT dv
ds = cv + R , (3.177)
T v
T v
s s0 = cv ln + R ln , (3.178)
T0 v0
s s0 T R RT /P
= ln + ln , (3.179)
cv T0 cv RT0 /P0
R/cv
T T P0
= ln + ln , (3.180)
T0 T0 P
1+R/cv R/cv
T P0
= ln + ln , (3.181)
T0 P
1+(cp cv )/cv (cp cv )/cv
T P0
= ln + ln , (3.182)
T0 P
k k1
T P0
= ln + ln , (3.183)
T0 P
k k1
T P0
s = s0 + cv ln + cv ln . (3.184)
T0 P
Now, for the calorically perfect ideal gas, one has
e = eo + cv (T To ). (3.185)
For the enthalpy, one gets
h = e + P v, (3.186)
3.5. VAN DER WAALS GAS 109
= e + RT, (3.187)
= eo + cv (T To ) + RT, (3.188)
= eo + cv (T To ) + RT + RTo RTo , (3.189)
= e + RT +c (T To ) + R(T To ), (3.190)
|o {z }o v
=ho
= ho + (cv + R)(T To ), (3.191)
| {z }
=cp
= ho + cp (T To ). (3.192)
For the Helmholtz free energy, one gets
a = e T s, (3.193)
k k1 !
T P0
= eo + cv (T To ) T s0 + cv ln + cv ln . (3.194)
T0 P
For the Gibbs free energy, one gets
g = h T s, (3.195)
k k1 !
T P0
= ho + cp (T To ) T s0 + cv ln + cv ln . (3.196)
T0 P
3.5 Van der Waals gas

A van der Waals gas is a common model for a non-ideal gas. It can capture some of the
behavior of a gas as it approaches the vapor dome. Its form is
RT a
P (T, v) = 2, (3.197)
vb v
where b accounts for the finite volume of the molecules, and a accounts for intermolecular
forces.
Example 3.9
Find a general expression for e(T, v) if
RT a
P (T, v) = . (3.198)
v b v2
Proceed as before: First we have

e e
de = dT + dv, (3.199)
T v v T
recalling that

e e P
= cv , = T P. (3.200)
T v v T T v
Now for the van der Waals gas, we have

P R
= , (3.201)
T v vb

P RT
T P = P, (3.202)
T v vb

RT RT a a
= 2 = 2. (3.203)
vb vb v v
So we have

e a
, = (3.204)
v T v2
a
e(T, v) = + f (T ). (3.205)
v
Here f (T ) is some as-of-yet arbitrary function of T . To evaluate f (T ), take the derivative with respect
to T holding v constant:

e df
= = cv . (3.206)
T v dT
Since f is a function of T at most, here cv can be a function of T at most, so we allow cv = cv (T ).

Integrating, we find f (T ) as
Z T
f (T ) = C + cv (T)dT, (3.207)
To
where C is an integration constant. Thus e is

Z T
a
e(T, v) = C + cv (T)dT . (3.208)
To v
Taking C = eo + a/vo , we get

Z T
1 1
e(T, v) = eo + cv (T)dT + a . (3.209)
To vo v
We also find
Z T
1 1
h = e + P v = eo + cv (T)dT + a
+ P v, (3.210)
To vo v
Z T
1 1 RT v a
h(T, v) = eo + cv (T)dT + a + . (3.211)
To vo v vb v
Example 3.10
Thermodynamic process with a van der Waals Gas
A van der Waals gas with

J
R = 200 , (3.212)
kg K
P a m6
a = 150 , (3.213)
kg 2
m3
b = 0.001 , (3.214)
kg

J
cv = (350 + 0.2(T 300)) , (3.215)
kg K
begins at T1 = 300 K, P1 = 1 105 P a. It is isothermally compressed to state 2 where P2 = 1 106 P a.
It is then isochorically heated to state 3 where T3 = 1000 K. Find w13 , q13 , and s3 s1 . Assume the
surroundings are at 1000 K. Recall
RT a
P = , (3.216)
v b v2
so at state 1
200 300 150
105 = 2 . (3.217)
v1 0.001 v1
Expanding, one gets
0.15 + 150v1 60100v12 + 100000v13 = 0. (3.218)
This is a cubic equation which has three solutions:
m3
v1 = 0.598 , physical, (3.219)
kg
m3
v1 = 0.00125 0.0097i not physical, (3.220)
kg
m3
v1 = 0.00125 + 0.0097i not physical. (3.221)
kg
Now at state 2, P2 and T2 are known, so v2 can be determined:
200 300 150
106 = 2 . (3.222)
v2 0.001 v2
The physical solution is v2 = 0.0585 m3 /kg. Now at state 3 it is known that v3 = v2 and T3 . Determine
P3 :
200 1000 150
P3 = = 3478261 43831 = 3434430 P a. (3.223)
0.0585 0.001 0.05852
R2 R3 R2
Now w13 = w12 + w23 = 1 P dv + 2 P dv = 1 P dv since 2 3 is at constant volume. So
Z v2
RT a
w13 = 2 dv, (3.224)
v1 vb v
Z v2 Z v2
dv dv
= RT1 a 2
, (3.225)
v1 v b v1 v

v2 b 1 1
= RT1 ln +a , (3.226)
v1 b v2 v1

0.0585 0.001 1 1
= 200 300 ln + 150 , (3.227)
0.598 0.001 0.0585 0.598
= 140408 + 2313, (3.228)
J
= 138095 , (3.229)
kg
kJ
= 138.095 . (3.230)
kg
The gas is compressed, so the work is negative. Since e is a state property:

Z T3
1 1
e3 e1 = cv (T )dT + a . (3.231)
T1 v1 v3
Now
1
cv = 350 + 0.2(T 300) = 290 + T. (3.232)
5
so
Z
T3
1 1 1
e3 e1 = 290 + T dT + a , (3.233)
T1 5 v1 v3

1 1 1
= 290 (T3 T1 ) + T32 T12 + a , (3.234)
10 v1 v3

1 1 1
= 290 (1000 300) + 10002 3002 + 150 , (3.235)
10 0.598 0.0585
= 203000 + 91000 2313, (3.236)
J
= 291687 , (3.237)
kg
kJ
= 292 . (3.238)
kg
Now from the first law
e3 e1 = q13 w13 , (3.239)

q13 = e3 e1 + w13 , (3.240)
q13 = 292 138, (3.241)
kJ
q13 = 154 . (3.242)
kg
The heat transfer is positive as heat was added to the system.
Now find the entropy change. Manipulate the Gibbs equation:
T ds = de + P dv, (3.243)
1 P
ds = de + dv, (3.244)
T T
1 a P
ds = cv (T )dT + 2 dv + dv, (3.245)
T v T
1 a 1 RT a
ds = cv (T )dT + 2 dv + 2 dv, (3.246)
T v T vb v
cv (T ) R
ds = dT + dv, (3.247)
T vb
Z T3
cv (T ) v3 b
s3 s1 = dT + R ln , (3.248)
T1 T v1 b
Z 1000
290 1 v3 b
= + dT + R ln , (3.249)
300 T 5 v1 b
1000 1 0.0585 0.001
= 290 ln + (1000 300) + 200 ln , (3.250)
300 5 0.598 0.001
= 349 + 140 468, (3.251)
J
= 21 , (3.252)
kg K
kJ
= 0.021 . (3.253)
kg K
Is the second law satisfied for each portion of the process?
First look at 1 2
e2 e1 = q12 w12 , (3.254)
q12 = e2 e1 + w12 , (3.255)
Z T2 !
1 1 v2 b 1 1
q12 = cv (T )dT + a + RT1 ln +a . (3.256)
T1 v1 v2 v1 b v2 v1
Recalling that T1 = T2 and canceling the terms in a, one gets

v2 b 0.0585 0.001 J
q12 = RT1 ln = 200 300 ln = 140408 . (3.257)
v1 b 0.598 0.001 kg
Since the process is isothermal,

v2 b
s2 s1 = R ln (3.258)
v1 b

0.0585 0.001
= 200 ln (3.259)
0.598 0.001
J
= 468.0 (3.260)
kg K
Entropy drops because heat was transferred out of the system.
Check the second law. Note that in this portion of the process in which the heat is transferred out
of the system, that the surroundings must have Tsurr 300 K. For this portion of the process let us
take Tsurr = 300 K.
q12
s2 s1 ? (3.261)
T
J
J 140408 kg
468.0 , (3.262)
kg K 300 K
J J
468.0 468.0 ok. (3.263)
kg K kg K
Next look at 2 3
q23 = e3 e2 + w23 , (3.264)

Z T3 ! Z v3
1 1
q23 = cv (T )dT + a + P dv , (3.265)
T2 v2 v3 v2
Z T3
since isochoric q23 = cv (T )dT, (3.266)
T2
Z 1000
T J
= 290 + dT = 294000 . (3.267)
300 5 kg
Now look at the entropy change for the isochoric process:

Z T3
cv (T )
s3 s2 = dT , (3.268)
T2 T
Z T3
290 1
= + dT , (3.269)
T2 T 5
1000 1
= 290 ln + (1000 300), (3.270)
300 5
J
= 489 . (3.271)
kg K
Entropy rises because heat transferred into system.
In order to transfer heat into the system we must have a different thermal reservoir. This one must
have Tsurr 1000 K. Assume here that the heat transfer was from a reservoir held at 1000 K to assess
the influence of the second law.
q23
s3 s2 ? (3.272)
T
J
J 294000 kg
489 , (3.273)
kg K 1000 K
J J
489 294 , ok. (3.274)
kg K kg K
3.6 Redlich-Kwong gas

The Redlich-Kwong equation of state is
RT a
P = . (3.275)
v b v(v + b)T 1/2
It is modestly more accurate than the van der Waals equation in predicting material behavior.
3.6. REDLICH-KWONG GAS 115
Example 3.11
For the case in which b = 0, find an expression for e(T, v) consistent with the Redlich-Kwong state
equation.
Here the equation of state is now
RT a
P = 2 1/2 . (3.276)
v v T
Proceeding as before, we have

e P
= T P, (3.277)
v T T
v
R a RT a
= T + 2 3/2 2 1/2 , (3.278)
v 2v T v v T
3a
= . (3.279)
2v 2 T 1/2
Integrating, we find
3a
e(T, v) = + f (T ). (3.280)
2vT 1/2
Here f (T ) is a yet-to-be-specified function of temperature only. Now the specific heat is found by the
temperature derivative of e:

e 3a df
cv (T, v) = = + . (3.281)
T v 4vT 3/2 dT
Obviously, for this material, cv is a function of both T and v.

Let us define cvo (T ) via
df
cvo (T ). (3.282)
dT
Integrating, then one gets
Z T
f (T ) = C + cvo (T) dT. (3.283)
To
1/2
Let us take C = eo + 3a/2/vo/To . Thus we arrive at the following expressions for cv (T, v) and e(T, v):
3a
cv (T, v) = cvo (T ) + , (3.284)
4vT 3/2 !
Z T
3a 1 1
e(T, v) = eo + cvo (T) dT + 1/2
(3.285)
To 2 vo To vT 1/2
3.7 Compressibility and generalized charts

A simple way to quantify the deviation from ideal gas behavior is to determine the so-called
compressibility Z, where
Pv
Z . (3.286)
RT
For an ideal gas, Z = 1. For substances with a simple molecular structure, Z can be
tabulated as functions of the so-called reduced pressure Pr and reduced temperature Tr . Tr
and Pr are dimensionless variables found by scaling their dimensional counterparts by the
specific materials temperature and pressure at the critical point, Tc and Pc :
T P
Tr , Pr . (3.287)
Tc Pc
Often charts are available which give predictions of all reduced thermodynamic properties.
These are most useful to capture the non-ideal gas behavior of materials for which tables are
not available.
3.8 Mixtures with variable composition

Consider now mixtures of N species. The focus here will be on extensive properties and
molar properties.
PN Assume that each species has ni moles, and the the total number of moles
is n = i=1 ni . Now one might expect the extensive energy to be a function of the entropy,
the volume and the number of moles of each species:
E = E(V, S, ni ). (3.288)
The extensive version of the Gibbs law in which all of the ni are held constant is
dE = P dV + T dS. (3.289)
Thus
E E
= P, = T. (3.290)
V S,ni S V,ni
In general, since E = E(S, V, ni ), one should expect, for systems in which the ni are allowed
to change that
N
E E X E
dE = dV + dS + dni . (3.291)
V S,ni S V,ni i=1
n i

S,V,n j
Defining the new thermodynamics property, the chemical potential i , as

E
i , (3.292)
ni S,V,nj
3.8. MIXTURES WITH VARIABLE COMPOSITION 117
one has the important Gibbs equation for multicomponent systems:

N
X
dE = P dV + T dS + i dni . (3.293)
i=1
Obviously, by its definition, i is on a per mole basis, so it is given the appropriate overline
notation. In Eq. (3.293), the independent variables and their conjugates are
x1 = V, 1 = P, (3.294)
x2 = S, 2 = T, (3.295)
x3 = n1 , 3 = 1 , (3.296)
x4 = n2 , 4 = 2 , (3.297)
.. ..
. .
xN +2 = nN , N +2 = N . (3.298)
Equation (3.293) has 2N +1 1 Legendre functions. Three are in wide usage: the extensive
analog to those earlier found. They are
H = E + P V, (3.299)
A = E T S, (3.300)
G = E + P V T S. (3.301)
A set of non-traditional, but perfectly acceptable additional Legendre functions would be
formed from E 1 n1 . Another set would be formed from E + P V 2 n2 . There are
many more, but one in particular is sometimes noted in the literature: the so-called grand
potential, . The grand potential is defined as
N
X
E TS i ni . (3.302)
i=1
Differentiating each defined Legendre function, Eqs. (3.299-3.302), and combining with
Eq. (3.293), one finds
N
X
dH = T dS + V dP + i dni , (3.303)
i=1
XN
dA = SdT P dV + i dni , (3.304)
i=1
XN
dG = SdT + V dP + i dni , (3.305)
i=1
XN
d = P dV SdT ni di . (3.306)
i=1
Thus, canonical variables for H are H = H(S, P, ni). One finds a similar set of relations as
before from each of the differential forms:

E H
T = = , (3.307)
S V,ni S P,ni

E A
P = = = , (3.308)
V S,ni V T,ni V T,i

H G
V = = , (3.309)
P S,ni P T,ni

A G
S = = = , (3.310)
T V,ni T P,ni T V,i

ni = . (3.311)
i V,S,j

E H A G
i = = = = (3.312)
ni S,V,nj ni S,P,nj ni T,V,nj ni T,P,nj
Each of these induces a corresponding Maxwell relation, obtained by cross differentiation.
These are

T P
= , (3.313)
V S,ni S V,ni

T V
= , (3.314)
P S,ni S P,ni

P S
= , (3.315)
T V,ni V T,ni

V S
= , (3.316)
T P,ni P T,ni

i S
= , (3.317)
T P,nj ni V,nj

i V
= , (3.318)
P T,nj ni V,nj

l k
= , (3.319)
nk T,P,nj nl T,P,nj

S P
= , (3.320)
V T,i T V,i

ni nk
= (3.321)
k V,S,j ,j6=k i V,S,j ,j6=i
3.9. PARTIAL MOLAR PROPERTIES 119
3.9 Partial molar properties

3.9.1 Homogeneous functions
In mathematics, a homogeneous function f (x1 , . . . , xN ) of order m is one such that
f (x1 , . . . , xN ) = m f (x1 , . . . , xN ). (3.322)
If m = 1, one has
f (x1 , . . . , xN ) = f (x1 , . . . , xN ). (3.323)
Thermodynamic variables are examples of homogeneous functions.
3.9.2 Gibbs free energy

Consider an extensive property, such as the Gibbs free energy G. One has the canonical
form
G = G(T, P, n1 , n2 , . . . , nN ). (3.324)
One would like to show that if each of the mole numbers ni is increased by a common factor,
say , with T and P constant, that G increases by the same factor :
G(T, P, n1 , n2 , . . . , nN ) = G(T, P, n1 , n2 , . . . , nN ). (3.325)
Differentiate both sides of Eq. (3.325) with respect to , while holding P , T , and nj constant,
to get
G(T, P, n1 , n2 , . . . , nN ) =

G d(n1 ) G d(n2 ) G d(nN )
+ + ...+ , (3.326)
(n1 ) nj ,P,T d (n2 ) nj ,P,T d (nN ) nj ,P,T d

G G G
= n1 + n2 + . . . + nN , (3.327)
(n1 ) nj ,P,T (n2 ) nj ,P,T (nN ) nj ,P,T
This must hold for all , including = 1, so one requires

G G G
G(T, P, n1, n2 , . . . , nN ) = n1 + n2 + . . . + nN ,
n1 nj ,P,T n2 nj ,P,T nN nj ,P,T
(3.328)
N
X G
= ni . (3.329)
i=1
ni nj ,P,T
Recall now the definition partial molar property, the derivative of an extensive variable with
respect to species ni holding nj , i 6= j, T , and P constant. Because the result has units per
mole, an overline superscript is utilized. The partial molar Gibbs free energy of species i, g i
is then
G
gi , (3.330)
ni nj ,P,T
so that
N
X
G= g i ni . (3.331)
i=1
Using the definition of chemical potential, Eq. (3.312), one also notes then that
N
X
G(T, P, n2 , n2 , . . . , nN ) = i ni . (3.332)
i=1
The temperature and pressure dependence of G must lie entirely within i (T, P, nj ), which
one notes is also allowed to be a function of nj as well. Consequently, one also sees that the
Gibbs free energy per unit mole of species i is the chemical potential of that species:
g i = i . (3.333)
Using Eq. (3.331) to eliminate G in Eq. (3.301), one recovers an equation for the energy:
N
X
E = P V + T S + i ni . (3.334)
i=1
3.9.3 Other properties

A similar result also holds for any other extensive property such as V , E, H, A, or S. One
can also show that
N
X V
V = ni , (3.335)
i=1
ni nj ,A,T
XN
E
E = ni (3.336)
i=1
ni nj ,V,S
N
X H
H = ni , (3.337)
i=1
ni nj ,P,S
N
X A
A = ni , (3.338)
i=1
ni nj ,T,V
N
X S
S = ni . (3.339)
i=1
ni n j ,E,T
3.9. PARTIAL MOLAR PROPERTIES 121
Note that these expressions do not formally involve partial molar properties since P and T
are not constant.
Take now the appropriate partial molar derivatives of G for an ideal mixture of ideal
gases to get some useful relations:
G = H T S, (3.340)

G H S
= T . (3.341)
ni T,P,nj ni T,P,nj ni T,P,nj
Now from the definition of an ideal mixture hi = hi (T, P ), so one has

N
X
H = nk hk (T, P ), (3.342)
k=1
N
!
H X
= nk hk (T, P ) , (3.343)
ni T,P,nj ni k=1
N
X nk
= hk (T, P ), (3.344)
ni
k=1 |{z}
=ik
N
X
= ik hk (T, P ), (3.345)
k=1
= hi (T, P ). (3.346)
Here, the Kronecker delta function ki has been again used. Now for an ideal gas one further
has hi = hi (T ). The analysis is more complicated for the entropy, in which
X N
o Pk
S = nk sT,k R ln , (3.347)
Po
k=1
X N
o yk P
= nk sT,k R ln , (3.348)
k=1
P o
N !!
X P n k
= nk soT,k R ln R ln PN , (3.349)
k=1
P o q=1 nq
N N
!
X P X nk
o
= nk sT,k R ln R nk ln PN , (3.350)
k=1
Po k=1 q=1 nq
N
S X o P
= nk sT,k R ln
ni T,P,nj ni k=1 Po
N
!!
X nk
R nk ln PN , (3.351)
ni T,P,nj k=1 q=1 nq
N
X
nk P
= soT,k R ln
n Po
k=1 |{z}i
=ik
N
!!
X nk
R nk ln PN , (3.352)
ni T,P,nj q=1 n q
k=1
N
!!
P X nk
o
= sT,i R ln R nk ln PN , (3.353)
Po ni T,P,nj k=1 q=1 nq
(3.354)
Evaluation of the final term on the right side requires closer examination, and in fact, after
tedious but straightforward analysis, yields a simple result which can easily be verified by
direct calculation:
N
!! !
X nk ni
nk ln PN = ln PN . (3.355)
ni T,P,nj k=1 q=1 nq q=1 nq
So the partial molar entropy is in fact

!
S P ni
= soT,i R ln R ln PN , (3.356)
ni T,P,nj Po q=1 nq

o P
= sT,i R ln R ln yi , (3.357)
Po

o Pi
= sT,i R ln , (3.358)
Po
= si (3.359)
Thus, one can in fact claim for the ideal mixture of ideal gases that
g i = hi T si . (3.360)
3.9.4 Relation between mixture and partial molar properties

A simple analysis shows how the partial molar property for an individual species is related
to the partial molar property for the mixture. Consider, for example, the Gibbs free energy.
The mixture-averaged Gibbs free energy per unit mole is
G
g= . (3.361)
n
Now take a partial molar derivative and analyze to get

g 1 G G n
= , (3.362)
ni T,P,nj n ni T,P,nj n2 ni T,P,nj
3.10. FROZEN SOUND SPEED 123
N
!
1 G G X
= 2 nk , (3.363)
n ni T,P,nj n ni T,P,nj k=1
N
1 G G X nk
= 2 , (3.364)
n ni T,P,nj n ni T,P,nj
k=1
N
1 G GX
= 2 ik , (3.365)
n ni T,P,nj n k=1

1 G G
= , (3.366)
n ni T,P,nj n2
1 g
= gi . (3.367)
n n
Multiplying by n and rearranging, one gets

g
gi = g + n . (3.368)
ni T,P,nj
A similar result holds for other properties.
3.10 Frozen sound speed

Let us develop a relation for what is known as the frozen sound speed, denoted as c. This
quantity is a thermodynamic property defined by the relationship

2 P
c . (3.369)
s,ni
Later we shall see that this quantity describes the speed of acoustic waves, but at this point
let us simply treat it as a problem in thermodynamics. The quantity is called frozen
because the derivative is performed with all of the species mole numbers frozen at a constant
value. Note that since v = 1/,

2 P P dv 1 P
2 P
c = = = 2 = v . (3.370)
s,ni v s,ni d v s,ni v s,ni
So we can say

2 P
(c) = . (3.371)
v s,ni
Now we know that the caloric equation of state for a mixture of ideal gases takes the
form, from Eq. (2.162):
N
X
e= Yiei (T ). (3.372)
i=1
Employing the ideal gas law, we could say instead

N
X
P
e= Yi ei . (3.373)
i=1
R
In terms of number of moles, ni , we could say

N
X
Mi ni P
e= ei (3.374)
V R
i=1 | {z }
=Yi
We can in fact generalize, and simply consider caloric equations of state of the form
e = e(P, , ni ), (3.375)
or in terms of the specific volume v = 1/,
e = e(P, v, ni). (3.376)
These last two forms can be useful when temperature is unavailable, as can be the case for
some materials.
Now reconsider the Gibbs equation for a multicomponent system, Eq. (3.293
N
X
dE = P dV + T dS + i dni . (3.377)
i=1
Let us scale Eq. (3.377) by a fixed mass m, recalling that e = E/m, v = V /m, and s = S/m,
so as to obtain
XN
i
de = P dv + T ds + dni . (3.378)
i=1
m
Considering Eq. (3.378) with dv = 0 and dni = 0, and scaling by dP , we can say

e s
=T . (3.379)
P v,ni P v,ni
Considering Eq. (3.378) with dni = 0, and dP = 0 and scaling by dv, we can say

e s
= P + T . (3.380)
v P,ni v P,ni
Applying now Eq. (3.59), we find

P s v
= 1. (3.381)
v s,ni P v,ni s P,ni
3.11. IRREVERSIBILITY IN A CLOSED MULTICOMPONENT SYSTEM 125
So we get

s
P v P,ni
= . (3.382)
v s,ni s
P v,ni
Substituting Eq. (3.382) into Eq. (3.371) we find

s
v P,n
(c)2 = s i . (3.383)
P v,ni
Substituting Eqs. (3.379,3.380) into Eq. (3.383) we get

1 e
T
P + v P,ni
(c)2 = 1 e
, (3.384)
T P v,ni

e
P + v
= P,ni .
e
(3.385)
P v,ni
So the sound speed c is

v
u e
u P + v
c = v t e
P,ni
. (3.386)

P v,ni
Eq. (3.386) is useful because we can identify the frozen sound speed from data for e, P , and
v alone. If

e e
> P, and > 0, (3.387)
v P,ni P v,ni
we will have a real sound speed. Other combinations are possible as well, but not observed
in nature.
3.11 Irreversibility in a closed multicomponent system

Consider a thermodynamic system closed to mass exchanges with its surroundings coming
into equilibrium. Allow the system to be exchanging work and heat with its surroundings.
Assume the temperature difference between the system and its surroundings is so small that
both can be considered to be at temperature T . If dQ is introduced into the system, then
the surroundings suffer a loss of entropy:
dQ
dSsurr = . (3.388)
T
The systems entropy S can change via this heat transfer, as well as via other internal
irreversible processes, such as internal chemical reaction. The second law of thermodynamics
requires that the entropy change of the universe be positive semi-definite:
dS + dSsurr 0. (3.389)
Eliminating dSsurr , one requires for the system that
dQ
dS . (3.390)
T
Consider temporarily the assumption that the work and heat transfer are both reversible.
Thus, the irreversibilities must be associated with internal chemical reaction. Now the first
law for the entire system gives
dE = dQ dW, (3.391)
= dQ P dV, (3.392)
dQ = dE + P dV. (3.393)
Note because the system is closed, there can be no species entering or exiting, and so there
is no change dE attributable to dni . While within the system the dni may not be 0, the net
contribution to the change in total internal energy is zero. A non-zero dni within the system
simply re-partitions a fixed amount of total energy from one species to another. Substituting
Eq. (3.393) into Eq. (3.390) to eliminate dQ, one gets
1
dS (dE + P dV ), (3.394)
T| {z }
=dQ
T dS dE P dV 0, (3.395)
dE T dS + P dV 0. (3.396)
Eq. (3.396) involves properties only and need not require assumptions of reversibility for
processes in its derivation. In special cases, it reduces to simpler forms.
For processes which are isentropic and isochoric, the second law expression, Eq. (3.396),
reduces to
dE|S,V 0. (3.397)
For processes which are isoenergetic and isochoric, the second law expression, Eq. (3.396),
reduces to
dS|E,V 0. (3.398)
Now using Eq. (3.293) to eliminate dS in Eq. (3.398), one can express the second law as

!
1 P 1X
N

dE + dV i dni 0, (3.399)
T T T i=1
| {z
}
=dS E,V
3.11. IRREVERSIBILITY IN A CLOSED MULTICOMPONENT SYSTEM 127
N
1X
dni 0. (3.400)
T i=1 i
| {z }
irreversibility
The irreversibility associated with the internal chemical reaction must be the left side of
Eq. (3.400).
Now, while most standard texts focusing on equilibrium thermodynamics go to great
lengths to avoid the introduction of time, it really belongs in a discussion describing the
approach to equilibrium. One can divide Eq. (3.400) by a positive time increment dt to get
N
1 X dni
0. (3.401)
T i=1 i dt
Since T 0, one can multiply Eq. (3.401) by T to get

N
X dni
i 0. (3.402)
i=1
dt
This will hold if a model for dn

dt
i
is employed which guarantees that the left side of Eq. (3.402)
is negative semi-definite. One will expect then for dn
dt
i
to be related to the chemical potentials
i .
Elimination of dE in Eq. (3.396) in favor of dH from dH = dE + P dV + V dP gives
{z V dP} T dS + P dV
|dH P dV 0, (3.403)
=dE
dH V dP T dS 0. (3.404)
Thus, one finds for isobaric, isentropic equilibration that
dH|P,S 0. (3.405)
For the Helmholtz and Gibbs free energies, one analogously finds
dA|T,V 0, (3.406)
dG|T,P 0. (3.407)
The expression of the second law in terms of dG is especially useful as it may be easy in an
experiment to control so that P and T are constant. This is especially true in an isobaric
phase change, in which the temperature is guaranteed to be constant as well.
Now one has
N
X
G = ni g i , (3.408)
i=1
XN
= ni i . (3.409)
i=1
P
One also has from Eq. (3.305): dG = SdT + V dP + N i=1 i dni , holding T and P constant
that
N
X
dG|T,P = i dni . (3.410)
i=1
Here the dni are associated entirely with internal chemical reactions. Substituting Eq. (3.410)
into Eq. (3.407), one gets the important version of the second law which holds that
N
X
dG|T,P = i dni 0. (3.411)
i=1
In terms of time rates of change, one can divide Eq. (3.411) by a positive time increment
dt > 0 to get
N
G X dni
= i 0. (3.412)
t T,P i=1
dt
3.12 Equilibrium in a two-component system

A major task of non-equilibrium thermodynamics is to find a functional form for dn dt
i
which
guarantees satisfaction of the second law, Eq. (3.412) and gives predictions which agree with
experiment. This will be discussed in more detail in the following chapter on thermochem-
istry. At this point, some simple examples will be given in which a nave but useful functional
form for dndt
i
is posed which leads at least to predictions of the correct equilibrium values. A
much better model which gives the correct dynamics in the time domain of the system as it
approaches equilibrium will be presented in the chapter on thermochemistry.
3.12.1 Phase equilibrium

Here, consider two examples describing systems in phase equilibrium.
Example 3.12
Consider an equilibrium two-phase mixture of liquid and vapor H2 O at T = 100 C, x = 0.5. Use
the steam tables to check if equilibrium properties are satisfied.
In a two-phase gas liquid mixture one can expect the following reaction:
H2 O(l) H2 O(g) . (3.413)
That is one mole of liquid, in the forward phase change, evaporates to form one mole of gas. In the
reverse phase change, one mole of gas condenses to form one mole of liquid.
Because T is fixed at 100 C and the material is a two phase mixture, the pressure is also fixed at a
constant. Here there are two phases at saturation; g for gas and l for liquid. Equation (3.411) reduces
to
l dnl + g dng 0. (3.414)
3.12. EQUILIBRIUM IN A TWO-COMPONENT SYSTEM 129
Now for the pure H2 O if a loss of moles from one phase must be compensated by the addition to
another. So one must have
dnl + dng = 0. (3.415)
Hence
dng = dnl . (3.416)
So Eq. (3.414), using Eq. (3.416) becomes
l dnl g dnl 0, (3.417)

dnl (l g ) 0. (3.418)
At this stage of the analysis, most texts, grounded in equilibrium thermodynamics, assert that l = g ,
ignoring the fact that they could be different but dnl could be zero. That approach will not be taken
here. Instead divide Eq. (3.418) by a positive time increment, dt 0 to write the second law as
dnl
( g ) 0. (3.419)
dt l
One convenient, albeit nave, way to guarantee second law satisfaction is to let
dnl
= (l g ), 0, convenient, but nave model (3.420)
dt
Here is some positive semi-definite scalar rate constant which dictates the time scale of approach to
equilibrium. Note that Eq. (3.420) is just a hypothesized model. It has no experimental verification; in
fact, other more complex models exist which both agree with experiment and satisfy the second law.
For the purposes of the present argument, however, Eq. (3.420) will suffice. With this assumption, the
second law reduces to
(l g )2 0, 0, (3.421)
which is always true.
Eq. (3.420) has two important consequences:
Differences in chemical potential drive changes in the number of moles.
The number of moles of liquid, nl , increases when the chemical potential of the liquid is less than that
of the gas, l < g . That is to say, when the liquid has a lower chemical potential than the gas, the
gas is driven towards the phase with the lower potential. Because such a phase change is isobaric and
isothermal, the Gibbs free energy is the appropriate variable to consider, and one takes = g. When
this is so, the Gibbs free energy of the mixture, G = nl l + ng g is being driven to a lower value. So
when dG = 0, the system has a minimum G.
The system is in equilibrium when the chemical potentials of liquid and gas are equal: l = g .
The chemical potentials, and hence the molar specific Gibbs free energies must be the same for
each constituent of the binary mixture at the phase equilibrium. That is
gl = gg . (3.422)
Now since both the liquid and gas have the same molecular mass, one also has the mass specific Gibbs
free energies equal at phase equilibrium:
gl = gg . (3.423)
This can be verified from the steam tables, using the definition g = h T s. From the tables

kJ kJ kJ
gl = hl T sl = 419.02 ((100 + 273.15) K) 1.3068 = 68.6 , (3.424)
kg kg K kg

kJ kJ kJ
gg = hg T sg = 2676.05 ((100 + 273.15) K) 7.3548 = 68.4 . (3.425)
kg kg K kg
The two values are essentially the same; the difference is likely due to table inaccuracies.
3.12.2 Chemical equilibrium: introduction

Here consider two examples which identify the equilibrium state of a chemically reactive
system.
3.12.2.1 Isothermal, isochoric system

The simplest system to consider is isothermal and isochoric. The isochoric assumption
implies there is no work in coming to equilibrium.
Example 3.13
At high temperatures, collisions between diatomic nitrogen molecules induce the production of
monatomic nitrogen molecules. The chemical reaction can be described by the model
N2 + N2 2N + N2 . (3.426)
Here one of the N2 molecules behaves as an inert third body. An N2 molecule has to collide with
something, to induce the reaction. Some authors leave out the third body and write instead N2 2N ,
but this does not reflect the true physics as well. The inert third body is especially important when
the time scales of reaction are considered. It plays no role in equilibrium chemistry.
Consider 1 kmole of N2 and 0 kmole of N at a pressure of 100 kP a and a temperature of 6000 K.
Using the ideal gas tables, find the equilibrium concentrations of N and N2 if the equilibration process
is isothermal and isochoric.
The ideal gas law can give the volume.
nN2 RT
P1 = , (3.427)
V
nN2 RT
V = , (3.428)
P1
kJ

(1 kmole) 8.314 kmole K (6000 K)
= , (3.429)
100 kP a
= 498.84 m3 . (3.430)
Initially, the mixture is all N2 , so its partial pressure is the total pressure, and the initial partial pressure
of N is 0.
Now every time an N2 molecule reacts and thus undergo a negative change, 2 N molecules are
created and thus undergo a positive change, so
1
dnN2 = dnN . (3.431)
2
This can be parameterized by a reaction progress variable , also called the degree of reaction, defined
such that
d = dnN2 , (3.432)
1
d = dnN . (3.433)
2
As an aside, one can integrate this, taking = 0 at the initial state to get
= nN2 |t=0 nN2 , (3.434)

1
= nN . (3.435)
2
Thus
nN 2 = nN2 |t=0 , (3.436)

nN = 2. (3.437)
One can also eliminate to get nN in terms of nN2 :
nN = 2 ( nN2 |t=0 nN2 ) . (3.438)
Now for the reaction, one must have, for second law satisfaction, that
N2 dnN2 + N dnN 0, (3.439)

N2 (d) + N (2d) 0, (3.440)

N2 + 2N d 0 (3.441)
d
N2 + 2N 0. (3.442)
dt
In order to satisfy the second law, one can usefully, but navely, hypothesize that the non-equilibrium
reaction kinetics are given by
d
= k(N2 + 2N ), k 0, convenient, but nave model (3.443)
dt
Note there are other ways to guarantee second law satisfaction. In fact, a more complicated model is
well known to fit data well, and will be studied later. For the present purposes, this nave model will
suffice. With this assumption, the second law reduces to
k(N2 + 2N )2 0, k 0, (3.444)
d
which is always true. Obviously, the reaction ceases when dt = 0, which holds only when
2N = N2 . (3.445)
d
Away from equilibrium, for the reaction to go forward, one must expect dt > 0, and then one must
have
N2 + 2N 0, (3.446)
2N N2 . (3.447)
The chemical potentials are the molar specific Gibbs free energies; thus, for the reaction to go forward,
one must have
2gN g N2 . (3.448)
Substituting using the definitions of Gibbs free energy, one gets

2 h N T sN h N 2 T sN 2 , (3.449)

yN P yN2 P
2 hN T soT,N R ln hN2 T soT,N2 R ln , (3.450)
Po Po

yN P yN2 P
2 hN T soT,N hN2 T soT,N2 2RT ln + RT ln , (3.451)
Po Po

yN P yN2 P
2 hN T soT,N + hN2 T soT,N2 2RT ln RT ln , (3.452)
Po Po
2 2
yN P Po
2 hN T soT,N + hN2 T soT,N2 RT ln , (3.453)
Po2 P yN2
2
yN P
2 hN T soT,N + hN2 T soT,N2 RT ln . (3.454)
yN2 Po
At the initial state, one has yN = 0, so the right hand side approaches , and the inequality holds.
At equilibrium, one has equality.
2
yN P
2 hN T soT,N + hN2 T soT,N2 = RT ln . (3.455)
yN2 Po
Taking numerical values from Table A.9:

kJ kJ
2 5.9727 105 (6000 K) 216.926 +
kmole kg K

kJ kJ
2.05848 105 (6000 K) 292.984
kmole kg K
2
kJ yN P
= 8.314 (6000 K) ln ,
kmole K yN2 Po
2
yN P
2.87635 = ln , (3.456)
| {z } yN2 Po
ln KP
2
yN P
0.0563399
| {z } = , (3.457)
yN2 Po
KP
2
nN
nN +nN2 RT
= (nN + nN2 ) , (3.458)
nN2 Po V
nN +nN2
n2N RT
= , (3.459)
nN2 Po V
2
(2 ( nN2 |t=0 nN2 )) RT
= , (3.460)
nN 2 Po V
2
(2 (1 kmole nN2 )) (8.314)(6000)
= . (3.461)
nN 2 (100)(498.84)
This is a quadratic equation for nN2 . It has two roots
nN2 = 0.888154 kmole physical; nN2 = 1.12593 kmole, non-physical (3.462)
The second root generates more N2 than at the start, and also yields non-physically negative nN =
0.25186 kmole. So at equilibrium, the physical root is
nN = 2(1 nN2 ) = 2(1 0.888154) = 0.223693 kmole. (3.463)
The diatomic species is preferred.
Note in the preceding analysis, the term KP was introduced. This is the so-called equilibrium
constant which is really a function of temperature. It will be described in more detail later, but one
notes that it is commonly tabulated for some reactions. Its tabular value can be derived from the more
fundamental quantities shown in this example. BSs Table A.11 gives for this reaction at 6000 K the
value of ln KP = 2.876. Note that KP is fundamentally defined in terms of thermodynamic properties
for a system which may or may not be at chemical equilibrium. Only at chemical equilibrium, can it
can further be related to mole fraction and pressure ratios.
The pressure at equilibrium is
(nN2 + nN )RT
P2 = , (3.464)
V
kJ
(0.888154 kmole + 0.223693 kmole) 8.314 kmole K (6000 K)
= , (3.465)
498.84
= 111.185 kP a. (3.466)
The pressure has increased because there are more molecules with the volume and temperature being
equal.
The molar concentrations i at equilibrium, are
0.223693 kmole kmole mole
N = 3
= 4.484 104 3
= 4.484 107 , (3.467)
498.84 m m cm3
N 2 = = 1.78044 103 = 1.78044 106 . (3.468)
498.84 m3 m3 cm3
Now consider the heat transfer. One knows for the isochoric process that Q = U2 U1 . The initial
energy is given by
U1 = nN2 uN2 , (3.469)
= nN2 (hN2 RT ), (3.470)

kJ kJ
= (1 kmole) 2.05848 105 8.314 (6000 K) , (3.471)
kmole kmole K
= 1.555964 105 kJ. (3.472)
The energy at the final state is
U2 = nN 2 u N 2 + nN u N , (3.473)
= nN2 (hN2 RT ) + nN (hN RT ), (3.474)

kJ kJ
= (0.888154 kmole) 2.05848 105 8.314 (6000 K) (3.475)
kmole kmole K

kJ kJ
+(0.223693 kmole) 5.9727 105 8.314 (6000 K) , (3.476)
kmole kmole K
= 2.60966 105 kJ. (3.477)
So
Q = U2 U1 , (3.478)
= 2.60966 105 kJ 1.555964 105 kJ, (3.479)
= 1.05002 105 kJ. (3.480)
Heat needed to be added to keep the system at the constant temperature. This is because the nitrogen
dissociation process is endothermic.
One can check for second law satisfaction in two ways. Fundamentally, one can demand that
Eq. (3.390), dS dQ/T , be satisfied for the process, giving
Z 2
dQ
S2 S1 . (3.481)
1 T
For this isothermal process, this reduces to
Q
S2 S1 , (3.482)
T
(nN2 sN2 + nN sN )|2
Q
(nN2 sN2 + nN sN )|1 , (3.483)
T
yN2 P yN P
o
nN2 sT,N2 R ln o
+ nN sT,N R ln
Po Po
2
yN2 P yN P Q
nN2 soT,N2 R ln + nN soT,N R ln , (3.484)
Po Po T
1
PN2 PN
nN2 soT,N2 R ln + nN soT,N R ln
Po Po
2
PN2 PN Q
nN2 soT,N2 R ln + nN soT,N R ln , (3.485)
Po Po T
1

nN2 RT nN RT
o
nN2 sT,N2 R ln o
+ nN sT,N R ln
Po V Po V
2

n N RT n N RT Q
o
nN2 sT,N2 R ln 2 o
+ nN sT,N R ln . (3.486)
Po V |{z} Po V T

=0 1
Now at the initial state nN = 0 kmole, and RT /Po /V has a constant value of
kJ

RT 8.314 kmole K (6000 K) 1
= =1 , (3.487)
Po V (100 kP a)(498.84 m3 ) kmole
so Eq. (3.486) reduces to
nN2 n
N
nN2 soT,N2 R ln + nN soT,N R ln
1 kmole 1 kmole 2
n Q
o N2
nN2 sT,N2 R ln ,
1 kmole 1 T
(3.488)
((0.888143) (292.984 8.314 ln (0.88143)) + (0.223714) (216.926 8.314 ln (0.223714)))|2
105002
((1) (292.984 8.314 ln (1)))|1 ,
6000
kJ kJ
19.4181 17.5004 . (3.489)
K K
Indeed, the second law is satisfied. Moreover the irreversibility of the chemical reaction is 19.4181
17.5004 = +1.91772 kJ/K.
For the isochoric, isothermal process, it is also appropriate to use Eq. (3.406), dA|T,V 0, to check
for second law satisfaction. This turns out to give an identical result. Since by Eq. (3.300), A = U T S,
A2 A1 = (U2 T2 S2 ) (U1 T1 S1 ). Since the process is isothermal, A2 A1 = U2 U1 T (S2 S1 ).

For A2 A1 0, one must demand U2 U1 T (S2 S1 ) 0, or U2 U1 T (S2 S1 ), or
S2 S1 (U2 U1 )/T . Since Q = U2 U1 for this isochoric process, one then recovers S2 S1 Q/T.
3.12.2.2 Isothermal, isobaric system

Allowing for isobaric rather than isochoric equilibration introduces small variation in the
analysis.
Example 3.14
Consider the same reaction
N2 + N2 2N + N2 . (3.490)
for an isobaric and isothermal process. That is, consider 1 kmole of N2 and 0 kmole of N at a pressure
of 100 kP a and a temperature of 6000 K. Using the ideal gas tables, find the equilibrium concentrations
of N and N2 if the equilibration process is isothermal and isobaric.
The initial volume is the same as from the previous example:
V1 = 498.84 m3 . (3.491)
Note the volume will change in this isobaric process. Initially, the mixture is all N2 , so its partial
pressure is the total pressure, and the initial partial pressure of N is 0.
A few other key results are identical to the previous example:
nN = 2 ( nN2 |t=0 nN2 ) , (3.492)
and
2gN g N2 . (3.493)
Substituting using the definitions of Gibbs free energy, one gets

2 h N T sN h N 2 T sN 2 , (3.494)

yN P yN2 P
2 hN T soT,N R ln hN2 T soT,N2 R ln , (3.495)
Po Po

yN P yN2 P
2 hN T soT,N hN2 T soT,N2 2RT ln + RT ln , (3.496)
Po Po

yN P yN2 P
2 hN T soT,N + hN2 T soT,N2 2RT ln RT ln , (3.497)
Po Po
2 2
yN P Po
2 hN T soT,N + hN2 T soT,N2 RT ln . (3.498)
Po2 P yN2
In this case Po = P , so one gets

2

yN
2 hN T soT,N + hN 2 T soT,N2 RT ln . (3.499)
yN2
At the initial state, one has yN = 0, so the right hand side approaches , and the inequality holds.
At equilibrium, one has equality.
2
yN
2 hN T soT,N + hN2 T soT,N2 = RT ln . (3.500)
yN2
Taking numerical values from Table A.9:

kJ kJ
2 5.9727 105 (6000 K) 216.926 +
kmole kg K

kJ kJ
2.05848 105 (6000 K) 292.984
kmole kg K
2
kJ yN
= 8.314 (6000 K) ln ,
kmole K yN2
2
yN
2.87635 = ln , (3.501)
| {z } yN2
ln KP
2
yN
0.0563399
| {z } = , (3.502)
yN2
KP
2
nN
nN +nN2
= , (3.503)
nN2
nN +nN2
n2N
= , (3.504)
nN2 (nN + nN2 )
2
(2 ( nN2 |t=0 nN2 ))
= , (3.505)
nN2 (2 ( nN2 |t=0 nN2 ) + nN2 )
2
(2 (1 kmole nN2 ))
= . (3.506)
nN2 (2 (1 kmole nN2 ) + nN2 )
This is a quadratic equation for nN2 . It has two roots
nN2 = 0.882147 kmole physical; nN2 = 1.11785 kmole, non-physical (3.507)
The second root generates more N2 than at the start, and also yields non-physically negative nN =
0.235706 kmole. So at equilibrium, the physical root is
nN = 2(1 nN2 ) = 2(1 0.882147) = 0.235706 kmole. (3.508)
Again, the diatomic species is preferred. As the temperature is raised, one could show that the
monatomic species would come to dominate.
The volume at equilibrium is
(nN2 + nN )RT
V2 = , (3.509)
P
kJ
(0.882147 kmole + 0.235706 kmole) 8.314 kmole K (6000 K)
= , (3.510)
100 kP a
= 557.630 m3 . (3.511)
The volume has increased because there are more molecules with the pressure and temperature being
equal.
The molar concentrations i at equilibrium, are

N = 3
= 4.227 104 3
= 4.227 107 , (3.512)
557.636 m m cm3
N 2 = = 1.58196 103 = 1.58196 106 . (3.513)
557.636 m3 m3 cm3
The molar concentrations are a little smaller than for the isochoric case, mainly because the volume is
larger at equilibrium.
Now consider the heat transfer. One knows for the isobaric process that Q = H2 H1 . The initial
enthalpy is given by

kJ
H1 = nN2 hN2 = (1 kmole) 2.05848 105 = 2.05848 105 kJ. (3.514)
kmole
The enthalpy at the final state is
H2 = nN2 hN2 + nN hN , (3.515)

kJ kJ
= (0.882147 kmole) 2.05848 105 + (0.235706 kmole) 5.9727 105 (3.516)
,
kmole kmole
= 3.22368 105 kJ. (3.517)
So
Q = H2 H1 = 3.22389 105 kJ 2.05848 105 kJ = 1.16520 105 kJ. (3.518)
Heat needed to be added to keep the system at the constant temperature. This is because the nitrogen
dissociation process is endothermic. Relative to the isochoric process, more heat had to be added to
maintain the temperature. This to counter the cooling effect of the expansion.
Lastly, it is a straightforward exercise to show that the second law is satisfied for this process.
3.12.3 Equilibrium condition

The results of both of the previous examples, in which a functional form of a progress
variables time variation, d
dt
, was postulated in order to satisfy the second law gave a condition
for equilibrium. This can be generalized so as to require at equilibrium that
N
X
i i = 0. (3.519)
i=1
| {z }

Here i represents the net number of moles of species i generated in the forward reaction.
This negation of the term on the left side of Eq (3.519) is sometimes defined as the chemical
affinity, :
N
X
i i . (3.520)
i=1
So in the phase equilibrium example, Eq. (3.519) becomes
l (1) + g (1) = 0. (3.521)
In the nitrogen chemistry example, Eq. (3.519) becomes
N2 (1) + N (2) = 0. (3.522)
This will be discussed in detail in the following chapter.

Chapter 4
Thermochemistry of a single reaction
This chapter will further develop notions associated with the thermodynamics of chemical
reactions. The focus will be on single chemical reactions. Several sources can be consulted
as references. 1 2 3 4 5
4.1 Molecular mass

The molecular mass of a molecule is a straightforward notion from chemistry. One simply
sums the product of the number of atoms and each atoms atomic mass to form the molecular
mass. If one defines L as the number of elements present in species i, li as the number of
moles of atomic element l in species i, and Ml as the atomic mass of element l, the molecular
mass Mi of species i
L
X
Mi = Ml li. (4.1)
l=1
In vector form, one would say
MT = MT , or M = T M. (4.2)
Here M is the vector of length N containing the molecular masses, M is the vector of length
L containing the elemental masses, and is the matrix of dimension L N containing the
number of moles of each element in each species. Generally, is full rank. If N > L, has
rank L. If N < L, has rank N. In any problem involving an actual chemical reaction, one
1
Edition, John Wiley, New York. See Chapters 14 and 15.
2
M. M. Abbott and H. C. van Ness, 1972, Thermodynamics, Schaums Outline Series in Engineering,
McGraw-Hill, New York. See Chapter 7.
3
D. Kondepudi and I. Prigogene, 1998, Modern Thermodynamics: From Heat Engines to Dissipative
Structures, John Wiley, New York. See Chapters 4, 5, 7, and 9.
4
S. R. Turns, 2000, An Introduction to Combustion, McGraw-Hill, Boston. See Chapter 2.
5
K. K. Kuo, 2005, Principles of Combustion, Second Edition, John Wiley, New York. See Chapters 1, 2.
139
140 CHAPTER 4. THERMOCHEMISTRY OF A SINGLE REACTION
will find N L, and in most cases N > L. In isolated problems not involving a reaction,
one may have N < L. In any case, M lies in the column space of T , which is the row space
of .
Example 4.1
Find the molecular mass of H2 O. Here, one has two elements H and O, so L = 2, and one species,
so N = 1; thus, in this isolated problem, N < L. Take i = 1 for species H2 O. Take l = 1 for
element H. Take l = 2 for element O. From the periodic table, one gets M1 = 1 kg/kmole for H,
M2 = 16 kg/kmole for O. For element 1, there are 2 atoms, so 11 = 2. For element 2, there is 1 atom
so 21 = 1. So the molecular mass of species 1, H2 O is

11
M1 = ( M1 M2 ) , (4.3)
21
= M1 11 + M2 21 , (4.4)

kg kg
= 1 (2) + 16 (1), (4.5)
kmole kmole
kg
= 18 . (4.6)
kmole
Example 4.2
Find the molecular masses of the two species C8 H18 and CO2 . Here, for practice, the vector matrix
notation is exercised. In a certain sense this is overkill, but it is useful to be able to understand a
general notation.
One has N = 2 species, and takes i = 1 for C8 H18 and i = 2 for CO2 . One also has L = 3 elements
and takes l = 1 for C, l = 2 for H, and l = 3 for O. Now for each element, one has M1 = 12 kg/kmole,
M2 = 1 kg/kmole, M3 = 16 kg/kmole. The molecular masses are then given by

11 12
( M1 M2 ) = ( M1 M2 M3 ) 21 22 , (4.7)
31 32

8 1
= ( 12 1 16 ) 18 0 , (4.8)
0 2
= ( 114 44 ) . (4.9)
That is for C8 H18 , one has molecular mass M1 = 114 kg/kmole. For CO2 , one has molecular mass
M2 = 44 kg/kmole.
4.2. STOICHIOMETRY 141
4.2 Stoichiometry
4.2.1 General development
Stoichiometry represents a mass balance on each element in a chemical reaction. For example,
in the simple global reaction
2H2 + O2 2H2 O, (4.10)
one has 4 H atoms in both the reactant and product sides and 2 O atoms in both the reactant
and product sides. In this section stoichiometry will be systematized.
Consider now a general reaction with N species. This reaction can be represented by
N
X N
X
i i i i . (4.11)
i=1 i=1
Here i is the ith chemical species, i is the forward stoichiometric coefficient of the ith
reaction, and i is the reverse stoichiometric coefficient of the ith reaction. Both i and i
are to be interpreted as pure dimensionless numbers.
In Equation (4.10), one has N = 3 species. One might take 1 = H2 , 2 = O2 , and
3 = H2 O. The reaction is written in more general form as
(2)1 + (1)2 + (0)3 (0)1 + (0)2 + (2)3 , (4.12)
(2)H2 + (1)O2 + (0)H2 O (0)H2 + (0)O2 + (2)H2 O. (4.13)
Here, one has
1 = 2, 1 = 0, (4.14)
2 = 1, 2 = 0, (4.15)
3 = 0, 3 = 2. (4.16)
It is common and useful to define another pure dimensionless number, the net stoichio-
metric coefficients for species i, i . Here i represents the net production of number if the
reaction goes forward. It is given by
i = i i . (4.17)
For the reaction 2H2 + O2 2H2 O, one has
1 = 1 1 = 0 2 = 2, (4.18)
2 = 2 2 = 0 1 = 1, (4.19)
3 = 3 3 = 2 0 = 2. (4.20)
With these definitions, it is possible to summarize a chemical reaction as
N
X
i i = 0. (4.21)
i=1
In vector notation, one would say

T = 0. (4.22)
For the reaction of this section, one might write the non-traditional form
2H2 O2 + 2H2 O = 0. (4.23)
It remains to enforce a stoichiometric balance. This is achieved if, for each element, l =
1, . . . , L, one has the following equality:
N
X
li i = 0, l = 1, . . . , L. (4.24)
i=1
That is to say, for each element, the sum of the product of the net species production and
the number of elements in the species must be zero. In vector notation, this becomes
= 0. (4.25)
One may recall from linear algebra that this demands that lie in the right null space of .
Example 4.3
Show stoichiometric balance is achieved for 2H2 O2 + 2H2 O = 0. Here again, the number of
elements L = 2, and one can take l = 1 for H and l = 2 for O. Also the number of species N = 3, and
one takes i = 1 for H2 , i = 2 for O2 , and i = 3 for H2 O. Then for element 1, H, in species 1, H2 , one
has
11 = 2, H in H2 . (4.26)
Similarly, one gets
12 = 0, H in O2 , (4.27)
13 = 2, H in H2 O, (4.28)
21 = 0, O in H2 , (4.29)
22 = 2, O in O2 , (4.30)
23 = 1, O in H2 O. (4.31)
PN
In matrix form then, i=1 li i = 0 gives

1
2 0 2 2 = 0 . (4.32)
0 2 1 0
3
This is two equations in three unknowns. Thus, it is formally underconstrained. Certainly the trivial
solution 1 = 2 = 3 = 0 will satisfy, but one seeks non-trivial solutions. Assume 3 has a known value
3 = . Then, the system reduces to

2 0 1 2
= . (4.33)
0 2 2
The inversion here is easy, and one finds 1 = , 2 = 21 . Or in vector form,

1
2 = 1 , (4.34)
2
3

1
= 12 , R1 (4.35)
1
Again, this amounts to saying the solution vector (1 , 2 , 3 )T lies in the right null space of the coefficient
matrix li .
Here is any real scalar. If one takes = 2, one gets

1 2
2 = 1 , (4.36)
3 2
This simply corresponds to

2H2 O2 + 2H2 O = 0. (4.37)
If one takes = 4, one still achieves stoichiometric balance, with

1 4
2 = 2 , (4.38)
3 4
which corresponds to the equally valid
4H2 + 2O2 4H2 O = 0. (4.39)
In summary, the stoichiometric coefficients are non-unique but partially constrained by mass conserva-
tion. Which set is chosen is to some extent arbitrary, and often based on traditional conventions from
chemistry. But others are equally valid.
There is a small issue with units here, which will be seen to be difficult to reconcile.
In practice, it will have little to no importance. In the previous example, one might be
tempted to ascribe units of kmoles to i . Later, it will be seen that in classical reaction
kinetics, i is best interpreted as a pure dimensionless number, consistent with the definition
of this section. So in the context of the previous example, one would then take to be
dimensionless as well, which is perfectly acceptable for the example. In later problems, it
will be more useful to give the units of kmoles. Note that multiplication of by any scalar,
e.g. kmole/(6.02 1026 ), still yields an acceptable result.
Example 4.4
Balance an equation for hypothesized ethane combustion
1 C2 H6 + 2 O2 3 CO2 + 4 H2 O. (4.40)
One could also say in terms of the net stoichiometric coefficients
1 C2 H6 + 2 O2 + 3 CO2 + 4 H2 O = 0. (4.41)
Here one takes 1 = C2 H6 , 2 = O2 , 3 CO2 , 4 = H2 O. So there are N = 4 species. There are also
L = 3 elements: l = 1 : C, l = 2 : H, l = 3 : O. One then has
11 = 2, C in C2 H6 , (4.42)
12 = 0, C in O2 , (4.43)
13 = 1, C in CO2 , (4.44)
14 = 0, C in H2 O, (4.45)
21 = 6, H in C2 H6 , (4.46)
22 = 0, H in O2 , (4.47)
23 = 0, H in CO2 , (4.48)
24 = 2, H in H2 O, (4.49)
31 = 0, O in C2 H6 , (4.50)
32 = 2, O in O2 , (4.51)
33 = 2, O in CO2 , (4.52)
34 = 1, O in H2 O, (4.53)
(4.54)
PN
So the stoichiometry equation, i=1 li i = 0, is given by
1
2 0 1 0 0

6 0 0 2 2 = 0. (4.55)
3
0 2 2 1 0
4
Here, there are three equations in four unknowns, so the system is underconstrained. There are many
ways to address this problem. Here, choose the robust way of casting the system into row echelon form.
This is easily achieved by Gaussian elimination. Row echelon form seeks to have lots of zeros in the
lower left part of the matrix. The lower left corner has a zero already, so that is useful. Now, multiply
the top equation by 3 and subtract the result from the second to get
1
2 0 1 0 0
0 0 3 2 2
= 0. (4.56)
3
0 2 2 1 0
4
Next switch the last two equations to get
1
2 0 1 0 0

0 2 2 1 2 = 0. (4.57)
3
0 0 3 2 0
4
Now divide the first by 2, the second by 2 and the third by 3 to get unity in the diagonal:
1
1 0 12 0 0
0 1 1 1 2 0.
2 = (4.58)
3
0 0 1 23 0
4
So-called basic variables have non-zero coefficients on the diagonal, so one can take the basic variables
to be 1 , 2 , and 3 . The remaining variables are free variables. Here one takes the free variable to be
4 . So set 4 = , and rewrite the system as

1 0 21 1 0
0 1 1 2 = 1 . (4.59)
2
2
0 0 1 3 3
Solving, one finds

1 1
1 3 3
2 76 76
= 2 = 2 , R1 . (4.60)
3 3 3
4 1
Again one finds a non-unique solution in the right null space of . If one chooses = 6, then one gets

1 2
2 7
= , (4.61)
3 4
4 6
which corresponds to the stoichiometrically balanced reaction
2C2 H6 + 7O2 4CO2 + 6H2 O. (4.62)
In this example, is dimensionless.
Example 4.5
Consider stoichiometric balance for a propane oxidation reaction which may produce carbon monox-
ide and hydroxyl in addition to carbon dioxide and water.
The hypothesized reaction takes the form
1 C3 H8 + 2 O2 3 CO2 + 4 CO + 5 H2 O + 6 OH. (4.63)
In terms of net stoichiometric coefficients, this becomes
1 C3 H8 + 2 O2 + 3 CO2 + 4 CO + 5 H2 O + 6 OH = 0. (4.64)
There are N = 6 species and L = 3 elements. One then has
11 = 3, C in C3 H8 , (4.65)
12 = 0, C in O2 , (4.66)
13 = 1, C in CO2 , (4.67)
14 = 1, C in CO, (4.68)
15 = 0, C in H2 O, (4.69)
16 = 0, C in OH, (4.70)
21 = 8, H in C3 H8 , (4.71)
22 = 0, H in O2 , (4.72)
23 = 0, H in CO2 , (4.73)
24 = 0, H in CO, (4.74)
25 = 2, H in H2 O, (4.75)
26 = 1, H in OH, (4.76)
31 = 0, O in C3 H8 , (4.77)
32 = 2, O in O2 , (4.78)
33 = 2, O in CO2 , (4.79)
34 = 1, O in CO, (4.80)
35 = 1, O in H2 O, (4.81)
36 = 1, O in OH. (4.82)
The equation = 0, then becomes

1
2
3 0 1 1 0 0 0

8 0 0 0 2 1 3 = 0. (4.83)
4
0 2 2 1 1 1 0
5
6
Multiplying the first equation by 38 and adding it to the second gives

1
2
3 0 1 1 0 0 0
0 0 8 8 2 1 3
3 3 = 0 . (4.84)
4
0 2 2 1 1 1 0
5
6
Trading the second and third rows gives

1
2
3 0 1 1 0 0 0

0 2 2 1 1 1 3 = 0. (4.85)
4
0 0 83 38 2 1 0
5
6
Dividing the first row by 3, the second by 2 and the third by 38 gives

1
1 1
2
1 0 3 3 0 0 0
0 1 1 1 1 1 3 0.
2 2 2 = (4.86)
4
0 0 1 1 43 38 0
5
6
Take basic variables to be 1 , 2 , and 3 and free variables to be 4 , 5 , and 6 . So set 4 = 1 , 5 = 2 ,
and 6 = 3 , and get

1 0 1
3 1 31
0 1 1 2 = 21 22 23 . (4.87)
0 0 1 3 1 + 43 2 + 38 3
Solving, one finds

1

1 8 (22 3 )
2 1 .
= 8 (41 102 73 ) (4.88)
1
3 8 (8 1 + 62 + 33 )
For all the coefficients, one then has
1
1 8 (22 3 ) 0 2 1
1
2 8 (41 102 73 ) 4 10 7
1

3 (8 1 + 62 + 3 )
3 1 8 2 6 3 3
= 8 = + + . (4.89)
4 1 8 8 8 0 8 0

5 2 0 8 0
6 3 0 0 8
Here, one finds three independent vectors in the right null space. To simplify the notation, take 1 = 81 ,
2 = 82 , and 3 = 83 . Then,

1 0 2 1
2 4 10 7

3 8 6 3
= 1 + 2 + 3 . (4.90)
4 8 0 0

5 0 8 0
6 0 0 8
The most general reaction that can achieve a stoichiometric balance is given by
(22 3 )C3 H8 + (41 102 73 )O2 + (81 + 62 + 33 )CO2 + 81 CO + 82 H2 O + 83 OH = 0. (4.91)
Rearranging, one gets
(22 + 3 )C3 H8 + (41 + 102 + 73 )O2 (81 + 62 + 33 )CO2 + 81 CO + 82 H2 O + 83 OH. (4.92)
This will be balanced for all 1 , 2 , and 3 . The values that are actually achieved in practice depend
on the thermodynamics of the problem. Stoichiometry only provides some limitations.
A slightly more familiar form is found by taking 2 = 1/2 and rearranging, giving
(1 + 3 ) C3 H8 + (5 41 + 73 ) O2 (3 81 + 33 ) CO2 + 4 H2 O + 81 CO + 83 OH. (4.93)

One notes that often the production of CO and OH will be small. If there is no production of CO or
OH, 1 = 3 = 0 and one recovers the familiar balance of
C3 H8 + 5 O2 3 CO2 + 4 H2 O. (4.94)
One also notes that stoichiometry alone admits unusual solutions. For instance, if 1 = 100 2 = 1/2,
and 3 = 1, one has
2 C3 H8 + 794 CO2 388 O2 + 4 H2 O + 800 CO + 8 OH. (4.95)
This reaction is certainly admitted by stoichiometry but is not observed in nature. To determine
precisely which of the P infinitely many possible final states are realized requires a consideration of the
N
equilibrium condition i=1 i i = 0.
Looked at in another way, we can think of three independent classes of reactions admitted by the
stoichiometry, one for each of the linearly independent null space vectors. Taking first 1 = 1/4, 2 = 0,
3 = 0, one gets, after rearrangement
2CO + O2 2CO2 , (4.96)
as one class of reaction admitted by stoichiometry. Taking next 1 = 0, 2 = 1/2, 3 = 0, one gets
C3 H8 + 5O2 3CO2 + 4H2 O, (4.97)
as the second class admitted by stoichiometry. The third class is given by taking 1 = 0, 2 = 0, 3 = 1,
and is
C3 H8 + 7O2 3CO2 + 8OH. (4.98)
In this example, both and are dimensionless.
In general, one can expect to find the stoichiometric coefficients for N species composed
of L elements to be of the following form:
N
X L
i = Dik k , i = 1, . . . , N. (4.99)
k=1
Here Dik is a dimensionless component of a full rank matrix of dimension N (N L)

and rank N L, and k is a dimensionless component of a vector of parameters of length
N L. The matrix whose components are Dik are constructed by populating its columns
with vectors which lie in the right null space of li . Note that multiplication of k by any
constant gives another set of i , and mass conservation for each element is still satisfied.
4.2.2 Fuel-air mixtures

In combustion with air, one often models air as a simple mixture of diatomic oxygen and
inert diatomic nitrogen in the

air: air (O2 + 3.76N2 ). (4.100)
The air-fuel ratio, A and its reciprocal, the fuel-air ratio, F , can be defined on a mass
and mole basis.
mair nair
Amass = , Amole = . (4.101)
mf uel nf uel
Via the molecular masses, one has
mair nair Mair Mair
Amass = = = Amole . (4.102)
mf uel nf uel Mf uel Mf uel
If there is not enough air to burn all the fuel, the mixture is said to be rich. If there is
excess air, the mixture is said to be lean. If there is just enough, the mixture is said to be
stoichiometric. The equivalence ratio, , is defined as the actual fuel-air ratio scaled by the
stoichiometric fuel-air ratio:
Factual Astoichiometric
= (4.103)
Fstoichiometric Aactual
The ratio is the same whether F s are taken on a mass or mole basis, because the ratio of
molecular masses cancel.
Example 4.6
Calculate the stoichiometry of the combustion of methane with air with an equivalence ratio of
= 0.5. If the pressure is 0.1 M P a, find the dew point of the products.
First calculate the coefficients for stoichiometric combustion:
1 CH4 + 2 (O2 + 3.76N2 ) 3 CO2 + 4 H2 O + 5 N2 . (4.104)
Or
1 CH4 + 2 O2 + 3 CO2 + 4 H2 O + (5 + 3.762 )N2 = 0. (4.105)
Here one has N = 5 species and L = 4 elements. Adopting a slightly more intuitive procedure for
variety, one writes a conservation equation for each element to get
1 + 3 = 0, C (4.106)
41 + 24 = 0, H (4.107)
22 + 23 + 4 = 0, O (4.108)
3.762 + 5 = 0, N. (4.109)
In matrix form this becomes

1
1 0 1 0 0 0
2
4 0 0 2 0 0
3 = . (4.110)
0 2 2 1 0 0
4
0 3.76 0 0 1 0
5
Now, one might expect to have one free variable, since one has five unknowns in four equations.
While casting the equation in row echelon form is guaranteed to yield a proper solution, one can often
use intuition to get a solution more rapidly. One certainly expects that CH4 will need to be present
for the reaction to take place. One might also expect to find an answer if there is one mole of CH4 . So
take 1 = 1. Realize that one could also get a physically valid answer by assuming 1 to be equal to
any scalar. With 1 = 1, one gets

0 1 0 0 2 1
0 0 2 0 3 4
= . (4.111)
2 2 1 0 4 0
3.76 0 0 1 5 0
One easily finds the unique inverse does exist, and that the solution is

2 2
3 1
= . (4.112)
4 2
5 7.52
If there had been more than one free variable, the four by four matrix would have been singular, and
no unique inverse would have existed.
In any case, the reaction under stoichiometric conditions is
CH4 2O2 + CO2 + 2H2 O + (7.52 + (3.76)(2))N2 = 0, (4.113)

CH4 + 2(O2 + 3.76N2) CO2 + 2H2 O + 7.52N2. (4.114)
For the stoichiometric reaction, the fuel-air ratio on a mole basis is
1
Fstoichiometric = = 0.1050. (4.115)
2 + 2(3.76)
Now = 0.5, so
Factual = Fstoichiometric , (4.116)

= (0.5)(0.1050), (4.117)
= 0.0525. (4.118)
By inspection, one can write the actual reaction equation as
CH4 + 4(O2 + 3.76N2 ) CO2 + 2H2 O + 2O2 + 15.04N2. (4.119)
Check:
1
Factual = = 0.0525. (4.120)
4 + 4(3.76)
For the dew point of the products, one needs the partial pressure of H2 O. The mole fraction of
H2 O is
2
yH2 O = = 0.0499 (4.121)
1 + 2 + 2 + 15.04
So the partial pressure of H2 O is
PH2 O = yH2 O P = 0.0499(100 kP a) = 4.99 kP a. (4.122)
From the steam tables, the saturation temperature at this pressure is Tsat = Tdew point = 32.88 C. If
the products cool to this temperature in an exhaust device, the water could condense in the apparatus.
4.3 First law analysis of reacting systems

One can easily use the first law to learn much about chemically reacting systems.
4.3.1 Enthalpy of formation

The enthalpy of formation is the enthalpy that is required to form a molecule from combining
its constituents at P = 0.1 MP a and T = 298 K. Consider the reaction (taken here to be
irreversible)
C + O2 CO2 . (4.123)
4.3. FIRST LAW ANALYSIS OF REACTING SYSTEMS 151
In order to maintain the process at constant temperature, it is found that heat transfer to
the volume is necessary. For the steady constant pressure process, one has
E2 E1 = Q12 W12 , (4.124)

Z 2
= Q12 P dV, (4.125)
1
= Q12 P (V2 V1 ), (4.126)
Q12 = E2 E1 + P (V2 V1 ), (4.127)
= H2 H1 , (4.128)
Q12 = Hproducts Hreactants , (4.129)
X X
= ni hi ni hi . (4.130)
products reactants
In this reaction, one measures that Q12 = 393522 kJ for the reaction of 1 kmole of C
and O2 . That is the reaction liberates such energy to the environment. So measuring the
heat transfer can give a measure of the enthalpy difference between reactants and products.
Assign a value of enthalpy zero to elements in their standard state at the reference state.
kJ
Thus, C and O2 both have enthalpies of 0 kmole at T = 298 K, P = 0.1 MP a. This enthalpy
is designated, for species i,
o o
hf,i = hTo ,i , (4.131)
and is called the enthalpy of formation. So the energy balance for the products and reactants,
here both at the standard state, becomes
o o o
Q12 = nCO2 hf,CO2 nC hf,C nO2 hf,O2 , (4.132)

o kJ kJ
393522 kJ = (1 kmole)hf,CO2 (1 kmole) 0 (1 kmole) 0 .
kmole kmole
(4.133)
o kJ
Thus, the enthalpy of formation of CO2 is hf,CO2 = 393522 kmole , since the reaction involved
creation of 1 kmole of CO2 .
Often values of enthalpy are tabulated in the forms of enthalpy differences hi . These
are defined such that
o o
hi = hf,i + (hi hf,i ), (4.134)
| {z }
=hi
o
= hf,i + hi . (4.135)
Lastly, one notes for an ideal gas that the enthalpy is a function of temperature only,
and so does not depend on the reference pressure; hence
o o
hi = hi , hi = hi , if ideal gas. (4.136)
Example 4.7
(SBvW, p. 575). Consider the following irreversible reaction in a steady state, steady flow process
confined to the standard state of P = 0.1 M P a, T = 298 K:
CH4 + 2O2 CO2 + 2H2 O(). (4.137)
The first law holds that

X X
Qcv = ni hi ni hi . (4.138)
products reactants
All components are at their reference states. Table A.10 gives properties, and one finds
Qcv = nCO2 hCO2 + nH2 O hH2 O nCH4 hCH4 nO2 hO2 , (4.139)

kJ kJ
= (1 kmole) 393522 + (2 kmole) 285830
kmole kmole

kJ kJ
(1 kmole) 74873 (2 kmole) 0 , (4.140)
kmole kmole
= 890309 kJ. (4.141)
A more detailed analysis is required in the likely case in which the system is not at the
reference state.
Example 4.8
(adopted from Moran and Shapiro6 ) A mixture of 1 kmole of gaseous methane and 2 kmole of
oxygen initially at 298 K and 101.325 kP a burns completely in a closed, rigid, container. Heat transfer
occurs until the final temperature is 900 K. Find the heat transfer and the final pressure.
The combustion is stoichiometric. Assume that no small concentration species are generated. The
global reaction is given by
CH4 + 2O2 CO2 + 2H2 O. (4.142)
The first law analysis for the closed system is slightly different:
E2 E1 = Q12 W12 . (4.143)
Since the process is isochoric, W12 = 0. So
Q12 = E2 E1 , (4.144)
= nCO2 eCO2 + nH2 O eH2 O nCH4 eCH4 nO2 eO2 , (4.145)
= nCO2 (hCO2 RT2 ) + nH2 O (hH2 O RT2 ) nCH4 (hCH4 RT1 ) nO2 (hO2 RT1 ), (4.146)
= hCO2 + 2hH2 O hCH4 2hO2 3R(T2 T1 ), (4.147)
o o o o
= (hCO2 ,f + hCO2 ) + 2(hH2 O,f + hH2 O ) (hCH4 ,f + hCH4 ) 2(hO2 ,f + hO2 )
6
M. J. Moran and H. N. Shapiro, 2003, Fundamentals of Engineering Thermodynamics, Fifth Edition,
John Wiley, New York. p. 619.
3R(T2 T1 ), (4.148)
= (393522 + 28030) + 2(241826 + 21937) (74873 + 0) 2(0 + 0)
3(8.314)(900 298), (4.149)
= 745412 kJ. (4.150)
For the pressures, one has
P1 V1 = (nCH4 + nO2 )RT1 , (4.151)

(nCH4 + nO2 )RT1
V1 = , (4.152)
P1

kJ
(1 kmole + 2 kmole) 8.314 kg K (298 K)
= , (4.153)
101.325 kP a
= 73.36 m3 . (4.154)
Now V2 = V1 , so
(nCO2 + nH2 O )RT2

P2 = , (4.155)
V2

kJ
(1 kmole + 2 kmole) 8.314 kg K (900 K)
= , (4.156)
73.36 m3
= 306.0 kP a. (4.157)
The pressure increased in the reaction. This is entirely attributable to the temperature rise, as the
number of moles remained constant here.
4.3.2 Enthalpy and internal energy of combustion

The enthalpy of combustion is the difference between the enthalpy of products and reactants
when complete combustion occurs at a given pressure and temperature. It is also known
as the heating value or the heat of reaction. The internal energy of combustion is related
and is the difference between the internal energy of products and reactants when complete
combustion occurs at a given volume and temperature.
The term higher heating value refers to the energy of combustion when liquid water is in
the products. Lower heating value refers to the energy of combustion when water vapor is
in the product.
4.3.3 Adiabatic flame temperature in isochoric stoichiometric sys-

tems
The adiabatic flame temperature refers to the temperature which is achieved when a fuel and
oxidizer are combined with no loss of work or heat energy. Thus, it must occur in a closed,
insulated, fixed volume. It is generally the highest temperature that one can expect to
achieve in a combustion process. It generally requires an iterative solution. Of all mixtures,

stoichiometric mixtures will yield the highest adiabatic flame temperatures because there is
no need to heat the excess fuel or oxidizer.
Here four examples will be presented to illustrate the following points.
The adiabatic flame temperature can be well over 5000 K for seemingly ordinary
mixtures.
Dilution of the mixture with an inert diluent lowers the adiabatic flame temperature.
The same effect would happen in rich and lean mixtures.
Preheating the mixture, such as one might find in the compression stroke of an engine,
increases the adiabatic flame temperature.
Consideration of the presence of minor species lowers the adiabatic flame temperature.
4.3.3.1 Undiluted, cold mixture
Example 4.9
A closed, fixed, adiabatic volume contains a stoichiometric mixture of 2 kmole of H2 and 1 kmole
of O2 at 100 kP a and 298 K. Find the adiabatic flame temperature assuming the irreversible reaction
2H2 + O2 2H2 O. (4.158)
The volume is given by
(nH2 + nO2 )RT1

V = , (4.159)
P1
kJ

(2 kmole + 1 kmole) 8.314 kmole K (298 K)
= , (4.160)
100 kP a
= 74.33 m3 . (4.161)
The first law gives
E2 E1 = Q W, (4.162)
E2 E1 = 0, (4.163)
nH2 O eH2 O nH2 eH2 nO2 eO2 = 0, (4.164)
nH2 O (hH2 O RT2 ) nH2 (hH2 RT1 ) nO2 (hO2 RT1 ) = 0, (4.165)
2hH2 O 2 hH2 hO2 +R(2T2 + 3T1 ) = 0, (4.166)
|{z} |{z}
=0 =0
2hH2 O + (8.314) ((2) T2 + (3) (298)) = 0, (4.167)
hH2 O 8.314T2 + 3716.4 = 0, (4.168)
o
hf,H2 O + hH2 O 8.314T2 + 3716.4 = 0, (4.169)
241826 + hH2 O 8.314T2 + 3716.4 = 0, (4.170)
238110 + hH2 O 8.314T2 = 0. (4.171)
At this point, one begins an iteration process, guessing a value of T2 and an associated hH2 O . When
T2 is guessed at 5600 K, the left side becomes 6507.04. When T2 is guessed at 6000 K, the left side
becomes 14301.4. Interpolate then to arrive at
T2 = 5725 K. (4.172)
This is an extremely high temperature. At such temperatures, in fact, one can expect other species to
co-exist in the equilibrium state in large quantities. These may include H, OH, O, HO2 , and H2 O2 ,
among others.
The final pressure is given by
nH2 O RT2
P2 = , (4.173)
V
kJ
(2 kmole) 8.314 kmole K (5725 K)
= , (4.174)
74.33 m3
= 1280.71 kP a. (4.175)
The final concentration of H2 O is

2 kmole kmole
H2 O = 3
= 2.69 102 . (4.176)
74.33 m m3
4.3.3.2 Dilute, cold mixture
Example 4.10
Consider a variant on the previous example in which the mixture is diluted with an inert, taken
here to be N2 . A closed, fixed, adiabatic volume contains a stoichiometric mixture of 2 kmole of H2 ,
1 kmole of O2 , and 8 kmole of N2 at 100 kP a and 298 K. Find the adiabatic flame temperature and
the final pressure, assuming the irreversible reaction
2H2 + O2 + 8N2 2H2 O + 8N2 . (4.177)
(nH2 + nO2 + nN2 )RT1

V = , (4.178)
P1
kJ

(2 kmole + 1 kmole + 8 kmole) 8.314 kmole K (298 K)
= , (4.179)
100 kP a
= 272.533 m3 . (4.180)
The first law gives
E2 E1 = Q W,
E2 E1 = 0,
nH2 O eH2 O nH2 eH2 nO2 eO2 + nN2 (eN2 2 eN2 1 ) = 0,
nH2 O (hH2 O RT2 ) nH2 (hH2 RT1 ) nO2 (hO2 RT1 ) + nN2 ((hN2 2 RT2 ) (hN2 1 RT1 )) = 0,
2hH2 O 2 hH2 hO2 +R(10T2 + 11T1 ) + 8( hN2 2 hN2 1 ) = 0,
|{z} |{z} | {z } | {z }
=0 =0 =hN2 =0
2hH2 O + (8.314) (10T2 + (11)(298)) + 8hN2 2 = 0,

2hH2 O 83.14T2 + 27253.3 + 8hN2 2 = 0,
o
2hf,H2 O + 2hH2 O 83.14T2 + 27253.3 + 8hN2 2 = 0,
2(241826) + 2hH2 O 83.14T2 + 27253.3 + 8hN2 2 = 0,
456399 + 2hH2 O 83.14T2 + 8hN22 = 0,
T2 is guessed at 2000 K, the left side becomes 28006.7. When T2 is guessed at 2200 K, the left side
becomes 33895.3. Interpolate then to arrive at
T2 = 2090.5 K. (4.181)
The inert diluent significantly lowers the adiabatic flame temperature. This is because the N2 serves as
a heat sink for the energy of reaction. If the mixture were at non-stoichiometric conditions, the excess
species would also serve as a heat sink, and the adiabatic flame temperature would be lower than that
of the stoichiometric mixture.
(nH2 O + nN2 )RT2

P2 = , (4.182)
V
kJ
(2 kmole + 8 kmole) 8.314 kmole K (2090.5 K)
= , (4.183)
272.533 m3
= 637.74 kP a. (4.184)
The final concentrations of H2 O and N2 are

2 kmole kmole
H2 O = 3
= 7.34 103 , (4.185)
272.533 m m3
8 kmole kmole
N2 = 3
= 2.94 102 . (4.186)
272.533 m m3
4.3.3.3 Dilute, preheated mixture
Example 4.11
Consider a variant on the previous example in which the diluted mixture is preheated to 1000 K.
One can achieve this via an isentropic compression of the cold mixture, such as might occur in an engine.
To simplify the analysis here, the temperature of the mixture will be increased, while the pressure will
be maintained. A closed, fixed, adiabatic volume contains a stoichiometric mixture of 2 kmole of H2 ,
1 kmole of O2 , and 8 kmole of N2 at 100 kP a and 1000 K. Find the adiabatic flame temperature and
the final pressure, assuming the irreversible reaction
2H2 + O2 + 8N2 2H2 O + 8N2 . (4.187)
(nH2 + nO2 + nN2 )RT1

V = , (4.188)
P1
kJ

(2 kmole + 1 kmole + 8 kmole) 8.314 kmole K (1000 K)
= , (4.189)
100 kP a
3
= 914.54 m . (4.190)
The first law gives
E2 E1 = Q W,
E2 E1 = 0,
nH2 O eH2 O nH2 eH2 nO2 eO2 + nN2 (eN2 2 eN2 1 ) = 0,
nH2 O (hH2 O RT2 ) nH2 (hH2 RT1 ) nO2 (hO2 RT1 ) + nN2 ((hN2 2 RT2 ) (hN2 1 RT1 )) = 0,
2hH2 O 2hH2 hO2 + R(10T2 + 11T1 ) + 8(hN2 2 hN2 1 ) = 0,
2(241826 + hH2 O ) 2(20663) 22703 + (8.314) (10T2 + (11)(1000)) + 8hN2 2 8(21463) = 0,
2hH2 O 83.14T2 627931 + 8hN2 2 = 0,
T2 is guessed at 2600 K, the left side becomes 11351. When T2 is guessed at 2800 K, the left side
becomes 52787. Interpolate then to arrive at
T2 = 2635.4 K. (4.191)
The preheating raised the adiabatic flame temperature. Note that the preheating was by 1000 298 =
702 K. The new adiabatic flame temperature is only 2635.4 2090.5 = 544.9 K greater.
(nH2 O + nN2 )RT2

P2 = , (4.192)
V
kJ
(2 kmole + 8 kmole) 8.314 kmole K (2635.4 K)
= , (4.193)
914.54 m3
= 239.58 kP a. (4.194)
The final concentrations of H2 O and N2 are

2 kmole kmole
H2 O = = 2.19 103 , (4.195)
914.54 m3 m3
8 kmole kmole
N 2 = 3
= 8.75 103 . (4.196)
914.54 m m3
4.3.3.4 Dilute, preheated mixture with minor species
Example 4.12
Consider a variant on the previous example. Here allow for minor species to be present at equilib-
rium. A closed, fixed, adiabatic volume contains a stoichiometric mixture of 2 kmole of H2 , 1 kmole
of O2 , and 8 kmole of N2 at 100 kP a and 1000 K. Find the adiabatic flame temperature and the final
pressure, assuming reversible reactions. Here, the details of the analysis are postponed, but the result
is given which is the consequence of a calculation involving detailed reactions rates. One can also solve
an optimization problem to minimize the Gibbs free energy of a wide variety of products to get the
same answer. In this case, the equilibrium temperature and pressure are found to be
T = 2484.8 K, P = 227.89 kP a. (4.197)
Equilibrium species concentrations are found to be
kmole
minor product H2 = 1.3 104 , (4.198)
m3
kmole
minor product H = 1.9 105 , (4.199)
m3
kmole
minor product O = 5.7 106 , (4.200)
m3
kmole
minor product O2 = 3.6 105 , (4.201)
m3
kmole
minor product OH = 5.9 105 , (4.202)
m3
kmole
major product H2 O = 2.0 103 , (4.203)
m3
kmole
trace product HO2 = 1.1 108 , (4.204)
m3
kmole
trace product H2 O2 = 1.2 109 , (4.205)
m3
kmole
trace product N = 1.7 109 , (4.206)
m3
kmole
trace product N H = 3.7 1010 , (4.207)
m3
kmole
trace product N H2 = 1.5 1010 , (4.208)
m3
kmole
trace product N H3 = 3.1 1010 , (4.209)
m3
kmole
trace product N N H = 1.0 1010 , (4.210)
m3
kmole
minor product N O = 3.1 106 , (4.211)
m3
kmole
trace product N O2 = 5.3 109 , (4.212)
m3
kmole
trace product N 2 O = 2.6 109 , (4.213)
m3
kmole
trace product HN O = 1.7 109 , (4.214)
m3
kmole
major product N 2 = 8.7 103 . (4.215)
m3
4.4. CHEMICAL EQUILIBRIUM 159
Note that the concentrations of the major products went down when the minor species were considered.
The adiabatic flame temperature also went down by a significant amount: 2635 2484.8 = 150.2 K.
Some thermal energy was necessary to break the bonds which induce the presence of minor species.
4.4 Chemical equilibrium

Often reactions are not simply unidirectional, as alluded to in the previous example. The
reverse reaction, especially at high temperature, can be important.
Consider the four species reaction
1 1 + 2 2 3 3 + 4 4 (4.216)
In terms of the net stoichiometric coefficients, this becomes
1 1 + 2 2 + 3 3 + 4 4 = 0. (4.217)
One can define a variable , the reaction progress. Take the dimension of to be kmoles.
When t = 0, one takes = 0. Now as the reaction goes forward, one takes d > 0. And a
forward reaction will decrease the number of moles of 1 and 2 while increasing the number
of moles of 3 and 4 . This will occur in ratios dictated by the stoichiometric coefficients of
the problem:
dn1 = 1 d, (4.218)
dn2 = 2 d, (4.219)
dn3 = +3 d, (4.220)
dn4 = +4 d. (4.221)
Note that if ni is taken to have units of kmoles, i , and i are taken as dimensionless, then
must have units of kmoles. In terms of the net stoichiometric coefficients, one has
dn1 = 1 d, (4.222)
dn2 = 2 d, (4.223)
dn3 = 3 d, (4.224)
dn4 = 4 d. (4.225)
Again, for arguments sake, assume that at t = 0, one has
n1 |t=0 = n1o , (4.226)
n2 |t=0 = n2o , (4.227)
n3 |t=0 = n3o , (4.228)
n4 |t=0 = n4o . (4.229)
Then after integrating, one finds
n1 = 1 + n1o , (4.230)
n2 = 2 + n2o , (4.231)
n3 = 3 + n3o , (4.232)
n4 = 4 + n4o . (4.233)
One can also eliminate the parameter in a variety of fashions and parameterize the
reaction one of the species mole numbers. Choosing, for example, n1 as a parameter, one
gets
n1 n1o
= . (4.234)
1
Eliminating then one finds all other mole numbers in terms of n1 :
n1 n1o
n2 = 2 + n2o , (4.235)
1
n1 n1o
n3 = 3 + n3o , (4.236)
1
n1 n1o
n4 = 4 + n4o . (4.237)
1
Written another way, one has
n1 n1o n2 n2o n3 n3o n4 n4o
= = = = . (4.238)
1 2 3 4
P
For an N-species reaction, N i=1 i i = 0, one can generalize to say
dni = = i d, (4.239)
ni = i + nio , (4.240)
ni nio
= . (4.241)
i
Note that
dni
= i . (4.242)
d
Now, from the previous chapter, one manifestation of the second law is Eq. (3.411):
N
X
dG|T,P = i dni 0. (4.243)
i=1
Now, one can eliminate dni in Eq. (4.243) by use of Eq. (4.239) to get
N
X
dG|T,P = i i d 0, (4.244)
i=1
N
G X
= i i 0, (4.245)
T,P i=1
= 0. (4.246)
Then for the reaction to go forward, one must require that the affinity be positive:
0. (4.247)
One also knows from the previous chapter that the irreversibility takes the form of Eq. (3.400):
N
1X
dni 0, (4.248)
T i=1 i
N
1 X
d i i 0, (4.249)
T i=1
N
1 d X
i 0. (4.250)
T dt i=1 i
PN
In terms of the chemical affinity, = i=1 i i , Eq. (4.250) can be written as
1 d
0. (4.251)
T dt
Now one straightforward, albeit nave, way to guarantee positive semi-definiteness of the
irreversibility and thus satisfaction of the second law is to construct the chemical kinetic rate
equation so that
X N
d
= k i i = k, k 0, provisional, nave assumption (4.252)
dt i=1
This provisional assumption of convenience will be supplanted later by a form which agrees
well with experiment. Here k is a positive semi-definite scalar. In general, it is a function of
temperature, k = k(T ), so that reactions proceed rapidly at high temperature and slowly at
low temperature. Then certainly the reaction progress variable will cease to change when
the equilibrium condition
XN
i i = 0, (4.253)
i=1
is met. This is equivalent to requiring

= 0. (4.254)
Now, while Eq. (4.253) is the most compact form of the equilibrium condition, it is not
the most commonly used form. One can perform the following analysis to obtain the form
in most common usage. Start by equating the chemical potential with the Gibbs free energy
per unit mole for each species i: i = g i . Then employ the definition of Gibbs free energy
for an ideal gas, and carry out a set of operations:
N
X
g i i = 0, at equilibrium (4.255)
i=1
N
X
(hi T si )i = 0. at equilibrium (4.256)
i=1
For the ideal gas, one can substitute for hi (T ) and si (T, P ) and write the equilibrium con-
dition as

X N Z T Z T

o c P i ( T ) y i P
h + c ( ) dT T so +
T d R ln
T i = 0,
To ,i P i
To ,i
T P
i=1 T
| o {z } | To
{z }
o

=hT,i
o
=soT,i
| {z } | {z }
o
=hT,i =hT,i =sT,i
(4.257)
Now writing the equilibrium condition in terms of the enthalpies and entropies referred to
the standard pressure, one gets
XN
o o yi P
hT,i T sT,i R ln i = 0, (4.258)
i=1
Po
X N XN
o o yi P
hT,i T sT,i i = RT i ln , (4.259)
i=1 | {z } i=1
Po
=g oT,i =oT,i
N
X N
X i
yi P
g oT,i i = RT ln , (4.260)
Po
|i=1 {z } i=1
Go
X N
Go yi P i
= ln , (4.261)
RT i=1
P o
YN i !
yi P
= ln , (4.262)
i=1
Po
Y N
Go yi P i
exp = , (4.263)
RT Po
| {z } i=1
KP
N
Y
yi P i
KP = , (4.264)
i=1
Po
PNi=1 i Y n
P
KP = yii , (4.265)
Po i=1
N
Y i
Pi
KP = . at equilibrium (4.266)
i=1
Po
Here KP is what is known as the pressure-based equilibrium constant. It is dimensionless.

Despite its name, it is not a constant. It is defined in terms of thermodynamic properties,
and for the ideal gas is a function of T only:

Go
KP exp . generally valid (4.267)
RT
Only at equilibrium does the property KP also equal the product of the partial pressures
as in Eq. (4.266). The subscript P for pressure comes about because it is also related to
the product of the ratio of the partial pressure to the reference pressure raised to the net
stoichiometric coefficients. Also, the net change in Gibbs free energy of the reaction at the
reference pressure, Go , which is a function of T only, has been defined as
N
X
o
G g oT,i i . (4.268)
i=1
The term Go has units of kJ/kmole; it traditionally does not get an overbar. If Go > 0,
one has 0 < KP < 1, and reactants are favored over products. If Go < 0, one gets KP > 1,
and products are favored over reactants. One can also deduce that higher pressures P push
the equilibrium in such a fashion that fewer moles are present, all else being equal. One can
also define Go in terms of the chemical affinity, referred to the reference pressure, as
Go = o . (4.269)
One can also define another convenient thermodynamic property, which for an ideal gas
is a function of T alone, the equilibrium constant Kc :
PNi=1 i
Po Go
Kc exp . generally valid (4.270)
RT RT
This property is dimensional, and the
PN
units depend on the stoichiometry of the reaction.
3 i
The units of Kc will be (kmole/m ) i=1 .
The equilibrium condition, Eq. (4.266), is often written in terms of molar concentrations
and Kc . This can be achieved by the operations, valid only at an equilibrium state:
YN i
i RT
KP = , (4.271)
i=1
P o
PNi=1 i Y
N
Go RT
exp = i i , (4.272)
RT Po i=1
PNi=1 i o
N
Y
Po G
exp = i i , (4.273)
RT RT
| {z } i=1
Kc
N
Y
Kc = i i . at equilibrium (4.274)
i=1
One must be careful to distinguish between the general definition of Kc as given in Eq. (4.270),
and the fact that at equilibrium it is driven to also have the value of product of molar species
concentrations, raised to the appropriate stoichiometric power, as given in Eq. (4.274).
4.5 Chemical kinetics of a single isothermal reaction

In the same fashion in ordinary mechanics that an understanding of statics enables an under-
standing of dynamics, an understanding of chemical equilibrium is necessary to understand
to more challenging topic of chemical kinetics. Chemical kinetics describes the time-evolution
of systems which may have an initial state far from equilibrium; it typically describes the
path of such systems to an equilibrium state. Here gas phase kinetics of ideal gas mixtures
that obey Daltons law will be studied. Important topics such as catalysis and solid or liquid
reactions will not be considered.
Further, this section will be restricted to strictly isothermal systems. This simplifies the
analysis greatly. It is straightforward to extend the analysis of this system to non-isothermal
systems. One must then make further appeal to the energy equation to get an equation for
temperature evolution.
The general form for evolution of species is taken to be

d i i
= . (4.275)
dt
Multiplying both sides of Eq. (4.275) by molecular mass Mi and using the definition of mass
fraction Yi then gives the alternate form
dYi i Mi
= . (4.276)
dt
4.5.1 Isochoric systems

Consider the evolution of species concentration in a system which is isothermal, isochoric
and spatially homogeneous. The system is undergoing a single chemical reaction involving
4.5. CHEMICAL KINETICS OF A SINGLE ISOTHERMAL REACTION 165
N species of the familiar form

N
X
i i = 0. (4.277)
i=1
Because the density is constant for the isochoric system, Eq. (4.275) reduces to
di
= i . (4.278)
dt
Then, experiment, as well as a more fundamental molecular collision theory, shows that the
evolution of species concentration i is given by
i
z }| {
Y !
di E
N 1 Y k
N

= i aT exp k k 1 k , isochoric system (4.279)
dt RT K c
| {z } | k=1{z } | k=1 {z }
k(T ) forward reaction reverse reaction
| {z }
r
This relation actually holds for isochoric, non-isothermal systems as well, which will not be
considered in any detail here. Here some new variables are defined as follows:
a: a kinetic rate constant called the collision frequency factor. Its units will depend
on the actual reaction and could involve various combinations of length, time, and
temperature. It is constructed so
PN
that di /dt has units of kmole/m3 /s; this requires it

to have units of (kmole/m3 )(1 k=1 k ) /s/K .
: a dimensionless parameter whose value is set by experiments, sometimes combined

with guiding theory, to account for weak temperature dependency of reaction rates.
E: the activation energy. It has units of kJ/kmole, though others are often used, and is
fit by both experiment and fundamental theory to account for the strong temperature
dependency of reaction.
Note that in Eq. (4.279) that molar concentrations are raised to the k and k powers. As
it does not make sense to raise a physical quantity to a power with units, one traditionally
interprets the values of k , k , as well as k to be dimensionless pure numbers. They are
also interpreted in a standard fashion: the smallest integer values that actually correspond
to the underlying molecular collision which has been modeled. While stoichiometric balance
can be achieved by a variety of k values, the kinetic rates are linked to one particular set
which is defined by the community.
Equation (4.279) is written in such a way that the species concentration production rate
increases when
The net number of moles generated in the reaction, measured by i increases,
The temperature increases; here, the sensitivity may be very high, as one observes in
nature,
The species concentrations of species involved in the forward reaction increase; this
embodies the principle that the collision-based reaction rates are enhanced when there
are more molecules to collide,
The species concentrations of species involved in the reverse reaction decrease.
Here, three intermediate variables which are in common usage have been defined. First one
takes the reaction rate to be

Y !
E
N 1 Y k
N

r aT exp k k
1 k , (4.280)
RT Kc k=1
| {z } | {z }
k=1
| {z }

1 Y k
N
Y N
E
= aT exp
k k k (4.281)
RT Kc
| {z } k=1
| {z } | k=1 {z }
k(T ), Arrhenius rate
| forward reaction{z reverse reaction }
law of mass action
The reaction rate r has units of kmole/m3 /s.

The temperature-dependency of the reaction rate is embodied in k(T ) is defined by what
is known as an Arrhenius rate law:

E
k(T ) aT exp . (4.282)
RT
This equation was advocated by vant Hoff in 1884; in 1889 Arrhenius gave a physical justifi-
cation. The units of k(T ) actually depend on the reaction. This is a weakness of the
P
theory,
3 (1 N
k=1 k ) /s.
and precludes a clean non-dimensionalization. The units must be (kmole/m )
In terms of reaction progress, one can also take
1 d
r= . (4.283)
V dt
The factor of 1/V is necessary because r has units of molar concentration per time and
has units of kmoles. The over-riding importance of the temperature sensitivity is illustrated
as part of the next example. The remainder of the expression involving the products of the
species concentrations is the defining characteristic of systems which obey the law of mass
action. Though the history is complex, most attribute the law of mass action to Guldberg
and Waage in 1864. Last, the overall molar production rate of species i, is often written as
i , defined as
i i r. (4.284)
As i is considered to be dimensionless, the units of i must be kmole/m3 /s.
Example 4.13
Study the nitrogen dissociation problem considered in an earlier example in which at t = 0 s,
1 kmole of N2 exists at P = 100 kP a and T = 6000 K. Take as before the reaction to be isothermal
and isochoric. Consider again the elementary nitrogen dissociation reaction
N2 + N2 2N + N2 , (4.285)
which has kinetic rate parameters of
cm3 K 1.6
a = 7.0 1021 , (4.286)
mole s
= 1.6, (4.287)
cal
E = 224928.4 . (4.288)
mole
In SI units, this becomes
3 1.6
3
21 cm K 1m 1000 mole m3 K 1.6
a = 7.0 10 = 7.0 1018 , (4.289)
mole s 100 cm kmole kmole s

cal J kJ 1000 mole kJ
E = 224928.4 4.186 = 941550 . (4.290)
mole cal 1000 J kmole kmole
At the initial state, the material is all N2 , so PN2 = P = 100 kP a. The ideal gas law then gives at
t=0

P |t=0 = PN2 |t=0 = N2 t=0 RT, (4.291)
P |t=0
N2 t=0 = , (4.292)
RT
100 kP a
= kJ
, (4.293)
8.314 kmole K (6000 K)
kmole
= 2.00465 103 . (4.294)
m3
Thus, the volume, constant for all time in the isochoric process, is
nN2 |t=0 1 kmole
V = = = 4.9884 102 m3 . (4.295)
N2 t=0 2.00465 103 kmole
m3
Now the stoichiometry of the reaction is such that

1
dnN2 = dnN , (4.296)
2
1
(nN2 nN2 |t=0 ) = (nN nN |t=0 ), (4.297)
| {z } 2 | {z }
=1 kmole =0
nN = 2(1 kmole nN2 ), (4.298)

nN 1 kmole nN2
= 2 , (4.299)
V V V

1 kmole
N = 2 N 2 , (4.300)
4.9884 102 m3

3 kmole
= 2 2.00465 10 N 2 . (4.301)
m3
3
k(T) (m /kmole/s)
10000
0.0001
1. x 10-12
1. x 10-20
1. x 10-28
1. x 10-36 T (K)
1000 1500 2000 3000 5000 7000 10000
Figure 4.1: k(T ) for Nitrogen dissociation example.
Now the general equation for kinetics of a single reaction, Eq. (4.279), reduces for N2 molar con-
centration to

dN2 E
N
N 1 N2 N
= N2 aT exp (N2 ) (N )
2 1 ( ) (N ) (4.302)
dt RT Kc N2

Realizing that N 2
= 2, N = 0, N2 = 1, and N = 2, one gets

dN2 E 1 2N
= aT exp 2N2 1 (4.303)
dt RT K c N 2
| {z }
=k(T )
Examine the primary temperature dependency of the reaction

E
k(T ) = aT exp , (4.304)
RT
!
3 1.6 kJ
18 m K 1.6 941550 kmole
= 7.0 10 T exp kJ
, (4.305)
kmole s 8.314 kmole KT

7.0 1018 1.1325 105
= exp (4.306)
T 1.6 T
Figure 4.1 gives a plot of k(T ) which shows its very strong dependency on temperature. For this
problem, T = 6000 K, so

7.0 1018 1.1325 105
k(6000) = exp , (4.307)
60001.6 6000
m3
= 40071.6 . (4.308)
kmole s
Now, the equilibrium constant Kc is needed. Recall
PN i=1 i

Po Go
Kc = exp (4.309)
RT RT
f ( N ) (kmole/m 3 /s)
2
unstable
equilibrium
1
0.0005 0.001 0.0015 0.002 0.0025
(kmole/m 3 )
N2
-1
unstable stable,
equilibrium physical
-2 equilibrium
Figure 4.2: Forcing function, f (N2 ), which drives changes of N2 as a function of N2 in

isothermal, isochoric problem.
PN
For this system, since i=1 i = 1, this reduces to

Po (2goN g oN2 )
Kc = exp , (4.310)
RT RT
o o !
Po (2(hN T soT,N ) (hN2 T soT,N2 ))
= exp , (4.311)
RT RT

100 (2(597270 (6000)216.926) (205848 (6000)292.984))
= exp ,(4.312)
(8.314)(6000) (8.314)(6000)
kmole
= 0.000112112 . (4.313)
m3
The differential equation for N2 evolution is then given by
kmole
2 !
dN2 m3 1 2 2.00465 103 N2
= 40071.6 2N2 1 m3
,
dt kmole 0.000112112 kmole
m3
N 2
| {z }
f (N2 )
(4.314)
= f (N2 ). (4.315)
The system is at equilibrium when f (N2 ) = 0. This is an algebraic function of N2 only, and can be
plotted. Figure 4.2 gives a plot of f (N2 ) and shows that it has three potential equilibrium points. It
is seen there are three roots. Solving for the equilibria requires solving
2 !
m3 2 1 2 2.00465 103 kmole
m3 N 2
0 = 40071.6 N 2 1 .
kmole 0.000112112 kmole
m3
N 2
(4.316)
The three roots are

kmole kmole kmole
N 2 = 0 , 0.00178121 , 0.00225611 . (4.317)
m3 }
| {z | {z m3 } | {z m3 }
unstable stable unstable
By inspection of the topology of Fig. 4.2, the only stable root is 0.00178121 kmole
m3 . This root agrees with
the equilibrium value found in an earlier example for the same problem conditions. Small perturbations
from this equilibrium induce the forcing function to supply dynamics which restore the system to its orig-
inal equilibrium state. Small perturbations from the unstable equilibria induce non-restoring dynamics.
For this root, one can then determine that the stable equilibrium value of N = 0.000446882 kmole m3 .
One can examine this stability more formally. Define an equilibrium concentration eq N2 such that
f (eq
N2 ) = 0. (4.318)
Now perform a Taylor series of f (N2 ) about N2 = eq

N2 :

df 1 d2 f
f (N2 ) f (eq ) + (N2 eq
N2 ) + ( eq 2
N2 ) + . . . (4.319)
} dN2 N 2 d2N2 N2
N2
| {z =eq
2 N 2
=0
Now the first term of the Taylor series is zero by construction. Next neglect all higher order terms as
small so that the approximation becomes

df
f (N2 ) ( eqN2 ) (4.320)
dN2 =eq N2
N2 N2
Thus, near equilibrium, one can write

dN2 df
(N2 eq
N2 ) (4.321)
dt dN2 eq
N2 =N2
Since the derivative of a constant is zero, one can also write the equation as

d df
(N2 eq ) ( eqN2 ) (4.322)
dt N2
dN2 =eq N2
N2 N2
This has a solution, valid near the equilibrium point, of

(N2 eq
N2 ) = C exp
df t , (4.323)
dN2 =eq
N2 N2

N2 = eq N2 + C exp
df t , (4.324)
dN2 =eq
N2 N2
(4.325)
Here C is some constant whose value is not important for this discussion. If the slope of f is positive,
that is,
df
> 0, unstable, (4.326)
dN2 =eq
N2 N2
concentration (kmole/m 3 ) 0.002

N2
0.0015
0.001
0.0005
N
t (s)
0.001 0.002 0.003 0.004 0.005
Figure 4.3: N2 (t) and N (t) in isothermal, isochoric nitrogen dissociation problem.
the equilibrium will be unstable. That is a perturbation will grow without bound as t . If the
slope is zero,
df
= 0, neutrally stable, (4.327)
dN2 =eq
N2 N2
the solution is stable in that there is no unbounded growth, and moreover is known as neutrally stable.
If the slope is negative,

df
< 0, asymptotically stable, (4.328)
dN2 =eq
N2 N2
the solution is stable in that there is no unbounded growth, and moreover is known as asymptotically
stable.
A numerical solution via an explicit technique such as a Runge-Kutta integration is found for
Eq. (4.314). The solution for N2 , along with N is plotted in Fig. 4.3.
Linearization of Eq. (4.314) about the equilibrium state gives rise to the locally linearly valid
d
N2 0.00178121 = 1209.39(N2 0.00178121) + . . . (4.329)
dt
This has local asymptotically stable solution
N2 = 0.00178121 + C exp (1209.39t) . (4.330)
Here C is some integration constant whose value is irrelevant for this analysis. The time scale of
relaxation is the time when the argument of the exponential is 1, which is
1
= = 8.27 104 s. (4.331)
1209.39 s1
One usually finds this time scale to have high sensitivity to temperature, with high temperatures giving
fast time constants and thus fast reactions.
The equilibrium values agree exactly with those found in the earlier example. Here the kinetics
provide the details of how much time it takes to achieve equilibrium. This is one of the key questions
of non-equilibrium thermodynamics.
4.5.2 Isobaric systems

The form of the previous section is the most important as it is easily extended to a Cartesian
grid with fixed volume elements in fluid flow problems. However, there is another important
spatially homogeneous problem in which the formulation needs slight modification: isobaric
reaction, with P equal to a constant. Again, in this section only isothermal conditions will
be considered.
In an isobaric problem, there can be volume change. Consider first the problem of isobaric
expansion of an inert mixture. In such a mixture, the total number of moles of each species
must be constant, so one gets
dni
= 0. inert, isobaric mixture (4.332)
dt
Now carry out the sequence of operations, realizing the the total mass m is also constant:
1 d
(ni ) = 0, (4.333)
m dt
d ni
= 0, (4.334)
dt m

d ni V
= 0, (4.335)
dt V m

d i
= 0, (4.336)
dt
1 di d
2i = 0, (4.337)
dt dt
di i d
= . (4.338)
dt dt
So a global density decrease of the inert material due to volume increase of a fixed mass
system induces a concentration decrease of each species. Extended to a material with a
single reaction rate r, one could say either
di d
= i r + i , or (4.339)
dt dt

d i 1
= i r, generally valid, (4.340)
dt
i
= . (4.341)

Equation (4.340) is consistent with Eq. (4.275) and is actually valid for general systems with
variable density, temperature, and pressure.
However, in this section, it is required that pressure and temperature be constant. Now
differentiate the isobaric, isothermal, ideal gas law to get the density derivative.
N
X
P = i RT, (4.342)
i=1
XN
di
0 = RT , (4.343)
i=1
dt
XN
di
0 = , (4.344)
i=1
dt
XN
i d
0 = i r + , (4.345)
i=1
dt
N
X N
1 d X
0 = r i + i , (4.346)
i=1
dt i=1
PN
d r i=1 i
= PN i . (4.347)
dt i=1
P
r N i
= PNi=1 , (4.348)
i=1 i
PN
r i=1 i
= P
, (4.349)
RT
PN
RT r i=1 i
= , (4.350)
P
PN
RT r k=1 k
= . (4.351)
P
PN
Note that if there is no net number change in the reaction, k=1 k = 0, the isobaric,
isothermal reaction also guarantees there would be no density or volume change. It is
convenient to define the net number change in the elementary reaction as n:
N
X
n k . (4.352)
k=1
Here n is taken to be a dimensionless pure number. It is associated with the number

change in the elementary reaction and not the actual mole change in a physical system; it
is, however, proportional to the actual mole change.
Now use Eq. (4.351) to eliminate the density derivative in Eq. (4.339) to get
P
di i RT r Nk=1 k
= i r , (4.353)
dt P

N

i RT X
= r i k , (4.354)
|{z}
reaction effect P
| {zk=1 }
expansion effect

= r i yin . (4.355)
|{z} | {z }
reaction effect expansion effect
There are two terms dictating the rate change of species molar concentration. The first, a
reaction effect, is precisely the same term that drove the isochoric reaction. The second is
due to the fact that the volume can change if the number of moles change, and this induces
an intrinsic change in concentration. Note that the term i RT /P = yi, the mole fraction.
Example 4.14
Study a variant of the nitrogen dissociation problem considered in an earlier example in which at
t = 0 s, 1 kmole of N2 exists at P = 100 kP a and T = 6000 K. In this case, take the reaction to be
isothermal and isobaric. Consider again the elementary nitrogen dissociation reaction
N2 + N2 2N + N2 , (4.356)
which has kinetic rate parameters of
cm3 K 1.6
a = 7.0 1021 , (4.357)
mole s
= 1.6, (4.358)
cal
E = 224928.4 . (4.359)
mole
In SI units, this becomes
3 1.6
3
21 cm K 1m 1000 mole m3 K 1.6
a = 7.0 10 = 7.0 1018 , (4.360)
mole s 100 cm kmole kmole s

cal J kJ 1000 mole kJ
E = 224928.4 4.186 = 941550 . (4.361)
mole cal 1000 J kmole kmole
At the initial state, the material is all N2 , so PN2 = P = 100 kP a. The ideal gas law then gives at
t=0
P = PN2 = N2 RT, (4.362)
P
N 2 = , (4.363)
t=0
RT
100 kP a
= kJ
, (4.364)
8.314 kmole K (6000 K)
kmole
= 2.00465 103 . (4.365)
m3
Thus, the initial volume is
nN2 |t=0 1 kmole
V |t=0 = = kmole
= 4.9884 102 m3 . (4.366)
i |t=0 2.00465 103 m3
In this isobaric process, one always has P = 100 kP a. Now, in general
P = RT (N2 + N ), (4.367)
therefore one can write N in terms of N2 :

P
N = N 2 , (4.368)
RT
100 kP a
= kJ
N2 , (4.369)
8.314 kmole K (6000 K)

kmole
= 2.00465 103 N 2 . (4.370)
m3
Then the equations for kinetics of a single isobaric isothermal reaction, Eq. (4.354) in conjunction
with Eq. (4.280), reduce for N2 molar concentration to
!
dN2 E
N 1 N2 RT
= aT exp (N2 ) 2 (N ) N 1 ( ) (N )
N 2 N
N2 (N2 + N ) .
dt RT Kc N2 P
| {z }
=r
(4.371)

Realizing that N 2
= 2, N = 0, N2 = 1, and N = 2, one gets
!
dN2 E 1 2N N2 RT
= aT exp 2N2 1 1 (4.372)
dt RT K c N 2 P
| {z }
=k(T )
The temperature dependency of the reaction is unchanged from the previous reaction:

E
k(T ) = aT exp , (4.373)
RT
!
3 1.6 kJ
18 m K 1.6 941550 kmole
= 7.0 10 T exp kJ
, (4.374)
kmole s 8.314 kmole KT

7.0 1018 1.1325 105
= exp (4.375)
T 1.6 T
For this problem, T = 6000 K, so

7.0 1018 1.1325 105
k(6000) = exp , (4.376)
60001.6 6000
m3
= 40130.2 . (4.377)
kmole s
The equilibrium constant Kc is also unchanged from the previous example. Recall
PN
i=1 i

Po Go
Kc = exp (4.378)
RT RT
PN
For this system, since i=1 i = n = 1, this reduces to

Po (2goN g oN2 )
Kc = exp , (4.379)
RT RT
f ( N ) (kmole/m 3 /s)
2
4
unstable
unstable 3
equilibrium
equilibrium
2
-0.002 -0.001 0.001 0.002 0.003 (kmole/m 3 )

N2
-1
stable,
stable,
physical
non-physical
equilibrium
equilibrium
Figure 4.4: Forcing function, f (N2 ), which drives changes of N2 as a function of N2 in

isothermal, isobaric problem.

Po (2goN g oN2 )
= exp , (4.380)
RT RT
o o !
Po (2(hN T soT,N ) (hN2 T soT,N2 ))
= exp , (4.381)
RT RT

100 (2(597270 (6000)216.926) (205848 (6000)292.984))
= exp ,(4.382)
(8.314)(6000) (8.314)(6000)
kmole
= 0.000112112 . (4.383)
m3
The differential equation for N2 evolution is then given by
2 !
dN2 m3 2 1 2.00465 103 kmole
m3 N 2
= 40130.2 N2 1
dt kmole 0.000112112 kmole
m3
N 2
!
kJ
8.314 kmole K (6000 K)
1 N2 ,
100 kP a
(4.384)
f (N2 ). (4.385)
The system is at equilibrium when f (N2 ) = 0. This is an algebraic function of N2 only, and can be
plotted. Figure 4.4 gives a plot of f (N2 ) and shows that it has four potential equilibrium points. It is
seen there are four roots. Solving for the equilibria requires solving
2 !
m3 2 1 2.00465 103 kmole
m3 N 2
0 = 40130.2 N 2 1
kmole 0.000112112 kmole
m3
N 2
!
kJ
8.314 kmole K (6000 K)
1 N2 ,
100 kP a
0.002
concentration (kmole/m 3 )

N2
0.0015
0.001
0.0005

N
t (s)
0.001 0.002 0.003 0.004 0.005
Figure 4.5: N2 (t) and N (t) in isobaric, isothermal nitrogen dissociation problem.
(4.386)
The four roots are

kmole kmole kmole kmole
N2 = 0.002005 , 0 , 0.001583 , 0.00254 . (4.387)
| {z m3 } m3 }
| {z |
3
{z m } |
3
{z m }
stable,nonphysical unstable stable,physical unstable
By inspection of the topology of Fig. 4.2, the only stable, physical root is 0.001583 kmole m3 . Small
perturbations from this equilibrium induce the forcing function to supply dynamics which restore the
system to its original equilibrium state. Small perturbations from the unstable equilibria induce non-
restoring dynamics. For this root, one can then determine that the stable equilibrium value of N =
0.000421 kmole
m3 .
A numerical solution via an explicit technique such as a Runge-Kutta integration is found for
Eq. (4.386). The solution for N2 , along with N is plotted in Fig. 4.5.
Linearization of Eq. (4.386) about the equilibrium state gives rise to the locally linearly valid
d
0.001583 = 967.073(N2 0.001583) + . . . (4.388)
dt N2
This has local solution
N2 = 0.001583 + C exp (967.073t) . (4.389)
Again, C is an irrelevant integration constant. The time scale of relaxation is the time when the
argument of the exponential is 1, which is
1
= = 1.03 103 s. (4.390)
967.073 s1
Note that the time constant for the isobaric combustion is about a factor 1.25 greater than for isochoric
combustion under the otherwise identical conditions.
The equilibrium values agree exactly with those found in the earlier example. Again, the kinetics
provide the details of how much time it takes to achieve equilibrium.
4.6 Some conservation and evolution equations

Here a few useful global conservation and evolution equations are presented for some key
properties. Only some cases are considered, and one could develop more relations for other
scenarios.
4.6.1 Total mass conservation: isochoric reaction

One can easily show that the isochoric reaction rate model, Eq. (4.279), satisfies the principle
of mixture mass conservation. Begin with Eq. (4.279) in a compact form, using the definition
of the reaction rate r, Eq. (4.281) and perform the following operations:
di
= i r, (4.391)
dt
d Yi
= i r, (4.392)
dt Mi
d
(Yi ) = i Mi r, (4.393)
dt
L
X
d
(Yi ) = i Ml li r, (4.394)
dt l=1
| {z }
=Mi
L
X
d
(Yi ) = Ml li i r, (4.395)
dt
l=1
N
X XN XL
d
(Yi ) = Ml li i r, (4.396)
i=1
dt i=1 l=1

d
N L X N
X X
Yi = Ml li i r, (4.397)
dt i=1 l=1 i=1
| {z }
=1
X L X N
d
= r Ml li i , (4.398)
dt
l=1
|i=1{z }
=0
d
= 0. (4.399)
dt
PN
Note the term i=1 li i = 0 because of stoichiometry, Eq. (4.24).
4.6. SOME CONSERVATION AND EVOLUTION EQUATIONS 179
4.6.2 Element mass conservation: isochoric reaction

Through a similar series of operations, one can show that the mass of each element, l =
1, . . . , L, in conserved in this reaction, which is chemical, not nuclear. Once again, begin
with Eq. (4.281) and perform a set of operations,
di
= i r, (4.400)
dt
d
li i = li i r, l = 1, . . . , L, (4.401)
dt
d
(li i ) = rli i , l = 1, . . . , L, (4.402)
dt
XN XN
d
(li i ) = rlii , l = 1, . . . , L, (4.403)
i=1
dt i=1
N
! N
d X X
li i = r li i , l = 1, . . . , L, (4.404)
dt i=1 i=1
| {z }
=0
N
!
d X
li i = 0, l = 1, . . . , L. (4.405)
dt i=1
P
The term N i=1 li i represents the number of moles of element l per unit volume, by the
following analysis
N
X XN
moles element l moles species i moles element l
li i = = l e . (4.406)
i=1 i=1
moles species i volume volume
Here the elemental mole density, l e , for element l has been defined. So the element concen-
tration for each element remains constant in a constant volume reaction process:
dl e
= 0, l = 1, . . . , L. (4.407)
dt
One can also multiply by the elemental mass, Ml to get the elemental mass density, el :
el Ml l e , l = 1, . . . , L. (4.408)
Since Ml is a constant, one can incorporate this definition into Eq. (4.407) to get
del
= 0, l = 1, . . . , L. (4.409)
dt
The element mass density remains constant in the constant volume reaction. One could also
simply say since the elements density is constant, and the mixture is simply a sum of the
elements, that the mixture density is conserved as well.
4.6.3 Energy conservation: adiabatic, isochoric reaction

Consider a simple application of the first law of thermodynamics to reaction kinetics: that
of a closed, adiabatic, isochoric combustion process in a mixture of ideal gases. One may
be interested in the rate of temperature change. First, because the system is closed, there
can be no mass change, and because the system is isochoric, the total volume is a non-zero
constant; hence,
dm
= 0, (4.410)
dt
d
(V ) = 0, (4.411)
dt
d
V = 0, (4.412)
dt
d
= 0. (4.413)
dt
For such a process, the first law of thermodynamics is
dE
= Q W
. (4.414)
dt
But there is no heat transfer or work in the adiabatic isochoric process, so one gets
dE
= 0, (4.415)
dt
d
(me) = 0, (4.416)
dt
de dm
m +e = 0, (4.417)
dt dt
|{z}
=0
de
= 0. (4.418)
dt
Thus for the mixture of ideal gases, e(T, 1 , . . . , N ) = eo . One can see how reaction rates
affect temperature changes by expanding the derivative in Eq. (4.418)
N
!
d X
Yiei = 0, (4.419)
dt i=1
XN
d
(Yiei ) = 0, (4.420)
i=1
dt
XN
dei dYi
Yi + ei = 0, (4.421)
i=1
dt dt
XN
dei dT dYi
Yi + ei = 0, (4.422)
i=1
dT dt dt
XN
dT dYi
Yi cvi + ei = 0, (4.423)
i=1
dt dt
N N
dT X X dYi
Yi cvi = ei , (4.424)
dt i=1 i=1
dt
| {z }
=cv
N
X
dT d Mi i
cv = ei , (4.425)
dt i=1
dt
XN
dT di
cv = ei Mi , (4.426)
dt i=1
dt
XN
dT
cv = ei Mi i r, (4.427)
dt i=1
PN
dT r i=1 i ei
= . (4.428)
dt cv
If one defines the net energy change of the reaction as
N
X
E i ei , (4.429)
i=1
one then gets

dT rE
= . (4.430)
dt cv
The rate of temperature change is dependent on the absolute energies, not the energy dif-
ferences. If the reaction is going forward, so r > 0, and that is a direction in which the net
molar energy change is negative, then the temperature will rise.
4.6.4 Energy conservation: adiabatic, isobaric reaction

Solving for the reaction dynamics in an adiabatic isobaric system requires some non-obvious
manipulations. First, the first law of thermodynamics says dE = dQdW . Since the process
is adiabatic, one has dQ = 0, so dE + P dV = 0. Since it is isobaric, one gets d(E + P V ) = 0,
or dH = 0. So the total enthalpy is constant. Then
d
H = 0, (4.431)
dt
d
(mh) = 0, (4.432)
dt
dh
= 0, (4.433)
dt
!
N
X
d
Yi hi = 0, (4.434)
dt i=1
XN
d
(Yi hi ) = 0, (4.435)
i=1
dt
N
X dhi dYi
Yi + hi = 0, (4.436)
i=1
dt dt
N
X dhi dT dYi
Yi + hi = 0, (4.437)
i=1
dT dt dt
N
X N
dT X dYi
Y i cP i + hi = 0, (4.438)
i=1
dt i=1
dt
N N
dT X X dYi
Y i cP i + hi = 0, (4.439)
dt i=1 i=1
dt
| {z }
=cP
N
dT X d i Mi
cP + hi = 0, (4.440)
dt i=1
dt
N
dT X d i
cP + hi Mi = 0. (4.441)
dt i=1
dt
Now use Eq (4.340) to eliminate the term in Eq. (4.441) involving molar concentration
derivatives to get
N
dT X i r
cP + hi = 0, (4.442)
dt i=1

PN
dT r i=1 hi i
= . (4.443)
dt cP
So the temperature derivative is known as an algebraic function. If one defines the net
enthalpy change as
XN
H hi i , (4.444)
i=1
one gets
dT rH
= . (4.445)
dt cP
or
dT
cP = rH. (4.446)
dt
Equation (4.446) is in a form which can easily be compared to a later form when we add
variable pressure and diffusion effects.
Now differentiate the isobaric ideal gas law to get the density derivative.
N
X
P = i RT, (4.447)
i=1
XN N
dT X d
0 = +
i R RT i , (4.448)
i=1
dt i=1
dt
N N
dT X X i d
0 = + T i r + , (4.449)
dt i=1 i i=1 dt
N N N
1 dT X X 1 d X
0 = +r i + , (4.450)
T dt i=1 i i=1
dt i=1 i
PN PN
d T1 dT
dt i=1 i r i=1 i
= PN i . (4.451)
dt i=1
dT
One takes dt
from Eq. (4.443) to get
PN PN PN
1 r hi i
i=1 i r i
i=1
d T cP i=1
= PN i . (4.452)
dt i=1
Now recall that = /M and cP = cP M, so cP = cP . Then Equation (4.452) can be

reduced slightly:
=1
z }| {
PN XN
hi i i PN
i=1
cP T
i=1 i
d i=1

= r PN , (4.453)
dt i=1 i
PN hi i PN
i=1 cP T i=1 i
= r PN , (4.454)

PN i=1 i
hi
i=1 i cP T 1
= r P
, (4.455)
RT
N
RT X hi
= r i 1 , (4.456)
P i=1 cP T
N
X
hi
= rM i 1 , (4.457)
i=1
cP T
PN
where M is the mean molecular mass. Note for exothermic reaction i=1 i hi < 0, so
exothermic reaction induces a density decrease as the increased temperature at constant
pressure causes the volume to increase.
Then using Eq. (4.457) to eliminate the density derivative in Eq. (4.339), and changing
the dummy index from i to k, one gets an explicit expression for concentration evolution:
XN
di i hk
= i r + rM k 1 , (4.458)
dt k=1
cP T

N

i X hk
= r i + M k 1 , (4.459)
cP T
| {z } k=1
=yi
N
X !
hk
= r i + yi k 1 . (4.460)
k=1
cP T
PN
Defining the change of enthalpy of the reaction as H k=1 k hk , and the change of
PN
number of the reaction as n k=1 k , one can also say

di H
= r i + yi n . (4.461)
dt cP T
Exothermic reaction, H < 0, and net number increases, n > 0, both tend to decrease the
molar concentrations of the species in the isobaric reaction.
Lastly, the evolution of the adiabatic, isobaric system, can be described by the simul-
taneous, coupled ordinary differential equations: Eqs. (4.443, 4.452, 4.460). These require
numerical solution in general. Note also that one could also employ a more fundamental
P
treatment as a differential algebraic system involving H = H1 , P = P1 = RT N i=1 i and
Eq. (4.339).
4.6.5 Non-adiabatic isochroic combustion

Consider briefly combustion in a fixed finite volume in which there is simple convective heat
transfer with the surroundings. In general, the first law of thermodynamics is
dE
= Q W
. (4.462)
dt
= 0. And using standard relations from simple convective
Because the system is isochoric W
heat transfer, one can say that
dE
= hA(T T ). (4.463)
dt
Here h is the convective heat transfer coefficient and A is the surface area associated with
the volume V , and T is the temperature of the surrounding medium. One can, much as
before, write E in detail and get an equation for the evolution of T .
4.6.6 Entropy evolution: Clausius-Duhem relation

Now consider whether the kinetics law that has been posed actually satisfies the second law
of thermodynamics. Consider again Eq. (3.400). There is an algebraic relation on the right
side. If it can be shown that this algebraic relation is positive semi-definite, then the second
law is satisfied, and the algebraic relation is known as a Clausius-Duhem relation.
Now take Eq. (3.400) and perform some straightforward operations on it:
N
1X
dS|E,V = dni 0, (4.464)
T i=1 i
| {z }
irreversibility
N
dS V X dni 1
= 0, (4.465)
dt E,V T i=1 i dt V
N
V X di
= 0, (4.466)
T i=1 i dt

Y !
V X
N
E
N 1 Y k
N
k
= i i aT exp k 1 k 0,(4.467)
T i=1 RT Kc
| {z } | k=1{z } | k=1 {z }
| {z }
r
N N
! N
!
V X Y k 1 Y k
= i k(T ) k 1 0, (4.468)
T i=1 i k=1
Kc k=1 k
N
! N
! N !
V Y k 1 Y k X
= k(T ) k 1 k i i 0, (4.469)
T Kc i=1
k=1 k=1
| {z }
=
Change the dummy index from k back to i,
N
! N
!
V Y i 1 Y i
= k(T ) i 1 0, (4.470)
T i=1
Kc i=1 i
V
= r, (4.471)
T
d
= . (4.472)
T dt
Consider now the affinity term in Eq. (4.470) and expand it so that it has a more useful
form:
N
X N
X
= i i = g i i , (4.473)
i=1 i=1
XN
Pi
= g oT,i + RT ln i , (4.474)
i=1
Po
XN X N i
Pi
= g oT,i i RT ln , (4.475)
P o
|i=1 {z } i=1
=Go

N i
Go X Pi

= RT ln , (4.476)
Po
| RT
{z } i=1
=ln KP
N !
Y Pi i
= RT ln KP ln , (4.477)
i=1
Po
!
N
Y
1 Pi i
= RT ln + ln , (4.478)
KP i=1
P o
N
Y !
1 Pi i
= RT ln , (4.479)
KP i=1 Po
PNi=1 i N
Po Y RT i
i
= RT ln RT , (4.480)
Kc i=1
P o
N
!
1 Y i
= RT ln . (4.481)
Kc i=1 i
Equation (4.481) is the common definition of affinity. Another form can be found by em-
ploying the definition of Kc from Eq. (4.270) to get
PNi=1 i Y
N
!
Po Go i
= RT ln exp i , (4.482)
RT RT i=1
PNi=1 i Y N
!!
Go Po
= RT + ln i i , (4.483)
RT RT i=1
PNi=1 i YN
!
P o
= Go RT ln i i (4.484)
RT i=1
To see clearly that the entropy production rate is positive semi-definite, substitute
Eq. (4.481) into Eq. (4.470) to get
N
! N
! N
!!

dS V Y 1 Y 1 Y
= k(T ) i i 1 i i RT ln i i 0,
dt E,V T i=1
Kc i=1 Kc i=1
(4.485)
N
! N
! N
!
Y i 1 Y 1 Y
= RV k(T ) i 1 i i ln i i 0. (4.486)
i=1
Kc i=1
Kc i=1
Define forward and reverse reaction coefficients, R , and R , respectively, as

N
Y
R k(T ) i i , (4.487)
i=1
YN
k(T )
R i i . (4.488)
Kc i=1
Both R and R have units of kmole/m3 /s. It is easy to see that

r = R R . (4.489)
Note that since k(T ) > 0, Kc > 0, and i 0, that both R 0 and R 0. Since
i = i i , one finds that
N N
1 Y i 1 k(T ) Y i i R
= = . (4.490)
Kc i=1 i Kc k(T ) i=1 i R
Then Eq. (4.486) reduces to

dS R R
= RV R 1 ln 0, (4.491)
dt E,V R R

R
= RV (R R ) ln 0. (4.492)
R
Obviously, if the forward rate is greater than the reverse rate R R > 0, ln(R /R ) > 0,
and the entropy production is positive. If the forward rate is less than the reverse rate,
R R < 0, ln(R /R ) < 0, and the entropy production is still positive. The production
rate is zero when R = R .
Note that the affinity can be written as

R
= RT ln . (4.493)
R
And so when the forward reaction rate exceeds the reverse, the affinity is positive. It is zero
at equilibrium, when the forward reaction rate equals the reverse.
4.7 Simple one-step kinetics

A common model in theoretical combustion is that of so-called simple one-step kinetics. Such
a model, in which the molecular mass does not change, is quantitatively appropriate only
for isomerization reactions. However, as a pedagogical tool as well as a qualitative model for
real chemistry, it can be valuable.
Consider the reversible reaction
AB (4.494)
where chemical species A and B have identical molecular masses MA = MB = M. Consider
further the case in which at the initial state, no moles of A only are present. Also take the
reaction to be isochoric and isothermal. These assumptions can easily be relaxed for more
general cases. Specializing then Eq. (4.240) for this case, one has
nA = A + nAo , (4.495)
|{z} |{z}
=1 =no
nB = B + nBo . (4.496)
|{z} |{z}
=1 =0
Thus
nA = + no , (4.497)
nB = . (4.498)
Now no is constant throughout the reaction. Scale by this and define the dimensionless
reaction progress as /no to get
nA
= + 1, (4.499)
no
|{z}
=yA
nB
= . (4.500)
no
|{z}
=yB
In terms of the mole fractions then, one has
yA = 1 , (4.501)
yB = . (4.502)
The reaction kinetics for each species reduce to

dA no
= r, A (0) = o , (4.503)
dt V
dB
= r, B (0) = 0. (4.504)
dt
4.7. SIMPLE ONE-STEP KINETICS 189
Addition of Eqs. (4.503) and (4.504) gives
d
( + B ) = 0, (4.505)
dt A
A + B = o , (4.506)
A
+ B = 1. (4.507)
o o
|{z} |{z}
=yA =yB
In terms of the mole fractions yi , one then has
yA + yB = 1. (4.508)
The reaction rate r is then

1 B
r = kA 1 , (4.509)
Kc A

A 1 B /o
= ko 1 , (4.510)
o Kc A /o

1 yB
= ko yA 1 , (4.511)
Kc y A

1
= ko (1 ) 1 . (4.512)
Kc 1
Now r = (1/V )d/dt = (1/V )d(no )/dt = (no /V )d()/dt = o d/dt. So the reaction
dynamics can be described by a single ordinary differential equation in a single unknown:

d 1
o = ko (1 ) 1 , (4.513)
dt Kc 1

d 1
= k(1 ) 1 . (4.514)
dt Kc 1
Equation (4.514) is in equilibrium when
1 1
= 1 1 + ... (4.515)
1 + Kc Kc
As Kc , the equilibrium value of 1. In this limit, the reaction is irreversible. That

is, the species B is preferred over A. Equation (4.514) has exact solution

1 exp k 1 + K1c t
= . (4.516)
1 + K1c
For k > 0, Kc > 0, the equilibrium is stable. The time constant of relaxation is
1
= . (4.517)
1
k 1+ Kc
For the isothermal, isochoric system, one should consider the second law in terms of the
Helmholtz free energy. Combine then Eq. (3.406), dA|T,V 0, with Eq. (3.304), dA =
P
SdT P dV + N i=1 i dni and taking time derivatives, one finds
!
XN

dA|T,V = SdT P dV + i dni 0, (4.518)

i=1 T,V
N
X dni
dA
= i 0, (4.519)
dt T,V i=1
dt
N
1 dA V X di
= 0. (4.520)
T dt T i=1 i dt
This is exactly the same form as Eq. (4.486), which can be directly substituted into Eq. (4.520)
to give
N
! N
! N
!
1 dA Y 1 Y 1 Y
= RV k(T ) i i 1 i i ln i i 0,
T dt T,V i=1
K c i=1 K c i=1
(4.521)
N
! N
! N
!
dA Y i 1 Y i 1 Y i
= RV T k(T ) i 1 ln 0.
dt T,V i=1
Kc i=1 i Kc i=1 i
(4.522)
For the assumptions of this section, Eq. (4.522) reduces to

dA 1 1
= RT ko V (1 ) 1 ln 0, (4.523)
dt T,V Kc 1 Kc 1

1 1
= kno RT (1 ) 1 ln 0. (4.524)
Kc 1 Kc 1
Since the present analysis is nothing more than a special case of the previous section,
Eq. (4.524) certainly holds. One questions however the behavior in the irreversible limit,
1/Kc 0. Evaluating this limit, one finds

dA
1
0.
lim = kn RT (1 ) ln +(1 ) ln (1 ) ln(1 ) + . . .
1/Kc 0 dt T,V
o | {z }
Kc
>0 | {z }

(4.525)
4.7. SIMPLE ONE-STEP KINETICS 191
Now, performing the distinguished limit as 1; that is the reaction goes to completion,
one notes that all terms are driven to zero for small 1/Kc . Recall that 1 goes to zero faster
than ln(1 ) goes to . Note that the entropy inequality is ill-defined for a formally
irreversible reaction with 1/Kc = 0.
Chapter 5
Thermochemistry of multiple
reactions
This chapter will extend notions associated with the thermodynamics of a single chemical
reactions to systems in which many reactions occur simultaneously. Some background is in
some standard sources. 1 2 3
5.1 Summary of multiple reaction extensions

Consider now the reaction of N species, composed of L elements, in J reactions. This section
will focus on the most common case in which J (N L), which is usually the case in large
chemical kinetic systems in use in engineering models. While much of the analysis will only
require J > 0, certain results will depend on J (N L). It is not difficult to study the
complementary case where 0 < J < (N L).
The molecular mass of species i is still given by
L
X
Mi = Ml li, i = 1, . . . , N. (5.1)
l=1
However, each reaction has a stoichiometric coefficient. The j th reaction can be summarized
in the following ways:
N
X N
X
i ij i ij , j = 1, . . . , J, (5.2)
i=1 i=1
N
X
i ij = 0, j = 1, . . . , J. (5.3)
i=1
1
S. R. Turns, 2000, An Introduction to Combustion, McGraw-Hill, Boston. Chapters 4-6.
2
K. K. Kuo, 2005, Principles of Combustion, Second Edition, John Wiley, New York. Chapters 1 and 2.
3
D. Kondepudi and I. Prigogene, 1998, Modern Thermodynamics: From Heat Engines to Dissipative
Structures, John Wiley, New York. Chapters 16 and 19.
193
194 CHAPTER 5. THERMOCHEMISTRY OF MULTIPLE REACTIONS
Stoichiometry for the j th reaction and lth element is given by
N
X
liij = 0, l = 1, . . . , L, j = 1, . . . , J. (5.4)
i=1
The net change in Gibbs free energy and equilibrium constants of the j th reaction are defined
by
N
X
Goj g oT,i ij , j = 1, . . . , J, (5.5)
i=1

Goj
KP,j exp , j = 1, . . . , J, (5.6)
RT
PNi=1 ij
Po Goj
Kc,j exp , j = 1, . . . , J. (5.7)
RT RT
The equilibrium of the j th reaction is given by
N
X
i ij = 0, j = 1, . . . , J, (5.8)
i=1
XN
g i ij = 0, j = 1, . . . , J. (5.9)
i=1
The multi-reaction extension for affinity is
N
X
j = i ij , j = 1, . . . , J. (5.10)
i=1
In terms of the chemical affinity of each reaction, the equilibrium condition is simply
j = 0, j = 1, . . . , J. (5.11)
At equilibrium, then the equilibrium constraints can be shown to reduce to
N
Y
Pi ij
KP,j = , j = 1, . . . , J, (5.12)
i=1
Po
YN
ij
Kc,j = i , j = 1, . . . , J. (5.13)
i=1
5.1. SUMMARY OF MULTIPLE REACTION EXTENSIONS 195
For isochoric reaction, the evolution of species concentration i due to the combined effect
of J reactions is given by
i
z }| {
Y !
di X
J
E j
N
1 Y kj
N
j kj
= ij aj T exp k 1 k , i = 1, . . . , N. (5.14)
dt RT Kc,j k=1
j=1 | {z } | k=1{z } | {z }
kj (T ) forward reaction reverse reaction
| {z }
rj =(1/V )dj /dt
The extension to isobaric reactions is straightforward, and follows the same analysis as for
a single reaction. Again, three intermediate variables which are in common usage have been
defined. First one takes the reaction rate of the j th reaction to be

Y !
E
N
1 Y kj
N
j j kj
rj aj T exp k 1 k , j = 1, . . . , J, (5.15)
RT Kc,j k=1
| {z } | k=1{z } | {z }
kj (T ) forward reaction reverse reaction

Y

N N
j E j

kj 1 Y kj
= aj T exp k k , j = 1, . . . , J, (5.16)
RT k=1 Kc,j k=1
| {z } | {z } | {z }
kj (T ), Arrhenius rate
| forward reaction {z reverse reaction }
law of mass action
1 dj
= . (5.17)
V dt
Here j is the reaction progress variable for the j th reaction.
Each reaction has a temperature-dependent rate function kj (T ), which is

j E j
kj (T ) aj T exp , j = 1, . . . , J. (5.18)
RT
The evolution rate of each species is given by i , defined now as
J
X
i ij rj , i = 1, . . . , N. (5.19)
j=1
The multi-reaction extension for mole change in terms of progress variables is

J
X
dni = ij dj , i = 1, . . . , N. (5.20)
j=1
One also has

N
X
dG|T,P = i dni , (5.21)
i=1
XN J
X
= i ik dk , (5.22)
i=1 k=1
N J
G X k X
= i ik , (5.23)
j p i=1 k=1
j
N
X J
X
= i ik kj , (5.24)
i=1 j=1
N
X
= i ij , (5.25)
i=1
= j , j = 1, . . . , J. (5.26)
In a very similar fashion to that shown for a single reaction, one can further sum over all
reactions and prove that mixture mass is conserved, element mass and number are conserved.
A similar expression is obtained for temperature changes.
Example 5.1
Show that element mass and number are conserved for the multi-reaction formulation.
Start with Eq. (5.14) and expand as follows:
J
X
di
= ij rj , (5.27)
dt j=1
J
X
di
li = li ij rj , (5.28)
dt j=1
XJ
d
(li i ) = li ij rj , (5.29)
dt j=1
X N XN X J
d
(li i ) = li ij rj , (5.30)
i=1
dt i=1 j=1
N
! J X N
d X X
li i = li ij rj , (5.31)
dt i=1 j=1 i=1
| {z }
=l e
J
X N
X
dl e
= rj li ij , (5.32)
dt j=1 i=1
| {z }
=0
5.1. SUMMARY OF MULTIPLE REACTION EXTENSIONS 197
dl e
= 0, l = 1, . . . , L, (5.33)
dt
d
(Ml l e ) = 0, l = 1, . . . , L, (5.34)
dt
dl e
= 0, l = 1, . . . , L. (5.35)
dt
It is also straightforward to show that the mixture density is conserved for the multi-reaction, multi-
component mixture:
d
= 0. (5.36)
dt
The proof of the Clausius-Duhem relationship for the second law is an extension of the
single reaction result. Start with Eq. (4.464) and operate much as for a single reaction model.
N
1X
dS|E,V = dni 0, (5.37)
T i=1 i
| {z }
irreversibility
N
dS V X dni 1
= 0, (5.38)
dt E,V T i=1 i dt V
N
V X di
= 0, (5.39)
T i=1 i dt
N J
V X X
= i ij rj 0, (5.40)
T i=1 j=1
N J
V XX
= ij rj 0, (5.41)
T i=1 j=1 i
J N
V X X
= rj ij 0, (5.42)
T j=1 i=1 i
J N N
! N
V X Y ij 1 Y ij X
= kj 1 i ij 0, (5.43)
T j=1 i=1 i Kc,j i=1 i i=1
J N N
! N
!!
V X Y ij 1 Y ij 1 Y ij
= kj 1 RT ln 0,
T j=1 i=1 i Kc,j i=1 i Kc,j i=1 i
(5.44)
J N N
! N
!
X Y 1 Y 1 Y
= RV kj i ij 1 i ij ln i ij 0,
j=1 i=1
Kc,j i=1
Kc,j i=1
(5.45)
Note that Eq. (5.42) can also be written in terms of the affinities (see Eq. (5.10)) and reaction
progress variables (see Eq. (5.17) as
J
dS 1 X dj
= j 0. (5.46)
dt E,V T j=1 dt
Similar to the argument for a single reaction, if one defines

N
Y
ij
Rj = kj i , (5.47)
i=1
N
kj Y ij
Rj = , (5.48)
Kc,j i=1 i
then it is easy to show that

rj = Rj Rj , (5.49)
and
J
dS X Rj
= RV Rj Rj ln 0. (5.50)
dt E,V j=1
Rj
Since kj (T ) > 0, R > 0, and V 0, and each term in the summation combines to be positive
semi-definite, one sees that the Clausius-Duhem inequality is guaranteed to be satisfied for
multi-component reactions.
5.2 Equilibrium conditions

For multicomponent mixtures undergoing multiple reactions, determining the equilibrium
condition is more difficult. There are two primary approaches, both of which are essentially
equivalent. The most straightforward method requires formal minimization of the Gibbs free
energy of the mixture. It can be shown that this actually finds the equilibrium associated
with all possible reactions.
5.2.1 Minimization of G via Lagrange multipliers

PN
Recall
Eq. (3.407), dG| T,P 0. Recall also Eq. (3.409), G = i=1 g i ni . Since i = g i =
G
PN PN
ni
, one also has G = i=1 i ni . From Eq. (3.410), dG|T,P = i=1 i dni . Now
P,T,nj
one must also demand for a system coming to equilibrium that the element numbers are
conserved. This can be achieved by requiring
N
X
li (nio ni ) = 0, l = 1, . . . , L. (5.51)
i=1
5.2. EQUILIBRIUM CONDITIONS 199
Here recall nio is the initial number of moles of species i in the mixture, and li is the number
of moles of element l in species i.
One can now use the method of constrained optimization given by the method of Lagrange
multipliers to extremize G subject to the constraints of element conservation. The extremum
will be a minimum; this will not be proved, but it will be demonstrated. Define a set of L
Lagrange multipliers l . Next define an augmented Gibbs free energy function G , which is
simply G plus the product of the Lagrange multipliers and the constraints:
L
X N
X

G =G+ l li (nio ni ). (5.52)
l=1 i=1
Now when the constraints are satisfied, one has G = G, so assuming the constraints can be
satisfied, extremizing G is equivalent to extremizing G . To extremize G , take its differential
with respect to ni , with P , T and nj constant and set it to zero for each species:
L
G G X
= l li = 0. i = 1, . . . , N. (5.53)
ni T,P,nj ni T,P,nj l=1
| {z }
=i
With the definition of the partial molar property i , one then gets
L
X
i l li = 0, i = 1, . . . , N. (5.54)
l=1
Next, for an ideal gas, one can expand the chemical potential so as to get
L
X
Pi
oT,i + RT ln l li = 0, i = 1, . . . , N, (5.55)
Po
| {z } l=1
=i

! L
ni P 1 X
oT,i
+ RT ln PN l li = 0, i = 1, . . . , N. (5.56)
k=1 nk
Po
l=1
| {z }
=Pi
PN
Recalling that k=1 nk = n, in summary then, one has N + L equations
L
X
ni P
oT,i + RT ln l li = 0, i = 1, . . . , N, (5.57)
n Po l=1
N
X
li (nio ni ) = 0, l = 1, . . . , L. (5.58)
i=1
in N + L unknowns: ni , i = 1, . . . , N, l , l = 1, . . . , L.
Example 5.2
Consider a previous example problem in which
N2 + N2 2N + N2 . (5.59)
Take the reaction to be isothermal and isobaric with T = 6000 K and P = 100 kP a. Initially one has
1 kmole of N2 and 0 kmole of N . Use the extremization of Gibbs free energy to find the equilibrium
composition.
First find the chemical potentials at the reference pressure of each of the possible constituents.
o o o
oT,i = goi = hi T soi = h298,i + hi T soi . (5.60)
For each species, one then finds
kJ
oN2 = 0 + 205848 (6000)(292.984) = 1552056 , (5.61)
kmole
kJ
oN = 472680 + 124590 (6000)(216.926) = 704286 . (5.62)
kmole
To each of these one must add
ni P
RT ln
nPo
to get the full chemical potential. Now P = Po = 100 kP a for this problem, so one only must consider
kJ
RT = 8.314(6000) = 49884 kmole . So, the chemical potentials are

nN 2
N2 = 1552056 + 49884 ln , (5.63)
nN + nN 2

nN
N = 704286 + 49884 ln . (5.64)
nN + nN 2
Then one adds on the Lagrange multiplier and then considers element conservation to get the
following coupled set of nonlinear algebraic equations:

nN 2
1552056 + 49884 ln 2N = 0, (5.65)
nN + nN 2

nN
704286 + 49884 ln N = 0, (5.66)
nN + nN 2
nN + 2nN2 = 2. (5.67)
These non-linear equations are solved numerically to get
nN 2 = 0.88214 kmole, (5.68)

nN = 0.2357 kmole, (5.69)
kJ
N = 781934 . (5.70)
kmole
These agree with results found in an earlier example problem.
Example 5.3
Consider a mixture of 2 kmole of H2 and 1 kmole of O2 at T = 3000 K and P = 100 kP a.
Assuming an isobaric and isothermal equilibration process with the products consisting of H2 , O2 ,
H2 O, OH, H, and O, find the equilibrium concentrations. Consider the same mixture at T = 298 K
and T = 1000 K.
The first task is to find the chemical potentials of each species at the reference pressure and
T = 3000 K. Here one can use the standard tables along with the general equation
o o o
oT,i = goi = hi T soi = h298,i + hi T soi . (5.71)
For each species, one then finds
kJ
oH2 = 0 + 88724 3000(202.989) = 520242 , (5.72)
kmole
kJ
oO2 = 0 + 98013 3000(284.466) = 755385 , (5.73)
kmole
kJ
oH2 O = 241826 + 126548 3000(286.504) = 974790 , (5.74)
kmole
kJ
oOH = 38987 + 89585 3000(256.825) = 641903 , (5.75)
kmole
kJ
oH = 217999 + 56161 3000(162.707) = 213961 , (5.76)
kmole
kJ
oO = 249170 + 56574 3000(209.705) = 323371 . (5.77)
kmole
To each of these one must add
ni P
RT ln
nPo
to get the full chemical potential. Now P = Po = 100 kP a for this problem, so one must only consider
kJ
RT = 8.314(3000) = 24942 kmole . So, the chemical potentials are

nH2
H2 = 520243 + 24942 ln , (5.78)
nH2 + nO2 + nH2 O + nOH + nH + nO

nO2
O2 = 755385 + 24942 ln , (5.79)

nH2 O
H2 O = 974790 + 24942 ln , (5.80)

nOH
OH = 641903 + 24942 ln , (5.81)

nH
H = 213961 + 24942 ln , (5.82)

nO
O = 323371 + 24942 ln . (5.83)
Then one adds on the Lagrange multipliers and then considers element conservation to get the following
coupled set of nonlinear equations:

nH2
520243 + 24942 ln 2H = 0, (5.84)

nO2
755385 + 24942 ln 2O = 0, (5.85)

nH2 O
974790 + 24942 ln 2H O = 0, (5.86)
nH + nO2 + nH2 O + nOH + nH + nO
2
nOH
641903 + 24942 ln H O = 0, (5.87)

nH
213961 + 24942 ln H = 0, (5.88)
nH + nO2 + nH2 O + nOH + nH + nO
2
nO
323371 + 24942 ln O = 0, (5.89)
2nH2 + 2nH2 O + nOH + nH = 4, (5.90)
2nO2 + nH2 O + nOH + nO = 2. (5.91)
These non-linear algebraic equations can be solved numerically via a Newton-Raphson technique.
The equations are sensitive to the initial guess, and one can use ones intuition to help guide the
selection. For example, one might expect to have nH2 O somewhere near 2 kmole. Application of the
Newton-Raphson iteration yields
nH2 = 3.19 101 kmole, (5.92)

1
nO2 = 1.10 10 kmole, (5.93)
nH2 O = 1.50 100 kmole, (5.94)
nOH = 2.20 101 kmole, (5.95)
nH = 1.36 101 kmole, (5.96)
nO = 5.74 102 kmole, (5.97)
kJ
H = 2.85 105 , (5.98)
kmole
kJ
O = 4.16 105 . (5.99)
kmole
At this relatively high value of temperature, all species considered have a relatively major presence.
That is, there are no truly minor species.
Unless a very good guess is provided, it may be difficult to find a solution for this set of non-
linear equations. Straightforward algebra allows the equations to be recast in a form which sometimes
converges more rapidly:
2
nH2 520243 H
= exp exp , (5.100)
nH2 + nO2 + nH2 O + nOH + nH + nO 24942 24942
2
nO2 755385 O
= exp exp , (5.101)
nH2 + nO2 + nH2 O + nOH + nH + nO 24942 24942
2
nH2 O 974790 O H
= exp exp exp , (5.102)
nH2 + nO2 + nH2 O + nOH + nH + nO 24942 24942 24942

nOH 641903 O H
= exp exp exp , (5.103)
nH2 + nO2 + nH2 O + nOH + nH + nO 24942 24942 24942

nH 213961 H
= exp exp , (5.104)
nH2 + nO2 + nH2 O + nOH + nH + nO 24942 24942

nO 323371 O
= exp exp , (5.105)
nH2 + nO2 + nH2 O + nOH + nH + nO 24942 24942
G() - G(0) (kJ)
40
30
20
10
-0.01 -0.005 0.005 0.01

(kmole)
Figure 5.1: Gibbs free energy variation as mixture composition is varied maintaining element
conservation for mixture of H2 , O2 , H2 O, OH, H, and O at T = 3000 K, P = 100 kP a.
2nH2 + 2nH2 O + nOH + nH = 4, (5.106)

2nO2 + nH2 O + nOH + nO = 2. (5.107)
Then solve these considering ni , exp (O /24942), and exp (H /24942) as unknowns. The same result
is recovered, but a broader range of initial guesses converge to the correct solution.
One can verify that this choice extremizes G by direct computation; moreover, this will show
that the extremum is actually a minimum. In so doing, one must exercise care to see that element
conservation is retained. As an example, perturb the equilibrium solution above for nH2 and nH such
that
nH2 = 3.19 101 + , (5.108)

1
nH = 1.36 10 2. (5.109)
Leave all other species mole numbers the same. In this way, when = 0, one has the original equilibrium
solution. For 6= 0, the solution
PN moves off the equilibrium value in such a way that elements are
conserved. Then one has G = i=1 i ni = G().
The difference G() G(0) is plotted in Fig. 5.1. When = 0, there is no deviation from the value
predicted by the Newton-Raphson iteration. Clearly when = 0, G() G(0), takes on a minimum
value, and so then does G(). So the procedure works.
At the lower temperature, T = 298 K, application of the same procedure yields very different
results:
nH2 = 4.88 1027 kmole, (5.110)

27
nO2 = 2.44 10 kmole, (5.111)
0
nH2 O = 2.00 10 kmole, (5.112)
nOH = 2.22 1029 kmole, (5.113)
nH = 2.29 1049 kmole, (5.114)

54
nO = 1.67 10 kmole, (5.115)
kJ
H = 9.54 104 , (5.116)
kmole
kJ
O = 1.07 105 . (5.117)
kmole
At the intermediate temperature, T = 1000 K, application of the same procedure shows the minor
species become slightly more prominent:
nH2 = 4.99 107 kmole, (5.118)

nO2 = 2.44 107 kmole, (5.119)
nH2 O = 2.00 100 kmole, (5.120)
nOH = 2.09 108 kmole, (5.121)
12
nH = 2.26 10 kmole, (5.122)
nO = 1.10 1013 kmole, (5.123)
kJ
H = 1.36 105 , (5.124)
kmole
kJ
O = 1.77 105 . (5.125)
kmole
5.2.2 Equilibration of all reactions

In another equivalent method, if one commences with a multi-reaction model, one can require
each reaction to be in equilibrium. This leads to a set of algebraic equations for rj = 0, which
from Eq. (5.16) leads to
PNi=1 ij N
Y
Po Goj kj
Kc,j = exp = k , j = 1, . . . , J. (5.126)
RT RT k=1
With some effort it can be shown that not all of the J equations are linearly independent.
Moreover, they do not possess a unique solution. However, for closed systems, only one of
the solutions is physical, as will be shown in the following section. The others typically
involve non-physical, negative concentrations.
Nevertheless, Eqs. (5.126) are entirely consistent with the predictions of the N + L equa-
tions which arise from extremization of Gibbs free energy while enforcing element number
constraints. This can be shown by beginning with Eq. (5.56), rewritten in terms of molar
concentrations, and performing the following sequence of operations:
! L
ni /V P X
o
T,i + RT ln PN l li = 0, i = 1, . . . , N, (5.127)
k=1 nk /V
Po l=1
! L
i P X
o
T,i + RT ln PN l li = 0, i = 1, . . . , N, (5.128)
P
k=1 k o l=1
X L
o i P
T,i + RT ln l li = 0, i = 1, . . . , N, (5.129)
Po
l=1
X L
RT
oT,i + RT ln i l li = 0, i = 1, . . . , N, (5.130)
Po l=1
XL
RT
ij oT,i + ij RT ln i ij l li = 0, i = 1, . . . , N,
Po l=1
j = 1, . . . , J, (5.131)
N
X N
X X N XL
RT
ij oT,i + ij RT ln i ij l li = 0, j = 1, . . . , J, (5.132)
Po
|i=1 {z } i=1 i=1 l=1
=Goj
N
X X L XN
RT
Goj + RT ij ln i l li ij = 0, j = 1, . . . , J, (5.133)
i=1
Po l=1 i=1
| {z }
=0
N
X
RT
Goj + RT ij ln i = 0, j = 1, . . . , J. (5.134)
i=1
Po
Here, the stoichiometry for each reaction has been employed to remove the Lagrange multi-
pliers. Continue to find
N
X ij
RT Goj
ln i = , j = 1, . . . , J, (5.135)
i=1
Po RT
X N ij !
RT Goj
exp ln i = exp , j = 1, . . . , J, (5.136)
i=1
Po RT
YN ij
RT Goj
i = exp , j = 1, . . . , J, (5.137)
i=1
Po RT
PN
i=1 ij YN
RT Goj
i ij = exp , j = 1, . . . , J, (5.138)
Po i=1
RT
N
Y PNi=1 ij
Po Goj
i ij = exp , j = 1, . . . , J, (5.139)
RT RT
i=1 | {z }
=Kc,j
N
Y
i ij = Kc,j , j = 1, . . . , J. (5.140)
i=1
Thus, extremization of Gibbs free energy is consistent with equilibrating each of the J
reactions.
5.2.3 Zeldovichs uniqueness proof*

Here a proof is given for the global uniqueness of the equilibrium point in the physically
accessible region of composition space following a procedure given in a little known paper by
the great Russian physicist Zeldovich. 4 The proof follows the basic outline of Zeldovich,
but the notation will be consistent with the present development. Some simplifications were
available to Zeldovich but not employed by him. For further background see Powers and
Paolucci. 5
5.2.3.1 Isothermal, isochoric case

Consider a mixture of ideal gases in a closed fixed volume V at fixed temperature T . For
such a system, the canonical equilibration relation is given by Eq. (3.406), dA|T,V 0. So A
must be always decreasing until it reaches a minimum. Consider then A. First, combining
Eqs. (3.300) and (3.301), one finds
A = P V + G. (5.141)
Now from Eq. (3.409) one can eliminate G to get
N
X
A = P V + ni i . (5.142)
i=1
PN
From the ideal gas law, P V = nRT , and again with n = i=1 ni , one gets
N
X
A = nRT + ni i , (5.143)
i=1
N
X N
X
Pi
= ni RT + ni oT,i + RT ln , (5.144)
i=1 i=1
Po
N
X
oT,i ni P
= RT ni 1 + ln , (5.145)
i=1
RT nPo
XN
oT,i RT
= RT ni 1 + ln ni . (5.146)
i=1
RT Po V
4
Zeldovich, Ya. B., 1938, A Proof of the Uniqueness of the Solution of the Equations for the Law of
Mass Action, Zhurnal Fizicheskoi Khimii, 11: 685-687.
5
Powers, J. M., and Paolucci, S., 2008, Uniqueness of Chemical Equilibria in Ideal Mixtures of Ideal
Gases, American Journal of Physics, 76(9): 848-855.
For convenience, define, for this isothermal isochoric problem no , the total number of moles
at the reference pressure, which for this isothermal isochoric problem, is a constant:
Po V
no . (5.147)
RT
So
N
X oT,i n
i
A = RT ni 1 + ln o
. (5.148)
i=1
RT n
Now recall that the atomic element conservation, Eq. (5.51), demands that
N
X
li (nio ni ) = 0, l = 1, . . . , L. (5.149)
i=1
As defined earlier, li is the number of atoms of element l in species i; note that li 0.

It is described by a L N non-square
PN matrix, typically of full rank, L. Defining the initial
number of moles of element l, l = i=1 li nio , one can rewrite Eq. (5.149) as
N
X
li ni = l , l = 1, . . . , L. (5.150)
i=1
Equation (5.150) is generally under-constrained, and one can find solutions of the form

n1 n1o D11 D12 D1 N L
n2 n2o D21 D D
. = . + . 1 + .22 2 + . . . + 2 .N L N L . (5.151)
.. .. .. .. ..
nN nN o DN 1 DN 2 DN N L
Here, Dik represents a dimensionless component of a matrix of dimension N (N L). Here,

in contrast to earlier analysis, i is interpreted as having the dimensions of kmole. Each of
the N L column vectors of the matrix whose components are Dik has length N and lies in
the right null space of the matrix whose components are li . That is,
N
X
li Dik = 0, k = 1, . . . , N L, l = 1, . . . , L. (5.152)
i=1
There is an exception to this rule when the left null space of ij is of higher dimension than
the right null space of li . In such a case, there are quantities conserved in addition to the
elements as a consequence of the form of the reaction law. The most common exception
occurs when each of the reactions also conserves the number of molecules. In such a case,
there will be N L 1 free variables, rather than N L. One can robustly form Dij from
the set of independent column space vectors of ij . These vectors are included in the right
P
null space of li since N i=1 li ij = 0.
One can have Dik (, ). Each of these is a function of li . In matrix form, one
can say

n1 n1o D11 D12 . . . D1 N L 1
n2 n2o D21 D22 . . . D2 N L 2
. = . + . .. .. .. . . (5.153)
.. .. .. . . . ..
nN nN o DN 1 DN 2 . . . DN N L N L
In index form, this becomes

N
X L
ni = nio + Dik k , i = 1, . . . , N. (5.154)
k=1
It is also easy to show that the N L column space vectors in Dik are linear combinations
of N L column space vectors that span the column space of the rank-deficient N J
components of ij . The N values of ni are uniquely determined once N L values of k
are specified. That is, a set of independent k , k = 1, . . . , N L, is sufficient to describe
the system. This insures the initial element concentrations are always maintained. One
can also develop a J-reaction generalization of the single reaction Eq. (4.99). Let kj ,
k = 1, . . . , N L, j = 1, . . . , J, be the extension of k . Then the appropriate generalization
of Eq. (4.99) is
N
X L
ij = Dik kj . (5.155)
k=1
In Gibbs notation, one would say

= D . (5.156)
One can find the matrix by the following operations
D = , (5.157)
D D = DT ,
T
(5.158)
1
= DT D DT . (5.159)
This calculation has only marginal utility.

Returning to the primary exercise, note that one can form the partial derivative of
Eq. (5.154):
N L
ni X k
= Dik , i = 1, . . . , N; p = 1, . . . , N L, (5.160)
p j k=1
p j
N
X L
= Dik kp , i = 1, . . . , N; p = 1, . . . , N L, (5.161)
k=1
= Dip , i = 1, . . . , N; p = 1, . . . , N L. (5.162)
Here kp is the Kronecker delta function.

Next, return to consideration of Eq. (5.146). It is sought to minimize A while holding
T and V constant. The only available variables are ni , i = 1, . . . , N. These are not fully
independent, but they are known in terms of the independent p , p = 1, . . . , N L. So one
can find an extremum of A by differentiating it with respect to each of the p and setting
each derivative to zero:
o !
A XN
ni T,i ni n n
i i
= RT 1 + ln o + ni ln o = 0,
p j ,T,V i=1
p

j
RT p

j
n p n
p = 1, . . . , N L, (5.163)
o !
N
X ni T,i ni n XN
nq n
i i
= RT 1 + ln + n ln = 0,
p j RT p j
i
i=1
no q=1
p n q no
p = 1, . . . , N L, (5.164)
!
N
X oT,i n N
X 1
i
= RT Dip 1 + Dip ln o
+ ni Dqp iq = 0,
i=1
RT n q=1
ni
p = 1, . . . , N L, (5.165)
N
X n
oT,i i
= RT Dip 1 + Dip ln + Dip = 0,
i=1
RT no
p = 1, . . . , N L, (5.166)
N
X n
oT,i i
= RT Dip + ln = 0, p = 1, . . . , N L. (5.167)
i=1
RT no
Following Zeldovich, one then rearranges Eq. (5.167) to define the equations for equilibrium:
N
X n N
X Dip oT,i
i
Dip ln o
= , p = 1, . . . , N L, (5.168)
i=1
n i=1
RT
XN n Dip XN
Dip oT,i
i
ln = , p = 1, . . . , N L, (5.169)
i=1
no i=1
RT
N
ni Dip
Y N
X Dip oT,i
ln o
= , p = 1, . . . , N L, (5.170)
i=1
n i=1
RT
N N L
! !Dip N
Y X 1 X Dip oT,i
ln nio + Dik k = , p = 1, . . . , N L. (5.171)
i=1 k=1
no i=1
RT
Equation (5.171) forms N L equations in the N L unknown values of p and can be solved
by Newton iteration. Note that as of yet, no proof exists that this is a unique solution. Nor
is it certain whether or not A is maximized or minimized at such a solution point.
One also notices that the method of Zeldovich is consistent with a more rudimentary
form. Rearrange Eq. (5.167) to get
N n
A X i
o
= Dip T,i + RT ln o = 0, p = 1, . . . , N L, (5.172)
p j ,T,V i=1
n
XN
o ni P
= Dip T,i + RT ln , (5.173)
i=1
nPo
XN
o Pi
= Dip T,i + RT ln , (5.174)
i=1
Po
N
X
= i Dip . (5.175)
i=1
Now, Eq. (5.175) is easily found via another method. Recall Eq. (3.304), and then operate
on it in an isochoric, isothermal limit, taking derivative with respect to p , p = 1, . . . , N L:
N
X
dA = SdT P dV + i dni , (5.176)
i=1
N
X
dA|T,V = i dni , (5.177)
i=1
N
A X ni
= i , (5.178)
p j ,T,V i=1
p
XN
= i Dip . (5.179)
i=1
One can determine whether such a solution, if it exists, is a maxima or minima by

examining the second derivative, given by differentiating Eq. (5.167). We find the Hessian,
H, to be
2A XN
n
i
H= = RT Dip ln o , (5.180)
j p i=1
j n
XN
o
= RT Dip ln ni ln n , (5.181)
i=1
j j
N
X 1 ni
= RT Dip , (5.182)
i=1
ni j
XN
Dip Dij
= RT . (5.183)
i=1
ni
Here j = 1, . . . , N L. Scaling each of the rows of Dip by a constant does not affect the rank.

So we can say that the N (N L) matrix whose entries are Dip / ni has rank N L, and
consequently the Hessian H, of dimension (N L) (N L), has full rank N L, and is
symmetric. It is easy to show by means of singular value decomposition, or other methods,
that the eigenvalues of a full rank symmetric square matrix are all real and non-zero.
Now consider whether 2 A/j p is positive definite. By definition, it is positive definite
if for an arbitrary vector zi of length N L with non-zero norm that the term T V , defined
below, be always positive:
N
X L N
X L
2A
T V = zj zp > 0. (5.184)
j=1 p=1
j p
Substitute Eq. (5.183) into Eq. (5.184) to find

N
X L N
X L N
X Dip Dij
T V = RT zj zp , (5.185)
j=1 p=1 i=1
ni
N
X N
X L N
X L
1
= RT Dip Dij zj zp , (5.186)
i=1
ni j=1 p=1
XN N L N L
1 X X
= RT Dij zj Dip zp , (5.187)
i=1
ni j=1 p=1
N N L
! N L !
X 1 X X
= RT Dij zj Dip zp . (5.188)
i=1
ni j=1 p=1
Define now
N
X L
yi Dij zj , i = 1, . . . , N. (5.189)
j=1
This yields
N
X y2 i
T V = RT . (5.190)
i=1
ni
Now, to restrict the domain to physically accessible space, one is only concerned with ni 0,
T > 0, so for arbitrary yi , one finds T V > 0, so one concludes that the second mixed partial
of A is positive definite globally.
We next consider the behavior of A near a generic point in the physical space where
A = A. If we let d represent , we can represent the Helmholtz free energy by the
Taylor series
1
A() = A + d T A + dT H d + . . . , (5.191)
2
where A and H are evaluated at . Now if = eq , where eq denotes an equilibrium point,

A = 0, A = Aeq , and
1
A() Aeq = d T H d + . . . . (5.192)
2
Because H is positive definite in the entire physical domain, any isolated critical point will
be a minimum. Note that if more than one isolated minimum point of A were to exist in
the domain interior, a maximum would also have to exist in the interior, but maxima are
not allowed by the global positive definite nature of H. Subsequently, any extremum which
exists away from the boundary of the physical region must be a minimum, and the minimum
is global.
Global positive definiteness of H alone does not rule out the possibility of non-isolated
multiple equilibria, as seen by the following analysis. Because it is symmetric, H can be
orthogonally decomposed into H = QT Q, where Q is an orthogonal matrix whose
columns are the normalized eigenvectors of H. Note that QT = Q1 . Also is a diagonal
matrix with real eigenvalues on its diagonal. We can effect a volume-preserving rotation of
axes by taking the transformation dw = Q d; thus, d = QT dw. Hence,
1 1 1
AAeq = (QT dw)T HQT dw = dwT QQT QQT dw = dwT dw. (5.193)
2 2 2
The application of these transformations gives in the neighborhood of equilibrium the quadratic
form
N L
eq 1X
AA = p (dwp )2 . (5.194)
2 p=1
For A to be a unique minimum, p > 0. If one or more of the p = 0, then the minimum
could be realized on a line or higher dimensional plane, depending on how many zeros are
present. The full rank of H guarantees that p > 0. For our problem the unique global
minimum which exists in the interior will exist at a unique point.
Lastly, one must check the boundary of the physical region to see if it can form an
extremum. Near a physical boundary given by nq = 0, one finds that A behaves as
X N n
T,i i
lim A RT ni 1 + ln o finite. (5.195)
nq 0
i=1
RT n
The behavior of A itself is finite because limnq 0 nq ln nq = 0, and the remaining terms in
the summation are non-zero and finite.
The analysis of the behavior of the derivative of A on a boundary of nq = 0 is more
complex. We require that nq 0 for all N species. The hyperplanes given by nq = 0 define
a closed physical boundary in reduced composition space. We require that changes in nq
which originate from the surface nq = 0 be positive. For such curves we thus require that
near nq = 0, perturbations dk be such that
N
X L
dnq = Dqk dk > 0. (5.196)
k=1
We next examine changes in A in the vicinity of boundaries given by nq = 0. We will restrict

our attention to changes which give rise to dnq > 0. We employ Eq. (5.172) and find that
N L N n NXL
X A X i
o
dA = dp = + RT ln Dip dp . (5.197)
p=1
p T,V,j6=p i=1
i
no
p=1
On the boundary given by nq = 0, the dominant term in the sum is for i = q, and so on this
boundary
n NX
L
q
lim dA = RT ln o Dqp dp . (5.198)
nq 0 n p=1
| {z }
>0
The term identified by the brace is positive because of Eq. (5.196). Because R and T > 0,
we see that as nq moves away from zero into the physical region, changes in A are large and
negative. So the physical boundary can be a local maximum, but never a local minimum in A.
Hence, the only admissible equilibrium is the unique minimum of A found from Eq. (5.171);
this equilibrium is found at a unique point in reduced composition space.
5.2.3.2 Isothermal, isobaric case

Next, consider the related case in which T and P are constant. For such a system, the
canonical equilibration relation is given by Eq. (3.407) holds that dG|T,P 0. So G must
be decreasing until it reaches a minimum. So consider G, which from Eq. (3.409) is
N
X
G = ni i , (5.199)
i=1
XN
Pi
= ni oT,i + RT ln , (5.200)
i=1
Po
XN
ni P
= ni + RT lnoT,i , (5.201)
i=1
nPo
N o !!
X T,i P ni
= RT ni + ln + ni ln PN . (5.202)
i=1
RT Po k=1 nk
Next, differentiate with respect to each of the independent variables p for p = 1, . . . , N L:

N !!!
G X ni oT,i P ni
= RT + ln + ni ln PN , (5.203)
p i=1
p RT Po p k=1 nk
X N
ni oT,i P
= RT + ln
i=1
p RT Po
N
!!!
X nq ni
+ ni ln PN , (5.204)
q=1
p nq k=1 nk
X N o
T,i P
= RT Dip + ln
i=1
RT Po
N
!!!!
X ni ni
+ Dqp PN + iq 1 + ln, PN (5.205)
q=1 k=1 nk k=1 nk
!
X N
oT,i P n n X N
i i
= RT Dip + ln + 1 + ln Dqp , (5.206)
i=1
RT Po n n q=1
X N o n
T,i P i
= RT Dip + ln + 1 + ln
i=1
RT Po n
N
X N
ni X
RT Dqp , (5.207)
i=1
n q=1
N
X n
oT,i P i
= RT Dip + ln + 1 + ln
i=1
RT Po n
XN N
X ni
RT Dqp , (5.208)
q=1 i=1
n
N
X n N
X
oT,i P i
= RT Dip + ln + 1 + ln RT Dqp , (5.209)
i=1
RT Po n q=1
N
X n N
X
oT,i P i
= RT Dip + ln + 1 + ln RT Dip , (5.210)
i=1
RT Po n i=1
XN n
P oT,i i
= RT Dip + ln + ln , (5.211)
i=1
RT Po n
XN
o ni P
= Dip T,i + RT ln , (5.212)
i=1
nPo
XN
o ni P
= Dip T,i + RT ln , (5.213)
i=1
nPo
N
X
= i Dip , p = 1, . . . , N L. (5.214)
i=1
Note this simple result is entirely consistent with a result

Pthat could have been deduced by
commencing with the alternative Eq. (3.411), dG|T,P = N i=1 i dni . Had this simplification
been taken, one could readily deduce that

N
G X ni
= i , p = 1, . . . N L, (5.215)
p j i=1
p
N
X
= i Dip , p = 1, . . . N L. (5.216)
i=1
Now, to equilibrate, one sets the derivatives to zero to get

N
X
ni P
Dip oT,i
+ RT ln = 0, p = 1, . . . , N L, (5.217)
i=1
nPo
XN Dip N
X
ni P oT,i
ln = Dip , (5.218)
i=1
nPo i=1
RT
YN Dip XN
ni P oT,i
ln = Dip , (5.219)
i=1
nPo i=1
RT
PN L Dip
YN nio + k=1 Dik k P N
X oT,i

ln PN PN L = Dip
RT
,
i=1 q=1 nqo + k=1 Dqk k Po i=1
p = 1, . . . , N L. (5.220)
These N L non-linear algebraic equations can be solved for the N L unknown values of p
via an iterative technique. One can extend the earlier analysis to show that the equilibrium
is unique in the physically accessible region of composition space.
We can repeat our previous analysis to show that this equilibrium is unique in the phys-
ically accessible region of composition space. By differentiating Eq. (5.216) it is seen that
2G X N
n
i
= RT Dip ln , (5.221)
j p i=1
j n
N N
!
X X
= RT Dip ln ni Dip ln n , (5.222)
i=1
j i=1
j
N N
!
X 1 ni X 1 n
= RT Dip Dip , (5.223)
i=1
ni j i=1
n j
N N N
!!
X 1 ni X 1 X
= RT Dip Dip nq , (5.224)
i=1
ni j i=1
n j q=1
N N N
!!
X 1 ni X X 1 nq
= RT Dip Dip , (5.225)
i=1
ni j i=1 q=1
n j
N N N
!
X Dip Dij 1 XX
= RT Dip Dqj . (5.226)
i=1
ni n i=1 q=1
Next consider the sum

N L N L N L N L XN N N
!
X X 2G X X Dip Dij 1 XX
T P = zj zp = RT Dip Dqj zj zp . (5.227)
j=1 p=1
j p j=1 p=1 i=1
ni n i=1 q=1
We use Eq. (5.189) and following a long series of calculations, reduce Eq. (5.227), to the
positive definite form
N N r r 2
RT X X nj ni
T P = yi yj > 0. (5.228)
n i=1 j=i+1 ni nj
It is easily verified by direct expansion that Eqs. (5.227) and (5.228) are equivalent.
On the boundary ni = 0, and as for the isothermal-isochoric case, it can be shown that
dG for changes with dni > 0. Thus, the boundary has no local minimum, and we
can conclude that G is minimized in the interior and the minimum is unique.
5.2.3.3 Adiabatic, isochoric case

One can extend Zeldovichs proof to other sets of conditions. For example, consider a case
which is isochoric and isoenergetic. This corresponds to a chemical reaction in an fixed
volume which is thermally insulated. In this case, one operates on Eq. (3.407):
N
X
dE = P |{z}
|{z} dV +T dS + i dni , (5.229)
=0 =0 i=1
N
X
0 = T dS + i dni , (5.230)
i=1
N
X
1
dS = i dni , (5.231)
T i=1
XN N
X L
1
= i Dik dk , (5.232)
T i=1 k=1
N
XXN L
1
= i Dik dk , (5.233)
T i=1 k=1
XN NXL
S 1 k
= i Dik , (5.234)
j T i=1 k=1
j
N N L
1 XX
= Dik kj , (5.235)
T i=1 k=1 i
N
1X
= Dij , (5.236)
T i=1 i
N
1X o Pi
= T,i + RT ln Dij , (5.237)
T i=1 Po
N
1X o ni P
= T,i + RT ln Dij , (5.238)
T i=1 nPo
N
1X o ni RT To
= T,i + RT ln Dij , (5.239)
T i=1 Po V To
N
1X o T ni RTo
= T,i + RT ln + RT ln Dij , (5.240)
T i=1 To Po V

N o
X T,i T ni RTo
= + R ln +R ln Dij , (5.241)
T T o P o V
i=1 | {z }
i (T )
N
X
ni RTo
= i (T ) + R ln Dij , (5.242)
i=1
P o V
XN
2S i (T ) T R ni
= + Dij , (5.243)
k j i=1
T k ni k
XN
i (T ) T R
= + Dik Dij . (5.244)
i=1
T k ni
Now, consider dT for the adiabatic system.
N
X
E = Eo = nq eq (T ), (5.245)
q=1

XN
eq
dE = 0 = nq dT + eq dnq
, (5.246)
T
q=1 |{z}
=cvq
N
X
0 = (nq cvq dT + eq dnq ) , (5.247)
q=1
PN
q=1 eq dnq
dT = PN , (5.248)
q=1 nq cvq
N N L
1 X X
= eq Dqp dp , (5.249)
ncv q=1 p=1
N N L
1 XX
= eq Dqp dp , (5.250)
ncv q=1 p=1
N N L
T 1 XX p
= eq Dqp , (5.251)
k ncv q=1 p=1 k
N N L
1 XX
= eq Dqp pk , (5.252)
ncv q=1 p=1
N
1 X
= eq Dqk . (5.253)
ncv q=1
Now return to Eq. (5.244), using Eq. (5.253) to expand:

N N
!
2S X i (T ) 1 X R
= eq Dqk Dik Dij , (5.254)
k j i=1
T ncv q=1 ni
XN N N
i (T ) 1 X X R
= eq Dqk Dij Dik Dij , (5.255)
i=1
T ncv q=1 i=1
ni
N N N
1 X X i (T ) X Dik Dij
= eq Dqk Dij R . (5.256)
ncv i=1 q=1 T i=1
n i
Now consider the temperature derivative of i (T ), where i (T ) is defined in Eq. (5.241):

oT,i T
i + R ln , (5.257)
T To
di oT,i 1 doT,i R
= 2 + + . (5.258)
dT T T dT T
Now
o
oT,i = hT,i T soT,i , (5.259)
Z T Z T !
o c Pi (
T )
= hTo ,i + cP i (T)dT T soTo ,i + dT , (5.260)
To To T
| {z } | {z }
o
=hT,i =soT,i
Z T
doT,i o cP i (T)
= cP i sTo ,i d T cP i , (5.261)
dT To T
Z T
o cP i (T)
= sTo ,i dT , (5.262)
To T
= soT,i . (5.263)
Note that Eq. 5.263 is a special case of the Gibbs equation given by Eq. 3.317. With this,
one finds that Eq. 5.258 reduces to
di oT,i soT,i R
= 2 + , (5.264)
dT T T T
1 o o

= 2 T,i + T sT,i RT , (5.265)
T
1
= 2 g oT,i + T soT,i RT , (5.266)
T
1 o o o
= 2 hT,i T sT,i +T sT,i RT , (5.267)
T | {z }
=goT,i
1 o
= 2 hT,i RT , (5.268)
T
1
= 2 ei + Pi v i RT , (5.269)
T
1
= 2 ei + RT RT , (5.270)
T
1
= 2 ei . (5.271)
T
So substituting Eq. (5.271) into Eq. (5.256), one gets
N N N
2S 1 XX X Dik Dij
= ei eq D qk D ij R . (5.272)
k j ncv T 2 i=1 q=1 i=1
n i
Next, as before, consider the sum

N
X L N
X L N L N L N N
2S 1 X X XX
zk zj = ei eq Dqk Dij zk zj
k=1 j=1
k j ncv T 2 k=1 j=1 i=1 q=1
N
X L N
X N
L X
Dik Dij
R zk zj , (5.273)
k=1 j=1 i=1
ni
N N N L N L
1 XX X X
= eq Dqk zk ei Dij zj
ncv T 2 i=1 q=1 k=1 j=1
N N
X X L N
X L
Dik zk Dij zj
R , (5.274)
i=1 k=1 j=1
ni
N N N L
! N L
!
1 XX X X
= eq Dqk zk ei Dij zj
ncv T 2 i=1 q=1 k=1 j=1
N N L
! N L
!
X 1 X X
R Dik zk Dij zj , (5.275)
i=1
ni j=1
k=1
N N L
! N N L
!
1 X X X X
= ei Dij zj eq Dqk zk
ncv T 2 i=1 j=1 q=1 k=1
N N L
! N L
!
X 1 X X
R Dik zk Dij zj , (5.276)
i=1
ni k=1 j=1
N N L
!2 N N L
!2
1 X X X 1 X
= ei Dij zj R Dik zk (5.277)
,
ncv T 2 i=1 j=1 i=1
ni k=1
N
!2 N
1 X X yi2
= ei yi R . (5.278)
ncv T 2 i=1 i=1
ni
Since cv > 0, T > 0, R > 0, ni > 0, and the other terms are perfect squares, it is obvious
that the second partial derivative of S < 0; consequently, critical points of S represent a
maximum. Again near the boundary of the physical region, S ni ln ni , so limni 0 S 0.
From Eq. (5.253), there is no formal restriction on the slope at the boundary. However, if a
critical point is to exist in the physical domain in which the second derivative is guaranteed
negative, the the slope at the boundary must be positive everywhere. This combines to
guarantee that if a critical point exists in the physically accessible region of composition
space, it is unique.
5.2.3.4 Adiabatic, isobaric case
A similar proof holds for the adiabatic-isobaric case. Here the appropriate Legendre trans-
formation is H = E + P V , where H is the enthalpy. We omit the details, which are similar
to those of previous sections, and find a term which must be negative semi-definite, HP :
N
X L N
X L
2S
HP = zk zj (5.279)
k=1 j=1
k j
N
!2 N N r r 2
1 X RX X nj ni
= hi yi yi yj . (5.280)
ncP T 2 i=1 n i=1 j=i+1 ni nj
5.3. CONCISE REACTION RATE LAW FORMULATIONS 221
Because cP > 0 and ni 0, the term involving hi yi is a perfect square, and the term
multiplying R is positive definite for the same reasons as discussed before. Hence, HP < 0,
and the Hessian matrix is negative definite.
5.3 Concise reaction rate law formulations

One can employ notions developed in the Zeldovich uniqueness proof to obtain a more effi-
cient representation of the reaction rate law for multiple reactions. There are two important
cases: 1) J (N L); this is most common for large chemical kinetic systems, and 2)
J < (N L); this is common for simple chemistry models.
The species production rate is given by Eq. (5.14), which reduces to
J
di 1 X dj
= ij , i = 1, . . . , N. (5.281)
dt V j=1 dt
Now differentiating Eq. (5.154), one obtains

N
X L
dni = Dik dk , i = 1, . . . , N. (5.282)
k=1
Comparing then Eq. (5.282) to Eq. (5.20), one sees that

J
X N
X L
ij dj = Dik dk , i = 1, . . . , N, (5.283)
j=1 k=1
J N L
1 X 1 X
ij dj = Dik dk , i = 1, . . . , N, (5.284)
V j=1 V k=1
(5.285)
5.3.1 Reaction dominant: J (N L)

Consider first the most common case in which J (N L). One can say the species
production rate is given
N L J
di 1 X dk X
= Dik = ij rj , i = 1, . . . , N. (5.286)
dt V k=1 dt j=1
One would like to invert for dk /dt. However, Dik is non-square and has
PNand solve directly P
no inverse. But since i=1 li Dip = 0, and N i=1 li ij = 0, L of the equations N equations
in Eq. (5.286) are redundant.
At this point, it is more convenient to go to a Gibbs vector notation, where there is an

obvious correspondence between the bold vectors and the indicial counterparts:
d 1 d
= D = r, (5.287)
dt V dt
d
DT D = V DT r, (5.288)
dt
d
= V (DT D)1 DT r. (5.289)
dt
Because of the L linear dependencies, there is no loss of information in this matrix projection.
This system of N L equations is the smallest number of differential equations that can be
solved for a general system in which J > (N L).
Lastly, one recovers the original system when forming
d
D = V D (DT D)1 DT r. (5.290)
dt | {z }
=P
Here the N N projection matrix P is symmetric, has norm of unity, has rank of N L,
has N L eigenvalues of value unity, and L eigenvalues of value zero. And, while in general,
application of a projection matrix to r loses some of the information in r, because of
the nature of the linear dependencies, no information is lost in Eq. (5.290) relative to the
original Eq. (5.287).
Note finally that it can be shown that D, of dimension N (N L), and , of dimension
N J, share the same column space, which is of dimension (N L); consequently, both
matrices map vectors into the same space.
5.3.2 Species dominant: J < (N L)

Next consider the case in which J < (N L). This often arises in models of simple chemistry,
for example one- or two-step kinetics.
The fundamental reaction dynamics are most concisely governed by the J equations
which form
1 d
= r. (5.291)
V dt
However, r is a function of the concentrations; one must therefore recover as a function
of reaction progress . In vector form, Eq. (5.281) is written as
d 1 d
= . (5.292)
dt V dt
Take as an initial condition that the reaction progress is zero at t = 0 and that there are an
appropriate set of initial conditions on the species concentrations :
= 0, t = 0, (5.293)
= o , t = 0. (5.294)
5.4. ONSAGER RECIPROCITY* 223
Then, since is a constant, Eq. (5.292) is easily integrated. After applying the initial
conditions, Eq. (5.294), one gets
1
= o + . (5.295)
V
Last, if J = (N L), either approach yields the same number of equations, and is equally
concise.
5.4 Onsager reciprocity*

There is a powerful result from physical chemistry which speaks to how systems behave near
equilibrium. The principle was developed by Lars Onsager in the early twentieth century,
and is known as the reciprocity principle. Where it holds, one can guarantee that on their
approach to equilibrium, systems will approach the equilibrium in a non-oscillatory manner.
It will be illustrated here.
Note that Eq. (5.17) can be written as
1 dj
= rj = Rj Rj , j = 1, . . . , J, (5.296)
V dt

Rj
= Rj 1 , j = 1, . . . , J. (5.297)
Rj
Note further that the definition of affinity gives

j Rj
exp = , j = 1, . . . , J. (5.298)
RT Rj
Therefore, one can say

1 dj j
= rj = Rj 1 exp , j = 1, . . . , J. (5.299)
V dt RT
Now, as each reaction comes to equilibrium, one finds that j 0, so a Taylor series
expansion of rj yields

1 dj j
= rj Rj 1 1 + ... , j = 1, . . . , J, (5.300)
V dt RT
j
Rj , j = 1, . . . , J. (5.301)
RT
Note that Rj > 0, while j can be positive or negative. Note also there is no summation on
j. Take now the matrix R to be the diagonal matrix with Rj populating its diagonal:

R1 0 . . . 0
0 R2 . . . 0
R ... .. .. .. . (5.302)
. . .
0 0 0 RJ
Adopt the vector notation

1
r= R . (5.303)
RT
Now the entropy production is given for multi-component systems by Eq. (5.46):
J
dS V X
= j rj 0. (5.304)
dt E,V T j=1
Cast this entropy inequality into Gibbs notation:

dS V T
= r 0. (5.305)
dt E,V T
Now consider the definition of affinity, Eq. (5.10), in Gibbs notation:
= T . (5.306)
Now T is of dimension J N with rank typically N L. Because T is typically not of
full rank, one finds only N L of the components of to be linearly independent. When
one recalls that T maps vectors into the column space of T , one recognizes that can
be represented as

= C . (5.307)
Here C is a J (N L)-dimensional matrix of full rank, N L, whose N L columns are
populated by the linearly independent vectors which form the column space of T , and

is a column vector of dimension (N L) 1. If J N L, one can explicitly solve for ,
starting by operating on both sides of Eq. (5.307) by CT :
CT = CT C , (5.308)
T
C C = C ,T
(5.309)
1
= CT C
CT , (5.310)
1
= CT C CT T , (5.311)
1
=C = C CT C CT T . (5.312)
| {z }
B
Here, in the recomposition of , one can employ the J J symmetric projection matrix B,
which has N L eigenvalues of unity and J (N L) eigenvalues of zero. The matrix B
has rank N L, and is thus not full rank.
Substitute Eqs. (5.303, 5.307) into Eq. (5.305) to get
T
= =
z }| {
dS V T 1 z }| {
= C R C 0, (5.313)
dt E,V T RT
| {z }
=r
! !
V T T
= C R C}
| {z 0. (5.314)
R T L
T
Since each of the entries of the diagonal R are guaranteed positive semi-definite in the
physical region of composition space, the entropy production rate near equilibrium is also
positive semi-definite. The constant square matrix L, of dimension (N L) (N L), is
given by
L CT R C. (5.315)
The matrix L has rank N L and is thus full rank. Because R is diagonal with positive semi-
definite elements, L is symmetric positive semi-definite. Thus its eigenvalues are guaranteed
to be positive and real. Note that off-diagonal elements of L can be negative, but that
the matrix itself remains positive semi-definite. Onsager reciprocity simply demands that
near equilibrium, the linearized version of the combination of the thermodynamic forces
(here the affinity ) and fluxes (here the reaction rate r) be positive semi-definite. Upon
linearization, one should always be able to find a positive semi-definite matrix associated
with the dynamics of the approach to equilibrium. Here that matrix is L, and by choices
made in its construction, it has the desired properties.
One can also formulate an alternative version of Onsager reciprocity using the projection
matrix B, which from Eq. (5.312), is
1
B C CT C CT . (5.316)
With a series of straightforward substitutions, it can be shown that the entropy production
rate given by Eq. (5.314) reduces to

dS V T T
= B
| R B}
{z 0. (5.317)
dt E,V R T T
L
Here, an alternative symmetric positive semi-definite matrix L, of dimension J J and rank

N L, has been defined as
L BT R B. (5.318)
One can also express the entropy generation directly in terms of the chemical potential
rather than the affinity by defining the J N matrix S as
S B T , (5.319)
= C (CT C)1 C T . (5.320)
The matrix S has rank N L and thus is not full rank. With a series of straightforward
substitutions, it can be shown that the entropy production rate given by Eq. (5.317) reduces
to
dS V T T
= R S}
|S {z 0. (5.321)
dt E,V R T T
L
Here, an alternative symmetric positive semi-definite matrix L, of dimension N N and

rank N L, has been defined as
L ST R S. (5.322)
Example 5.4
Find the matrices associated with Onsager reciprocity for the reaction mechanism given by
H2 + O2 2OH, (5.323)
H2 + OH H + H2 O, (5.324)
H + O2 O + OH, (5.325)
H2 + O H + OH, (5.326)
H +H H2 , (5.327)
2OH O + H2 O, (5.328)
H2 H + H, (5.329)
O2 O + O, (5.330)
H + OH H2 O. (5.331)
Here there are N = 6 species (H, H2 , O, O2 , OH, H2 O), composed of L = 2 elements (H, O), reacting
in J = 9 reactions. Here J N L, so the analysis of this section can be performed.
Take species i = 1 as H, i = 2 as H2 ,. . ., i = N = 6 as H2 O. Take element l = 1 as H and element
l = L = 2 as O. The full rank stoichiometric matrix , of dimension 2 6 = L N and rank 2 = L, is

1 2 0 0 1 2
= . (5.332)
0 0 1 2 1 1
The rank-deficient matrix of stoichiometric coefficients , of dimension 69 = N J and rank 4 = N L

is

0 1 1 1 2 0 2 0 1
1 1 0 1 1 0 1 0 0

0 0 1 1 0 1 0 2 0
= . (5.333)
1 0 1 0 0 0 0 1 0

2 1 1 1 0 2 0 0 1
0 1 0 0 0 1 0 0 1
Check for stoichiometric balance:

0 1 1 1 2 0 2 0 1

1 1 0 1 1 0 1 0 0

1 2 0 0 1 2 0 0 1 1 0 1 0 2 0
= , (5.334)
0 0 1 2 1 1 1 0 1 0 0 0 0 1 0

2 1 1 1 0 2 0 0 1
0 1 0 0 0 1 0 0 1

0 0 0 0 0 0 0 0 0
= . (5.335)
0 0 0 0 0 0 0 0 0
So the number and mass of every element is conserved in every reaction; every vector in the column
space of is in the right null space of .
The detailed version of the reaction kinetics law is given by
d
= r, (5.336)
dt

r1
r2
0 1 1 1 2 0 2 0 1
r
1 1 0 1 1 0 1 0 0 3
r4
0 0 1 1 0 1 0 2 0
= r , (5.337)
1 0 1 0 0 0 0 1 0 5
r6
2 1 1 1 0 2 0 0 1
r7
0 1 0 0 0 1 0 0 1
r8
r9

r2 r3 + r4 2r5 + 2r7 r9
r1 r2 r4 + r5 r7

r3 r4 + r6 + 2r8
= . (5.338)
r1 r3 r8

2r1 r2 + r3 + r4 2r6 r9
r2 + r6 + r9
The full rank matrix D, of dimension 6 4 = N (N L) and rank 4 = N L, is composed of

vectors in the right null space of . It is non-unique, as linear combinations of right null space vectors
suffice. It is equivalently composed by casting the N L linearly independent vectors of the column
space of in its columns. Recall that some of the columns of are linearly dependent. In the present
example, the first N L = 4 column vectors of happen to be linearly dependent, and thus will not
suffice. Other sets are not; the last N L = 4 column vectors of happen to be linearly independent
and thus suffice for the present purposes. Take then

0 2 0 1
0 1 0 0

1 0 2 0
D= . (5.339)
0 0 1 0

2 0 0 1
1 0 0 1
It is easily verified by direct substitution that D is in the right null space of :

0 2 0 1
0 1 0 0

1 2 0 0 1 2 1 0 2 0 0 0 0 0
D= = . (5.340)
0 0 1 2 1 1 0 0 1 0 0 0 0 0

2 0 0 1
1 0 0 1
Since = 0 and D = 0, one concludes that the column spaces of both and D are one and the
same.
The non-unique concise version of the reaction kinetics law is given by
d
= V (DT D)1 DT r, (5.341)
dt
1
0 2 0 1

0 0 1 0 2 1 0 1 0 0 0 0 1 0 2 1

2 1 0 0 0 0 1 0 2 0 2 1 0 0 0 0
= V
0 0 2 1 0 0 0 0 1 0 0 0 2 1 0 0

1 0 0 0 1 1 2 0 0 1 1 0 0 0 1 1
1 0 0 1

r1
r2
0 1 1 1 2 0 2 0 1
r
1 1 0 1 1 0 1 0 0 3
r4
0 0 1 1 0 1 0 2 0
r , (5.342)
1 0 1 0 0 0 0 1 0 5
r6
2 1 1 1 0 2 0 0 1
r7
0 1 0 0 0 1 0 0 1
r8
r9

2r1 r3 r4 + r6
r + r2 + r4 r5 + r7
= V 1 . (5.343)
r1 + r3 + r8
2r1 + r2 + r3 + r4 + r9
The rank-deficient projection matrix P, of dimension 6 6 = N N and rank 4 = N L, is
1
P = D DT D DT , (5.344)
54 14 3 6 4 11
61 61 61 61 61 61
33 6 12
14 8 22

61 61 61 61 61 61
3 6 51 20 7 4

= 61
6
61
12
61
20
61
21
61
14
61
8 . (5.345)
61 61 61 61 61 61
4 8 7 14 50 15
61 61 61 61 61 61
11 22 4 8 15 35
61 61 61 61 61 61
The projection matrix P has 4 = N L eigenvalues of unity and 2 = L eigenvalues of zero.

The affinity vector , of dimension 9 1 = J 1, is given by
= T , (5.346)
T
0 1 1 1 2 0 2 0 1 1
1 1 0 1 1 0 1 0 0 2

0 0 1 1 0 1 0 2 0 3
= , (5.347)
1 0 1 0 0 0 0 1 0 4

2 1 1 1 0 2 0 0 1 5
0 1 0 0 0 1 0 0 1 6

0 1 0 1 2 0
1 1 0 0 1 1
1
1 0 1 1 1 0

1 1 1 0 1 0 2

= 2 1 0 0 0 0 3 , (5.348)

0 0 1 0 2 1 4
5
2 1 0 0 0 0
6
0 0 2 1 0 0
1 0 0 0 1 1

2 + 4 25
1 + 2 + 5 6

1 3 + 4 5

1 + 2 + 3 5

= 21 2 . (5.349)

3 + 25 6

21 + 2

23 + 4
1 + 5 6
The full rank matrix C, of dimension 9 4 = J (N L) and rank 4 = N L, is composed of

the set of N L = 4 linearly independent column space vectors of T ; thus they also comprise the
N L = 4 linearly independent row space vectors of . It does not matter which four are chosen, so
long as they are linearly independent. In this case, the first four column vectors of T suffice:

0 1 0 1
1 1 0 0

1 0 1 1

1 1 1 0

C = 2 1 0 0 . (5.350)

0 0 1 0

2 1 0 0

0 0 2 1
1 0 0 0
When J > N L, not all of the components of are linearly independent. In this case, one can
of dimension 4 1 = (N L) 1 via
form the reduced affinity vector, ,
=
(CT C)1 CT , (5.351)
1
0 1 0 1
1 1 0 0

1 0 1 1
0 1 1 1 2 0 2 0 1
1 1 1 0
1 1 0 1 1 0 1 0 0
= 2 1 0 0
0 0 1 1 0 1 0 2 0
0 0 1 0
1 0 1 0 0 0 0 1 0
2 1 0 0

0 0 2 1
1 0 0 0

2 + 4 25
1 + 2 + 5 6

1 3 + 4 5
0 1 1 1 2 0 2 0 1
1 + 2 + 3 5
1 1 0 1 1 0 1 0 0
21 2 , (5.352)
0 0 1 1 0 1 0 2 0
3 + 25 6
1 0 1 0 0 0 0 1 0
21 + 2

23 + 4
1 + 5 6

1 5 + 6
2 25 + 26
= . (5.353)
3 + 25 6
4 + 45 26
The rank-deficient projection matrix B, of dimension 9 9 = J J and rank 4 = N L, is found to be
B = C (CT C)1 CT , (5.354)

57 9 31 26 1 17 1 6 8
94 94 94 94 94 94 94 94 94
9 41 5 14 15 27 15 4 26
94 94 94 94 94 94 94 94 94
31 5 35 4 11 1 11 28 6
94 94 94 94 94 94 94 94 94
26 14 4 30 12 16 12 22 2
94 94 94 94 94 94 94 94 94
1 15 11 12 33 3 33 10 18
= 94 94 94 94 94 94 94 94 94 . (5.355)
17 27 1 16 3 43 3 18 24
94 94 94 94 94 94 94 94 94
1 15 11 12 33 3 33 10 18
94 94 94 94 94 94 94 94 94
6 4 28 22 10 18 10 60 14
94 94 94 94 94 94 94 94 94
8 26 6 2 18 24 18 14 44
94 94 94 94 94 94 94 94 94
The projection matrix B has a set of 9 = J eigenvalues, 4 = N L of which are unity, and 5 = J (N L)
of which are zero. One can recover by the operation = C = B T .
The square full rank Onsager matrix L, of dimension 44 = (N L)(N L) and rank 4 = N L,
is given by
L = CT R C, (5.356)

R1 0 0 0 0 0 0 0 0
0 R2 0 0 0 0 0 0 0

0 0 R3 0 0 0 0 0 0
0 1 1 1 2 0 2 0 1
0 0 0 R4 0 0 0 0 0
1 1 0 1 1 0 1 0 0
= 0 0 0 0 R5 0 0 0 0
0 0 1 1 0 1 0 2 0
0 0 0 0 0 R6 0 0 0
1 0 1 0 0 0 0 1 0
0 0 0 0 0 0 R7 0 0

0 0 0 0 0 0 0 R8 0

0 0 0 0 0 0 0 0 R9

0 1 0 1
1 1 0 0

1 0 1 1

1 1 1 0

2 1 0 0 ,

0 0 1 0

2 1 0 0

0 0 2 1
1 0 0 0
(5.357)

R2 + R3 + R4 + 4R5 + 4R7 + R9 R2 R4 2R5 2R7 R3 R4 R3
R2 R4 2R5 2R7 R1 + R2 + R4 + R5 + R7 R4 R1

= .
R3 R4 R4 R3 + R4 + R6 + 4R8

R3 2R8
R3 R1 R3 2R8
R1 + R3 + R8
(5.358)
Obviously L is symmetric, and thus has all real eigenvalues. It is also positive semi-definite. With a
similar effort, one can obtain the alternate rank-deficient square Onsager matrices L and L. Recall that
L, L, and L each have rank N L, while L has the smallest dimension, (N L) (N L), and so
forms the most efficient Onsager matrix.
5.5. IRREVERSIBILITY PRODUCTION RATE* 231
5.5 Irreversibility production rate*

This section will be restricted to isothermal, isochoric reaction. It is known that Gibbs free
energy and irreversibility production rates reach respective minima at equilibrium. Consider
the gradient of the irreversibility production rate in the space of species progress variables
k . As discussed extensively by Prigogene, there is a tendency for systems to relax to a
state which, near equilibrium, minimizes the production rate of irreversibility. Moreover,
in the neighborhood of equilibrium, the irreversibility production rate can be considered a
Lyapunov function.
First, recall Eq. (5.216) for the gradient ofPGibbs free energy with respect to the inde-
pendent species progress variables, G/k = N i=1 i Dik . Now, beginning from Eq. (4.464),
define the differential irreversibility dI as
N
1X
dI = dni . (5.359)
T i=1 i
In terms of an irreversibility production rate, one has

N
dI 1 X dni
=I= . (5.360)
dt T i=1 i dt
Take now the time derivative of Eq. (5.154) to get

N
X L
dni dk
= Dik . (5.361)
dt dt
k=1
Substitute from Eq. (5.361) into Eq. (5.360) to get

N
X N
X L
1 dk
I = i Dik , (5.362)
T i=1 k=1
dt
N L
X dk X N
1
= i Dik , (5.363)
T k=1
dt i=1
| {z }
G
=
k
N
X L
1 dk G
= . (5.364)
T dt k
k=1
Now Eq. (5.289) gives an explicit algebraic formula for dk /dt. Define then the constitutive
k :
function
k (1 , . . . , N L) dk = V (DT D)1 DT r. (5.365)
dt
So the irreversibility production rate is

N
X L
1 k G .
I = (5.366)
T k=1
k
The gradient of this field is given by

N L
!
I 1 X k G
2
k G
= + . (5.367)
p T k=1 p k p k
The Hessian of this field is given by

N L
!
I
2
1 X 2
k G k 2 G k 2 G
3G
= + + + k (5.368)
l p T k=1 l p k p l k l p k l p k
Now at equilibrium, k = ke , k = 0 as well as G/k = 0. Thus

we have

I = 0, (5.369)
= e
k k
I
= 0, (5.370)
k
k =ke
!
2 I
N L
1 X k 2 G
k 2 G

= + . (5.371)
l p T p l k l p k
k =ke k=1 k =ke
It can be shown for arbitrary matrices that the first term in the right hand side of Eq. (5.371)
is the transpose of the second; moreover, for arbitrary matrices, the two terms are in general
asymmetric. But, their sum is a symmetric matrix, as must be since the Hessian of I must
be symmetric. But for arbitrary matrices, we find no guarantee that the Hessian of I is
positive semidefinite, as the eigenvalues can take any sign. However, it can be verified by
direct expansion for actual physical systems that the two terms in the Hessian evaluated
at equilibrium are identical. This must be attributable to extra hidden symmetries in the
system. Moreover for actual physical systems, the eigenvalues of the Hessian of I are positive
semi-definite. So one can say

2
k
NXL
I 2 2 G
= . (5.372)
l p e
T k=1 p e
l k k =e
k =k k =k k
Where does the hidden symmetry arise? The answer lies in the fact that the k and G
are not independent. Using notions from Onsager reciprocity, let us see why the Hessian of
I is indeed positive semi-definite. Consider, for instance, Eq. (5.314), which is valid in the
neighborhood of equilibrium:
! !
V T
I = L valid only near equilibrium. (5.373)
R T T
has a
Recall that by construction L is positive semi-definite. Now, near equilibrium,
Taylor series expansion

=
+ ( |eq ) + .... (5.374)
eq
|{z} eq
=0
= 0. Let us also define the Jacobian of

Now recall that at equilibrium that as J . Thus
Eq. (5.373) can be rewritten as
V T
I = J ( | eq ) L J ( | eq ) , (5.375)
RT 2
V
= ( |eq )T JT L J ( |eq ). (5.376)
RT 2
By inspection, then we see that the Hessian of I is
2V T
HI = J L J . (5.377)
RT 2
Moreover L has the same eigenvalues as the similar matrix JT L J , so HI is positive
semi-definite.
Returning to the less refined formulation, in Gibbs notation, we can write Eq. (5.372) as
2
HI = HG J, (5.378)
T
where J is the Jacobian matrix of k . Consider the eigenvectors and eigenvalues of the various
matrices here. The eigenvalues of J give the time scales of reaction in the neighborhood of
equilibrium. They are guaranteed real and negative. The eigenvectors of J give the directions
of fast and slow modes. Near equilibrium, the dynamics will relax to the slow mode, and the
motion towards equilibrium will be along the eigenvector associated with the slowest time
scale. Now, in the unlikely circumstance that HG were the identity matrix, one would have
the eigenvalues of HI equal to the product of 2/T and the eigenvalues of HG . So they
would be positive, as expected. Moreover, the eigenvectors of HI would be identical to those
of J, so in this unusual case, the slow dynamics could be inferred from examining the slowest
descent down contours of I. Essentially the same conclusion would be reached if HG had a
diagonalization with equal eigenvalues on its diagonal. This would correspond to reactions
proceeding at the same rate near equilibrium. However, in the usual case, the eigenvalues
of HG are non-uniform. Thus the action of HG on J is to stretch it non-uniformly in such
a fashion that HI does not share the same eigenvalues or eigenvectors. Thus it cannot be
used to directly infer the dynamics.
Now consider the behavior of I in the neighborhood of an equilibrium point. In the
neighborhood of a general point k = k , I has a Taylor series expansion

N
X L

I 1 NXL N
X L 2 I

I = I
+ k k + l l p p + . . .
k =k
k=1
k 2 l=1 p=1 l p
k =k k =k
(5.379)
Near equilibrium, the first two terms of this Taylor series are zero, and I has the behavior

N
X L N
X L 2
1 I
I = (l le ) p pe + . . . (5.380)
2 l=1 p=1
l p e k =k
Substituting from Eq. (5.372), we find near equilibrium that

1
N
X L N
X L N
X L

2
G
k
I = (l le ) e
p + ... (5.381)
l k k =e
p
T l=1 p=1 k=1 p e k =k k
Lastly, let us study whether I is a Lyapunov function. We can show that I > 0, p 6= pe ,
and I = 0, p = pe . Now to determine whether or not the Lyapunov function exists, we

must study dI/dt:
N L
dI X I dp
= , (5.382)
dt p=1
p dt
N L
X I
= p , (5.383)
p=1
p
N L N L
!!
X 1 X k G
2
= k G
+ p ,
(5.384)
p=1
T k=1 p k p k
N L N L
!
1 XX k G
2
p G
p + k
= . (5.385)
T p=1 k=1 p k p k
At the equilibrium state,

k = 0, so obviously dI/dt = 0, at equilibrium. Away from
equilibrium, we know from our uniqueness analysis that the term 2 G/p k is positive

semi-definite; so this term contributes to rendering dI/dt < 0. But the other term does not
transparently contribute.

Let us examine the gradient of dI/dt. It is
N L N L
dI 1 XX k G
2 2
p + k G

p + k G p
=
m dt T p=1 m p k p m k p k m
k=1
!
p 2 G
3G 2
+ k +
p p G k
k + . (5.386)
m p k m p k p k m

At equilibrium, all terms in the gradient of dI/dt are zero, so it is a critical point.

Let us next study the Hessian of dI/dt to ascertain the nature of this critical point.
Because there are so many terms, let us only write those terms which will be non-zero at
equilibrium. In this limit, the Hessian is

2 dI
N L N L
1 X X k 2 G p k 2 G
p
= +
n m dt T p=1 p m k n p n k m
eq k=1
!
p 2 G
k p 2 G k
+ + . (5.387)
m p k n n p k m
With some effort, it may be possible to show the total sum is negative semi-definite. This

would guarantee that dI/dt < 0 away from equilibrium. If so, then it is true that I is a
Lyapunov function in the neighborhood of equilibrium. Far from equilibrium, it is not clear
whether or not a Lyapunov function exists.
Chapter 6
Reactive Navier-Stokes equations
Here we will present the compressible reactive Navier-Stokes equations for an ideal mixture
of N gases. We will not give detailed derivations. We are guided in part by the excellent
derivations given by Aris 1 and Merk. 2
6.1 Evolution axioms

6.1.1 Conservative form
The conservation of mass, linear momenta, and energy, and the entropy inequality for the
mixture are expressed in conservative form as

+ (u) = 0, (6.1)
t

(u) + (uu + P I ) = 0, (6.2)
t
1 1 q
e+ uu + u e + u u + j + (P I ) u = 0, (6.3)
t 2 2

jq
(s) + su + 0. (6.4)
t T
New variables here are the velocity vector u, the viscous shear tensor , the diffusive heat
flux vector jq . Note that these equations are precisely the same one would use for a single
fluid. We have neglected body forces.
The evolution of molecular species is dictated by the evolution axiom

(Yi ) + (Yi u + jm
i ) = Mi
i, i = 1, . . . , N 1. (6.5)
t
1
R. Aris, 1962, Vectors, Tensors, and the Basic Equations of Fluid Mechanics, Dover, New York.
2
H. J. Merk, 1959, The macroscopic equations for simultaneous heat and mass transfer in isotropic
continuous and closed systems, Applied Scientific Research, 8(1): 73-99.
237
238 CHAPTER 6. REACTIVE NAVIER-STOKES EQUATIONS
Here, the diffusive mass flux vector of species i is jm

i . Note that together, Eqs. (6.1) and the
N 1 of Eq. (6.5) form N equations for the evolution of the N species. We insist that the
species diffusive mass flux be constrained by
N
X
jm
i = 0. (6.6)
i=1
Recall from our earlier definitions of Mi and i that Eq. (6.5) can be rewritten as
L
! J
X X
(Yi ) + (Yi u + jm
i ) = M
l li ij rj . (6.7)
t j=1
l=1
Let us sum Eq. (6.7) over all species to get

N N N L
! J
X X X X X
(Yi ) + (Yi u + jm
i ) = M l li ij rj , (6.8)
i=1
t i=1 i=1 l=1 j=1

X
N N N N X L X J
X X
m
X
Yi + u Yi + ji = Ml li ij rj , (6.9)
t i=1 i=1 i=1
i=1 l=1 j=1
| {z } | {z } | {z }
=1 =1 =0
J
X L
X N
X
= rj Ml liij , (6.10)
| {z }
j=1 l=1 i=1
=0

+ (u) = 0. (6.11)
t
So, the summation over all species gives a redundancy with Eq. (6.1).
We can get a similar relation for the elements. Let us multiply Eq. (6.5) by our stoichio-
metric matrix li to get

li (Yi ) + li (Yi u + jm i ) = li Mi i , (6.12)
t
J
X
li Yi li Yi jm
i
+ u + li = li ij rj , (6.13)
t Mi Mi Mi j=1
XN X N N
X J
X
li Yi li Yi jm
i
+ u + li = li ij rj , (6.14)
i=1
t M i i=1
M i Mi i=1 j=1
N
! N N
! J N
X li Yi X li Yi X jm X X
+ u+ li i = rj li ij , (6.15)
t i=1 Mi Mi Mi | {z }
i=1 i=1 j=1 i=1
=0
N
! N N
!
X li Yi X li Yi X jm
+ u+ li i = 0. (6.16)
t i=1 Mi i=1
Mi i=1
Mi
6.1. EVOLUTION AXIOMS 239
PN
We recall that i = Yi /Mi and that i=1 li i = el , the element concentration. So, taking
the diffusive element flux to be
N
X
e jm
jl li , (6.17)
i=1
Mi
we get
e e

e
( ) + l u + jl = 0. (6.18)
t l
e
Taking the element mass density as el = Ml el and jel = Ml jl , we get
el
+ (el u + jel ) = 0. (6.19)
t
We can also insist that
N
X
jel = 0, (6.20)
i=1
so as to keep the total mass of each element constant. In summary, we have L 1 conserva-
tion equations for the elements, one global mass conservation equation, and N L species
evolution equations, in general. These add to form N equations for the overall species
evolution.
6.1.2 Non-conservative form

It is often convenient to have an alternative non-conservative form of the governing equations.
Let us define the material derivative as
d
= + u . (6.21)
dt t
6.1.2.1 Mass
Using the product rule to expand the mass equation, Eq. (6.1), we get

+ (u) = 0, (6.22)
t

+ u + u = 0, (6.23)
|t {z }
=d/dt
d
+ u = 0. (6.24)
dt
6.1.2.2 Linear momentum

We again use the product rule to expand the linear momentum equation, Eq. (6.2):
u
+ u + u (u) + u u + P = 0, (6.25)
t t
u
+ u u +u + (u) +P = 0, (6.26)
t t
| {z } | {z }
=du/dt =0
du
+ P = 0. (6.27)
dt
The key simplification was effected by using the mass equation Eq. (6.1).
6.1.2.3 Energy
Now use the product rule to expand the energy equation, Eq. (6.3).

1 1 1
e + u u + u e + u u + e + u u + (u)
t 2 2 2 t
| {z }
=0
q
+ j + (P u) ( u) = 0, (6.28)

1 1
e + u u + u e + u u + jq + (P u) ( u) = 0. (6.29)
t 2 2
We have once again used the mass equation, Eq. (6.1), to simplify. Let us expand more using
the product rule:

e u
+ u e +u + u u + jq + (P u) ( u) = 0, (6.30)
t t
| {z }
=de/dt

de u
+u + u u + P + jq + P u : u = 0, (6.31)
dt t
| {z }
=0
de
+ jq + P u : u = 0. (6.32)
dt
We have used the linear momentum equation, Eq. (6.27) to simplify.
From the mass equation, Eq. (6.24), we have u = (1/)d/dt, so the energy equation,
Eq. (6.32), can also be written as
de P d
+ jq : u = 0. (6.33)
dt dt
6.2. MIXTURE RULES 241
This energy equation can be formulated in terms of enthalpy. Use the definition h =
e + P/ to get an expression for dh/dt:
P 1
dh = de
2
d + dP, (6.34)

dh de P d 1 dP
= + , (6.35)
dt dt 2 dt dt
de P d dh 1 dP
2 = , (6.36)
dt dt dt dt
de P d dh dP
= . (6.37)
dt dt dt dt
So the energy equation, Eq. (6.33), in terms of enthalpy is
dh dP
+ jq : u = 0. (6.38)
dt dt
6.1.2.4 Second law

The second law in non-conservative form is, by inspection
ds jq
+ 0. (6.39)
dt T
6.1.2.5 Species
The species evolution equation in non-conservative form, is by inspection
dYi
+ jm
i = Mi
i, i = 1, . . . , N 1. (6.40)
dt
6.2 Mixture rules

We adopt the following rules for the ideal mixture:
N
X
P = Pi , (6.41)
i=1
XN
1 = Yi , (6.42)
i=1
XN XN
i
= i = , (6.43)
i=1 i=1
Mi
XN
e = Yi ei , (6.44)
i=1
N
X
h = Yi hi , (6.45)
i=1
XN
s = Y i si , (6.46)
i=1
XN
cv = Yi cvi , (6.47)
i=1
N
X
cP = Y i cP i , (6.48)
i=1
V = Vi , (6.49)
T = Ti . (6.50)
6.3 Constitutive models

The evolution axioms do not form a complete set of equations. Let us supplement these by
a set of constitutive model equations appropriate for a mixture of calorically perfect ideal
gases that react according the to the law of mass action with an Arrhenius kinetic reaction
rate. We have seen many of these models before, and repeat them here for completeness.
For the thermal equation of state, we take the ideal gas law for the partial pressures:
Yi i
Pi = RT i = RT = RT = Ri T i . (6.51)
Mi Mi
So the mixture pressure is
N
X N
X XN
Yi i
P = RT i = RT = RT . (6.52)
i=1 i=1
Mi i=1
M i
For the ideal gas, the enthalpy and internal energy of each component is a function of T at
most. We have for the enthalpy of a component
Z T
o
hi = hTo ,i + cP i(T)dT. (6.53)
To
So the mixture enthalpy is

N
X Z T
h = o
Yi hTo ,i +
cP i (T )dT . (6.54)
i=1 To
We then use the definition of enthalpy to recover the internal energy of component i:
Pi
ei = hi = hi Ri T. (6.55)
i
6.3. CONSTITUTIVE MODELS 243
So the mixture internal energy is

XN Z T
e = o
Yi hTo ,i Ri T + cP i(T )dT , (6.56)
i=1 To
XN Z T
= Yi hoTo ,i Ri (T To ) Ri (To ) +
cP i (T )dT , (6.57)
i=1 To
XN Z T
= Yi hoTo ,i Ri (To ) +
(cP i(T ) Ri )dT , (6.58)
i=1 To
XN Z T
= Yi hoTo ,i Ri (To ) +
cvi (T )dT , (6.59)
i=1 To
N
X Z T
= Yi eoTo ,i + cvi (T)dT . (6.60)
i=1 To
The mixture entropy is

N Z N
X T
cP (T) X Pi
s = Yi soTo ,i + dT Yi Ri ln . (6.61)
i=1 To T i=1
Po
The viscous shear stress for an isotropic Newtonian fluid which satisfies Stokes assump-
tion is

u + (u) 1
= 2 ( u)I . (6.62)
2 3
Here
is the mixture viscosity coefficient which is determined from a suitable mixture rule
averaging over each component.
The energy flux vector jq is written as
XN X N
q m DiT yi Mi P
j = kT + ji hi RT + 1 . (6.63)
i=1 i=1
Mi yi M P
Here k is a suitably mixture averaged thermal conductivity. The parameter DiT is the so-
called thermal diffusion coefficient. Recall yi is the mole fraction.
We consider a mass diffusion vector with multicomponent diffusion coefficient Dik :
XN
m Mi Dik Yk yk Mk P T
ji = 1 DiT . (6.64)
k=1, k6=i
M yk M P T
We adopt, as before, the reaction rate of creation of species i

J
X
i = ij rj . (6.65)
j=1
Here rj is given by the law of mass action:
N N
!
Y 1 Y kj
rj = k j kkj 1 . , (6.66)
k=1
Kc,j k=1 k
kj is given by the Arrhenius kinetics rule

j E j
kj = aj T exp , (6.67)
RT
and the equilibrium constant Kc,j for the ideal gas mixture is given by
PNi=1 ij
Po Goj
Kc,j = exp (6.68)
RT RT
6.4 Temperature evolution

Because temperature T has strong physical meaning, let us formulate our energy conservation
principle as a temperature evolution equation by employing a variety of constitutive laws.
Let us begin with Eq. (6.38) coupled with our constitutive rule for h, Eq. (6.54):

Z T
d
N
X dP
o
Yi hTo ,i + cP i(T )dT + jq : u = 0, (6.69)
dt i=1 To dt
| {z }
=hi
| {z }
=h
N
d X dP
Yi hi + jq : u = 0, (6.70)
dt i=1 dt
XN
dhi dYi dP
Yi + hi + jq : u = 0, (6.71)
i=1
dt dt dt
XN
dT dYi dP
Yi cP i + hi + jq : u = 0, (6.72)
i=1
dt dt dt
N N
dT X X dYi dP
Y i cP i + hi + jq : u = 0, (6.73)
dt i=1 i=1 | {z dt} dt
| {z } =Mi i jm i
=cP
N
dT X dP
cP + hi (Mi i jm q
i )+j : u = 0. (6.74)
dt i=1
dt
6.4. TEMPERATURE EVOLUTION 245
Now let us define the thermal diffusion flux jT as

XN
T DiT yi Mi P
j RT + 1 (6.75)
i=1
M i y i M P
With this our total energy diffusion flux vector jq , Eq. (6.63), becomes
N
X
q
j = kT + jm T
i hi + j . (6.76)
i=1
Now substitute Eq. (6.76) into Eq. (6.74) and rearrange to get
N N
!
dT X X
cP + hi (Mi i jm
i )+ kT + jm
i hi + j
T
dt i=1 i=1
dP
= + : u,(6.77)
dt
N
dT X dP
cP + (hi Mi i hi jm m T
i + (ji hi )) = (kT ) j + + : u,(6.78)
dt i=1
dt
N
dT X dP
cP + (hi Mi i + jm T
i hi )) = (kT ) j + + : u,(6.79)
dt i=1
dt
N
dT X dP
cP + (hi Mi i + cP i jm T
i T ) = (kT ) j + + : u.(6.80)
dt i=1
dt
So the equation for the evolution of a material fluid particle is

X N
dT dP
cP = (hi Mi i + cP ijm T
i T ) + (kT ) j + + : u. (6.81)
dt i=1
dt
Let us impose some more details from the reaction law. First, we recall that hi = hi Mi ,
so
X N X N
dT dP
cP = hi i cP i jm T
i T + (kT ) j + + : u. (6.82)
dt i=1 i=1
dt
Impose now Eq. (6.65) to expand i

X XN J X N
dT dP
cP = hi ij rj cP i jm T
i T + (kT ) j + + : u, (6.83)
dt i=1 j=1 i=1
dt
XJ XN XN
dT dP
cP = rj hi ij cP i jm T
i T + (kT ) j + + : u. (6.84)
dt j=1 i=1 i=1
dt
Now let us extend the definition of heat of reaction, Eq. (4.444) to account for multiple
reactions,
N
X
Hj hi ij . (6.85)
i=1
With this, Eq. (6.84) becomes, after small rearrangement,
XJ N
dT dP X
cP = rj Hj + cP i jm T
i T + (kT ) j + : u . (6.86)
dt j=1
dt
|i=1 {z }
diffusion effects
By comparing Eq. (6.86) with Eq. (4.446), it is easy to see the effects of multiple reactions,
variable pressure, and diffusion on how temperature evolves. Interestingly, mass, momen-
tum, and energy diffusion all influence temperature evolution. The non-diffusive terms are
combinations of advection, reaction, and spatially homogeneous effects.
Chapter 7
Simple solid combustion:

Reaction-diffusion
Here we will modify the simple thermal explosion theory which balanced unsteady evolution
against reaction to include the effects of diffusion. Starting from a fully unsteady formulation,
we will focus on cases which are steady, resulting in a balance between reaction and diffusion;
however, we will briefly consider a balance between all three effects. Such a theory is known
after its founder as Frank-Kamenetskii theory.
In particular, we will look for transitions from a low temperature reaction to a high
temperature reaction, mainly in the context of steady state solutions, i.e. solutions with no
time dependence.
We draw upon the work of Buckmaster and Ludford for guidance. 1
7.1 Simple planar model

Let us consider a slab of solid fuel/oxider mixture. The material is modelled of infinite extent
in the y and z directions and has length in the x direction of 2L. The temperature at each
end, x = L, is held fixed at T = To .
The slab is initially unreacted. Exothermic conversion of solid material from reactants to
products will generate an elevated temperature within the slab T > To , for x (L, L). If
the thermal energy generated diffuses rapidly enough, the temperature within the slab will be
T To , and the reaction rate will be low. If the energy generated by local reaction diffuses
slowly, it will accumulate in the interior, and accelerate the local reaciton rate, inducing
rapid energy release and high temperature.
Let us assume
The material is an immobile, incompressible solid with constant specific heat,

1
J. D. Buckmaster and G. S. S. Ludford, 1983, Lectures on Mathematical Combustion, SIAM, Philadelphia.
See Chapter 1.
247
248 CHAPTER 7. SIMPLE SOLID COMBUSTION: REACTION-DIFFUSION
Thus is constant,
Thus cP = cv is a constant,
Thus, there is no advective transport of mass, momentum, or energy: u = 0.
The material has variation only with x and t,
The reaction can be modelled as
A B, (7.1)
where A and B have identical molecular masses.
The reaction is irreversible,
The reaction is exothermic,
Initially only A is present,
The thermal conductivity, k is constant.
Let us take, as we did in thermal explosion theory,
YA = 1 , (7.2)
YB = . (7.3)
We interpret as a reaction progress variable which has [0, 1]. For = 0, the material
is all A; when = 1, the material is all B.
7.1.1 Model equations

Our simple model for reaction is

= aeE/R/T (1 ), (7.4)
t
e 2T
= k 2, (7.5)
t x
e = cv T q. (7.6)
Equation (7.4) is our reaction kinetics law; Eq. (7.5) is our energy conservation expression;
and Eq. (7.6) is our caloric equation of state; it is Eq. (1.303) with q = eoTo ,A eoTo ,B and
YB = . Equations (7.4-7.6) are completed by initial and boundary conditions, which are
T (L, t) = T (L, t) = To , T (x, 0) = To , (x, 0) = 0. (7.7)

7.1. SIMPLE PLANAR MODEL 249
7.1.2 Simple planar derivation

Let us perform a simple control volume derivation of the planar energy equation, Eq. (7.6).
Consider a small volume of dimension A by x. At the left boundary x, we have heat
flux q, in which we notate as qx . At the right boundary, x + x, we have heat flux out of
qx+x . Recall the units of heat flux are J/m2 /s.
The first law of thermodynamics is
change in total energy = heat in work

| {z out} . (7.8)
=0
There is no work for our system. But there is heat flux over system boundaries. In a
combination of symbols and words, we can say
total energy @ t + t total energy @ t = energy flux in energy flux out . (7.9)
| {z } | {z }
unsteady advection
| {z } and diffusion
=0
Mathematically, we can say
E|t+t E|t = (Ef lux out Ef lux in ) , (7.10)

Ax e|t+t e|t = (qx+x qx ) |{z}At , (7.11)
| {z } | {z } | {z } 2
kg J/m2 /s (m s)
J/kg

e|t+t e|t qx+x qx
= (7.12)
t x
Now, let x 0 and t 0, and we get
e qx
= . (7.13)
t x
Now standard constitutive theory gives Fouriers law to specify the heat flux:
T
q = k . (7.14)
x
So Eq. (7.14) along with the thermal state equation, (7.6) when substituted into Eq. (7.13)
yield
2T
(cv T q) = k 2 , (7.15)
t x
T 2T
cv q = k 2, (7.16)
t t x
2
T T
cv qaeE/R/T (1 ) = k 2 , (7.17)
t x
T 2T
cv = k 2 + qaeE/R/T (1 ). (7.18)
t x
So our complete system is two equations in two unknowns with apprpriate initial and bound-
ary conditions:
T 2T
cv = k 2 + qaeE/R/T (1 ), (7.19)
t x

= aeE/R/T (1 ), (7.20)
t
T (L, t) = T (L, t) = To , T (x, 0) = 0, (x, 0) = 0. (7.21)
7.1.3 Ad hoc approximation

Let us consider an ad hoc approximation to a system much like Eqs. (7.19-7.21), but which
has the advantage of being one equation and one unknown.
7.1.3.1 Planar formulation

If there were no diffusion, Eq. (7.13) would yield e/t = 0 and would lead us to conclude
that e(x, t) = e(x). And because we have nothing now to introduce a spatial inhomogeneity,
there is no reason to take e(x) to be anything other than a constant eo . That would lead us
to
e(x, t) = eo , (7.22)
so
eo = cv T q. (7.23)
Now at t = 0, = 0, and T = To , so eo = cv To ; thus,
cv To = cv T q, (7.24)
cv (T To )
= . (7.25)
q
We also get the final temperature at = 1 to be
q
T ( = 1) = To + . (7.26)
cv
We shall adopt Eq. (7.25) as our model for in place of Eq. (7.20). Had we admitted species
diffusion, we could more rigorously have arrived at a similar result, but it would be more
difficult to justify treating the material as a solid.
Let us use Eq. (7.25) to eliminate in Eq. (7.19) so as to get a single equation for T :

T 2T E/R/T cv (T To )
cv = k 2 + qae 1 , (7.27)
t x q
| {z }
reaction source term
7.2. NON-DIMENSIONALIZATION 251
The first two terms in Eq. (7.27) are nothing more than the classical heat equation. The
non-classical term is an algebraic source term due to chemical reaction. Note when T = To ,
the reaction source term is qa exp(E/R/T ) > 0, so at the initial state there is a tendency
to increase the temperature. When = 1, so T = To + q/cv , the reaction source term is
zero, so there is no local heat release. This effectively accounts for reactant depletion.
Scaling the equation by cv and employing the well known formula for thermal diffusivity
= k//cP = k//cv , we get

T 2T q
= 2 +a (T To ) eE/R/T , (7.28)
t x cv
T (L, t) = T (L, t) = To , T (x, 0) = To . (7.29)
From here on, we shall consider Eqs. (7.29) and cylindrical and spherical variants to be the
full problem. We shall consider solutions to it in various limits. We will not return here to
the original problem without the ad hoc assumption, but that would be a straightforward
exercise.
7.1.3.2 More general coordinate systems

We note that Eq. (7.29) can be extended to general coordinate systems via

T 2 q
= T + a (T To ) eE/R/T . (7.30)
t cv
Appropriate initial and boundary conditions for the particular coordinate system would be
necessary.
For one-dimensional solutions in planar, cylindrical, and spherical coordinates, one can
summarize the formulations as

T 1 m T q
= m x +a (T To ) eE /R/T , (7.31)
t x x x cv
(7.32)
Here we have m = 0 for a planar coordinate system, m = 1 for cylindrical, and m = 2 for
spherical. For cylindrical and spherical systems, the Dirichlet boundary condition at x = L
would be replaced with a boundedness condition on T at x = 0.
7.2 Non-dimensionalization
Let us non-dimensionalize Eqs. (7.30). The scaling we will choose is not unique.
Let us define non-dimensional variables
cv (T To ) x
T = , x = , t = at. (7.33)
q L
With these choices, Eqs. (7.30) transform to

qa T q 1 m m T q q E 1
= x x +a T exp (7.34)
,
cv t cv L2 x x cv cv R To + (q/cv )T

T m T E 1
= 2
x xm + (1 T ) exp , (7.35)
t aL x x RTo 1 + q T
cv To
(7.36)
Now both sides are dimensionless. Let us define the dimensionless parameters
aL2 E q
D= , = , Q= . (7.37)
RTo cv To
Here D is a so-called Damkoler number. If we recall a diffusion time scale d is
L2
d = , (7.38)

and a first estimate (which will be shown to be crude) of the reaction time scale, r is
1
r = , (7.39)
a
We see that the Damkoler number is the ratio of the thermal diffusion time to the reaction
time (ignoring activation energy effects!):
L2 / d thermal diffusion time

D= = = . (7.40)
1/a r reaction time
We can think of and Q as ratios as well with
activation energy exothermic heat release
= , Q= . (7.41)
ambient energy ambient energy
The initial and boundary conditions scale to
T (1, t ) = T (1, t ) = 0, T (x , 0) = 0, if m = 0, (7.42)

T (1, t ) = 0, T (0, t ) < , T (x , 0) = 0, if m = 1, 2. (7.43)
7.2.1 Diffusion time discussion

As an aside, let us consider in more depth the thermal diffusion time. Consider the dimen-
sional heat equation
T 2T
= 2, (7.44)
t x
7.2. NON-DIMENSIONALIZATION 253
with a complex representation of a sinusoidal initial condition:

T (x, 0) = Aeikx . (7.45)
Here, A is a temperature amplitude, k is the wave number, and i2 = 1. If one wants, one
can insist we only worry about the real part of the solution, but that will not be important
for this analysis. Note that since
+ Ai sin(kx),

Aeikx = A cos(kx)
Note when the
we can think of the initial condition as a signal with wavelength = 2/k.
wavenumber k is large, the wavelength is small, and vice versa.
Let us assume a separation of variables solution of the form

T (x, t) = B(t)eikx . (7.46)
Thus
T dB ikx

= e , (7.47)
t dt
and
T ikx ,
= ikBe (7.48)
x
2T
= k2 Beikx . (7.49)
x2
So our partial differential equation reduces to
dB
eikx = k 2 Beikx , (7.50)
dt
dB
= k 2 B, B(0) = A, (7.51)
dt
2
B(t) = Aek t (7.52)
Note that as t B 0 with diffusion time constant

1
d = .
k 2
So, a large thermal diffusivity causes modes to relax quickly. But high wavenumber k
also induces modes to relax quickly. Note also that in terms of wavelength of the initial
disturbance,
1 2
d = .
4 2
So small wavelength disturbances induce fast relaxation.
Importantly, recall that an arbitrary signal with have a Fourier decomposition into an
infinite number of modes, each with a different wave number. The above analysis shows that
each mode will decay with its own diffusion time constant. Indeed some of the modes, even
at the initial state, may have negligibly small amplitude. But many will not, and we cannot
know a priori which ones will be large or small.
Lastly, we note that in a finite domain, we will find a discrete spectrum of wavenumbers;
while in an infinite domain, we will find a continuous spectrum.
7.2.2 Final form

Let us now drop the notation and understand that all variables are dimensionless. So our
equations become

T 1 m m T
= x x + (1 T ) exp ,
t D x x 1 + QT
T (1, t) = T (1, t) = 0, T (x, 0) = 0, if m = 0,
T (1, t) = 0, T (0, t) < , T (x, 0) = 0, if m = 1, 2.
(7.53)
Note that the initial and boundary conditions are homogeneous. The only inhomogeneity
lives in the exothermic reaction source term, which is non-linear due to the exp(1/T ) term.
Also, it will prove to be the case that a symmetry boundary condition at x = 0 suffices,
though our original formulation is more rigorous. Such equivalent boundary conditions are
T
T (1, t) = 0, (0, t) = 0, T (x, 0) = 0, m = 1, 2, 3. (7.54)
x
7.2.3 Integral form

As an aside, let us consider the evolution of total energy within the domain. To do so we
integrate a differential volume element through the entire volume. We recall dV xm dx,
for m = 0, 1, 2, (planar, cylindrical, spherical).

m T 1 m m m T m
x dx = x x x dx + x (1 T ) exp dx,
t D x x 1 + QT
(7.55)
Z 1 Z 1 Z 1
T 1 T
xm dx = xm dx + xm (1 T ) exp dx,
0 t 0 D x x 0 1 + QT
(7.56)
Z 1 Z 1

1 T
xm T dx = + xm (1 T ) exp dx .
t 0 D x x=1 0 1 + QT
| {z } | {z } | {z }
thermal energy change boundary heat flux internal conversion
(7.57)
7.3. STEADY SOLUTIONS 255
Note that the total thermal energy in our domain changes due to two reasons: 1) diffusive
energy flux at the isothermal boundary, 2) internal conversion of chemical energy to thermal
energy.
7.2.4 Infinite Damk

oler limit
Note for D diffusion becomes unimportant, and we recover a balance between unsteady
effects and reaction:

dT
= (1 T ) exp , T (0) = 0. (7.58)
dt 1 + QT
This is the problem we have already considered in thermal explosion theory. Let us here
commence on a different route than that taken in the earlier chapter. We recall that thermal
explosion theory predicts significant acceleration of reaction when
e
t . (7.59)
Q
7.3 Steady solutions

Let us seek solutions to the planar (m = 0) version of Eqs. (7.54) that are formally steady,
so that /t = 0, and a balance between reaction and energy diffusion is attained. In that
limit, Eqs.(7.53) reduce to the following two point boundary value problem:

1 d2 T
0 = + (1 T ) exp , (7.60)
D dx2 1 + QT
0 = T (1) = T (1). (7.61)
This problem is difficult to solve analytically because of the strong non-linearity in the
reaction source term.
7.3.1 High activation energy asymptotics

Motivated by our earlier success in getting approximate solutions to a similar problem in
spatially homogeneous thermal explosion theory, let us take a similar approach here.
Let us seek low temperature solutions where the non-linearity may be weak. So let us
define a small parameter with 0 << 1. And now let us assume a power series expansion
of T of the form
T = T1 + 2 T2 + . . . (7.62)
Here we assume T1 (x) O(1), T2 (x) O(1), . . .. We will focus attention on getting an
approximate solution for T1 (x).
With assumption Eq. (7.62), Eq. (7.60) expands to
d2 T1
+ . . . = D(1 T1 . . .) exp((1 QT1 + . . .)). (7.63)
dx2
Ignoring higher order terms and simplifying we get
d2 T1
= D(1 T1 ) exp() exp(QT1 ). (7.64)
dx2
Now let us once again take the high activation energy limit and insist that be defined such
that
1
. (7.65)
Q
This gives us
d2 T1 D
2
= (1 T1 ) exp() exp(T1 ). (7.66)
dx
We have gained an analytic advantage once again by moving the temperature into the nu-
merator of the argument of the exponential.
Now let us neglect T1 as small relative to 1 and define a new parameter such that
D
exp() = DQ exp(). (7.67)

Our governing equation system then reduces to
d2 T1
= eT1 , (7.68)
dx2
T1 (1) = T1 (1) = 0. (7.69)
It is still not clear how to solve this. It seems reasonable to assume that T1 should have
symmetry about x = 0. If so, we might also presume that the gradient of T1 is zero at x = 0:

dT1
= 0, (7.70)
dx x=0
which induces T1 (0) to take on an extreme value, say T1 (0) = T1m , where m could denote
maximum or minimum. Certainly T1m is unknown at this point.
Let us explore the appropriate phase space solution. To aid in this, we can define q as
dT1
q . (7.71)
dx
We included the minus sign so that q has a physical interpretation of heat flux. With this
assumption, our equation system can be written as two autonomous ordinary differential
equations in two unknowns:
dq
= eT1 , q(0) = 0, (7.72)
dx
dT1
= q, T1 (0) = T1m . (7.73)
dx
This is slightly unsatisfying because we do not know T1m . But we presume it exists, and is
a constant. Let us see if the analysis can reveal it by pressing forward.
Let us scale Eq. (7.73) by Eq. (7.72) to get
dT1 q
= T1 , T1 |q=0 = T1m . (7.74)
dq e
Now Eq. (7.74) can be solved by separating variables. Doing so and solving we get
eT1 dT1 = qdq, (7.75)

q2
eT1 = + C, (7.76)
2
applying the initial condition, (7.77)
T1m
e = C, (7.78)
2
q m
= (eT1 eT1 ), (7.79)
2 q
m
q = 2(eT1 eT1 ). (7.80)
The plus square root is taken here. This will correspond to x [0, 1]. The negative square
root will correspond to x [1, 0].
We can now substitute Eq. (7.80) into Eq. (7.73) to get
q
dT1 m
= 2(eT1 eT1 ) (7.81)
dx
Once again separate variables to get
dT1
p m
= dx. (7.82)
2(eT1 eT1 )
Computer algebra reveals this can be integrated to form

q T

2 tanh1 1 eT1m1
e
m /2
= x + C. (7.83)
eT1
Now, when x = 0, we have T1 = T1m . The inverse hyperbolic tangent has an argument of
zero there, and it evaluates to zero. Thus we must have C = 0, so
q
T
2 tanh1 1 eT1m1
e
m /2
= x, (7.84)
eT1
r ! r
e T1 m
tanh1 1 Tm = eT1 /2 x. (7.85)
e 1 2
Now recall that

sech1 t = tanh1 1 t2 ,
so Eq. (7.85) becomes
r
T1 T1m m
sech1 e 2 = eT1 /2 x, (7.86)
2
r!
T1 T1m
T1m /2
e 2 = sech e x , (7.87)
2
r !
T1 m m
e 2 = eT1 /2 sech eT1 /2 x , (7.88)
2
r !!
T1m /2 T1m /2
T1 = 2 ln e sech e x . (7.89)
2
So we have an exact solution for T1 (x). This is a nice achievement, but we are not sure it
satisfies the boundary condition at x = 1, nor do we know the value of T1m . We must choose
T1m such that T1 (1) = 0, which we have not yet enforced. So
r !!
m m
0 = 2 ln eT1 /2 sech eT1 /2 . (7.90)
2
Only when the argument of a logarithm is unity does it map to zero. So we must demand
that
r !
m m
eT1 /2 sech eT1 /2 = 1, (7.91)
2
r
m m
eT1 /2 = sech1 eT1 /2 , (7.92)
r 2
T1m /2 m
e = cosh1 eT1 /2 , (7.93)
r 2
m m
= eT1 /2 cosh1 eT1 /2 . (7.94)
2
T1 m
8

0.2 0.4 0.6 0.8
Figure 7.1: Plot of maximum temperature perturbation T1m versus reaction rate constant
for steady reaction-diffusion in the high activation energy limit
Equation (7.94) gives a direct relationship between and T1m . Given T1m , we get explicitly;
thus we can easily generate a plot; see Fig 7.1. The inverse cannot be done analytically,
but can be done numerically via iterative techniques. We notice a critical value of , c =
0.878458. When = c = 0.878458, we find T1m = 1.118684. For < c , there are two
admissible values of T1m . That is to say the solution is non-unique for < c . Moreover,
both solutions are physical. For = c there is a single solution. For > c there are no
steady solutions.
Presumably re-introduction of neglected processes would aid in determining which solu-
tion is realized in nature. We shall see that transient stability analysis aids in selecting the
correct solution when there are choices. We shall also see that reintroduction of reaction
depletion into the model induces additional physical solutions, including those for > c .
For = 0.4 < c we get two solutions. Both are plotted in Fig 7.2. The high temperature
solution has T1m = 3.3079, and the low temperature solution has T1m = 0.24543. At this point
we are not sure if either solution is temporally stable to small perturbations. We shall later
prove that the low temperature solution is stable, and the higher temperature solution is
unstable. Certainly for > c , there are no low temperature solutions. We will see that upon
reintroduction of missing physics, there are stable high temperature solutions. To prevent
high temperature solutions, we need < c .
Now recall that
= DQe . (7.95)
T1
3.0
2.5
2.0
1.5
1.0
0.5
x
-1.0 -0.5 0.5 1.0
Figure 7.2: Plot of T1 (x) for = 0.4 showing the two admissible steady solutions.
Now to prevent the high temperature solution, we demand
= DQe < 0.878458. (7.96)
Let us bring back our dimensional parameters to examine this criteria:
DQe < 0.878458, (7.97)

L2 a q E E
exp < 0.878458 (7.98)
cv To RTo RTo
Thus the factors that tend to prevent thermal explosion are
High thermal diffusivity; this removes the thermal energy rapidly,
Small length scales; the energy thus has less distance to diffuse,
Slow reaction rate kinetics,
High activation energy.
7.3.2 Method of weighted residuals

The method of the previous section was challenging. Let us approach the problem of calculat-
ing the temperature distribution with a powerful alternate method: the method of weighted
residuals, using so-called Dirac delta functions as weighting functions. We will first briefly
review the Dirac delta function, then move on to solving the Frank-Kamenetskii problem
with the method of weighted residuals.
The Dirac D -distribution (or generalized function, or simply function), is defined by

Z
0 if a 6 [, ]
f (x)D (x a)dx = (7.99)
f (a) if a [, ]
From this it follows by considering the special case in which f (x) = 1 that
D (x a) = 0 if x 6= a (7.100)
Z
D (x a)dx = 1. (7.101)

Let us first return to Eq. (7.68) rearranged as:
d2 T1
+ eT1 = 0, (7.102)
dx2
T1 (1) = T1 (1) = 0. (7.103)
We shall approximate T1 (x) by

N
X
T1 (x) = Ta (x) = ci fi (x). (7.104)
i=1
At this point we do not know the so-called trial functions fi (x) nor the constants ci .
7.3.2.1 One-term collocation solution

Let us choose fi (x) to be linearly independent functions which satisfy the boundary condi-
tions. Moreover let us consider the simplest of approximations for which N = 1. A simple
function which satisfies the boundary conditions is a polynomial. The polynomial needs to
be at least quadratic to be non-trivial. So let us take
f1 (x) = 1 x2 . (7.105)
Note this gives f1 (1) = f1 (1) = 0. So our one-term approximate solution takes the form
Ta (x) = c1 (1 x2 ). (7.106)
We still do not yet have a value for c1 . Let us choose c1 so as to minimize an error of our
approximation. The error of our approximation e(x) will be
d2 Ta
e(x) = 2
+ eTa (x) , (7.107)
dx
d2 2

= c 1 (1 x ) + exp(c1 (1 x2 )), (7.108)
dx2
= 2c1 + + exp(c1 (1 x2 )). (7.109)
If we could choose c1 in such a way that e(x) was exactly 0 for x [1, 1], we would be
done. That will not happen. So let us choose c1 to drive a weighted domain-averaged error
to zero. That is let us demand that
Z 1
1 (x)e(x) dx = 0. (7.110)
1
We have introduced here a weighting function 1 (x). Many choices exist for the weighting
function. If we choose 1 (x) = Tm (x) our method is known as a Galerkin method. Let
us choose instead another common weighting, 1 (x) = D (x). This method is known as a
collocation method. We have chosen a single collocation point at x = 0. With this choice,
Eq. (7.110) becomes
Z 1

D (x) 2c1 + exp(c1 (1 x2 )) dx = 0. (7.111)
1
The evaluation of this integral is particularly simple due to the choice of the Dirac weighting.
We simply evaluate the integrand at the collocation point x = 0 and get
2c1 + + exp(c1 ) = 0. (7.112)
Thus
= 2c1 ec1 . (7.113)
Note here is a physical parameter with no relation the the Dirac delta function D . Note
that with our approximation Ta = c1 (1 x2 ), that the maximum value of Ta is Tam = c1 . So
we can say
m
= 2Tam eTa . (7.114)
We can plot Tam as a function of ; see Fig 7.3. Note the predictions of Fig. 7.3 are
remarkably similar to those of Fig. 7.1. For example, when = 0.4, numerical solution of
Eq. (7.114) yields two roots:
Tam = 0.259171, Tam = 2.54264. (7.115)
Thus we get explicit approximations for the high and low temperature solutions for a one-
term collocation approximation:
Ta = 0.259171(1 x2 ), (7.116)
Ta = 2.54264(1 x2 ). (7.117)
Plots of the one-term collocation approximation Ta (x) for high and low temperature solutions
are given in Fig 7.4.
We can easily find c for this approximation. Note that differentiating Eq. (7.114) gives
d
= 2 exp T1m (1 T1m ). (7.118)
dT1m
A critical point exists when d/dT1m = 0. We find this exists when T1m = 1. So c =
2(1)e1 = 0.735769.
Ta m
8

0.1 0.2 0.3 0.4 0.5 0.6 0.7
Figure 7.3: Plot of maximum temperature perturbation Tam versus reaction rate constant
for steady reaction-diffusion using a one-term collocation method.
Ta
3.5
3.0
2.5
2.0
1.5
1.0
0.5
x
-1.0 -0.5 0.5 1.0
Figure 7.4: Plots of high and low temperature distributions Ta (x) using a one-term colloca-
tion method.
7.3.2.2 Two-term collocation solution

We can improve our accuracy by including more basis functions. Let us take N = 2. We
have a wide variety of acceptable choices for basis function that 1) satisfy the boundary
conditions, and 2) are linearly independent. Let us focus on polynomials. We can select the
first as before with f1 = 1 x2 as the lowest order non-trivial polynomial that satisfies both
boundary conditions. We could multiply this by an arbitrary constant, but that would not
be of any particular use. Leaving out details, if we selected the second basis function as a
cubic polynomial, we would find the coefficient on the cubic basis function to be c2 = 0.
That is a consequence of the oddness of the cubic basis function and the evenness of the
solution we are simulating. So it turns out the lowest order non-trivial basis function is a
quartic polynomial, taken to be of the form
f2 (x) = a0 + a1 x + a2 x2 + a3 x3 + a4 x4 . (7.119)
We can insist that f2 (1) = 0 and f2 (1) = 0. This gives
a0 + a1 + a2 + a3 + a4 = 0, (7.120)
a0 a1 + a2 a3 + a4 = 0. (7.121)
We solve these for a0 and a1 to get a0 = a2 a4 , a1 = a3 . So our approximation is
f2 (x) = (a2 a4 ) a3 x + a2 x2 + a3 x3 + a4 x4 . (7.122)
Motivated by the fact that a cubic approximation did not contribute to the solution, let us
select a3 = 0 so to get
f2 (x) = (a2 a4 ) + a2 x2 + a4 x4 . (7.123)

R1
Let us now make the choice that 1 f1 (x)f2 (x) dx = 0. This guarantees an orthogonal
basis, although these basis functions are not eigenfunctions of any relevant self-adjoint linear
operator for this problem. Often orthogonality of basis functions can lead to a more efficient
capturing of the solution. This results in a4 = 7a2 /8. Thus

1 2 7 4
f2 (x) = a2 + x x (7.124)
8 8
Let us select a2 = 8 so that
f2 (x) = 1 8x2 + 7x4 = (1 x2 )(1 7x2 ). (7.125)
and
f1 (x) = 1 x2 , (7.126)
f2 (x) = (1 x2 )(1 7x2 ). (7.127)
So now we seek approximate solutions TA (x) of the form

TA (x) = c1 (1 x2 ) + c2 (1 x2 )(1 7x2 ). (7.128)
This leads to an error e(x) of
d2 TA
e(x) = + eTA (x) , (7.129)
dx2
d2 2 2 2

= c 1 (1 x ) + c 2 (1 x )(1 7x )
dx2
+ exp(c1 (1 x2 ) + c2 (1 x2 )(1 7x2 )), (7.130)
= 2c1 + 56c2 x2 2c2 (1 7x2 ) 14c2 (1 x2 )
+ exp(c1 (1 x2 ) + c2 (1 x2 )(1 7x2 )). (7.131)
Now we drive two weighted errors to zero:
Z 1
1 (x)e(x) dx = 0, (7.132)
1
Z 1
2 (x)e(x) dx = 0. (7.133)
1
Let us once again choose the weighting functions i (x) to be Dirac delta functions so that we
have a two-term collocation method. Let us choose unevenly distributed collocation points
so as to generate independent equations taking x = 0 and x = 1/2. Symmetric choices would
lead to a linearly dependent set of equations. Other unevenly distributed choices would work
as well. So we get
Z 1
D (x)e(x) dx = 0, (7.134)
1
Z 1
D (x 1/2)e(x) dx = 0. (7.135)
1
or
e(0) = 0, (7.136)
e(1/2) = 0. (7.137)
Expanding, these equations are
2c1 16c2 + exp (c1 + c2 ) = 0, (7.138)

3 9
2c1 + 5c2 + exp c1 c2 = 0. (7.139)
4 16
For = 0.4, we find a high temperature solution via numerical methods:
c1 = 3.07054, c2 = 0.683344, (7.140)
TA
4
x
-1.0 -0.5 0.5 1.0
Figure 7.5: Plots of high and low temperature distributions TA (x) using a two-term colloca-
tion method.
and a low temperature solution as well

c1 = 0.243622, c2 = 0.00149141 (7.141)
So the high temperature distribution is
TA (x) = 3.07054(1 x2 ) + 0.683344(1 x2 )(1 7x2 ). (7.142)
The peak temperature of the high temperature distribution is TAm = 3.75388. This is an
improvement over the one term approximation of Tam = 2.54264 and closer to the high
temperature solution found via exact methods of T1m = 3.3079.
The low temperature distribution is
TA (x) = 0.243622(1 x2 ) + 0.00149141(1 x2 )(1 7x2 ). (7.143)
The peak temperature of the low temperature distribution is TAm = 0.245113. This in
an improvement over the one term approximation of Tam = 0.259171 and compares very
favorably to the low temperature solution found via exact methods of T1m = 0.24543.
Plots of the two-term collocation approximation TA (x) for high and low temperature
solutions are given in Fig 7.5. While the low temperature solution is a very accurate repre-
sentation, the high temperature solution exhibits a small negative portion near the boundary.
7.3.3 Steady solution with depletion

Let us return to the version of steady state reaction with depletion without resorting to the
high activation energy limit, Eq.(7.61):

1 d2 T
0 = + (1 T ) exp , (7.144)
D dx2 1 + QT
0 = T (1) = T (1). (7.145)

Rearrange to get

d2 T
= D (1 T ) exp , (7.146)
dx2 1 + QT

Q exp()
= D (1 T ) exp , (7.147)
Q exp() 1 + QT
(7.148)
Now simply adapting our earlier definition of = DQ exp(), which we note does not
imply we have taken any high activation energy limits, we get

d2 T exp()
= (1 T ) exp , (7.149)
dx2 Q 1 + QT
T (1) = T (1) = 0. (7.150)
Equations (7.149-7.150) can be solved by a numerical trial and error method where we
demand that dT /dx(x = 0) = 0 and guess T (0). We keep guessing T (0) until we have
satisfied the boundary conditions.
When we do this with = 0.4, = 15, Q = 1 (so D = e /Q/ = 87173.8), we find
three steady solutions. One is at low temperature with T m = 0.016. This is a somewhat lower
than of high activation energy limit estimate of T1m = 0.24542. We find a second intermediate
temperature solution with T m = 0.417. Again this is lower than the high activation limit
value of T1m = 3.3079. And we find a high temperature solution with T m = 0.987. There is
no counterpart to this solution from the high activation energy limit analysis. Plots of T (x)
for high, low, and intermediate temperature solutions are given in Fig 7.6.
We can use a one-term collocation approximation to estimate the relationship between
and T m . Let us estimate that
Ta (x) = c1 (1 x2 ). (7.151)
With that choice, we get an error of

e(x) = 2c1 + exp (1 c1 (1 x2 )) (7.152)
Q 1 + c1 Q(1 x2 )
R1
We choose a one term collocation method with 1 (x) = D (x). Then setting 1 1 (x)e(x)dx =
0 gives

e(0) = 2c1 + exp (1 c1 ) = 0. (7.153)
Q 1 + c1 Q
We solve for and get

2c1 Q
= exp (7.154)
1 c1 e 1 + c1 Q
T
1.0
0.8
0.6
0.4
0.2
x
-1.0 -0.5 0.5 1.0
Figure 7.6: Plots of high, low, and intermediate temperature distributions T (x) for = 0.4,
Q = 1, = 15.
Ta m
1.0
0.8
0.6
0.4
0.2

0.2 0.4 0.6 0.8 1.0 1.2 1.4
Figure 7.7: Plots of Tam versus , with Q = 1, = 15 from a one-term collocation approxi-
mate solution.
7.4. UNSTEADY SOLUTIONS 269
The maximum temperature of the approximation is given by Tam = c1 and occurs at x = 0.

A plot of Tam versus is given in Fig 7.7. For < c1 0.2, one low temperature solution
exists. For c1 < < c2 0.84, three solutions exist. For > c2 , one high temperature
solution exists.
7.4 Unsteady solutions

Let us know study the effects of time-dependency on our combustion problem. Let us
consider the planar, m = 0, version of Eqs. (7.53):

T 1 2T
= + (1 T ) exp ,
t D x2 1 + QT
T (1, t) = T (1, t) = 0, T (x, 0) = 0. (7.155)
7.4.1 Linear stability

We will first consider small deviations from the steady solutions found earlier and see if those
deviations grow or decay with time. This will allow us to make a definitive statement about
the linear stability of those steady solutions.
7.4.1.1 Formulation
First, recall that we have independently determined three exact numerical steady solutions
to the time-independent version of Eq. (7.155). Let us call any of these Te (x). Note that by
construction Te (x) satisfies the boundary conditions on T .
Let us subject a steady solution to a small perturbation and consider that to be our
initial condition for an unsteady calculation. Take then
T (x, 0) = Te (x) + A(x), A(1) = A(1) = 0, (7.156)
A(x) = O(1), 0 < << 1. (7.157)
Here A(x) is some function which satisfies the same boundary conditions as T (x, t).
Now let us assume that
T (x, t) = Te (x) + T (x, t). (7.158)
with
T (x, 0) = A(x). (7.159)
Here T is an O(1) quantity. We then substitute our Eq. (7.158) into Eq. (7.155) to get
1 2
(Te (x) + T (x, t)) = (Te (x) + T (x, t))
t D x2

+ (1 Te (x) T (x, t)) exp ,
1 + QTe (x) + QT (x, t)
(7.160)
From here on we will understand that Te is Te (x) and T is T (x, t). Now consider the
exponential term:
!
1
exp = exp Q
, (7.161)
1 + QTe + QT 1 + QTe 1 + 1+QT T
e

Q
exp 1 T , (7.162)
1 + QTe 1 + QTe

Q
exp exp T , (7.163)
1 + QTe (1 + QTe )2

Q
exp 1+ T . (7.164)
1 + QTe (1 + QTe )2
So our Eq. (7.160) can be rewritten as
1 2
(Te + T ) = (Te + T )
t D x2
Q
+ (1 Te T ) exp 1+ T ,
1 + QTe (1 + QTe )2
1 2
= (Te + T )
D x2
Q
+ exp (1 Te T ) 1 + T ,
1 + QTe (1 + QTe )2
1 2
= 2
(Te + T )
D x
(1 Te )Q 2
+ exp (1 Te ) + T 1 + + O( )
1 + QTe (1 + QTe )2

2T 1 2 Te
= + + exp (1 Te )
D x2 D x2 1 + QTe
| {z }
=0

(1 Te )Q 2
+ exp T 1 + + O( )
1 + QTe (1 + QTe )2
(7.165)
Now, we recognize the bracketed term as zero because Te (x) is constructed to satisfy the
steady state equation. We also recognize that Te (x)/t = 0. So our equation reduces to,
neglecting O(2 ) terms, and canceling

T 1 2T (1 Te )Q
= + exp 1 + T
t D x2 1 + QTe (1 + QTe )2
(7.166)
Equation (7.166) is a linear partial differential equation for T (x, t). It is of the form
T 1 2T
= + B(x)T , (7.167)
t D x2
with

(1 Te (x))Q
B(x) exp 1 + (7.168)
1 + QTe (x) (1 + QTe (x))2
7.4.1.2 Separation of variables

Let us use the standard technique of separation of variables to solve Eq. (7.167). We first
assume that
T (x, t) = H(x)K(t). (7.169)
So Eq. (7.167) becomes

dK(t) 1 d2 H(x)
H(x) = = K(t) + B(x)H(x)K(t), (7.170)
dt D dx2
1 dK(t) 1 1 d2 H(x)
= + B(x) = . (7.171)
K(t) dt D H(x) dx2
Since the left side is a function of t and the right side is a function of x, the only way the
two can be equal is if they are both the same constant. We will call that constant .
Now Eq. (7.171) really contains two equations, the first of which is
dK(t)
+ K(t) = 0. (7.172)
dt
This has solution
K(t) = C exp(t), (7.173)
where C is some arbitrary constant. Clearly if > 0, this solution is stable, with time
constant of relaxation = 1/.
The second differential equation contained within Eq. (7.171) is
1 d2 H(x)
2
+ B(x)H(x) = H(x), (7.174)
D
dx
1 d2
+ B(x) H(x) = H(x). (7.175)
D dx2
This is of the classical eigenvalue form for a linear operator L; that is L(H(x)) = H(x).
We also must have
H(1) = H(1) = 0, (7.176)

to satisfy the spatially homogenous boundary conditions on T (x, t).

This eigenvalue problem is difficult to solve because of the complicated nature of B(x).
Let us see how the solution would proceed in the limiting case of B as a constant. Then we
will generalize later.
If B is a constant, we have
d2 H
+ (B + )DH = 0, H(1) = H(1) = 0. (7.177)
dx2
The following mapping simplifies the problem somewhat:
x+1
y= . (7.178)
2
This takes our domain of x [1, 1] to y [0, 1]. By the chain rule
dH dH dy 1 dH
= = .
dx dy dx 2 dy
So
d2 H 1 d2 H
= .
dx2 4 dy 2
So our eigenvalue problem transforms to
d2 H
+ 4D(B + )H = 0, H(0) = H(1) = 0. (7.179)
dy 2
This has solution
p p
H(y) = C1 cos ( 4D(B + ) )y + C2 sin ( 4D(B + )) y . (7.180)
At y = 0 we have then
H(0) = 0 = C1 (1) + C2 (0), (7.181)
so C1 = 0. Thus
p
H(y) = C2 sin ( 4D(B + )) y . (7.182)
At y = 1, we have the other boundary condition:

p
H(1) = 0 = C2 sin ( 4D(B + )) . (7.183)
Since C2 6= 0 to avoid a trivial solution, we must require that

p
sin ( 4D(B + )) = 0. (7.184)
For this to occur, the argument of the sin function must be an integer multiple of :
p
4D(B + ) = n, n = 1, 2, 3, . . . (7.185)
Thus
n2 2
= B. (7.186)
4D
We need > 0 for stability. For large n and D > 0, we have stability. Depending on the
value of B, low n, which corresponds to low frequency modes, could be unstable.
7.4.1.3 Numerical eigenvalue solution

Let us return to the full problem where B = B(x). Let us solve the eigenvalue problem via
the method of finite differences. Let us take our domain x [1, 1] and discretize into N
points with
2
x = , xi = (i 1)x 1. (7.187)
N 1
Note that when i = 1, xi = 1, and when i = N, xi = 1. Let us define B(xi ) = Bi and
H(xi ) = Hi .
We can rewrite Eq. (7.174) as
d2 H(x)
+ D(B(x) + )H(x) = 0, H(1) = H(1) = 0. (7.188)
dx2
Now let us apply an appropriate equation at each node. At i = 1, we must satisfy the
boundary condition so
H1 = 0. (7.189)
At i = 2, we discretize Eq. (7.188) with a second order central difference to obtain
H1 2H2 + H3
+ D(B2 + )H2 = 0. (7.190)
x2
We get a similar equation at a general interior node i:
Hi1 2Hi + Hi+1

+ D(Bi + )Hi = 0. (7.191)
x2
At the i = N 1 node, we have
HN 2 2HN 1 + HN
+ D(BN 1 + )HN 1 = 0. (7.192)
x2
At the i = N node, we have the boundary condition

HN = 0. (7.193)
These represent a linear tridiagonal system of equations of the form

2
( Dx 2 + B2 )
1
0 0 ... 0 H2 H2
Dx2
1 2
( Dx 2 + B3 )
1
0 ... 0 H H3
.3
. = ... (7.194)
Dx2 Dx2
0 1
. . . . . . . . . . . . .
Dx 2 . .
... ... . . . . . . . . . . . . .. ..
0 0 ... ... ... ... .. ..
| {z } | {z. .
}
=L =h
This is of the classical linear algebraic eigenvalue form L h = h. All one need do is
discretize and find the eigenvalues of the matrix L. These will be good approximations to the
eigenvalues of the differential operator L. The eigenvectors of L will be good approximations
of the eigenfunctions of L. To get a better approximation, one need only reduce x.
Note because the matrix L is symmetric, the eigenvalues are guaranteed real, and the
eigenvectors are guaranteed orthogonal. This is actually a consequence of the original prob-
lem being in Sturm-Liouville form, which is guaranteed to be self-adjoint with real eigenvalues
and orthogonal eigenfunctions.
7.4.1.3.1 Low temperature transients For our case of = 0.4, Q = 1, = 15 (so

D = 87173.8), we can calculate the stability of the low temperature solution. Choosing
N = 101 points to discretize the domain, we find a set of eigenvalues. They are all positive,
so the solution is stable. The first few are
= 0.0000232705, 0.000108289, 0.000249682, 0.000447414, . . . . (7.195)
The first few eigenvalues can be approximated by inert theory with B(x) = 0, see Eq. (7.186):
n2 2
= 0.0000283044, 0.000113218, 0.00025474, 0.00045287, . . . . (7.196)
4D
The first eigenvalue is associated with the longest time scale = 1/0.0000232705 =
42972.9 and a low frequency mode, whose shape is given by the associated eigenvector,
plotted in Fig 7.8. This represents the fundamental mode. The first harmonic mode is
associated with the next eigenfunction, which is plotted in Fig 7.9. The second harmonic
mode is associated with the next eigenfunction, which is plotted in Fig 7.10.
7.4.1.3.2 Intermediate temperature transients For the intermediate temperature

solution with T m = 0.417, we find the first few eigenvalues to be
= 0.0000383311, 0.0000668221, 0.000209943, . . . (7.197)
Except for the first, all the eigenvalues are positive. The first eigenvalue of = 0.0000383311
is associated with an unstable fundamental mode. All the harmonic modes are stable. We
plot the first three modes in Figs. 7.11-7.13.
0.14
0.12
0.10
0.08
0.06
0.04
0.02
x
-1.0 -0.5 0.5 1.0
Figure 7.8: Plot of lowest frequency, fundamental mode, eigenfunction versus x, with = 0.4,
Q = 1, = 15, low temperature steady solution Te (x).
0.10
0.05
x
-1.0 -0.5 0.5 1.0
-0.05
-0.10
Figure 7.9: Plot of first harmonic eigenfunction versus x, with = 0.4, Q = 1, = 15, low
temperature steady solution Te (x).
0.10
0.05
x
-1.0 -0.5 0.5 1.0
-0.05
-0.10
Figure 7.10: Plot of second harmonic eigenfunction versus x, with = 0.4, Q = 1, = 15,
low temperature steady solution Te (x).
0.14
0.12
0.10
0.08
0.06
0.04
0.02
x
-1.0 -0.5 0.5 1.0
Figure 7.11: Plot of fundamental eigenfunction versus x, with = 0.4, Q = 1, = 15,

intermediate temperature steady solution Te (x).
0.10
0.05
x
-1.0 -0.5 0.5 1.0
-0.05
-0.10
Figure 7.12: Plot of first harmonic eigenfunction versus x, with = 0.4, Q = 1, = 15,
0.10
0.05
x
-1.0 -0.5 0.5 1.0
-0.05
-0.10
x
-1.0 -0.5 0.5 1.0
-0.05
-0.10
-0.15
Figure 7.14: Plot of fundamental eigenfunction versus x, with = 0.4, Q = 1, = 15, high
temperature steady solution Te (x).
0.15
0.10
0.05
x
-1.0 -0.5 0.5 1.0
-0.05
-0.10
-0.15
Figure 7.15: Plot of first harmonic eigenfunction versus x, with = 0.4, Q = 1, = 15,
high temperature steady solution Te (x).
7.4.1.3.3 High temperature transients For the high temperature solution with T m =
0.987, we find the first few eigenvalues to be
= 0.000146419, 0.00014954, 0.000517724, . . . (7.198)
All the eigenvalues are positive, so all modes are stable. We plot the first three modes in
Figs. 7.14-7.16.
7.4.2 Full transient solution

We can get a full transient solution to Eqs. (7.155) with numerical methods. We omit details
of such numerical methods, which can be found in standard texts.
7.4.2.1 Low temperature solution

For our case of = 0.4, Q = 1, = 15 (so D = 87173.8), we show a plot of the full transient
solution in Fig. 7.17. We see in Fig. 7.18 that the centerline temperature T (0, t) relaxes to
0.15
0.10
0.05
x
-1.0 -0.5 0.5 1.0
-0.05
-0.10
high temperature steady solution Te (x).
300 000
200 000
100 000
0
0.015
0.010
0.005
0.000
-1.0
-0.5
0.0
0.5
1.0
Figure 7.17: Plot of T (x, t), with = 0.4, Q = 1, = 15.

0.015
0.010
0.005
t
50 000 100 000 150 000 200 000 250 000 300 000
Figure 7.18: Plot of T (0, t) along with the long time exact low temperature solution, Te (0),
with = 0.4 Q = 1, = 15.
1.5 106
6
1. 10
500 000
0
1.0
0.5
0.0
-1.0
-0.5
0.0
0.5
1.0
Figure 7.19: Plot of T (x, t), with = 1.2, Q = 1, = 15.
the long time value predicted by the low temperature steady solution:
lim T (0, t) = 0.016. (7.199)

t
7.4.2.2 High temperature solution

We next select a value of = 1.2 > c . This should induce transition to a high temperature
solution. We maintain = 15, Q = 1. We get D = e //Q = 261521. The full transient
solution is shown in Fig. 7.19
Figure 7.20 shows the centerline temperature T (0, t). We see it relaxes to the long time
value predicted by the high temperature steady solution:
lim T (0, t) = 0.9999185. (7.200)

t
T
1.0
0.8
0.6
0.4
0.2
t
200 000 400 000 600 000 800 000 1. 106 1.2 106 1.4 106
Figure 7.20: Plot of T (0, t) along with the long time exact high temperature exact solution
with = 1.2, Q = 1, = 15.
It is clearly seen that there is a rapid acceleration of the reaction for t 106 . This compares
with the prediction of the induction time from the infinite Damkoler number, D ,
thermal explosion theory of explosion to occur when
e e15
t = = 2.17934 105 . (7.201)
Q (1)(15)
The estimate underpredicts the value by a factor of five. This is likely due to 1) cooling
of the domain due to the low temperature boundaries at x = 1, and 2) effects of finite
activation energy.
Chapter 8
Simple laminar flames:

Reaction-advection-diffusion
Here we will consider premixed one-dimensional steady laminar flames. Background is avail-
able in many sources. 1 2 3
This topic is broad, but we will restrict attention to the simplest cases. We will consider
a simple reversible kinetics model
A B, (8.1)
where A and B have identical molecular masses, MA = MB = M, and are both calorically
perfect ideal gases with the same specific heats, cP A = cP B = cP . Because we are modeling
the system as premixed, we consider species A to be composed of molecules which have their
own fuel and oxidizer. This is not common in hydrocarbon kinetics, but is more so in the
realm of explosives. More common in gas phase kinetics are situations in which cold fuel and
cold oxidizer react together to form hot products. In such a situation, one can also consider
non-premixed flames in which streams of fuel first must mix with streams of oxidizer before
significant reaction can commence. Such flames will not be considered in this chapter.
We shall see that the introduction of advection introduces some unusual mathematical
difficulties in properly modeling cold unreacting flow. This will be overcome in a way which is
aesthetically unappealing, but nevertheless useful: we shall introduce an ignition temperature
Tig in our reaction kinetics law to suppress all reaction for T < Tig . This will serve to render
the cold boundary to be a true mathematical equilibrium point of the model; this is not a
traditional chemical equilibrium, but it is an equilibrium in the formal mathematical sense
nonetheless.
1
J. D. Buckmaster and G. S. S. Ludford, 2008, Theory of Laminar Flames, Cambridge, Cambridge.
2
F. A. Williams, 1985, Combustion Theory, Benjamin-Cummings, Menlo Park, California.
3
C. K. Law, 2006, Combustion Physics, Cambridge, Cambridge.
281
282CHAPTER 8. SIMPLE LAMINAR FLAMES: REACTION-ADVECTION-DIFFUSION
8.1 Governing Equations

8.1.1 Evolution equations
8.1.1.1 Conservative form
Let us commence with the one-dimensional versions of Eqs. (6.1-6.3) and (6.5):

+ (u) = 0, (8.2)
t x

(u) + u2 + P = 0, (8.3)
t x
1 2 1 2 q
e+ u + u e + u + j + (P )u = 0, (8.4)
t 2 x 2

(YB ) + (uYB + jBm ) = M B . (8.5)
t x
8.1.1.2 Non-conservative form

Using standard reductions similar to those made to achieve Eqs. (6.24),(6.27),(6.32), (6.40),
the non-conservative form of the governing equations, Eqs. (8.2-8.5) is
u
+u + = 0, (8.6)
t x x
u u P
+ u + = 0, (8.7)
t x x x
e e j q u u
+ u + +P = 0, (8.8)
t x x x x
YB YB jBm
+ u + = M B . (8.9)
t x x
8.1.1.3 Formulation using enthalpy

It will be more useful to formulate the equations using enthalpy. First rewrite Eqs. (8.6-8.9)
using the material derivative, d/dt = /t + u/x:
d u
+ = 0, (8.10)
dt x
du P
+ = 0, (8.11)
dt x x
de j q u u
+ +P = 0, (8.12)
dt x x x
dYB jBm
+ = M B . (8.13)
dt x
8.1. GOVERNING EQUATIONS 283
Now recall the definition for enthalpy, Eq. (3.78), h = e + P v = e + P/, to get an
expression for dh:
P 1
dh = de 2
d + dP, (8.14)

dh de P d 1 dP
= + , (8.15)
dt dt 2 dt dt
de P d dh 1 dP
2 = , (8.16)
dt dt dt dt
de P d dh dP
= . (8.17)
dt dt dt dt
Now using the mass equation, Eq. (8.10) to eliminate P u/x in the energy equation,
Eq. (8.12), the energy equation can be rewritten as
de j q P d u
+ = 0, (8.18)
dt x dt x
Next use Eq. (8.17) to simplify Eq. (8.18):
dh j q dP u
+ = 0, (8.19)
dt x dt x
(8.20)
8.1.1.4 Low Mach number limit

In the limit of low Mach number, one can do a formal asymptotic expansion with the re-
ciprocal of the Mach number squared as a perturbation parameter. All variables take the
form = o + (1/M 2 )1 , where is a general variable. In this limit, the linear momentum
equation can be shown to reduce at leading order to Po /x = 0, giving rise to Po = Po (t).
We shall ultimately be concerned only with time-independent flows where P = Po . We
adopt the constant pressure assumption now. Also in the low Mach number limit, it can be
shown that viscous work is negligible, so u/x 0. Our evolution equations then in the
low Mach number, constant pressure limit are
u
+u + = 0, (8.21)
t x x
h h j q
+ u + = 0, (8.22)
t x x
YB YB jBm
+ u + = M B . (8.23)
t x x
Note that we have effectively removed the momentum equation. It would re-appear in a
non-trivial way at the next order.
Equations (8.21-8.23) take on the conservative form

+ (u) = 0, (8.24)
t x

(h) + (uh + j q ) = 0, (8.25)
t x

(YB ) + (uYB + jBm ) = M B . (8.26)
t x
8.1.2 Constitutive models

Equations (8.21-8.23) are supplemented by simple constitutive models:
Po = RT, (8.27)
h = cP (T To ) YB q, (8.28)
YB
jBm = D , (8.29)
x
T YB
j q = k + Dq , (8.30)
x x

E/(RT ) 1 YB
B = aT e (1 YB ) 1 H(T Tig ), (8.31)
|M {z } Kc 1Y
| {z B }
=A =B /A
q/(RT )
Kc = e (8.32)
Thus we have nine equations for the nine unknowns, , u, h, T , j q , YB , jBm , B , Kc .

Many of these are obvious. Some are not. First, we note the new factor H(T Tig ) in
our kinetics law, Eq. (8.31). Here H is a Heaviside unit step function. For T < Tig it takes
a value of zero. For T Tig , it takes a value of unity. Next, let us see how to get Eq. (8.29)
from the more general Eq. (6.64). We first take the thermal diffusion coefficient to be zero,
DiT = 0. We also note since MA = MB = M that yk = Yk ; that is mole fractions are the
same as mass fractions. So Eq. (6.64) simplifies considerably to
N
X Yk
jim = Dik . (8.33)
k=,1k6=i
x
Next we take Dik = D and write for each of the two diffusive mass fluxes that
YB
jAm = D , (8.34)
x
YA
jBm = D . (8.35)
x
Since YA + YB = 1, we have also
YA
jAm = D , (8.36)
x
YB
jBm = D . (8.37)
x
Note that jAm + jBm = 0, as required.

Under the same assumptions, Eq. (6.63) reduces to
T
j q = k + jAm hA + jBm hB , (8.38)
x
T
= k + jAm (hA,To + cP (T To )) + jBm (hB,To + cP (T To )), (8.39)
x
T
= k + jAm hA,To + jBm hB,To + (jAm + jBm ) cP (T To ), (8.40)
x | {z }
=0
T
= k jBm hA,To + jBm hB,To , (8.41)
x
T
= k jBm (hA,To hB,To ), (8.42)
x | {z }
=q
T
j q = k jBm q, (8.43)
x
T YB
jq = k + Dq , (8.44)
x x
(8.45)
For the equilibrium constant Kc , we recall that

Go
Kc = exp , (8.46)
RT

(g oB g oA )
= exp , (8.47)
RT
o
gA gBo
= exp , (8.48)
RT
o
hA T soA (hoB T soB )
= exp , (8.49)
RT
o o
hA hoB sB soA
= exp exp , (8.50)
RT R
o o
hA,To hoB,To sB,To soA,To
= exp exp (8.51)
RT R
Here because of constant specific heats for A and B, many terms have cancelled. Let us take
now soB,To = soA,To and our definition of the heat release q to get
q
Kc = exp . (8.52)
RT
8.1.3 Alternate forms

Further analysis of the evolution equations combined with the constitutive equations can
yield forms which give physical insight.
8.1.3.1 Species equation

If we combine our species evolution equation, Eq. (8.23) with the constitutive law, Eq. (8.29),
we get

YB YB YB
+ u D = M B . (8.53)
t x x x
Note that for flows with no advection (u = 0), constant density ( = constant), and no
reaction ( B = 0), this reduces to the classical heat equation from mathematical physics
YB /t = D( 2 YB /x2 ).
8.1.3.2 Energy equation

Consider the energy equation, Eq. (8.22), using Eqs. (8.28,8.30) to eliminate h and j q :

T YB
(cP (T To ) YB q) +u (cP (T To ) YB q) + k + Dq = 0,
t | {z } x | {z } x x x
=h =h | {z }
=j q
(8.54)

q q k T YB q
T YB + u T YB D = 0,
t cP x cP x cP x x cP
(8.55)

T T k 2T q YB YB YB
+ u 2
+ u D = 0,
t x cP x cP t x x x
| {z }
=M B
(8.56)
We notice that the terms involving YB simplify via Eq. (8.53) to yield an equation for
temperature evolution
T T k 2T q
+ u M B = 0, (8.57)
t x cP x2 cP
T T k 2T q
+u = 2
+ M B , (8.58)
t x cP x cP
|{z}
=
We note that thermal diffusivity is

k
= . (8.59)
cP
We note here that this definition, convenient and in the combustion literature, is slightly
non-traditional as it involves a variable property . So the reduced energy equation is
T T 2T q
+u = 2
+ M B , (8.60)
t x x cP
Note that for exothermic reaction, q > 0 accompanied with production of product B, B > 0,
induces a temperature rise of a material particle. The temperature change is modulated by
energy diffusion. In the inert zero advection limit, we recover the ordinary heat equation
T /t = ( 2 T /x2 ).
8.1.3.3 Schwab-Zeldovich form

Let us now adopt the assumption that mass and energy diffuse at the same rate. This is
not difficult to believe as both are molecular collision phenomena in gas flames. We note
the dimensionless ratio of energy diffusivity to mass diffusivity D is known as the Lewis 4
number, Le:

Le = . (8.61)
D
If we insist that mass and energy diffuse at the same rate, we have Le = 1, which gives
k
D== . (8.62)
cP
With this assumption, the energy equation, Eq. (8.55), takes the form

q q k T k YB q
T YB + u T YB = 0,
t cP x cP x cP x cP x cP
(8.63)

q q k q
T YB + u T YB T YB = 0,
t cP x cP cP x x cP
(8.64)

q q k q
T YB +u T YB T YB = 0,
t cP x cP cP x x cP
(8.65)
4
named after Warren K. Lewis, 1882-1975, MIT chemical engineering professor.
We rewrite in terms of the material derivative

d q k 2 q
T YB = T YB . (8.66)
dt cP cP x2 cP
Equation (8.66) holds that for a material fluid particle, the quantity T YB q/cP changes
only in response to local spatial gradients. Now if we consider an initial value problem in
which T and YB are initially spatially uniform and there are no gradients of either T or
YB at x , then there will no tendency for any material particle to have its value of
T YB q/cP change. Let us assume that at t = 0, we have T = To and no product, so YB = 0.
Then we can conclude that the following relation holds for all space and time:
q
T YB = To . (8.67)
cP
Solving for YB , we get
cP (T To )
YB = . (8.68)
q
For this chapter, we are mainly interested in steady waves. We can imagine that our steady
waves are the long time limit of a situation just described which was initially spatially
uniform.
Compare Eq. (8.68) to our related ad hoc assumption for reactive solids, Eq. (7.25). They
are essentially equivalent, especially when one recalls that plays the same role as YB and
for the reactive solid cP cv .
We can use Eq. (8.68) to eliminate YB in the species equation, Eq. (8.53) to get a single
equation for temperature evolution. First adopt the equal diffusion assumption, Eq. (8.62),
in Eq. (8.53):

YB YB k YB
+ u = M B . (8.69)
t x x cP x
Next eliminate YB :

cP (T To ) cP (T To ) k cP (T To )
+ u = M B .
t q x q x cP x q
(8.70)
Simplifying, we get

T T 2T
cP +u k = qM B . (8.71)
t x x2
Now let us use Eq. (8.31) to eliminate B :

T T 2T
cP +u k 2
t x x

E/(RT ) 1 YB
= qaT e (1 YB ) 1 H(T Tig ). (8.72)
Kc 1 Y B
Now use Eq. (8.52) to eliminate Kc and Eq. (8.68) to eliminate YB so to get

T T 2T
cP +u k 2
t x x
cP (T To )
!
c P (T To ) q
= qaT eE/(RT ) 1 1 eq/(RT ) cP (T To )
H(T Tig ). (8.73)
q 1 q
Equation (8.73) is remarkably similar to our earlier Eq. (7.27) for temperature evolution in
a heat conducting reactive solid. The differences are as follows:
We use cP instead of cv ,
We have included advection,
We have accounted for finite ,
We have accounted for reversible reaction.
We have imposed and ignition temperature.
8.1.4 Equilibrium conditions

We can gain insights examining when Eq. (8.73) is in equilibrium. There is one potential
equilibria, the state of chemical equilibrium where Kc = YB /(1 YB ) = YB /YA . This occurs
when
cP (T To )
q/(RT ) q
1e cP (T To )
= 0. (8.74)
1 q
This is a transcendental equation for T . Let us examine how the equilibrium T varies for
some sample parameters. For this and later calculations, we give a set of parameters in
Table 8.1. Let us consider how the equilibrium temperature varies as q is varied from near
zero to its maximum value of Table 8.1. The chemical equilibrium flame temperature is
plotted as a function of q in Figure 8.1. For q = 0, the equilibrium temperature is the
ambient temperature. As q increases, the equilibrium temperature increases monotonically.
The corresponding equilibrium mass fraction as a function of q is given in Fig. 8.2. When
q = 0, the equilibrium mass fraction is YB = 0.5; that is, there is no tendency for either
products or reactants. As q increases, the tendency for product increases and YB 1.
There is also the state of complete reaction when YB = 1, which corresponds to
cP (T To )
1 = 0, (8.75)
q
q
T = To + . (8.76)
cP
Parameter Value Units

Fundamental Dimensional
To 300 K
cP 1000 J/kg/K
R 287 J/kg/K
q 1.5 106 J/kg
E 1.722 105 J/kg
o 105 m2 /s
a 105 1/s
uo 1.4142 m/s
Po 105 Pa
Tig 600 K
Secondary Dimensional
o 1.1614 kg/m3
k 0.0116114 J/m/s
Fundamental Dimensionless
1.4025
Q 5
2
D 2
TIG 0.2
Table 8.1: Numerical values of parameters for simple laminar flame calculation.
8.2. STEADY BURNER-STABILIZED FLAMES 291
T eq HKL
2000
1500
1000
500
q HJkgL
0 750 000 1 500 000
Figure 8.1: Variation of chemical equilibrium temperature with heat release.
YB eq
1.0
0.8
0.6
0.4
0.2
q HJkgL
0 750 000 1 500 000
Figure 8.2: Variation of chemical equilibrium product mass fraction with heat release.
Numerically, at our maximum value of q = 1.5 106 J/kg, we get complete reaction when
T = 1800 K. Note this state is not an equilibrium state unless Kc . As x ,
we expect a state of chemical equilibrium and a corresponding high temperature. At some
intermediate point, the temperature will have its value at ignition, Tig . We will ultimately
transform our domain so the ignition temperature is realized at x = 0, which we can associate
with a burner surface. WE say that the flame is thus anchored to the burner. Without such
a prescribed anchor, our equations are such that any translation in x will yield an invariant
solution, so the origin of x is arbitrary.
8.2 Steady burner-stabilized flames

Let us consider an important problem in laminar flame theory: that of a burner-stabilized
flame. We shall consider a doubly-infinite domain, x (, ). We will assume that as
x , we have a fresh unburned stream of reactant A, YA = 1, YB = 0, with known
velocity uo , density o and temperature To . The values of o and To will be consistent with
a state equation so that the pressure has a value of Po .
8.2.1 Formulation
Let us consider our governing equations in the steady wave frame where there is no variation
with t: /t = 0. This, coupled with our other reductions, yields the system
d
(u) = 0, (8.77)
dx
dT d2 T E/(RT ) cP (T To )
ucP k 2 = qaT e 1
dx dx q
cP (T To )
!
q
1 eq/(RT ) cP (T To )
H(T Tig ), (8.78)
1 q
Po = RT. (8.79)
Now we have three equations for the three remaining unknowns, , u, T .

For a burner-stabilized flame, we assume at x that we know the velocity, uo , the
temperature, To , and the density, o . We can integrate the mass equation to get
u = o uo . (8.80)
We then rewrite the remaining differential equation as

dT d2 T Po 1 E/(RT ) cP (T To )
o uo cP k 2 = qaT e 1
dx dx R q
cP (T To )
!
q
1 eq/(RT ) cP (T To )
H(T Tig ). (8.81)
1 q
Our boundary conditions for this second order differential equation are
dT
0, x . (8.82)
dx
We will ultimately need to integrate this equation numerically, and this poses some difficulty
because the boundary conditions are applied at . We can overcome this in the following
way. Note that our equations are invariant under a translation in x. We will choose some
large value of x to commence numerical integration. We shall initially assume this is near
the chemical equilibrium point. We shall integrate backwards in x. We shall note the value
of x where T = Tig . Beyond that point, the reaction rate is zero, and we can obtain an
exact solution for T . We will then translate all our results so that the ignition temperature
is realized at x = 0.
Let us scale temperature so that

q
T = To 1 + T . (8.83)
cP To
Inverting, we see that

cP (T To )
T = . (8.84)
q
We get a dimensionless ignition temperature TIG of
cP (Tig To )
TIG = . (8.85)
q
Let us also define a dimensionless heat release Q as
q
Q= . (8.86)
cP To
Thus
T
= 1 + QT . (8.87)
To
Note that T = 1 corresponds to our complete reaction point where YB = 1. Let us also
restrict ourselves, for convenience, to the case where = 0. With these scalings, and with
E
= , (8.88)
RTo
our differential equations become

dT d2 T Po 1 T
o uocP To Q kTo Q 2 = qa exp
dx dx R 1 + QT
To (1 + QT )

! !
1
Q T
1 exp H(T TIG ),(8.89)
1 + QT 1 T
dT
0, x . (8.90)
dx
Let us now scale both sides by o uo cP To to get

dT k 1 d2 T Po q a 1 T
Q Q 2 = exp
dx o cP uo dx o RTo cP To uo 1 + QT 1 + QT
| {z } | {z }
=1 =Q

! !
1
Q T
1 exp H(T TIG ), (8.91)
1 + QT 1 T
Realizing that Po = o RTo , the ambient thermal diffusivity, o = k/(o cP ) and cancelling Q,
we get

dT 1 d2 T a 1 T
o = exp
dx uo dx2 uo 1 + QT 1 + QT

! !
1
Q T
1 exp H(T TIG ), (8.92)
1 + QT 1 T
Now let us scale x by a characteristic length, L = uo /a. This length scale is dictated by
a balance between reaction and advection. When the reaction is fast, a is large, and the
length scale is reduced. When the incoming velocity is fast, uo is large, and the length scale
is increased.
x ax
x = = . (8.93)
L uo
With this choice of length scale, Eq. (8.92) becomes

dT a d2 T 1 T
o 2 = exp
dx uo dx2 1 + QT 1 + QT

! !
1
Q T
1 exp H(T TIG ), (8.94)
1 + QT 1 T
Now, similar to Eq. (7.37), we take the Damkohler number D to be
aL2 u2
D= = o . (8.95)
o ao
In contrast to Eq. (7.37) the Damkohler number here includes the effects of advection. With
this choice, Eq. (8.94) becomes

dT 1 d2 T 1 T
= exp
dx D dx2 1 + QT 1 + QT

! !
1
Q T
1 exp H(T TIG ), (8.96)
1 + QT 1 T
Lastly, let us dispose with the notation and understand that all variables are dimen-
sionless. Our differential equation and boundary conditions become

dT 1 d2 T 1T
= exp
dx D dx2 1 + QT 1 + QT

! !
1
Q T
1 exp H(T TIG ), (8.97)
1 + QT 1 T

dT
0. (8.98)
dx x
Equation (8.97) is remarkably similar to our Frank-Kamenetskii problem embodied in
Eq. (7.60). The major differences are reflected in that Eq. (8.97) accounts for
advection,
variable density,
reversible reaction,
doubly-infinite spatial domain.
8.2.2 Solution procedure

There are some unusual challenges in the numerical solution of the equations for laminar
flames. Formally we are solving a two-point boundary value problem on a doubly-infinite
domain. The literature is not always coherent on solutions methods or the interpretations
of results. At the heart of the issue is the cold boundary difficulty. We have patched
this problem via the use of an ignition temperature built into a Heavside function; see
Eq. (8.31). We shall see that this patch has advantages and disadvantages. Here we will
not dwell on nuances, but will present a result which is mathematically sound and offer
interpretations. Our result will use standard techniques of non-linear analysis from dynamic
systems theory. We will pose the problem as a coupled system of first order non-linear
differential equations, find their equilibria, use local linear analysis to ascertain the stability
of the chemical equilibrium fixed point, and use numerical integration to calculate the non-
linear laminar flame structure.
8.2.2.1 Model linear system

Before commencing with the difficult Eq. (8.97), let us consider a related linear model system,
given here:

dT d2 T dT
3 2 = 4(1 T )H(T TIG ), 0. (8.99)
dx dx dx x
The forcing is removed when either T = 1 or T < TIG .

Defining q dT /dx, we rewrite our model equation as a linear system of first order
equations:
dq
= 3q + 4(1 T )H(T TIG ), q() = 0, (8.100)
dx
dT
= q, q() = 0. (8.101)
dx
The system has an equilibrium point at q = 0 and T = 1. The system is also in equilibrium
when q = 0 and T < TIG , but we will not focus on this. Taking T = T 1, the system near
the equilibrium point is

d q 3 4 q
= . (8.102)
dx T 1 0 T
The local Jacobian matrix has two eigenvalues = 4 and = 1. Thus the equilibrium is
a saddle.
Considering the solution away from the equilibrium point, but still for T > TIG , and
returning from T to T , the solution takes the form

q 0 1 x C1 + 4C2 1 4x 4C1 4C2
= + e + e (8.103)
T 1 5 C1 + 4C2 5 C1 + C2
We imagine as x that T approaches unity from below. We wish this equilibrium to be

achieved. Thus, we must suppress the unstable = 4 mode; this is achieved by requiring
C1 = C2 . This yields

q 0 1 x 5C1 0 x 1
= + e = + C1 e (8.104)
T 1 5 5C1 1 1
We can force T (xIG ) = TIG by taking C1 = (1 TIG )exIG . The solution for x > xIG is
T = 1 (1 TIG )e(xxIG ) , (8.105)
q = (1 TIG )e(xxIG ) . (8.106)
At the interface x = xIG , we have T (xIG ) = TIG and q(xIG ) = (1 TIG ).
For x < xIG , our system reduces to

dT d2 T dT
3 2 = 0, T (xIG ) = TIG , = 0. (8.107)
dx dx dx x
Solutions in this region take the form

T = TIG e3(xxIG ) + C 1 e3(xxIG ) , (8.108)
q = 3TIG e3(xxIG ) + 3Ce3(xxIG ) . (8.109)
All of them have the property that q 0 as x . One is faced with the question of how
to choose the constant C. Let us choose it to match the energy flux predicted at x = xIG .
At the interface x = xIG we get T = TIG and q = 3TIG + 3C. Matching values of q at the
interface, we get
(1 TIG ) = 3TIG + 3C. (8.110)

Solving for C, we get
4TIG 1
C= . (8.111)
3
So we find for x < xIG that

3(xxIG ) 4TIG 1
T = TIG e + 1 e3(xxIG ) , (8.112)
3
q = 3TIG e3(xxIG ) + (4TIG 1) e3(xxIG ) . (8.113)
8.2.2.2 System of first order equations

Let us apply this technique to the full non-linear Eq. (8.97). First, again define the non-
dimensional Fourier heat flux q as
dT
q= . (8.114)
dx
Note this is a mathematical convenience. It is not the full diffusive energy flux as it makes no
account for mass diffusion affects. However, it is physically intuitive. With this definition,
we can rewrite Eq. (8.114) along with Eq. (8.97) as

dq 1T
= Dq + D exp
dx 1 + QT 1 + QT

! !
1
Q T
1 exp H(T TIG ), (8.115)
1 + QT 1 T
dT
= q. (8.116)
dx
We have two boundary conditions for this problem:
q() = 0. (8.117)
We are also going to fix our coordinate system so that T = TIG at x = xIG 0. We will
require continuity of T and dT /dx at x = xIG = 0.
Our non-linear system has the form

d q f (q, T )
= (8.118)
dx T g(q, T )
8.2.2.3 Equilibrium
Our equilibrium condition is f (q, T ) = 0, g(q, T ) = 0. By inspection, equilibrium is found
when
q = 0, (8.119)

!
1
Q T
1 exp = 0 or T < TIG . (8.120)
1 + QT 1T
We will focus most attention on the isolated equilibrium point which corresponds to tradi-
tional chemical equilibrium. We have considered the dimensional version of the same equilib-
rium condition for T in an earlier section. The solution is found via solving a transcendental
equation for T .
8.2.2.4 Linear stability of equilibrium

Linear stability of the equilibrium point can be found by examining the eigenvalues of the
Jacobian matrix J:
f f
D f
J = g g =
q T T eq (8.121)
q T eq
1 0
This Jacobian has eigenvalues of

s !
D 4 f
= 1 1 2 (8.122)
2 D T eq
p
For D >> f /T , the eigenvalues are approximated well by

1 f
1 D, 2 . (8.123)
D T eq
p
This gives rise to a stiffness ratio, valid in the limit of D >> f /T , of

1 2
D (8.124)
2 f
T
We adopt the parameters of Table 8.1. With these, we find the dimensionless chemical
equilibrium temperature at
T eq = 0.953. (8.125)
This corresponds to a dimensional value of 1730.24 K. Our parameters induce a dimensional

length scale of uo /a = 1.4142 105 m. This is smaller than actual flames, which actually
have much larger values of D. So as to de-stiffen the system, we have selected a smaller than
normal value for D. For these values, we get a Jacobian matrix of

2 0.287
J= . (8.126)
1 0
We get eigenvalues near the equilibrium point of
= 2.134, = 0.134. (8.127)
The equilibrium is a saddle point. Note the negative value of f /T at equilibrium gives rise
to the saddle character of the equilibrium. The actual stiffness ratio is |2.134/(0.134)| =
15.9. This is well predicted by our simple formula which gives |22 /(0.287)| = 13.9.
In principle, we could also analyze the set of equilibrium points along the cold boundary,
where q = 0 and T < TIG . Near such points, linearization techniques fail because of the
nature of the Heaviside function. We would have to perform a more robust analysis. As an
alternative, we shall simply visually examine the results of calculations and infer the stability
of this set of fixed points.
8.2.2.5 Laminar flame structure

8.2.2.5.1 TIG = 0.2. Let us get a numerical solution by integrating backwards in space
from the equilibrium point. Here we will use the parameters of Table 8.1, in particular
to contrast with a later calculation, the somewhat elevated ignition temperature of TIG =
0.2. We shall commence the solution near the isolated equilibrium point corresponding to
chemical equilibrium. We could be very careful and choose the initial condition to lie just
off the saddle along the eigenvector associated with the negative eigenvalue. It will work
just as well to choose a point on the correct side of the equilibrium. Let us approximate
x by xB , require q(xB ) = 0, T (xB ) = 1 , where 0 < << 1. That is we perturb the
temperature to be just less than its equilibrium value. We integrate from the large positive
x = xB back towards x = . When we find T = TIG , we record the value of x. We
translate the plots so that the ignition point is reached at x = 0 and give results.
Predictions of T (x) are shown in in Figure 8.3. We see the temperature has T (0) =
T
1.0
T = T eq =0.953
0.8
0.6
0.4
0.2
x
-5 0 5 10 15 20 25
Figure 8.3: Dimensionless temperature as a function of position for a burner-stabilized

premixed laminar flame, TIG = 0.2.
0.2 = TIG . As x , the temperature approaches its equilibrium value. For x < 0, the
temperature continues to fall until it comes to a final value of T 0.132. Note that this
value cannot be imposed by the boundary conditions, since we enforce no other condition
on T except to anchor it at the ignition temperature at x = 0.
Predictions of q(x) are shown in in Figure 8.4. The heat flux is always negative. And it
clearly relaxes to zero as x . This indicates that energy released in combustion makes
its way back into the fresh mixture, triggering combustion of the fresh material. This is the
essence of the laminar diffusion flame. Predictions of (x) 1/(1 + QT (x)) are shown in in
Figure 8.5. We are somewhat troubled because (x) nowhere takes on dimensionless value of
unity, despite it being scaled by o . Had we anchored the flame at an ignition temperature
very close to zero, we in fact would have seen approach unity at the anchor point of the
flame. We chose an elevated ignition temperature so as to display the actual idiosyncrasies
of the model. Even for a flame anchored at Tig To , we would find a significant decay of
density in the cold region of the flow for x < xig .
q
x
-5 5 10 15 20 25
-0.02
-0.04
-0.06
-0.08
-0.10
-0.12
Figure 8.4: Dimensionless Fourier heat flux as a function of distance for a burner-stabilized

0.8
0.6
0.4
0.2
x
-5 0 5 10 15 20 25 30
Figure 8.5: Dimensionless density as a function of distance for a burner-stabilizedpremixed

laminar flame, TIG = 0.2.
Predictions of u(x) 1+QT (x) are shown in in Figure 8.6. Similar to , we are somewhat
troubled that the dimensionless velocity does not take a value near unity. Once again, had
we set Tig To , we would have found the u at x = xig would have taken a value very
near unity. But it also would modulate significantly in the cold region x < xig . We also
would have discovered that had Tig To that the temperature in the cold region would
drop significantly below To , which is a curious result. Alternatively, we could iterate on the
parameter Tig until we found a value for which limx T 0. For such a value, we would
also find u and to approach unity as x .
We can better understand the flame structure by considering the (T, q) phase plane as
shown in Fig. 8.7. Here, green denotes equilibria. We see the isolated equilibrium point at
(T, q) = (0.953, 0). And we see a continuous one-dimensional set of equilibria for {(T, q)|q =
0, T (, TIG ]}. Blue lines are trajectories in the phase space. The arrows have been
associated with movement in the x direction. This is because this is the direction in which
it is easiest to construct the flame structure.
u
6
x
-5 0 5 10 15 20 25 30
Figure 8.6: Dimensionless fluid particle velocity u as a function of distance for a burner-
stabilized premixed laminar lame, TIG = 0.2.
We know from dynamic systems theory that special trajectories, known as heteroclinic,
are those that connect one equilibrium point to another. These usually have special meaning.
The heteroclinic orbit shown here in the thick blue line is that of the actual flame structure. It
connects the saddle at (T, q) = (0.953, 0) to a stable point on the one-dimensional continuum
of equlibria at (T, q) = (0.132, 0). It originates on the eigenvector associated with the
negative eigenvalue at the equilibrium point. Moveover, it appears that the heteroclinic
trajectory is an attracting trajectory, as points that begin on nearby trajectories seem to be
drawn into the heteroclinic trajectory.
The orange dashed line represents our chosen ignition temperature, TIG = 0.2. Had
we selected a different ignition temperature, the heteroclinic trajectory originating from the
saddle would be the same for T > TIG . However it would have turned at a different location
and relaxed to a different cold equilibrium on the one-dimensional continuum of equilibria.
Had we attempted to construct our solution by integrating forward in x, our task would
have been more difficult. We would likely have chosen the initial temperature to be just
greater than TIG . But we would have to had guessed the initial value of q. And because of
the saddle nature of the equilibrium point, a guess on either side of the correct value would
cause the solution to diverge as x became large. It would be possible to construct a trial and
error procedure to hone the initial guess so that on one side of a critical value, the solution
diverged to , while on the other it diverged to . Our procedure, however, has the clear
advantage, as no guessing is required.
As an aside, we note there is one additional heteroclinic orbit admitted mathematically;
however, its physical relevance is far from clear. It seems there is another attracting trajec-
tory for T > 0.953. This trajectory is associated with the same eigenvector as the physical
trajectory. Along this trajectory, T increases beyond unity, at which point the mass fraction
becomes greater than unity, and is thus non-physical. Mathematically this trajectory contin-
ues until it reaches an equilibrium at (T, q) (, 0). These dynamics can be revealed using
the mapping of the Poincare sphere; see Fig. 8.8. Details can be found in some dynamic
T
0.2 0.4 0.6 0.8 1.0 1.2
-0.05
-0.10
-0.15
Figure 8.7: (T ,q) phase plane for a burner-stabilized premixed laminar flame, TIG = 0.2.
5
systems texts. In short the mapping from (T, q) (T , q ) via
T
T = p , (8.128)
1 + T 2 + q2
q
q = p , (8.129)
1 + T 2 + q2
p
is introduced. Often a third variable in the mapping is introduced Z = 1/ 1 + T 2 + q2 .
This induces T 2 + q2 + Z 2 = 1, the equation of a sphere, known as the Poincare sphere.
As it is not clear that physical relevance can be found for this, we leave out most of the
mathematical details and briefly describe the results. This mapping takes points at infinity
in physical space onto the unit circle.
Equilibria are marked in green. Trajectories are in blue. While the plot is somewhat
incomplete, we notice in Fig. 8.8 that the saddle is evident around (T , q ) (0.7, 0). We
also see new equilibria on the unit circle, which corresponds to points at infinity in the
original space. One of these new equilibria is at (T , q ) = (1, 0). It is a sink. The other
two, one in the second quadrant, the other in the fourth, are sources. The heteroclinic
trajectories which connect to the saddle from the sources in the second and fourth quadrants
5
L. Perko, 2001, Differential Equations and Dynamical Systems, Third Edition, Springer, New York.
q'
1.0
0.5
T
-1.0 -0.5 0.5 1.0
-0.5
-1.0
Figure 8.8: Projection of trajectories on Poincare sphere onto the (T , q ) phase plane for a
burner-stabilized premixed laminar flame, TIG = 0.2.
form the boundaries for the basins of attraction. To the left of this boundary, trajectories
are attracted to the continuous set of equilibria. To the right, they are attracted to the point
(1, 0).
8.2.2.5.2 TIG = 0.076. We can modulate the flame structure by altering the ignition
temperature. In particular, it is possible to iterate on the ignition temperature in such a
way that as x , T 0, 1, and u 1. Thus the cold flow region takes on
its ambient value. This has some aesthetic appeal. It is however unsatisfying in that one
would like to think that an ignition temperature, if it truly existed, would be a physical
property of the system, and not just a parameter to adjust to meet some other criterion.
That duly noted, we present flame structures using all the parameters of Table 8.1 except
we take Tig = 414 K, so that TIG = 0.076.
Predictions of T (x) are shown in in Figure 8.9. We see the temperature has T (0) =
0.076 = TIG . As x , the temperature approaches its chemical equilibrium value. For
x < 0, the temperature continues to fall until it comes to a final value of T 0. Thus in
dimensional terms, the temperature has arrived at To in the cold region.
Predictions of q(x) are shown in in Figure 8.10. There is really nothing new here. Pre-
dictions of (x) 1/(1 + QT (x)) are shown in in Figure 8.11. We see that this special
T
1.0
T = T eq =0.953
0.8
0.6
0.4
0.2
x
0 5 10 15 20 25
Figure 8.9: Dimensionless temperature as a function of position for a burner-stabilized

q
x
5 10 15 20 25
-0.02
-0.04
-0.06
-0.08
-0.10
-0.12
-0.14
Figure 8.10: Dimensionless Fourier heat flux as a function of distance for a burner-stabilized
value of TIG has allowed the dimensionless density to take on a value of unity as x .
Predictions of u(x) 1+QT (x) are shown in in Figure 8.12. In contrast to our earlier result,
for this special value of TIG , we have been able to allow u = 1 as x . We can better
understand the flame structure by considering the (T, q) phase plane as shown in Fig. 8.13.
Figure 8.13 is essentially the same as Figure 8.7 except the ignition point has been moved.
8.2.3 Detailed H2-O2 -N2 kinetics

We give brief results for a premixed laminar flame in a mixture of calorically imperfect
ideal gases which obey mass action kinetics with Arrhenius reaction rates. Multicomponent
diffusion is modelled as is thermal diffusion. We consider the kinetics model of Table 1.2.
The solution method is described in detail in a series of reports from Reaction Design, Inc. 6
6
Kee, R. J., et al., 2000, Premix: A Fortran Program for Modeling Steady, Laminar, One Dimensional
Premixed Flames, from the Chemkin Collection, Release 3.6, Reaction Design, San Diego, CA.

1.2
1.0
0.8
0.6
0.4
0.2
x
-5 0 5 10 15 20 25 30
Figure 8.11: Dimensionless density as a function of distance for a burner-stabilized pre-

mixed laminar flame, TIG = 0.076.
u
6
x
-5 0 5 10 15 20 25 30
Figure 8.12: Dimensionless fluid particle velocity u as a function of distance for a burner-
stabilized premixed laminar flame, TIG = 0.076.
7 8 9
The methods employed in the solution here are slightly different. In short, a large, but
finite domain defined. Then the ordinary differential equations describing the flame structure
are discretized. This leads to a large system of non-linear algebraic equations. These are
solved by iterative methods, seeded with an appropriate initial guess. The temperature is
pinned at To at one end of the domain, which is a slightly different boundary condition than
we employed previously. At an intermediate value of x, x = xf , the temperature is pinned
at T = Tf . Full details are in the report.
7
Kee, R. J., et al., 2000, Chemkin: A Software Package for the Analysis of Gas-Phase Chemical and
Plasma Kinetics, from the Chemkin Collection, Release 3.6, Reaction Design, San Diego, CA.
8
R. J. Kee, et al., 2000, The Chemkin Thermodynamic Data Base, part of the Chemkin Collection
Release 3.6, Reaction Design, San Diego, CA.
9
Kee, R. J., Dixon-Lewis, G., Warnatz, J., Coltrin, M. E., and Miller, J. A., 1998, A Fortran Com-
puter Code Package for the Evaluation of Gas-Phase Multicomponent Transport Properties, Sandia National
Laboratory, Report SAND86-8246, Livermore, CA.
T
0.2 0.4 0.6 0.8 1.0 1.2
-0.05
-0.10
-0.15
Figure 8.13: (T ,q) phase plane for a burner-stabilized premixed laminar flame, TIG = 0.076.
For To = 298 K, Po = 1.01325 105 P a, and a stoichiometric unreacted mixture of

2H2 + O2 + 3.76 N2 , along with a very small amount of minor species, we give plots of Y( x)
in Fig. 8.14. There is, on a log-log scale, an incubation period spanning a few orders of
magnitude of length. Just past x = 2 cm, a vigorous reaction commences, and all species
relax to a final equilibrium. Temperature as a function of distance is shown in Fig. 8.15.
Density and particle velocity are shown in Figs. 8.16 and 8.17, respectively.
0
10
5
10
10 H2
10
Yi
O2
H2O
OH
15 H
10 O
HO2
H2O2
N2
20
10 2 1 0 1
10 10 10 10
x [cm]
Figure 8.14: Mass fractions as a function of distance in a premixed laminar flame with
detailed H2 -O2-N2 kinetics.
2500
2000
[K ]
1500
T
1000
500
0
4 2 0 2 4 6 8 10 12 14
x xf [cm]
Figure 8.15: Temperature as a function of distance in a premixed laminar flame with detailed
H2 -O2 -N2 kinetics.
4
x 10
9
8
7
[ g/cm3 ]
6
5

4
3
2
1
4 2 0 2 4 6 8 10 12 14
x xf [cm]
Figure 8.16: Density as a function of distance in a premixed laminar flame with detailed
H2 -O2 -N2 kinetics.
1600
1400
1200
u [cm/s]
1000
800
600
400
200
4 2 0 2 4 6 8 10 12 14
x xf [cm]
Figure 8.17: Fluid particle velocity as a function of distance in a premixed laminar flame
with detailed H2 -O2-N2 kinetics.
Chapter 9
Simple detonations:
Reaction-advection
Plato say, my friend, that society cannot be saved until either the Professors
of Greek take to making gunpowder, or else the makers of gunpowder become
Professors of Greek.
Undershaft to Cusins in G. B. Shaws Major Barbara, Act III.
-
Let us consider aspects of the foundations of detonation theory. Detonation is defined
as a shock-induced combustion process. It is well modeled by considering the mechanisms
of advection and reaction. Diffusion plays a higher order role; its neglect is not critical to
the main physics. The definitive text is that of Fickett and Davis. 1 The present chapter is
strongly influenced by this monograph.
9.1 Reactive Euler equations

9.1.1 One-step irreversible kinetics
Let us focus here on one-dimensional planar solutions in which all diffusion processes are
neglected. We shall also here only be concerned with simple one-step irreversible kinetics in
which
A B. (9.1)
We will adopt the assumption that A and B are materials with identical properties, thus
MA = MB = M, c P A = cP B = cP , (9.2)
1
W. Fickett and W. C. Davis, 1979, Detonaton, U. California, Berkeley.
309
310 CHAPTER 9. SIMPLE DETONATIONS: REACTION-ADVECTION
but A is endowed with chemical energy which is released as it forms B. So here we have
N = 2 and J = 1. The number of elements will not be important for this analysis. Extension
to multi-step kinetics is straightforward, but involves a sufficient number of details to obscure
many of the key elements of the analysis. At this point, we will leave the state equations
relatively general, but we will soon extend them to simple ideal, calorically perfect relations.
Since we have only a single reaction, our reaction rate vector rj , is a scalar, which we will
call r. Our stoichiometric matrix ij is of dimension 2 1 and is

1
ij = . (9.3)
1
Here the first entry is associated
PJ with A, and the second with B. Thus our species production
rate vector, from i = j=1 ij rj , reduces to

A 1 r
= (r) = . (9.4)
B 1 r
Now the right side of Eq. (6.5) takes the form
Mi i . (9.5)
Let us focus on the products and see how this form expands.
MB B = MB r, (9.6)
= MB kA , (9.7)
= MA kA , (9.8)

YA
= MA k , (9.9)
MA
= kYA , (9.10)
= k(1 YB ). (9.11)
For a reaction which commences with all A, let us define the reaction progress variable as
= YB = 1 YA . And let us define r for this problem, such that
r = k(1 YB ) = k(1 ). (9.12)
Expanding further, we could say

E E
r = aT exp (1 ) = aT exp (1 ), (9.13)
RT RT
but we will delay introduction of temperature T . Here we have defined E = E/M. Note that
the units of r for this problem will be 1/s, whereas the units for r must be mole/cm3 /s, and
that

r= r. (9.14)
MB
The use of two different forms of the reaction rate, r and r is unfortunate. One is nearly
universal in the physical chemistry literature, r; the other is nearly universal in the one-step
detonation chemistry community r.
9.1. REACTIVE EULER EQUATIONS 311
9.1.2 Thermicity
Let us specialize the definition of frozen sound speed, Eq. (3.386), for our one-step chemistry:
v
u e
u P + v
c = v t e
P,
. (9.15)
P v,
This gives
e

P+ v P,
2 c2 =
e
(9.16)
P v,
Let us, as is common in the detonation literature, define the thermicity via the relation

e
P,v
c2 = e . (9.17)

P v,
We now specialize our general mathematical relation, Eq. (3.59), to get

P e
= 1, (9.18)
e v, p,v P v,e
so that one gets
e

P,v P
= . (9.19)
e
P v,
v,e
Thus we can rewrite Eq. (9.17) as

2 P
c = , (9.20)
v,e

1 P
= . (9.21)
c2 v,e
For our one step reaction, we see that as the reaction moves forward, i.e. as increases, that
is a measure of how much the pressure increases, since > 0, c2 > 0.
9.1.3 Parameters for H2-Air

Let us postulate some parameters which loosely match results of the detailed kinetics cal-
culation of Powers and Paolucci 2 for for H2 -air detonations. Rough estimates which allow
one-step kinetics models with calorically perfect ideal gas assumptions to approximate the
results of detailed kinetics models with calorically imperfect ideal gas mixtures are given in
Table 9.1
2

1.4
M 20.91 kg/kmole
R 397.58 J/kg/K
Po 1.01325 105 Pa
To 298 K
o 0.85521 kg/m3
vo 1.1693 m3 /kg
q 1.89566 106 J/kg
E 8.29352 106 J/kg
a 5 109 1/s
0
Table 9.1: Numerical values of parameters which roughly model H2 -air detonation.
9.1.4 Conservative form

Let us first consider a three-dimensional conservative form of the governing equations in the
inviscid limit:

+ (u) = 0, (9.22)
t

(u) + (uu + P I) = 0, (9.23)
t

1 1 P
e+ uu + u e + u u + = 0, (9.24)
t 2 2

() + (u) = r, (9.25)
t
e = e(P, , ), (9.26)
r = r(P, , ). (9.27)
For mass conservation, Eq. (9.22) is identical to the eariler Eq. (6.1). For linear mo-
mentum conservation, Eq. (9.23) is Eq. (6.2) with the viscous stress, = 0. For energy
conservation, Eq. (9.24) is Eq. (6.3) with viscous stress = 0, and diffusive energy trans-
port jq = 0. For species evolution, Eq. (9.27) is Eq. (6.5) with diffusive mass flux jmi = 0.
Equations (9.22-9.27) form eight equations in the eight unknowns, , u, P , e, , r. Note that
u has three unknowns, and Eq. (9.23) gives three equations. Note also that we assume the
functional forms of e and r are given.
The conservative form of the equations is the most useful and one of the more fundamental
forms. It is the form which arises from the even more fundamental integral form, which
admits discontinuities. We shall take advantage of this in later forming shock jump equations.
The disadvantage of the conservative form is that it is unwieldy and masks simpler causal
relations.
9.1.5 Non-conservative form

We can reveal some of the physics described by Eqs. (9.22-9.24) by writing them in what is
known as a non-conservative form.
9.1.5.1 Mass
We can use the product rule to expand Eq. (9.22) as

+ u + u = 0. (9.28)
|t {z }
=d/dt
We recall the well known material derivative, also known as the derivative following a material
particle, the total derivative, or the substantial derivative:
d
+ u .
dt t
Using the material derivative, we can rewrite the mass equation as
d
+ u = 0. (9.29)
dt
Often, Newtons notation for derivatives, the dot, is used to denote the material derivative.
We can also recall the definition of the divergence operator, div, to be used in place of ,
so that the mass equation can be written compactly as
+ div u = 0. (9.30)
The density of a material particle changes in response to the divergence of the velocity field,
which can be correlated with the rate of volume expansion of the material region.
9.1.5.2 Linear momenta

We can use the product rule to expand the linear mometa equations, Eq. (9.23), as
u
+u + u( (u)) + u u + P = 0, (9.31)
t t

u
+u + (u) +u u + P = 0, (9.32)
t t
| {z }
=0

u
+ +u u + P = 0, (9.33)
t
du
+ P = 0. (9.34)
dt
Or in terms of our simplified notation with v = 1/, we have
u + v grad P = 0. (9.35)
The fluid particle accelerates in response to a pressure gradient.
9.1.5.3 Energy
We can apply the product rule to the energy equation, Eq. (9.24) to get

1 1 1
e+ uu + e+ uu + e + u u (u)
t 2 2 t 2

1
+u e + u u + (P u) = 0, (9.36)
2

1 1
e+ uu + e+ uu + (u)
t 2 2 t
| {z }
=0

1
+u e + u u + (P u) = 0, (9.37)
2

1 1
e + u u + u e + u u + (P u) = 0. (9.38)
t 2 2
This is simpler, but there is more that can be done by taking advantage of the linear momenta
equations. Let us continue working with the product rule once more along with some simple
rearrangements

e 1 1
+ u e + u u + u u u + (P u) = 0, (9.39)
t t 2 2
e u
+ u e + u + u (u )u + u P + P u = 0, (9.40)
t t
e u
+ u e +u + (u )u + P +P u = 0, (9.41)
| t {z } |
t
{z }
= de/dt =0
de
+ P u = 0. (9.42)
dt
Now from Eq. (9.29), we have u = (1/)d/dt. Eliminating the divergence of the
velocity field from Eq. (9.42), we get
de P d
= 0, (9.43)
dt dt
de P d
= 0. (9.44)
dt 2 dt
Recalling that = 1/v, so d/dt = (1/v 2 )dv/dt = 2 dv/dt, our energy equation becomes
de dv
+p = 0. (9.45)
dt dt
Or in terms of our dot notation, we get
e + pv = 0. (9.46)
The internal energy of a fluid particle changes in response to the work done by the
pressure force.
9.1.5.4 Reaction
Using the product rule on Eq. (9.27) we get

+ + u + (u) = r, (9.47)
t t

+ u + + (u) = r, (9.48)
t t
| {z } | {z }
=d/dt =0
d
= r, (9.49)
dt
= r. (9.50)
The mass fraction of a fluid particles product species changes according to the forward
reaction rate.
9.1.5.5 Summary
In summary, our non-conservative equations are
+ div u =
0, (9.51)
u + v grad P =
0, (9.52)
e + P v =
0, (9.53)
=
r, (9.54)
e =
e(P, , ), (9.55)
r =
r(P, , ), (9.56)
1
= . (9.57)
v
9.1.6 One-dimensional form

Here we consider the equations to be restricted to one-dimensional planar geometries.

In the one-dimensional planar limit, Eqs. (9.22-9.27) reduce to

+ (u) = 0, (9.58)
t x

(u) + u2 + P = 0, (9.59)
t x

1 2 1 2 P
e+ u + u e + u + = 0, (9.60)
t 2 x 2

() + (u) = r, (9.61)
t x
e = e(P, , ), (9.62)
r = r(P, , ). (9.63)
9.1.6.2 Non-conservative form
The non-conservative Eqs. (9.51-9.57) reduce to the following in the one-dimensional planar
limit:
u
+ = 0, (9.64)
x
P
u + v = 0, (9.65)
x
e + P v =
0, (9.66)
=
r, (9.67)
e =
e(P, , ), (9.68)
r =
r(P, , ), (9.69)
1
= . (9.70)
v
Equations (9.64-9.70) can be expanded using the definition of the material derivative:

u
+u + = 0, (9.71)
t x x

u u P
+u +v = 0, (9.72)
t x x

e e v v
+u +P +u = 0, (9.73)
t x t x

+u = r, (9.74)
t x
e = e(P, , ), (9.75)
r = r(P, , ), (9.76)
1
= . (9.77)
v
9.1.6.3 Reduction of energy equation

Let us use standard results from calculus of many variables to expand the caloric equation
of state, Eq. (9.75):

e e e
de = dP + dv + d. (9.78)
P v, v P, P,v
Taking the time derivative gives

e + e
v + e
.
e = P (9.79)
P v, v P, P,v
Note this is simply a time derivative of the caloric state equation; it says nothing about energy
conservation. Next let us use Eq. (9.79) to eliminate e in the first law of thermodynamics,
Eq. (9.66) to get

e e e
P+ v + +P v = 0, (9.80)
P v, v P, P,v
| {z }
=e
!
e

e
P + v
P + e
P, v + P,v = 0,
e
(9.81)
P v, P v,
| {z } | {z }
=2 c2 =c2
P + c v c2 = 0,
2 2
(9.82)

P = 2 c2 v + c2 , (9.83)
P = c2 (r v) . (9.84)
Now since v = (1/2 ),

we get
P = c2 + c2 r. (9.85)
Note that if either r = 0 or = 0, the pressure changes will be restricted to those from
classical isentropic thermo-acoustics: P = c2 .
If r > 0, reaction induces positive pressure
changes. If r < 0, reaction induces negative pressure changes. Moreover note that Eq. (9.85)
is not restricted to calorically perfect ideal gases. It is valid for general state equations.
9.1.7 Characteristic form

Let us obtain a standard form known as characteristic form for the one-dimensional un-
steady equations. For this form let us use Eq. (9.85) and generalized forms for sound speed
c and thermicity to recast our governing equations, Eq. (9.71-9.77) as
u
+u + = 0, (9.86)
t x x
u u 1 P
+u + = 0, (9.87)
t x x

P P 2
+u c +u = c2 r, (9.88)
t x t x

+u = r, (9.89)
t x
c2 = c2 (P, ), (9.90)
r = r(P, , ), (9.91)
= (P, , ). (9.92)
Let us write the differential equations in matrix form:

1 0 0 0 t
u 0 0 x
0
0 u 0 1
2 1 0 0 Pt +
u
0 u
x
0
= 2
c 0 1 0 c2 u 0 u (9.93)
t 0 P
x
c r

0 0 0 1 t 0 0 0 u x
r
These take the general form
wj wj
Aij + Bij = Ci . (9.94)
t x
Let us attempt to cast this the left hand side of this system in the form wj /t + wj /x.
Here is a scalar which has the units of a velocity. To do so, we shall seek vectors i such
that

wj wj wj wj
i Aij + i Bij = i Ci = mj + (9.95)
t x t x
For i to have the desired properties, we will insist that
i Aij = mj , (9.96)
i Bij = mj . (9.97)
Using Eq. (9.96) to eliminate mj in Eq. (9.97), we get
i Bij = i Aij , (9.98)

0 = i (Aij Bij ). (9.99)
Equation (9.99) has the trivial solution i = 0. For a non-trivial solution, standard linear
algebra tells us that we must enforce the condition that the determinant of the coefficient
matrix be zero:
|Aij Bij | = 0. (9.100)

Specializing Eq. (9.100) for Eq. (9.93), we find

u 0 0

0 u 1
0
= 0. (9.101)
c2 ( u) 0 u 0

0 0 0 u
Let us employ standard co-factor expansion operations to reduce the determinant:

u 0

( u) 0 u 1 = 0, (9.102)
c2 ( u) 0 u
2
2
c ( u)
( u) ( u) ( u) + = 0, (9.103)

( u)2 ( u)2 c2 = 0 (9.104)
There are four roots to this equation; two are repeated:
= u, (9.105)
= u, (9.106)
= u + c, (9.107)
= u c. (9.108)
Let us find the eigenvector i associated with the eigenvalues = uc. So Eq. (9.99) reduces
to

(u c) u 0 0
0 (u c) u 1 0
( 1 2 3 4 ) c2 ((u c) u)
= (0 0 0 0).

0 (u c) u 0
0 0 0 (u c) u
(9.109)
Simplifying,

c 0 0
0 c 1 0
( 1 2 3 4 )
c3
= (0 0 0 0). (9.110)
0 c 0
0 0 0 c
We thus find four equatons, with linear dependencies:
c1 c3 3 = 0, (9.111)
1 c2 = 0, (9.112)
1
2 c3 = 0, (9.113)

c4 = 0. (9.114)
Now c 6= 0, so Eq. (9.114) insists that

4 = 0. (9.115)
We expect a linear dependency, which implies that we are free to set at least one of the
remaining i to an arbitrary value. Let us see if we can get a solution with 3 = 1. With
that Eqs. (9.111-9.113) reduce to
c1 c3 = 0, (9.116)
1 c2 = 0, (9.117)
1
2 c = 0. (9.118)

Solving Eq. (9.116) gives
1 = c2 . (9.119)
Then Eq. (9.117) becomes c2 c2 = 0. Solving gives
2 = c. (9.120)
This is redundant with solving Eq. (9.118), which also gives 2 = c. So we have for
= u c that
i = ( c2 c 1 0 ) . (9.121)
So Eq. (9.93) becomes after multiplication by i from Eq. (9.121):

1 0 0 0 t
u 0 0 x
0 1 0 0 u 0 u 1 0 u
( c2 c 1 0 ) c2
P t + ( c2
c 1 0 ) 2 x
0 1 0 t c u 0 u 0 Px

0 0 0 1 t 0 0 0 u
x
0
0
= ( c2 c 1 0 )
c2 r
r
(9.122)
Carrying out the vector-matrix multiplication operations, we get

t x
u u
( 0 c 1 0 )

t
P
+ ( 0 c(u c) u c 0 )

x
P

t x

t x

0
0
= ( c2 c 1 0 )
c2 r .
r
(9.123)
Simplifying,

u u P P
c + (u c) + + (u c) = c2 r. (9.124)
t x t x
Now let us confine our attention to lines in x t space on which
dx
= u c. (9.125)
dt
On such lines, Eq. (9.124) can be written as

u dx u P dx P
c + + + = c2 r, (9.126)
t dt x t dt x
Consider now a variable, say u, which is really u(x, t). From calculus of many variables, we
have
u u
du = dt + dx, (9.127)
t x
du u dx u
= + . (9.128)
dt t dt x
If we insist dx/dt = u c, let us call the derivative du/dt , so that Eq. (9.126) becomes
du dP
c + = c2 r, (9.129)
dt dt
In the inert limit, after additional analysis, Eq. (9.129) reduces to the form d/dt = 0, which
shows that is maintained as a constant on lines where dx/dt = u c. Thus one can say
that a signal is propagated in x t space at speed u c. So we see from the characteristic
analysis how the thermodynamic property c has the added significance of influencing the
speed at which signals propagate.
Let us now find i for = u. For this root, we find

0 0 0
0 0 1 0
( 1 2 3 4 ) 0 0
= (0 0 0 0), (9.130)
0 0
0 0 0 0
It can be seen by inspection that two independent solutions i satisfy Eq. (9.130):
i = ( 0 0 1 0 ) , (9.131)
i = ( 0 0 0 1 ) . (9.132)
These two eigenvectors induce the characteristic form which was already obvious from the
initial form of the energy and species equations:

P P
+u c2 +u = c2 r, (9.133)
t x t x

+u = r. (9.134)
t x
On lines where dx/dt = u, that is to say on material particle pathlines, these reduce to
P c2 = c2 r and = r.
Because we all of the eigenvalues are real, and because we were able to find a set of
four linearly independent right eigenvectors i so as to transform our four partial differential
equations into characteristic form, we can say that our system is strictly hyperbolic. Thus it
is well posed for initial value problems given that initial data is provided on a non-
characteristic curve,
admits discontinuous solutions described by a set of -Hugoniot jump conditions which

arise from a more primitive form of the governing equations.
In summary we can write our equations in characteristic form as
dP du dx
+ c = c2 r, on = u + c, (9.135)
dt+ dt+ dt
dP du dx
c = c2 r, on = u c, (9.136)
dt dt dt
dP d dx
c2 = c2 r, on =u (9.137)
dt dt dt
d dx
= r on = u. (9.138)
dt dt
9.1.8 Rankine-Hugoniot jump conditions

As described by LeVeque, 1992, the proper way to arrive at the jump equations describing
discontinuities is to use a more primitive form of the conservation laws, expressed in terms
of integrals of conservative form quantities balanced by fluxes and source terms of those
quantities. If q is a set of conservative form variables, and f(q) is the flux of q (e.g. for mass
conservation, is a conserved variable and u is the flux), and s(q) is the internal source
term, then the primitive form of the conservation form law can be written as
Z x2 Z x2
d
q(x, t)dx = f(q(x1 , t)) f(q(x2 , t)) + s(q(x, t))dx. (9.139)
dt x1 x1
Here we have considered flow into and out of a one-dimensional box for x [x1 , x2 ]. For our
reactive Euler equations we have

u 0
2
u u + P
q= , f(q) =
, s(q) = 0 . (9.140)
e + u 1 2 u e + u + 1 2 P 0
2 2
u r
If we assume there is a discontinuity in the region x [x1 , x2 ] propagating at speed U, we

can break up the integral into the form
Z Z
d x1 +U t d x2
q(x, t)dx + q(x, t)dx
dt x1 dt x1 +U t+
Z x2
= f(q(x1 , t)) f(q(x2 , t)) + s(q(x, t))dx. (9.141)
x1
+
Here x1 + Ut lies just before the discontinuity and x1 + Ut lies just past the discontinuity.
Using Leibnitzs rule, we get
Z x1 +U t Z x2
q + q
q(x1 + Ut , t)U + 0 + dx + 0 q(x1 + Ut , t)U + dx (9.142)
x1 t x1 +U t+ t
Z x2
= f(q(x1 , t)) f(q(x2 , t)) + s(q(x, t))dx.
x1
Now if we assume that x2 x1 0 and that on either side of the discontinuity the volume of
integration is sufficiently small so that the time and space variation of q is negligibly small,
we get
q(x1 )U q(x2 )U = f(q(x1 )) f(q(x2 )), (9.143)
U (q(x1 ) q(x2 )) = f(q(x1 )) f(q(x2 )). (9.144)
Note that the contribution of the source term s is negligible as x2 x1 0. Defining next
the notation for a jump as
Jq(x)K q(x2 ) q(x1 ), (9.145)
the jump conditions are rewritten as
U Jq(x)K = Jf(q(x))K . (9.146)
If U = 0, as is the case when we transform to the frame where the wave is at rest, we
simply recover
0 = f(q(x1 )) f(q(x2 )), (9.147)
f(q(x1 )) = f(q(x2 )), (9.148)
Jf(q(x))K = 0. (9.149)
That is the fluxes on either side of the discontinuity are equal. We also get a more general
result for U 6= 0, which is the well-known
f(q(x2 )) f(q(x1 )) Jf(q(x))K
U= = . (9.150)
q(x2 ) q(x1 ) Jq(x)K
The general Rankine-Hugoniot equation then for the one-dimensional reactive Euler equa-
tions across a non-stationary jump is given by

2 1 2 u2 1 u1
2 u2 1 u1 2 u22 + P2 1 u21
P1
U 1 2

1 2 =
1 2 P2 1 2

P1 .
2 e2 + 2 u2 1 e1 + 2 u1 2 u2 e2 + 2 u2 + 2 1 u1 e1 + 2 u1 + 1
2 2 1 1 2 u2 2 2 u1 1
(9.151)
Note that if there is no discontinuity, Eq. (9.139) reduces to our partial differential equa-
tions which are the reactive Euler equations. We can rewrite Eq. (9.139) as
Z x2 Z x2
d
q(x, t)dx + (f(q(x2 , t)) f(q(x1 , t))) = s(q(x, t))dx, (9.152)
dt x1 x1
Now if we assume continuity of all fluxes and variables, we can use Taylor series expansion
and Leibnizs rule to say
Z x2 Z x2
f
q(x, t)dx + f(q(x1 , t)) + (x2 x1 ) + . . . f(q(x1 , t)) = s(q(x, t))dx,
x1 t x x1
let x2 x1
Z x2 Z x2
f
q(x, t)dx + (x2 x1 ) = s(q(x, t))dx,
x1 t x x1
Z x2 Z x2 Z x2
f
q(x, t)dx + dx = s(q(x, t))dx,
x1 t x1 x x1
(9.153)
Combining all terms under a single integral, we get
Z x2
q f
+ s dx = 0. (9.154)
x1 t x
Now this integral must be zero for an arbitrary x1 and x2 , so the integrand itself must be
zero, and we get our partial differential equation:
q f
+ s = 0, (9.155)
t x

q(x, t) + f(q(x, t)) = s(q(x, t)), (9.156)
t x
which applies away from jumps.
9.1.9 Galilean transformation

We know that Newtonian mechanics have been constructed so as to be invariant under a so-
called Galilean transformation which takes one from a fixed laboratory frame to a constant
velocity frame with respect to the fixed frame. The Galilean transformation is such that
our original coordinate system in the laboratory frame, (x, t), transforms to a steady wave
x, t), via
frame, (
x = x Dt, (9.157)
t = t. (9.158)
We thus get differentials
x x
d
x = dx + dt = dx Ddt, (9.159)
x t
t t
dt = dx + dt = dt. (9.160)
x t
x by dt gives us then
Scaling d
d
x dx
= D. (9.161)
dt dt
Taking as usual the particle velocity in the fixed frame to be u = dx/dt and defining the
x/dt, we see that
particle velocity in the laboratory frame to be u = d
u = u D. (9.162)
Now a dependent variable has a representation in the original space of (x, t), and in the
transformed space as (x, t). And they must both map to the same value of and the same
point. And there is a differential of in both spaces which must be equal:

d = dx + dt = d
x+ dt, (9.163)
x t t x x t t x

= (dx Ddt) + dt, (9.164)
x t t x

= dx + D dt. (9.165)
x t t x x t
Now consider Eq. (9.165) for constant x, thus dx = 0, and divide by dt:

= D . (9.166)
t x t x x t
More generally the partial with respect to t becomes

= D . (9.167)
t x t x x t
Now consider Eq. (9.165) for constant t, thus dt = 0, and divide by dx:

= . (9.168)
x t x t
More generally,

= . (9.169)
x t x t
Let us consider how a material derivative transforms then

+u = D + (
u + D) , (9.170)
t x x t tx x t x t

= + u . (9.171)
tx x t
Thus we can quickly write our non-conservative form, Eqs. (9.64-9.67) in the transformed
frame
u
+ u + = 0, (9.172)
t x x
u u 1 P
+ u + = 0, (9.173)
t x x

e e v v
+ u P + u = 0, (9.174)
t x t x

+ u = r. (9.175)
t x
Moreover, we can write these equations in conservative form. Leaving out the details, which
amounts to reversing our earlier steps which led to the non-conservative form, we get

+ (u) = 0, (9.176)
t x

(
u) + u2 + P
= 0, (9.177)
t x
1 2 1 2 P
e + u +
u e + u + = 0, (9.178)
t 2 x 2

() + (
u) = r. (9.179)
t x
9.2 One-dimensional, steady solutions

Let us consider a steadily propagating disturbance in a one dimensional flow field. In the
laboratory frame, the disturbance is observed to propagate with constant velocity D. Let us
analyze such a disturbance.
9.2. ONE-DIMENSIONAL, STEADY SOLUTIONS 327
9.2.1 Steady shock jumps

In the steady frame, our jump conditions, Eq. (9.151) have U = 0 and reduce to

0 2 u2 1 u1
0 2 u22 + P2 1 u21 P1
=
0 2 u2 e2 + 1 u22 + P2 1 u1 e1 + 1 u21 + P1 . (9.180)
2 2 2 1
0 2 u2 2 2 u1 1
The mass jump equation can be used to quickly simplify the energy and species jump equa-
tions to get a revised set
2 u2 = 1 u1
2 u22 + P2 = 1 u21 + P1
1 P2 1 P1
e2 + u22 + = e1 + u21 +
2 2 2 1
2 = 1 (9.181)
9.2.2 Ordinary differential equations of motion

Let us now assert that in the steady laboratory frame that no variable has dependence on
t, so / t = 0, and / x = d/d
x. With this assumption our partial differential equations of
motion, Eqs. (9.176-9.179), become ordinary differential equations:
d
(u) = 0, (9.182)
d
x
d
u2 + P
= 0, (9.183)
d x
d 1 2 P

u e + u + = 0, (9.184)
d
x 2
d
(
u) = r, (9.185)
d
x
e = e(P, , ), (9.186)
r = r(P, , ). (9.187)
We shall take the following conditions for the undisturbed fluid just before the shock at
x = 0 :
(
x = 0 ) = o , (9.188)
u(
x = 0 ) = D, (9.189)
P (
x = 0 ) = Po , (9.190)
e(
x = 0 ) = eo , (9.191)
(x = 0 ) = 0. (9.192)
Note that in the laboratory frame, this corresponds to a material at rest since u = u + D =
(D) + D = 0.
9.2.2.2 Unreduced non-conservative form

We can gain insights into how the differential equations, Eqs. (9.182-9.185) behave by writing
them in a non-conservative form. Let us in fact write them, taking advantage of the reduc-
tions we used to acquire the characteristic form, Eqs. (9.86-9.89) to write those equations
after transformation to the steady Galilean transformed frame as
du d
+ u = 0, (9.193)
dx d
x
du 1 dP
u + = 0, (9.194)
dx d x
dP d
u c2 u = c2 r, (9.195)
d
x d
x
d
u = r. (9.196)
d
x
Let us attempt next to get explicit representations for the first derivatives of each equation.
In matrix form, we can say our system is
d
u 0 0 d
x
0
0
u 1 0 du 0
dPx
d
c2 u 0 u 0 = c2 r . (9.197)
d
x
d
0 0 0 u d
x
r
We can use Cramers rule to invert the coefficient matrix to solve for the evolution of each
state variable. This first requires the determinant of the coefficient matrix .

u 0 0

0 u
1
0
= 2
,
(9.198)
c u 0 u 0
0 0 0 u

u 0

= u 0 u ,
1
(9.199)
c2 u 0 u

2 c2 u
= u u( u ) , (9.200)

= u2 (u2 c2 ). (9.201)
As appears in the denominator after application of Cramers rule, we see immediately that
when the waveframe velocity u becomes locally sonic (
u = c) that our system of differential
equations is potentially singular.
Now by Cramers rule, d/d

x is found via

0 0 0

0 u 1 0

c2 r 0 u 0

d r 0 0 u
= , (9.202)
d
x
0 0

u 0 u 1
c2 r 0 u
= , (9.203)
u2 (
u 2 c2 )
u()(c2 r)
= , (9.204)
u2 (
u 2 c2 )
c2 r
= . (9.205)
u(u 2 c2 )
Similarly, solving for d

u/d
x we get

u 0 0 0

0 0 1
0

c2 u c2 r u 0

d
u 0 r 0 u
= , (9.206)
d
x
u 0 0

u 0 0 1

c2 u c2 r u
= , (9.207)
u2 (
u 2 c2 )
u2 (c2 r)
= , (9.208)
u2 (u 2 c2 )
c2 r
= 2 , (9.209)
u c2
(9.210)
For dP/d
x, we get

u 0 0

0 u 0 0

c2 u 0 c2 r 0

dP 0 0 r u
= , (9.211)
d
x

u 0

u 0 u 0

c2 u 0 c2 r
= , (9.212)
u2 (u 2 c2 )
uc2 ru2
= 2 2 , (9.213)
u ( u c2 )
uc2 r
= 2 , (9.214)
u c2
(9.215)
Lastly, we have by inspection

d r
= . (9.216)
d
x u
We can employ the local Mach number in the steady wave frame,
2
2 u ,
M (9.217)
c2
to write our system of ordinary differential equations as
d r
= , (9.218)
d
x u(1 M 2)
d
u r
= , (9.219)
d
x 1M 2
dP
ur
= , (9.220)
d
x 1M 2
d r
= . (9.221)
d
x u
We note the when the flow is locally sonic, M = 1, that our equations are singular. Recall-
ing an analogy from compressible flow with area change, which is really an application of
lHopitals rule, in order for the flow to be locally sonic, we must insist that simultaneously
the numerator must be zero; thus, we might demand that at a sonic point, r = 0. So an
end state with r = 0 may in fact be a sonic point. For multistep reactions, each with their
own thermicity and reaction progress, we require the generalization r = 0 at a local
sonic point. Here is the vector of thermicities and r is the vector of reaction rates. This
condition will be important later when we consider so-called eigenvalue detonations.
9.2.2.3 Reduced non-conservative form

Let us use the mass equation, Eq. (9.182) to simplify the reaction equation, Eq. (9.185), then
integrate our differential equations for mass momentum and energy conservation, apply the
initial conditions, and thus reduce our system of four differential and two algebraic equations
to one differential and five algebraic equations:
u = o D, (9.222)
P + u2 = Po + o D 2 , (9.223)

1 2 P 1 2 Po
u e + u +
= o D eo + D + , (9.224)
2 2 o
d 1
= r, (9.225)
dx u
e = e(P, , ), (9.226)
r = r(P, , ). (9.227)
With some effort, we can unravel these equations to form one ordinary differential equation in
one unknown. But let us delay that analysis until after we have examined the consequences
of the algebraic constraints
9.2.3 Rankine-Hugoniot analysis

Let us first analyze our steady mass, momentum, and energy equations, Eqs. (9.222-9.224).
Our analysis here will be valid both within, (0 < < 1) and at the end of the reaction zone
( = 1).
9.2.3.1 Rayleigh line

Let us get what is known as the Rayleigh line by considering only the mass and linear
momentum equations, Eqs. (9.222-9.223). Let us first rewrite Eq. (9.223) as
2 u2 2o D 2
P+ = Po + . (9.228)
o
Then, the mass equation, Eq. (9.22) allows us to rewrite the momentum equation, Eq. (9.228)
as
2o D 2 2 D 2
P+ = Po + o . (9.229)
o
Rearranging to solve for P , we find

1 1
P = Po 2o D 2 . (9.230)
o
In terms of v = 1/, and a slight rearrangement, Eq. (9.230) can be re-stated as
D2
P = Po (v vo ) , (9.231)
vo2

P D2 v
= 1 1 . (9.232)
Po Po vo vo
P HPaL
8. 106
6. 106
4. 106
2. 106
O
1 Hm3 kgL
0.5 1.0
Figure 9.1: Plot of Rayleigh line for parameters of Table 9.1 and D = 2800 m/s. Slope is
2o D 2 < 0.
Note:
This is a line in (P, 1/) space, the Rayleigh line.
The slope of the Rayleigh line is strictly negative.
The magnitude of the slope of the Rayleigh line is proportional the the square of the
wave speed; high wave speeds induce steep slopes.
The Rayleigh line passes through the ambient state (Po , 1/o ).
Small volume leads to high pressure.
These conclusions are a consequence of mass and momentum conservation alone. No

consideration of energy has been made.
The Rayleigh line equation is valid at all stages of the reaction: the inert state, a
shocked stated, an intermediate reacted state, and a completely reacted state. It is
always the same line.
A plot of a Rayleigh line for D = 2800 m/s and the parameters of Table 9.1 is shown in
Fig. 9.1. Here the point labeled O is the ambient state (1/o , Po).
9.2.3.2 Hugoniot curve

Let us next focus on the energy equation, Eq. (9.224). We shall use the mass and momentum
equations (9.222,9.223) to cast the energy equation in a form which is independent of both
velocities and wave speeds. From Eq. (9.222), we can easily see that Eq. (9.224) first reduces
to
1 P 1 Po
e + u2 + = eo + D 2 + , (9.233)
2 2 o
(9.234)
Now the mass equation (9.222) also tells us that

o
u = D, (9.235)

so Eq. (9.233) can be recast as

2
1 o P 1 Po
e+ D2 + = eo + D 2 + , (9.236)
2 2 o
2 !
1 o P Po
e eo + D 2 1 + = 0, (9.237)
2 o

1 2 (o )(o + ) P Po
e eo + D 2
+ = 0. (9.238)
2 o
Now the Rayleigh line, Eq. (9.230) can be used to solve for D 2 ,
1
2Po P 1 1 Po P o
D = = . (9.239)
2o o 2o o
Now use Eq. (9.239) to eliminate D 2 in Eq. (9.238):

1 Po P o (o )(o + ) P Po
e eo + + = 0, (9.240)
2 2o 2 o
| {z o }
=D 2

1 Po P
o + P Po
e eo + + = 0, (9.241)
2 o o

1 1 1 P Po
e eo + (Po P ) + + = 0, (9.242)
2 o o
1 Po 1 Po 1 P 1P P Po
e eo + + + = 0, (9.243)
2 2 o 2 2 o o

1 Po Po P P
e eo + + = 0, (9.244)
2 o o

1 1 1
e eo + (P + Po ) = 0. (9.245)
2 o
Rearranging, we get

P + Po 1 1
e eo = . (9.246)
| {z }
| {z2 } o
| {z }
change in energy -average pressure
change in volume
This is the Hugnoniot equation for a general material. It applies for solid, liquid, or gas.
Note that
This form of the Hugoniot does not depend on the state equation.
The Hugoniot has no dependency on particle velocity or wave speed.
The Hugoniot is valid for all e(); that is it is valid for inert, partially reacted, and
totally reacted material. Different degrees of reaction will induce shifts in the curve.
Now let us specify an equation of state. For convenience and to more easily illustrate
the features of the Rankine-Hugoniot analysis, let us focus on the simplest physically-based
state equation, the calorically perfect ideal gas for our simple irreversible kinetics model.
With that we adopt the perfect caloric state equation studied earlier, Eq. (7.6):
e = cv T q. (9.247)
We also adopt an ideal gas assumption
P = RT. (9.248)
Solving for T , we get T = P//R. Thus cv T = (cv /R)(P/). Recalling that R = cP cv , we

then get cv T = (cv /(cP cv ))(P/). And recalling that for the calorically perfect ideal gas
cP
= , (9.249)
cv
we get
1 P
cv T = . (9.250)
1
Thus our caloric equation of state, Eq. (9.247), for our simple model becomes
1 P
e(P, , ) = q. (9.251)
1
In terms of v, Eq. (9.251) is

1
e(P, v, ) = P v q. (9.252)
1
As an aside, we note that for this equation of state, the sound speed can be deduced from
Eq. (9.16) as

e 1
2 2
P + v P, P + 1 P P
c =
e
= 1 = . (9.253)
P v,
v
1
v
P 1 P
c2 = 2
= P v = . (9.254)
v
For the thermicity, we note that
R R e + q
P = RT = T = , (9.255)
v v c v
P Rq
= = ( 1)q, (9.256)
v,e cv v

1 P 1 q 1 q
= 2 = 2 ( 1)q = ( 1) 2 = . (9.257)
c v,e c c P
Now at the initial state, we have = 0, and so Eq. (9.251) reduces to

1 Po
eo = . (9.258)
1 o
We now use our caloric state relations, Eqs. (9.251,9.258) to specialize our Hugoniot
relation, Eq. (9.246), to

1 P Po P + Po 1 1
q = . (9.259)
1 o 2 o
Employing v = 1/, we get a more compact form:
1 1
(P v Po vo ) q = (P + Po ) (v vo ) . (9.260)
1 2
Let us next operate on Eq. (9.259) so as to see more clearly how such the Hugoniot is
represented in the (P, v) = (P, 1/) plane.
1 1
(P v Po vo ) + (P v + Po v P vo Po vo ) = q, (9.261)
1 2

1 1 1 1 1 1
P v + v vo Po vo v + vo = q, (9.262)
1 2 2 1 2 2

P +1 Po + 1
v vo vo v = q, (9.263)
2 1 2 1

+1
2q + Po 1 vo v
P = . (9.264)
+1
1
v v o
Note that as
1
v vo , P . (9.265)
+1
For = 7/5, this gives v vo /6 induces and infinite pressure. In fact an ideal gas cannot
be compressed beyond this limit, known as the strong shock limit. Note also that as
1
v , Po < 0. (9.266)
+1
So very large volumes induce non-physical pressures.
Let us continue to operate to get a more compact form for the calorically perfect ideal
gas Hugoniot curve. Let us add a common term to both sides of Eq. (9.264):

+1
1 2q + Po 1 vo v 1
P+ Po = + Po , (9.267)
+1 +1
v v +1
1 o

1 +1 +1
P+ Po v vo = 2q + Po vo v
+1 1 1

1 +1
+ Po v vo , (9.268)
+1 1
+1
= 2q + Po vo Po v
1
1
+Po v Po vo , (9.269)
+1
+1 1
= 2q + Po vo Po vo , (9.270)
1 +1
4
= 2q + 2 Po vo , (9.271)
1

P 1 +1 v 2q 4
+ 1 = + 2 , (9.272)
Po + 1 1 vo Po vo 1

P 1 v 1 1 2q 4
+ = + . (9.273)
Po + 1 vo + 1 + 1 Po vo ( + 1)2
Equation (9.273) represents a hyperbola in the (P, v) = (P, 1/) plane. As proceeds from
0 to 1 the Hugoniot moves. A plot of a series of Hugoniot curves for values of = 0, 1/2, 1
along with two Rayleigh lines for D = 2800 m/s and D = 1991.1 m/s are shown in Fig. 9.2.
The D = 1991.1 m/s Rayleigh line happens to be exactly tangent to the = 1 Hugoniot
curve. We will see this has special significance.
2 as
Let us define, for convenience, the parameter
1
2
. (9.274)
+1
P HPaL
8. 106
6. 106
N S
4. 106
N
2. 106 C
W
1 Hm3 kgL
0.5 1.0 O
Figure 9.2: Plot of = 0, 1/2, 1 Hugoniot curves and two Rayleigh lines for D = 2800 m/s,
D = 1991.1 m/s and parameters of Table 9.1.
With this definition, we note that

2
4 1 2 + 2 + 1 2 + 2 1 4
1 = 1 = = . (9.275)
+1 ( + 1)2 ( + 1)2
With this definition, Eq. (9.273), the Hugoniot, becomes

P 2 v 2 q
+
= 2 2 4 .
+1 (9.276)
Po vo Po vo
Now let us seek to intersect the Rayliegh line with the Hugoniot curve and find the points
of intersection. To do so, let us use the Rayleigh line, Eq. (9.232), to eliminate pressure in
the Hugoniot, Eq. (9.276):

D2 v 2 v 2 q
1 1 + = 22 +1 4. (9.277)
Po vo vo vo Po vo
We now structure Eq. (9.277) so that it can be solved for v:

2 D2 v D2 v 2 q
1+ + = 2 2 4 .
+1 (9.278)
Po vo vo Po vo vo Po vo
Now, let us regroup to form
2
v D2 v 2 D2 2 2 D2
+ (1 +
) 1+
1+ +
vo Po vo vo Po vo Po vo
q
2
2 4 = 0,
1+
Po vo
(9.279)
2
v D2 v 2 D2 2 D2 q
+ (1 +
) 1+
1+ 2
2 1 = 0.
vo Po vo vo Po vo Po vo Po vo
(9.280)
This quadratic equation has two roots:

D2
1 + Po vo (1 + 2 )
v
= 2
vo 2 PDo vo
r 2
2 )2 4 PDo vo 1 + 1 + 2 Pq
2 2 D2
1 + PDo vo (1 + Po vo
2 2
o vo
2 (9.281)
2 PDo vo
For a given wave speed D, initial undisturbed conditions, Po , vo , and material properties,
,
q, Eq. (9.281) gives the specific volume as a function of reaction progress . Depending on
these parameters, we can mathematically expect two distinct real solutions, two repeated
solutions, or two complex solutions.
9.2.4 Shock solutions

We know from Eq. (9.181) that does not change through a shock. So if = 0 before
the shock, it has the same value after. So we can get the shock state by enforcing the
Rankine-Hugoniot jump conditions with = 0:

1+ D2
2 )
(1 +
v Po vo
= 2
vo 2 PDo vo
r 2
2 )2 4 PDo vo 1 + 1 +
2 2 D2
1 + PDo vo (1 + Po vo
2

2 (9.282)
2 PDo vo
With considerable effort, or alternatively, by direct calculation via computational algebra,

the two roots of Eq. (9.282) can be shown to reduce to:
v
= 1, (9.283)
vo
v Po vo
2 + 2 (1 +
= 2 ). (9.284)
vo D
The first is the ambient solution; the second is the shocked solution. The shock solution can
also be expressed as
v 1 2 Po vo
= + . (9.285)
vo + 1 + 1 D2
In the limit as Po vo /D2 0, the so-called strong shock limit, we find
v 1
. (9.286)
vo +1
The reciprocal gives the density ratio in the strong shock limit:
+1
. (9.287)
o 1
With the solutions for v/vo , we can employ the Rayleigh line, Eq. (9.232) to get the
pressure. Again, we find two solutions:
P
= 1, (9.288)
Po
P D2
= 2 )
(1 2 . (9.289)
Po Po vo
The first is the inert solution, and the second is the shock solution. The shock solution is
rewritten as
P 2 D2 1
= . (9.290)
Po + 1 Po vo + 1
In the strong shock limit, Po vo /D2 0, the shock state reduces to
P 2 D2
. (9.291)
Po + 1 Po vo
Note, this can also be rewritten as
P 2 D 2
, (9.292)
Po + 1 Po vo
2 D 2
. (9.293)
+ 1 c2o
Lastly, the particle velocity can be obtained via the mass equation, Eq. (9.235):
u o
= , (9.294)
D
v
= , (9.295)
vo
1 2 Po vo
= . (9.296)
+ 1 + 1 D2
In the strong shock limit, Po vo /D2 0, we get
u 1
. (9.297)
D +1
Note that in the laboratory frame, one gets
uD 1
, (9.298)
D +1
u 1
1 , (9.299)
D +1
u 2
. (9.300)
D +1
9.2.5 Equilibrium solutions

Our remaining differential equation, Eq. (9.225) is in equilibrium when r = 0, which for
one-step irreversible kinetics, Eq. (9.12), occurs when = 1. So for equilibrium end states,
we enforce = 1 in Eq. (9.281) and get

1 + D2
2 )
(1 +
v Po vo
= 2
vo 2 PDo vo
r 2
2 )2 4 PDo vo 1 + 1 + PDo vo 2 Poqvo
2 2 2
1 + PDo vo (1 + 2 2
2 (9.301)
2 PDo vo
This has only one free parameter, D. There are two solutions for v/vo at complete reaction.
They can be distinct and real, repeated and real, or complex, depending on the value of
D. We are most interested in D for which the solutions are real; these will be physically
realizable.
9.2.5.1 Chapman-Jouguet solutions

Let us first consider solutions for which the two roots of Eq. (9.301) are repeated. This is
known as a Chapman-Jouguet (CJ) solution. For a CJ solution, the Rayleigh line is tangent
to the Hugoniot at = 1 if the reaction is driven by one-step irreversible kinetics.
We can find values of D for which the solutions are CJ by requiring the discriminant
under the square root operator in Eq. (9.301) to be zero. We label such solutions with a CJ
subscript and say
2
2 2
2

DCJ
2 2 DCJ DCJ 2 2 q
1+ 4
1+ 1+ 1+ 2
= 0. (9.302)
Po vo Po vo Po vo Po vo
2
Equation (9.302) is quartic in DCJ and quadratic in DCJ . It has solutions
q
2
DCJ 1 + 4 4 2 P2q
2 Poqvo o vo
2 (1 + 2
2
4 )
= (9.303)
Po vo 2 1)2
(
The + root corresponds to a large value of DCJ . This is known as the detonation branch.
For the parameter values of Table 9.1, we find by substitution that DCJ = 1991.1 m/s
for our H2 -air mixture. It corresponds to a pressure increase and a volume decrease. The
- root corresponds to a small value of DCJ . It corresponds to a pressure decrease and
a volume increase. It is known as the deflagration branch. For our H2 -air mixture, we
find DCJ = 83.306 m/s. Here we are most concerned with the detonation branch. The
deflagration branch may be of interest, but neglected mechanisms, such as diffusion, may
be of more importance for this branch. In fact laminar flames in hydrogen move much
slower than that predicted by the CJ deflagration speed. We also note that for q 0, that
2
DCJ 2 )/(1
/(Po vo ) = (1 + 2 ) = . Thus for q 0, we have DCJ
2
Po vo , and the wave
speed is the ambient sound speed.
2
Taylor series expansion of the detonation branch in the strong shock limit, Po vo /DCJ 0
shows that
2
DCJ 2q( 2 1), (9.304)

vCJ vo , (9.305)
+1
q
PCJ 2( 1) , (9.306)
vo
p
uCJ 2q( 2 1), (9.307)
+1
p
2q( 2 1)
uCJ . (9.308)
+1
Importantly the Mach number in the wave frame at the CJ state is
CJ u2
M 2
= 2CJ , (9.309)
cCJ
u2CJ
= , (9.310)
PCJ vCJ
2
(+1)2
2q( 2 1)
= , (9.311)
v2qo ( 1)vo +1
2
(+1)2
( 2 1)
= , (9.312)
( 1) +1
2
(+1)2
( + 1)( 1)
= , (9.313)
( 1) +1
= 1. (9.314)
In the strong shock limit, the local Mach number in the wave frame is sonic at the end of
the reaction zone. This can be shown to hold away from the strong shock limit as well.
9.2.5.2 Weak and strong solutions

For D < DCJ there are no real solutions. For D > DCJ , there are two real solutions. These
are known as the weak and strong solution. These solutions are represent the intersection of
the Rayleigh line with the complete reaction Hugoniot at two points. The higher pressure
solution is known as the strong solution. The lower pressure solution is known as the weak
solution. Equilibrium end state analysis cannot determine which of the solutions, strong or
weak, is preferred if D > DCJ .
We can understand much about detonations, weak, strong, and CJ, by considering how
they behave as the final velocity in the laboratory frame is changed. We can think of the final
velocity in the laboratory frame as that of a piston which is pushing the detonation. While
we could analyze this on the basis of the theory we have already developed, the algebra
is complicated. Let us instead return to a more primitive form. Consider the Rankine-
Hugoniot jump equations, Eqs. (9.222-9.224) with caloric state equation, Eq. (9.251) with
the final state, denoted by the subscript f , being = 1 and the initial state being = 0,
Eq. (9.222) being used to simplify Eq. (9.224), and the laboratory frame velocity u used in
place of u:
(uf D) = o D, (9.315)
Pf + (uf D)2 = Po + o D 2 , (9.316)
1 Pf 1 Pf 1 Po 1 2 Po
q + (uf D)2 + = + D + . (9.317)
1 f 2 f 1 o 2 o
| {z } | {z }
ef =eo
Let us consider the unknowns to be Pf , f , and D. Computer algebra solution of these three
equations yields two sets of solutions. The relevant physical branch has a solution for D of
v
u !2
+ 1 1 q u Po +1 1 q
D = uf + + t + + (9.318)
2 2 2
4 2 uf o uf 4 2 uf
We give a plot of D as a function of the supporting piston velocity uf in Figure 9.3.

We notice on Fig. 9.3 that there is a clear minimum D. This value of D is the CJ value of
DCJ = 1991.1 m/s. It corresponds to a piston velocity of uf = 794.9 m/s.
A piston driving at uf = 794.9 m/s will just drive the wave at the CJ speed. At this
piston speed, all of the energy to drive the wave is supplied by the combustion process itself.
As uf increases beyond 794.9 m/s, the wave speed D increases. For such piston velocities,
the piston itself is supplying energy to drive the wave. For uf < 794.9 m/s, our formula
predicts an increase in D, but that is not what is observed in experiment. Instead, a wave
propagating at DCJ is observed.
D HmsL
6000
5000
4000
3000
2000
DCJ
1000
u f HmsL
0 1000 2000 3000 4000
Figure 9.3: Plot of detonation wave speed D versus piston velocity uf ; parameters from
Table 9.1 loosely model H2 -air detonation.
that in the inert, q = 0, small piston velocity, uf 0, limit that Eq. (9.318)
We last note p
reduces to D = Po /o . That is, the wave speed is the ambient sound speed.
9.2.5.3 Summary of solution properties

Here is a summary of the properties of solutions for various values of D that can be obtained
by equilibrium end state analysis:
D < DCJ : No Rayleigh line intersects a complete reaction Hugoniot in real space.
There is no real equilibrium solution.
D = DCJ : There is a two repeated solutions at a single point, which we will call C, the
Chapman-Jouguet point. At C, the Rayleigh line is tangent to the complete reaction
Hugoniot. Some properties of this solution are
uCJ /cCJ = 1; the flow is sonic in the wave frame at complete reaction.
This is the unique speed of propagation of a wave without piston support if the
reaction is one-step irreversible.
At this wave speed, D = DCJ , all the energy from the reaction is just sufficient
to drive the wave forward.
Because the end of the reaction zone is sonic, downstream acoustic disturbances
cannot overtake the reaction zone.
D > DCJ ; Two solutions are admitted at the equilibrium end state, the strong solution
at point S, and the weak solution at point W .
The strong solution S has

u/c < 1, subsonic.
piston support is required to drive the wave forward
some energy to drive the wave comes from the reaction, some comes from the
piston.
if the piston support is withdrawn, acoustic disturbances will overtake and
weaken the wave.
The weak solution W has
u/c > 1, supersonic
often thought to be non-physical, at least for one-step irreversible kinetics
because of no initiation mechanism.
exceptions exist.
9.2.6 ZND solutions: One-step irreversible kinetics

We next consider the structure of the reaction zone. Equation (9.281) gives v() for either
the strong or weak branches of the solution. Knowing v() and thus (), since v = 1/, we
can use the integrated mass equation, Eq. (9.222) to write an explicit equation for u(). Our
Rankine-Hugoniot analysis also give T (). These can be employed in the reaction kinetics
equation, Eq. (9.225, 9.13) to form a single ordinary differential equation for the evolution
of of the form

E
d a(T ()) exp RT ()
(1 )
= , (0) = 0. (9.319)
d
x u()
We consider the initial unshocked state to be labeled O. We label the point after the shock
N for the Neumann point, named after John von Neumann, one of the pioneers of detonation
theory. Recall = 0 both at O and at N. But the state variables, e.g. P , , u, change from
O to N.
Before we actually solve the differential equations, we can learn much by considering
how P varies with in the reaction zone by using Rankine-Hugoniot analysis. Consider
the Rankine-Hugoniot equations, Eqs. (9.222-9.224) with caloric state equation, Eq. (9.251),
Eq. (9.222) being used to simplify Eq. (9.224):
u = o D,
(9.320)
P + u2 = Po + o D 2 , (9.321)
1 P 1 P 1 Po 1 2 Po
q + u2 + = + D + . (9.322)
1 2 1 o 2 o
| {z } | {z }
e =eo
Computer algebra reveals the solution for P () to be

s
2 2
1 Po 2( 1)q
P () = Po + o D 2 1 1 (9.323)
+1 o D 2 D2
P HPaL
N
5. 106 D > DCJ
S
4. 106
3. 106 N
N D = DCJ
2. 106
D < DCJ
C
1. 106
W
D > DCJ
O

0.2 0.4 0.6 0.8 1.0
Figure 9.4: P versus from Rankine-Hugoniot analysis for one-step irreversible reaction
for D = 2800 m/s > DCJ , D = 1991.1 m/s = DCJ , D = 1800 m/s < DCJ for H2 /air-based
parameters of Table 9.1.
In Fig. (9.4), we plot P versus for three different values of D: D = 2800 m/s > DCJ ,
D = 1991.1 m/s = DCJ , D = 1800 m/s < DCJ .
as a function
We can also, via detailed algebraic analysis get an algebraic expression for M
of . We omit that here, but do get a plot for our system. In Fig. (9.5), we plot M
versus for three different values of D: D = 2800 m/s > DCJ , D = 1991.1 m/s = DCJ ,
D = 1800 m/s < DCJ . The plot of M versus is log-log so that the sonic condition may be
clearly exhibited.
For D = 2800 m/s > DCJ , there are two branches, the strong and the weak. The
strong branch commences at N where = 0 and proceeds to decrease to = 1 where the
equilibrium point S is encountered at a subsonic state. There other branch commences at
O and pressure increases until the supersonic W is reached at = 1.
For D = 1991.1 m/s = DCJ , the behavior is similar, except that the branches commenc-
ing at N and O both reach complete reaction at the same point C. The point C can be shown
to be sonic with M = 1. We recall from our earlier discussion regarding Eqs. (9.218-9.221)
that sonic points are admitted only if r = 0. For one-step irreversible reaction, r = 0 when
= 1, so a sonic condition is admissible.
For D = 1800 m/s < DCJ , the strong and weak branches merge at a point of incomplete
reaction. At the point of merger, near = 0.8, the flow is locally sonic; however, this is not
a point of complete reaction, so there can be no real-valued detonation structure for this
value of D.
Finally, we write an alternate differential-algebric equations which can be integrated for
`
M
O
5.0 O
D > DCJ
O
3.0
D < DCJ D = DCJ W
2.0
1.5
1.0 C
D < DCJ
S
N D > DCJ

0.01 0.02 0.05 0.10 0.20 0.50 1.00
Figure 9.5: M versus from Rankine-Hugoniot analysis for one-step irreversible reaction
for D = 2800 m/s > DCJ , D = 1991.1 m/s = DCJ , D = 1800 m/s < DCJ for H2 /air-based
parameters of Table 9.1.
the detonation structure:

u = o D,
(9.324)
P + u2 = Po + o D 2 , (9.325)
1 P 1 Po
e + u2 + = eo + D 2 + , (9.326)
2 2 o
1 P
e = q, (9.327)
1
P = RT, (9.328)

d 1 E
= a exp (9.329)
d
x u RT
We need the condition (0) = 0. These form six equations for the six unknowns , u, P , e,
T , and .
9.2.6.1 CJ ZND structures

We now fix D = DCJ = 1991.1 m/s and integrate Eq. (9.319) from the shocked state N
to a complete reaction point, C, the Chapman-Jouguet detonation state. We could also
integrate from O to C, but this is not observed in nature. After obtaining ( x), we can use
our Rankine-Hugoniot analysis results to plot all state variables as functions of x.
Hkgm3 L
5
`
HmL x
-0.015 -0.010 -0.005 0.000
Figure 9.6: ZND structure of ( x) for D = DCJ = 1991.1 m/s for one-step irreversible
reaction for H2 /air-based parameters of Table 9.1.
P HPaL
2.5 106
2. 106
1.5 106
1. 106
500 000
`
HmL x
-0.010 -0.005 0.000
Figure 9.7: ZND structure of P ( x) for D = DCJ = 1991.1 m/s for one-step irreversible
In Fig. (9.6), we plot the density versus x. The density first jumps discontinuously from
O to its shocked value at N. From there it slowly drops through the reaction zone until it
relaxes near x 0.01 m to its equilibrium value at complete reaction. Thus the reaction
zone has a thickness of roughly 1 cm. Similar behavior is seen for the pressure in Fig. 9.7.
The wave frame velocity is shown in Fig. 9.8. Since the unshocked fluid is at rest in
the laboratory frame with u = 0 m/s, the fluid in the wave frame has velocity u = 0
1991.1 m/s = 1991.1 m/s. It is shocked to a lower velocity and then relaxes to its
equilibrium value.
The structure of the velocity profiles is easier to understand in the laboratory frame, as
shown in Fig. 9.9. Here we see the unshocked fluid velocity of 0 m/s. The fluid is shocked
to a high velocity, which then decreases to a value at the end of the reaction zone. The final
velocity can be associated with that of a supporting piston, uf = 794.9 m/s.
The temperature is plotted in Fig. 9.10. It is shocked from 298 K to a high value, then
continues to mainly increase through the reaction zone. Near the end of the reaction zone,
there is a final decrease as it reaches its equilibrium value.
The Mach number calculated in the wave frame, M is shown in Fig. 9.11. It goes from an
`
HmsL u
`
HmL x
-0.010 -0.005 0.000
-500
-1000
-1500
-2000
Figure 9.8: ZND structure of wave frame-based fluid particle velocity u(x) for D = DCJ =
1991.1 m/s for one-step irreversible reaction for H2 /air-based parameters of Table 9.1.
u HmsL
1500
1000
500
0
`
HmL x
-0.010 -0.005 0.000
Figure 9.9: ZND structure of laboratory frame-based fluid particle velocity u( x) for D =
DCJ = 1991.1 m/s for one-step irreversible reaction for H2 /air-based parameters of Table 9.1.
T HKL
2500
2000
1500
1000
500
`
HmL x
-0.010 -0.005 0.000
Figure 9.10: ZND structure of T ( x) for D = DCJ = 1991.1 m/s for one-step irreversible
`
M
5
`
HmL x
-0.015 -0.010 -0.005 0.000
Figure 9.11: ZND structure of wave frame-based Mach number M (

x) for D = DCJ =
1991.1 m/s for one-step irreversible reaction for H2 /air-based parameters of Table 9.1.

1.0
0.8
0.6
0.4
0.2
0.0
`
HmL x
-0.015 -0.010 -0.005 0.000
Figure 9.12: ZND structure of ( x) for D = DCJ = 1991.1 m/s for one-step irreversible
= 4.88887, which we call the CJ detonation Mach number, MCJ = 4.88887,

initial value of M
to a post-shock value of M = 0.41687. Note this result confirms a standard result from
compressible flow that a standing normal shock must bring a flow from a supersonic state to
a subsonic state. At equilibrium it relaxes to M = 1. This relaxation to a sonic state when
= 1 is what defines the CJ state. We recall that this result is similar to that obtained
in so-called Rayleigh flow of one-dimensional gas dynamics. Rayleigh flow admits heat
transfer to a one-dimensional channel, and it is well known that the addition of heat always
induces the flow to move to a sonic (or choked) state. So we can think of the CJ detonation
wave as a thermally choked flow.
The reaction progress variable is plotted in Fig. 9.12. Note that it undergoes no shock
jump and simply relaxes to its equilibrium value of = 1 near x = 0.01 m.
Lastly, we plot T (
x) on a log-log scale in Fig. (9.13). The sign of x is reversed so as to
avoid the plotting of the logarithms of negative numbers. We notice on this scale that the
temperature is roughly that of the shock until x = 0.001 m, at which point a steep rise
begins. We call this length the induction length, ind . When we compare this figure to Fig. 2
T HKL
2600.
2400.
2200.
2000.
1800.
`
-HmL x
10-5 10-4 0.001 0.01
Figure 9.13: ZND structure on a log-log scale of T ( x) for D = DCJ = 1991.1 m/s for
one-step irreversible reaction for H2 /air-based parameters of Table 9.1.
parameter simple detailed

2
rxn 10 m 102 m
3
ind 10 m 104 m
DCJ 1991.1 m/s 1979.7 m/s
Ps 2.80849 10 P a 2.8323 106 P a
6
PCJ 1.4553 106 P a 1.6483 106 P a

Ts 1664.4 K 1542.7 K
TCJ 2570.86 K 2982.1 K
s 4.244 kg/m3 4.618 kg/m3
CJ 1.424 kg/m3 1.5882 kg/m3
Mo 4.88887 4.8594
Ms 0.41687 0.40779
M CJ 1 0.93823
Table 9.2: Numerical values of parameters which roughly model CJ H2 -air detonation.
of Powers and Paolucci, 2005, we see a somewhat similar behavior. However the detailed
kinetics model shows ind 0.0001 m. The overall reaction zone length rxn is predicted well
by the simple model. Its value of rxn 0.01 m is also predicted by the detailed model. Some
of the final values at the end state are different as well. This could be due to a variety of
factors, especially differences in the state equations. Comparisons between values predicted
by the detailed model of Powers and Paolucci, 2005, against those of the simple model here
are given in Table 9.2.
9.2.6.2 Strong ZND structures

We increase the detonation velocity D to D > DCJ and obtain strong detonation structures.
These have a path from O to N to the equilibrium point S. These detonations require piston
support to propagate, as the energy supplied by heat release alone is insufficient to maintain
Hkgm3 L
5
`
HmL x
-0.0002 -0.0001 0
Figure 9.14: ZND structure of (x) for strong D = 2800 m/s > DCJ for one-step irreversible
P HPaL
5. 106
4. 106
3. 106
2. 106
1. 106
`
HmL x
-0.0001 0
Figure 9.15: ZND structure of P (

x) for strong D = 2800 m/s > DCJ for one-step irreversible
their steady speed.

Similar to our plots of the CJ structures, we give plots of the strong, D = 2800 m/s
structures of (
x), P (
x), u(
x), u(
x), T ( x), (
x), M( x) in Figs. 9.14-9.20, respectively.
The behavior of the plots is qualitatively similar to that for CJ detonations. We see
however that the reaction zone has become significantly thinner, rxn 0.0001 m. This is
because the higher temperatures associated with the stronger shock induce faster reactions,
thus thinning the reaction zone. Comparison with Fig. 9.2 reveals that the shocked and final
values of pressure agree with those of the Rankine-Hugoniot jump analysis. We also note
is subsonic. This allows information to propagate from the supporting
the final value of M
piston all the way to the shock front.
9.2.6.3 Weak ZND structures

For the simple one-step irreversible kinetics model, there is no path from O through the
shocked state N to the weak solution W . There is a direct path from O to W ; however,
it is physically unrealistic. For D = 2800 m/s, we plot versus x in Fig. 9.21. Numerical
`
HmsL u
`
HmL x
-0.0001 0
-500
-1000
-1500
-2000
-2500
Figure 9.16: ZND structure of wave frame fluid particle velocity u( x) for strong D =
2800 m/s > DCJ for one-step irreversible reaction for H2 /air-based parameters of Table 9.1.
u HmsL
2000
1500
1000
500
0
`
HmL x
-0.0001 0
Figure 9.17: ZND structure of laboratory frame fluid particle velocity u(

x) for strong
D = 2800 m/s > DCJ for one-step irreversible reaction for H2 /air-based parameters of
Table 9.1.
T HKL
4000
3000
2000
1000
`
HmL x
-0.0001 0
Figure 9.18: ZND structure of T (

x) for strong D = 2800 m/s > DCJ for one-step irreversible
`
M
7
1
`
HmL x
-0.0002 -0.0001 0
Figure 9.19: ZND structure of wave frame Mach number M (

x) for strong D = 2800 m/s >
DCJ for one-step irreversible reaction for H2 /air-based parameters of Table 9.1.

1.0
0.8
0.6
0.4
0.2
0.0
`
HmL x
-0.0002 -0.0001 0
Figure 9.20: ZND structure of (x) for strong D = 2800 m/s > DCJ for one-step irreversible
0.08
0.06
0.04
0.02
`
HmL x
-3.2249 1021 -1.61245 1021
Figure 9.21: ZND structure of ( x) for unshocked, weak D = 2800 m/s > DCJ for one-step
irreversible reaction for H2 /air-based parameters of Table 9.1. The galactic distance scales
are far too large to be realistic representations of reality!
solution was available only until 0.02. Numerical precision issues arose at this point.
Note, importantly, that rxn 1021 m is unrealistically large! Note the distance from Earth
to the Large Magellanic Cloud, a dwarf galaxy orbiting the Milky Way, is 1021 m. Our
combustion model is not well calibrated to those distances, so it is entirely unreliable to
predict this class of weak detonation!
9.2.6.4 Piston problem

We can understand the physics of the one-step kinetics problem better in the context of
a piston problem, where the supporting piston connects to the final laboratory frame fluid
x ). Let us consider pistons with high velocity and then lower them
velocity up = u(
and examine the changes of structure.
up > up,CJ . This high velocity piston will drive a strong shock into the fluid at a speed
D > DCJ . The solution will proceed from O to N to S and be subsonic throughout.
Therefore changes at the piston face will be able to be communicated all the way to
the shock front. The energy to drive the wave comes from a combination of energy
released during combustion and energy supplied by the piston support.
up = up,CJ At a critical value of piston velocity, up,CJ , the solution will go from O to
N to C, and where it is locally sonic.
up < up,CJ For such flows, the detonation wave is self-supporting. There is no means to
communicate with the supporting piston. The detonation wave proceeds at D = DCJ .
We note that DCJ is not a function of the specific kinetic mechanism. So for a one-step
irreversible kinetics model, the conclusion the DCJ is the unique speed of propagation of
an unsupported wave is verified. It should be noted that nearly any complication added
to the model, e.g. reversibility, multi-step kinetics, multi-dimensionality, diffusion, etc.,
will alter this conclusion.
9.2.7 Detonation structure: Two-step irreversible kinetics

Let us consider a small change to the one-step model of the previous sections. We will now
consider a two-step irreversible kinetics model. The first reaction will be exothermic and
the second endothermic. Both reactions will be driven to completion, and when they are
complete, the global heat release will be identical to that of the one-step reaction. All other
parameters will remain the same from the one-step model.
We will see that this simple modification has profound effects on what is a preferred
detonation structure. In particular, we will see that for such a two-step model
the CJ structure is no longer the preferred state of an unsupported detonation wave,
the steady speed of the unsupported detonation wave is unique and greater than the
CJ speed,
there is a path from the unshocked state O to the shocked state N through a sonic
incomplete reaction pathological point P to the weak equilibrium end state W ,
there is a strong analog to steady compressible one-dimensional inert flow with area
change, i.e. rocket nozzle flow.
Let us pose the two step irreversible kinetics model of
1 : A B, (9.330)
2 : B C. (9.331)
Let us insist that A, B, and C each have the same molecular mass, MA = MB = MC = M,
and the same constant specific heats, cP A = cP B = cP C = cP . Let us also insist that both
reactions have the same kinetic parameters, E1 = E2 = E, a1 = a2 = a, 1 = 2 = 0.
Therefore the reaction rates are such that
k1 = k2 = k. (9.332)
Let us assume at the initial state that we have all A and no B or C: A (0) = A , B (0) = 0,
C (0) = 0. Since J = 2, we have a reaction vector rj of length 2:

r1
rj = . (9.333)
r2
Our stoichiometric matrix ij has dimension 3 2 since N = 3 and J = 2:

1 0
ij = 1 1 . (9.334)
0 1
P
For species production rates, from i = Jj=1 ij rj we have

A / A 1 0 r1
d 1 1 r1 1
B / = B = 1 1 = r1 r2 . (9.335)
dt r2
C / C 0 1 r2
We recall here that d/dt denotes the material derivative following a fluid particle, d/dt =
/t + u/ x. For our steady waves, we will have d/dt = ud/d x. We next recall that
i / = Yi /Mi, which for us is Yi/M, since the molecular masses are constant. So we have

YA A 1 0 r1
d M M r1 M
YB = B = 1 1 = r1 r2 . (9.336)
dt r2
YC C 0 1 r2
Elementary row operations gives us the row echelon form

YA 1 0
d M 0 1 r1 .
YA + YB = (9.337)
dt r2
YA + YB + YC 0 0
We can integrate the homogeneous third equation and apply the initial condition to get
YA + YB + YC = 1. (9.338)
This can be thought of as an unusual matrix equation:

YA
( 1 1 1 ) YB = ( 1 ) . (9.339)
YC
+ D and get
We can perform an analogous exercise to finding the form =

YA 1 1 0
YB = 0 + 1 1 1 . (9.340)
2
YC 0 0 1
| {z }
=F
The column vectors of F are linearly independent and lie in the right null space of the
coefficient matrix (1, 1, 1). The choices for F are not unique, but are convenient. We can
think of the independent variables 1 , 2 as reaction progress variables. Thus, for reaction
1, we have 1 , and for reaction 2, we have 2 . Both 1 (0) = 0 and 2 (0) = 0. The mass
fraction of each species can be related to the reaction progress via
Y A = 1 1 , (9.341)
Y B = 1 2 , (9.342)
Y C = 2 . (9.343)
When the reaction is complete, we have 1 1, 2 1, and YA 0, YB 0, YC 1.
Now our reaction law is

YA r1 kA YA
d M M
YB = r1 r2 = kA kB = k YA YB (9.344)
dt
YC r2 kB YB
Eliminating YA , YB and YC in favor of 1 and 2 , we get

1 1 (1 1 )
d
1 2 = k (1 1 ) (1 2 ) (9.345)
dt
2 1 2
This reduces to

1 1 1
d
1 2 = k 1 21 + 2 (9.346)
dt
2 1 2
The second of these equations is the difference of the first and the third, so it is redundant
and we need only consider

d 1 1 1
=k (9.347)
dt 2 1 2
In the steady wave frame, this is written as
d1
u = (1 1 )k, (9.348)
d
x
d2
u = (1 2 )k. (9.349)
d
x
Because the rates k1 = k2 = k have been taken identical, we can actually get 2 (1 ). Dividing
our two kinetic equations gives
d2 1 2
= . (9.350)
d1 1 1
Because 1 (0) = 0 and 2 (0) = 0, we can say that 2 (1 = 0) = 0. We rerrange this
differential equation to get
d2 1 1
+ 2 = . (9.351)
d1 1 1 1 1
This equation is first order and linear. It has an integrating factor of

Z
d1 1
exp = exp ( ln(1 1 )) = .
1 1 1 1
Multiplying both sides by the integrating factor, we get

2
1 d2 1 1
+ 2 = . (9.352)
1 1 d1 1 1 (1 1 )2
Using the product rule, we then get

d 2 1
= , (9.353)
d1 1 1 (1 1 )2
Z
2 1
= d1 , (9.354)
1 1 (1 1 )2
(9.355)
Taking u = 1 and dv = d1 /(1 1 )2 and integrating the right side by parts, we get
Z
2 1 d1
= , (9.356)
1 1 1 1 1 1
1
= + ln(1 1 ) + C, (9.357)
1 1
2 = 1 + (1 1 ) ln(1 1 ) + C(1 1 ). (9.358)
Now since 2 (1 = 0) = 0, we get C = 0, so
2 ( x) + (1 1 (
x) = 1 ( x)) ln(1 1 (
x)). (9.359)
Leaving out details of the derivation, our state equation becomes
e(T, 1 , 2 ) = cv (T To ) 1 q1 2 q2 . (9.360)
We find it convenient to define Q(1 , 2 ) as
Q(1 , 2 ) 1 q1 + 2 q2 . (9.361)
So the equation of state can be written as
e(T, 1 , 2 ) = cv (T To ) Q(1 , 2 ). (9.362)
The frozen sound speed remains

P
c2 = P v = . (9.363)


1.4
M 20.91 kg/kmole
R 397.58 J/kg/K
Po 1.01325 105 Pa
To 298 K
o 0.85521 kg/m3
vo 1.1693 m3 /kg
q1 7.58265 106 J/kg
q2 5.68698 106 J/kg
E 8.29352 106 J/kg
a 5 109 1/s
0
Table 9.3: Numerical values of parameters for two-step irreversible kinetics.
There are now two thermicities:

1 P q1
1 = = , (9.364)
c2 1 v,e,2 1 P

1 P q2
2 = 2 = , (9.365)
c 2 v,e,1 1 P
(9.366)
Parameters for our two step model are identical to those of our one step model, except
for the heat releases. The parameters are listed in Table 9.3. Note that at complete reaction
Q(1 , 2 ) = Q(1, 1) = q1 + q2 = 1.89566 106 J/kg. Thus, the overall heat release at
complete reaction 1 = 2 = 1 is identical to our earlier one-step kinetic model.
Let us do some new Rankine-Hugoniot analysis. We can write a set of mass, momentum,
energy, and state equations as
u = o D,
(9.367)
P + u2 = Po + o D 2 , (9.368)
1 P 1 Po
e + u2 + = eo + D 2 + , (9.369)
2 2 o
1 P
e = 1 q1 2 q2 , (9.370)
1
2 = 1 + (1 1 ) ln(1 1 ). (9.371)
Let us consider D and 1 to be unspecified but known parameters for this analysis. These
equations are five equations for the five unknowns, , u, P , e, and 2 . They can be solved
for (D, 1 ), u(D, 1 ), P (D, 1), e(D, 1 ), and 2 (1 ).
P HPaL
N
5. 106 N S

D>D
N S
4. 106

D=D D=D
3. 106
D<D P

D<D
D=D
2. 106

D>D
1. 106
W
O
1
0.2 0.4 0.6 0.8 1.0
Figure 9.22: Pressure versus 1 for two-step kinetics problem for three different values of
D: D = 2800 m/s > D, D = 2616.5 m/s = D, D = 2500 m/s < D; parameters are from
Table 9.3.
The solution is lengthy, but the plot is revealing. For three different values of D, pressure
as a function of 1 is shown in in Fig. 9.22. There are three important classes of D, each
shown in Fig. 9.22, depending on how D compares to a critical value we call D.
D > D. There are two potentially paths here. The important physical branch starts
at point O, and is immediately shocked to state N, the Neumann point. From N the
pressure first decreases as 1 increases. Near 1 = 0.75, the pressure reaches a local
minimum, and then increases to the complete reaction point at S. This is a strong
solution. There is a second branch which commences at O and is unshocked. On this
branch the pressure increases to a maximum, then decreases to the end state at W .
While this branch is admissible mathematically, its length scales are unphysically long,
and this branch is discarded.
D = D. Let us only consider branches which are shocked from O to N. The unshocked
branches are again non-physical. On this branch, the pressure decreases from N to
the pathological point P . At P , the flow is locally sonic, with M = 1. Here the
pressure can take two distinct paths. The one chosen will depend on the velocity of
the supporting piston at the end state. On one path the pressure increases to its final
value at the strong point S. On the other the pressure decreases to its final value at
the weak point W .
D < D. For such values of D, there is no physical structure for the entire reaction
zone 0 < 1 < 1. This branch is discarded.
`
M
O
5.0 O
3.0

D>D
W
2.0

D<D
1.5
D=D
P
1.0

D=D
D<D

D>D S
N

0.01 0.02 0.05 0.10 0.20 0.50 1.00 1
Figure 9.23: Mach number M versus 1 for two-step kinetics problem for three different
D = 2616.5 m/s = D,
values of D: D = 2800 m/s > D, D = 2500 m/s < D; parameters
are from Table 9.3.
Mach number in the wave frame, M as a function of 1 is shown in in Fig. 9.23. The
results here are similar to those in Fig. 9.22. Note the ambient point O is always supersonic,
and the Neumann point N is always subsonic. For flows originating at N, if D > D, the flow

remains subsonic throughout until its termination at S. For D = D, the flow can undergo
a subsonic to supersonic transition at the pathological point P . The weak point W is a
supersonic end state.
The important Fig. 9.22 bears remarkable similarity to curves of P (x) in compressible
inert flow in a converging-diverging nozzle. We recall that for such flows, a subsonic to
supersonic transition is only realized at an area minimum. This can be explained because the
equation for evolution of pressure for such flows takes the form dP/dx (dA/dx)/(1 M 2 ).
So if the flow is locally sonic, it must encounter a critical point in area, dA/dx = 0 in order
to avoid infinite pressure gradients. This is what is realized in actual nozzles.
For us the analogous equation is the two-step version of Eq. (9.220), which can be shown
to be
dP
u(1 r1 + 2 r2 )
= . (9.372)
d
x 1M 2
Because the first reaction is exothermic, we have 1 > 0, and because the second reaction
is endothermic, we have 2 < 0. With r1 > 0, r2 > 0, this gives rise to the possibility that
1 r1 + 2 r2 = 0 at a point where the reaction is incomplete. The point P is just such a point;
it is realized when D = D.
Finally, we write our differential-algebric equations which are integrated for the detona-
tion structure:
u = o D,
(9.373)
P + u2 = Po + o D 2 , (9.374)
1 P 1 Po
e + u2 + = eo + D 2 + , (9.375)
2 2 o
1 P
e = 1 q1 2 q2 , (9.376)
1
P = RT, (9.377)
2 = 1 + (1 1 ) ln(1 1 ), (9.378)

d1 1 1 E
= a exp (9.379)
d
x u RT
We need the condition 1 (0) = 0. These form seven equations for the seven unknowns ,
u, P , e, T , 1 , and 2 . We also realize that the algebraic solutions are multi-valued and
must take special care to be on the proper branch. This becomes particularly important for
solutions which pass through P .
9.2.7.1 Strong structures

and D = D.
Here we consider strong structures for two cases: D > D All of these will
proceed from O to N through a pressure minimum, and finish at the strong point S.
9.2.7.1.1 D > D Structures for a strong detonation with D = 2800 m/s > D are given
in Figs. 9.24-9.31. The structure of all of these can be compared directly to those of the one-
step kinetics model at the same D = 2800 m/s, Figs. 9.14-9.20. Note the shock values are
identical. The reaction zone thicknesses are quite similar as well at rxn 0.0001 m. The
structures themselves have some differences; most notably, the two-step model structures
display interior critical points before complete reaction.
We take special note of the pressure plot of Fig. 9.25, which can be compared with
Fig. 9.22. We see in both figures the shock from O to N, followed by a drop of pressure
to a minimum, followed by a final relaxation to an equilibrium value at S. Note that the
two curves have the opposite sense of direction as 1 commences at 0 and goes to 1, while x
commences at 0 and goes to 0.0002 m.
We can also compare the M( x) plot of Fig. 9.29 with that of M
(1 ) of Fig. 9.23. In both
the supersonic O is shocked to a subsonic N. The Mach number rises slightly then falls in
the reaction zone to its equilibrium value at S. It never returns to a supersonic state.
9.2.7.1.2 D = D For D = D = 2616.5 m/s, we can find a strong structure with a path
from O to N to S. Pressure P and Mach number M are plotted in Figures 9.32-9.33. Note
that at an interior point in the structure, a cusp in the P and M profile is seen. At this
= 1.
point, the flow is locally sonic with M
Hkgm3 L
5
`
HmL x
-0.0002 -0.0001 0
Figure 9.24: ZND structure of ( for two-step irreversible

x) for strong D = 2800 m/s > D
reaction with parameters of Table 9.3.
P HPaL
5. 106
4. 106
3. 106
2. 106
1. 106
`
HmL x
-0.0001 0
Figure 9.25: ZND structure of P ( for two-step irreversible

`
HmsL u
`
HmL x
-0.0001 0
-500
-1000
-1500
-2000
-2500
Figure 9.26: ZND structure of u( for two-step irreversible

u HmsL
2000
1500
1000
500
0
`
HmL x
-0.0001 0

T HKL
5000
4000
3000
2000
1000
`
HmL x
-0.0001 0
Figure 9.28: ZND structure of T ( for two-step irreversible

`
M
7
1
`
HmL x
-0.0002 -0.0001 0

Figure 9.29: ZND structure of M( for two-step irreversible
1
1.0
0.8
0.6
0.4
0.2
0.0
`
HmL x
-0.0002 -0.0001 0
Figure 9.30: ZND structure of 1 ( for two-step irreversible

2
1.0
0.8
0.6
0.4
0.2
0.0
`
HmL x
-0.0002 -0.0001 0

P HPaL
5. 106
4. 106
3. 106
2. 106
1. 106
`
HmL x
-0.0001 0

x) for strong D = 2616.5 m/s = D
`
M
7
1
`
HmL x
-0.0002 -0.0001 0
(
Figure 9.33: ZND structure of M for two-step irreversible
x) for strong D = 2616.5 m/s = D
Hkgm3 L
5
`
HmL x
-0.000292911 0
Figure 9.34: ZND structure of ( for two-step irreversible

x) for weak D = 2616.5 m/s = D
9.2.7.2 Weak, eigenvalue structures

Let us now consider weak structures with D = D = 2616.5 m/s. Special care must be
taken in integrating the governing equations. In general one must integrate to very near the
pathological point P , then halt. While there are more sophisticated techniques involving
further coordinate transformations, one can record the values near P on its approach from
N. Then one can perturb slightly all state variables so that the M is just greater than unity
and recommence the integration.
Figures of the structures which commence at O, are shocked to N, pass through sonic
point P , and finish at the supersonic W are shown in Figs. 9.34-9.41. One may again compare
the pressure and Mach number plots of Figs. 9.35,9.39 with those of Figs. 9.22,9.23.
9.2.7.3 Piston problem

We can understand the physics of the two-step kinetics problem better in the context of
a piston problem, where the supporting piston connects to the final laboratory frame fluid
P HPaL
5. 106
4. 106
3. 106
2. 106
1. 106
`
HmL x
-0.000585822 -0.000292911 0

`
HmsL u
`
HmL x
-0.000585822 -0.000292911 0
-500
-1000
-1500
-2000
-2500

x ). Let us consider pistons with high velocity and then lower them
velocity up = u(
and examine the changes of structure.
up > ups . This high velocity piston will drive a strong shock into the fluid at a speed
D > D. The solution will proceed from O to N to S and be subsonic throughout.
Therefore changes at the piston face will be able to be communicated all the way to
the shock front. The energy to drive the wave comes from a combination of energy
released during combustion and energy supplied by the piston support.
up = ups . At a critical value of piston velocity, ups , the solution will go from O to N
to P to S, and be locally sonic. This is analogous to the subsonic design condition
for a converging-diverging nozzle.
up [
ups , upw ]. Here the flow can be complicated. Analogous to flow in a nozzle, there
can be standing shock waves in the supersonic portion of the flow which decelerate
the flow so as to match the piston velocity at the end of the reaction. Such flows will
proceed from O to N through P , and then are shocked back onto the subsonic branch
to terminate at S.
u HmsL
2000
1500
1000
500
0
`
HmL x
-0.000585822 -0.000292911 0

x) forweak D = 2616.5 m/s = D
T HKL
4000
3000
2000
1000
`
HmL x
-0.000585822 -0.000292911 0
Figure 9.38: ZND structure of T ( for two-step irreversible

up = upw . This state is analogous to the supersonic design condition of flow in a

converging-diverging nozzle. The fluid proceeds from O to N through P and terminates
at W . All of the energy to propagate the wave comes from the reaction.
up < upw . For such flows, the detonation wave is self-supporting. There is no means
to communicate with the supporting piston. The detonation wave proceeds at D = D.

We note that D is a function of the specific kinetic mechanism. This non-classical result
contradicts the conclusion from CJ theory with simpler kinetics in which the wave speed
of an unsupported detonation is independent of the kinetics. Note for our problem that
D = 2616.5 m/s. This stands in contrast to the CJ velocity, independently computed
of DCJ = 1991.1 m/s for the same mixture.
9.2.8 Detonation structure: Detailed H2 O2 N2 kinetics

These same notions for detonation with simple kinetics and state equations can easily be
extended to more complex models. Let us consider a one-dimensional steady detonation in a
stoichiometric hydrogen air mixture with the detailed kinetics model of Table 1.2. We shall
`
M
7
1
`
HmL x
-0.000585822 -0.000292911 0
Figure 9.39: ZND structure of M ( for two-step irreversible

1
1.0
0.8
0.6
0.4
0.2
0.0
`
HmL x
-0.000585822 -0.000292911 0

consider a case almost identical to that studied by Powers and Paolucci. 3 The mixture is
a stoichiometric hydrogen-air mixture of 2H2 + O2 + 3.76N2 . As we did in our modeling of
the same mixture under spatially homogeneous isochoric, adiabatic conditions in an earlier
chapter, we will take the number of moles of each of the minor species to be a small number
near machine precision. This has the effect of removing some numerical roundoff errors in
the very early stages of reaction. Our model will be the steady one-dimensional reactive
Euler equations, obtained by considering the reactive Navier-Stokes equations in the limit
as , jm q
i , j all go to zero.
Our model then is 1) the integrated mass, momentum, and energy equations, Eqs. (9.222-
9.224) with an opposite sign on D to account for the left-running wave, 2) the one-dimensional
steady diffusion-free version of species evolution, Eq. (6.40), 3) the calorically imperfect ideal
gas state equations of Eqs. (6.52), (6.60), and 4) the law of mass action with Arrhenius
kinetics of Eqs. (6.65-6.68).
3
2
1.0
0.8
0.6
0.4
0.2
0.0
`
HmL x
0

u = o D, (9.380)
P + u2 = Po + o D 2 , (9.381)

1 2 P 1 2 Po
u e + u +
= o D eo + D + , (9.382)
2 2 o
dYi

u = Mi i , (9.383)
dx
XN
Yi
P = RT , (9.384)
i=1
Mi
XN Z T
e = o
Yi eTo ,i +
cvi (T )dT , (9.385)
i=1 To
J
X
i = ij rj , (9.386)
j=1
N N
!
Y 1 Y kj
rj = kj 1 kkj . , (9.387)
k=1
Kc,j k=1 k

j E j
kj = aj T exp , (9.388)
RT
PNi=1 ij
Po Goj
Kc,j = exp . (9.389)
RT RT
In order to cleanly plot the results on a log scale, we will, in contrast to the previous
right-running detonations, consider left running waves with detonation velocity D. This
differential-algebraic system must be solved numerically. One can use standard differential-
algebraic methods to achieve this. Alternatively, and with significant effort, one can remove
all of the algebraic constraints. Part of this requires a numerical iteration to find certain
roots. After this effort, one can in principle write a set of N L differential equations for
evolution of the N L independent species.
Here, we chose D DCJ = 1979.70 m/s. Because this model is not a simple one-step
model, we cannot expect to find the equilibrium state to be exactly sonic. However, it is
possible to slightly overdrive the wave and achieve a nearly sonic state at the equilibrium
state. Here, our final Mach number was M = 0.9382. Had we weakened the overdrive
further, we would have encountered an interior sonic point at a non-equilibrium point, thus
inducing a non-physical sonic singularity.
Numerical solution for species mass fraction is given in Fig. (9.42). The figure is plotted on
0
10

10
Yi
OH
H 2O
H2
O2
10 H
O
HO 2
H 2O2
N2
0
10 10 10
x (m)
Figure 9.42: Detailed kinetics ZND structure of species mass fractions for near CJ detonation,
DCJ D = 1979.70 m/s, in 2H2 + O2 + 3.76N2 , Po = 1.01325 105 P a, To = 298 K,
Ts = 1542.7 K,Ps = 2.83280 106 P a
a log-log scale because of the wide range of length scales and mass fraction scales encountered.
Note that the minor species begin to change at a very small length scale. At a value of
x 2.6 104 m, a significant event occurs, known as a thermal explosion. This length is
known as the induction length, ind = 2.6104 m. We get a rough estimate of the induction
time by the formula tind ind /us = 7.9 107 s. Here us is the post shock velocity in the
wave frame. Its value is us = 330.54 m/s. All species contribute to the reaction dynamics
here. This is followed by a relaxation to chemical equilibrium, achieved around 0.1 m.
The pressure profile is given in Fig. (9.2.8). We artificially located the shock just away
6
10
P (Pa)
5
10
10 5 0
10 10 10
x (m)
Figure 9.43: Detailed kinetics ZND structure of pressure for near CJ detonation, DCJ
D = 1979.70 m/s, in 2H2 + O2 + 3.76N2 , Po = 1.01325 105 P a, To = 298 K, Ts =
1542.7 KPs = 2.83280 106 P a.
from x = 0, so as to ease the log-log plot. The pressure is shocked from it atmospheric value
to 2.83280 106 P a (see Table 9.2). After the shock, the pressure holds nearly constant for
several decades of distance. Once the thermal explosion commences, the pressure relaxes to
its equilibrium value. This figure can be compared with its one-step equivalent of Fig. 9.15.
Similar behavior is seen in the temperature plot of Fig. 9.44. The temperature is shocked
to Ts = 1542.7 K, stays constant in the induction zone, and then increases to its equilibrium
value after the thermal explosion. This figure can be compared with its one-step equivalent
of Fig. 9.18.
It is very interesting to compare these results to those obtained in an earlier chapter.
Earlier, an isochoric, adiabatic combustion of precisely the same stoichiometric hydrogen-air
mixture with precisely the same kinetics was conducted. The adiabatic/isochoric mixture
had an initial temperature and pressure identical to the post-shock pressure and temperature
here. We compare the induction time of the spatially homogeneous problem, tind = 6.6
107 s, Eq. (1.362) to our estimate from the detonation found earlier, tind ind / us =
7
7.9 10 s. The two are remarkably similar!
We compare some other relevant values in Table 9.4. Note that these mixtures are
identical at the onset of the calculation. The detonating mixture has reached the same initial
state after the shock. And a fluid particle advecting through the detonation reaction zone
with undergo a thermal explosion at nearly the same time a particle that was stationary
in the closed vessel will. After that the two fates are different. This is because there is
Parameter Spatially Homogeneous Detonation

To 1542.7 K 298 K
Ts 1542.7 K
Po 2.83280 106 P a 1.01325 105 P a
Ps 2.83280 106 P a
tind 6.6 107 s 7.9 107 s
T eq 3382.3 K 2982.1 K
P eq 5.53 106 P a 1.65 106 P a
eq 4.62 kg/m3 1.59 kg/m3
ueq 0 m/s 1066 m/s
YOeq2 1.85 102 1.38 102
YHeq 5.41 104 2.71 104
eq
YOH 2.45 102 1.48 102
YOeq 3.88 103 1.78 103
YHeq2 3.75 103 2.57 103
YHeq2 O 2.04 101 2.22 101
eq
YHO 2
6.84 105 2.23 105
YHeq2 O2 1.04 105 3.08 106
YNeq2 7.45 101 7.45 101
Table 9.4: Comparison of relevant predictions of a spatially homogeneous model with those
of a near CJ detonation in the same mixture.
3500
3000
2500
2000
T (K)
1500
1000
500
0 0
10 10 10
x (m)
Figure 9.44: Detailed kinetics ZND structure of temperature for near CJ, DCJ D =
1979.70 m/s, detonation in 2H2 + O2 + 3.76N2 , Po = 1.01325 105 P a, To = 298 K,
Ts = 1542.7 K, Ps = 2.83280 106 P a.
no kinetic energy in the spatially homogeneous problem. Thus all the chemical energy
is transformed into thermal energy. This is reflected in the higher final temperature and
pressure of the spatially homogeneous problem relative to the detonating flow. Because the
final temperature is different, the two systems relax to a different chemical equilibrium, as
reflected in the different mass fractions. For example, note that the cooler detonating flow
has a higher final mass fraction of H2 O.
Chapter 10
Blast waves
Here we will study the Taylor-Sedov blast wave solution. We will follow most closely two
papers of Taylor from 1950. 1 2 Taylor notes that the first of these was actually written in
1941, but was classified. One may also consult other articles by Taylor for background. 3 4
Consider a point source of energy which at t = 0 is released into a calorically perfect ideal
gas. The point source could be the combustion products of an intense reaction event. We
shall follow Taylors analysis and obtain what is known as self-similar solutions. Though
there are more general approaches which may in fact expose more details of how self-similar
solutions are obtained, we will confine ourselves to Taylors approach and use his notation.
The self-similar solution will be enabled by studying the Euler equations in what is known
as the strong shock limit for a spherical shock wave. Now a shock wave will raise both the
internal and kinetic energy of the ambient fluid into which it is propagating. We would
like to consider a scenario in which the total energy, kinetic and internal, enclosed by the
strong spherical shock wave is a constant. The ambient fluid is initially at rest, and a point
source of energy exists at r = 0. For t > 0, this point source of energy is distributed to the
mechanical and thermal energy of the surrounding fluid.
Let us follow now Taylors analysis from his 1950 Part I Theoretical Discussion paper.
We shall
write the governing inert one-dimensional unsteady Euler equations in spherical coor-
dinates,
1
Taylor, G. I., 1950, The Formation of a Blast Wave by a Very Intense Explosion. I. Theoretical
Discussion, Proceedings of the Royal Society of London. Series A, Mathematical and Physical Sciences
Vol. 201, No. 1065, pp. 159-174.
2
Taylor, G. I., 1950, The Formation of a Blast Wave by a Very Intense Explosion. II. The Atomic
Explosion of 1945, Proceedings of the Royal Society of London. Series A, Mathematical and Physical
Sciences, 201(1065): 175-186.
3
Taylor, G. I., 1950, The Dynamics of the Combustion Products Behind Plane and Spherical Detonation
Fronts in Explosives, Proceedings of the Royal Society of London. Series A, Mathematical and Physical
Sciences, 200(1061): 235-247.
4
Taylor, G. I., 1946, The Air Wave Surrounding an Expanding Sphere, Proceedings of the Royal Society
of London. Series A, Mathematical and Physical Sciences, 186(1006): 273-292.
375
376 CHAPTER 10. BLAST WAVES
reduce the partial differential equations in r and t to ordinary differential equations in

an appropriate similarity variable,
solve the ordinary differential equations numerically, and
show our transformation guarantees constant total energy in the region r [0, R(t)],
where R(t) is the locus of the moving shock wave.
We shall also refer to specific equations in Taylors first 1950 paper.
10.1 Governing equations

The non-conservative formulation of the governing equations is as follows:
u 2u
+u + = , (10.1)
t r r r
u u 1 P
+u + = 0, (10.2)
t r r

e e P
+u 2 +u = 0, (10.3)
t r t r
1 P
e = , (10.4)
1
P = RT. (10.5)
For review, lets look at the energy equation in a little more detail. Recall the material
derivative is d/dt = /t + u/r, so the energy equation is
de P d
= 0. (10.6)
dt 2 dt
Let us now substitute the thermal energy equation, Eq. (10.5) into the energy equation,
Eq. (10.6):

1 d P P d
2 = 0, (10.7)
1 dt dt
1 P d 1 1 dP P d
2
+ 2 = 0, (10.8)
1 dt 1 dt dt
P d 1 dP P d
2 + ( 1) 2 = 0, (10.9)
dt dt dt
1 dP P d
2 = 0, (10.10)
dt dt
dP P d
= 0, (10.11)
dt dt
10.2. SIMILARITY TRANSFORMATION 377
1 dP P d

+1 = 0, (10.12)
dt dt

d P
= 0, (10.13)
dt

P P

+u = 0. (10.14)
t r
10.2 Similarity transformation

We shall next make some non-intuitive and non-obvious choices for a transformed coordinate
system and transformed dependent variables. These choices can be systematically studied
with the techniques of group theory, not discussed here.
10.2.1 Independent variables

So, let us transform the independent variables (r, t) (, ) with
r
= , (10.15)
R(t)
= t. (10.16)
We will seek solutions such that the dependent variables are functions of , the distance
relative to the time-dependent shock, only. We will have little need for the transformed time
since it is equivalent to the original time t.
10.2.2 Dependent variables

Let us also define new dependent variables as
P
= y = R3 f1 (), (10.17)
Po

= (), (10.18)
o
u = R3/2 1 (). (10.19)
These amount to definitions of a scaled pressure, f1 , a scaled density and a scaled velocity
1 , with the assumption that each is a function of only. Here Po , and o are constant
ambient values of pressure and density, respectively.
We also assume the shock velocity to be of the form
dR
U= = AR3/2 . (10.20)
dt
The constant A is to be determined.
10.2.3 Derivative transformations

Now by the chain rule we have

= + (10.21)
t t t
Now by Eq. (10.15) we get
r dR
= 2 , (10.22)
t R dt
dR
= , (10.23)
R(t) dt

= AR3/2 , (10.24)
R
A
= 5/2 . (10.25)
R
From Eq. (10.16) we simply get

= 1. (10.26)
t
Thus the chain rule, Eq. (10.21), can be written as
A
= 5/2 + . (10.27)
t R
As we are insisting the / = 0, we get
A d
= 5/2 , (10.28)
t R d
In the same way, we get

= + , (10.29)
r r r |{z}

=0
1 d
= . (10.30)
R d
10.3 Transformed equations

Let us now apply our rules for derivative transformation Eqs. (10.25,10.30), and our trans-
formed dependent variables, Eqs. (10.17-10.19) to the governing equations.
10.3. TRANSFORMED EQUATIONS 379
10.3.1 Mass
First, we shall consider the mass equation, Eq. (10.1). We get
A d 1 d 1 d 2
5/2 (o ) + R3/2 1 (o ) + o R3/2 1 = o R3/2 1 . (10.31)
R d | {z } | {z } R d | {z } |{z} R d | {z } r |{z} | {z }
| {z } = =u | {z } = = | {z } =u = =u
=/t =/r =/r
Realizing the R(t) = R( ) is not a function of , canceling the common factor of o , and
eliminating r with Eq. (10.15), we can write
A d 1 d d1 2 1
5/2
+ 5/2 + 5/2 = , (10.32)
R d R d R d R5/2
d d d1 2
A + 1 + = 1 , (10.33)
d d d

d d d1 2
A + 1 + + 1 = 0. mass (10.34)
d d d
Equation (10.34) is number 9 in Taylors paper, which we will call here Eq. T(9).
10.3.2 Linear momentum

Now consider the linear momentum equation, Eq. (10.2), and apply the same transforma-
tions:
1
R3/2 1 + R3/2 1 R3/2 1 + Po R3 f1 = 0, (10.35)
t | {z } | {z } r | {z } o r | {z }
=u =u =u |{z} =P
=1/
1 3 5/2 dR 1
R3/2 R 1 +R3/2 1 R3/2 1 + Po R3 f1 = 0, (10.36)
| t {z2 dt } r o r
=u/t

3/2 A
d1 3 5/2
R 5/2 R AR3/2 1 + R3/2 1 R3/2 1
R
d 2 r
1
+ Po R3 f1 = 0, (10.37)
o r
A d1 3 A 1
4 4
1 + R3/2 1 R3/2 1 + Po R3 f1 = 0, (10.38)
R d 2R r o r
A d1 3 A 3/2 1 d 3/2
1 1 d 3

4 1 + R 1 R 1 + P o R f1 = 0, (10.39)
R d 2 R4 R d o R d
A d1 3 A 1 d1 Po 1 df1
4 4
1 + 4 + = 0, (10.40)
R d 2R R d o R4 d
d1 3 d1 Po df1
A A1 + 1 + = 0. (10.41)
d 2 d o d
Our final form is

3 d1 d1 Po 1 df1
A 1 + + 1 + = 0. linear momentum (10.42)
2 d d o d
Equation (10.42) is T(7).
10.3.3 Energy
Let us now consider the energy equation. It is best to begin with a form in which the
equation of state has already been imposed. So we will start by expanding Eq. (10.11) in
terms of partial derivatives:

P P P
+u +u = 0, (10.43)
|t {z r} t
| {z
r
}
=dP/dt =d/dt

Po R3 f1 + R3/2 1 Po R3 f1
t r
Po R3 f1 3/2
(o ) + R 1 (o ) = 0, (10.44)
o t r

R3 f1 + R3/2 1 R3 f1
t r
3
R f1 3/2
+R 1 = 0, (10.45)
t r
f1 dR
R3 3R4 f1 + R3/2 1 R3 f1
t dt r
R3 f1 3/2
+R 1 = 0, (10.46)
t r

A df1
R3 5/2 3R4 (AR3/2 )f1 + R3/2 1 R3 f1
R d r
3

R f1 3/2
+R 1 = 0. (10.47)
t r
Carrying on, we have
A df1 A 1 df1
11/2
3 11/2 f1 + R3/2 1 R3
R d R R d
3

R f1 A d 3/2 1 d
5/2 +R 1 = 0, (10.48)
R d R d

A df1 A 1 df1 f1 d d
11/2 3 11/2 f1 + 11/2 A + 1 = 0, (10.49)
R d R R d R11/2 d d

df1 df1 f1 d d
A 3Af1 + 1 A + 1 = 0. (10.50)
d d d d
10.4. DIMENSIONLESS EQUATIONS 381
Our final form is

df1 f1 d df1
A 3f1 + + (A + 1 ) 1 = 0. energy (10.51)
d d d
Equation (10.51) is T(11), correcting for a typographical error replacing a r with .
10.4 Dimensionless equations

Let us now write our conservation principles in dimensionless form. We take the constant
ambient sound speed co to be defined for the calorically perfect ideal gas as
Po
c2o . (10.52)
o
Note, we have used our notation for sound speed here; Taylor uses a instead.
Let us also define
c 2
o
f f1 , (10.53)
A
1
. (10.54)
A
10.4.1 Mass
With these definitions, the mass equation, Eq. (10.34) becomes

d d d 2
A + A + A + A = 0, (10.55)
d d d

d d d 2
+ + + = 0, (10.56)
d d d

d d 2
( ) = + , (10.57)
d d
d
1 d d
+ 2

= . mass (10.58)
d
Equation (10.58) is T(9a).
10.4.2 Linear momentum

With the same definitions, the momentum equation, Eq. (10.42) becomes

3 d d Po 1 A2 df
A A + A + A2 + = 0, (10.59)
2 d d o c2o d

3 d d 1 1 df
+ + + = 0, (10.60)
2 d d d
d 3 1 df
( ) + = 0, (10.61)
d 2 d
d 1 df 3
( ) = . momentum(10.62)
d d 2
10.4.3 Energy
The energy equation, Eq. (10.51) becomes
2
A A2 df f A2 d A2 df
A 3 2 f + 2 + (A + A) A = 0, (10.63)
co co d c2o d c2o d
df f d df
3f + + ( + ) = 0, (10.64)
d d d
df 1 d df
3f + + f ( + ) = 0. energy(10.65)
d d d
10.5 Reduction to non-autonomous form

Let us eliminate d/d and d/d from Eq. (10.65) with use of Eqs. (10.58,10.62).
d 2
!
df d
+ df
3f + + f ( + ) = 0, (10.66)
d d
1 df 3
d 2
df
2
+
3f + + f
( + ) df = 0, (10.67)
d d
1 df 3
!
df d
2
2
3f + ( ) f + = 0, (10.68)
d

2 df 1 df 3 2
3f ( ) + ( ) f + ( ) = 0, (10.69)
d d 2

2 f df 3 2
( ) f 3( ) + ( ) = 0, (10.70)
d 2

2 f df 3 22
( ) + f 3 3 + 2 + = 0, (10.71)
d 2

2 f df 1 22
( ) + f 3 3 + + = 0. (10.72)
d 2
10.5. REDUCTION TO NON-AUTONOMOUS FORM 383
Rearraning, we get

2 f df 1 22
( ) = f 3 + 3 + . (10.73)
d 2
(10.74)
Equation (10.73) is T(14).
We can thus write an explicit non-autonomous ordinary differential equation for the
evolution of f in terms of the state variables f , , and , as well as the independent variable
.
22
1
df f 3 + 3 + 2

= (10.75)
d ( )2 f
Eq. (10.75) can be directly substituted into the momentum equation, Eq. (10.62) to get
1 df
d d
32
= . (10.76)
d
Then Eq. (10.76) can be substituted into Eq. (10.58) to get
d
d d
+ 2

= . (10.77)
d
Equations (10.75-10.77) form a non-autonomous system of first order differential equa-
tions of the form
df
= g1 (f, , , ), (10.78)
d
d
= g2 (f, , , ), (10.79)
d
d
= g3 (f, , , ). (10.80)
d
They can be integrated with standard numerical software. One must of course provide
conditions of all state variables at a particular point. We apply conditions not at = 0, but
at = 1, the locus of the shock front. Following Taylor, the conditions are taken from the
Rankine-Hugoniot equations applied in the limit of a strong shock. We take the subscript s
to denote the shock state at = 1. For the density, one applies Eq. (9.287) and finds
s +1
= , (10.81)
o 1
o s +1
= , (10.82)
o 1
+1
s = ( = 1) = . (10.83)
1
For the pressure, one finds, by slight modification of Eq. (9.293) (taking D 2 = (dR/dt)2 ),
that
dR 2
dt + 1 Ps
= , (10.84)
c2o 2 Po
2 3
AR + 1 3
= R f1s , (10.85)
c2o 2
A2 R3 + 1 3 A2
= R fs , (10.86)
c2o 2 c2o
+1
1 = fs , (10.87)
2
2
fs = f ( = 1) = . (10.88)
+1
For the velocity, using Eq. (9.300), one finds
us 2
dR
= , (10.89)
dt
+1
R3/2 1s 2
3/2
= , (10.90)
AR +1
R3/2 As 2
3/2
= , (10.91)
AR +1
2
s = ( = 1) = . (10.92)
+1
Equations (10.83, 10.88,10.92) form the appropriate set of initial conditions for the integra-
tion of Eqs. (10.75-10.77).
10.6 Numerical solution

Solutions for f (), () and () are shown for = 7/5 in Figs. 10.1-10.3, respectively.
So we now have a similarity solution for the scaled variables. We need to relate this to
physical dimensional quantitites. Let us assign some initial conditions for t = 0, r > 0; that
is, away from the point source. Take
u(r, 0) = 0, (r, 0) = o , P (r, 0) = Po . (10.93)
We also have from the ideal gas law
Po
T (r, 0) = = To . (10.94)
o R
For the calorically perfect gas we further have
1 Po
e(r, 0) = = eo . (10.95)
1 o
10.6. NUMERICAL SOLUTION 385
f
1.2
1.0
0.8
0.6
0.4
0.2

0.0 0.2 0.4 0.6 0.8 1.0
Figure 10.1: Scaled pressure f versus similarity variable for = 7/5 in Taylor-Sedov blast
wave.

1.0
0.8
0.6
0.4
0.2

0.0 0.2 0.4 0.6 0.8 1.0
Figure 10.2: Scaled velocity versus similarity variable for = 7/5 in Taylor-Sedov blast
wave.

0.0 0.2 0.4 0.6 0.8 1.0
Figure 10.3: Scaled density versus similarity variable for = 7/5 in Taylor-Sedov blast
wave.
10.6.1 Calculation of total energy

Now as the point source expands, it will generate a strong shock wave. Material which has
not been shocked is oblivious to the presence of the shock. Material which the shock wave
has reached has been influenced by it. It stands to reason from energy conservation principles
that we want the total energy, internal plus kinetic, to be constant in the shocked domain,
r (0, R(t)], where R(t) is the shock front location.
Let us recall some spherical geometry so this energy conservation principle can be properly
formulated. Consider a thin differential spherical shell of thickness dr located somewhere in
the shocked region: r (0, R(t)]. The volume of the thin shell is
dV = 4r 2
|{z} dr
|{z} (10.96)
(surface area) (thickness)
The differential mass dm of this shell is
dm = dV, (10.97)
= 4r 2 dr. (10.98)
Now recall the mass-specific internal energy is e and the mass-specific kinetic energy is u2 /2.
So the total differential energy, internal plus kinetic, in the differential shell is

1 2
dE = e + u dm, (10.99)
2

1 2 2
= 4 e + u r dr. (10.100)
2
Now, the total energy E within the shock is the integral through the entire sphere,
Z R(t) Z R(t)
1 2 2
E= dE = 4 e + u r dr, (10.101)
0 0 2
Z R(t)
1 P 1 2 2
= 4 + u r dr, (10.102)
0 1 2
Z R(t) Z R(t)
4 2
= P r dr + 2 u2 r 2 dr . (10.103)
1 0 0
| {z } | {z }
thermal energy kinetic energy
We introduce variables from our similarity transformations next:
Z 1 Z 1
4 2 2
E = 3
Po R f1 R Rd +2 o R3 21 R2 2 Rd , (10.104)
1 0 | {z } | {z } |{z} 0
|{z} | {z } | {z } |{z}
P r2 dr u2 r2 dr
Z 1 Z 1
4
= Po f1 2 d + 2 o 21 2 d, (10.105)
1 0 0
10.6. NUMERICAL SOLUTION 387
Z 1 Z 1
4 Po A2 2
= 2
f d + 2 o A2 2 2 d, (10.106)
1 c
0 o
Z 1
0
Z
2 P o 2 o 1 2 2
= 4A f d + d , (10.107)
c2o ( 1) 0 2 0
Z 1 Z
2 1 2 1 1 2 2
= 4A o f d + d , (10.108)
( 1) 0 2 0
| {z }
dependent on only
(10.109)
The term inside the parentheses is dependent on only. So, if we consider air with = 7/5,
we can, using our knowledge of f (), (), and (), which only depend on , to calculate
once and for all the value of the integrals. For = 7/5, we obtain via numerical quadrature

2 1 1
E = 4A o (0.185194) + (.185168) , (10.110)
(7/5)(2/5) 2
= 5.3192o A2 . (10.111)
Now from Eq. (10.17, 10.52,10.53, 10.111) with = 7/5, we get
3 A2
P = Po R f 2 , (10.112)
co
3 o 2
= Po R f A , (10.113)
Po
1
= R3 f o A2 , (10.114)

1 E
= R3 f 7 , (10.115)
5
5.3192
= 0.1343R3Ef, (10.116)

E r
P (r, t) = 0.1343 3 f . (10.117)
R (t) R(t)
The peak pressure occurs at = 1, where r = R, and where
2 2(1.4)
f ( = 1) = = = 1.167. (10.118)
+1 1.4 + 1
So at = 1, where r = R, we have
E
P = (0.1343)(1.167)R3E = 0.1567 . (10.119)
R3
The peak pressure decays at a rate proportional to 1/R3 in the strong shock limit.
Now from Eqs. (10.19,10.54,10.111) we get for u:
u = R3/2 A, (10.120)
s
E
= R3/2 , (10.121)
5.319o
s
E 1 r
u(r, t) = . (10.122)
5.319o R3/2 (t) R(t)
Let us now explicitly solve for the shock position R(t) and the shock velocity dR/dt. We
have from Eqs. (10.20, 10.111) that
dR
= AR3/2 , (10.123)
dt s
E 1
= 3/2
, (10.124)
5.319o R (t)
s
E
R3/2 dR = dt, (10.125)
5.319o
s
2 5/2 E
R = t + C. (10.126)
5 5.319o
Now since R(0) = 0, we get C = 0, so

s
2 5/2 E
R = t, (10.127)
5 5.319o
2 5/2 p
t = R 5.319o E 1/2 , (10.128)
5
= 0.9225R5/2 1/2
o E
1/2
. (10.129)
Equation (10.129) is T(38). Solving for R, we get
1
R5/2 = t1/2 E 1/2 , (10.130)
0.9225 o
R(t) = 1.032791/5
o E 1/5 t2/5 . (10.131)
Thus, we have a prediction for the shock location as a function of time t, as well as point
source energy E. If we know the position as a function of time, we can easily get the shock
velocity be direct differentiation:
dR
= 0.41311/5
o E 1/5 t3/5 . (10.132)
dt
10.7. CONTRAST WITH ACOUSTIC LIMIT 389
If we can make a measurement of the blast wave location R at a given known time t, and
we know the ambient density o , we can estimate the point source energy E. Let us invert
Eq. (10.131) to solve for E and get
o R5
E = , (10.133)
(1.03279)5t2
o R5
= 0.85102 2 . (10.134)
t
10.6.2 Comparison with experimental data

Now Taylors Part II paper from 1950 gives data for the 19 July 1945 atomic explosion at
the Trinity site in New Mexico. We choose one point from the photographic record which
finds the shock from the blast to be located at R = 185 m when t = 62 ms. Let us assume
the ambient air has a density of o = 1.161 kg/m3 . Then we can estimate the energy of the
device by Eq. (10.134) as

kg
1.161 m 2 (185 m)5
E = 0.85102 , (10.135)
(0.062 s)2
= 55.7 1012 J. (10.136)
Now 1 ton of the high explosive TNT is know to contain 4.25 109 J of chemical energy. So
the estimated energy of the Trinity site device in terms of a TNT equivalent is
55.7 1012 J
T NTequivalent = 9 J
= 13.1 103 ton. (10.137)
4.25 10 ton
In common parlance, the Trinity site device was a 13 kiloton bomb by this simple estimate.
Taylor provides some nuanced corrections to this estimate. Modern estimates are now around
20 kiloton.
10.7 Contrast with acoustic limit

We saw in Eq. (10.119) that in the expansion associated with a strong shock, the pressure
decays as 1/R3 . Let us see how that compares with the decay of pressure in the limit of a
weak shock.
Let us first rewrite the governing equations. Here we 1) rewrite Eq. (10.1) in a conserva-
tive form, using the chain rule to absorb the source term inside the derivative, 2) repeat the
linear momentum equation, Eq. (10.2), and 3) re-cast the energy equation for a calorically
perfect ideal gas, Eq. (10.11) in terms of the full partial derivatives:
1 2
+ 2 r u = 0, (10.138)
t r r
u u 1 P
+u + = 0, (10.139)
t r r

P P P
+u +u = 0. (10.140)
t r t r
Now let us consider the acoustic limit, which corresponds to perturbations of a fluid at
rest. Taking 0 < << 1, we recast the dependent variables , P , and u as
= o + 1 + . . . , (10.141)
P = Po + P1 + . . . , (10.142)
u = uo +u1 + . . . . (10.143)
|{z}
=0
Here o and Po are taken to be constants. The ambient velocity uo = 0. Strictly speaking,
we should non-dimensionalize the equations before we introduce an asymptotic expansion.
However, so doing would not change the essence of the argument to be made.
We next introduce our expansions into the governing equations:
1 2
(o + 1 ) + 2 r (o + 1 ) (u1 ) = 0, (10.144)
t r r
1
(u1 ) + (u1 ) (u1 ) + (Po + P1 ) = 0, (10.145)
t r o + 1 r

(Po + P1 ) + (u1 ) (Po + P1 )
t r
Po + P1
(o + 1 ) + (u1 ) (o + 1 ) = 0. (10.146)
o + 1 t r
Now derivatives of constants are all zero, and so at leading order the constant state satisfies
the governing equations. At O(), the equations reduce to
1 1
+ 2 (r 2 o u1 ) = 0, (10.147)
t r r
u1 1 P1
+ = 0, (10.148)
t o r
P1 Po 1
= 0. (10.149)
t o t
Now, adopt as before c2o = Po /o , so the energy equation, Eq. (10.149), becomes
P1 1
= c2o . (10.150)
t t
Now substitute Eq. (10.150) into the mass equation, Eq. (10.147), to get
1 P1 1
2
+ 2 (r 2 o u1 ) = 0. (10.151)
co t r r
10.7. CONTRAST WITH ACOUSTIC LIMIT 391
We take the time derivative of Eq. (10.151) to get

1 2 P1 1 2
+ r o u1 = 0, (10.152)
c2o t2 t r 2 r

1 2 P1 1 2 u1
+ 2 r o = 0. (10.153)
c2o t2 r r t
We next use the momentum equation, Eq. (10.148), to eliminate u1 /t in Eq. (10.153):

1 2 P1 1 2 1 P1
+ 2 r o = 0, (10.154)
c2o t2 r r o r

1 2 P1 1 2 P1
2 r = 0, (10.155)
c2o t2 r r r

1 2 P1 1 2 P1
= 2 r . (10.156)
c2o t2 r r r
This second-order linear partial differential equation has a well-known solution of the
DAlembert form:

1 r 1 r
P1 = g t + h t+ . (10.157)
r co r co
Here g and h are arbitrary functions which are chosen to match the initial conditions. Let
us check this solution for g; the procedure can easily be repeated for h.
If P1 = (1/r)g(t r/co), then

P1 1 r
= g t , (10.158)
t r co

2 P1 1 r
= g t , (10.159)
t2 r co
and

P1 11 r 1 r
= g t 2g t . (10.160)
r co r co r co
With these results, let us substitute into Eq. (10.156) to see if it is satisfied:

1 1 r 1 2 11 r 1 r
g t = 2 r g t 2g t , (10.161)
c2o r co r r co r co r co

1 r r r
= 2 g t +g t , (10.162)
r r co co co

1 r r 1 r 1 r
= 2 2g t + g t g t (10.163)
,
r co co co co co co

1 r r
= 2 g t , (10.164)
r c2o co

1 1 r
= 2 g t . (10.165)
co r co
Indeed, our form of P1 (r, t) satisfies the governing partial differential equation. Moreover we
can see by inspection of Eq. (10.157) that the pressure decays as does 1/r in the limit of
acoustic disturbances. This is a much slower rate of decay than for the blast wave, which
goes as the inverse cube of radius.
Bibliography
Here we give a summary of some of the major books relevant to these lecture notes. Specific
journal articles are not cited, though the main text of the lecture notes occasionally draws
on such articles where relevant. The emphasis is on combustion, physical chemistry, and
thermodynamics, with some mathematics. Some general works of historic importance are
also included. The list is by no means comprehensive.
M. M. Abbott and H. C. van Ness, 1972, Thermodynamics, Schaums Outline Series in
Engineering, McGraw-Hill, New York.
This is written in the style of all the Schaums series. It has extensive solved problems and a crisp
rigorous style that is readable by undergraduate engineers. It has a chemical engineering emphasis,
but is also useful for all engineers.
R. Aris, 1962, Vectors, Tensors, and the Basic Equations of Fluid Mechanics, Dover, New
York.
This book is a jewel of continuum mechanics theory. It is brief and incisive. While the focus is on
the inert Navier-Stokes equations, there is a short, excellent chapter on the reactive Navier-Stokes
equaitons. Highly recommended.
J. Bebernes and D. Eberly, 1989, Mathematical Problems from Combustion Theory, Springer,
Berlin.
This monograph focuses on mathematical analysis of combustion systems, focusing on blowup phe-
nomena.
A. Bejan, 2006, Advanced Engineering Thermodynamics, Third Edition, John Wiley, Hobo-
ken, New Jersey.
This is an advanced undergraduate text. It gives a modern treatment of the science of classical
thermodynamics. It does not confine its attention to traditional engineering problems, and considers
applications across biology and earth sciences as well. The thermodynamics of irreversible processes
are discussed in detail.
Berry, R. S., Rice, S. A., and Ross, J., 2000, Physical Chemistry, Second Edition, Oxford
University Press, Oxford.
This is a rigorous general text in physical chemistry at a senior or first year graduate level. It has a
full treatment of classical and statistical thermodynamics as well as quantum mechanics.
393
R. B. Bird, W. E. Stewart, and E. N. Lightfoot, 2006, Transport Phenomena, Second
Edition, Wiley, New York.
This acknowledged classic of the chemical engineering literature does not focus on reaction, but has
an exemplary treatment of mass, momentum, and energy advection-diffusion phenomena.
L. Boltzmann, 1995, Lectures on Gas Theory, Dover, New York.

This is a detailed monograph by the founding father of statistical thermodynamics.
R. Boyle, 2003, The Sceptical Chymist, Dover, New York.

This is the original work of the famous early figure of the scientific revolution.
J. D. Buckmaster and G. S. S. Ludford, 1983, Lectures on Mathematical Combustion, SIAM,

Philadelphia.
This is set of lecture notes which serves as an excellent graduate-level introduction to some of the
mathematical challenges of combustion.
J. D. Buckmaster and G. S. S. Ludford, 2008, Theory of Laminar Flames, Cambridge,

Cambridge.
This seminal monograph, first published in 1982, highlights rigorous mathematical methods applied
to laminar flames.
H. B. Callen, 1985, Thermodynamics and an Introduction to Thermostatistics, Second Edi-

tion, John Wiley, New York.
This advanced undergraduate text has an emphasis on classical physics applied to thermodynamics,
with a few chapters devoted to quantum and statistical foundations.
S. Carnot, 2005, Reflections on the Motive Power of Fire, Dover, New York.
This is the original source of Carnots foundational work on the heat engines. Also included is a paper
of Clausius.
N. Chigier, 1981, Energy, Combustion, and Environment, McGraw-Hill, New York.

This advanced undergraduate text gives an integrated science-based discussion of the topic matter
stated in its title.
R. Courant and K. O. Friedrichs, 1999, Supersonic Flow and Shock Waves, Springer, New
York.
This seminal mathematical treatment of inviscid compressible flow gives an outstanding and rigorous
discussion of the application of partial differential equations to fluid mechanics. There is only a small
discussion of combustion, but the methods are of high relevance in high speed reactive flow.
394
S. R. de Groot and P. Mazur, 1984, Non-Equilibrium Thermodynamics, Dover, New York.
This is an influential monograph which summarizes much of the work of the famous Belgian school
of thermodynamics. It is written at a graduate level and has a strong link to fluid mechanics and
chemical reactions.
E. Fermi, 1936, Thermodynamics, Dover, New York.

This short classic clearly and efficiently summarizes the fundamentals of thermodynamics. It is based
on a series of lectures given by this Nobel prize winner.
W. Fickett and W. C. Davis, 1979, Detonaton, U. California, Berkeley.

This distinctive monograph has a rich and focused discussion on detonation science. It opens with
the relevant physical chemistry, then gives detailed exposition on the theory of detonation. Careful
discussion of experiment is also included.
I. Glassman and R. Yetter, 2008, Combustion, Fourth Edition, Academic Press, New York.
This well known monograph give a good description of important combustion phenomena.
J. F. Griffiths and J. A. Barnard, 1995, Flame and Combustion, Third Edition, Chapman
and Hall, Glasgow.
This is a graduate level text in combustion.
E. P. Gyftopoulos and G. P. Beretta, 1991, Thermodynamics: Foundations and Applica-

tions, Macmillan, New York.
This beginning graduate text has a rigorous development of classical thermodynamics.
J. O. Hirschfelder, C. F. Curtis, and R. B. Bird, 1954, Molecular Theory of Gases and

Liquids, Wiley, New York.
This comprehensive tome is a valuable addition to any library of thermal science. Its wide ranging
text covers equations of state, molecular collision theory, reactive hydrodynamics, reaction kinetics,
and many other topics all from the point of view of careful physical chemistry. Much of the work
remains original.
W. Hundsdorfer and J. G. Verwer, 2003, Numerical Solution of Time-Dependent Advection-

Diffusion-Reaction Equations, Springer, Berlin.
This nice graduate level monograph focuses on modern numerical techniques for paradigm reactive
flow systems. The systems themselves are easy to pose which allows effective exposition of the key
challenges in numerical modeling.
A. M. Kanury, 1975, Introduction to Combustion Phenomena, Gordon and Breach, Ams-

terdam.
This introductory graduate level text gives practical problems and good discussion of combustion
phenomena.
395
A. K. Kapila, 1983, Asymptotic Treatment of Chemically Reacting Systems, Pitman, Boston.
This short introductory monograph gives lecture notes on asymptotic analysis as applied to combus-
tion.
E. L. Keating, 2007, Applied Combustion, Second Edition, CRC Press, Boca Raton, FL.
This book is an engineering text for seniors or first year graduate students with a general coverage of
topics.
J. Kestin, 1966, A Course in Thermodynamics, Blaisdell, Waltham, Massachusetts.

This is a foundational textbook that can be read on many levels. All first principles are reported in a
readable fashion. In addition the author makes a great effort to expose the underlying mathematical
foundations of thermodynamics.
R. J. Key, M. E. Coltrin, and P. Glarborg, 2003, Chemically Reacting Flow: Theory and
Practice, John Wiley, New York.
This comprehensive text gives an introductory graduate level discussion of fluid mechanics, thermo-
chemistry, and finite rate chemical kinetics. The focus is on low Mach number flows, and there is
significant discussion of how to achieve computational solutions.
C. Kittel and H. Kroemer, 1980, Thermal Physics, Second Edition, Freeman, San Francisco.
This is a classic undergraduate text for physics students. It has a good introduction to statistical
thermodynamics and a short effective chapter on classical thermodynamics.
D. Kondepudi and I. Prigogene, 1998, Modern Thermodynamics: From Heat Engines to

Dissipative Structures, John Wiley, New York.
Detailed modern exposition which exploits the authors unique vision of thermodynamics with both
a science and engineering flavor.
V. P. Korobeinikov, 1989, Unsteady Interaction of Shock and Detonation Waves in Gases,

Taylor and Francis, New York.
This is a graduate level monograph.
K. K. Kuo, 2005, Principles of Combustion, Second Edition, John Wiley, New York.
This is a readable graduate level engineering text for combustion fundamentals. It includes a full
treatment of reacting thermodynamics as well as discussion of links to fluid mechanics.
K. J. Laidler, 1965, Chemical Kinetics, McGraw-Hill, New York.

This is a standard advanced undergraduate chemistry text on the dynamics of chemical reactions.
L. D. Landau and E. M. Lifshitz, 1987, Fluid Mechanics, Second Edition, Butterworh-

Heinemann, Oxford.
This classic comprehensive text of fluid physics contains a small discussion of reactive flow.
396
B. H. Lavenda, 1978, Thermodynamics of Irreversible Processes, John Wiley, New York.
This is a lively and opinionated monograph describing and commenting on irreversible thermodynam-
ics. The author is especially critical of the Prigogene school of thought on entropy production rate
minimization.
A. Lavoisier, 1984, Elements of Chemistry, Dover, New York.

This is the classic treatise which gives the first explicit statement of mass conservation in chemical
reactions.
H. W. Liepmann and A. Roshko, 1957, Elements of Gasdynamics, John Wiley, New York.
This is an influential text in compressible aerodynamics that is appropriate for seniors or beginning
graduate students. It has a strong treatment of the physics and thermodynamics of compressible flow
along with elegant and efficient text. Its treatment of both experiment and the underlying theory is
outstanding.
C. K. Law, 2006, Combustion Physics, Cambridge, Cambridge.

This modern text is a comprehensive graduate level discussion of combustion theory.
J. H. S. Lee, 2008, The Detonation Phenomenon, Cambridge, Cambridge.

This is a graduate level monograph focusing on detonation.
J. C. Maxwell, 2001, Theory of Heat, Dover, New York.

This is a short readable book by the 19th century master. Here the mathematics is minimized in favor
of more words of explanation.
B. Lewis and G. von Elbe, 1961, Combustion, Flames and Explosions of Gases, Academic
Press, New York.
For many years this was the graduate text in combustion. It has a fine description of scientific
combustion experiments and good discussion of the key mechanisms of combustion.
M. J. Moran and H. N. Shapiro, 2003, Fundamentals of Engineering Thermodynamics, Fifth

Edition, John Wiley, New York.
This is a standard undergraduate thermodynamics text, and one of the more popular. Note: some
examples in this set of notes have been adopted from the 1992 Second Edition of this text.
I. M
uller and T. Ruggeri, 1998, Rational Extended Thermodynamics, Springer-Verlag, New
York.
This modern monograph gives a rigorous treatment of some of the key issues in modern continuum
mechanics and classical thermodynamics.
I. M
uller and W. Weiss, 2005, Entropy and Energy, Springer-Verlag, New York.
This is a unique treatise on fundamental concepts in thermodynamics. The author provide mathe-
matical rigor, historical perspective, and examples from a diverse set of scientific fields.
397
I. M
uller, 2007, A History of Thermodynamics: The Doctrine of Energy and Entropy,
Springer-Verlag, New York.
The author gives a readable text at an undergraduate level which highlights some of the many con-
troversies of thermodynamics, both ancient and modern.
E. S. Oran and J. P. Boris, 2001, Numerical Simulation of Reactive Flow, Second Edition,
Cambridge, Cambridge.
This monograph by two of the leading figures of computational combustion gives a good discussion
of numerical challenges as well as physically relevant problems.
W. Pauli, 2000, Thermodynamics and the Kinetic Theory of Gases, Dover, New York.
This is a monograph on thermodynamics by a well known physicist.
L. Perko, 2001, Differential Equations and Dynamical Systems, Third Edition, Springer,
New York.
This monograph gives no explicit discussion of combustion but does give an outstanding treatment of
dynamic systems theory.
M. Planck, 1945, Treatise on Thermodynamics, Dover, New York.

This brief book gives many unique insights from a great physicist.
T. Poinsot and D. Veynante, 2005, Theoretical and Numerical Combustion, Edwards, Flour-
town, PA.
I. Prigogine, 1967, Introduction to Thermodynamics of Irreversible Processes, Third Edition,

Interscience, New York.
This is a famous book that summarizes the essence of the work of the Belgian school for which the
author was awarded the Nobel prize.
L. E. Reichl, 1998, A Modern Course in Statistical Physics, John Wiley, New York.
This full service graduate text has a good summary of key concepts of classical thermodynamics and
a strong development of modern statistical thermodynamics.
W. C. Reynolds, 1968, Thermodynamics, Second Edition, McGraw-Hill, New York.

This is an unusually good undergraduate text written for mechanical engineers. The author has
wonderful qualitative problems in addition to the usual topics in such texts. A good introduction to
statistical mechanics is included as well. Highly recommended.
D. E. Rosner, 2000, Transport Processes in Chemically Reacting Flow Systems, Dover, New
York.
This graduate level monograph gives a chemical engineering perspective on reaction engineering in
fluid systems.
398
S. I. Sandler, 1998, Chemical and Engineering Thermodynamics, Third Edition, John Wiley,
New York.
Advanced undergraduate text in thermodynamics from a chemical engineering perspective with a
good mathematical treatment.
E. Schrodinger, 1989, Statistical Thermodynamics, Dover, New York.

This is a short monograph written by the one of the pioneers of quantum physics.
A. H. Shapiro, 1953, The Dynamics and Thermodynamics of Compressible Fluid Flow,

Vols. I and II, John Wiley, New York.
This classic two volume set has a comprehensive treatment of the subject of its title. It has numerous
worked example problems, and is written from an engineers perspective.
R. E. Sonntag, C. Borgnakke, and G. J. von Wylen, 2003, Fundamentals of Thermodynam-

ics, Sixth Edition, John Wiley, New York.
This classic and popular undergraduate mechanical engineering text has stood the test of time and
has a full treatment of most classical problems.
J. M. Smith, H. C. van Ness, and M. Abbott, 2004, Introduction to Chemical Engineering

Thermodynamics, Seventh Edition, McGraw-Hill, New York.
This is probably the most common undergraduate text in thermodynamics in chemical engineering.
It is rigorous and has went through many revisions.
R. A. Strehlow, 1984, Combustion Fundamentals, McGraw-Hill, New York.

This graduate level text has a nice discussion of physical chemistry followed by good descriptions of
the authors seminal work in detonation experiments, among other standard combustion topics.
J. W. Tester and M. Modell, 1997, Thermodynamics and Its Applications, Third Edition,
Prentice Hall, Upper Saddle River, New Jersey.
This entry level graduate text in thermodynamics is written from a chemical engineers perspective.
It has a strong mathematical development for both classical and statistical thermodynamics.
K. Terao, 2007, Irreversible Phenomena: Ignitions, Combustion, and Detonation Waves,

Springer, Berlin.
This unique monograph links irreversible thermodynamics with phase transition, ignition and deto-
nation.
T.-Y. Toong, 1982, Combustion Dynamics: The Dynamics of Chemically Reacting Fluids,
McGraw-Hill, New York.
This is an entry level graduate text on combustion science.
399
C. A. Truesdell, 1980, The Tragicomic History of Thermodynamics, 1822-1854, Springer
Verlag, New York.
This idiosyncratic monograph has a lucid description of the history of nineteenth century thermal
science. It is written in an erudite fashion and the reader who is willing to dive into a difficult subject
will be rewarded for diligence by many new insights.
C. A. Truesdell, 1984, Rational Thermodynamics, Springer-Verlag, New York.

This is a modern update on the evolution of classical thermodynamics in the twentieth century. It is
at a graduate level and contains several articles by some of the present leaders of the field.
S. R. Turns, 2000, An Introduction to Combustion, McGraw-Hill, Boston.

This is a popular senior-level undergraduate text on combustion which uses many notions from ther-
modynamics of mixtures.
H. C. van Ness, 1983, Understanding Thermodynamics, Dover, New York.

This is a short readable monograph from a chemical engineering perspective.
W. G. Vincenti and C. H. Kruger, 1965, Introduction to Physical Gas Dynamics, John

Wiley, New York.
This graduate text on high speed non-equilibrium flows contains some nice descriptions of the interplay
of classical and statistical mechanics. There is an emphasis on aerospace applications.
J. Warnatz, U. Maas, and R. W. Dibble, 2006, Combustion, Fourth Edition, Springer, New
York.
This short graduate level text is nonetheless comprehensive and gives a good introduction to combus-
tion science.
F. A. Williams, 1985, Combustion Theory, Benjamin-Cummings, Menlo Park, California.

This influential monograph was first published in 1965. It was one of the first texts to give a compre-
hensive theoretical treatment of reactive fluid mechanics for model and practical systems.
L. C. Woods, 1975, The Thermodynamics of Fluid Systems, Clarendon, Oxford.

This graduate text gives a good, detailed survey of the thermodynamics of irreversible processes,
especially related to fluid systems in which convection and diffusion play important roles.
I. B. Zeldovich and A. S. Kompaneets, 1960, Theory of Detonation, Academic Press, New

York.
This is a readable monograph whose first author is one of the founding fathers of modern detonation
theory.
Ya. B. Zeldovich and Yu. P. Raizer, 2002, Physics of Shock Waves and High-Temperature
Hydrodynamic Phenomena, Dover, New York.
This classic tome of the Russian school is a tour de force of reactive flow physics.
400

Combustion Notes PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Combustion Notes PDF

Uploaded by

Copyright:

Available Formats

LECTURE NOTES ON

Department of Aerospace and Mechanical Engineering

1.2.1.5 High activation energy asymptotics . . . . . . . . . . . . . . 62

3 Mathematical foundations of thermodynamics* 93

4 Thermochemistry of a single reaction 139

4.3.1 Enthalpy of formation . . . . . . . . . . . . . . . . . . . . . . . . . . 150

5 Thermochemistry of multiple reactions 193

6 Reactive Navier-Stokes equations 237

6.1.2.3 Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240

7 Simple solid combustion: Reaction-diffusion 247

8 Simple laminar flames: Reaction-advection-diffusion 281

8.1.1.3 Formulation using enthalpy . . . . . . . . . . . . . . . . . . 282

9 Simple detonations: Reaction-advection 309

9.2.2.1 Conservative form . . . . . . . . . . . . . . . . . . . . . . . 327

10 Blast waves 375

10.7 Contrast with acoustic limit . . . . . . . . . . . . . . . . . . . . . . . . . . . 389

Notre Dame, Indiana; USA

Copyright c 2010 by Joseph M. Powers.

1.1 Isothermal, isochoric kinetics

1.1.1.1 Pair of irreversible reactions

In general, we will have be a column vector of dimension N 1, will be a rectangular

The symbol R is the universal gas constant, where

What we have done is the standard linear algebra decomposition of = P1 L U.

We see here this is true:

exp(- E j /(R T))

Figure 1.1: Plot of exp(E j /R/T ) versus T ; transition occurs at T E j /R.

for T E j /R, the reaction rate is extremely sensitive to temperature,

So the units of a13 are

1.1.1.1.2 Example calculation Let us consider an example problem. Let us take T =

Then the differential equation system becomes

1.1.1.1.2.1 Species concentration versus time A solution was obtained numeri-

Substituting from Eq. (1.59) into Eq. (1.60), we get

For our example, we only have two species, so

The pressure at the initial state t = 0 is

Figure 1.4: Pressure versus time for oxygen dissociation example.

Figure 1.5: Equilibria for oxygen dissociation example.

33948.3 (1.78439 1011 )(O )2 (5.8 1013 )(O )3 = f (O ) = 0. (1.80)

We find df /dO by differentiating Eq. (1.80) to get

The equilibrium concentration of O dropped by two orders of magnitude relative to

T = 1000 K. Here we reduce the temperature more. This temperature is common in

The O concentration at equilibrium is greatly diminished to the point of being difficult

T = 300 K. This is obviously near room temperature. We find

T = 10000 K. Such high temperature could be achieved in an atmospheric re-entry

1.1.1.2 Single reversible reaction

1.1.1.2.1 Mathematical model

For this system, we have N = 2 molecular species in L = 1 elements reacting in J = 1

For this reversible reaction, we slightly modify the kinetics equations to

In matrix form, we can simplify to

Here the N J or 2 1 matrix is

So here the N N or 2 2 matrix L1 is

Once again the stoichiometric matrix is

Thus, we once again find

As before, we can say

1.1.1.2.1.2 Thermodynamics Equations (1.102-1.103) are supplemented by an ex-

Go13 = 2g oO g oO2 . (1.120)

We rearrange Eq. (1.122) to find the familiar

Thus, by Eq. (1.120), we have

1.01326 106 dyne

Using Eq. (1.116) we find this corresponds to

1.1.2 Zeldovich mechanism of N O production

1.1.2.1 Mathematical model

We have L = 2 with N and O as the 2 elements. The stoichiometric matrix of dimension

And we have J = 2 reactions. The reaction vector of length J = 2 is