Turbulent Flows Stephen B. Pope Cornell University = CAMBRIDGE ‘8 UNIVERSITY PRESSContents List of tables Preface Nomenclature PART ONE: FUNDAMENTALS 1 Introduction 11 12 The nature of turbulent flows The study of turbulent flows 2 The equations of fluid motion 21 2.2 2.3 24 2.5 2.6 27 2.8 29 Continuum fluid properties Eulerian and Lagrangian fields The continuity equation The momentum equation The role of pressure Conserved passive scalars The vorticity equation Rates of strain and rotation Transformation properties 3 The statistical description of turbulent flows 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 The random nature of turbulence Characterization of random variables Examples of probability distributions Joint random variables Normal and joint-normal distributions Random processes Random fields Probability and averaging page xv xvii xxi wwe ~ 10 12 14 16 18 21 22 23 24 34 34 37 43 61 65 74 79vu Contents 4 Mean-flow equations 83 4.1 Reynolds equations 83 42 Reynolds stresses 86 4.3. The mean scalar equation 91 4.4 Gradient-diffusion and turbulent-viscosity hypotheses 92 5 Free shear flows 96 5.1 The round jet: experimental observations 96 5.1.1 A description of the flow 96 5.1.2. The mean velocity field 97 5.1.3. Reynolds stresses 105 5.2 The round jet: mean momentum 11 5.2.1. Boundary-layer equations 111 5.2.2 Flow rates of mass, momentum, and energy 115 5.2.3 Self-similarity 116 5.2.4 Uniform turbulent viscosity 118 5.3 The round jet: kinetic energy 122 5.4 Other self-similar flows 134 5.4.1 The plane jet 134 5.4.2 The plane mixing layer 139 5.4.3 The plane wake 147 5.4.4 The axisymmetric wake 151 5.4.5 Homogeneous shear flow 154 5.4.6 Grid turbulence 158 5.5 Further observations 161 5.5.1 A conserved scalar 161 5.5.2 Intermittency 167 5.5.3 PDFs and higher moments 173 5.5.4 Large-scale turbulent motion 178 6 — The scales of turbulent motion 182 61 The energy cascade and Kolmogorov hypotheses 182 6.1.1 The energy cascade 183 6.1.2 The Kolmogorov hypotheses 184 6.1.3 The energy spectrum 188 6.1.4 Restatement of the Kolmogorov hypotheses 189 6.2 Structure functions 191 6.3 Two-point correlation 195 6.4 Fourier modes 207 6.4.1 Fourier-series representation 207 6.4.2 The evolution of Fourier modes 211Contents 64.3 The kinetic energy of Fourier modes 6.5 Velocity spectra 6.5.1 Definitions and properties 6.5.2. Kolmogorov spectra 6.5.3 A model spectrum 6.5.4 Dissipation spectra 6.5.5 The inertial subrange 6.5.6 The energy-containing range 6.5.7 Effects of the Reynolds number 6.5.8 The shear-stress spectrum 6.6 The spectral view of the energy cascade 67 Limitations, shortcomings, and refinements 6.7.1 The Reynolds number 6.7.2 Higher-order statistics 6.7.3 Internal intermittency 6.7.4 Refined similarity hypotheses 6.7.5 Closing remarks Wall flows 7.1 Channel flow 7.1.1 A description of the flow 7.1.2 The balance of mean forces 7.1.3 The near-wall shear stress 7.1.4 Mean velocity profiles 7.1.5 The friction law and the Reynolds number 7.1.6 Reynolds stresses 7.1.7 Lengthscales and the mixing length 7.2 Pipe flow 7.2.1 The friction law for smooth pipes 7.2.2 Wall roughness 73 Boundary layers 74 7.3.1 A description of the flow 7.3.2. Mean-momentum equations 7.3.3. Mean velocity profiles 7.3.4 The overlap region reconsidered 7.3.5 Reynolds-stress balances 7.3.6 Additional effects Turbulent structures 215 219 220 229 232 234 238 240 242 246 249 254 254 255 258 260 263 264 264 264 266 268 271 278 281 288 290 290 295 298 299 300 302 308 313 320 322Contents 10 11 PART TWO: MODELLING AND SIMULATION An introduction to modelling and simulation 8.1 The challenge 8.2. An overview of approaches 8.3 Criteria for appraising models Direct numerical simulation 9.1 Homogeneous turbulence 9.1.1 Pseudo-spectral methods 9.1.2. The computational cost 9.1.3. Artificial modifications and incomplete resolution 9.2 Inhomogeneous flows 9.2.1. Channel flow 9.2.2 Free shear flows 9.2.3 Flow over a backward-facing step 9.3 Discussion Turbulent-viscosity models 10.1 The turbulent-viscosity hypothesis 10.1.1. The intrinsic assumption 10.1.2. The specific assumption 10.2 Algebraic models 10.2.1 Uniform turbulent viscosity 10.2.2 The mixing-length model 10.3 Turbulent-kinetic-energy models 10.4 The k-e model 10.4.1 An overview 10.4.2. The model equation for ¢ 10.4.3. Discussion 10.5 Further turbulent-viscosity models 10.5.1. The k-w model 10.5.2. The Spalart-Allmaras model Reynolds-stress and related models 11.1 Introduction 11.2. The pressure-rate-of-strain tensor 11.3 Return-to-isotropy models 11.3.1 Rotta’s model 11.3.2. The characterization of Reynolds-stress anisotropy 11.3.3. Nonlinear return-to-isotropy models 11.4 Rapid-distortion theory 11.4.1 Rapid-distortion equations 333 335 335 336 336 344 344 344 346 352 353 353 354 355 356 358 359 359 364 365 365 366 369 373 323 375 382 383 383 385 387 387 388 392 302 393 398 405Contents xi 12 11.5 11.6 11.7 118 11.9 11.4.2, The evolution of a Fourier mode 11.4.3 The evolution of the spectrum 11.4.4 Rapid distortion of initially isotropic turbulence 11.4.5 Final remarks Pressure-rate-of-strain models 11.5.1. The basic model (LRR-IP) 11.5.2. Other pressure-rate-of-strain models Extension to inhomogeneous flows 11.6.1 Redistribution 11.6.2. Reynolds-stress transport 11.6.3 The dissipation equation Near-wall treatments 11.7.1 Near-wall effects 11.7.2 Turbulent viscosity 11.7.3. Model equations for k and ¢ 11.7.4 The dissipation tensor 11.7.5 Fluctuating pressure 11.7.6 Wall functions Elliptic relaxation models Algebraic stress and nonlinear viscosity models 11.9.1 Algebraic stress models 11.9.2 Nonlinear turbulent viscosity 11.10 Discussion PDF methods 12.1 12.2 12.3 12.4 The Eulerian PDF of velocity 12.1.1. Definitions and properties 12.1.2 The PDF transport equation 12.1.3. The PDF of the fluctuating velocity The model velocity PDF equation 12.2.1. The generalized Langevin model 12.2.2. The evolution of the PDF 12.2.3 Corresponding Reynolds-stress models 12.2.4 Eulerian and Lagrangian modelling approaches 12.2.5 Relationships between Lagrangian and Fulerian PDFs Langevin equations 12.3.1 Stationary isotropic turbulence 12.3.2. The generalized Langevin model Turbulent dispersion 406 41 415 421 422 423 425 428 428 429 432 433 433 434 435 436 439 442 445 448 448 452 457 463 464 464 465 467 468 469 470 475 479 480 483 484 489 494Contents 12.5 The velocity—frequency joint PDF 506 12.5.1. Complete PDF closure 506 12.5.2. The log-normal model for the turbulence frequency 507 12.5.3 The gamma-distribution model 511 12.5.4 The model joint PDF equation 514 12.6 The Lagrangian particle method 516 12.6.1 Fluid and particle systems 516 12.6.2 Corresponding equations 519 12.6.3 Estimation of means 523 12.64 Summary 526 12.7 Extensions 529 12.7.1 Wall functions 529 12.7.2 The near-wall elliptic-relaxation model 534 12.7.3. The wavevector model 540 12.7.4 Mixing and reaction 545 12.8 Discussion 555 Large-eddy simulation 558 13.1 Introduction 558 13.2 Filtering 561 13.2.1 The general definition 561 13.2.2 Filtering in one dimension 562 13.2.3 Spectral representation 565 13.2.4 The filtered energy spectrum 568 13.2.5 The resolution of filtered fields S71 13.2.6 Filtering in three dimensions 575 13.2.7 The filtered rate of strain 578 13.3 Filtered conservation equations 581 13.3.1 Conservation of momentum 581 13.3.2. Decomposition of the residual stress 582 13.3.3 Conservation of energy 585 13.4 The Smagorinsky model 587 13.4.1. The definition of the model 587 13.4.2 Behavior in the inertial subrange 587 13.4.3. The Smagorinsky filter 590 13.4.4 Limiting behaviors 594 13.4.5 Near-wall resolution 598 13.4.6 Tests of model performance 601 13.5 LES in wavenumber space 604 13.5.1 Filtered equations 604Contents xiii 13.5.2. Triad interactions 13.5.3. The spectral energy balance 13.5.4 The spectral eddy viscosity 13.5.5 Backscatter 13.5.6 A statistical view of LES 13.5.7 Resolution and modelling 13.6 Further residual-stress models 13.6.1. The dynamic model 13.6.2 Mixed models and variants 13.6.3. Transport-equation models 13.6.4 Implicit numerical filters 13.6.5 Near-wall treatments 13.7 Discussion 13.7.1 An appraisal of LES 13.7.2 Final perspectives PART THREE: APPENDICES Appendix A Cartesian tensors Al Cartesian coordinates and vectors A.2_ The definition of Cartesian tensors A.3 Tensor operations A4_ The vector cross product A.5 A summary of Cartesian-tensor suffix notation Appendix B Properties of second-order tensors Appendix C Dirac delta functions Cl The definition of 5(x) C2 Properties of 5(x) C3 Derivatives of 5(x) C4 Taylor series C5 The Heaviside function C.6 Multiple dimensions Appendix D Fourier transforms Appendix E Spectral representation of stationary random processes E.1 Fourier series E.2_ Periodic random processes E.3 Non-periodic random processes E4_ Derivatives of the process Appendix F The discrete Fourier transform 606 609 610 611 612 615 619 619 627 629 631 634 635 635 638 641 643 643 647 649 654 659 661 670 670 672 673 675 675 677 678 683 683 686 689 690 692xiv Contents Appendix G Power-law spectra Appendix H Derivation of Eulerian PDF equations Appendix I Characteristic functions Appendix J Diffusion processes Bibliography Author index Subject index 696 702 707 713 727 749 754List of tables Spreading rate parameters of turbulent round jets Timescales in turbulent round jets Spreading parameters of turbulent axisymmetric wakes Statistics in homogeneous turbulent shear flow Characteristic scales of the dissipation spectrum Characteristic scales of the energy spectrum Tail contributions to velocity-derivative moments Wall regions and layers and their defining properties Stati in turbulent channel flow Computational difficulty of different turbulent flows Numerical parameters for DNS of isotropic turbulence Numerical parameters for DNS of channel flow Numerical parameter for DNS of the flow over a backward-facing step The turbulent Reynolds number of self-similar free shear flows Definition of variables in two-equation models Special states of the Reynolds-stress tensor Mean velocity gradients for simple deformations Tensors used in pressure-rate-of-strain models Coefficients in pressure—rate-of-strain models Coefficients in algebraic stress models Integrity basis for turbulent viscosity models Attributes of different RANS turbulence models Comparison between fluid and particle systems Different levels of PDF models Resolution in DNS and in some variants of LES Filter functions and transfer functions Estimates of filtered and residual quantities in the inertial subrange Definition of the different types of triad interactions Operations between first- and second-order tensors Fourier-transform pairs Spectral properties of random processes Power-law spectra and structure functions Relationships between characteristic functions and PDFs 101 131 153 157 238 240 259 275 283 338 349 355 356 366 384 394 415 426 427 452 453 457 518 555 560. 563 589 608 662 679 689 700 710XX Preface knowledge of probability theory, and consequently the necessary material is provided in the text (e.g. Sections 3.2-3.5). For a less demanding pace, Parts I and II can be covered in two semesters — there is ample material. Alternatively, if a coverage of modelling is not required, Part I by itself provides a reasonably complete introduction to turbulent flows. Many of the exercises ask the reader to ‘show that ...,’ and thereby intro- duce additional results and observations. Consequently, it is recommended that all the exercises be read, even if they are not performed. The book is designed to be a self-contained text, but sufficient references are given to provide an entry into the research literature. However much care is taken in the preparation of a book of this nature, it is inevitable that there will be errors in the first printing. A list of known corrections is given at http://mae.cornell.edu/~pope/TurbulentFlows. The reader is asked to report any further corrections to the author at pope@mae. cornell. edu. I am profoundly grateful to many people for their help in the prepa- ration of this work. For their support and technical input I thank my colleagues at Cornell, David Caughey, Sidney Leibovich, John Lumley, Di- etmar Rempfer, and Zellman Warhaft. For their valuable suggestions based on reading draft chapters, I am grateful to Peter Bradshaw, Paul Durbin, Rodney Fox, Kemo Hanjali¢, Charles Meneveau, Robert Moser, Blair Perot, Ugo Piomelli, P. K. Yeung, and Norman Zabusky. Similarly, I am grateful to the following Cornell graduates for their feedback on drafts of the book: Bertrand Delarue, Thomas Dreeben, Matthew Overholt, Paul Van Slooten, Jun Xu, Cem Albukrek, Dawn Chamberlain, Timothy Fisher, Laurent Myd- larski, Gad Reinhorn, Shankar Subramaniam, and Walter Welton. The first five mentioned are also thanked for their assistance in producing the figures. Most of the typescript was prepared by June Meyermann, whose patience, accuracy, and enthusiasm are greatly appreciated. The accuracy of the bib- liography has been much improved by the careful checking performed by Sarah Pope. Above all, I wish to thank my wife, Linda, for her patience, support, and encouragement during this project and over the years.Nomenclature The notation used is given here in the following order: upper-case Roman, lower-case Roman, upper-case Greek, lower-case Greek, superscripts, sub- scripts, symbols, and abbreviations. Then the symbols O( ), o( ), and ~ that are used to denote the order of a quantity are explained. Upper-case Roman At van Driest constant (Eq. (7.145)) A control surface bounding V B log-law constant (Eq. (7.43)) B, constant in the velocity-defect law (Eq. (7.50)) B, Loitsyanskii integral (Eq. (6.92)) By log-law constant for fully-rough walls (Eq. (7.120)) B(s/d,) log-law constant for rough walls (Eq. (7.121)) Cc Kolmogoroy constant related to E(x) (Eq. (6.16)) Cy coefficient in the Langevin equation (Eqs. (12.26) and (12.100)) cq Kolmogorov constant related to Ey1(«:) (Eq. (6.228)) Cc Kolmogorov constant related to E22(x;) (Eq. (6.231)) Cy Kolmogorov constant related to Di, (Eq. (6.30)) Q constant in the IP model (Eq. (11.129)) C3 constant in the model equation for w" (Eq. (12.194)) Ce LES dissipation coefficient (Eq. (13.285)) Cr skin-friction coefficient (tw/(3pU?)) Cr Rotta constant (Eq. (11.24)) Cs Smagorinsky coefficient (Eq. (13.128)) G constant in Reynolds-stress transport models (Eq. (11.147))XXII C, Ca, Car Cy Nomenclature constant in the model equation for ¢ (Eq. (11.150)) constants in the model equation for ¢ (Eq. (10.53)) turbulent-viscosity constant in the k-e model (Eq. (10.47)) LES eddy-viscosity coefficient (Eq. (13.286)) constant in the IEM mixing model (Eq. (12.326)) constant in the definition of Q (Eq. (12.193)) constants in the model equation for « (Eq. (10.93)) Kolmogorov constant (Eq. (12.96)) cross stress (Eq. (13.101)) pipe diameter second-order velocity structure function (Eq. (6.23)) second-order Lagrangian structure function (Eq. (12.95)) longitudinal second-order velocity structure function longitudinal third-order velocity structure function (Eq. (6.86)) transverse second-order velocity structure function nth-order longitudinal velocity structure function (Eq. (6.304) substantial derivative (0/ét-+ U-V) mean substantial derivative (@/ét + (U)+V) substantial derivative based on filtered velocity Cartesian coordinate system with basis vectors e; Cartesian coordinate system with basis vectors 2; kinetic energy (}U + U) kinetic energy of the mean flow (}(U) - (U)) kinetic energy flow rate of the mean flow energy-spectrum function (Eq. (3.166)) one-dimensional energy spectrum (Eq. (6.206)) energy-spectrum function of filtered velocity (Eq. (13.62)) frequency spectrum (defined for positive frequencies, Eq. (3.140)) frequency spectrum (defined for positive and negative frequencies, Eq. (E.31)) determinant of the normalized Reynolds stress (Eq. (11.52)) cumulative distribution function (CDF) of U (Eq. (3.7)) velocity-defect law (Eq. (7.46)) Fourier transform (Eq. (D.1)) inverse Fourier transform (Eq. (D.2))'Nomenciature AAU Fourier integral operator (Eq. (6.116)) coefficient in the GLM (Eggs. (12.26) and (12.110)) LES filter function LES filter transfer function shape factor (5°/@) Heaviside function (Eq. (C.33)) identity matrix indicator function for intermittency (Eq. (5.299)) principal invariants of the second-order tensor s (Eqs. (B.31)-(B.33)) kurtosis of the longitudinal velocity derivative kurtosis of Knudsen number modified Bessel function of the second kind lengthscale (k3/e) lengthscale (u’*/e) longitudinal integral lengthscale (Eq. (3.161)) lateral integral lengthscale (Eq. (6.48)) characteristic lengthscale of the flow length of side of cube in physical space resolved stress (Eq. (13.252)) Leonard stress (Eq. (13.100)) momentum flow rate of the mean flow scaled composite rate-of-strain tensor (Eq. (13.255)) normalized nth moment of the longitudinal velocity derivative (Eq. (6.303)) Mach number normal distribution with mean y and variance co” quantity of big order h quantity of little order h pressure (Eq. (2.32)) probability of event 4 particle pressure (Eq. (12.225)) projection tensor (Eq. (6.133)) production: rate of production of turbulent kinetic energy (Eq. (5.133)) rate of production of Reynolds stress (Eq. (7.179)) rate of production of residual kinetic energy (Eq. (13.123) rate of production of scalar variance (Eq. (5.282))