Professional Documents
Culture Documents
Fluid PDF
Fluid PDF
Contents
1
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
1
1
2
4
6
7
9
10
10
11
12
14
14
17
18
19
20
24
25
27
28
Hydrostatics
2.1
Introduction . . . . . . . . . . . . .
2.2
Hydrostatic Pressure . . . . . . . .
2.3
Buoyancy . . . . . . . . . . . . . .
2.4
Equilibrium of Floating Bodies . . .
2.5
Vertical Stability of Floating Bodies
2.6
Angular Stability of Floating Bodies
2.7
Determination of Metacentric Height
2.8
Energy of a Floating Body . . . . .
2.9
Curve of Buoyancy . . . . . . . . .
2.10 Rotational Hydrostatics . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
33
33
33
34
35
37
37
39
43
44
49
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3
2.11
2.12
2.13
2.14
2.15
3
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
51
54
58
62
68
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
73
73
74
76
77
78
80
81
86
92
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
95
95
95
96
98
98
100
102
103
105
108
110
111
112
113
114
117
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
119
119
119
122
123
124
126
130
132
136
Surface Tension
3.1
Introduction . . . . . . . . .
3.2
Young-Laplace Equation . .
3.3
Spherical Interfaces . . . . .
3.4
Capillary Length . . . . . .
3.5
Angle of Contact . . . . . .
3.6
Jurins Law . . . . . . . . .
3.7
Capillary Curves . . . . . .
3.8
Axisymmetric Soap-Bubbles
3.9
Exercises . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
4
5.10
5.11
5.12
5.13
6
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
141
143
144
147
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
149
149
149
150
152
153
154
160
165
168
179
181
186
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
193
193
193
193
195
196
197
198
199
200
203
205
210
213
215
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
219
219
219
220
224
229
234
239
240
244
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
5
8.10
8.11
9
Incompressible Aerodynamics
9.1
Introduction . . . . . . . . . . . .
9.2
Theorem of Kutta and Zhukovskii
9.3
Cylindrical Airfoils . . . . . . . .
9.4
Zhukovskiis Hypothesis . . . . .
9.5
Vortex Sheets . . . . . . . . . . .
9.6
Induced Flow . . . . . . . . . . .
9.7
Three-Dimensional Airfoils . . . .
9.8
Aerodynamic Forces . . . . . . .
9.9
Ellipsoidal Airfoils . . . . . . . .
9.10 Simple Flight Problems . . . . . .
9.11 Exercises . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
257
257
258
260
263
269
270
271
275
278
281
283
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
287
287
287
289
291
292
293
296
299
299
300
306
309
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
313
313
313
315
317
318
320
323
328
329
331
332
334
335
336
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
6
11.15 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338
12 Terrestrial Ocean Tides
12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . .
12.2 Tide Generating Potential . . . . . . . . . . . . . . . .
12.3 Decomposition of Tide Generating Potential . . . . . .
12.4 Expansion of Tide Generating Potential . . . . . . . .
12.5 Surface Harmonics and Solid Harmonics . . . . . . . .
12.6 Planetary Rotation . . . . . . . . . . . . . . . . . . . .
12.7 Total Gravitational Potential . . . . . . . . . . . . . .
12.8 Planetary Response . . . . . . . . . . . . . . . . . . .
12.9 Laplace Tidal Equations . . . . . . . . . . . . . . . . .
12.10 Harmonics of Forcing Term in Laplace Tidal Equations
12.11 Response to Equilibrium Harmonic . . . . . . . . . . .
12.12 Global Ocean Tides . . . . . . . . . . . . . . . . . . .
12.13 Non-Global Ocean Tides . . . . . . . . . . . . . . . .
12.14 Useful Lemma . . . . . . . . . . . . . . . . . . . . . .
12.15 Transformation of Laplace Tidal Equations . . . . . .
12.16 Another Useful Lemma . . . . . . . . . . . . . . . . .
12.17 Basis Eigenfunctions . . . . . . . . . . . . . . . . . .
12.18 Auxilliary Eigenfunctions . . . . . . . . . . . . . . . .
12.19 Gyroscopic Coecients . . . . . . . . . . . . . . . . .
12.20 Proudman Equations . . . . . . . . . . . . . . . . . .
12.21 Hemispherical Ocean Tides . . . . . . . . . . . . . . .
12.22 Exercises . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
341
341
341
343
344
345
346
347
348
353
357
359
360
366
367
369
371
371
373
374
376
380
385
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
393
393
394
394
396
397
404
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
407
407
407
411
411
414
415
416
420
425
427
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
7
14.11 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 430
15 Two-Dimensional Compressible Inviscid Flow
15.1 Introduction . . . . . . . . . . . . . . . .
15.2 Oblique Shocks . . . . . . . . . . . . . .
15.3 Supersonic Flow in Corner or over Wedge
15.4 Weak Oblique Shocks . . . . . . . . . . .
15.5 Supersonic Compression by Turning . . .
15.6 Supersonic Expansion by Turning . . . .
15.7 Detached Shocks . . . . . . . . . . . . .
15.8 Shock-Expansion Theory . . . . . . . . .
15.9 Thin-Airfoil Theory . . . . . . . . . . . .
15.10 Croccos Theorem . . . . . . . . . . . . .
15.11 Homenergic Homentropic Flow . . . . . .
15.12 Small-Perturbation Theory . . . . . . . .
15.13 Subsonic Flow Past a Wave-Shaped Wall .
15.14 Supersonic Flow Past a Wave-Shaped Wall
15.15 Linearized Subsonic Flow . . . . . . . . .
15.16 Linearized Supersonic Flow . . . . . . .
15.17 Exercises . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
437
437
437
441
443
445
447
450
451
453
458
459
460
464
466
468
470
472
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
483
483
483
484
486
487
489
491
492
495
496
498
498
499
502
502
504
505
506
510
511
514
516
519
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
8
A.24
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 520
B Cartesian Tensors
B.1
Introduction . . . . . . . . .
B.2
Tensors and Tensor Notation
B.3
Tensor Transformation . . .
B.4
Tensor Fields . . . . . . . .
B.5
Isotropic Tensors . . . . . .
B.6
Exercises . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
525
525
525
528
531
533
536
C Non-Cartesian Coordinates
C.1
Introduction . . . . . . . . . . . . .
C.2
Orthogonal Curvilinear Coordinates
C.3
Cylindrical Coordinates . . . . . . .
C.4
Spherical Coordinates . . . . . . . .
C.5
Exercises . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
539
539
539
543
546
548
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
557
557
557
559
562
562
565
1
Mathematical Models of Fluid Motion
1.1 Introduction
This chapter sets forth the mathematical models commonly used to describe the
equilibrium and dynamics of fluids. Unless stated otherwise, all of the analysis is
performed using a standard right-handed Cartesian coordinate system: (x1 , x2 , x3 ).
Moreover, the Einstein summation convention is employed (which means that repeated roman subscripts are assumed to be summed from 1 to 3). (See Appendix B.)
More information on the fundamental assumptions that underlie conventional fluid
mechanics, as well as the basic equations that describe fluid motion, can be found in
Batchelor 2000.
dierence between these two types of fluid lies in their relative compressibility. To be
more exact, gases can be compressed much more easily than liquids. Consequently,
any motion that involves significant pressure variations is generally accompanied by
much larger changes in mass density in the case of a gas than in the case of a liquid.
A macroscopic fluid ultimately consists of a huge number of individual molecules.
However, most practical applications of fluid mechanics are concerned with behavior on lengthscales that are far longer than the typical intermolecular spacing. Under
these circumstances, it is reasonable to suppose that the bulk properties of a given
fluid are the same as if it were completely continuous in structure. A corollary of this
assumption is that when, in the following, we talk about infinitesimal volume elements, we really mean elements that are suciently small that the bulk fluid properties (such as mass density, pressure, and velocity) are approximately constant across
them, but are still suciently large that they contain a very great number of molecules
(which implies that we can safely neglect any statistical variations in the bulk properties). The continuum hypothesis also requires infinitesimal volume elements to be
much larger than the molecular mean-free-path between collisions.
In addition to the continuum hypothesis, our study of fluid mechanics is premised
on three major assumptions:
1. Fluids are isotropic media. In other words, there is no preferred direction in a
fluid.
2. Fluids are Newtonian. In other words, there is a linear relationship between the
local shear stress and the local rate of strain, as first postulated by Isaac Newton
(1642-1727). It is also assumed that there is a linear relationship between the
local heat flux density and the local temperature gradient.
3. Fluids are classical. In other words, the macroscopic motion of ordinary fluids is
well described by Newtonian dynamics, and both quantum and relativistic eects
can be safely ignored.
It should be noted that the previous assumptions are not valid for all fluid types (e.g.,
certain liquid polymers, which are non-isotropic; thixotropic fluids, such as jelly or
paint, which are non-Newtonian; and quantum fluids, such as liquid helium, which
exhibit non-classical eects on macroscopic lengthscales). However, most practical
applications of fluid mechanics involve the equilibrium and motion of bodies of water or air, extending over macroscopic lengthscales, and situated relatively close to
the Earths surface. Such bodies are very well described as isotropic, Newtonian,
classical fluids.
(1.1)
The second type of force is short range in nature, and is most conveniently modeled as momentum transport within the fluid. Such transport is generally due to a
combination of the mutual forces exerted by contiguous molecules, and momentum
fluxes caused by relative molecular motion. Suppose that x (r, t) is the net flux density of x-directed fluid momentum due to short-range forces at position r and time
t. In other words, suppose that, at position r and time t, as a direct consequence
of short-range forces, x-momentum is flowing at the rate of | x | newton-seconds per
meter squared per second in the direction of vector x . Consider an infinitesimal
plane surface element, dS = n dS , located at point r. Here, dS is the area of the
element, and n its unit normal. (See Section A.7.) The fluid which lies on that side
of the element toward which n points is said to lie on its positive side, and vice versa.
The net flux of x-momentum across the element (in the direction of n) is x dS
newtons, which implies (from Newtons second law of motion) that the fluid on the
positive side of the surface element experiences a force x dS in the x-direction due
to short-range interaction with the fluid on the negative side. According to Newtons
third law of motion, the fluid on the negative side of the surface experiences a force
x dS in the x-direction due to interaction with the fluid on the positive side. Shortrange forces are often called surface forces, because they are directly proportional to
the area of the surface element across which they act. Let y (r, t) and z (r, t) be the
net flux density of y- and z-momentum, respectively, at position r and time t. By
a straightforward extension of previous argument, the net surface force exerted by
the fluid on the positive side of some planar surface element, dS, on the fluid on its
negative side is
(1.2)
f = ( x dS, y dS, z dS).
In tensor notation (see Appendix B), the previous equation can be written
fi = i j dS j ,
(1.3)
first-order tensor (as all forces are proper vectors), and the dS i are the components
of an arbitrary first-order tensor (as surface elements are also proper vectorssee
Section A.7and Equation (1.3) holds for surface elements whose normals point in
any direction), so application of the quotient rule (see Section B.3) to Equation (1.3)
reveals that the i j transform under rotation of the coordinate axes as the components
of a second-order tensor.] We can interpret i j (r, t) as the i-component of the force
per unit area exerted, at position r and time t, across a plane surface element normal
to the j-direction. The three diagonal components of i j are termed normal stresses,
because each of them gives the normal component of the force per unit area acting
across a plane surface element parallel to one of the Cartesian coordinate planes. The
six non-diagonal components are termed shear stresses, because they drive shearing
motion in which parallel layers of fluid slide relative to one another.
where the first term on the right-hand side is the integrated volume force acting
throughout V, whereas the second term is the net surface force acting across S . Making use of the tensor divergence theorem (see Section B.4), the previous expression
becomes
i j
fi =
Fi dV +
dV.
(1.5)
V
V x j
In the limit V 0, it is reasonable to suppose that the Fi and i j /x j are approximately constant across the element. In this situation, both contributions on the
right-hand side of the previous equation scale as V. According to Newtonian dynamics, the i-component of the net force acting on the volume element is equal to
the i-component of the rate of change of its linear momentum. However, in the limit
V 0, the linear acceleration and mass density of the fluid are both approximately
constant across the element. In this case, the rate of change of the elements linear
momentum also scales as V. In other words, the net volume force, surface force, and
rate of change of linear momentum of an infinitesimal fluid element all scale as the
volume of the element, and consequently remain approximately the same order of
magnitude as the volume shrinks to zero. We conclude that the linear equation of
motion of an infinitesimal fluid element places no particular restrictions on the stress
tensor.
The i-component of the total torque, taken about the origin O of the coordinate
system, acting on a fluid element that consists of a fixed volume V enclosed by a
(1.6)
where the first and second terms on the right-hand side are due to volume and surface forces, respectively. [Here, i jk is the third-order permutation tensor. See Equation (B.7).] Making use of the tensor divergence theorem (see Section B.4), the
previous expression becomes
(x j kl )
i jk x j Fk dV +
i jk
dV,
(1.7)
i =
xl
V
V
which reduces to
i =
i jk x j Fk dV +
V
i jk k j dV +
V
i jk x j
V
kl
dV,
xl
(1.8)
because xi /x j = i j . [Here, i j is the second-order identity tensor. See Equation (B.9).] Assuming that point O lies within the fluid element, and taking the limit
V 0 in which the Fi , i j , and i j /x j are all approximately constant across the
element, we deduce that the first, second, and third terms on the right-hand side of
the previous equation scale as V 4/3 , V, and V 4/3 , respectively (because x V 1/3 ).
According to Newtonian dynamics, the i-component of the total torque acting on
the fluid element is equal to the i-component of the rate of change of its net angular
momentum about O. Assuming that the linear acceleration and density of the fluid
are approximately constant across the element, we deduce that the rate of change of
its angular momentum scales as V 4/3 (because the net rate of change of the linear
momentum scales as V, so the net rate of change of the angular momentum scales
as x V, and x V 1/3 ). Hence, it is clear that the rotational equation of motion of
a fluid element, surrounding a general point O, becomes completely dominated by
the second term on the right-hand side of Equation (1.8) in the limit that the volume
of the element approaches zero (because this term is a factor V 1/3 larger than the
other terms). It follows that the second term must be identically zero (otherwise an
infinitesimal fluid element would acquire an absurdly large angular velocity). This is
only possible, for all choices of the position of point O, and the shape of the element,
if
(1.9)
i jk k j = 0
throughout the fluid. The previous relation shows that the stress tensor must be symmetric: that is,
ji = i j .
(1.10)
It immediately follows that the stress tensor only has six independent components
(i.e., 11 , 22 , 33 , 12 , 13 , and 23 ).
It is always possible to choose the orientation of a set of Cartesian axes in such a
manner that the non-diagonal components of a given symmetric second-order tensor
field are all set to zero at a given point in space. (See Exercise B.6.) With reference to such principal axes, the diagonal components of the stress tensor i j become
so-called principal stresses11 , 22 , 33 , say. In general, the orientation of the
principal axes varies with position. The normal stress 11 acting across a surface
element perpendicular to the first principal axis corresponds to a tension (or a compression if 11 is negative) in the direction of that axis. Likewise, for 22 and 33 .
Thus, the general state of the fluid, at a particular point in space, can be regarded as
a superposition of tensions, or compressions, in three orthogonal directions.
The trace of the stress tensor, ii = 11 + 22 + 33 , is a scalar, and, therefore,
independent of the orientation of the coordinate axes. (See Appendix B.) Thus, it
follows that, irrespective of the orientation of the principal axes, the trace of the stress
tensor at a given point is always equal to the sum of the principal stresses: that is,
ii = 11 + 22 + 33 .
(1.11)
1
3 ii ,
0,
0
1
0 ,
(1.12)
0,
3 ii ,
1
0,
0,
3 ii
and
11 13 ii ,
0,
0
1
0,
22 3 ii ,
0
0,
0,
33 13 ii
(1.13)
The tensor (1.12) is isotropic (see Section B.5), and corresponds to the same
normal force per unit area acting inward (because the sign of ii /3 is invariably
negative) on each face of the volume element. This uniform compression acts to
change the elements volume, but not its shape, and can easily be withstood by the
fluid within the element.
The tensor (1.13) represents the departure of the stress tensor from an isotropic
form. The diagonal components of this tensor have zero sum, in view of Equation (1.11), and thus represent equal and opposite forces per unit area, acting on
opposing faces of the volume element, which are such that the forces on at least
one pair of opposing faces constitute a tension, and the forces on at least one pair
constitute a compression. Such forces necessarily act to change the shape of the volume element, either elongating or compressing it along one of its symmetry axes.
Moreover, this tendency cannot be oset by any volume force acting on the element,
because such forces become arbitrarily small compared to surface forces in the limit
that the elements volume tends to zero (because the ratio of the net volume force to
the net surface force scales as the volume to the surface area of the element, which
tends to zero in the limit that the volume tends to zerosee Section 1.4). We previously defined a fluid as a material that is incapable of withstanding any tendency
of applied forces to change its shape. (See Section 1.2.) It follows that if the diagonal components of the tensor (1.13) are non-zero anywhere inside the fluid then it
is impossible for the fluid at that point to be at rest. Hence, we conclude that the
principal stresses, 11 , 22 , and 33 , must be equal to one another at all points in a
static fluid. This implies that the stress tensor takes the isotropic form (1.12) everywhere in a stationary fluid. Furthermore, this is true irrespective of the orientation of
the coordinate axes, because the components of an isotropic tensor are rotationally
invariant. (See Section B.5.)
Fluids at rest are generally in a state of compression, so it is convenient to write
the stress tensor of a static fluid in the form
i j = p i j ,
(1.14)
(1.15)
where p = ii /3 is the static pressure: that is, minus the normal stress acting in
any direction. The normal stress at a given point in a moving fluid generally varies
with direction. In other words, the principal stresses are not equal to one another.
However, we can still define the mean principal stress as (11 + 22 + 33 )/3 =
ii /3. Moreover, given that the principal stresses are actually normal stresses (in a
coordinate frame aligned with the principal axes), we can also regard ii /3 as the
mean normal stress. It is convenient to define pressure in a moving fluid as minus the
mean normal stress: that is,
1
(1.16)
p = ii .
3
Thus, we can write the stress tensor in a moving fluid as the sum of an isotropic
part, p i j , which has the same form as the stress tensor in a static fluid, and a
remaining non-isotropic part, di j , which includes any shear stresses, and also has
diagonal components whose sum is zero. In other words,
i j = p i j + di j ,
(1.17)
dii = 0.
(1.18)
where
Moreover, because i j and i j are both symmetric tensors, it follows that di j is also
symmetric: that is,
(1.19)
d ji = di j .
It is clear that the so-called deviatoric stress tensor, di j , is a consequence of fluid
motion, because it is zero in a static fluid. Suppose, however, that we were to view a
static fluid both in its rest frame and in a frame of reference moving at some constant
velocity relative to the rest frame. We would expect the force distribution within the
fluid to be the same in both frames of reference, because the fluid does not accelerate
in either. However, in the first frame, the fluid appears stationary and the deviatoric
stress tensor is therefore zero, while in the second it has a spatially uniform velocity
field and the deviatoric stress tensor is also zero (because it is the same as in the
rest frame). We, thus, conclude that the deviatoric stress tensor is zero both in a
stationary fluid and in a moving fluid possessing no spatial velocity gradients. This
suggests that the deviatoric stress tensor is driven by velocity gradients within the
fluid. Moreover, the tensor must vanish as these gradients vanish.
Let the vi (r, t) be the Cartesian components of the fluid velocity at point r and
time t. The various velocity gradients within the fluid then take the form vi /x j .
The simplest possible assumption, which is consistent with the previous discussion,
is that the components of the deviatoric stress tensor are linear functions of these
velocity gradients: that is,
vk
.
(1.20)
di j = Ai jkl
xl
Here, Ai jkl is a fourth-order tensor (this follows from the quotient rule because di j
and vi /x j are both proper second-order tensors). Any fluid in which the deviatoric
stress tensor takes the previous form is termed a Newtonian fluid, because Newton
was the first to postulate a linear relationship between shear stresses and velocity
gradients.
In an isotropic fluidthat is, a fluid in which there is no preferred directionwe
would expect the fourth-order tensor Ai jkl to be isotropicthat is, to have a form in
which all physical distinction between dierent directions is absent. As demonstrated
in Section B.5, the most general expression for an isotropic fourth-order tensor is
Ai jkl = i j kl + ik jl + il jk ,
(1.21)
where , , and are arbitrary scalars (which can be functions of position and time).
v j
vk
vi
i j +
+
.
xk
x j
xi
(1.22)
1 vi v j
+
2 x j xi
(1.24)
1
di j = 2 ei j ekk i j ,
(1.25)
3
where = . We, thus, conclude that the most general expression for the stress
tensor in an isotropic Newtonian fluid is
1
i j = p i j + 2 ei j ekk i j ,
(1.26)
3
where p(r, t) and (r, t) are arbitrary scalars.
1.7 Viscosity
The significance of the parameter , appearing in the previous expression for the
stress tensor, can be seen from the form taken by the relation (1.25) in the special case
of simple shearing motion. With v1 /x2 as the only non-zero velocity derivative, all
of the components of di j are zero apart from the shear stresses
d12 = d21 =
v1
.
x2
(1.27)
Thus, is the constant of proportionality between the rate of shear and the tangential
force per unit area when parallel plane layers of fluid slide over one another. This
constant of proportionality is generally referred to as viscosity. It is a matter of
experience that the force acting between layers of fluid undergoing relative sliding
motion always tends to oppose the motion, which implies that > 0.
The viscosities of dry air and pure water at 20 C and atmospheric pressure are
about 1.8 105 kg/(m s) and 1.0 103 kg/(m s), respectively. In neither case does
the viscosity exhibit much variation with pressure. However, the viscosity of air
increases by about 0.3 percent, and that of water decreases by about 3 percent, per
degree Centigrade rise in temperature (Batchelor 2000).
10
where the integral is taken over all elements of V. Let dS be an outward directed
element of the bounding surface of V. Suppose that this element is located at point
r. The volume of fluid that flows per second across the element, and so out of V,
is v(r, t) dS. Thus, the amount of the fluid property under consideration that is
convected across the element per second is (r, t) v(r, t) dS. It follows that the net
amount of the property that is convected out of volume V by fluid flow across its
bounding surface S is
=
v dS,
(1.29)
where the integral is taken over all outward directed elements of S . Suppose, finally,
that the property in question is created within the volume V at the rate S per second.
The conservation equation for the fluid property takes the form
d
= S .
(1.30)
dt
In other words, the rate of increase in the amount of the property contained within
V is the dierence between the creation rate of the property inside V, and the rate at
which the property is convected out of V by fluid flow. The previous conservation
law can also be written
d
+ = S .
(1.31)
dt
Here, is termed the flux of the property out of V, whereas S is called the net
generation rate of the property within V.
11
dV +
v dS = 0,
(1.35)
V t
S
because V is non-time-varying. Making use of the divergence theorem (see Section A.20), the previous equation becomes
+ ( v) dV = 0.
(1.36)
V t
However, this result is true irrespective of the size, shape, or location of volume V,
which is only possible if
+ ( v) = 0
(1.37)
t
throughout the fluid. The previous expression is known as the equation of fluid continuity, and is a direct consequence of mass conservation.
=
+v =
+ vi
.
Dt t
t
xi
(1.39)
Clearly, the so-called convective time derivative, D/Dt, represents the time derivative
seen in the local rest frame of the fluid.
12
(1.40)
Moreover, the flux of i-momentum across S , and out of V, is [see Equation (1.29)]
i =
vi v j dS j .
(1.44)
S
Finally, the i-component of the net force acting on the fluid within V is
fi =
Fi dV +
i j dS j ,
V
(1.45)
where the first and second terms on the right-hand side are the contributions from
volume and surface forces, respectively.
Momentum conservation requires that the rate of increase of the net i-momentum
of the fluid contained within V, plus the flux of i-momentum out of V, is equal to the
rate of i-momentum generation within V. Of course, from Newtons second law of
13
motion, the latter quantity is equal to the i-component of the net force acting on the
fluid contained within V. Thus, we obtain [cf., Equation (1.31)]
dPi
+ i = fi ,
dt
which can be written
( vi )
dV +
vi v j dS j =
Fi dV +
i j dS j ,
t
V
S
V
S
(1.46)
(1.47)
i j
( vi ) ( vi v j )
+
dV =
Fi +
dV.
t
x j
x j
V
(1.48)
However, the previous result is valid irrespective of the size, shape, or location of
volume V, which is only possible if
i j
( vi ) ( vi v j )
+
= Fi +
t
x j
x j
(1.49)
everywhere inside the fluid. Expanding the derivatives, and rearranging, we obtain
v j
i j
vi
vi
+ vj
+ vj
+
.
vi +
= Fi +
t
x j
x j
t
x j
x j
(1.50)
+ vj
+
= 0.
t
x j
x j
(1.51)
So, combining Equations (1.50) and (1.51), we obtain the following fluid equation of
motion,
i j
vi
vi
+ vj
.
(1.52)
= Fi +
t
x j
x j
An alternative form of this equation is
Fi 1 i j
Dvi
=
+
.
Dt
x j
(1.53)
The previous equation describes how the net volume and surface forces per unit mass
acting on a co-moving fluid element determine its acceleration.
14
vi v j
Dvi
p
2 v j
= Fi
+
+
.
(1.54)
Dt
xi x j
x j xi
xi 3 x j
This equation is generally known as the Navier-Stokes equation, and is named after
Claude-Louis Navier (1785-1836) and George Gabriel Stokes (1819-1903). In situations in which there are no strong temperature gradients in the fluid, it is a good
approximation to treat viscosity as a spatially uniform quantity, in which case the
Navier-Stokes equation simplifies somewhat to give
2
vi
p
1 2v j
Dvi
= Fi
+
+
.
(1.55)
Dt
xi
x j x j 3 xi x j
When expressed in vector form, the previous expression becomes
v
1
Dv
2
+ (v ) v = F p + v + ( v) ,
Dt
t
3
(1.56)
bi
,
x j
( 2 v)i = 2 vi .
(1.57)
(1.58)
Note, however, that the previous identities are only valid in Cartesian coordinates.
(See Appendix C.)
1
1
E =
E + vi vi v j dS j =
E + vi vi v j dV,
(1.60)
2
2
S
V x j
15
where use has been made of the tensor divergence theorem. According to the first
law of thermodynamics, the rate of increase of the energy contained within V, plus
the net energy flux out of V, is equal to the net rate of work done on the fluid within
V, minus the net heat flux out of V: that is,
. .
dE
+ E = W Q,
dt
.
(1.61)
where W is the net rate of work, and Q the net heat flux. It can be seen that W Q is
the eective energy generation rate within V [cf., Equation (1.31)].
The net rate at which volume and surface forces do work on the fluid within V is
.
(vi i j )
W=
vi Fi dV +
vi i j dS j =
vi F i +
dV,
(1.62)
x j
V
S
V
where use has been made of the tensor divergence theorem.
Generally speaking, heat flow in fluids is driven by temperature gradients. Let
the qi (r, t) be the Cartesian components of the heat flux density at position r and
time t. It follows that the heat flux across a surface element dS, located at point r, is
q dS = qi dS i . Let T (r, t) be the temperature of the fluid at position r and time t.
Thus, a general temperature gradient takes the form T/xi. Let us assume that there
is a linear relationship between the components of the local heat flux density and the
local temperature gradient: that is,
q i = Ai j
T
,
x j
(1.63)
where the Ai j are the components of a second-rank tensor (which can be functions
of position and time). In an isotropic fluid we would expect Ai j to be an isotropic
tensor. (See Section B.5.) However, the most general second-order isotropic tensor
is simply a multiple of i j . Hence, we can write
Ai j = i j ,
(1.64)
where (r, t) is termed the thermal conductivity of the fluid. It follows that the most
general expression for the heat flux density in an isotropic fluid is
T
,
xi
(1.65)
q = T.
(1.66)
qi =
or, equivalently,
.
T
T
Q=
dS i =
dV,
(1.67)
xi
xi
S
V xi
where use has been made of the tensor divergence theorem.
16
Equations (1.59)(1.62) and (1.67) can be combined to give the following energy
conservation equation:
1
1
E + vi vi v j dV
E + vi vi +
2
x j
2
V t
T
vi F i +
vi i j +
dV.
(1.68)
=
x j
x j
V
However, this result is valid irrespective of the size, shape, or location of volume V,
which is only possible if
1
1
E + vi vi v j = vi F i +
vi i j +
(1.69)
E + vi vi +
t
2
x j
2
x j
x j
everywhere inside the fluid. Expanding some of the derivatives, and rearranging, we
obtain
T
D
1
vi i j +
,
(1.70)
E + vi vi = vi F i +
Dt
2
x j
x j
where use has been made of the continuity equation, (1.40). The scalar product of v
with the fluid equation of motion, (1.53), yields
i j
D 1
Dvi
=
vi vi = vi F i + vi
.
(1.71)
vi
Dt
Dt 2
x j
Combining the previous two equations, we get
DE
vi
=
i j +
.
Dt
x j
x j x j
(1.72)
Finally, making use of Equation (1.26), we deduce that the energy conservation equation for an isotropic Newtonian fluid takes the general form
p vi 1
T
DE
=
+ +
.
(1.73)
Dt
xi
x j x j
Here,
vi vi
vi
1
vi v j 2 vi v j
=
di j = 2 ei j ei j eii e j j =
+
(1.74)
x j
3
x j x j x j xi 3 xi x j
is the rate of heat generation per unit volume due to viscosity. When written in vector
form, Equation (1.73) becomes
p
( T )
DE
= v+ +
.
Dt
(1.75)
According to the previous equation, the internal energy per unit mass of a co-moving
fluid element evolves in time as a consequence of work done on the element by
pressure as its volume changes, viscous heat generation due to flow shear, and heat
conduction.
17
(1.76)
(1.77)
(1.79)
where
(1.80)
speaking, momentum diuses a distance of order t meters in t seconds as a consequence of viscosity. The kinematic viscosity of water at 20 C is about 1.0106 m2 /s
(Batchelor 2000). It follows that viscous momentum diusion in water is a relatively
slow process.
The complete set of equations governing incompressible flow is
=
v = 0,
p
Dv
=
+ 2 v.
Dt
(1.81)
(1.82)
18
Here, and are regarded as known constants, and (r, t) as a known function.
Thus, we have four equationsnamely, Equation (1.81), plus the three components
of Equation (1.82)for four unknownsnamely, the pressure, p(r, t), plus the three
components of the velocity, v(r, t). Note that an energy conservation equation is
redundant in the case of incompressible fluid flow.
(1.83)
(1.84)
Dp p D
M 2 p
1
=+
,
1 Dt
Dt
R
(1.85)
where
=
cp
cV + R
=
cV
cV
(1.86)
is the ratio of the molar specific heat at constant pressure, c p , to that at constant
volume, cV . [Incidentally, the result that c p = cV + R for an ideal gas is a standard
theorem of thermodynamics (Reif 1965).] (See Section 14.2.) The ratio of specific
heats of dry air at 20 C is 1.40 (Batchelor 2000).
19
The complete set of equations governing compressible ideal gas flow are
D
= v,
Dt
p
2
1
Dv
=
+
v + ( v) ,
Dt
1
Dp p D
M 2 p
=+
,
1 Dt
Dt
R
(1.87)
(1.88)
(1.89)
(1.90)
2
1
1
Dv
v
,
= 1+ 2 +
p 2 +
Re
Re
Dt
Fr
Fr
(1.91)
V0
Fr =
.
(g L)1/2
Re =
(1.92)
(1.93)
Here, the dimensionless quantities Re and Fr are known as the Reynolds number
and the Froude number, respectively. [After Osborne Reynolds (18421912) and
William Froude (18101879), respectively.] The Reynolds number is the typical ratio
of inertial to viscous forces within the fluid, whereas the square of the Froude number
is the typical ratio of inertial to gravitational forces. Thus, viscosity is relatively
important compared to inertia when Re 1, and vice versa. Likewise, gravity is
20
relatively important compared to inertia when Fr 1, and vice versa. Note that, in
principal, Re and Fr are the only quantities in Equations (1.90) and (1.91) that can be
significantly greater or smaller than unity.
For the case of water at 20 C, located on the surface of the Earth (Batchelor
2000),
Re 1.0 106 L(m) V0 (m s1 ),
(1.94)
(1.95)
(1.96)
(1.97)
For the case of lubrication oil at 20 C, located on the surface of the Earth,
1.0 104 m2 s1 (i.e., oil is about 100 times more viscous than water), and
so (Batchelor 2000)
Re 1.0 104 L(m) V0 (m s1 ),
(1.98)
(1.99)
Suppose that oil is slowly flowing down a narrow lubrication channel such that L
103 m and V0 101 m s1 . It follows, from the previous expressions, that Re 1
and Fr 1. In this situation, the inertial term on the left-hand side of (1.91) becomes
negligible, and the (unnormalized) incompressible fluid flow equations reduce to the
following inertia-free, incompressible, fluid flow equations:
v = 0,
0=
(1.100)
p
+ 2 v.
(1.101)
21
a typical gravitational acceleration (assuming that represents a gravitational potential energy per unit mass). Furthermore, p0 corresponds to atmospheric pressure at
ground level, and is a uniform constant. It follows that p represents deviations from
atmospheric pressure. All barred quantities are dimensionless, and are designed to
be comparable with unity. The normalized equations of compressible ideal gas flow
take the form
D
= v,
(1.102)
Dt
2
1
1 p
1
1
Dv
2 +
= 1+ 2 +
v ( v) ,
(1.103)
Re
Re
3
Dt
Fr
Fr
p0 + p D
1
Dp
1
2 p0 + p
+
, (1.104)
=
1 Dt
Dt
1 + Re (1 + 1/Fr 2 ) Re Pr
p0 =
1
,
Ma (1 + 1/Fr 2 + 1/Re)
2
(1.105)
where D/Dt /t + v ,
L V0
,
V0
Fr =
,
(g L)1/2
Pr =
,
H
V0
Ma =
,
p0 /0
Re =
(1.106)
(1.107)
(1.108)
(1.109)
and
,
0
M
H =
.
R 0
=
(1.110)
(1.111)
Here, the dimensionless numbers Re, Fr, Pr, and Ma are known as the Reynolds
number, Froude number, Prandtl number, and Mach number, respectively. [The
latter two numbers are named after Ludwig Prandtl (18751953) and Ernst Mach
(18381916), respectively.] The Reynolds number is the typical ratio of inertial to
viscous forces within the gas, the square of the Froude number is the typical ratio
of inertial to gravitational forces, the Prandtl number is the typical ratio of the momentum and thermal diusion rates, and the Mach number is the typical ratio of the
gas flow and sound propagation speeds. Thus, thermal diusion is far faster than
momentum diusion when Pr 1, and vice versa. Moreover, the gas flow is termed
subsonic when
Ma 1, supersonic when Ma
1, and transonic when Ma O(1).
Note that p0 /0 is the speed of sound in the undisturbed gas (Reif 1965). The
22
quantity H is called the thermal diusivity of the gas, and has units of
meters squared
per second. Thus, heat typically diuses through the gas a distance H t meters in
t seconds. The thermal diusivity of dry air at atmospheric pressure and 20 C is
about H = 2.1 105 m2 s1 (Batchelor 2000). It follows that heat diusion in air
is a relatively slow process. The kinematic viscosity of dry air at atmospheric pressure and 20 C is about = 1.5 105 m2 s1 (Batchelor 2000). Hence, momentum
diusion in air is also a relatively slow process.
For the case of dry air at atmospheric pressure and 20 C (Batchelor 2000),
Re 6.7 104 L(m) V0 (m s1 ),
Fr 3.2 10
V0 (m s )/[L(m)]
Pr 7.2 101 ,
Ma 2.9 103 V0 (m s1 ).
(1.112)
1/2
(1.113)
(1.114)
(1.115)
(1.116)
v 1.
(1.117)
(1.118)
(1.119)
It follows that the equations which govern subsonic gas dynamics close to the surface
of the Earth are essentially the same as those that govern the flow of water.
Suppose that L 1 m and V0 300 m s1, as is typically the case for transonic
air dynamics (e.g., air flow over the wing of a fighter jet). In this situation, Equations (1.105) and (1.112)(1.115) yield Re, Fr
1 and Ma, Pr, p0 O(1). It follows
that the final two terms on the right-hand sides of Equations (1.103) and (1.104) can
be neglected. Thus, the (unnormalized) compressible ideal gas flow equations reduce
23
D
= v,
Dt
p
Dv
= ,
Dt
D p
= 0.
Dt
(1.120)
(1.121)
(1.122)
p
cV
ln .
S=
M
(1.123)
D
= v,
Dt
p
Dv
= ,
Dt
=
.
p0
0
(1.124)
(1.125)
(1.126)
24
v x
,
x
vy
,
= p + 2
y
xx = p + 2
(1.127)
yy
(1.128)
zz = p + 2
v x
+
xy = yx =
y
v x
+
xz = zx =
z
vy
+
yz = zy =
z
vz
,
z
vy
,
x
vz
,
x
vz
,
y
(1.129)
(1.130)
(1.131)
(1.132)
where v is the velocity, p the pressure, and the viscosity. The equations of compressible fluid flow, (1.87)(1.89) (from which the equations of incompressible fluid
flow can easily be obtained by setting = 0), become
D
= ,
Dt
1 p 2
1
Dv x
=
+
vx +
Dt
x
x
3
Dvy
1 p 2
1
=
+
vy +
Dt
y
y
3
1 p 2
1
Dvz
=
+
vz +
Dt
z
z
D p D
M 2 p
1
=+
,
1 Dt
Dt
R
(1.133)
,
x
,
y
,
z
(1.134)
(1.135)
(1.136)
(1.137)
25
where is the mass density, the ratio of specific heats, the heat conductivity, M
the molar mass, and R the molar ideal gas constant. Furthermore,
v x vy vz
+
+
,
x
y
z
=
+ vx
+ vy
+ vz ,
Dt t
x
y
z
2
2
2
+
+
,
x 2 y 2 z 2
2
v x 2
vy 2
vz
1 v x vy
= 2
+
+
+
+
x
y
z
2 y
x
2
1 v x vz
1 vy vz
+
+
+
+
.
2 z
x
2 z
y
2 =
(1.138)
(1.139)
(1.140)
(1.141)
1 v vr
= p + 2
+
,
r
r
rr = p + 2
vz
,
zz = p + 2
z
1 vr v v
+
r = r =
,
r
r
r
vr vz
+
,
rz = zr =
z
r
1 vz v
+
,
z = z =
r
z
(1.142)
(1.143)
(1.144)
(1.145)
(1.146)
(1.147)
26
D
= ,
Dt
Dvr v2
1 p
Dt
r
r
r
2
vr
2 v 1
+
+
vr 2 2
,
r
r 3 r
1 p 1
Dv vr v
+
=
Dt
r
r r
2
2 vr v
1
+
+
v + 2
,
r r 2 3 r
1 p 2
1
Dvz
=
+
vz +
,
Dt
z
z
3 z
D p D
M 2 p
1
=+
,
1 Dt
Dt
R
(1.148)
(1.149)
(1.150)
(1.151)
(1.152)
where
1 (r vr ) 1 v vz
+
+
,
(1.153)
r r
r
z
D
=
+ vr
+
+ vz ,
(1.154)
Dt t
r
r
z
1
2
1 2
2 =
+ 2,
(1.155)
r
+ 2
2
r r r
r
z
2
2
2
vr
1 v vr
vz
1 1 vr v v
+
+
= 2
+
+
+
r
r
r
z
2 r
r
r
2
1 vr vz
1 v 1 vz
+
+
+
+
(1.156)
.
2 z
r
2 z
r
=
27
1 v vr
= p + 2
+
,
r
r
1 v vr cot v
+ +
,
= p + 2
r sin
r
r
1 vr v v
+
,
r = r =
r
r
r
1 vr v v
+
,
r = r =
r sin
r
r
1 v 1 v cot v
= =
+
,
r sin r
r
rr = p + 2
(1.157)
(1.158)
(1.159)
(1.160)
(1.161)
(1.162)
2
2
Dvr v + v
1 p 2
2 vr
2 v
+
vr 2 2
Dt
r
r
r
r
r
v 1
2 cot v
2
+
2
,
(1.164)
r2
r sin 3 r
2
1 p 1 2
2 vr
Dv vr v cot v
+
=
+
v + 2
Dt
r
r r
r
2 cot v
1
v
+
,
(1.165)
r 2 sin2 r 2 sin 3 r
Dv vr v + cot v v
v
1
1 2
p
+
=
+
v
2
Dt
r
r sin r sin
r sin2
vr
2
2 cot v
1
+
+
+
, (1.166)
r 2 sin2 r 2 sin 3 r sin
D p D
M 2 p
1
=+
,
(1.167)
1 Dt
Dt
R
28
where
1 (r 2 vr )
1 (sin v )
1 v
+
+
,
2
r
r sin
r sin
r
v
D
=
+ vr
+
+
,
Dt t
r
r r sin
2
1 2
1
2 = 2
,
r
+ 2
sin
+
r
r r
r sin
r 2 sin2 2
2
vr
1 v vr
1 v vr cot v
+
+ +
= 2
+
+
r
r
r
r sin
r
r
2
1 vr v v
1 1 vr v v
1
+
+
+
2 r
r
r
2 r sin
r
r
2
1 v 1 v cot v
1
+
+
.
2 r sin r
r
=
(1.168)
(1.169)
(1.170)
(1.171)
1.21 Exercises
1.1 Equations (1.66), (1.75), and (1.87) can be combined to give the following
energy conservation equation for a non-ideal compressible fluid:
DE p D
= q,
Dt
Dt
where is the mass density, p the pressure, E the internal energy per unit
mass, the viscous energy dissipation rate per unit volume, and q the heat
flux density. We also have
D
= v,
Dt
q = T,
where v is the fluid velocity, T the temperature, and the thermal conductivity. According to a standard theorem in thermodynamics (Reif 1965),
T dS = dE
p
d,
2
where S is the entropy per unit mass. Moreover, the entropy flux density at a
given point in the fluid is (Hazeltine and Waelbroeck 2004)
s = Sv +
q
,
T
29
where the first term on the right-hand side is due to direct entropy convection
by the fluid, and the second is the entropy flux density associated with heat
conduction.
Derive an entropy conservation equation of the form
dS
+ S = S ,
dt
where S is the net amount of entropy contained in some fixed volume V, S
the entropy flux out of V, and S the net rate of entropy creation within V.
Give expressions for S , S , and S . Demonstrate that the entropy creation
rate per unit volume is
qq
= +
.
T
T2
Finally, show that 0, in accordance with the second law of thermodynamics.
1.2 The Navier-Stokes equation for an incompressible fluid of uniform mass density takes the form
p
Dv
=
+ 2 v,
Dt
where v is the fluid velocity, p the pressure, the potential energy per unit
mass, and the (uniform) kinematic viscosity. The incompressibility constraint requires that
v = 0.
Finally, the quantity
v
is generally referred to as the fluid vorticity.
Derive the following vorticity evolution equation from the Navier-Stokes
equation:
D
= ( ) v + 2 .
Dt
1.3 Consider two-dimensional incompressible fluid flow. Let the velocity field
take the form
v = v x (x, y, t) e x + vy (x, y, t) ey.
Demonstrate that the equations of incompressible fluid flow (see Exercise 1.2)
can be satisfied by writing
vx =
vy =
,
y
,
x
30
+ [, ] = 2 ,
t
and
= 2 .
Here, [A, B] ez A B, and 2 = 2 /x 2 + 2 /y 2 . Furthermore, the
quantity is termed a stream function, because v = 0. In other words,
the fluid flow is everywhere parallel to contours of .
1.4 Consider incompressible irrotational flow: that is, flow that satisfies
v = 0,
as well as
Dv
p
=
+ 2 v,
Dt
v = 0.
Here, v is the fluid velocity, the uniform mass density, p the pressure, the
potential energy per unit mass, and the (uniform) kinematic viscosity.
Demonstrate that the previous equations can be satisfied by writing
v = ,
where
2 = 0,
and
1 2 p
+ v + + = C(t).
t 2
Here, C(t) is a spatial constant. This type of flow is known as potential flow,
because the velocity field is derived from a scalar potential.
1.5 The equations of inviscid adiabatic ideal gas flow are
D
= v,
Dt
Dv
p
=
,
Dt
D p
= 0.
Dt
Here, is the mass density, v the flow velocity, p the pressure, the potential
energy per unit mass, and the (uniform) ratio of specific heats. Suppose
that the pressure and potential energy are both time independent: that is,
p/t = /t = 0.
31
1 2
p
v +
+
2
1
DE
p D
= 0.
Dt 2 Dt
Here, is the mass density, v the flow velocity, p the pressure, the potential
energy per unit mass, and E the internal energy per unit mass. Suppose that
the pressure and potential energy are both time independent: that is, p/t =
/t = 0. Demonstrate that
H=
1 2
p
v +E+ +
2
1 2 p
v + + .
2
32
2
Hydrostatics
2.1 Introduction
This chapter discusses the mechanical equilibrium of incompressible fluids, a topic
that is conventionally termed hydrostatics. More information on the fundamentals of
hydrostatics can be found in Lamb 1928. More information on Maclaurin spheroids,
Jacobi ellipsoids, and Roche ellipsoids can be found in Chandrasekhar 1969, and
Lamb 1993.
(2.1)
where p is the static fluid pressure, the mass density, = g z the gravitational
potential energy per unit mass, and g the (approximately uniform) acceleration due
to gravity. Now,
0
z>0
,
(2.2)
(z) =
z0
0
where 0 is the (approximately uniform) mass density of water. Here, the comparatively small mass density of air has been neglected. Because = (z) and = (z),
it immediately follows, from Equation (2.1), that p = p(z), where
dp
= g.
dz
(2.3)
We conclude that constant pressure surfaces in a stationary body of water take the
form of horizontal planes. Making use of Equation (2.2), the previous equation can
33
34
be integrated to give
p(z) =
z>0
,
z0
p0
p0 0 g z
(2.4)
2.3 Buoyancy
Consider the air/water system described in the previous section. Let V be some volume, bounded by a closed surface S , that straddles the plane z = 0, and is thus partially occupied by water, and partially by air. The i-component of the net force acting
on the fluid (i.e., either water or air) contained within V is written (see Section 1.3)
i j dS j +
Fi dV,
(2.5)
fi =
S
where
i j = p i j
(2.6)
is the stress tensor for a static fluid (see Section 1.5), and
F = g ez
(2.7)
the gravitational force density. (Recall that the indices 1, 2, and 3 refer to the x-, y-,
and z-axes, respectively. Thus, f3 fz , et cetera.) The first term on the right-hand
side of Equation (2.5) represents the net surface force acting across S , whereas the
second term represents the net volume force distributed throughout V. Making use
of the tensor divergence theorem (see Section B.4), Equations (2.5)(2.7) yield the
following expression for the net force:
f = B + W,
where
Bi =
p
dV,
xi
(2.8)
(2.9)
and
W x = Wy = 0,
g dV.
Wz =
V
(2.10)
(2.11)
Hydrostatics
35
Here, B is the net surface force, and W the net volume force.
It follows from Equations (2.4) and (2.9) that
B = M0 g ez ,
(2.12)
where M0 = 0 V0 . Here, V0 is the volume of that part of V which lies below the
waterline, and M0 the total mass of water contained within V. Moreover, from Equations (2.2), (2.10), and (2.11),
W = M0 g ez.
(2.13)
It can be seen that the net surface force, B, is directed vertically upward, and exactly balances the net volume force, W, which is directed vertically downward. Of
course, W is the weight of the water contained within V. On the other hand, B,
which is generally known as the buoyancy force, is the resultant pressure of the water immediately surrounding V. We conclude that, in equilibrium, the net buoyancy
force acting across S exactly balances the weight of the water inside V, so that the
total force acting on the contents of V is zero, as must be the case for a system in
mechanical equilibrium. We can also deduce that the line of action of B (which is
vertical) passes through the center of gravity of the water inside V. Otherwise, a net
torque would act on the contents of V, which would contradict our assumption that
the system is in mechanical equilibrium.
(2.14)
where W0 = M0 g and M0 are the weight and mass of the displaced water, respectively. The fact that the buoyancy force is unchanged also implies that the vertical
line of action of B passes through the center of gravity, H (say), of the displaced
water. Incidentally, H is generally known as the center of buoyancy.
36
(2.15)
and the buoyancy force, B = W0 ez , due to the pressure of the surrounding water.
Of course, the line of action of W passes through the bodys center of gravity, G
(say). So as to remain in equilibrium, the body must be subject to zero net force and
zero net torque. The requirement of zero net force yields W0 = W. In other words,
in equilibrium, the weight of the water displaced by a floating body is equal to the
weight of the body, or, alternatively, in equilibrium, the magnitude of the buoyancy
force acting on a floating body is equal to the weight of the displaced water. This
famous result is known as Archimedes principle, after Archimedes of Syracuse (c.
278212 BCE). The requirement of zero net torque implies that, in equilibrium, the
center of gravity, G, and center of buoyancy, H, of a floating body must lie on the
same vertical straight-line.
Consider a floating body of mass M and volume V. Let = M/V be the bodys
mean mass density. Archimedes principle implies that, in equilibrium,
V0
= s,
V
where
s=
(2.16)
(2.17)
is termed the bodys specific gravity. (Recall, that V0 is the submerged volume, and
0 the mass density of water.) We conclude, from Equation (2.16), that the volume
fraction of a floating body that is submerged is equal to the bodys specific gravity.
Obviously, the specific gravity must be less than unity, because the submerged volume fraction cannot exceed unity. In fact, if the specific gravity exceeds unity then
it is impossible for the buoyancy force to balance the bodys weight, and the body
consequently sinks.
Consider a body of volume V and specific gravity s that floats in equilibrium.
It follows, from Equation (2.16), that the submerged volume is V0 = s V. Hence,
the volume above the waterline is V1 = V V0 = (1 s) V. Suppose that the body
is inverted such that its previously submerged part is raised above the waterline, and
vice versa: that is, V0 V1 . According to Equation (2.16), the body can only remain
in equilibrium in this configuration if its specific gravity is
s =
V1 V V0
=
= 1 s.
V
V
(2.18)
Hydrostatics
37
(2.20)
38
G
M
H
STABLE
H
UNSTABLE
Figure 2.1
Stable and unstable configurations for a floating body.
vertical, straight-line that passes through the bodys center of gravity, G, and original
center of buoyancy, H. Owing to the altered shape of the volume of displaced water,
the center of buoyancy is shifted to some new position H . Let the vertical straightline passing through H meet GH at M. (See Figure 2.1.) Point M (in the limit
0) is called the metacenter. In the disturbed state, the bodys weight W acts
downward through G, and the buoyancy 0 V0 g acts upward through M. Let us
assume that the submerged volume, V0 , is unchanged from the equilibrium state
(which excludes vertical oscillations from consideration). It follows that the weight
and the buoyancy force are equal and opposite, so that there is no net force on the
body. However, as can be seen from Figure 2.1, the weight and the buoyancy force
generate a net torque of magnitude = W sin . Here, is the length MG: that
is, the distance between the metacenter and the center of gravity. This distance is
generally known as the metacentric height, and is defined such that it is positive
when M lies above G, and vice versa. Moreover, as is also clear from Figure 2.1,
when M lies above G the torque acts to reduce , and vice versa. In the former case,
the torque is known as a righting torque. We conclude that a floating body is stable
to small angular displacements about some horizontal axis lying in the plane z = 0
provided that its metacentric height is positive: that is, provided that the metacenter
lies above the center of gravity. Because we have already demonstrated that a floating
body is unconditionally stable to small vertical displacements (and as it is also fairly
obvious that such a body is neutrally stable to both horizontal displacements and
angular displacements about a vertical axis passing through its center of gravity), it
follows that a necessary and sucient condition for the stability of a floating body to
a general small perturbation (made up of arbitrary linear and angular components) is
Hydrostatics
39
that its metacentric height be positive for angular displacements about any horizontal
axis.
A0
A0
where the integrals are taken over the whole cross-section at z = 0. Let (x, y) be
the bodys draft: that is, the vertical distance between the surface of the water and
the bodys lower boundary. It follows, from symmetry, that (x, y) = (x, y) and
(x, y) = (x, y). Moreover, the submerged volume is
dx dy dz =
(x, y) dx dy.
(2.23)
V0 =
A0
A0
(2.24)
A0
The depth of the unperturbed center of buoyancy below the surface of the water is
z dx dy dz
2 A0
1
A 0
h= 0
=
2 (x, y) dx dy = 0 ,
(2.25)
V0
2 V0 A 0
2 V0
where
1/2
A 2 (x, y) dx dy
0
.
0 =
A0
(2.26)
A0
where use has been made of Equations (2.22) and (2.23). Thus, the submerged volume is unchanged, as should be the case for a purely angular displacement. The new
40
h =
A0 0
z dx dy dz
V0
1
2 V0
2 (x, y) 2 y (x, y) + O( 2 ) dx dy = h,
A0
(2.28)
where use has been made of Equations (2.24) and (2.25). Thus, the depth of the
center of buoyancy is also unchanged. Moreover, from symmetry, it is clear that the
center of buoyancy still lies at x = 0. Finally, the new y-coordinate of the center of
buoyancy is
A0 0
y dx dy dz
V0
=
A0
y [(x, y) y] dx dy
V0
x2 A0
,
V0
(2.29)
A0
(2.30)
x2 A0
+ zH ,
V0
(2.31)
x2 A0
zGH ,
V0
(2.32)
where zGH = zG zH . Note that, because x2 A0 /V0 > 0, the metacenter always lies
above the center of buoyancy.
A simple extension of the previous argument reveals that if the body turns through
a small angle about the y-axis then the metacentric height is
=
y2 A0
V0
zGH ,
(2.33)
Hydrostatics
41
where
1/2
2
A x dx dy
0
,
y =
A0
(2.34)
(2.36)
Thus, the minimum value of 2 is the lesser of x2 and y2 . It follows that the equilibrium state in question is unconditionally stable provided it is stable to small amplitude angular displacements about horizontal axes normal to its two vertical symmetry
planes (i.e., the x = 0 and y = 0 planes).
As an example, consider a uniform rectangular block of specific gravity s floating
such that its sides of length a, b, and c are parallel to the x-, y-, and z-axes, respectively. Such a block can be thought of as a very crude model of a ship. The volume
of the block is V = a b c. Hence, the submerged volume is V0 = s V = s a b c. The
cross-sectional area of the block at the waterline (z = 0) is A0 = a b. It is easily
demonstrated that (x, y) = 0 = V0 /A0 = s c. Thus, the center of buoyancy lies a
depth h = 02 A0 /2 V0 = s c/2 below the surface of the water. [See Equation (2.25).]
Moreover, by symmetry, the center of gravity is a height c/2 above the bottom surface of the block, which is located a depth s c below the surface of the water. Hence,
zH = h = s c/2, zG = c/2 s c, and zGH = c (1 s)/2. Consider the stability of
the block to small amplitude angular displacements about the x-axis. We have
a/2 b/2
x2
a/2 b/2
y 2 dx dy
ab
b2
.
12
(2.37)
b2
c
(1 s).
12 s c 2
(2.38)
(2.39)
Because the maximum value that s (1 s) can take is 1/4, it follows that the block is
stable for all specific gravities when
c < c0 =
2
b.
3
(2.40)
42
On the other hand, if c > c0 then the block is unstable for intermediate specific
gravities such that s < s < s+ , where
1 1 c02 /c 2
,
(2.41)
s =
2
and is stable otherwise. Assuming that the block is stable, its angular equation of
motion is written
d 2
I 2 = W sin W ,
(2.42)
dt
where
a/2 b/2 cs c
W 2
W
(y 2 + z 2 ) dx dy dz =
I=
b + 4 [(1 s)3 + s 3 ] c 2
g V a/2 b/2 s c
12 g
(2.43)
is the moment of inertia of the block about the x-axis. Thus, we obtain the the simple
harmonic equation
d 2
= 2 ,
(2.44)
dt 2
where
c02 4 s (1 s) c 2
g
W
=
.
(2.45)
2 =
I
s c c02 + (8/3) [(1 s)3 + s 3 ] c 2
We conclude that the block executes small amplitude angular oscillations about the xaxis at the angular frequency . For the case of rotation about the y-axis, the previous
analysis is unchanged except that a b. The previous analysis again neglects the
phenomenon of added mass, and, therefore, underestimates the eective inertia of
the block. (See Sections 5.9 and 7.10.)
The metacentric height of a conventional ship whose length greatly exceeds its
width is typically much less for rolling (i.e., rotation about a horizontal axis running
along the ships length) than for pitching (i.e., rotation about a horizontal axis perpendicular to the ships length), because the radius of gyration for pitching greatly
exceeds that for rolling. As is clear from Equation (2.45), a ship with a relatively
small metacentric height (for rolling) has a relatively long roll period, and vice versa.
An excessively low metacentric height increases the chances of a ship capsizing if
the weather is rough, if its cargo/ballast shifts, or if the ship is damaged and partially
flooded. For this reason, maritime regulatory agencies, such as the International Maritime Organization, specify minimum metacentric heights for various dierent types
of sea-going vessel. A relatively large metacentric height, on the other hand, generally renders a ship uncomfortable for passengers and crew, because the ship executes
short period rolls, resulting in large g-forces. Such forces also increase the risk that
cargo may break loose or shift.
We saw earlier, in Section 2.4, that if a body of specific gravity s floats in vertical equilibrium in a certain position then a body of the same shape, but of specific
gravity 1 s, can float in vertical equilibrium in the inverted position. We shall now
demonstrate that these positions are either both stable, or both unstable, provided the
Hydrostatics
43
body is of uniform density. Let V1 and V2 be the volumes that are above and below
the waterline, respectively, in the first position. Let H1 and H2 be the mean centers of
these two volumes, and H that of the whole volume. It follows that H2 is the center
of buoyancy in the first position, H1 the center of buoyancy in the second (inverted)
position, and H the center of gravity in both positions. Moreover,
V1 V2
H1 H2 ,
(2.46)
V 1 H1 G = V 2 H2 G =
V1 + V2
where H1G is the distance between points H1 and G, et cetera. The metacentric
heights in the first and second positions are
V1 V2
2 A
1
1 =
H1 G =
A 2
H1 H2 ,
(2.47)
V1
V1
V1 + V2
2 A
1
V1 V2
2
A
H1 H2 ,
(2.48)
H2 G =
2 =
V2
V2
V1 + V2
respectively, where A and are the area and radius of gyration of the common waterline section, respectively. Thus,
2
V1 V2
1
2
1 2 =
A
H1 H2 0,
V1 V2
V1 + V2
(2.49)
which implies that 1 0 as 2 0, and vice versa. It follows that the first and
second positions are either both stable, both marginally stable, or both unstable.
44
r
p
H0
H
B
Figure 2.2
Curve of buoyancy for a floating body.
an equilibrium is only stable when the potential energy is minimized. Thus, it follows
that a stable equilibrium configuration of a floating body is such as to minimize the
height of the bodys center of gravity relative to its center of buoyancy.
Hydrostatics
45
and the center of buoyancy is equal to the perpendicular distance, p, between G and
the tangent to the curve of buoyancy at H. An equilibrium configuration therefore
corresponds to a maximum or a minimum of p as point H moves along the curve
of buoyancy. However, the equilibrium is only stable if p is minimized. If R is the
radius of curvature of the curve of buoyancy then, according to a standard result in
dierential calculus (Lamb 1928),
dr
.
dp
(2.50)
R
p,
r
(2.51)
R=r
Writing this result in the form
r =
it can be seen that maxima and minima of p, which are the points on the curve of
buoyancy where p = 0, correspond to the points where r = 0, and are, thus, coincident with maxima and minima of r. In other words, an equilibrium configuration
corresponds to a point of maximum or minimum r on the curve of buoyancy: that is,
a point at which GH meets the curve at right-angles. At such a point, r = p, and the
potential energy consequently takes the value W r.
Let H0 be a point on the curve of buoyancy, and let r0 , p0 , and R0 be the corresponding values of r, p, and R. For neighboring points on the curve, we can write
dr
(p p0 ),
r r0 =
(2.52)
d p H0
or
r r0 =
R0
(p p0 ).
r0
(2.53)
It follows that p p0 has the same sign as r r0 (because R0 and r0 are both positive).
[The fact that R0 is positive (i.e., dr/d p > 0) follows from the previously established
result that the metacenter, which is the center of curvature of the curve of buoyancy,
always lies above the center of buoyancy, implying that the curve of buoyancy is
necessarily concave upwards.] Hence, the minima and maxima of r occur simultaneously with those of p. Consequently, a stable equilibrium configuration corresponds
to a point of minimum r on the curve of buoyancy: that is, a minimum in the distance
GH between the center of gravity and the center of buoyancy.
We can use the previous result to determine the stable equilibrium configurations
for a beam of square cross-section, and uniform specific gravity s, that floats with its
length horizontal. In order to achieve this goal, we must calculate the distance GH for
all possible configurations of the beam that are in vertical force balance. However,
we need only consider cases where s < 1/2, because, according to the analysis of
Section 2.6, for every stable equilibrium configuration with s = s0 < 1/2 there is a
corresponding stable inverted configuration with s = 1 s0 > 1/2, and vice versa.
Let us define fixed rectangular axes, x and y, passing through the center of the
middle section of the beam, and running parallel to its sides. Let us start with the
46
x
Q
P
P
B
y
Figure 2.3
Beam of square cross-section floating with two corners immersed.
case where the waterline PQ is parallel to a side. (See Figure 2.3.) If the length of a
side is 2 a then Equation (2.16) yields
AP = BQ = 2 a s.
(2.54)
Suppose that the beam is turned through an angle > 0 such that the waterline
assumes the position P Q in Figure 2.3, but still intersects two opposite sides. The
lengths AP and BQ satisfy
BQ AP = 2 a tan .
(2.55)
Moreover, the area of the trapezium P ABQ must match that of the rectangle PABQ
in order to ensure that the submerged volume remains invariant (otherwise, the beam
would not remain in vertical force balance): that is,
(AP + BQ ) a = 4 a 2 s.
(2.56)
AP = a (2 s tan ),
(2.57)
It follows that
BQ = a (2 s + tan ).
(2.58)
The constraint that the waterline intersect two opposite sides of the beam implies that
AP > 0, and, hence, that
tan < 2 s.
(2.59)
Hydrostatics
47
The coordinates of the center of buoyancy, H, which is the mean center of the trapezium P ABQ , are
a a
x dx dy (2/3) a 3 tan
a
a ah(x)
tan ,
(2.60)
x = a a
=
=
2
6
s
4a s
dx dy
a ah(x)
a a
y dx dy
a ah(x)
y = a a
dx dy
a ah(x)
=
4 a 3 s (1 s) (1/3) a 3 tan2
a
tan2 ,
= (1 s) a
12 s
4 a2 s
(2.61)
where
h(x) = 2 a s + x tan .
(2.62)
2
t2
t2
+
(1
s)
,
12 s
36 s 2
(2.63)
t 2 12 s (1 s) + 2 ,
(2.64)
3 t 2 12 s (1 s) + 2 .
(2.65)
The minima and maxima of u occur when du/dt = 0, d 2 u/dt 2 > 0 and du/dt = 0,
d 2 u/dt 2 < 0, respectively. It follows that the symmetrical position, t = 0, in which
the sides of the beam are either parallel or perpendicular to the waterline, is always
an equilibrium, but is only stable when
s2 s +
1
>0:
6
(2.66)
that is, when s < 1/2 1/ 12 = 0.2113. It is also possible to obtain equilibria in
asymmetric positions such that t is the root of
t 2 = 12 s (1 s) 2.
(2.67)
Such equilibria only exist for s > 0.2113, and are stable. Finally, in order to satisfy
the constraint (2.59), we must have t < 2 s, which, in combination with the previous
equation, implies that
(2.68)
8 s 2 6 s + 1 > 0,
or s < 0.25.
48
Q
O
B
y
Figure 2.4
Beam of square cross-section floating with one corner immersed.
Suppose that the constraint (2.59) is not satisfied, so that the immersed portion of
the beams cross-section is triangular. (See Figure 2.4.) It is clear that
BQ
= tan .
BP
(2.69)
Moreover, the area of the triangle P BQ in Figure 2.4 must match that of the rectangle PABQ in Figure 2.3, in order to ensure that the submerged volume remain
invariant: that is,
1
BP BQ = 4 a 2 s.
(2.70)
2
It follows that
BP = (8 s/ tan )1/2 a,
(2.71)
(2.72)
(2.73)
BQ = 3 z a.
(2.74)
The coordinates of the center of buoyancy, H, which is the mean center of triangle
Hydrostatics
49
P BQ , are
x = a BP /3 = a (1 z 1 ),
y = a BQ /3 = a (1 z),
(2.75)
(2.76)
because the perpendicular distance of the mean center of a triangle from one of its
sides is one third of the perpendicular distance from the side to the opposite vertex.
Thus, if u = r 2 /a 2 = ( x 2 + y 2 )/a 2 then
u = (1 z1 )2 + (1 z)2 ,
du 2 2 (z 2 1) (z 2 1 z + 1)
=
,
dz
z3
d 2 u 2 2 (z 4 2 1 z + 3)
=
.
dz 2
z4
(2.77)
(2.78)
(2.79)
3
.
2
(2.80)
The stable and unstable equilibria correspond to du/dz = 0, d 2 u/dz 2 > 0 and du/dz,
d 2 u/dz 2 < 0, respectively. It follows that the symmetrical position, z = 1, in which
the diagonals of the beam are either parallel or perpendicular to the waterline is an
equilibrium provided < 2/3, or s < 1/2, but is only stable when > 1/2, or
s > 9/32 = 0.28125. It is also possible to obtain equilibria in asymmetric positions
such that z is the root of
(2.81)
z 2 1 z + 1 = 0.
Such equilibria only exist for 2/3 < < 1/2, or 1/4 < s < 9/32, and are stable.
In summary, the stable equilibrium configurations of a beam of square crosssection, floating with its length horizontal, are such that the sides are either parallel
or perpendicular to the waterline for s < 0.2113, such that two corners are immersed
but the sides and diagonals are neither parallel nor perpendicular to the waterline for
0.2113 < s < 0.25, such that only one corner is immersed but the sides and diagonals
are neither parallel nor perpendicular to the waterline for 0.25 < s < 0.28125, and
such that the diagonals are either parallel or perpendicular to waterline for 0.28125 <
s < 0.5. For s > 0.5, the stable configurations are the same as those for a beam with
the complimentary specific gravity 1 s.
50
(2.82)
where p is the static fluid pressure, the mass density, the gravitational potential
energy per unit mass, and r a position vector (measured with respect to an origin
that lies on the axis of rotation). The final term on the right-hand side of the previous
equation represents the fictitious centrifugal force density. Without loss of generality,
we can assume that = ez . It follows that
0 = p + ( + ),
(2.83)
where
1
(2.84)
= 2 (x 2 + y 2 )
2
is the so-called centrifugal potential. Recall, incidentally, that is a uniform constant
in an incompressible fluid.
As an example, consider the equilibrium of a body of water, located on the Earths
surface, that is uniformly rotating about a vertical axis at the fixed angular velocity
. It is convenient to adopt cylindrical coordinates (see Section C.3), r, , z, whose
symmetry axis coincides with the axis of rotation. Let z increase upward. It follows
that = g z and = (1/2) 2 r 2 . Assuming that the pressure distribution is
axisymmetric, so that p = p(r, z), the force balance equation, (2.83), reduces to
p
( + )
+
= 0,
r
r
( + )
p
+
= 0,
z
z
(2.85)
(2.86)
or
p
2 r = 0,
r
p
+ g = 0.
z
The previous two equations can be integrated to give
1 2 2
p(r, z) = p0 + r g z ,
2
(2.87)
(2.88)
(2.89)
Hydrostatics
51
pressure matches that of the atmosphere. It follows that the surface of the water is
the paraboloid of revolution
2 2
z=
r ,
(2.90)
2g
where r is the perpendicular distance from the axis of rotation, and z = 0 the on-axis
height of the surface.
According to the analysis of Section 2.3, the buoyancy force acting on any corotating solid body, which is wholly or partially immersed in the water, is the same
as that which would maintain the mass of water displaced by the body in relative
equilibrium. In the case of a floating body, this mass is limited by the continuation of
the waters curved surface through the body. Let points G and H represent the centers
of gravity and buoyancy, respectively, of the body. Of course, the latter point is
simply the center of gravity of the displaced water. Suppose that G and H are located
perpendicular distances rG and rH from the axis of rotation, respectively. Finally,
let M be the mass of the body, and M0 the mass of the displaced water. It follows
that the buoyancy force has an upward vertical component M0 g, and an outward
horizontal component M0 2 rH . Thus, according to standard Newtonian dynamics
(Fitzpatrick 2012), the equation of horizontal motion of a general co-rotating body is
..
M (r 2 rG ) = M0 2 rH ,
(2.91)
..
r = 2 (rH rG ) < 0.
(2.92)
In other words, a floating body drifts radially inward towards the rotation axis. On
the other hand, M0 < M for a fully submerged body that is more dense than water.
However, if the body is of uniform density then its centers of gravity and buoyancy
coincide with one another, so that rH = rG . Hence, we obtain
..
r = (M M0 ) 2 rG > 0.
(2.93)
In other words, a fully submerged body drifts radially outward from the rotation axis.
The previous analysis accounts for the common observation that objects heavier than
water, such as grains of sand, tend to collect on the outer side of a bend in a fast
flowing river, while floating objects, such as sticks, tend to collect on the inner side.
52
bounding surface? This famous theoretical problem had its origins in investigations
of the figure of a rotating planet, such as the Earth, that were undertaken by Newton, Maclaurin, Jacobi, Meyer, Liouville, Dirichlet, Dedekind, Riemann, and other
celebrated scientists, in the 17th, 18th, and 19th centuries (Chandrasekhar 1969).
Incidentally, it is reasonable to treat the Earth as a liquid, for the purpose of this
calculation, because the shear strength of the solid rock out of which the terrestrial
crust is composed is nowhere near sucient to allow the actual shape of the Earth to
deviate significantly from that of a hypothetical liquid Earth (Fitzpatrick 2012).
In a co-rotating reference frame, the shape of a self-gravitating, rotating, liquid
planet is determined by a competition between fluid pressure, gravity, and the fictitious centrifugal force. The latter force opposes gravity in the plane perpendicular
to the axis of rotation. Of course, in the absence of rotation, the planet would be
spherical. Thus, we would expect rotation to cause the planet to expand in the plane
perpendicular to the rotation axis, and to contract along the rotation axis (in order to
conserve volume).
For the sake of simplicity, we shall restrict our investigation to a rotating planet
of uniform density whose outer boundary is ellipsoidal. An ellipsoid is the threedimensional generalization of an ellipse. Let us adopt the right-handed Cartesian
coordinate system x1 , x2 , x3 . An ellipse whose principal axes are aligned along the
x1 - and x2 -axes satisfies
x12
a12
x22
a22
= 1,
(2.94)
where a1 and a2 are the corresponding principal radii. Moreover, as is easily demonstrated,
dA = a1 a2 ,
A=
xi2 dA =
1 2
a A,
4 i
x1 x2 dA = 0,
(2.95)
(2.96)
(2.97)
where A is the area, dA an element of A, and the integrals are taken over the whole
interior of the ellipse. Likewise, an ellipsoid whose principal axes are aligned along
the x1 -, x2 -, and x3 -axes satisfies
x12
a12
x22
a22
x32
a32
= 1,
(2.98)
Hydrostatics
53
demonstrated,
V=
dV =
4
a1 a2 a3 ,
3
1 2
a V,
5 i
x2 x3 dV = 0,
x1 x2 dV =
xi2 dV =
(2.99)
(2.100)
(2.101)
where V is the volume, dV an element of V, and the integrals are taken over the
whole interior of the ellipsoid.
Suppose that the planet is rotating uniformly about the x3 -axis at the fixed angular
velocity . The planets moment of inertia about this axis is [cf., Equation (2.100)]
I33 =
1
M (a12 + a22 ),
5
(2.102)
1
M (a12 + a22 ) ,
5
(2.103)
1
1
I33 2 =
M (a12 + a22 ) 2 .
2
10
(2.104)
According to Equations (2.83) and (2.84), the fluid pressure distribution within
the planet takes the form
1
p = p0 2 (x12 + x22 ) ,
(2.105)
2
where is the gravitational potential (i.e., the gravitational potential energy of a
unit test mass) due to the planet, = M/V the uniform planetary mass density, and
p0 a constant. However, it is demonstrated in Appendix D that the gravitational potential inside a homogeneous self-gravitating ellipsoidal body can be written (Chandrasekhar 1969; Lamb 1993)
3
2
= G M 0
i xi ,
(2.106)
4
i=1,3
where G is the universal gravitational constant (Yoder 1995), and
du
,
0 =
0
du
i =
,
2
0 (ai + u)
= (a12 + u)1/2 (a22 + u)1/2 (a32 + u)1/2 .
(2.107)
(2.108)
(2.109)
54
Thus, we obtain
3
3
1
3
p = p0
G M 1 2 x12 + G M 2 2 x22 + G M 3 x32 , (2.110)
2
2
2
2
where p0 is the central fluid pressure. The pressure at the planets outer boundary
must be zero, otherwise there would be a force imbalance across the boundary. In
other words, we require
3
3
1
3
2
2
2
2
2
G M 1 x1 + G M 2 x2 + G M 3 x3 = p0 , (2.111)
2
2
2
2
whenever
x12
a12
x22
a22
x32
a32
= 1.
(2.112)
(2.113)
(2.114)
a22 )
du
= 0,
2
2
2
(a1 + u) (a2 + u) (a3 + u)
a12 a22
a32
(2.115)
3
1
M (a12 + a22 ) 2
G M 2 0 .
10
10
(2.117)
Hydrostatics
55
(2.118)
Thus, if e13 = 0 then there is no flattening, and the planet is consequently spherical,
whereas if e13 1 then the flattening is complete, and the planet consequently
collapses to a disk in the x1 -x2 plane.
2
Let u = a12 and = e13
/z 2 1. Setting a2 = a1 in Equation (2.114), we obtain
2
d
2 1/2 2
= (1 e13 ) e13
2 (1 + e 2 )3/2
2 G
(1
+
)
0
13
e13
2 1/2 e13
2
2 (1 e13 )
z dz
z 2 dz
2
=
(1 e13 )
. (2.119)
3
(1 z 2 )1/2
(1 z 2 )3/2
e13
0
0
Performing the integrals, which are standard (Speigel, Liu, and Lipschutz 1999), we
find that
2
3 2 e13
3
2
2 1/2
1
2
2 (1 e13
=
)
sin
e
).
(2.120)
(1
e
13
13
3
2 G
e13
e13
This famous result was first obtained by Colin Maclaurin (1698-1746) in 1742. Finally, in order to calculate the potential energy, (2.116), we need to evaluate
e13
1
2
d
dz
=
0 =
2
1/2
a1 0 (1 + ) (1 + e13 )
a1 e13 0 (1 z 2 )1/2
=
2 sin1 e13
.
a1
e13
(2.121)
2
2
3 cos ,
(2.122)
=
(1 + 2 cos )
sin
sin2
1/2
6 (cos )1/6
3 cos
L=
,
(2.123)
(1 + 2 cos2 )
5
sin
sin
3 (cos )1/3
2
+ 3 cos .
(2.124)
E=
(1 4 cos )
10 sin2
sin
56
0.00000
0.02582
0.05168
0.07758
0.10357
0.12967
0.15591
0.18231
0.20889
0.23567
0.26267
0.28989
L
0.00000
0.01266
0.02540
0.03830
0.05144
0.06491
0.07882
0.09329
0.10846
0.12450
0.14163
0.16013
E
0.60000
0.59980
0.59919
0.59817
0.59672
0.59479
0.59236
0.58936
0.58572
0.58135
0.57612
0.56986
e13
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
0.96
0.97
0.98
0.99
0.31729
0.34484
0.37239
0.39967
0.42612
0.45046
0.46932
0.47045
0.46472
0.45418
0.43475
0.39389
L
0.18037
0.20286
0.22834
0.25792
0.29345
0.33833
0.39994
0.50074
0.53194
0.57123
0.62486
0.71209
E
0.56233
0.55320
0.54200
0.52800
0.51001
0.48587
0.45107
0.39272
0.37485
0.35273
0.32351
0.27916
Table 2.1
Properties of the Maclaurin spheroids.
These properties are set out in Table 2.1.
In the limit, 0, in which the planet is relatively slowly rotating (i.e., 1),
and its degree of flattening consequently slight, Equations (2.122)(2.124) reduce to
15
,
(2.125)
e13
2
6
,
(2.126)
L
5
3
E .
(2.127)
5
In other words, in the limit of relatively slow rotation, when the planet is almost
spherical, its eccentricity becomes directly proportional to its angular velocity. In
this case, it is more conventional to parameterize angular velocity in terms of
m=
2 a0 3 2
= ,
g0
2
(2.128)
5
m.
4
(2.130)
Hydrostatics
57
0.2
2
0.1
Figure 2.5
Normalized angular velocity squared of a Maclaurin spheroid (solid) and a Jacobi
ellipsoid (dashed) versus the eccentricity e13 in the x1 -x3 plane.
For the case of the Earth ( = 7.27 105 rad. s1 , a0 = 6.37 106 m, g0 =
9.81 m s1Yoder 1995), we obtain
m
1
.
291
(2.131)
Thus, it follows that, were the Earth homogeneous, its figure would be a spheroid,
flattened at the poles, of ellipticity
1
5 1
.
4 291 233
(2.132)
This result was first obtained by Newton. The actual ellipticity of the Earth is about
1/294 (Yoder 1995), which is substantially smaller than Newtons prediction. The
discrepancy is due to the fact that the Earth is strongly inhomogeneous, being much
denser at its core than in its outer regions.
Figures 2.5 and 2.6 illustrate the variation of the normalized angular velocity,
, and angular momentum, L, of a Maclaurin spheroid with its eccentricity, e13 ,
as predicted by Equations (2.122)(2.124). It can be seen, from Figure 2.5, that
there is a limit to how large the normalized angular velocity of such a spheroid can
become. The limiting value corresponds to = 0.47399, and occurs when e13 =
0.92995. For values of lying below 0.47399 there are two possible Maclaurin
spheroids, one with an eccentricity less than 0.92995, and one with an eccentricity
58
1.2
1.1
1
0.9
0.8
0.7
L 0.6
0.5
0.4
0.3
0.2
0.1
0
Figure 2.6
Normalized angular momentum of a Maclaurin spheroid (solid) and a Jacobi ellipsoid (dashed) versus the eccentricity e13 in the x1 -x3 plane.
greater than 0.92995. Note, however, from Figure 2.6, that despite the fact that the
angular velocity, , of a Maclaurin spheroid varies in a non-monotonic manner with
the eccentricity, e13 , the angular momentum, L, increases monotonically with e13 ,
becoming infinite in the limit e13 1. It follows that there is no upper limit to the
angular momentum of a Maclaurin spheroid.
a12 a22
(a12 + u) (a22 + u)
du
= 0.
2
(a3 + u)
a32
(2.133)
(2.134)
a3 = a1 cos ,
(2.135)
Hydrostatics
59
where . It follows that the cross-sections of the planets outer boundary in the
x1 -x2 and x1 -x3 planes are ellipses of eccentricities
e12 = (1 a32 /a22 )1/2 = sin ,
(2.136)
(2.137)
(2.138)
(2.139)
F(, ) E(, ) cos E(, )
cos2
+
,
=2
tan sin tan
tan3 cos2 tan2 cos2
(2.140)
and
respectively, where
E(, ) =
(2.141)
d
,
(1 sin2 sin2 )1/2
(2.142)
F(, ) =
are special functions known as incomplete elliptic integrals (Abramowitz and Stegun
1965). The integral 0 , defined in Equation (2.107), transforms to
0 =
(2.143)
Finally, making use of some of the analysis in the previous two sections, the normalized angular momentum, and normalized mechanical energy, of the planet can be
written
6 1 + cos2
L=
,
(2.144)
10 (cos cos )2/3
E=
3
1 + cos2
3 (cos cos )1/3
F(, ) +
2,
5
sin
20 (cos cos )2/3
(2.145)
respectively.
The constraint (2.139) is obviously satisfied in the limit 0, because this
60
e12
e13
e12
e13
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.81267
0.81293
0.81372
0.81504
0.81691
0.81934
0.82237
0.82603
0.83037
0.83544
0.84131
0.84808
0.43257
0.43257
0.43257
0.43256
0.43253
0.43248
0.43237
0.43220
0.43191
0.43146
0.43078
0.42976
0.30375
0.30375
0.30375
0.30377
0.30380
0.30388
0.30402
0.30427
0.30468
0.30532
0.30628
0.30772
0.50452
0.50459
0.50459
0.50458
0.50457
0.50453
0.50445
0.50432
0.50410
0.50376
0.50326
0.50250
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
0.96
0.97
0.98
0.99
0.85585
0.86480
0.87510
0.88705
0.90102
0.91761
0.93778
0.96340
0.96950
0.97605
0.98317
0.99101
0.42827
0.42609
0.42288
0.41807
0.41069
0.39879
0.37787
0.33353
0.31776
0.29691
0.26722
0.21809
0.30984
0.31296
0.31760
0.32462
0.33562
0.35390
0.38783
0.46860
0.50078
0.54672
0.62003
0.76872
0.50138
0.49975
0.49734
0.49372
0.48814
0.47908
0.46295
0.42782
0.41499
0.39771
0.37241
0.32842
Table 2.2
Properties of the Jacobi ellipsoids.
Hydrostatics
61
0.6
0.5
0.4
0.3
0.2
0.1
0
0.2
0.4
0.6
0.8
1.2
Figure 2.7
Normalized mechanical energy of a Maclaurin spheroid (solid) and a Jacobi ellipsoid
(dashed) versus the normalized angular momentum.
ing, liquid planet, the relevant constraints are that the mass, volume, and net angular momentum of the system cannot spontaneously change. Thus, we expect such
a planet to evolve toward the equilibrium state with the lowest energy for a given
mass, volume, and angular momentum. This suggests, from Figure 2.7, that at relatively high angular momentum (i.e., L > 0.30375, e13 > 0.81267), when the Jacobi
ellipsoid solutions exist, they are stable equilibrium states (because there is no lower
energy state to which the system can evolve), whereas the Maclaurin spheroids are
unstable. On the other hand, at relatively low angular momentum (i.e., L < 0.30375,
e13 < 0.81267), when there are no Jacobi ellipsoid solutions, the Maclaurin spheroids
are stable equilibrium states (again, because there is no lower energy state to which
they can evolve). These predictions are borne out by the results of direct stability
analysis performed on the Maclaurin spheroids and Jacobi ellipsoids (Chandrasekhar
1969). In fact, such stability studies demonstrate that the Maclaurin spheroids are
unstable in the presence of weak dissipation for e13 > 0.81267, and unconditionally
unstable for e13 > 0.95289. The Jacobi ellipsoids, on the other hand, are unconditionally stable for e13 < 0.93858, but are unconditionally unstable for e13 > 0.93858,
evolving toward lower energy pear shaped equilibria (which are, themselves, unstable in the presence of weak dissipation).
62
G (M + M )
.
R3
(2.146)
In other words, in the inertial frame, the moon and the planet orbit in a fixed plane
about their common center of mass at the angular velocity . It is convenient to
transform to a non-inertial reference frame that rotates (with respect to the inertial
frame), about an axis passing through C , at the angular velocity . It follows that
points C, C , and C appear stationary in this frame. It is also convenient to adopt the
standard right-handed Cartesian coordinates, x1 , x2 , x3 , and to choose the coordinate
axes such that = e3 , C = (0, 0, 0), C = (R, 0, 0), and C = ([M /(M +
M )] R, 0, 0). Thus, in the non-inertial reference frame, the orbital rotation axis runs
parallel to the x3 -axis, and the centers of the moon and the planet both lie on the
x1 -axis.
Suppose that the moon does not rotate (about an axis passing through its center
of mass) in the non-inertial reference frame. This implies that, in the inertial frame,
the moon appears to rotate about an axis parallel to the x3 -axis, and passing through
C, at the same angular velocity as it orbits about C . This type of rotation is termed
synchronous, and ensures that the same hemisphere of the moon is always directed
toward the planet. Such rotation is fairly common in the solar system. For instance,
the Moon rotates synchronously in such a manner that the same hemisphere is always
visible from the Earth. Synchronous rotation in the solar system is a consequence of
process known as tidal locking (Murray and Dermott 1999).
Because a synchronously rotating moon is completely stationary in the aforementioned non-inertial frame, its internal pressure, p, is governed by a force balance
equation of the form [cf., Equation (2.83)]
0 = p + ( + + ),
(2.147)
where is the uniform internal mass density, the gravitational potential due to the
moon, the gravitational potential due to the planet, and
1 2
M
2
= x1
R
+
x
(2.148)
2
2
M + M
the centrifugal potential due to the fact that the non-inertial frame is rotating (about
Hydrostatics
63
an axis parallel to the x3 -axis and passing through point C ) at the angular velocity
[cf., Equation (2.84)]. Suppose that the moon is much less massive that the planet
(i.e., M/M 1). In this limit, the centrifugal potential (2.148) reduces to
1/2
x 2 + x22 + x32
G M
x
1 2 1 + 1
R
R
R2
G M
x1 x12 (1/2) x22 (1/2) x32
+
+ ,
1 +
R
R
R2
=
3 2 1 2
x1 x3 ,
+
2
2
(2.150)
(2.151)
where
G M
,
(2.152)
R3
and any constant terms have been neglected. Thus, the net force field experienced
by the moon due to the combined action of the fictitious centrifugal force and the
gravitational force field of the planet is
=
( + ) = (3 x1 , 0, x3 ).
(2.153)
The previous type of force field is known as a tidal force field, and clearly acts to
elongate the moon along the axis joining the centers of the moon and planet (i.e.,
the x1 -axis), and to compress it along the orbital rotation axis (i.e., the x3 -axis).
Moreover, the magnitude of the tidal force increases linearly with distance from the
center of the moon. The tidal force field is a consequence of the dierent spatial
variation of the centrifugal force and the planets gravitational force of attraction.
This dierent variation causes these two forces, which balance one another at the
center of the moon, to not balance away from the center (Fitzpatrick 2012). As a
result of the tidal force field, we expect the shape of the moon to be distorted from a
sphere. Of course, the moon also generates a tidal force field that acts to distort the
shape of the planet. However, we are assuming that the tidal distortion of the planet
is much smaller than that of the moon (which justifies our earlier statement that
the planet is essentially spherical). As will be demonstrated later, this assumption
is reasonable provided the mass of the moon is much less than that of the planet
(assuming that the planet and moon have similar densities).
64
x22
a22
x32
a32
= 1,
(2.154)
3
2
= G M 0
i xi ,
(2.155)
4
i=1,3
where the integrals i , for i = 0, 3, are defined in Equations (D.30) and (D.31).
Hence, from Equations (2.147) and (2.151), the pressure distribution within the moon
is given by
3
3
1
3
p = p0
G M 1 3 x12 + G M 2 x22 + G M 3 + x32 , (2.156)
2
2
2
2
where p0 is the central pressure. The pressure must be zero on the moons bounding
surface, otherwise this surface would not be in equilibrium. Thus, in order to achieve
equilibrium, we require
3
3
3
1
x22
a22
x32
a32
= 1.
1
a12 = 2 a22 = 3 +
a 2.
(3/2) G M
(3/2) G M 3
(2.158)
(2.159)
2 = 2
,
cos2 cos
a1 sin3 sin2 cos2
sin2
2
cos sin
E(, )
3 = 2
+
,
cos2
cos2 cos
a1 sin3
1 =
(2.160)
(2.161)
(2.162)
Hydrostatics
65
0.05
e13 e12
0.04
0.03
0.02
0.01
Figure 2.8
Properties of the Roche ellipsoids.
where the incomplete elliptic integrals E(, ) and F(, ) are defined in Equations (2.141) and (2.142), respectively. Thus, Equation (2.159) yields
cos2
1
F(, ) (1 + cos2 ) E(, ) 1 +
=
sin tan tan
cos2
sin sin cos cos
+
,
(2.163)
cos2
subject to the constraint
cos2 sin sin cos cos
2
0 = cos F(, ) 2 E(, ) + E(, )
cos2
cos2
cos2
+ (3 + cos2 ) E(, ) + [F(, ) cos2 2 E(, )]
cos2
2 sin sin cos cos
,
(2.164)
+
cos2
where
=
M a03
,
M R3
(2.165)
and a0 = (a1 a2 a3 )1/3 is the mean radius of the moon. The dimensionless parameter
66
e13
0.00000
0.04613
0.09223
0.13809
0.18364
0.22879
0.27346
0.31756
0.36104
0.40383
0.44588
0.48718
0.52769
0.00000
0.00213
0.00852
0.01913
0.03392
0.05282
0.07573
0.10253
0.13308
0.16721
0.20470
0.24528
0.28865
e12
0.52
0.56
0.60
0.64
0.68
0.72
0.76
0.80
0.84
0.88
0.92
0.96
1.00
e13
0.56740
0.60632
0.64445
0.68182
0.71848
0.75446
0.78984
0.82472
0.85923
0.89353
0.92793
0.96294
1.00000
0.33440
0.38204
0.43094
0.48027
0.52890
0.57532
0.61729
0.65150
0.67265
0.67151
0.62978
0.50135
0.00000
Table 2.3
Properties of the Roche ellipsoids.
measures the strength of the tidal distortion field, generated by the planet, that acts
on the moon. There is an analogous parameter,
=
3
M a0
,
M R 3
(2.166)
where a0 is the mean radius of the planet, which measures the tidal distortion field,
generated by the moon, that acts on the planet. We previously assumed that the
former distortion field is much stronger than the latter, allowing us to neglect the
tidal distortion of the planet altogether, and so to treat it as a sphere. This assumption
is only justified if
, which implies that
M
,
M
(2.167)
where = M/[(4/3) a03] and = M /[(4/3) a0 3 ] are the mean densities of the
moon and the planet, respectively. Assuming that these densities are similar, the
previous condition reduces to M M , or, equivalently, a0 < a0 . In other words,
neglecting the tidal distortion of the planet, while retaining that of the moon, is generally only reasonable when the mass of the moon is much less than that of the planet,
as was previously assumed to be the case.
Equations (2.163) and (2.164), which describe the ellipsoidal equilibria of a synchronously rotating, relatively low mass, liquid moon due to the tidal force field of
Hydrostatics
67
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0
Figure 2.9
Properties of the Roche ellipsoids.
It can be seen, from Table 2.3 and Figure 2.8, that the eccentricity e12 = sin
of a Roche ellipsoid in the x1 -x2 plane is almost equal to its eccentricity e13 = sin
in the x1 -x3 plane. In other words, Roche ellipsoids are almost spheroidal in shape,
being elongated along the x1 -axis (i.e., the axis joining the centers of the moon and
the planet), and compressed by almost equal amounts along the x2 - and x3 -axes. In
the limit 1, in which the tidal distortion field due to the planet is weak, it is
easily shown that
2
2
e13
e12
15
.
2
(2.168)
For the case of the tidal distortion field generated by the Earth, and acting on the
Moon, which is characterized by M/M = 0.01230 and R/a0 = 221.29, we obtain
= 7.50 106 (Yoder 1995). It follows that e13 = 7.50 103 , and (a1 a3 )/a1
2
e13
/2 = 2.81 105. In other words, were the Moon a homogeneous liquid body, the
elongation generated by the tidal field of the Earth would be about 50 m.
It can be seen, from Table 2.3 and Figure 2.9, that the parameter attains a
maximum value as the eccentricity of a Roche ellipsoid varies from 0 to 1. In fact,
this maximum value, = 0.06757, occurs when e12 = 0.8594 and e13 = 0.8759. It
follows that there is a maximum strength of the tidal distortion field, generated by a
planet, that is consistent with an ellipsoidal equilibrium of a synchronously rotating,
homogeneous, liquid moon in a circular orbit about the planet. It is plausible that if
this maximum strength is exceeded then the moon is tidally disrupted by the planet.
68
1/3
R
>
2.455
,
a0
(2.169)
where = M/[(4/3) a03] and = M /[(4/3) a0 3 ] are the mean densities of the
moon and the planet, respectively. According to the previous expression, there is
a minimum orbital radius of a moon circling a planet. Below this radius, which is
called the Roche radius, the moon is presumably torn apart by tidal eects. The
Roche radius for a synchronously rotating, self-gravitating, liquid moon in a circular
orbit about a spherical planet is about 2.5 times the planets radius (assuming that
the moon and the planet have approximately the same mass density). Of course, relatively small objects, such as artificial satellites, which are held together by internal
tensile strength, rather than gravity, can orbit inside the Roche radius without being
disrupted.
2.15 Exercises
2.1 A hollow vessel floats in a basin. If, as a consequence of a leak, water flows
slowly into the vessel, how will the level of the water in the basin be aected?
(Lamb 1928.)
2.2 A hollow spherical shell made up of material of specific gravity s > 1 has
external and internal radii a and b, respectively. Demonstrate that the sphere
will only float in water if
1/3
1
b
> 1
.
a
s
2.3 Show that the equilibrium of a solid of uniform density floating with an edge
or corner just emerging from the water is unstable. (Lamb 1928.)
2.4 Prove that if a solid of uniform density floats with a flat face just above the
waterline then the equilibrium is stable. (Lamb 1928.)
2.5 Demonstrate that a uniform solid cylinder floating with its axis horizontal
is in a stable equilibrium provided that its length exceeds the breadth of the
waterline section. [Hint: The cylinder is obviously neutrally stable to rotations about its axis, which means that the corresponding metacentric height
is zero.] (Lamb 1928.)
2.6 Show that a uniform solid cylinder of radius
a and height h can float in stable
equilibrium, with its axis vertical, if h/a < 2. If the ratio h/a exceeds this
Hydrostatics
69
value, prove that the equilibrium is only stable when the specific gravity of
the cylinder lies outside the range
1
a 2
1 1 2 2 .
2
h
2.7 A uniform, thin, hollow cylinder of radius a and height h is open at both
ends. Assuming that h > 2 a, prove that the cylinder cannot float upright if
its specific gravity lies in the range
1 a2
1
.
2
4 h2
(Lamb 1928.)
2.8 Show that the cylinder of the preceding exercise can float with its axis horizontal provided
h
> 3 sin(s ),
2a
where s is the specific gravity of the cylinder. (Lamb 1928.)
2.9 Prove that any segment of a uniform sphere, made up of a substance lighter
than water, can float in stable equilibrium with its plane surface horizontal
and immersed. (Lamb 1928.)
2.10 A vessel carries a tank of oil, of specific gravity s, running along its length.
Assuming that the surface of the oil is at sea level, show that the eect of
the oils fluidity on the rolling of the vessel is equivalent to a reduction in the
metacentric height by A 2 s/V, where V is the displacement of the ship, A
the surface-area of the tank, and the radius of gyration of this area. In what
ratio is the eect diminished when a longitudinal partition bisects the tank?
(Lamb 1928.)
2.11 Find the stable equilibrium configurations of a cylinder of elliptic crosssection, with major and minor radii a and b < a, respectively, made up of
material of specific gravity s, which floats with its axis horizontal.
2.12 A cylindrical tank has a circular cross-section of radius a. Let the center of
gravity of the tank be located a distance c above its base. Suppose that the
tank is pivoted about a horizontal axis passing through its center of gravity,
and is then filled with fluid up to a depth h above its base. Demonstrate that
the position in which the tanks axis is upright is unstable for all filling depths
provided
1
c 2 < a 2.
2
Show that if c 2 > (1/2) a 2 then the upright position is stable when h lies in
the range
c c 2 a 2 /2.
70
2.13 A thin cylindrical vessel of cross-sectional area A floats upright, being immersed to a depth h, and contains water to a depth k. Show that the work
required to pump out the water is 0 A k (h k) g. (Lamb 1928.)
2.14 A sphere of radius a is just immersed in water that is contained in a cylindrical
vessel of radius R whose axis is vertical. Prove that if the sphere is raised just
clear of the water then the waters loss of potential energy is
2 a2
,
Wa 1
3 R2
where W is the weight of the water originally displaced by the sphere. (Lamb
1928.)
2.15 A sphere of radius a, weight W, and specific gravity s > 1, rests on the bottom
of a cylindrical vessel of radius R whose axis is vertical, and which contains
water to a depth h > 2 a. Show that the work required to lift the sphere out of
the vessel is less than if the water had been absent by an amount
2 a3 W
.
ha
3 R2 s
(Lamb 1928.)
2.16 A lead weight is immersed in water that is steadily rotating at an angular
velocity about a vertical axis, the weight being suspended from a fixed
point on this axis by a string of length l. Prove that the position in which the
weight hangs vertically downward is stable or unstable depending on whether
l < g/ 2 or l > g/ 2 , respectively. Also, show that if the vertical position
is unstable then there exists a stable inclined position in which the string is
normal to the surface of equal pressure passing though the weight.
2.17 A thin cylindrical vessel of radius a and height H is orientated such that its
axis is vertical. Suppose that the vessel is filled with liquid of density to
some height h < H above the base, spun about its axis at a steady angular
velocity , and the liquid allowed to attain a steady state. Demonstrate that,
provided 2 a 2 /g < 4 h and 2 a 2 /g < 4 (H h), the net radial thrust on the
vertical walls of the vessel is
2
2 a2
.
a h2 g 1 +
4gh
2.18 A thin cylindrical vessel of radius a with a plane horizontal lid is just filled
with liquid of density , and the whole rotated about a vertical axis at a fixed
angular velocity . Prove that the net upward thrust of the fluid on the lid is
1
a 4 2.
4
Hydrostatics
71
2.19 A liquid-filled thin spherical vessel of radius a spins about a vertical diameter
at the fixed angular velocity . Assuming that the liquid co-rotates with the
vessel, and that 2 > g/a, show that the pressure on the wall of the vessel is
greatest a depth g/ 2 below the center. Also prove that the net normal thrusts
on the lower and upper hemispheres are
5
3
Mg+
M 2 a,
4
16
and
1
3
Mg
M 2 a,
4
16
respectively, where M is the mass of the liquid.
2.20 A closed cubic vessel filled with water is rotating about a vertical axis passing
through the centers of two opposite sides. Demonstrate that, as a consequence
of the rotation, the net thrust on a side is increased by
1 4
a 2,
6
where a is the length of an edge of the cube, and the angular velocity of
rotation.
2.21 A closed vessel filled with water is rotating at constant angular velocity
about a horizontal axis. Show that, in the state of relative equilibrium, the
constant pressure surfaces in the water are circular cylinders whose common
axis is a height g/ 2 above the axis of rotation. (Batchelor 2000.)
2.22 Verify Equations (2.95)(2.97) and (2.99)(2.101).
2.23 Consider a homogeneous, rotating, liquid body of mass M, mean radius a0 ,
and angular velocity , whose outer boundary is a Maclaurin spheroid of
eccentricity e.
(a) Demonstrate that
e
5
2 (G M/a03 )1/2
72
3
Surface Tension
3.1 Introduction
As is well known, small drops of water in air, and small bubbles of gas in water, tend
to adopt spherical shapes. This observation, and a host of other natural phenomena,
can only be accounted for on the hypothesis that an interface between two dierent
media is associated with a particular form of energy whose magnitude is directly
proportional to the interfacial area. To be more exact, if S is the interfacial area then
the contribution of the interface to the Helmholtz free energy of the system takes the
form S , where only depends on the temperature and chemical composition of the
two media on either side of the interface. It follows, from standard thermodynamics,
that S is the work that must be performed on the system in order to create the
interface via an isothermal and reversible process (Reif 1965). However, this work
is exactly the same as that which we would calculate on the assumption that the
interface is in a state of uniform constant tension per unit length . Thus, can be
interpreted as both a free energy per unit area of the interface, and a surface tension.
This tension is such that a force of magnitude per unit length is exerted across
any line drawn on the interface, in a direction normal to the line, and tangential to
the interface. More information on surface tension can be found in Lamb 1928, and
Batchelor 2000.
Surface tension originates from intermolecular cohesive forces. The average free
energy of a molecule in a given isotropic medium possessing an interface with a
second medium is independent of its position, provided that the molecule does not lie
too close to the interface. However, the free energy is modified when the molecules
distance from the interface becomes less than the range of the cohesive forces (which
is typically 109 m). Because this range is so small, the number of molecules in a
macroscopic system whose free energies are aected by the presence of an interface
is directly proportional to the interfacial area. Hence, the contribution of the interface
to the total free energy of the system is also proportional to the interfacial area. If
only one of the two media in question is a condensed phase then the parameter
is invariably positive (i.e., such that a reduction in the surface area is energetically
favorable). This follows because the molecules of a liquid or a solid are subject to
an attractive force from neighboring molecules. However, molecules that are near to
an interface with a gas lack neighbors on one side, and so experience an unbalanced
cohesive force directed toward the interior of the liquid/solid. The existence of this
73
74
n
2
1
S
C
t
tn
Figure 3.1
Interface between two immiscible fluids.
force makes it energetically favorable for the interface to contract (i.e., > 0). On
the other hand, if the interface separates a liquid and a solid, or a liquid and another
liquid, then the sign of cannot be predicted by this argument. In fact, it is possible
for both signs of to occur at liquid/solid and liquid/liquid interfaces.
The surface tension of a water/air interface at 20 C is = 7.28 102 N m1
(Batchelor 2000). The surface tension at most oil/air interfaces is much lower
typically, 2102 N m1 (Batchelor 2000). On the other hand, interfaces between
liquid metals and air generally have very large surface tensions. For instance, the
surface tension of a mercury/air interface at 20 C is 4.87 101 N m1 (Batchelor
2000).
For some pairs of liquids, such as water and alcohol, an interface cannot generally
be observed because it is in compression (i.e., < 0). Such an interface tends to
become as large as possible, leading to complete mixing of the two liquids. In other
words, liquids for which > 0 are immiscible, whereas those for which < 0 are
miscible.
Finally, the surface tension at a liquid/gas or a liquid/liquid interface can be affected by the presence of adsorbed impurities at the interface. For instance, the
surface tension at a water/air interface is significantly deceased in the presence of
adsorbed soap molecules. Impurities that tend to reduce surface tension at interfaces
are termed surfactants.
Surface Tension
75
(3.1)
(In fact, the net force would be zero even in the absence of equilibrium, because the
interface has zero mass.)
Applying the curl theorem (see Section A.22) to the curve C, we find that
F dr =
F dS,
(3.3)
C
(3.4)
(3.5)
(3.6)
(3.7)
(3.9)
76
t n dr = ( n) n dS ,
C
(3.10)
(3.11)
Finally, given that S is arbitrary, the previous expression reduces to the pressure
balance constraint
p = n,
(3.12)
where p = p1 p2 . The previous relation is generally known as the Young-Laplace
equation, and is named after Thomas Young (1773-1829), who developed the qualitative theory of surface tension in 1805, and Pierre-Simon Laplace (1749-1827) who
completed the mathematical description in the following year. The Young-Laplace
equation can also be derived by minimizing the free energy of the interface. (See
Section 3.8.) Note that p is the jump in pressure seen when crossing the interface in
the opposite direction to n. Of course, a plane interface is characterized by n = 0.
On the other hand, a curved interface generally has n 0. In fact, n measures
the local mean curvature of the interface. Thus, according to the Young-Laplace
equation, there is a pressure jump across a curved interface between two immiscible
fluids, the magnitude of the jump being proportional to the surface tension.
Surface Tension
77
2
.
R
(3.14)
Thus, given that p is the pressure jump seen crossing the interface in the opposite
direction to n, we conclude that the pressure inside a droplet or bubble exceeds that
outside by an amount proportional to the surface tension, and inversely proportional
to the droplet or bubble radius. This explains why small bubbles are louder that large
ones when they burst at a free surface: for instance, champagne fizzes louder than
beer. Note that soap bubbles in air have two interfaces defining the inner and outer
extents of the soap film. Consequently, the net pressure dierence is twice that across
a single interface.
l=
1/2
(3.16)
is known as the capillary length, and takes the value 2.7 103 m for pure water
at 20 C (Batchelor 2000). We conclude that the eect of surface tension on the
shape of an liquid/air interface is likely to dominate the eect of gravity when the
interfaces radius of curvature is much less than the capillary length, and vice versa.
78
Figure 3.2
Interface between a liquid (1), a gas (2), and a solid (3).
(3.17)
However, an equilibrium state is one which minimizes the free energy, implying that
the previous expression is zero for arbitrary (small) r: that is,
cos =
23 13
.
12
(3.18)
We conclude that, in equilibrium, the angle of contact, , between the liquid and
the solid takes a fixed value that depends on the free energies per unit area at the
liquid/air, liquid/solid, and air/solid interfaces. Note that the previous formula could
also be obtained from the requirement that the various surface tension forces acting
at the edge balance one another, assuming that it is really appropriate to interpret 13
and 23 as surface tensions when one of the media making up the interface is a solid.
Surface Tension
79
air
a
liquid
glass tube
air
liquid
free surface z = 0
Figure 3.3
Elevation of liquid level in a capillary tube.
23 + 12 > 13 > 23 12 .
(3.19)
If 13 > 23 + 12 then the angle of contact is 180, which corresponds to the case
where the free energy at the liquid/solid interface is so large that the liquid does not
wet the solid at all, but instead breaks up into beads on its surface. On the other hand,
if 13 < 23 12 then the angle of contact is 0 , which corresponds to the case where
the free energy at the liquid/solid interface is so small that the liquid completely wets
the solid, spreading out indefinitely until it either covers the whole surface, or its
thickness reaches molecular dimensions.
The angle of contact between water and glass typically lies in the range 25 to
29 , whereas that between mercury and glass is about 127 (Batchelor 2000).
80
gh =
2 cos
,
a
(3.22)
2 cos
.
ga
(3.23)
This result, which relates the height, h, to which a liquid rises in a capillary tube of
radius a to the liquids surface tension, , is known as Jurins law, and is named after
its discoverer, James Jurin (1684-1750). The assumption that the radius of the tube
is much less than the capillary length is equivalent to the assumption that the height
of the interface above the free surface of the liquid is much greater than the radius of
the tube. This follows, from Equations (3.16) and (3.23), because
l2
h
= 2 cos 2 .
(3.24)
a
a
Thus, the ordering a l implies that h
a.
For the case of water at 20 , assuming a contact angle of 25 , Jurins law yields
h(mm) = 13.5/a(mm) (Batchelor 2000). Thus, water rises a height 13.5 mm in a
capillary tube of radius 1 mm, but rises 13.5 cm in a capillary tube of radius 0.1 mm.
In the case of a liquid, such a mercury, that has an oblique angle of contact with
glass, so that cos < 0, the liquid level in a capillary tube is depressed below that of
the free surface (i.e., h < 0).
Surface Tension
81
ez f x e x
(z f )
=
,
|(z f )| (1 + f x2 )1/2
(3.25)
where f x = d f /dx. Hence, the mean curvature of the interface is (Riley 1974)
n =
f xx
,
(1 + f x2 )3/2
(3.26)
where f xx = d 2 f /dx 2 . According to Equations (3.15) and (3.16), the shape of the
interface is governed by the nonlinear dierential equation
f =
l 2 f xx
.
(1 + f x2 )3/2
(3.27)
where the vertical height, f , of the interface is measured relative to its equilibrium
height in the absence of surface tension. Multiplying the previous equation by f x /l 2 ,
and integrating with respect to x, we obtain
f2
1
= C 2,
2
1/2
(1 + f x )
2l
(3.28)
f2
1,
2 l2
C f 2 /2 l 2
1
=
.
fx
[1 (C f 2 /2 l 2 )2 ]1/2
Let
C=
2
1,
k2
(3.29)
(3.30)
(3.31)
f2
= cos(2 ),
2 l2
(3.32)
(3.33)
82
2
z/l
1
0
2
0
x/l
Figure 3.4
Capillary curves for /4 3/4 and (in order from the top to the bottom)
k = 0.6, 0.7, 0.8, 0.9, and 0.99.
and so the constraint (3.29) implies that /4 3/4. Moreover, Equations (3.30)
and (3.33) reduce to
1
dx
1
=
.
(3.34)
=
fx d f
tan(2 )
It follows from Equations (3.32) and (3.34) that
dx dx d f
l k cos(2 )
=
=
,
d d f d
(1 k 2 sin2 )1/2
which can be integrated to give
/2
x
k cos(2 )
=
d,
l
(1 k 2 sin2 )1/2
x
2
2
= k
F(, k) + E(,
k),
l
k
k
(3.35)
(3.36)
(3.37)
where
k) = E(/2, k) E(, k),
E(,
(3.38)
(3.39)
Surface Tension
83
0.8
0.7
0.6
0.5
z/l 0.4
0.3
0.2
0.1
0
1
0.5
0
x/l
0.5
Figure 3.5
Liquid/air interface for a liquid trapped between two vertical parallel plates located
at x = l. The contact angle of the interface with the plates is = 30 .
and
E(, k) =
(1 k 2 sin2 )1/2 ,
(3.40)
(1 k 2 sin2 )1/2 ,
(3.41)
F(, k) =
are types of incomplete elliptic integral (Abramowitz and Stegun 1965). In conclusion, the interface shape is determined parametrically by
2
2
x
= k
F(, k) + E(,
k),
(3.42)
l
k
k
2
z
= (1 k 2 sin2 )1/2 ,
l
k
(3.43)
where /4 3/4. Here, the parameter k is restricted to lie in the range 0 < k <
1.
Figure 3.4 shows the capillary curves predicted by Equations (3.42) and (3.43) for
various dierent values of k. Here, we have chosen the plus sign in Equation (3.43).
However, if the minus sign is chosen then the curves are simply inverted: that is,
x x and z z. In can be seen that all of the curves shown in the figure are
84
E(,
4
2
k) + k ( + sin cos ),
F(,
4
(3.44)
(3.45)
sin(2 ),
l
2
z 2 k
[1 cos(2 )].
l
k 2
(3.46)
(3.47)
It follows that the interface is a segment of the curved surface of a cylinder whose
axis runs parallel to the y-axis. If the distance between the plates is 2 d, and the
contact angle is , then we require x = d when = 3/4 /2 (which corresponds
to = /4 + /2). From Equation (3.46), this constraint yields
d k
cos .
l
2
(3.48)
Thus, the height that the liquid rises between the two platesthat is, h = z(x = 0) =
z( = /2) 2 l/kis given by
h
cos
.
gd
(3.49)
Surface Tension
85
1.1
1
0.9
0.8
0.7
z/l
0.6
0.5
0.4
0.3
0.2
0.1
0
x/l
Figure 3.6
Liquid/air interface for a liquid in contact with a vertical plate located at x = 0. The
contact angle of the interface with the plate is = 25 .
This result is the form taken by Jurins law, (3.23), for a liquid drawn up between
two parallel plates of spacing 2 d.
Consider the case k = C = 1, which is such that the distance between the two
plates is infinite. Let the leftmost plate lie at x = 0, and let us completely neglect the
rightmost plate, because it lies at infinity. Suppose that h = z(x = 0) is the height of
the interface above the free surface of the liquid at the point where the interface meets
the leftmost plate. If is the angle of contact of the interface with the plate then we
require f x = 1/ tan at x = 0. Because C = 1, it follows from Equation (3.28) that
h2
= 1 sin ,
2 l2
(3.50)
h = 2 l sin(/4 /2).
(3.51)
or
Furthermore, again recalling that C = 1, Equation (3.30) can be integrated to give
x=
z
df
=l
fx
h/2l
z/2l
1 2 y2
dy,
y (1 y 2 )1/2
(3.52)
where we have chosen the minus sign, and y = f /(2 l). Making the substitution
86
sin1 (h/2 l)
sin1 (z/2 l)
sin1 (h/2 l)
1 + cos u
1
2 sin u du = ln
, (3.53)
+ 2 cos u
sin u
sin u
sin1 (z/2 l)
which reduces to
1/2
1/2
z2
x
h2
1 2 l
1 2 l
= cosh
4 2
,
cosh
+ 4 2
l
z
h
l
l
(3.54)
because cosh1 (z) ln[z + (z 2 1)1/2 ]. Thus, Equations (3.51) and (3.54) specify
the shape of a liquid/air interface that meets an isolated vertical plate at x = 0. In
particular, Equation (3.51) gives the height that the interface climbs up the plate
(relative to the free surface) due to the action of surface tension. Note that this height
is restricted to lie in the range 2 l h 2 l, irrespective of the angle of contact.
Figure 3.6 shows an example interface calculated for = 25 .
er fz ez
(r f )
=
,
|(r f )| (1 + fz2 )1/2
(3.55)
where fz d f /dz. Hence, from Equation (C.39), the mean curvature of the surface
is given by
f
1 d
.
(3.56)
n=
f fz dz (1 + fz2 )1/2
The Young-Laplace equation, (3.12), then yields
d
f
f fz
=
,
a
dz (1 + fz2 )1/2
where
a=
.
p0
(3.57)
(3.58)
Here, is the net surface tension, including the contributions from the internal and
external soap/air interfaces. Moreover, p0 = p is the pressure dierence between
the interior and the exterior of the bubble. Equation (3.57) can be integrated to give
f
f2
+ C,
=
2
1/2
2a
(1 + fz )
(3.59)
Surface Tension
87
where C is a constant.
Suppose that the bubble occupies the region z1 z z2 , where z1 < z2 , and
has a fixed radius at its two end-points, z = z1 and z = z2 . This could most easily
be achieved by supporting the bubble on two rigid parallel co-axial rings located at
z = z1 and z = z2 . The net free energy required to create the bubble can be written
E = S p0 V,
(3.60)
where S is area of the bubble surface, and V the enclosed volume. The first term on
the right-hand side of the previous expression represents the work needed to overcome surface tension, whereas the second term represents the work required to overcome the pressure dierence, p0 , between the exterior and the interior of the bubble. From the general principles of statics, we expect a stable equilibrium state of
a mechanical system to be such as to minimize the net free energy, subject to any
dynamical constraints (Fitzpatrick 2012). It follows that the equilibrium shape of the
bubble is such as to minimize
z2
z2
E=
2 f (1 + fz2 )1/2 dz p0
f 2 dz,
(3.61)
z1
z1
where
L( f, fz ) = 2 f (1 + fz2 )1/2 p0 f 2 ,
(3.63)
subject to the constraint that f is fixed at the limits. This is a standard problem
in the calculus of variations. (See Appendix E.) In fact, because the functional
L( f, fz ) does not depend explicitly on z, the minimizing function is the solution of
[see Equation (E.14)]
L
L fz
= C,
(3.64)
fz
where C is an arbitrary constant. Thus, we obtain
f
f2
2
= C ,
(1 + fz2 )1/2 2 a
(3.65)
which can be rearranged to give Equation (3.59). Hence, we conclude that application of the Young-Laplace equation does indeed lead to a bubble shape that minimizes the net free energy of the soap/air interfaces.
Consider the case p0 = 0, in which there is no pressure dierence across the
surface of the bubble. In this situation, writing C = b > 0, Equation (3.59) reduces
to
(3.66)
f = b (1 + fz2 )1/2 .
88
1
0.9
0.8
0.7
0.6
r/c 0.5
0.4
0.3
0.2
0.1
0
0
z/c
0.2
0.4
0.6
Figure 3.7
Radius versus axial distance for a catenoid soap bubble supported by two parallel
co-axial rings of radius c located at z = 0.65 c.
Moreover, according to the previous discussion, the bubble shape specified by Equation (3.66) is such as to minimize the surface area of the bubble (because the only
contribution to the free energy of the soap/air interfaces is directly proportional to
the bubble area). The previous equation can be rearranged to give
2
1/2
f
,
(3.67)
fz = 2 1
b
which leads to
z z0 =
b
df
=
fz
b
(f
df
= b cosh1 (r/b),
1)1/2
2 /b 2
(3.68)
or
r = b cosh(|z z0 |/b),
(3.69)
Surface Tension
89
(3.71)
For example, if d = 0.65 c then b = 0.6416 c, and the resulting bubble shape is
illustrated in Figure 3.7.
Let d/c = and d/b = u, in which case the previous equation becomes
G(u) = u cosh u = 0.
(3.72)
1
,
tanh u0
(3.73)
when u0 = sinh1 (1/). Moreover, if G(u0 ) > 0 then Equation (3.72) possesses
two roots. It turns out that the root associated with the smaller value of u minimizes the interface system energy, whereas the other root maximizes the free energy.
Hence, the former root corresponds to a stable equilibrium state, whereas the latter
corresponds to an unstable equilibrium state. On the other hand, if G(u0 ) < 0 then
Equation (3.72) possesses no roots, implying the absence of any equilibrium state.
The critical case G(u0 ) = 0 corresponds to u = uc and = c , where uc tanh uc = 1
and c = 1/ sinh uc . It is easily demonstrated that uc = 1.1997 and c = 0.6627. We
conclude that a stable equilibrium state of a catenoid bubble only exists when c ,
which corresponds to d 0.6627 c. If the relative ring spacing d exceeds the critical
value 0.6627 c then the bubble presumably bursts.
Consider the case p0 0, in which there is a pressure dierence across the
surface of the bubble. In this situation, writing
2 a = + ,
2 a C = ,
Equation (3.59) becomes
( + ) f
= f 2 + ,
(1 + fz2 )1/2
(3.74)
(3.75)
(3.76)
( 2 f 2 )1/2 ( f 2 2 )1/2
.
f 2 +
(3.77)
We can assume, without loss of generality, that || > ||. It follows, from the previous
expression, that || f ||. Hence, we can write
f 2 = 2 cos2 + 2 sin2 ,
k2 =
2 2
,
2
(3.78)
(3.79)
90
1.1
1
0.9
0.8
r/
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
2
0
z/
Figure 3.8
Radius versus axial distance for an unduloid soap bubble calculated with k = 0.95.
where 0 /2 and 0 < k 1. It follows that
f = || (1 k 2 sin2 )1/2 ,
(3.80)
= sgn() || (1 k )
(3.81)
2 1/2
and
dz
1 df
=
= f +
,
d
fz d
f
(3.82)
(3.83)
where E(, k) and F(, k) are incomplete elliptic integrals [see Equations (3.40) and
(3.41)]. Here, we have assumed that = 0 when z = 0. There are three cases of
interest.
In the first case, > 0 and > 0. It follows that (1 k 2 )1/2 r/ 1 for
/2 0, and 0.5 /(p0 ) < 1 for 1 k > 0, where
r = (1 k 2 sin2 ),
|z| = E(, k) + (1 k 2 )1/2 F(, k) .
(3.84)
(3.85)
Surface Tension
91
1.1
1
0.9
0.8
r/
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.5
0
z/
0.5
Figure 3.9
Radius versus axial distance for a positive pressure nodoid soap bubble calculated
with k = 0.95.
(3.86)
(3.87)
92
0.6
0.5
r/||
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
z/||
0.4
0.5
Figure 3.10
Radius versus axial distance for a negative pressure nodoid soap bubble calculated
with k = 0.95.
0 < k 1, where
r = || (1 k 2 sin2 ),
|z| = || E(, k) (1 k 2 )1/2 F(, k) .
(3.88)
(3.89)
3.9 Exercises
3.1 Show that if N equal spheres of water coalesce so as to form a single spherical
drop then the surface energy is decreased by a factor 1/N 1/3 . (Lamb 1928.)
3.2 A circular cylinder of radius a, height h, and specific gravity s floats upright
in water. Show that the depth of the base below the general level of the water
Surface Tension
93
surface is
2
cos ,
a
where is the surface tension at the air/water interface, and the contact
angle of the interface with the cylinder. (Lamb 1928.)
sh +
3.3 A film of water is held between two parallel plates of glass a small distance
2 d apart. Prove that the apparent attraction between the plates is
2 A cos
+ L sin ,
d
where is the surface tension at the air/water interface, the angle of contact
of the interface with glass, A the area of the film, and L the circumference of
the film. (Lamb 1928.)
3.4 Show that if the surface of a sheet of water is slightly corrugated then the
surface energy is increased by
2
dx
per unit breadth of the corrugations. Here, x is measured horizontally, perpendicular to the corrugations. Moreover, denotes the elevation of the surface above the mean level. Finally, is the surface tension at an air/water
interface. If the corrugations are sinusoidal, such that
= a sin(k x),
show that the average increment of the surface energy per unit area is
(1/4) a 2 k 2 .
(Lamb 1928.)
3.5 A mass of liquid, which is held together by surface tension alone, revolves
about a fixed axis at a small angular velocity , so as to assume a slightly
spheroidal shape of mean radius a. Prove that the ellipticity of the spheroid
is
2 a3
,
=
8
where is the uniform mass density, and the surface tension. [If r+ is the
maximum radius, and r the minimum radius, then a = (r+2 r )1/3 , and the
ellipticity is defined = (r+ r )/r+ .] (Lamb 1928.)
3.6 A liquid mass rotates, in the form of a circular ring of radius a and small
cross-section, with a constant angular velocity , about an axis normal to
94
2
a c2
1/2
,
where is the density, the surface tension, and c the radius of the crosssection. (Lamb 1928.)
3.7 Two spherical soap bubbles of radii a1 and a2 are made to coalesce. Show
that when the temperature of the gas in the resulting bubble has returned to
its initial value the radius a of the bubble satisfies
p0 a 3 + 4 a 2 = p0 (a13 + a23 ) + 4 (a12 + a22 ),
where p0 is the ambient pressure, and the surface tension of the soap/air
interfaces. (Batchelor 2000.)
3.8 A rigid sphere of radius a rests on a flat rigid surface, and a small amount of
liquid surrounds the contact point, making a concave-planar lens whose diameter is small compared to a. The angle of contact of the liquid/air interface
with each of the solid surfaces is zero, and the surface tension of the interface
is . Show that there is an adhesive force of magnitude 4 a acting on the
sphere. (It is interesting to note that the force is independent of the volume
of liquid.) (Batchelor 2000.)
3.9 Two small solid bodies are floating on the surface of a liquid. Show that the
eect of surface tension is to make the objects approach one another if the
liquid/air interface has either an acute or an obtuse angle of contact with both
bodies, and to make them move away from one another if the interface has an
acute angle of contact with one body, and an obtuse angle of contact with the
other. (Batchelor 2000.)
4
Incompressible Inviscid Flow
4.1 Introduction
This chapter introduces some of the fundamental concepts that arise in the theory of
incompressible, inviscid (or, to be more exact, high Reynolds number) fluid motion.
Further information on these concepts can be found in Faber 1995, Batchelor 2000,
and Milne-Thompson 2011.
96
D
D
C
C
B
A
B
A
Figure 4.1
Bernoullis theorem.
implies that
v1 S 1 = v2 S 2 .
(4.1)
is constant along a streamline, where p is the pressure, the density, and T the total
energy per unit mass.
The proof is straightforward. Consider the body of fluid bounded by the crosssectional areas AB and CD of the stream filament pictured in Figure 4.1. Let us
97
(4.3)
Let T denote the total energy of the section of the fluid lying between A B and CD.
Thus, the increase in energy of the fluid body in the time interval t is
(m T2 + T ) (m T1 + T ) = m (T2 T1 ).
(4.4)
In the absence of viscous energy dissipation, this energy increase must equal the net
work done on the fluid by the pressures at AB and CD, which is
p1 p2
p1 S 1 v1 t p2 S 2 v2 t = m
.
(4.5)
1 2
Equating expressions (4.4) and (4.5), we find that
p2
p1
+ T1 =
+ T2 ,
1
2
(4.6)
which demonstrates that p/ + T has the same value at any two points on a given
stream filament, and is therefore constant along the filament. Note that Bernoullis
theorem has only been proved for the case of the steady motion of an inviscid fluid.
However, the fluid in question may either be compressible or incompressible.
For the particular case of an incompressible fluid, moving in a conservative forcefield, the total energy per unit mass is the sum of the kinetic energy per unit mass,
(1/2) v 2, and the potential energy per unit mass, , and Bernoullis theorem thus
becomes
p 1 2
+ v + = constant along a streamline.
(4.7)
2
If we focus on a particular streamline, 1 (say), then Bernoullis theorem states that
p 1 2
+ v + = C1 ,
2
(4.8)
where C1 is a constant characterizing that streamline. If we consider a second streamline, 2 (say), then
p 1 2
+ v + = C2 ,
(4.9)
2
where C2 is another constant. It is not generally the case that C1 = C2 . If, however,
the fluid motion is irrotational then the constant in Bernoullis theorem is the same
for all streamlines (see Section 4.15), so that
p 1 2
+ v + =C
(4.10)
2
throughout the fluid.
98
99
S1
V
S2
V
Figure 4.2
The dAlembert paradox.
external forces, such as gravity, F is the resultant in the direction of the flow of the
pressure thrusts acting on the boundary of A.
Consider two cross-sections, S 1 and S 2 , a great distance upstream and downstream of A, respectively. The fluid between these sections can be split into a great
many stream filaments to each of which Eulers momentum theorem is applicable.
The outer filaments are bounded by the walls of the tube, and so the thrust components acting on these are directed perpendicular to the flow. The walls of the filaments
in contact with A are acted on by the obstacle, which exerts on them a force whose
component in the direction of the flow is F. By Eulers theorem, the resultant of all
of the thrusts acting on the fluid in the tube is
S 1 V 2 + S 2 V 2 ,
(4.11)
(4.12)
F = 0.
(4.13)
100
B
A
Figure 4.3
Outflow through an orifice.
flow does not alter the dynamical conditions. Therefore, we conclude that the resistance to a solid body moving with uniform velocity through an unbounded inviscid
fluid, otherwise at rest, is zero.
101
that the outflow through the orifice is not suciently strong to cause the surface level
to drop at a significant rate.) Let v be the uniform fluid velocity in the vena contracta.
Of course, the pressure is atmospheric both at the surface of the fluid and in the vena
contracta. It follows that
1
g h = v 2,
(4.14)
2
or
v = (2 g h)1/2.
(4.15)
In other words, the eux velocity of the fluid from the orifice is the same as that
it would have acquired by falling a height h under gravity. This result is known as
Torricellis law, after Evangelista Torricelli (1608-1647). Finally, the discharge rate
of fluid flowing through the orifice is
Q = C S v = C S (2 g h)1/2.
(4.16)
(4.17)
where use has been made of Equation (4.15). Thus, we conclude that the contraction
coecient takes the value 1/2.
In reality, Bernoullis theorem suggests that when the orifice is opened the pressure on the walls in the neighborhood of the orifice will fall below the hydrostatic
value, which implies that the accelerating thrust is actually greater than S (p p0 ).
Consequently, C > 1/2. Obviously, C cannot exceed unity, so we conclude that, in
general, 1/2 < C < 1. For instance, if the orifice is a circular hole punched in a thin
plate then the contraction coecient is observed to take the value 0.62 (Batchelor
2000).
Suppose, however, that we fit a small cylindrical nozzle projecting inward from
the orifice, as shown in Figure 4.4. In this case, the original assumption that the
pressure on the walls in the neighborhood of the orifice is hydrostatic is essentially
correct. This follows because the region where the lines of flow are converging on the
orifice is far removed from the walls, and the velocity of the fluid in contact with the
walls is negligible. Thus, the contraction coecient is exactly 1/2. This arrangement
is known as a Borda mouthpiece, after Jean-Charles Borda (17331799).
102
Figure 4.4
A Borda mouthpiece.
Q v2
+
.
v 2g
(4.20)
(4.21)
It is easily demonstrated that the function F(v) attains its minimum value,
Fc =
1/3
3 Q2
,
2 g
(4.22)
103
when v = vc , where
vc = (Q g)1/3 .
(4.23)
We conclude that, as long as H > (3/2) (Q /g) , Equation (4.20) possesses two
possible solutions that are consistent with a given head height and flow rate. In one
solution, the stream flows at a relatively slow velocity, v , which is such that v < vc .
In the other, the stream flows at a relatively fast velocity, v+ , which is such that
v+ > vc . The corresponding depths are h = Q/v and h+ = Q/v+ , respectively.
It is helpful to introduce the dimensionless Froude number,
2
1/3
v
Fr = .
gh
(4.24)
(See Section 1.15.) Note that g h is the characteristic propagation velocity of a
gravity wave in shallow water of depth h. (See Section 11.4.) Hence, if Fr < 1
then the streams flow velocity falls below the wave speedsuch flow is termed
sub-critical. On the other hand, if Fr > 1 then the flow velocity exceeds the wave
speedsuch flow is termed super-critical.
We can combine Equations (4.18), (4.19), and (4.24) to give
H = G(Fr),
where
G(Fr) =
Q2
g
1/2
1
Fr 4/3
+
.
2
Fr 2/3
(4.25)
(4.26)
1/3
3 Q2
,
Gc =
2 g
(4.27)
Frc = 1.
(4.28)
Hence, we again conclude that, as long as H > (3/2) (Q 2/g)1/3 , there are two possible flow velocities of the stream (parameterized by two dierent Froude numbers)
that are consistent with a given head height and flow rate. However, it is now clear
that the smaller velocity is sub-critical (i.e., Fr < 1), whereas the larger velocity is
super-critical (i.e., Fr > 1).
104
V
H
d
Figure 4.5
Flow over a shallow bump.
stream passes over the top of the bump, its velocity is v, and its surface rises a height
h H above the unperturbed surface.
Fluid continuity yields
H V = (H + h d) v.
(4.29)
Furthermore, application of Bernoullis equation to a streamline lying on the surface
of the water (where the pressure is atmospheric) gives
gH+
1
1 2
V = g (H + h) + v 2 .
2
2
(4.30)
(4.31)
hd
hd
h
1+
1 = 2
.
Fr 1 +
H
H
H
2
(4.32)
(4.33)
(4.34)
Here,
V
Fr =
(4.35)
gH
is the Froude number of the unperturbed flow. Finally, given that h/H 1 and
d/H 1, Equation (4.34) reduces to
h
d
.
1 1/Fr 2
(4.36)
105
h2
h1
v2
v1
Figure 4.6
A hydraulic jump.
It follows, from the previous expression, that if the flow is super-critical, so that
Fr > 1, then h is positive. On the other hand, if the flow is sub-critical, so that
Fr < 1, then h is negative. Thus, if a super-critical shallow stream passes over a very
shallow bump on its bed then the surface of the stream becomes slightly elevated.
On the other hand, if a sub-critical stream passes over such a bump then the surface
of the stream becomes slightly depressed. A similar eect occurs when there is a
narrowing of the channel in the horizontal direction. A more sophisticated version
of the previous calculation, which does not necessarily assume that the stream is
shallow, can be found in Section 11.10.
106
well as the Euler momentum theorem, neither of which depend on the conservation
of mechanical energy.
Fluid continuity yields
Q = h 1 v1 = h 2 v2 ,
(4.37)
where Q is the fixed flow rate per unit width. Furthermore, the Euler momentum
theorem (see Section 4.4) implies that
F21 =
0
h2
(p2 + v22 ) dz
h1
(4.38)
Here, F21 is the horizontal thrust per unit width exerted by the channel bed on the
water lying between points 1 and 2. Moreover, z measured vertical height above the
bed. Assuming the usual linear pressure variation with depth, we can write
p1 (z) = p0 + g (h1 z),
(4.39)
(4.40)
(4.41)
The neglect of frictional drag at the channel bed implies that F21 = 0. Thus, we
obtain
1 2 1 2 Q
h h + (v2 v1 ) = 0.
(4.42)
2 2 2 1 g
The previous equation possesses the trivial solution h1 = h2 and v1 = v2 , which
corresponds to the absence of a hydraulic jump. Eliminating v1 and v2 between Equations (4.37) and (4.42), and canceling a common factor h2 h1 , we obtain the nontrivial solution
2 Q2
.
(4.43)
h2 h1 (h2 + h1 ) =
g
Now, the upstream Froude number is defined
v1
Q
Fr1 =
=
.
1/2
g h1 g h13/2
(4.44)
h2
h1
2
+
h2
2 Fr12 = 0,
h1
(4.45)
h2 1
=
1 + 1 + 8 Fr12 .
h1 2
(4.46)
107
Here, we have neglected an unphysical solution in which h2 /h1 is negative. Note that
h2
1
h1
Fr1 1.
as
(4.47)
(4.48)
h1
h2
2
+
h1
2 Fr22 = 0,
h2
(4.49)
h1 1
2
=
1 + 1 + 8 Fr2 .
h2 2
(4.50)
(4.51)
1
1 + 1 + 8 x 2 .
2
(4.52)
Note that
G(x) 1
as
x 1,
(4.53)
Fr2 1.
(4.54)
as
In other words, if the upstream flow is super-critical then the downstream flow is
sub-critical, and h2 > h1 . On the other hand, if the upstream flow is sub-critical then
the downstream flow is super-critical, and h2 < h1 .
By analogy with Equation (4.18), we can define the head heights of the flows
upstream and downstream of the jump as
H1 = h 1 +
1 v12
,
2 g
(4.55)
H2 = h 2 +
1 v22
,
2 g
(4.56)
and
respectively. Of course, in the absence of any transfer of mechanical energy from the
108
laminar to the turbulent component of the flow, within the jump, we would expect
H1 = H2 . (Because this is what Bernoullis equation predicts.) In the presence of
the transfer, we expect H1 > H2 . In other words, we expect there to be a head loss
across the jump. Note that it is impossible for H2 to exceed H1 , because this would
imply a transfer of mechanical energy from the turbulent to the laminar component
of the flow, within the jump, which violates the second law of thermodynamics. Let
H L = H1 H2 = h 1 h 2 +
v2
v12
2
2g 2g
(4.57)
be the positive definite head loss. Making use of some previous definitions, we can
write
h12
h2
1 2
HL
=
1 + Fr1 1 2 .
(4.58)
h1
h1
2
h2
Now, according to Equation (4.45),
Fr12
1 h2 h2
=
+1 .
2 h1 h1
h2
1 h1 h2
h2
HL =
1 4
+
+ 1 ,
4 h2 h1
h1
h1
or
HL =
(4.59)
(4.60)
(4.61)
109
Earth frame
V
u2
h2
h1
u1
Co-moving frame
V u2
h2
h1
V + u1
Figure 4.7
A tidal bore.
bore is of depth h2 , and is flowing upstream at the velocity u2 . The lower part of
the figure shows the same phenomenon in the rest frame of the bore. In this frame,
we observe a stationary hydraulic jump with an upstream depth and flow velocity
h1 and v1 = V + u1 , respectively, and a downstream depth and flow velocity h2 and
v2 = V u2 , respectively. Making use of Equations (4.44) and (4.45), we obtain
h2
h1
2
+
h2 2 (u1 + V)2
= 0,
h1
g h1
(4.62)
g h2 (h1 + h2 )
V = u1 +
2 h1
1/2
.
(4.63)
Thus, we deduce that the speed of the bore, relative to the unperturbed river, is a
simple function of the upstream and downstream depths. Note that, in the limit
h1 h2 h, the previous equation reduces to V u1 + g h. In other words, a
weak bore degenerates intoan ordinary shallow water gravity wave propagating at
the characteristic velocity g h relative to the stream.
Tidal bores are found in river estuaries where a funneling eect causes the speed
of the incoming tide to increase to such a point that the flow becomes super-critical.
For example, bores can be observed daily on the River Severn in England.
110
v(x)
h(x)
d(x)
x
Figure 4.8
Flow over a broad-crested weir.
v2
,
2g
(4.64)
and
Q = h v,
(4.65)
where (via continuity) the flow rate, Q, is not a function of x. It follows that
H d(x) = F(v),
(4.66)
where F(v) is defined in Equation (4.21). Suppose that the weir attains its maximum
height, d(0), at x = 0. It follows that H d(x) passes through a minimum value at
x = 0. Hence, we would expect F(v) to also pass through its minimum value at x = 0.
It follows that
H d(0) = Fc =
1/3
3 Q2
,
2 g
v(0) = vc = (Q g)1/3,
(4.67)
(4.68)
111
where use has been made of Equations (4.22) and (4.23). However,
Q = h(0) v(0),
(4.69)
where h(0) is the depth of the stream as its passes over the highest point of the weir.
Thus, we deduce that
Q = [g h (0)]
3
1/2
3/2
1/2
2
=
.
g [H d(0)] 3
3
(4.70)
In other words, the flow rate (per unit width) is very simply related to the depth of
the stream as it passes over the highest point of the weir, or, alternatively, the depth
of the highest point of the weir below the water surface in the reservoir feeding the
stream. For this reason, weirs are commonly used as devices to both measure and
control flow rates in streams.
(4.71)
A vortex line is a line whose tangent is everywhere parallel to the local vorticity
vector. The vortex lines drawn through each point of a closed curve constitute the
surface of a vortex tube. Finally, a vortex filament is a vortex tube whose crosssection is of infinitesimal dimensions.
Consider a section AB of a vortex filament. The filament is bounded by the
curved surface that forms the filament wall, as well as two plane surfaces, whose
vector areas are S1 and S2 (say), which form the ends of the section at points A and B,
respectively. (See Figure 4.9.) Let the plane surfaces have outward pointing normals
that are parallel (or anti-parallel) to the vorticity vectors, 1 and 2 , at points A and
B, respectively. The divergence theorem (see Section A.20), applied to the section,
yields
dS =
dV,
(4.72)
(4.74)
112
S2
B
A
S1
Figure 4.9
A vortex filament.
Now, dS = 0 on the curved surface of the filament, because is, by definition,
tangential to this surface. Thus, the only contributions to the surface integral come
from the plane areas S1 and S2 . It follows that
dS = S 2 2 S 1 1 = 0.
(4.75)
This result is essentially an equation of continuity for vortex filaments. It implies
that the product of the magnitude of the vorticity and the cross-sectional area, which
is termed the vortex intensity, is constant along the filament. It follows that a vortex
filament cannot terminate in the interior of the fluid. For, if it did, the cross-sectional
area, S , would have to vanish, and, therefore, the vorticity, , would have to become infinite. Thus, a vortex filament must either form a closed vortex ring, or must
terminate at the fluid boundary.
Because a vortex tube can be regarded as a bundle of vortex filaments whose net
intensity is the sum of the intensities of the constituent filaments, we conclude that
the intensity of a vortex tube remains constant along the tube.
113
where dr is an element of C, and the integral is taken around the whole curve, is
termed the circulation of the flow around the curve. The sense of circulation (i.e.,
either clockwise or counter-clockwise) is arbitrary.
Let S be a surface having the closed curve C for a boundary, and let dS be an
element of this surface (see Section A.7) with that direction of the normal which is
related to the chosen sense of circulation around C by the right-hand circulation rule.
(See Section A.8.) According to the curl theorem (see Section A.22),
C =
v dr =
dS.
(4.77)
C
Thus, we conclude that circulation and vorticity are intimately related to one another.
In fact, according to the previous expression, the circulation of the fluid around loop
C is equal to the net sum of the intensities of the vortex filaments passing through
the loop and piercing the surface S (with a filament making a positive, or negative,
contribution to the sum depending on whether it pierces the surface in the direction
determined by the chosen sense of circulation around C and the right-hand circulation
rule, or in the opposite direction). One important proviso to Equation (4.77) is that
the surface S must lie entirely within the fluid.
However, for a loop that is co-moving with the fluid, we have dv = d(dr/dt) =
d(dr)/dt. Thus,
dC
dv
=
dr +
v dv.
(4.79)
dt
C dt
C
By definition, dv/dt = Dv/Dt for a co-moving loop. (See Section 1.10.) Moreover, the equation of motion of an incompressible inviscid fluid can be written [see
Equation (1.79)]
p
Dv
=
+ ,
(4.80)
Dt
p 1
dC
=
v 2 + dr = 0,
(4.81)
dt
2
C
because v dv = d(v 2 /2) = (v 2 /2) dr (see Section A.18), and p/ v 2 /2 + is
obviously a single-valued function.
114
C
Figure 4.10
A vortex tube.
One corollary of the Kelvin circulation theorem is that the fluid particles that
form the walls of a vortex tube at a given instance in time continue to form the
walls of a vortex tube at all subsequent times. To prove this, imagine a closed loop
C that is embedded in the wall of a vortex tube but does not circulate around the
interior of the tube. (See Figure 4.10.) The normal component of the vorticity over
the surface enclosed by C is zero, because all vorticity vectors are tangential to this
surface. Thus, from Equation (4.77), the circulation around the loop is zero. By
Kelvins circulation theorem, the circulation around the loop remains zero as the
tube is convected by the fluid. In other words, although the surface enclosed by C
deforms, as it is convected by the fluid, it always remains on the tube wall, because
no vortex filaments can pass through it.
Another corollary of the circulation theorem is that the intensity of a vortex tube
remains constant as it is convected by the fluid. This can be proved by considering
the circulation around the loop C pictured in Figure 4.10.
115
v
p
+ (v ) v =
+ ,
t
(4.82)
Thus, we obtain
(v ) v = (v 2 /2) v .
(4.83)
p 1 2
v
=
+ v + + v .
t
2
(4.84)
Taking the curl of this equation, and making use of the vector identities 0
[see Equation (A.176)], A 0 [see Equation (A.173)], as well as the identity
(A.179), and the fact that v = 0 in an incompressible fluid, we obtain the vorticity
evolution equation
D
= ( ) v.
(4.85)
Dt
Thus, if = 0, initially, then D/Dt = 0, and, consequently, = 0 at all subsequent
times.
Suppose that O is a fixed point, and P an arbitrary movable point, in an irrotational fluid. Let O and P be joined by two dierent paths, OAP and OBP (say). It
follows that OAPBO is a closed curve. Because the circulation around such a curve
in an irrotational fluid is zero, we can write
v dr +
v dr = 0,
(4.86)
OAP
PBO
v dr =
OAP
v dr = P
(4.87)
OBP
(say). It is clear that P is a scalar function whose value depends on the position of P
(and the fixed point O), but not on the path taken between O and P. Thus, if O is the
origin of our coordinate system, and P an arbitrary point whose position vector is r,
P
then we have eectively defined a scalar field (r) = O v dr.
Consider a point Q that is suciently close to P that the velocity v is constant
along PQ. Let be the position vector of Q relative to P. It then follows that (see
Section A.18)
Q
= Q + P =
v dr v .
(4.88)
The previous equation becomes exact in the limit that || 0. Because Q is arbitrary (provided that it is suciently close to P), the direction of the vector is also
arbitrary, which implies that
v = .
(4.89)
We, thus, conclude that if the motion of a fluid is irrotational then the associated
velocity field can always be expressed as minus the gradient of a scalar function of
116
position, (r). This scalar function is called the velocity potential, and flow which
is derived from such a potential is known as potential flow. Note that the velocity
potential is undefined to an arbitrary additive constant.
We have demonstrated that a velocity potential necessarily exists in a fluid whose
velocity field is irrotational. Conversely, when a velocity potential exists the flow is
necessarily irrotational. This follows because [see Equation (A.176)]
= v = = 0.
(4.90)
Incidentally, the fluid velocity at any given point in an irrotational fluid is normal to
the constant- surface that passes through that point.
If a flow pattern is both irrotational and incompressible then we have
v =
(4.91)
v = 0.
(4.92)
and
These two expressions can be combined to give (see Section A.21)
2 = 0.
(4.93)
p 1 2
+ v +
= 0,
(4.94)
2
t
which implies that
p 1 2
+ v +
= C(t),
(4.95)
2
t
where C(t) is uniform in space, but can vary in time. In fact, the time variation of C(t)
can be eliminated by adding the appropriate function of time (but not of space) to the
velocity potential, . Note that such a procedure does not modify the instantaneous
velocity field v derived from . Thus, the previous equation can be rewritten
p 1 2
+ v +
= C,
2
t
(4.96)
117
4.16 Exercises
4.1 Liquid is led steadily through a pipeline that passes over a hill of height h
into the valley below, the speed at the crest being v. Show that, by properly
adjusting the ratio of the cross-sectional areas of the pipe at the crest and
in the valley, the pressure may be equalized at these two places. (MilneThomson 1958.)
4.2 Water of mass density and pressure p flows through a curved pipe of uniform cross-sectional area S , whose radius of curvature is R, at the uniform
speed v. Demonstrate that there is a net force per unit length S (pp0 + v 2 )/R
acting on the pipe, and that this force is everywhere directed away from the
pipes local center of curvature. Here, p0 is atmospheric pressure.
4.3 Water is held in a right circular conical tank whose apex lies vertically below
the center of its base. The water initially fills the tank to a height h above the
vertex. Let a be the initial radius of the surface of the water inside the tank.
A small hole of area S (that is much less than a 2) is made at the bottom of
the tank. Demonstrate that the time required to empty the tank is at least
1/2
5 a2 h
.
2 S 2g
4.4 Water is held in a spherical tank of radius a, and initially fills the tank to a
height h < 2 a above its lowest point. A small hole of area S (that is much
less than a 2 ) is made at the bottom of the tank. Demonstrate that the time
required to empty the tank is at least
2
2
1 5/2
3/2
a
h
h
.
5
S (2 g)1/2 3
4.5 Water is held in two contiguous tanks whose cross-sectional areas, A1 and
A2 , are independent of height. A small hole of area S (where S A1 , A2 )
is made in the wall connecting the tanks. Assuming that the initial dierence
in water level between the two tanks is h, show that the time required for the
water levels to equilibrate is at least
1/2
h
2 A1 A2
.
S A1 + A2 2 g
4.6 For a channel of width W, having a discharge rate Q, show that there is a
critical depth hc , where
2 1/3
Q
hc =
,
gW2
which must be exceeded before a hydraulic jump is possible.
118
4.7 Show that for a stationary hydraulic jump in a rectangular channel, the upstream Froude number Fr1 , and the downstream Froude number Fr2 , are related by
8 Fr12
Fr22 =
.
[(1 + 8 Fr12 )1/2 1] 3
4.8 Consider a simply-connected volume V whose boundary is the surface S .
Suppose that V contains an incompressible fluid whose motion is irrotational.
Let the velocity potential be constant over S . Prove that has the same
constant value throughout V. [Hint: Consider the identity (A A) A
A + A 2 A.]
4.9 In Exercise 4.8, suppose that, instead of taking a constant value on the
boundary, the normal velocity is everywhere zero on the boundary. Show
that is constant throughout V.
4.10 An incompressible fluid flows in a simply-connected volume V bounded by a
surface S . The normal flow at the boundary is prescribed. Show that the flow
pattern with the lowest kinetic energy is irrotational. This result is known
as the Kelvin minimum energy theorem. [Hint: Try writing v = + v,
where is the velocity potential of the irrotational flow pattern. Let v = 0
throughout V, and v dS = 0 on S . Show that the kinetic energy is lowest
when v = 0 throughout V.]
5
Two-Dimensional Incompressible Inviscid
Flow
5.1 Introduction
This chapter investigates two-dimensional, incompressible, inviscid flow. More information on this subject can be found in Batchelor 2000, and Milne-Thompson
2011.
(5.1)
Let A be a fixed point in the x-y plane, and let ABP and ACP be two curves,
also in the x-y plane, that join A to an arbitrary point P. (See Figure 5.1.) Suppose
that fluid is neither created nor destroyed in the region, R (say), bounded by these
curves. Because the fluid is incompressible, which essentially means that its density
is uniform and constant, fluid continuity requires that the rate at which the fluid flows
into the region R, from right to left (in Figure 5.1) across the curve ABP, is equal to
the rate at which it flows out the of the region, from right to left across the curve
ACP. The rate of fluid flow across a surface is generally termed the flux. Thus, the
flux (per unit length parallel to the z-axis) from right to left across ABP is equal to
the flux from right to left across ACP. Because ACP is arbitrary, it follows that the
flux from right to left across any curve joining points A and P is equal to the flux
from right to left across ABP. In fact, once the base point A has been chosen, this
flux only depends on the position of point P, and the time t. In other words, if we
denote the flux by then it is solely a function of the location of P and the time.
Thus, if point A lies at the origin, and point P has Cartesian coordinates (x, y), then
119
120
P
C
B
A
Figure 5.1
Two-dimensional flow.
we can write
= (x, y, t).
(5.2)
The function is known as the stream function. Moreover, the existence of a stream
function is a direct consequence of the assumed incompressible nature of the flow.
Consider two points, P1 and P2 , in addition to the fixed point A. (See Figure 5.2.)
Let 1 and 2 be the fluxes from right to left across curves AP1 and AP2 . Using
similar arguments to those employed previously, the flux across AP2 is equal to the
flux across AP1 plus the flux across P1 P2 . Thus, the flux across P1 P2 , from right to
left, is 2 1 . If P1 and P2 both lie on the same streamline then the flux across P1 P2
is zero, because the local fluid velocity is directed everywhere parallel to P1 P2 . It
follows that 1 = 2 . Hence, we conclude that the stream function is constant along
a streamline. The equation of a streamline is thus = c, where c is an arbitrary
constant.
Let P1 P2 = s be an infinitesimal arc of a curve that is suciently short that it
can be regarded as a straight-line. The fluid velocity in the vicinity of this arc can
be resolved into components parallel and perpendicular to the arc. The component
parallel to s contributes nothing to the flux across the arc from right to left. The
component perpendicular to s contributes v s to the flux. However, the flux is
equal to 2 1 . Hence,
2 1
.
(5.3)
v =
s
In the limit s 0, the perpendicular velocity from right to left across ds becomes
v =
d
.
ds
(5.4)
121
P2
P1
A
Figure 5.2
Two-dimensional flow.
and y-axes, we deduce that
vx =
vy =
,
y
.
x
(5.5)
(5.6)
(5.7)
Note that when the fluid velocity is written in this form then it immediately becomes
clear that the incompressibility constraint v = 0 is automatically satisfied [because
(AB) 0see Equations (A.175) and (A.176)]. It is also clear that the stream
function is undefined to an arbitrary additive constant.
The vorticity in two-dimensional flow takes the form
= z ez ,
where
z =
vy v x
.
x
y
(5.8)
(5.9)
2 2
+
= 2 .
x 2
y 2
(5.10)
122
(5.11)
When expressed in terms of cylindrical coordinates (see Section C.3), Equation (5.7) yields
v = vr (r, , t) er + v (r, , t) e ,
(5.12)
where
1
,
r
.
v =
r
vr =
(5.13)
(5.14)
1
1 2
z =
.
r
+ 2
r r r
r 2
(5.15)
,
x
vy = .
y
vx =
(5.17)
(5.18)
,
y
.
x
(5.19)
(5.20)
123
Finally, if the flow is both irrotational and incompressible then Equations (5.17)
(5.18) and (5.19)(5.20) hold simultaneously, which implies that
=
,
x
y
(5.21)
= .
x
y
(5.22)
or
2
2
2
2
=
= 2,
=
2
x y y x
x
y
(5.23)
2 2
+
= 0.
x 2 y 2
(5.24)
(5.25)
We conclude that, for two-dimensional, irrotational, incompressible flow, the velocity potential and the stream function both satisfy Laplaces equation. Equations (5.21)
and (5.22) also imply that
= 0.
(5.26)
In other words, the contours of the velocity potential and the stream function cross at
right-angles.
(5.27)
which corresponds to flow at the uniform speed V0 in a fixed direction that subtends
a (counter-clockwise) angle 0 with the x-axis. It follows, from Equations (5.5) and
(5.6), that the stream function for steady uniform flow takes the form
(x, y) = V0 (sin 0 x cos 0 y) .
(5.28)
(5.29)
124
(5.31)
where r = (x 2 + y 2 )1/2 . Consider a cylindrical surface S of unit height (in the zdirection) and radius r that is co-axial with the source. In a steady state, the rate at
which fluid crosses this surface must be equal to the rate at which the section of the
source enclosed by the surface emits fluid. Hence,
v dS = 2 r vr (r) = Q,
(5.32)
S
(r, ) =
Q
.
2
(5.34)
Note that the streamlines, = c, are directed radially away from the z-axis, as illustrated in Figure 5.3. Note, also, that the stream function associated with a line source
is multivalued. However, this does not cause any particular diculty, because the
stream function is continuous, and its gradient is single valued.
It follows from Equation (5.34) that /r = 2 / 2 = 0. Hence, according to
Equation (5.15), z = 2 = 0. In other words, the steady flow pattern associated
with a uniform line source is irrotational, and can, thus, be derived from a velocity
potential. In fact, it is easily demonstrated that this potential takes the form
(r, ) =
Q
ln r.
2
(5.35)
A uniform line sink, coincident with the z-axis, which absorbs fluid isotropically
125
Figure 5.3
Streamlines of the flow generated by a line source coincident with the z-axis.
at the steady rate of Q unit volumes per unit length per unit time has an associated
steady flow pattern
Q
er ,
(5.36)
v=
2 r
whose stream function is
Q
(r, ) =
.
(5.37)
2
This flow pattern is also irrotational, and can be derived from the velocity potential
(r, ) =
Q
ln r.
2
(5.38)
Consider a line source and a line sink of equal strength, which both run parallel
to the z-axis, and are located a small distance apart in the x-y plane. Such an arrangement is known as a doublet or dipole line source. Suppose that the line source, which
is of strength Q, is located at r = d/2 (where r is a position vector in the x-y plane),
and that the line sink, which is also of strength Q, is located at r = d/2. Let the
function
Q
Q
tan1 (y/x)
(5.39)
Q (r) = =
2
2
be the stream function associated with a line source of strength Q located at r = 0.
Thus, Q (r r0 ) is the stream function associated with a line source of strength Q
126
(5.40)
D sin( 0 )
,
2
r
(5.41)
where D = Q d is termed the strength of the dipole source. The previous stream
function is antisymmetric across the line = 0 joining the source to the sink. It
follows that the associated dipole flow pattern,
D cos( 0 )
,
(5.42)
2
r2
D sin( 0 )
v (r, ) =
,
(5.43)
2
r2
is symmetric across this line. Figure 5.4 shows the streamlines associated with a
dipole flow pattern characterized by D > 0 and 0 = 0. Note that the flow speed in a
dipole pattern falls o like 1/r 2 .
A dipole flow pattern is necessarily irrotational because it is a linear superposition
of two irrotational flow patterns. The associated velocity potential is
vr (r, ) =
(r, ) =
D cos( 0 )
.
2
r
(5.44)
127
1
0.8
0.6
0.4
0.2
y
0
0.2
0.4
0.6
0.8
1
1
0.5
0
x
0.5
Figure 5.4
Streamlines of the flow generated by a dipole line source (with D > 0) coincident
with the z-axis, and aligned along the x-axis. The flow is outward along the positive x-axis and inward along the negative x-axis. Positive and negative contours are
shown as solid and dashed lines, respectively.
or
.
(5.47)
2 r
According to Equations (5.13) and (5.14), the stream function associated with a
vortex filament of intensity that is coincident with the z-axis is
v (r) =
ln r.
(5.48)
2
Note that the streamlines, = c, circulate around the z-axis, as illustrated in Figure 5.5.
It can be seen, from Equation (5.48), that (/r)(r /r) = / = 0. Hence,
it follows from Equation (5.15) that z = 2 = 0. In other words, the flow
pattern associated with a straight vortex filament is irrotational. This is a somewhat
surprising result, because there is a net circulation of the flow around the filament,
and, according to Section 4.13, non-zero circulation implies non-zero vorticity. The
paradox can be resolved by supposing that the filament has a small, but finite, radius.
In fact, let the filament have the finite radius a, and be such that the vorticity is
uniform inside this radius, and zero outside: that is,
/( a 2)
ra
z =
.
(5.49)
0
r>a
(r, ) =
128
Figure 5.5
Streamlines of the flow generated by a line vortex coincident with the z-axis.
Note that the intensity of the filament (i.e., the product of its vorticity and crosssectional area) is still . According to Equation (5.15), and assuming that = (r),
ra
/( a 2)
1 d d
.
(5.50)
r
=
r r dr
0
r>a
The solution that is well behaved at r = 0, and continuous (up to its first derivative)
at r = a, is
(/4) (r 2/a 2 1)
ra
(r, ) =
.
(5.51)
(/2) ln(r/a)
r>a
This expression is equivalent to Equation (5.48) (apart from an unimportant additive
constant) outside the filament, but diers inside. The associated circulation velocity,
v (r) = /r, is
ra
(/2) (r/a 2)
,
(5.52)
v (r) =
(/2) (1/r)
r>a
whereas the circulation, r (r) = 2 r v (r), is written
ra
(r/a)2
r (r) =
.
r>a
(5.53)
Thus, we conclude that the flow pattern associated with a straight vortex filament is
129
irrotational outside the filament, but has finite vorticity inside the filament. Moreover,
the non-zero internal vorticity generates a constant net circulation of the flow outside
the filament. In the limit in which the radius of the filament tends to zero, the vorticity
within the filament tends to infinity (in such a way that the product of the vorticity
and the cross-sectional area of the filament remains constant), and the region of the
fluid in which the vorticity is non-zero becomes infinitesimal in extent.
Let us determined the pressure profile in the vicinity of a vortex filament of finite
radius. Assuming, from symmetry, that p = p(r), Equation (1.149), yields
v2
dp
= ,
dr
r
(5.54)
v2
dr,
r
(5.55)
ra
.
r>a
(5.56)
It follows that the minimum pressure occurs at the center of the vortex (r = 0), and
takes the value
$ %2
.
(5.57)
p0 = p
2 a
Under normal circumstances, the pressure in a fluid must remain positive, which
implies that a vortex filament of intensity , embedded in a fluid of density and
background pressure p , has a minimum radius of order
amin
1/2 $
%
.
2
(5.58)
Finally, because the flow pattern outside a straight vortex filament is irrotational,
it can be derived from a velocity potential. It is easily demonstrated that the appropriate potential takes the form
(r, ) =
.
2
(5.59)
Note that the previous potential is multivalued. However, this does not cause any
particular diculty, because the potential is continuous, and its gradient is single
valued.
130
(5.60)
In cylindrical coordinates, because = (r, , t), this expression implies that (see
Section C.3)
1
1 2
= 0.
(5.61)
r
+ 2
r r r
r 2
Let us search for a separable solution of Equation (5.61) of the form
(5.62)
r d dR
1 d 2
,
r
=
R dr dr
d 2
(5.63)
d dR
r
r
= m 2 R,
dr dr
d 2
= m 2 ,
d 2
(5.64)
(5.65)
where m , m , and m are arbitrary constants. We can recognize the first few terms
on the right-hand side of the previous expression. The constant term 0 has zero
gradient, and, therefore, does not give rise to any flow. The term 0 ln r is the flow
131
5
4
3
2
1
y/a
0
1
2
3
4
5
5 4 3 2 1 0 1
x/a
Figure 5.6
Streamlines of the flow generated by a cylindrical obstacle of radius a, whose axis
runs along the z-axis, placed in the uniform flow field v = V0 e x . The normalized
circulation is = 0.
1
= .
r
r
(5.67)
(5.68)
It follows that
(r, ) = 0 0 +
m>0
(m r m m rm ) cos[m ( 0 )].
(5.69)
132
(5.73)
a
a r
where
=
,
2 a V0
(5.74)
and is the circulation of the flow around the cylinder. (Note that the velocity field
can be irrotational, but still possess nonzero circulation around the cylinder, because
a loop that encloses the cylinder cannot be spanned by a surface lying entirely within
the fluid. Thus, zero fluid vorticity does not necessarily imply zero circulation around
such a loop from the curl theorem.) Let us assume that 0, for the sake of
definiteness.
Figure 5.65.8 show streamlines of the flow calculated for various dierent values of the normalized circulation, . For < 2, there exist a pair of points on the
133
5
4
3
2
1
y/a
0
1
2
3
4
5
5 4 3 2 1 0 1
x/a
Figure 5.7
Streamlines of the flow generated by a cylindrical obstacle of radius a, whose axis
runs along the z-axis, placed in the uniform flow field v = V0 e x . The normalized
circulation is = 1.
surface of the cylinder at which the flow speed is zero. These are known as stagnation
points, and can be located in Figures 5.6 and 5.7 as the points at which streamlines
intersect the surface of the cylinder at right-angles. The tangential fluid velocity at
the surface of the cylinder is
= V0 ( + 2 sin ).
(5.75)
vt () = v (a, ) =
r r=a
The stagnation points correspond to the points at which vt = 0 (because the normal
velocity is automatically zero at the surface of the cylinder). Thus, the stagnation
points lie at = sin1 (/2). When > 2, the stagnation points coalesce and
move o the surface of the cylinder, as illustrated in Figure 5.8 (the stagnation point
corresponds to the point at which two streamlines cross at right-angles).
The irrotational form of Bernoullis theorem, (4.97), can be combined with the
boundary condition v V0 as r/a , as well as the fact that is constant in the
present case, to give
1
p = p0 + V02 v 2 ,
(5.76)
2
where p0 is the constant static fluid pressure a large distance from the cylinder. In
particular, the fluid pressure on the surface of the cylinder is
1
(5.77)
P() = p(a, ) = p0 + V02 vt2 = p1 + V02 (cos 2 2 sin ) ,
2
134
where p1 = p0 (1/2) V02 (1 + 2 ). The net force per unit length exerted on the
cylinder by the fluid has the Cartesian coordinates
F x = P cos a d,
(5.78)
(5.79)
Fy = P sin a d.
Thus, it follows from Equation (5.77) that
F x = 0,
Fy =
2 V02
(5.80)
a = V0 ().
(5.81)
The component of the force that a moving fluid exerts on an obstacle, placed in its
path, in a direction parallel to that of the unperturbed flow is usually called drag. On
the other hand, the component of the force that the fluid exerts in a direction perpendicular to that of the unperturbed flow is usually called lift. Hence, the previous
equations imply that if a cylindrical obstacle is placed in a uniformly flowing inviscid
fluid then there is zero drag. On the other hand, as long as there is net circulation of
the flow around the cylinder, the lift is non-zero. Lift is generated because (negative)
circulation tends to increase the fluid speed directly above, and to decrease it directly
below, the cylinder. Thus, from Bernoullis theorem, the fluid pressure is decreased
above, and increased below, the cylinder, giving rise to a net upward force (i.e., a
force in the +y-direction).
Suppose that the cylinder is placed in a fluid which is initially at rest, and that the
fluids uniform flow velocity, V0 , is then very slowly ramped up (in such a manner
that no vorticity is induced in the upstream flow at infinity). Because the flow pattern
is initially irrotational, and because the flow pattern well upstream of the cylinder is
assumed to remain irrotational, the Kelvin circulation theorem indicates that the flow
pattern around the cylinder
also remains irrotational. Consider the time evolution of
#
the circulation, = C v dr, around some fixed curve C that lies entirely within the
fluid, and encloses the cylinder. We have
p 1 2
v
dC
=
dr =
+ v + v dr =
v dr, (5.82)
dt
2
C t
C
C
where use has been made of Equation (4.84) (with assumed constant). However,
= z ez in two-dimensional flow, and dr ez = dS, where dS is an outward surface
element of a unit depth (in the z-direction) surface whose normal lies in the x-y plane,
and that cuts the x-y plane at C. In other words,
dC
= z v dS.
(5.83)
dt
S
We, thus, conclude that the rate of change of the circulation around C is equal to
minus the flux of the vorticity across S [assuming that vorticity is convected by the
flow, which follows from Equation (4.85), the fact that = z ez , and the fact that
135
5
4
3
2
1
y/a
0
1
2
3
4
5
5 4 3 2 1 0 1
x/a
Figure 5.8
Streamlines of the flow generated by a cylindrical obstacle of radius a, whose axis
runs along the z-axis, placed in the uniform flow field v = V0 e x . The normalized
circulation is = 2.5.
136
r
V
a
Figure 5.9
Cylinder moving through an inviscid fluid
(5.84)
where v is the fluid velocity. Moreover, because the fluid is incompressible, we have
v = 2 = 0.
(5.85)
Let x, y, z be Cartesian coordinates in the initial rest frame of the fluid, and let r,
be cylindrical coordinates in a frame of reference that co-moves with the cylinder,
as shown in Figure 5.9. In the following, all calculations are performed in the rest
frame. We expect the fluid a long way from the cylinder to remain stationary. In
other words,
(r, , t) 0
as r .
(5.86)
(5.87)
137
= V x cos .
(5.88)
r r=a
It is easily demonstrated that the solution to Equation (5.85), subject to the boundary
conditions (5.86) and (5.88), is
or
(r, , t) = V x (t)
a2
cos .
r
(5.89)
Hence,
a2
cos ,
r2
a2
v (r, , t) = V x (t) 2 sin .
r
vr (r, , t) = V x (t)
(5.90)
(5.91)
where is the uniform fluid mass density, and p0 the fluid pressure at infinity. Thus,
the pressure distribution at the surface of the cylinder can be written
1
V |r=a .
(5.93)
p(a, , t) = p0 v 2 +
2
t r=a
r=a
The final term on the right-hand side of the previous equation arises because
= V .
(5.94)
t x,y t r,
Hence, we obtain
p(a, , t) = p0
1
dV x
V x2 + V x2 cos(2 ) + a
cos .
2
dt
(5.95)
The net force per unit length exerted on the cylinder by the fluid has the Cartesian
components
(5.96)
F x (t) = p(a, , t) cos a d,
(5.97)
Fy (t) = p(a, , t) sin a d.
It follows that
F x (t) = a 2
Fy (t) = 0,
dV x
,
dt
(5.98)
(5.99)
138
or
F = m
dV
,
dt
(5.100)
where m = a 2 is the mass per unit length of the fluid displaced by the cylinder.
Suppose that the cylinder is subject to an external (i.e., not due to the fluid) force
per unit length Fext . The equation of motion of the cylinder is thus
m
dV
= F + Fext ,
dt
(5.101)
where m is the cylinders mass per unit length. The previous two equations can be
combined to give
dV
= Fext .
(5.102)
(m + m )
dt
In other words, the cylinder moves under the action of the external force, Fext , as
if its mass per unit length were m + m , rather than m. Here, m + m is commonly
referred to as the cylinders virtual mass (per unit length), whereas m is termed the
added mass (per unit length).
The origin of added mass is easily explained. According to Equations (5.90) and
(5.91), the total kinetic energy per unit length of the fluid surrounding the cylinder is
$ a %4
1 2
1
1
v dV = 2
V x2
r dr = m V x2 .
(5.103)
Kfluid =
r
2
fluid 2
a 2
However, the kinetic energy per unit length of the cylinder is
Kcylinder =
1
m V x2 .
2
(5.104)
1
(m + m ) V x2 .
2
(5.105)
In other words, the kinetic energy of the fluid surrounding the cylinder can be accounted for by supposing that a mass (per unit length) m of the fluid co-moves with
the cylinder, and that the remainder of the fluid remains stationary. This entrained
fluid mass accounts for the added mass of the cylinder. Note that the added mass
is independent of the speed of the cylinder (i.e., it is the same whether the cylinder
moves slowly or rapidly.) In the present case, the added mass is equal to the mass of
the displaced fluid. However, this is not a general rule. (In general, the added mass
of a object moving through an inviscid fluid is proportional to the displaced mass,
but the constant of proportionality is not necessarily unity, and depends on the shape
of the object.)
Let us generalize the previous calculation to allow the cylinder to move in any
direction in the x-y plane: that is,
V(t) = V x (t) e x + Vy (t) ey .
(5.106)
139
Furthermore, let the fluid possess the initial circulation in the x-y plane. According
to the Kelvin circulation theorem, this circulation remains constant in time. Thus, we
must now solve Equation (5.85) subject to the boundary conditions
2 r
as r .
(5.107)
= V x cos Vy sin .
r r=a
It is easily demonstrated that the appropriate solution is
(5.108)
(r, , t)
and
(r, , t) =
a2
a2
+ V x (t)
cos + Vy (t)
sin .
2 r
r
r
(5.109)
Hence,
a2
a .2
cos
+
V
(t)
sin ,
y
r2
r2
a2
a2
+ V x (t) 2 sin Vy (t) 2 cos .
v (r, , t) =
2 r
r
r
vr (r, , t) = V x (t)
(5.110)
(5.111)
(5.112)
where we have assumed that the fluid and cylinder are both situated in a gravitational
field of uniform acceleration g = g ey . Thus, the pressure distribution at the surface
of the cylinder can be written
1
V |r=a ,
p(a, , t) = p0 v 2 g a sin
(5.113)
2
t r=a
r=a
which yields
$ %2
1
2
p(a, , t) = p0 V +
+ (V x2 Vy2 ) cos(2 ) + 2 V x Vy sin(2 )
2
2 a
dVy
dV x
cos + a
sin g a sin
dr
dr
V x sin Vy cos .
a
+ a
(5.114)
It follows from Equations (5.96) and (5.97) that the force per unit length exerted on
the cylinder by the fluid has the Cartesian components
dV x
,
dt
dVy
+ m g.
Fy (t) = V x m
dt
F x (t) = Vy m
(5.115)
(5.116)
140
Here, the first terms on the right-hand sides of the previous two equations are the
components of the lift (per unit length) acting on the cylinder, due to the fluid circulation, whereas the final term on the right-hand side of the second equation is the
buoyancy force (per unit length) acting on the cylinder. The cylinders equation of
motion,
dV x
= Fx,
dt
dVy
= Fy m g,
m
dt
(5.117)
(5.118)
leads to
d 2X
= 2 Y,
dt 2
m m
d 2Y
2
=
X
+
g,
m + m
dt 2
(5.119)
(5.120)
where = /(m + m ), and (X, Y) are the Cartesian components of the cylinders
axis. Let us assume that X = Y = dX/dt = dY/dt = 0 at t = 0. It follows that
m m g
X(t) =
[sin( t) t] ,
(5.121)
m + m 2
m m g
[1 cos( t)] .
(5.122)
Y(t) =
m + m 2
Consider, first, the case in which there is no circulation of the flow: that is, = 0.
In this case, the previous two equations reduce to
X(t) = 0,
Y(t) =
mm
g t 2.
m + m
(5.123)
(5.124)
In other words, the cylinder moves vertically (i.e., in the y-direction) with the constant acceleration
s1
g,
(5.125)
ay =
s+1
where s = m/m is the cylinders specific gravity. It follows that if the cylinder is
much denser than the fluid (i.e., s
1) then it accelerates downward at the acceleration due to gravity: that is, ay = g. However, if the cylinder is much less dense
than the fluid (i.e., 0 < s 1) then it accelerates upward at the acceleration due to
gravity: that is ay = +g. Note that, in the latter case, the upward acceleration is limited by the cylinders added mass (i.e., in the absence of added mass, the acceleration
would be infinite.)
In the general case, in which the fluid circulation is non-zero, the trajectory of
the cylinder is a cycloid. In particular, assuming that m > m , the lift acting on
141
Figure 5.10
Inviscid flow past a wedge.
the cylinder prevents it from falling through the fluid a distance greater than 2 [(m
m )/(m + m )] (g/ 2). Once the cylinder has fallen through this distance, it starts
to rise again, until it attains its original height, and the motion then repeats itself
ad infinitum. Moreover, the cylinder simultaneously moves horizontally (i.e. in the
x-direction) at the mean velocity [(m m )/(m + m )] (g/).
(5.126)
The first two boundary conditions ensure that the normal velocity at the surface of
the wedge is zero. The third boundary condition follows from the observation that,
142
3
2
1
y
0
1
2
3
3
0
x
Figure 5.11
Streamlines of inviscid incompressible irrotational flow past a 90 wedge.
by symmetry, the streamline that meets the apex of the wedge splits in two, and then
flows along its top and bottom boundaries, combined with the well-known result that
is constant on a streamline. It is easily demonstrated that
(r, ) =
A
r 1+m sin [(1 + m) ( )]
1+m
(5.127)
is a solution of Equation (5.61). Moreover, this solution satisfies the boundary conditions provided (1 + m) (1 /2) = 1, or
m=
.
2
(5.128)
Because, as is well known, the solutions to Laplaces equation (for problems with
well-posed boundary conditions) are unique (Riley 1974), we can be sure that Equation (5.127) is the correct solution to the problem under investigation. According to
this solution, the tangential velocity on the surface of the wedge is given by
vt (r) = A r m ,
(5.129)
where m 0. Note that the tangential velocity is zero at the apex of the wedge.
Because the normal velocity is also zero at this point, we conclude that the apex is
a stagnation point of the flow. Figure 5.11 shows the streamlines of the flow for the
case = 1/2.
143
Figure 5.12
Inviscid flow over a wedge.
(5.130)
These boundary conditions ensure that the normal velocity at the surface of the
wedge is zero. It is easily demonstrated that
(r, ) =
A
r 1m cos [(1 m) ( /2)]
1m
(5.131)
,
1 +
(5.132)
144
3
2
1
y
0
1
2
3
3
0
x
Figure 5.13
Streamlines of inviscid incompressible irrotational flow over a 90 wedge.
where = 1 . Because the solutions to Laplaces equation are unique, we
can again be sure that Equation (5.131) is the correct solution to the problem under
investigation. According to this solution, the tangential velocity on the surface of the
wedge is given by
(5.133)
vt (r) = A rm ,
where m 0. Note that the tangential velocity, and hence the flow speed, is infinite
at the apex of the wedge. However, this singularity in the flow can be eliminated by
slightly rounding the apex. Figure 5.13 shows the streamlines of the flow for the case
= 1/2.
145
h1
v
v
h2
Figure 5.14
A two-dimensional jet impinging on a rigid plate.
longer accelerating (and, hence, their widths are constant). It immediately follows
from Bernoullis theorem that the asymptotic flow speed in the two streams is v, as
indicated in Figure 5.14.
Fluid continuity demands that h v = h1 v + h2 v, or
h = h1 + h2 .
(5.134)
Moreover, because the plate can only exert a normal force on the fluid, it follows
that the parallel (to the plate) momentum flux of the incident jet must equal the net
parallel momentum flux of the two streams. In other words,
h v 2 sin = h1 v 2 h2 v 2 ,
(5.135)
h sin = h1 h2 .
(5.136)
or
Equations (5.134) and (5.136) can be combined to give
h
(1 + sin ),
2
h
h2 = (1 sin ).
2
h1 =
(5.137)
(5.138)
Let us apply the Euler momentum theorem (see Section 4.4) to the flux tube
shown in Figure 5.14, treating the plate as one of the walls of the tube. By symmetry,
146
h1 v 2
h1 /2
hv2
h2 /2
h2 v 2
Figure 5.15
Determination of the center of pressure.
the net force (per unit width perpendicular to the page) exerted across the walls and
ends of the tube is the resultant of the three forces shown in Figure 5.15. It is easily
demonstrated that this net force, F, is normal to the plate, as shown in the figure, and
of magnitude
(5.139)
F = h v 2 cos .
Obviously, the force in question corresponds to the force exerted by the plate on the
fluid. An equal and opposite force is exerted by the fluid on the plate. (Note that there
is no net pressure force on the flux tube because it is entirely surrounded by fluid at
the same pressure: i.e., atmospheric pressure.) The line of action of the resultant
force passes through the so-called center of pressure, C. This is displaced slightly
from the point, O, where the middle of the jet strikes the plate. (See Figure 5.15.)
Let d be the distance OC. We can calculate d by demanding that the net moment of
the three forces shown in Figure 5.15 about point C be zero. In other words,
h v 2 cos d h1 v 2
h1
h2
+ h2 v 2
= 0,
2
2
(5.140)
which yields
d=
h
tan .
2
(5.141)
147
5.13 Exercises
5.1 For the case of the two-dimensional motion of an incompressible fluid, determine the condition that the velocity components
v x = a x + b y,
vy = c x + d y
satisfy the equation of continuity. Show that the magnitude of the vorticity is
c b.
5.2 For the case of the two-dimensional motion of an incompressible fluid, show
that
v x = 2 c x y,
vy = c (a 2 + x 2 y 2 )
are the velocity components of a possible flow pattern. Determine the stream
function and sketch the streamlines. Prove that the motion is irrotational, and
find the velocity potential.
5.3 A cylindrical vortex in an incompressible fluid is co-axial with the z-axis, and
such that z takes the constant value for r a, and is zero for r > a, where
r is a cylindrical coordinate. Show that
1 dp 2 r
= 4 ,
dr
a
where p(r) is the pressure at radius r inside the vortex, and the circulation of
the fluid outside the vortex is 2 . Deduce that
p(r) =
2 r2
+ p0 ,
2 a4
2
+ p ,
2r2
148
5.6 Prove that in the two-dimensional motion of a liquid the mean tangential fluid
velocity around any small circle of radius r is r, where 2 is the value of
vy v x
x
y
at the center of the circle. Neglect terms of order r 3 .
5.7 Show that the equation of continuity for the two-dimensional motion of an
incompressible fluid can be written
(r vr ) v
+
= 0,
r
where r, are cylindrical coordinates. Demonstrate that this equation is satisfied when vr = a k r n exp[k (n + 1) ] and v = a r n exp[k (n + 1) ].
Determine the stream function, and show that the fluid speed at any point is
(n + 1) 1 + k 2 /r,
where is the stream function at that point (defined such that = 0 at r = 0).
5.8 Demonstrate that streamlines cross at right-angles at a stagnation point in
two-dimensional, incompressible, irrotational flow.
5.9 Consider two-dimensional, incompressible, inviscid flow. Demonstrate that
the fluid motion is governed by the following equations:
+ [, ] = 0,
t
2 = ,
2 = + 2 ,
where v = ez , [A, B] = ez A B, and = p/ + (1/2) v 2 + .
5.10 For irrotational, incompressible, inviscid motion in two dimensions show that
q q = q 2 q,
where q = |v|.
6
Two-Dimensional Potential Flow
6.1 Introduction
This chapter describes the use of complex analysis to facilitate calculations in twodimensional, incompressible, irrotational fluid dynamics. Incidentally, incompressible, irrotational flow is usually referred to as potential flow, because the associated velocity field can be represented in terms of a velocity potential that satisfies
Laplaces equation. (See Section 4.15.) In the following, all flow patterns are assumed to be such that the z-coordinate is ignorable. In other words, the fluid velocity is everywhere parallel to the x-y plane, and /z = 0. It follows that all line
sources and vortex filaments run parallel to the z-axis. Moreover, all solid surfaces
are of infinite extent in the z-direction, and have uniform cross-sections. Hence, it
is only necessary to specify the locations of line sources, vortex filaments, and solid
surfaces in the x-y plane. More information on the use of complex analysis in twodimensional fluid mechanics can be found in Batchelor 2000, Milne-Thomson 1958,
Milne-Thomson 2011, and Lamb 1993.
(6.1)
where i represents the square root of minus one. Here, x and y are both real, and are
identified with the corresponding Cartesian coordinates. (Incidentally, z should not
be confused with a z-coordinate: this is a strictly two-dimensional discussion.) We
can also write
z = r e i ,
(6.2)
where r = x 2 + y 2 and = tan1 (y/x) are the modulus and argument of z, respectively, but can also be identified with the corresponding plane polar coordinates.
Finally, Eulers theorem (Riley 1974),
e i = cos + i sin ,
(6.3)
149
150
implies that
x = r cos ,
(6.4)
y = r sin .
(6.5)
We can define functions of the complex variable, F(z), in the same way that we
define functions of a real variable. For instance,
F(z) = z 2 ,
(6.6)
1
.
z
(6.7)
F(z) =
(6.8)
(6.9)
F(x + i y) = (x + i y)2 = (x 2 y 2 ) + 2 i x y,
(6.10)
(x, y) = x 2 y 2 ,
(6.11)
(x, y) = 2 x y.
(6.12)
then
giving
lim |z|
F(z + z) F(z)
.
z
(6.13)
(6.14)
151
=
+i
.
x
x
dF
=
dz
lim |x|0
(6.15)
Suppose that we now approach z along the imaginary axis, so that z = i y. We get
dF
=
dz
lim |y|0
= i
+
.
y y
(6.16)
But, if F(z) is a well-behaved function then its derivative must be well-defined, which
implies that the previous two expressions are equivalent. This requires that
=
,
x
y
(6.17)
= .
x
y
(6.18)
These expressions are called the Cauchy-Riemann relations, and are, in fact, sucient to ensure that all possible ways of taking the limit (6.13) give the same result
(Riley 1974).
The Cauchy-Riemann relations can be combined to give
2 2 2 2
+
= 0.
+
=
+
=
x x y y
x 2 y 2
x 2 y 2
(6.19)
In other words,
2 = 0,
(6.20)
= 0,
(6.21)
= 0.
(6.22)
It follows that the real and imaginary parts of a well-behaved function of the complex
variable both satisfy Laplaces equation. Furthermore, the contours of these functions
cross at right-angles.
152
(6.24)
(r, ) = V0 r sin .
(6.25)
It can be seen, by comparison with the analysis of Section 5.4, that the complex velocity potential (6.23) corresponds to uniform flow of speed V0 directed along the
x-axis. Furthermore, as is easily demonstrated, the complex velocity potential associated with uniform flow of speed V0 whose direction subtends a (counter-clockwise)
angle 0 with the x-axis is F(z) = V0 z ei 0 .
Suppose that
Q
ln z,
(6.26)
F(z) =
2
where Q is real. Because ln z = ln r + i (Riley 1974), it follows that
Q
ln r,
2
Q
(r, ) = .
2
(r, ) =
(6.27)
(6.28)
Thus, according to the analysis of Section 5.5, the complex velocity potential (6.26)
corresponds to the flow pattern of a line source, of strength Q, located at the origin.
(See Figure 5.3.) As a simple generalization of this result, the complex potential of a
line source, of strength Q, located at the point (x0 , y0 ), is F(z) = (Q/2) ln(z z0 ),
where z0 = x0 + i y0 . It can be seen, from Equation (6.26), that the complex velocity
potential of a line source is singular at the location of the source.
Suppose that
ln z,
(6.29)
F(z) = i
2
where is real. It follows that
,
2
(r, ) =
ln r.
2
(r, ) =
(6.30)
(6.31)
Thus, according to the analysis of Section 5.6, the complex velocity potential (6.29)
153
$z%
a2
ln
F(z) = V0 z +
+i
,
(6.32)
z
2
a
where V0 , a, and , are real. It follows that
a2
,
(r, ) = V0 r +
cos
r
2
$r%
a2
sin +
ln
.
(r, ) = V0 r
r
2
a
(6.33)
(6.34)
Thus, according to the analysis of Section 5.8, the complex velocity potential (6.32)
corresponds to uniform flow of unperturbed speed V0 , running parallel to the x-axis,
around an impenetrable circular cylinder of radius a, centered on the origin. (See
Figures 5.6, 5.7, and 5.8.) Here, is the circulation of the flow about the cylinder.
It can be seen that = 0 on the surface of the cylinder (r = a), which ensures
that the normal velocity is zero on this surface (because the surface corresponds to a
streamline), as must be the case if the cylinder is impenetrable.
=
+i
= v x + i vy .
dz
x
x
Consequently, dF/dz is termed the complex velocity. It follows that
2
dF = v 2 + v 2 = v 2 ,
x
y
dz
(6.35)
(6.36)
(6.37)
at a stagnation point. For instance, the stagnation points of the flow pattern produced
154
a2
dF
= V0 1 2 + i
= 0.
(6.38)
dz
2 z
z
z
= i 1 2 ,
(6.39)
a
where = /(4 V0 a), with the proviso that |z|/a > 1, because the region |z|/a < 1
is occupied by the cylinder. Thus, if
1 then there are two stagnation points on
the surface of the cylinder at x/a = 1 2 and y/a = . On the other hand,
if > 1 then
there is a single stagnation point below the cylinder at x/a = 0 and
y/a = 2 1.
According to Section 4.15, Bernoullis theorem in an steady, irrotational, incompressible fluid takes the form
The solutions are
p+
1 2
v = p0 ,
2
(6.40)
where p0 is a uniform constant. Here, gravity (and any other body force) has been
neglected. Thus, the pressure distribution in such a fluid can be written
1 dF 2
p = p0 .
2 dz
(6.41)
155
Q
a
x
a
Q
Figure 6.1
Two line sources.
the resulting flow pattern is the sum of the complex potentials of each source taken
in isolation. Hence, from Section 6.4,
F(z) =
Q
Q
Q
ln(z i a)
ln(z + i a) =
ln(z 2 + a 2 ).
2
2
2
(6.42)
Thus, the stream function of the flow pattern (which is the imaginary part of the
complex potential) is
(x, y) =
2 xy
Q
tan1 2
.
2
x y2 + a2
(6.43)
Note that (x, 0) = 0, which implies that there is zero flow normal to the plane y = 0.
Hence, in the region y > 0, we could interpret the previous stream function as that
generated by a single line source of strength Q, located at the point (0, a), in the
presence of a planar rigid boundary at y = 0. This follows because the stream function satisfies 2 = 0 everywhere in the region y > 0, has the requisite singularity
(corresponding to a line source of strength Q) at (0, a), and satisfies the physical
boundary condition that the normal velocity be zero at the rigid boundary. Moreover,
as is well-known, the solutions of Laplaces equation are unique. The streamlines
of the resulting flow pattern are shown in Figure 6.2. Incidentally, we can think of
the two sources in Figure 6.1 as the images of one another in the boundary plane.
Hence, this method of calculation is usually referred to as the method of images.
156
6
5
4
y/a 3
2
1
0
3
0
x/a
Figure 6.2
Stream lines of the two-dimensional flow pattern due to a line source at (0, a) in the
presence of a rigid boundary at y = 0.
The complex velocity associated with the complex velocity potential (6.42) is
dF
Q
z
=
.
2
dz
z + a2
Hence, the flow speed at the boundary is
|x|/a
Q
dF
.
v(x, 0) = (x, 0) =
dz
a 1 + x 2 /a 2
(6.44)
(6.45)
It follows from Equation (6.41) (and the fact that the flow speed at infinity is zero)
that the excess pressure on the boundary, due to the presence of the source, is
x 2 /a 2
$ Q %2
dF 2
.
(6.46)
p(x, 0) = =
2 dz y=0
2 a (1 + x 2 /a 2 )2
Thus, the excess force per unit length (in the z-direction) acting on the boundary in
the y-direction is
2
Q2
Q2
.
(6.47)
Fy =
p(x, 0) dx = 2
d
=
4 a
2 a (1 + 2 )2
The fact that the force is positive implies that the boundary is attracted to the source,
and vice versa.
157
a
x
a
Figure 6.3
Two vortex filaments.
As a second example, consider the situation, illustrated in Figure 6.3, in which
there are two vortex filaments of intensities and situated at (0, a). As before,
the complex velocity potential of the resulting flow pattern is the sum of the complex
potentials of each filament taken in isolation. Hence, from Section 6.4,
z ia
ln(z i a) i
ln(z + i a) = i
ln
F(z) = i
.
(6.48)
2
2
2
z+ ia
Thus, the stream function of the flow pattern is
2
x + (y a)2
ln 2
.
(x, y) =
2
x + (y + a)2
(6.49)
As before, (x, 0) = 0, which implies that there is zero flow normal to the plane y = 0
(because the plane corresponds to a streamline). Hence, in the region y > 0, we could
interpret the previous stream function as that generated by a single vortex filament
of intensity , located at the point (0, a), in the presence of a planar rigid boundary
at y = 0. The streamlines of the resulting flow pattern are shown in Figure 6.4. We
conclude that a vortex filament reverses its sense of rotation (i.e., ) when
reflected in a boundary plane.
As a final example, consider the situation, illustrated in Figure 6.5, in which
there is an impenetrable circular cylinder of radius a, centered on the origin, and a
line source of strength Q located at (b, 0), where b > a. Consider the so-called analog
problem, also illustrated in Figure 6.5, in which the cylinder is replaced by a source
158
6
5
4
y/a 3
2
1
0
3
0
x/a
Figure 6.4
Stream lines of the two-dimensional flow pattern due to a vortex filament at (0, a) in
the presence of a rigid boundary at y = 0.
of strength Q, located at (c, 0), where c < a, and a source of strength Q, located at
the origin. We can think of these two sources as the images of the external source
in the cylinder. Moreover, given that the solutions of Laplaces equation are unique,
if the analog problem can be adjusted in such a manner that r = a is a streamline then
the flow in the region r > a will become identical to that in the actual problem. The
complex velocity potential in the analog problem is simply
(z b) (z c)
Q
Q
Q
Q
ln(z b)
ln(z c) +
ln z = ln
. (6.50)
F(z) =
2
2
2
2
z
Hence, writing z = r e i , the corresponding stream function takes the form
(r b c/r) sin
Q
1
(r, ) =
tan
.
2
(r + b c/r) cos (b + c)
(6.51)
159
Q
a
Q
c
Figure 6.5
A line source in the presence of an impenetrable circular cylinder.
4
3
2
y/a
1
0
1
2
3
4
4 3 2 1
0
x/a
Figure 6.6
Stream lines of the two-dimensional flow pattern due to a line source at (2 a, 0) in
the presence of a rigid circular cylinder of radius a centered on the origin.
160
(6.53)
0 + d = f (z0 + dz ).
(6.54)
In other words, the points z0 + dz and z0 + dz in the complex z-plane map to the
points 0 + d and 0 + d in the complex -plane, respectively. If |dz |, |dz | 1
then
d = f (z0 ) dz ,
(6.55)
d = f (z0 ) dz ,
(6.56)
d dz
= .
d
dz
(6.57)
|d | |dz |
=
,
|d |
|dz |
(6.58)
(6.59)
and
We can think of dz and dz as infinitesimal vectors connecting neighboring points
in the complex z-plane to the point z = z0 . Likewise, d and d are infinitesimal
vectors connecting the corresponding points in the complex -plane. It is clear, from
the previous two equations, that, in the vicinity of z = z0 , the mapping from the
complex z-plane to the complex -plane is such that the lengths of dz and dz expand
or contract by the same factor, and the angle subtended between these two vectors
remains the same. (See Figure 6.7.) This type of mapping is termed conformal.
Suppose that F() = (, ) + i (, ) is a well-behaved function of the complex
variable . It follows that 2 = 2 = = 0. Hence, the functions (, ) and
(, ) can be interpreted as the velocity potential and stream function, respectively,
of some two-dimensional, incompressible, irrotational flow pattern, where and
are Cartesian coordinates. However, if = f (z), where f (z) is well-behaved, then
((x, y), where G(z) is also well-behaved. It
F() = F[ f (z)] = G(z) = (
(x, y) + i
( = (
( = 0. In other words, the functions (
= 2
(x, y) and
follows that 2(
161
d
d
0
x
dz
z0
dz
Figure 6.7
A conformal map.
((x, y) can be interpreted as the velocity potential and stream function, respectively,
(6.60)
ln
(6.61)
F() = i
2
+i
corresponds to the flow pattern generated by a vortex filament of intensity , located
at the point = (0, 1), in the presence of a rigid plane at = 0. Hence,
$z%
ln tanh
G(z) = F(i e z/a ) = i
,
(6.62)
2
2a
corresponds to the flow pattern generated by a vortex filament of intensity , located
at the origin, in the presence of two rigid planes at y = a/2. This follows because
the line = 0 is mapped to the lines y = a/2, and the point = (0, 1) is mapped
162
A
a
A
C
D
1
C
A A
Figure 6.8
The conformal map = i e z/a .
to the origin. Moreover, if the line = 0 is a streamline in the -plane then the lines
y = a/2 are also streamlines in the z-plane. Thus, these lines could all correspond to
rigid boundaries. The stream function associated with the previous complex velocity
potential,
cosh( x a) cos( y/a)
(x, y) = ln
,
(6.63)
(6.64)
This maps the positive -axis to the positive x-axis, the negative -axis to the positive
y-axis, the region > 0 to the region x > 0, y > 0, and the point = (0, 2 a 2) to the
point z = (a, a). As we saw in Section 6.6, in the region > 0, the velocity potential
F() =
Q
ln( 2 + 4 a 4 ),
2
(6.65)
163
y/a
0.5
0
0.5
1
0.5
0
x/a
0.5
Figure 6.9
Stream lines of the two-dimensional flow pattern due to a vortex filament at the origin
in the presence of two rigid planes at y = a/2.
2
y/a
1
x/a
Figure 6.10
Stream lines of the two-dimensional flow pattern due to a line source at (a, a) in the
presence of two rigid planes at x = 0 and y = 0.
164
l2
,
(6.68)
(6.69)
(6.70)
Thus, the map converts the circle + = c in the -plane, where c > l, into the
ellipse
$ x %2 $ y %2
+
=1
(6.71)
a
b
in the z-plane, where
2
a = 2 l cosh[ln(c/l)] = c + l 2 /c,
(6.72)
b = 2 l sinh[ln(c/l)] = c l /c.
(6.73)
Note that the center of the ellipse lies at the origin, and its major and minor axes
run parallel to the x- and the y-axes, respectively. As we saw in Section 6.4, in the
-plane, the complex velocity potential
c2
,
(6.74)
F = V0 +
c2
(x, y) = V0 r
sin ,
(6.75)
r
where
r = l exp(cosh1 p),
p
1 y
= tan
,
x [p 2 1]1/2
1/2
x 2 /l 2 + y 2 /l 2 + 4 + [x 2 /l 2 + y 2 /l 2 + 4]2 16 x 2 /l 2 1/2
p =
.
8
Figure 6.11 shows the streamlines of the flow pattern calculated for c = 1.5 l.
(6.76)
(6.77)
(6.78)
165
4
3
2
1
y/l
0
1
2
3
4
4 3 2 1
0
x/l
Figure 6.11
Stream lines of the two-dimensional flow pattern due to uniform flow parallel to the
x-axis around an elliptical cylinder.
(6.79)
where K is a constant, and a, b, c, are the (real) values of that correspond to the
vertices of the polygon (Milne-Thompson 2011).
It is often convenient to take the point in the -plane that corresponds to one
of the vertices of the polygonsay, that given by = ato be at infinity. In this
case, the factor ( a) in Equation (6.79) becomes eectively constant, and can be
absorbed into a new constant of proportionality, K .
Consider, for example, a semi-infinite strip in the z-plane, for which = 0,
= /2, and = /2. This is mapped onto the upper half of the -plane, with the
zero-angle vertex corresponding to a point at infinity in the -plane, by means of the
166
y
z0 + i /K
z-plane
z0
-plane
c
Figure 6.12
Conformal transformation of a semi-infinite strip in the z-plane into the upper half of
the -plane.
transformation
d
= K ( b)1/2 ( c)1/2 ,
dz
(6.80)
where K is real and positive. We can integrate the previous expression to give
=
1
1
(b + c) + (b c) cosh[K (z z0 )].
2
2
(6.81)
Here, the points = b and = c in the -plane correspond to the vertices z = z0 and
z = z0 + i /K , respectively, in the z-plane. (See Figure 6.12.)
Suppose that b = 0, c = w, z0 = 0, and /K = w. In this case, the transformation (6.81) becomes
$ z%
'
w&
=
cosh
1 ,
(6.82)
2
w
which implies that
$ x%
$ y%
'
w&
=
cosh
cos
1 ,
(6.83)
2
w
w
$
%
$
%
x
y
w
sin
.
(6.84)
= sinh
2
w
w
The transformation (6.82) maps the semi-infinite strip in the z-plane that is bounded
by the lines y = 0, x = 0, and y = w into the upper half of the -plane. The
transformation also maps the origin of the -plane to the origin of the z-plane, and
the -axis to the lines y = w, x = 0, and y = 0. Now, in the -plane,
F=
Q
ln
(6.85)
167
y/w
0.5
1
x/w
Figure 6.13
Stream function due to a line source located in the left-hand corner of a semi-infinite
strip bounded by rigid planes at y = 0, x = 0, and y = w.
is the complex potential associated with a line source of strength 2 Q, located at the
origin. By symmetry, the -axis corresponds to a streamline, and can, therefore, be
replaced by a rigid boundary. It follows that, in the z-plane, the same potential is
that due to a line source, of strength Q, located in the lower left-hand corner of a
semi-infinite strip bounded by rigid planes at y = 0, x = 0, and y = w. (Incidentally,
the source strength in the z-plane is Q, rather than 2 Q, because, by symmetry, half
the output from the source in the -plane goes below the -axis and, therefore, does
not map to the semi-infinite strip in the z-plane.) The corresponding stream function
is
Q
Q
1
= arg() = tan
,
(6.86)
(6.87)
(zz0 )
(6.88)
This transformation maps the point = b + 1 to the point z = z0 , the -axis to the two
lines y = Im(z0 ) and y = Im(z0 )+/K , and the upper half of the -plane to the region
between the lines y = Im(z0 ) and y = Im(z0 ) +/K , as illustrated in Figure 6.14. It is
clear that the transformation (6.60), studied in the preceding section, is just a special
case of the transformation (6.88).
168
A
/K
z-plane
B
x
z0
A
-plane
b
A A
b+1
B
Figure 6.14
Conformal transformation of an infinite strip in the z-plane into the upper half of the
-plane.
1
1
dz
= ln
= ln
= ln
+ i ,
(6.89)
v x i vy
v
d F
and v and are the magnitude and direction (relative to
where z = x + i y, F = + i ,
the x-axis) of the velocity vector (v x , vy ). (See Section 6.5.) Here, F = F, = ,
and = . It follows that the real part of is constant on each free streamline,
whereas the imaginary part is constant on each straight segment of the rigid boundary.
Thus, the whole boundary of the fluid is represented in the -plane by a polygon.
We know, from the SchwarzChristoel theorem (see Section 6.8), that it is always
possible to map the interior (or exterior) of a polygon in one complex plane onto the
upper half of another complex plane. Hence, it is possible to find conformal relations
between and a new complex variable , and between F and , such that the flow
region is mapped onto the upper half of the -plane in both cases. In this way, we
from which an expression for F in terms of z
can find a relation between and F,
follows via integration.
169
F -plane
= 1
= 0
z-plane
A
B
= 1 C
B
-plane
= 1
C
/2
b
= 0
B
A
C
C
ln(v/V )
A
C
A
= 0
B
-plane
= 1
A
B
C C
Figure 6.15
Conformal transformations required for the determination of the flow from an orifice
in a plane wall in two dimensions.
As an example of the application of free streamline theory, let us calculate the
contraction ratio of a two-dimensional jet of liquid emerging from an orifice. Suppose that the orifice in question is a long thin slot in a plane wall of small thickness,
and that the wall forms part of a large vessel containing liquid. The speed of the
liquid on the free streamlines that emerge from the edges of the orifice is uniform,
and equal to V, say. This is also the speed of the interior of the jet far downstream
of the orifice, where (neglecting the eects of gravity) the streamlines are straight
and parallel. (See Figure 6.16.) Let the two streamlines bounding the flow, on which
= + 1 and = 1 , saybe ABC and A BC , respectively, where A, A , C, and
C all represent points at infinity. Figure 6.15 shows the corresponding straight-line
and -planes, where is now defined in a more convenient
boundaries in the Fmanner as
$V %
dz
= ln
= ln V
+ i .
(6.90)
v
d F
We, first, need to map the semi-infinite strip A BC CBA in the -plane to the
upper half of the -plane. The conformal transformation of a semi-infinite strip onto
the upper half of another complex plane is achieved by the relation (6.81). Adapting
this relation to our current needs, we take K = 1, z0 = i /2, and b = c = 1, so
that
= i sinh
(6.91)
170
to the upper
Next, we need to map the infinite strip ABCA BC in the F-plane
half of the -plane. The conformal transformation of a semi-infinite strip onto the
upper half of another complex plane is achieved by the relation (6.88). Adapting this
relation to our current needs, we take K = /(2 1 ), b = 0, z0 = i 1 , so that
F
= i exp
.
(6.92)
2 1
The flow field has now been mapped onto the upper half of the -plane in two
coincident ways, giving
F
1 d F
i
dz
= i exp
= i sinh =
V
.
(6.93)
2 1
2
d F V dz
Hence,
dz
= i + (1 2 )1/2 ,
(6.94)
d F
where, with cuts in the z-plane at AB and A B , the correct branch of (1 2)1/2 is that
which is real and positive when = 0. Integration, with the aid of Equation (6.92),
yields
V
V
(z z1 ) = i ( 1) (1 2 )1/2 + tanh1 [(1 2 )1/2 ],
2 1
(6.95)
(6.96)
Finally, the required relationship between z and F is obtained by eliminating between Equations (6.92) and (6.96).
On the free streamline BC, we have
= 1 ,
= V s,
= i ,
(6.97)
where s denotes distance measured along the streamline from B. It follows from
Equations (6.91) and (6.92) that
V
= sin = exp
s .
(6.98)
2 1
Thus, making use of Equation (6.96), the equation of streamline BC can be written,
in parametric form, as
2 1
tanh1 (cos ) cos ) ,
V
2 1
(1 + sin ) .
y=d
V
x=
(6.99)
(6.100)
171
1
0.8
0.6
0.4
y/d
0.2
0
0.2
0.4
0.6
0.8
1
1
x/d
Figure 6.16
Free streamlines of a liquid jet emerging from an orifice in a plane wall in two dimensions.
2 1
.
V
(6.101)
(6.102)
In other words, as expected, the velocity profile becomes uniform across the jet far
downstream of the orifice, which implies that 1 = b V. It follows, from Equation (6.101), that the contraction ratio of a two-dimensional liquid jet emerging from
an orifice in a plane wall takes the value
b
=
= 0.61.
d +2
(6.103)
Somewhat surprisingly, this value is very close to the observed contraction ratio of
a three-dimensional jet emerging from a circular hole in a plane wall. (See Section 4.6.) According to Equations (6.99), (6.100), and (6.103), the equation of the
172
0.9
b/d
0.8
0.7
0.6
0.5
1
(2/)
Figure 6.17
Contraction ratio of a liquid jet emerging from a two-dimensional orifice formed by
a gap between two semi-infinite plane walls that subtend an angle 2 .
free streamline BC can be written
2
x
=
tanh1 (cos ) cos ) ,
d +2
y + 2 sin
=
,
d
+2
(6.104)
(6.105)
$ %
F
,
(6.106)
= i exp
= i sinh
2 1
2
which implies that
V
2 /
dz
.
= i + (1 2 )1/2
dF
(6.107)
173
Far from the orifice, which corresponds to 0, the previous expression yields
d F/dz
1/V. In other words, a long way downstream of the orifice, the velocity
profile is again uniform across the jet. Hence, we can write 1 = b V, where b is the
semi-width of the jet far from the orifice. Combining the previous two equations, we
obtain
2 /
dz d
=
.
(6.108)
i + (1 2 )1/2
2 b
2 /
dk
.
i k + (1 k 2 )1/2
k
(6.109)
d
2
.
exp i
tan
sin1 ()
/2
(6.110)
As before, we have = exp[(/2) (s/b)] on the free streamline BC, where s denotes
distance measured along the streamline from B. Far downstream of the orifice, 0
and Im(z) b. Hence, we obtain the following expression for the contraction ratio
of the jet:
1
b
2 /2 sin[(2 /) ]
= 1+
d .
(6.111)
d
0
tan
This contraction ratio is plotted as a function of in Figure 6.17. It can be seen that
the ratio takes the value 1 when = 0, which corresponds to a two-dimensional jet
emerging from a gap between two parallel planes. Not surprisingly, there is no contraction in this case, because the streamlines of the flow are already parallel before
they emerge from the orifice. On the other hand, the ratio takes the value 1/2 when
= , which corresponds to the two-dimensional equivalent to the Borda mouthpiece discussed in Section 4.6. In this case, the contraction ratio is exactly the same
as that of a three-dimensional Borda mouthpiece, which is not surprising, because
there is nothing in the discussion of Section 4.6 that depends crucially on the shape
of the mouthpiece cross-section.
As a second example of the use of free streamline theory, consider the flow past
a flat plate normal to a liquid stream of infinite extent. The pressure in the cavity
formed behind the plate is assumed to be equal to that of the undisturbed stream.
Thus, the speed of the liquid on the free streamlines bounding the cavity is equal to
V, which is the uniform flow speed far upstream of the plate. Let us take = 0 on
the central streamline that divides at the stagnation point O, and later becomes the
two free streamlines. (See Figure 6.18.)
The correspondence between the various points on the streamline = 0 in the
and -planes is specified in Figure 6.18. The flow occupies the whole of
z-, F-,
the F-plane,
apart from a thin slit running along the positive section of the real axis.
As previously, the procedure is to find transformations that map the flow regions in
174
F -plane
y
B
C
z-plane
F 1/2-plane
C
O
= 0
B
-plane
B
/2
C A
C
B
B
ln(V /v)
-plane
C
C, A, C
B
Figure 6.18
Conformal transformations required for the determination of the flow past a flat plate
with a cavity at ambient pressure.
and -planes coincidentally onto the upper half of the -plane. The
both the Fsemi-infinite strip in the -plane has the same width, position, and orientation as that
shown in Figure 6.15. Thus, the appropriate relationship between and is again
= i sinh ,
(6.112)
with the locations of the points B and B again being = 1 and = +1, respectively.
The appropriate relationship between F and can be recognized by noting, first, that
the flow occupies the upper half of the F 1/2 -plane, and, second, that an inversion and
change of sign are needed to bring corresponding points on the two real axes into
coincidence. Thus, we obtain
kV
=
F
1/2
,
(6.113)
kV
F
1/2
= i sinh =
i
dz
1 d F
V
.
2
d F V dz
(6.114)
175
1/2
1/2
kV
1 d F
kV
= i
+ 1
,
V dz
F
F
(6.115)
1/2 is that which is positive on AO. Integrawhere the relevant branch of (1 k V/F)
0
V
F/k
du
=
1/2
i u
+ (1 1/u)1/2
V
F/k
i u 1/2 + (1 1/u)1/2 du, (6.116)
or
1/2
1/2
1/2
1/2
F 1/2
F
z
F
F
F
=
1
1 +2 i
ln
+
+i . (6.117)
k
kV
kV
kV
kV
kV
2
We can now evaluate the constant k making use of the fact that at the point B, where
= 1,
F
= 1.
kV
z = i b,
(6.118)
2b
.
+4
(6.119)
= i ,
(6.120)
where s represents distance along the streamline measured from B. Thus, on this
streamline,
1/2
k
= sin =
.
(6.121)
k+s
Hence, according to Equation (6.117), the equation of the streamline BC can be
written, in parametric form, as
$
s %1/2 $ s %1/2
1/2 1/2
x = (k + s) s k ln 1 +
+
,
(6.122)
k
k
(6.123)
y = 2 (k + s)1/2 k1/2 + k.
2
By symmetry, the equation of the free streamline BC is obtained from the previous equation via the transformation x x, y y. We deduce that the cavity
downstream of the plate extends to infinity, and that its boundary asymptotes to the
parabola
8b
x.
(6.124)
y2 = 4 k x =
+4
176
From Bernoullis theorem, the net force per unit length exerted by the fluid on
the plate, which is obviously normal to the plate, and, therefore, directed parallel to
the x-axis, is
b
b
1
1
V 2 v 2 dy
D=
(p p0 ) x=0 dy =
(6.125)
2
b
b 2
kV
d .
(6.126)
= V2b
y
0
In conventional parlance, this force constitutes a drag, because it is directed parallel
V) = /(k
V) < 1 on OB, so
to the undisturbed flow. (See Section 5.8.) Now, F/(k
1/2
1
1 d F
= i 1/2 + i
1
,
(6.127)
V dz
1/2
1
1
1
2
2
D = V b V k
1 d,
1/2
0
which yields
2
V 2 b.
D=
+4
(6.128)
(6.129)
Finally, if we define the drag coecient, in the conventional manner (see Section 8.8),
then we obtain
f
2
CD =
= 0.88.
(6.130)
=
2
V b + 4
We can extend the previous calculation to determine the flow past a flat plate
inclined at an angle to a liquid stream of infinite extent. Let the plate lie in the
plane x = 0. The flow field then occupies a semi-infinite strip in the -plane that
has the same width, position, and orientation as that in the previous calculation.
Consequently, the appropriate relation between and is again
dz
i
1 d F
V
.
(6.131)
= i sinh =
2
d F V dz
It follows that
1 d F
= i + 1 2.
(6.132)
V dz
As before, the two edges of the plate, B and B , correspond to = 1 and = +1,
respectively. Furthermore, the stagnation point, O, on the front surface of the plate
corresponds to || . (See Figure 6.18.) However, the latter point is no longer
necessarily located at the plates midpoint.
The transformation (6.113) generalizes to give
F =
kV
.
( cos )2
(6.133)
177
= i V /y.
Hence,
On the front surface of the plate, B OB, || 1 and d F/dz
from Equation (6.132),
1
1
= 2 1 =
,
V y
2 1
(6.134)
where the upper/lower signs correspond to > 0 and < 0, respectively. (The signs
kV
.
( cos )2
dy y d
2k
21 .
=
=
d ( cos )3
(6.135)
(6.136)
Writing
1 cos cos
,
(6.137)
cos cos
the points B, O, and B correspond to = 0, = , and = , respectively.
Furthermore,
cos cos
d
=
sin d,
(6.138)
( cos )3
sin4
sin sin
.
(6.139)
2 1 =
cos cos
=
2k
(1 cos cos + sin sin ) sin d,
sin4
(6.140)
%'
$
k &
2
, (6.141)
+
sin
sin
cos
+
sin
2
cos
+
cos
sin
2
sin4
where we have chosen the constant of integration so as to ensure that the midpoint
of the plate lies at y = 0. We expect y = b when = 0, where b is the semi-width of
the plate. We, thus, deduce, from the previous expression, that
k=
2 b sin4
.
sin + 4
(6.142)
178
(6.143)
(6.144)
(6.145)
1 d F
i
,
1+
V dz
sin
(6.146)
(6.147)
Now, is real and negative on the free streamline BC, and real and positive on the
free streamline BC . (See Figure 6.18.) We conclude that, in the small-|| limit, the
equations of these streamlines can be written, in parametric form, as
x
k
,
2
y
2k
.
sin
(6.148)
(6.149)
In other words, far downstream of the plate, the free streamlines, which form the
boundaries of the cavity behind the plate, asymptote to the parabola
4k
8 b sin2
2
x =
x.
y =
(6.150)
sin + 4
sin2
From Bernoullis theorem, the net force per unit length exerted by the fluid on
the plate, which is obviously normal to the plate, and, therefore, directed parallel to
the x-axis, is
b
b
vy2
1
2
f =
(p p0 ) x=0 dy = V
1 2 dy
2
V x=0
b
b
1
vy2 y d
1
d
= V2
1 2
2
V d
+1
1
V vy
d
= k V 2
,
(6.151)
vy V ( cos )3
+1
179
where use has been made of Equation (6.135). However, according to Equations (6.134)
and (6.139),
V vy
y 1
2 sin sin
=V
= 2 2 1 =
.
vy V
cos cos
V y
Thus, making use of Equation (6.138), we obtain
2kV2 2
kV2
f =
.
sin
d
=
sin3 0
sin3
Hence, it follows from Equation (6.142) that
2 sin
f =
V 2 b.
sin + 4
(6.152)
(6.153)
(6.154)
The drag, D, is the component of the force acting on the plate in the direction of the
unperturbed flow: that is,
2 sin2
D = f sin =
V 2 b.
(6.155)
sin + 4
On the other hand, the lift, L, is the component of the force acting perpendicular to
the direction of the unperturbed flow. (See Section 5.8.) Thus,
2 sin cos
L = f cos =
(6.156)
V 2 b.
sin + 4
Finally, it can be seen, from a comparison between Equations (6.124), (6.129),
(6.150), and (6.155), that there appears to be a connection between the drag force
exerted on the plate, and the shape of the parabola to which the cavity formed behind
the plate asymptotes. In fact,
2
y
D
= lim
,
(6.157)
x 4 x
V2
cavity boundary
where the unperturbed flow runs parallel to the x-axis. It turns out that this relation
is a general one for two-dimensional flow past a body of arbitrary shape with an
attached cavity at ambient pressure (Batchelor 2000).
180
Because dz = dx + i dy, and writing F(z) = (x, y) + i (x, y), where (x, y) and
(x, y) are real functions, it follows that J = Jr + i Ji , where
Jr = ( dx dy),
(6.159)
C
Ji =
( dx + dy).
(6.160)
However, we can also write the previous expressions in the two-dimensional vector
form
Jr =
A dr,
(6.161)
C
Ji =
B dr,
(6.162)
B dr =
( B)z dS ,
(6.164)
+
Jr =
dS ,
y
S x
dS .
Ji =
y
S x
Let
(6.165)
(6.166)
J=
F(z) dz,
(6.167)
F(z) dz,
(6.168)
J =
C
where C is a closed curve in the complex plane that completely surrounds the smaller
curve C. Consider
(6.169)
J = J J .
Writing J = Jr + i Ji , a direct generalization of the previous analysis reveals that
Jr =
+
dS ,
(6.170)
y
S x
dS ,
(6.171)
Ji =
y
S x
181
where S is now the region of the x-y plane lying between the curves C and C . Suppose that F(z) is well-behaved (i.e., finite, single-valued, and dierentiable) throughout S . It immediately follows that its real and imaginary components, and , respectively, satisfy the Cauchy-Riemann relations, (6.17)(6.18), throughout S . However, if this is the case then it is apparent, from the previous two expressions, that
Jr = Ji = 0. In other words, if F(z) is well-behaved throughout S then J = J .
The circulation of the flow about some closed curve C in the x-y plane is defined
dF
dz,
(6.172)
= (v x dx + vy dy) = Re
C
C dz
where F(z) is the complex velocity potential of the flow, and use has been made of
Equation (6.35). Thus, the circulation can be evaluated by performing a line integral
in the complex z-plane. Moreover, as is clear from the previous discussion, this
integral can be performed around any loop that can be continuously deformed into
the loop C while still remaining in the fluid, and not passing over a singularity of the
complex velocity, dF/dz.
1 dF 2
.
2 dz
(6.173)
Let us evaluate the resultant force (per unit length), and the resultant moment (per
unit length), acting on the fluid within the curve as a consequence of this pressure
distribution.
Consider a small element of the curve C, lying between x, y and x + dx, y + dy,
which is suciently short that it can be approximated as a straight-line. Let P be
the local fluid pressure on the outer (i.e., exterior to the curve) side of the element.
As illustrated in Figure 6.19, the pressure force (per unit length) acting inward (i.e.,
toward the inside of the curve) across the element has a component P dy in the minus
x-direction, and a component P dx in the plus y-direction. Thus, if X and Y are
the components of the resultant force (per unit length) in the x- and y-directions,
respectively, then
dX = P dy,
(6.174)
dY = P dx.
(6.175)
The pressure force (per unit length) acting across the element also contributes to a
182
P dx
dx
P dy
dy
dl
P dl
x
Figure 6.19
Force acting across a short section of a curve.
(6.176)
Thus, the x- and y-components of the resultant force (per unit length) acting on the
of the fluid within the curve, as well as the resultant moment (per unit length) about
the z-axis, are given by
X=
Y=
P dy,
(6.177)
P dx,
(6.178)
P (x dx + y dy),
(6.179)
M=
respectively, where the integrals are taken (counter-clockwise) around the curve C.
Finally, given that the pressure distribution on the curve takes the form (6.173), and
183
that a constant pressure obviously yields zero force and zero moment, we find that
2
dF dy,
C dz
2
1
dF dx,
Y=
2
C dz
2
1
dF (x dx + y dy).
M=
dz
2
X=
(6.180)
(6.181)
(6.182)
1
i
2
2
dF d z.
C dz
(6.183)
However,
2
dF d z = dF d F d z = dF d F,
(6.184)
dz
dz d z
dz
where dF = d + i d and d F = d i d. Suppose that the curve C corresponds to
a streamline of the flow, in which case = constant on C. Thus, d = 0 on C, and
so d F = dF. Hence, on C,
2
dF
dF dF
dF =
dF =
dz,
dz
dz
dz
1
i
2
dF
dz
(6.185)
2
dz.
(6.186)
This result is known as the Blasius theorem, after Paul Blasius (18831970).
Now, x dx + y dy = Re(z d z). Hence,
2
1
dF z d z ,
M = Re
2
C dz
or, making use of an analogous argument to that employed previously,
1 dF 2
M = Re
z dz ,
2
C dz
(6.187)
(6.188)
In fact, Equations (6.186) and (6.188) hold even when is not constant on the
curve C, as long as C can be continuously deformed into a constant- curve without
leaving the fluid or crossing over a singularity of (dF/dz) 2 .
As an example of the use of the Blasius theorem, consider again the situation,
discussed in Section 6.6, in which a line source of strength Q is located at (0, a), and
184
y
C
x
C
Figure 6.20
Source in the presence of a rigid boundary.
there is a rigid boundary at y = 0. As we have seen, the complex velocity in the
region y > 0 takes the form
Q
z
dF
=
.
(6.189)
dz
z2 + a2
Suppose that we evaluate the Blasius integral, (6.188), about the contour C shown in
Figure 6.20. This contour runs along the boundary, and is completed by a semi-circle
in the upper half of the z-plane. As is easily demonstrated, in the limit in which the
radius of the semi-circle tends to infinity, the contribution of the curved section of
the contour to the overall integral becomes negligible. In this case, only the straight
section of the contour contributes to the integral. Note that the straight section corresponds to a streamline (because it is coincident with a rigid boundary). In other
words, the contour C corresponds to a streamline at all constituent points that make a
finite contribution to the Blasius integral, which ensures that C is a valid contour for
the application of the Blasius theorem. In fact, the Blasius integral specifies the net
force (per unit length) exerted on the whole fluid by the boundary. Observe, however,
that the contour C can be deformed into the contour C , which takes the form of a
small circle surrounding the source, without passing over a singularity of (dF/dz) 2 .
(See Figure 6.20.) Hence, we can evaluate the Blasius integral around C without
changing its value. Thus,
2
dF
1
z2
1 $ Q %2
X iY = i
dz = i
dz,
(6.190)
2
2 2
2
2
C dz
C (z + a )
or
1
2
1
1 $ Q %2
+
+
dz.
(6.191)
X iY = i
2
8
(z + i a) (z i a) (z + i a)
C (z i a)
185
i Q2
.
4 a
(6.192)
Q
ln z V0 z ei 0 .
2
(6.193)
The net force (per unit length) acting on the source (which is calculated by performing the Blasius integral around a large loop that follows streamlines, and then
shrinking the loop to a small circle centered on the source) is (see Exercise 6.1)
F = Q V.
(6.194)
This force acts in the opposite direction to the flow. Thus, an external force F,
acting in the same direction as the flow, must be applied to the source in order for
it to remain stationary. In fact, the previous result is valid even in a non-uniformly
flowing fluid, as long as V is interpreted as the fluid velocity at the location of the
source (excluding the velocity field of the source itself).
Finally, consider a vortex filament of intensity placed at the origin in a uniformly flowing fluid whose velocity is V = V0 (cos 0 , sin 0 ). From Section 6.4, the
complex velocity potential of the net flow is
F(z) = i
ln z V0 z ei 0 .
2
(6.195)
The net force (per unit length) acting on the filament (which is calculated by performing the Blasius integral around a small circle centered on the filament) is (see
Exercise 6.2)
F = V ez .
(6.196)
This force is directed at right-angles to the direction of the flow (in the sense obtained
by rotating V through 90 in the opposite direction to the filaments direction of
rotation). Again, the previous result is valid even in a non-uniformly flowing fluid, as
long as V is interpreted as the fluid velocity at the location of the filament (excluding
the velocity field of the filament itself).
186
6.12 Exercises
6.1 Demonstrate that a line source of strength Q (running along the z-axis) situated in a uniform flow of (unperturbed) velocity V (lying in the x-y plane)
and density experiences a force per unit length
F = Q V.
6.2 Demonstrate that a vortex filament of intensity (running along the z-axis)
situated in a uniform flow of (unperturbed) velocity V (lying in the x-y plane)
and density experiences a force per unit length
F = V ez .
6.3 Show that two parallel line sources of strengths Q and Q , located a perpendicular distance r apart, exert a radial force per unit length Q Q /(2 r) on
one another, the force being attractive if Q Q > 0, and repulsive if Q Q < 0.
6.4 Show that two parallel vortex filaments of intensities and , located a perpendicular distance r apart, exert a radial force per unit length /(2 r)
on one another, the force being repulsive if > 0, and attractive if <
0.
6.5 A vortex filament of intensity runs parallel to, and lies a perpendicular
distance a from, a rigid planar boundary. Demonstrate that the boundary experiences a net force per unit length 2 /(4 a) directed toward the filament.
6.6 Two rigid planar boundaries meet at right-angles. A line source of strength
Q runs parallel to the line of intersection of the planes, and is situated a perpendicular distance a from each. Demonstrate that the source is subject to a
force per unit length
3 2 Q2
8 a
directed towards the line of intersection of the planes.
6.7 A line source of strength Q is located a distance b from an impenetrable
circular cylinder of radius a < b (the axis of the cylinder being parallel to the
source). Demonstrate that the cylinder experiences a net force per unit length
Q2
a2
2 b (b 2 a 2 )
directed toward the source.
187
6.8 A dipole line source consists of a line source of strength Q, running parallel to
the z-axis, and intersecting the x-y plane at (d/2) (cos , sin ), and a parallel
source of strength Q that intersects the x-y plane at (d/2)( cos , sin ).
Show that, in the limit d 0, and Q d D, the complex velocity potential
of the source is
D ei
.
F(z) =
2 z
Here, D e i is termed the complex dipole strength.
6.9 A dipole line source of complex strength D e i is placed in a uniformly
flowing fluid of speed V0 whose direction of motion subtends a (counterclockwise) angle 0 with the x-axis. Show that, while no net force acts on
the source, it is subject to a moment (per unit length) M = D V0 sin( 0 )
about the z-axis.
6.10 Consider a dipole line source of complex strength D1 e i 1 running along the
z-axis, and a second parallel source of complex strength D2 e i 2 that intersects the x-y plane at (x, 0). Demonstrate that the first source is subject to a
moment (per unit length) about the z-axis of
M=
D1 D2
sin(1 + 2 ),
2 x 2
X=
respectively. Show that the second source is subject to the same moment, but
an equal and opposite force.
6.11 A dipole line source of complex strength D e i runs parallel to, and is located a perpendicular distance a from, a rigid planar boundary. Show that the
boundary experiences a force per unit length
D2
8 a 3
acting toward the source.
6.12 Demonstrate that a conformal map converts a line source into a line source
of the same strength, and a vortex filament into a vortex filament of the same
intensity.
6.13 Consider the conformal map
z = i c cot(/2),
188
c sinh
,
cosh cos
y=
c sin
.
cosh cos
2 c i V0
cot(/n),
n
where
z = i c cot(/2).
Here, V0 , n, and c are real and positive. Show that
Hence, deduce that the flow at |z| is uniform, parallel to the x-axis, and
of speed V0 . Demonstrate that
(, ) =
2 V0 c
sin(2 /n)
.
n cosh(2 /n) cos(2 /n)
Hence, deduce that the streamline = 0 runs along the x-axis for |x| > c, but
along a circular arc connecting the points (c, 0) for |x| < c. Furthermore,
show that if 1 < n < 2 then this arc lies above the x-axis, and is of maximum
height
cos( n/2) + 1
h=c
,
sin( n/2)
but if 2 < n < 3 then the arc lies below the x-axis, and is of maximum depth
cos( n/2) + 1
.
d=c
| sin( n/2)|
Hence, deduce that if 1 < n < 2 then the complex velocity potential under
investigation corresponds to uniform flow of speed V0 , parallel to a planar
boundary that possesses a cylindrical bump (whose axis is normal to the flow)
of height h and width 2 c, but if 2 < n < 3 then the potential corresponds to
189
2 V0
[1 cos( n/2)] .
n2
a V0
tanh(a /z)
c2 i
e ,
F(z) = V ei +
where
l2
.
(l < c) represents uniform flow of unperturbed speed V, whose direction subtends a (counter-clockwise) angle with the x-axis, around an impenetrable
elliptic cylinder of major radius a = c + l 2 /c, aligned along the x-axis, and
minor radius b = c l 2 /c, aligned along the y-axis. Demonstrate that the
moment per unit length (about the z-axis) exerted on the cylinder by the flow
is
M = V 2 (a 2 b 2 ) sin(2 ).
2
Hence, deduce that the moment acts to turn the cylinder broadside-on to the
flow (i.e., = /2 is a dynamically stable equilibrium state), and that the
equilibrium state in which the cylinder is aligned with the flow (i.e., = 0)
is dynamically unstable.
z=+
190
l2
.
1
V 2 a 2 sin2 + b 2 cos2 ,
2
where is the fluid mass density. Hence, show that the cylinders added mass
per unit length is
madded = a 2 sin2 + b 2 cos2 .
6.20 Demonstrate from Equation (6.110) that the equation of the free streamline
BC, in the case of a liquid jet emerging from a two-dimensional orifice of
semi-width d formed by a gap between two semi-infinite plane walls that
subtend an angle 2 , can be written parametrically as:
1
d
2 /2
d
2
2
1+
sin
,
tan
0
tan
sin1 ()
1
d
d
2
2
y
2 /2
2 /2
=1
sin
sin
,
1+
d
sin1 ()
tan
0
tan
x 2
=
d
/2
cos
191
flow a long way from the orifice is in the +y-direction. Show that for the
case of a two-dimensional Borda mouthpiece, = , the previous equations
reduce to
x 1
=
ln() 1 + 2 ,
d
y 1 1 1
= + sin () + (1 2 )1/2 .
d 2
Finally, show that the previous equations predict that the free streamline is
re-entrant, with x/d attaining its minimum value (1 ln 2)/(2) = 0.153
when y/d = 3/4 + 1/(2) = 0.909.
192
7
Axisymmetric Incompressible Inviscid Flow
7.1 Introduction
This chapter investigates axisymmetric, incompressible, inviscid flow. More information on this subject can be found in Batchelor 2000, and Milne-Thompson 2011.
194
R
s
R
P
Q3
Q2
Q1
Figure 7.1
Definition of the Stokes stream function.
position of the arbitrary point P, and, possibly, on that of the fixed point A. In fact,
if we take another fixed point B on the symmetry axis, and draw the meridian curve
BQ3 P, then the flux across the surface generated by BQ3 P will be the same as that
across the surface generated by AQ1 P, because, by symmetry, there is no flow across
AB. (See Figure 7.1.) It follows that the value of does not depend on the particular
fixed point that is used in its definition, as long as this point lies on the symmetry
axis. Hence, we conclude that the value of the stream function at P depends solely
on the position of P. Furthermore, if P lies on the axis then = 0.
Consider two neighboring points, R and R , lying in the meridian plane. (See
Figure 7.1.) The flux from right to left across the surface generated by revolving any
line joining R to R about the symmetry axis is 2 R 2 R . If the distance RR
takes the infinitesimal value s then we can write
2 (R R ) = 2 s v ,
(7.1)
where v is the normal flow velocity (from right to left) across the straight-line RR
in the meridian plane. (See Figure 7.1.) It follows that
v =
1
.
s
(7.2)
In particular, if we suppose that ds is, in turn, equal to d, dz, dr, and r d then we
obtain the following expressions for the in-plane velocity components in cylindrical
195
1
,
vr =
r sin r
v =
1
,
1
v =
.
r sin r
vz =
(7.3)
(7.4)
(7.5)
When the fluid velocity is written in this form it becomes obvious that the incompressibility constraint v = 0 is satisfied [because (A B) 0see Equations (A.175) and (A.176)]. It is also clear that the Stokes stream function, , is
undefined to an arbitrary additive constant.
In fact, the most general expression for an axisymmetric incompressible flow
pattern is
v = + ,
(7.6)
where = (, z) is the angular velocity of flow circulating about the z-axis.
(This follows because ( ) = 0 when / = 0. See Appendix C.) The
previous expression implies that v = 0 (because = 0 when / =
0). In other words, when plotted in the meridian plane, streamlines in a general
axisymmetric flow pattern correspond to contours of .
Making use of the vector identities (A.176) and (A.178), we can also write Equation (7.6) in the form
v = ( ) + .
(7.7)
(7.8)
(7.9)
196
= e 2 ( ) = 2
3
1
1 2
=
+
z 2
1
1 2
1
.
=
+
r sin r 2 r 3 sin
(7.10)
(7.12)
1 2
1
r
+
sin
= 0.
(7.13)
r
r 2 r
r 2 sin
Let us search for a separable solution of Equation (7.13) of the form
(r, ) = R(r) ().
(7.14)
d
1 d 2 dR
1
d
r
=
sin
,
R dr
dr
sin d
d
(7.15)
d 2 dR
r
l (l + 1) R = 0,
dr
dr
d
d
(1 2 )
+ l (l + 1) = 0,
d
d
(7.16)
(7.17)
197
where = cos , and l (l + 1) is a constant. The solutions to Equation (7.17) that are
well behaved for in the range 1 to +1 are known as the Legendre polynomials,
and are denoted the Pl (), where l is a non-negative integer (Jackson 1962). (If l is
non-integer then the solutions are singular at = 1.) In fact,
Pl () =
(1) l d l
2 l
.
1
2 l l! d l
(7.18)
Hence,
P0 () = 1,
(7.19)
P1 () = ,
(7.20)
1 2
P2 () =
3 1 ,
2
1 3
P3 () =
5 3 ,
2
(7.21)
(7.22)
where the l and l are arbitrary coecients. It follows from Equations (7.4) that the
corresponding expression for the Stokes stream function is
%
$ l
dPl
l
r l+1 r l 1 2
,
(7.24)
(r, ) = 0 +
l+1
l
d
l=1,
where = cos .
(7.25)
(7.26)
198
Of course, uniform flow is irrotational [this is clear from a comparison of Equations (7.10) and (7.25)], so we can also represent the flow pattern in terms of a velocity potential: that is (see Section 5.4),
(, z) = V z,
(7.27)
(r, ) = V r cos .
(7.28)
or
It follows, from the previous analysis, that the velocity field of a uniform stream,
running parallel to the z-axis, can either be written v = , with specified by
Equations (7.25)(7.26), or v = , with specified by Equations (7.27)(7.28).
Q
.
4 r 2
(7.31)
Of course, v = 0.
According to Equations (7.4), the Stokes stream function associated with a point
source at the origin is such that = (), and is obtained by integrating
vr =
It follows that
Q
1
.
= 2
4 r 2
r sin
(7.32)
Q
z
Q
cos =
.
(7.33)
4
4 ( 2 + z 2 )1/2
It is clear, from a comparison of Equations (7.10) and (7.33), that the previously
specified flow pattern is irrotational. Hence, this pattern can also be derived from a
velocity potential. In fact, by symmetry, we expect that = (r). The potential itself
is obtained by integrating
Q
(7.34)
= .
vr =
2
r
4 r
=
199
P
r1
r
1
d/2
r2
2
d/2
Figure 7.2
A dipole source.
It follows that
=
Q
Q
=
.
4 r 4 ( 2 + z 2 )1/2
(7.35)
Q
(cos 2 cos 1 ) ,
4
Q 1
1
=
,
4 r2 r1
(7.36)
(7.37)
(7.38)
200
d (sin 2 sin 1 )
.
2 sin[(2 1 )/2] cos[(2 1 )/2]
(7.39)
d cos[(2 + 1 )/2]
.
cos[(2 1 )/2]
(7.40)
Q r1 r2
D cos
=
.
4 r1 r2
4 r 2
Equation (7.36) implies that
Q z d/2 z + d/2
1
1
Q
d 1
1
=
+
=
.
z
4
r2
r1
4
r2 r1
2 r2 r1
Thus, in the limit 2 1 and r2 r1 r, we obtain
D
2
Q d cos2 d
D sin2
=
,
=
2
4
r
r
4 r
4 ( + z 2 )3/2
(7.41)
(7.42)
(7.43)
(7.44)
where use has been made of Equation (7.41), as well as the fact that z = r cos .
Figure 7.3 shows the stream function of a dipole point source located at the origin.
Incidentally, Equations (7.26), (7.28), (7.33), (7.35), (7.42), and (7.44) imply that
the terms in the expansions (7.23) and (7.24) involving the constants 0 , 1 , and 1
correspond to a point source at the origin, uniform flow parallel to the z-axis, and a
dipole point source at the origin, respectively. Of course, the term involving 0 is
constant, and, therefore, gives rise to no flow.
201
1
0.8
0.6
0.4
0.2
x
0
0.2
0.4
0.6
0.8
1
1
0.5
0
z
0.5
Figure 7.3
Contours of the stream function of a dipole source located at the origin.
a uniform flow pattern is automatically irrotational), it follows from the Kelvin circulation theorem that the velocity field remains irrotational as it is convected past the
sphere. (See Section 4.14.) Hence, we can write v = , where
2 = 0
(7.45)
as r ,
(7.46)
and
(a, , )
= 0.
(7.47)
r
The latter constraint arises because the surface of the sphere is impenetrable, which
implies that v er = 0 at r = a.
Let us search for an axisymmetric solution of Equation (7.45) of the form
(r, ) = V r P1 (cos ) + 1 r 2 P1 (cos ).
(7.48)
It can be seen, by comparison with Equation (7.23), that the previous expression definitely solves Equation (7.45). Moreover, the expression also automatically satisfies
the boundary condition (7.46) [because P1 (cos ) = cos ]. The remaining boundary
condition, (7.47), yields 1 = V/(2 a 3). Hence, we obtain
r 1 $ a %2
cos ,
(7.49)
(r, ) = V a +
a 2 r
202
5
4
3
2
x/a
1
0
1
2
3
4
5
5
0
z/a
Figure 7.4
Contours of the stream function generated by a spherical obstacle of radius a placed
in the uniform flow field v = V ez .
or
a3
vr (r, ) = V 1 3 cos ,
r
1 a3
sin .
v (r, ) = V 1 +
2 r3
(7.50)
(7.51)
Because the solutions of Laplaces equation, subject to well-posed boundary conditions, are unique (Riley 1974), we can be sure that the previous axisymmetric solution is the most general solution to the problem.
It is clear, by comparison with Equations (7.28) and (7.42), that the velocity potential (7.49) is the superposition of that associated with uniform flow with velocity
V, parallel to the z-axis, and a dipole point source of strength D = 2 V a 3 , located
at the origin. Thus, making use of Equations (7.26) and (7.44), the associated stream
function takes the form
1
1
a3
a3
= V r 2 sin2 1 3 = V 2 1
.
(7.52)
2
r
2
( 2 + z 2 )3/2
Figure 7.4 show the contours of this stream function.
Bernoullis theorem yields (see Section 4.3)
p 1 2 p0 1
+ v =
+ V 2,
2
(7.53)
203
where is the uniform fluid mass density, and p0 the fluid pressure at infinity. Thus,
making use of Equations (7.50) and (7.51), the pressure distribution on the surface
of the sphere can be written
p(a, , ) = p0
1
9
V2 +
V 2 cos(2 ).
16
16
(7.54)
The net force exerted on the sphere by the fluid has the Cartesian components
F x = p(a, , ) a 2 sin cos d,
(7.55)
(7.56)
Fy = p(a, , ) a 2 sin sin d,
(7.57)
Fz = p(a, , ) a 2 cos d,
where the integrals are over all solid angle. Thus, it follows that
F x = Fy = Fz = 0.
(7.58)
In other words, the fluid exerts zero net force on the sphere, in accordance with
dAlemberts paradox. (See Section 4.5.)
(7.59)
as r .
(7.60)
(7.61)
204
= Vz cos .
(7.62)
r r=a
It is easily demonstrated that the solution to Equation (7.59), subject to the boundary
conditions (7.60) and (7.62), is
or
(r, , t) =
a3
1
Vz (t) 2 cos .
2
r
(7.63)
Hence,
a3
cos ,
r3
1
a3
v (r, , t) = Vz (t) 3 sin .
2
r
vr (r, , t) = Vz (t)
(7.64)
(7.65)
where is the uniform fluid mass density, p0 the fluid pressure at infinity, and it is
assumed that fluid and cylinder are both situated in a gravitational field of uniform
acceleration g ez. Thus, the pressure distribution at the surface of the sphere can be
written
1 2
V |r=a .
p(a, , t) = p0 v g a cos +
(7.67)
2
t r=a
r=a
The final term on the right-hand side of the previous equation arises because
= V .
(7.68)
t x,y,z t r,
Hence, we obtain
p(a, , , t) = p0
dVz
1
9
1
Vz2 + Vz2 cos(2 ) g a cos + a
cos . (7.69)
16
16
2
dt
The net force exerted on the sphere by the fluid is specified by Equations (7.55)
(7.57). It follows that
F x = Fy = 0,
(7.70)
and
1 dVz
m
,
(7.71)
2
dr
where m = (4/3) a 3 is the mass of the fluid displaced by the sphere. The equation
of vertical (i.e., in the z-direction) motion of the sphere is thus,
F z = m g
dVz
= Fz m g,
dt
(7.72)
1 dVz
= m g + m g.
m+ m
2
dt
205
(7.73)
It follows that the sphere moves under the action of its weight, and the buoyancy
force exerted on it by the fluid (i.e., the first and second terms, respectively, on the
right-hand side of the previous equation), as if it had the virtual mass m + (1/2) m. In
other words, as it moves, the sphere eectively entrains an added mass of fluid equal
to half of the displaced fluid mass. The net vertical acceleration of the sphere can be
written
s1
g,
(7.74)
az =
s + 1/2
where s = m/m is the spheres specific gravity. For the case of a bubble of gas
in a liquid, we expect 0 < s 1, because the gas is inevitably much less dense
than the liquid. It follows that the bubble accelerates vertically upward at twice the
acceleration due to gravity.
The origin of the spheres added mass is easily explained. According to Equations (7.64) and (7.65), the total kinetic energy of the fluid surrounding the sphere
is
1 2
v dV
Kfluid =
fluid 2
$ %6
a
1
1
(7.75)
= V2
cos2 + sin2 2 r 2 sin dr d,
2
r
4
a
0
which reduces to
1 2
m V .
4
However, the kinetic energy of the sphere is
Kfluid =
Ksphere =
Thus, the total kinetic energy is
K = Kfluid + Ksphere
1
m V 2.
2
1
1 2
=
m+ m V .
2
2
(7.76)
(7.77)
(7.78)
In other words, the kinetic energy of the fluid surrounding the sphere can be accounted for by supposing that a mass m /2 of the fluid co-moves with the sphere, and
that the remainder of the fluid remains stationary.
206
(7.79)
(7.80)
z
= .
(7.81)
It follows, from the previous two expressions, that = 0. In other words, and
are orthogonal coordinates in the meridian plane. [Incidentally, we are assuming
that (, , ) are a right-handed set of coordinates.] Furthermore, it can also be
shown from the Cauchy-Riemann relations that
2 1/2
2
2 1/2
z
z
=
,
h = h =
+
+
(7.82)
1
1
=
,
h
h
(7.83)
v =
1
1
=
.
h h
(7.84)
Of course, writing the velocity field in terms of a Stokes stream function ensures
that the field is incompressible, which also implies that 2 = 0. The additional
requirement that the field be irrotational yields = 0. Making use of the analysis
of Appendix C, this requirement reduces to
or
(h v ) (h v )
= 0,
(7.85)
1
1
+
= 0.
(7.86)
207
E
= 0
V
F =
C =0
Figure 7.5
An axisymmetric solid body moving through an incompressible irrotational fluid.
at infinity implies that asymptotes to a constant a long way from the body. Without
loss of generality, we can chose this constant to be zero. Thus, one constraint on the
system is that
( , ) = 0.
(7.87)
The appropriate constraint at the surface of the body is that
v = V e ,
(7.88)
when = 0 . Integrating, making use of the constraint (7.87) (which implies that
= 0 on the z-axis, where is constant, by symmetry), we obtain
1
(0 , ) = V 2 .
2
(7.90)
We can also set the velocity potential, , to zero at infinity, and on the z-axis.
The total kinetic energy of the fluid surrounding the moving body is
1
1
1
(7.91)
K = v 2 dV = ()2 dV = ( ) dV,
2
2
2
where we have made use of the fact that 2 = 0. Here, is the fluid mass density,
208
and dV is an element of the volume obtained by rotating the area ABCDEFA, shown
in Figure 7.5, about the z-axis. Making use of the divergence theorem, we obtain
1
1
1
K = n 2 ds =
2 ds.
(7.92)
2
2 ABC h
where ds is an element of the curve ABCDEFA, and n is an outward pointing, unit,
normal vector to the area ABCDEFA. Here, we have made use of the fact that the
velocity potential is zero at infinity (i.e., on DEF), and also on the z-axis (i.e., on CD
and FA). On the curve ABC, we can write ds = h d. Furthermore, it follows from
Equations (7.82) and (7.83) that h = h , and / = 1 /. Thus,
d
K =
,
(7.93)
0
=0
or
K =
=0
d
(7.94)
=0
(7.95)
(7.96)
= c e sin ,
(7.97)
z 2 + 2 = r 2,
(7.98)
r() = c e .
(7.99)
Thus, the constant- surfaces are concentric spheres of radius r(). If we set
a = c e 0
(7.100)
then the problem reduces to that of a sphere, of radius a, moving through a fluid that
is at rest at infinity. This problem was solved, via dierent methods, in Section 7.10.
The constraints (7.87) and (7.90) yield
( , ) = 0,
1
(0 , ) = V c2 e 2 0 sin2 ,
2
(7.101)
(7.102)
where use has been made of Equation (7.97). This suggests that we can write
(, ) = F() sin2 .
(7.103)
209
d dF
e
2 e F = 0,
d
d
(7.104)
(7.105)
(7.106)
1
B = V c 2 e 3 0 ,
2
(7.107)
sin2
1
(, ) = V a 3
.
2
c e
(7.108)
sin
=
= V a3
,
2
(c e )2
(7.109)
1
cos
V a3
.
2
(c e )2
(7.110)
Note that the previous expression is formally the same as expression (7.63), as long
as we make the identifications V Vz (t), c e r, and .
On the surface of the sphere, = 0 , we obtain
1
(0 , ) = V a 2 sin2 ,
2
1
(0 , ) = V a cos .
2
Thus,
K =
=0
d
=0
1
= a3 V 2
2
2 d =
(7.111)
(7.112)
3
a V 2.
3
(7.113)
As is clear from the analysis of Section 7.10, the spheres added mass can be written
madded =
K
2 3
a .
=
2
3
(1/2) V
(7.114)
Hence, we arrive at the standard result that the added mass is half the displaced mass
[i.e., half of (4/3) a 3 ].
210
(7.115)
(7.116)
= c cosh sin .
(7.117)
a = c cosh 0 ,
(7.118)
b = c sinh 0 .
(7.119)
Let
(7.120)
(7.121)
(7.122)
respectively. Setting (, ) = F() sin2 , and substituting into the governing equation, (7.86), we obtain
1 dF
d
2
F = 0,
(7.123)
d cosh d
cosh
which can be rearranged to give
d
dF
cosh
2 sinh F = 0.
d
d
(7.124)
On integration, we get
cosh
dF
2 sinh F = B,
d
(7.125)
d
F
B
=
.
2
d cosh
cosh3
(7.126)
211
It follows that
F() = B cosh2
d
cosh3
sinh
1
1
2
1
= B cosh
tan
2
sinh
cosh2
1
1
= B sinh cosh2 tan1
,
2
sinh
(7.127)
where use has been made of the constraint (7.121). Let e = (1 b 2 /a 2 )1/2 be the
eccentricity of the spheroid. Thus, cosh 0 = 1/e, sinh 0 = (1 e 2 )1/2 /e, c = e a,
and tan1 (1/ sinh 0 ) = sin1 e. The constraint (7.122) yields
(1 e 2 )1/2 sin1 e
1
1
2
V a = F(0 ) = B
,
(7.128)
2
2
e
e2
V a2 e2
B=
.
e (1 e 2 )1/2 sin1 e
or
(7.129)
Hence,
2
1
1
2 2 sinh cosh tan (1/ sinh )
(, ) = V a e
sin2 .
2
e (1 e 2 )1/2 sin1 e
(7.130)
=
,
c cosh sin
(7.131)
1 sinh tan1 (1/ sinh )
(, ) = V a e
cos .
e (1 e 2 )1/2 sin1 e
(7.132)
(7.133)
e (1 e 2 )1/2 sin1 e
(0 , ) = V a
cos .
sin1 e e (1 e 2 )1/2
(7.134)
Thus,
K =
=0
d
=0
2 1/2
2
sin1 e
3 2 e (1 e )
.
= a V
3
sin1 e e (1 e 2 )1/2
(7.135)
212
3
2
x/a
1
0
1
2
3
3
0
z/a
Figure 7.6
Contours of the Stokes stream function for the case of a disk of radius a, lying in the
x-y plane, and placed in a uniform, incompressible, irrotational flow directed parallel
to the z-axis.
It follows that the added mass of the spheroid is
madded =
where
G(e) =
4
a 3 G(e),
3
e (1 e 2 )1/2 sin1 e
sin1 e e (1 e 2 )1/2
(7.136)
(7.137)
and
(7.138)
(7.139)
8
a 3,
(7.140)
3
respectively. Here, z = sinh cos and = cosh sin . It follows that, in the
madded =
213
(7.142)
(7.143)
= c sinh sin .
(7.144)
a = c cosh 0 ,
(7.145)
b = c sinh 0 .
(7.146)
Let
(7.147)
(7.148)
(7.149)
respectively. Setting (, ) = F() sin2 , and substituting into the governing equation, (7.86), we obtain
1 dF
d
2
F = 0.
(7.150)
d sinh d
sinh
The solution that satisfied the constraint (7.148) is
1
!
"
F() = B cosh + sinh2 ln tanh(/2) .
2
(7.151)
214
(7.153)
=
,
c sinh sin
which can be integrated to give
!
"
1 + cosh ln tanh(/2)
(, ) = V a
cos .
(1 e 2 ) 1 + (1/2 e) ln [(1 e)/(1 + e)]
(7.154)
(7.155)
(7.156)
and
(0 , ) = V a
1 + (1/2 e) ln [(1 e)/(1 + e)]
cos .
(1 e 2 ) 1 + (1/2 e) ln [(1 e)/(1 + e)]
(7.157)
Thus,
K =
d
=0
=0
1 + (1/2 e) ln [(1 e)/(1 + e)]
2
= a b2 V 2
.
3
(1 e 2 ) 1 + (1/2 e) ln [(1 e)/(1 + e)]
(7.158)
where
G(e) =
4
a b 2 G(e),
3
1 + (1/2 e) ln [(1 e)/(1 + e)]
.
(1 e 2 ) 1 + (1/2 e) ln [(1 e)/(1 + e)]
(7.159)
(7.160)
is a monotonic function that takes the value 1/2 when e = 0, and asymptotes to
(1 e) ln[1/(1 e)] as e 1.
215
7.14 Exercises
7.1 Demonstrate that the streamlines that pass through a stagnation point in an
incompressible, irrotational, axisymmetric flow pattern cross at right angles.
7.2 A point source of strength Q is placed at the origin in a uniform stream of
incompressible fluid of velocity v = V ez . Show that the stream function of
the resultant flow pattern is
1
Q
cos ,
(7.161)
(r, ) = V r 2 sin2 +
2
4
where r, , are spherical coordinates. Demonstrate that the flow pattern
possesses a single stagnation point at r = a, = , where
$ Q %1/2
a=
,
4 V
and that the streamline, other than = 0, that passes through this stagnation
point satisfies
1
r
=
.
a | sin(/2)|
The flow pattern (7.161) can be reinterpreted as that which results when
a blunt obstacle lying to the right of the previously specified streamline is
placed in a uniform stream of velocity v = V ez . Show that the obstacle in
question has the asymptotic (i.e., as z ) thickness 4 a. Demonstrate that
the pressure distribution over the surface of the obstacle is
3
2
2
2
sin (/2) 1 ,
p = p0 + V sin (/2)
2
where is the fluid mass density, and p0 the pressure at infinity. Show that the
maximum pressure, p = p0 + (1/2) V 2 , on the surface occurs at = ,and
that the minimum pressure, p = p0 (1/6) V 2 , occurs at = 2 sin1 (1/ 3).
Finally, demonstrate that these are, respectively, the maximum and minimum
pressures attained in the whole flow pattern.
7.3 A and B are point sources of strengths Q and Q , respectively, in an infinite incompressible fluid. Here, Q > Q > 0. Show that the equation of a
streamline is
Q cos Q cos = constant,
where , are the angles that AP, BP make with AB. P being a general point.
In addition, show that the cone defined by the equation
cos = 1
2 Q
Q
divides the streamlines issuing from A into two sets, one extending to infinity,
and the other terminating at B. (Milne-Thompson 2011.)
216
7.4 A and B are point sources of equal strength Q (where Q > 0) located at
(, z) = (0, a) and (0, +a), respectively in a uniform stream of incompressible fluid of velocity v = V ez . Show that the stream function of the
resultant flow pattern is
Q
1
(cos 1 cos 2 ),
= V 2 +
2
4
(7.162)
where , are the angles that AP, BP make with AB. P = (, z) being a
general point. Demonstrate that the flow pattern has two stagnation points,
located at C = (0, l) and D = (0, +1), where
(l 2 a 2 )2 = 2 a b 2 l,
and b 2 = Q/(2 V). Show that the streamline (other that = 0) that passes
through these points satisfies
2 = b 2 (cos 1 cos 2 ),
and also passes through the points E = (h, 0) and F = (h, 0), where
h2
2a
= 2
.
2
b
(h + a 2 )1/2
The flow pattern (7.162) can be reinterpreted as that which results when an
axisymmetric solid body of oval cross-section CEDF, lying inside the previously specified streamline, is placed in a uniform stream of velocity v = V ez .
Such an obstacle is known as a Rankine solid.
7.5 Verify that
=
$A
%
cos + B r 2 sin2
r2
is a possible stream function for an axisymmetric, incompressible, irrotational flow pattern, and find the corresponding velocity potential. (MilneThompson 2011.)
7.6 A sphere, a great depth below the surface of an incompressible fluid, is projected with velocity V at an inclination of 45 to the horizontal. If the density
of the sphere is twice that of the fluid, prove that the greatest height above the
point of projection attained by the sphere is 5 V 2 /(8 g). (Milne-Thompson
2011.)
7.7 A sphere of radius a is placed in an incompressible fluid flowing with the
uniform velocity V = V ez . Show that the streamlines of the resultant flow
pattern satisfy
sin2
= constant,
a3 r3
r
where r, are spherical coordinates whose origin coincides with the center
217
of the sphere. If the sphere is divided into two halves by a diametral plane
lying in the x-y plane, show that the resultant force between the two parts is
less than it would have been if the fluid were at rest, the pressure at infinity
remaining the same, by an amount a 2 V 2 /16, where is the fluid density.
(Milne-Thompson 2011.)
7.8 A sphere of radius a is moving along the z-axis through an incompressible
fluid with the variable speed V. Show that the pressure on the surface of the
sphere is least at the intersection of the surface with the plane
z=
2 a 2 dV
,
9 V 2 dt
1 2
c V cos
2
218
8
Incompressible Boundary Layers
8.1 Introduction
Previously, in Sections 4.5, 5.8 and 7.9, we saw that a uniformly flowing incompressible fluid that is modeled as being completely inviscid is incapable of exerting a drag
force on a rigid stationary obstacle placed in its path. This resultwhich is known
as dAlemberts paradoxis surprising because, in practice, a stationary obstacle
experiences a significant drag when situated in such a fluid, even in the limit that the
Reynolds number tends to infinity (which corresponds to the inviscid limit). In this
chapter, we shall attempt to reconcile these two results by introducing the concept
of a boundary layer. This is a comparatively thin layer that covers the surface of
an obstacle placed in a high Reynolds number incompressible fluid. Viscosity is assumed to have a significant eect on the flow inside the layer, but a negligible eect
on the flow outside. For the sake of simplicity, we shall restrict our discussion to
the two-dimensional boundary layers that form when a high Reynolds number fluid
flows transversely around a stationary obstacle of infinite length and uniform crosssection. More information on such boundary layers can be found in Batchelor 2000
and Schlichting 1987.
220
In reality, all physical fluids possess finite viscosity. Moreover, when a viscous
fluid flows around a rigid stationary obstacle both the normal and the tangential components of the fluid velocity are found to be zero at the obstacles surface. The additional constraint that the tangential fluid velocity be zero at a rigid stationary boundary is known as the no slip condition, and is ultimately justified via experimental
observations.
The concept of a boundary layer was first introduced into fluid mechanics by
Ludwig Prandtl (18751953) in order to account for the modification to the flow
pattern of a high Reynolds number irrotational fluid necessitated by the imposition of
the no slip condition on the surface of an impenetrable stationary obstacle. According
to Prandtl, the boundary layer covers the surface of the obstacle, but is relatively thin
in the direction normal to this surface. Outside the layer, the flow pattern is the
same as that of an idealized inviscid fluid, and is thus generally irrotational. This
implies that the normal fluid velocity is zero on the outer edge of the layer, where
it interfaces with the irrotational flow, but, in general, the tangential velocity is nonzero. However, the no slip condition requires the tangential velocity to be zero on
the inner edge of the layer, where it interfaces with the rigid surface. It follows that
there is a very large normal gradient of the tangential velocity across the layer, which
implies the presence of intense internal vortex filaments trapped within the layer.
Consequently, the flow within the layer is not irrotational. In the following, we shall
attempt to make the concept of a boundary layer more precise.
221
irrotational fluid
U (x)
boundary layer
x
solid surface
Figure 8.1
A boundary layer.
(see Section 1.14)
v x vy
+
= 0,
x
y
2
vx 2vx
1 p
v x
v x
+ vy
=
+
+
,
x
y
x
x 2
y 2
2
vy
vy
vy 2 vy
1 p
vx
,
+
+ vy
=
+
x
y
y
x 2
y 2
vx
(8.1)
(8.2)
(8.3)
where is the (constant) density, and the kinematic viscosity. Here, Equation (8.1)
is the equation of continuity, whereas Equations (8.2) and (8.3) are the x- and ycomponents of the fluid equation of motion, respectively. The boundary conditions
at the outer edge of the layer, where it interfaces with the irrotational fluid, are
v x (x, y) U(x),
(8.4)
p(x, y) P(x)
(8.5)
as y/ . Here, P(x) is the fluid pressure at the outer edge of the layer, and
U
dU
1 dP
=
dx
dx
(8.6)
(because vy = 0, and viscosity is negligible, just outside the layer). The boundary
conditions at the inner edge of the layer, where it interfaces with the impenetrable
surface, are
v x (x, 0) = 0,
(8.7)
vy (x, 0) = 0.
(8.8)
222
x
,
L
y
,
vx
,
U0
vy
,
U1
p
,
p0
(8.9)
(8.10)
(8.11)
(8.12)
(8.13)
where U1 and p0 are constants. All of these variables are designed to be O(1) inside
the layer. Equation (8.1) yields
U0 V x U1 Vy
+
= 0.
L X
Y
(8.14)
In order for the terms in this equation to balance one another, we need
U1 =
U0 .
L
(8.15)
In other words, within the layer, continuity requires the typical flow velocity in the
y-direction, U1 , to be much smaller than that in the x-direction, U0 .
Equation (8.2) gives
U02
V x
V x
p 0 P $ U 0 % $ %2 2 V x 2 V x
+ Vy
+
+
.
(8.16)
Vx
=
L
X
Y
L X
L
2
X 2
Y 2
In order for the pressure term on the right-hand side of the previous equation to be of
similar magnitude to the advective terms on the left-hand side, we require that
p0 = U02 .
(8.17)
Furthermore, in order for the viscous term on the right-hand side to balance the other
terms, we need
U1
1
=
=
,
(8.18)
L U0 Re1/2
where
U0 L
(8.19)
Re =
223
is the Reynolds number of the flow external to the layer. (See Section 1.16.) The
assumption that /L 1 can be seen to imply that Re
1. In other words, the
normal thickness of the boundary layer separating an irrotational flow pattern from
a rigid surface is only much less than the typical variation length-scale of the pattern
when the Reynolds number of the flow is much greater than unity.
Equation (8.3) yields
Vy
Vy
1
1 1 2 Vy 2 Vy
P
+ Vy
+
+
.
(8.20)
Vx
=
Re
X
Y
Y Re Re X 2
Y 2
In the limit Re
1, this reduces to
P
= 0.
Y
(8.21)
dP
dU
= U
,
dX
dX
(8.22)
U(X) = U/U0 , and use has been made of Equation (8.6). In other words, the pressure
is uniform across the layer, in the direction normal to the surface of the obstacle, and
is thus the same as that on the outer edge of the layer.
Retaining only O(1) terms, our final set of normalized layer equations becomes
V x Vy
+
= 0,
X
Y
Vx
Vy
d U 2 Vy
V x
+ Vy
=U
+
,
X
Y
X
Y 2
(8.23)
(8.24)
(8.25)
V x (X, 0) = 0,
(8.26)
Vy (X, 0) = 0.
(8.27)
and
dU
2vx
v x
v x
+ vy
=U
+
,
x
y
dx
y 2
(8.28)
(8.29)
(8.30)
224
(8.31)
vy (x, 0) = 0.
(8.32)
,
y
.
x
(8.33)
(8.34)
3 2 2
dU
=U
,
+
dx
y 3 x y 2 y x y
(8.35)
(8.36)
and
(x, 0) = 0,
(8.37)
(x, 0)
= 0.
y
(8.38)
2
,
y 2
(8.39)
whereas the x-component of the viscous force per unit area acting on the surface of
the obstacle is written (see Section 1.18)
v x
2
=
xy y=0 =
(8.40)
.
y y=0
y 2 y=0
225
that, at a given distance, x, along the layer, the tangential flow profile, v x (x, y),
is a scaled version of some common profile: that is, v x (x, y) = U(x) F[y/(x)],
where (x) is a scale-factor, and F(z) a dimensionless function. It follows that
(x, y) = U(x) (x) f [y/(x)], where f (z) = F(z).
Let us search for a self-similar solution to Equation (8.35) of the general form
1/2
1/2
2 U0 x m+1
2
(x, y) =
f () = U0 x m
f (), (8.41)
m+1
(m + 1) U0 x m1
where
(m + 1) U0 x m1
2
1/2
y.
(8.42)
This implies that (x) = [2 /(m + 1) U0 x m1 ]1/2 , and U(x) = U0 x m . Here, U0 and
m are constants. Moreover, U0 x m has dimensions of velocity, whereas m, , and f ,
are dimensionless. Transforming variables from x, y to x, , we find that
m 1
=
+
,
(8.43)
x y x
2 x
x
1/2
(m + 1) U0 x m1
=
.
y x
2
x
(8.44)
Hence,
1/2
U0 x m1
=
[(m + 1) f + (m 1) f ],
x
2 (m + 1)
= U0 x m f ,
y
1/2
(m + 1) U03 x 3 m1
2
f ,
2
y 2
(8.45)
(8.46)
(8.47)
2
U0 x m1
=
[2 m f + (m 1) f ],
x y
2
(m + 1) U02 x 2 m1
3
f ,
=
y 3
2
(8.48)
(8.49)
U02
dU 2
1
.
x 2 m1 dx
(8.50)
226
then
U(x) = U0 x m
(8.52)
(8.53)
where
2m
.
(8.54)
m+1
Expression (8.53) is known as the Falkner-Skan equation. The solutions to this equation that satisfy the physical boundary conditions (8.36)(8.38) are such that
=
f (0) = f (0) = 0,
(8.55)
f () = 1,
(8.56)
and
f () = 0.
(8.57)
(The final condition corresponds to the requirement that the vorticity tend to zero at
the edge of the layer.) Note, from Equations (8.39), (8.42), (8.47), (8.52), (8.55), and
(8.56), that the normally integrated vorticity within the boundary layer is
dy = U(x).
(8.58)
0
Furthermore, from Equations (8.40), (8.47), and (8.52), the x-component of the viscous force per unit area acting on the surface of the obstacle is
$ %1/2
1
xy y=0 = U 2
(m + 1)1/2 2 f (0).
(8.59)
2
Ux
It is convenient to parameterize this quantity in terms of a skin friction coecient,
xy y=0
cf =
.
(8.60)
(1/2) U 2
It follows that
c f (x) =
(m + 1)1/2 2 f (0)
,
[Re(x)]1/2
(8.61)
where
U(x) x
(8.62)
is the eective Reynolds number of the flow on the outer edge of the layer at position
x. Hence, c f (x) x (m+1)/2 . Finally, according to Equation (8.41), the width of the
boundary layer is approximately
Re(x) =
(x)
1
,
x
[Re(x)] 1/2
(8.63)
227
1.5
f (0)
0.5
0
0
Figure 8.2
f (0) calculated as a function of for solutions of the Falkner-Skan equation.
which implies that (x) x (m1)/2 .
If m > 0 then the external tangential velocity profile, U(x) = U0 x m , corresponds
to that of irrotational inviscid flow incident, in a symmetric fashion, on a semi-infinite
wedge whose apex subtends an angle , where = 2 m/(m + 1). (See Section 5.10,
and Figure 5.10.) In this case, U(x) can be interpreted as the tangential velocity
a distance x along the surface of the wedge from its apex (in the direction of the
flow). By analogy, if m = 0 then the external velocity profile corresponds to that of
irrotational inviscid flow parallel to a semi-infinite flat plate (which can be thought
of as a wedge whose apex subtends zero angle). In this case, U(x) can be interpreted
as the tangential velocity a distance x along the surface of the plate from its leading
edge (in the direction of the flow). (See Section 8.5.) Finally, if m < 0 then the
external velocity profile is that of symmetric irrotational inviscid flow over the back
surface of a semi-infinite wedge whose apex subtends an angle (1 ) , where
= m/(1 + m). (See Section 5.11, and Figure 5.11.) In this case, U(x) can be
interpreted as the tangential velocity a distance x along the surface of the wedge
from its apex (in the direction of the flow).
Unfortunately, the Falkner-Skan equation, (8.53), possesses no general analytic
solutions. However, this equation is relatively straightforward to solve via numerical methods. Figure 8.2 shows f (0), calculated numerically as a function of =
2 m/(m + 1), for the solutions of Equation (8.53) that satisfy the boundary conditions
(8.55)(8.57). In addition, Figure 8.3 shows f () versus , calculated numerically
for various dierent values of m. Because 2 as m , solutions of the Falkner-
228
1
0.9
0.8
0.7
0.6
f
0.5
0.4
0.3
0.2
0.1
0
0.1
0
9 10
Figure 8.3
Solutions of the Falkner-Skan equation. In order from the left to the right, the various
solid curves correspond to forward flow solutions calculated with m = 4, 1, 1/3, 1/9,
0, 0.05, and 0.0904, respectively. The dashed curve shows a reversed flow solution
calculated with m = 0.05.
Skan equation with > 2 have no physical significance. For 0 < < 2, it can be
seen, from Figures 8.2 and 8.3, that there is a single solution branch characterized by
f () > 0 and f (0) > 0. This branch is termed the forward flow branch, because it is
such that the tangential velocity, v x () f (), is in the same direction as the external
tangential velocity [i.e., v x ()] across the whole layer (i.e., 0 < < ). The forward flow branch is characterized by a positive skin friction coecient, c f f (0).
It can also be seen that for < 0 there exists a second solution branch, which is
termed the reversed flow branch, because it is such that the tangential velocity is in
the opposite direction to the external tangential velocity in the region of the layer
immediately adjacent to the surface of the obstacle (which corresponds to = 0).
The reversed flow branch is characterized by a negative skin friction coecient. The
reversed flow solutions are probably unphysical, because reversed flow close to the
wall is generally associated with a phenomenon known as boundary layer separation
(see Section 8.10) that invalidates the boundary layer orderings. It can be seen that
the two solution branches merge together at = = 0.1989, which corresponds
to m = m = 0.0905. Moreover, there are no solutions to the Falkner-Skan equation with < or m < m . The disappearance of solutions when m becomes too
negative (i.e., when the deceleration of the external flow becomes too large) is also
related to boundary layer separation.
229
y
boundary layer
plate
U0
x
L
wake
Figure 8.4
Flow over a flat plate.
230
1
0.9
0.8
vx / U0
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
5 4 3 2 1 0 1
y/
Figure 8.5
Tangential velocity profile across the boundary layers located above and below a flat
plate of negligible thickness located at y = 0.
the upper one, the tangential velocity profile across the both layers is written
v x (x, y) = U0 f (),
where
=
Here, f () is the solution of
$ U %1/2
0
|y|.
2x
f + f f = 0
(8.64)
(8.65)
(8.66)
(8.67)
f () = 1,
(8.68)
and
f () = 0.
(8.69)
231
which can be interpreted as the distance through which streamlines just outside the
layer are displaced laterally due to the retardation of the flow within the layer. (Of
course, the thickness of the lower boundary layer is the same as that of the upper
layer.) It follows that
(x) =
x
U0
1/2
2
[1 f ()] d.
(8.71)
In fact, the numerical solution of Equation (8.66), subject to the boundary conditions (8.67)(8.69), yields
1/2
x
(x) = 1.72
.
(8.72)
U0
Hence, the thickness of the boundary layer increases as the square root of the distance
from the leading edge of the plate. In particular, the thickness at the trailing edge of
the plate is
1.72
(L)
=
,
(8.73)
L
Re1/2
where
U0 L
(8.74)
Re =
is the appropriate Reynolds number for the interaction of the flow with the plate.
Note that if Re
1 then the thickness of the boundary layer is much less than its
length, as was previously assumed.
The tangential velocity profile across the both boundary layers, which takes the
form
|y|
v x (x, y) = U0 f 1.22
,
(8.75)
(x)
is plotted in Figure 8.5. In addition, the vorticity profile across the layers, which is
written
|y|
U0
f 1.22
(x, y) = sgn(y) 1.22
,
(8.76)
(x)
(x)
is shown in Figure 8.6. The vorticity is negative in the upper boundary layer (i.e.,
y > 0), positive in the lower boundary layer (i.e., y < 0), and discontinuous across
the plate (which is located at y = 0). Finally, the net viscous drag force per unit width
(along the z-axis) acting on the plate in the x-direction is
D=2
0
xy y=0 dx,
(8.77)
where the factor of 2 is needed to take into account the presence of boundary layers
both above and below the plate. It follows from Equation (8.59) (with m = 0) that
D = U02
U0
1/2
2 f (0)
0
x1/2 dx = U02
L
U0
1/2
2 2 f (0).
(8.78)
/ (U0/)
232
0.6
0.5
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
5 4 3 2 1 0 1
y/
Figure 8.6
Vorticity profile across the boundary layers located above and below a flat plate of
negligible thickness located at y = 0.
In fact, the numerical solution of (8.66) yields
D = 1.33
U02 L
Re1/2
= 1.33 U0 ( U0 L)1/2 .
(8.79)
The previous discussion is premised on the assumption that the flow in the upper (or lower) boundary layer is both steady and z-independent. It turns out that
this assumption becomes invalid when the Reynolds number of the layer, U0 /, exceeds a critical value which is about 600 (Batchelor 2000). In this case, small-scale
z-dependent disturbances spontaneously grow to large amplitude, and the layer becomes turbulent. Because x1/2 , if the criterion for boundary layer turbulence is
not satisfied at the trailing edge of the plate, x = L, then it is not satisfied anywhere
else in the layer. Thus, the previous analysis, which neglects turbulence, remains
valid provided U0 (L)/ < 600. According to Equation (8.73), this implies that the
analysis is valid when 1 Re < 1.2 105 , where Re = U0 L/ is the Reynolds
number of the external flow.
Consider, finally, the situation illustrated in Figure 8.7 in which an initially irrotational fluid passes between two flat parallel plates. Let d be the perpendicular
distance between the plates. As we have seen, the finite viscosity of the fluid causes
boundary layers to form on the inner surfaces of the upper and lower plates. The
flow within these layers possesses non-zero vorticity, and is significantly aected by
233
boundary layer
plate
U0
potential flow
Figure 8.7
Flow between two flat parallel plates.
viscosity. On the other hand, the flow outside the layers is irrotational and essentially inviscidthis type of flow is usually termed potential flow (because it can be
derived from a velocity potential satisfying Laplaces equation). The thickness of the
two boundary layers increases like x1/2 , where x represents distance, parallel to the
flow, measured from the leading edges of the plates. It follows that, as x increases, the
region of potential flow shrinks in size, and eventually disappears. (See Figure 8.7.)
Assuming that, prior to merging, the two boundary layers do not significantly aect
one another, their thickness, (x), is given by formula (8.72), where U0 is the speed
of the incident fluid. The region of potential flow thus extends from x = 0 (which
corresponds to the leading edge of the plates) to x = l, where
(l) =
d
.
2
(8.80)
It follows that
l
= 11.8 Re,
d
(8.81)
where
Re =
U0 d
.
(8.82)
Thus, when an irrotational high Reynolds number fluid passes between two parallel
plates then the region of potential flow extends a comparatively long distance between the plates, relative to their spacing (i.e., l/d
1). By analogy, if an irrotational
high Reynolds number fluid passes into a pipe then the fluid remains essentially irrotational until it has travelled a considerable distance along the pipe, compared to its
diameter. Obviously, these conclusions are modified if the flow becomes turbulent.
234
(8.83)
(8.84)
where
|u| U0 .
(8.85)
1
,
x x
1
,
y
(8.86)
(8.87)
u.
x
(8.88)
The flow external to the boundary layers, and the wake, is both uniform and essentially inviscid. Hence, according to Bernoullis theorem, the pressure in this region
is also uniform. [See Equation (8.22).] However, as we saw in Section 8.3, there is
no y-variation of the pressure across the boundary layers. It follows that the pressure
is uniform within the layers. Thus, it is reasonable to assume that the pressure is
also uniform within the wake, because the wake is formed via the convection of the
boundary layers downstream of the plate. We conclude that
p(x, y) p0
everywhere in the fluid, where p0 is a constant.
(8.89)
235
2
vx 2vx
1 p
v x
v x
+ vy
=
+
vx
+
.
x
y
x
x 2
y 2
(8.90)
2u
u
2.
x
y
(8.91)
(8.92)
ensures that the flow outside the wake remains unperturbed. Note that Equation (8.91)
has the same mathematical form as a conventional diusion equation, with x playing the role of time, and /U0 playing the role of the diusion coecient. Hence,
by analogy with the standard solution of the diusion equation, we would expect
( x/U0 )1/2 (Riley 1974).
As can easily be demonstrated, the self-similar solution to Equation (8.91), subject to the boundary condition (8.92), is
2
y
Q
exp 2 ,
u(x, y) =
(8.93)
where
(x) = 2
x
U0
1/2
,
(8.94)
u dy = Q,
(8.95)
(8.96)
is plotted in Figure 8.8. In addition, the vorticity profile across the wake, which is
written
Q 2 y
(x, y)
=
exp(y 2 / 2 )
(8.97)
U0 /
U0
is shown in Figure 8.9. It can be seen that the profiles pictured in Figures 8.8 and
8.9 are essentially smoothed out versions of the boundary layer profiles shown in
Figures 8.5 and 8.6, respectively.
Suppose that the plate and a portion of its trailing wake are enclosed by a cuboid
236
1
0.9
0.8
vx / U0
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
3
0
y/
Figure 8.8
Tangential velocity profile across the wake of a flat plate of negligible thickness
located at y = 0. The profile is calculated for Q/(U0 ) = 0.5.
0.5
0.4
0.3
/ (U0/)
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
3
0
y/
Figure 8.9
Vorticity profile across the boundary layers above and below a flat plate of negligible
thickness located at y = 0. The profile is calculated for Q/(U0 ) = 0.5.
237
v(x)
y
y=h
plate
wake
U0 u(y)
U0
y = h
x=l
x = l
v(x)
Figure 8.10
Control volume surrounding a flat plate and its trailing wake.
control volume of unit depth (in the z-direction) that extends from x = l to x = +l
and from y = h to y = h. (See Figure 8.10.) Here, l
L and h
(l), where L
is the length of the plate, and (x) the width of the wake. Hence, the control volume
extends well upstream and downstream of the plate. Moreover, the volume is much
wider than the wake.
Let us apply the integral form of the fluid equation of continuity to the control
volume. For a steady state, this reduces to (see Section 1.9)
v dS = 0,
(8.98)
S
where S is the bounding surface of the control volume. The normal fluid velocity
is U0 at x = l, U0 u(y) at x = l, and v(x) at y = h, as indicated in the figure.
Hence, Equation (8.98) yields
or
U0 dy +
[U0 u(y)] dy + 2
u(y) dy = 2
v(x) dx.
v(x) dx = 0,
(8.99)
(8.100)
238
u(y) dy = 2
v(x) dx Q,
(8.101)
where Q is independent of x. Note that the slight retardation of the flow inside the
wake, due to the presence of the plate, which is parameterized by Q, necessitates a
small lateral outflow, v(x), in the region of the fluid external to the wake.
Let us now apply the integral form of the x-component of the fluid equation of
motion to the control volume. For a steady state, this reduces to (see Section 1.11)
v x v dS = F x +
x j dS j ,
(8.102)
S
where F x is the net x-directed force exerted on the fluid within the control volume
by the plate. It follows, from Newtons third law of motion, that F x = D, where D
is the viscous drag force per unit width (in the z-direction) acting on the plate in the
x-direction. In an incompressible fluid (see Section 1.6),
vi v j
i j = p i j +
+
.
(8.103)
x j xi
Hence, we obtain
h
2
U0 dy +
h
!
"
U0 u(y) 2 dy
+2
U0 v(x) dx = D 2
d
dl
u(y) dy,
h
(8.104)
because the pressure within the fluid is essentially uniform, the tangential fluid velocity at y = h is U0 , and v is assumed to be negligible at x = l. Making use of
Equation (8.101), as well as the fact that Q is independent of l, we get
D = U0 Q.
(8.105)
Here, we have neglected any terms that are second order in the small quantity u. A
comparison with Equation (8.79) reveals that
or
Q = 1.33 ( U0 L)1/2 ,
(8.106)
$ L %1/2
Q
= 0.664
.
U0
x
(8.107)
Hence, from Equations (8.96) and (8.97), the velocity and vorticity profiles across
the layer are
$ L %1/2
v x (x, y)
= 1 0.375
exp(y 2 / 2 ),
(8.108)
U0
x
239
$ L %1/2 y
(x, y)
= 0.749
exp(y 2 / 2 ),
U0 /
x
(8.109)
respectively, where (x) = 2 ( x/U0 )1/2 . Finally, because the previous analysis is
premised on the assumption that |1 v x /U0 | = |u|/U0 1, it is clear that the previous
three expressions are only valid when x
L (i.e., well downstream of the plate).
The previous analysis only holds when the flow within the wake is non-turbulent.
Let us assume, by analogy with the discussion in the previous section, that this is the
case as long as the Reynolds number of the wake, U0 (x)/, remains less than some
critical value that is approximately 600. Because the Reynolds number of the wake
can be written 2 Re1/2 (x/L)1/2 , where Re = U0 L/ is the Reynolds number of the
external flow, we deduce that the wake becomes turbulent when x/L > 9 104 /Re.
Hence, the wake is always turbulent suciently far downstream of the plate. Our
analysis, which eectively assumes that the wake is non-turbulent in some region,
immediately downstream of the plate, whose extent (in x) is large compared with L,
is thus only valid when 1 Re 9 104 .
u
u
dU
2u
+v
U
= 2,
x
y
dx
y
(8.110)
(8.111)
(8.112)
(8.113)
240
Here, U(x) is the external tangential fluid velocity at the edge of the layer. Integrating
(8.111) across the layer, making use of the boundary conditions (8.113), leads to
u
u
u
dU
=
u
v
U
dy
y y=0
dx
x
y
0
(U u)
(U u)
dU
+u
+v
(U u)
=
dy
dx
x
y
0
(U u)
v
dU
+u
(U u)
(U u)
dy
=
dx
x
y
0
(U u)
u
dU
+u
+ (U u)
(U u)
=
dy
dx
x
x
0
d
dU
(U u) dy +
u (U u) dy.
(8.114)
=
dx 0
dx 0
Here, we have integrated the final term on the right-hand side by parts, making use
of Equations (8.110), (8.112), and (8.113). Let us define the displacement thickness
of the layer [see Equation (8.70)]
$
u%
1 (x) =
dy,
(8.115)
1
U
0
as well as the so-called momentum thickness
$
u%
u
2 (x) =
1
dy.
U
0 U
It follows from Equation (8.114) that
u
dU
d
= U 2 2 + U
(1 + 2 2 ).
y y=0
dx
dx
(8.116)
(8.117)
This important result is known as the von Karman momentum integral, and is fundamental to many of the approximation methods commonly employed to calculate boundary layer thicknesses on the surfaces of general obstacles placed in high
Reynolds number flows. (See Section 8.10.)
241
separation point
potential flow streamlines
boundary layer
obstacle
wake
stagnation point
Figure 8.11
Boundary layer separation.
number of the external flow. Suppose, however, that the obstacle is not streamlined:
that is, the surface of the obstacle is not closely aligned with the streamlines of the
unperturbed flow pattern. In this case, the typically observed behavior is illustrated
in Figure 8.11, which shows the flow pattern of a high Reynolds number irrotational
fluid around a cylindrical obstacle (whose axis is normal to the direction of the unperturbed flow). It can be seen that a stagnation point, at which the flow velocity is
locally zero, forms in front of the obstacle. Moreover, a thin boundary layer covers
the front side of the obstacle. The thickness of this layer is smallest at the stagnation
point, and increases towards the back side of the obstacle. However, at some point
on the back side, the boundary layer separates from the obstacles surface to form a
vortex-filled wake whose transverse dimensions are similar to those of the obstacle
itself. This phenomenon is known as boundary layer separation.
Outside the boundary layer, and the wake, the flow pattern is irrotational and
essentially inviscid. So, from Section 5.8, the tangential flow speed just outside the
boundary layer (neglecting any circulation of the external flow around the cylinder)
is
(8.118)
U() = 2 U0 sin ,
where U0 is the unperturbed flow speed, and is a cylindrical coordinate defined
such that the stagnation point corresponds to = 0. Note that the tangential flow
accelerates (i.e., increases with increasing arc-length, along the surface of the obstacle, in the direction of the flow) on the front side of the obstacle (i.e., 0 /2),
and decelerates on the back side. Boundary layer separation is always observed to
242
take place at a point on the surface of an obstacle where there is deceleration of the
external tangential flow. In addition, from Section 5.8, the pressure just outside the
boundary layer (and, hence, on the surface of the obstacle, because the pressure is
uniform across the layer) is
P() = p1 + U02 cos(2 ),
(8.119)
D
,
U02 a
(8.120)
where is the fluid density, and a the typical transverse size of the obstacle (in the
present example, the radius of the cylinder). The drag force that acts on a nonstreamlined obstacle placed in a high Reynolds number flow, as a consequence of
boundary layer separation, is generally characterized by a drag coecient of order
unity. The exact value of the coecient depends strongly on the shape of the obstacle,
but only relatively weakly on the Reynolds number of the flow. Consequently, this
type of drag is termed form drag, because it depends primarily on the external shape,
or form, of the obstacle. Form drag scales roughly as the cross-sectional area (per
unit width) of the vortex-filled wake that forms behind the obstacle.
Boundary layer separation is associated with strong adverse pressure gradients,
or, equivalently, strong flow deceleration, on the back side of an obstacle placed in a
243
/2
3/2
P () p1
stagnation point
separation points
Figure 8.12
Pressure variation over surface of a cylindrical obstacle in a high Reynolds number
flow both with (dashed curve) and without (solid curve) boundary layer separation.
high Reynolds number flow. Such gradients can be significantly reduced by streamlining the obstacle: that is, by closely aligning its back surface with the unperturbed
streamlines of the external flow. Indeed, boundary layer separation can be delayed,
or even completely prevented, on the surface of a suciently streamlined obstacle,
thereby significantly decreasing, or even eliminating, the associated form drag (essentially, by reducing the cross-sectional area of the wake). However, even in the
limit that the form drag is reduced to a negligible level, there is still a residual drag
acting on the obstacle due to boundary layer viscosity. This type of drag is called
friction drag. As is clear from a comparison of Equations (8.79) and (8.120), the
drag coecient associated with friction drag is O(Re1/2 ), where Re is the Reynolds
number of the flow. Friction drag thus tends to zero as the Reynolds number tends to
infinity.
The phenomenon of boundary layer separation allows us to resolve dAlemberts
paradox. Recall, from Section 5.8, that an idealized fluid that is modeled as inviscid
and irrotational is incapable of exerting a drag force on a stationary obstacle, despite
the fact that very high Reynolds number, ostensibly irrotational, fluids are observed
to exert significant drag forces on stationary obstacles. The resolution of the paradox
lies in the realization that, in such fluids, viscosity can only be neglected (and the
flow is consequently only irrotational) in the absence of boundary layer separation.
In this case, the region of the fluid in which viscosity plays a significant role is localized to a thin boundary layer on the surface of the obstacle, and the resultant friction
244
drag scales as Re1/2 , and, therefore, disappears in the inviscid limit (essentially, because the boundary layer shrinks to zero thickness in this limit). On the other hand,
if the boundary layer separates then viscosity is important both in a thin boundary
layer on the front of the obstacle, and in a wide, low-pressure, vortex-filled, wake
that forms behind the obstacle. Moreover, the wake does not disappear in the inviscid limit. The presence of significant fluid vorticity within the wake invalidates
irrotational fluid dynamics. Consequently, the pressure on the back side of the obstacle is significantly smaller than that predicted by irrotational fluid dynamics. Hence,
the resultant pressure force on the front side is larger than that on the back side, and
a significant drag is exerted on the obstacle. The drag coecient associated with this
type of drag is generally of order unity, and does not tend to zero as the Reynolds
number tends to infinity.
(8.121)
245
(8.123)
(8.124)
1
2
(y ) dy .
(8.125)
(8.126)
u
dU
2u
u
+v
=U
+ 2.
x
y
dx
y
(8.127)
As is clear from Equation (8.123), the derivative 2 u/y 2 does not become infinite
as x x0 . The same is true of the function U dU/dx, which is determined from the
flow outside the boundary layer. However, both terms on the left-hand side of the
previous expression become infinite as x x0 . Hence, in the immediate vicinity of
the separation point,
u
u
+v
0.
(8.128)
u
x
y
Because u/x = v/y, we can rewrite this equation in the form
u
v
u
$v%
+v
= u 2
0.
y
y
y u
(8.129)
(8.130)
In other words, v/u is a function of x only. From Equations (8.123) and (8.125),
(y)
v
=
+ O(1).
u u0 (y) x0 x
(8.131)
246
Hence, if this ratio is a function of x alone then (y) = (1/2) A u0(y), where A is a
constant: that is,
A u0 (y)
v(x, y)
.
(8.132)
2 x0 x
Finally, because Equation (8.126) yields = 2 d/dy = A du0 /dy, we obtain
u(x, y) u0 (y) + A
du0
x0 x.
dy
(8.133)
The previous two expressions specify u and v as functions of x and y near the point
of separation. Beyond the point of separation, that is for x > x0 , the expressions
are physically meaningless, because the square roots become imaginary. This implies that the solutions of the boundary layer equations cannot sensibly be continued
beyond the separation point.
The standard boundary conditions at the surface of the obstacle require that u =
v = 0 at y = 0. It, therefore, follows from Equations (8.132) and (8.133) that
u0 (0) = 0,
du0
= 0.
dy
(8.134)
(8.135)
y=0
Thus, we obtain the important prediction that both the tangential velocity, u, and its
first derivative, u/y, are zero at the separation point (i.e., x = x0 and y = 0). This
result was originally obtained by Prandtl, although the argument we have used to
derive it is due to L. D. Landau (19081968) (Landau and Lifshitz 1987).
If the constant A in expressions (8.132) and (8.133) happens to be zero then the
point x = x0 and y = 0, at which the derivative u/y vanishes, has no particular
properties, and is not a point of separation. However, there is no reason, in general,
why A should take the special value zero. Thus, in practice, a point on the surface of
an obstacle at which u/y = 0 is always a point of separation.
Incidentally, if there were no separation at the point x = x0 (i.e., if A = 0) then we
would have u/y < 0 for x > x0 . In other words, u would become negative as we
move away from the surface, y being still small. That is, the fluid beyond the point
x = x0 would move tangentially, in the region of the boundary layer immediately
adjacent to the surface, in the direction opposite to that of the external flow: that is,
there would be back-flow in this region. In practice, the flow separates from the
surface at x = x0 , and the back-flow migrates into the wake.
The dimensionless boundary layer equations, (8.23)(8.27), are independent of
the Reynolds number of the external flow (assuming that this number is much greater
than unity). Thus, it follows that the point on the surface of the obstacle at which
u/y = 0 is also independent of the Reynolds number. In other words, the location
of the separation point is independent of the Reynolds number (as long as this number
is large, and the flow in the boundary layer is non-turbulent).
At y = 0, the equation of tangential motion in the boundary layer, (8.111), is
written
1 dU 1 dP
2 u
=
,
(8.136)
=
2
U
dx
dx
y y=0
247
where P(x) is the pressure just outside the layer, and use has been made of Equation (8.6). Because u is positive, and increases away from the surface (upstream of
the separation point), it follows that ( 2 u/y 2)y=0 > 0 at the separation point itself,
where (u/y)y=0 = 0. Hence, according to the previous equation,
dU
< 0,
(8.137)
dx x=x0
dP
> 0.
(8.138)
dx x=x0
In other words, we predict that the external tangential flow is always decelerating at
the separation point, whereas the pressure gradient is always adverse (i.e., such as to
decelerate the tangential flow), in agreement with experimental observations.
u
u
dU
2u
+v
U
= 2,
x
y
dx
y
(8.139)
(8.140)
(8.141)
u(x, 0) = 0,
(8.142)
v(x, 0) = 0.
(8.143)
= U
y 2 y=0
dx
(8.144)
(8.145)
248
(8.146)
2 dU
.
dx
Finally, let us assume that f , f , and f are continuous at = 1: that is,
=
f (1) = 1,
f (1) = 0,
f (1) = 0.
(8.148)
(8.149)
(8.150)
(8.151)
These constraints corresponds to the reasonable requirements that the velocity, vorticity, and viscous stress tensor, respectively, be continuous across the layer. Given
that a = 0, Equations (8.146), (8.147), and (8.149)(8.151) yield
f () = F() + G()
(8.152)
F() = 1 (1 ) 3 (1 + ),
(8.153)
for 0 1, where
1
(1 ) 3 .
(8.154)
6
Thus, the tangential velocity profile across the layer is a function of a single parameter, , which is termed the Pohlhausen parameter. The behavior of this profile is
illustrated in Figure 8.13. Note that, under normal circumstances, the Pohlhausen
parameter must lie in the range 12 12. For > 12, the profile is such that
f () > 1 for some < 1, which is not possible in a steady-state solution. On the other
hand, for < 12, the profile is such that f (0) < 0, which implies flow reversal
close to the wall. As we have seen, flow reversal is indicative of separation. Indeed,
the separation point, f (0) = 0, corresponds to = 12. Expression (8.152) is only
an approximation, because it satisfies some, but not all, of the boundary conditions
satisfied by the true velocity profile. For instance, dierentiation of Equation (8.140)
with respect to y reveals that ( 3 u/y 3)y=0 f (0) = 0, which is not the case for
expression (8.152).
It follows from Equations (8.115), (8.116), and (8.152)(8.154) that
1
,
(8.155)
1 (x) =
(1 f ) d =
10 120
0
1
37
2
2 (x) =
f (1 f ) d =
.
(8.156)
315 945 9072
0
G() =
249
1
0.9
0.8
0.7
0.6
f 0.5
0.4
0.3
0.2
0.1
0
0.2
0.4
0.6
0.8
Figure 8.13
Pohlhausen velocity profiles for = 12 (solid curve) and = 12 (dashed curve).
Furthermore,
u
U
U$
%
=
f (0) =
2+
.
y y=0
(8.157)
d2 22 dU
U
2 u
1
.
2
+
=
2+
dx
dx
2
U y y=0
(8.158)
Defining
(x) =
22 dU
,
dx
(8.159)
we obtain
U
(8.160)
250
1
F
0
0.15
0.1
0.05
0.05
0.1
Figure 8.14
The function F() (solid curve) and the linear function 0.47 6 (dashed line).
where
2
37
,
(8.161)
=
315 945 9072
37
2
1
F1 () =
=
,
(8.162)
2
10 120
315 945 9072
$
2
2 u
% 37
= 2 +
,
(8.163)
F2 () =
U y y=0
6 315 945 9072
2
2
1
37
2
116
+
+
3 .
F() = 2
2
2 +
315 945 9072
315
945 120
9072
(8.164)
It is generally necessary to integrate Equation (8.158) from the stagnation point at
the front of the obstacle, through the point of maximum tangential velocity, to the
separation point on the back side of the obstacle. At the stagnation point we have
U = 0 and dU/dx 0, which implies that F() = 0. Furthermore, at the point of
maximum tangential velocity we have dU/dx = 0 and U 0, which implies that
= = 0. Finally, as we have already seen, = 12 at the separation point, which
implies, from Equation (8.161), that = 0.1567.
As was first pointed out by Walz (Schlichting 1987), and is illustrated in Figure 8.14, it is a fairly good approximation to replace F() by the linear function
251
d U 22
dU 22
= 0.47 5
,
(8.165)
dx
dx
which can be integrated to give
22 0.47
=
U6
U 5 (x ) dx ,
(8.166)
(8.167)
(8.168)
0.47 m
0.47
=
.
1+5m
/2 4
(8.172)
252
0.2
0.1
20
40
60
()
80
Figure 8.15
The function () for flow around a circular cylinder.
Figure 8.16
Flow over the back surface of a semi-infinite wedge.
100
120
253
We expect boundary layer separation on the back surface of the wedge when <
s = 0.1567. This corresponds to > s , where
s =
( s )
13 .
2 0.47 + 4 ( s)
(8.173)
Hence, boundary layer separation can be prevented by making the wedges angle of
dip suciently shallow: that is, by streamlining the wedge, which has the eect of
reducing the deceleration of the flow on the wedges back surface. The critical value
of m (i.e., m s = 0.0125) at which separation occurs in our approximate solution
is very similar to the critical value of m (i.e., m = 0.0905) at which the exact
self-similar solutions described in Section 8.4 can no longer be found. This suggests
that the absence of self-similar solutions for m < m is related to boundary layer
separation.
8.11 Exercises
8.1 Fluid flows between two non-parallel plane walls, towards the intersection
of the planes, in such a manner that if x is measured along a wall from the
intersection of the planes then U(x) = U0 /x, where U0 is a positive constant.
Verify that a solution of the boundary layer equation (8.35) can be found such
that is a function of y/x only. Demonstrate that this solution yields
$ %1/2
U0
y
u(x, y)
=F
,
U(x)
x
where u = /y, and
F F 2 = 1,
z
2
F(z) = 3 tanh + 2
2
is a suitable solution of the previous dierential equation, where tanh2 =
2/3.
8.2 A jet of water issues from a straight narrow slit in a wall, and mixes with
the surrounding water, which is at rest. On the assumption that the motion is
non-turbulent and two-dimensional, and that the approximations of boundary
layer theory apply, the stream function satisfies the boundary layer equation
3 2 2
= 0.
+
y 3 x y 2 y x y
Here, the symmetry axis of the jet is assumed to run along the x-direction,
254
,
y
where u(x, y) = u(x, y), and u(x, y) 0 as y . We expect the momentum flux of the jet parallel to its symmetry axis,
M=
u 2 dy,
to be independent of x.
Consider a self-similar stream function of the form
(x, y) = 0 x p F(y/x q ).
Demonstrate that the boundary layer equation requires that p+q = 1, and that
M is only independent of x when 2 p q = 0. Hence, deduce that p = 1/3
and q = 2/3.
Suppose that
(x, y) = 6 x 1/3 F(y/x 2/3).
Demonstrate that F(z) satisfies
F + 2 F F + 2 F 2 = 0,
subject to the constraints that F (z) = F (z), and F (z) 0 as z . Show
that
F(z) = tanh( z)
is a suitable solution, and that
M = 48 2 3 .
8.3 The growth of a boundary layer can be inhibited by sucking some of the fluid
through a porous wall. Consider conventional boundary layer theory. As a
consequence of suction, the boundary condition on the normal velocity at the
wall is modified to v(x, 0) = v s , where v s is the (constant) suction velocity.
Demonstrate that, in the presence of suction, the von Karman velocity integral
becomes
dU
d
u
= U2 2 +U
(1 + 2 2 ) + U v s .
y y=0
dx
dx
Suppose that
u(x, y) = U(x)
sin( y)
1
0 y /(2 )
,
y > /(2 )
255
1/2
1/2
8
x
1 = (/2 1)
,
4
U0
but that in the presence of suction
1 =
(/2 1)
.
vs
,
U0
Re1/2
where Re = U0 L/.
256
9
Incompressible Aerodynamics
9.1 Introduction
This chapter investigates the forces exerted on a stationary obstacle situated in a uniform, high Reynolds number, subsonic (and, therefore, eectively incompressible
see Section 1.17) wind, on the assumption that the obstacle is suciently streamlined
that there is no appreciable separation of the boundary layer from its back surface.
Such an obstacle is termed an airfoil (or aerofoil). Obviously, airfoil theory is fundamental to the theory of flight. Further information on this subject can be found in
Milne-Thomson 1958.
The flow around an airfoil is essentially irrotational and inviscid everywhere apart
from a thin boundary layer localized to its surface, and a thin wake emitted by its
trailing edge. (See Sections 8.5 and 8.6.) It follows that, for the flow external to the
boundary layer and wake, we can write
v = ,
(9.1)
which automatically ensures that the flow is irrotational. Assuming that the flow is
also incompressible, so that v = 0, the velocity potential, , satisfies Laplaces
equation: that is,
(9.2)
2 = 0.
The appropriate boundary condition at the surface of the airfoil is that the normal
velocity be zero. In other words, n = 0, where n is a unit vector normal to the
surface. In general, the tangential velocity at the airfoil surface, obtained by solving
2 = 0 in the external region, subject to the boundary condition n = 0 on
the surface, is non-zero. Of course, this is inconsistent with the no slip condition,
which demands that the tangential velocity be zero at the surface. (See Section 8.2.)
However, as described in the previous chapter, this inconsistency is resolved by the
boundary layer, across which the tangential velocity is eectively discontinuous, being non-zero on the outer edge of the layer (where it interfaces with the irrotational
flow), and zero on the inner edge (where it interfaces with the airfoil). The discontinuity in the tangential velocity across the layer implies the presence of bound vortices
covering the surface of the airfoil (see Section 9.7), and also gives rise to a friction
drag acting on the airfoil in the direction of the external flow. However, the magnitude of this drag scales as Re 1/2 , where Re is the Reynolds number of the wind.
(See Section 8.5.) Hence, such drag becomes negligibly small in the high Reynolds
257
258
number limit. In the following, we shall assume that any form drag, due to the
residual separation of the boundary layer at the back of the airfoil, is also negligibly
small. Moreover, for the sake of simplicity, we shall initially restrict our discussion
to two-dimensional situations in which a high Reynolds number wind flows transversely around a stationary airfoil of infinite length (in the z-direction) and uniform
cross-section (parallel to the x-y plane).
(9.3)
According to Equation (6.172), the circulation, , of air about the airfoil is determined by performing the integral
dF
Re
dz =
(9.4)
C dz
around a loop C that lies just above the airfoil surface. However, as discussed in
Section 6.10, the value of this integral is unchanged if it is performed around any
loop that can be continuously deformed onto C, while not passing through the airfoil
surface, or crossing a singularity of the complex velocity, dF/dz (i.e., a line source or
a z-directed vortex filament). Because (in the high Reynolds number limit in which
the boundary layer and the wake are infinitely thin) there are no line sources or zdirected vortex filaments external to the airfoil, we can evaluate the integral around
a large circle of radius R, centered on the origin. It follows that z = R e i and
dz = i R e i d = i z d. Hence,
i (+)
1
= Re i
V Re
+ A + O(R ) d = 2 Im(A),
(9.5)
which implies that
dF
B
= V ei + i
+
+
dz
2 z z 2
(9.6)
at large |z|.
As discussed in Section 6.11, the net force (per unit length) acting on the airfoil,
Incompressible Aerodynamics
L = X e x + Y ey , is determined by performing the Blasius integral,
2
dF
1
i
dz = X i Y,
2
C dz
259
(9.7)
around a loop C that lies just above the airfoil surface. However, as before, the value
of the integral is unchanged if instead we perform it around a large circle of radius
R, centered on the origin. Far from the airfoil,
2
V e i 82 V B e i 2
dF
= V 2 e2i + i
+ O(z 3 ).
(9.8)
+
dz
z
42 z 2
So, we obtain
X iY =
or
1 2
i
2
V ei
+ O(R 1 ) d = i e i V , (9.9)
V 2 R e i (+2 ) + i
X + i Y = i ei V = e i (/2) V .
(9.10)
In other words, the resultant force (per unit length) acting on the airfoil is of magnitude V , and has the direction obtained by rotating the wind vector through a rightangle in the sense opposite to that of the circulation. This type of force is known as
lift, and is responsible for flight. The result (9.10) is known as the theorem of Kutta
and Zhukovskii, after the German scientist M. W. Kutta (18671944), and the Russian scientist N. E. Zhukovskii (18471921), who discovered it independently. Note
that (at fixed circulation) the lift is independent of the shape of the airfoil. Furthermore, according to the KuttaZhukovskii theorem, there is zero drag acting on the
airfoil (i.e., zero force acting in the direction of the wind). In reality, there is always
a small friction drag due to air viscosity, as well as a (hopefully) small form drag
due to residual separation of the boundary layer from the back of the airfoil. There is
actually a third type of drag, known as induced drag, that is discussed in Section 9.8.
As discussed in Section 6.11, the net moment per unit length (about the origin),
M, acting on the airfoil is determined by performing the integral
2
dF
1
z dz = M
Re
(9.11)
2
C dz
around a loop C that lies just above the airfoil surface. As before, we can deform C
into a circle of radius R, centered on the origin, without changing the value of the
integral. Hence, we obtain
V R e i (+)
1
V 2 R 2 e 2 i (+) + i
M = Re i
2
2
i
2
8 V B e
+
+ O(R 1 ) d ,
(9.12)
42
or
M = Re 2 V B e i (/2) .
(9.13)
260
$ %
a 2 i
e
ln
+i
F() = V e i +
,
(9.14)
2
a
whereas the associated stream function takes the form
$r%
a2
ln
(r, ) = V r
sin( + ) +
,
r
2
a
(9.15)
= V ei + i
.
d
2
2
(9.16)
(9.17)
(9.18)
(9.19)
(9.20)
where V is the wind vector, L the lift vector, and M the moment of the lift vector
about the origin. We conclude that, for a cylindrical airfoil of circular cross-section,
the lift vector is normal to the wind vector, and the line of action of the lift passes
through the centroid of the cross-section (because the lift generates zero moment
about the origin). (See Figure 9.1.)
Of course, a cylindrical airfoil of circular cross-section is completely unrealistic,
because its back side (i.e., the side opposite to that from which the wind is incident)
is not suciently streamlined to prevent boundary layer separation. (See Chapter 8.)
However, as described in Section 6.7, we can use the conformal transformation
z=+
l2
(9.21)
Incompressible Aerodynamics
261
Figure 9.1
A cylindrical airfoil of circular cross-section.
to transform a cylinder of circular cross-section in the -plane to a cylinder of elliptical cross-section in the z-plane. (Note that both cross-sections have centroids located
at the origin.) Moreover, a cylindrical airfoil of elliptical cross-section that is suciently elongated, and whose major axis subtends a suciently small angle with the
incident wind direction, constitutes a realistic airfoil, because its back side is, for the
most part, closely aligned with the external flow.
An elliptical airfoil of width c and thickness < c, as shown in Figure 9.2, is
obtained when the parameters a and l are given the following values:
1
(c + ),
4
1
l = (c 2 2 )1/2 .
4
a=
(9.22)
(9.23)
In this case, the surface of the airfoil satisfies the parametric equations
c
cos ,
2
y = sin .
2
x=
(9.24)
(9.25)
262
V
d
F
c
M
Figure 9.2
A cylindrical airfoil of elliptical cross-section.
given by
2
dF dF d
V a 2 e i
i
=
= Ve +i
.
dz
d dz
2
2
2 l2
(9.26)
(c + )
= i V sin( + ) +
.
dz
(c + ) ( cos + i c sin )
(9.27)
A long way from the airfoil, z l 2 /z, so that Equation (9.26) reduces to
V (l 2 e i a 2 ei )
dF
V ei + i
+
.
dz
2 z
z2
(9.28)
A comparison with Equation (9.6) reveals that the circulation of air around the airfoil
takes the same value, , in both the complex - and z-planes. In other words, the
conformal transformation (9.21) does not modify the circulation. The transformation
also does not modify the external wind speed or direction [because, from (9.16) and
(9.28), dF/d = dF/dz = V e i at very large || and |z|]. On the other hand, it is
clear that the constant B, which takes the value zero in the complex -plane, takes
the value
B = V (l 2 e i a 2 ei )
(9.29)
Incompressible Aerodynamics
263
in the complex z-plane. Hence, Equations (9.10) and (9.13) reveal that
V = V e ,
(9.30)
L = V e ,
M = V 2 (c 2 2 ) sin(2 ),
8
(9.31)
(9.32)
where V is the wind vector, L the lift vector, and M the moment of the lift vector
about the origin. Here,
(9.33)
e = cos e x + sin ey
is a unit vector parallel to the incident wind direction, and
e = e ez = sin e x + cos ey
(9.34)
is a unit vector perpendicular to the wind direction. We conclude that, for a cylindrical airfoil of elliptic cross-section, the lift vector is normal to the wind vector, but the
line of action of the lift intersects the major axis of the airfoil a distance
d=
1
sin
M
= (c )
L cos 4
(9.35)
264
dS
C
V
Wake
Boundary Layer
U
Trailing Edge
Airfoil
Figure 9.3
The boundary layer and wake of a streamlined airfoil. Only shaded regions posses
non-zero vorticity.
Here, v is the wind velocity, the wind vorticity, and dS an outward surface element
(of unit depth in the z-direction) lying on C. We expect the vorticity flux to be
independent of the size and shape of C, otherwise the circulation of the flow, ,
about the airfoil would not have a unique value. In the limit that C becomes very
large, v V, where V is the incident wind velocity. Thus,
d
= V z ,
dt
(9.37)
where V is the wind speed, and z the vortex intensity per unit length in the wake
(at the point where it crosses the curve C). Here, we are assuming that the vorticity
within the wake is convected by the flow, giving rise to a net flux of vorticity across C.
Because the wake is essentially an extension of the boundary layer, it is reasonable to
assume that its vortex intensity per unit length is proportional to that in the boundary
layer at the airfoils trailing edge, where the wake and boundary layer intersect. In
other words, z = k U0 , where U0 is the tangential velocity immediately above the
trailing edge of the airfoil, and k is a constant. It follows that
d
= k V U0 .
dt
(9.38)
According to Equation (9.27), the tangential velocity just above the surface of the
Incompressible Aerodynamics
airfoil is
265
c sin i cos
dF
(c 2 sin2 + 2 cos2 )1/2 dz ||=a
(c + )
= V sin( + ) +
.
(c + ) (c 2 sin2 + 2 cos2 )1/2
U() = Re
U0 =
V sin
.
(c + )
(9.39)
(9.40)
(9.41)
where = /[ V (c + )], t = t/t0 , and t0 = /(k V). Assuming that the circulation
of the flow about the airfoil is initially zero (i.e., = 0 at t = 0), the previous
equation can be solved to give
!
"
= sin 1 exp(t) .
(9.42)
(9.43)
(9.44)
According to Equation (9.40), when the circulation, , takes the value the tangential velocity at the airfoils training edge, U0 , is zero. In other words, the steady-state
circulation set up around the airfoil is such as to render its trailing edge a stagnation point of the flow. This conclusion is known as Zhukovskiis hypothesis, after its
discoverer N. E. Zhukovskii (18471921).
Incidentally, it should be clear, from the previous discussion, that the air circulation about the airfoil is only able to change its value because of the presence of the
boundary layer, and the associated wake that trails from the airfoils trailing edge.
This follows because the flow is irrotational everywhere except within the boundary
layer and the wake. Moreover, as we have seen, a change in circulation is necessarily
associated with a net vorticity flux away from the airfoil, and such a flux cannot be
carried by an irrotational wind. Thus, in the absence of the boundary layer and the
wake, the air circulation about the airfoil would be constrained to remain zero (assuming that it was initially zero), in accordance with the Kelvin circulation theorem.
(See Section 5.8.) This implies, from Equation (9.31), that zero lift would act on
the airfoil, irrespective of its shape, and irrespective of the incident wind speed or
266
y/c
2
2
0
x/c
Figure 9.4
Streamlines around a slender cylindrical airfoil of elliptic cross-section situated in
a uniform, high Reynolds number wind. The parameters for this calculation are
/c = 0.1, = /12, and = 0.
direction. In other words, flight would be impossible. Fortunately, as long as the tangential air velocity at the trailing edge of the airfoil is non-zero, the wake that trails
behind the airfoil carries a net flux of z-directed vorticity, which causes the airfoil
circulation to evolve in time. This process continues until the circulation becomes
such that the tangential velocity at the airfoils trailing edge is zero: that is, such
that the trailing edge is a stagnation point. Thereafter, the circulation remains constant (assuming that the wind speed and direction remain constant). Figures 9.4 and
9.5 show the streamlines of the flow around a slender cylindrical airfoil of elliptic
cross-section situated in a uniform, high Reynolds number wind whose direction of
incidence is slightly inclined to the airfoils major axis. In the first figure, the air
circulation about the airfoil is zero. In the second figure, the circulation is such as to
make the trailing edge of the airfoil a stagnation point.
According to Equations (9.31) and (9.44), when the air circulation about the airfoil has attained its steady-state value, , the lift acting on the airfoil becomes
L = V = V 2 (c + ) sin .
(9.45)
The lift is positive (i.e., upward) when > 0 (i.e., when the wind is incident on the
airfoils bottom surface), negative (i.e., downward) when < 0 (i.e., when the wind
is incident on the airfoils top surface), and zero when = 0 (i.e., when the wind is
incident parallel to airfoils major axis). Incidentally, the angle is conventionally
Incompressible Aerodynamics
267
y/c
2
2
0
x/c
Figure 9.5
Streamlines around a slender cylindrical airfoil of elliptic cross-section situated in
a uniform, high Reynolds number wind. The parameters for this calculation are
/c = 0.1, = /12, and = .
termed the angle of attack. Finally, from Equations (9.35) and (9.43), the focus of
the airfoil is located a distance
d=
sin 1
1
(c )
= (c )
4
4
(9.46)
in front of the centroid of its cross-section. In the limit that the airfoil becomes very
thin (i.e., c), this distance asymptotes to c/4. Thus, we conclude that the focus
of a thin airfoil, which is defined as the point of action of the lift, is located one
quarter of the way along the airfoil from its leading edge.
The previous analysis is premised on the assumption that there is no appreciable
separation of the boundary layer from the back of the airfoil, which implies the neglect of form drag. We can check that this assumption is reasonable by calculating
the approximate locations of the boundary layer separation points using the analysis
of Section 8.10. Let s represent arc-length along the surface of the airfoil, measured
from the front stagnation point. Assuming that, in accordance with Zhukovskiis
hypothesis, the circulation is such that = , this stagnation point is located at
= 0 , where 0 = 2 . [See Equation (9.49).] It follows that
ds = (dx 2 + dy 2 )1/2 =
1
h() d,
2
(9.47)
268
()
180
10
()
20
Figure 9.6
The angular locations of the boundary layer separation points, , calculated as a
function of the angle of attack, , for a cylindrical airfoil of elliptic cross-section
situated in a uniform, high Reynolds number wind. The solid, dashed, short-dash
dotted, and long-dashdotted curves correspond to airfoils of ellipticity /c = 1.0,
0.5, 0.25, and 0.125, respectively. The trailing edge of the airfoil is located at =
180 .
where
h() = (c 2 sin2 + 2 cos2 )1/2 .
(9.48)
Moreover, from Equations (9.39) and (9.44), the tangential air speed just above the
surface of the airfoil can be written
U() = V (c + )
f ()
,
h()
(9.49)
with
f () = sin( + ) + sin .
(9.50)
f
d
ln
= g(),
d
h
where
g() =
.
f ()
h 2 ()
(9.51)
(9.52)
Incompressible Aerodynamics
269
y
vz (y = 0+)
z
vz (y = 0)
dz
Figure 9.7
Side view of a vortex sheet.
According to the analysis of Section 8.10, the separation points are located at = ,
where ( ) = 0.1567, and
h 4 () g() f 5 ( )
d .
() = 0.47
(9.53)
4
f 5 ()
0 h ( )
Here, > + > 0 and 0 > > . Moreover, the x- and y-coordinates of the separation points are x = (c/2) cos and y = (/2) sin , respectively. Figure 9.6
shows the angular locations of the separation points, calculated as a function of the
angle of attack, for cylindrical airfoils of various dierent ellipticity, /c. (Note that
has been re-expressed as an angle in the range 0 to 2.) It can be seen that for a
blu airfoil (e.g., /c = 1) the angular distance between the separation points is large,
indicating the presence of a wide wake, and a high associated form drag (because the
magnitude of form drag is roughly proportional to the width of the wake). On the
other hand, for a slender airfoil (e.g., /c = 0.125) the angular distance between the
separation points is much smaller, indicating the presence of a narrow wake, and a
low associated form drag. However, in the latter case, as the angle of attack is gradually increased from zero, there is an initial gradual increase in the angular distance
between the separation points, followed by an abrupt, and very large, increase. We
would expect there to be a similar gradual increase in the form drag, followed by an
abrupt, and very large, increase. The value of the angle of attack at which this abrupt
increase occurs is termed the critical angle of attack. We conclude that the previous
analysis, which neglects form drag, is valid only for slender airfoils whose angles of
attack do not exceed the critical value (which is generally only a few degrees).
(9.54)
270
where (y) is a Dirac delta function. Here, = x e x is the sheets vortex intensity
per unit length. Let vz (y = 0+ ) and vz (y = 0 ) be the z-component of the fluid velocity
immediately above and below the sheet, respectively. Consider a small rectangular
loop in the y-z plane that straddles the sheet, as shown in the figure. Integration of
= v around the loop (making use of the curl theorem) yields
vz vz (y = 0+ ) vz (y = 0 ) = x .
(9.55)
In other words, a vortex sheet induces a discontinuity in the tangential flow across
the sheet. The previous expression can easily be generalized to give
= n v,
(9.56)
where is the sheets vortex intensity per unit length, n is a unit vector normal to
the sheet, and v is the jump in tangential velocity across the sheet (traveling in the
direction of n). Furthermore, it is reasonable to assume that the previous relation
holds locally for non-planar and non-uniform vortex sheets.
(9.58)
Here, j is the current density, and B the magnetic field-strength. By analogy, given
that vorticity is related to fluid velocity via the familiar relation
= v,
we can write
v(r) =
1
4
(r ) (r r ) 3
d r.
|r r | 3
(9.59)
(9.60)
This expression allows us to determine the flow field induced by a given vorticity
distribution. In particular, for a vortex filament of intensity the previous expression
reduces to
1
(r ) (r r )
v(r) =
dl ,
(9.61)
4
|r r | 3
Incompressible Aerodynamics
271
where dl is an element of length along the filament. Likewise, for a vortex sheet of
intensity per unit length , we obtain
1
(r ) (r r )
v(r) =
dS ,
(9.62)
4
|r r | 3
where dS is an element of area of the sheet.
272
V
x
airfoil
c(z)
z = b/2
z=0
z = b/2
Figure 9.8
Top view of a three-dimensional airfoil of finite size.
y
b
Y
Y
+
(z)
Figure 9.9
Back view of a three-dimensional airfoil of finite size, indication the pressure variation over its surface.
Incompressible Aerodynamics
273
x
P
Q
Figure 9.10
Top view of the airflow over the top (left) and bottom (right) surfaces of a threedimensional airfoil of finite size.
Bernoullis theorem, depends on the air speed) must be continuous at Q. Thus, we
conclude that there is a discontinuity in the direction of the air emitted by the trailing edge of a wing. This implies that the interface, , between the two streams of
air that travel over the upper and lower surfaces of the wing is a vortex sheet. (See
Section 9.5.) Of course, this vortex sheet constitutes the wake that trails behind the
airfoil. Moreover, we would generally expect the wake to be convected by the incident wind. It follows that the vorticity per unit length in the wake can be written
= I(z) e ,
(9.63)
where I(z) = vz , and vz is tangential velocity discontinuity across the wake. [See
Equation (9.56).]
As we saw previously, the boundary layer that covers the airfoil is such that the
tangential velocity U just outside the layer is sharply reduced to zero at the airfoil surface. Actually, the nature of the substance enclosed by the surface is irrelevant to our
argument, and nothing is changed in our analysis if we suppose that this region contains air at rest. Thus, we can replace the airfoil by air at rest, and the boundary layer
by a vortex sheet, S , with a vortex intensity per unit length S that is determined by
the velocity discontinuity U between the air just outside the boundary layer and that
at rest in the region where the airfoil was previously located. In fact, Equation (9.56)
yields
S = n U,
(9.64)
274
x
z
V
Figure 9.11
Vortex structure around a wing.
(9.65)
where V is the external wind velocity, v the velocity field induced by the free vortex
filaments that constitute , and vS the velocity field induced by the bound filaments
that constitute S .
Consider some point P that lies on S . Let P+ and P be two neighboring points
that are equidistant from P, where P+ lies just outside S , and P lies just inside S ,
and the line P P+ is normal to S . We can write
v(P+ ) = V + v (P+ ) + vS (P+ ),
(9.66)
v(P ) = V + v (P ) + vS (P ).
(9.67)
However, v(P+ ) = U(P), where U(P) is the tangential air velocity just above point
P on the airfoil surface, and v(P ) = 0, as the air within S is stationary. Moreover,
v (P+ ) = v (P ) = v (P), because we expect v to be continuous across S . On the
other hand, we expect the tangential component of vS to be discontinuous across S .
Let us define
1
vS (P) = [vS (P+ ) + vS (P )] .
(9.68)
2
This quantity can be identified as the velocity induced at point P by the bound vortices on S , excluding the contribution from the local bound vortex at P (because this
vortex induces equal and opposite velocities at P+ and P ). Finally, taking half the
sum of Equations (9.66) and (9.67), we obtain
1
U(P) = V + v (P) + vS (P).
2
(9.69)
Incompressible Aerodynamics
275
where the integral is taken over the surface of the airfoil, S . Here, n is an outward
unit normal vector on S , dS is an element of S , and p is the air pressure. From
Bernoullis theorem (in an irrotational fluid), we can write p = p0 (1/2) v 2, where
p0 is a constant pressure. Because a constant pressure exerts no net force on a closed
surface, we get
1
(9.71)
A = U 2 n dS ,
2
S
where U is the tangential air velocity just above the surface of the airfoil. Now,
U (n U) = U 2 n (n U) U = U 2 n,
because n U = 0 on the surface. Hence,
1
A = U (n U) dS .
2
S
(9.72)
(9.73)
Making use of Equations (9.64) and (9.69), the previous expression can be written
A = (V + v + vS ) S dS = L + D + F,
(9.74)
S
where
L = V
D=
S dS ,
(9.75)
v S dS ,
(9.76)
vS S dS .
(9.77)
F=
Here, V, v , and vS are the incident wind velocity, the velocity induced by the free
vortices in the wake, and the velocity induced by the bound vortices covering the
surface of the airfoil, respectively. The forces L and D are called the lift and the
induced drag, respectively. (Note, that L now represents a net force, rather than a
force per unit length.) We shall presently demonstrate that the force F is negligible.
Let us assume that
(9.78)
S z ez :
that is, that the bound vortices covering the surface of the airfoil run parallel to the
z-axis. This assumption is exactly correct for an airfoil of infinite wingspan and
276
where the closed curve C is the intersection of the airfoil surface with the plane z = z,
and (z) is the air circulation about the airfoil in this plane. Thus, it follows from
Equation (9.75) that
b/2
(z) dz e .
(9.80)
L = V
b/2
This expression is the generalization of Equation (9.31) for a three-dimensional airfoil of finite size. As before, the lift is at right-angles to the incident wind direction.
Let us make the further assumptionknown as the lifting line approximation
(because the lifting action of the wing is eectively concentrated onto a line)that
v = w(z) e
(9.81)
throughout S , where w(z) e is the induced velocity due to the free vortices in ,
evaluated at the trailing edge of the airfoil. Here, the velocity w(z) is called the
downwash velocity. It follows from Equation (9.76) that
D=
b/2
b/2
w(z) (z) dz e .
(9.82)
Note that the induced drag is parallel to the incident wind direction. The origin of
induced drag is as follows. It takes energy to constantly resupply free vortices to
the wake, as they are swept downstream by the wind (note that a vortex filament
possesses energy by virtue of the kinetic energy of its induced flow pattern), and this
energy is supplied by the work done in opposing the induced drag. The drag acting
on a well-designed airfoil (i.e., an airfoil with an aerodynamic shape that minimizes
form drag) situated in a high Reynolds number wind (which implies that the friction
drag is negligible) is generally dominated by induced drag.
According to Equations (9.62) and (9.77), the force F is written
[S (r ) (r r )] S (r)
dS dS .
(9.83)
F=
4 S S
|r r | 3
We can interchange primed and unprimed variables without changing the value of
the integral. Hence,
Incompressible Aerodynamics
277
r r
Figure 9.12
Semi-infinite vortex filament.
Taking the half the sum of the previous two equations, we obtain
dS dS .
(9.86)
F=
8 S S
|r r | 3
But, the assumption (9.78) implies that S (r ) S (r) 0. Hence, F is negligible,
as was previously stated.
Consider a closed surface covering the small section of the airfoil lying between
the parallel planes z = z and z = z + dz. The flux of vorticity into the surface due
to bound vortices at z is (z). The flux of vorticity out of the surface due to bound
vortices at z + dz is (z + dz). Finally, the flux of vorticity out of the surface due to the
free vortices in the part of the wake lying between z and z + dz is I(z) dz. However,
the net flux of vorticity out of a closed surface is zero, because vorticity is divergence
free. Hence,
(z) = (z + dz) + I(z) dz,
(9.87)
which implies that
d
.
(9.88)
dz
Finally, consider a semi-infinite straight vortex filament of vortex intensity =
e x that terminates at the origin, O, as shown in Figure 9.12. Let us calculate the
flow velocity induced by this filament at the point P = (0, 0, z). From the diagram
l = z tan , dl = z sec2 d, |r r | = z sec , and |r r | = z ey . Hence, from
Equation (9.61), the induced velocity at P is v = vy ey , where
I(z) =
vy =
4 z
/2
cos d =
0
.
4 z
(9.89)
278
This result allows us to calculate the downwash velocity, w(z) = vy (z), induced at
the trailing edge of the airfoil by the semi-infinite free vortices in the wake. The
vortex intensity in the small section of the wake lying between z and z + dz is I(z) dz,
so we obtain
b/2
d(z )
I(z ) dz
1
1
w(z) =
=
,
(9.90)
4 b/2 z z
4
z z
where use has been made of Equation (9.88).
(9.91)
(9.92)
(9.93)
(9.94)
0 V c0 .
(9.95)
where
Here, the angle of attack, , is assumed to be small. From Equations (9.90) and
(9.94), the downwash velocity in the region |z| < b/2 is given by
0
cos d
d
0
cos
=
w() =
1+
.
(9.96)
2 b 0 cos cos 2 b
0 cos cos
The integrand appearing in the integral
0
d
cos cos
(9.97)
Incompressible Aerodynamics
279
is singular when = . However, we can still obtain a finite value for the integral
by taking its Cauchy principal part: that is,
lim
0
d
+
cos cos
+
d
.
cos cos
(9.98)
Physically, this is equivalent to omitting the contribution of the local free vortex at
a given point on the airfoils trailing edge to the downwash velocity induced at that
point, which is reasonable because a vortex induces zero velocity at its center. Hence,
we obtain
sin (1/2) ( + )
d
1
=
lim
ln
sin
0
sin (1/2) ( ) 0
0 cos cos
+
sin (1/2) ( + )
1
ln
+
sin
sin (1/2) ( )
sin( /2)
1
ln
= 0,
(9.99)
= lim
0 sin
sin( + /2)
which implies that
w() =
0
.
2b
0
d
0
||
w() =
1
=
.
1
2b
0 cos +
2b
2 1
(9.100)
(9.101)
280
1.2
0.8
w/(0/2 b)
0.4
0
0.4
0.8
1.2
1.6
2
2
0
z/b)
Figure 9.13
Downwash velocity profile induced at the trailing edge by an ellipsoidal airfoil.
b/2
b/2
(z) dz =
0 b
2
sin2 d =
0
0 b.
4
(9.103)
Hence, Equation (9.80), (9.82), and (9.100) yield the following expression for the lift
and induced drag acting on an ellipsoidal airfoil,
V b 0 ,
4
D = 02 .
8
L=
(9.104)
(9.105)
b c0 .
4
(9.106)
b2 4 b
=
.
S
c0
(9.107)
Incompressible Aerodynamics
281
(9.108)
2
L0 2 ,
A
(9.109)
L0 = V 2 S .
(9.110)
D=
where
(9.111)
(9.112)
Let us assume that the angles and are both small. According to Equations (9.108)
and (9.109), L L0 and D (2 L0 /A) 2 . Thus, L O() and D O( 2 ).
Moreover, it is clear from Equations (9.111) and (9.112) that T O( 2 ) and W
O(). Thus, to lowest order in , Equation (9.111) yields
W
,
L0
(9.113)
2 W
T
.
W A L0
(9.114)
whereas Equation (9.112) gives
Expression (9.113) relates the angle of attack to the ratio of the aircrafts weight to
its (theoretical) maximum lift (at a given airspeed). Expression (9.114) relates the
aircrafts angle of controlled (i.e., at constant airspeed) ascent to the thrust developed
282
V
x
D
W
Figure 9.14
Side view of a fixed wing aircraft in flight.
by its engine. An unpowered aircraft, such as a glider, has zero thrust. For such
an aircraft, Equation (9.114) reveals that the angle of controlled decentwhich is
usually termed the glide angletakes the value
g=
2 W
.
A L0
(9.115)
At fixed airspeed, V, and wing surface area, S , (which implies that L0 is fixed) this
angle can be minimized by making the wing aspect-ratio, A, as large as possible.
This result explains accounts for the fact that gliders (and albatrosses) have long thin
wings, rather than short stubby ones. For a powered aircraft, the critical thrust to
weight ratio required to maintain level flight (i.e., = 0) is
T
= g.
(9.116)
W
Hence, this ratio is minimized by minimizing the glide angle, which explains why
long-haul aircraft, which generally need to minimize fuel consumption, tend to have
long thin wings. Finally, as we saw in Section 9.4, if the angle of attack exceeds
some (generally small) critical value c then boundary layer separation occurs on
the back sides of the wings, giving rise to a greatly increased level of drag acting on
the aircraft. In aerodynamics, this phenomenon is called a stall. As is clear from
Equations (9.110) and (9.113), the requirement < c is equivalent to
1/2
W
V > Vs =
.
(9.117)
S c
Incompressible Aerodynamics
283
In other words, a stall can be avoided by keeping the airspeed above the critical
value V s , which is known as the stall speed. Note that the stall speed decreases with
decreasing altitude, as the air becomes denser.
9.11 Exercises
9.1 Consider the integral
In () =
0
cos(n ) d
,
cos cos
sin(n )
.
sin
9.2 Suppose that an airfoil of negligible thickness, and wingspan b, has a width
whose z variation is expressed parametrically as
c() =
c sin( ),
=1,3,5,
for 0 , where
b
cos .
2
Show that the air circulation about the airfoil takes the form
() =
sin( ),
z=
=1,3,5,
284
V b 1 ,
4
D=
2,
8 =1,3,5,
L=
c 2
2
D
,
= 1 +
L
A
c2
=3,5,7, 1
where A is the aspect ratio. Hence, deduce that the airfoil shape (in the x-y)
plane that minimizes this ratio (at fixed aspect ratio) is an ellipse (i.e., such
that c = 0 for > 1).
9.3 Consider a plane that flies with a constant angle of attack, and whose thrust is
adjusted such that it cancels the induced drag. The plane is eectively subject
to two forces. First, its weight, W = W ey , and second its lift L = k v vy e x +
k v v x ey . Here, x and y are horizontal and vertical coordinates, respectively, v
is the planes instantaneous velocity, and k is a positive constant. Note that the
lift is directed at right angles to the planes instantaneous direction of motion,
and has a magnitude proportional to the square of its airspeed. Demonstrate
that the planes equations of motion can be written
v vy
dv x
=
,
dt
h
dvy v v x
=
g,
dt
h
where h = k g/W is a positive constant with the dimensions of length. Show
that
1 2
v + g y = E,
2
where E is a constant. Suppose that v x = g h (1 + u) and vy = g h w, where
|u|, |w| 1. Demonstrate that, to first order in perturbed quantities,
g
du
w,
dt
h
dw
g
2
u.
dt
h
Incompressible Aerodynamics
285
Hence, deduce that if the plane is flying horizontally at some speed v0 , and
is subject to a small perturbation,
then its altitude oscillates sinusoidally at
286
10
Incompressible Viscous Flow
10.1 Introduction
This chapter investigates incompressible flow in which viscosity plays a significant
role throughout the bulk of the fluid. Such flow generally takes place at relatively
low Reynolds number. From Section 1.14, the equations governing incompressible
viscous fluid motion can be written
v = 0,
Dv
= P + 2 v,
Dt
(10.1)
(10.2)
(10.3)
which is a combination of the actual fluid pressure, p, and the gravitational potential
energy per unit volume, , is known as the eective pressure. Here, is the fluid
mass density, the fluid viscosity, and the gravitational potential. More information on this topic can be found in Batchelor 2000.
288
y=d
vx(y)
y=0
Figure 10.1
Viscous flow between parallel plates.
dierence to the final result. Suppose that the fluid velocity profile between the plates
takes the form
(10.5)
v = v x (y) e x .
From Section 1.18, this profile automatically satisfies the incompressibility constraint v = 0, and is also such that Dv/Dt = 0. Hence, Equation (10.2) reduces
to
P
,
(10.6)
2v =
dy
If the two plates are stationary then the solution that satisfies the no slip constraint
(see Section 8.2), v x (0) = v x (d) = 0, at each plate is
v x (y) =
G
y (d y).
2
(10.8)
Thus, steady, two-dimensional, viscous flow between two stationary parallel plates
is associated with a parabolic velocity profile that is symmetric about the midplane,
y = d/2. The net volume flux (per unit width in the z-direction) of fluid between the
plates is
d
Gd3
.
(10.9)
Q=
v x dy =
12
0
Note that this flux is directly proportional to the eective pressure gradient, inversely
proportional to the fluid viscosity, and increases as the cube of the distance between
the plates.
Suppose that the upper plate is stationary, but that the lower plate is moving in
the x-direction at the constant speed U. In this case, the no slip boundary condition
at the lower plate becomes v x (0) = U, and the modified solution to Equation (10.7)
is
G
dy
v x (y) =
y (d y) + U
.
(10.10)
2
d
289
Figure 10.2
Viscous flow down an inclined plane.
Hence, the modified velocity profile is a combination of parabolic and linear profiles.
This type of flow is known as Couette flow, in honor of Maurice Couette (1858
1943). The net volume flux (per unit width) of fluid between the plates becomes
Q=
Gd3 1
+ U d.
12 2
(10.11)
(10.12)
where g is the acceleration due to gravity. In this case, there is no gradient in the
actual pressure in the x-direction, and the flow down the plane is driven entirely by
gravity. As before, we can write
v = v x (y) e x ,
(10.13)
290
dy
(10.14)
G = g sin .
(10.15)
where
Application of the no slip condition at the surface of the plane, y = 0, yields the
standard boundary condition v x (0) = 0. However, the appropriate physical constraint
at the
fluid/air interface, y = h, is that the normal viscous stress there be zero (i.e.,
xy y=h = 0), because there is nothing above the interface that can exchange momentum with the fluid (assuming that the finite inertia and viscosity of air are both
negligible.) Hence, from Section 1.18, we get the boundary condition
dv x
= 0.
(10.16)
dy y=h
The solution to Equation (10.14) that satisfies the boundary conditions is
v x (y) =
G
y (2 h y).
2
(10.17)
Thus, the profile is again parabolic. In fact, it is the same as the lower half of the
profile obtained when fluid flows between two (stationary) parallel plates situated a
perpendicular distance 2 h apart.
The net volume flux (per unit width in the z-direction) of fluid down the plane is
Q=
v x dy =
0
G h 3 g sin h 3
=
,
3
3
(10.18)
where use has been made of Equation (10.15). Here, = / is the kinematic
viscosity of the fluid. Thus, given the rate Q that fluid is poured down the plane, the
depth of the layer covering the plane becomes
3Q
h=
g sin
1/3
.
(10.19)
Suppose that the rate at which fluid is poured down the plane is suddenly increased slightly from Q to Q + Q. We would expect an associated change in depth
of the layer covering the plane from h to h + h, where
dh
h =
Q.
(10.20)
dQ
Let the interface between the layers of dierent depth propagate in the x-direction at
the constant velocity V. In a frame of reference that co-moves with this interface, the
volume fluxes (per unit width) immediately to the right and to the left of the interface
291
1/3
dQ G h 2
9 Q 2 g sin
=
=
.
dh
(10.21)
(10.22)
As can easily be verified, this velocity is twice the maximum fluid velocity in the
layer.
(10.23)
(10.24)
the time independent velocity profile driven by this gradient. It follows from Section 1.19 that v = 0 and Dv/Dt = 0. Hence, Equation (10.2) reduces to
2v =
G
ez .
1 d dvz
G
r
= ,
r dr dr
(10.25)
(10.26)
where use has been made of Equation (1.155). The most general solution of the
previous equation is
G 2
r + A ln r + B,
(10.27)
vz (r) =
4
where A and B are arbitrary constants. The physical constraints are that the flow
velocity is non-singular at the center of the pipe (which implies that A = 0), and is
292
zero at the edge of the pipe [i.e., vz (a) = 0], in accordance with the no slip condition.
Thus, we obtain
G
vz (r) =
(a 2 r 2 ).
(10.28)
4
The volume flux of fluid down the pipe is
a
G a4
Q=
.
(10.29)
2 r vz dr =
8
0
According to the previous analysis, the quantity Q/a 4 should be directly proportional
to the eective pressure gradient along the pipe. The accuracy with which experimental observations show that this is indeed the case (at relatively low Reynolds
number) is strong evidence in favor of the assumptions that there is no slip at the
pipe walls, and that the flow is non-turbulent. In fact, the result (10.29), which is
known as Poiseuilles law, is valid experimentally provided the Reynolds number of
the flow, Re = U a/, remains less than about 6.5 103 . Here, U = Q/( a 2) is the
mean flow speed. On the other hand, if the Reynolds number exceeds the critical
value 6.5 103 then the flow in the pipe becomes turbulent, and Poiseuilles law
breaks down.
(10.31)
(r2 ) = 2 .
(10.32)
It again follows from Section 1.19 that v = 0 and Dv/Dt = 0. Hence, Equation (10.2) reduces to
P
.
(10.33)
2v =
293
Assuming that P = 0 within the fluid, because any flow is driven by the angular
rotation of the two shells, rather than by pressure gradients or gravity, and again
making use of the results quoted in Section 1.19, the previous expression yields
1 d v
v
(10.34)
r
2 = 0,
r dr dr
r
or
1 d 3 d
r
= 0.
dr
r 2 dr
1 1 2 2 r12 1 r22
(r) = 2 2
+
r r1 r22
r12 r22
(10.35)
(10.36)
It can be seen that this angular velocity profile is a combination of the solid body
rotation profile = constant, and the irrotational rotation profile r 2 .
From Section 1.19, the only non-zero component of the viscous stress tensor
within the fluid is
d
d $ v %
.
(10.37)
= r
r = r
dr r
dr
Thus, the viscous torque (acting in the -direction) per unit height (in the z-direction)
exerted on the inner cylinder is
1 2
2
.
1 = 2 r1 r (r1 ) = 4 2
(10.38)
r1 r22
Likewise, the torque per unit height exerted on the outer cylinder is
1 2
2
.
2 = 2 r2 r (r2 ) = 4 2
r1 r22
(10.39)
As expected, these two torques are equal and opposite, and act to make the two
cylinders rotate at the same angular velocity (in which case, the fluid between them
rotates as a solid body).
(10.40)
(10.41)
294
As we saw in Sections 10.2 and 10.4, for the case of flow along a straight channel of
uniform cross-section, v and (v ) v are both identically zero, and the governing
equations consequently reduce to the simple relation
2v =
P
.
(10.42)
Suppose, however, that the cross-section of the channel varies along its length. As
we shall demonstrate, provided this variation is suciently slow, the flow is still
approximately described by the previous relation.
Consider steady, two-dimensional, viscous flow, that is predominately in the xdirection, between two plates that are predominately parallel to the y-z plane. Let the
spacing between the plates, d(x), vary on some length scale l
d. Suppose that
P = G(x) e x ,
(10.43)
where G(x) also varies on the same lengthscale. Assuming that /x O(1/l) and
/y O(1/d), it follows from Equation (10.40) that
vy
d
.
(10.44)
O
vx
l
Hence,
v x2
[(v ) v] x O
,
l
2
2 v x
d
2
( v) x =
1
+
O
l
y 2
The x-component of Equation (10.41) reduces to
d
2 v x
vx d 2
G
+
O
1
+
O
= .
2
l
l
(10.45)
(10.46)
(10.47)
Thus, if
d
1,
l
(10.48)
vx d 2
1
l
(10.49)
in other words, if the channel is suciently narrow, and its cross-section varies
suciently slowly along its lengththen Equation (10.47) can be approximated as
2vx
G
.
y 2
(10.50)
This, of course, is the same as the equation governing steady, two-dimensional, viscous flow between exactly parallel plates. (See Section 10.2.) Assuming that the
295
plates are located at y = 0 and y = d(x), and making use of the analysis of Section 10.2, the appropriate solution to the previous equation is
v x (x, y) =
G(x)
y [d(x) y].
2
The volume flux (per unit width) of fluid between the plates is thus
d
G(x) d 3 (x)
Q=
.
v x dy =
12
0
(10.51)
(10.52)
However, for steady incompressible flow, this flux must be independent of x, which
implies that
dP
= 12 Q d 3(x).
(10.53)
G(x) =
dx
Suppose that a constant dierence in eective pressure, P, is established between
the fixed points x = x1 and x = x2 , where x2 x1 = l. Integration of the previous
equation between these two points yields
P = 12 Q l d 3 ,
(10.54)
x2
where = x ( ) dx/l. Hence, the volume flux (per unit width) of fluid between
1
the plates that is driven by the eective pressure dierence becomes
Q=
1
P
.
l 12 d 3
(10.55)
1
dP P
=
,
dx
l d 3 (x) d 3
(10.56)
which allows us to determine the velocity profile at that point from Equation (10.51).
Thus, given the average eective pressure gradient, P/l, and the variable separation,
d(x), we can fully specify the flow between the plates.
Using analogous arguments to those employed previously, but adapting the analysis of Section 10.4, rather than that of Section 10.1, we can easily show that steady
viscous flow down a straight pipe of circular cross-section, whose radius a varies
slowly with distance, z, along the pipe, is characterized by
vz (r, z) =
G(z) 2
a (z) r 2 ,
4
1
P
,
4
l a (z) a 4
P
.
Q=
l 8 a 4
G(z) =
(10.57)
(10.58)
(10.59)
Here, Q is the volume flux of fluid down the pipe, P = P(z2 ) P(z1 ), l = z2 z1 ,
296
y
d1
d2
l
Figure 10.3
The lubrication layer between two planes in relative motion.
z2
and = z ( ) dz/l. The approximations used to derive the previous results are
1
valid provided
a
1,
l
(10.60)
vz a 2
1.
l
(10.61)
(10.62)
297
where
d1 d2
.
l
As discussed in the previous section, provided that
=
(10.63)
d
1,
l
(10.64)
U d2
1,
l
(10.65)
the cross-section of the channel between the two bodies is suciently slowly varying
in the x-direction that the channel can be treated as eectively uniform at each point
along its length. Thus, it follows from Equation (10.10) that the velocity profile
within the channel takes the form
"
d(x) y
G(x) !
v x (x, y) =
y d(x) y + U
,
(10.66)
2
d
where
dp
(10.67)
dx
is the pressure gradient. Here, we are neglecting gravitational forces with respect to
both pressure and viscous forces. The volume flux per unit width (in the z-direction)
of fluid along the channel is thus
G(x) =
Q=
v x dy =
G(x) d 3 (x) 1
+ U d(x).
12
2
1
1
6 1
1
U
p(x) p0 =
Q 2 2 ,
d d1
d
d1
(10.68)
(10.69)
(10.70)
where p0 = p(0). Assuming that the sliding block is completely immersed in fluid of
uniform ambient pressure p0 , we would expect the pressures at the two ends of the
lubricating layer to both equal p0 , which implies that p(l) = p0 . It follows from the
previous equation that
d1 d2
,
(10.71)
Q=U
d1 + d2
and
p(x) p0 =
d 2 (x) (d1 + d2 )
(10.72)
298
Note that if d1 > d2 then the pressure increment p(x) p0 is positive throughout the
layer, and vice versa. In other words, a lubricating layer sandwiched between two
solid bodies in relative motion only generates a positive pressure, that is capable of
supporting a normal load, when the motion is such as to drag (by means of viscous
stresses) fluid from the wider to the narrower end of the layer. The pressure increment
has a single maximum in the layer, and its value at this point is of order l U/d 2 ,
assuming that (d1 d2 )/d1 is of order unity. This suggests that very large pressures
can be set up inside a thin lubricating layer.
The net normal force (per unit width in the z-direction) acting on the lower plane
is
l
d1
d1 d2
6U
2
.
(10.73)
fy = [p(x) p1 ] dx = 2 ln
d2
d1 + d2
0
Moreover, the net tangential force (per unit width) acting on the lower plane is
fx =
v x
y
y=0
dx =
2U
d1 d2
d1
3
.
2 ln
d2
d1 + d2
(10.74)
1+k
U l2 3
= fx =
ln
2k ,
1k
d02 2 k2
1+k
U l 1
2 ln
3k ,
fy = fy =
d0 k
1k
f x
(10.75)
(10.76)
act on the upper plane. Here, d0 = (d1 + d2 )/2 is the mean channel width, and
k = (d1 d2 )/(d1 + d2 ). It can be seen that if 0 < d2 < d1 then 0 < k < 1. The
eective coecient of friction, C f , between the two sliding bodies is conventionally
defined as the ratio of the tangential to the normal force that they exert on one another.
Hence,
f 4 d0
fx
H(k),
(10.77)
= x =
Cf =
fy
fy
3 l
where
1+k
1+k
3k
H(k) = k ln
ln
2k .
1k
2
1k
(10.78)
The function H(k) is a monotonically decreasing function of k in the range 0 < k < 1.
In fact, H(k 0) 3/(4 k), whereas H(k 1) 1. Thus, if k O(1) [i.e., if
(d1 d2 )/d1 O(1)] then C f O(d0 /l) 1. In other words, the eective coecient
of friction between two solid bodies in relative motion that are separated by a thin
fluid layer is independent of the fluid viscosity, and much less than unity. This result
is significant because the coecient of friction between two solid bodies in relative
motion that are in direct contact with one another is typical of order unity. Hence,
the presence of a thin lubricating layer does indeed lead to a large reduction in the
frictional drag acting between the bodies.
299
(10.79)
0 = P + 2 v.
(10.80)
(10.81)
where = v, and use has been made of Equation (A.177). Taking the curl of
this expression, we obtain
2 = 0,
(10.82)
which is the governing equation for Stokes flow. Here, use has been made of Equations (A.173), (A.176), and (A.177).
(10.84)
where (r, ) is the Stokes stream function (i.e., v = 0). It follows that
1
,
sin
1
v (r, ) =
.
r sin r
vr (r, ) =
r2
(10.85)
(10.86)
300
L()
1 (r v ) 1 vr
=
,
r r
r
r sin
(10.87)
sin 1
2
.
+ 2
2
r
r sin
(10.88)
(10.89)
(10.90)
Thus, by analogy with Equations (10.84) and (10.89), and making use of Equations (A.173) and (A.177), we obtain
( ) = 2 = L 2 () .
(10.91)
(10.92)
which is the governing equation for axisymmetric Stokes flow. In addition, Equations (10.81) and (10.90) yield
P = [L()].
(10.93)
is much less than unity, the flow around the sphere is an example of axisymmetric
Stokes flow. Let us transform to a frame of reference in which the sphere is stationary,
and centered at the origin. Adopting the standard spherical coordinates r, , , the
surface of the sphere corresponds to r = a, and the surrounding fluid occupies the
region r > a. By symmetry, the flow field outside the sphere is axisymmetric (i.e.,
301
/ = 0), and has no toroidal component (i.e., v = 0). The physical boundary
conditions at the surface of the sphere are
vr (a, ) = 0,
(10.95)
v (a, ) = 0 :
(10.96)
that is, the normal and tangential fluid velocities are both zero at the surface. A long
way from the sphere, we expect the fluid velocity to asymptote to v = V ez . In other
words,
vr (r , ) V cos ,
(10.97)
v (r , ) V sin .
(10.98)
Let us write
v = ,
(10.99)
where (r, ) is the Stokes stream function. As we saw in the previous section, axisymmetric Stokes flow is characterized by
L 2 () = 0.
(10.100)
(10.104)
In this case,
2 cos f (r)
,
r2
sin d f
,
v (r, ) =
r dr
and Equations (10.100)(10.103) reduce to
2
2
d
2
f = 0,
dr 2 r 2
d f
= 0,
f (a) =
dr r=a
vr (r, ) =
f (r )
1
V r 2.
2
(10.105)
(10.106)
(10.107)
(10.108)
(10.109)
302
3
2
1
z/a
0
1
2
3
3
0
x/a
Figure 10.4
Contours of the stream function in the x-z plane for Stokes flow around a solid
sphere.
Let us try a test solution to Equation (10.107) of the form f (r) = r n . We find
that
[n (n 1) 2] [(n 2) (n 3) 2] = 0,
(10.110)
which implies that n = 1, 1, 2, 4. Hence, the most general solution to Equation (10.107) is
A
f (r) = + B r + C r 2 + D r 4 ,
(10.111)
r
where A, B, C, D are arbitrary constants. However, the boundary condition (10.109)
yields C = (1/2) V and D = 0, whereas the boundary condition (10.108) gives
A = (1/4) V a 3 and B = (3/4) V a. Thus, we conclude that
f (r) =
V (r a)2 (2 r + a)
,
4r
(10.112)
V (r a)2 (2 r + a)
.
4r
3 V a sin
L()
2
sin d 2
(r, ) =
f =
.
=
r sin
r
dr 2 r 2
2r2
(10.113)
(10.114)
303
3
2
1
z/a
0
1
2
3
3
0
x/a
Figure 10.5
Contours of the vorticity, , in the x-z plane for Stokes flow around a solid sphere.
Solid/dashed lines correspond to opposite signs of .
(See Figure 10.5.) Moreover, from Equation (10.81),
P = = ( r sin ).
(10.115)
Hence,
3 V a cos
P
=
,
r
r3
3 V a sin
P
=
,
2r2
(10.116)
(10.117)
which implies that the eective pressure distribution within the fluid is
P(r, ) = p0 +
3 V a cos
,
2r2
(10.118)
3V a
cos .
2r2
(10.119)
From Section 1.20, the radial and tangential components of the force per unit
304
3
2
1
z/a
0
1
2
3
3
0
x/a
Figure 10.6
Contours of the eective pressure, P p0 , in the x-z plane for Stokes flow around a
solid sphere. Solid/dashed lines correspond to opposite signs of P p0 .
vr
fr () = rr (a, ) = p + 2
,
r r=a
1 vr v v
f () = r (a, ) =
+
.
r
r
r r=a
(10.120)
(10.121)
Now, vr (a, ) = v (a, ) = 0. Moreover, because v = 0, it follows from Equation (1.168) that (vr /r)r=a = 0. Finally, Equation (10.87) yields (v /r)r=a =
(a, ). Hence,
fr () = p(a, ) = p0 + g a cos
f () = (a, ) =
3V
cos ,
2a
3V
sin .
2a
(10.122)
(10.123)
3V
ez + (p0 + g a cos ) er .
2a
(10.124)
305
It follows that the net vertical force exerted on the sphere by the fluid is
Fz =
f ez dS
S
3V
4 a 2 + 2 a 2
=
2a
(10.125)
which reduces to
Fz = 6 a V +
4 3
a g.
3
(10.126)
By symmetry, the horizontal components of the net force both average to zero. We
can recognize the second term on the right-hand side of the previous equation as
the buoyancy force due to the weight of the fluid displaced by the sphere. (See
Chapter 2.) Moreover, the first term can be interpreted as the viscous drag acting on
the sphere. Note that this drag acts in the opposite direction to the relative motion of
the sphere with respect to the fluid, and its magnitude is directly proportional to the
relative velocity.
Vertical force balance requires that
Fz = M g,
(10.127)
where M is the spheres mass. In other words, in a steady state, the weight of the
sphere balances the vertical force exerted by the surrounding fluid. If the sphere is
composed of material of mean density then M = (4/3) a 3 . Hence, in the frame
in which the fluid a large distance from the sphere is stationary, the steady vertical
velocity with which the sphere moves through the fluid is
2 a2 g
V=
1
,
(10.128)
9
9 2
For the case of a grain of sand falling through water at 20 C, we have / 2
and = 1.0 106 m2 /s (Batchelor 2000). Hence, Re = (a/6 105 ) 3 , where a
is measured in meters. Thus, expression (10.128), which is strictly speaking only
valid when Re 1, but which turns out to be approximately valid for all Reynolds
numbers less than unity, only holds for sand grains whose radii are less than about
60 microns. Such grains fall through water at approximately 8 103 m/s. For the
case of a droplet of water falling through air at 20 C and atmospheric pressure, we
have / = 780 and = 1.5 105 m2 /s (Batchelor 2000). Hence, Re = (a/4
105 ) 3 , where a is measured in meters. Thus, expression (10.128) only holds for
306
water droplets whose radii are less than about 40 microns. Such droplets fall through
air at approximately 0.2 m/s.
At large values of r/a, Equations (10.105), (10.106), and (10.112) yield
$ a %2
a
3
V cos + O
,
2
r
r
$ a %2
a
3
v (r, ) = V sin V sin + O
.
4
r
r
vr (r, ) = V cos +
(10.130)
(10.131)
It follows that
vr v vr v2 V 2 a
+
,
[ (v ) v]r = vr
r
r
r
r2
and
( 2 v)r
Hence,
2 vr V a
3 .
r 2
r
$r%
[ (v ) v]r V r
Re
a
( 2 v)r
(10.132)
(10.133)
(10.134)
where Re is the Reynolds number of the flow in the immediate vicinity of the sphere.
[See Equation (10.94).] Our analysis is based on the assumption that advective inertia is negligible with respect to viscosity. However, as is clear from the previous
expression for the ratio of inertia to viscosity within the fluid, even if this ratio is
much less than unity close to the spherein other words, if Re 1it inevitably
becomes much greater than unity far from the sphere: that is, for r
a/Re. In other
words, inertia always dominates viscosity, and our Stokes flow solution therefore
breaks down, at suciently large r/a.
$A
r
%
+ Br + C r2 + Dr4 ,
(10.135)
307
2
4
(r, ) = sin + B r + C r + D r ,
(10.136)
r
respectively. Here, A, B, C, et cetera, are arbitrary constants. Likewise, the previous
analysis also allows us to deduce that
$A
%
B
(10.137)
vr (r, ) = 2 cos 3 + + C + D r 2 ,
r
r
$ A
%
B
(10.138)
v (r, ) = sin 3 + + 2 C + 4 D r 2 ,
r
r
6A
(10.139)
r (r, ) = sin 4 + 6 D r ,
r
12 A 6 B
+ 2 + 12 D r
(10.140)
rr (r, ) = p0 + g r cos + cos
r4
r
and
in the region r > a, with analogous expressions in the region r < a. Here, , , p0
are the viscosity, density, and ambient pressure of the fluid surrounding the drop. Let
, , and p0 be the corresponding quantities for the fluid that makes up the drop.
In the region outside the drop, the fluid velocity must asymptote to v = V ez at
large r/a. This implies that C = (1/2) V and D = 0. Furthermore, vr (a, ) = 0
that is, the normal velocity at the drop boundary must be zerootherwise, the drop
would change shape. This constraint yields A/a 3 + B/a + (1/2) V = 0.
Inside the drop, the fluid velocity must remain finite as r 0. This implies
that A = B = 0. Furthermore, we again require that vr (a, ) = 0, which yields
C + D a 2 = 0.
Two additional physical constraints that must be satisfied at the interface between the two fluids are, firstly, continuity of tangential velocitythat is, v (a , ) =
v (a+ , )and, secondly, continuity of tangential stressthat is, r (a , ) = r (a+ , ).
These constraints yield A/a 3 + B/a V = 2 C + 4 D a 2 and 6 A/a 4 = 6 D a,
respectively.
At this stage, we have enough information to determine the values of A, B, C,
and D. In fact, the stream functions outside and inside the drop can be shown to take
the form
$ r %2
1
a
2+ 3 r
,
(10.141)
(r, ) = V a2 sin2
+2
4
+ r
+ a
a
and
(r, ) =
$ r %2
$ r %2
1
V a2 sin2
1
,
4
+ a
a
(10.142)
(10.143)
308
z/a
2
2
0
x/a
Figure 10.7
Contours of the Stokes stream function in the x-z plane for Stokes flow in and around
a fluid sphere. Here, / = 10.
z/a
2
2
0
x/a
Figure 10.8
Contours of the Stokes stream function in the x-z plane for Stokes flow in and around
a fluid sphere. Here, / = 1/10.
309
2
,
a
(10.144)
where is the surface tension of the interface between the two fluids. (See Section 3.3.) Hence, we obtain
2
,
(10.145)
p0 p0 =
a
and
a2 g
+
V=
1
,
(10.146)
3
+ (3/2)
where = / is the kinematic viscosity of the surrounding fluid. The fact that we
have been able to completely satisfy all of the physical constraints at the interface
between the two fluids, as long as the drop moves at the constant vertical velocity V,
proves that our previous assumptions that the interface is spherical, and that the drop
moves vertically through the surrounding fluid at a constant speed without changing
shape, were correct. In the limit,
, in which the drop is much more viscous
than the surrounding fluid, we recover Equation (10.128): that is, the drop acts like a
solid sphere. On the other hand, in the limit , and , which is appropriate
to an air bubble rising through a liquid, we obtain
V=
a2 g
.
3
(10.147)
10.12 Exercises
10.1 Consider viscous fluid flow down a plane that is inclined at an angle to the
horizontal. Let x measure distance along the plane (i.e., along the path of
steepest decent), and let y be a transverse coordinate such that the surface of
the plane corresponds to y = 0, and the free surface of the fluid to y = h.
Show that within the fluid (i.e., 0 y h)
g sin
y (2 h y),
2
p(y) = p0 + g cos (h y),
v x (y) =
a4
(G + g sin ) ,
8
310
10.3 Viscous fluid flows steadily, parallel to the axis, in the annular region between
two coaxial cylinders of radii a and n a, where n > 1. Show that the volume
flux of fluid flow is
G a4 4
(n2 1)2
Q=
n 1
,
8
ln n
where G is the eective pressure gradient, and the viscosity. Find the mean
flow speed.
10.4 Consider viscous flow along a cylindrical pipe of elliptic cross-section. Suppose that the pipe runs parallel to the z-axis, and that its boundary satisfies
x2 y2
+
= 1.
a2 b2
Let
v = vz (x, y) ez.
Demonstrate that
2
2
G
+
vz = ,
2
2
x
y
where G is the eective pressure gradient, and the fluid viscosity. Show
that
G a2 b2 b2 x2 a2 y2
vz (x, y) =
2
a2 + b2
is a solution of this equation that satisfies the no slip condition at the boundary. Demonstrate that the flow rate is
Q=
G a3 b3
.
4 a2 + b2
Finally, show that a pipe with an elliptic cross-section has lower flow rate
than an otherwise similar pipe of circular cross-section that has the same
cross-sectional area.
10.5 Consider a velocity field of the form
v(r) = r 2 (r) sin2 ,
where r, , are spherical coordinates. Demonstrate that this field satisfies
the equations of steady, incompressible, viscous fluid flow (neglecting advective inertia) with uniform pressure (neglecting gravity) provided that
d 4 d
r
= 0.
dr
dr
311
Suppose that a solid sphere of radius a, centered at the origin, is rotating about
the z-axis, at the uniform angular velocity 0 , in a viscous fluid, of viscosity
, that is stationary at infinity. Demonstrate that
(r) = 0
a3
,
r3
for r a. Show that the torque that the sphere exerts on the fluid is
= 8 0 a 3 .
10.6 Consider a solid sphere of radius a moving through a viscous fluid of viscosity at the fixed velocity V = V ez . Let r, , be spherical coordinates
whose origin coincides with the instantaneous location of the spheres center. Show that, if inertia and gravity are negligible, the fluid velocity, and the
radial components of the stress tensor, a long way from the sphere, are
a
3
V cos ,
2
r
a
3
v V sin ,
4
r
a
9
rr p0 V cos 2 ,
2
r
r 0,
vr
respectively. Hence, deduce that the net force exerted on the fluid lying inside
a large spherical surface of radius r, by the fluid external to the surface, is
F = 6 a V,
independent of the surface radius.
312
11
Waves in Incompressible Fluids
11.1 Introduction
This chapter investigates low amplitude waves propagating through incompressible
fluids. More information on this topic can be found in Lighthill 1978, and MilneThomson 2011.
v
+ (v ) v = p g ez + 2 v,
t
(11.1)
(11.2)
where is the (uniform) mass density, the (uniform) viscosity, and g the (uniform)
acceleration due to gravity. (See Section 1.14.) Let us write
p(r, t) = p0 g z + p1 (r, t),
(11.3)
where p0 is atmospheric pressure, and p1 the pressure perturbation due to the wave.
Of course, in the absence of the wave, the water pressure a depth h below the surface
is p0 + g h. (See Chapter 2.) Substitution into Equation (11.2) yields
v
p1 + 2 v,
t
(11.4)
where we have neglected terms that are second order in small quantities (i.e., terms
of order v 2 ).
313
314
Let us also neglect viscosity, which is a good approximation provided that the
wavelength is not ridiculously small. [For instance, for gravity waves in water, viscosity is negligible as long as
(2 /g)1/3 5 105 m.] It follows that
v
p1 .
t
(11.5)
0,
t
(11.6)
(11.8)
.
t
(11.9)
We now need to derive the physical constraints that must be satisfied at the waters upper and lower boundaries. It is assumed that the water is bounded from below
by a solid surface located at z = d. Because the water must always remain in contact with this surface, the appropriate physical constraint at the lower boundary is
vz |z=d = 0 (i.e., the normal velocity is zero at the lower boundary), or
= 0.
(11.10)
z z=d
The waters upper boundary is a little more complicated, because it is a free surface.
Let represent the vertical displacement of this surface due to the wave. It follows
that
.
= vz |z=0 =
(11.11)
t
z z=0
The appropriate physical constraint at the upper boundary is that the water pressure
there must equal atmospheric pressure, because there cannot be a pressure discontinuity across a free surface (in the absence of surface tensionsee Section 11.11).
Accordingly, from Equation (11.3), we obtain
p0 = p0 g + p1 |z=0 ,
(11.12)
315
or
g = p1 |z=0 ,
(11.13)
p1
=
,
= g
(11.14)
t
z z=0
t z=0
where use has been made of Equation (11.11). The previous expression can be combined with Equation (11.9) to give the boundary condition
2
1
= g
(11.15)
.
z z=0
t 2 z=0
g
(11.16)
(11.18)
where A and B are arbitrary constants. The boundary condition (11.10) is satisfied
provided that B = A exp(2 k d), giving
(x, z, t) = A e k z + ek (z+2 d) cos( t k x),
(11.19)
The boundary condition (11.15) then yields
2
A k 1 e2 k d cos( t k x) = A
1 + e2 k d cos( t k x).
g
(11.20)
(11.21)
(11.22)
316
in which the depth, d, of the water greatly exceeds the wavelength, = 2/k, of the
wave. In this limit, the gravity wave dispersion relation (11.21) reduces to
= (g k)1/2 ,
(11.23)
= v,
t
(11.27)
where v(r, t) is the perturbed velocity. For a plane wave of wave number k = k e x , in
the limit k d
1, Equation (11.19) yields
(x, z, t) = A e k z cos( t k x).
(11.28)
(11.29)
(11.30)
(11.31)
vz (x, z, t) = a e
(11.32)
kz
cos( t k x),
and
p1 = g z ,
(11.33)
317
surface
z
x
z=0
Figure 11.1
Motion of water particles associated with a deep water gravity wave propagating in
the x-direction.
where use has been made of Equations (11.7), (11.9), and (11.27). Here, a is the
amplitude of the vertical oscillation at the waters surface. According to Equations (11.29)(11.32), the passage of the wave causes a water particle located a depth
h below the surface to execute a circular orbit of radius a e k h about its equilibrium
position. The radius of the orbit decreases exponentially with increasing depth. Furthermore, whenever the particles vertical displacement attains a maximum value the
particle is moving horizontally in the same direction as the wave, and vice versa.
(See Figure 11.1.)
Finally, if we define h(x, z, t) = z (x, z, t) z as the equilibrium depth of the water
particle found at a given point and time then Equations (11.3) and (11.33) yield
p(x, z, t) = p0 + g h(x, z, t).
(11.34)
In other words, the pressure at this point and time is the same as the unperturbed
pressure calculated at the equilibrium depth of the water particle.
(11.35)
in which the depth, d, of the water is much less than the wavelength, = 2/k, of
the wave. In this limit, the gravity wave dispersion relation (11.21) reduces to
= (g d)1/2 k,
(11.36)
318
(11.37)
(11.39)
(11.40)
(11.41)
(11.42)
Here, a is again the amplitude of the vertical oscillation at the waters surface. According to the previous expressions, the passage of a shallow water gravity wave
causes a water particle located a depth h below the surface to execute an elliptical orbit, of horizontal radius a/(k d), and vertical radius a (1 h/d), about its equilibrium
position. The orbit is greatly elongated in the horizontal direction. Furthermore, its
vertical radius decreases linearly with increasing depth such that it becomes zero at
the bottom (i.e., at h = d). As before, whenever the particles vertical displacement
attains a maximum value the particle is moving horizontally in the same direction as
the wave, and vice versa.
1/2
tanh(k d)
kd
1/2
.
(11.43)
(11.44)
319
It follows that neither the phase velocity nor the group velocity of a gravity wave
can ever exceed the critical value (g d)1/2 . It is also easily demonstrated that the
displacement and velocity fields associated with a plane gravity wave of wavenumber
k e x , angular frequency , and surface amplitude a, are
cosh[k (z + d)]
cos( t k x),
sinh(k d)
sinh[k (z + d)]
z (x, z, t) = a
sin( t k x),
sinh(k d)
cosh[k (z + d)]
sin( t k x),
v x (x, z, t) = a
sinh(k d)
sinh[k (z + d)]
vz (x, z, t) = a
cos( t k x).
sinh(k d)
x (x, z, t) = a
(11.45)
(11.46)
(11.47)
(11.48)
The mean kinetic energy per unit surface area associated with a gravity wave is
defined
1 2
K=
v dz,
(11.49)
d 2
where
(x, t) = a sin( t k x)
(11.50)
is the vertical displacement at the surface, and
2
d(k x)
=
( )
2
0
(11.51)
is an average over a wavelength. Given that cos2 ( tk x) = sin2 ( tk x) = 1/2,
it follows from Equations (11.47) and (11.48) that, to second order in a,
0
1
cosh[2 k (z + d)]
2
1
dz = g a 2
.
(11.52)
K = a2 2
2
4
4
g k tanh(k d)
sinh (k d)
d
Making use of the general dispersion relation (11.21), we obtain
K=
1
g a 2.
4
(11.53)
The mean potential energy perturbation per unit surface area associated with a
gravity wave is defined
1
U=
g z dz + g d 2 ,
(11.54)
2
d
which yields
1
1
1
U = g ( 2 d 2 ) + g d 2 = g 2 ,
2
2
2
or
U=
1
g a 2.
4
(11.55)
(11.56)
320
In other words, the mean potential energy per unit surface area of a gravity wave is
equal to its mean kinetic energy per unit surface area.
Finally, the mean total energy per unit surface area associated with a gravity wave
is
1
E = K + U = g a 2.
(11.57)
2
This energy depends on the wave amplitude at the surface, but is independent of the
wavelength, or the water depth.
(11.58)
Suppose, for the sake of argument, that the wave train is of uniform transverse width
w. Consider a fixed line drawn downstream of the ship perpendicular to its path. The
rate at which the length of the train is increasing ahead of this line is V. Therefore, the
rate at which the energy of the train is increasing ahead of the line is (1/2) g a 2 w V,
where a is the typical amplitude of the transverse waves in the train. As is well
known (Fitzpatrick 2013), wave energy travels at the group velocity, rather than the
phase velocity. Thus, the energy flux per unit width of a propagating gravity wave
is simply E vg . Wave energy consequently crosses our fixed line in the direction of
the ships motion at the rate (1/2) g a 2 w vg . Finally, the ship does work against the
drag force, which goes to increase the energy of the train in the region ahead of our
line, at the rate D V. Energy conservation thus yields
1
1
g a 2 w V = g a 2 w vg + D V.
2
2
However, because V = v p , we obtain
vg
1
2kd
1
D = g a2 w 1
= g a2 w 1
,
2
vp
4
sinh(2 k d)
(11.59)
(11.60)
321
where use has been made of Equation (11.44). Here, k d is determined implicitly in
terms of the ship speed via expression (11.58). Note that this expression cannot be
satisfied when the speed exceeds the critical value (g d)1/2, because gravity waves
cannot propagate at speeds in excess of this value. In this situation, no transverse
wave train can keep up with the ship, and the drag associated with such waves consequently disappears. In fact, we can see, from the previous formulae, that when
V (g d)1/2 then k d 0, and so D 0. However, the transverse wave amplitude,
a, generally increases significantly as the ship speed approaches the critical value.
Hence, the drag due to transverse waves actually peaks strongly at speeds just below
the critical speed, before eectively falling to zero as this speed is exceeded. Consequently, it usually requires a great deal of propulsion power to force a ship to travel
at speeds faster than (g d)1/2.
In the deep water limit k d
1, Equation (11.60) reduces to
D=
1
g a 2 w.
4
(11.61)
322
1.1
1
0.9
0.8
D(a.u.)
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.2
0.4
0.6
Fr
0.8
1.2
Figure 11.2
Variation of wave drag with Froude number for a ship traveling through deep water.
V
(g l)1/2
(11.65)
323
(11.66)
This characteristic speed is sometimes called the hull speed. It can be seen that the
hull speed increases with the length of the ship. In other words, long ships tend to
have higher hull speeds than short ones.
vp
.
V
(11.67)
This, therefore, is the condition that must be satisfied in order for an obliquely propagating gravity wave to maintain a constant phase relation with respect to the ship.
In shallow water, all gravity waves propagate at the same phase velocity: that is,
v p = (g d)1/2,
where d is the water depth. Hence, Equation (11.67) yields
1/2
1 (g d)
= sin
.
V
(11.68)
(11.69)
(11.70)
324
D
D
E
vp t
A
Vt
C
Figure 11.3
An oblique plane wave generated on the surface of the water by a moving ship.
In other words, the ship must be traveling faster than the critical speed (g d)1/2.
Moreover, if this is the case then there is only one value of that satisfies Equation (11.69). This implies the scenario illustrated in Figure 11.4. Here, the ship
is instantaneously at A, and the wave maxima that it previously generatedwhich
all propagate obliquely, subtending a fixed angle with the x-axishave interfered
constructively to produce a single strong wave maximum DAE. In fact, the wave
maxima generated when the ship was at A have travelled to B and C , the wave
maxima generated when the ship was at A have travelled to B and C , et cetera.
We conclude that a ship traveling over shallow water produces a V-shaped wake
whose semi-angle, , is determined by the ships speed. Indeed, as is apparent from
Equation (11.69), the faster the ship travels over the water, the smaller the angle
becomes. Shallow water wakes are especially dangerous to other vessels, and particularly destructive of the coastline, because all of the wave energy produced by the
ship is concentrated into a single large wave maximum. Note, finally, that the wake
contains no transverse waves, because, as we have already mentioned, such waves
cannot keep up with a ship traveling faster than the critical speed ( g)1/2 .
Let us now discuss the wake generated by a ship traveling over deep water. In this
case, the phase velocity of gravity waves is v p = (g/k)1/2. Thus, Equation (11.67)
yields
v p $ g %1/2
=
.
(11.71)
sin =
V
kV2
It follows that in deep water any obliquely propagating gravity wave whose wave
number exceeds the critical value
k0 =
g
V2
(11.72)
325
D
B
B
A
A
C
C
E
Figure 11.4
A shallow water wake.
can keep up with the ship, as long as its direction of propagation is such that Equation (11.71) is satisfied. In other words, the ship continuously excites gravity waves
with a wide range of dierent wave numbers and propagation directions. The wake
is essentially the interference pattern generated by these waves. As is well known
(Fitzpatrick 2013), an interference maximum generated by the superposition of plane
waves with a range of dierent wave numbers propagates at the group velocity, vg .
Furthermore, as we have already seen, the group velocity of deep water gravity waves
is half their phase velocity: that is, vg = v p /2.
Consider Figure 11.5. The curve APD corresponds to a particular interference
maximum in the wake. Here, A is the ships instantaneous position. Consider a point
P on this curve. Let x and y be the coordinates of this point, relative to the ship.
The interference maximum at P is part of the plane wavefront BC emitted some
time t earlier, when the ship was at point A . Let be the angle subtended between
this wavefront and the x-axis. Because interference maxima propagate at the group
velocity, the distance A P is equal to vg t. Of course, the distance AA is equal to V t.
Simple trigonometry reveals that
x = V t vg t sin ,
(11.73)
y = vg t cos .
(11.74)
Moreover,
dy
= tan ,
dx
(11.75)
because BC is the tangent to the curve APDthat is, the curve y(x)at point P. It
326
D C
P
vg t
B
Vt
A
Figure 11.5
Formation of an interference maximum in a deep water wake.
follows from Equation (11.71), and the fact that vg = v p /2, that
1
x = X 1 sin2 ,
2
(11.76)
1
X sin cos ,
(11.77)
2
where X() = V t. The previous three equations can be combined to produce
y=
sin
,
cos
(11.78)
which reduces to
X
dX
=
.
d
tan
This expression can be solved to give
X = X0 sin ,
(11.79)
(11.80)
1
(11.81)
x = X0 sin 1 sin2 ,
2
y=
1
X0 sin2 cos .
2
(11.82)
327
0.3
0.2
y/X0
0.1
0
0.1
0.2
0.3
0
0.1
0.2
0.3 0.4
x/X0
0.5
0.6
Figure 11.6
Locus of an interference maximum in a deep water wake.
Here, the angle ranges from /2 to +/2. The curve specified by the previous
equations is plotted in Figure 11.6. As usual, A is the instantaneous position of the
ship. It can be seen that the interference maximum essentially consists of the transverse maximum BCD, and the two radial maxima AB and AD. As is easily demonstrated, point C, which corresponds to = 0, lies at x = X0 /2,
y = 0. Moreover,
the two cusps, B and D, which correspond to = tan1 (1/ 8) = 19.47, lie at
x = (8/27)1/2 X0 , y = (1/27)1/2 X0 .
The complete interference pattern that constitutes the wake is constructed out of
many dierent wave maximum curves of the form shown in Figure 11.6, corresponding to many dierent values of the parameter X0 . However, these X0 values must be
chosen such that the wavelength of the pattern along the x-axis corresponds to the
wavelength 0 = 2/k0 = 2 V 2 /g of transverse (i.e., = 0) gravity waves whose
phase velocity matches the speed of the ship. This implies that X0 = 2 j 0 , where j
is a positive integer. A complete deep water wake pattern is shown in Figure 11.7.
This pattern, which is made up of interlocking transverse and radial wave maxima,
fills a wedge-shaped
regionknown as a Kelvin wedgewhose semi-angle takes the
value tan1 (1/ 8) = 19.47. This angle is independent of the ships speed. Finally,
our initial assumption that the gravity waves that form the wake are all deep water
waves is valid provided k0 d
1, which implies that
V (g d)1/2.
(11.83)
y/0
328
6
5
4
3
2
1
0
1
2
3
4
5
6
1 0 1 2 3 4 5 6 7 8 9 10 11
x/0
Figure 11.7
A deep water wake.
In other words, the ship must travel at a speed that is much less than the critical speed
(g d)1/2. This explains why the wake contains transverse wave maxima.
(11.84)
(11.85)
where v1 and p1 are the small velocity and pressure perturbations, respectively, due
to the wave. To first order in small quantities, the fluid equations of motion, (11.1)
and (11.2), reduce to
v1 = 0,
p1
+ V v1 =
,
t
(11.86)
(11.87)
329
respectively. We can also define the displacement, (r, t), of a fluid particle due to
the passage of the wave, as seen in a frame co-moving with the fluid, as
+ V = v1 .
(11.88)
t
The curl of Equation (11.87) implies that v1 = 0. Hence, we can write
v1 = , and Equation (11.87) yields
p1
+V = .
(11.89)
t
(11.90)
The most general traveling wave solution to Equation (11.90), with wave vector
k = k e x , and angular frequency , is
(x, z, t) = [A cosh(k z) + B sinh(k z)] cos( t k x).
(11.91)
(11.92)
(11.93)
(11.94)
be the small vertical displacement of the interface due to the wave. In the lower
fluid, the perturbed velocity potential must be of the form (11.91), with the constants
330
z = d
V
z=0
z
z = d
x
Figure 11.8
Gravity waves at an interface between two immiscible fluids.
A and B chosen such that vz |z=d = 0 and z (x, 0, t) = (x, t). It follows that A =
(V c) 0 / tanh(k d) and B = (V c) 0 , so that
(x, z, t) = (V c) 0
cosh[k (z + d)]
cos( t k x).
sinh(k d)
(11.95)
In the upper fluid, the perturbed velocity potential must again be of the form (11.91),
with the constants A and B chosen such that vz |z=d = 0 and z (x, 0, t) = (x, t). It
follows that A = (V c) 0 / tanh(k d ) and B = (V c) 0 , so that
(x, z, t) = (V c) 0
cosh[k (z d )]
cos( t k x).
sinh(k d )
(11.96)
Here, c = /k is the phase velocity of the wave. From Equations (11.85) and (11.92),
the fluid pressure just below the interface is
p(x, 0, t) = p0 g + p1 (x, 0 , t)
= p0 g 0 sin( t k x) k (V c) A sin( t k x)
k (V c)2
= p0 g
(11.97)
0 sin( t k x).
tanh(k d)
Likewise, the fluid pressure just above the interface is
p(x, 0+ , t) = p0 g + p1 (x, 0+ , t)
= p0 g 0 sin( t k x) k (V c) A sin( t k x)
k (V c)2
= p0 g +
(11.98)
0 sin( t k x).
tanh(k d )
In the absence of surface tension at the interface, these two pressure must equal one
another: that is,
! "z=0+
(11.99)
p z=0 = 0.
331
k (V c)2 k (V c)2
+
,
tanh(k d)
tanh(k d )
(11.100)
which takes the form of a quadratic equation for the phase velocity, c, of the wave.
We can see that:
i. If = 0 and V = 0 then the dispersion relation reduces to (11.43) (with v p = c).
ii. If the two fluids are of infinite depth then the dispersion relation simplifies to
( ) g = k (V c)2 + k (V c)2 .
(11.101)
iii. In general, there are two values of c that satisfy the quadratic equation (11.100).
These are either both real, or form a complex conjugate pair.
iv. The condition for stability is that c is real. The alternative is that c is complex,
which implies that is also complex, and, hence, that the perturbation grows or
decays exponentially in time. Because the complex roots of a quadratic equation
occur in complex conjugate pairs, one of the roots always corresponds to an
exponentially growing mode. In other words, an instability.
v. If both fluids are at rest (i.e., V = V = 0), and of infinite depth, then the dispersion reduces to
g ( )
.
(11.102)
c2 =
k ( + )
It follows that the configuration is only stable when > : that is, when the
heavier fluid is underneath.
As a particular example, suppose that the lower fluid is water, and the upper fluid
is the atmosphere. Let s = / = 1.225 103 be the specific density of air at s.t.p.
(relative to water). Putting V = V = 0, d , and making use of the fact that s is
small, the dispersion relation (11.100) yields
1/2
tanh(k d)
1
c (g d)1/2
(11.103)
1 s [1 + tanh(k d)] .
kd
2
Comparing this with Equation (11.43), we can see that the presence of the atmosphere tends to slightly diminish the phase velocities of gravity waves propagating
over the surface of a body of water.
332
where a is much smaller than d. Let the elevation of the free surface of the water be
z = b sin(k x). We wish to determine the relationship between a and b.
We expect the velocity potential, perturbed pressure, and vertical displacement of
the water to be of the form (11.91), (11.92), and (11.93), respectively, with = c = 0,
because we are looking for a stationary (i.e., non-propagating) perturbation driven by
the static corrugations in the bottom. The boundary condition at the bottom is
which yields
(11.104)
(11.105)
(11.106)
b = V 1 B.
(11.107)
which gives
(11.108)
g b = k V A.
(11.109)
which leads to
Hence, from Equations (11.105), (11.107), and (11.109),
b=
a
,
cosh(k d) (g/k V 2 ) sinh(k d)
(11.110)
a
,
cosh(k d) (1 c 2 /V 2 )
(11.111)
or
b=
where c = [(g/k) tanh(k d)]1/2 is the phase velocity of a gravity wave of wave number
k. [See Equation (11.21).] It follows that the peaks and troughs of the free surface
coincide with those of the bottom when |V| > |c|, and the troughs coincide with the
peaks, and vice versa, when |V| < |c|. If |V| = |c| then the ratio b/a becomes infinite,
implying that the oscillations driven by the corrugations are not of small amplitude,
and, therefore, cannot be described by linear theory.
333
be interpreted as a surface tension. Let us now incorporate surface tension into our
analysis. Suppose that the interface lies at
z = (x, t),
(11.112)
(z )
.
|(z )|
(11.113)
It follows that
nx
,
x
nz 1.
(11.114)
(11.115)
(11.116)
where p is the jump in pressure seen crossing the interface in the opposite direction
to n. However, from Equations (11.114) and (11.115), we have
n
2
.
x 2
(11.117)
2
.
x 2
(11.118)
This expression is the generalization of Equation (11.99) that takes surface tension
into account.
Suppose that the interface in question is that between a body of water, of density
and depth d, and the atmosphere. Let the unperturbed water lie between z = d
and z = 0, and let the unperturbed atmosphere occupy the region z > 0. In the limit
in which the density of the atmosphere is neglected, the pressure in the atmosphere
takes the fixed value p0 , whereas the pressure just below the surface of the water is
p0 g + p1 |z=0 . Here, p1 is the pressure perturbation due to the wave. The relation
(11.118) yields
2
g p1 |z=0 = 2 ,
(11.119)
x
where is the surface tension at an air/water interface. However, /t = (/z)z=0,
where is the perturbed velocity potential of the water. Moreover, from Equation (11.9), p1 = (/t). Hence, the previous expression gives
2
3
+
(11.120)
g
=
.
z z=0
t 2 z=0 z 2 x z=0
334
This relation, which is a generalization of Equation (11.15), is the condition satisfied at a free surface in the presence of non-negligible surface tension. Applying
this boundary condition to the general solution, (11.19) (which already satisfies the
boundary condition at the bottom), we obtain the dispersion relation
k3
2 = g k +
tanh(k d),
(11.121)
which is a generalization of Equation (11.21) that takes surface tension into account.
k3
.
1/2
l=
.
g
(11.122)
(11.123)
(See Section 3.4.) The capillary length of an air/water interface at s.t.p. is 2.7103 m
(Batchelor 2000). The associated capillary wavelength is c = 2 l = 1.7 102 m.
Roughly speaking, surface tension is negligible for waves whose wavelengths are
much larger than the capillary wavelength, and vice versa. It is also helpful to introduce the critical phase velocity
vc = (2 g l)1/2.
(11.124)
This critical velocity takes the value 0.23 m/s for an air/water interface at s.t.p.
(Batchelor 2000). It follows from Equation (11.122) that the phase velocity, v p =
/k, of a surface water wave can be written
1/2
vp
1
1
=
.
(11.125)
kl+
vc
2
kl
Moreover, the ratio of the phase velocity to the group velocity, vg = d/dk, becomes
vg
1 1 + 3 (k l)2
=
.
(11.126)
v p 2 1 + (k l)2
In the long wavelength limit
c (i.e., k l 1), we obtain
vp
1
,
v0
(2 k l)1/2
(11.127)
335
vg
1
.
vp 2
(11.128)
We can identify this type of wave as the deep water gravity wave discussed in Section 11.3.
In the short wavelength limit c (i.e., k l
1), we get
1/2
vp
kl
,
(11.129)
vc
2
and
vg
3
.
vp 2
(11.130)
This corresponds to a completely new type of wave known as a capillary wave. Such
waves have wavelengths that are much less than the capillary wavelength. Moreover,
Equation (11.129) can be rewritten
1/2
k
vp
,
(11.131)
which demonstrates that gravity plays no role in the propagation of a capillary wave.
In fact, its place is taken by surface tension. Finally, it is easily seen that the phase
velocity (11.125) attains the minimum value v p = vc when = c (i.e., when k l = 1).
Moreover, from Equation (11.126), vg = v p at this wavelength. It follows that the
phase velocity of a surface wave propagating over a body of water can never be less
than the critical value, vc .
k (V c)2 k (V c)2
+
,
tanh(k d)
tanh(k d )
(11.132)
which is a generalization of the dispersion relation (11.100) that takes surface tension
into account. Here, c = /k is the phase velocity of a wave propagating along the
interface.
336
For the case in which both fluids are at rest, and of infinite depth, the previous
dispersion relation simplifies to give
( ) g + k 2 = ( + ) k c 2 .
(11.133)
Suppose that s = / is the specific gravity of the upper fluid with respect to the
lower. In the case in which s < 1 (i.e., the upper fluid is lighter than the lower one),
it is helpful to define
1/2
l=
(11.134)
g (1 s)
1/2
1s
.
(11.135)
c0 = 2 g l
1+s
It follows that
c2 1 1
+
k
l
.
=
c02 2 k l
(11.136)
Thus, we conclude that the phase velocity of a wave propagating along the interface
between the two fluids achieves its minimum value, c = c0 , when k l = 1. Furthermore, waves of all wavelength are able to propagate along the interface (i.e., c 2 > 0
for all k). In the opposite case, in which s > 1 (i.e., the upper fluid is heavier than the
lower one), we can redefine the capillary length as
1/2
l=
.
(11.137)
g (s 1)
The dispersion relation (11.133) then becomes
s1
1
2
c = gl
kl
.
s+1
kl
(11.138)
It is apparent that c 2 < 0 for k l < 1, indicating instability of the interface for waves
whose wavelengths exceed the critical value c = 2 l. On the other hand, waves
whose wavelengths are less than the critical value are stabilized by surface tension.
This result is exemplified by the experiment in which water is retained by atmospheric pressure in an inverted glass whose mouth is closed by a gauze of fine mesh
(the purpose of which is to put an upper limit on the wavelengths of waves that can
exist at the interface.)
337
the surface of the water move horizontally at the constant velocity V . Suppose that
is the density of water, s = / the specific gravity of air with respect to water,
and the surface tension at an air/water interface. With V = 0, k d , k d ,
the dispersion relation (11.132) reduces to
(1 s) g + k 2 = k c 2 + s k (V c)2 .
This expression can be rearranged to give
2V s
V 2 s
2
c
= c12 ,
+
1+s
1+s
(11.139)
(11.140)
which is a quadratic equation for the phase velocity, c, of the wave. Here,
g 1s
k
2
c1 =
,
(11.141)
+
k 1+s
(1 + s)
where c1 is the phase velocity that the wave would have in the absence of the wind.
In fact, we can write
1
c 2
c12 =
+
(11.142)
c
2 c
0
where c = 2 l is the capillary wavelength, and l and c0 are defined in Equations (11.134) and (11.135), respectively.
For a given wavelength, , the wave velocity, c, attains its maximum value, cm ,
when dc/dV = 0. According to the dispersion relation (11.140), this occurs when
V = cm = (1 + s)1/2 c1 .
(11.143)
If the wind has any other velocity, greater or less than cm , then the wave velocity is
less than cm .
According to Equation (11.140), the wave velocity, c, becomes complex, indicating an instability, when
(1 + s)2 2 (1 + s)2 c 2
2
c1 =
+
(11.144)
c .
V >
s
2s
c
0
We conclude that if the wind speed exceeds the critical value
Vc =
(1 + s)
c0 = 6.6 m/s = 12.8 kts
s1/2
(11.145)
then waves whose wavelengths fall within a certain range, centered around c , are
unstable and grow to large amplitude.
The two roots of Equation (11.140) are
c=
1/2
V s
s V 2
.
c12
1+s
(1 + s)2
(11.146)
338
Moreover, if
V < (1 + s 1 )1/2 c1
(11.147)
then these roots have opposite signs. Hence, the waves can either travel with the
wind, or against it, but travel faster when they are moving with the wind. If V
exceeds the value given previously then the waves cannot travel against the wind.
Because c1 has the minimum value c0 , it follows that waves traveling against the
wind are completely ruled out when
V > (1 + s 1 )1/2 c0 = 6.6 m/s = 12.8 kts.
(11.148)
11.15 Exercises
11.1 Find the velocity potential of a standing gravity wave in deep water for which
the associated elevation of the free surface is
z = a cos( t) cos(k x).
Determine the paths of water particles perturbed by the wave.
11.2 Deep water fills a rectangular tank of length l and breadth b. Show that the
resonant frequencies of the water in the tank are
(g )1/2 (n 2 l 2 + m 2 b 2 )1/4 ,
where n and m are integers. Neglect surface tension.
11.3 Demonstrate that a sinusoidal gravity wave on deep water with surface elevation
= a cos( t k x)
possesses a mean momentum per unit surface area
1
a 2.
2
11.4 A seismic wave passes along the bed of an ocean of uniform depth d such
that the vertical perturbation of the bed is a cos[k (x V t)]. Show that the
amplitude of the consequent gravity waves at the surface is
1
c2
a 1 2 cosh(k d) ,
V
where c is the phase velocity of waves of wavenumber k.
339
11.5 A layer of liquid of density and depth d has a free upper surface, and lies
over liquid of infinite depth and density > . Neglecting surface tension,
show that two possible types of wave of wavenumber k, with phase velocities
c 2 = k g,
c2 =
k g ( )
,
coth(k d) +
1
( g + k 2 ) a 2 ,
4
1
( g + k 2 ) a 2 .
4
11.7 Show that in water of uniform depth d the phase velocity of surface waves
can only attain a stationary (i.e., maximum or minimum) value as a function
of wavenumber, k, when
1/2
sinh(2 k d) 2 k d
k=
kc ,
sinh(2 k d) + 2 k d
where kc = ( g/)1/2 . Hence, deduce that the phase velocity has just one
stationary value (a minimum) for any depth greater than 31/2 kc1 4.8 mm,
but no stationary values for depths less than that.
11.8 Unlike gravity waves in deep water, whose group velocities are half their
phase velocities, the group velocities of capillary waves are 3/2 times their
phase velocities. Adapt the analysis of Section 11.7 to investigate the generation of capillary waves by a very small object traveling across the surface
of the water at the constant speed V. Suppose that the unperturbed surface
corresponds to the x-y plane. Let the object travel in the minus x-direction,
such that it is instantaneously found at the origin. Find the present position
of waves that were emitted, traveling at an angle to the objects direction of
motion, when it was located at (X, 0). Show that along a given interference
maximum the quantities X and vary in such a manner that X cos3 takes
a constant value, X1 (say). Deduce that the interference maximum is given
parametrically by the equations
1
x = X1 sec tan2
,
2
y=
3
X1 sec tan .
2
340
12
Terrestrial Ocean Tides
12.1 Introduction
This chapter outlines the classical theory of terrestrial ocean tides, the study of which
was pioneered by Pierre-Simon Laplace (1749-1827). Our treatment of ocean tides
takes all major harmonics of the tide generating potential into account, as well as
the finite elasticity of the Earth, and the self-gravity of the oceans. As will become
apparent, the theory of terrestrial ocean tides is extremely complicated. Hence, for
the sake of simplicity and brevity, we shall only consider two relatively idealized
scenarios. First, tides in a frictionless ocean of constant depth that covers the whole
surface of the Earth, and, second, tides in a frictionless ocean of constant depth
that only covers one hemisphere of the Earth (and lies between two meridians of
longitude). It turns out that these two scenarios (in particular, the latter) exhibit most
of the observed features of terrestrial ocean tides. More information on terrestrial
tides can be found in Lamb 1993 and Cartwright 1999.
m
(12.1)
r =
r
m + m
is the vector position of the center of mass, relative to the center of the planet. Moreover, the planet and moon both orbit the center of mass with angular velocity 1 ,
where (Fitzpatrick 2012)
G (m + m )
12 =
.
(12.2)
r 3
Here, G is the universal gravitational constant (Yoder 1995).
Neglecting the planets axial rotation (which is introduced into our analysis in
341
342
Section 12.6), the planets orbital motion about the center of mass causes its constituent points to execute circular orbits of radius r with angular velocity 1 . The
centers of these orbits have the same spatial relation to the center of mass that the
constituent points have to the center of the planet. Moreover, the orbits lie in parallel planes. Thus, the common centripetal acceleration of the constituent points is
(Fitzpatrick 2012)
(12.3)
g centripetal = 12 r = centripetal .
It follows, from the previous three equations, that
G (m + m )
G m
G m
r
r
=
r
=
r cos
r 3
r 3
r 2
G m r
= P1 (cos ),
r r
centripetal (r) =
(12.4)
r r
.
r r
(12.5)
Here,
1 dn
[(x 2 1)n ],
(12.6)
2 n n! dx n
for n 0, denotes a Legendre polynomial (Abramowitz and Stegun 1965). In particular,
Pn (x) =
P0 (x) = 1,
(12.7)
P1 (x) = x,
(12.8)
P2 (x) =
1
(3 x 2 1).
2
(12.9)
The gravitational force per unit mass exerted by the moon on one of the planets
constituent points, located at position vector r, is written
g moon (r) = moon ,
(12.10)
G m $ r %n
G m
=
Pn (cos ),
|r r |
r n=0, r
(12.11)
assuming that r > r. The tidal force per unit mass acting on the point (i.e., the force
per unit mass that is not compensated by the centripetal acceleration) is
g tide (r) = g moon (r) g centripetal = tide ,
(12.12)
(12.13)
where
343
is termed the tide generating potential. Thus, according to Equations (12.4) and
(12.11),
$ r %2
$ r %3
G m
tide (r) = 1 + P2 (cos ) + O .
(12.14)
r
r
r
The truncation of the expansion is ultimately justified by the fact that a/r 1.
(12.15)
(12.16)
(12.17)
(12.18)
(12.19)
Here,
d m [Pn (x)]
,
(12.21)
dx m
for n 0 and m n, denotes an associated Legendre function (Abramowitz and
Stegun 1965). In particular,
Pnm (x) = (1)m (1 x 2 )m/2
P20 (x) =
1
(3 x 2 1),
2
(12.22)
(12.23)
P22 (x) = 3 (1 x 2 ).
(12.24)
344
$ r %n
(12.25)
(12.26)
(12.27)
G m a 2 0
P2 (cos ) R2(0) (r, )
r 3
1
+ P21 (cos ) cos R2(1) (r, , )
3
1
+ P21 (cos ) sin R2(+1) (r, , )
3
1 2
P (cos ) cos(2 ) R2(2) (r, , )
+
12 2
1
+ P22 (cos ) sin(2 ) R2(+2) (r, , ) .
12
(12.28)
tan
,
cos
(12.29)
(12.30)
where
= 1 t + 2 e sin[(1 2 ) t] + O(e 2 ),
1
2
r = R 1 e cos[(1 2 ) t] + O(e 2 ) .
(12.31)
(12.32)
345
Here, it is assumed that the closest point on the moons orbit to the center of the planet
corresponds to = 2 t, where 2 is the uniform precession rate of this point. [It is
necessary to include such precession in our analysis because the Moons perigee precesses steadily in such a manner that it completes an orbit about the Earth once every
8.85 years. This eect is caused by the perturbing influence of the Sun (Fitzpatrick
2013).] Suppose that the inclination of the moons orbit to the planets equatorial
plane, = /2, is relatively small, so that sin 1. It follows that
cos = sin cos(1 t) + O(e sin ),
= 1 t + 2 e sin[(1 2 ) t] + O(e 2 ) + O(sin2 ).
(12.33)
(12.34)
G m a 2 1
3
2
tide (r, , ) =
1 + 3 e cos [(1 2 ) t] sin cos[2 1 t] R2(0) (r, )
2
2
R3
1
sin R2(1) (r, , )
2
1
+ sin R2(1) (r, , 2 1 t)
2
7 e (2)
(r, , [(3/2) 1 (1/2) 2] t)
R
8 2
1
R2(2) (r, , 1 t)
4
e
+ R2(2) (r, , [(1/2) 1 + (1/2) 2] t)
8
+O(e 2 ) + O(e sin ) + O(sin2 ) .
+
(12.35)
Here, we have retained a term proportional to sin2 in the previous expression, despite the fact that we are formally neglecting O(sin2 ) terms, because the term in
question gives rise to important fortnightly tides on the Earth.
(12.36)
346
cnm
where the
and dnm are arbitrary coecients, and the Pnm (x) associated Legendre
functions.
A solid harmonic of degree n (where n is a non-negative integer), denoted Rn (r, , ),
is defined as a well-behaved solution to
2 Rn = 0
(12.38)
in the interior of a sphere (i.e., the region r < constant). It follows that (Love 1927)
Rn (r, , ) r n Sn (, ).
(12.39)
In particular, the functions Rn(m) (r, , ) and Rn(0) (r, ), introduced in Section 12.3,
are solid harmonics of degree n. Note that the Cartesian coordinates xi (where i runs
from 1 to 3) are solid harmonics of degree 1. Moreover, Rn /xi is a solid harmonic
of degree n 1. Finally, tide (r), specified in Equation (12.35), is a solid harmonic
of degree 2.
The following results regarding solid harmonics are useful (Love 1927):
xi
Rn
Rn
= n Rn ,
=r
xi
r
(12.40)
2 (xi Rn ) = (Rn xi + xi Rn ) = 2 xi Rn = 2
2 (r m Rn ) = 2 (r m+n Sn ) =
Rn
,
xi
(12.41)
1 d 2 d m+n
m+n2
(r
S
)
Sn
r
n n (n + 1) r
dr
r 2 dr
= m (m + 2 n + 1) r m2 Rn .
(12.42)
Here, use has been made of the Einstein summation convention (Riley 1974).
(12.43)
= t.
(12.44)
Let
It follows that r, , are spherical coordinates in a frame of reference that co-rotates
with the planet.
347
b n (, ),
(12.46)
n=1,
(12.47)
t n (, ),
(12.48)
n=1,
(12.49)
where (r) is the total gravitational potential. In the limit d/a 1, we can write
n (r, , ) + tide (r, , ),
(12.50)
(r, , ) = 0 (r) +
n=1,
where
0 =
2
4 G + 4 G d (r a)
0
ra
,
r>a
(12.51)
and
2 n>0 = 4 G ( b n + n ) (r a).
(12.52)
(, ) = t (, ) b (, ),
(12.53)
Here,
is the net change in ocean depth due to planetary rotation and tidal eects, and n =
t n b n . Moreover, (x) is a Dirac delta function (Riley 1974). The boundary
conditions are that the n be well behaved as r 0, and
n (r) 0
(12.54)
348
as r . It follows that
g
(3 a 2 r 2 ) 3 g d ,
2a
g a2
d
0 (r > a) =
1+3
,
r
a
0 (r a) =
(12.55)
(12.56)
and
3
n>0 (r a, , ) = g
b n +
2n+1
3
n>0 (r > a, , ) = g
b n +
2n+1
$ %n
r
n
,
a
$ %(n+1)
r
n
,
(12.57)
(12.58)
where
G m 4
Ga
(12.59)
=
3
a2
is the mean gravitational acceleration at the planets surface. Note that, inside the
planet (i.e., r a), n (r) is a solid harmonic of degree n. We can identify the
three terms appearing on the right-hand side of Equation (12.50) as the equilibrium
gravitational potential generated by the planet and the ocean, the potential generated
by non-spherically-symmetric radial displacements of the planet and ocean surfaces,
and the tide generating potential, respectively.
g=
i j
+
i j = p i j +
,
(12.60)
x j xi
and subject to the incompressibility constraint
= 0.
(12.61)
Here, i j (r) is the stress tensor, i j the identity tensor, (r) the elastic displacement,
p(r) the pressure, and the (uniform) rigidity of the material making up the planet
(Riley 1974).
The planets equation of elastic motion in the co-rotating frame is (Fitzpatrick
2012)
2 i
k 1 i j
=
+ 2 i jk j
,
(12.62)
2
t
x j
xi xi
t
349
where i jk is the totally antisymmetric tensor (Riley 1974), and (Fitzpatrick 2012)
1
(r, ) = 2 r 2 sin2 = 0 (r) + 2 (r, ),
2
2 a 2 $ r %2
0 (r) =
,
3
a
2 a 2 $ r %2 0
2 (r, ) =
P2 (cos ).
3
a
(12.63)
(12.64)
(12.65)
Note that 2 (r) is a solid harmonic of degree 2. The second term on the left-hand side
of Equation (12.62) is the Coriolis acceleration (due to planetary rotation), whereas
the final term on the right-hand side is the centrifugal acceleration (likewise, due to
planetary rotation). The first two terms on the right-hand side are the forces per unit
mass due to internal stresses and gravity, respectively.
Equations (12.60), (12.61), and (12.62) can be combined to give
p
2
2
=
++ .
(12.66)
+2
t
t 2
/
:
(12.67)
a
that is, the typical timescale on which the tide generating potential varies (in the corotating frame) is much longer than the transit time of an elastic shear wave through
the interior of the planet. In this case, we can neglect the left-hand side of Equation (12.66), and write
p
2
+ + 0,
(12.68)
which is equivalent to saying that the interior of the planet always remains in an
equilibrium state.
Let
!
"
(12.69)
p(r, , ) = p0 (r) +
pn (r, , ) n (r, , ) ext (r, , ),
n=1,
where
ext (r, , ) = tide (r, , ) + 2 (r, ),
(12.70)
is the sum of the tide generating potential and the fictitious centrifugal potential due
to planetary rotation. Note that ext is a solid harmonic of degree 2. It follows from
Equations (12.50), (12.63), and (12.68) that
d p0
+ 0 + 0 = 0,
(12.71)
dr
and
2
n=1,
pn = 0.
(12.72)
350
Taking the divergence of the previous equation, and making use of Equations (12.61),
we find that 2 pn = 0, which implies that pn (r) is a solid harmonic (of degree n).
Equation (12.71) can be integrated to give
2
(12.73)
g 2 a (a 2 r 2 ) + g d + patm ,
p0 (r)
2a
3
where use has been made of Equations (12.55) and (12.64), as well as
p0 (a) = g d + patm .
(12.74)
Here, patm is the atmospheric pressure at sea level. The previous boundary condition
ensures that the mean pressure at the surface of the planet is able to support the
mean weight of the ocean, as well as the weight of the atmosphere. Incidentally, it is
assumed that g
2 a, and we have neglected terms of order 2 a d with respect
to terms of order g d.
The elastic displacement in the interior of the planet satisfies Equations (12.61)
and (12.72). It is helpful to define the radial component of the displacement
r =
xi i
,
r
(12.75)
(12.76)
where use has been made of Equation (12.60). The radial displacement at r = a is
equivalent to the displacement of the planets surface. Hence, according to Equation (12.45),
(12.77)
r (a, , ) = b (, ).
The stress at any point on the surface r = a must be entirely radial (because a fluid
ocean cannot withstand a tangential stress), and such as to balance the weight of the
column of rock and ocean directly above the point in question. In other words,
!
" $ xi %
.
(12.78)
Xi (a, , ) = g b + (d + )
r r=a
It follows from Equation (12.57), (12.69), (12.70), (12.74), (12.76), and (12.78) that
xi i
(r r )
i +
pn
r
r
r
r
xi
n=1,
r=a
$ %
2
n
2
2 n 2
xi
n
= g
,
2 n 2 +
b n +
2n+1
2n+1
r r=a
n=1,
where
(12.79)
(12.80)
351
n
pn
+ B n p n xi +
.
xi
xi
(12.82)
Here, An and Bn are spatial constants, and n (r) is an arbitrary solid harmonic of
degree n. It follows that
r r n xi i n = (n An + Bn ) r 2 pn + n n ,
(12.83)
(12.84)
n
i n
pn
i n = n An r 2
+ n Bn pn xi + (n 2)
.
r
xi
xi
(12.85)
and
r
(2 An + Bn ) r 2 pn + n n
r=a
= a b n ,
(12.86)
and
pn
n
(1 [2 n An + (2 + n) Bn] ) (pn xi )r=a (2 n An + Bn ) r
+ 2 (n 1)
xi
xi r=a
2n2
2 n 2
n (xi )r=a , (12.87)
= g 2 n 2 +
b n +
2n+1
2n+1
2
(12.88)
(12.89)
352
and
(1 [2 n An + (2 + n) Bn] ) pn |r=a =
2n2
2 n 2
n ,
g 2 n 2 +
b n +
2n+1
2n+ 1
(12.90)
respectively.
The expression for i n given in Equation (12.82) satisfies Equations (12.61) and
(12.72) provided that
2 n An + (3 + n) Bn = 0,
(12.91)
2 (2 n + 1) An + 2 Bn = 1,
(12.92)
respectively, where use has been made of Equations (12.40)(12.42). It follows that
(3 + n)
,
2 (2 n + 3) (n + 1)
2n
.
Bn =
2 (2 n + 3) (n + 1)
An =
(12.93)
(12.94)
(12.95)
where
,
2n+1
)
2n+1
(2 n + 1) (2 n 2 + 4 n + 3)
hn =
1+
,
2n2
ga
(6 + n 2 )
)
2n+ 1
(2 n + 1) (2 n 2 + 4 n + 3)
hn =
1+
.
3
ga
(6 + n 2 )
n =
(12.96)
(12.97)
(12.98)
(12.99)
n=1,
This quantity can be recognized as the total perturbing potential at the planets surface due to the combination of tidal, rotational, and ocean self-gravity, eects. It
follows from Equation (12.57), (12.80), and (12.95) that
=
(1 + k2 ) 2 n 2 + (1 + kn ) n n ,
(12.100)
g r=a n=1,
353
where
)
3
(2 n + 1) (2 n 2 + 4 n + 3)
1+
,
2n2
ga
(6 + n 2 )
1
(2 n + 1) (2 n 2 + 4 n + 3)
.
kn = 1 +
ga
(6 + n 2 )
kn =
(12.101)
(12.102)
Here, the dimensionless quantities hn , hn , kn , and kn are known as Love numbers
(Love 1911).
p
v
+ 2 v =
++ ,
t
(12.103)
(12.104)
where v(r) is the fluid velocity, and p(r) the fluid pressure. Here, (r) is the total
gravitational potential, (r) the centrifugal potential (due to planetary rotation), and
2 v the Coriolis acceleration (likewise, due to planetary rotation). Furthermore,
we have neglected the finite compressibility of the fluid that makes up the ocean. Let
r, , be spherical coordinates in the co-rotating frame, and let z = r a. It is helpful
to define u = v , v = v , and w = vr . Assuming that |z| a, the previous equations of
motion reduce to
1
v w
1
(sin u) +
+
0,
a sin
a sin z
(12.105)
and
u
1 p
2 cos v =
++ ,
t
a
1
p
v
+ 2 cos u + 2 sin w =
++ ,
t
a sin
p
w
2 sin v =
++ .
t
z
(12.106)
(12.107)
(12.108)
In accordance with Equations (12.45), (12.47), and (12.53), the lower and upper
surfaces of the ocean lie at
z = b (, , t),
(12.109)
354
and
z = d + b (, , t) + (, , t),
(12.110)
|b |, || d a.
(12.111)
respectively, where
It follows that
b
,
t
(b + )
.
=
t
w|z=0 =
(12.112)
w|z=d
(12.113)
Let us assume that u = u(, , t) and v = v(, , t): that is, the horizontal components
of the fluid velocity are independent of z. Integration of Equation (12.105) from z = 0
to z = d, making use of the previous two boundary conditions, yields
v
=
(sin u) +
.
(12.114)
t
a sin
d
d
w
u,
v u, v.
a
a
In accordance with the analysis of Section 12.8, we can write
2 2
+ = g a (z d) + (, , t),
3
(12.115)
(12.116)
where is the total perturbing potential due to tidal, rotation, and ocean selfgravity, eects. We have neglected an unimportant constant term. We have also
neglected the z-dependence of the perturbing potential, because the variation lengthscale of this potential is a, rather than d, so that
() d
d
,
z
a
(12.117)
p
+ + 0.
(12.118)
z
It follows that
p patm
+ + = P(, , t),
(12.119)
where patm is the (uniform and constant) pressure of the atmosphere. However, pressure balance at the oceans upper surface requires that
p|z=d+b + = patm .
(12.120)
355
(12.121)
to first order in small quantities [i.e., b /a, /a, /(g a), and 2 a/g]. Thus, Equations (12.106), (12.107), and (12.115) yield
1
u
2 cos v =
t
a
v
+ 2 cos u =
t
a
"
!
g (b + ) + ,
"
1
!
g (b + ) + .
sin
(12.122)
(12.123)
Let us justify our previous assumption that the ocean is in approximate vertical
force balance. Equations (12.108), (12.115), and (12.119) imply that
p patm
+ + = P(, , t) + O(d v),
(12.124)
where we have assumed that /t O() and /z d 1 . However, Equations (12.121)(12.123) yield
P g a v
d v.
(12.125)
(12.126)
,
a
() d
d
u
u u,
a
a a a
u
(12.127)
(12.128)
h2 2 n 2 + hn n n ,
(12.129)
n=1,
g 1 =
(1 + k2 ) 2 n 2 + (1 + kn ) n n .
n=1,
(12.130)
356
v
=
(sin u) +
,
(12.131)
t
a sin
g
u
(1 + k2 h2 ) 2
2 cos v =
(1 + kn hn ) n n ,
t
a
n=1,
(12.132)
g
v
(1 + k2 h2 ) 2
+ 2 cos u =
(1 + kn hn ) n n .
t
a sin
n=1,
(12.133)
We can write
(, ) =
n (, ),
(12.134)
(12.135)
n=1,
n (, ) =
m=0,n
and the cnm are arbitrary complex constants. Now, (Abramowitz and Stegun 1965)
2
(n + m)!
n n .
Pnm (cos ) Pnm (cos ) sin d =
(12.136)
2 n + 1 (n m)!
0
Hence,
(1 + kn hn ) n n =
n=1,
G F (, ; , ) ( , ) sin d d , (12.137)
where
G F (, ; , ) =
(2 n + 1) (n m)!
(1 + kn hn )
4
(n + m)!
n=1, m=0,n
=
(sin u) +
,
t
a sin
g
u
2 cos v =
(1 + k2 h2 ) 2 G F d ,
t
a
g
v
+ 2 cos u =
(1 + k2 h2 ) 2 G F d ,
t
a sin
(12.138)
(12.139)
(12.140)
(12.141)
357
j
(1 + 2 )
3 e 2
+(3/2) sin2 2
sin 2
sin 2
+(7 e/4) 2
+(1/2) 2
mj
0
0
0
1
1
2
2
j
0
1 2
2 1
2 1
2 3 1 + 2
2 2 1
Classification
Equilibrium
Long Period
Long Period
Diurnal
Diurnal
Semi-diurnal
Semi-diurnal
Table 12.1
The principal harmonics of the forcing term in the Laplace tidal equations.
j (, , t),
(12.142)
where
m
j (, , t) = j P2 j (cos ) e i (m j + j t) .
(12.143)
358
j
( + M + E )
+(3/2) sin2 E
3 e M M
+(3/2) sin2 M
sin M
sin E
sin ( M + E )
+(7 e M /4) M
+(1/2) M
+(1/2) E
mj
0
0
0
0
1
1
1
2
2
2
j
0
2 E
M M
2 M
2 M
2 E
2 3 M + M
2 2 M
2 2 E
Classification
Equilibrium
Long Period
Long Period
Long Period
Diurnal
Diurnal
Diurnal
Semi-diurnal
Semi-diurnal
Semi-diurnal
Symbol
S 0 , M0
S sa
Mm
Mf
O1
P1
K1
N2
M2
S2
Table 12.2
The principal harmonics of the forcing term in the Laplace tidal equations for the
Earth.
The amplitudes, j , azimuthal mode numbers, m j , and frequencies, j , of the principal harmonics of the forcing term are specified in Table 12.1, where
1 =
2 a2
,
3g
(12.144)
2 =
1 m a 4
.
2 m R3
(12.145)
Here, m is the planetary mass, m the mass of the moon, and R the moons orbital
major radius. As indicated in the table, the harmonics of the forcing term can be
divided into four classes: an equilibrium term that is time independent; two long
period terms that oscillate with periods much longer than the planets diurnal (i.e.,
rotational) period; two diurnal terms that oscillate at periods close to the planets
diurnal period; and two semi-diurnal terms that oscillate at periods close to half
the planets diurnal period. Here, we have neglected a semi-diurnal term that is
significantly smaller than the other two.
For the Earth-Moon system, m = 5.97 1024 kg, m = 7.35 1022 kg, a =
6.37 106 m, R = 3.84 108 m, = 7.29 105 rad. s1 (Yoder 1995). It follows
that g = 9.82 m s2 , as well as 1 = , 2 = M , where
= 7.32 103 m,
(12.146)
M = 1.79 101 m.
(12.147)
359
j (m) m j
2
+1.93 10
2.95 102
+4.25 102
7.12 102
3.24 102
10.36 102
+1.72 102
+8.95 102
+4.07 102
0
0
0
1
1
1
2
2
2
f j = j /2 Period
3
2.73 10
1.81 102
3.65 102
4.63 101
4.97 101
5.00 101
9.45 101
9.63 101
9.97 101
Symbol
182.62 days S sa
27.555 days Mm
13.661 days M f
25.819 hours O1
24.066 hours P1
23.934 hours K1
12.658 hours N2
12.421 hours M2
12.000 hours S 2
Table 12.3
The principal harmonics of the forcing term in the Laplace tidal equations for the
Earth, excluding the equilibrium harmonic.
R, e M , M , and M are the Moons mass, orbital major radius, orbital eccentricity,
mean orbital angular velocity, and rate of perigee precession, respectively. Here, we
have neglected the small (i.e., about 5 ) inclination of the Moons orbital plane to the
ecliptic.
For the Earth-Sun system, m = 1.99 1030 kg and R = 1.50 1011 m (Yoder
1995). It follows that 1 = and = E , where
E = 8.14 102 m.
(12.148)
360
(12.149)
(12.150)
(12.151)
(12.152)
We deduce that the Earths rotation causes a planetary equatorial (i.e., = /2) bulge
of about 9.2 km, and a polar (i.e., = 0) flattening of 18.3 km, but does not give rise
to any spatial variation in ocean depth. The observed equatorial bulge and polar flattening of the Earth are 7.1 km and 14.3 km, respectively (Yoder 1995). Our estimates
for these values are too large because, for the sake of simplicity, we are treating the
Earth as a uniform body. In reality, the Earth possesses a mass distribution that is
strongly concentrated in its core.
1
(sin ) +
+ = 0,
(12.153)
sin
g d ( F)
,
a2
.
g d ( F )
..
+ 2 cos = 2
a sin
..
2 cos =
where
F (, , t) = (1 + k2 h2 ) 2 (, , t) +
(12.154)
,
G F (, ; , ) ( , , t) d ,
(12.155)
(12.156)
361
and
a
d
a
v=
d
u=
,
t
.
t
(12.157)
(12.158)
Here, /t.
It is convenient to write (Love 1913)
=
,
sin
(12.159)
1
+
.
sin
(12.160)
1 2
1
sin
+
.
sin
sin 2
(12.161)
(12.162)
.
.
D +2
+ 2 cos D sin
= 0.
(12.163)
. gd
.
+ 2 cos D sin
= 2 D ( F ).
(12.164)
D+2
t
a
Let = cos . Consider the response, j (, , t), of the ocean to a particular
harmonic of the forcing term that takes the form [see Equation (12.143)]
m
2 (, , t) = j P2 j () e i (m j + j t) ,
zn
m
Pn j () e i (m j + j t) ,
n
(n
+
1)
n=1,
m
n Pn j () e i (m j + j t) ,
(, , t) = i
(12.165)
(12.166)
(12.167)
n=1,
where the Pnm (x) are associated Legendre functions, and the zn and n are constants.
Now, by definition (Abramowitz and Stegun 1965),
D Pnm () e i m = n (n + 1) Pnm() e i m .
(12.168)
362
(12.169)
n=1,
(12.170)
It follows from Equations (12.163) and (12.164) that
m
(1 2 ) d
m
n f j n (n + 1) + m j Pn j () + zn
Pn j () = 0,
n (n + 1) d
n=1,
(12.171)
and
n (n + 1)
fj mj
m
2
(1 [1 + kn hn ] n ) +
f j Pn j ()
zn
n
(n
+
1)
n=1,
d
6
m
m
+ f j n n (n + 1) (1 2 )
Pn j () = (1 + k2 h2 ) j P2 j (), (12.172)
d
fj =
(12.173)
(12.174)
$ n+m %
nm+1 m
m
(12.175)
Pn () =
Pn+1 () +
P m (),
2n+1
2 n + 1 n1
m
n (n m + 1) m
(n + 1) (n + m) m
2 dPn
=
(1 )
Pn+1 () +
Pn1 (). (12.176)
d
2n+1
2n+1
Equations (12.171), (12.172), (12.175), and (12.176) can be combined to give
n + 1 n mj
n n + mj + 1
m j n (n + 1) f j n +
zn1 +
zn+1 = 0, (12.177)
n 2n 1
n+ 1 2n+3
and
mj fj
n (n + 1)
2
(1 [1 + kn hn ] n ) +
f j zn
n (n + 1)
(n 1) (n + 1) (n m j )
+ fj
n1
2n1
n (n + 2) (n + m j + 1)
6
+ fj
n+1 = (1 + k2 h2 ) j n 2 .
2n+3
(12.178)
363
Finally, Equations (12.177) and (12.178) yield the tridiagonal matrix equation (Love
1913)
6
a(n) zn2 + b(n) zn + c(n) zn+2 = (1 + k2 h2 ) j n 2
(12.179)
for n = 2, 4, 6, , where
a(n) =
b(n) =
f j n (n + 1) (n m j 1) (n m j )
,
(2 n 3) (2 n 1) [m j n (n 1) f j ]
(12.180)
mj fj
4
n (n + 1) 3
f j2
1 [1 + kn hn ] n +
n (n + 1)
f j (n 1)2 (n + 1) (n m j ) (n + m j )
n (2 n 1) (2 n + 1) [m j n (n 1) f j ]
f j (n + 2)2 n (n m j + 1) (n + m j + 1)
,
(n + 1) (2 n + 1) (2 n + 3) [m j (n + 1) (n + 2) f j ]
(12.181)
f j n (n + 1) (n + m j + 1) (n + m j + 2)
.
(2 n + 3) (2 n + 5) [m j (n + 1) (n + 2) f j ]
(12.182)
c(n) =
(12.183)
The j , m j , and f j values for the various harmonics of the forcing term in the Laplace
tidal equations for the Earth are specified in Table 12.3. Note that, on the longitudinal
meridian = 0, the Moon and Sun culminate simultaneously at t = 0. Moreover, at
this time, the Sun and Moon are both passing though the summer solstice, and the
Moon is passing though its perigee. The mean density of the Earth is such that / =
5.5 (Yoder 1995). Moreover, the mean depth of the Earths oceans is d = 3.8 km
(Yoder 1995), which implies that = 23.1. Finally, the Earth responds elastically to
the relatively low-amplitude lunar and solar tidal gravitational fields in such a manner
that /( g a) 0.35 (Fitzpatrick 2012).
Figures 12.1, 12.2, and 12.3 show the long-period, diurnal, and semi-diurnal
solutions of the Laplace tidal equations together with the corresponding equilibrium
tides (i.e., the tides calculated in the absence of ocean inertia),
1 + k2 h 2
m
j P2 j (sin ).
(12.185)
j () =
1 (1 + k2 h2 ) 2
364
0.04
0.03
0.005
0.02
0.01
0
0.005
0.01
0.02
0.01
0
0.2
0.4
0.6
sin()
0.8
0.03
0.2
0.4
0.6
sin()
0.8
Figure 12.1
Long-period solutions to the Laplace tidal equations (left panel) together with the
corresponding equilibrium tides (right panel). The solid, dashed, and dash-dotted
lines correspond to j = 1, j = 2, and j = 3, respectively. Here, j is measured in
meters.
It can be seen from Figure 12.1 that the long-period solutions are direct (i.e.,
they have the same signs as the corresponding equilibrium tides). However, the amplitudes of these solutions are about 4 times lower than those of the equilibrium tides.
The j = 1 (S sa ) and j = 3 (M f ) solutions cause a slight increase in the equatorial
bulge of the Earths oceans when the Sun and Moon, respectively, pass through the
equinoxes, and a slight reduction when they pass through the solstices. The j = 2
(Mm ) solution causes a slight increase in the equatorial bulge of the Earths oceans
when the Moon passes through its perigee, and a slight reduction when it passes
through its apogee.
It can be seen from Figure 12.2 that the diurnal solutions are inverted (i.e., they
have the opposite signs to the corresponding equilibrium tides). The amplitude of
the j = 4 solution is about 6 times lower than that of the j = 4 equilibrium tide.
The amplitudes of the j = 5 and j = 6 solutions are negligibly small. (In fact, the
amplitude of the j = 6 solution is identically zero.) The j = 4 (O1 ) solution causes
the Earths oceans to displace slightly toward the Moon when it passes through the
equinoxes, and away from the Moon when it passes through the solstices.
Finally, it can be seen from Figure 12.3 that the semi-diurnal solutions are inverted in equatorial regions, and direct in polar regions. Moreover, the amplitudes of
these solutions are similar to those of the corresponding equilibrium tides. The dominant j = 8 (M2 ) solution generates a high tide at a particular point on the Earths
surface every 12.421 hours. The j = 9 (S 2 ) solution causes an enhancement of the
high tidea so-called spring tideat the new and full moons (i.e., every half a synodic month, or 14.77 days). Finally, the j = 7 (N2 ) solution causes an enhancement
365
0.12
0.005
0.09
j
0.06
0.01
0.03
0.015
0
0.2
0.4
0.6
sin()
0.8
0.2
0.4
0.6
sin()
0.8
Figure 12.2
Diurnal solutions to the Laplace tidal equations (left panel) together with the corresponding equilibrium tides (right panel). The solid, dashed, and dash-dotted lines
correspond to j = 4, j = 5, and j = 6, respectively. Here, j is measured in meters.
0.1
0.05
0.2
0
0.05
j
0.1
0.15
0.1
0.2
0.25
0.3
0.35
0.2
0.4
0.6
sin()
0.8
0.2
0.4
0.6
sin()
0.8
Figure 12.3
Semi-diurnal solutions to the Laplace tidal equations (left panel) together with the
corresponding equilibrium tides (right panel). The solid, dashed, and dash-dotted
lines correspond to j = 7, j = 8, and j = 9, respectively. Here, j is measured in
meters.
366
of the spring tidea so-called perigean spring tidewhenever the new or full moon
coincides with the passage of the moon through its perigee.
The previously described global solutions to the Laplace tidal equations were
first obtained (in simplified form) by Laplace (Lamb 1993). Let us consider how
these solutions compare with the observed terrestrial tides. One obvious prediction
of the global solutions is that the time variation in ocean depth at a given point on
the Earths surface should be expressible as a superposition of terms that oscillate at
constant amplitude (but not necessarily in phase with one another) with the periods
specified in Table 12.3. This is indeed found to be the case (Cartwright 1999). Another prediction is that the amplitudes of the long-period solutions should be small
compared to the amplitudes of the semi-diurnal solutions. This is also found to be
the case. Yet another prediction is that amplitudes of the diurnal solutions should
be small compared to the amplitudes of the semi-diurnal solutions. This turns out
not to be the case (Cartwright 1999). In fact, this predication is an artifact of the
fact that we are assuming that the ocean covering the Earths surface is of constant
mean depth. If we relax this assumption, and allow the ocean depth to vary in a
realistic manner, then the amplitudes of the semi-diurnal solutions become comparable to those of the diurnal solutions (Lamb 1993). The final prediction is that a
given solution is such that the tide at a given point on the Earth either oscillates in
phase or in anti-phase with the corresponding equilibrium solution (i.e., the tide is
either direct or inverted). In particular, for the dominant M2 tide, this implies that,
at a given point on the Earths surface, there should be either a tidal maximum or a
tidal minimum whenever the Moon culminates. It turns out that this is not the case
(Cartwright 1999). Indeed, the observed phase dierence between the M2 tide and
the corresponding equilibrium tide varies from place to place on the Earths surface,
and can take any value. In other words, high tide at a given point on the Earth can
occur three hours before the Moon culminates, or two hours after, et cetera, and the
time lag between high tide and the culmination of the Moon varies from place to
place. We conjecture that this eect is caused by the continents impeding the flow of
tidal waves around the Earth. Hence, we now need to consider tides in oceans that
do not cover the whole surface of the Earth.
367
(sin ) +
+ = 0,
sin
gd
( F ),
a 2
.
gd
..
( F ),
+ 2 cos = 2
a sin
..
2 cos =
where
F (, , t) = (1 + k2 h2 ) 2 (, , t) +
G F (, ; , ) ( , , t) d .
(12.186)
(12.187)
(12.188)
(12.189)
The finite extent of the ocean in the direction introduces the additional constraint
(, , t) = 0
(12.190)
Y
1
(sin X) +
DP =
,
(12.191)
sin
where
1 P(, , t)
= Y(, , t),
sin
(12.192)
X
(sin Y)
,
sin
(12.193)
(12.194)
1 Q
P
,
sin
(12.195)
Y=
1 P Q
+
.
sin
(12.196)
368
Let
X = X +
1 Q
P
+
,
sin
(12.197)
Y = Y +
1 P Q
.
sin
(12.198)
(12.199)
Y
(sin X ) +
= 0,
(12.200)
X
(sin Y )
= 0.
(12.201)
1 P
,
Y =
sin
X =
(12.202)
(12.203)
and
1 P (, , t)
= 0.
sin
(12.204)
(12.205)
Multiplying by P , integrating over the ocean, and making use of the boundary condition (12.204), we obtain
2
2
1 P
P
+
d = 0.
sin
(12.206)
Here, is the surface of the Earth that is covered by ocean, and d = sin d d.
It follows that P is a constant. Thus, Equations (12.202) and (12.203) imply that
X = Y = 0, and, hence, that Equations (12.195) and (12.196) are valid.
369
1
(sin )
,
sin
(12.207)
(12.208)
where
1 (, , t)
= 0,
sin
(12.209)
(, , t) = 0.
(12.210)
Here, use has been made of Equations (12.186) and (12.190). Furthermore [cf.,
Equations (12.159) and (12.160)],
=
,
sin
(12.211)
1
+
.
sin
(12.212)
1
(sin cos )
(cos ) ,
(12.213)
D =
sin
1
(sin cos ) +
(cos ) ,
(12.214)
D =
sin
where
1 (, , t)
= cos (, , t),
sin
(, , t) = 0.
(12.215)
(12.216)
Furthermore,
cos =
1
sin
(12.217)
cos =
1
+
.
sin
(12.218)
370
Substitution of Equations (12.211), (12.212), (12.217), and (12.218) into Equations (12.187) and (12.188) yields
A
1 B
+
= 0,
sin
(12.219)
B
1 A
= 0,
sin
(12.220)
where
..
..
A = 2
gd
( F),
a2
B = 2 .
(12.221)
(12.222)
(12.223)
D B = 0.
(12.224)
(12.225)
B(, , t) = 0.
(12.226)
We have already seen that the solution of Equation (12.205), subject to the boundary condition (12.204), is P = constant. It follows that the solution of Equation (12.223), subject to the boundary condition (12.225), is A = constant. Analogous arguments reveal that the solution of Equation (12.224), subject to the boundary condition (12.226), is B = 0. Hence, we deduce that the Laplace tidal equations,
(12.186)(12.188), are equivalent to the following set of equations:
..
2
D = ,
(12.227)
gd
( F ) = constant,
a2
(12.228)
2 = 0,
(12.229)
..
.
371
(12.230)
1 g(, )
= 0,
sin
(12.231)
f (, ) = 0,
(12.232)
g(, ) = 0.
(12.233)
or
1 g f
g f
+
d
=
g D f d =
f D g d.
sin2
(12.234)
(12.235)
(12.236)
(12.237)
Hence, we deduce that the r are real. It follows that we can choose the r to be real
functions. Equations (12.234) and (12.235) also yield
2
2
1 r
r
2
r D r d = r
r d =
+ sin d, (12.238)
which implies that the r are positive. Integration of Equation (12.235), subject to
the boundary condition (12.236), gives
r d = 0.
(12.239)
r
372
(12.240)
(12.241)
It follows that
s r d = 0
(12.242)
This argument can be extended to multiple linearly independent solutions corresponding to the same eigenvalue. Hence, we conclude that it is possible to choose
the r such that they satisfy the orthonormality condition
r
r s d = s
r s d = r s .
(12.244)
(12.245)
(12.246)
r=1,
(12.248)
(12.249)
373
Using analogous arguments to those employed previously, we can show that the r
are real and positive, and that the r can be chosen so as to satisfy the orthonormality
constraint
r s d = s
r s d = r s .
(12.250)
r
(12.251)
(12.252)
r=1,
F r d.
(12.253)
cos r
,
sin
Y = cos
r
,
(12.254)
(12.255)
P = r ,
(12.256)
r
(12.257)
Q=
1
r
r
D r =
cos
cos
,
sin
1
r
1
r
sin cos
+
cos
,
D r =
sin
sin
(12.258)
(12.259)
and
r
1 r (, )
= cos
,
sin
r (, ) = 0,
(12.260)
(12.261)
374
as well as
cos r
1 r
= r
,
sin
sin
(12.262)
r
1 r r
=
+
.
sin
(12.263)
cos
Let
r
,
cos r
,
Y=
sin
X = cos
(12.264)
(12.265)
P = r ,
Q=
(12.266)
r
(12.267)
1
r
1
r
D r =
sin cos
+
cos
,
sin
sin
1
r
r
cos
cos
,
D r =
sin
(12.268)
(12.269)
and
cos r (, )
1 r (, )
=
,
sin
sin
r (, ) = 0,
(12.270)
(12.271)
as well as
1 r
r
= r
,
sin
(12.272)
r
cos r
1 r
=
+
.
sin
sin
(12.273)
cos
r, s = s
r
s d =
r D s d
r s
1 r s
+
d,
sin2
(12.274)
375
where use has been made of Equations (12.235) and (12.236). It follows from Equations (12.262) and (12.263) that
cos r s r s
r, s =
d
sin
1 r s r s
d.
(12.275)
sin
However, the second term on the right-hand side of the previous equation integrates
to zero with the aid of Equation (12.261). Hence, we are left with
cos r s r s
d.
(12.276)
r, s =
sin
Let
r, s = s
r s d =
r D s d
r s
1 r s
+
d,
sin2
(12.277)
where use has been made of Equations (12.235) and (12.236). It follows from Equations (12.272) and (12.273) that
r s
1 r s
+
cos
r, s =
d
sin2
1 r s r s
d.
(12.278)
sin
However, the second term on the right-hand side of the previous equation integrates
to zero with the aid of Equation (12.271). Hence, we are left with
r s
1 r s
+
cos
d.
(12.279)
r, s =
sin2
Let
r, s = s
r s d =
r D s d
r s
1 r s
+
d,
sin2
(12.280)
where use has been made of Equations (12.248) and (12.261). It follows from Equations (12.262) and (12.263) that
r s
1 r s
+
cos
r, s =
d
sin2
1 r s r s
d.
(12.281)
+
sin
376
However, the second term on the right-hand side of the previous equation integrates
to zero with the aid of Equation (12.249). Hence, we are left with
r, s =
cos
r s
1 r s
+
d.
sin2
(12.282)
Finally, let
r, s = s
r s d =
r D s d
r s
1 r s
+
d,
sin2
(12.283)
where use has been made of Equations (12.248) and (12.271). It follows from Equations (12.272) and (12.273) that
cos r s r s
=
d
sin
1 r s r s
d.
+
sin
r, s
(12.284)
However, the second term on the right-hand side of the previous equation integrates
to zero with the aid of Equation (12.249). Hence, we are left with
r, s =
cos r s r s
d.
sin
(12.285)
Incidentally, the r,s , r,s , r,s , and r,s are known collectively as gyroscopic
coecients (Proudman 1916).
pr (t) r (, ).
(12.286)
r=1,
r=1,
pr (t) r (, ).
(12.287)
377
sin
r=1,
r
1 r
pr
=
pr
,
sin
r=1,
r pr r .
=
(12.288)
(12.289)
(12.290)
r=1,
Let
A r = r
r d =
D r d
1 r
r
+
d,
sin2
(12.291)
(12.292)
where use has been made of Equations (12.235) and (12.236). It follows from Equations (12.217) and (12.218) that
r
r
Ar =
+
cos
d
sin
1 r r
d.
(12.293)
sin
However, the second term on the right-hand side of the previous equation integrates
to zero with the aid of Equation (12.216). Hence, we are left with
r
r
+
cos
d.
(12.294)
Ar =
sin
Finally, Equations (12.276), (12.282), (12.288), and (12.289) can be combined with
the previous equation to give
3
4
r d =
r, s p s + r, s ps .
(12.295)
r
s=1,
Let
Ar = r
r d =
D r d
r
1 r
+
d,
sin2
(12.296)
where use has been made of Equations (12.216) and (12.248). It follows from Equa-
378
r
r
Ar =
+
cos
d
sin
1 r r
d.
+
sin
(12.297)
However, the second term on the right-hand side of the previous equation integrates
to zero with the aid of Equation (12.249). Hence, we are left with
r
r
+
cos
Ar =
d.
(12.298)
sin
Finally, Equations (12.279), (12.285), (12.288), and (12.289) can be combined with
the previous equation to give
3
4
r, s p s + r, s ps .
(12.299)
r d =
r
s=1,
Operating on Equations (12.228) and (12.229) with r ( ) r d and with
r ( ) r d, respectively, yields the so-called Proudman equations (Proudman
1916):
..
pr + 2
(r, s p s + r, s ps ) +
s=1,
..
pr + 2
gd
gd
r pr = 2 Fr ,
a2
a
(r, s p s + r, s ps ) = 0,
(12.300)
(12.301)
s=1,
where
F r = r
F r d.
(12.302)
Here, use has been made of Equations (12.240), (12.244), (12.250), (12.286), (12.287),
(12.290), (12.295), and (12.299). Finally, it follows from Equations (12.138), (12.189),
and (12.290) that
r, s p s ,
(12.303)
F r = Pr
s=1,
where
P r = r
and
r, s =
(1 + k2 h2 ) 2 r d,
(12.304)
(12.305)
n=1, m=0,n
with
n,mr = r
Pnm (cos ) e i m r (, ) d.
(12.306)
379
Here,
Pnm (x) = bnm Pnm (x),
where
bnm
2 n + 1 (n m)!
=
4 (n + m)!
(12.307)
1/2
.
(12.308)
(12.309)
Consider the response of the ocean to a particular harmonic of the tide generating
potential for which
m
2 (, , t) = j P2 j (cos ) e i (m j + j t) .
(12.310)
(i f j r, s 1 r, s ) p s + i f j r, s ps
s=1,
m
=
and
f j2 pr + i
1 (1 + k2 h2 ) j 2, rj
mj
b2
( f j r, s p s + f j r, s ps ) = 0,
(12.311)
(12.312)
s=1,
380
/2
cos(m ) cos(m ) d =
m = m = 0
m = m 0 .
m m
(12.313)
(12.314)
nm = n (n + 1),
(12.315)
and
where
anm =
(4/nm )1/2
(2/nm )1/2
m>0
.
m=0
(12.316)
Likewise, the solutions to the eigenvalue problem (12.248) and (12.249), subject to
the orthonormality constraint (12.250), is such that r nm and r nm , where
nm (, ) = anm Pnm (cos ) sin(m ),
(12.317)
nm = n (n + 1).
(12.318)
and
nm nm
cos nm nm
d,
sin
nm nm
1 nm nm
m, m
d,
+
n, n =
cos
sin2
nm nm
1 nm nm
m, m
d,
+
n, n =
cos
sin2
nm nm
cos nm nm
m, m
n, n =
d,
sin
n,m,nm
(12.319)
(12.320)
(12.321)
(12.322)
m
n,m,n
m, m
n,
n
381
nm nm
nm nm
d d,
(12.323)
0
1
1
nm nm
1 nm nm
2
d d,
+
=
(1 )
1 2
0
1
(12.324)
1
m
m
nm n
1 nm n
d d, (12.325)
(1 2 )
+
=
1 2
0
1
1
nm nm
nm nm
d d,
=
(12.326)
0
1
m + m odd
.
m + m even
(12.327)
m, m
Hence, if m + m is even then the n,
n are zero. Otherwise, we obtain
n,m,nm
cnm cnm
m, m
n,
n
cnm cnm
m, m
n,
n
cnm cnm
2 m dPnm
dPnm m
m Pn
+ m 2
Pn d,
(12.328)
d
d
1
1
dPnm dPnm
m 2
2m
2
m
m
+
= 2
P P d, (12.329)
(1 )
d d
m m 2 1
1 2 n n
1
m dPnm
dPnm m
2 m m
+
P d,
= 2
(12.330)
P
d n
m m 2 1 n d
=
2
m 2 m 2
as well as
m
m , m
= n
n,m,n
, n ,
(12.331)
(12.332)
where
Let
In,m,nm =
Pnm () Pnm () d.
(12.333)
In,m,nm =
m
n,p
nm ,p
p=0,pmax p =0,pmax
(12.334)
382
and
pmax =
(1)m+p (n + m)!
,
2 m+2p (m + p)! p! (n m 2p)!
(12.335)
(n m)/2
(n m 1)/2
(12.336)
n m even
,
n m odd
cnm cnm
m
n,m,n
n (n + 1) + m (m + 1)
n (n + 1) n (n + 1) + m
2m
=
In,m,nm
m + 1
m 2 + m 2
1
+
(12.337)
I m, m +2 ,
m + 1 n, n
cnm cnm
(m 2
2m
(n 1) (n + 1) (n + m ) In,m,nm 1
2
m ) (2 n + 1)
+n (n + 2) (n m + 1) In,m,nm +1 ,
m, m
n,
n
cnm cnm
(12.338)
2 m m m, m
I .
m 2 m 2 n, n
(12.339)
m,m
m
n,n
According to Equation (12.306), we can also write n,r
, where
m,m
m,m
m,m
n,n
= (anm )1 n,n
,
+ i n,n
and
(12.340)
n,m,nm = n n m m .
(12.341)
2m
=
nm cnm cnm In,m,nm ,
2
2
m m
(12.342)
(12.343)
m,m
n,m,nm = (1 + kn hn ) n (anm )2 n,n
(1 + kn hn ) n (anm )2 nm ,n,m
(12.344)
383
odd
m
(12.345)
even
m
(12.346)
m m
+ 1 (n,m,nm n,m,nm ) pnm f j n,m,n
pn
=
and
n =1, m =0,n
1 (1 + k2 h2 ) j
m
c2 j
m ,m
m ,m
2, nj + 2, nj ,
m, m
m
m, m pnm = 0,
f j2 n,m,nm + f j n,
pn
+ fj
n
n, n
(12.347)
(12.348)
respectively. Here,
n,m,nm = i n+n n,m,nm ,
(12.349)
m, m
i n+n +1 n,
n ,
=i
n+n +1
n,m,nm ,
n,m,nm = i n+n n,m,nm ,
(12.351)
(12.352)
(12.353)
(12.354)
(12.355)
(12.350)
(12.356)
By symmetry, n,m,nm and n,m,nm are only non-zero when m + m is even, and n + n is
m, m
m, m
m, m
even; n,m,nm , n,
n , n, n , and n, n are only non-zero when m + m is odd, and
m,
m
m,
m
n + n is odd; and n, n and n, n are only non-zero when m + m is odd, and n + n
is even. It follows that all quantities appearing in Equations (12.347) and (12.348)
384
are real. Once we have solved these equations to obtain the pnm (which is a relatively
straightforward numerical task), we can reconstruct the tidal elevation as follows:
(, , t) = c (, ) cos( j t) + s (, ) sin( j t),
(12.357)
where
c (, ) =
n
even
nm m
n (, ) cos(n /2),
(12.358)
n=1, m=0,n
s (, ) =
n
odd
nm m
n (, ) sin(n /2),
(12.359)
n=1, m=0,n
and
nm = nm pnm .
(12.360)
(12.361)
where
t(, ) (, )
=
,
T
2
(12.362)
1 s (, )
(, ) = m j + tan
.
2
c (, )
(12.363)
385
12.22 Exercises
12.1 Consider a tidal wave whose wavelength is very much less than the radius of
the Earth. This corresponds to the limit n
1, where n is an azimuthal mode
number. Suppose, however, that the ocean depth, d, is allowed to vary with
386
Figure 12.4
Relative amplitude, ||/3 , of the ( j = 3) M f long-period tide in a hemispherical
ocean.
Figure 12.5
Phase-lag, , of the ( j = 3) M f long-period tide in a hemispherical ocean.
387
Figure 12.6
Relative amplitude, ||/|6 |, of the ( j = 6) K1 diurnal tide in a hemispherical ocean.
Figure 12.7
Phase-lag, , of the ( j = 6) K1 diurnal tide in a hemispherical ocean.
388
Figure 12.8
Relative amplitude, ||/8 , of the ( j = 8) M2 semi-diurnal tide in a hemispherical
ocean.
Figure 12.9
Phase-lag, , of the ( j = 8) M2 semi-diurnal tide in a hemispherical ocean.
389
position. Show that, in this limit, the Laplace tidal equations are written
(v d)
=
(sin u d) +
,
t
a sin
g
u
2 cos v =
t
a
v
+ 2 cos u =
t
a
( ),
( ),
sin
where
= (1 + k2 h2 ) 2 .
12.2 Consider short-wavelength tidal waves in a region of the ocean that is suciently localized that it is a good approximation to treat cos and sin as constants. We can define local Cartesian coordinates, x, y, z, such that dx = a d,
dy = a sin d, and dz = r a. It follows that the x-axis is directed southward, the y-axis is directed eastward, and the z-axis is directed vertically
upward. Show that, when expressed in terms of this local coordinate system,
the Laplace tidal equations derived in the previous exercise reduce to
(u d) (v d)
,
t
x
y
u
( )
f v = g
,
t
x
( )
v
+ f u = g
,
t
y
where
f = 2 cos
is a constant. In fact, the previous equations are the linearized equations of
motion of a body of shallow water confined to a tangent plane that touches
the Earth at the angular coordinates , . This plane is known as the f -plane
because, as a consequence of the Earths diurnal rotation, it rotates about ez
at the angular velocity f .
12.3 Demonstrate that the set of equations derived in the previous exercise can be
written in the coordinate-free (in the x-y plane) form
= (d v),
t
v
+ f ez v = g ( ),
t
where v = u e x + v ey . Let = ez v and = v. Assuming that d and
390
2
2
2
2
+
f
g
d
=
g
,
2
t
t
where 2 denotes a two-dimensional Laplacian (in the x-y plane).
k x i f ky
u0 = g
0 ,
2 f 2
ky + i f k x
v0 = g
0 , ,
2 f 2
and
where c =
2 = f 2 + c 2 k 2,
d g. This type of wave is known as a Poincare wave.
where c =
(x, y, t) = Z1 (x + c t, y) + Z2 (x c t, y),
d g, and
Z1 (x + c t, y) = Z1 (x + c t, 0) e+s y/l ,
Z2 (x c t, y) = Z2 (x c t, 0) es y/l .
Here, Z1 (x, 0) and Z2 (x, 0) are arbitrary functions, l = d g/| f |, and s =
sgn( f ). These solutions are known as Kelvin waves. Deduce that Kelvin
waves propagate along coastlines, at the speed c, in such a manner as to keep
the coastline to the right of the direction of propagation in the Earths northern
hemisphere, and to the left in the southern hemisphere.
391
12.6 We can take into account the latitude dependence of the parameter f by writing f = f0 x, where f0 = 2 cos and = 2 sin /a. Let us assume that
the ocean is of constant depth, d. Furthermore, let us search for an almost incompressible, free solution of the Laplace tidal equations which is such that
u /y and v /x. By eliminating from the final two f -plane
equations, show that
0.
+
2
t
y
By searching for a wavelike solution of the previous equation of the form
= 0 e i ( tkr) , deduce that
k 2 + ky = 0.
This is the dispersion relation of a so-called Rossby wave. Demonstrate that
Rossby waves always travel with a westward component of phase velocity.
Finally, show that it is reasonable to neglect compression provided that
| f0 |.
392
13
Equilibrium of Compressible Fluids
13.1 Introduction
In this chapter, we investigate the equilibria of compressible fluids, such as gases.
As is the case for an incompressible fluid (see Chapter 2), a compressible fluid in
mechanical equilibrium must satisfy the force balance equation
0 = p + ,
(13.1)
where p is the static fluid pressure, the mass density, and the gravitational potential energy per unit mass. In an ideal gas, the relationship between p and is
determined by the energy conservation equation, (1.89), which can be written
D p
M 2 p
=
.
1 Dt
R
(13.2)
Here, is the ratio of specific heats, the thermal conductivity, M the molar mass,
and R the molar ideal gas constant. Note that the viscous heat generation term has
been omitted from the previous equation (because it is zero in a stationary gas). The
limits in which the left- and right-hand sides of Equation (13.2) are dominant are
termed the adiabatic and isothermal limits, respectively. In the isothermal limit, in
which thermal transport is comparatively large, so that Equation (13.2) can only be
satisfied when 2 (p/) 0, the temperature (recall that T p/ in an ideal gas)
distribution in the gas becomes uniform, and the pressure and density are consequently related according to the isothermal gas law,
p
= constant.
(13.3)
On the other hand, in the adiabatic limit, in which thermal transport is negligible, so
that Equation (13.2) can only be satisfied when D/Dt(p/ ) 0, the pressure and
density are related according to the adiabatic gas law,
p
= constant.
(13.4)
393
394
(13.6)
(13.7)
p(z) = p0 exp(z/H),
(13.8)
Hence, we obtain
where p0 10 N m
5
RT
gM
(13.9)
is known as the isothermal scale height of the atmosphere. Using the values T =
273 K (0 C), M = 29 103 kg, and g = 9.8 m s2 , which are typical of the Earths
atmosphere (at ground level), as well as R = 8.315 J mol1 K1 , we find that H =
7.99 km. Equations (13.6) and (13.8) yield
(z) = 0 exp(z/H),
(13.10)
where 0 = p0 /(g H) is the atmospheric mass density at z = 0. According to Equations (13.8) and (13.10), in an isothermal atmosphere, the pressure and density both
decrease exponentially with increasing altitude, falling to 37% of their values at
ground level when z = H, and to only 5% of these values when z = 3 H.
395
convection currents. When a packet of air ascends, under the influence of such currents, the diminished pressure at higher altitudes causes it to expand. Because this
expansion generally takes place far more rapidly than heat can diuse into the packet,
the work done against the pressure of the surrounding gas, as the packet expands,
leads to a reduction in its internal energy, and, hence, in its temperature. Assuming
that the atmosphere is in a continually mixed state, while remaining in approximate
vertical force balance (such a state is known as a convective equilibrium), and that
the eect of heat conduction is negligible (because the mixing takes place too rapidly
for thermal diusion to aect the temperature), we would expect the adiabatic gas
law, (13.4), to oer a much more accurate description of the relationship between
atmospheric pressure and density than the isothermal gas law, (13.3).
Let p = p/p0 , = /0 , and T = T/T 0 , where p0 , 0 , and T 0 are the pressure,
mass density, and temperature of the atmosphere, respectively, at ground level. The
adiabatic gas law, (13.4), can be combined with the ideal gas equation of state, (13.6),
to give
(13.11)
p = = T /(1) .
The isothermal scale height of the atmosphere is conveniently redefined as [cf., Equation (13.9)]
p0
R T0
=
H=
.
(13.12)
gM
g 0
Equations (13.5), (13.11), and (13.12) yield
d p
= p 1/ ,
dz
where z = z/H, or, from Equation (13.11),
dT
1
=
.
dz
1 z
T (z) = T 0 1
.
H
(13.13)
(13.14)
(13.15)
/(1)
1 z
,
(13.16)
p(z) = p0 1
H
1/(1)
1 z
.
(13.17)
(z) = 0 1
H
396
(13.19)
397
(13.20)
(13.22)
where (r) is the gravitational potential energy per unit mass, and d/dr the radial
gravitational acceleration. The force balance criterion (13.1) yields
d
dp
+
= 0,
dr
dr
(13.23)
where p(r) is the pressure. The previous three equations can be combined to give
1 d r 2 dp
= 4 G .
(13.24)
r 2 dr dr
In order to make any further progress, we need to determine the relationship
between the Suns internal pressure and density. Unfortunately, this relationship is
ultimately controlled by energy transport, which is a very complicated process in a
star. In fact, a stars energy is ultimately derived from nuclear reactions occurring
deep within its core, the details of which are extremely complicated. This energy is
398
then transported from the core to the outer boundary via a combination of convection and radiation. (Conduction plays a much less important role in this process.)
Unfortunately, an exact calculation of radiative transport requires an understanding
of the opacity of stellar material, which is an exceptionally dicult subject. Finally,
once the energy reaches the boundary of the star, it is radiated away. The following
ingenious model, due to Eddington (Eddington 1926), is appropriate to a star whose
internal energy transport is dominated by radiation. This turns out to be a fairly good
approximation for the Sun. The main advantage of Eddingtons model is that it does
not require us to know anything about stellar nuclear reactions or opacity.
Now, the temperature inside the Sun is suciently high that radiation pressure
cannot be completely neglected with respect to conventional gas pressure. In other
words, we must write the solar equation of state in the form
p = pg + pr ,
where
pg =
kT
mp
(13.25)
(13.26)
is the gas pressure (modeling the plasma within the Sun as an ideal gas of free electrons and ions), and (Chandrasekhar 1967)
pr =
1
T 4
3
(13.27)
the radiation pressure (assuming that the radiation within the Sun is everywhere in
local thermodynamic equilibrium with the plasma). Here, T (r) is the Suns internal
temperature, k the Boltzmann constant, m p the mass of a proton, and the relative
molecular mass (i.e., the ratio of the mean mass of the free particles making up the
solar plasma to that of a proton). Note that the electron mass has been neglected
with respect to that of a proton. Furthermore, = 4 /c, where is the StefanBoltzmann constant, and c the velocity of light in a vacuum. Incidentally, in writing
Equation (13.26), we have expressed M/R in the equivalent form m p /k.
Let
pg = (1 ) p,
(13.28)
pr = p,
(13.29)
where the parameter is assumed to be uniform. In other words, the ratio of the radiation pressure to the gas pressure is assumed to be the same everywhere inside the
Sun. This fairly drastic assumption turns outperhaps, somewhat fortuitously (Mestel 1999)to lead to approximately the correct internal pressure-density relation for
the Sun. In fact, Equations (13.26)(13.29) can be combined to give
p = K 4/3 ,
where
1/3
3
.
K =
4
4
m p (1 )
(13.30)
(13.31)
399
, y 1
Figure 13.1
The functions () (solid) and y() (dashed).
It can be seen, by comparison with Equation (13.4), that the previous pressuredensity relation takes the form of an adiabatic gas law with an eective ratio of
specific heats = 4/3. Note, however, that the actual ratio of specific heats for a
fully ionized hydrogen plasma, in the absence of radiation, is = 5/3. Hence, the
4/3 exponent, appearing in Equation (13.30), is entirely due to the non-negligible
radiation pressure within the Sun.
Let T c = T (0), c = (0), and pc = p(0), be the Suns central temperature,
density, and pressure, respectively. It follows from Equation (13.30) that
pc = K 4/3
c ,
(13.32)
pc m p
(1 ) .
c k
(13.33)
Suppose that
T
= ,
Tc
(13.34)
400
log10 (T [K])
= 3,
c
p
= 4.
pc
(13.35)
(13.36)
Moreover, it is clear, from the previous expressions, that = 1 at the center of the
Sun, r = 0, and = 0 at the edge, r = R (say), where the temperature, density, and
pressure are all assumed to fall to zero. Suppose, finally, that
r = a ,
(13.37)
a=
K
G c2/3
1/2
.
(13.38)
Thus, the center of the Sun corresponds to = 0, and the edge to = 1 (say), where
(1 ) = 0, and
R = 1 a.
(13.39)
Equations (13.35)(13.38) can be used to transform the equilibrium relation (13.24)
401
Figure 13.3
Solar mass density versus mass fraction obtained from the Eddington Solar Model
(solid) and the Standard Solar Model (dashed).
into the non-dimensional form
1 d 2 d
= 3 .
2 d
d
(13.40)
Moreover, Equation (13.21) can be integrated, with the aid of Equations (13.35),
(13.37), and (13.40), and the physical boundary condition m(0) = 0, to give
m = 4 c a 3 y,
(13.41)
where
d
.
(13.42)
d
Equation (13.40) is known as the Lane-Emden equation (of degree 3), and can, unfortunately, only be solved numerically (Chandrasekhar 1967). The appropriate solution
takes the form = 1 2 /6 + O( 4 ) when 1, and must be integrated to = 1 ,
where (1 ) = 0. Figure 13.1 shows (), and the related function y(), obtained via
numerical methods. Note that 1 = 6.897, and y1 = y(1 ) = 2.018.
According to Equation (13.41), the solar mass, M = m(R), can be written
y() = 2
M = 4 c a 3 y1 ,
(13.43)
= 2,
(1 )4
(13.44)
402
where
= 2
and
M0 =
M
,
M0
4 1/2
k
3
1
4 y1 = 3.586 1031 kg.
( G)3 m p
(13.45)
(13.46)
G M2/3
.
(4 y1 )2/3
(13.47)
According to Equations (13.44) and (13.45), the ratio, /(1 ), of the radiation
pressure to the gas pressure in a radiative star is a strongly increasing function of the
stellar mass, M , and mean molecular weight, . In the case of the Sun, for which
1, Equation (13.44) can be inverted to give the approximate solution
= 2
1 4
+ O( 6 ).
4
(13.48)
Using the observed solar mass M = 1.989 1030 kg, and the value = 0.68 (which
represents the best fit to the Standard Solar Model mentioned in the following), we
find that = 6.58 104 . In other words, the radiation pressure inside the Sun is
only a very small fraction of the gas pressure. This immediately implies that radiative
energy transport is much less ecient than convective energy transport. Indeed, in
regions of the Sun in which convection occurs, we would expect the convective transport to overwhelm the radiative transport, and so to drive the local pressure-density
relation toward an adiabatic law with an exponent 5/3. Fortunately, convection only
takes place in the Suns outer regions, which contain a minuscule fraction of its mass.
Equations (13.32), (13.33), (13.39), (13.43), and (13.47) yield
Tc =
G M m p
1
= 1.34 107 K,
(1 )
4 y1
Rk
(13.49)
c =
13 M
= 7.63 104 kg m3,
4 y1 R 3
(13.50)
pc =
14
16 y12
G M2
= 1.24 1016 N m2 ,
R4
(13.51)
where the solar radius R has been given the observed value 6.960 108 m. The actual values of the Suns central temperature, density, and pressure, as determined by
the so-called Standard Solar Model (SSM),1 which incorporates detailed treatments
of nuclear reactions and opacity, are T c = 1.58 107 K, c = 15.6 104 kg m3 ,
and pc = 2.38 1016 N m2 , respectively. It can be seen that the values of T c , c ,
1 http://en.wikipedia.org/wiki/Standard\_Solar\_Model
403
17
log10 (p [N m2])
16
15
14
13
12
11
10
Figure 13.4
Solar pressure versus mass fraction obtained from the Eddington Solar Model (solid)
and the Standard Solar Model (dashed).
and pc obtained from the Eddington model lie within a factor of two of those obtained from the much more accurate SSM. Figures 13.2, 13.3, and 13.4 show the
temperature, density, and pressure profiles, respectively, obtained from the SSM 2
and the Eddington model. The profiles are plotted as functions of the mass fraction,
m(r)/M = y()/y1 , where = r/a. It can be seen that there is fairly good agreement
between the profiles calculated by the two models. Finally, Figure 13.5 compares
the ratio, /(1 ), of the radiation pressure to the gas pressure obtained from the
SSM and the Eddington model. Recall, that it is a fundamental assumption of the
Eddington model that this pressure ratio is uniform throughout the Sun. In fact, it
can be seen that the pressure ratio calculated by the SSM is not spatially uniform.
On the other hand, the spatial variation of the ratio is fairly weak, except close to the
edge of the Sun, where convection sets in, and the Eddington model, thus, becomes
invalid. We conclude that, despite its simplicity, the Eddington solar model does a
remarkably good job of accounting for the Suns internal structure.
404
log10[/(1 )]
2.5
3.5
0.2
0.4
0.6
m/M
0.8
Figure 13.5
Ratio of radiation pressure to gas pressure calculated from the Eddington Solar
Model (solid) and the Standard Solar Model (dashed).
13.6 Exercises
13.1 Prove that the fraction of the whole mass of an isothermal atmosphere that
lies between the ground and a horizontal plane of height z is
1 ez/H .
Evaluate this fraction for z = H, 2 H, 3 H, respectively.
13.2 If the absolute temperature in the atmosphere diminishes upwards according
to the law
T
z
=1 ,
T0
c
where c is a constant, show that the pressure varies as
$
p
z %c/H
= 1
.
p0
c
13.3 If the absolute temperature in the atmosphere diminishes upward according
to the law
1
T
,
=
T0 1 + z
405
p
1 z2
z
= exp
.
p0
H 2 H
13.4 Show that if the absolute temperature, T , in the atmosphere is any given function of the altitude, z, then the vertical distribution of pressure in the atmosphere is given by
T 0 z dz
p
.
=
ln
p0
H 0 T
13.5 Show that if the Earth were surrounded by an atmosphere of uniform temperature then the pressure a distance r from the Earths center would be
2
p
a 1 1
= exp
,
p0
H r a
where a is the Earths radius.
13.6 Show that if the whole of space were occupied by air at the uniform temperature T then the densities at the surfaces of the various planets would be
proportional to the corresponding values of
exp
$g M a%
RT
where a is the radius of the planet, and g its surface gravitational acceleration.
13.7 Prove that in an atmosphere arranged in horizontal strata the work (per unit
mass) required to interchange two thin strata of equal mass without disturbance of the remaining strata is
p1
p2
1 1
,
2 11
1
1 2
where the suxes refer to the initial states of the two strata. Hence, show that
for stability the ratio p/ must increase upwards.
13.8 A spherically symmetric star is such that m(r) is the mass contained within radius r. Show that the stars total gravitational potential energy can be written
in the following three alternative forms:
U =
0
1
Gm
dm =
r
2
0
dm = 3
p dV
0
Here, M is the total mass, R the radius, (r) the gravitational potential per
unit mass (defined such that 0 as r ), p(r) the pressure, and
dV = 4 r 2 dr.
406
13.9 Suppose that the pressure and density inside a spherically symmetric star are
related according to the polytropic gas law,
p = K (1+n)/n ,
where n is termed the polytropic index. Let = c n , where c is the central
mass density. Demonstrate that satisfies the Lane-Emden equation
1 d 2 d
= n ,
d
2 d
where r = a , and
1/2
(n + 1) K (n1)/n
=
c
.
4 G
Show that the physical solution to the Lane-Emden equation, which is such
that (0) = 1 and (1 ) = 0, for some 1 > 0, is
=1
for n = 0,
=
for n = 1, and
2
6
sin
(1 + 2 /3)1/2
for n = 5. Determine the ratio of the central density to the mean density
in all three cases. Finally, demonstrate that, in the general case, the total
gravitational potential energy can be written
U=
3 G M2
,
5n R
5 G M2
,
4 R
where M is the mass of the star. Finally, show that the total gravitational
potential energy of the star can be written
U=
26 G M2
.
35 R
14
One-Dimensional Compressible Inviscid Flow
14.1 Introduction
This chapter investigates one-dimensional, compressible, inviscid flow. Flow is said
to be one-dimensional when the fluid properties only depend on a single Cartesian
coordinate. Flow is said to be compressible when there is a significant variation in
the mass density along a given streamline. Generally speaking, compressible flow is
much more common in gases than in liquids. As described in Section 1.17, subsonic
gas dynamics (i.e., dynamics in which the typical flow speed is much less than the
propagation speed of sound waves in the gas) is essentially incompressible, and is
governed by the same equations that govern the incompressible flow of liquids. On
the other hand, sonic gas dynamics (i.e., dynamics in which the typical flow speed is
comparable with the sound speed) exhibits significant dierences to subsonic dynamics. Hence, this chapter will concentrate on sonic gas dynamics. More information
on this subject can be found in Liepmann & Roshko 1957, Hughes & Brighton 1999,
and Milne-Thomson 2011.
408
(14.1)
where the constant R is termed the specific gas constant. This quantity is related to
the molar gas constant, R, introduced in Section 1.15, according to R = R/M, where
M is the molar massthat is, the mass of NA = 6.022 1023 gas molecules. Here,
NA is Avogadros number (Reif 1965).
For all gases, whether in mean motion, or not, there exists an internal energy
function, E, that is independent of the mean motion, and dependent only on the
variables of state, p, , and T . Moreover, this function is such that when a small
quantity of heat, q, is added to the system (Reif 1965),
q = dE + p dV.
(14.2)
Thus, the quantity dE is the excess of the energy supplied over the mechanical work
done by the pressure. Of course, internal energy is an extensive quantity. The corresponding intensive quantity is the specific internal energy, E = E/m. In an ideal
gas, the specific internal energy, E, is a function of the absolute temperature, T , alone
(Reif 1965). It follows that
dE = k(T ) dT,
(14.3)
and Equation (14.2) becomes
q = m (k dT + p dv).
The heat capacity at constant volume of the gas is defined
q
.
C V =
dT v
(14.4)
(14.5)
(14.6)
C V
= k.
m
(14.7)
Finally, the molar specific heat capacity at constant volume, which is the heat capacity of Avogadros number of gas molecules, takes the form
cV = M C V .
(14.8)
cV
dT.
M
(14.9)
409
(14.10)
v d p + p dv = R dT.
(14.11)
E = CV T =
Equation (14.1) yields
Combining the previous equation with Equations (14.4) and (14.7), we obtain
!
"
q = m (CV + R) T v d p .
(14.12)
Now, the heat capacity at constant pressure of the gas is defined
q
C p =
dT p
(14.13)
(14.14)
C p
= CV + R.
m
(14.15)
Finally, the molar specific heat capacity at constant pressure, which is the heat capacity of Avogadros number of gas molecules, takes the form [cf., Equation (1.86)]
c p = M C p = cV + R.
(14.16)
The molar specific heat capacity at constant pressure of an ideal gas is a constant,
independent of the temperature.
The enthalpy of a gas is defined (Reif 1965)
H = E + p V.
(14.17)
cp
T,
M
(14.18)
where use has been made of Equations (14.1), (14.10), (14.15), and (14.16).
In the preceding analysis, we denoted a small quantity of heat by q, rather than
dQ, because there is, in general, no function Q of which q is an exact dierential.
However, we can write (Reif 1965)
q = T dS ,
(14.19)
410
p
dv
dT
dT
dS = m CV
+ dv = m CV
+R
.
T
T
T
v
(14.20)
(14.21)
Hence, we obtain
dv
dp
dS = m C p
+ CV
,
v
p
(14.22)
where use has been made of Equation (14.15). In an ideal gas, CV and C p are constants, independent of T , so the previous expression can be integrated to give
S = m CV ln(p v ) = m CV ln
p
,
(14.23)
where
=
Cp
CV
(14.24)
is the constant ratio of specific heats. Obviously, the specific entropy is defined
S=
p
S
= CV ln .
m
(14.25)
(14.26)
dH = V d p + T dS ,
(14.27)
or, equivalently,
dE = p dv + T dS,
dH = v d p + T dS.
(14.28)
(14.29)
411
D p
= 0,
Dt
(14.30)
(14.31)
(14.32)
where v is the flow velocity, and the potential energy per unit mass. It is clear from
Equations (14.25) and (14.32) that the specific entropy is constant along a streamline,
but not necessarily the same constant on dierent streamlines. Such flow is said to
be isentropic. From Equation (14.29), isentropic flow is characterized by
dH =
dp
(14.33)
+u
+
= 0,
t
x
x
u 1 p
u
+u
+
= 0.
t
x x
(14.34)
(14.35)
Equation (14.32) implies that the flow is isentropic. In other words, p/ = constant.
Thus, it follows that
p
= c2 ,
(14.36)
x
x
where
c2
dp p
=
.
d
(14.37)
412
u
+u
+
= 0,
t
x
x
u c 2
u
+u
+
= 0.
t
x
x
(14.38)
(14.39)
Unfortunately, these equations are dicult to solve exactly, because of the presence
of nonlinear terms such as u u/x. (See Example 14.9.)
Let us write u(x, t) = u1 (x, t), (x, t) = 0 + 1 (x, t), and p(x, t) = p0 + p1 (x, t),
where quantities with the subscript 1 are assumed to be much smaller that corresponding quantities with the subscript 0. Thus, we are now considering the evolution
of small-amplitude pressure and density perturbations in a stationary gas of uniform
density and pressure 0 and p0 , respectively. To first order in small quantities, the
previous two equations yield
1 1 u1
+
= 0,
0 t
x
(14.40)
u1 c02 1
+
= 0,
t
0 x
(14.41)
where
c0 =
p0
.
0
(14.42)
The general solution to Equations (14.40) and (14.41) is well known (Fitzpatrick
2013):
1 (x, t)
= F(x c0 t),
0
(14.43)
(14.44)
where use has been made of Equation (14.1). It follows that the speed of sound is
solely determined by the temperature of an ideal gas.
Let us now consider the eect of finite wave amplitude on the propagation of a
sound wave through an isentropic ideal gas. The previous analysis suggest that, in a
413
frame of reference
moving with the local flow velocity, u, a sound wave propagates
at the speed c = p/. Thus, the net wave propagation speed is
c = u + c.
Using the isentropic law, p/ = constant, to eliminate p from c =
deduce that
(1)/2
c = c0
.
0
(14.46)
p/, we
(14.47)
Now, Equations (14.43) and (14.44) suggest that, in the presence of a sound wave
propagating in the +x-direction, the general dierential relationship between the local flow velocity and the density is
du = c
d
.
(14.48)
Making use of Equation (14.47), we can integrate the previous expression to give
(1)/2
2 c0
d
2
u=
=
(c c0 ),
c
1 =
(14.49)
1
0
0
or
c = c0 +
1
u.
2
It follows that
c = c + u = c0 +
or
+1
u,
2
(1)/2
+ 1
c = c0
1
1+
.
1
0
(14.50)
(14.51)
(14.52)
Thus, we conclude that the net wave propagation speed is higher than c0 in regions
of compression (i.e., > 0 ), and lower in regions of rarefaction (i.e., < 0 ). This
implies that a finite-amplitude sound wave distorts as it propagates in such a manner that the regions of compression tend to catch up with the regions of rarefaction,
leading to a monotonic increase with time of the velocity and temperature gradients
in the former regions, and a monotonic decrease in the latter regions. Eventually, the
velocity and temperature gradients in the regions of compression become so large
that it is no longer valid to neglect the eects of viscosity and heat conduction, leading to a local breakdown of the isentropic gas law. Viscosity and heat conduction
prevent any further steepening of the velocity and temperature gradients in the regions of compression, and the wave subsequently propagates without additional distortion. However, it has now eectively been transformed into a shock wave. (See
Section 14.8.)
It is clear, from the previous discussion, that, at finite amplitude, there is an important dierence in the behavior of compression and rarefaction waves propagating
414
through a gas. A compression wave tends to steepen (i.e., the temperature and velocity gradients tend to increase), and eventually becomes a shock wave, in which
case it is no longer isentropic. A rarefaction wave, on the other hand, always remains isentropic, because it tends to flatten, thereby, further reducing the velocity
and temperature gradients. Hence, there are no rarefaction, or expansion, shocks.
1 2
1
v + = Cp T + v 2 +
2
2
(14.54)
is constant along a streamline, where use has been made of Equation (14.18).
Consider a solid object moving through an ideal gas. Generally speaking, there is
at least one stagnation point in front of the object, where the gas comes to rest relative
to it. At this point, the gas is adiabatically compressed, and there is an associated rise
in temperature. In the frame of reference in which the object is at rest, and the
gas a long way from it moves with constant speed and temperature, application of
Bernoullis theorem yields
Cp T +
1 2
v = Cp T0,
2
(14.55)
where T 0 is the stagnation point temperature, and T the temperature at some general
point where the flow speed is v. Thus, the total temperature rise due to adiabatic
compression is
T = T 0 T =
v2
,
2 Cp
(14.56)
where v and T are the asymptotic flow speed and temperature, respectively. It can
be seen that the stagnation temperature rise only depends on the velocity dierence
between the object and the gas, and is independent of the gass density, temperature,
or pressure. Note, however, that a lower molar mass implies a higher specific heat,
and, hence, a smaller stagnation temperature rise.
415
(14.57)
where c = R T is the local sound speed. [See Equation (14.45).] Setting v 2 =
Ma 2 c 2 in Equation (14.55), we obtain
1
1
T
2
= 1 + ( 1) Ma
,
T0
2
(14.58)
where use has been made of Equations (14.15) and (14.24), which imply that
1
R,
1
R.
CP =
1
CV =
(14.59)
(14.60)
Incidentally, the relation (14.58) is valid for any streamline, because the stagnation temperature, T 0 , can be defined, by means of Equation (14.55), even when the
streamline in question does not pass through a stagnation point. We can combine
Equation (14.58) with the isentropic relation, p / = constant along a streamline
(see Section 14.3), as well as the ideal gas law, (14.1), to give
/(1)
T
p
=
= 1+
p0
T0
1/(1)
T
=
= 1+
0
T0
1
( 1) Ma 2
2
1
( 1) Ma 2
2
/(1)
(14.61)
(14.62)
1/(1)
Here, p0 and 0 are the pressure and density, respectively, at the stagnation point. In
principle, the stagnation values, T 0 , p0 , and 0 , can be dierent on dierent streamlines. However, if a solid object moves through a homogeneous ideal gas that is
asymptotically at rest then the stagnation parameters become true constants, independent of the streamline. Such flow is said to be homentropic.
A point where the speed of a steadily flowing ideal gas equals the local speed of
sound, v = c, is termed a sonic point. The sonic temperature, T 1 , pressure, p1 , and
density, 1 , are simply related to the stagnation values. In fact, setting Ma = 1 in
416
2
T1
=
,
T0
+1
/(1)
2
p1
=
,
p0
+1
1/(1)
2
1
=
.
0
+1
(14.63)
(14.64)
(14.65)
1
1 2
T
= 1+
,
(14.66)
Ma 1
T1
+1
/(1)
1 2
p
= 1+
,
(14.67)
Ma 1
p1
+1
1/(1)
1 2
= 1+
.
(14.68)
Ma 1
1
+1
(14.69)
In particular, A u = 1 A1 u1 , where the subscript 1 refers to the sonic point (assuming that such a point exists). Using u/c = Ma and u1 /c1 = 1, we find that
A
1 c1 1
1 $ T 1 %1/2 1
1 u1
=
=
=
,
(14.70)
A1
u
Ma c
Ma T
where use has been made of Equation (14.45). It follows from Equations (14.66) and
(14.68) that
(+1)/[2 (1)]
1
1 2
A
=
.
(14.71)
1+
Ma 1
A1 Ma
+1
417
Ma
Figure 14.1
Simple model of a de Laval nozzle for = 7/5. The solid, dashed, long-dashed,
dash-dotted, and dotted curves show A/A1 , p/p1 , /1 , T/T 1 , and u/u1 , respectively.
Moreover, the previous two equations can be combined with Equation (14.68) to give
1/2
1 2
u
Ma 1
= Ma 1 +
.
u1
+1
(14.72)
Figure 14.1 shows A/A1 , p/p1 , /1 , T/T 1 , and u/u1 , plotted as functions of the
local Mach number, Ma, for an ideal gas with = 7/5. Here, use has been made
of Equations (14.71), (14.67), (14.68), (14.66), and (14.72), respectively. Inspecting
the curves, we can see, somewhat surprisingly, that the cross-sectional area function,
A/A1 , attains a minimum value when Ma = 1. In fact, the figure indicates that,
for subsonic flow (Ma < 1), a decreasing cross-sectional area of the nozzle in the
direction of the gas flow leads to an increasing flow speed, and decreasing pressure,
density, and temperature. However, for supersonic flow (Ma > 1), this behavior is
reversed, and an increasing cross-sectional area of the nozzle leads to an increasing
flow speed, and decreasing pressure, density, and temperature. We conclude that
local Mach number of gas flowing through a converging nozzle (i.e., a nozzle whose
cross-sectional area decreases monotonically in the direction of the gas flow) can
never exceed unity. Moreover, the maximum Mach number (i.e., unity) is achieved
on exit from the nozzle, where the cross-sectional area is smallest. On the other hand,
the local Mach number of gas flowing through a converging-diverging nozzle (i.e., a
nozzle whose cross-sectional area initially decreases in the direction of the gas flow,
attains a minimum value, and then increases) can exceed unity. For this to happen, the
418
flow conditions must be arranged such that the sonic point corresponds precisely to
the narrowest point of the nozzle, which is generally known as the throat. In this case,
as the gas flows through the converging part of the nozzle, the local cross-sectional
area, A, travels down the left-hand, subsonic branch of the A/A1 curve shown in
Figure 14.1, while the flow speed, u, simultaneously increases. After passing through
the throat at the sonic speed, the gas flows through the diverging part of the nozzle,
and the cross-sectional area travels up the right-hand, supersonic branch of the A/A1
curve, while the flow speed continues to increase. The type of converging-diverging
nozzle just described is known as a de Laval nozzle, after its inventor, Gustaf de Laval
(18451913).
Consider a de Laval nozzle whose gas supply is derived from a large reservoir.
Assuming that the gas in the reservoir is essentially at rest, it follows that the temperature, pressure, and density of the gas in the reservoir correspond to the stagnation
temperature, pressure, and densityT 0 , p0 , and 0 , respectively. Equation (14.63)
(14.65) then specify the temperature, pressure, and density of the gas at the throat of
the nozzleT 1, p1 , and 1 , respectivelyin terms of the temperature, pressure, and
density of the gas in the reservoir.
Suppose that a de Laval nozzle exhausts gas into the atmosphere, whose pressure
is patm . Now, for the case of incompressible flow, the pressure of the gas exhausted
from a nozzle, pe (say), must match the ambient pressure, patm . The reason for this is
that any mismatch between the exhaust and ambient pressures is instantly communicated to the whole fluid by means of sound waves that travel infinitely fast (because
an incompressible fluid corresponds to the limit ). Of course, the sound speed
is finite in a compressible gas. However, in subsonic compressible flow, upstream
communication is still possible, because the local sound speed exceeds the local flow
speed. However, in supersonic compressible flow, upstream communication is impossible, because sound waves cannot catch up with the flow. Consequently, in the
case of a nozzle with a subsonic exhaust speed, we would generally expect the exhaust pressure to match the ambient pressure. However, in the case of a nozzle with
a supersonic exhaust speed, the exhaust pressure can be significantly dierent to the
ambient pressure.
Equation (14.61) yields
5
Ma e =
2
1
(1)/
p0
1,
pe
(14.73)
where Ma e is the Mach number of the gas exhausted by the nozzle, pe the exhaust
pressure, and p0 the reservoir pressure. Suppose that M e < 1that is, the exhaust
speed of the gas is subsonic. In this case, we would expect pe = patm , so that
5
Ma e =
2
1
p0 (1)/
1.
patm
(14.74)
It follows that the critical value of p0 /patm at which the exhaust speed becomes sonic
p0
patm
crit
+1
=
2
419
/(1)
.
(14.75)
Thus, in order for a de Laval nozzle to achieve supersonic exhaust speeds, p0 /patm
must exceed this critical value. For the case of a gas with = 7/5, we find that
(p0 /patm )crit = 1.89. Note that if p0 /patm does not exceed the critical value then,
as the gas flows through the converging part of the nozzle, its local cross-sectional
area, A, travels down the left-hand, subsonic branch of the A/A1 curve shown in
Figure 14.1. However, when the gas passes through the throat of the nozzle, the
area turns around, and then backtracks up the left-hand branch while the gas passes
through the diverging part of the nozzle. In this case, the Mach number never reaches
unity. In fact, in the converging part of the nozzle, the flow speed increases, while
the pressure, density, and temperature decrease. The flow speed attains its maximum,
subsonic value at the throat of the nozzle, while the pressure, density, and pressure
simultaneously attain minimum values. Finally, in the diverging part of the nozzle,
the flow speed decreases, while the pressure, density, and temperature increase.
Equations (14.64) and (14.67) yield
(1)/
2
p0
1
2
.
(14.76)
(Ma e 1) =
1+
+1
+ 1 pe
Moreover, Equations (14.45) and (14.63) imply that
5
2
u1 = R T 0
.
+1
(14.77)
Hence, Equations (14.72), (14.73), (14.76), and (14.77) give the following expression
for the exhaust speed from a de Laval nozzle,
5
(1)/
2
pe
.
ue = R T 0
(14.78)
1
1
p0
Because pe cannot be negative, we deduce that there is an upper limit to the exhaust
speed: namely,
1/2
2
ue max =
c0 = 2 C p T 0 ,
(14.79)
1
where use has been made of Equation (14.60). Here, c0 is the reservoir sound speed.
If the exhaust pressure of a de Laval nozzle is higher than the ambient pressure,
pe > patm , then the gas is said to be under-expanded. In this case, a pattern of
standing shock waves, called shock diamonds, forms in the exhaust plume external
to the nozzle. On the other hand, if the exhaust pressure is lower than the ambient
pressure, pe < patm , then the gas is said to be over-expanded. In this case, a static
shock front forms inside the diverging part of the nozzle. (See Section 14.8.) As the
gas passes through the front, its speed drops abruptly from a supersonic to a subsonic
value, whereas the pressure, density, and temperature all increase abruptly. As the
subsonic gas flows through the remainder of the nozzle, its velocity decreases further.
420
(14.81)
(14.82)
In writing the previous equations, we have neglected viscosity, heat conduction, and
potential energy. Note that we have used the more fundamental energy conservation
equation, (1.75), rather than Equation (14.32), because the latter equation incorporates the ideal gas law, and this law is not valid inside the shock, because the gas
there is not in thermodynamic equilibrium.
Consider a compressible gas flowing steadily down a duct of constant crosssectional area A. Let x measure distance along the duct. The gass temperature,
T (x), density, (x), pressure p(x), specific internal energy E(x), and normal velocity
u(x) = v x , are all assumed to be constant across any cross-section, and independent
of time. Let the shock be situated at x = 0. Suppose that in the region upstream of the
shock, x < 0, the temperature, density, pressure, specific internal energy, and flow
speed of the gas take the constant values T 1 , 1 , p1 , E1 , and u1 , respectively. Likewise, suppose that in the region downstream of the shock, x > 0, the temperature,
density, pressure, specific internal energy, and flow speed of the gas take the constant
values T 2 , 2 , p2 , E2 , and u2 , respectively. Of course, in the immediate vicinity of
the shock, T (x), (x), p(x), E(x), and u(x) are all rapidly-varying functions of x. We
wish to find the relationship between the upstream and downstream gas parameters.
We can achieve this goal by invoking conservation of mass, momentum, and energy
across the shockin other words, by making use of Equations (14.80)(14.82).
The mass continuity equation (14.71) yields
u
du
d
= ,
dx
dx
(14.83)
(14.84)
421
du
1 dp
=
,
dx
dx
(14.85)
du d p
+
= 0,
dx dx
(14.86)
or
d 2
(14.87)
u + p = 0,
dx
where use has been made of Equation (14.84). Finally, the energy conservation equation (14.73) yields
p du
dE
=
,
(14.88)
u
dx
dx
which can be rearranged to give
u
or
d
dp
dE
+
(p u) = u
,
dx dx
dx
1
du 2
d 1
d
2 du
3
( u E + p u) = u
= u
=
u ,
dx
dx
2
dx
dx 2
(14.89)
(14.90)
where use has been made of Equations (14.84) and (14.86). Thus, we obtain
1
d
u E + p u + u 3 = 0,
(14.91)
dx
2
which reduces to
d
p 1
E + + u2 = 0
dx
2
(14.92)
with the aid of Equation (14.84). The previous equation can also be written
d
1
(14.93)
H + u 2 = 0,
dx
2
where H = E + p/ is the specific enthalpy of the gas. [See Equation (14.18).]
Integrating Equations (14.84), (14.87), and (14.93) across the shock, we obtain
1 u1 = 2 u2 ,
2 u22
(14.94)
1 u12
+ p1 =
+ p2 ,
(14.95)
H1 +
1 2
1
u1 = H2 + u22 ,
2
2
(14.96)
422
downstream of the shock obeys the ideal gas law, and has the same ratio of specific
heats, . It follows that
p1,2 = R 1,2 T 1,2 ,
H1,2 =
R T 1,2 .
1
(14.97)
(14.98)
where use has been made of Equations (14.1), (14.18), and (14.60). Let Ma1 and Ma2
be the Mach numbers upstream and downstream of the shock, respectively. Thus,
u1,2
,
Ma1,2 =
R T 1,2
(14.99)
where use has been made of Equations (14.45) and (14.57). Equations (14.96),
(14.98), and (14.99) can be combined to give
T 2 1 + (1/2) ( 1) Ma12
=
.
T 1 1 + (1/2) ( 1) Ma22
(14.100)
T 1 Ma1
,
T 2 Ma2
T 2 Ma1
,
T 1 Ma2
(14.101)
(14.102)
where use has been made of Equation (14.97). Finally, Equations (14.95), (14.97),
and (14.99) give
p2 1 + Ma12
=
.
(14.103)
p1 1 + Ma22
Eliminating T 2 , 2 , and p2 between Equations (14.100), (14.101), and (14.103),
we obtain
1/2
2 + ( 1) Ma12
.
Ma2 =
(14.104)
1 + 2 Ma12
Note that
dMa22
dMa12
1+
< 0.
2
1 + 2 Ma1
(14.105)
Moreover, it is easily verified that Ma2 = 1 when Ma1 = 1. Hence, we deduce that
Ma2 1
as
Ma1 1.
(14.106)
In other words, if the downstream flow is subsonic then the upstream flow is supersonic, and vice versa. Equations (14.1), (14.100), (14.101), (14.103), and (14.104)
423
(14.107)
( + 1) Ma12
2 u1
=
=
,
1 u2 2 + ( 1) Ma12
(14.108)
p2 1
T2
=
.
T1
p1 2
(14.109)
The previous three equations completely describe the conditions downstream of the
shock in terms of the upstream conditions. We can rearrange these equations to give
Ma12 =
1 + ( + 1) (p2 /p1 )
,
2
(14.110)
2 u1 1 + ( + 1) (p2 /p1 )
,
=
=
1 u2 + 1 + ( 1) (p2 /p1 )
p2 + 1 + ( 1) (p2/p1 )
T2
=
.
T1
p1 1 + ( + 1) (p2/p1 )
(14.111)
(14.112)
1 + 2x
G(x) =
+1
2 + ( 1) x
( + 1) x
(14.114)
.
(14.115)
(14.116)
as
Ma1 1.
(14.117)
Now, the second law of thermodynamics forbids a spontaneous decrease in the specific entropy of the gas as it passes through the shock (Reif 1965). In other words, the
second law of thermodynamics demands that S2 S1 0. It follows that Ma1 0.
However, it is easily seen from Equations (14.107)(14.109) that p2 /p1 = 2 /1 =
424
u2 /u1 = T 2 /T 1 = 1 when Ma1 = 1. In other words, there is no shock (i.e., no discontinuity in the properties of the gas at x = 0) when the upstream Mach number
is exactly unity. Thus, we conclude that the only type of normal shock that is consistent with the second law of thermodynamics is one in which the upstream flow
is supersonic, and the downstream flow subsonicthat is, Ma1 > 1 and Ma2 < 1.
It is apparent from Equations (14.107)(14.109) that if Ma1 > 1 then p2 /p1 > 1,
2 /1 > 1, and T 2 /T 1 > 1. In other words, the passage of the gas through the shock
front leads to both compression and heating.
The dimensionless parameter
p2 p1
p
=
p1
p1
(14.118)
1 p
T
,
T1
p1
3
S ( + 1) p
,
R
12 2 p1
(14.119)
(14.120)
(14.121)
where = 2 1 , et cetera, and use has been made of Equation (14.59). It can be
seen that the flow across the shock is isentropic (i.e., p/ and T /p 1 are constant)
to first order in the shock strength. This is the case because the increase in specific
entropy across the shock is only third order in the shock strength.
Consider a strong shock. It can be seen from Equations (14.111) and (14.112)
that
+1
2
,
1
1
T2
1 p
,
T1
+ 1 p1
(14.122)
(14.123)
425
transformed into a shock wave. (See Section 14.4.) The pressure, density, and temperature behind the wave are p2 , 2 , and T 2 , respectively. Also, the gas behind the
wave follows the shock front at the speed u1 u2 . We can regard Ma1 as the Mach
number of the shock wave in the unperturbed gas. It follows from Equation (14.110)
that
+ 1 p
2
Ma1 = 1 +
.
(14.124)
2 p1
Thus, for a weak shock wave (i.e., p/p1 1),
Ma1 1.
(14.125)
In other words, a weak shock wave propagates through the unperturbed gas at the
local sound speed. Indeed, such a wave is essentially indistinguishable from a conventional sound wave. On the other hand, for a strong shock wave (i.e., p/p1
1),
5
+ 1 p
.
(14.126)
Ma1
2 p1
We conclude that a strong shock wave propagates through the unperturbed gas at a
speed that greatly exceeds the local sound speed.
(14.127)
(14.128)
426
on p
ath
x = Vp t
pist
sh
k
oc
pa
x = Vs t
th
Vp
Vs
region 2
region 1
Figure 14.2
A piston-generated shock wave.
where is the gass ratio of specific heats. We can define the Mach number of the
shock wave as Ma s = V s /c1 = u1 /c1 , where c1 is the sound speed in the unperturbed
gas. According to Equation (14.110),
1/2
+1
.
Ma s = 1 +
(14.129)
Finally, we can define the Mach number of the piston as Ma p = V p /c1 . Thus,
Ma p =
u1 u2 u1
u2
1
=
1
= Ma s 1
,
c1
c1
u1
2
(14.130)
where use has been made of Equation (14.111). Equations (14.127) and (14.129) can
be combined with the previous equation to give
1
1
+1
.
= 2+
2
2
Ma p
(14.131)
427
1),
Ma p ,
+1
Ma p ,
Ma s 1 +
4
(14.132)
(14.133)
( + 1) Ma p2
2
+1
Ma s
Ma p .
2
(14.134)
(14.135)
Thus, a weak shock is associated with a subsonic piston (i.e., Ma p 1), whereas a
strong shock is associated with a supersonic piston (i.e., Ma p
1).
428
x = c1 t
x = Vp t
expansion front
pist
on p
ath
x
Vp
c1
region 2
region 1
Figure 14.3
A piston-generated expansion wave.
the local sonic speed and the local gas flow speed. The terminating characteristic on
the right has slope dx/dt = c1 , where c1 is the region-1 sound speed. The terminating
characteristic on the left has the slope dx/dt = c2 V p , where c2 is the region-2 sound
speed. The latter slope may be either positive or negative, depending on whether c2
is greater than, or less than, V p .
Making use of isentropic relationships (see Section 14.3), we can explicitly evaluate the structure of the fan as follows. In terms of c1 , the local sonic speed in the
fan can be written [cf., Equation (14.47)]
(1)/2
c(x) = c1
.
(14.136)
1
Moreover, the relationship between the gas speed, u(x), and the density in the wave
is [cf., Equation (14.48)]
d
(14.137)
du = c .
1
( 1) u(x).
2
(14.138)
Hence, the absolute disturbance speed in the fan is [cf., Equation (14.51)]
c (x) = c(x) + u(x) = c1 +
1
( + 1) u(x).
2
(14.139)
429
In particular, the terminating characteristic on the left (for which u = V p ) has the
slope
dx
1
= c2 V p = c1 ( + 1) V p .
(14.140)
dt
2
It follows that
1
c2 = c1 ( 1) V p .
(14.141)
2
Standard isentropic relations (see Section 14.3) then yield the density and pressure
changes across the fan:
V p 2/(1)
2
1
= 1 ( 1)
,
(14.142)
1
2
c1
V p 2 /(1)
p2
1
= 1 ( 1)
.
(14.143)
p1
2
c1
Here, use has been made of the fact that c T 1/2 . At the critical piston withdrawal
speed V p = [2/( 1)] c1, the pressure and density in region 2 are both reduced to
zero, and the terminating characteristic on the left co-moves with the piston. Any
further increase in the withdrawal speed makes no dierence to the flow.
The equation of motion of the right edge of the expansion wave is
x1 = c1 t.
(14.144)
(14.145)
x x2
Vp
u(x) = 1 +
x1 x2
(14.147)
for x2 x x1 . In other words, the flow speed varies linearly with x inside the
expansion wave. It follows from Equations (14.138) and (14.141) that
c(x) c2
x x2
=
c1 c2
x1 x2
(14.148)
for x2 x x1 . Thus, the sound speed also varies linearly with x inside the expansion wave. However, according to Equation (14.136),
(1)/2 2(1)/2
1(1)/2
2(1)/2
x x2
x1 x2
(14.149)
430
p (1)/2 p2
p1(1)/2
p2(1)/2
x x2
x1 x2
(14.150)
for x2 x x1 . Thus, neither the density nor the pressure vary linearly with x inside
the expansion wave.
14.11 Exercises
14.1 Prove the following useful theorems regarding partial derivatives:
)
Y
X
=1
,
Y Z
X Z
X
X
Y
=
,
Y Z
W Z W Z
Z
Z
X
=
Y Z
Y X X Y
14.2
(a) The specific internal energy of a (not necessarily ideal) gas is defined by
dE = T dS p dv.
Demonstrate that
E
,
T=
S v
E
p=
v
and
T
v
,
S
p
.
S v
H
T=
S
and
H
v=
p
,
p
T
p
v
=
S
.
p
,
S
431
F
S=
T
F
p=
v
,
v
and
S
v
=
T
p
T
,
T
.
v
S=
and
G
T
v=
S
p
v
=
T
G
p
,
T
.
p
14.3 Demonstrate that the specific heat at constant volume of a (not necessarily
ideal) gas can be written
E
S
=T
.
cv =
T v
T v
Likewise, show that the specific heat at constant pressure takes the form
H
S
=T
.
cp =
T v
T p
14.4 The quantities
1 v
,
v T p
1 v
,
T =
v p T
1 v
S =
,
v p S
=
432
T v 2
,
T
T S =
T v 2
.
cp
(14.151)
c p T
= .
cv
S
Show that for the special case of an ideal gas, = 1/T , T = 1/p, and S =
1/( p). Hence, obtain the following standard results for an ideal gas:
cp
= ,
cv
c p cv = R,
cp =
R,
1
14.5 Show that for an ideal gas
2 1/2
1 1 ( 1) v
,
2
c0
2
v
1
= 1 ( 1)
,
2
c0
2 /(1)
v
1
= 1 ( 1)
,
2
c0
2 1/(1)
v
1
= 1 ( 1)
,
2
c0
v
Ma =
c0
T
T0
p
p0
where Ma is the Mach number, v the flow speed, and T 0 , p0 , 0 , and c0 , are the
temperature, pressure, density, and sound speed, respectively, at the stagnation
point.
14.6 Consider the flow of an isentropic ideal gas down a straight nozzle with a
slowly-varying cross-sectional area, A(x), where x measures distance along
the nozzle. Let u(x), (x), c(x), and Ma(x) be the flow speed, density, sonic
speed, and Mach number, respectively. Demonstrate that
c2
d
+ u du = 0,
433
diaphragm
p1
shock
|Vs|
region 1
p4
contact surface
|u2|
region 2
p4 > p1
expansion front
c4
region 3
region 4
Figure 14.4
A shock tube.
and
d dA du
+
+
= 0.
A
u
434
|u2 | = c1
and
p2
1
p1
2/1
2 1 + (1 + 1) (p2 /p1 1)
1/2
,
(4 1)/4
2 c4
p2
,
|u2 | =
1
4 1
p4
where c1 and c4 are the sound speeds in Regions 1 and 4, respectively. Hence,
obtain
2 4 /(4 1)
p2
p4
(4 1) (c1 /c4 ) (p2 /p1 1)
=
.
1
p1
p1
2 1 2 1 + (1 + 1) (p2/p1 1)
This expression give the shock strength, p2 /p1 , implicitly as a function of
the diaphragm pressure ratio, p4 /p1 . All other quantities of interest can be
expressed in terms of the shock strength. Show that
1/2
1 + 1 p2
|V s | = c1 1 +
1
,
2 1
p1
2 2 1 + (1 + 1) (p2 /p1 1)
,
=
1 2 1 + (1 1) (p2 /p1 1)
T2
p2 2 1 + (1 1) (p2 /p1 1)
=
,
T1
p1 2 1 + (1 + 1) (p2 /p1 1)
1/4
p2 /p1
3
=
,
4
p4 /p1
and
( 1)/4
p2 /p1 4
T3
=
.
T4
p4 /p1
14.8 Show that the maximum shock strength and shock speed attainable in a (uniform) shock tube, in the limit p4 /p1 , are
p 2 c4
=
p 1 c1
2 1 (1 + 1)
,
(4 1)
435
1 + 1
|V s | =
c4 ,
4 1
respectively. Show, further, that the contact surface moves at the speed
2
|u2 | =
c4 .
4 2
14.9 Show that Equations (14.38) and (14.39) can be written in the form
c
c
u
2
+u
+c
= 0,
1 t
x
x
u
2 a c
u
+u
+
= 0,
t
x 1 x
where c is the sound speed. By adding and subtracting the previous equations,
obtain
2c
+ (u + c)
u+
= 0,
t
x
1
2c
+ (u c)
u
= 0.
t
x
1
These equations indicate that the quantities P = u + 2 c/( 1) and Q =
u 2 c/( 1) are constant on curves that have the slope dx/dt = u + c and
dx/dt = u c, respectively. These curves are called characteristics, and P
and Q are known as Riemann invariants. In situations in which all the Q
characteristics originate from regions where the gas is at rest, we expect Q to
be constant throughout the gas. Deduce that [cf., Equation (14.50)]
1
u,
c = c0 +
2
where c0 is the stagnation sound speed. Show that
+ (u + c)
(u + c) = 0.
t
x
Hence, conclude that the P characteristics are straight-lines.
14.10 Consider a one-dimensional sound wave propagating through an ideal gas
whose unperturbed sound speed is c0 . At time t = 0, the velocity perturbation, u(x, t), due to the wave, has the form u0 sin(k x), where u0 and k are both
positive. Demonstrate that shocks form after a time
t
2
( + 1) k u0
436
14.11 An ideal gas is initially at rest in a uniform tube, and occupies the region to
the right of a tight-fitting piston whose position is X(t) = (1/2) a t 2 for t > 0.
Here, a > 0. Show that a shock first forms at t = t and x = x , where
2
c0
t =
,
+1 a
2
c0
2
.
x =
+1 a
Here, c0 is the stagnation sound speed, and the ratio of specific heats.
15
Two-Dimensional Compressible Inviscid Flow
15.1 Introduction
This chapter investigates two-dimensional, supersonic, compressible, inviscid flow.
Flow is said to be two-dimensional when the fluid properties only depend on two
Cartesian coordinates. More information on this subject can be found in Liepmann
& Roshko 1957, Hughes & Brighton 1999, Emanuel 2000, and Anderson 2003.
438
u1
v
w1
u2
Figure 15.1
An oblique shock.
Equations (14.107), (14.108), (14.109), and (14.113), yield
p2 1 + 2 Ma12 sin2
,
=
p1
+1
(15.1)
( + 1) Ma12 sin2
2 u1
=
=
,
(15.2)
1 u2 2 + ( 1) Ma12 sin2
1 + 2 Ma12 sin2 2 + ( 1) Ma12 sin2
T2
,
(15.3)
=
T1
( + 1) 2 Ma12 sin2
1
1
2 1
2
2
2
S2 S1
1 + 2 Ma1 sin ( + 1) Ma1 sin
= ln
, (15.4)
2
2
+
1
2 + ( 1) Ma1 sin
respectively. Here, p1 , 1 , T 1 , and S1 are the upstream pressure, density, temperature, and specific entropy, respectively, p2 , 2 , T 2 , and S2 are the corresponding
downstream quantities, is the ratio of specific heats, R the specific gas constant,
and use has been made of Equation (14.59). As before (see Section 14.8), the second
law of thermodynamics demands that S2 S1 0, which implies that Ma1 sin 1.
This sets a minimum inclination of the upstream flow to the shock front for a given
upstream Mach number. The maximum inclination is, of course, = /2. Thus,
,
2
(15.5)
1
Ma1
439
(15.6)
is termed the Mach angle. The downstream Mach number can be calculated by noting that Ma2 = w2 /c2 , rather that u2 /c2 , where c2 is the downstream sound speed,
and u2 = w2 sin( ). In other words, we simply need to make the transformation Ma2 Ma2 sin( ) in the previous analysis. Thus, it follows from Equation (14.104) that
Ma22 sin2 ( ) =
2 + ( 1) Ma12 sin2
1 + 2 Ma12 sin2
u1
,
v
(15.7)
(15.8)
and
u2
.
v
Eliminating v, and then making use of Equation (15.2), we obtain
tan( ) =
tan tan
,
1 + tan tan
(15.9)
(15.10)
(15.11)
(15.12)
Here, use has been made of the identity cos(2 ) cos2 sin2 . The previous
expression implies that = 0 at = and = /2, which are the limits of the range
of allowed values for defined in Equation (15.5). Within this range, is positive,
and, must, therefore, have a maximum value, max . This is illustrated in Figure 15.2,
where the relationship between and , for an ideal gas with = 1.4, is plotted for
various values of Ma1 .
If < max then, for each value of and Ma1 , there are two possible solutions,
corresponding to two dierent values of . The larger value of corresponds to the
stronger shock [because, according to Equation (15.1), the shock strength, p2 /p1 1,
is a monotonically increasing function of ].
Also shown in Figure 15.2 is the locus of solutions for which Ma2 = 1. In the
solution with the stronger shock, the downstream flow always becomes subsonic. On
the other hand, in the solution with the weaker shock, the downstream flow remains
supersonic, except for a small range of values of that are slightly smaller that max .
440
40
( )
30
20
10
10 20 30 40 50 60 70 80 90
( )
Figure 15.2
Oblique shock solutions for = 1.4. The solid curves show solutions for Ma1 = 1.2,
1.4, 1.6, 2, 3, 4, 5, 10, and , in order from the innermost to the outermost curve.
The dashed curve shows the locus of solutions for which = max . The dash-dotted
curve shows the locus of solutions for which Ma2 = 1. Solutions to the left and right
of the dash-dotted curve correspond to Ma2 > 1 and Ma2 < 1, respectively.
+ 1 tan( )
1
1
=
,
2
tan
2
Ma12 sin2
(15.13)
Ma12
+1
sin sin
.
sin 1 =
Ma12
2
cos( )
2
(15.14)
Ma12
sin 1
2
+1
2
Ma1 tan .
2
(15.15)
441
shock front
Ma2
Ma1
wall
wall
Figure 15.3
Supersonic flow in concave corner.
442
Ma2
Ma1
wedge
Ma1
Ma1
Ma2
shock front
Figure 15.4
Supersonic flow over symmetric wedge.
local Mach number of the flow, according to the relation (15.6). As we have already
mentioned, is known as the Mach angle. Note that 0 /2 for Ma1 1.
The lines of inclination (with respect to the upstream directed streamline) that
may be drawn at any point in the flow are known as Mach lines. At a general point,
P, there are always two lines that intersect the local streamline at the angle . (In
three-dimensional flow, the Mach lines define a conical surface with apex P, known
as a Mach cone.) Thus, a two-dimensional supersonic flow pattern is associated with
two families of Mach lines. These are conventionally distinguished by the labels
(+) and (). Those lying in the (+) set run to the right of the (upstream directed)
streamline, whereas those in the () set run to the left. (See Figure 15.5.) Mach lines
are also sometimes called characteristics. It is clear, from our previous discussion,
that the flow conditions at point P can only aect those at some other point Q if the
latter point lies between the (+) and () Mach lines passing through point P. (See
Figure 15.5, as well as Exercise 15.2.) Hence, we deduce that, unlike subsonic flow,
there is no upstream influence in supersonic flow.
Oblique shocks can also be distinguished by the labels (+) and (), according to
which set of characteristics they asymptote to in the limit of zero shock strength.
443
()
streamline
(+)
Figure 15.5
Right- and left-running Mach lines [labeled (+) and (), respectively] emanating
from an arbitrary point, P, in a supersonic flow field.
+1
2
2
2
Ma1 sin 1
(15.16)
Ma1 tan .
2
It follows that is either close to or /2, depending on whether Ma2 > 1 or
Ma2 < 1. (See Figure 15.2.) For the present, we are only considering the former case
(Ma2 > 1), for which we may use the approximation [see Equation (15.6)]
tan tan =
(15.17)
Ma12 1
+ 1 Ma1
2
2
.
Ma1 sin 1
2
2
Ma1 1
(15.18)
This is the basic relation that is needed to obtain all other approximate expressions,
because these expressions only depend on the normal component of the upstream
flow, Ma1 sin . Thus, Equation (15.1) yields
p Ma12
,
(15.19)
p1
Ma12 1
444
where p = p2 p1 . It is apparent that the shock strength, p/p1 , is directly proportional to the deflection angle, .
The changes in the other flow quantities, except the specific entropy, across the
shock front are also directly proportional to . (See Exercise 15.3.) The change in the
entropy, on the other hand, is proportional to the third power of the shock strength
(see Section 14.8), and, hence, to the third power of the deflection angle.
To explicitly find the deviation of the wave angle, , from the Mach angle, , we
write
= + ,
(15.20)
where . Thus, we have
sin = sin( + ) sin + cos .
(15.21)
However, by definition, sin = 1/Ma1 , and cos = (Ma12 1)1/2 /Ma1 [see Equation (15.6)], so we obtain
Ma1 sin 1 +
or
Ma12
Ma12 1,
sin 1 + 2 Ma12 1.
2
(15.22)
(15.23)
+ 1 Ma12
.
4
Ma12 1
(15.24)
In other words, for a finite deflection angle, , the wave angle, , diers from the
Mach angle, , by an amount, , that is of the same order of magnitude as .
The change in flow speed across the shock front is obtained from the ratio
w22
w12
u22 + v 2
u12
v2
cos2
(u2 /v)2 + 1 tan2 ( ) + 1
=
=
,
2
2
(u1 /v) + 1
cos2 ( )
tan + 1
where use has been made of Equations (15.8) and (15.9). However,
Ma
1
2
1
2
2
,
cos = 1 sin
1
2
Ma1
Ma12 1
where use has been made of Equation (15.23). Similarly,
Ma
1
2
(
1
.
cos2 ( )
1
Ma1
Ma12 1
(15.25)
(15.26)
(15.27)
445
shock
weak shocks
(a)
(b)
Ma1
wall
Ma1
(c)
Mach lines
Ma1
Figure 15.6
Compression of supersonic flow that is turned through an angle . Case (a) shows a
single turn through an angle . Case (b) shows the turn subdivided into a number of
smaller turns of magnitude . Case (c) shows a smooth continuous turn through an
angle .
Hence, to first order in ,
w2
,
1
w1
Ma12 1
(15.28)
,
w1
Ma12 1
(15.29)
where w = w2 w1 .
446
(15.30)
S ()3 .
(15.31)
(15.32)
(15.33)
3
.
n2
(15.34)
(15.35)
Let us now consider what happens to the flow further away from the wall, where
the weak shocks and Mach lines of Figure 15.6(b) and Figure 15.6(c), respectively,
447
shock front
Ma2
b
Ma1
Ma2
envelope of
Mach lines
a
Ma1
wall
Figure 15.7
Convergence of Mach lines in continuous compression by turning.
converge. Consider Figure 15.7. Because of the convergence of the Mach lines,
the change from the initial Mach number, Ma1 , to the final Mach number, Ma2 , is
achieved in a much shorter distance on streamline b than on streamline a. Thus,
the velocity gradients on streamline b are higher than those on streamline a. Now,
an intersection of Mach lines would imply infinitely high velocity gradients, because
there would be two dierent values of the Mach number at a single point. In practice,
this does not occur, because, in the region where the Mach lines converge, and before
they cross, the velocity gradients become high enough that the conditions are no
longer isentropic. In fact, a shock wave develops, as illustrated in the figure.
448
shock
Ma1
Ma2
(b)
Ma1
1
1
wall
Ma2
(c)
Ma1
Ma2
Figure 15.8
Expansion of supersonic flow via turning. Case (a) shows a single turn through an
oblique shock, which is forbidden on thermodynamic grounds. Case (b) shows a
Prandtl-Mayer expansion fan. Case (c) shows smooth continuous expansion.
However, although this would satisfy the equations of fluid motion, it would lead
to a spontaneous decrease in specific entropy across the shock front, and is, thus,
forbidden by the second law of thermodynamics. (See Section 14.8.)
What actually happens is as follows. The nonlinear mechanism that tends to
steepen a compression (see Sections 14.4 and 15.5) produces the opposite eect in
an expansion. In particular, instead of being convergent, the Mach lines are divergent,
as shown in Figure 15.8(b) and (c). Consequently, there is a tendency for velocity
gradients to decrease with increasing distance from the wall. Thus, an expansion is
isentropic throughout the fluid.
The expansion at a corner takes place via a fan of straight Mach lines that is
generally known as a Prandtl-Mayer expansion fan. [See Figure 15.8(b).] The arguments that lead to this conclusion are as follows. First, the flow up to the corner
is uniform, at Mach number Ma1 (say), and, thus, the leading Mach line must be
straight, and inclined to the flow at the Mach angle 1 = sin1 (1/Ma1 ). Because of
the limited upstream influence in supersonic flow, the same argument may be applied
to each succeeding portion of the flow. The terminating Mach line stands at the angle
2 = sin1 (1/Ma2 ) to the wall. Here, Ma2 is the Mach number of the downstream
flow. Second, because there is no characteristic length to define a scale in this problem, any variation of flow parameters can only be with respect to angular position,
449
measured relative to the corner. That is, flow conditions must be constant along rays
emanating from the corner.
Figure 15.8(c) shows a typical expansion over a continuous convex turn. Because
the flow is isentropic, it is also reversible (i.e., if the direction of the flow is reversed
then the fluid is isentropicaly compressed).
Equation (15.35), which specifies the relationship between and Ma in an isentropic compression or expansion by turning, can be written
dw
,
(15.36)
d = Ma 2 1
w
or
Ma
dw
+ const. =
(Ma).
(15.37)
Ma 2 1
w
In order to evaluate the integral, and, thus, derive an explicit form for the function
(Ma), we need to rewrite w in terms of Ma using the definition
w = c Ma.
(15.38)
Here, c is the local sound speed. Given that c T [see Equation (14.45)], it follows
from Equation (14.58) that
c02
1
=1+
(15.39)
Ma 2 ,
2
c2
where c0 is the stagnation sound speed. Thus,
1
dMa
dw dMa dc
=
+
=
.
2
w
Ma
c
Ma
1 + [( 1)/2] Ma
(15.40)
The function (Ma), which is known as the Prandtl-Mayer function, is then given by
(see Exercise 15.10)
Ma
M2 1
dM
(Ma)
(15.41)
2
1 + [( 1)/2] M M
1
1/2
1/2
$
1/2 %
1
+1
1
2
(Ma 1) tan1 Ma 2 1
=
.
tan
1
+1
Here, the constant of integration has been arbitrarily chosen such that = 0 corresponds to Ma = 1. A supersonic Mach number, Ma, is thus associated with a
definite value of the function . In fact, as Ma increases from 1 to , increases
monotonically from 0 to max , where
1/2
+ 1
max =
1 .
(15.42)
2 1
Equation (15.37) yields
= (Ma) (Ma1 ),
(15.43)
450
d
(a)
(b)
c
c
Ma > 1
= max
afterbody
Ma < 1
Ma1
Ma = Ma1
Ma = 1
d
/2
shoulder
Ma > 1
Figure 15.9
Detached shock waves. (a) Detached shock on wedge with afterbody. [At lettered
points on the shock, flow deflection is given by corresponding points in (b).] (b)
Deflection angle versus wave angle for fixed Ma1 .
as the relationship between the deflection angle, , and the Mach number, Ma, in an
isentropic compression or expansion by turning. In a compressive turn, decreases,
whereas, in an expansive turn, it increases: in each case, by an amount equal to the
change in . Here, we have set 1 = 0, because only the change in matters. Note
that is positive in compressive turns, and negative in expansive turns.
451
illustrated in Figure 15.9(b), which shows one curve from Figure 15.2. The complete
description of the bow shock is complicated by the fact that its location cannot be
found explicitly, because of the subsonic influence upstream, which implies that the
entire flow field must be solved as one. There is an absolute limit of , just greater
than 45 , beyond which the shock will always detach, no matter how large the value
of the upstream Mach number, Ma1 . Hence, if the wedge is replaced by a blunt-nosed
obstacle then the shock will always detach (because = 90 at the nose of such an
obstacle).
For a given wedge angle, , the sequence of events with decreasing upstream
Mach number, Ma1 , is as follows. When Ma1 is suciently high, the shock front
is attached to the wedge apex, and its straight portion is independent of both the
shoulder and the afterbody. [See Figure 15.9(a).] The straight portion of the shock
front lies between the wedge apex and the point where the first Mach line, emanating
from the shoulder, intersects it. As Ma1 decreases, the wave angle, , increases. With
a further decrease in the Mach number, a value is reached at which the fluid behind
the shock becomes subsonic. The shoulder now has an eect on the whole shock
front, which may become curved, although still remaining attached to the wedge
apex. These conditions correspond to the region between the lines Ma2 = 1 and
= max in Figure 15.2. At the Mach number corresponding to = max , the shock
front starts to detach from the apex. This Mach number is called the detachment
Mach number. With a further decrease of Ma1 , the detached shock moves upstream
of the wedge apex.
452
expansion fan
p1
Ma1
p2
p2
shock
plate
0
p1
slipstream
c
shock
expansion fan
Figure 15.10
A flat lifting plate. Ma1 is the upstream Mach number, p1 , p2 , et cetera, denote
pressures, and 0 is the angle of attack.
(p2
p2 ) c sin 0 ,
(15.44)
(15.45)
respectively, where c is the chord-length (i.e., width). (See Figure 15.10.) The increase in entropy of the flow along the upper surface of the airfoil is not the same as
that for the flow along the lower surface, because the upper and lower shock waves
occur at dierent Mach numbers. Consequently, the streamline attached to the trailing edge of the airfoil is a slipstreamthat is, it separates flows with the same pressures, but slightly dierent speeds, temperatures, and densitiesinclined at a small
angle relative to the free stream. (The angle of inclination is determined by the requirement that the pressures on both sides of the slipstream be equal to one another.)
Note that the drag that develops on the airfoil is of a completely dierent nature
to the previously discussed (see Chapter 9) drags that develop on subsonic airfoils,
such as friction drag, form drag, and induced drag. This new type of drag is termed
supersonic wave drag, and exists even in an idealized, inviscid fluid. It is ultimately
due to the trailing shock waves attached to the airfoil.
Comparatively far from the airfoil, the attached shock waves and expansion fans
intersect one another. The expansion fans then attenuate the oblique shocks, making them weak and curved. At very large distances, the shock waves asymptote to
free-stream Mach lines. However, this phenomenon does not aect the previous calculation of the lift and drag on a flat-plate airfoil.
453
p Ma 2
(15.46)
,
p
Ma 2 1
specifies the relative change in pressure across either a weak oblique shock (see Section 15.4) or a weak expansion fan (see Exercise 15.6) that deflects flow of Mach
number Ma through an angle . Because, in the weak wave approximation, the
pressure, p, never greatly diers from the upstream pressure, p1 , and the Mach number, Ma, never diers appreciably from the upstream Mach number, Ma1 , we can
write
p p1 Ma12
,
(15.47)
p1
Ma 2 1
1
which is correct to first order in . Here, is the deflection angle relative to the
upstream flow.
It is convenient to define a dimensionless quantity known as the pressure coecient:
p p1
,
(15.48)
Cp =
q1
where q1 = (1/2) 1 w12 , and 1 and w1 are the upstream density and flow speed,
respectively. Given that the upstream sound speed is c1 = p1 /1 , and Ma1 =
w1 /c1 , we obtain
2 p p1
Cp =
,
(15.49)
Ma12 p1
which yields
2
Cp =
.
Ma12 1
(15.50)
This is the fundamental formula of thin-airfoil theory. It states that the pressure coefficient is proportional to the local deflection of the flow from the upstream direction.
Consider the flat-plate airfoil shown in Figure 15.10. The deflection angle is 0
454
on the upper surface of the airfoil, and +0 on the lower surface. (A positive deflection angle corresponds to compression, and a negative deflection angle to expansion.)
Thus, the pressure coecients on the upper and lower surfaces are
C pU =
and
2 0
(15.51)
Ma12 1
2 0
C pL =
,
Ma12 1
(15.52)
(15.53)
D
.
q1 c
(15.54)
CL =
(pL pU ) c cos 0
= (C pL C pU ) cos 0 ,
q1 c
(15.55)
CD =
(pL pU ) c sin 0
= (C pL C pU ) sin 0 .
q1 c
(15.56)
CL =
and the dimensionless coecient of drag,
CD =
It follows that
Making use of Equations (15.51) and (15.52), as well as the conventional small angle
approximations cos 0 1 and sin 0 0 , we obtain
4 0
CL =
,
Ma12 1
(15.57)
4 02
.
CD =
Ma12 1
(15.58)
The focus (see Section 9.3) of the airfoil is at the midchord. Moreover, the ratio
D/L 2 = (Ma12 1)1/2 /4 is independent of 0 .
As a second example, consider the diamond-section airfoil pictured in Figure 15.11.
This airfoil has a nose angle 2 , and zero angle of attack. The pressure coecient
on the front face of the airfoil is
2
,
C pF =
Ma12 1
(15.59)
455
expansion fan
p2
Ma1
p3
2
2
p2
t
p3
Figure 15.11
A diamond-section airfoil. Ma1 is the upstream Mach number. p2 and p3 denote
pressures.
whereas that on the rear face is
2
C pR =
.
2
Ma1 1
(15.60)
(15.61)
42
q1 c,
D = (p2 p3 ) t = (p2 p3 ) c =
Ma12 1
(15.62)
giving a drag
where t and c are the thickness and chord-length of the airfoil, respectively. (See
Figure 15.11.) Thus,
CD =
$ t %2
42
4
D
=
=
.
q1 c
c
2
2
Ma1 1
Ma1 1
(15.63)
456
yU (x)
yC (x)
yL(x)
xL
xR
x
Figure 15.12
An arbitrary airfoil.
is assumed to be uniform in the z-direction, with the upstream flow parallel to the
x-axis. The upper surface of the airfoil corresponds to the curve y = yU (x), the lower
surface to the curve y = yL (x), and the camber line (i.e., the centerline) to the curve
y = yC (x). Furthermore, the leading and trailing edges of the airfoil lie at x = xL
and x = xR , respectively. Hence, yU (xL ) = yC (xL ) = yL (xL ) and yU (xR ) = yC (xR ) =
yL (xR ). By definition,
yU (x) + yL (x)
,
(15.64)
yC (x) =
2
for xL x xR . It is helpful to define the half-width of the airfoil,
h(x) =
yU (x) yL (x)
,
2
(15.65)
for xL x xR . Note that h(xL ) = h(xR ) = 0. We can also define the mean angle of
attack of the airfoil:
0 =
(15.66)
(15.67)
(15.68)
457
is the local angle of attack of the camber line. Thus, the camber function, C (x),
parameterizes the deviations of (x) from 0 across the width of the airfoil. Equations (15.64), (15.65), and (15.67) yield
yU (x) = yC (x) + h(x) = yC (xL ) 0 (x xL ) C (x) + h(x),
(15.69)
(15.70)
Hence, the airfoil shape is completely specified by the thickness function, h(x), the
camber function, C (x), and the mean angle of attack, 0 .
The pressure coecients on the upper and lower surfaces of the airfoil are [see
Equations (15.50)]
dyU
2
C pU =
,
dx
Ma12 1
2
dyL
C pL =
,
dx
Ma12 1
(15.71)
(15.72)
dh
,
dx
dh
.
dx
(15.73)
(15.74)
Now, the lift and drag per unit transverse length acting on the airfoil are given by
[cf., Equations (15.53)(15.56)]
xR
L = q1
C pL C pU dx,
(15.75)
D = q1
xL
xR
xL
dyU
dyL
C pL
+ C pU
dx.
dx
dy
2
xR
dyU
2 q1
dyL
L=
dx + dx dx
Ma12 1 xL
xR
2
4 q1
dh
2
+ dx.
=
dx
Ma12 1 xL
(15.76)
(15.77)
(15.78)
458
(15.79)
(15.80)
(15.81)
(15.82)
Hence, the coecients of lift and drag, L/(q1 c) and D/(q1 c), are written
4 0
4
CL =
=
,
Ma12 1
Ma12 1
dh 2
4
2
CD =
+
dx
2
Ma1 1
dh 2
4
,
2
2
=
0
C
dx
2
Ma1 1
(15.83)
(15.84)
respectively. Thus, in thin-airfoil theory, the lift only depends on the mean angle
of attack, whereas the drag splits into three components. Namely, a drag due to
thickness, a drag due to lift, and a drag due to camber.
1 2
v = H0 ,
2
(15.85)
where H is the specific enthalpy, H0 the stagnation enthalpy, and v the flow velocity.
(See Section 14.5.) The fluid equation of motion, (14.31), reduces to
(v ) v =
p
.
(15.86)
459
(15.87)
p
.
(15.88)
p
+ T S,
(15.89)
where T is the temperature, and S the specific entropy. Moreover, Equation (15.85)
yields
H + (v 2 /2) = H0 .
(15.90)
Thus, we arrive at Croccos theorem:
v = H0 T S.
(15.91)
In most aerodynamic flows, the fluid originates from a common reservoir, which
implies that the stagnation enthalpy, H0 , is the same on all streamlines. Such flow is
termed homenergic. It follows from Equation (14.18) that the stagnation temperature,
T 0 , is the same on all streamlines in homenergic flow. According to Croccos theorem, an irrotational (i.e., = 0 everywhere) homenergic (i.e., H0 = 0 everywhere)
flow pattern is also homentropic (i.e., S = 0 everywhere). Conversely, a homenergic, homentropic flow pattern is also irrotational (at least, in two dimensions, where
v and cannot be parallel to one another).
( u) +
( v) = 0,
x
y
u
u
u
1 p
u
+v
=
,
x
y
x
v
v
1 p
+v
=
,
x
y
y
p
=
,
p0
0
(15.92)
(15.93)
(15.94)
(15.95)
460
where p0 and 0 are the uniform stagnation pressure and density, respectively. (Note
(15.97)
u2
u
u
c 2
+uv
= u ,
x
y
x
(15.98)
uv
v
c 2
v
+ v2
= v .
x
y
y
(15.99)
Summing the previous two equations, and then making use of Equation (15.97), we
obtain
u v
2
2 u
2
2 v
u c
+ v c
+uv
+
= 0.
(15.100)
x
y
y x
1
( 1) u 2 + v 2 ,
2
(15.101)
(15.102)
v = 0.
(15.103)
Suppose that a solid body, such as an airfoil, is placed in the aforementioned flow
pattern. The cross-section of the body is assumed to be independent of the Cartesian
461
coordinate z. The body disturbs the flow pattern, and changes its velocity field, which
is now written
u = U + u ,
v=v,
(15.104)
(15.105)
1
( 1) 2 U u + u 2 + v 2 .
2
(15.106)
2
2
+
1
1
v
u
v u
u
2 u
2
+
= Ma ( + 1) +
+
(1 Ma )
x
y
U
2
U
2
U x
+ 1 v
1 u v
u
2
+
+ Ma ( 1) +
U
2
U
2
U y
u u v
2 v
+ Ma
+
1+
.
(15.107)
U
U y
x
The previous equation is exact. However, if u /U and v /V are small then it becomes
possible to neglect many of the terms on the right-hand side. For instance, neglecting
terms that are third-order in small quantities, we obtain
(1 Ma2 )
u v
u u
u v
+
Ma2 ( + 1)
+ Ma2 ( 1)
x
y
U x
U y
v u v
+
+ Ma2
.
(15.108)
U y
x
u v
+
0.
x
y
(15.109)
u v
u u
+
Ma2 ( + 1)
.
x
y
U x
(15.110)
462
On the other hand, subsonic (i.e., Ma < 1) and supersonic flow (i.e., Ma > 1)
are both governed by Equation (15.109). Another situation in which certain terms
on the right-hand side of Equation (15.108) must be retained is hypersonic flow (i.e.,
Ma
1). This follows because, although u /U and v /V are small, their products
with Ma2 can still be non-negligible. Roughly speaking, Equation (15.109) is valid
for 0 Ma 0.8 and 1.2 Ma 5. In other words, transonic flow corresponds
to 0.8 < Ma < 1.2, and hypersonic flow to Ma > 5 (Anderson 2003).
The pressure coecient is defined
Cp =
p p
p p
2
=
.
(1/2) U 2 Ma2 p
(15.111)
1
1
( 1) q 2 = c2 + ( 1) U 2 ,
2
2
(15.112)
(15.113)
/(1)
p
1
q2
= 1 + ( 1) Ma2 1 2
.
p
2
U
(15.114)
Hence, we obtain
2
Cp =
Ma2
/(1)
2
q
1
2
1 + ( 1) Ma 1
1 ,
2
2
U
(15.115)
which reduces to
2
Cp =
Ma2
2
2 /(1)
1 1 ( 1) Ma2 2 u + u + v
.
1
2
U
U2
(15.116)
Using the binomial expansion on the expression in square brackets, and neglecting
terms that are third-order, or higher, in small quantities, we obtain
2
2
2 u
u
v
2
+ (1 Ma )
+
C p
(15.117)
.
U
U
U
For two-dimensional flows, in the limit in which Equation (15.109) is valid, it is
consistent to retain only first-order terms in the previous equation, so that
Cp
2 u
.
U
(15.118)
463
Let
f (x, y) = 0
(15.119)
be the equation of the surface of the solid body that perturbs the flow. At the surface,
the velocity vector of the flow must be perpendicular to the local normal: that is, the
flow must be tangential to the surface. In other words,
q f = 0,
(15.120)
which reduces to
f
f
+ v
= 0.
x
y
Neglecting u with respect to U, we obtain
(U + u )
(15.121)
f /x dy
v
=
,
U
f /y dx
(15.122)
where dy/dx is the slope of the surface, and v /U the approximate slope of a streamline.
Now, the body has to be thin in order to satisfy our assumption that the induced
velocities are relatively small. This implies that the coordinate y diers little from
zero (say) on the surface of the body. Hence, we can write
v
y +
(15.123)
v (x, y) = v (x, 0) +
y y=0
in the immediate vicinity of the surface. Within the framework of small-perturbation
theory, it is consistent to neglect all terms on the right-hand side of the previous
equation after the first. Hence, the boundary condition (15.122) reduces to
dy
v (x, 0) = U
,
(15.124)
dx body
respectively.
Because a homenergic, homentropic flow pattern is necessarily irrotational (see
Section 15.10), we can write
q = U e x ,
(15.125)
where is the perturbed velocity potential. (See Section 4.15.) It follows that
u =
,
x
v =
.
y
(15.126)
2 2
+
0,
x 2 y 2
2
,
U x
dy
(x, 0)
U
,
y
dx body
Cp
respectively.
(15.127)
(15.128)
(15.129)
464
(15.130)
2 2
+
0,
x 2 y 2
(15.131)
(15.132)
We also expect the perturbed flow to become vanishingly small far from the wall,
which implies that
(x, ) (x, )
=
= 0.
(15.133)
x
y
Consider subsonic flow, for which 1 Ma2 m 2 > 0. Equation (15.131) is of
the elliptic type
1 2
2
+ 2
=0
(15.134)
2
x
m y 2
(Arfken 1985). Let us search for a separable solution of the form
(x, y) = F(x) G(y).
(15.135)
We obtain
F
1 G
= 2
,
(15.136)
F
m G
where denotes derivative with respect to argument. Given that the left-hand side of
the previous equation is a function of x only, whereas the right-hand side is a function
of y only, the two sides must equal the same constant. In other words,
F
= k 2 ,
F
G
= +k 2 .
G
(15.137)
(15.138)
465
Here, the sign of the constant has been chosen so as to allow the boundary condition
(15.133) to be satisfied. The most general solution of Equation (15.137) is
F(x) = A1 sin(k x) + A2 cos(k x),
(15.139)
where A1 and A2 are arbitrary constants. Likewise, the most general solution of
Equation (15.138) is
(15.140)
G(y) = B1 e m k y + B2 e+m k y ,
where B1 and B2 are arbitrary constants.
The boundary condition (15.133) implies that B2 = 0. The boundary condition (15.132) then yields
m k B1 [A1 sin(k x) + A2 cos(k x)] = U cos( x).
(15.141)
= U cos( x) em y .
v =
y
(x, y) =
(15.142)
(15.143)
(15.144)
According to Equation (15.128), the pressure coecient at the wall can be written
2
2
=
Cp
sin( x).
(15.145)
U x y=0
2
1 Ma
It immediately follows that there is zero net drag force acting on the wall, because the
pressure variations are in phase with the walls sinusoidal modulations, and, hence,
symmetrical about the crests and troughs.
Let us now consider under what circumstances the approximations made in the
small-perturbation theory developed in the previous section are valid for the problem
under investigation. In fact, small-perturbation theory is premised on three assumptions. First, that u /U 1 and v /V 1. It is evident from Equations (15.143) and
(15.144) that this assumption is valid provided
1.
1 Ma2
(15.146)
u
u u
Ma2 ( + 1)
,
x
U x
(15.147)
466
1.
(15.148)
Finally, in Equation (15.123), it is assumed that the second term on the right-hand
side is negligible compared to the first, when evaluated at the surface of the wall,
which implies that
(15.149)
1 Ma2 1.
Hence, we deduce that the analysis described in this section is valid provided
(1 Ma2 ) 3/2
1
1 Ma2 ,
.
,
2
Ma ( + 1)
2
1 Ma
(15.150)
0
x 2 2 y 2
(15.151)
(Arfken 1985). As is easily verified, the previous equation has the general solution
(x, y) = f (x y) + g(x + y),
(15.152)
where f and g are arbitrary functions (Fitzpatrick 2013). It is clear that f is constant
along lines x y = const., whereas g is constant along lines x + y = const.. These
lines are inclined at the Mach angle, = cot1 [(Ma2 1) 1/2 ], to the undisturbed
flow. They are, in fact, the Mach lines, or characteristics, of the unperturbed flow.
(See Section 15.3.) The former characteristics are inclined downstream. In other
words, they originate at the wall. The latter characteristics are inclined upstream. In
other words, they originate at infinity. Because there are no sources at infinity, the
latter characteristics carry no perturbation, which implies that g = 0.
The boundary condition at the wall, (15.129), yields
f (x) = U cos( x),
(15.153)
U
sin( x),
(15.154)
467
U
=
cos[ (x y)],
u =
x
= U cos[ (x y)].
v =
y
(x, y) = f (x y) =
(15.155)
(15.156)
(15.157)
Note that, unlike the case of subsonic flow, the amplitude of the perturbed velocity
does not decrease with increasing distance from the wall.
The pressure coecient at the wall is
2
2
=
cos( x).
(15.158)
Cp
U x y=0
Ma2 1
It can be seen, by comparison with Equation (15.145), that the maxima and minima
of the pressure are now phase-shifted by /2 radians with respect to the corresponding maxima and minima in the subsonic case. Hence, the pressure distribution at
the wall is anti-symmetric with respect to the crests and troughs of the wall. Consequently, a net drag force is exerted on the wall. The mean coecient of drag per
wavelength is [see Equation (15.56)]
1 l
dy
dx,
(15.159)
Cp
CD =
l 0
dx
which is obtained by replacing the sine of the slope of the wall by the tangent, dy/dx.
This approximation is valid within the context of the small-perturbation theory. However, according to Equations (15.130) and (15.158), C p can be written
dy
2
,
Cp =
dx
Ma2 1
which implies that
CD =
Ma2 1
2
dy
.
dx
(15.160)
(15.161)
dy
dx
1
=
l
l
dy
dx
2
dx.
(15.162)
Equations (15.160) and (15.161) are valid for any small-amplitude, periodic modulation of the wall. For the particular sinusoidal modulation under consideration in this
section, we find that
( ) 2
.
(15.163)
CD =
Ma2 1
468
The discussion of the range of validity of the approximations used in deriving the
previous results follows the same lines as in the subsonic case.
(15.164)
1 Ma2 .
(15.165)
(15.166)
= m y.
(15.167)
= 1,
x
= 0,
y
= 0,
x
= m,
y
(15.169)
so
=
+
= ,
x x x
(15.170)
=
+
=m .
y y y
(15.171)
It follows that
1
=
,
x m
=
,
y
2
1 2
=
,
2
m 2
x
2
2
= m 2.
2
y
(15.172)
469
(15.173)
However, this is simply Laplaces equation, which governs incompressible flow. (See
) represents an incompressible flow in (, ) space that
Section 4.15.) Hence, (,
is related to a compressible flow, represented by (x, y), in (x, y) space.
Suppose that the surface of the airfoil is given by y = f (x) in (x, y) space, and
= q() in (, ) space. According to Equation (15.129), the boundary condition on
the flow at the airfoil surface is
U
(x, 0)
df
dx
y
(15.174)
0)
dq
(,
.
d
(15.175)
=
,
y
(15.176)
df
dq
=
.
dx d
(15.177)
In other words, the shape of the airfoil in (x, y) space is the same as that in (, )
space. Hence, the coordinate transformation (x, y) (, ) transforms compressible
flow over an airfoil in (x, y) space to incompressible flow over the same airfoil in
(, ) space.
According to Equation (15.128), the pressure coecient in (x, y) space is written
Cp
2
2
=
.
U x U m
(15.178)
2
.
U
(15.179)
C p0
(15.180)
1 Ma2
This result is known as the Prandtl-Glauert rule: it is a similarity rule that relates
incompressible flow over a given two-dimensional airfoil to subsonic compressible
470
flow over the same airfoil. Given that the coecient of lift is directly related to the
pressure coecient (see Section 15.9), we can also deduce that
C L0
CL =
,
1 Ma2
(15.181)
where C L is the coecient of lift for compressible flow, and C L0 the corresponding
coecient for incompressible flow. It is clear that compressibility has the eect of
increasing the lift on a given airfoil.
Equation (15.181) indicates that the coecient of lift goes to infinity as Ma approaches unity, which is an impossible result. The quandary is resolved by recalling
that linearized theory breaks down in the transonic regime (where Ma 1). In fact,
the Prandlt-Glauert rule is only accurate up to Mach numbers of approximately 0.8.
The eect of compressibility on subsonic flow-fields can be seen by noting that
u =
1
u
=
=
,
x
m
1 Ma2
(15.182)
where u = /
is the incompressible velocity perturbation parallel to the unperturbed flow. It can be seen that as Ma increases, the parallel velocity perturbation,
u , also increases relative to u . In other words, compressibility strengthens the disturbance of the flow introduced by the airfoil.
In conventional inviscid, incompressible aerodynamic theory, a two-dimensional
airfoil experiences zero aerodynamic drag (assuming that there is no separation of
the boundary layer). (See Chapters 8 and 9.) This is due to the fact that, in the
absence of friction, the pressure distributions over the forward and rearward portions
of the airfoil exactly cancel in the direction of the unperturbed flow. Note that the
compressible pressure coecient, C p , only diers from the incompressible pressure
coecient, C p0 , by a constant scale-factor. Thus, if the distribution of C p0 over the
surface of the airfoil results in zero drag then the distribution of C p will also result
in zero drag. In other words, in subsonic flow, compressibility does not give rise to
additional aerodynamic drag.
471
0,
x 2 2 y 2
where
=
(15.183)
Ma2 1.
(15.184)
(15.185)
where f and g are arbitrary functions. Because disturbances only propagate along
downstream-running Mach lines (i.e., Mach lines that originate at the airfoil), we
only need the function f for the upper surface, and the function g for the lower
surface. Thus,
(x, y) = f (x y)
y > 0,
(15.186)
(x, t) = g(x + y)
y < 0.
(15.187)
dy
(x, 0+)
= f (x),
=
U
dx U
x
which implies that
U dy
.
f (x) =
dx U
(15.188)
(15.189)
Similarly,
U dy
.
(15.190)
g (x) =
dx L
According to Equation (15.128), the pressure coecient on the surface of the airfoil
is given by
2
.
(15.191)
Cp
U x y=0
Hence, we obtain
Cp =
2
f (x)
U
(15.192)
Cp =
2
g (x)
U
(15.193)
dyU
2
C pU =
,
dx
Ma2 1
2
dyL
.
C pL =
dx
Ma2 1
(15.194)
(15.195)
472
This is the same result as that obtained in Section 15.9, where the pressure on thin
airfoils was obtained by an approximation to the shock-expansion method. In fact,
for the purposes of calculating the velocity and pressure perturbations on the surface
of the airfoil, the linearized theory discussed in this section is equivalent to the weak
wave approximations of Section 15.9.
15.17 Exercises
15.1 Show that Equation (15.12) can be written in the form
X 3 + b X 2 + c X + d = 0,
(15.196)
where
X = sin2 ,
Ma12 + 2
b =
+ sin ,
2
Ma1
2 Ma12 + 1 + 1 2 1 2
sin ,
+
+
c=
2
Ma14
Ma12
d=
cos2
.
Ma14
Let
1
1
c b 2,
3
9
1
1 3
R = (b c 3 d)
b ,
6
27
Q=
D = Q 3 + R 2,
Demonstrate that the oblique shock solution only exists for D < 0 [i.e., when
Equation (15.196) possesses three real roots.] Show that the strong shock solution, s , and the weak shock solution, w , are given by
s
,
1 s
w
1
,
w = tan
1 w
s = tan1
473
where
1
s = b + 2 Q cos ,
3
1
w = b Q cos 3 sin ,
3
1 1 D
= tan
+ .
3
R
Here, = 0 if R 0, and = if R < 0.
15.2 Assuming that information propagates with respect to a two-dimensional supersonic flow pattern at the local sound speed, show that, in order for the flow
at some point P to aect the flow at some other point Q, the latter point must
lie between the (+) and () characteristics that pass through P.
15.3 Show that for a weak oblique shock with ,
1 p
,
1
p1
1 p
T
,
T1
p1
3
S ( + 1) p
,
R
12 2 p1
where = 2 1 , et cetera, and
p Ma12
.
p1
Ma 2 1
1
Here, is the wave angle, the Mach angle, 1 the deflection angle, the
ratio of specific heats, R the specific gas constant, and Ma1 the upstream Mach
number. Furthermore, p1 , 1 , T 1 , and S1 are the upstream pressure, density,
temperature, and specific entropy, respectively, whereas p2 , 2 , T 2 , and S2 are
the corresponding downstream quantities. Show, also, that
1
+
[(
1)/2]
Ma
Ma
1
.
Ma1
2
Ma1 1
15.4 Show that for a weak oblique shock
1 + ,
2 ,
474
Mach line
Ma1
wall
|2|
wall
Figure 15.13
Flow of Mach number Ma1 over convex corner of deflection angle |2 |.
where 1. Here, 1 and 2 are the Mach angles upstream and downstream
of the shock front, respectively. Moreover, is the wave angle, and 1
the deflection angle. Hence, deduce that the shock front subtends the same
angle, , with the Mach lines upstream and downstream of it. In other words,
the shock position is the average of the Mach line positions on either side
of it. Consider supersonic flow incident on a wedge of small nose angle, with
an afterbody, as illustrated in Figure 15.9(a). Assume that the shock front is
attached to the apex of the wedge, and that the flow downstream of the shock
is supersonic. Use the result just proved to show that the shape of the shock
front in the region of attenuation by expansion (i.e., in the region in which
the shock front is intersected by Mach lines emanating from the shoulder) is
parabolic. [Hint: Use the well-known optical result that a parabolic mirror
perfectly focuses a parallel beam of light rays.] (Leipmann & Roshko 1957.)
15.5 Show that, to second order in the deflection angle, , the relative change in
pressure across a weak oblique shock is written
2
Ma12
p Ma1
+
( + 1) Ma14 4 (Ma12 1) 2 ,
2
p1
4 (Ma1 1) 2
Ma12 1
where p = p2 p1 . Here, p1 is the upstream pressure, p2 the downstream
pressure, Ma1 the upstream Mach number, and the ratio of specific heats.
(Leipmann & Roshko 1957.)
15.6 An ideal gas of pressure, density, temperature, and Mach number p1 , 1 , T 1 ,
and Ma1 , respectively, flows over a convex corner that turns through an angle
475
|2 |, as shown in Figure 15.13. Consider a particular Mach line in the PrandtlMayer expansion fan that subtends an angle with the continuation of the
upstream wall, as shown in the figure. Let p, , T , Ma, ||, , and be the
pressure, density, temperature, Mach number, magnitude of the deflection angle, Mach angle, and Prandtl-Mayer function, respectively, on the Mach line
in question. Furthermore, let
1/2
+1
z,
= (Ma1 ) +
2
1
1 1/2
1
2
1/2
z1 = tan
(Ma1 1) ,
+1
where is the ratio of specific heats. Show that, inside the fan,
1
2
1/2 tan z
,
= cot (Ma1 1)
tan z1
1/2
+1
1
2
1/2
1
2
1/2 tan z
(z z1 ),
|| = tan (Ma1 1)
+
tan (Ma1 1)
tan z1
1
tan 2 z
,
tan 2 z1
1
2
1/2
1
2
1/2 tan z
= (Ma1 ) + tan (Ma1 1)
tan (Ma1 1)
tan z1
1/2
+1
+
(z z1 ),
1
2 /(1)
cos z
p = p1
,
cos z1
2/(1)
cos z
= 1
,
cos z1
2
cos z
T = T1
,
cos z1
Ma 2 = 1 + (Ma12 1)
1/2
+1
(z2 z1 ).
2 = 1
1
Show that , p, 0, Ma ,
max
1/2
+ 1
1 ,
2 1
476
incident shock
reected shock
Ma1
Ma2
Ma3
1
1
wall
Figure 15.14
Reflection of oblique shock by wall. Here, Ma1 , Ma2 , et cetera, are Mach numbers.
and
|| |max | max (Ma1 ),
in the limit that z /2. Hence, deduce that if |2 | > |max | then the the fan
only extends over the region |max | 1 , and the region |2 | < <
|max | is occupied by a vacuum (i.e., a gas with zero pressure and density).
Assuming that || 1, show that
1/2
Ma12 1
1
2
(z z1 ),
||
1 +1
Ma12
1 Ma12
||,
1
2
Ma12 1
and
Ma
p
1
||,
p1
Ma 2 1
1
1 p
,
1
p1
T
1 p
,
T1
p1
1 + [( 1)/2] Ma12 p
Ma
,
Ma1
p1
Ma12
477
same as those for a weak shock. (See Exercise 15.3.) Why is this not surprising? (Hint: The jump in specific entropy across a weak shock is third order in
the deflection angle.) Deduce that to second order in the deflection angle,
2
Ma12
p
Ma1
4
2
4
(Ma
1)
|| 2
(
+
1)
Ma
||
+
1
1
2
2
p1
4
(Ma
1)
2
1
Ma1 1
for a weak Prandtl-Mayer fan. (See Exercise 15.5.)
15.7 If an oblique shock is intercepted by a wall then it is reflected, as illustrated in
Figure 15.14. Calculate 1 1 , assuming that the shocks are suciently weak
that the approximate expressions of Section 15.4 can be used. Demonstrate
that, in this limit,
+ 1 Ma12
1 1 2
1 1 ,
2
4
Ma1 1
where 1 1 is the deflection angle of the incident shock, and the ratio
of specific heats. Show that 1 1 > 0 if 1 < Ma1 < [4/(3 )]1/2 and
1 1 < 0 if Ma1 > [4/(3 )]1/2 . Demonstrate that Ma2 Ma1 (1 ) and
Ma3 Ma1 (1 2 ), where
1 + [( 1)/2] Ma12
1 .
=
Ma12 1
15.8 Figure 15.15 shows a situation in which two oblique shocks of the same family
[in this case, the () family], produced by successive concave corners in a wall,
merge together to form a single stronger shock [of the () family]. Assuming
that the shocks are suciently weak that the approximate expressions of Section 15.4 can be used (which implies that 1 1 and 2 1), demonstrate
that
1 + 2 ,
where
Ma12
1 ,
Ma12 1
+ 1 Ma12
2 ,
2 =
4
Ma12 1
1 =
+1
4
478
shock
Ma1
Ma4
slipstream
shock shock
Ma3
Ma1
1
Ma2
2
wall
Figure 15.15
Merging of two oblique shocks of the same family produced by successive concave
corners of deflection angles 1 and 2 . Here, Ma1 , Ma2 , et cetera, are Mach numbers.
and also that Ma2 Ma1 (1 1 ) and Ma3 Ma4 Ma1 (1 1 2 ), where
2
1 + [( 1)/2] Ma1
1 =
1 ,
Ma12 1
1 + [( 1)/2] Ma12
2 .
2 =
Ma12 1
Demonstrate that the strength of the merged shock is approximately the sum
of the strengths of the two component shocks, and, hence, that the pressures
on either side of the slipstream shown in the figure are equal (at least, to first
order in 1 and 2 ). Finally, show that
Ma16
S4 S3 ( + 1)
1 2 (1 + 2 ),
R
4
(Ma12 1)3/2
where S is specific entropy, and R the specific gas constant. It is, thus, clear
that the specific entropy is not quite the same on either side of the slipstream.
15.9 If two shocks of opposite families intersect then they pass through one another,
but are slightly bent in the process, as illustrated in Figure 15.16. Assuming
that the shocks are suciently weak that the approximate expressions of Section 15.4 can be used (which implies that 1 1, 2 1, et cetera), show
479
(+)
()
Ma3
Ma2
2
Ma1
2
1
Ma3
Ma2
()
(+)
Figure 15.16
Crossing of two oblique shocks of dierent families. Here, Ma1 , Ma2 , et cetera, are
Mach numbers, and 1 , 2 , et cetera, are deflection angles.
that
p2 p1
p1
p3 p1
p1
p2 p1
p1
p3 p1
p1
2
Ma1
1 ,
2
Ma1 1
Ma12
(1 + 2 ),
Ma 2 1
1
2
Ma1
1 ,
Ma12 1
Ma12
(1 + 2 ),
Ma 2 1
1
480
c2
1
2
sin +
cos2 = 12 ,
+1
w
where is the Mach angle, the ratio of specific heats, c1 the sound speed at
the sonic point, and w the flow speed. Deduce that
b2 1
dw
= 2
Ma 2 1
d,
w
b + tan2
where
b2 =
and, hence, that
Ma 2 1
1
,
+1
1
1
dw
= tan1
tan + const.
w
b
b
Ma
1/2
1/2
1
+1
dw
1
2
2
=
Ma 1
tan
(Ma 1)
w
1
+1
$
1/2 %
.
tan1 Ma 2 1
where
1
1 n2
tan
.
n=
Ma12 1
481
Here, is the sweepback angle, and Ma1 the upstream Mach number. [Hint:
Resolve Ma1 into components normal and parallel to the leading edge. The
flow may then be studied in the plane normal to the leading edge using standard
thin-airfoil theory.] (Leipmann & Roshko 1957.)
15.13 Consider the problem of subsonic flow past a wave-shaped wall that was discussed in Section 15.13. Show that if the flow is bounded by a second wall
that lies at y = b (where b > 0) then
(x, y) =
cosh[m (b y)]
U
cos( x)
.
m
sinh(m b)
Show, on the other hand, that if the flow is bounded by a free surface at y = b
(where p = p ) then
(x, y) =
U
sinh[m (b y)]
cos( x)
.
m
cosh(m b)
482
A
Vectors and Vector Fields
A.1 Introduction
This appendix outlines those aspects of vector algebra, vector calculus, and vector
field theory that are useful in the study of fluid dynamics. Most of the material appearing in this appendix is reproduced from Fitzpatrick 2008. For more information
on vector algebra, vector calculus, and vector field theory see Schey 1992 and MilneThomson 2011.
straight-line, PQ (say), where the arrow indicates the direction of the displacement
(i.e., from point P to point Q). (See Figure A.1.) The magnitude of the vector is
represented by the length of the straight-line.
It is conventional to denote vectors by bold-faced symbols (e.g., a, F) and scalars
by non-bold-faced symbols (e.g., r, S ). The magnitude of a general vector, a, is
denoted |a|, or just a, and is, by definition, always greater than or equal to zero. It
is convenient to define a vector with zero magnitudethis is denoted 0, and has no
direction. Finally, two vectors, a and b, are said to be equal when their magnitudes
and directions are both identical.
483
484
Suppose that the displacements PQ and QR, shown in Figure A.2, represent the
vectors a and b, respectively. It can be seen that the result of combining these two
displacements is to give the net displacement PR. Hence, if PR represents the vector
c then we can write
c = a + b.
(A.1)
This defines vector addition. By completing the parallelogram PQRS , we can also
see that
PR = PQ + QR = PS + S R .
(A.2)
However, PS has the same length and direction as QR, and, thus, represents the
same vector, b. Likewise, PQ and S R both represent the vector a. Thus, the previous
equation is equivalent to
c = a + b = b + a.
(A.3)
We conclude that the addition of vectors is commutative. It can also be shown that
the associative law holds: that is,
a + (b + c) = (a + b) + c.
(A.4)
(A.5)
where a is a general vector. The negative of a is defined as that vector which has the
same magnitude, but acts in the opposite direction, and is denoted a. The sum of a
and a is thus the null vector: that is,
a + (a) = 0.
(A.6)
P
Figure A.1
A vector.
485
R
b
c=a+b
S
a
b
P
Figure A.2
Vector addition.
c=ab
b
a
c
Figure A.3
Vector subtraction.
(A.7)
486
z
y
O
Figure A.4
A right-handed Cartesian coordinate system.
opposite to a. These definitions imply that if n and m are two scalars then
n (m a) = n m a = m (n a),
(A.8)
(n + m) a = n a + m a,
(A.9)
n (a + b) = n a + n b.
(A.10)
487
(A.14)
(A.15)
(A.16)
represented by moveable line elements in space (i.e., in Figure A.2, PQ and S R represent the same vector), it follows that the components of a general vector are not
aected by a simple shift in the origin of a Cartesian coordinate system. On the other
hand, the components are modified when the coordinate axes are rotated.
Suppose that we transform to a new coordinate system, Ox y z , which has the
same origin as Oxyz, and is obtained by rotating the coordinate axes of Oxyz through
an angle about Oz. (See Figure A.5.) Let the coordinates of a general point P be
(x, y, z) in Oxyz and (x , y , z ) in Ox y z . According to simple trigonometry, these
two sets of coordinates are related to one another via the transformation
x = x cos + y sin ,
(A.17)
y = x sin + y cos ,
(A.18)
z = z.
(A.19)
Consider the vector displacement r OP. Note that this displacement is represented
488
y
y
z
P
x
O
Figure A.5
Rotation of the coordinate axes about Oz.
by the same symbol, r, in both coordinate systems, because the magnitude and direction of r are manifestly independent of the orientation of the coordinate axes. The
coordinates of r do depend on the orientation of the axes: that is, r (x, y, z) in
Oxyz, and r (x , y , z ) in Ox y z . However, they must depend in a very specific
manner [i.e., Equations (A.17)(A.19)] which preserves the magnitude and direction
of r.
The components of a general vector a transform in an analogous manner to Equations (A.17)(A.19): that is,
a x = a x cos + ay sin ,
(A.20)
(A.21)
az = az .
(A.22)
Moreover, there are similar transformation rules for rotation about Ox and Oy. Equations (A.20)(A.22) eectively constitute the definition of a vector: in other words,
the three quantities (a x , ay , az ) are the components of a vector provided that they
transform under rotation of the coordinate axes about Oz in accordance with Equations (A.20)(A.22). (And also transform correctly under rotation about Ox and Oy.)
Conversely, (a x , ay , az ) cannot be the components of a vector if they do not transform in accordance with Equations (A.20)(A.22). Of course, scalar quantities are
invariant under rotation of the coordinate axes. Thus, the individual components of
a vector (a x , say) are real numbers, but they are not scalars. Displacement vectors,
and all vectors derived from displacements (e.g., velocity and acceleration), automatically satisfy Equations (A.20)(A.22). There are, however, other physical quantities
that have both magnitude and direction, but are not obviously related to displacements. We need to check carefully to see whether these quantities are really vectors.
(See Sections A.7 and A.9.)
489
(A.23)
axes through
45 about Oz. In the new coordinate system, a (1/ 2, 1/ 2, 0) and
b (1/ 2, 1/ 2, 0), giving a % b = 1/2. Clearly, a % b is not invariant under
rotational transformation, so the previous definition is a bad one.
Consider, now, the dot product or scalar product:
a b a x b x + ay by + az bz = scalar number.
(A.24)
Let us rotate the coordinate axes though degrees about Oz. According to Equations (A.20)(A.22), a b takes the form
a b = (a x cos + ay sin ) (b x cos + by sin )
+ (a x sin + ay cos ) (b x sin + by cos ) + az bz
= a x b x + ay by + az bz
(A.25)
in the new coordinate system. Thus, a b is invariant under rotation about Oz. It
is easily demonstrated that it is also invariant under rotation about Ox and Oy. We
conclude that a b is a true scalar, and that the definition (A.24) is a good one.
Incidentally, a b is the only simple combination of the components of two vectors
that transforms like a scalar. It is readily shown that the dot product is commutative
and distributive: that is,
a b = b a,
a (b + c) = a b + a c.
(A.26)
The associative property is meaningless for the dot product, because we cannot have
(a b) c, as a b is scalar.
We have shown that the dot product a b is coordinate independent. But what is
the geometric significance of this property? In the special case where a = b, we get
a b = a x2 + ay2 + az2 = |a| 2 = a 2 .
(A.27)
490
ba
Figure A.6
A vector triangle.
Let us now investigate the general case. The length squared of AB in the vector
triangle shown in Figure A.6 is
(b a) (b a) = |a| 2 + |b| 2 2 a b.
(A.28)
(A.29)
(A.30)
ab
.
|a| |b|
(A.31)
(A.32)
491
Figure A.7
A vector area.
492
If we approach a limit, by letting the number of plane facets increase, and their areas
reduce, then we obtain a continuous surface denoted by the resultant vector area
S=
Si .
(A.35)
i
It is clear that the area of the rim seen looking down the x-axis is just S x . Similarly,
for the areas of the rim seen looking down the other coordinate axes. Note that it
is the rim of the surface that determines the vector area, rather than the nature of
the surface spanning the rim. So, two dierent surfaces sharing the same rim both
possess the same vector area.
In conclusion, a loop (not all in one plane) has a vector area S which is the
resultant of the component vector areas of any surface ending on the loop. The
components of S are the areas of the loop seen looking down the coordinate axes. As
a corollary, a closed surface has S = 0, because it does not possess a rim.
(A.36)
Is a b a proper vector? Suppose that a = (0, 1, 0), b = (1, 0, 0). In this case,
a b = 0. However,
if we rotate
axes through 45 about Oz then
the coordinate
(A.37)
Does this rather unlikely combination transform like a vector? Let us try rotating the
coordinate axes through an angle about Oz using Equations (A.20)(A.22). In the
new coordinate system,
c x = (a x sin + ay cos ) bz az (b x sin + by cos )
= (ay bz az by ) cos + (az b x a x bz ) sin
= c x cos + cy sin .
(A.38)
Thus, the x-component of a b transforms correctly. It can easily be shown that the
other components transform correctly as well, and that all components also transform
493
thumb
ab
middle finger
b
index finger
Figure A.8
The right-hand rule for cross products. Here, is less that 180.
correctly under rotation about Ox and Oy. Thus, a b is a proper vector. Incidentally, a b is the only simple combination of the components of two vectors that
transforms like a vector (which is non-coplanar with a and b). The cross product is
anti-commutative,
a b = b a,
(A.39)
distributive,
a (b + c) = a b + a c,
(A.40)
a (b c) (a b) c.
(A.41)
(A.42)
(A.43)
494
Thus,
|a b| = |a| |b| sin ,
(A.44)
where is the angle subtended between a and b. Clearly, a a = 0 for any vector,
because is always zero in this case. Also, if a b = 0 then either |a| = 0, |b| = 0, or
b is parallel (or antiparallel) to a.
Consider the parallelogram defined by the vectors a and b. (See Figure A.9.) The
scalar area of the parallelogram is a b sin . By convention, the vector area has the
magnitude of the scalar area, and is normal to the plane of the parallelogram, in the
sense obtained from a right-hand circulation rule by rotating a on to b (through an
acute angle). In other words, if the fingers of the right-hand circulate in the direction
of rotation then the thumb of the right-hand indicates the direction of the vector area.
So, the vector area is coming out of the page in Figure A.9. It follows that
S = a b,
(A.45)
(A.46)
(A.47)
a
S
b
a
Figure A.9
A vector parallelogram.
495
P
r
r sin
Figure A.10
A torque.
A.9 Rotation
Let us try to define a rotation vector whose magnitude is the angle of the rotation,
, and whose direction is parallel to the axis of rotation, in the sense determined by
a right-hand circulation rule. Unfortunately, this is not a good vector. The problem is that the addition of rotations is not commutative, whereas vector addition
is commuative. Figure A.11 shows the eect of applying two successive 90 rotations, one about Ox, and the other about the Oz, to a standard six-sided die. In the
left-hand case, the z-rotation is applied before the x-rotation, and vice versa in the
right-hand case. It can be seen that the die ends up in two completely dierent states.
In other words, the z-rotation plus the x-rotation does not equal the x-rotation plus
the z-rotation. This non-commuting algebra cannot be represented by vectors. So,
although rotations have a well-defined magnitude and direction, they are not vector
quantities.
But, this is not quite the end of the story. Suppose that we take a general vector
a, and rotate it about Oz by a small angle z . This is equivalent to rotating the
coordinate axes about Oz by z . According to Equations (A.20)(A.22), we have
a a + z ez a,
(A.48)
where use has been made of the small angle approximations sin and cos 1.
The previous equation can easily be generalized to allow small rotations about Ox
and Oy by x and y , respectively. We find that
a a + a,
(A.49)
= x e x + y ey + z ez .
(A.50)
where
496
z
x
z -axis
x-axis
x-axis
z -axis
Figure A.11
Eect of successive rotations about perpendicular axes on a six-sided die.
Clearly, we can define a rotation vector, , but it only works for small angle rotations
(i.e., suciently small that the small angle approximations of sine and cosine are
good). According to the previous equation, a small z-rotation plus a small x-rotation
is (approximately) equal to the two rotations applied in the opposite order. The fact
that infinitesimal rotation is a vector implies that angular velocity,
= lim
t0
,
t
(A.51)
497
c
b
Figure A.12
A vector parallelepiped.
vectors a, b, and c. (See Figure A.12.) This volume is independent of how the triple
product is formed from a, b, and c, except that
a b c = a c b.
(A.53)
So, the volume is positive if a, b, and c form a right-handed set (i.e., if a lies above
the plane of b and c, in the sense determined from a right-hand circulation rule by
rotating b onto c), and negative if they form a left-handed set. The triple product is
unchanged if the dot and cross product operators are interchanged,
a b c = a b c.
(A.54)
The triple product is also invariant under any cyclic permutation of a, b, and c,
a b c = b c a = c a b,
(A.55)
(A.56)
The scalar triple product is zero if any two of a, b, and c are parallel, or if a, b, and c
are coplanar.
If a, b, and c are non-coplanar then any vector r can be written in terms of them:
that is,
r = a + b + c.
(A.57)
Forming the dot product of this equation with b c, we then obtain
r b c = a b c,
so
(A.58)
rbc
.
(A.59)
abc
Analogous expressions can be written for and . The parameters , , and are
uniquely determined provided a b c 0: that is, provided the three vectors are
non-coplanar.
=
498
(A.61)
Let us try to prove the first of the previous theorems. The left-hand side and
the right-hand side are both proper vectors, so if we can prove this result in one
particular coordinate system then it must be true in general. Let us take convenient
axes such that Ox lies along b, and c lies in the x-y plane. It follows that b
(b x , 0, 0), c (c x , cy , 0), and a (a x , ay , az ). The vector b c is directed along
Oz: in fact, b c (0, 0, b x cy ). Hence, a (b c) lies in the x-y plane: in fact,
a (b c) (ay b x cy , a x b x cy , 0). This is the left-hand side of Equation (A.60) in
our convenient coordinate system. To evaluate the right-hand side, we need a c =
a x c x + ay cy and a b = a x b x . It follows that the right-hand side is
RHS = ( [a x c x + ay cy ] b x , 0, 0) (a x b x c x , a x b x cy , 0)
= (ay cy b x , a x b x cy , 0) = LHS,
(A.62)
da x day daz
da
,
,
.
(A.64)
dt
dt dt dt
Suppose that a is, in fact, the product of a scalar (t) and another vector b(t).
What now is the time derivative of a? We have
d
db x
d
da x
= ( b x ) =
bx +
,
dt
dt
dt
dt
(A.65)
db
da d
=
b+ .
dt
dt
dt
(A.66)
499
P
O
Q l
P
.
Figure A.13
A line integral.
Moreover, it is easily demonstrated that
da
db
d
(a b) =
b+a
,
dt
dt
dt
(A.67)
and
d
da
db
(a b) =
b+a
.
(A.68)
dt
dt
dt
Hence, it can be seen that the laws of vector dierentiation are analogous to those in
conventional calculus.
where dl = dx 2 + dy 2 .
For example, consider the integral of f (x, y) = x y 2 between P and Q along
the
two routes indicated in Figure A.14. Along route 1, we have x = y, so dl = 2 dx.
Thus,
1
Q
2
.
(A.70)
x y 2 dl =
x 3 2 dx =
4
P
0
500
Q = (1, 1)
2
2
P = (0, 0)
Figure A.14
An example line integral.
The integration along route 2 gives
1
x y 2 dx +
x y 2 dy
x=1
0
0
y=0
1
1
=0+
y 2 dy = .
3
0
x y 2 dl =
(A.71)
Note that the integral depends on the route taken between the initial and final points.
The most common type of line integral is that in which the contributions from dx
and dy are evaluated separately, rather that through the path-length element dl: that
is,
Q
!
"
f (x, y) dx + g(x, y) dy .
(A.72)
P
y dx + x 3 dy
(A.73)
along the two routes indicated in Figure A.15. Along route 1, we have x = y + 1 and
dx = dy, so
y dx + x 3 dy =
17
.
y dy + (y + 1)3 dy =
4
(A.74)
2
7
x dy +
y dx = .
x=1
4
1
y=1
(A.75)
1
0
Along route 2,
Q
P
y dx + x dy =
3
501
2
2
Q = (2, 1)
P = (1, 0)
Figure A.15
An example line integral.
Suppose that we have a line integral that does not depend on the path of integration. It follows that
Q
( f dx + g dy) = F(Q) F(P)
(A.76)
P
for some function F. Given F(P) for some point P in the x-y plane,
F(Q) = F(P) +
( f dx + g dy)
(A.77)
defines F(Q) for all other points in the plane. We can then draw a contour map of
F(x, y). The line integral between points P and Q is simply the change in height in
the contour map between these two points:
( f dx + g dy) =
(A.78)
Thus,
dF(x, y) = f (x, y) dx + g(x, y) dy.
(A.79)
Q
P
Q
3 x 2 y dx + x 3 dy = x 3 y P
(A.80)
502
For instance, if A is a force-field then the line integral is the work done in going from
P to Q.
For example, consider the work done by a repulsive inverse-square central field,
F = r/|r 3 |. The element of work done is dW = F dr. Take P = (, 0, 0) and
Q = (a, 0, 0). The first route considered is along the x-axis, so
a
a
1
1
1
(A.82)
W=
= .
2 dx =
x
a
x
The second route is, firstly, around a large circle (r = constant) to the point (a, , 0),
and then parallel to the y-axis. (See Figure A.16). In the first part, no work is done,
because F is perpendicular to dr. In the second part,
W=
0
1
y dy
1
=
= .
(a 2 + y 2 )3/2
(y 2 + a 2 )1/2 a
(A.83)
In this case, the integral is independent of the path. However, not all vector line
integrals are path independent.
S i 0
is a surface integral. For instance, the volume of water in a lake of depth D(x, y) is
V=
D(x, y) dS .
(A.85)
503
2
2
Q
O a
P
x
Figure A.16
An example vector line integral.
To evaluate this integral, we must split the calculation into two ordinary integrals.
The volume in the strip shown in Figure A.17 is
x2
D(x, y) dx dy.
(A.86)
x1
Note that the limits x1 and x2 depend on y. The total volume is the sum over all
strips: that is,
y2 x2 (y)
dy
D(x, y) dx
D(x, y) dx dy.
(A.87)
V=
y1
x1 (y)
Of course, the integral can be evaluated by taking the strips the other way around:
that is,
x2 y2 (x)
V=
dx
D(x, y) dy.
(A.88)
y1 (x)
x1
Interchanging the order of integration is a very powerful and useful trick. But great
care must be taken when evaluating the limits.
For example, consider
x y 2 dx dy,
(A.89)
where S is shown in Figure A.18. Suppose that we evaluate the x integral first:
1y
2
1y
x
y2
2
2
dy
(1 y)2 dy.
x y dx = y dy
=
(A.90)
2 0
2
0
Let us now evaluate the y integral:
1
2
y
y4
1
y3 +
dy =
.
2
2
60
0
(A.91)
504
y
y2
dy
y1
x1
x2
Figure A.17
Decomposition of a surface integral.
We can also evaluate the integral by interchanging the order of integration:
1x
1
1
x
1
2
(1 x)3 dx =
.
x dx
y dy =
3
60
0
0
0
(A.92)
In some cases, a surface integral is just the product of two separate integrals. For
instance,
x 2 y dx dy
(A.93)
1
1
1 1
1 1 1
2
2
= ,
dx
x y dy =
x dx
y dy =
3 2 6
0
0
0
0
(A.94)
dS 0
where is the angle subtended between the normal to the surface and the flow velocity.
505
y
(0, 1)
1y =x
(0, 0)
(1, 0)
Figure A.18
An example surface integral.
Analogously to line integrals, most surface integrals depend both on the surface
and the rim. But some (very important) integrals depend only on the rim, and not
on the nature of the surface which spans it. As an example of this, consider incompressible fluid flow between two surfaces S 1 and S 2 that end on the same rim. (See
Figure A.23.) The volume between the surfaces is constant, so what goes in must
come out, and
v dS =
v dS.
(A.96)
S1
It follows that
S2
v dS
(A.97)
depends only on the rim, and not on the form of surfaces S 1 and S 2 .
(A.98)
506
line joining the centroid to the apex. In fact, the height of the center of mass is given
by
)
z=
z dV
dV.
(A.99)
The bottom integral is just the volume of the pyramid, and can be written
a
a (az)/2 (az)/2
dz
dy
dx =
(a z)2 dz
dV =
(az)/2
=
(az)/2
a 1
(a 2 2 a z + z 2 ) dz = a 2 z a z 2 + z 3 /3 = a 3 .
0
3
(A.100)
Here, we have evaluated the z-integral last because the limits of the x- and y- integrals
are z-dependent. The top integral takes the form
(az)/2 (az)/2
a
a
z dz
dy
dx =
z (a z)2 dz
z dV =
(az)/2
0
a
(az)/2
a
(z a 2 2 a z 2 + z 3 ) dz = a 2 z 2 /2 2 a z 3 /3 + z4 /4
0
=
Thus,
1 4
a .
12
(A.101)
)
1 4 1 3 1
z =
a
a = a.
12
3
4
(A.102)
In other words, the center of mass of a pyramid lies one quarter of the way between
the centroid of the base and the apex.
A.18 Gradient
A one-dimensional function f (x) has a gradient d f /dx which is defined as the slope
of the tangent to the curve at x. We wish to extend this idea to cover scalar fields in
two and three dimensions.
Consider a two-dimensional scalar field h(x, y) that represents (say) height above
sea-level in a hilly region. Let dr (dx, dy) be an element of horizontal distance.
Consider dh/dr, where dh is the change in height after moving an infinitesimal distance dr. This quantity is somewhat like the one-dimensional gradient, except that
dh depends on the direction of dr, as well as its magnitude. In the immediate vicinity of some point P, the slope reduces to an inclined plane. (See Figure A.19.) The
largest value of dh/dr is straight up the slope. It is easily shown that for any other
direction
dh
dh
=
cos ,
(A.103)
dr
dr max
507
contours of h(x, y)
high
dr
low
O
Figure A.19
A two-dimensional gradient.
where is the angle shown in Figure A.19. Let us define a two-dimensional vector,
grad h, called the gradient of h, whose magnitude is (dh/dr)max, and whose direction is the direction of steepest ascent. The cos variation exhibited in the previous
expression ensures that the component of grad h in any direction is equal to dh/dr
for that direction.
The component of dh/dr in the x-direction can be obtained by plotting out the
profile of h at constant y, and then finding the slope of the tangent to the curve at
given x. This quantity is known as the partial derivative of h with respect to x at
constant y, and is denoted (h/x)y. Likewise, the gradient of the profile at constant
x is written (h/y) x. Note that the subscripts denoting constant x and constant y are
usually omitted, unless there is any ambiguity. It follows that in component form
h h
,
grad h
.
(A.104)
x y
Now, the equation of the tangent plane at P = (x0 , y0 ) is
hT (x, y) = h(x0 , y0 ) + (x x0 ) + (y y0 ).
(A.105)
h
,
x
h
.
y
(A.106)
508
(A.108)
Incidentally, the previous equation demonstrates that grad h is a proper vector, because the left-hand side is a scalar, and, according to the properties of the dot product,
the right-hand side is also a scalar provided that dr and grad h are both proper vectors
(dr is an obvious vector, because it is directly derived from displacements).
Consider, now, a three-dimensional temperature distribution T (x, y, z) in (say)
a reaction vessel. Let us define grad T , as before, as a vector whose magnitude is
(dT/dr)max, and whose direction is the direction of the maximum gradient. This
vector is written in component form
T T T
grad T
,
,
.
(A.109)
x y z
Here, T/x (T/x)y,z is the gradient of the one-dimensional temperature profile
at constant y and z. The change in T in going from point P to a neighboring point
oset by dr (dx, dy, dz) is
dT =
T
T
T
dx +
dy +
dz.
x
y
z
(A.110)
dT = grad T dr.
(A.111)
(A.112)
Q
This integral is clearly independent of the path taken between P and Q, so P grad T
dr must be path independent.
Q
Consider a vector field A(r). In general, the line integral P A dr depends on
the path taken between the end points. However, for some special vector fields the
integral is path independent. Such fields are called conservative fields. It can be
shown that if A is a conservative field then A = grad V for some scalar field V. The
proof of this is straightforward. Keeping P fixed, we have
Q
A dr = V(Q),
(A.114)
P
509
T = constant
grad T
dr
isotherms
Figure A.20
Isotherms.
where V(Q) is a well-defined function, due to the path independent nature of the line
integral. Consider moving the position of the end point by an infinitesimal amount
dx in the x-direction. We have
V(Q + dx) = V(Q) +
Q+dx
A dr = V(Q) + A x dx.
(A.115)
Hence,
V
= Ax,
x
(A.116)
(A.117)
The force field due to gravity is a good example of a conservative field. Now, if
A(r) is a force-field then A dr is the work done in traversing some path. If A is
conservative then
A dr = 0,
(A.118)
#
where corresponds to the line integral around a closed loop. The fact that zero net
work is done in going around a closed loop is equivalent to the conservation of energy
(which is why conservative fields are called conservative). A good example of a
non-conservative field is the force field due to friction.
Clearly, a frictional system
#
loses energy in going around a closed cycle, so A dr 0.
510
,
,
,
x y z
(A.119)
which is usually called the grad or del operator. This operator acts on everything
to its right in a expression, until the end of the expression or a closing bracket is
reached. For instance,
f f f
,
,
.
(A.120)
grad f = f
x y z
For two scalar fields and ,
grad ( ) = grad + grad
(A.121)
(A.122)
Suppose that we rotate the coordinate axes through an angle about Oz. By
analogy with Equations (A.17)(A.19), the old coordinates (x, y, z) are related to
the new ones (x , y , z ) via
x = x cos y sin ,
y = x sin + y cos ,
z=z.
Now,
(A.123)
(A.124)
(A.125)
y
z
x
+
+
,
=
x
x y ,z x
x y ,z y
x y ,z z
(A.126)
+ sin ,
= cos
x
x
y
(A.127)
x = cos x + sin y .
(A.128)
giving
and
It can be seen, from Equations (A.20)(A.22), that the dierential operator transforms in an analogous manner to a vector. This is another proof that f is a good
vector.
511
y + dy
z + dz
z
y
z
x + dx
Figure A.21
Flux of a vector field out of a small box.
A.20 Divergence
#
Let us start with a vector field A(r). Consider S A dS over some closed surface
S , where dS denotes an outward pointing surface element. This surface integral is
#usually called the flux of A out of S . If A represents the velocity of some fluid then
A dS is the rate of fluid flow out of S .
S
If A is constant in space then it is easily demonstrated that the net flux out of S
is zero. In fact,
A dS = A
dS = A S = 0,
(A.129)
A x
A x
dx dy dz =
dV,
x
x
(A.130)
where dV = dx dy dz is the volume element. (See Figure A.21.) There are analogous contributions from the sides normal to the y- and z-axes, so the total of all the
contributions is
A x Ay Az
+
+
A dS =
dV.
(A.131)
x
y
z
The divergence of a vector field is defined
div A = A =
A x Ay Az
+
+
.
x
y
z
(A.132)
512
S
interior contributions cancel
.
Figure A.22
The divergence theorem.
Divergence is a good scalar (i.e., it is coordinate independent), because it is the dot
product of the vector operator with A. The formal definition of A is
#
A dS
.
(A.133)
A = lim
dV0
dV
This definition is independent of the shape of the infinitesimal volume element.
One of the most important results in vector field theory is the so-called divergence
theorem. This states that for any volume V surrounded by a closed surface S ,
A dS =
A dV,
(A.134)
S
for any closed surface S . So, for two surfaces, S 1 and S 2 , on the same rim,
A dS =
A dS,
(A.136)
S1
S2
513
S2
S1
rim
Figure A.23
Two surfaces spanning the same rim (right), and the equivalent closed surface (left).
as illustrated in Figure A.23. (Note that the direction of the surface elements on
S 1 has been reversed relative to those on the closed surface. Hence, the sign of the
associated surface integral is also reversed.) Thus, if A = 0 then the surface
integral depends on the rim, but not on the nature of the surface that spans it. On the
other hand, if A 0 then the integral depends on both the rim and the surface.
#
Consider an incompressible fluid whose velocity field is v. It is clear that v
dS = 0 for any closed surface, because what flows into the surface must flow out
again. Thus, according to the divergence theorem, v dV = 0 for any volume.
The only way in which this is possible is if v is everywhere zero. Thus, the velocity
components of an incompressible fluid satisfy the following dierential relation:
v x vy vz
+
+
= 0.
x
y
z
(A.137)
A dS =
S
A dV.
(A.138)
If A = 0 then there is no net flux of lines out of any surface. Such a field is called
a solenoidal vector field. The simplest example of a solenoidal vector field is one in
which the lines of force all form closed loops.
514
,
,
,
x y z
(A.139)
A x Ay Az
+
+
,
x
y
z
(A.140)
which is a scalar field formed from a vector field. There are two ways in which we
can combine gradient and divergence. We can either form the vector field ( A)
or the scalar field (). The former is not particularly interesting, but the scalar
field () turns up in a great many physical problems, and is, therefore, worthy
of discussion.
Let us introduce the heat flow vector h, which is the rate of flow of heat energy
per unit area across a surface perpendicular to the direction of h. In many substances,
heat flows directly down the temperature gradient, so that we can write
h = T,
(A.141)
#
where is the thermal conductivity. The net rate of heat flow S h dS out of some
closed surface S must be equal to the rate of decrease of heat energy in the volume
V enclosed by S . Thus, we have
h dS =
c T dV ,
(A.142)
t
S
where c is the specific heat. It follows from the divergence theorem that
h = c
1
Figure A.24
Divergent lines of force.
T
.
t
(A.143)
515
Taking the divergence of both sides of Equation (A.141), and making use of
Equation (A.143), we obtain
( T ) = c
T
.
t
(A.144)
c T
.
t
(A.145)
T
T
T
(T ) =
+
+
x x
y y
z z
=
2 T 2 T 2T
+
+
2 T.
x 2
y 2
z 2
(A.146)
2
2
2
+ 2+ 2
2
x
y
z
(A.147)
is called the Laplacian. The Laplacian is a good scalar operator (i.e., it is coordinate
independent) because it is formed from a combination of divergence (another good
scalar operator) and gradient (a good vector operator).
What is the physical significance of the Laplacian? In one dimension, 2 T reduces to 2 T/x 2 . Now, 2 T/x 2 is positive if T (x) is concave (from above), and
negative if it is convex. So, if T is less than the average of T in its surroundings then
2 T is positive, and vice versa.
In two dimensions,
2T 2T
2T =
+
.
(A.148)
x 2
y 2
Consider a local minimum of the temperature. At the minimum, the slope of T
increases in all directions, so 2 T is positive. Likewise, 2 T is negative at a local
maximum. Consider, now, a steep-sided valley in T . Suppose that the bottom of the
valley runs parallel to the x-axis. At the bottom of the valley 2 T/y 2 is large and
positive, whereas 2 T/x 2 is small and may even be negative. Thus, 2 T is positive,
and this is associated with T being less than the average local value.
Let us now return to the heat conduction problem:
2T =
c T
.
t
(A.149)
It is clear that if 2 T is positive in some small region then the value of T there is less
than the local average value, so T/t > 0: that is, the region heats up. Likewise,
if 2 T is negative then the value of T is greater than the local average value, and
heat flows out of the region: that is, T/t < 0. Thus, the previous heat conduction
equation makes physical sense.
516
A.22 Curl
Consider a vector field A(r),
# and a loop that lies in one plane. The integral of A
around this loop is written A dr, where
dr is a line element of the loop. If A is
#
a conservative field then
A
=
and
A
Ay
dy dz.
y
Az Ay
A dr =
dS ,
y
z
(A.152)
(A.153)
Az Ay
(A.154)
dS x
A dr =
y
z
is valid for a small loop dS = (dS x , 0, 0) of any shape in the y-z plane. Likewise, we
can show that if the loop is in the x-z plane then dS = (0, dS y , 0) and
A x Az
A dr =
(A.155)
dS y .
z
x
517
z
4
z + dz
1
z
3
y
y + dy
y
Figure A.25
A vector line integral around a small rectangular loop in the y-z plane.
Finally, if the loop is in the x-y plane then dS = (0, 0, dS z ) and
Ay A x
dS z .
A dr =
x
y
(A.156)
A dr = curl A dS = |curl A| |dS| cos ,
where
(A.158)
Az Ay A x Az Ay A x
curl A =
,
y
z z
x x
y
(A.159)
and is the angle subtended between the directions of curl A and dS. Note that
curl A = A.
(A.160)
This demonstrates that A is a good vector field, because it is the cross product of
the operator (a good vector operator) and the vector field A.
Consider a solid body rotating about the z-axis. The angular velocity is given by
= (0, 0, ), so the rotation velocity at position r is
v = r.
(A.161)
518
dS
3
x
Figure A.26
Decomposition of a vector area into its Cartesian components.
[See Equation (A.52).] Let us evaluate
v on the axis of rotation. The x-component
#
is proportional to the integral v dr around a loop in the y-z plane.
# This is plainly
zero. Likewise, the y-component is also zero. The z-component
is
v dr/dS around
#
some loop in the x-y plane. Consider a circular loop. We have v dr = 2 r r with
dS = r2 . Here, r is the perpendicular distance from the rotation axis. It follows
that ( v)z = 2 , which is independent of r. So, on the axis, v = (0 , 0 , 2 ).
O the axis, at position r0 , we can write
v = (r r0 ) + r0 .
(A.162)
The first part has the same curl as the velocity field on the axis, and the second part
has zero curl, because it is constant. Thus, v = (0, 0, 2 ) everywhere in the
body. This allows us to form a physical picture of A. If we imagine A(r) as
the velocity field of some fluid then A at any given point is equal to twice the
local angular rotation velocity: that is, 2 . Hence, a vector field with A = 0
everywhere is said to be irrotational.
Another important result of vector field theory is the curl theorem:
A dr =
A dS,
(A.163)
C
for some (non-planar) surface S bounded by a rim C. This theorem can easily be
proved by splitting the loop up into many small rectangular loops, and forming the
integral around all of the resultant loops. All of the contributions from the interior
loops cancel, leaving just the contribution from the outer rim. Making use of Equation (A.158) for each of the small loops, we can see that the contribution from all of
519
the loops is also equal to the integral of A dS across the whole surface. This
proves the theorem.
One immediate consequence of the curl theorem is that A is incompressible.
Consider any two surfaces, S 1 and S 2 , that share the same rim. (See Figure A.23.) It
is clear from the
# curl theorem that A dS is the same for both surfaces. Thus,
it follows that A #dS = 0 for anyclosed surface. However, we have from the
divergence theorem that A dS = ( A) dV = 0 for any volume. Hence,
( A) 0.
(A.164)
() 0.
(A.165)
520
(A.166)
(a b) c = (c a) b (c b) a,
(A.167)
(a b) (c d) = (a c) (b d) (a d) (b c),
(a b) (c d) = (a b d) c (a b c) d,
( ) = + ,
(A.168)
(A.169)
(A.170)
(A B) = A ( B) + B ( A) + (A ) B + (B ) A,
= 2 ,
(A.171)
(A.172)
A = 0,
(A.173)
( A) = A + A ,
(A B) = B A A B,
= 0,
(A.174)
(A.175)
(A.176)
( A) = ( A) A,
2
( A) = A + A,
(A B) = ( B) A ( A) B + (B ) A (A ) B.
(A.177)
(A.178)
(A.179)
A.24 Exercises
A.1 The position vectors of the four points A, B, C, and D are a, b, 3 a + 2 b, and
521
A.5 Find the scalar product a b and the vector product a b when
(a) a = e x + 3 ey ez ,
b = 3 e x + 2 ey + ez ,
(b) a = e x 2 ey + ez ,
b = 2 e x + ey + ez .
A.6 Which of the following statements regarding the three general vectors a, b,
and c are true?
(a) c (a b) = (b a) c.
(b) a (b c) = (a b) c.
(c) a (b c) = (a c) b (a b) c.
(d) d = a + b implies that (a b) d = 0.
(e) a c = b c implies that c a c b = c |a b|.
(f) (a b) (c b) = [b (c a)] b.
A.7 Prove that the length of the shortest straight-line from point a to the straightline joining points b and c is
|a b + b c + c a|
.
|b c|
A.8 Identify the following surfaces:
(a) |r| = a,
(b) r n = b,
(c) r n = c |r|,
(d) |r (r n) n| = d.
Here, r is the position vector, a, b, c, and d are positive constants, and n is a
fixed unit vector.
A.9 Let a, b, and c be coplanar vectors related via
a + b + c = 0,
where , , and are not all zero. Show that the condition for the points
with position vectors u a, v b, and w c to be colinear is
+ + = 0.
u v w
A.10 If p, q, and r are any vectors, demonstrate that a = q + r, b = r + p,
and c = p + q are coplanar provided that = 1, where , , and are
scalars. Show that this condition is satisfied when a is perpendicular to p, b
to q, and c to r.
522
A.11 The vectors a, b, and c are non-coplanar, and form a non-orthogonal vector
base. The vectors A, B, and C, defined by
A=
bc
,
abc
523
524
B
Cartesian Tensors
B.1
Introduction
As we saw in Appendix A, many physical entities can be represented mathematically as either scalars or vectors, depending on their transformation properties under
rotation of the coordinate axes. However, it turns out that scalars and vectors are particular types of a more general class of mathematical constructs known as tensors.
In fact, a scalar is a tensor of order zero, and a vector is a tensor of order one. In
fluid mechanics, certain important physical entities (i.e., stress and rate of strain) are
represented mathematically by tensors of order greater than one. It is therefore necessary to supplement our investigation of fluid mechanics with a brief discussion of
the mathematics of tensors. For the sake of simplicity, we shall limit this discussion
to Cartesian coordinate systems. Tensors expressed in such coordinate systems are
known as Cartesian tensors. For more information on Cartesian tensors see Jeries
1961, Riley 1974, and Temple 2004.
B.2
526
tensor is not simply another name for an n-dimensional matrix. A matrix is merely an
ordered set of numbers. A tensor, on the other hand, is an ordered set of components
that have specific transformation properties under rotation of the coordinate axes.
(See Section B.3.)
Consider two vectors a and b that are represented as ai and bi , respectively, in
tensor notation. According to Section A.6, the scalar product of these two vectors
takes the form
(B.1)
a b = a1 b1 + a2 b2 + a3 b3 .
The previous expression can be written more compactly as
a b = ai bi .
(B.2)
Here, we have made use of the Einstein summation convention, according to which,
in an expression containing lower case roman subscripts, any subscript that appears
twice (and only twice) in any term of the expression is assumed to be summed
from 1 to 3 (unless stated otherwise). Thus, ai bi = a1 b1 + a2 b2 + a3 b3 , and
ai j b j = ai1 b1 + ai2 b2 + ai3 b3 . Note that when an index is summed it becomes a
dummy index and can be written as any (unique) symbol: that is, ai j b j and aip b p
are equivalent. Moreover, only non-summed, or free, indices count toward the order
of a tensor expression. Thus, aii is a zeroth-order tensor (because there are no free
indices), and ai j b j is a first-order tensor (because there is only one free index). The
process of reducing the order of a tensor expression by summing indices is known as
contraction. For example, aii is a zeroth-order contraction of the second-order tensor
ai j . Incidentally, when two tensors are multiplied together without contraction the
resulting tensor is called an outer product: for instance, the second-order tensor ai b j
is the outer product of the two first-order tensors ai and bi . Likewise, when two tensors are multiplied together in a manner that involves contraction then the resulting
tensor is called an inner product: for instance, the first-order tensor ai j b j is an inner
product of the second-order tensor ai j and the first-order tensor bi . It can be seen
from Equation (B.2) that the scalar product of two vectors is equivalent to the inner
product of the corresponding first-order tensors.
According to Section A.8, the vector product of two vectors a and b takes the
form
(a b)1 = a2 b3 a3 b2 ,
(B.3)
(a b)2 = a3 b1 a1 b3 ,
(B.4)
(a b)3 = a1 b2 a2 b1
(B.5)
Here,
i jk
+1
1
=
(a b)i = i jk a j bk .
(B.6)
if i, j, k is an even permutation of 1, 2, 3
if i, j, k is an odd permutation of 1, 2, 3
otherwise
(B.7)
Cartesian Tensors
527
is known as the third-order permutation tensor (or, sometimes, the third-order LeviCivita tensor). Note, in particular, that i jk is zero if one of its indices is repeated: for
instance, 113 = 212 = 0. Furthermore, it follows from Equation (B.7) that
i jk = jki = ki j = k ji = jik = ik j .
(B.8)
It is helpful to define the second-order identity tensor (also known as the Kroenecker delta tensor),
1
if i = j
i j =
.
(B.9)
0
otherwise
It is easily seen that
i j = ji ,
(B.10)
ii = 3,
(B.11)
ik k j = i j ,
(B.12)
i j a j = ai ,
(B.13)
i j ai b j = ai bi ,
(B.14)
i j aki b j = aki bi ,
(B.15)
et cetera.
The following is a particularly important tensor identity:
i jk ilm = jl km jm kl .
(B.16)
In order to establish the validity of the previous expression, let us consider the various
cases that arise. As is easily seen, the right-hand side of Equation (B.16) takes the
values
+1
if j = l and k = m j,
(B.17)
if j = m and k = l j,
(B.18)
otherwise.
(B.19)
Moreover, in each product on the left-hand side of Equation (B.16), i has the same
value in both factors. Thus, for a non-zero contribution, none of j, k, l, and m
can have the same value as i (because each factor is zero if any of its indices are
repeated). Because a given subscript can only take one of three values (1, 2, or 3),
the only possibilities that generate non-zero contributions are j = l and k = m, or
j = m and k = l, excluding j = k = l = m (as each factor would then have repeated
indices, and so be zero). Thus, the left-hand side of Equation (B.16) reproduces
Equation (B.19), as well as the conditions on the indices in Equations (B.17) and
(B.18). The left-hand side also reproduces the values in Equations (B.17) and (B.18)
because if j = l and k = m then i jk = ilm and the product i jk ilm (no summation)
is equal to +1, whereas if j = m and k = l then i jk = iml = ilm and the product
528
i jk ilm (no summation) is equal to 1. Here, use has been made of Equation (B.8).
Hence, the validity of the identity (B.16) has been established.
In order to illustrate the use of Equation (B.16), consider the vector triple product
identity (see Section A.11)
a (b c) = (a c) b (a b) c.
(B.20)
(B.21)
where use has been made of Equation (B.6). Employing Equations (B.8) and (B.16),
this becomes
(B.22)
[a (b c)]i = ki j klm a j bl cm = il jm im jl a j bl cm ,
which, with the aid of Equations (B.2) and (B.13), reduces to
[a (b c)]i = a j c j bi a j b j ci = [(a c) b (a b) c]i .
(B.23)
Thus, we have established the validity of the vector identity (B.20). Moreover, our
proof is much more rigorous than that given earlier in Section A.11.
B.3
Tensor Transformation
As we saw in Appendix A, scalars and vectors are defined according to their transformation properties under rotation of the coordinate axes. In fact, a scalar is invariant under rotation of the coordinate axes. On the other hand, according to Equations (A.49) and (B.6), the components of a general vector a transform under an
infinitesimal rotation of the coordinate axes according to
ai = ai + i jk j ak .
(B.24)
Here, the ai are the components of the vector in the original coordinate system, the ai
are the components in the rotated coordinate system, and the latter system is obtained
from the former via a combination of an infinitesimal rotation through an angle 1
about coordinate axis 1, an infinitesimal rotation through an angle 2 about axis 2,
and an infinitesimal rotation through an angle 3 about axis 3. These three rotations
can take place in any order. Incidentally, a finite rotation can be built up out of a great
many infinitesimal rotations, so if a vector transforms properly under an infinitesimal
rotation of the coordinate axes then it will also transform properly under a finite
rotation.
Equation (B.24) can also be written
ai = Ri j a j ,
(B.25)
Cartesian Tensors
529
where
Ri j = i j k ki j
(B.26)
is a rotation matrix (which is not a tensor, because it is specific to the two coordinate
systems it transforms between). To first order in the i , Equation (B.25) can be
inverted to give
ai = R ji aj .
(B.27)
This follows because, to first order in the i ,
Rik R jk = (ik l lik ) ( jk m m jk ) = ik jk l jk lik m ik m jk
= i j l li j l l ji = i j ,
(B.28)
where the dummy index m has been relabeled l, and use has been made of Equations (B.8), (B.10), and (B.12). Likewise, it is easily demonstrated that
Rki Rk j = i j .
(B.29)
(B.30)
(B.31)
Here, there has been much relabeling of dummy indices, and use has been made of
Equations (B.10) and (B.16). It can similarly be shown that
i jk Ril R jm Rkn = lmn .
(B.32)
(B.34)
(B.35)
530
(B.36)
Incidentally, because all tensors of the same order transform in the same manner, it
immediately follows that two tensors of the same order whose components are equal
in one particular Cartesian coordinate system will have their components equal in all
coordinate systems that can be obtained from the original system via rotation of the
coordinate axes. In other words, if
ai j = bi j
(B.37)
(B.38)
in all Cartesian coordinate systems (with the same origin and system of units as the
original system). Conversely, it does not make sense to equate tensors of dierent
order, because such an equation would only be valid in one particular coordinate
system, and so could not have any physical significance (because the laws of physics
are coordinate independent).
It can easily be shown that the outer product of two tensors transforms as a tensor
of the appropriate order. Thus, if
ci jk = ai b jk ,
(B.39)
and
ai = Ri j a j ,
(B.40)
(B.41)
then
ci jk = ai bjk = Ril al R jm Rkn bmn = Ril R jm Rkn al bmn
= Ril R jm Rkn clmn ,
(B.42)
(B.43)
where use has been made of Equation (B.14). Again, the previous proof is more rigorous than that given in Section A.6. The proof also indicates that the inner product
of two tensors transforms as a tensor of the appropriate order.
The result that both the inner and outer products of two tensors transform as
Cartesian Tensors
531
tensors of the appropriate order is known as the product rule. Closely related to this
rule is the so-called quotient rule, according to which if (say)
ci j = aik b jk ,
(B.44)
(B.45)
where use has been made of the fact that ci j and bi j transform as tensors, as well as
Equation (B.28). Rearranging, we obtain
(aik Ril Rkm alm ) bjk = 0.
(B.46)
However, the bi j are arbitrary, so the previous equation can only be satisfied, in general, if
(B.47)
aik = Ril Rkm alm ,
which is the correct transformation rule for a tensor. Incidentally, the quotient rule
applies to any type of valid tensor product.
The components of the second-order identity tensor, i j , have the special property
that they are invariant under rotation of the coordinate axes. This follows because
i j = Rik R jl kl = Rik R jk = i j ,
(B.48)
where use has been made of Equation (B.28). The components of the third-order
permutation tensor, i jk , also have this special property. This follows because
ijk = Ril R jm Rln lmn = i jk ,
(B.49)
where use has been made of Equation (B.31). The fact that i jk transforms as a proper
third-order tensor immediately implies, from the product rule, that the vector product
of two vectors transforms as a proper vector: that is, i jk a j bk is a first-order tensor
provided that ai and bi are both first-order tensors. This proof is much more rigorous
that that given earlier in Section A.8.
B.4
Tensor Fields
We saw in Appendix A that a scalar field is a set of scalars associated with every
point in space: for instance, (x), where x = (x1 , x2 , x3 ) is a position vector. We
also saw that a vector field is a set of vectors associated with every point in space:
532
for instance, ai (x). It stands to reason, then, that a tensor field is a set of tensors
associated with every point in space: for instance, ai j (x). It immediately follows that
a scalar field is a zeroth-order tensor field, and a vector field is a first-order tensor
field.
Most tensor fields encountered in physics are smoothly varying and dierentiable. Consider the first-order tensor field ai (x). The various partial derivatives of
the components of this field with respect to the Cartesian coordinates xi are written
ai
.
x j
(B.50)
Moreover, this set of derivatives transform as the components of a second-order tensor. In order to demonstrate this, we need the transformation rule for the xi , which is
the same as that for a first-order tensor: that is,
xi = Ri j x j .
(B.51)
xi
= Ri j .
x j
(B.52)
xi
= R ji .
xj
(B.53)
ai xk (Ril al )
ai
al
=
R jk = Ril R jk
,
=
x j xk x j
xk
xk
(B.54)
Thus,
Now,
which is the correct transformation rule for a second-order tensor. Here, use has been
made of the chain rule, as well as Equation (B.53). [Note, from Equation (B.26),
that the Ri j are not functions of position.] It follows, from the previous argument,
that dierentiating a tensor field increases its order by one: for instance, ai j /xk
is a third-order tensor. The only exception to this rule occurs when dierentiation
and contraction are combined. Thus, ai j /x j is a first-order tensor, because it only
contains a single free index.
The gradient (see Section A.18) of a scalar field is an example of a first-order
tensor field (i.e., a vector field):
()i =
.
xi
(B.55)
The divergence (see Section A.20) of a vector field is a contracted second-order tensor field that transforms as a scalar:
ai
a =
.
(B.56)
xi
Finally, the curl (see Section A.22) of a vector field is a contracted fifth-order tensor
that transforms as a vector
ak
.
(B.57)
( a)i = i jk
x j
Cartesian Tensors
533
The previous definitions can be used to prove a number of useful results. For
instance,
2
( )i = i jk
= i jk
= 0,
(B.58)
x j xk
x j xk
which follows from symmetry because ik j = i jk whereas 2 /xk x j = 2 /x j xk .
Likewise,
ak
ak
= 0,
(B.59)
i jk
= i jk
( a) =
xi
x j
xi x j
which again follows from symmetry. As a final example,
a j
bk
bk + i jk a j
i jk a j bk = i jk
xi
xi
xi
ak
bk
= bi i jk
ai i jk
= b ( a) a ( b).
x j
x j
(a b) =
(B.60)
(B.61)
or
(B.63)
ai j
dV,
x j
(B.64)
a
dV.
xi
ai j dS j =
S
or even
a dS i =
B.5
Isotropic Tensors
A tensor which has the special property that its components take the same value
in all Cartesian coordinate systems is called an isotropic tensor. We have already
encountered two such tensors: namely, the second-order identity tensor, i j , and the
third-order permutation tensor, i jk . Of course, all scalars are isotropic. Moreover,
as is easily demonstrated, there are no isotropic vectors (other than the null vector).
It turns out that the most general isotropic Cartesian tensors of second-, third-, and
fourth-order are i j , i jk , and i j kl + ik jl + il jk , respectively, where ,
, , , and are scalars. Let us prove these important results (Hodge 1961).
534
(B.65)
for arbitrary rotations of the coordinate axes. It follows from Equation (B.24) that,
to first order in the i ,
m mis a s j + m js ais = 0.
(B.66)
However, the i are arbitrary, so we can write
mis a s j + m js ais = 0.
(B.67)
(B.68)
2 ai j + a ji = a ss i j .
(B.69)
which reduces to
Interchanging the labels i and j, and then taking the dierence between the two
equations thus obtained, we deduce that
ai j = a ji .
Hence,
ai j =
a ss
i j ,
3
(B.70)
(B.71)
(B.72)
(B.73)
(B.74)
(B.75)
2 ai jk + ak ji + aik j = a s js ik + aiss jk ,
(B.76)
(B.77)
(B.78)
(B.79)
Cartesian Tensors
535
(B.80)
2 ai jk + a jik + ak ji = 0,
(B.81)
2 ai jk + a jik + aik j = 0,
(B.82)
2 ai jk + ak ji + aik j = 0.
(B.83)
(B.84)
(B.85)
(B.86)
(B.87)
(B.88)
(B.89)
(B.90)
(B.91)
a s jsl = jl ,
(B.92)
a s jks = jk .
(B.93)
(B.94)
aisks = a sisk ,
(B.95)
aissl = a sils .
(B.96)
536
Thus, we get
2 ai jkl + a jikl + ak jil + al jki = i j kl + ik jl + il jk ,
(B.97)
(B.98)
(B.99)
(B.100)
(B.101)
(B.102)
(B.103)
(B.104)
The latter two equations are obtained from the first via cyclic permutation of j, k,
and l, with i remaining unchanged. Summing Equations (B.102)(B.104), we get
2 (ai jkl +aikl j +ail jk )+3 (ai jlk +aik jl +ailk j ) = (++) (i j kl +ik jl +il jk ). (B.105)
It follows from symmetry that
ai jkl + aikl j + ail jk = ai jlk + aik jl + ailk j =
1
( + + ) (i j kl + ik jl + il jk ). (B.106)
5
This can be seen by swapping the indices k and l in the previous expression. Finally,
substitution into Equation (B.102) yields
ai jkl = i j kl + ik jl + il jk ,
(B.107)
= (4 )/10,
(B.108)
= (4 )/10,
(B.109)
= (4 )/10.
(B.110)
where
B.6
Exercises
B.1 Show that a general second-order tensor ai j can be decomposed into three
tensors
ai j = ui j + vi j + si j ,
Cartesian Tensors
537
538
B.5 If
A i j x j = xi
then and x j are said to be eigenvalues and eigenvectors of the second-order
tensor Ai j , respectively. The eigenvalues of Ai j are calculated by solving the
related homogeneous matrix equation
(Ai j i j ) x j = 0.
Now, it is a standard result in linear algebra that an equation of the previous
form only has a non-trivial solution when (Riley 1974)
det(Ai j i j ) = 0.
Demonstrate that the eigenvalues of Ai j satisfy the cubic polynomial
3 tr(A) 2 + (A) det(A) = 0,
where tr(A) = Aii and (A) = (Aii A j j Ai j A ji )/2. Hence, deduce that Ai j
possesses three eigenvalues1, 2 , and 3 (say). Moreover, show that
tr(A) = 1 + 2 + 3 ,
det(A) = 1 2 3 .
B.6 Suppose that Ai j is a (real) symmetric second-order tensor: that is, A ji = Ai j .
(a) Demonstrate that the eigenvalues of Ai j are all real, and that the eigenvectors can be chosen to be real.
(b) Show that eigenvectors of Ai j corresponding to dierent eigenvalues
are orthogonal to one another. Hence, deduce that the three eigenvectors of Ai j are, or can be chosen to be, mutually orthogonal.
(c) Demonstrate that Ai j takes the diagonal form Ai j = i i j (no sum) in
a Cartesian coordinate system in which the coordinate axes are each
parallel to one of the eigenvectors.
B.7 In an isotropic elastic medium under stress, the displacement ui satisfies
i j
2 ui
= 2 ,
x j
t
1 uk ul
+
i j = ci jkl
,
2 xl xk
where i j is the stress tensor (note that ji = i j ), the mass density (which
is a uniform constant), and
ci jkl = K i j kl + [ik jl + il jk (2/3) i j kl ].
the isotropic stiness tensor. Here, K and are the bulk modulus and shear
modulus of the medium, respectively. Show that the divergence and the
curl of u both satisfy wave equations. Furthermore, demonstrate that the
characteristic wave velocities of the divergence and curl waves are [(K +
4/3)/]1/2 and (/)1/2 , respectively.
C
Non-Cartesian Coordinates
C.1 Introduction
Non-Cartesian coordinates are often employed in fluid mechanics to exploit the symmetry of particular fluid systems. For example, it is convenient to employ cylindrical
coordinates to describe systems possessing axial symmetry. In this appendix, we
investigate a particularly useful class of non-Cartesian coordinates known as orthogonal curvilinear coordinates. The two most commonly occurring examples of this
class in fluid mechanics are cylindrical and spherical coordinates. (Incidentally, the
Einstein summation convention is not used in this appendix.) For more information
on orthogonal curvilinear coordinates see Milne-Thomson 2011.
(C.1)
In this case, u1 , u2 , u3 are said to constitute an orthogonal coordinate system. Suppose, further, that
(C.2)
e1 e2 e3 = 1
at all points in space, so that u1 , u2 , u3 also constitute a right-handed coordinate
539
540
(C.3)
Finally, a general vector A, associated with a particular point in space, can be written
A=
A i ei ,
(C.4)
i=1,3
where the ei are the local basis vectors of the u1 , u2 , u3 system, and Ai = ei A is
termed the ith component of A in this system.
Consider two neighboring points in space whose coordinates in the u1 , u2 , u3
system are u1 , u2 , u3 and u1 + du1 , u2 + du2 , u3 + du3 . It is easily shown that the
vector directed from the first to the second of these points takes the form
du2
du3
du1
hi dui ei .
(C.5)
dx =
e1 +
e2 +
e3 =
|u1 |
|u2 |
|u3 |
i=1,3
Hence, from (C.1), an element of length (squared) in the u1 , u2 , u3 coordinate system
is written
hi2 dui2 .
(C.6)
dx dx =
i=1,3
Here, the hi , which are generally functions of position, are known as the scale
factors of the system. Elements of area that are normal to e1 , e2 , and e3 , at a
given point in space, take the form dS 1 = h2 h3 du2 du3 , dS 2 = h1 h3 du1 du3 , and
dS 3 = h1 h2 du1 du2 , respectively. Finally, an element of volume, at a given point in
space, is written dV = h du1 du2 du3 , where
h = h1 h2 h3 .
(C.7)
(C.8)
i ui = 0.
h
(C.9)
The latter result follows from Equations (A.175) and (A.176) because (h12 /h) u1 =
u2 u3 , et cetera. Finally, it is easily demonstrated from Equations (C.1) and (C.3)
that
ui u j = hi2 i j ,
ui u j uk = h
i jk .
(C.10)
(C.11)
Consider a scalar field (u1 , u2 , u3 ). It follows from the chain rule, and the relation ei = hi ui , that
1
ui =
ei .
(C.12)
=
ui
h ui
i=1,3
i=1,3 i
Non-Cartesian Coordinates
541
1
.
hi ui
(C.13)
hi2
h
A=
ui
(Ai ei ) =
(hi Ai ui ) =
Ai
hi
h
i=1,3
i=1,3
i=1,3
1 h
h2
h
i
ui
Ai =
Ai ,
(C.14)
=
h
hi
h ui hi
i=1,3
i=1,3
where use has been made of Equations (A.174), (C.9), and (C.10). Thus, the divergence of A in the u1 , u2 , u3 coordinate system takes the form
1
h
Ai .
A=
(C.15)
h ui hi
i=1,3
We can write
(Ak ek ) =
(hk Ak uk ) =
(hk uk ) uk
A=
k=1,3
k=1,3
k=1,3
(hk Ak )
=
u j uk ,
u j
j,k=1,3
(C.16)
where use has been made of Equations (A.178), (C.8), and (C.12). It follows from
Equation (C.11) that
(hk Ak )
hi (hk Ak )
hi
ui u j uk =
i jk
.
( A)i = ei A =
u j
h u j
j,k=1,3
j,k=1,3
(C.17)
Hence, the components of A in the u1 , u2 , u3 coordinate system are
hi (hk Ak )
( A)i =
i jk
.
(C.18)
h u j
j,k=1,3
Now, 2 = [see Equation (A.172)], so Equations (C.12) and (C.15) yield
the following expression for 2 in the u1 , u2 , u3 coordinate system:
1 h
.
2 =
(C.19)
h ui hi2 ui
i=1,3
The vector identities (A.171) and (A.179) can be combined to give the following
expression for (A ) B that is valid in a general coordinate system:
(A ) B =
1
[(A B) (A B) ( A) B + ( B) A
2
A ( B) B ( A)] .
(C.20)
542
Making use of Equations (C.13), (C.15), and (C.18), as well as the easily demonstrated results
AB=
Ai Bi ,
(C.21)
i=1,3
AB=
i jk A j Bk ,
(C.22)
j,k=1,3
and the tensor identity (B.16), Equation (C.20) reduces (after a great deal of tedious
algebra) to the following expression for the components of (A ) B in the u1 , u2 , u3
coordinate system:
A j Bi A j B j h j Ai B j hi
[(A ) B]i =
+
.
h j u j
hi h j ui
hi h j u j
j=1,3
(C.23)
Note, incidentally, that the commonly quoted result [(A ) B]i = A Bi is only
valid in Cartesian coordinate systems (for which h1 = h2 = h3 = 1).
Let us define the gradient A of a vector field A as the tensor whose components
in a Cartesian coordinate system take the form
(A)i j =
Ai
.
x j
(C.24)
(C.25)
It thus follows from Equation (C.23), and the relation (ei ) j = ei e j = i j , that
(A)i j =
Ak hi
A j h j
1 Ai
+ i j
.
h j u j hi h j ui
h h uk
k=1,3 i k
(C.26)
The vector identity (A.177) yields the following expression for 2 A that is valid
in a general coordinate system:
2 A = ( A) ( A).
(C.27)
Making use of Equations (C.15), (C.18), and (C.19), as well as (C.21) and (C.22),
and the tensor identity (B.16), the previous equation reduces (after a great deal of
tedious algebra) to the following expression for the components of 2 A in the u1 , u2 ,
Non-Cartesian Coordinates
543
u3 coordinate system:
2
1 hi
1 h j
Aj
( A)i = Ai +
hi h j hi u j ui h j ui u j
j=1,3
2
A j h j hi2 Ai hi h j2
2
hi ui u j h
h j2 u j u j h
2
2
2
2 h 2
A j h 1 h j h j h h j h j
j
+
+
hi h j3 h j ui u j h h j2 ui h u j h ui u j h
2 hi
Ai 2 hi
.
(C.28)
2
2
hi h hi u j
u
h
+
hi h j2
Note, again, that the commonly quoted result ( 2 A)i = 2 Ai is only valid in Cartesian coordinate systems (for which h1 = h2 = h3 = h = 1).
In the cylindrical coordinate system, u1 = r, u2 = , and u3 = z, where r = x 2 + y 2 ,
= tan1 (y/x), and x, y, z are standard Cartesian coordinates. Thus, r is the perpendicular distance from the z-axis, and the angle subtended between the projection of
the radius vector (i.e., the vector connecting the origin to a general point in space)
onto the x-y plane and the x-axis. (See Figure C.1.)
A general vector A is written
A = A r er + A e + A z ez ,
(C.29)
where er = r/|r|, e = /||, and ez = z/|z|. Of course, the unit basis vectors
er , e , and ez are mutually orthogonal, so Ar = A er , et cetera.
As is easily demonstrated, an element of length (squared) in the cylindrical coordinate system takes the form
dx dx = dr 2 + r 2 d 2 + dz 2 .
(C.30)
Hence, comparison with Equation (C.6) reveals that the scale factors for this system
are
hr = 1,
(C.31)
h = r,
(C.32)
hz = 1.
(C.33)
544
e
r
er
z
x
Figure C.1
Cylindrical coordinates.
Thus, surface elements normal to er , e , and ez are written
dS r = r d dz,
(C.34)
dS = dr dz,
(C.35)
dS z = r dr d,
(C.36)
(C.37)
According to Equations (C.13), (C.15), and (C.18), gradient, divergence, and curl
in the cylindrical coordinate system are written
1
er +
e +
ez ,
(C.38)
r
r
z
1
1 A Az
(r Ar ) +
+
,
(C.39)
A=
r r
r
z
1 Az A
Ar Az
1
1 Ar
(r A )
A=
er +
e +
ez ,
r
z
z
r
r r
r
=
(C.40)
respectively. Here, (r) is a general scalar field, and A(r) a general vector field.
According to Equation (C.19), when expressed in cylindrical coordinates, the
Non-Cartesian Coordinates
545
1 2 2
=
+ 2.
r
+ 2
r r r
r 2
z
2
(C.41)
(C.42)
(C.43)
(C.44)
(C.45)
Here, the superscript T denotes a transpose. Thus, if the i j element of some secondorder tensor S is S i j then the corresponding element of ST is S ji . According to
6 in the cylindrical coordinate system are
Equation (C.26), the components of A
6 rr =
(A)
6 =
(A)
6 zz =
(A)
6 r = (A)
6 r =
(A)
6 zr =
6 rz = (A)
(A)
6 z =
6 z = (A)
(A)
Ar
,
r
1 A Ar
+ ,
r
r
Az
,
z
1 1 Ar A A
+
,
2 r
r
r
1 Ar Az
+
,
2 z
r
1 A 1 Az
+
.
2 z
r
(C.46)
(C.47)
(C.48)
(C.49)
(C.50)
(C.51)
Finally, from Equation (C.28), the components of 2 A in the cylindrical coordinate system are
Ar
2 A
,
r 2 r 2
2 Ar A
2,
( 2 A) = 2 A + 2
r
r
( 2 A)r = 2 Ar
( 2 A)z = 2 Az .
(C.52)
(C.53)
(C.54)
546
z
r
Figure C.2
Spherical coordinates.
In the spherical coordinate system, u1 = r, u2 = , and u3 = , where r = x 2 + y 2 + z 2 ,
= cos1 (z/r), = tan1 (y/x), and x, y, z are standard Cartesian coordinates. Thus,
r is the length of the radius vector, the angle subtended between the radius vector
and the z-axis, and the angle subtended between the projection of the radius vector
onto the x-y plane and the x-axis. (See Figure C.2.)
A general vector A is written
A = A r er + A e + A e ,
(C.55)
(C.56)
Hence, comparison with Equation (C.6) reveals that the scale factors for this system
are
hr = 1,
(C.57)
h = r,
(C.58)
h = r sin .
(C.59)
Non-Cartesian Coordinates
547
(C.60)
dS = r sin dr d,
(C.61)
dS = r dr d,
(C.62)
(C.63)
According to Equations (C.13), (C.15), and (C.18), gradient, divergence, and curl
in the spherical coordinate system are written
=
1
1
er +
e +
e ,
r
r
r sin
1
1 A
1 2
(r Ar ) +
(sin A ) +
,
r sin
r sin
r 2 r
1
1 A
A=
(sin A )
er
r sin
r sin
1 Ar 1
(r A ) e
+
r sin
r r
1
1 Ar
(r A )
+
e ,
r r
r
A=
(C.64)
(C.65)
(C.66)
respectively. Here, (r) is a general scalar field, and A(r) a general vector field.
According to Equation (C.19), when expressed in spherical coordinates, the Laplacian of a scalar field becomes
2
1 2
1
.
(C.67)
2 = 2
r
+ 2
sin
+
2
r
r r
r sin
r 2 sin 2
Moreover, from Equation (C.23), the components of (A )A in the spherical
coordinate system are
[(A ) A]r = A Ar
[(A ) A] = A A +
A2 + A2
,
r
Ar A cot A2
,
r
Ar A + cot A A
[(A ) A] = A A +
.
r
(C.68)
(C.69)
(C.70)
548
Ar
,
r
1 A Ar
+ ,
r
r
1 A Ar cot A
+
+
,
r sin
r
r
1 1 Ar A A
+
,
2 r
r
r
1 Ar A A
1
+
,
2 r sin
r
r
1 A 1 A cot A
1
+
.
2 r sin
r
r
(C.71)
(C.72)
(C.73)
(C.74)
(C.75)
(C.76)
A
2Ar
2 A 2 cot A
2
,
2
2
2
2
r
r
r
r sin
A
2
A
2 Ar
2
,
2
2
r 2
r
sin
r sin
A
2
2 cot A
Ar
( 2 A) = 2 A
+
+ 2
.
2
2
2
2
r
sin
r sin r sin
( 2 A) = 2 A +
(C.77)
(C.78)
(C.79)
C.5 Exercises
C.1 Find the Cartesian components of the basis vectors er , e , and ez of the cylindrical coordinate system. Verify that the vectors are mutually orthogonal.
Do the same for the basis vectors er , e , and e of the spherical coordinate
system.
C.2 Use cylindrical coordinates to prove that the volume of a right cylinder of radius a and length l is a 2 l. Demonstrate that the moment of inertia of a uniform cylinder of mass M and radius a about its symmetry axis is (1/2) M a 2 .
C.3 Use spherical coordinates to prove that the volume of a sphere of radius a is
(4/3) a 3. Demonstrate that the moment of inertia of a uniform sphere of
mass M and radius a about an axis passing through its center is (2/5) M a 2 .
C.4 For what value(s) of n is (r n er ) = 0, where r is a spherical coordinate?
Non-Cartesian Coordinates
549
550
D
Ellipsoidal Potential Theory
D.1 Introduction
Let us adopt the right-handed Cartesian coordinate system x1 , x2 , x3 . Consider a
homogeneous ellipsoidal body whose outer boundary satisfies
x12
a12
x22
a22
x32
a32
= 1,
(D.1)
where a1 , a2 , and a3 are the principal radii along the x1 -, x2 -, and x3 -axes, respectively. Let us calculate the gravitational potential (i.e., the potential energy of a unit
test mass) at some point P (x1 , x2 , x3 ) lying within this body. More information
on ellipsoidal potential theory can be found in Chandrasekhar 1969.
D.2 Analysis
Consider the contribution to the potential at P from the mass contained within a
double cone, whose apex is P, and which is terminated in both directions at the
bodys outer boundary. (See Figure D.1.) If the cone subtends a solid angle d
then a volume element is written dV = r 2 dr d, where r measures displacement
from P along the axis of the cone. Thus, from standard classical gravitational theory
(Fitzpatrick 2012), the contribution to the potential takes the form
0
r
G
G
dV
dV,
(D.2)
d =
r
0
r (r)
where r = |PQ|, r = |PR|, and is the constant mass density of the ellipsoid.
Hence, we obtain
r
r
1
r dr +
r dr d = G (r 2 + r 2 ) d.
(D.3)
d = G
2
0
0
The net potential at P is obtained by integrating over all solid angle, and dividing the
result by two to adjust for double counting. This yields
1
(D.4)
= G (r 2 + r 2 ) d.
4
551
552
Q
d
O
r
P
R
Figure D.1
Calculation of ellipsoidal gravitational potential.
From Figure D.1, the position vector of point Q, relative to the origin, O, is
x = x + r n,
(D.5)
where x = (x1 , x2 , x3 ) is the position vector of point P, and n a unit vector pointing
from P to Q. Likewise, the position vector of point R is
x = x + r n.
(D.6)
However, Q and R both lie on the bodys outer boundary. It follows, from Equation (D.1), that r and r are the two roots of
xi + r n i 2
= 1,
(D.7)
ai
i=1,3
which reduces to the quadratic
A r 2 + B r + C = 0,
(D.8)
where
A=
n2
i
,
ai2
xi n i
(D.9)
i=1,3
B=2
i=1,3
C=
x2
i
i=1,3
ai2
(D.10)
1.
(D.11)
ai2
553
According to standard polynomial equation theory (Riley 1974), r +r = B/A, and
r r = C/A. Thus,
r 2 + r 2 = (r + r )2 2 r r =
B2
C
2 ,
2
A
A
2
2 2
1 i=1,3 xi2 /ai2
i=1,3 xi ni /ai
1
d.
= G
2 +
2
2
2
2
2
i=1,3 ni /ai
i=1,3 ni /ai
The previous expression can also be written
2
2 2
1 i=1,3 xi2 /ai2
i, j=1,3 xi x j ni n j /(ai a j )
1
d.
+
= G
2
2
2
2
i=1,3 ni /ai
n 2 /a 2
i=1,3
(D.12)
(D.13)
(D.14)
However, the cross terms (i.e., i j) integrate to zero by symmetry, and we are left
with
2 2
4
2
2
x
n
/a
x
/a
1
1
i=1,3 i i
i=1,3 i
i
i
d.
+
(D.15)
= G
2
2
2
2
n
/a
2
2
i=1,3 i
i
n
/a
i=1,3 i
i
Let
J=
It follows that
1 J
=
ai ai
d
.
2
2
i=1,3 ni /ai
2 ni2/ai4
i=1,3
ni2 /ai2
2 d.
(D.16)
(D.17)
1
2
Ai xi ,
= G J
2
i=1,3
where
Ai =
J
1 J
.
ai2 ai ai
(D.18)
(D.19)
a12
a22
a32
0
0
(D.20)
(D.21)
554
/2
sin d
dt
= 4
a + b t2
/2
sin d
,
(a b)1/2
(D.22)
where
a=
sin2 cos2
+
,
a12
a32
(D.23)
b=
sin2 cos2
+
.
a22
a32
(D.24)
Hence, we obtain
J=
4 a1 a2 a32
/2
sin sec2 d
.
(a12 + a32 tan2 )1/2 (a22 + a32 tan2 )1/2
(D.25)
du
,
(D.26)
where
= (a12 + u)1/2 (a22 + u)1/2 (a32 + u)1/2 .
Now, from Equations (D.19), (D.26), and (D.27),
du
2 a1 a2 a3 du 1
Ai =
a
a
2
a
1 2 3
ai ai
ai2
0
0
2 a1 a2 a3
1
du
=
.
du = 2 a1 a2 a3
2
ai
0 ai
0 (ai + u)
Thus, Equations (D.18), (D.26), and (D.28) yield
3
2
i xi ,
= G M 0
4
i=1,3
where
0 =
i =
(D.27)
(D.28)
(D.29)
du
,
(D.30)
du
.
(ai2 + u)
(D.31)
Here, M = V and V = (4/3) a1 a2 a3 are the bodys mass and volume, respectively.
555
The total gravitational potential energy of the body is written (Fitzpatrick 2012)
1
U=
dV,
(D.32)
2
where the integral is taken over all interior points. It follows from Equation (D.29)
that
1
3
2
2
i ai .
U = G M 0
(D.33)
8
5 i=1,3
In writing
the previous expression, use has been made of the easily demonstrated
result xi2 dV = (1/5) ai2 V. Now,
2
ai2
1
d $u%
= +
,
du
i=1,3 (ai2 + u)
(D.34)
so
i ai2 =
i=1,3
i=1,3
ai2 du
(ai2 + u)
2
0
d $u% 1
+
du = 0 .
du
(D.35)
Hence, we obtain
U=
3
G M 2 0 .
10
(D.36)
D.3 Exercises
D.1 Demonstrate that the volume of an ellipsoid whose bounding surface satisfies
x12 x22 x32
+
+
= 1,
a12 a22 a32
is V = (4/3) a1 a2 a3 .
D.2 Demonstrate that the moments of inertia about the three Cartesian axes of a
homogeneous ellipsoidal body of mass M, whose bounding surface satisfies
(x1 /a1 )2 + (x2 /a2 )2 + (x3 /a3 )2 = 1, are
M 2
(a + a32 ),
5 2
M 2
(a + a32 ),
I2 =
5 1
M 2
I3 =
(a + a22 ).
5 1
I1 =
556
D.3 According to MacCullaghs formula (Fitzpatrick 2012), the gravitational potential a relatively long way from a body of mass M whose center of mass
coincides with the origin, and whose principal moments of inertial are I1 , I2 ,
and I3 (assuming that the principal axes coincide with the Cartesian axes),
takes the form
(x1 , x2 , x3 )
r
2r3
2r5
where r = (x12 + x22 + x32 )1/2 . Demonstrate that if the body in question is a homogeneous ellipsoid whose bounding surface satisfies (x1 /a1 )2 + (x2 /a2 )2 +
(x3 /a3 )2 = 1 then
GM GM 2 2 1 2
2
a x (x + x3 )
(D.37)
(x1 , x2 , x3 )
r
5r5 1 1 2 2
GM 2 2 1 2
GM 2 2 1 2
2
2
(x
(x
a
+
x
)
+
x
)
x
x
3
2 .
5r5 2 2 2 1
5r5 3 3 2 1
D.4 Show that the gravitational potential external to a homogeneous ellipsoidal
body of mass M, whose outer boundary satisfies (x1 /a1 )2 +(x2 /a2 )2 +(x3 /a3 )2 =
1, takes the form
3
i xi2 ,
(x1 , x2 , x3 ) = G M 0
(D.38)
4
i=1,3
where
0 =
i =
du
,
(ai2
du
,
+ u)
and = (a12 + u)1/2 (a22 + u)1/2 (a32 + u)1/2 . Here, is the positive root of
i=1,3
xi2
ai2 +
= 1.
E
Calculus of Variations
E.1
Indroduction
This appendix gives a brief overview of the calculus of variations. More information
on this topic can be found in Riley 1974.
E.2
Euler-Lagrange Equation
l=
(dx + dy )
2
2 1/2
[1 + y 2 (x)]1/2 dx,
(E.1)
(E.3)
557
558
Figure E.1
Dierent paths between points A and B.
where the end points of the integration are fixed. Suppose that y(x) y(x) + y(x).
The first-order variation in I is written
b
F
F
(E.4)
y + y dx,
I =
y
y
a
where y = d(y)/dx. Setting I to zero, we obtain
b
F
F
y + y dx = 0.
y
y
a
(E.5)
This equation must be satisfied for all possible small perturbations y(x).
Integrating the second term in the integrand of the previous equation by parts, we
get
b
b
F
F
d F
y
= 0.
(E.6)
y
dx
+
y dx y
y
a
a
However, if the end points are fixed then y = 0 at x = a and x = b. Hence, the last
term on the left-hand side of the previous equation is zero. Thus, we obtain
b
F
d F
y dx = 0.
(E.7)
y dx y
a
The previous equation must be satisfied for all small perturbations y(x). The only
way in which this is possible is for the expression enclosed in square brackets in the
integral to be zero. Hence, the functional I attains an extremum value whenever
F
d F
= 0.
(E.8)
dx y
y
Calculus of Variations
559
(E.9)
Next, suppose that F does not depend explicitly on x. Multiplying Equation (E.8) by
y , we obtain
d F
F
= 0.
(E.10)
y
y
dx y
y
However,
F
d F
d F
(E.11)
=y
+ y .
y
dx
y
dx y
y
Thus, we get
d F
F
F
+ y .
(E.12)
= y
y
dx
y
y
y
Now, if F is not an explicit function of x then the right-hand side of the previous
equation is the total derivative of F, namely dF/dx. Hence, we obtain
d F
dF
,
(E.13)
=
y
dx
y
dx
which yields
y
F
F = const.
y
(E.14)
Returning to the case under consideration, we have F = 1 + y 2 , according to
Equation (E.1) and (E.3). Hence, F is not an explicit function of y, so Equation (E.9)
yields
F
y
=
= c,
(E.15)
y
1 + y 2
where c is a constant. So,
c
y =
= const.
(E.16)
1 c2
Of course, y = constant is the equation of a straight-line. Thus, the shortest distance
between two fixed points in a plane is indeed a straight-line.
E.3
Conditional Variation
Suppose that we wish to find the function y(x) which maximizes or minimizes the
functional
b
F(y, y, x) dx,
(E.17)
I=
a
560
J=
G(y, y , x) dx
(E.18)
remains constant. We can achieve our goal by finding an extremum of the new functional K = I + J, where (x) is an undetermined function. We know that J = 0,
because the value of J is fixed, so if K = 0 then I = 0 as well. In other words,
finding an extremum of K is equivalent to finding an extremum of I. Application of
the Euler-Lagrange equation yields
d [ G]
d F
F
[ G]
+
= 0.
(E.19)
dx y
y
dx y
y
In principle, the previous equation, together with the constraint (E.18), yields the
functions (x) and y(x). Incidentally, is generally termed a Lagrange multiplier. If
F and G have no explicit x-dependence then is usually a constant.
As an example, consider the following famous problem. Suppose that a uniform
chain of fixed length l is suspended by its ends from two equal-height fixed points
that are a distance a apart, where a < l. What is the equilibrium configuration of the
chain?
Suppose that the chain has the uniform density per unit length . Let the x- and
y-axes be horizontal and vertical, respectively, and let the two ends of the chain lie
at (a/2, 0). The equilibrium configuration of the chain is specified by the function
y(x), for a/2 x +a/2, where y(x) is the vertical distance of the chain below its
end points at horizontal position x. Of course, y(a/2) = y(+a/2) = 0.
According to standard Newtonian dynamics, the stable equilibrium state of a
conservative dynamical system is one that minimizes the systems potential energy
(Fitzpatrick 2012). Now, the potential energy of the chain is written
U = g
y ds = g
a/2
a/2
y (1 + y 2 )1/2 dx,
(E.20)
where ds = dx 2 + dy 2 is an element of length along the chain, and g is the acceleration due to gravity. Hence, we need to minimize U with respect to small variations
in y(x). However, the variations in y(x) must be such as to conserve the fixed length
of the chain. Hence, our minimization procedure is subject to the constraint that
l=
ds =
a/2
a/2
(1 + y 2 )1/2 dx
(E.21)
remains constant.
It follows, from the previous discussion, that we need to minimize the functional
K = U + l =
a/2
a/2
( g y + ) (1 + y 2 )1/2 dx,
(E.22)
Calculus of Variations
561
where is an, as yet, undetermined constant. Because the integrand in the functional
does not depend explicitly on x, we have from Equation (E.14) that
y 2 ( g y + ) (1 + y 2 )1/2 ( g y + ) (1 + y 2 )1/2 = k,
(E.23)
(E.24)
(E.25)
(E.26)
x
z = + c,
h
(E.27)
Hence,
(E.28)
(E.30)
The final unknown constant, h, is determined via the application of the constraint
(E.21). Thus,
l=
a/2
a/2
(1 + y 2 )1/2 dx =
a/2
a/2
(E.31)
Hence, the equilibrium configuration of the chain is given by the curve (E.30), which
is known as a catenary (from the Latin for chain), where the parameter h satisfies
$ a %
l
= sinh
.
2h
2h
(E.32)
562
E.4
Multi-Function Variation
(E.33)
Here, the integrand F is now a functional of the F independent functions yi (x), for
i = 1, F . A fairly straightforward extension of the analysis in Section E.2 yields F
separate Euler-Lagrange equations,
F
d F
= 0,
dx yi
yi
(E.34)
for i = 1, F , which determine the F functions yi (x). If F does not explicitly depend
on the function yk then the kth Euler-Lagrange equation simplifies to
F
= const.
yk
(E.35)
E.5
Exercises
E.1 Find the extremal curves y = y(x) of the following constrained optimization
problems, using the method of Lagrange multipliers:
1
1
y 2 + x2 dx, such that 0 y 2 dx = 2.
0
(b) 0 y 2 dx, such that y(0) = y() = 0, and 0 y 2 dx = 2.
1
1 + y 2 dx = 2/3.
(c) 0 y dx, such that y(0) = y(1) = 1, and
(a)
E.2 Suppose P and Q are two points lying in the x-y plane, which is orientated
vertically such that P is above Q. Imagine there is a thin, flexible wire
connecting the two points and lying entirely in the x-y plane. A frictionless
bead travels down the wire, impelled by gravity alone. Show that the shape
Calculus of Variations
563
of the wire that results in the bead reaching the point Q in the least amount
of time is a cycloid, which takes the parametric form
x() = k ( sin ) ,
y() = k (1 cos ) ,
where k is a constant.
E.3 Find the curve y(x), in the interval 0 x p, which is of length , and
maximizes
p
y dx.
0
564
565
Bibliography
Abramowitz, M. (ed.), and Stegun, I. (ed.) 1965. Handbook of Mathematical Functions: with Formulas, Graphs, and Mathematical Tables. Dover.
Anderson, J.D. Jr. 2003. Modern Compressible Flow, with Historical Perspective,
3rd Edition. McGraw-Hill.
Arfken, G. 1985. Mathematical Methods for Physicists, 3rd Edition. Academic
Press.
Batchelor, G.K. 2000. An Introduction to Fluid Dynamics. Cambridge.
Cartwright, D.E. 1999. Tides. A Scientific History. Cambridge.
Chandrasekhar, S. 1967. An Introduction to the Study of Stellar Structure. Dover.
Chandrasekhar, S. 1969. Ellipsoidal Figures of Equilibrium. Yale.
Darwin, G.H. 1886. Proceedings of the Royal Society of London 41, 319.
Emanuel, G. 2000. Analytical Fluid Dynamics. CRC Press.
Eddington, A.S. 1926. The Internal Constitution of the Stars. Cambridge.
Faber, T.E. 1995. Fluid Dynamics for Physicists, 1st Edition. Cambridge.
Farrell, W.E. 1972. Reviews of Geophysics 10, 761.
Fitzpatrick, R. 2008. Maxwells Equations and the Principles of Electromagnetism.
Jones & Bartlett.
Fitzpatrick, R. 2012. An Introduction to Celestial Mechanics. Cambridge.
Fitzpatrick, R. 2013. Oscillations and Waves: An Introduction. CRC.
Hazeltine, R.D., and Waelbroeck, F.L. 2004. The Framework of Plasma Physics.
Westview.
Hodge, P.G., Jr. 1961. American Mathematical Monthly 68, 793.
Hughes, W., and Brighton, J. 1999. Schaums Outline of Fluid Dynamics, 3rd Edition. McGraw-Hill.
Jackson, J.D. 1962. Classical Electrodynamics, 2nd Edition. Wiley.
Jereys, H. 1961. Cartesian Tensors. Cambridge.
Lamb, H. 1928. Statics, Including Hydrostatics and the Elements of the Theory of
Elasticity, 3rd Edition. Cambridge.
566
Lamb, H. 1993 Hydrodynamics, 6th Edition. Cambridge.
Landau, L.D., and Lifshitz, E.M. 1987. Fluid Mechanics, 2nd Edition. ButterworthHeinemann.
Lighthill, J, 1978. Waves in Fluids. Cambridge.
Liepmann, H.W., and Roshko, A. 1957. Elements of Gasdynamics. Wiley.
Longuet-Higgins, M.S., and Pond, G.S. 1970. Philosophical Transactions of the
Royal Society of London A 266, 193.
Love, A.E.H. 1911. Some Problems of Geodynamics: Being an Essay to Which the
Adams Prize in the University of Cambridge was Adjudged in 1911. Cambridge.
Love, A.E.H. 1913. Proceedings of the London Mathematical Society 12, 309.
Love, A.E.H. 1927. A Treatise on the Mathematical Theory of Elasticity, 4th Edition. Cambridge.
Mestel, L. 1999. Phys. Reports 311, 295.
Milne-Thomson, L.M. 1958. Theoretical Aerodynamics, 4th Edition, Revised and
enlarged. Dover.
Milne-Thomson, L.M. 2011. Theoretical Hydrodynamics, 5th Edition. Dover.
Murray, C.D., and Dermott, S.F. 1999. Solar System Dynamics. Cambridge.
Press, W.H., Teukolsky, S.A., Vetterling, W.T., and Flannery, B.P. 2007. Numerical
Recipes: The Art of Scientific Computing, 3rd Edition. Cambridge.
Proudman, J. 1916. Proceedings of the London Mathematical Society 18, 51.
Reif, F. 1965. Fundamentals of Statistical and Thermal Physics. McGraw-Hill.
Riley, K.F. 1974. Mathematical Methods for the Physical Sciences. Cambridge.
Schey, H.M. 1992. Div, Grad, Curl, and All That, 2nd Edition. Norton.
Schlichting, H. 1987. Boundary Layer Theory, 7th Edition. McGraw-Hill.
Spiegel, M.R., Liu, J., and Lipschutz, S. 1999. Mathematical Handbook of Formulas and Tables, 2nd Edition. McGraw-Hill.
Temple, G. 2004. Cartesian Tensors. Dover.
Wong, B.R. 1998. Journal of Physics A: Mathematical and General 31, 1101.
Yoder, C.F., 1995. Astrometric and Geodetic Properties of Earth and the Solar
System, in Global Earth Physics: A Handbook of Physical Constants, Ahrens, T.
(ed.). American Geophysical Union.