Stokes' Theorem Notes

V13. Stokes' Theorem 1. Introduction; statement of the theorem. The normal form of Green's theorem generalizes in 3-space to the divergence theorem. What is the generalization to space of the tangential form of Green's theorem? It says where C is a simple closed curve enclosing the plane region R. Since the left side represents work done going around a closed curve in the plane, its natural generalization to space would be the integral $ F dr representing work done going around a closed curve in 3-space. In trying to generalize the right-hand side of (I), the space curve C can only be the boundary of some piece of surface S - which of course will no longer be a piece of a plane. So it is natural to look for a generalization of the form gjc F . dr = /L(something derived from F ) d S The surface integral on the right should have these properties: a) If curl F = 0 in Bspace, then the surface integral should be 0; (for F is then a gradient field, by V12, (4), so the line integral is 0, by V l l , (12)). b) If C is in the xy-plane with S as its interior, and the field F does not depend on z and has only a k-component, the right-hand side should be JL curl F d S . These things suggest that the theorem we are looking for in space is Stokes' theorem For the hypotheses, first of all C should be a closed curve, since it is the boundary of S , and it should be oriented, since we have to calculate a line integral over it. S is an oriented surface, since we have to calculate the flux of curl F through it. This means that S is two-sided, and one of the sides designated as positive; then the unit normal n is the one whose base is on the positive side. (There is no "standard" choice for positive side, since the surface S is not closed.) cubical surface: no boundary It is important that C and S be compatibly oriented. By this we mean that the right-hand rule applies: when you walk in the positive direction on C , keeping S to your left, then your head should point in the direction of n. The pictures give some examples. V. VECTOR INTEGRAL CALCLUS 2 + cu + The field F = M i N j P k should have continuous first partial derivatives, so that we will be able to integrate curl F. For the same reason, the piece of surface S should be piecewise smooth and should be finite- i.e., not go off to infinity in any direction, and have finite area. 2. Examples and discussion. i\ Example 1. Verify the equality in Stokes' theorem when S is the half of the sphere centered at the origin on which y 2 0, oriented so n makes an acute angle with the positive y-axis; take F = y i 22j x k . + : A : :, , ,' ........ ...,. ................. ". + I:: i I j/ Solution. The picture illustrates C and S . Notice how C must be directed to make its orientation compatible with that of S. We turn to the line integral first. C is a circle in the xz-plane, traced out clockwise in the plane. We select a parametrization and calculate: hydx+2xdy+zdz -c0s2tdt = = t s i n 2t [-2-4] 2lT = -IT. 0 For the surface S , we see by inspection that n = x i x2 y2 z2 = 1 on S . We calculate + + i curl F = j k ax ay a, y 2x + yj + z k ; this is a unit vector since +k; = -j (curlF).n=-y+z x Integrating in spherical coordinates, we have y = sin 4 sin 0, z = cos 4, d S = sin 4 d4 dB, since p = 1 on S; therefore / L C U r l F . ~ S= f l (s- y + z ) d ~ = l'llT(-sinmsin~+cos~)sin~d~d~; inner integral = sin 0 ($) ,sin24 - IT outer integral = -- cos 9 2 = -IT + 51 sin2,]: , = IT sin, which checks. n + + Example 2. Suppose F = x2 i x j z2 k and S is given as the graph of some function z = g(x, y), oriented so n points upwards. Show that h F . d r = area of R, where C is the boundary of S , com- patibly oriented, and R is the projection of S onto the xy-plane. i Solution. We have curl F = j k ax a, a, x2 x z = k . By Stokes' theorem, (cf. V9, (12)) STOKES' THEOREM V13. since n . k > 0, In . k 1 = n . k ; therefore h F . dr = SR dA = area of R . The relation of Stokes' theorem to Green's theorem. Suppose F is a vector field in space, having the form F = M(x, y) i + N(x, y) j , and C a simple closed curve in the xy-plane, oriented positively (so the interior is on your left you walk upright in the positive direction). Let S be its interior, compatibly oriented this means that the unit normal n to S is the vector k , and d S = dA. Then we get by the usual determinant method curl F = (N, - My)k ; since n = k , Stokes theorem becomes which is Green's theorem in the plane. The same is true for other choices of the two variables; the most interesting one is F = M(x, z) i + P ( x , z) k , where C is a simple closed curve in the xz-plane. If careful attention is paid to the choice of normal vector and the orientations, once again Stokes' theorem becomes just Green's theorem for the xz-plane. (See the Exercises.) Interpretation of curl F. Suppose now that F represents the velocity vector field for a three-dimensional fluid flow. Drawing on the interpretation we gave for the two-dimensional curl in Section V4, we can give the analog for 3-space. The essential step is to interpret the u-component of (curl F)O at a point Po, , where u is a given unit vector, placed so its tail is at Po. Put a little paddlewheel of radius a in the flow so that its center is at Po and its axis points in the direction u . Then by applying Stokes' theorem to a little circle C of radius a and center at P o , lying in the plane through Po and having normal direction u, we get just as in Section V4 (p. 4) that angular velocity of the paddlewheel = 1 - by Stokes' theorem, S being the circular disc having C as boundary; 1 27ra2 (curl F ) o . u (7ra2), since curl F . u is approximately constant on S if a is small, and S has area 7ra2; passing to the limit as a 4 0, the approximation becomes an equality: 1 angular velocity of the paddlewheel = - (curl F) . u . 2 yu ... V. VECTOR INTEGRAL CALCLUS 4 The preceding interprets (curl F ) o . u for us. Since it has its maximum value when u has the direction of (curl F)o, we conclude direction of (curl F)o = axial direction in which wheel spins fastest magnitude of (curl F)o = twice this maximum angular velocity. 3. Proof of Stokes' Theorem. We will prove Stokes' theorem for a vector field of the form P ( x , y, z) k . That is, we will show, with the usual notations, We assume S is given as the graph of z = f (x, y) over a region R of the xy-plane; we let C be the boundary of S , and C' the boundary of R. We take n on S to be pointing generally upwards, so that In.k 1 = n . k . To prove (3), we turn the left side into a line integral around C', and the right side into a double integral over R, both in the xy-plane. Then we show that these two integrals are equal by Green's theorem. To calculate the line integrals around C and C', we parametrize these curves. Let be a parametrization of the curve C' in the xy-plane; then gives a corresponding parametrization of the space curve C lying over it, since C lies on the surface z = f (x, y). Attacking the line integral first, we claim that This looks reasonable purely formally, since we get the right side by substituting into the left side the expressions for z and dz in terms of x and y: z = f (x, y), dz = fxdx f,dy. To justify it more carefully, we use the parametrizations given above for C and C' to calculate the line integrals. + = i, P ( x , y, f (x, y))(f.dx + fydy), the right side of (4). We now calculate the surface integral on the right side of (3), using x and y as the variables. In the calculation, we must distinguish carefully between such expressions as PI(x, y, f ) and V13. 5 STOKES' THEOREM &P(x, y, f ) . The first of these means: calculate the partial derivative with respect to the first variable x, treating x, y, z as independent; then substitute f (x, y) for z. The second means: calculate the partial with respect to x, after making the substitution z = f (x, y); the answer is a -P(x,y,f) = P1(x,y,f) + P 3 ( ~ , ~ , f ) .f i ax (We use PI rather than Px since the latter would be ambiguous - when you use numerical subscripts, everyone understands that the variables are being treated as independent.) With this out of the way, the calculation of the surface integral is routine, using the standard procedure of an integral over a surface having the form z = f (x, y) given in Section V9. We get dS = (-fxi - f y j i ~ u r l ( P ( x , ~ , z ) k=) a x O + k)dxdy, by V9, (13); j k dy dZ = P z ( x , y , z ) i - P l ( x , y , z ) j O P We have now turned the line integral into an integral around C' and the surface integral into a double integral over R. As the final step, we show that the right sides of (4) and (5) are equal by using Green's theorem Namely, we have Therefore, since fxy = fyx, four terms cancel, and we end up with which is precisely the integrand on the right side of (5). This completes the proof of Stokes' theorem when F = P ( x , y, z) k . In the same way, if F = M(x, y, z) i and the surface is x = g(y, z), we can reduce Stokes' theorem to Green's theorem in the yz-plane. If F = N(x, y, z) j and y = h(x, z) is the surface, we can reduce Stokes' theorem to Green's theorem in the xz-plane. + + Since a general field F = M i N j P k can be viewed as a sum of three fields, each of a special type for which Stokes' theorem is proved, we can add up the three Stokes' theorem equations of the form (3) to get Stokes' theorem for a general vector field. A difficulty arises if the surface cannot be projected in a 1-1 way onto each of three coordinate planes in turn, so as to express it in the three forms needed above: 6 V. VECTOR INTEGRAL CALCLUS In this case, it can usually be divided up into smaller pieces which can be so expressed (if some of these are parallel to one of the coordinate planes, small modifications must be made in the argument). Stokes' theorem can then be applied t o each piece of surface, then the separate equalities can be added up to get Stokes' theorem for the whole surface (in the addition, line integrals over the cut-lines cancel out, since they occur twice for each cut, in opposite directions). This completes the argument, manus undulans, for Stokes' theorem. Exercises: Section 6F