Calculus 4 Course Notes for AMATH 231 J. Wainwright? and G. Tenti Department of Applied Mathematics University of Waterloo April 22, 2010 1@J. Wainwright, revised December 2005Contents Preface Acknowledgements 1 Curves & Vector Fields 1.1 Curves in R” 1.1.1 Curves as vector-valued functions 1.1.2 Reparametrization of a curve 1.1.3 Limits 1.1.4 Derivatives 1.1.5 Arclength 1.2. Vector fields 1.2.1 Examples from physics . 1.2.2 Field lines of a vector field . 2 Line Integrals & Green’s Theorem 2.1 Line integral of a scalar field 2.1.1 Motivation and definition 2.1.2 Applications . 2.2 Line integral of a vector field 2.2.1 Motivation and definition 2.2.2 Applications . 2.2.3 Some technical ‘matters . 2.3. Path-independent line integrals : 2.3.1 First Fundamental Theorem for Line Integrals 2.3.2 Second Fundamental Theorem for Line Integrals 2.3.3 Conservative (i.e. gradient) vector fields 24 Green’s Theorem cee ee 2.4.1 The theorem. 242 Existence of a potential in 2.5 Vorticity and circulation 3 Surfaces & Surface Integrals 3.1 Parametrized surfaces 3.1.1 Surfaces as vector-valued functions 65 65 653.2 Surface Integrals 3.12 3.13 314 3.2.1 3.2.2 3.2.3 The tangent plane Surface area Orientation of a surface Scalar fields Vector fields Properties of surface integrals Gauss’ and Stokes’ Theorems 4.1. The vector differential operator V . 42 43 44 44d 412 4.13 Divergence and cutl of a vector field Identities involving V. . . . Expressing V in curvilinear coordinates Gauss’ Theorem 421 4.22 4.2.3 ‘The theorem Conservation laws and PDEs ‘The Generalized Divergence Theorem Stokes’ Theorem 43.1 43.2 43.3 ‘The theorem . Faraday’s law . ‘The physical interpretation of V x F ‘The Potential Theorems 441 4.4.2 Irrotational vector fields Divergence-free vector fields Fourier Series and Fourier Transforms Fourier Series 5.1 52 5.3 54 5.1 5.12 5.13 5.14 Calculating Fourier coefficients Pointwise convergence of a Fourier series Symmetry properties Complex form of the Fourier series Convergence of series of functions . 5.21 5.2.2 5.2.3, 5.24 A Second Look at Fourier Series 5.3.1 5.3.2 ‘The Fourier transform & Fouri 541 542 54.3, 5.44 545 ‘A deficiency of pointwise convergence . ‘The maximum norm and mean square norm ...........-~ Uniform and Mean Square Convergence. . ‘Termwise integration of series . Uniform and mean square convergence . Integration of Fourier series : integral ; The definition Calculating Fourier transforms using the definition Properties of the Fourier transfcrm . . Parseval's formula for a non-periodic function Relation between the continuous spectrum and the discrete spectrum ii 69 7 73 74 74 77 79 81 81 81 83 86 93 93 98 101 102 102 106 108 108 109 ul 115 115 uz 121 126 131 135 136 138 142 148, 150 150 152 154 154 157 160 164 1655.4.6 ‘Things are simpler in the frequency domain 5.5. Digitized signals ~ a glimpse cee 5.5.1 The sampling theorem . 5.5.2 The discrete Fourier transform (DFT). References Problem Sets Sample Tests & Exams iii 167 168 169 170 175 177 219Preface This course has two main themes: i) the calculus of veetor fields, and ii) Fourier analysis. ‘The first theme, which includes the famous integral theorems associated with the names of Green, Gauss and Stokes, represents the culmination of the traditional Calculus sequence. ‘This material provides the mathematical foundation for continuum mechanics (AMATH 361), fluid dynamics (AMATH 463), and electromagnetic theory (PHYS 252 & 253) and is of importance for partial differential equations (AMATH 353). The second theme, Fourier analysis, is built on the remarkable idea that a variety of complicated functions can be synthesized from pure sine and cosine functions. This material provides the mathematical foundation for signal and image processing (course under development) and is of importance for PDEs (AMATH 353) and quantum mechanics (AMATH 373).Acknowledgements ‘Thanks are expressed to Woei Chet Lim: for assistance with the figures. Stanley Lipshitz: for many discussions and suggestions concerning the material in Chap- ter 5, and for comments on the remainder of the notes, Bev Marshman: for detailed comments on the notes, ‘Ann Puncher: for excellent typesetting Dan Wolezuk: for creating the index. viiChapter 1 Curves & Vector Fields ‘The notion of a curve is of fundamental importence in Applied Mathematics courses because it arises in many physical and mathematical contexts. The path or trajectory of a particle moving in space, e.g. a satellite orbiting the earth, is a curve in R®, A section of a power line suspended between two pylons is another example of a curve in R°. Looking ahead in this course, the field lines of a vector field F(z, y, 2) are a family of curves in R° (more on this in Section 1.2). More generally, if one considers a physical system whose state at time t is a vector x(t) € R", then the evolution in time of the system will be described by a curve in R® (the value of n will depend on the complexity of the system) ‘The first mathematical description of a curve that one encounters is the graph of a function f :R + R, described by an equation y=f(2), aszsb. ‘This way of describing curves has a number of limitations: it is only valid for curves in R?, and it doesn’t give a simple description of the motion of particles (e.g. a particle moving in a circle). So one introduces vector-valued functions and uses a parametric description of curves (MATH 138), x=g(t) ast R®, it is important to make a distinction between the range of g, ie. the subset {g(i)|a R? defined by g(t) =(cost,sint), 0 R" by In general, let A : [o, 4] — [a,b] be function. Given g : [a,b] + R”, we define ar) =e(h(7)), a R" which gives a parametrization of this curve. At the same time, we know that there are infinitely many other functions & that give reparametrizations of C. Which parametrization we use for doing a calculation will be a matter of convenience. 1.1.3. Limits Given a vector-valued function g : [a, 6] + R", the Newton quotient at ¢ is Flot + a4 - st) (a7) for a non-zero change At. The derivative of g et t, denoted g/(t), is defined to be the vector that is the limit of the Newton quotient as At 0. So we need to think about the limit of fa vector-valued function Definition: Given a vector-valued function g defined in a neighbourhood of tp and a constant vector L, the statement limg(!) = L means that iim | e()-L|-0. Comments. 1) Since || g(t) — L || is Euclidean distance between the vectors g(t) and L, the definition captures the usual idea of “limit”, namely that the vector g(t) gets arbitrarily close to the unique vector L as t approaches to 2) When doing calculations with vector-valued functions one works with components. So we need to express the basic concepts in terms of components. Let B(t) = (gi(t),--..gn(tl), Lb = (Li,---, Ln), ‘Then p2pees Me (18) lim g(t) Li i to L = limatt)This equivalence follows on noting that I w(t) = I= [(gu(t) — £1)? +--+ Galt) = Ln)"], which implies lo(t)- Lil in terms of one-sided limits, with L+ # L-, then the curve would have a “break” in it. 0 1.1.4 Derivatives Definition Given a vector-valued function g : [a,8] > R", the derivative of g at t, denoted g'(t), is defined by 80) = Jim, 2 late + 40) - g(0)], (19) provided this limit exists. Comments: i) If g'(t) exists we say that g is differentiable at t, ii) In terms of component functions alt) =((t,--- gut), the Newton quotient is given by (eae ml sd ceil) 1 Ble Ad — 80] Ae arene Ae 10It follows from (1.8) and (1.9) that Bit) exists <> glit) exists, 1,2,...,%% and 8 (t) = (Hit),---s9n(t)) (1.10) In practice one calculates the derivative g/(t) using (1.10) and routine differentiation. o Physical interpretation of the derivative: Consider a curve C in R°, x = g(t), which represents the path of a particle. We write Ax = g(t + At) — g(t). (=) 4 ‘Thus by (1.9), g'(t) represent the rate of change of position with respect: to time t, Le. g(t) is the velocity v(t) of the particle. We write Then the Newton quotient (1.7) is (a) tete+ 40) ~ a0) = _ ax v= et) or v= = (uu) Writing the position function g(t) in terms of its component functions, g(t) = (x(t), w(t), 28), equations (1.10) and (1.11) imply that v(t) = (2'(t), uO), 2"), (1.12) giving the components of the velocity vector. The magnitude of the velocity vector I ¥( l= Vet? + oO? + 20)? (1.13) is of course the speed of the particle. If g’(t) is itself differentiable, the function g will have a second derivative denoted g"(t), and B"(t) = 2"), 9"), 2") in terms of the component functions. The secon¢ derivative represents the acceleration vector of the particle, and we write @x a(t)=g"(t), or alt) =F (1.14) uGeometrical interpretation of the derivative: Given a curve git), with g'(to) # 0, then g/(to) és a vector in the direction of the tangent line at the point x = g(to). This conclusion follows from Figure 1.10. Note that as At — 0 the increment ‘Ag tends to zero and the secant line approaches the position of the tangent line. ‘This result can also be confirmed on physical grounds: the velocity vector v(to) = g/(to) gives the instantaneous direction of motion and hence must be tangent to the path of the particle when t = to. x the secant line / _-— the tangent line AB = a(t, + AL)— g(t.) Figure 1.10: The secant line and tangent line to a curve, Equation of the tangent line: Given a curve x = g(t) with g’(fo) #0, the equation of the tangent line at x = g(tv) is x = B(to) + (t- to)B'(to), (1.15) as follows from Figure 1.11. Referring to Figure 1.11, we note that. 0Q = OP + PQ, and PQ is a multiple of g/(to) which we write as (t — to)g’(to). This choice of parameter on the tangent line ensures that t = to gives the point of tangency. In equation (1.15) we have defined a vector-valued function which we denote by Ly (t): Lro(t) = gto) + (t — to)a'(to) (1.16) 12Figure 1.11: The tangent line to a curve x = g(t) called the linearization of g at to Since the tangent line approximates the curve for t suffi ciently close to to, we obtain the linear approximation a(t) © Delt), or in full, g(t) © g(to) + (t — tog (to), (1.17) for ¢ sufficiently close to fy. We can write (1.17) more concisely in terms of the increments Ax = g(t) (to), At=t—to. Rearranging (1.17) gives Ax = (Aig'(to), (1.18) for At sufficiently close to zero. Terminology: i) When we say that a vector-valued function g : [a,4] + R” is of class C (or more briefly, is C}) we mean that g has a derivative function g’ that is itself a continuous function: think of C? as “C for continuous” and “1 for first derivative”. We shall usually work with C! curves, ie. curves described by C! functions, although one can imagine situations where the path of a particle would be a continuous but not a C' curve ‘There is a subtle point concerning C? curves. Based on experience with the graphs of scalar functions one might expect that a C? curve would look “smooth”. A C! curve g(t) can, however, have cusps (i.e. sharp spikes, or corners). See Examples 1.5 and 1.6 below. Physically, these occur when a particle stops, and changes its direction of motion. 13,ii) A function g : [a,6] + R” is piecewise C' means that there is a partition of [a,6], Sto 0, x is non-decreasing, ii) y > 0, y is periodic. = 0,41, 42,.. This curve is C!, but is not smooth. It is the path of a point on a uniformly rolling circle. (See Figure 1.12) 0 Comment: ‘A curve x = g(t), a 0 is a constant. This law is plausible physically, since one expects that the larger is the spatial temperature graclient, represented ty the gradient Vu, the larger will be the rate of heat transfer. The constant f, called the thermal conductivity, depends on the material, being bigger for a good conductor of heat than for a poor conductor (¢.8. Keopper/ Kyla > 1) a Having given some physical examples of vestor fields we now introduce the terminology formally. Let U be an open subset of R”. A vector field on U is a function F : U + R", whose domain is{/ and whose range is in R", i.e. F assigns to each x € / a unique vector F(x) € R". The vector fields we work with will usually be of class C', which means that the n component functions, F(x) = (Fil), +» Fa(x)) are C? functions. ‘We shall also assume that the domain U is a connected set, which means that U is not, the union of two or more disjoint sets, 1.2.2 Field lines of a vector field One visualizes a vector field F on an open set Uf C R® as a “field of vectors”, represented by arrows, attached to the points of /. The length of the vector at. a point gives the strength of the field at the point, and the arrow gives the direction of the field. It is also helpful to think of the family of curves in U with the property that at each point P the tangent to the curve through P equals the vector field evaluated at P. These curves are called the field lines' of the vector field. One thinks of a field line threading its way through the vector field, always following the direction of the vector field (see Figure 1.18). Figure 1.18: Two field lines of a vector field. "TThese curves are also called integral curves of the vector field. 21Figure 1.19: Field lines of a magnetic field, In the case of the velocity field of a fluid the field lines are simply the paths of the fluid particles (see Figures 1.16 and 1.18). In the case of a magnetic field one can do a simple experiment to visualize the field lines. ‘Take a sheet of paper and sprinkle it with iron filings ‘Then put a bar magnet under the paper and shake the paper slightly. You will observe the iron filings arranging themselves in lines going from one magnetic pole to the other. The strength of the magnetic field is revealed by the density of packing of the filings. In this way one obtains a picture of the field lines of the magnetic field We now consider the problem of determining the field lines of a given vector field F(x). Let the curve x = g(t), assumed C’, be a field line. Its tangent vector is g/(t), and thus the defining condition of a field line is written g'(t) = F(e(t)). (1.31) (see Figure 1.20). Equation (1.31) is a differential equation for the unknown vector-valued function g(t). If you want to find the field line through a given point xo then you should impose the initial condition B(to) = Xo. (1.32) DBs such as (1.31) are studied in depth in the course AM 451. In general they can only be solved numerically using a computer. For this course, however, it will be enough to concentrate on simple types that can be solved explicitly, as in the examples to follow. Example 1.12: Find the field lines of the vector field (1.33)2) = Fg) SN x= g(t) x, Figure 1.20: The field line of a vector field F through a given point xo. in R®, and sketch the field portrait. Solution: A field line x = g(t) satisfies 8'(0) = F(e(s)). In terms of components g(t) = (2(t),y(t)), this reads (o'(),y'(0) = (-v(), 20), giving de dy _ BIO a= el (1.34) These two coupled DEs can be written as a single DE by using the chain rule: dy _ dy de dt dr dt’ giving ay dey by equations (1.34). Solving this separable DE yields [uty~- fxd, giving where Cis a constant The conclusion is that any field line of the given vector field is a circle centred on the origin. Equations (1.34) show that the circles are traversed counterclockwise, giving Figure 1.21. We note that the field line through a given point (zo, yo) is the circle of radius V/z3 + yp specifying a point on the field line fixes the value of the constant C. 23Figure 1.21: The field lines x? + y? = C of the vector field F = (—y, 2). Exercise 1.3: The vector field in Example 1.12 can be interpreted physically as the velocity field of a rigidly rotating dise with unit angular velocity. Verify that do Figure 1.22: Velocity of a point on a rotating disc. Example 1.13: Find the field lines of the vector field = =u a F(x,y) = (=e 7) (1.35) on the open set U = R® — {(0,0)}, and sketch the “portrait” 24Solution: We have to solve the DEs dy yoy it a Pty dt ey Proceeding as in Example 1.12, these equations lead to the same DE at acy giving P+y=C, where C is a constant, as the field lines. 0 Figure 1.23: Field lines of the vector field (1.35). Comment: ‘The vector field (1.35) could represent the velocity field of a fluid swirling down a drain, Note that for Example 1.12, I F(a) I= V2? +9, i.e. the speed equals the distance from the origin, while for Example 1.13, I F@y) IF (0,0), Fa owe i.e. the speed equals the reciprocal of the distance from the origin. We have indicated this difference in Figures 1.21 and 1.23 by the size of the arrows. We note that Examples 1.12 and 1.13 illustrate that different vector fields can have the same field lines. Exercise 1.4: Find the field lines of the vector field F(x, y) = (,2y) in R®, and sketch the field portrait. 25Gradient vector fields: ‘The field lines of a vector field F(x) = Vu(x) in R? that is the gradient of a scalar field can be drawn without solving a DE. We know (Calculus 3) that the gradient Vu of a scalar field wis orthogonal to the level curves u = constant of the scalar field. It follows that the field lines of the vector field F = Vu are the orthogonal trajectories of the family of level curves of w. Figure 1.24: The field lines of F = Vu intersect the level curves u = constant orthogonally. 26Chapter 2 Line Integrals & Green’s Theorem In this chapter we define two types of integral that are associated with a curve in R”. 2.1 Line integral of a scalar field 2.1.1 Motivation and definition Consider a nuclear fuel rod, with linear mass density p (i.e. the physical dimensions are [p] = ML") and length ¢. If p is a constant, the mass m of the rod is simply m = pf. Suppose that due to manufacturing defects, the density depends on position on the rod, say p= 02), 0 Fx)Asi, (22) d a provided the limit exists (At; = ti — ti-1 as before) Thinking of the scalar field f as representing the linear density of the hanging wire, the term f(x;)As; approximates the mass of the ¢” segment of the wire, and so we expect the limit of the sum to give the total mass of the wire. Figure 2.2: The i segment of the hanging wire. Substituting x; = g(t) and using the approximation As; ~|| g/(t) || At: (see equation (1.23), we can approximate the sum in (2.2) as Not DY Feeds, = Y7 (a(t) Het) |] Ate, (23) 28for At; sufficiently small. In the limit as N + +00 and |At;| + 0 the right side of (2.3) equals the Riemann integral / f(g(¢)) || g(t) || dt, and we expect the approximation in (2.3) to become increasingly accurate. Comperison of (2.2) and (2.3) then motivates the definition to follow. Definition: Consider a curve C in R" given by x = g(t), a < t 0 for r € (a, J], the curve is described by with (7) = B(h(r)) (2.6)Then 5 e [Hee ico ee= [180 10) Nar (22) Proof: (outline) Differentiate (2.6) with respect to 7 and use the Chain Rule to get, Br) = Hr)e(A(7))- Aside: See Problem Set 1, #20. Since A'(r) > 0 it follows that WC) Ill (7) A"). (2.8) Substitute (2.6) and (2.8) into the integral on the right in (2.7). Since t = h(r) and a = h(a), = (A), the Change of Variable Theorem now implies the integral on the right equals the integral on the left. Exercise 2.2: Repeat Example 2.1 using a different parametrization of C, for example x = &(r) = (bcos2r,bsin2r), O<7 <4, and confirm that you get the same value for the line integral 2.1.2 Applications When working with a line integral in a physical context it is essential to keep in mind that an integral is the limit of a sum! (as in equation (2.2). An important special case arises if f(x) = 1 for all x € C. Then the definition (2.4) becomes ; fes= [reo ia Ler) which by equation (1.24) equals the arclength of the curve C. In words, the line integral of the constant scalar field f(x) = 1 along a curve C equals the arclength of C, i.e. one thinks of the line integral J ds as summing the elements of arclength along the curve C to give the z total arclength. We now give a glimpse of some other applications of the line integral of a scalar field i) An “everyday” example. ‘The base of a vertical curved fence is a curve C in the xy-plane, and its height at position (x,y) is h(x,y). What is the total area of the fence? "This statement applies to ANY tegral 31Consideration of the preliminary definition (2.2) leads to the conclusion that the area A is given by : | / (xy) G N\ Figure 2.3: A curved fence whose base is a plane curve C. One thinks of the line integral as summing the product of height h and element of arclength As. O Exercise 2.3: ‘The base of a vertical fence is given by x = g(t) is a positive constant, and the height at position the area of the fence is 27. 0 (bcos t, bsin® t), 0 < t < m, where b (x,y) is h(x,y) = b+ 3y. Show that, Reference: Marsden & Tromba, page 417, example 2, ii) Line integral of a linear density function It is helpful to think of the physical dimensions of quantities when interpreting a line integral | ds. (29) 2 In terms of the preliminary definition (2.2), we have wa T= jim Ls where As; is the arclength of the i” segment of C. Since? [LPGa)J[Asi] = YGa)]L, We use the symbol [J] to denote the dimensions of a physical quantity I. [feaAs, 32it follows that the dimensions of the integral J are (= [fle (2.10) In the hanging wire problem, f represents the linear mass density, i.e. [f] = ME, and the line integral I in (2.9) represents the total mass of the wire. Equation (2.10) gives U]=MI“L=M, which is consistent with the interpretation of I. As another example, think of the curve C as representing a conductor (e.g a copper wire) which is charged, with f being the linear charge density, i.e. [f] = (charge]L~!. The line integral I will give the total charge on the conductor. Equation (2.10) implies (0) = [charge}Z-L? = [charge which is again consistent with the interpretaticn of I. In general we can think of the scalar field f as representing the linear density of some “physical stuff”, which is distributed on a curve C. Then the line integral I in (2.9) gives the total amount of “physical stuff” on the curve. As before, equation (2.10) guarantees dimensional consistency. iii) Average value of a scalar field on a curve For a continuous function f : [a,6] + R, the average value of f over the interval [a, 8] is defined by be (fy Sei (e)dx fidz * i.e. the average value is the integral of f over [a,] divided by the length of the interval. We can generalize the concept of “average of a furction” to the case of a scalar field f that is continuous on a curve C in R”. In this case we define the average value by fas f= . (2.11) ie, the average value of f is the line integral of f along C divided by the arclength of C. Comment: If we choose a partition so that the NV’ curve segments of C are of equal length (i.e. As; = As, i=0,1,...,.N = 1), it follows from (2.11) and (2.2) that (Ne EY sor id ive. the “continuous average of {” is approximated by the “discrete average” of V values of the scalar field on C. 33Exercise 2.4: The steady state temperature of a circular metal plate of radius b centred on the origin in the xy-plane is given by u u(y) = ee? —¥ where uo is a constant. Show that the average temperature along the diameter y = (tan 8) is given by 2.2 Line integral of a vector field 2.2.1 Motivation and definition Consider a particle moving along the z-axis from « =a to = b under the action of a force F. If F is constant the work done on the particle is simply F W =(F-ij(b-a), where Fi is the component of F in the direction of motion (i is a unit vector in the 2 direction). If F = F(z) then the work done can be calculated as an integral (the limit of a Riemann sum), we [-ne. A similar but more difficult problem is to calculate the work done by a force field F(x) acting on a particle moving in a complicated way in space. Evidently we will need some sort of integral along the curve C that represents the path of the particle in space. This new type of integral is called the line integral of a vector field. Given a curve C in R", defined by x = g(t), a < t < b, where g is of class C', and a vector field F continuous on C, the line integral of F along C is denoted by [Pax We first give a tentative definition, motivated by the problem of calculating the work done by a force field. The development: parallels that of Section 2.1, 2.1.1 very closely. We introduce a partition of [a, 6), a=ty R" be a continuous vector field whose line integral is path independent in U. ir x (x) [ F-dx, (2.24) ho where Xp is a specified point, then, Vo(x) = F(x) (2.25) for all x EU. Proof: For simplicity we give the proof in R?. Wit @ defined by (2.24), we have to prove that a6 Ob an 7 Ph By Fr, where Fj and Fp are the components of F ‘The key idea is this: since the line integral is path-independent, we are free to make a special choice of the curve joining xp = (1,4) to x = (z,y), ie. to choose a “custom- designed” curve. Figure 2.6 shows the curve we need. Suitable parametrizations for C; and Care x= Bilt) = (tot) yoStsy x= g(t)= (ty), mo Sts 43Figure 2.6: A piecewise smooth curve C = C, UC; joining (zo, yo) to (x,y). Using these equations and the definition of ste) [rds= [eter fae y 2 = [Pena f° Fi(tyudde integral, It now follows from the first FTC for Riemann integrals (see equations (2.22) and (2.23)) that a6 je Ot Az), since we are treating y as a constant. Similarly we get the result for #2 by choosing a different path (do it!). 0 Terminology: ‘The significance of Theorem 2.1 is this: any vector field F whose line integral is path- independent can be written as the gradient of « C! scalar field. Such a vector field is called a gradient field. The scalar field y is called a potential for F, for physical reasons that we'll soon see. The level sets w(x) = C of the potential y are called equipotentials. In R®, we have eguipotential lines (2, y) = C and in B® we have equipotential surfaces p(a,y, 2) = C. ‘As an example, we note that the vector field F(2,y) given by (2.21) (the electric field due to a point of charge q at the origin) is derivable from the potential ka Very ie. E(x,y) = Vo(, y) (verify this!). The equipotential lines are given by (2, y) = constant, ie. vey) (2.26) a? +y? = const, 442.3.2 Second Fundamental Theorem for Line Integrals Continuing the train of thought from the previous subsection we ask Q: In elementary calculus we learned that if G,g : [a,8] > R are such that g is continuous and G’ = g, then [ oerde= 0 - cto, (221) (the second FTC). Is there a way to extend this result to line integrals? A: Yes, provided the vector field F is a gradient field, ie. F = V¢. ‘This generalization is the Second Fundamental Theorem for line integrals. ‘Theorem 2.2: Let F : U — R” be a continuous vector field on a connected open set U CR", and let x1, x2 be two points in U. If F = Vd, where @: U > R is a C! scalar field, and C is any curve in U joining x: to Xo, then [ Petx= ote) ~ 66) (2.28) 2 Proof: Let C be given by x=e(t) hstSt, so that Xi =g(ti), x2 = g(t). (2.29) By the hypothesis, ee [vo dx e = [ Vd(g(t)) + 8i(t)dt (by definition of line integral) = * A aga(oniae (by the Chain Rule) = G(g(t2)) — o(g(ts)) (by the second FTC) = 9(%2) — $04). (by (2.29) O 45Comment: ‘The significance of Theorem 2.2 is two-fold: i) if F is a gradient field, the line integral f F-dx only depends on the end points of the 2 curve C, and hence is path-independent, ii) if the potential ¢ is known, then equation (2.28) gives the value of the line integral immediately. Exercise 2.10: ‘We have seen that the vector field __ x y Rew =-4 (aa ara) on U = R? — {(0,0)} is a gradient field with potential kq Oty) = Va In exercise 9 you showed that the line integral of B along each of 2 curves joining (1,0) to (0,1) equalled zero. Use Theorem 2.2 to verify this result. 0 Looking ahead: So far (Theorems 2.1 and 2.2) we have established that [Pas is path-independent if and é only if F is a gradient field. Moreover, if F is a gradient field and we can find a potential $, then the line integral f F-dx can be quickly evaluated. Now we are faced with two problems: i) if we are given a vector field F, how can we tell quickly whether it is a gradient field? ii) if we know F is a gradient field, how do we find a potential 4? Answering the first question requires the famous Green's theorem, while the second is more straightforward. But before dealing with these questions we first discuss the physical significance of the Second Fundamental Theorem for line integrals. 2.3.3 Conservative (i.e. gradient) vector fields Thinking of the vector field F in Theorem 2.2 as a force field, the line integral [ra equals é the work done by the force field on a particle as the particle moves along the curve C from xX) to x2. The theorem asserts that if the force field is a gradient field, F = V@, then the work done depends only on the potential at the end points x; and x2 46In this physical context, it is customary, to define a scalar field V -@, so that F=-vv. (2.30) Theorem 2.2 then has the form [rex = -V (x2) + V(x). (2.31) o Since “work” is the same as “energy”, physicists call V(x) the potential energy of the particle at position x, when moving under the action of F. Comment: ‘The minus sign in equation (2.30) becomes appropriate when one thinks, for example of the force field due to the earth’s gravitational field. The potential energy of a particle increases if its distance from the earth’s centre increases i.e. VV points radially outwards, while the gravitational force field F acts radially inwards. One can also relate the work done to the kinetic energy K of the particle, defined by mv. (2.32) Describing the path C of the particle by x = r(), th <¢ < ta, the work done can be written J Fax = [" Fer(t) (ade. (2.33) 2 But Newton's second law tells us that mr"(t) = F(r(é)) It follows that F(r(¢)) + ¥'(t) = mv'(t)- v(0) {since ¥(@) = v(t) Amlv(t)- v())’ (property of the derivative) = SKw) (by (2.32). Thus, by (2.33) and the PCH, [Pix= Ke) Ku) (2.34) e Equation (2.34) and (2.31) gives K(t) + V(r) = K(b) + V(r), (2.35) for any two times t, and fy. In words, for a gradient force field one can define a potential energy V of a particle in such a way that the sum of the potential energy and kinetic energy K is constant, i.e. conservation of energy holds. a7It is for this reason that when vector fields are thought of a force fields, gradient fields are also called conservative fields. Comment: In many applications in the real world, conservation of energy does not have the simple form of (2.35), because, for example, of energy losses due to friction - think of the space shuttle re-entering the atmosphere. Dissipative, i.e. non-conservative forces have also to be considered. It is nevertheless important to be eble to find out whether a given force field is conservative, and this is the problem we now consider. 2.4 Green’s Theorem In this section we introduce Green's theorem, and discuss a number of applications, including how to spot conservative/gradient, vector fields in R?, 2.4.1 The theorem ‘We need some additional terminology related to curves. Consider a curve C in R" given by x=g(t), oStSd, g continuous. i) C is a closed curve means that g(a) = g(0). ii) C is a simple closed curve means that g(a) = g(d) and g is a one-to-one function on the interval a < ¢ F isa gradient field in UCR" in UCR" (2.47) We begin the final stage of the discussion by establishing that s Hi Fedx_ is path-independent in Ut” equivalent to + [Pa =0 for all simple closed curves in U”. ‘The reason for doing this is that Green’s theorem deals with line integrals around simple closed curves. Proposition 2.3: Let U be an open subset of R". A continuous vector field F : U — R” is path independent, in U if and only if [Pate = 0 for every simple closed curve in U. Proof: 1) Suppose f F-dx is path-independent in YU. Let C be a simple closed curve in U. c Decompose C into C; and C2 as in figure 2.10. Then [-ff[-f-L ca & A -t since the line integral is path independent 53.x x 4 - 0) O x %, Figure 2.10: Decomposing a simple closed curve C into C; and C2 2) Suppose f F-dx = 0 for every simple closed curve in Uf. Let x1,x2 be any two points z in U, and let C; and Cp be two curves joining x; to x2 which do not intersect each other. Then C; U (C2) is a simple closed curve C. It follows that FEEL & & & a & x, x fc -<, q ey x, x, Figure 2.11: Two curves C;,C2 joining x; to xz form a simple closed curve € = C, U(—Ca). If C, and Cy intersect each other, introduce a third curve Cy that does not intersect either C; or Cp. It follows as before that [-f~ [-] a & a & We have thus shown that | = / for any two curves joining x1 to x2, i.e. the line a & integral is path-independent. 0Figure 2.12: A third curve Cs avoids intersections, Combining Proposition 2.3 with the result (2.47) gives the following: [ra closed curve in UCR" in UCR (2.48) 0 forevery simple <> F isa gradient field We now restrict our considerations to R®. Suppose that the C? vector field F : U — R? is a gradient field, i.e. F = V¢, or in component form, war (2) It follows that, Om OF, _ Fe a6 dz Oy dady dyox since ¢ is of class C?. This result, with (2.48), means that if F is a gradient field in R?, the formula in Green’s theorem,namely os [] (@-B)ew a is identically satisfied. ‘This formnla also suggests that if 9% — 25 = 0 in U, then f Fdx =0 oD for any simple closed curve 9D in U, so that by (2.48), F is a gradient field. The following example, however, shows that the situation is rot as simple as this. Example 2.6: Let UU = R? — {(0,0)}. Show that the vector field F : — R? defined by F(x) (+ zy) (2.49) satisfies ar ah _y a Te By in U, but that F is not a gradient field in U 55Solution: A simple calculation gives OF _ OF _ oz Oy According to (2.48) we can show that F is not a gradient field by giving a simple closed curve C inU such that f Fedx 40. a ty wry c Consider the unit circle C given by =(cost,sint), 0S t < 2m, A routine calculation gives [rex =2n £0. (doit!) O z Comment: ‘The essential point is that Green's theorem cannot be applied to the vector field (2.49) on the set D = {(x,y) | x? +y? <1} whose boundary is the circle C, 2? + y? = 1, since F is not C! on D - it is not even defined at (0,0). Another way of looking at the difficulty is that the set & = R?—{(0,0)} on which F is C" has a “hole” in it~ the point (0,0) has been deleted. So we need to introduce a “no holes” restriction on the set Yon which F is C? Definition A connected open set C R” is simply-connected means that every simple close curve in U can be shrunk continuously to a point while remaining in U. eg. i) D= {(x,y)| 1 <2? +y? <4} is not simply-connected in R?, The circle drawn cannot be shrunk continuously to a point. ii) D=R? — {(z,0)| « <0} is simply-connected in R’. iii) R? minus a finite number of points is simply-connected. iv) R® minus an infinite line is not simply-connected. We are now ready to state the theorem on detecting gradient vector fields in R®. Having done the preparatory work, the proof is short! 56Theorem 2.4 (test for conservative fields) 1 i) U is a simply-connected open subset of R?, ii) F is a C' vector field that satisfies (2.50) in U, then there exists a single-valued C? potential ¢ in U, i.e. F=Vé in U. Proof: Let C be any simple closed curve in U4. Since U/ is simply-connected the interior D of C belongs to U¢ and thus (2.50) holds in DUC. By Green's theorem, / F-dx = 0. c Since C is arbitrary it follows from (2.48) that F is a gradient field in. 0 Example 2.7: ‘Test: the vector field F(x) = (ye™™,ne™™ + 2y) (2.51) it is, find a potential ¢. for being conservative, an Solution: A routine calculation shows that 92 — 2& = 0 on R?. Since R? is simply- connected, F is conservative by Theorem 2.4. To find a potential, we have to integrate the equations 6 88 | py = ne fa ye™, 7 a +2y. (2.52) The first. gives 6(e,y) = e+ Ky), (2.53) where the “constant of integration” depends on y. Differentiate (2.53) with respect to y and use the second equation in (2.52): ze + 2y = ae + Ky), giving K’(y) = 2y, and hence Kw)= +6, 87where C is constant. By (2.53) the potential is day) =e +9 +6, unique up to an additive constant C. O Exercise 2.12: Test whether the vector field F : U — R? is conservative on the set U C R?, and if so, find a potential ¢: i) F=(4-#), U={(ev)|y>0}, ii) F = (ycos(ey), x cos(zy)), UW =R?, iii) F= Spat) U= RP {(0,0)} Fa eP Answers: i) Yes; 6=£+C ii) NO iit) Yes; 6 = jln(z? +y?) + Comment: Exercise 2.12 ili) shows that even if U is not simply-connected the vector field may be conservative. 2.5 Vorticity and circulation Theorem 2.4 (test for conservative fields) shows that given a vector field F = (F,, F:) in R?, the quantity on OF 2 OF Oe Oy. is of fundamental importance: if it is zero on a simply-connected set 4 C R?, then F is a gradient conservative field in U. ‘This quantity also plays an important role when the vector field is the velocity field v = (0,02) of a fluid flow in two dimensions* and in this context it is called the vorticity of the fluid, denoted by 9: a Y _ du avy coe ~ Oe” By" We now describe the physical significance of the vorticity. Think of a small wooden paddle wheel that is carried along by the fluid. The question is: will the motion of the fluid cause the paddle wheel to gotate about its aris? In order to illustrate the problem we consider two simple vector fields (2.54) a= (u,0), “By a fiuid flow in two dimensions we mean a three dimensional flow which is “stratified”, ie. the fluid velocity is the same in each of a family of planes. Without loss of generality v(x) = (vi(z,y),va(.4)+0). 58ZB =— axis Figure 2.13: A paddle wheel. where u is a positive constant with [u] = LT-?, and v= (ay,0), where a is a positive constant with [a] = T-1. It follows from figure 2.14 that the velocity field u will not cause the paddle wheel to rotate, while figure 2.15 shows that the paddle wheel will rotate under the action of v. Figure 2.14: The Figure 2.15: The velocity field v = (ay, 0) acting on a paddle wheel, ‘This difference between u and v can be characterized by considering the line integral of the velocity fields around the circle C of radius p that represents the circumference of the paddle wheel. By (2.54) the vorticity for each velocity field is uy dus a, = Om _ an 8 Oe Oy Be Oy 59and hence Green’s theorem applied to the paddle wheel disc gives [va =0, [vw =-—np’a. e c ‘The line integral of a velocity field v along a curve is in fact equal to the line integral of the tangential component v - T along the curve, where T is the unit tangent vector to the curve. This result is seen as follows: . [vax=[ v(g(t)) -8'()dt Aside:g'(t) = T | e'(t) || Q : = [ vew)- Teo le (255) = I (v-T)ds, 2 i f vedx <0, ie. v-'T is on balance negative along C, the paddle wheel will rotate in the e opposite direction to ‘T. On the other hand, if f'v-dx = 0, ie. v-'T is on balance zero along J c C, the paddle wheel will not rotate. We now summarize the conclusion. Let the curve C, oriented counter-clockwise as usual, be the circumference of a paddle wheel. If <0 [ve =o, lc >0 then the fluid flow will cause the paddle wheel to rotate clockwise not rotate rotate counter-clockwise. The quantity f v-dx is called the circulation of the fluid around the simple closed curve C. c Green's theorem leads to a relation betweer the vorticity 0 and the circulation f vedx. é Let D, be the disc of radius p centred at x with boundary OD, oriented counter-clockwise.By Green’s theorem where & is some point in D,. The last. step follows from the Mean Value Theorem for integrals. Divide by the area mp? and let p + 0 giving (2.56) In words, the vorticity at x equals the circulation per unit area at x. For our purposes it is helpful to write an approximation, 2x0) x 4 / vedx, ae ° WDy for p sufficiently close to 0. ‘We thus obtain the desired physical interpretation of the vorticity: <0 rotate clockwise Q(x) 4=0 = apaddle wheel at x will {not rotate >0 rotate counter-clockwise. ‘We conclude with two classic vector fields in R? that illustrate vorticity and circulation. Example 2.8: Consider the vector field v=a(-y,z), (2.57) 61where @ is a positive constant with [a] = T-!. The flow lines are circles 2? + y? = c?, traversed counter-clockwise (see example 1.12 in Chapter 1). The vorticity 9 is non-zero, v2 Iu St - Sh =2 ae Oy Since the domain of v is R®, Green’s theorem can be applied to any circle of radius p, giving / v-dx = 2a(mp") > for the circulation. It follows that a paddle wheel will rotate counter-clockwise (see Figure 2.16) Example 2.9: Consider the vector field (a “vortex field”) a Fae #00) (258) ‘The flow lines are again circles x? + y? = c? (see example 1.13 in Chapter 1). The vorticity Q is, however, zero: (verify). In this case the domain of v is Rp—{(C,0)}, which is not simply-connected. Green's theorem can be applied to any circle that does not enclose or pass through the origin, giving [va =0 (since Q = 0). Thus a paddle wheel will not rotate as it moves with the fluid (see Figure 2.17). However, for a circle of radius b that encloses the origin the circulation is non-zero: / vedx = 2rab. c (Verify using the definition of line integral.) Thus in this case, although the fluid is locally non-rotating (i.e. the paddle wheel does not rotate), it does rotate globally, i.e. there are simple closed curves with non-zero circulation. 62Figure 2.16: The velocity field (2.57) causes the paddle wheel to rotate. Figure 2.17: The velocity field (2.58) does not cause the paddle wheel to rotate. 63Chapter 3 Surfaces & Surface Integrals The notion of a surface arises in various physical and mathematical contexts, e.g. the wing or fuselage of an aircraft, an ocean wave at an instant of time, or a soap film formed by dipping a closed wire loop into a soap solution. In a mathematical context, when deriving the equation that governs heat transfer one considers an arbitrary finite chunk of the conducting medium that is bounded by a surface ©, and one writes the heat flux through © as a surface integral. 3.1 Parametrized surfaces 3.1.1 Surfaces as vector-valued functions ‘The first mathematical description of a surface that one encounters is the graph of a function ff: RR. The graph is a surface in R%, described by the equation z=flay, (nye DCR’ ‘This way of describing surfaces is somewhat limited: it cannot describe a surface that “folds over” such as a sphere or a torus. So we think about vector-valued functions and generalize the way they are used to describe curves by introducing two parameters u and v, and writing x=g(u,v), (uv) € Duo. (3.1) Here Dyy is a subset of R? (the wo-plane), x = (x,y, 2) € R* (yz-space), and g : Dy > R® is a vector-valued function. Often Dyy will be a rectangle in the uv-plane, but in general it will be a bounded subset of R?, whose boundary is a simple closed curve. Figure 3.1 gives ‘a schematic representation of the domain Dy, and the surface 5 : as the point (u,v) moves through the set Dy, the image point g(u,v) sweeps out the surface D in R° One thinks of the function g as a map from R? —+ RS that acts on the rectangle Dus, bending and stretching it so as to form the surface ©. It is helpful to think of the surface © as being generated by two families of curves namely the images of the two families of straight lines u = constant and v = constant. ‘The two families of curves form a “grid” or “fish-net” that covers the surface. We think of u and v as being coordinates on the surface B and we refer to the curves on © that are defined by u = constant and v = constant as coordinate curues which form a coordinate grid. 65x igure 3.1: The vector-valued function g maps the domain D,, onto the surface 5. Example 3.1: ‘The equation x=g(u,v)=atuer+ver, (u,v) € Dus (3.2) where Dy, = {(u,v) | -1 < u,v < 1}, and e),e» are two linearly independent vectors in R, describes a surface which is a piece of the plane through the point a, and containing the vectors @, and e2. Referring to Figure 3.2, the vector x — a lies in the plane and hence is a linear combination of e; and ep < a= g(0.0) Figure 3.2: Parametric representation of a plane. One can obtain the equation of the plane in standard form m,(x — a) + no(y — b) +n3(z—c) =0 (3.3) by calculating a normal vector n. Since e; and es lie in the plane the vector product e; x e2 is a vector normal to the plane: n=e, xe. 66Then, taking the scalar product of equation (3.2) with n gives n(x —a) =0, which is the standard form (3.3) Recall: ‘The vector product a x b can be evaluated using a “symbolic determinant”: ijk axb=|a; a2 a3), (3.4) by ba bs which, when formally expanded by the first row, gives a x b= (agby — aab)i + (aabi ~ aibs)j + (aibo — abi )k. O , Tk . gine Rep 8 Exercise 3.1: d ‘ 1 A plane in R® is given by RAM +R) x= (12,3) +u(1, 3,0) +0(1,-1,1), B= O -D- Mya) Y2 De with (u,v) €R2, Find the equation of the plaue in standard form. |" d= Myr) 22D)-0 Answer: (x -1)—(y—2)-2(z-3) =0. Exercise 3.2: Find a vector-valued function to describe the plane x — 3y +2 = Hint: Let u=a,0=y. X- ay tR2) Answer: x = g(u,v) = (0,0,2) + u(1,0,—1) +0(0,1,3). Example 3.2: ‘The vector-valued function g : Dy — R* defined by (sin ucosv, sin usin v, cos), (35) B(u,v) with Dw = {(uyv)|OSusm, Ov <2}, describes the surface of the unit sphere in R°. This can be verified by writing x = g(u,v) and verifying that Ix P=2t+y+2=1 The vector-valued function (3.5) is obtained from the formulas that relate spherical polar coordinates to Cartesian coordinates: r=rsin9cosd y=rsin9sing z= rosé. o7Figure 3.3: Spherical polar coordinates. cmrcles = const. circles Figure 3.4: The coordinate grid on the sphere created by the spherical polar angles 0 and ¢. To obtain a unit sphere we choose r = 1, and then let u = 0, v = ¢. The coordinate lines u = constant on the sphere are circles of constant latitude, and the coordinate lines » = constant are circles of constant longitude, as sown in Figure 3.4. —eoooer Comment: Consider the surface © defined by z=S(2,y), (ty)E DCR? One can write a vector equation for © by introducing u = and v = y as parameters. Then 2 = f(u.v), and the vector equation for the surface is x= g(u,v) = (u,v, fv), (u,v) € D. (36) Exercise 3.3: NOUSUY& Give a vector-valued function to describe the cone 2? +9? = 2? with O0<2<1. (uv) = (cosy) Answer: x = g(u,v) = (wcosv, usin, u), x with (u,v) € Duy = {(u,v) | OS u<1,00. The function g is of class C', but the surface does not have a tangent plane at the point g(0,v) = (0,0,0). The tangent vectors are og _ Og _ - Fe = (cosmrsine,), FE = (—usin v, weosu, 0), and are linearly dependent when u = 0, since + hireaw Telependlent, GH) +0 Boo (0, 0,0). This requirement of linear independence is analogous to the requirement that g/(t) £0 for a curve: if g/(to) =0 the curve may have a busy at g(to). UuT™ For a vector-valued function g : Dy, — R° the linear approximation corresponds to using ‘the tangent plane to approximate th surface x = g(u,v). Using (3.8) we have as Os (1,0) * Bo, to) + (tu to) 5E (os 20) + (0 — 00) (tos 0), for (u,v) sufficiently close to (uo, vo). Equivalently, introducing the increments Ag = g(u,v) —B(uo,r), Au wwe obtain og Age a for Au, Av sufficiently close to zer’ Example 3.3: ‘The surface $ defined by z=flay), (zy) ED, (3.12) can be described equivalently by x= g(u,v) = (u,v, f(u,v)), (uv) ED (see (3.6)). The tangent vectors (3.7) are given by 3 (0.94), % Cr) (3.13) 70