PARTIAL DIFFERENTIAL EQUATIONS Prof. Fine 1 PARTIAL DIFFERENTIAL EQUATIONS Prof. Benjamin Fine Text: M. Coleman Partial Differential Equations Prentice Hall Course Goals: To provide an introduction to the mathematical techniques that are used in the study of partial differential equations as well some very important applications. There will three projects and three exams. Prerequisites: Knowledge of one-variable and multi-variable Calculus,linear Algebra and basic ordinary differential equations. There will be three projectsthis semester. Course Outline A. Ordinary Differential Equations 1. Basic ideas, order and solution techniques 2. Linear Differential Equations - fundamental existence and uniqueness theorem B. Partial Differential Equations 1.PDE’s and Classification 2. Elementary Solution Technqiues 3. Boundary Value Problems - the big three: Heat,Wave and Laplace equations C. Vector and Scalar Field Theory 1. Vector and Scalar Fields - gradient, divergence and curl 2 2. Line, Surface and Volume Integrals - Stoke’s and Divergence Theorem 3. Derivation of the Heat and Wave Equation D. Fourier Series Methods 1. Fourier Series and Fourier Expansions 2. Separation of Variables E. Solutions of the Big Three 1. Solution of the Heat Equation 2. Solution of the Wave Equation a. D’Alembert’s Method 3. Solution of the Laplace Equation F. Transform Methods 1. Basic ideas of Integral transforms 2. The Laplace transform and Applications 3. The Fourier transform and Applications - Fourier Inversion Theorem 3 PARTIAL DIFFERENTIAL EQUATIONS (1) PARTIAL DIFFERENTIAL EQUATIONS - BASIC IDEAS A. Differential Equations 1.Definition A differential equation is an equation which involves a variable, (or variables), a function of that variable and one or more of the derivatives of that variable. We express this as f (x, y, y ′ , y ′′ , ..., y (n) ) = 0 a.If the equation involves only ordinary derivatives its called an ordinary differential equation or ODE while if there are partial derivatives it a partial differential equation or PDE. b. The order of a differential equation is the order of the highest derivative which appears in the equation. EXAMPLES (1) y ′ = y. This is a first order ODE. (2) x2 ex dy + 2x sin xdx = 0. This is again a first order ODE (3) y ′′ + (x2 + cos x)y = ex . This is a second order ODE (4) ∂2u ∂x2 + ∂2u ∂y 2 = 0. This is a second order PDE. 2. Solutions 4 a.Definition A function y = y(x) defined on some interval I which satisfies the equation is a solution. (1) Generally there will be infinitely many solutions. Sometimes all solutions to an ODE will be given in terms of some real variables called parameters. The solution given in terms of these parameters is called the general solution. Particular values of these parameters lead to particular solutions. EXAMPLES (1) y ′ = y. The general solution is given by y = cex The value c is the parameter. Any value of c gives a particular solution so for example y = 4ex is a particular solution. (2) y ′′ + y = 0. The general solution is given by y = c1 sin x + c2 cos x The parameters are c1 , c2 . Specific values for c1 , c2 gives a particular solution. (2) If the solution y = y(x) is given explicitly then we have an explicit solution. If the solution is given as f (x, y) = 0 then its an implicit solution (3) If we have the ODE f (x, y, y ′ , ..., y (n) ) = 0 5 and we specify particular values at a given point y(x0 ) = y0 , y ′ (x0 ) = y1 , ....., y (n) (x0 ) = yn then its an initial value problem. Generally under many conditions an initial value problem will have a unique solution. (4) If we have the ODE f (x, y, y ′ , ..., y (n) ) = 0 and we specify y(a) = y0 , y(b) = y1 at two different points then we have a boundary value problem. 2. Linear Differential Equations a. Definition A linear differential equation is one of the form an (x)y (n) + an−1 (x)y (n−1) + ... + a0 (x)y = g(x) That is in a linear differential equation no derivative appears with an exponent higher than one. Otherwise it is non-linear. b. Fundamental Existence and Uniqueness Theorem (1) Theorem Consider the initial value problem an (x)y (n) + an−1 (x)y (n−1) + ... + a0 (x)y = g(x) y(x0 ) = y0 , y ′ (x0 ) = y1 , ..., y n (x0 ) = yn+1 with x0 ∈ I. Then there exists a unique solution to the initial value problem. 6 c. Homeogenous and Nonhomogeneous Equations (1) Theorem The set of solutions to a homogeneous linear differential equation of order n from a vector space of dimension n. A basis is called a fundamental set of solutions B. Partial Differential Equations 1.We will concentrate for the most part on first and second order linear PDE’s a.Definition [1] A first order linear partial differential equation is a PDE of the form aux + buy + cu = g (1) where a, b, c, g are all functions of x, y. We will assume that these functions are defined and continuous over some common region R in two-space. If a first-order PDE is not of this form it is non-linear. [2] A second order linear partial differential equation is a PDE of the form auxx + 2buxy + cuyy + dux + euy + f u = g (2) where a, b, c, d, e, f, g are all functions of x, y. We will assume that these functions are defined and continuous over some common region R in two-space. If a second-order PDE is not of this form it is non-linear. (1) If g(x, y) ≡ 0 then the equation is homogeneous. Otherwise it is non-homogeneous 7 (2) We let L be the linear differential operator ∂2 ∂2 ∂ ∂2 ∂ L = a 2 + 2b +c 2 +d + e + f. ∂x ∂x∂y ∂y ∂x ∂y Then the second order PDE (1) can be expressed in compact form as Lu = g (3) Lemma If g ≡ 0 so that (1) is homogeneous then if u1 , u2 , ..., un are solutions then any linear combination c1 u1 + c2 u2 + ... + cn un is also a solution. Further if g ̸≡ 0 and up is any particular solution to (1) then up + c1 u1 + c2 u2 + ... + cn un is also a solution, where u1 , ..., un are solutions to the associated homogeneous equation. 2. Classification of Second Order PDE’s a.Definition Consider a second order linear partial differential equation auxx + 2buxy + cuyy + dux + euy + f u = g Then this is (i) elliptic if b2 − ac < 0 (ii) parabolic if b2 − ac = 0 (iii) hyperbolic if b2 − ac > 0 8 (1) b.Mixed Type Equations (1) Note that if a, b, c are not constants then the equation can be of mixed type that is different in different regions. EXAMPLE Tricomi’s Equation which arises in the theory of transonic flow is uxx + xuyy = 0 Here b2 − ac = −x so the equation is elliptic in the right half plane x > 0 and hyperbolic in the left half-plane x < 0. C. The Big Three PDE’s 1. Laplace Equation a.Definition The Laplace Equation is uxx + uyy = 0 (1) Here a = 1, b = 0, c = 1 so it is an elliptic equation (2) Roughly Laplace equations arise is modeling equilibrium problems 2. Diffusion Equation or Heat Equation a.Definition The Diffusion Equation or Heat Equation is α2 uxx = ut 9 where α2 is the diffusivity constant and t is time. (1) Here a = α2 , b = 0, c = 0 so it is a parabolic equation (2) Roughly diffusion equations arise is modeling diffusion problems such as heat transfer 3. Wave Equation a.Definition The Wave Equation is β 2 uxx = utt where β 2 is a constant and t is time. (1) Here a = β 2 , b = 0, c = −1 so it is a hyperbolic equation (2) Roughly wave equations arise in modeling wave phenomena D. Elementary Solution Techniques 1. Solution by Integration a. Many linear PDE’s can be solved either by direct integration or using methods of ordinary differential equations. Recall however that in partial integration the constant of integration is an arbitrary function of the other variable. EXAMPLE 1 10 ∂ 2u =0 ∂y 2 Integrating with respect to y we get ∂u = f (x) ∂y integrating again =⇒ u(x, y) = yf (x) + g(x) where f (x), g(x) are arbitrary functions of x. EXAMPLE 2 ∂ 2u ∂u + =1 ∂x∂y ∂y Letting v = ∂u ∂y we get vx + v = 1 This is a linear ODE with an integrating factor e ∫ 1dx = ex =⇒ (vex )′ = ex =⇒ ex v = ex + h(y) ∂u =⇒ v = 1 + e−x h(y) =⇒ = 1 + e−x h(y) ∂y ∫ =⇒ u(x, y) = y + e−x h(y)dy + g(x) where f (x) is an arbitrary function of x and h(y) is an arbitrary function of y EXAMPLE 3 11 ∂ 2u − y 2 u = ex 2 ∂x Treating y as a constant this is a linear ODE in X with constant coefficients uxx − y 2 u = ex The corresponding homogeneous equation is uxx − y 2 u = 0 which has an auxillary polynomial m2 − y 2 = 0 =⇒ m = ±y =⇒ uc = c1 exy + c2 e−xy However c1 , c2 are functions of y =⇒ uc = f (y)exy + g(y)e−xy This is the general solution of the corresponding homogeneous equation. To find a particular solution we’ll use the method of undetermined coefficients Let up (x) = A(y)ex =⇒ A(y)ex − y 2 A(y)ex = ex 1 =⇒ A(y) = 1 − y2 ex 1 − y2 where f (y) and g(y) are an arbitrary functions of y =⇒ u(x, y) = f (y)exy + g(y)e−xy + EXAMPLE 4 A first order equation aux + buy = f (x) 12 can be solved in the following manner. method works if the right side is g(y). (An analogous Recall that du = ux dx + uy dy Now suppose that b dx dy = =⇒ dy = dx a b a =⇒ bx − ay = c a constant Then b du du = ux dx + uy dy = ux dx + uy dx =⇒ a = aux + buy a dx 1 =⇒ du = f (x)dx a ∫ 1 f (x)dx + C =⇒ u(x, y) = a where C is a constant. Then C = g(bx − ay) where g(t) is an arbitrary function of a single variable. Consider for example 3ux + 4uy = ex Then 1 u(x, y) = ex + g(4x − 3y) 3 2. Method of Separation of Variables a.For a homogeneous linear PDE the method of separation of variables attempts to find a solution of the form u(x, y) = X(x)Y (y) 13 EXAMPLE ∂ 2u ∂u = 4 ∂x2 ∂y Let u = XY where X = X(x), Y = Y (y) then =⇒ X ′′ Y = 4XY ′ Y′ X ′′ = =⇒ 4X Y Since X is a function of x alone and Y is a function of y alone this equality is possible only if both sides equal the same constant c. The constant c is called the separation constant. Case 1 c = λ2 > 0 then X ′′ − 4λ2 X = 0 Y ′ − λ2 Y = 0 =⇒ X = c1 e2λx + c2 e−2λx and Y = c3 eλ 2 y =⇒ u(x, y) = Aeλ y (Be2λx + Ce−2λx ) 2 Case 2 c = −λ2 < 0 then X ′′ + 4λ2 X = 0 Y ′ + λ2 Y = 0 =⇒ X = c1 cos(2λx) + c2 sin(2λx) and Y = c3 e−λ 2 =⇒ u(x, y) = Ae−λ y (B sin(2λx) + C cos(−2λx)) 2 Case 3 c = 0 then X ′′ = 0 14 y Y′ =0 =⇒ X = c1 + c2 x and Y = c3 =⇒ u(x, y) = Ax + B 3. Superposition Principle a. Theorem If u1 , u2 , ..., un are solutions to a linear homogeneous PDE then anyy linear combination is also a solution. (1) We assume that if u1 , u2 ..., un , ... are an infinite sequence of solutions then ∞ ∑ ui i=1 is also a solution 15 PARTIAL DIFFERENTIAL EQUATIONS (2) SCALAR AND VECTOR FIELD THEORY A. Scalar and Vector Fields 1.Scalar Fields a.Definition: A scalar field is a real-valued function f (x, y, z) defined on a region in R3 . b.Isotimic Surfaces (1) Definition If f (x, y, z) is a scalar field then a surface defined by f (x, y, z) = c is called an isotimic surface. If f is gravitational potential or electric potential called an equipotential surface. If f is temperature its called an isothermal surface. EXAMPLE f (x, y, z) = x3 z 2 − xy 2 z is a scalar field. The surface defined by x3 z 2 − xy 2 z = 9 is an isotimic surface. 2.Vector Fields a. Definiiton A vector field is a function F which assigns a tangent vector at each point p ∈ R3 . Thus a vector field is a function F : R3 → R3 , F = F1 (x, y, z)i + F2 (x, y, z)j + F3 (x, y, z)k F1 , F2 , F3 are the scalar field components of F. F is a differentiable vector field if its components are differentiable scalar fields. We will assume that our vector fields are differentiable. 16 EXAMPLE F = (x3 y 2 z 2 )i + (2x − z 3 y)j + (3y 3 ez )k is a differentiable vector field. The functions F1 = x3 y 2 z 2 ,F2 = 2x − z 3 y, F3 = 3y 3 ez are the scalar field components. b. Velocity Fields and Flow Lines (1) It can be useful to think of a vector field as a velocity vector field for some fluid moving through R3 . Thus at each point p, F(p) = velocity of the fluid at p. (2) A curve is a flow line for a vector field F if at each point of the curve F defines a tangent vector to the curve. (a) If R(t) = x(t)i + y(t)j + z(t)k is a curve then it is a flow line for the vector field F if there exists a β ∈ R such that dx dy dz βF1 = , βF2 = , βF3 = dt dt dt B. Differential Operators on Scalar and Vector Fields 1. The Derivatives of Scalar Fields a.The Gradient of a Scalar Field (1) Definition If f is a differentiable scalar field then its gradient is the vector field defined by ∂f ∂f ∂f gradf = i+ j+ k. ∂x ∂y ∂z 17 EXAMPLE The gradient of the scalar field f (x, y, z) = x3 z 2 − xy 2 z is the vector field gradf = (3x2 z 2 − y 2 z)i + (−2xyz)j + (2x3 z − xy 2 )k b.The Directional Derivative (1) Definition Let p be a point, f (x, y, z) a scalar field and v a unit vector with p considered as its endpoint. Let g(t) = f (p + tv). Then vp [f ] = g ′ (0) is called the directional derivative of f in the direction of v at p. Note that v is taken as a unit vector - that is the directional derivative depends on the direction not on the magnitude of the vector. (a) vp [f ] measures the rate of change of f at p measured in the v direction. (b) Lemma If v is a tangent vector at p ∈ R3 then the directional derivative of a scalar field f in the direction of v at p is given by vp [f ] = v.gradf (p). EXAMPLE Find the directional derivative of f (x, y, z) = x3 z 2 − xy 2 z in the direction of v = 2i + 3j − k at the point (2, 0, 1). 18 Here a unit vector in the direction of v is u= v 3 1 2 = √ i+ √ j− √ k |v| 14 14 14 gradf (2, 01, ) = (3x2 z 2 −y 2 z)i+(−2xyz)j+(2x3 z−xy 2 )k|(2,0,1) = 12i+16k 3 1 8 2 =⇒ vp [f ] = u.gradf (p) = √ i+ √ j− √ k).(12i+16k) = √ 14 14 14 14 (2) Lemma gradf points in the direction of maximum rate of increase of the scalar field f . (Since gradf = v.gradf then |gradf | = |v||gradf | cos θ. Since v is a unit vector this will be greatest when cos θ = 1 or when the angle is zero, that is, v is in the direction of the gradient.) (3) Lemma |gradf | = maximum rate of increase of f per unit distance. (4) Theorem Through any point (x0 , y0 , z0 ) where gradf ̸= 0 there passes an isotimic surface f (x, y, z) = c. gradf is normal to this surface and therefore defines a set of attitude numbers for the tangent plane. 2. The Derivatives of Vector Fields a.The Rate of Change of a Vector Field (1) There are two fundamental measures of the rate of change of a vector field - the divergence and the curl. The divergence of a vector field is a scalar field which roughly measures the extent to which the field diverges from that point. The curl of a vector field is a vector field which 19 roughly measures the extent to which the field swirls about that point. b. The Divergence of a Vector Field (1) The Divergence Operator (a) Definition Consider a region B ⊂ R3 bounded by a surface S with surface element dA, outward normal n and volume V . The flux or flow of a vector field F across a surface element is n.FdA. The divergence at a point p ∈ B is ∫ n.FdA DivF (p) = lim S V →0 V [1] The divergence of a vector field is then a scalar field [2] Note that the above is a coordinate independent definition. (2) Coordinate Expression for the Divergence (a) A coordinate expression for the divergence is obtained by considering B to be a paralleopiped. (b) Theorem If F = F1 i + F2 j + F3 k is a vector field then its divergence is the scalar field given by DivF = ∂F1 ∂F2 ∂F3 + + . ∂x ∂y ∂z EXAMPLE The divergence of the vector field F = (x3 y 2 z 2 )i + (2x − z 3 y)j + (3y 3 ez )k 20 is the scalar field DivF = 3x2 y 2 z 2 − z 3 + 3y 3 ez (c) Derivation of the Coordinate Expression (1) Consider a region B ⊂ R3 bounded by a surface S with surface element dA,outward normal n and volume V . The flux or flow of a vector field F across a surface element is n.FdA. The divergence at a point p ∈ B is ∫ n.FdA DivF (p) = lim S V →0 V (2) Now consider B to be a rectangular box alligned with the coordinate axes and let P0 = (x0 , y0 , z0 ) be the center of the box. Then an outward normal across the front face is n = i while an outward normal across ther back face is n = −i. Then if the force is given by F = F1 i + F2 j + F3 k it follows that n.F = ±F1 Further across the front face F1 (x, y, z) = F1 (x0 + ∆x , y, z) 2 and across the back face F1 (x, y, z) = F1 (x0 − 21 ∆x , y, z) 2 Then: ∫ ∫ n.FdA = front face F1 (x0 + front face ∆x , y, z)dydz 2 Using the mean value theorem it follows that = F1 (x0 + ∆x , y1 , z1 )∆y∆z 2 for some y1 , z1 . Similarly across the back face ∫ ∆x n.FdA = F1 (x0 − , y1 , z1 )∆y∆z 2 back face further V = ∆x∆y∆z Hence: ∫ lim V →0 front + back faces n.FdA V = ∂F1 ∂x Now the other faces follow in a similar manner. c.The Curl of a Vector Field (1) The Curl Operator (a) Definition Consider a region B ⊂ R3 bounded by a surface S with surface element dA,outward normal n and volume V . The swirl of a vector field F about a point P across a surface element is n×FdA. The curl at a point p ∈ B is ∫ n × FdA CurlF(p) = lim S V →0 V [1] The curl of a vector field is then also a vector field 22 [2] Note that the above is a coordinate independent definition. (2) Coordinate Expression for the Curl (a) A coordinate expression for the curl is obtained by considering B to be a paralleopiped. (b) Theorem If F = F1 i + F2 j + F3 k is a vector field then its curl is the vector field given by i j k ∂ ∂ ∂ CurlF = det ∂x ∂y ∂z F1 F2 F3 EXAMPLE The curl of the vector field F = (x3 y 2 z 2 )i + (2x − z 3 y)j + (3y 3 ez )k is the vector field i j k ∂ ∂ ∂ CurlF = det ∂x ∂y ∂z F1 F2 F3 i j k ∂ ∂ ∂ =⇒ CurlF = det ∂x ∂y ∂z (x3 y 2 z 2 2x − z 3 y 3y 3 ez CurlF = (9y 2 ez − 3z 2 y(i + (2x3 y 2 z)j + (2 − 2x3 yz 2 )k (3) Curl as Angular Velocity 23 (a) Considering F as a velocity vector field for a moving fluid then CurlF = 1 angular velocity 2 (b) Derivation as Angular Velocity Consider a planar fluid with velocity vector field F = F1 i + F2 j + F3 k swirling counterclockwise about the z-axis around a point P at the center of a rectangle with corners O, A, B. Here F3 = 0 since its planar. Then the angular velocity of the fluid about P is the average of the angular velocity of the segment OA and the angular velocity of the segment OB. We then have angular velocity of OA = F2 (A) − F2 (O) ∂F2 k→ k as ∆x → 0 ∆x ∂x 24 angular velocity of OB = F1 (O) − F1 (B) ∂F1 k→− k as ∆y → 0 ∆y ∂y Then 1 ∂F2 ∂F1 ω= ( − )k 2 ∂x ∂y Putting all three directions together gives that the curl is twice the angular velocity. 3.The Del Notation a. We define the del operator by ▽= ∂ ∂ ∂ i+ j+ k ∂x ∂y ∂z It can operate on either scalar or vector fields as described below. (1)Lemma If f is a scalar field and F is a vector field then ▽f = gradf ▽.F = DivF ▽ × F = CurlF 4.The Laplacian 25 a. Definition If f is a scalar field then its Laplacian is the scalar field defined by ∂ 2f ∂ 2f ∂ 2f + + . ▽ f = Div(gradf ) = ∂x2 ∂y 2 ∂z 2 2 (1) Roughly the Laplacian measures the difference between the average value of the field in an immediate neighborhood of a point and the precise value of the field at that point b. The Laplacian in Mathematical Physics (1) ▽2 f = 0 is Laplaces equation and a solution is a harmonic function (2) ▽2 f = (3) ▽2 f = 1 ∂2f a2 ∂t2 is the wave equation 1 ∂f a2 ∂t is the heat equation 5.Some Vector Operator Identities a. Suppose f1 , f2 are scalar fields and F, G are vector fields then (1)▽(f1 f2 ) = f1 ▽f2 + f2 ▽f1 (2)▽.f F = f (▽.F) + F.▽f (3)▽ × f F = f (▽ × F) + ▽f × F ∂f (4)▽f (u) = ▽u ∂u (5)▽.(F × G) = G.▽ × F − F.(▽ × G) (6)▽ × (F × G) = (g.▽)F − (F.▽)G + (▽.G)F − (▽.F)G (7)▽ × (▽ × F) = ▽(▽.F) − ▽2 F 26 (8)▽(F.G) = (F.▽)G+(G.▽)F+F×(▽×G)+G×(▽×F) (9)▽ × ▽f = 0 ⇒ Curl(gradf ) = 0 (10)▽.(▽ × F) = 0 ⇒ Div(CurlF) = 0 (11)▽.(▽f1 × ▽f2 ) = 0 6.Summary of Vector Operators VECTOR OPERATORS ▽= ∂ ∂ ∂ i+ j+ k ∂x ∂y ∂z Name Interpretation gradf = ▽f Maximum rate of change of f in the maximal direction DivF = ▽.F Net outflux of F per unit volume CurlF = ▽ × F swirl of F per unit area 2 Laplacian(f ) = ▽ f difference between f (P ) and the average of f around P C. Vector Field Integration 1.Line Integrals a.Definition: Let F be a continuous vector field defined on a domain D containing a curve C. The the line integral of F over C is ∫ ∑ F.dR = lim|∆Ri |→0 F(Pi )∆Ri C where R = R(t) is a parametrization of the curve. 27 In terms of the parameter t this is given by ∫ ∫ ∫ F.dR = F.Tds = Ft ds c C C where Ft is the tangential component of F over C. This can also be written as ∫ ∫ F.dR = F1 dx + F2 dy + F3 dz C C F1 dx + F2 dy + F3 dz is a first order differential form and thus the line integral is the integral of this differential form over C. b. Therefore for a line integral we need a vector field and an oriented curve and the result is a scalar. c. Computed in terms of the parameter t ∫ t2 dy dz dx F.dR = (F1 (x(t), y(t), z(t)) +F2 (x(t), y(t), z(t)) +F3 (x(t), y(t), z(t)) ) dt dt dt C t1 ∫ 2.Conservative Vector Fields a. A vector field F is conservative if there exists a scalar field ϕ(x, y, z) such that F = gradϕ. ϕ is a potential function or potential for F. (1) Lemma. If F is a conservative vector field then CurlF = 0. ∫ b. C F.dR is independent of path if the value of the line integral depends only on the endpoints of the curve C. (1) Theorem. Suppose F is a continuous vector field on a domain D. Then the following are equivalent: 28 (i) F is conservative ∫ (ii) C F.dR is independent of path H (iii) C F.dR = 0 for any regular closed curve C in D. If the domain D is simply connected then the following is also equivalent to the above: (iv) CurlF = 0 ( If CurlF = 0, F is called irrotational. In the case of the above theorem we have ∫ Q F.dR = ϕ(P ) − ϕ(Q) P where ϕ is a potential for F. 3.Solenoidal Vector Fields and Vector Potentials a. If F is a vector field such that Div F = 0 then F is solenoidal. If F = Curl G then F is solenoidal and G is called a vector potential for F. (1) Theorem.A vector field F continuously differentiable in a star-shaped region is solenoidal if and only if there exists a vector potential. 4.Surface Area a. The surface area of a regular surface element with parametrization R(u, v) over a region D ⊂ R2 is ∫ ∫ S= |Ru × Rv |dudv D 29 The area element is then dS = |dS| = |Ru × Rv |dudv. The vector dS = Ru × Rv dudv = Ru du × Rv dv is a a vector normal to the surface whose magnitude dS = |dS| is the area element. Then ∫ ∫ ∫ ∫ ∫ ∫ S= |dS| = dS = n.dS D D D where n is the outward unit normal in the same direction as dS. (1) Two Special Cases (a) Suppose z = 0 so that the surface lies entirely in the xy plane. Then: dS = | ∂(x, y) |dudv ∂(u, v) This is the area element for a change of variables. In the case where x = r cos θ, y = r sin θ, z = 0 we have the change to polar coordinates {where r = u, θ = v}. Then dS = rdrdθ. (b) Suppose z = f (x, y) defines a surface over D ⊂ R . Then: ∫ ∫ √ S= 1 + fx2 + fy2 dxdy. 2 D Let γ be the angle between dS and k. Then | cos γ| = √ 12 2 ⇒ 1+fx +fy ∫ ∫ dxdy S= D | cos γ| 30 In the case where z = f (x, y) is actually given by F (x, y, z) = 0 then ▽F.k cos γ = |▽F | EXAMPLE Show that the surface area of a sphere of radius R is 4πR2 . We can place the sphere centered on the origin and hence a parametrization of the upper half sphere would be √ z = f (x, y) = R2 − (x2 + y 2 ) Then −y x , fy = √ fx = √ R2 − (x2 + y 2 ) R2 − (x2 + y 2 ) R2 x2 + y 2 =⇒ 1+fx2 +fy2 = 1+ √ =√ R2 − (x2 + y 2 ) R2 − (x2 + y 2 ) ∫ ∫ Then √ A= D R2 R2 − (x2 + y 2 ) dxdy where D is the unit circle. Changing to polar coordinates ∫ 2π ∫ R R2 √ A= rdrdθ = 2πR2 2 2 R −r 0 ) This is the surface area for the upper hemisphere and hence the surface area for the sphere is 4πR2 . 5.Surface Integrals 31 a. If f (x, y, z) is a continuous scalar field defined in a region containing a regular surface element S then the surface integral of f over S is ∫ ∫ ∑ f dS = lim f (xi , yi , zi )∆Si |∆Si |→0 S Computationally if S is parametrized by R(u, v) then ∫ ∫ ∫ ∫ dS = |Ru × Rv |dudv so f (x(u, v), y(u, v), z(u, v))|Ru × Rv |dudv f dS = S D b. The Flux of a Vector Field If F(x, y, z) is a vector field over a surface S then the flux of F over S is ∫ ∫ ∫ ∫ F.ndS = F.dS S S (1) If S is specified by z = f (x, y) then ∫ ∫ ∫ ∫ dxdy . F.dS = F.n | cos γ| S D c. Some Applications in Physics (1) Suppose we have a steady state temperature distribution with scalar field T (x, y, z) over a region filled with a homogenuous material with thermal conductivity k. Then − → the vector Q = −k▽T gives the direction in which heat is flowing. If S is a surface contained in this region then ∫ ∫ (−k▽T ).ndS S 32 gives the total number of calories per second flowing across the surface S. (2) If F is the velocity field of a fluid and ρ its density then: ∫ ∫ ρF.ndS s is the rate of flow of liquid across the surface S expressed as mass per unit time. (3) Gauss’ Law of Electrostatics If E is an electrostatic field and S is a closed surface then ∫ ∫ q E.ndS = ϵ0 S where q = total charge enclosed by the surface S and ϵ0 = a constant depending on the units. D. The Divergence Theorem and Stokes Theorem 1.Introduction to the Divergence Theorem and Stokes Theorem a.The general idea behind the various versions of Stokes Theorem is that the integral of the ”derivative” of a differential form over a region D (suitably defined for the dimensionality of the integral) is equal to a one-dimension lower integral of the corresponding function on the boundary of D - again suitably defined. This is: ∫ ∫ dω = ω D ∂D . 33 (1) For ordinary integrals this is just the Fundamental Theorem of Calculus: ∫ b f ′ (x)dx = f (b) − f (a) a (2) In two dimensions this is Green’s Theorem: I ∫ ∫ ∂Q ∂P P dx + Qdy ( − )dxdy = ∂y D ∂x ∂D (3) In three dimensions this is the Divergence Theorem and Stokes Theorem representing the two types of differentiation of vector fields. (a) The Divergence Theorem: ∫ ∫ ∫ ∫ ∫ divFdV = F.dS S ∂S (b) Stokes Theorem: ∫ ∫ ∫ CurlF.dS = S F.dR ∂S 2. The Divergence Theorem a.Theorem (The Divergence Theorem) Let D ⊂ R3 be a region so that each straight line through any interior point of D cuts the boundary in exactly two points and so that the boundary ∂D is a piecewise smooth, closed oriented surface. Let F = F1 i + F2 j + F3 k be a C 1 vector field on D. Then: ∫ ∫ ∫ ∫ ∫ divFdV = F.dS S ∂S 34 b. Restatement of the Divergence Theorem is: The total divergence within a region D equals the net flux emerging from D. 3. Stokes Theorem a.Theorem (Stokes Theorem) Let S be a smooth oriented surface with boundary ∂S a piecewise smooth, closed oriented curve. Let F be a C 1 vector field. Then: ∫ ∫ ∫ CurlF.dS = F.dR S ∂S (1) Corollary: Let the swirl of the vector field F at → P in the direction − n be defined by ∫ 1 LmA→0 F.dR A C where C is the circumference of a circle of area A centered → at P with unit normal − n . Then: → (a) CurlF.− n = swirl (b) CurlF is in the direction of maximum swirl and has magnitude equal to this maximum swirl. This gives a coordinate free definition of the curl of a vector field. 4. Green’s Theorem 1.Theorem (Green’s Theorem) Let D be a region in R2 with boundary ∂D a simple closed curve. Let P (x, y), Q(x, y) 35 be C 1 functions on a region containing D and its boundary. Then: ∫ ∫ I ∂Q ∂P − )dxdy = P dx + Qdy ( ∂y ∂D D ∂x a. Two Interpretations (1)If F = P (x, y)i + qQ(x, y)j is a planar vector ∂Q field its divergence is then divF = ∂P ∂x + ∂y . If γ is a simple closed smooth planar curve with outward directed normal n then I I −Qdx + P dy = F.nds γ γ Green’s Theorem can then be interpreted as ∫ ∫ ∫ divFdA = F.Nnds d ∂D a planar version of the Divergence Theorem (2)If F = P (x, y)i + q(x, y)j and D is a region in R2 . Consider D then as a surface in R3 with z = 0. ∂P Then CurlF = ( ∂Q ∂x − ∂y )k and Green’s Theorem can be interpreted as ∫ ∫ ∫ curlF.dS = F.R D ∂D which is a restricted version of Stokes Theorem 5. Summary of the Integral Theorems 36 Operator Interpretation gradf = ▽f Maximum rate of change of f DivF = ▽.F Net outflux of F per unit volume CurlF = ▽ × F swirl of F per unit area Integral Theorem ∫Q ∫ ∫P ∫▽f.dR = f (Q) ∫ ∫− f (P ) ∫ ∫ D ▽.FdV = ∫ ∂D F.d§ S ▽ × F.d§ = ∂S F.dR E. Some Basic PDE’s and their Derivations 1. Fourier’s Law of Heat Conduction a. The heat flux q ( calories per sec) across a plane area element A is proportional to the temperature gradient ∂T ∂n normal to A and to the area itself q = −kA ∂T = −kA▽T.n ∂n b. This is a special case of Fick’s first law of diffusion ∂C J = −DA = −DA▽C.n ∂n where J = mass flux - mass per unit time: D = mass diffusivity, A = control surface area , C = concentration of the diffusing species. 2. Continuity Equation of Fluid Mechanics a. Let q be the velocity vector field of a fluid with mass density σ(x, y, z, t). Then ∂σ + ▽(σ.q) = 0 ∂t 37 is the continuity equation of fluid mechanics (1).For fluids this follows from the conservation of mass. (2) If σ = constant its an imcompressible fluid. Then ▽q = div(q) = 0 b. Derivation of the Continuity Equation Consider a velocity vector field q of fluid of mass density σ(x, y, z, t) going into a control volume V . The mass at any time t is then ∫ ∫ ∂σ dM = dV M= σdV =⇒ dt ∂t V V Further dM = mass entering + mass created dt ∫ mass entering = − σq.ndA ∂V ∫ while mass created = f.dV V where f = mass generator. Therefore ∫ ∫ ∫ ∂σ dM = dV = − σq.ndA + f dV dt V ∂t ∂V V applying the divergence theorem ∫ ∫ − σq.ndA = − ▽.(σq)dV ∂V V 38 ∫ ∫ ∫ ∂σ =⇒ dV = − ▽.(σq)dV + f dV V ∂t V v ∫ ∂σ =⇒ ( + ▽(σq) − f )dV = 0 V ∂t Now since V was arbitrary ∂σ + ▽(σq) − f = 0 ∂t By the conservation of mass f = 0 so the basic PDE is =⇒ ∂σ + ▽(σq) = 0 ∂t This is the basic continuity equation for fluid mechanics. Note: If this were an electric field f can be a charge source and can be non-zero. Further if this is an incompressible fluid then it is time independent so σ = k so the equation becomes ▽q = 0 3. Unsteady Heat Conduction a. Consider an unsteady heat conduction with a region D ⊂ R3 Then the temperature field T (x, y, z, t) satisfies (for no source term) α 2 ▽2 T = ∂T ∂t known as the heat equation or diffusion equation. α2 is the diffusivity constant. 39 (1) Steady state conduction when ∂F ∂t = 0 leads to ∂ 2T =0 ∂x2 which is the Laplace Equation. b. Derivation of the Heat Equation The heat in cals contained in a mass m at temperature T with specific heat c is then mcT If σ = σ(x, y, z) is the mass density then m = Therefore the accumulation of heat is given by ∫ ∫ d ∂T cT σdV = c σdV dt V ∂t ∫ σdV . By Fourier’s law of heat conduction the heat flux is ∫ ∫ ∂T k▽T.ndA k σdA = V V ∂n where k is the conductivity of the material. Suppose there is a heat generator f (x, y, z, t) so the heat generated is ∫ f dV ∫ ∫ ∂T k▽T.ndA + f dV c σdV = ∂V V V ∂t Now apply the divergence theorem. Then: ∫ ∫ ∫ ∂T c σdV = ▽.(k▽T )dV + f dV V ∂t V V Then ∫ 40 Again since V is arbitrary ∂T − ▽.(k▽T ) − f = 0 ∂t If k is independent of x, y, z then =⇒ cσ α 2 ▽2 T = where F = f cσ and α2 = k cσ ∂T −F ∂t is the diffusivity. If F = 0 then ∂T ∂t which is called the heat equation. α 2 ▽2 T = 4. Maxwell’s Equations a.Let E be the electric field intensity,(newtons/coulomb), H the magnetic field intensity (amperes, per meter) and J the currect density (amperes per meter squared). Then Maxwell’s equations are ▽×E =0 ▽×H =J b. Derivation of Maxwell’s Equations (a) By Faradays’ law the emf around a closed curve equals the time rate of change of the magnetic flux through the curve. Then: ∫ ∫ ∫ d ∂B E.dR = − B.ndA = − .ndA dt S ∂S S ∂t 41 for every fixed surface S with boundary ∂S where E = electric field intensity B = magnetic flux density Now apply Stokes Theorem ∫ ∫ =⇒ E.dR = ▽ × E.ndA ∂S S ∂B ∂t This is Maxwell’s equation for time varying electric fields. For steady state fields =⇒ ▽ × E = − ▽×E=0 . (b) By Ampere’s law ∫ H.dR = I ∂S where H is the magnetic field intensity and I is the current through any surface which has ∂S as a boundary. Further ∫ I= σq.ndA S where σ is the charge density and q is the velocity vector field. Let J = σq = current density 42 ∫ then: ∫ H.dR = ∂S σ.q.ndA S Apply Stokes Theorem ∫ ∫ ∫ H.dR = ▽ × (H).dA = J.ndA ∂S S S Since S was arbitrary ▽×H=J 5. The Wave equation a.Wave Equations (1) The One-Dimensional Wave Equation is β 2 uxx = utt (2) Roughly wave equations arise is modeling wave phenomena (3) This arises from problems in mechanical vibrations. A solution u(x, t) will represent the small displacements of an idealized vibrating string (a) The wave equation also appears in the theory of high frequency transmission lines, fluid mechanics, acoustics and electricity. b.Derivation of the Wave Equation 43 (1) The Basic Assumptions (a) Let y(x, t) the transverse vibrations of a string stretched between x = 0 and x = L. The solution y(x, t) measures the displacements from the x-axis for t > 0. f (x) represents the initial configuration of the string and g(x) represents the initial velocity at each point. (2) Conditions for One-Dimensional Wave Equation (a) The string is perfectly flexible (b) The displacements are small relative to the ∂y length of the string - hence y and ∂x are small (c) The tension of the string is constant (d) The string is homogeneous, that is its mass per unit length is constant (e) The tension is large compared to the force of gravity (f) No other forces act on the string (3) Derivation (a) Assume that the tension is τ and the density is ρ. Then using Newton’s law we get ∂ 2y τ sin θ(x+∆x, t)−τ sin θ(x, t)−f (x+α∆x, t)∆x = (ρ∆s) 2 (x+β∆x, t) ∂t That is the sum of the vertical forces is equal to the vertical acceleration, 44 (i) Under the assumptions θ is small and therefore sin θ u θ u tan θ = ∂y and ∆s u ∆x ∂x Therefore τ ∂y ∂y ∂ 2y ( (x+∆x, t))− (x, t))−f (x+α∆x, t) = ρ 2 (x+β∆x, t) ∆x ∂x ∂x ∂t Letting ∆x → 0 yields ∂ 2y ∂ 2y τ 2 − f (x, t) = ρ 2 ∂x ∂t (ii) If f = ρg just the force of gravity then τ ∂ 2y ∂ 2y = ρ + ρg ∂x2 ∂t2 Further if gravity is negligible then ∂ 2y ∂ 2y c2 2 = 2 where c = ∂x ∂t 45 √ τ ρ PARTIAL DIFFERENTIAL EQUATIONS (3) TAYLOR SERIES AND POWER SERIES SOLUTIONS A. Power Series and Taylor Series 1. Taylor’s Theorem a. Theorem If f (x) is a C n+1 [a, b] function then for each x in [a, b] f (x) = f (a) + f ′ (a)(x − a) + ..... + f (n) (a) (x − a)n + Rn (x) n! where Rn (x) = f (n+1) (c) (x − a)n+1 for some c, a < c < x (n + 1)! . Tn (x) = f (a) + f ′ (a)(x − a) + ..... + f (n) (a) (x − a)n n! is the degree n Taylor polynomial centered on a for f (x) while Rn (x) is the remainder. If a = 0 its called a Maclaurin polynomial (1) Thus a C n+1 function can be approximated by its degree n Taylor polynomial with error term given by Rn (x). If we can get a uniform bound on the error we get a uniform approximation for f (x). 2. Power Series ∑ n a.Definiiton P (x) = ∞ n=0 an (x − x0 ) is a power series with center x0 . For each value of x we get an 46 infinite series and for those x values for which the series converges we get a function. ∑ n (1) Theorem For P (x) = ∞ n=0 an (x−x0 ) there is a number r ≥ 0 called the radius of convergence such that P (x) converges for all x such that |x − x0 | < r and diverges if |x − x0 | > r. Generally 1 |an+1 | = L−1 m r |an | (a) Corollary A power series can be differentiated and integrated within its circle of convergence. Further an = P (n) (x0 ). 3. Taylor Series a. Definition If the function f (x) can be written as a power series f (x) = ∞ ∑ f (n) n=0 n! (x − x0 )n then this power series is the Taylor series for f (x) centered at x0 . If x0 = 0 it is called a Maclaurin series. If f (x) has a Taylor series about x0 then we say that f (x) is analytic at x0 (1) Theorem If f (x) is a C ∞ function on [a, b] with x0 in (a, b) and if the remainder Rn (x) goes to zero uniformly in an interval about xx0 then f (x) is analytic at x0 . Note: Not every C ∞ function is analytic. 47 4. Some Common Taylor Series a. P (x) = P (x) if P (x) is a polynomial ∑ 1 n b. 1−x = ∞ n=0 x , −1 < x < 1 ∑ xn c. ex = ∞ n=0 n! , all x ∑ n x2n+1 d. sin x = ∞ n=0 (−1) (2n+1)! , all x ∑ n x2n e. cos x = ∞ n=0 (−1) (2n)! , all x ∑ n xn+1 f. Ln(x + 1) = ∞ n=0 (−1) n+1 , −1 < xLe1 5. Manipulating Series - often other series can be found by manipulating exisiting series. bf EXAMPLE ∞ ∞ 1 d 1 d ∑ n ∑ n−1 nx x = = ( )= (1 − x)2 dx 1 − x dx n=0 n=1 B. Power Series Solutions to ODE’s 1.Power Series Methods a. Consider a second order linear ODE y ′′ + P (x)y ′ + Q(x)y = 0 (1) Definition A point x = x0 is an ordinary point for equation (1) if both P (x) and Q(x) are analytic 48 at x0 . Otherwise it is a singular point. We only consider here ordinary points. b. Procedure for Power Series Solutions Step 1: Consider a solution of the form y(x) = n n=0 an (x − x0 ) ∑∞ Step 2: Substitute this solution type into equation (1) to get a recurrence relation among the coefficients an . Step 3: There will generally be two free parameters. Values for these will give the two independent solutions. EXAMPLE Solve y ′′ − 2xy = 0. Let y = a0 + a1 x + a2 x2 + a3 x3 + ....... then y ′ = a1 + 2a2 x + 3a3 x2 + 4a4 x3 + ..... y ′′ = 2a2 + (3)(2)a3 x + (4)(3)a4 x2 = .... so y ′′ − 2xy = 2a2 + [(3)(2)a3 − 2a1 ]x + ..... = 0 In general then 2a2 + ∞ ∑ [(k + 2)(k + 1)ak+2 − 2ak−1 ]xk = 0 k=1 Therefore the coefficients must be zero. This implies a2 = 0 ak+2 = 2ak−1 (k + 2)(k + 1) 49 substtituing in for the first few values we get that a0 , a1 are arbitrary and 1 1 1 a2 = 0, a3 = a0 , a4 = a1 , a5 = 0, a6 = a0 3 6 45 Therefore the solution is given by 1 1 y = a0 + a1 x + a0 x3 + a1 x4 + ..... 3 6 2.Taylor Series Methods a. Consider a second order linear initial value problem y ′′ + P (x)y ′ + Q(x)y = 0 : y(x0 ) = y0 , y ′ (x0 ) = y1 b. Procedure for Taylor Series Solutions Step 1: Consider a Taylor series solution 1 1 y(x) = y(x0 )+y ′ (x0 )(x−x0 )+ y ′′ (x0 )(x−x0 )2 + y ′′′ (x0 )(x−x0 )3 +.... 2 3! Therefore we have a solution once we have the values y(x0 ), y ′ (x0 ), y ′′ (x0 ) etc. Step 2: Substitute into the equation to get the above values EXAMPLE (a) Solve y ′ = y : y(0) = 1 Know y ′ = y =⇒ y ′ (0) = y(0) = 1 50 differentiating =⇒ y ′′ = y ′ =⇒ y ′′ (0) = y ′ (0) = 1 Continuing we get that y(x) = 1 + x + 1 2 1 3 x + x + ... = ex 2! 3! (b) Solve y ′′ + y = 0 : y(0) = 0, y ′ (0) = 1 Know y ′′ = −y =⇒ y ′′ (0) = −y(0) = 0 differentiating =⇒ y ′′′ = −y ′ =⇒ y ′′′ (0) = −y ′ (0) = −1 Continuing we get that y(x) = x − 1 3 x + ... = sin x 3! (c) Solve y ′′ − 2xy = sin x : y(0) = 0, y ′ (0) = 1 Know y ′′ = 2xy + sin x =⇒ y ′′ (0) = 0 differentiating =⇒ y ′′′ = 2xy ′ +2y+cos x =⇒ y ′′′ (0) = 1 differentiating =⇒ y (4) = 2xy ′′ +2y ′ +2y ′ −sin x =⇒ y (4) (0) = 4 Therefore y(x) = x + 1 3 4 4 x + x + ..... 3! 4! 51 PARTIAL DIFFERENTIAL EQUATIONS (4) FOURIER SERIES AND FOURIER ANALYSIS A. Fourier Series of a Function 1.General Preliminaries a.The Idea of a Fourier Series (1) Definition A function f (x) is periodic of period m if f (x + m) = f (x) for all x. (a) Lemma sin x and cos x are both periodic of period 2π. (b) Trigonometric functions are periodic and periodic behavior appears quite often in physical applications. The basic idea in a Fourier series or Fourier expansion is to express a general periodic function as a sum or infinite series of trigonometric functions. b.Orthogonality Relations ∫ (1) Theorem (i) l cos( −l mπx nπx ) cos( )dx = 0, m ̸= n = 1, m = n ̸= 0, m = n ̸= 2l, m = n = 0 l l (ii) ∫ l sin( −l (iii) mπx nπx ) sin( )dx = 0, m ̸= n = 1, m = n ̸= 0 l l ∫ l cos( −l nπx mπx ) sin( )dx = 0, ∀m, n l l 52 2. The Fourier Series of a Function a.General Trigonometric Series (1) Definition A trigonometric series is an infinite series of the form ∞ ∑ (an cos(knx) + bn sin(knx) n=1 b. The Fourier Series of a Periodic Function (1) Definition Let f (x) be a periodic function with fundamental period 2l. Then its Fourier series is a0 + ∞ ∑ (an cos( n=1 nπx nπx ) + bn sin( ) l l where 1 an = l 1 bn = l 1 a0 = 2l ∫ ∫ l f (x)dx −l l f (x) cos( −l ∫ l f (x) sin( −l nπx )dx, n = 1, 2, ... l nπx )dx, n = 1, 2, ... l The an , bn are called the Fourier coefficients while cos( nπx l ) nπx and sin( l ) are the harmonics 53 (a) Note A trigonometric series clearly is periodic. It is a Fourier series if it is the Fourier series of some periodic function (b) Note a0 = average value of f (x) on (−l, l) (c) Lemma There exists trigonometric series which are not Fourier series. For example ∞ ∑ n=1 1 sin(nx) ln(1 + n) converges for each x but is not a Fourier series. (d) Note If f (x) is an odd fucntion then its Fourier series has no cosine terms while if f (x) is an even function f (x) has no sine terms. c. Complex Form of a Fourier Series (1) In terms of complex variables inπx l nπx e cos( )= l + e− 2 inπx l − e− l 2i ∞ ∑ inπx cn e l =⇒ f (x) = nπx e sin( )= l inπx inπx l n=−∞ where 1 cn = 2l ∫ l f (x)e− inπx l dx −l d. The Fourier Convergence Theorem 54 (1) Theorem Let f (x) be periodic of period 2l. Then: (i) If both f (x) and f ′ (x) are piecewise continuous on (−l, l) then the Fourier series converges point+ (x− ) wise to the mean value f (x )+f . 2 (ii) If both f (x) and f ′ (x) are continuous on (−l, l) then the Fourier series converges uniformly to f (x). (a) Corollary At each jump discontinuity the Fourier series converges to the average value of the jump. (b) Corollary If f (x) is a continuously differentiable periodic function then f (x) is represented everywhere by its Fourier series. EXAMPLE Let f (x) be the square wave function defined by f (x) = 0, −π < x < 0 and = 4, 0 < x < π and = 2, x = −π, 0, π and then periodic of period 2π. We determine the Fourier expansion of f (x). Now 1 a0 = 2l Then 1 an = l ∫ l 1 f (x)dx = 2π −l ∫ ∫ π 1 f (x)dx = 2π −π ∫ π 4dx = 2 0 ∫ 1 π nπx )dx = f (x) cos( f (x) cos(nx)dx l π −π −l ∫ 1 π 4 cos(nx)dx = 0 = π 0 l 55 Finally ∫ nπx 1 π f (x) sin( f (x) sin(nx)dx )dx = l π −π −l ∫ 4 1 π 4 sin(nx)dx = (1 − cos(nπ)) = π 0 nπ Further since cos(nπ) = (−1)n we get 1 bn = l ∫ l 4 8 (1 − (−1)n ) = , n = 1, 3, .... nπ nπ Therefore the Fourier expansion of f (x) is given by bn = ∞ 8 ∑ 1 f (x) = 2 + sin(nx) π n=1,3,.. n By the Fourier convergence theorem this series will converge to f (x) at each point of continuity of f (x) - hence if x ̸= nπ. e.Finite Interval Half-Range Expansions (1) Suppose f (x) is defined on (0, l). Then a periodic extension f (x) can be created on (−l, l). The Fourier series of f (x) will then converge to f (x) on (0, l). (a) If f (−x) = f (x) that is we define f (x) = f (−x) on (−l, 0) then we get the even periodic extension. The resulting Fourier series has only cosine terms and is the half-range cosine expansion. (b) If f (−x) = −f (x) we get the odd extension and the resulting Fourier series is the half-range sine expansion. 56 (2) We can also define quarter-range sine and cosine expansions (3) Formulas for Periodic Expansions - f (x) defined on 0 < x < L (a) Half- Range Sine f (x) = ∞ ∑ bn sin( n=1 2 bn = L ∫ L f (x) sin( 0 nπx ) L nπx )dx L (b) Half- Range Cosine f (x) = a0 + ∞ ∑ an cos( n=1 nπx ) L ∫ 1 L f (x)dx a0 = L 0 ∫ 2 L nπx an = f (x) cos( )dx L 0 L (c) Quarter- Range Sine ∞ ∑ f (x) = bn sin( n=1,3,... 2 bn = L ∫ L f (x) cos( 0 57 nπx ) 2L nπx )dx 2L (d) Quarter- Range Cosine ∞ ∑ f (x) = an cos( n=1,3... 2 an = L ∫ L f (x) cos( 0 nπx ) 2L nπx )dx 2L EXAMPLE Consider the function f (x) = k, 0 < x < L. We find the Fourier half-range cosine and half-range sine expansion. First the half-range cosine expansion. From the formulas above we have ∫ 1 L a0 = kdx = k L 0 ∫ 2 L nπx 2k nπx L an = k cos( )dx = sin( )| = 0 l 0 L nπ L 0 Therefore the Fourier half-range cosine expansion is just f (x) = k. Next the half-range sine expansion. Again from the formulas above we have ∫ nπx 2k 2 L bn = k sin( )dx = − (cos(nπ) − 1) l 0 L nπ Therefore the Fourier half-range sine expansion is ∞ 2k ∑ 1 − cos(nπ) nπx f (x) = sin( ) π n=1 n L 58 3. Bessel’s Inequality and the Parseval Identity a.Theorem If the periodic function f (x) has the Fourier series ∞ ∑ nπx nπx a0 + (an cos( ) + bn sin( ) l l n=1 then: k ∑ 1 (i) 2a20 + (a2n + b2n ) ≤ l n=1 ∫ l −l |f (x)|2 dx for all k = 1, 2, .... This is called Bessel’s Inequality. (ii) 2a20 + ∞ ∑ (a2n + b2n ) n=1 1 = l ∫ l −l |f (x)|2 dx This is called Parseval’s Identity . 4. Manipulation of Fourier Series a.Uniqueness of Fourier Series (1)Theorem If two trigonometric series each of the form a0 + ∞ ∑ n=1 (an cos( nπx nπx ) + bn sin( ) l l converge to the same sum for all x then the corresponding coefficients are equal. b.Termwise Integration of Fourier Series 59 (1)Theorem Let f (x) be a piecewise continuous periodic function of period 2l with Fourier series a0 + ∞ ∑ (an cos( n=1 nπx nπx ) + bn sin( ). l l Suppose that a0 = 0. Then the Fourier series of the integral function ∫ x F (x) = f (t)dt −l can be obtained by termwise integration and the resulting series converges to F (x) for all x. c.Termwise Differentiation of Fourier Series (1)Theorem Let f (x) be a piecewise continuous periodic function of period 2l with Fourier series a0 + ∞ ∑ n=1 (an cos( nπx nπx ) + bn sin( ). l l Suppose that f ′ (x) is also piecewise continuous on (−l, l). Then the Fourier series for f ′ (x) can be obtained by termwise differentiation of the Fourier series for f (x). B. Sturm-Liouville Theory 1.Sturm-Liouville Problems a. Definition A Sturm-Liouville Problem is a second-order linear boundary value ODE of the form (−p(x)y ′ (x))′ + q(x)y(x) = λr(x)g(x), a < x < b 60 c1 y(a) + c2 y ′ (a) = 0 c3 y(b) + c4 y ′ (b) = 0 and p(x), p′ (x), g(x), r(x) ∈ C 0 (a, b), p(x) > 0, r(x) > 0 on (a, b) and c21 + c22 ̸= 0, c23 + c24 ̸= 0. (1) y(x) ≡ 0 solves this equation. Each value of λ which allows a non-trivial solution is called an eigenvalue and the corresponding function is an eigenfunction. b. Definition The differential operator 1 d d L = − ( (p + q) r dx dx is defined on the space of C 2 functions satisfying the boundary conditions. The Sturm-Liouville problem then has the form Ly = λy (1) Lemma On the inner product space of L2 functions with inner product ∫ b < u, v >= u(x)v(x)r(x)dx a the operator L is a self-adjoint operator. That is < Lu, v >=< u, Lv > c. The Spectrum of L (1) Theorem The eigenvalues and eigenfunctions of L have the following properties. 61 (i) All eigenvalues are real and compose a countable infinite discrete set λ1 < λ2 < ..... → ∞ (ii) To each eigenvalue λn corresponds only one independent eigenfunction wn (x) (iii) The normalized eigenfunctions (relative to the weight function r(x)) form a complete orthonormal family in L2 . 2.Eigenfunction Expansions a. Definition If {ϕn (x)} is a collection of eigenfunctions and f (x) is an L2 function then its eigenfunction expansion is ∞ ∑ < f (x), ϕn (x) > ϕn (x) < ϕ (x), ϕ (x) > n n i=1 (1) Pointwise Convergence of Eigenfunction Expansions (a) Theorem (i) If both f (x) and f ′ (x) are piecewise continuous in (a, b) then the eigenfunction expansion + (x− ) converges pointwise to f (x )+f for all x ∈ (a, b). 2 (ii) If f (x), f ′ (x) are continuous in (a, b) and f ′′ (x) is piecewise continuous and f (x) satisfies the boundary conditions of the base Sturm-Liouville problem then the eigenfunction expansion converges uniformly to f (x) on (a, b) 62 PARTIAL DIFFERENTIAL EQUATIONS (5) FOURIER INTEGRAL AND FOURIER TRANSFORMS A. Integral Transforms 1.General Idea of an Integral Transform a. The basic idea of an integral transform is to transform a function f (t) into another function f (w) via integration. This transformation will be linear on functions. Integral transforms will behave relative to differential equations much the way logarithms do for multiplicative equations they linearize them and simplify them. b. Definition Let K(s, t) be a function of the variables s, t defined for t ∈ I where I is some interval. K(s, t) is called the kernel. If f (t) is a function of t defined on I then its integral transform rerlative to the kernel K(s, t) is the function ∫ f (s) = f (t)K(s, t)dt I (1) There are many important integral transforms. However for our purposes in differential equations we will examine two - the Laplace transform and the Fourier transform. B. The Laplace Transform 1. The general idea of Laplace transforms is to change a linear differential equation into an ordinary algebraic equation. It plays a very similar role to logarithms which change multiplicative equations into additive equations. 63 2. The Laplace Transform a.Definition Let f (t) be defined on [0, ∞). Then its Laplace transform is the function F (s) = L(f (t)) defined by ∫ ∞ F (s) = L(f (t)) = e−st f (t)dt 0 provided this integral exists. Recall that ∫ ∞ ∫ b g(x)dx = lim g(x)dx b→∞ a a EXAMPLES (a) Evaluate L(1) ∫ L(1) = ∞ −1 −st ∞ 1 e |0 = if s > 0 s s e−st dt = 0 (b) Evaluate L(t) ∫ L(t) = ∞ −st te 0 −te−st ∞ 1 dt = | + s 0 s ∫ ∞ e−st dt by integration by parts 0 1 1 = 0 + L(1) = 2 if s > 0 s s (c) Evaluate L(e−bt ) L(e −bt ∫ )= ∞ e 0 −bt −st e ∫ ∞ dt = 0 64 e−(b+s)t dt = 1 if s > −b s+b (1) Definition A function f (t) is of exponential order if there exist constants M, c, T such that |f (t)| ≤ M ect for all t > T (a) Theorem If f (t) is of exponential order on [0, ∞) then L(f (t)) exists for t > c (b) From now on we will be concerned only with functions that have Laplace transforms. b. Theorem The Laplace transform is a linear operator on functions. That is L(af (t) + bg(t)) = aL(f (t)) + bL(g(t)) c. The Common Transforms (1) Theorem (i) L(1) = 1 s (ii) L(tn ) = n! sn+1 , n = 1, 2, .... 1 (iii) L(eat ) = s−a k (iv) L(sin kt) = s2 +k 2 s (v) L(cos kt) = s2 +k 2 k (vi) L(sinh kt) = s2 −k 2 s (vii) L(cosh kt) = s2 −k 2 3. The Inverse Transform a. Definition If f (t) is a function such that L(f (t)) = F (s) the f (t) is the inverse Laplace transform of F (s), =⇒ f (t) = L−1 (F (s)) 65 (1) The Common Inverses (a) Theorem (i) 1 = L−1 ( 1s ) n! (ii) tn = L−1 ( sn+1 )n = 1, 2, .... 1 (iii) eat = L−1 ( s−a) k (iv) sin kt = L−1 ( s2 +k 2) s (v) cos kt = L−1 ( s2 +k 2) k (vi) sinh kt = L−1 ( s2 −k 2) s (vii) cosh kt = L−1 ( s2 −k 2) EXAMPLE Evaluate L−1 ( 3s+5 s2 +7 ), 3s + 5 3s 5 ) = L−1 ( 2 ) + L−1 ( 2 )= 2 s +7 s +7 s +7 √ √ √ 5 7 5 s −1 −1 √ √ )+ ) = 3 cos( 7t) + L ( 2 sin( 7t) 3L ( 2 s +7 s +7 7 7 L−1 ( (b) Use of Partial fractions 1 EXAMPLE Evaluate Lv( (s−1)(s+2)(s+4) ) Method is to try to find A, B, C such that 1 A B C = + + (s − 1)(s + 2)(s + 4) s − 1 s + 2 s + 4 Combining the three fractions we get 1 A(s + 2)(s + 4) + B(s − 1)(s + 4) + C(s − 1)(s + 2) = = (s − 1)(s + 2)(s + 4) (s − 1)(s + 2)(s + 4) 66 (A + B + C)s2 + (6A + 3B + C)s + 8A − 4B + 2C (s − 1)(s + 2)(s + 4) Equating the tops gives three linear equations for A, B, C = A+B+C =0 6A + 3B + C = 0 8A − 4B + 2C = 1 Solving we get A = L−1 ( 1 15 , B = −1 6 ,C = 1 10 . Therefore 1 1 1 1 1 1 1 ) = L−1 ( )− L−1 ( )+ L−1 ( ) (s − 1)(s + 2)(s + 4) 15 (s − 1) 6 (s + 2) 10 (s + 4) = 1 t 1 −2t 1 e − e + e−4t 15 6 10 4. Some Properties of Laplace Transforms a. The Translation Theorem (1) Theorem L(eat f (t)) = F (s−a) if F (s) = L(f (t)) EXAMPLE Evaluate L(e5t t3 ) L(e5t t3 ) = F (s − 5) where F (s) = L(t3 ) L(t3 ) = 6 6 5t 3 =⇒ L(e t ) = s4 (s − 5)4 (2) Theorem eat f (t) = L(F (s−a)) = L(F (s))|s→s−a if F (s) = L(f (t)) s EXAMPLE Evaluate L−1 ( s2 +6s+11 ) L−1 ( s s s+3−3 −1 −1 ) = L ( ) = L ( ) s2 + 6s + 11 (s + 3)2 + 2 (s + 3)2 + 2 67 s+3 3 −1 ) − L ( ) (s + 3)2 + 2 (s + 3)2 + 2 Let v = s + 3 then √ 3 2 v − √ L−1 ( 2 ) = L−1 ( 2 v +2 v +2 2 = L−1 ( √ √ 3 = e−3t cos( 2t) − √ sin( 2t) 2 b. Derivatives of a Transform (1) Theorem 1, 2, .... L(tn f (t)) = (−1)n d n L(f (t)) dsn for n = EXAMPLES (a) Evaluate L(te3t ) L(te3t ) = − d d 1 1 (L(e3t )) = − ( )= ds ds s − 3 (s − 3)2 (b) Evaluate L(t sin kt) L(t sin kt) = − 2ks d k d ) = (L(sin kt)) = − ( 2 ds ds s + k 2 (s2 + k 2 )2 c. Transforms of Derivatives (1) Theorem L(f (n) (t)) = sn F (s) − sn−1 f (0) − sn−2 f ′ (0)..... − f (n−1) (0) where F (s) = L(f (t)) 68 d. The Convolution Theorem (1) Definition If f (x), g(x) are defined on [0, ∞) then their convolution if the function F (t) = f ∗ g(t) defined by ∫ ∞ f ∗ g(t) = f (u)g(t − u)du 0 (a) Theorem ∫ ∞ L( f (u)g(t − u))du = L(f (t))L(g(t)) 0 In words this says that the Laplace transform of the convolution of two functions is the product of the Laplace transforms. 5.Solving Differential Equations with Laplace Transforms a. The General Method (1) Consider the initial value problem an y (n) + an−1 y (n−1) + ... + a0 y = g(t) y(0) = y0 , y ′ (0) = y1 , ..., y (n−1) (0) = yn (2) Apply the Laplace Transform L(an y (n) + an−1 y (n−1) + ... + a0 y) = L(g(t)) = an L(y (n) ) + ... + a0 L(y) = L(g(t)) 69 Now applying the results on transforms of derivatives an [sn Y (s) − sn−1 y(0)] + ... = G(s) where Y (s) = L(y), G(s) = L(g). Then (an sn +...+a0 )Y (s) = an [sn−1 y0 +...+yn ]+an−1 [sn−2 y0 +...+yn−1 ]+...+G(s) Now Y (s) = L(y(t)) so the solution is then y(t) = L−1 (Y (s)) EXAMPLE Solve y ′ − 3y = e2t : y(0) = 1. y ′ = 3y + e2t =⇒ L(y ′ ) = 3L(y) + L(e3t ) Now let Y = L(y) so L(y ′ ) = sY − y(0) = sY − 1 and 1 L(e2t ) = s−2 so L(y ′ ) = 3L(y)+L(e3t ) =⇒ sY −1 = 3Y − 1 s−1 =⇒ Y = s−2 (s − 2)(s − 3) Therefore the solution is y(t) = L−1 ( s−1 ) (s − 2)(s − 3) By partial fractions s−1 1 2 =− + (s − 2)(s − 3) s−2 s−3 =⇒ y(t) = L−1 (− 1 2 + ) = −e2t + 2e3t s−2 s−3 6.The Heaviside and Dirac Delta Functions a.The Heaviside Function 70 (1) Definition The Heaviside step function is defined by H(x) = 1 : t ≥ 0 H(x) = 0 : t < 0 b. The Dirac Delta Function (1) Definition Consider the unit impulse function 1 : −ϵ < t < ϵ 2ϵ = 0 : elsewhere δϵ (t) = Then for each ϵ we have ∫ ∞ −∞ δϵ (t)dt = 1 The Dirac Delta Function is the limit of the unit impulse functions. δ(t) = Lmϵ→0 δϵ (t) (a) Lemma For the Dirac delta function we have ∫ ∞ δ(t)dt = 1 −∞ Hence δ(t) can be characterized as being infinite at 0, have the value 0 everywhere else and the total integral is one. (b) Lemma The Laplace transforms of δ(t), t ≥ 0 and the shifted dleta function δ(t − t0 ) are L(δ(⊔)) = ∞ 71 L(δ(⊔ − ⊔′ )) = ⌉−∫ ⊔′ C. The Fourier Integral Theorem 1. The Fourier Integral Theorem a.Definition If f (x) ∫ ∞is defined and absolutely integrable on (−∞, ∞), (that is −∞ |f (x)|dx converges) then its Fourier integral expansion is ∫ ∞ f (x) = (A(α) cos(αx) + B(α) sin(αx))dα 0 where 1 A(α) = π B(α) = 1 π ∫ ∞ f (x) cos(αx)dx −∞ ∫ ∞ f (x) sin(αx)dx −∞ (1) Theorem (The Fourier Integral Theorem) Suppose f (x) is defined and piecewise continuous and absolutely integrable on (−∞, ∞) and f ′ (x) is also piecewise continuous. Then its Fourier integral expansion converges to f (x + 0) + f (x − 0) 2 at each x and hence to f (x) at each point of continuity. b.Equivalent Forms of the Integral Theorem 72 (1) Theorem The Fourier integral theorem can be written in the following equivalent forms: ∫ ∫ 1 ∞ ∞ (i)f (x) = f (u) cos α(x − u)dudα π α=0 u=−∞ ∫ ∞∫ ∞ 1 (ii)f (x) = f (u)eiα(x−u) dudα 2π −∞ −∞ ∫ ∞ ∫ ∞ 1 (iii)f (x) = eiαx ( f (u)eiα(x−u) du)dα 2π −∞ −∞ where f (x) = f (x) at each point of continuity and = otherwise. f (x+0)+f (x−0) 2 The last form (iii) is the basis for the Fourier Inversion Theorem. D. The Fourier Transform 1. Fourier Transforms a.Definition If f (x) is a function satisfying the conditions of the Fourier Integral Theorem then its Fourier transform is ∫ ∞ F(f (x)) = f (w) = f (u)e−iwu du −∞ (1) Theorem (The Fourier Inversion Theorem) If f (x) is a function satisfying the conditions of the Fourier Integral Theorem and f (w) is its Fourier transform then ∫ ∞ 1 f (x) = f (w)eiwx du 2π −∞ 73 (Restatement of the Fourier integral theorem) 2. Properties of Fourier Transforms a.Linearity (1) Theorem The Fourier transform and its inverse are linear operators on the space of functions F(αf (x) + βg(x)) = αF(f (x)) + βF(g(x)) F −1 (αf (w) + βg(w)) = αF −1 (f (w)) + βF −1 (g(w)) b.Transforms of Derivatives (1) Theorem We have for a differentiable function with a Fourier transform F(f (n) (x)) = (iw)n f (w) c.Convolution Theorem (1) Definition The Fourier convolution of f (x), g(x) denoted f ⋆ g(x) is ∫ ∞ f ⋆ g(x) = f (x − u)g(u)du −∞ (a) Theorem (The Convolution Theorem) Suppose f (x), g(x) are functions satisfying the Fourier Integral Theorem and f (w), g(w) their Fourier transforms. Then the Fourier transform of the convolution is the product of the Fourier transforms. That is F(f ⋆ g) = f (w)g(w) 74 F −1 (f g) = f ⋆ g d.Parseval’s Identities for Fourier Integrals (1) Theorem If f (w), g(w) are the Fourier transforms of f (x), g(x) respectively then ∫ ∞ ∫ ∞ 1 f (x)G(x)dx = f (w)g(w)dw 2π −∞ −∞ ∫ ∞ ∫ ∞ 1 2 |f (x)| dx = |f (w)|2 dw 2π −∞ −∞ 75 PARTIAL DIFFERENTIAL EQUATIONS (7) THE BASIC BOUNDARY VALUE PROBLEMS A. Method of Separation of Variables 1.For a homogeneous linear PDE the method of separation of variables attempts to find a solution of the form u(x, y) = X(x)Y (y) EXAMPLE ∂ 2u ∂u = 4 ∂x2 ∂y Let u = XY where X = X(x), Y = Y (y) then =⇒ X ′′ Y = 4XY ′ X ′′ Y′ = 4X Y Since X is a function of x alone and Y is a function of y alone this equality is possible only if both sides equal the same constant c. =⇒ Case 1 c = λ2 > 0 then X ′′ − 4λ2 X = 0 Y ′ − λ2 Y = 0 =⇒ X = c1 e2λx + c2 e−2λx and Y = c3 eλ 2 =⇒ u(x, y) = Aeλ y (Be2λx + Ce−2λx ) 2 76 y Case 2 c = −λ2 < 0 then X ′′ + 4λ2 X = 0 Y ′ + λ2 Y = 0 =⇒ X = c1 cos(2λx) + c2 sin(2λx) and Y = c3 e−λ 2 y =⇒ u(x, y) = Ae−λ y (B sin(2λx) + C cos(−2λx)) 2 Case 3 c = 0 then X ′′ = 0 Y′ =0 =⇒ X = c1 + c2 x and Y = c3 =⇒ u(x, y) = Ax + B 2. Superposition Principle a. Theorem If u1 , u2 , ..., un are solutions to a linear homogeneous PDE then any linear combination is also a solution. (1) We assume that if u1 , u2 ..., un , ... are an infinite sequence of solutions then ∞ ∑ ui i=1 is also a solution B. Boundary Value Problems 1.Basic Boundary Value Problems 77 a. Recall our three basic prototype PDE’s (1) Diffusion or Heat Equation α2 uxx = ut (a) This is the one-dimensional heat equation. Recall in 3-variables the heat equation if α2 ▽2 u = ut . (b) This arises in the flow of heat in a solid in one direction. u(x, t) represents the temperature. This equation can also represent the flow of electricity in a cable - then its known as the telegraph equation (2) Wave Equation β 2 uxx = utt (a) This is the one-dimensional wave equation. Roughly wave equations arise is modeling wave phenomena (b) This arises from problems in mechanical vibrations. A solution u(x, t) will represent the small displacements of an idealized vibrating string [1] The wave equation also appears in the theory of high frequency transmission lines, fluid mechanics, acoustics and electricity. (3) Laplace Equation 78 uxx + uyy = 0 (a) Roughly Laplace equations arise is modeling equilibrium problems. A solution u(x, t) represents steady state temperature in a thin flat plate. Also arises in problems delaing with electrostatic potentials, gravitational potentials, and velocity potentials in fluid dynamics. b. Consider these basic equations subject to: (1) Boundary conditions (a) u or ux specified at x = x0 or u,uy specified at y = y0 . (2) Initial conditions (a) Specifying u(x, 0) = f (x) in the heat equation or u(x, 0) = f (x), ut (x, 0) = g(x) in the wave equation (3) The PDE together with these side conditions, both boundary and initial is called a boundary value problem. 2.The Heat Equation a. Basic Boundary Value Problem kuxx = ut : k > 0, 0 < x < L, t > 0 u(0, t) = 0, u(L, t) = 0 : t > 0 u(x, 0) = f (x) : 0 < x < L 79 b. Physical Interpretation (1) This represents the flow of heat in the x direction along a thin rod pf length L with an initial temperature of f (x) with ends held constant at 0 for all t. k > 0 is the diffusivity and is proportional to the thermal conductivity. The solution u(x, t) is the temperature. (2) Conditions for One-Dimensional Heat Equation (a) The flow of heat is only in the x-direction (b) No heat escapes from the lateral surface of the rod (c) No heat is being generated in the rod (d) The rod is homogeneous, that is its density per length is constant (e) The specific heat and thermal conductivity are constant c. The Solution (1) Attempt a separation of variables solution u = XT . We then arrive at X ′′ T′ = = −λ2 X kT Here we choose the constant c = −λ2 < 0 indicating that the heat is dying exponentially as it goes to the boundary. =⇒ X ′′ + λ2 X = 0 80 X(0) = 0, X(L) = 0 T ′ + kλ2 T = 0 The solution for T (t) is then T (t) = Ae−kλ 2 t The solution for X(x) is a Sturm-Liouville problem with eigenvalues nπ λn = , n = 1, 2, 3, .... L and eigenfucntions ϕn (x) = sin( nπx ), n = 1, 2, 3... L Therefore for each n = 1, 2, 3, .... we have a solution un (x, t) = An e−kλ t sin( 2 nπ x) L We have to be able to solve for the initial condition so by the superposition principle we also have the solution u(x, t) = ∞ ∑ An e−kλn t sin( 2 n=1 nπ x) L Now the initial condition u(x, 0) = f (x) =⇒ f (x) = ∞ ∑ n=1 An sin( nπ x) L We recognize this as the Fourier half-range sine expansion of f (x) defined on (0, L). Therefore 81 2 An = L ∫ L f (x) sin( 0 nπ x)dx L d. Insulated Boundaries (1) The heat equation is modified if the ends are insulated. Here the normal derivative of the temperature is zero. (2) For insulated boundaries the boundary value problem is kuxx = ut : k > 0, t > 0, 0 < x < L ux (0, t) = 0, ux (L, t) = 0 : t > 0 u(x, 0) = f (x); 0 < x < l 3.The Wave Equation a. Basic Boundary Value Problem a2 uxx = utt : 0 < x < L, t > 0 u(0, t) = 0, u(L, t) = 0 : t > 0 u(x, 0) = f (x), ut (x, 0) = g(x) : 0 < x < L f (0) = f (L) = 0 b. Physical Interpretation (1) This represents the transverse vibrations of a string stretched bewteen x = 0 and x = L. The solution u(x, t) measures the displacements from the x-axis for t > 0. 82 f (x) represents the initial configuration of the string and g(x) represents the initial velocity at each point. (2) Conditions for One-Dimensional Wave Equation (a) The string is perfectly flexible (b) The displacements are small relative to the length of the string (c) The tension of the string is constant (d) The string is homogeneous, that is its mass per unit length is constant (e) The tension is large compared to the force of gravity (f) No other forces act on the string c. The Solution (1) Attempt a separation of variables solution u = XT . We then arrive at T ′′ X ′′ = 2 = −λ2 X aT =⇒ X ′′ + λ2 X = 0 X(0) = 0, X(L) = 0 T ′′ + a2 λ2 T = 0 The solution for T (t) is then T (t) = c1 cos(aλt) + c2 sin(aλt) 83 The solution for X(x) is a Sturm-Liouville problem with eigenvalues nπ , n = 1, 2, 3, .... λn = L and eigenfucntions nπx ϕn (x) = sin( ), n = 1, 2, 3... L Therefore for each n = 1, 2, 3, .... we have a solution nπ nπ nπ un (x, t) = (An cos(a t) + Bn sin(a t) sin( x) L L L We have to be able to solve for the initial condition so by the superposition principle we also have the solution u(x, t) = ∞ ∑ (An cos(a n=1 nπ nπ nπ t) + Bn sin(a t) sin( x) L L L Now the initial conditions u(x, 0) = f (x) =⇒ f (x) = ∞ ∑ An sin( n=1 nπ x) L We recognize this as the Fourier half-range sine expansion of f (x) defined on (0, L). Therefore 2 An = L ∫ L f (x) sin( 0 nπ x)dx L Differentiating with respect to t we get ut = ∞ ∑ n=1 (−An a nπ nπ nπ nπ nπ sin(a t) + Bn a cos(a t) sin( x) L L L L L 84 ut (x, 0) = g(x) =⇒ g(x) = ∞ ∑ Bn a n=1 nπ nπ sin( x) L L We recognize this also as the Fourier half-range sine expansion of g(x) defined on (0, L). Therefore ∫ nπ 2 L nπ =⇒ Bn a = g(x) sin( x)dx L L 0 L ∫ L 2 nπ =⇒ Bn = g(x) sin( x)dx anπ 0 L Note that if the string is released from rest so that g(x) = 0 then each Bn = 0. 4.Laplace’s Equation a. Basic Boundary Value Problem uxx + uyy = 0 : 0 < x < a, 0 < y < b ux (0, y) = 0, ux (a, y) = 0 : 0 < y < b u(x, 0) = 0 : 0 < x < a u(x, b) = f (x) : 0 < x < a b. Physical Interpretation (1) This represents the steady state temperature u(x, y) in a rectangular plate with the prescribed boundary conditions c. The Solution 85 (1) Attempt a separation of variables solution u = XY . We then arrive at X ′′ + λ2 X = 0 X ′ (0) = 0, X ′ (a) = 0 Y ′′ − λ2 Y = 0 Y (0) = 0 The solution for Y (y) are then Y (y) = c1 cosh(λy) + c2 sinh(λy) Y (0) = 0 =⇒ Y = c2 sinh(λy) The solution for X(x) is a Sturm-Liouville problem with eigenvalues nπ λ = 0 and λn = , n = 1, 2, 3, .... a and eigenfunctions X = c for λ = 0 and ϕn (x) = cos( nπ x), n = 1, 2, 3... a Therefore by the superposition principle we also have the solution u(x, y) = A0 y + ∞ ∑ An sinh( n=1 nπ nπ y) cos( x) a a Now the initial conditions u(x, b) = f (x) = A0 b + ∞ ∑ n=1 86 An sinh( nbπ nπ ) cos( x) a a We recognize this as the Fourier half-range cosine expansion of f (x) defined on (0, a). Therefore ∫ 2 a f (x)dx =⇒ A0 = f (x)dx ab 0 0 ∫ nbπ 2 a nπ An sinh( )= f (x) cos( x)dx a a 0 a ∫ a nπ 2 f (x) cos( x)dx =⇒ An = nbπ a a sinh( a ) 0 2 2A0 b = a ∫ a 5.D’alembert’s Solution to The Wave Equation a. There is an alternative method for solving the wave equation due to D’Alembert. A change of variables simplifies the wave equation so you can directly get the most general solution. b. Consider the wave PDE; c2 uxx = utt . Make the change of variables ζ = x + ct, η = x − ct. This transforms the orginal equation to ∂ 2u =0 ∂ζ∂η (1) This can be solved directly to give u(x, y) = F (ζ) + G(η) = F (x + ct) + G(x − ct) this is the most general solution to the wave equation. (2) If we impose initial conditons u(x, 0) = f (x) and ut (x, 0) = g(x) then we further find that ∫ 1 x+ct f (x + ct) + F (x − ct) + g(u)du u(x, t) = 2 2c x−ct 87 PARTIAL DIFFERENTIAL EQUATIONS (8) HARMONIC AND VIBRATORY MOTION 1. Definition Periodic motion or harmonic motion is any motion that repeats itself in equal intervals of time. It is vibratory or oscillatory if it moves back and forth over the same path. If the vibratory motion is slowed by friction it is called damped harmonic motion. a. Note Mechanical and electromagnetic oscillations are described by the same mathematical equations. b. Definition The period T of a harmonic motion is the time to complete one cycle. The frequency ν is the number of oscillations (cycles) per unit time =⇒ ν = 1 T The equilibrium position is the position at which no net force acts on the oscillatory particle. The displacement is the distance linear or angular from the equilibrium position (at a particular time). 2. The Simple Harmonic Oscillator a.Definition An oscillating particle moving back and forth about an equilibrium position so that the potential energy is 1 U (x) = kx2 2 is a simple harmonic oscillator and the motion is simple harmonic motion. 88 (1) Lemma The force acting on a simple harmonic oscillator is F (x) = −kx where k is the force constant. This is derived from F (x) = − dUdx(x) . It is entirely analogous to Hooke’s law for a stretched spring. (2) Definition called the amplitude. The maximum displacement is (3) Note Hooke’s law holds up to the elastic limit for many common materials. 3. Simple Harmonic Motion a.Theorem motion is The basic ODE for simple harmonic k d2 x + x = 0. dt2 m The solution to this ODE can be given by x(t) = A cos(ωt + δ) where A = the amplitude 2π ω = 2πν = T where ν = the frequency and T = the period . ωt + δ = phase of the motion and − 89 δ = phase shift . ω 4. Energy in Simple Harmonic Motion a.Theorem For simple harmonic motion [1] The potential energy at time t is given by 1 U = kA2 cos2 (ωt + δ) 2 It has a maximum value of 21 kA2 . [2] The kinetic energy at time t is given by 1 U = kA2 sin2 (ωt + δ) 2 It has a maximum value of 21 kA2 . [3] The total energy at time t is given by 1 E = K + U = kA2 2 Hence the total energy in simple harmonic motion is proportional to the square of the amplitude. (1) Corollary During one period the average kinetic energy is equal to the average potential energy and is given by 14 kA2 . (2) Corollary Speed is a maximum at the equilibrium position x = 0 and zero at the maximum displacement x = A. In general √ k 2 v=± (A − x2 ) m 90 4. The Simple Pendulum a. Consider a mass m suspended from a cord with tension T as below (2) The restoring force is given by F = −mg sin θ (a) If θ is small then sin θ u θ =⇒ F = −mg sin θ = − mg x l Hence this is simple harmonic motion with k = − mg l (b) The period is then given by √ √ m l T = 2π = 2π k g Note that the period is independent of the mass. 91 II. Oscillations A. Oscillations 1. Periodic Motion a.Definition Periodic motion or harmonic motion is any motion that repeats itself in equal intervals of time. It is vibratory or oscillatory if it moves back and forth over the same path. If the vibratory motion is slowed by firction it is called damped harmonic motion. (1) Note Mechanical and electromagnetic oscillations are described by the same mathematical equations. (2) Definition The period T of a harmonic motion is the time to complete one cycle. The frequency ν is the number of oscillations (cycles) per unit time 1 =⇒ ν = T The equilibrium position is the position at which no net force acts on the oscillatory particle. The displacement is the distance linear or angular from the equilibrium position (at a particular time). B. Harmonic Motion 1. The Simple Harmonic Oscillator a.Definition An oscillating particle moving back and forth about an equilibrium position so that the potential energy is 1 U (x) = kx2 2 92 is a simple harmonic oscillator and the motion is simple harmonic motion. (1) Lemma The force acting on a simple harmonic oscillator is F (x) = −kx where k is the force constant. This is derived from F (x) = − dUdx(x) . It is entirely analogous to Hooke’s law for a stretched spring. (2) definition The maximum displacement is called the amplitude. (3) Note Hooke’s law holds up to the elastic limit for many common materials. 2. Simple Harmonic Motion a.Theorem The basic ODE for simple harmonic motion is k d2 x + x = 0. 2 dt m The solution to this ODE can be given by x(t) = A cos(ωt + δ) where A = the amplitude 2π ω = 2πν = T where ν = the frequency and T = the period . 93 ωt + δ = phase of the motion and − δ = phase shift . ω 3. Energy in Simple Harmonic Motion a.Theorem For simple harmonic motion [1] The potential energy at time t is given by 1 U = kA2 cos2 (ωt + δ) 2 It has a maximum value of 21 kA2 . [2] The kinetic energy at time t is given by 1 U = kA2 sin2 (ωt + δ) 2 It has a maximum value of 21 kA2 . [3] The total energy at time t is given by 1 E = K + U = kA2 2 Hence the total energy in simple harmonic motion is proportional to the square of the amplitude. (1) Corollary During one period the average kinetic energy is equal to the average potential energy and is given by 14 kA2 . (1) Corollary Speed is a maximum at the equilibrium position x = 0 and zero at the maximum displacement x = A. In general √ k 2 (A − x2 ) v=± m 94 4. Some Examples of Simple Harmonic Motion a.The Simple Pendulum (1) Consider a mass m suspended from a chord with tension T as below (2) The restoring force is given by F = −mg sin θ (a) If θ is small then sin θ u θ mg =⇒ F = −mg sin θ = − x l Hence this is simple harmonic motion with k = − mg l (b) The period is then given by √ √ m l = 2π T = 2π k g 95 Note that the period is independent of the mass. b.The Torsional Pendulum (1) Consider a disc attached at its center as below. Twist the wire (2) The restoring torque is given by τ = −κθ where κ is the torsional constant of the wire. (3) If κ is the torsional constant of the wire and I is th rotational inertia of the disc then the ODE here is given by κ dθ + θ=0 dt2 I (a) This implies that the solution is θ(t) = θm cos(ωt + δ) 96 The period is then √ T = 2π I κ c.The Physical Pendulum (1) Consider any rigid body mounted so that it can swing in a vertical plane about some axis (2) The restoring torque for an angular displacement θ is given by τ = −M gd sin θ where d is the distance from the pivot point to the center of mass (3) If θ is small then sin θ u θ and then τ = −M gdθ = −κθ where κ = M gd Hence it is harmonic motion. (a) The period of a physical pendulum is then given by √ T = 2π √ I = 2π κ I M gd (b) This implies that the inertia of the body is given by T 2 M gd I= 4π 2 97 Everyhting on the right is measurable so this can be used to determine the inertia I. (4) Lemma The length of a simple pendulum whose period is equal to that of a particular physical pendulum is I l= Md (a) For determining the period of oscillation the mass of a physical pendulum may be considered to be concentrated at a point whose distance from the pivot is l = MI d . This point is called the center of oscillation and depends on the pivot point. 5. Uniform Circular Motion a.Theorem Simple harmonic motion can be described as the projection along the diameter of unifrom circular motion. Conversely uniform circular motion can be described as a combination of two simple harmonic motions occurring along perpendicular lines having the same amplitude and frequency but differeing in phase by 90o . 6. Combinations of Harmonic Motions a.Theorem All combinations of two simple harmonic motions at right angles having the same frequency correspond to elliptical paths. The shape of the ellipse A depends only on the ratio of the amplitudes Axy and the difference between the phases. 98 7. Two Body Oscillations a. Consider a two-body oscillation, that is two masses m1 , m2 connected by a massless spring with spring constant k. This is then equivalent to a one-body oscillation by redefining a reduced mass. (1) The relevant ODE’s here are m1 d2 x1 d2 x2 = −kx and m = kx2 1 2 dt2 dt2 Defining x = (x1 − x2 ) − l to be the relative displacement this becomes d2 x d2 x k µ 2 + kx = 0 =⇒ + x=0 dt dt2 µ where m1 m2 1 1 1 or = + m1 + m2 µ m1 m2 is the reduced mass. µ= (2) The solution to the two-body oscillation is then x(t) = A cos(ωt + δ) where x = (x1 − x2 ) − l and ω 2 = k µ (a) The period and frequency are then √ √ 1 k µ and ν = . T = 2π k 2π µ 99 8. Damped Harmonic Motion a. If the oscillation is damped by friction so that it slows down it is called damped harmonic motion. (1) The damping force is proportional to the velocity so that it is given by Fdamp == −b dx dt (2) The resulting ODE is then dx d2 x m 2 + b + kx = 0 dt dt (a) If the frictional force b is small then the solution is given by x(t) = Ae− m cos(ω ′ t + δ) bt √ where ω ′ = 2πν ′ = k b 2 −( ) m 2m (b) When friction is present the frequency is smaller and the period is longer. Eventually the energy will dissipate to zero. 9. Forced Oscillations and Resonance a. Consider a harmonic oscillator subject to both friction and an external periodic driving force given by Fm cos(ω ′′ t). 100 (1) The ODE is given by m dx d2 x + b + kx = Fm cos(ω ′′ t) 2 dt dt (a) The solution is given by x(t) = where Fm sin(ω ′′ t − δ) G √ G = m2 (ω ′′2 − ω 2 ) + b2 ω ′′2 and bω ′′ ) δ = cos ( G −1 [1] The system resonantes with the frequency ω of the driving force rather than the natural frequency ω. ′′ [2] There is a characteristic value for ω ′′ for which FGm is a maximum. This is called the resonant frequency and the condition is called resonance. [a] If b = 0 then as ω ′′ → ω the amplitude becomes infinite. This is the basis for the Tacoma Bridge disaster and why soldiers break step on a bridge. [b] The displacement caused by a constant driving force Fm is just Fkm . EXAMPLE (a) A 40 lb. weight stretches a spring 4 ft. initally the weight starts from rest 2 ft. below the equilibrium 101 position. What is the equation of the resulting motion? What is the amplitude and period of the resulting motion? (b) Suppose the spring of part (a) is placed in a medium with a damping force equal to 1/5 the instantaneous velocity. What is the equation of motion? (c) Suppose the spring of part (a) is placed in a medium with a damping force equal to 1/5 the instantaneous velocity and is subjected to a driving force given by f (t) = 3 sin(2t). What is the equation of motion? SOLUTION (a) The basic ODE is y ′′ + Here m= k y=0 m 40 5 W = = g 32 4 From 40 = 4k =⇒ k = 10 =⇒ k =8 m Therefore the initial value problem is y ′′ + 8y = 0; y(0) = 2, y ′ (0) = 0 The solution is then √ y = 2 cos( 8t) 102 and hence the amplitude is 2 and the period is 2π √ . 8 (b) The basic ODE is my ′′ = −ky = βy ′ =⇒ y ′′ + Here β ′ k y + y=0 m m k =8 m Further 1 β 1 =⇒ = 5 m 40 Therefore the initial value problem is β= y ′′ + 1 ′ y + 8y = 0; y(0) = 2, y ′ (0) = 0 40 The auxillary equation is m2 + 1 m+8=0 40 This has solution 1 m=− ± 20 √ 1279 i 80 Therefore the solution to the ODE is √ √ 1 1279 1279 y = e− 20 t (c1 cos( t) + c2 sin( t) 80 80 From the initial conditions we get that c1 = 2, c2 = Therefore the equation of motion is 2 2√ 2√ y = e− 9 t (2 cos( 215t) − c2 sin( 215t) 9 9 103 (c) From the previous two parts the basic initial value problem is y ′′ + 1 ′ y + 8y = 3 sin(2t); y(0) = 2, y ′ (0) = 0 40 The solution to the homogeneous part is then √ √ 1 1279 1279 t) + c2 sin( t) y = e− 20 t (c1 cos( 80 80 To find a particular solution we use the method of undetermined coefficients. We try for a solution of the form y(t) = A sin(2t) + B cos(2t) 104 PARTIAL DIFFRENTIAL EQUATIONS (8) SPECIAL FUNCTIONS A.Special Functions are particular mathematical functions that have more or less established names and notations due to their importance in mathematical analysis, functional analysis, geometry, physics, or other applications. The term is defined by consensus, and thus lacks a general formal definition, but the List of mathematical functions contains functions that are commonly accepted as special. 1. Certain special differential equations occur frequently in applied mathematics and physics. The solutions are often complicated functions whcih have been extensively studied. These functions have been called special functions. We will look at two of these. The first appears in many statistical applications and is used in the definition of the second. 2. The Gamma Function (1) The Gamma Function is defined for x > 0 by ∫ ∞ Γ(x) = tx−1 e−t dt 0 (a) Lemma Γ(x + 1) = xΓ(x) (b) Lemma For a positive integer n we have Γ(n + 1) = n! For this reason the Gamma function is often called the generalized factorial function 105 (c) Lemma Γ( 12 ) = √ π (d) By using Γ(x) = x1 Γ(x + 1) the Gamma function can be defined for all negative reals except for n = 0, −1, −2, ...... B. The Bessel Function 1. A Bessel equation is an ODE x2 y ′′ + xy ′ + (x2 − v 2 )y = 0 : v ≥ 0 a.Definition The Bessel Function of the first kind of order v is Jv (x) = ∞ ∑ i=0 (−1)n x ( )2n+v n!Γ(1 + v + n) 2 [1] Theorem Jv (x) is always one solution to the Bessel equation. [i] If 2v is not an integer then the 2 independent solutions are Jv (x) and J−v (x) [ii] If 2v is an integer then a second solutions is given by ∫ J(x) = Jv (x) 106 dx x(Jv (x))2