150 2. BOUNDARY-VALUE PROBLEMS IN RECTANGULAR COORDINATES To solve the equation for n = N, we use the following observation. For any Wi =1= W, (;; + w.) cosw't = [w' - (w')'] cosw't while (:. + w' ) Therefore Taking the limit Wi cos wt = 0 f12 2) COS wit - cos wt w2 _ (W/)2 ( dt 2 + W ~ w, we have I = cosw t . cos wit - coswt t . 11m = smwt 2 w'-+w w - (W')2 2w Thus we have the solution 1 - (-l)N t . . N1rs 2 ' " sin (n1rs/ L) 1 - (-l)n y(Sjt) = A N -2 smwtsm- +-Acoswt L..J ( /L)2 2 L W 7r n:l;N n7rC - w n This solution is not a periodic function of time, but oscillates with increasing amplitude as time progresses. This is the phenomenon of resonance.• EXERCISES 2.4 1. Consider the initial-value problem for the symmetric plucked string. Ytt = y(O; t) 2 C YS8 = 0 = y(L; t) y(s; 0) = s y(s; 0) = L - S t > 0, 0 < S < L t>0 0 < s 5: L/2 L/2 < S < L 0<s<L Make a graphical representation of the solution for ct = L/4, L/2, 3L/4, L. At what time is y(s; t) = 0 for all 0 < S < L? 2. Let y(s; t) = 2:~1 Bn cos (n1rct/ L) sin(n1rs/ L) be a solution of the vibrating string problem. Suppose that the string is further constrained at its midpoint, so that y(L/2, t) = 0 for all t. What condition does this impose on the coefficients Bn? 3. Let y(s, t) = 2::=1 Bn cos (n7rct/ L) sin(n1rs/ L) be a solution of the vibrating string problem. Suppose that the string is constrained so that y(L/3, t) = 0 for all t. What condition does this impose on the coefficients Bn? 4. The energy of a vibrating string of tension To and density p = m/ L is defined by L 1 2 2 E = 2" (pyt + Toys)ds Yt(s; 0) = 0 Jor 2.4. THE VIBRATING STRING 151 Let 00 - coswnt + En - sinwnt) sin L nn8 Y(8; t) = "" ~(An n=l be a solution of the wave equation with Wn = nnc/ L. Use Parseval's theorem to write E as an infinite series involving An, En and verify the law of conservation of energy. 5. Let y(s; t) be a solution of the wave equation Ytt = c?Yss satisfying the boundary conditions y(O; t) = 0 = y(L; t). By differentiating under the integral sign, show directly that dE / dt = 0, where E is the energy defined in Exercise 4. 6. Consider the initial-value problem for the plucked string of Example 2.4.2. Compute the total energy corresponding to normal modes n ;f= 1 and show that this is less than half of the total energy. 7. Suppose that 12(8), -00 < s < 00, is an odd, 2L-periodic function. Define y(s; t) = (1/2c) fs8~~t 12(z) dz; show that y(O; t) = 0, y(Lj t) = 0 for all t. 8. Let I(s), -L < 8 < L, be a piecewise smooth function. Extend 1 to a 2L-periodic function defined for -00 < S < 00. Show that the resulting function is piecewise smooth on every interval a < s < b. 9. Let 1(8), -L :::; s :::; L, be an odd function. Extend 1 to a 2L-periodic function defined for -00 < 8 < 00. Show that the resulting function is again odd. 10. Let 1(8), -00 < 8 < 00, be an odd function with the property that I(L8) = 1(8). Show that 1(8 + L/2) + 1(8 - L/2) = 0 for all 8. 11. Let I(s), 0 :::; 8 < L, satisfy 1(8) = I(L - s). Let y be the solution (2.4.12) satisfying y(s; 0) = 1(8), Yt(S; 0) = O. Show that y(8, L/2e) = 0 for 0 < 8 < L. 12. Show that the Fourier series solution obtained in Example 2.4.2 converges uniformly for 0 ~ s :::; L. 13. Consider the following initial-value problem for the wave equation Ytt = c?Y88 for t > 0, 0 < s < L: y(Oj t) = y(Lj t) = 0 for t > 0; Y(8j 0) = 0 and Yt(8j 0) = 1 for 0 < 8 < L. (a) Find the Fourier representation of the solution. (b) Find the explicit representation of the solution and graph the solution for t = 0, L/4e, L/2e, 3L/4e, L/e. Exercises 14 to 16 are designed to review the techniques from calculus that are used in establishing d'Alembert's formula. Recall that the fundamental theorem of calculus states that (d/dx) I(z)dz = I(x) for any continuous function I. The chain rule for differentiating composite functions states that (d/dx)F(G(x» = F'(G(X»G'(x). 14. Let 1 be a continuous function and set F(x) = foX J(z) dz. Show that f:~~t J(z) dz = F(x + et) - F(x - et). f; 152 2. BOUNDARY-VALUE PROBLEMS IN RECTANGULAR COORDINATES 15. Use the chain rule and the fundamental theorem of calculus to show that (d/dx) f:~~t J(z) dz = J(x + ct) - J(x - ct). 16. Use the chain rule and the fundamental theorem of calculus to show that (d/dt) f:~~t J(z) dz = cJ(x + ct) + cJ(x - ct). 17. The voltage u(x; t) in a transmission cable is known to satisfy the partial differential equation Utt + 2aut + a2 u = c2uxx , where a and c are positive constants. Let y(x; t) = eatu(x; t) and show that y satisfies the wave equation Ytt = c'lyxx' 18. Use Exercise 17 and d'Alembert's formula to solve the initial-value problem utt+2aUt+a2u = r?u xx for t > 0, -00 < x < 00, with the initial conditions u(x;O) = gl(X), Ut(x;O) = O. 19. Use Exercise 17 and d' Alembert's formula to solve the initial-value problem 2 Utt + 2aut + a u = r?uxx for t > 0, -00 < x < 00, with the initial condtions u(x; 0) = 0, Ut(x; 0) = g2(X). 20. A vibrating string with friction in a periodic force field is described by the equation Ytt + 2aYt - c2 yss = Acoswt and boundary conditions y(O; t) = = y(L; t), where A, a, and ware positive constants. Find a particular solution that is also periodic in time. ° 2.5. Applications of Multiple Fourier Series In this section we consider boundary-value problems in rectangular coordinates (x, y, z) where more than one of these variables appears in the solution. This is in contrast to the previous sections, where the solution depended on z alone. We will solve initial-value problems for the heat equation, boundary-value problems for Laplace's equation, and the wave equation for a vibrating membrane. The key idea in our work is a double Fourier series. To illustrate this, we display a double Fourier sine series, 00 '" A . m1rX . n1rY L...J mnslllL slllL m ,n=1 1 2 which may be used in problems involving a rectangle or column, described by the inequalities 0 < x < L 1 , 0 < Y < L 2 • Similarly, a double Fourier cosine series is of the form 2: Amn cos (m1rx/ L 1 ) cos{n1rY/ L2). Clearly we could consider other combinations of these, where we mix sines and cosines, for example. All double series of this type are of the form L:m.n l/Jm{x)'l/ln(Y), where l/Jm, 'l/ln are the eigenfunctions of a Sturm-Liouville eigenvalue problem. Accordingly, the corresponding functions of two variables obey suitable orthogonality relations.