MATH 3150: PDE FOR ENGINEERS MIDTERM TEST #2 1. Consider d’Alembert’s solution u(x, t) = 1 1 (f (x − ct) + f (x + ct)) + 2 2c Z x+ct g(s) ds x−ct of the wave equation for a vibrating string, where f (x) is the odd function with period 2L which on the interval 0 ≤ x ≤ L gives the initial position of the string, and similarly g(x) gives the initial velocity on 0 ≤ x ≤ L and is odd and 2L periodic. Suppose that f (x) = sin πx g(x) = 0 with L = 1 and c = 1. What is the first positive time t that the string returns to its original shape? Solution: 1 (sin π(x − t) + sin π(x + t)) 2 and the hump of the first sine wave sits at u(x, t) = π(x − t) = π/2 which is x = t + 1/2 while the other sine wave has hump sitting at x = −t + 1/2 so that it goes back to its original shape when the two humps both align again, at x = 1/2 up to multiples of 2 (the period). This first happens at time t = 2. 2. Use separation of variables to find the general solution of the heat equation in a rectangular plate with edges of lengths a and b, and with all edges insulated. Solution: The equations to solve are 2 ∂u ∂ u ∂2u = c2 + ∂t ∂x2 ∂y 2 ∂u = 0 at x = 0 & x = a ∂x ∂u = 0 at y = 0 & y = b ∂y Date: November 12, 2000. 1 2 MATH 3150: PDE FOR ENGINEERS MIDTERM TEST #2 The last two equations use Fourier’s law to ensure that there is no heat flux through the boundary. Separating variables: u(x, y, t) = E(x)F (y)G(t) (we just split u into a product of functions, one of each variable) we find the heat equation becomes EF G0 = c2 (E 00 F G + EF 00 G) which, once we divide by EF G, is just 00 G0 E F 00 2 =c + . G E F The left side involves only t, and the right side only x and y. So there is a constant γ so that 00 G0 E F 00 2 = −γ = c + . G E F Therefore G(t) = e−γt while E 00 F 00 γ + − 2 = c E F or E 00 γ F 00 − = 2+ . E c F The left side depends only on x, and the right only on y, so that there is some constant β so that E 00 γ F 00 − =β= 2 + E c F or E 00 − =β E γ F 00 − = 2 −β F c We see that E(x) satisfies a harmonic oscillator type equation, and so does F (y). Our conditions on the edges force E 0 (x) = 0 for x = 0 and x = a and similarly F 0 (y) = 0 for y = 0 and y = b . Since the derivatives E 0 and F 0 have to have two zeros, our analysis of harmonic oscillator equations says that β and γ −β c2 must both be positive. In particular, γ must be positive, so therefore G(t) is an exponential decay, and up to rescaling p E(x) = cos βx r γ − βy . F (y) = cos c2 MATH 3150: PDE FOR ENGINEERS MIDTERM TEST #2 3 But the vanishing of the derivatives at x = a and at y = b gives p πm β= a r γ πn −β = c2 b for some integers m and n. In terms of m and n, everything looks like πmx E(x) = cos a πny F (y) = cos b G(t) = e−γmn t where γmn = c2 π 2 m2 n2 + 2 2 a b So πmx πmy m2 n2 + t cos cos u(x, y, t) = exp −c π a2 b2 a b and the general solution of the heat equation with insulated boundary is 2 ∞ X ∞ πmx πmy X m n2 u(x, y, t) = Amn exp −c2 π 2 + t cos cos 2 2 a b a b m=0 n=0 3. 2 2 Under rescaling a region by a factor of ε in the x and y variables, by rescaling x → εx y → εy the operators ∂ ∂ and ∂x ∂y get rescaled by ∂ 1 ∂ → ∂x ε ∂x ∂ 1 ∂ → ∂y ε ∂y Calculate the rescaling that must occur to the time variable t in the heat equation 2 ∂u ∂ u ∂2u 2 =c + 2 ∂t ∂x2 ∂y in order that the constant c stay the same, and the heat equation still be satisfied by the same function with the rescaled arguments. Recalling the basic idea that a mode consists of an exponential decay in time multiplied by a function of x and y, say exp(−κt)f (x, y) 4 MATH 3150: PDE FOR ENGINEERS MIDTERM TEST #2 how does this rescaling affect κ, the rate of decay of the mode? Given a replica of a plate of 1/20 the scale of the original, made of the same material, how many times faster will it cool or heat than the original? Solution: The heat equation tells us that the time variable has to be scaled like t → ε2 t so that the κ constant will be scaled like κ . ε2 Therefore the rate of cooling is faster by 1/ε2 ; in particular a κ→ ε = 1/20 scale replica cools at 1 1/20 2 = 400 times the rate of the original. 4. Show that the function un (r, θ) = r n R (an cos(nθ) + bn sin(nθ)) satisfies the Laplace equation ∇2 u = 0 on the disk of radius R, where in polar coordinates ∇2 u = ∂ 2 u 1 ∂u 1 ∂2u + + 2 2 . 2 ∂r r ∂r r ∂θ Now why does the function u(r, θ) = a0 + ∞ n X r n=1 R (an cos (nθ) + bn sin (nθ)) satisfy the Laplace equation? How should we pick the a0 and an and bn so that u(R, θ) = f (θ) is some specified temperature function f (θ) around the edge? Solution: The first derivatives are ∂ rn−1 un = n n (an cos nθ + bn sin nθ) ∂r R n = un r r n ∂ un = n (−an sin nθ + bn cos nθ) ∂θ R The second derivatives are ∂2 n(n − 1) un = un 2 ∂r r2 ∂2 un = −n2 un ∂θ2 MATH 3150: PDE FOR ENGINEERS MIDTERM TEST #2 5 1 0.8 0.6 0.4 0.2 –1 –0.5 y0 0.5 1 1 0.5 0 x –0.5 –1 Figure 1. A round peak Putting these together, we find 2 1 ∂ 1 ∂2 n(n − 1) n n2 ∂ + + u = + − un = 0 n ∂r2 r ∂r r2 ∂θ2 r2 r2 r2 Adding up the un we find, by linearity of ∇2 , that u satisfies ∇2 u = 0 . Moreover if we pick these a0 , an and bn according to Z 2π 1 a0 = f (θ) dθ 2π 0 Z 2π 1 an = f (θ) cos(nθ) dθ π 0 Z 1 2π bn = f (θ) sin(nθ) dθ π 0 (as Fourier amplitudes of f (θ)) then u(R, θ) = f (θ) (plug in r = R) so that this is the steady state of the heat equation in the disk with boundary held fixed at temperature u(R, θ) = f (θ). 5. Looking at the wave equation 2 ∂2u ∂ u ∂2u 2 =c + 2 ∂t2 ∂x2 ∂y and recalling second derivatives from calculus, explain why a round peak in the graph of u (as in figure 1) will be accelerated downward. Solution: The second derivatives in x and in y will be negative, so the right hand side of the wave equation is negative. Therefore the left hand side, the acceleration, must be negative too.