1a. We need to solve the characteristic equations x′ (s) = x, y ′ (s) = 1, u′ (s) = u subject to the initial conditions x(0) = r, y(0) = 0, u(0) = f (r). Solving this system of equations, one easily finds that x = x(0)es = res , u = u(0)es = f (r)es . y = s, Once we now eliminate the auxiliary variables r and s, we may conclude that u = f (r)es = f (xe−s )es = f (xe−y )ey . 1b. An initial condition of the form u(r, 0) = f (r) is not suitable because it gives 1 1 1 1 =0 = det det 0 0 0 y2 along the initial curve. Let us then impose the condition u(0, r) = f (r), where the function f is arbitrary. The associated characteristic equations are x′ (s) = 1, y ′ (s) = y 2 , u′ (s) = u and we need to solve these equations subject to the initial conditions x(0) = 0, y(0) = r, u(0) = f (r). Solving this system of equations, one easily finds that x = s, y= r , 1 − rs u = u(0)es = f (r)es . In particular, r = y(1 − rs) = y − rxy and this implies that y s r(1 + xy) = y =⇒ u = f (r)e = f ex . 1 + xy Since u(x, 0) = f (0)ex and the function f is arbitrary, there exists an infinite number of solutions such that u(x, 0) = ex . 1c. This is a fully nonlinear PDE with F = p2 + q − y and characteristic equations x′ = 2p, u′ = 2p2 + q, y ′ = 1, p′ = 0, q ′ = 1. Three of these equations can be integrated directly to give y = y0 + s, p = p0 , q = q0 + s and this allows us to determine the remaining two variables s2 + u0 . 2 u = (2p20 + q0 )s + x = 2p0 s + x0 , We note that y0 = 0 and u0 = x0 by assumption. When it comes to p0 , we have u(x, 0) = x =⇒ ux (x, 0) = 1 =⇒ p0 = 1, so we can use the given PDE to also find q0 = y0 − p20 = −1. In particular, the solution that we obtained above can be expressed in the form x = 2s + x0 , y = s, u=s+ s2 + x0 , 2 p = 1, q = s − 1. Once we now eliminate x0 and s, we may finally conclude that s + x0 = x − s = x − y =⇒ u= y2 s2 +x−y = + x − y. 2 2 2a. This problem is a special case of the first part of the extended maximum principle. Were the set B = {(x, y) ∈ A : u(x, y) > 0} nonempty, we would have uxx + uyy ≥ −cu ≥ 0 in B =⇒ max u = max u ≤ 0 B∪∂B ∂B by the standard maximum principle. This implies that u ≤ 0 within B, contrary to the definition of B. In particular, the set B is empty and so u ≤ 0 within A. 2b. Consider the difference w = u − v. This is easily seen to satisfy the equation wxx + wyy = uxx + uyy − vxx − vyy = u3 − v 3 = (u − v)(u2 + uv + v 2 ) = −cw, where c = −(u2 + uv + v 2 ). To see that c ≤ 0 at all points, we note that v 2 3v 2 2 2 −c = u + uv + v = u + ≥ 0. + 2 4 We may thus apply the result of part (a). Since w = u − v ≤ 0 on the boundary of A, we must have w ≤ 0 at all points and this means that u ≤ v at all points. 2c. Consider the function u(x, y) = (sin x)(sin y). It is easily seen to satisfy uxx + uyy = −(sin x)(sin y) − (sin x)(sin y) = −2u, while u = 0 on the boundary of the set A = (0, 2π) × (0, 2π). Since u attains positive values within A, the result of part (a) does not hold in the case that c = 2. 3a. Since y(x) is an extremal of J(y), the directional derivative Z b ′ Ly ϕ + Ly′ ϕ′ dx J (y)ϕ = a must be zero for all functions ϕ. Integrating by parts, we conclude that h ib Z b d ′ J (y)ϕ = Ly′ ϕ + Ly − Ly′ ϕ dx = 0 dx a a for all functions ϕ. When it comes to test functions ϕ, this gives Z b d d Ly ′ Ly − Ly′ ϕ dx = 0 =⇒ Ly = dx dx a by the fundamental lemma of variational calculus. In particular, we must have h ib ′ J (y)ϕ = Ly′ ϕ = 0 a for any function ϕ, so it easily follows that Ly′ = 0 when x = a, b. 3b. First of all, we need to solve the Euler-Lagrange equation d Ly ′ = Ly dx =⇒ 2y ′′ (x) + 4y ′ (x) = 4y ′ (x) − 2y(x) + 6 =⇒ y ′′ (x) + y(x) = 3. This is a second-order linear equation with constant coefficients whose solution is y(x) = c1 sin x + c2 cos x + 3. To ensure that the natural boundary conditions hold, we need to ensure that Ly′ = 2y ′ (x) + 4y(x) = (4c1 − 2c2 ) sin x + (2c1 + 4c2 ) cos x + 12 vanishes when x = 0 and also when x = π/3. In other words, we need to have √ √ 2c1 + 4c2 = −12 = (1 + 2 3)c1 + (2 − 3)c2 . Solving this system for c1 and c2 , we conclude that the only critical point is √ √ 4 3+6 2 3 − 12 y(x) = − sin x + cos x + 3. 5 5 3c. We check the sufficient condition for a local minimum. First, we note that Z π/3 Lyy ϕ2 + 2Lyy′ ϕϕ′ + Ly′ y′ (ϕ′ )2 dx 0 Z Z π/3 Z π/3 ′ 2 ϕ(x)ϕ (x) dx + 2 ϕ(x) dx + 8 = −2 ′′ J (y)ϕ = 0 0 π/3 ϕ′ (x)2 dx. 0 When it comes to the first integral, we can use the Poincaré inequality to get Z π/3 (π/3)2 ϕ(x) dx ≤ 2 2 0 Z π/3 π2 ϕ (x) dx = 18 2 ′ 0 Z π/3 ϕ′ (x)2 dx. 0 In particular, the second variation of J(y) satisfies the inequality ′′ J (y)ϕ ≥ π2 2− 9 2 Z π/3 0 π/3 ′ 2 ϕ (x) dx + 8 Z π/3 ϕ(x)ϕ′ (x) dx 0 iπ/3 18 − π ′ 2 2 = ϕ (x) dx + 4ϕ(x) 9 0 0 2 Z π/3 18 − π ϕ′ (x)2 dx = 9 0 Z h for any test function ϕ. This proves the sufficient condition for a local minimum.