Separation of Variables — Question 10

PDF ↗

Question 10

For ut=uxx+ruu_t=u_{xx}+ru on 0<x<π0<x<\pi, t≥0t\ge 0, impose homogeneous Dirichlet endpoints. The separated spatial modes are sin⁡(nx)\sin(nx) with eigenvalues n2n^2, n=1,2,…n=1,2,\ldots. Consider 1<r<41<r<4 and the finite-mode initial field u(x,0)=sin⁡(2x)+εsin⁡x,0<ε<1.u(x,0)=\sin(2x)+\varepsilon\sin x,\qquad 0<\varepsilon<1. Only this finite-mode construction is requested; no arbitrary-data expansion or infinite-series convergence theorem is needed.

Tasks

  1. Derive the time equation for each mode and construct a solution with the given initial field. Verify its PDE and endpoints.

  2. Compare the behavior when ε=0\varepsilon=0 with every fixed ε>0\varepsilon>0. Prove that in the latter case e−(r−1)tue^{-(r-1)t}u converges uniformly to εsin⁡x\varepsilon\sin x.

  3. Using ∥v∥22=∫0πv2dx\|v\|_2^2=\int_0^\pi v^2\,dx, compute the squared norm of each modal component and determine the time when the growing component overtakes the decaying component in this norm.

  4. For r=2r=2, ε=1/8\varepsilon=1/8, sketch the two positive modal amplitudes and mark their crossing. Explain why finding a decaying separated solution alone does not show that all solutions decay, and state the first-mode threshold in rr.

Original worksheet page 1: question and worked solution for 9-4-010
Show solutionHide solution

Question 10 – Solution

Strategy. Separation reveals individual growth rates; superposition shows how a small unstable component can dominate a decaying one.

Step 1: Construct and verify the finite sum. For Xn=sin⁡nxX_n=\sin nx, Xn″=−n2XnX_n''=-n^2X_n, so the factor equation is Tn′=(r−n2)TnT_n'=(r-n^2)T_n. Consequently u=e(r−4)tsin⁡2x+εe(r−1)tsin⁡x.\boxed{u=e^{(r-4)t}\sin 2x+\varepsilon e^{(r-1)t}\sin x.} Each time derivative has coefficient r−n2r-n^2, matching uxx+ruu_{xx}+ru. Both sine factors vanish at 0,π0,\pi, and their initial coefficients are 1,ε1,\varepsilon. Linearity justifies adding these two verified modes.

Step 2: Compare exact cancellation with small contamination. For ε=0\varepsilon=0, the solution decays uniformly because r−4<0r-4<0. For fixed ε>0\varepsilon>0, the first mode grows because r−1>0r-1>0. Moreover e−(r−1)tu=εsin⁡x+e−3tsin⁡2x,∥e−(r−1)tu−εsin⁡x∥∞=e−3t.e^{-(r-1)t}u=\varepsilon\sin x+e^{-3t}\sin 2x, \qquad\|e^{-(r-1)t}u-\varepsilon\sin x\|_\infty=e^{-3t}. This proves uniform convergence of the normalized shape. The unnormalized solution grows without bound in sup norm, as is also seen at x=π/2x=\pi/2.

Step 3: Compare modal norms exactly. Since ∫0πsin⁡2(nx)dx=π/2\int_0^\pi\sin^2(nx)\,dx=\pi/2, the squared component norms are N22=π2e2(r−4)t,N12=π2ε2e2(r−1)t.N_2^2=\frac\pi 2 e^{2(r-4)t},\qquad N_1^2=\frac\pi 2\varepsilon^2e^{2(r-1)t}. Their equality is ε2e6t=1\varepsilon^2e^{6t}=1, giving tc=13log⁡(1/ε)\boxed{t_c=\tfrac 13\log(1/\varepsilon)}. The growing component is larger for t>tct>t_c. The crossing time is independent of rr in the stated range because the difference of modal rates is always three. Orthogonality also gives ∥u∥22=N12+N22\|u\|_2^2=N_1^2+N_2^2.

Step 4: Interpret the crossing and the threshold. For r=2r=2, ε=1/8\varepsilon=1/8, the amplitudes are a1=et/8a_1=e^t/8 and a2=e−2ta_2=e^{-2t}. They cross at tc=log⁡2t_c=\log 2, with common value 1/41/4. A decaying mode only describes its own initial shape. A nonzero first-mode component decays for r<1r<1, is stationary for r=1r=1, and grows for r>1r>1. Thus one decaying example cannot establish decay for every admissible field.

See the diagram in the original worksheet below.

Original worksheet page 2: question and worked solution for 9-4-010

Original worksheet layout. Use Enlarge or open the PDF for a closer view.