Systems of Differential Equations — Question 8

Section 26-6 · Calculus pages 4695–4696

Show solutionHide solution

Question 8

The functions 1,t,et1,t,e^t form a scalar basis for y‴−y″=0y'''-y''=0. For the derivative state X=(y,y′,y″)TX=(y,y',y'')^T, let A=(010001001).A=\begin{pmatrix}0&1&0\\0&0&1\\0&0&1\end{pmatrix}. A fundamental matrix need not equal the identity at the chosen base time.

Tasks

  1. Form the fundamental matrix F(t)F(t) from the given scalar basis and its derivatives. Check F′=AFF'=AF and compute its determinant.

  2. Derive the transition matrix Φ(t,s)=F(t)F(s)−1\Phi(t,s)=F(t)F(s)^{-1} explicitly. Verify its normalization, composition law and determinant.

  3. Use this transition matrix to solve y‴−y″=1y'''-y''=1 with (y,y′,y″)(1)=(0,0,0)(y,y',y'')(1)=(0,0,0) for every real tt. Treat the integral orientation when t<1t<1.

  4. Explain why using F(t)F(t) directly as the transition from time zero gives wrong initial data in general. Prove that replacing FF by FCFC, for any constant invertible CC, leaves the true transition matrix unchanged.

Question 8 – Solution

Strategy. Normalize a scalar-derived fundamental matrix at the actual base time before transporting states or forcing.

Step 1: Lift the scalar basis. The derivative columns give F(t)=(1tet01et00et),det⁡F=et≠0.F(t)=\begin{pmatrix}1&t&e^t\\0&1&e^t\\0&0&e^t\end{pmatrix},\qquad \det F=e^t\ne 0. Differentiating the rows gives F′=AFF'=AF, including the last row because (et)′=et(e^t)'=e^t. Thus FF is fundamental on the entire real line.

Step 2: Normalize and compose. With r=t−sr=t-s, direct multiplication gives Φ(t,s)=(1rer−1−r01er−100er).\boxed{\Phi(t,s)=\begin{pmatrix}1&r&e^r-1-r\\0&1&e^r-1\\0&0&e^r\end{pmatrix}.} It equals II at t=st=s, and det⁡Φ=et−s\det\Phi=e^{t-s}. The identity F(t)F(s)−1F(s)F(a)−1=F(t)F(a)−1F(t)F(s)^{-1}F(s)F(a)^{-1}=F(t)F(a)^{-1} proves Φ(t,s)Φ(s,a)=Φ(t,a)\Phi(t,s)\Phi(s,a)=\Phi(t,a) for all real base times.

Step 3: Transport the forcing in both directions. The zero initial vector at time one gives X(t)=∫1tΦ(t,v)e3dv,y(t)=et−1−1−(t−1)−12(t−1)2.X(t)=\int_1^t\Phi(t,v)e_3\,dv,\qquad \boxed{y(t)=e^{t-1}-1-(t-1)-\tfrac 12(t-1)^2.} Indeed its first component integrates et−v−1−(t−v)e^{t-v}-1-(t-v). For t<1t<1 the integral is oriented: it is minus the integral from tt to 11. Writing r=t−1r=t-1, we obtain y′=er−1−ry'=e^r-1-r, y″=er−1y''=e^r-1, y‴=ery'''=e^r. These give the zero state at t=1t=1 and residual y‴−y″=1y'''-y''=1 on all of ℝ\mathbb R.

Step 4: Separate basis from transition. Here F(0)≠IF(0)\ne I. For example, F(t)e3=(et,et,et)TF(t)e_3=(e^t,e^t,e^t)^T starts at (1,1,1)T(1,1,1)^T, not at e3e_3. The correct expression is F(t)F(0)−1e3F(t)F(0)^{-1}e_3. For any constant invertible CC, (F(t)C)(F(s)C)−1=F(t)CC−1F(s)−1=Φ(t,s).(F(t)C)(F(s)C)^{-1}=F(t)CC^{-1}F(s)^{-1}=\Phi(t,s). A basis change alters coordinates of solution columns, but not the transport of a specified physical state.