Systems of Differential Equations — Question 10

PDF ↗

Question 10

For the stable scalar equation (D+1)3y=0(D+1)^3y=0, introduce scaled state coordinates q1=y,q2=(y′+y)/4,q3=(y″+2y′+y)/16.q_1=y,\qquad q_2=(y'+y)/4,\qquad q_3=(y''+2y'+y)/16. Let ∥⋅∥2\|\cdot\|_2 be the Euclidean norm and consider t≥0t\geq 0.

Tasks

  1. Derive the constant system for qq and the inverse map to (y,y′,y″)(y,y',y''). Explain why all characteristic roots are still −1-1.

  2. Solve the system for q(0)=(0,0,1)Tq(0)=(0,0,1)^T. Show explicitly that its Euclidean norm can exceed its initial value even though every solution tends to zero.

  3. Set w=(q1,4q2,16q3)Tw=(q_1,4q_2,16q_3)^T. Prove a quadratic norm estimate that is strictly decreasing for nonzero states, using 2w2(w1+w3)≤2∥w∥22.2w_2(w_1+w_3)\leq\sqrt 2\|w\|_2^2. Obtain an exponential bound for ∥q(t)∥2\|q(t)\|_2 valid for every initial state.

  4. Explain why transient growth is consistent with asymptotic stability. For the prescribed initial state, sketch ∥q(t)∥2\|q(t)\|_2 and ∥w(t)∥2/16\|w(t)\|_2/16, labeling the normalization, and distinguish the two quantities.

Original worksheet page 1: question and worked solution for 7-6-010
Show solutionHide solution

Question 10 – Solution

Strategy. Compare Euclidean growth in scaled coordinates with decay in a suitable weighted norm.

Step 1: Derive the scaled cascade. With NN having ones just above the diagonal, q′=(−I+4N)q,y=q1,y′=4q2−q1,y″=16q3−8q2+q1.\boxed{q'=(-I+4N)q,\qquad y=q_1,\quad y'=4q_2-q_1,\quad y''=16q_3-8q_2+q_1.} The coordinate map is invertible, so this is a similarity transformation of the scalar companion system. Its triangular matrix has three eigenvalues −1-1.

Step 2: Compute an actual transient. Since N3=0N^3=0, e(−I+4N)t=e−t(I+4tN+8t2N2)e^{(-I+4N)t}=e^{-t}(I+4tN+8t^2N^2). Therefore q=e−t(8t2,4t,1)T,∥q(t)∥2=e−t64t4+16t2+1.\boxed{q=e^{-t}(8t^2,4t,1)^T,\qquad \|q(t)\|_2=e^{-t}\sqrt{64t^4+16t^2+1}.} The initial norm is one, but ∥q(1)∥2=9/e>1\|q(1)\|_2=9/e>1. Every state still tends to zero because its components are polynomials times e−te^{-t}.

Step 3: Prove a uniform weighted estimate. The new state satisfies w′=(−I+N)ww'=(-I+N)w, so ddt∥w∥22=−2∥w∥22+2w2(w1+w3)≤−(2−2)∥w∥22.\frac{d}{dt}\|w\|_2^2=-2\|w\|_2^2+2w_2(w_1+w_3) \leq-(2-\sqrt 2)\|w\|_2^2. The stated inequality follows from Cauchy–Schwarz and 22|w2|w12+w32≤2∥w∥222\sqrt 2|w_2|\sqrt{w_1^2+w_3^2}\leq\sqrt 2\|w\|_2^2. With α=1−1/2>0\alpha=1-1/\sqrt 2>0, integration and ∥q∥2≤∥w∥2≤16∥q∥2\|q\|_2\leq\|w\|_2\leq 16\|q\|_2 give ∥q(t)∥2≤16e−αt∥q(0)∥2.\boxed{\|q(t)\|_2\leq 16e^{-\alpha t}\|q(0)\|_2.}

Step 4: Interpret coordinate-dependent growth. Asymptotic stability allows a finite transient; it does not require every norm to decrease at every instant. For this trajectory, ∥w∥2/16=e−tt4/4+t2+1\|w\|_2/16=e^{-t}\sqrt{t^4/4+t^2+1} decreases strictly, whereas ∥q∥2\|q\|_2 has a pronounced transient. They measure the same state with different weights.

See the diagram in the original worksheet below.

Original worksheet page 2: question and worked solution for 7-6-010

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