Review : Matrices & Vectors — Question 10

PDF ↗

Question 10

Work with real matrices and column vectors. Write InI_n for the n×nn\times n identity, ATA^T for transpose, and ∥v∥=vTv\|v\|=\sqrt{v^Tv} for Euclidean length. Show the reasoning behind every classification; do not use eigenvalue methods.

Let N=(010001000),A=I3+N.N=\begin{pmatrix}0&1&0\\0&0&1\\0&0&0\end{pmatrix},\qquad A=I_3+N. A discrete update is vm+1=Avmv_{m+1}=Av_m, starting from v0=(0,0,1)Tv_0=(0,0,1)^T. The index mm is a nonnegative integer, and each update uses the entire old vector.

Tasks

  1. Compute N2,N3N^2,N^3, det⁡A\det A and A−1A^{-1}. Verify the inverse without a general inverse algorithm.

  2. Prove a finite formula for AmA^m for all nonnegative integers mm, using induction or the binomial identity with justification.

  3. Find vmv_m and describe its coordinate growth. Explain why det⁡A=1\det A=1 does not imply bounded iterates.

  4. A program first updates y←y+zy\leftarrow y+z and then x←x+yx\leftarrow x+y, using the already updated yy. Find the matrix it actually applies and its trajectory from the same initial vector. Compare the two after two steps.

Original worksheet page 1: question and worked solution for 5-2-010
Show solutionHide solution

Question 10 – Solution

Strategy. The powers of the strictly upper triangular part terminate. A matrix update is simultaneous unless a different order is explicitly encoded.

Step 1: Exploit the terminating powers. Multiplication gives N2=(001000000)N^2=\begin{pmatrix}0&0&1\\0&0&0\\0&0&0\end{pmatrix} and N3=0N^3=0. The upper triangular matrix AA has determinant 11. Moreover A−1=I3−N+N2\boxed{A^{-1}=I_3-N+N^2} because (I3+N)(I3−N+N2)=I3+N3=I3(I_3+N)(I_3-N+N^2)=I_3+N^3=I_3; the reverse product is the same.

Step 2: Derive the finite power formula. The identity matrix commutes with NN, so the binomial expansion terminates: Am=I3+mN+m(m−1)2N2.\boxed{A^m=I_3+mN+\frac{m(m-1)}2N^2.} Alternatively, the formula is true at m=0m=0. Multiplication by I3+NI_3+N increases the coefficient of NN to m+1m+1 and that of N2N^2 to m(m−1)/2+m=m(m+1)/2m(m-1)/2+m=m(m+1)/2, establishing the induction step.

Step 3: Find the exact iterates. Applying the power formula to v0v_0 gives vm=(m(m−1)/2,m,1)T.\boxed{v_m=\bigl(m(m-1)/2,\ m,\ 1\bigr)^T.} The first coordinate grows quadratically, the second linearly and the third stays fixed. Determinant 11 preserves oriented volumes of three-dimensional sets, not individual lengths or boundedness under repetition. The plot marks integer-time values only; coincident coordinate values share a location.

Step 4: Diagnose the sequential-update error. The program computes ynew=y+zy_{\mathrm{new}}=y+z and then xnew=x+y+zx_{\mathrm{new}}=x+y+z. It actually applies B=(111011001)=I3+N+N2.B=\begin{pmatrix}1&1&1\\0&1&1\\0&0&1\end{pmatrix}=I_3+N+N^2. Starting from v0v_0, its third coordinate remains 11, its second becomes mm, and its first accumulates 1+2+⋯+m=m(m+1)/21+2+\cdots+m=m(m+1)/2. After two steps the correct vector is (1,2,1)T(1,2,1)^T, while the program gives (3,2,1)T(3,2,1)^T. Both matrices are invertible with determinant 11, so checking only the determinant would not expose this order-of-update mistake.

See the diagram in the original worksheet below.

Original worksheet page 2: question and worked solution for 5-2-010

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