MATH 240

Fri. November 1st, 2019


Change of Basis Review

Let B={b1,...,bn}B=\{b_1,...,b_n\} and C={c1,...cn}C=\{c_1,...c_n\} be bases of a vector space VV. Then there is a unique n×nn\times n matrix PC←BP_{C\leftarrow B} such that [x]C=PC←B[x]B[x]_C=P_{C\leftarrow B}[x]_B.

The columns of PC←BP_{C\leftarrow B} are the CC-coordinate vectors of the vectors in the basis BB, i.e. PC←B=[[b1]C,...,[bn]C]P_{C\leftarrow B}=\begin{bmatrix} [b_1]_C,...,[b_n]_C \end{bmatrix}.


Ex. Let b1=(−91), b2=(−51), c1=(1−4), c2=(3−5)b_1=\begin{pmatrix} -9\\1 \end{pmatrix},\ b_2=\begin{pmatrix} -5\\1 \end{pmatrix},\ c_1=\begin{pmatrix} 1\\-4 \end{pmatrix},\ c_2=\begin{pmatrix} 3\\-5 \end{pmatrix}.

B={b1,b2}, C={c1,c2}B=\{b_1,b_2\},\ C=\{c_1,c_2\} form bases of R2\R^2. Find the change of basis matrix PC←BP_{C\leftarrow B}.

[c1c2][b1]C=b1[c1c2][b2]C=b2\begin{aligned} \begin{bmatrix} c_1&c_2 \end{bmatrix}[b_1]_C&=b_1\\ \begin{bmatrix} c_1&c_2 \end{bmatrix}[b_2]_C&=b_2\\ \end{aligned} The operations to reduce [c1c2]\begin{bmatrix} c_1&c_2 \end{bmatrix} to the identity will be the same for both equations. Therefore to simplify computations we can place [b1b2]\begin{bmatrix} b_1&b_2 \end{bmatrix} to the right of it, reduce it to the identity, and then whatever ends up on the right will be PC←BP_{C\leftarrow B} (similar to the technique we used before to find the inverse of a matrix).


Eigenvalues and Eigenvectors

Given an n×nn\times n matrix AA, what points will remained fixed under the transformation x→Axx\rightarrow Ax?

This question is significant because the result of applying the transformation to that point any number of times will be easily computable.

Definition. Let AA be an n×nn\times n matrix. A scalar λ\lambda is called an eigenvalue of AA if there is a non-trivial (non-zero) solution xx of Ax=λxAx=\lambda x. Such a vector xx is called an eigenvector of λ\lambda.


Ex.
If xx is an eigenvector of AA with an eigenvalue of 22, then Ax=2xAx=2x.
A2(x)=A(Ax)=A(2x)=2A(x)=4xA^2(x)=A(Ax)=A(2x)=2A(x)=4x.
A3(x)=A(A(Ax))=A(4x)=4A(x)=8x.A^3(x)=A(A(Ax))=A(4x)=4A(x)=8x.
An(x)=2nxA^n(x)=2^nx.
This point moves away from the origin in a straight line under repeated multiplication with AA.

If xx is an eigenvector of AA with an eigenvalue of 12\frac{1}{2}, then Ax=12xAx=\frac{1}{2}x.
A2(x)=A(Ax)=A(12x)=12A(x)=14xA^2(x)=A(Ax)=A(\frac{1}{2}x)=\frac{1}{2}A(x)=\frac{1}{4}x.
A3(x)=A(A(Ax))=A(14x)=14A(x)=18x.A^3(x)=A(A(Ax))=A(\frac{1}{4}x)=\frac{1}{4}A(x)=\frac{1}{8}x.
An(x)=2−nxA^n(x)=2^{-n}x.
This point moves towards the origin in a straight line under repeated multiplication with AA.


Ex.
Let A=(1652)A=\begin{pmatrix} 1&6\\5&2 \end{pmatrix}, u=(6−5)u=\begin{pmatrix} 6\\-5 \end{pmatrix}, and v=(3−2)v=\begin{pmatrix} 3\\-2 \end{pmatrix}. Are either uu or vv eigenvectors of AA?

Au=(−2420)=−4(6−5)=−4u=λuAv=(−911)≠λv\begin{aligned} Au&=\begin{pmatrix} -24\\20 \end{pmatrix}\\ &=-4\begin{pmatrix} 6\\-5 \end{pmatrix}\\ &=-4u\\ &=\lambda u\\ \\ Av&=\begin{pmatrix} -9\\11 \end{pmatrix}\\ &\neq \lambda v \end{aligned}
uu is an eigenvector of AA with an eigenvalue −4-4 because Au=−4uAu=-4u.
vv is not an eigenvector of AA because there is no scalar λ\lambda such that Av=λvAv=\lambda v.


Ex.
Show that 77 is an eigenvalue of (1652)\begin{pmatrix} 1&6\\5&2 \end{pmatrix}.

Ax=7xAx−7x=0(A−7I)x=0(−665−5)x=0\begin{aligned} Ax&=7x\\ Ax-7x&=0\\ (A-7I)x&=0\\ \begin{pmatrix} -6&6\\5&-5 \end{pmatrix}x&=0 \end{aligned}