there is matrix $ B=\left(\begin{array}{ccc}1 & -1 \\ 1 & 3\\ \end{array}\right)$ ,

for vector $ v_1=\left(\begin{array}{ccc}1 \\ -1\\ \end{array}\right), B v_1=2v_1$ holds.
using $v_1$ and $v_2=\left(\begin{array}{ccc}a \\ b\\ \end{array}\right) $ which is a linear independent vector of magnitude 1, we can express $B.v_2$ as linear combination of $v_1$ & $v_2$ such as $B.v_2=\alpha v_1+2v_2$.
a.) find $v_2$ & $\alpha$

b.) $P=\left(\begin{array}{ccc}1 & a \\ -1 & b\\ \end{array}\right)$, then we can express $BP=PC$, find matrix C, and find $B^n$

first try to find $v_2$ & $\alpha$, $(B-2E)v_2=v_1$ $ B=[\left(\begin{array}{ccc}1 & -1 \\ 1 & 3\\ \end{array}\right)$ - $ \left(\begin{array}{ccc}2 & 0 \\ 0 & 2\\ \end{array}\right) ] .v_2 =\left(\begin{array}{ccc}1 \\ -1\\ \end{array}\right)$
got $v_2=\left(\begin{array}{ccc}-b-1 \\ b \end{array}\right)$ = $\left(\begin{array}{ccc}-1 \\ 1\\ \end{array}\right)s + \left(\begin{array}{ccc}-1 \\ 0\\ \end{array}\right)t$, since $v_2$ is vector with magnitude 1 , i chose $ \left(\begin{array}{ccc}-1 \\ 0\\ \end{array}\right)$ with $\alpha =1$, is my answer right?
then find matrix $C$ , i know matrix $C$ is matrix diagonal, $P^{-1}BP=C$, so with $P=\left(\begin{array}{ccc}1 & -1\\ -1 & 0\\ \end{array}\right)$ , after some calculation $ C=\left(\begin{array}{ccc}2 & 0 \\ 1 & 2\\ \end{array}\right)$

But i wasnt so sure to find $B^n$ ?

  • $\begingroup$ Actually,$$P^{-1}BP=\begin{pmatrix}2&1\\0&2\end{pmatrix}.$$ $\endgroup$ – José Carlos Santos Dec 8 '17 at 18:17
  • $\begingroup$ @JoséCarlosSantos thankyou so much!!! will calculate it once again $\endgroup$ – fiksx Dec 10 '17 at 5:20

Credit: José Carlos Santos for pointing out that $C= \begin{bmatrix} 2 & 1 \\ 0 & 2\end{bmatrix}$




$$C=\begin{bmatrix} 2 & 1 \\ 0 & 2 \end{bmatrix}$$

$$C^2 = \begin{bmatrix} 4 & 4 \\ 0 & 4 \end{bmatrix}=\begin{bmatrix} 2^2 & 2(2) \\ 0 & 2^2 \end{bmatrix}$$

$$C^3 = \begin{bmatrix} 8 & 0 \\ 12 & 8 \end{bmatrix}=\begin{bmatrix} 2^3 & 3(2^2) \\ 0 & 2^3 \end{bmatrix}$$

Compute a few more terms, spot a pattern and prove the pattern of $C^n$ perhaps by induction.


$\begin{bmatrix} a \\ b\end{bmatrix}= \begin{bmatrix} -1 \\ 1\end{bmatrix} b + \begin{bmatrix} -1 \\ 0\end{bmatrix}$

Note that there is no $t$ at the end.

Since you were solving $$(B-2I)v_2 = v_1$$

You have chosen $\alpha = 1$.

If you want to impose the condition that $v_2$ is of length $1$.

You just have to solve for $b$ in $\left\|\begin{bmatrix} -1 \\ 1\end{bmatrix} b + \begin{bmatrix} -1 \\ 0\end{bmatrix}\right\|^2=1$, no guessing is required.

  • $\begingroup$ thanks!!, but i saw some problem of computing power of matrix, with this relationship $P^{-1}B^nP=C^n$ , then $C=\left(\begin{array}{ccc}2^n & 0 \\ 0 & 2^n\\ \end{array}\right)$ , why it not applicable here? and also suppose it is big matrix like 3X3 then to compute $C^n$ will be hard, and hard to find the pattern too. so is there another way(?) $\endgroup$ – fiksx Dec 8 '17 at 10:24
  • $\begingroup$ It is not applicable because $C$ is not a diagonal matrix for this particular problem. We know the general format for the power of Jordan block. For example, here $\endgroup$ – Siong Thye Goh Dec 8 '17 at 10:29
  • $\begingroup$ C is not diagonal matrix, because B rank is 1 but the geometry multiplicity is two for $\lambda =2$? so it is not diagonal matrix? since it is diagonal matrix if gm=am, but here gm<am, so i have to use jordan block? but is this matrix diagonalizable(?) $\endgroup$ – fiksx Dec 8 '17 at 11:48
  • $\begingroup$ rank of $B$ is $2$. the question is what is the nullity of $B-\lambda I$. It is $1$ in this case. The geometric multiplicity is $1$, but the algebraic multiplicity is $2$, hence it is not diagonalizable. Every matrix has a jordan canonical form, which is more general than diagonalization. I am guessing you might be asked this question on a diagonalization question to test if you can figure out the relation of $P^{-1}B^nP=C^n$. $\endgroup$ – Siong Thye Goh Dec 8 '17 at 12:07
  • $\begingroup$ yes, i mean rank P is 1 and the nullity is 1 same as geometry multiplicity , but because algebraic multiplicity is 2, so it has 1 eigen vector only, it is not diagonalizable, but suppose the nullity of $B-\lambda I=4$ and rank= 0 (is it possible?), then it is diagonalizable with $\lambda $ in the diagonal and all $0$ , is my assumption true? yes, i never heard jordan block before. $\endgroup$ – fiksx Dec 8 '17 at 12:51

It looks like this exercise is taking you step-by-step through calculating what’s known as the Jordan normal form of $B$ (in this exercise, the matrix $C$) and then using that to compute $B^n=(PCP^{-1})^n=PC^nP^{-1}$. It’s not too hard to work out the pattern for $C^n$, as illustrated in other answers, but you can save yourself some guesswork by decomposing $C$ into the sum of the diagonal matrix $2I$ and a nilpotent matrix $N$, i.e., $$C=\begin{bmatrix}2&1\\0&2\end{bmatrix} = \begin{bmatrix}2&0\\0&2\end{bmatrix}+\begin{bmatrix}0&1\\0&0\end{bmatrix} = 2I+N.$$ Now, use the Binomial theorem and the fact that $2I$ and $N$ commute to expand $$(2I+N)^n = 2^nI+n2^{n-1}N+\binom n 2 2^{n-2}N^2 + \cdots$$ but $N^2=0$, as you can verify for yourself, therefore $$C^n = 2^nI+n2^{n-1}N = \begin{bmatrix}2^n & n2^{n-1} \\ 0 & 2^n\end{bmatrix}.$$

In fact, you can use a similar decomposition to compute $B^n$ more directly, without finding the matrix $P$ first. We have $$B = PCP^{-1} = P(2I+N)P^{-1} = 2PIP^{-1}+PNP^{-1} = 2I+PNP^{-1}.$$ Now, $$(PNP^{-1})^2 = PNP^{-1}PNP^{-1} = PN^2P^{-1} = 0,$$ so by a similar computation to that above we get $$B^n = 2^nI + n2^{n-1}(B-2I) = \begin{bmatrix}2^n&0\\0&2^n\end{bmatrix}+\begin{bmatrix}-n2^{n-1}&-n2^{n-1}\\n2^{n-1}&n2^{n-1}\end{bmatrix} = \begin{bmatrix}2^n-n2^{n-1} & -n2^{n-1} \\ n2^{n-1} & 2^n+n2^{n-1}\end{bmatrix}.$$ This is a lot less work than working out generalized eigenvectors of $B$ (the matrix $P$ in this exercise) and then doing all of those matrix multiplications. You can use this method to compute powers and exponentials of any $2\times2$ matrix with a repeated eigenvalue that isn’t a multiple of the identity (and the method can be generalized to some extent to larger matrices).

  • $\begingroup$ wow this method really effective!! $(PNP^{-1})^2 = PNP^{-1}PNP^{-1} = PN^2P^{-1} = 0,$, is this mean that $B^2=0$ and so on will be 0 right? is this method only applicable to matrix with repeated eigen value in diagonal? and not applicable to for example $(\lambda -2)^2 ( \lambda-1)$ ? $\endgroup$ – fiksx Dec 10 '17 at 6:20
  • $\begingroup$ @Vixf No, it’s the square and higher powers of $(B-\lambda I)$ that vanish. The method depends on the fact that $\lambda I$ commutes with everything so that you can rearrange the products to bring powers of $N$ together. A general diagonal matrix doesn’t have this property, so this doesn’t work for mixed eigenvalues. There are other short-cuts that you can use for other cases, but at some point it’s just easier to compute the full Jordan decomposition. $\endgroup$ – amd Dec 10 '17 at 18:18
  • $\begingroup$ ok thankyou so much for the clear explanation!!!! $\endgroup$ – fiksx Dec 13 '17 at 11:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.