3
$\begingroup$

One can take a diagonal matrix to a certain power by just taking diagonal elements that power. There is a similar polynomial-time (in respect to the matrix dimensions) shortcut for triangular matrices.

Is there a shortcut for exponentiating a square matrix of the form

$$ \left( \begin{array}{cccc} a & b & c & d \\ 1 & 0 & 0 & 0 \\ 0 & 1 & 0 & 0 \\ 0 & 0 & 1 & 0 \end{array} \right) $$ , where $a$, $b$, $c$, and $d$ are constants?

$\endgroup$
3
  • $\begingroup$ looks very similar to the transpose of the companion matrix en.wikipedia.org/wiki/Companion_matrix (which is diagonalizable, and hence the power is easily computed). $\endgroup$
    – Andreas H.
    Commented Feb 4, 2014 at 2:57
  • $\begingroup$ @AndreasH ... Not true in either regard, sadly. $\endgroup$ Commented Feb 4, 2014 at 3:04
  • $\begingroup$ @TedShifrin: yes, not exactly, but apparently similar enough (see my answer). $\endgroup$
    – Andreas H.
    Commented Feb 4, 2014 at 3:30

2 Answers 2

4
$\begingroup$

By using Cayley Hamilton Theorem. If you are doing the calculations by hand, then the problem is simple if the characteristic polynomial has simple roots.

The general approach is to divide $x^n$ by $x^4-a x^3 - b x^2 - c x - d$ find the remainder. Thus suppose $$ x^n = Q(x) (x^4-a x^3 - b x^2 - c x - d) + { \alpha x^3 + \beta x^2 + \gamma x + \delta } \tag 1 $$ Here we do not need the quotient $Q(x)$ but just the remainder $\alpha x^3 + \beta x^2 + \gamma x + \delta $

Then, replacing $x$ by $A$ we have

$$ A^n = Q(A) (A^4-a A^3 - b A^2 - c A- d I) + { \alpha A^3 + \beta A^2 + \gamma A + \delta I } \tag 2 $$ In (2) we have $$(A^4-a A^3 - b A^2 - c A- d I) =0 ~\text{Verify this by direct calculation!}$$ so $$ A^n = { \alpha A^3 + \beta A^2 + \gamma A + \delta I }$$

One can easily calculate $$A^2=\pmatrix{b+a^2&c+a\,b&d+a\,c&a\,d\cr a&b&c&d\cr 1&0&0&0\cr 0&1& 0&0\cr }$$ $$A^3=\pmatrix{c+2\,a\,b+a^3&d+a\,c+b^2+a^2\,b&a\,d+b\,c+a^2\,c& \left(b+a^2\right)\,d\cr b+a^2&c+a\,b&d+a\,c&a\,d\cr a&b&c&d\cr 1&0& 0&0\cr }$$

The key to the answer is finding the remainder. One can use recursion (synthetic division) or remainder theorem to find the remainder.

Here is a simple example on finding the remainder when the characteristic equation has simple roots.

Example: Let $$a=2 , b=1 , c=-2 , d=0 $$ Then the roots of $x^4-a x^3 - b x^2 - c x - d$ are $x=0$, $x=1$, $x=-1$ and $x=-2$.

Now let $$x^n = Q(x) (x^4-a x^3 - b x^2 - c x - d) + \alpha x^3 + \beta x^2 + \gamma x + \delta$$

We substitute different roots we got as

Set $x=0$ $$ 0 = \delta \tag 3$$ Set $x=1$: $$ 1 = \alpha+\beta+\gamma+\delta \tag 4$$ Set $x=-1$ $$ (-1)^n = -\alpha+\beta-\gamma+\delta \tag 5$$ Set $x = 2$ $$ 2^n = 8 \alpha + 4 \beta + 2 \gamma + \delta \tag6$$

We can solve to get $$ \alpha={{2^{n}-\left(-1\right)^{n}-3}\over{6}} , \beta={{ \left(-1\right)^{n}+1}\over{2}} , \gamma=-{{2^{n}+2\,\left(-1\right) ^{n}-6}\over{6}} , \delta=0 $$

So we have $$ A^n = {{2^{n}-\left(-1\right)^{n}-3}\over{6}} A^3 + {{ \left(-1\right)^{n}+1}\over{2}} A^2 + {{2^{n}+2\,\left(-1\right) ^{n}-6}\over{6}} A$$

$\endgroup$
3
$\begingroup$

The problem can be solved when the polynomial $x^4 - ax^3 - bx^2 - cx^1 - d$ has 4 distinct (complex) roots.

First we note, that if $\lambda$ is a root, the matrix has eigenvalue $\lambda$ with eigenvector $( \lambda^3, \lambda^2, \lambda, 1 )^T$:

$$ \left( \begin{array}{cccc} a & b & c & d \\ 1 & 0 & 0 & 0 \\ 0 & 1 & 0 & 0 \\ 0 & 0 & 1 & 0 \end{array} \right) \left( \begin{array}{cccc} \lambda^3 \\ \lambda^2 \\ \lambda \\ 1 \end{array} \right) = \left( \begin{array}{cccc} \lambda^4 \\ \lambda^3 \\ \lambda^2 \\ \lambda \end{array} \right) = \lambda \left( \begin{array}{cccc} \lambda^3 \\ \lambda^2 \\ \lambda \\ 1 \end{array} \right) $$

Since it has 4 distinct eigenvalues (and linear independent eigenvectors) we can then diagonalize the matrix.

Let $$ V = \left( \begin{array}{cccc} \lambda_1^3 & \lambda_2^3 & \lambda_3^3 & \lambda_4^3 \\ \lambda_1^2 & \lambda_2^2 & \lambda_3^2 & \lambda_4^2\\ \lambda_1 & \lambda_2 & \lambda_3 & \lambda_4\\ 1 & 1 & 1 & 1 \end{array} \right) $$ then we we have $$ D = V^{-1}M V $$ with the original matrix denoted as $M$ and $$ D = \left( \begin{array}{cccc} \lambda_1 & 0 & 0& 0 \\ 0 & \lambda_2 & 0 & 0\\ 0& 0 & \lambda_3 & 0\\ 0 & 0 & 0 & \lambda_4 \end{array} \right) $$

Finally, we can compute arbitrary powers of $M$ by $$ M^n = V D^n V^{-1} $$

This derivation is analogous to the one of the companion matrix, which is very similar to the present matrix (here different permutation and negative $a,b,c,d$). See http://en.wikipedia.org/wiki/Companion_matrix

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .