Exponentiation with Matrices

177 Views Asked by At

How would exponentiation be done with matrices?

For example:

$$ \left[\begin{array}{cc}1&2\\8&7\end{array}\right]^{\left[\begin{array}{cc}7&4\\2&9\end{array}\right]}$$

Where both base and exponent are matrices.

$$ \left[\begin{array}{cc}1&2\\8&7\end{array}\right]^6$$

Where the base is a matrix, and the exponent is not.

$$6^{\left[\begin{array}{cc}7&4\\2&9\end{array}\right]}$$

Where the base is not, and the exponent is a matrix.


Wolfram Alpha doesn't recognize it, which makes me wonder if it is even possible.

2

There are 2 best solutions below

3
On

Using abel and Peter's comments, I was able to solve my first question (matrix exponentiation with two matrices).

Please let me know if I am incorrect in any of my logic.

$$\text{Using a random example:} \\ a = \begin{bmatrix} 1 & 2\\ 8 & 7 \end{bmatrix} \\ b = \begin{bmatrix} 7 & 4\\ 2 & 9 \end{bmatrix} \\ a^{b} = e^{b\ln(a)} \\ \ln(a) = \begin{bmatrix}\ln(1)&\ln(2)\\\ln(8)&\ln(7)\end{bmatrix} \\ \\ a^b = e^{\begin{bmatrix}7&4\\2&9\end{bmatrix}\times\begin{bmatrix}\ln(1)&\ln(2)\\\ln(8)&\ln(7)\end{bmatrix}} \\ a^b = \begin{bmatrix} 7 & 16\\ 256 & 4782969 \end{bmatrix} $$

Wolfram Alpha corroborates the final exponential arithmetic, $e^{\begin{bmatrix}7&4\\2&9\end{bmatrix}\times\begin{bmatrix}\ln(1)&\ln(2)\\\ln(8)&\ln(7)\end{bmatrix}}$, link here.

4
On
  • Concerning the logarithm of a matrix
    The logarithm (understood as the inverse of exp) of a matrix is not the log of each component: example $$ \exp \left( {\left( {\begin{array}{*{20}c} 0 & 0 & 0 \\ 1 & 0 & 0 \\ 0 & 1 & 0 \\ \end{array} } \right)} \right) = \left( {\begin{array}{*{20}c} 0 & 0 & 0 \\ 1 & 0 & 0 \\ {1/2} & 1 & 0 \\ \end{array} } \right)\quad \quad \exp \left( {\left( {\begin{array}{*{20}c} 1 & 0 & 0 \\ 1 & 1 & 0 \\ 0 & 1 & 1 \\ \end{array} } \right)} \right) = \left( {\begin{array}{*{20}c} e & 0 & 0 \\ e & e & 0 \\ {1/2e} & e & e \\ \end{array} } \right) $$ To my knowledge, it can instead be defined through the Taylor expansion of $ln(\mathbf X)$ or better, of $ln(\mathbf I + (\mathbf X - \mathbf I))$ when this is convergent.
  • Concerning the matrices commutation

    You shall be aware that the product of two Matrices $\mathbf A$ and $\mathbf B$, in general does not commute. Same for the product of $ln(\mathbf A)$ and $\mathbf B$.
    You can "define" $\mathbf A ^{\mathbf B}$ as you like, as far as you state the properties you expect therefrom.