2

Consider the equation: $$ C \equiv HMH^{-1} \pmod{p}, $$ where $C,M, H$ are, say, $2\times 2$ matrices, and $p$ is an odd prime. The elements of the matrices $C, M$ are integers. The elements of the matrix $H$ are the unknowns (call them $h_{11}, h_{12}, h_{21}, h_{22}$). I am trying to solve this system for the unknowns $h_{11}, h_{12}, h_{21}, h_{22}$.

Here is my approach:

$$ C \equiv HMH^{-1} \pmod{p} $$ implies that $$CH = HM.$$ This will result in four linear equations in $h_{11}, h_{12}, h_{21}, h_{22}$. That is, we have

$$W \> h \equiv 0 \pmod{p},$$ where ${h} = transpose(h_{11}, h_{12}, h_{21}, h_{22}). $ Now for the system to have non-trvial solution the matrix $W$ must be singular in $Z_p$ (i.e. $\det(W) \equiv 0 \pmod{p}$). This also would imply that the system $$W \> { h} \equiv 0 \pmod{p},$$ has at least one free variable, and hence, this system has at least $p-1$ non-trivial solutions.

Is the argument above correct? Any help with this will be highly appreciated. Thanks in advance!

user26857
  • 52,094
user244486
  • 21
  • $W$ is a $4\times 4$ matrix with elements in $Z_{p}$. – user244486 Jun 02 '15 at 01:02
  • I have Found this post http://math.stackexchange.com/questions/275746/solving-matrix-equation-xa-ay-with-known-x-and-y?rq=1 and this link http://en.wikipedia.org/wiki/Sylvester_equation which might be useful. – user244486 Jun 09 '15 at 23:49

2 Answers2

1

@user244486,

  1. There is no general formula.

  2. Clearly, the associated theory is too complicated for you.

  3. Seek the solutions $H\in M_2(F)$ over $F=\mathbb{Z}/3\mathbb{Z}$ when $C=\begin{pmatrix}1&2\\1&1\end{pmatrix},\ M=\begin{pmatrix}a&1\\1&2\end{pmatrix}$. In particular, how to choose $a$ so that there are solutions ?

EDIT. Answer to user244486. It remains to find explicitly all the solutions in $H$ (there are $4$; why ?). Note that there are no eigenvalues of $C,M$ in $F$; indeed they are in an algebraic extension of $F$. Yet $C,M$ are similar over $F$ (this is a theorem !).

  • In this particular example, choose $a$ such that characteristic polynomial of $C$ and $M$ are identical. Setting $a=0$ results in trace($M$) = Trace($C$) = 2, $\det(M)=\det(C) = -1$. – user244486 Jun 02 '15 at 01:18
  • Do you mind sharing the associated theory? – user244486 Jun 02 '15 at 01:22
0

As far I see you are trying to decide if the matrix $C$ and $M$ are similar. See for example http://en.wikipedia.org/wiki/Matrix_similarity. A necessary condition for the existence of $H$ is going to be that the so called carateristic polinomial of $C$ and $M$ are the same, i.e. $\det(C) = \det(M)$ and ${\rm trace}(C) = {\rm trace}(M)$. But this conditions are not sufficient. Take for example $C = \left( \begin{array}{cc} 0 & 1 \\ 0 & 0 \\ \end{array} \right)$ and $M = \left( \begin{array}{cc} 0 & 0 \\ 0 & 0 \\ \end{array} \right)$.

Turning to your question. I think the condition $\det(W) = 0$ you got is equivalent to $\det(C) = \det(M)$ (but I did not check it). If this is so then of course the system $W.h = 0$ has a non trivial solution but could happen that the nonzero matrix $H$ you got is not invertible hence does not give you a solution of your original question i.e. if $C = H M H^{-1}$.

Summing up what you have to realize is that you are doing linear algebra in a vector space over a finite field. Then you can apply the standard techniques to solve the problem : given $C$ and $M$ does there exists $H$ such that $C = H M H^{-1}$ ? If you want I can give you the necessary and sufficient conditions for the existence of such $H$.

user26857
  • 52,094
Holonomia
  • 2,543
  • My guess is that the necessary and sufficient condition is that both of the matrices $M$ and $C$ have the same Jordan normal form? This is just a guess? – user244486 Jun 02 '15 at 01:06
  • The OP question was about $2 \times 2$ matrices so in this case the necessary and sufficient condition is : either the characteristic polynomial has two different roots or in case that the characteristic polynomial has a double root then both matrices are already diagonal or both matrices are non diagonal. Of course saying $M$ and $C$ had the same Jordan normal form is also a correct assertion but it preassume that you know the Jordan normal form. In such a case perhaps it is better to use the so called Rational Normal Form. – Holonomia Jun 02 '15 at 09:54