Let $V$ be vector space of $n\times n$ matrices .

$\langle A,B \rangle = {\rm tr}(A^T B)$

I wanted to find orthonormal basis of it.

I know that if there small say 3 dimension vector say then I would have find first matrix of bilinear form

But here dimension are $n$ so how to tackle such problem


There are $n\times n$ elements in this basis, let's label each element as $M^{(k,l)}$, for $k,l = 1,\cdots, n$

$$ M^{(k,l)}_{i,j} = \delta^k_i \delta^l_j $$

So the matrix $M^{(k,l)}$ is full of zeros, except for the row $k$ and column $l$, where the value is $1$. The product of any two elements of this basis is

\begin{eqnarray} \langle M^{(k,l)}, M^{(p,q)} \rangle &=& \sum_{i,j} M^{(k,l)}_{ji} M^{(p,q)}_{ji} \\ &=& \sum_{i,j} \delta^k_j\delta^l_i \delta^p_j \delta^q_i \\ &=& \delta^k_p \delta^l_q \end{eqnarray}

That is, the product is $1$ if $k = p $ and $l = q$, and is $0$ otherwise. So the set of matrices $\{M^{(k,l)} \}_{k,l = 1}^{n}$ is orthonormal

  • $\begingroup$ Sir I could not understand how to algebra of delta is working? it is very new for me.Please can you tell me ? $\endgroup$ – MathLover Nov 25 '18 at 15:42
  • $\begingroup$ @MathLover $$\delta^i_j = \begin{cases}1 &,& i = j \\ 0 &,& {\rm otherwise}\end{cases}$$ That means that in a sum of the form $\sum_i a_i \delta^i_j$ all terms vanish, except the one for which $i = j$, $$\sum_i a_i \delta^i_j = a_j$$ $\endgroup$ – caverac Nov 25 '18 at 15:45

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.