In my linear algebra text (Strang), an orthogonal matrix is defined to be a square matrix whose columns are orthonormal. In other words, an orthogonal matrix is a matrix $Q = [q_1 \cdots q_n]$ where each $q_i$ is a unit column vector of length $n$ with $$q_i^Tq_j = \begin{cases} 0 & \text{ when }i \neq j \\ 1 & \text{ when } i = j \end{cases}.$$

My book says that an orthogonal matrix $Q$ has the properties

  1. $Q^TQ = I$ and
  2. $QQ^T = I$,

from which it follows that $Q^T = Q^{-1}$. I can see how (1) follows from orthonormal columns, but I don't see why (2) is necessarily true. Can anyone provide some insight?

  • 1
    $\begingroup$ Each element of $Q^TQ$ is the scalar product of two columns of $Q$. These vectors are orthonormal, so... $\endgroup$ – amd Jun 19 '16 at 16:56
  • 2
    $\begingroup$ @amd My question was about $QQ^T$ though, not $Q^TQ$. $\endgroup$ – tddevlin Jun 19 '16 at 17:04
  • $\begingroup$ Oops. Missed that. The other follows from the properties of square matrices, as per the accepted answer below. $\endgroup$ – amd Jun 20 '16 at 20:42

From $\mathrm Q^{\top} \mathrm Q = \mathrm I_n$, we conclude that $\mathrm Q^{\top}$ is the left inverse of $\mathrm Q$. Since $\mathrm Q$ is square, the left inverse is also the right inverse and, thus, $\mathrm Q \mathrm Q^{\top} = \mathrm I_n$. Hence, the inverse of $\mathrm Q$ is $\mathrm Q^{-1} = \mathrm Q^{\top}$.


If we know that $Q^TQ=I$ we may note that $$ QQ^T = (QQ^T)(QQ^{-1}) = Q(Q^TQ)Q^{-1} = QIQ^{-1} = I. $$ Here we used the fact that matrix $Q^{-1}$ exist which is follows from $Q^TQ = I$ (indeed, $\det (QQ^T) = 1 \Rightarrow (\det Q)^2 = 1\Rightarrow \det Q \neq 0)$ .


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.