Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

How to prove a unit norm matrix is the average of two unitary matrix

share|improve this question
which norm is unit? –  Igor Rivin Jan 2 '11 at 17:49
Do you know this is true? How many dimensions are you working in? –  Peter Shor Jan 2 '11 at 17:55
I think the only norm where this question makes sense is the spectral norm. I see the votes to close piling up, but since nobody has jumped in and said that this is a well-known elementary fact, I think if you add some details to your question, and flag it to be reopened, it probably will be. There's an easy proof for 2 dimensions and the spectral norm. My first reaction is that I wouldn't think it would be true for arbitrarily many dimensions, and the spectral norm, but I could easily be wrong. –  Peter Shor Jan 2 '11 at 18:24
Further to Peter Shor's comment: mathoverflow.net/howtoask –  Yemon Choi Jan 2 '11 at 21:41
Moderators: Please rollback to the original statement of the question! –  Andres Caicedo Jan 6 '11 at 17:15
show 3 more comments

1 Answer

I assume that the norm is the spectral norm.

By the polar decomposition, any unit norm matrix can be written as $UD$ where $U$ is unitary and $D$ is matrix which, in some basis $E$, is diagonal with non-negative entries not greater than $1$. The diagonal entries can thus be written as $\cos(\theta_1),...,\cos(\theta_n)$ for real $\theta_i$'s. $UD$ is then the average of $U V$ and $U V^*$ where $V$ is the matrix which, in the basis $E$, is diagonal with diagonal entries $e^{i\theta_1},...,e^{i\theta_n}$.

share|improve this answer
thanks for your answering this question. However I still doubt whether the basis E is orthogonal(unitary) basis. –  yaoxiao Jan 2 '11 at 20:41
@yaoxiao: the matrix $D$ is hermitian (it is even positive semid-efinite), and every hermitian matrix can be diagonalized in an orthonormal basis. –  Mikael de la Salle Jan 2 '11 at 20:48
I see what you have mean,great! I think it is really interesting, when we compare it with the similar claim “every matrix can be decompsed into linear combinations of four orthogonal matrix –  yaoxiao Jan 3 '11 at 4:27
add comment

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.