Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am stuck with an exercise that I found in a textbook by Conway. First, I would like to clarify what is meant by a semi-inner product.

Definition. Suppose that $\mathscr X$ is a vector space over the complex field $\mathbb C$. A semi-inner product on $\mathscr X$ is a function $u:\mathscr X\times\mathscr X\to\mathbb C$ such that for all $\alpha,\beta$ in $\mathbb C$, and $x,y,z$ in $\mathscr X$, the following are satisfied:

  • $u(\alpha x+\beta y,z)=\alpha u(x,z)+\beta u(y,z)$,
  • $u(x,x)\ge 0$,
  • $u(x,y)=\overline{u(y,x)}$,

where $\bar\alpha$ is the complex conjugate of $\alpha$.

The difference between an inner product and a semi-inner product is that an inner product also satisfies the following:

  • if $u(x,x)=0$, then $x=0$.

Now I formulate the exercise from the textbook.

Let $u(\cdot,\cdot)$ be a semi-inner product on $\mathscr X$. Then $$\left|u(x,y)\right|^2=u(x,x)u(y,y)$$ if and only if there are $\alpha$ and $\beta$ in $\mathbb C$, not both $0$, such that $u(\beta x+\alpha y,\beta x+\alpha y)=0$.

How can I show that if there are $\alpha$ and $\beta$ in $\mathbb C$, not both $0$, such that $u(\beta x+\alpha y,\beta x+\alpha y)=0$, then $\left|u(x,y)\right|^2=u(x,x)u(y,y)$?

share|cite|improve this question
I rather strongly suspect that the requirement should say “not both $0$”, not “both not $0$”. – Harald Hanche-Olsen Oct 19 '12 at 11:17
Have you tried expanding out $u(\beta x+\alpha y,\beta x+\alpha y)$ and look at the result? Note that the result must always be nonnegative for all choices of $\alpha$, $\beta$, and zero for the given pair. This should be useful. – Harald Hanche-Olsen Oct 19 '12 at 11:20
The second condition in the definition should be $u(x,x) \geq 0$ not $u(x,y) \geq 0$. (If $u(x,y) \geq 0$ for all $x$ and $y$ then $u$ is identically equal to $0$). – Yury Oct 19 '12 at 12:00
Thank you, Yury, for correcting my typo. – Cm7F7Bb Oct 19 '12 at 12:03
Reply to Harald's remark: the requirement in the textbook says "both not $0$". I guess that otherwise the proposition is false. If $\alpha=0$ and $\beta\ne0$, then $u(x,x)=0$. Assume that $x\ne0$ (the equality is valid if $x=0$), but $u(x,x)=0$. Unless we show that $x(x,y)=0$ for all $y\in\mathscr X$ if $u(x,x)=0$, we get a contradiction. Is it possible to construct an example of a semi-inner product that for some $x\ne 0$ $u(x,x)=0$, but there exists $y\in\mathscr X$ such that $u(x,y)\ne0$? – Cm7F7Bb Oct 19 '12 at 12:27

Instead of $\mu \langle x, y\rangle $ I have used $ \langle x, y\rangle.$

Let $ \langle\mathcal{X}, .\rangle$ be semi inner product. Let $x $, $y $ be fixed vectors in $\mathcal{X}$ and $\gamma$ be scalar. Consider $$ \langle x - \gamma y, x - \gamma y\rangle = \langle x, y\rangle - \gamma\langle y,x\rangle - \bar\gamma\langle x, y\rangle + |\gamma|^2\langle y, y\rangle $$ Put $\langle y,x\rangle = b \mathrm{e}^{i\lambda} (b \ge 0) $ , $ \gamma = t\mathrm{e}^{-i\lambda}$ (t is real), $ a = \langle y, y\rangle, c = \langle x, x\rangle $ Note here that $\lambda, a, c $ are constants whereas $ t $ is real variable. With this, we have $$ \langle x - \gamma y, x - \gamma y\rangle = c - 2bt + at^2 \tag{1} $$ Now, $$ |\langle x, y\rangle|^2 = \langle x, x\rangle\langle y, y\rangle \iff b^2 - ac = 0 \iff 4b^2 - 4ac = 0 $$ $\iff c - 2bt + at^2 = 0 $ have equal roots. This is true if and only if $ c - 2bt + at^2 = 0$ has unique real root, say $ t_0 $. So by taking $\gamma_0 = t_0\mathrm{e}^{-i\lambda}$, from the equation $(1)$, we obtain that $$ \langle x - \gamma_0 y, x - \gamma_0 y\rangle = c - 2bt_0 + at_0^2 = 0 $$ Thus, the required scalars in your problem are $\beta = 1 $ and $\alpha = -\gamma_0 $

share|cite|improve this answer
If $\langle y,x\rangle=be^{i\lambda}$, how do we know that $\gamma=t e^{-i\lambda}$? We only know that $\langle x-\gamma y,x-\gamma y\rangle=0$, where $\gamma\in\mathbb C$. We can express any complex number as $\gamma=te^{i\varphi}$, but how do we know that $\varphi=-\lambda$? This trick is used in the proof of the CBS inequality, but there we can choose $\gamma\in\mathbb C$ freely and here $\gamma\in\mathbb C$ is fixed. – Cm7F7Bb Feb 3 '15 at 14:30

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.