10
$\begingroup$

Let $V$ be a finite-dimensional vector space and $\beta$ be a bilinear form. According to the definition of the bilinear forms is possible to do the following: $$\begin{align}\beta(u,v)&=\beta(u+w-w,v)=\beta(u+w,v)-\beta(w,v)\\ &=\beta(u+w,v)+\beta(w,-v)=\beta(u+2w,v-v)\\ &= \beta(u+2w,0)=0\end{align}$$

I find it curious, I think there's some mistake but I can not identify it, I ask for your help in this matter.

$\endgroup$
  • 14
    $\begingroup$ $\beta(a+b,c+d)=\beta(a,c)+\beta(b,d)$ doesn't hold. $\beta(a+b,c+d)=\beta(a,c)+\beta(a,d)+\beta(b,c)+\beta(b,d)$ does. $\endgroup$ – Yves Daoust Sep 2 '16 at 16:27
  • 12
    $\begingroup$ Note: this is a valid proof that all bilinear, linear functions are 0. $\endgroup$ – Solomonoff's Secret Sep 2 '16 at 21:28
30
$\begingroup$

The step $$\beta(u+w,v) + \beta(w,-v) = \beta(u+2w,v-v)$$is wrong. You must have one of the arguments fixed in both terms in order to apply linearity in the other one. That is, $\beta$ being bilinear in $u$ and $v$ is not the same as being linear in the pair $(u,v)$. For example, the multiplication in $\Bbb R$ is bilinear. You'd be saying that $$(u+w)v + w(-v) = (u+2w)(v-v)$$for all $u,v,w \in \Bbb R$, which is false (take $u=v=w=1$, say).

$\endgroup$
6
$\begingroup$

If the function $\gamma\colon V\times V\to V$ satisfies \begin{align} &\gamma(u,v)+\gamma(x,y)=\gamma(u+x,v+y),\tag{1} \\ &\gamma(au,v)=a\gamma(u,v),\tag{2} \\ &\gamma(u,av)=a\gamma(u,v),\tag{3} \end{align} for all $u,v,x,y\in V$ and all scalars $a$, then $\gamma(u,v)=0$ for all $u,v$.

Indeed, $\gamma(u,0)=\gamma(u,0\cdot0)=0\gamma(u,0)=0$ and similarly $\gamma(0,v)=0$; therefore $$ \gamma(u,v)=\gamma(u,0)+\gamma(0,v)=0+0=0 $$

Note that this is basically the statement you proved. However, for a bilinear form property $(1)$ is not required. Rather, it is required that $$ \beta(u,x+y)=\beta(u,x)+\beta(u,y),\qquad \beta(u+v,x)=\beta(u,x)+\beta(v,x) $$ which is very different from the other property.

Are there maps $f\colon V\times V\to V$ that satisfy property $(1)$? Yes, for instance all linear maps $V\oplus V\to V$ do. However, they don't satisfy $(2)$ and $(3)$, but only $$ f(au,av)=af(u,v) $$

$\endgroup$

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.