I'm going to asking a question about the polarization formula of an $n$-form.

Given an $n$-homogeneous form $q:V\rightarrow \mathbb{R}$ on a vector space $V,$ i.e., $\forall v\in V, r\in \mathbb{R},$ we have $q(rv) = r^nq(v),$ the polarization formula tells us that we can have an $n$-multilinear map $f:V^n\rightarrow\mathbb{R},$ $(v_1,...,v_n)\mapsto \frac{1}{n!}(\partial_{\alpha_1}...\partial_{\alpha_n} q(\alpha_1v_1+...+\alpha_nv_n))|_{(\alpha_1,...,\alpha_n)=(0,...,0)}$ such that $\forall v\in V,$ $f(v,...,v) = q(v).$

The equality $f(v,...,v) = q(v)$ is obvious, while my problem is that the $n$-multilinearity of this $f.$ This may be related to the linearity of the differential operators, but I cannot solve it by following the definition to prove $$\frac{1}{n!}(\partial_{\alpha_1}...\partial_{\alpha_n} q(\alpha_1(av_1+v)+...+\alpha_nv_n))|_{\alpha=(0,...,0)}=\\a\cdot\frac{1}{n!}(\partial_{\alpha_1}...\partial_{\alpha_n} q(\alpha_1v_1+...+\alpha_nv_n))|_{\alpha=(0,...,0)}+\frac{1}{n!}(\partial_{\alpha_1}...\partial_{\alpha_n} q(\alpha_1v+...+\alpha_nv_n))|_{\alpha=(0,...,0)}.$$ Or may my work be not feasible?

Thanks for any ideas in advance!


For any $v\in V$, write $\alpha\mapsto q(\alpha v)$ as the composition $q\circ g_v$ of the maps $q:V\to\mathbb{R}$ and $g_v:\mathbb{R}\to V, g_v(\alpha)=\alpha v$. Now $g_v'(0) = v$ and by the chain rule $$\partial_\alpha q(\alpha v)\vert_{\alpha=0} = dq(g_v(0))g_v'(0) = dq(0)v.$$ Since the derivative $dq(0):V\to \mathbb{R}$ of $q$ at 0 is a linear map, $$\partial_\alpha q(\alpha (av+w))\vert_{\alpha=0} = dq(0)(av+w) = a\cdot dq(0)v+dq(0)w = a\cdot\partial_\alpha q(\alpha v)\vert_{\alpha=0}+\partial_\alpha q(\alpha w)\vert_{\alpha=0}.$$

  • $\begingroup$ I got your ideas, thsnks a lot! $\endgroup$ – tommy xu3 Mar 28 '17 at 8:59

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.