Let us suppose I have a continuous smooth function

$f : \mathbb R^n\rightarrow \mathbb R$


$f(\lambda x_1,\lambda x_2, \ldots ,\lambda x_n)= \lambda f( x_1,x_2, \ldots , x_n),\space \forall\lambda\in\mathbb R \space \ldots(linear \space condition)$

My objective is to find the function $f$


Expand in Taylor series the function $f$ for $n$ variables $(i.e.\space x_i)\space in \space the\space equation\space of \space linear \space condition$

1)The constant term is zero as $f(0,...,0)=0$.

2)The linear terms cancel either side.

3)With the higher order terms of coefficients c, $$\lambda^{k-1}c_k=c_k$$

This condition holds true $\forall k,\lambda \in \mathbb R$

This implies that all coefficients of higher order(>1) terms are zero.

Hence I conclude that the function is $$f( x_1,x_2, \ldots , x_n)= a_1 x_1 + a_2 x_2 \ldots +a_n x_n$$

for some constants $a_i(1\leq i \leq n)$

My doubt is:

1) Is my reasoning correct, or is there a better way to do it?

2) Is this the only unique solution for the function?

  • $\begingroup$ For $(x_1, \dots,x_n)$ fixed, define $g(\lambda) = f(\lambda x_1, \dots, \lambda x_n)$. Then differentiate $g$ to get that $f$ is linear. $\endgroup$ – Alan Muniz May 31 '18 at 14:17
  • $\begingroup$ @AlanMuniz Yea, thanks for it. It works good. Can I claim that the solution is unique? $\endgroup$ – Prasanna May 31 '18 at 14:28
  • $\begingroup$ Yes, provided that $f$ is smooth. $\endgroup$ – Alan Muniz May 31 '18 at 14:33
  • $\begingroup$ I see....thanks $\endgroup$ – Prasanna May 31 '18 at 14:40

This is incorrect. Your approach is correct if $f$ is equal to its Taylor series, but this is not the case for most smooth functions.

What you can do instead is consider the behavior of $f$ along any line through the origin. Clearly the given homogeneity condition says $f$ is linear when restricted to any line through the origin (keep $(x_1,\dots,x_n)$ fixed and let $\lambda$ vary). Moreover, by computing the directional derivative at $0$ along such a line and using the fact that the directional derivative is the appropriate linear combination of the partial derivatives, you can conclude that in fact $f(x_1,\dots,x_n)=a_1x_1+\dots+a_nx_n$ where the $a_i$ are the partial derivatives of $f$ at $0$.

  • $\begingroup$ The method you said works out well. But I fail to see why some smooth functions cannot be represented in taylor series? $\endgroup$ – Prasanna Jun 1 '18 at 10:25

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.