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

How to prove that if a function $f\colon\mathbb R\to\mathbb R$ is of the class $C^{2n}$ and even then there exists a function $g\colon\mathbb R \to\mathbb R$ of the class $C^n$ such that $f(x)=g(x^2)$.

It is obvious, by using power series expansion in a symmetric neighbourhood of zero, that it holds if $f$ is analytic and even. But how to show that in general?


share|cite|improve this question
The title does not reflect the question. – lhf Aug 5 '11 at 12:44
up vote 3 down vote accepted

Taylor expand $f(x) = P_{2n}(x) + R_{2n}(x)$ at $x = 0$, where $P_{2n}(x)$ is of degree $2n$ and $R_{2n}(x)$ is the remainder. Because $f(x)$ is even, all odd derivatives of $f(x)$ at $x = 0$ are zero and therefore $P_{2n}(x)$ only contains even powers of $x$ and thus can be written as $g_1(x^2)$ for some polynomial $g_1$.

As for $R_{2n}(x)$, since it's the difference $f(x) - P_{2n}(x)$ of two even functions it is even as well. Thus one can write $R_{2n}(x) = g_2(x^2)$ for some $g_2$. The key is to show that $g_2(x)$ is actually $C^n$. On ${\mathbb R} - \{0\}$ it's actually $C^{2n}; g_2(x^2) = f(x) - P_{2n}(x)$ is the difference of two $C^{2n}$ functions and this is preserved under change of variables $x \rightarrow \sqrt{x}$ or $x \rightarrow -\sqrt{x}$.

Thus we must consider $x = 0$ only. Since $R_{2n}(x)$ is the remainder term of the Taylor expansion, it satisfies $|R_{2n}(x)| = o(|x|^{2n})$ as $|x| \rightarrow 0$. So $g_2(x) = R_{2n}(\sqrt{|x|})$ satisfies $g_2(x) = o(|x|^n)$ as $|x| \rightarrow 0$, and $g_2(x) = 0$. Similarly, using the corresponding properties of derivatives of $R_{2n}(x)$ one can inductively show that for any $k \leq n$, $g_2^{(k)}(x) = o(|x|^{n-k})$ as $|x| \rightarrow 0$. So taking successive difference quotients in the definition of the derivative shows that $g_2^k(0)$ exists and equals zero for all $k \leq n$.

So we can conclude $g_2(x)$ is a $C^n$ function. Therefore $g(x) = g_1(x) + g_2(x)$ is $C^n$ too.

share|cite|improve this answer
Thank you very much. – Richard Aug 5 '11 at 15:08
One question, bacause I did not understand: How to prove for example that $g_2^{''}(x)=o(|x|^{n-2})$ as $|x| \rightarrow 0$. – Richard Aug 8 '11 at 14:08
Use the chain rule on $R_{2n}(\sqrt{|x|})$ and the fact that the $k$th derivative of $R_{2n}(x)$ is $o(|x|^{2n - k})$ as $x \rightarrow 0$. So the $k$th derivative of $g_2(x)$ is $o(|x|^{n-k})$. – Zarrax Aug 8 '11 at 17:27
and the bounds on the $k$th derivative of $R_{2n}(x)$ just come from the fact $R_{2n}(x)$ has its first $2n$ derivatives all zero at $x = 0$. – Zarrax Aug 8 '11 at 17:34
You just need that $R_{2n}(x)$ is $C^{2n}$ on all of ${\mathbb R}$, which is true. So for example suppose you have established $g_2'(0) = 0$ from the definitions. Then for $x \neq 0$, $g_2'(x) = \pm {1 \over 2\sqrt{|x|}} R_{2n}'(\sqrt{|x|})$. You can take difference quotients of $g_2'(x)$, using the bounds on $R_{2n}'(x)$, and you'll get that $g_2''(0) = 0$. You can repeat this idea $n$ times. – Zarrax Aug 8 '11 at 18:42

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.