Let $f \, \colon \mathbb{R}^d \rightarrow \mathbb{R}^d$ be a smooth and convex function. Assume $f$ behaves asymptotically as a cone at infinity, i.e.,

$ \lim_{R \rightarrow \infty} \frac{f(R x)}{R} = F (x), $

for every $x \in \mathbb{R}^d$, where $F$ is a homogeneous function of degree one which is smooth away from the origin.

I would like to show that the second derivatives of $f$ are bounded.

So far, I was able to show that $f$ is Lipschitz and therefore I know that the gradient $\nabla f$ is bounded. But I have no idea how to prove boundedness of the second derivatives. I would be really grateful for any suggestions. Thanks!


This is not true. Here's a one-dimensional counterexample (that also works in higher dimensions). Begin with the function $g(x) = \sqrt{x^2+1}$ which is smooth, convex, asymptotic to $|x|$, and has $g''(0)=1$. Define $$f(x) = \sum_{n=0}^\infty 3^{-n} g(2^n(x-n))$$ Since $|x|\le g(x)\le |x|+1$, it follows that $$\sum_{n=0}^\infty (2/3)^{-n} |x-n| \le f(x) \le 3+ \sum_{n=0}^\infty (2/3)^{-n} |x-n|$$ hence $3|x|-C_1\le f(x)\le 3|x|+C_2 $ for some constants $C_1,C_2$. So, $f$ is asymptotic to a cone.

On the other hand, $$f''(n) \ge 3^{-n} 4^n g''(0) = (4/3)^n$$

  • $\begingroup$ Nice one! When I saw you were claiming it was false, I thought of this: Take any continuous positive unbounded $h$ on $[0,\infty)$ such that $\int_0^\infty h = L <\infty.$ Then $f(y) = \int_0^y\int_0^x h(t)\,dt\,dx$ is a counterexample. $\endgroup$ – zhw. May 31 '15 at 2:51
  • $\begingroup$ Thank you for this nice example! But I have one question: why is the function f twice differentiable? $\endgroup$ – Mike Jul 1 '15 at 7:52

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.