This problem is an extension to the simpler problem which deals with $f(x) + f'(x) \to A$ as $x \to \infty$ (see problem 2 on my blog).

If $f$ is twice continuously differentiable in some interval $(a, \infty)$ and $f(x) + f'(x) + f''(x) \to A$ as $x \to \infty$ then show that $f(x) \to A$ as $x \to \infty$.

However, the approach based on considering sign of $f'(x)$ for large $x$ (which applies to the simpler problem in the blog) does not seem to apply here. Any hints on this problem?

I believe that a similar generalization concerning expression $\sum\limits_{k = 0}^{n}f^{(k)}(x) \to A$ is also true, but I don't have a clue to prove the general result.

  • 11
    $\begingroup$ It doesn't work for all $n$, $$\sin x + \sin' x + \sin'' x + \sin''' x \equiv 0.$$ $\endgroup$ – Daniel Fischer Mar 25 '14 at 11:28
  • $\begingroup$ @DanielFischer: Cool Daniel! This counterexample was simple enough to guess but I did not think enough about it. I have updated my post. $\endgroup$ – Paramanand Singh Mar 25 '14 at 11:35

You may factorize the polynomial $\lambda^2+\lambda+1 = (\lambda -c)(\lambda-\bar{c})$ (where $c= (1+\sqrt{3}i)/2$ has positive real part) and accordingly the differential operator $D^2+D+id= (D+c\ id)\circ (D+\bar{c}\ id)$. It is then enough to show the following claim: If $c\in\mathbb C$ has positive real part, $g:(a,\infty)\to \mathbb C$ is continuous with $g(x)\to A \in\mathbb C$ then every solution $f$ of $f'(x)+c f(x) =g(x)$ satisfies $f(x)\to A/c$. (Applying this twice you get what you want since $c\ \bar c =|c|^2=1$.)

To prove the claim you just use the formula for the general solution of a linear differential equation: For every $x_0$ the unique solution with given $f(x_0)$ is $$ f(x)= e^{-c(x-x_0)}(f(x_0)+\int_{x_0}^x e^{c(t-x_0)}g(t) dt).$$ If $x_0$ is large enough you can replace $g(t)$ by $A$ and you obtain the desired limit $A/c$ by calculating the integral (to make this precise you need that the real part of $c$ is strictly positive).

This method gives also something for polynomials of higher degree as long as the roots have strictly positive real parts. In Daniel Fisher's example you have a root $-1$.

  • $\begingroup$ This is really amazing. I was not expecting anything based on differential equations. But the argument you have presented is very simple and clear enough. +1 $\endgroup$ – Paramanand Singh Mar 25 '14 at 13:43
  • $\begingroup$ By the way your solution needs the continuity of $g$ which implies the continuity of $f''$. I will add that in my question. I don't know if this can be avoided but let's assume it to be so. $\endgroup$ – Paramanand Singh Mar 25 '14 at 13:48

The idea in Jochen's answer provides the following:

Generalization. Let $p(x)=x^n+a_{n-1}x^{n-1}+\cdots+a_1x+a_0$, and assume that all the roots of $p$ have negative real parts (which implies that $a_0\ne 0$). If $f: (b,\infty)\to\mathbb C$ is $n$ times differentiable and $$ \lim_{t\to\infty} f^{(n)}(t)+a_{n-1}f^{(n-1)}(t)+\cdots+a_0 f(t)=A, $$ then $$ \lim_{t\to\infty}f(t)=\frac{A}{a_0}. $$

  • $\begingroup$ Is it easy to prove ? $\endgroup$ – user119228 Apr 2 '14 at 20:17
  • 1
    $\begingroup$ Factoring $p$, and using Jochen's idea $n$ times. $\endgroup$ – Yiorgos S. Smyrlis Apr 2 '14 at 20:18
  • $\begingroup$ Thank you, it's a nice result. $\endgroup$ – user119228 Apr 2 '14 at 20:20
  • $\begingroup$ it's not positive real parts ? $\endgroup$ – user119228 Apr 2 '14 at 20:21
  • $\begingroup$ The real parts of the roots are negative. In the original exercises the roots are $-\frac{1}{2}\pm \frac{i\sqrt{3}}{2}$. $\endgroup$ – Yiorgos S. Smyrlis Apr 2 '14 at 20:23

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.