2

we define

$$f^{[n]}(x)=\lim_{h\to 0}\sum_{k=0}^{n}\binom{n}{k}\dfrac{(-1)^kf[(x+(n-2k)h]}{(2h)^n}$$ if $f(x)$ is continuous on $[a,b]$,and such $f^{[n]}(x)=0$ for all $x$.

prove or disprove :$f(x)$ is polynomial function,

My try: I find sometimes,and I find this following therom:

(Schwarz,1870):If $F$ is continuous and $DF(x)=0$ for all $x$, then $F$ is a linear function where $$DF(x):=\lim_{h\to 0}\dfrac{F(x-h)-2F(x)+F(x+h)}{h^2}$$ and this solution can see: http://math.depaul.edu/mash/newharder.pdf

I think my problem is true.But I can't prove it,Thank you

  • I think this is a duplicate : http://math.stackexchange.com/questions/248313/convergence-of-finite-differences-to-zero-and-polynomials/252916#252916 – Ewan Delanoy Oct 21 '13 at 02:41
  • No,I think this is different problem, – math110 Oct 21 '13 at 03:21
  • 2
    @EwanDelanoy That question involved a forward difference operator. This question is about symmetric difference. The counter example from the linked question does not converge to zero in this case. – WimC Jan 09 '14 at 12:25

1 Answers1

4

The statement in its current form is false for $n\geq 3$.

For odd $n\geq 3$ odd, $f(x)=\left |x\right |$ can serve as counterexample. This function is locally linear for $n\not =0$ and thus the corresponding limit is zero. For $x=0$, we can make use of the symmetry of binomial coefficients $\binom{n}{k} = \binom{n}{n-k}$ and the fact that for odd $n$, we have $(-1)^{n-k} = -(-1)^k$, to rewrite the limit as follows:

$$\lim_{h\rightarrow 0} \frac{1}{(2h)^n}\sum_{k=0}^{\lfloor n/2\rfloor} (-1)^k\binom{n}{k}\left[f((n-2k)h) - f((2k-n)h)\right]$$

Since we also have $f(x)=f(-x)$ for all $x$, the expression in square brackets is identically equal to zero; so is the sum and thus also the limit.

For $n\geq 4$ even, a counterexample is provided by $f(x)=x\cdot \left|x\right|$. This function is locally quadratic for $x\not =0$, so again, $x=0$ is the only interesting point. In this case, the symmetry of binomial coefficients and identity $f(-x) = -f(x)$, in addition to $f(0)=0$, are sufficient to show the limit to be equal to zero too.

Both counterexamples are clearly continuous and the latter is actually differentiable. In fact, the same approach can be used to obtain $(n-3)$-times differentiable counterexamples by considering $$f(x)=\left| x\right|\cdot x^{n-3}$$