Show that the centered finite difference approximation for the first derivative of a function on a uniform mesh yields the exact derivative for any quadratic polynomial $P_2(x) = a+bx+cx^2$

Using the taylor series:

\begin{align} f(x+h) &= f(x) +f'(x)h + f''(x) \frac{h^2}{2}+f'''(x)\frac{h^3}{6}+ \dots \\ f(x-h) &= f(x) -f'(x)h + f''(x) \frac{h^2}{2}-f'''(x)\frac{h^3}{6}+ \dots \end{align}

From the first equation, we get the forward difference approximation: $$f'(x) = \frac{f(x+h)-f(x)}{h}+O(h).$$

From the second equation, we get the backward difference approximation $$f'(x) = \frac{f(x)-f(x-h)}{h}+O(h),$$

and if we subtract the second equation from the first equation we get $$f'(x) = \frac{f(x+h)-f(x-h)}{2h}+O(h^2),$$

which is the central difference formula.

But how do I use this to solve my problem? Is this formula the same as the centered finite difference approximation mentioned in the question?

up vote 3 down vote accepted

You've well applied Taylor's expansion.

\begin{align} f(x+h) &= f(x) +f'(x)h + f''(x) \frac{h^2}{2}+f'''(x)\frac{h^3}{6}+ o(h^3) \tag1 \label1 \\ f(x-h) &= f(x) -f'(x)h + f''(x) \frac{h^2}{2}-f'''(x)\frac{h^3}{6}+ o(h^3) \tag2 \label2 \end{align}

$$\frac{\eqref{1} - \eqref{2}}{2} : \frac{f(x+h)-f(x-h)}{2} = f'(x)h + f'''(x)\frac{h^3}{6} + o(h^3) \tag{3}\label{3}$$

When the above equality is applied to a quadratic polynomial, we have $f^{(n)} \equiv 0$ for $n \ge 3$, so only the first derivative remains. This should give you the desired result

$$f'(x) = \frac{f(x+h)-f(x-h)}{2h}. \tag{4}\label{4}$$


(Edit in response to OP's request)

Apply given formula $$f'(x) = \frac{f(x+h)-f(x-h)}{2h}+O(h^2) \tag5\label5$$ to $P_2(x)$.

$$\begin{aligned} P_2'(x) &= \frac{P_2(x+h)-P_2(x-h)}{2h}+O(h^2) \\ &= \frac{(a-a)+b[(x+h)-(x-h)] + c[(x+h)^2-(x-h)^2]}{2h}+O(h^2) \\ &= \frac{b(2h) + c[(x+h)+(x-h)][(x+h)-(x-h)]}{2h}+O(h^2) \\ &= \frac{2bh + c(2x)(2h)}{2h}+O(h^2) \\ &= b+2cx + O(h^2) \end{aligned} \tag6\label6$$

Since \eqref{5} actually comes from dividing \eqref{3} by $2h$, and the third and higher derivatives in \eqref{3} vanishes when $f = P_2$, therefore, we can discard $O(h^2)$ in \eqref{6} and conclude that the central finite difference approximation for the first derivative is exact when $f = P_2$.


  • When $O(h)$ is used to describe a particular function, it's used as an object. In symbols: $f = O(h)$.
  • When $O(h)$ is used as a subject, it describes a class of functions bounded by $Mh$ for a certain constant $M>0$.

When $P_{2}'(x) = b+2cx +O(h^2)$, $$P_{2}'(x) - (b+2cx) = (b+2cx +O(h^2)) - (b+2cx) = O(h^{2}).$$ We won't say $O(h^2) = 0$ in general. Instead, we need to make use of the given conditions to conclude further.

  • is this a complete solution? thanks for your post btw! I had an extra x on the ax in my question – fr14 Nov 8 at 20:08
  • 1
    @fr14 I've just read the statement to be proved above. Daniel Moraes has completed his solution quicker than I do, so there's no point to duplicate the effort. – GNUSupporter 8964民主女神 地下教會 Nov 8 at 20:18
  • okay and his looks right in your opinion? – fr14 Nov 8 at 20:23
  • @fr14 I agree with his solution except the last equality $$O(h^2) = 0.$$ – GNUSupporter 8964民主女神 地下教會 Nov 8 at 20:29
  • could you complete your solution? – fr14 Nov 8 at 20:31

This is the centered finite difference approximation method, basically you need to apply it to the general polynomial:
$f′(x)=\dfrac{f(x+h)−f(x−h)}{2h}+O(h^2)$ is what you already got

$P_{2}'(x)=\dfrac{(a+b(x+h)+c(x+h)^2) -(a+b(x-h)+c(x-h)^2)}{2h} +O(h^2)$

$P_{2}'(x)=\dfrac{a(1-1)+b(x+h-x+h)+c(x^2+2hx+h^2-x^2+2xh-h^2)}{2h} +O(h^2)$

$P_{2}'(x)=\dfrac{b(2xh)+c(4xh)}{2h} +O(h^2)$

$P_{2}'(x)=b+2cx +O(h^2)$
Hence :
$P_{2}'(x) - (b+2cx) = b+2cx - b-2cx = 0 = O(h^{2})$, here we stating that the error goes to 0, it is also possible to see that $P_{2}''(x+h)-P_{2}''(x+h)=0$ and the higher order derivatives are $0$, so this statement is equivalent

  • sorry yes I edited the question, could you show me how this is done? – fr14 Nov 8 at 20:04
  • ok, will open a lit bit more – Daniel Moraes Nov 8 at 20:04
  • if you could include all the details I can accept it! sorry for the typo – fr14 Nov 8 at 20:05
  • i rewrited it, if you don't get a step i can exaplin it further – Daniel Moraes Nov 8 at 20:14
  • 1
    i see no need, but i added it – Daniel Moraes Nov 8 at 20:44

Your Answer

 

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.