I'm reading Mathematics: It's Content, Methods, and Meanings and I am in a chapter about Taylor Series. It made sense until I came across the remainder part of the theorem. In order to prove the theorem it uses ${f(x)-P_n(x) \over -(x-a)^{n+1}}$ to be able to use Cauchy's Mean Value theorem to prove the Langrange form of the remainder. Could someone explain why they choose $-(x-a)^{n+1}$. Is it to make the proof easier or is there an underlying meaning? If so please explain why. Also, if it is just to make the proof easier is there a better (more intuitive) alternative proof?



EDIT: Here is part of a proof from http://davidlowryduda.com/?p=1804 Proof


Two things to say here: firstly, $G$ as defined in the proof you give has $G'(c) = -(n+1)(x-c)^{n}$, so the minus signs cancel out, and you have $$ R_n = (x-a)^{n+1} \frac{F'(c)}{(n+1)(x-c)^n}, $$ and presumably the rest of the proof is observing that $F'(c) = (x-c)^n f^{(n+1)}(c)/n!$, giving $$ R_n = \frac{(x-a)^{n+1}}{(n+1)!} f^{(n+1)}(c) .$$ Basically you have to choose it this way if you work with the polynomial $F$ with $(x-t)$ in it.

This is essentially the only way to prove Taylor's theorem with the mean-value remainders (I explained in my answer to a previous question of yours how to derive the integral form). Suppose we use the same $F$. Then the expression we have for the remainder is $$ R_n = \frac{(x-c)^n f^{(n+1)}(c)}{n!} \frac{G(x)-G(a)}{G'(c)}. $$ This is quite general, and you can get a variety of remainders from using different functions. The most basic thing you might try is $G(t) = At+B$. Then $$ G'(c) = A, \quad G(x)-G(a) = A(x-a), $$ so whatever linear function you use, you find $$ R_n = \frac{(x-c)^n f^{(n+1)}(c)}{n!} (x-a) $$ (this is called the Cauchy form of the remainder, in fact). Suppose now we try a quadratic polynomial. The argument above shows that we can just use a monic polynomial, and the constant is irrelevant, and we know about completing the square, so we can try $G(x) = (A-t)^2$ as the most general. $$ G'(c) = -2(A-t), \quad G(x)-G(a) = (x-a)(x+a-2A), $$ and we get $$ R_n = \frac{(x-c)^n f^{(n+1)}(c)}{n!} \frac{(x-a)(2A-x-a)}{2(A-c)}. $$ Well, this is not a very nice expression: but if we set $A=x$, the numerator simplifies to $(x-a)^2$ and the denominator cancels an $x-c$ from the $F'(c)$. You can do similar things with other degrees of polynomial: in general, you can consider $$ G(t) = \sum_{k=0}^{N} A_k (x-t)^k, $$ but the simplest one to take is $(x-t)^k$, which gives you $$ R_n = \frac{(x-c)^{n-k+1} f^{(n+1)}(c)}{n!k} (x-a)^k; $$ finally, taking $k=n+1$ gives the Lagrange form of the remainder, $$ R_n = \frac{f^{(n+1)}(c)}{(n+1)!} (x-a)^{n+1} $$

(Also, from the above discussion we conclude that $G(t) = -(x-t)^n$ would also have been a perfectly legitimate choice for the Lagrange form of the remainder.)

  • $\begingroup$ Nicely done. +1 $\endgroup$
    – Simon S
    May 22 '15 at 14:58

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.