If you are not familiar with continuous functions or closed intervals, start here.
If you would like a review of derivatives, start here.
Theorem: Taylor's Formula
Let f(x) be a continuous function over the closed interval [a,b] that has an (n+1) derivative which is referenced by f(n+1)(x) where n is a positive integer.
Then:
f(b) = f(a) + f'(a)(b-a) + ... + [f(n)(a)/n!](b-a)n + [f(n+1)(z)/(n+1)!](b-a)n+1)
for some number z which lies between a and b.
Proof:
(1) Let H = f(b) - f(a) - f'(a)(b-a) - [f''(a)/2!](b-a)2 - ... - [f(n)(a)/n!](b-a)n
(2) Let K = H/(b-a)n+1
(3) Let g(x) be a function such that:
g(x) = f(b) - f(x) - f'(x)(b-x) -[f''(x)/2!](b-x)2 - ... - [f(n)(x)/n!](b-x)n - K(b-x)n+1
(4) We can see that g(x) is a continuous function since:
(a) f(x) is continuous over [a,b] by the given.
(b) f(n) is a continuous function since by the given we know that f(x) has an (n+1) derivative and since f(x) is differentiable at x, then it is continuous at x (see here)
(c) (b-x)n is continuous over [a,b] since f(x) is continuous since:
Let h(x)=b-x
h(x) is continuous since it is the addition of two continuous functions i(x)=b and j(x)=-x [See here for proof of the Addition Law]
(b-x)n is continuous by the Multiplication Law [See here for proof of the Multiplication Law]
(b-x)n = (b - x)*(b-x)*... since n is a positive integer.
(d) f(b) is continuous since it is a constant. (See here)
(e) Each f(n)(x)/n! is continuous since (-1/n!) can be thought of as a constant function h(x)=(-1/n!) and the multiplication of two continuous functions is itself continuous (see here)
(f) Finally, g(x) is continuous because the addition of a set of continuous functions is itself continuous (see here)
(5) We can see that g(a) = 0 since:
g(a) = H - H/(b-a)n+1*(b-a)n+1 = H - H = 0
(6) We can also see that g(b) = 0 since:
g(b) = f(b) - f(b) - f'(b)(b-b) - [f''(b)/2!](b-b)2 - ... - [f(n)(b)/n!](b-b)n - K(b-b)n+1
(7) By Rolle's Theorem, since g(x) is continuous on [a,b], we know that there exists a value z such that z ∈ [a,b] and g'(z) = 0. [See here for Rolle's Theorem]
(8) If we differentiate on g(x), we get:
g'(x) = -f'(x) + f'(x) -f(2)(x)(b-x) + f(2)(x)(b-x) - (1/2!)f(3)(x)(b-x)2 + (1/2!)f(3)(x)(b-x)2 - (1/3!)f(4)(x)(b-x)3 + ... + [1/(n-1)!]f(n)(x)(b-x)n-1 - (1/n!)f(n+1)(x)(b-x)n + (n+1)K(b-x)n since:
(a) g(x) = f(b) - f(x) - f'(x)(b-x) -[f''(x)/2!](b-x)2 - ... - [f(n)(x)/n!](b-x)n - K(b-x)n+1
(b) By Lemma 3 here, we can differentiate each individual product in the sum.
(c) f(b) is a constant so d/dx(f(b)) = 0 [See here for Constant Rule]
(d) d/dx(-f(x)) = -f'(x) [See here for details]
(e) d/dx[-f'(x)(b-x)] = -f''(x)(b-x) - f'(x)(-1) = f'(x) -f''(x)(b-x) [See here for the Product Rule]
(f) d/dx[(-f''(x)/2!)(b-x)2] = (-f(3)(x)/2!)(b-x)2 + [-f''(x)/2!](2)(b-x)(-1)] =
= (-f(3)/2!)(b-x)2 + f''(x)(b-x) [See here for Generalized Power Rule]
(g) d/dx([-f(n)(x)/n!](b-x)n) = (-f(n+1)(x)/n!)(b-x)n + [-f(n)(n)/n!](n)(b-x)n-1*(-1) =
= (-f(n+1)(x)/n!)(b-x)n +[ f(n)/(n-1)!](b-x)n-1
(h) Since K is a constant
d/dx(-K(b-x)n+1) = (n+1)(-K)(b-x)n(-1) = (n+1)K(b-x)n
(9) We see that most of the terms cancel out so that we get:
g'(x) = (n+1)K(b-x)n - (1/n!)f(n+1)(x)(b-x)n
(10) Applying the fact that g'(z) = 0 from step #7 gives us:
g'(z) = (n+1)K(b-z)n - (1/n!)f(n+1)(z)(b-z)n = 0
(11) We can divide both sides by (b-z)n to get:
(n+1)K - (1/n!)f(n+1)(z) = 0
(12) Now we can rearrange (#11) to get:
K = [(1/n!)f(n+1)(z)]/(n+1) = [f(n+1)(z)]/(n+1)!
(13) Now, using (#12) and (#3) with x=a, we get:
g(a) = 0 = f(b) - f(a) - f'(a)(b-a) - [f''(a)/2!](b-a)2 - ... - [f(n)(a)/n!](b-a)n - [(b-a)n+1][f(n+1)(z)/(n+1)!]
(14) Now, if we move over all the elements after f(b), we get:
f(b) = f(a) + f'(a)(b-a) + [f''(a)/2!](b-a)2 + ... + [f(n)(a)/n!](b-a)n + [(b-a)n+1][f(n+1)(z)/(n+1)!]
QED
References
- Edwards & Penny, Calculus and Analytic Geometry
A question (rather a dumb one, I guess), what is the need for proving g(x) is a continuous function?
ReplyDeleteHi Jagadeesh,
ReplyDeleteWe need to show that g(x) is a continuous function in order to apply Rolle's Theorem which only applies to continuous functions.
-Larry
Wow, thanks. I still don't get it really good, but at least I was able to read through the proof.
ReplyDeleteBit of nitpicking about a part that confused me:
(8) (e):
(e) d/dx[-f'(x)(b-x)] = -f'''(x)(b-x) - f'(x)(-1) = f'(x) -f''(x)(b-x) [See here for the Product Rule]
You accidentally have written the third derivative of f(x) instead of second:
It should be:
(e) d/dx[-f'(x)(b-x)] = -f''(x)(b-x) - f'(x)(-1) = f'(x) -f''(x)(b-x) [See here for the Product Rule]
Hi Rauni,
ReplyDeleteThanks for your comment. I'll try to simplify the explanation of this proof at a future time.
In the mean time, I've fixed the typo you found (nitpicking encouraged) :-)
-Larry
This is a nice proof, thanks. I see a minor typo on 8a, the "n+1" at the end should be an exponent.
ReplyDeleteThis is difficult to understand. The proof given in wikipedia was simpler.
ReplyDeleteA silly question but how can one know that imaginary numbers can be applied in lemma 2? Is there a way of proving it?
ReplyDeleteAnyway your blog is amazing. just amazing.