Skip to content
Topos Uranos

Taylor series of the elementary functions

Summary

The lesson on the Taylor polynomial approximated a function near a point by a polynomial whose error is small compared with (x−a)n(x - a)^n, and the lesson on the Taylor remainder bounded that error; the lesson on power series proved that the sum of a power series can be differentiated and integrated term by term, and that its coefficients are its derivatives at the center divided by the factorials. This lesson brings the two lines together. An infinitely differentiable function has a Taylor series, whose partial sums are its Taylor polynomials, and the series converges to the function at a point if and only if the Taylor remainder tends to 00 at that point. With the bound for the remainder, the exponential, the sine and the cosine are the sum of their Taylor series on the whole line; with term-by-term integration and Abel’s theorem, the logarithm of 1+x1 + x is too, on (−1,1]\ioc{-1}{1}, and the arctangent, on [−1,1]\icc{-1}{1}; with term-by-term differentiation and an equation that only the power satisfies, the binomial series gives (1+x)α(1 + x)^{\alpha} on (−1,1)\ioo{-1}{1}, and also at the endpoints if α>0\alpha > 0. The Taylor series does not always represent the function: the function e−1/x2e^{-1/x^2}, completed with the value 00 at the origin, is infinitely differentiable, all its derivatives at 00 are zero, and its Taylor series, identically zero, coincides with it only at the origin.

Learning objectives

  1. Form the Taylor series of an infinitely differentiable function, distinguish the Taylor remainder from the remainder of a series, and prove that the series represents the function exactly where the Taylor remainder tends to 00.
  2. Prove the series for the exponential, the sine and the cosine on the whole line, and those for the logarithm and the arctangent with their endpoints.
  3. Prove the binomial series by term-by-term differentiation, and decide its convergence at the endpoints of the interval.
  4. Recognize, with the function e−1/x2e^{-1/x^2}, that an infinitely differentiable function may fail to be the sum of its Taylor series, and obtain new series from the known ones.

The Taylor series of a function

The lesson on power series proved, in the corollary on higher-order derivatives and uniqueness of the coefficients, that the sum SS of a power series with positive radius has derivatives of all orders and that its coefficients are the numbers S(k)(a)k!\frac{S^{(k)}(a)}{k!}. It also announced that this lesson would travel the reverse path: start from a function with derivatives of all orders, form the series with those coefficients and study when its sum is the function. A function ff is infinitely differentiable on an open interval II if for every natural number nn it is nn times differentiable on II.

DefinitionTaylor series

Let ff be an infinitely differentiable function on an open interval II and aa a point of II. The Taylor series of ff at aa is the following power series with center aa:

∑n=0∞f(n)(a)n!(x−a)n\sum_{n=0}^{\infty} \frac{f^{(n)}(a)}{n!} (x - a)^n

If a=0a = 0, it is also called the Maclaurin series of ff.

The partial sum of index nn of the Taylor series is the Taylor polynomial Pn,a(x)P_{n,a}(x) of the lesson on the Taylor polynomial, and the difference f(x)−Pn,a(x)f(x) - P_{n,a}(x) is the Taylor remainder Rn,a(x)R_{n,a}(x), defined at every xx of II, whether or not the series converges. The lesson on power series called Rn(x)=S(x)−Sn(x)R_n(x) = S(x) - S_n(x) the remainder of a series, which is defined only where the series converges and compares the partial sums with the sum S(x)S(x) of the series, not with f(x)f(x); the two remainders coincide at the points where the series converges with sum f(x)f(x), and only there. To avoid confusing them, in this lesson the Taylor remainder is always written with its two indices, also when a=0a = 0, as Rn,0(x)R_{n,0}(x), and the remainder of a series is written as the difference S(x)−Sn(x)S(x) - S_n(x), with no letter of its own.

PropositionRepresentation of a function by its Taylor series

Let ff be an infinitely differentiable function on an open interval II, and aa and xx points of II. The Taylor series of ff at aa converges at xx with sum f(x)f(x) if and only if Rn,a(x)→0R_{n,a}(x) \to 0 as n→∞n \to \infty. The difference between the function and each partial sum of its Taylor series is, by definition, the Taylor remainder.

Proof

  1. Sn(x)=∑k=0nf(k)(a)k!(x−a)k=Pn,a(x)→f(x)−Sn(x)=Rn,a(x)S_n(x) = \sum_{k=0}^{n} \frac{f^{(k)}(a)}{k!} (x - a)^k = P_{n,a}(x) \rightarrow \dato{dif}{f(x) - S_n(x) = R_{n,a}(x)}

    The partial sums of the Taylor series at xx are the Taylor polynomials, and by the definition of the Taylor polynomial and remainder, from the lesson on the Taylor polynomial, their difference from f(x)f(x) is the Taylor remainder.

  2. Sn(x)→f(x)⇔Rn,a(x)→0S_n(x) \to f(x) \Leftrightarrow \resaltar{R_{n,a}(x) \to 0}

    With xx fixed, f(x)f(x) is a number. By the definition of series, partial sums and sum, from the lesson on numerical series and their convergence, the series converges with sum f(x)f(x) if and only if Sn(x)→f(x)S_n(x) \to f(x), and by the algebra of limits, from the theorem stating that the limit theorems hold for sequences, from the lesson on sequences and their limit, this is equivalent to the difference tending to 00.

Conversely, a function that is the sum of a power series near a point is infinitely differentiable there, and the series is its Taylor series.

CorollaryThe power series of a function is its Taylor series

Let ff be a function defined on an open interval II, aa a point of II and δ>0\delta > 0 a number with (a−δ,a+δ)⊆I\ioo{a - \delta}{a + \delta} \subseteq I, and suppose that, for every xx of (a−δ,a+δ)\ioo{a - \delta}{a + \delta}, the series ∑n=0∞cn(x−a)n\sum_{n=0}^{\infty} c_n (x - a)^n converges with sum f(x)f(x). Then ff is infinitely differentiable on (a−δ,a+δ)\ioo{a - \delta}{a + \delta}, and for every integer n≥0n \geq 0:

cn=f(n)(a)n!c_n = \frac{f^{(n)}(a)}{n!}

so that the series is the Taylor series of ff at aa, and converges to ff on that interval. The function coincides with the sum of the series on an open interval, and the derivative at a point depends only on the nearby values.

Proof

  1. ρ≥δ>0\resaltar{\rho \geq \delta > 0}

    Let ρ\rho be the radius of convergence of the series. If ρ<δ\rho < \delta held, by the theorem on the existence of the radius of convergence, from the lesson on power series, the series would diverge at x=a+ρ+δ2x = a + \frac{\rho + \delta}{2}, which lies in (a−δ,a+δ)\ioo{a - \delta}{a + \delta}, contrary to the hypothesis.

  2. ck=S(k)(a)k!\dato{coef}{c_k = \frac{S^{(k)}(a)}{k!}}

    Let SS be the sum of the series. By the corollary on higher-order derivatives and uniqueness of the coefficients, from the lesson on power series, SS has derivatives of all orders on the interval of convergence, which contains (a−δ,a+δ)\ioo{a - \delta}{a + \delta}, and ck=S(k)(a)k!c_k = \frac{S^{(k)}(a)}{k!}.

  3. f(k)(x)=S(k)(x)→ck=f(k)(a)k!f^{(k)}(x) = S^{(k)}(x) \rightarrow \resaltar{c_k = \frac{f^{(k)}(a)}{k!}}

    On (a−δ,a+δ)\ioo{a - \delta}{a + \delta} the functions ff and SS are equal. By the lemma on the local character of the derivative, from the lesson on differentiation rules, applied by induction on kk to the derivatives of order kk on that open interval, ff has on it derivatives of all orders, equal to those of SS.

The proposition on representation reduces the problem to deciding whether the Taylor remainder tends to 00, and the corollary on the bound for the Taylor remainder, from the lesson on the Taylor remainder and its applications, bounds it with a factorial in the denominator. One limit is still needed, which that lesson did not prove.

LemmaPowers divided by the factorial

For every real number yy, ynn!→0\frac{y^n}{n!} \to 0 as n→∞n \to \infty. On passing from nn to n+1n + 1, the term is multiplied by yn+1\frac{y}{n + 1}, which eventually becomes small.

Proof

  1. A=∣y∣+2>1→Ann!→0A = \abs{y} + 2 > 1 \rightarrow \dato{ord}{\frac{A^n}{n!} \to 0}

    Let A=∣y∣+2A = \abs{y} + 2, greater than 11 also when y=0y = 0. By the theorem on orders of growth, from the lesson on sequences, functions and uniform continuity, An≪n!A^n \ll n!, that is, the quotient tends to 00.

  2. −Ann!≤ynn!≤Ann!→ynn!→0-\frac{A^n}{n!} \leq \frac{y^n}{n!} \leq \frac{A^n}{n!} \rightarrow \resaltar{\frac{y^n}{n!} \to 0}

    Since 0≤∣y∣<A0 \leq \abs{y} < A, by induction on nn, multiplying inequalities between nonnegative numbers, ∣y∣n≤An\abs{y}^n \leq A^n. The squeeze theorem, from the theorem stating that the limit theorems hold for sequences, gives the limit.

TheoremBounded derivatives criterion

Let ff be an infinitely differentiable function on an open interval II, aa and xx points of II, and M≥0M \geq 0 and B≥0B \geq 0 two numbers such that, for every natural number nn and every tt strictly between aa and xx:

∣f(n)(t)∣≤MBn\abs{f^{(n)}(t)} \leq M B^n

Then Rn,a(x)→0R_{n,a}(x) \to 0, and the Taylor series of ff at aa converges at xx with sum f(x)f(x). The derivatives may grow with their order, but not faster than a geometric progression, and the factorial in the denominator of the remainder prevails.

Proof

  1. ∣Rn,a(x)∣≤MBn+1(n+1)!∣x−a∣n+1=M (B∣x−a∣)n+1(n+1)!\dato{cota}{\abs{R_{n,a}(x)} \leq \frac{M B^{n + 1}}{(n + 1)!} \abs{x - a}^{n + 1} = M \, \frac{\left(B \abs{x - a}\right)^{n + 1}}{(n + 1)!}}

    If x=ax = a, for every nn we have Pn,a(a)=f(a)P_{n,a}(a) = f(a) and Rn,a(a)=0R_{n,a}(a) = 0. Let x≠ax \neq a, and let n≥0n \geq 0 be an integer. The function is n+1n + 1 times differentiable on II, and the corollary on the bound for the Taylor remainder, with MBn+1M B^{n + 1} in place of its MM, bounds the remainder.

  2. −M (B∣x−a∣)n+1(n+1)!≤Rn,a(x)≤M (B∣x−a∣)n+1(n+1)!→Rn,a(x)→0-M \, \frac{\left(B \abs{x - a}\right)^{n + 1}}{(n + 1)!} \leq R_{n,a}(x) \leq M \, \frac{\left(B \abs{x - a}\right)^{n + 1}}{(n + 1)!} \rightarrow \resaltar{R_{n,a}(x) \to 0}

    By the lemma on powers divided by the factorial, with y=B∣x−a∣y = B \abs{x - a}, and the lemma on shifting the indices and changing finitely many terms, from the lesson on sequences and their limit, the last quotient tends to 00; the algebra of limits and the squeeze theorem give Rn,a(x)→0R_{n,a}(x) \to 0, and the proposition on the representation of a function by its Taylor series, the conclusion.