Ma · Mathematics

When Does the Taylor Series for Logarithm Converge to the Logarithm?

Published By the 9to5Science editorial team

At a glance

  1. 1. Choose the expansion

    Write the logarithm as ln(1 + u), where u is the displacement from the expansion point.

  2. 2. Check the interval

    The series converges to ln(1 + u) for −1 < u < 1, and also converges at u = 1.

  3. 3. Inspect the endpoints

    At u = 1 it sums to ln 2; at u = −1 it diverges, matching the singularity of ln(1 + u).

  4. 4. Estimate the error

    Inside the interval, the alternating-series remainder gives a simple bound after truncation.

A Taylor series need not converge to its function throughout its interval of convergence, so the interval matters. For the Maclaurin series of ln⁡(1+u)\ln(1+u), the series gives the function for −1<u<1-1<u<1, and it also converges at the right endpoint u=1u=1. It diverges at u=−1u=-1.

This distinction matters because a function can be defined at a point where its Taylor series does not converge, and a convergent Taylor series is not automatically guaranteed to represent the function. Here, you can verify the equality directly.

The Taylor series for ln⁡(1+u)\ln(1+u)

The series is

ln⁡(1+u)=u−u22+u33−u44+⋯=∑n=1∞(−1)n+1unn.\ln(1+u)=u-\frac{u^2}{2}+\frac{u^3}{3}-\frac{u^4}{4}+\cdots =\sum_{n=1}^{\infty}(-1)^{n+1}\frac{u^n}{n}.

It is the Maclaurin series, meaning the Taylor series centered at u=0u=0. Its radius of convergence is 11. Thus, the power series converges absolutely when ∣u∣<1|u|<1, and it cannot converge for ∣u∣>1|u|>1. The two endpoints need separate checks.

A common equivalent form is obtained by setting x=1+ux=1+u:

ln⁡x=∑n=1∞(−1)n+1(x−1)nn.\ln x=\sum_{n=1}^{\infty}(-1)^{n+1}\frac{(x-1)^n}{n}.

This is the Taylor series for ln⁡x\ln x centered at x=1x=1. Its interior interval is 0<x<20<x<2, and it also converges at x=2x=2.

Why the series equals the logarithm

For ∣t∣<1|t|<1, the geometric series gives

11+t=1−t+t2−t3+⋯ .\frac{1}{1+t}=1-t+t^2-t^3+\cdots.

Integrate from 00 to uu. Since the antiderivative of 1/(1+t)1/(1+t) is ln⁡(1+t)\ln(1+t), integrating the terms gives

ln⁡(1+u)=u−u22+u33−⋯ .\ln(1+u)=u-\frac{u^2}{2}+\frac{u^3}{3}-\cdots.

For a more explicit check of convergence, the finite geometric identity is

11+t=∑k=0N−1(−t)k+(−t)N1+t.\frac{1}{1+t}=\sum_{k=0}^{N-1}(-t)^k+\frac{(-t)^N}{1+t}.

After integrating from 00 to uu, the difference between ln⁡(1+u)\ln(1+u) and the first NN terms is the integral of the remainder. If ∣u∣<1|u|<1, its absolute value is at most

∣u∣N+1(N+1)(1−∣u∣),\frac{|u|^{N+1}}{(N+1)(1-|u|)},

which tends to zero as NN increases. This proves that the partial sums approach the logarithm throughout the open interval −1<u<1-1<u<1.

What happens at the endpoints?

At u=1u=1, the series becomes the alternating harmonic series:

1−12+13−14+⋯ .1-\frac12+\frac13-\frac14+\cdots.

It converges conditionally, and its sum is ln⁡2\ln 2. One way to see the endpoint equality is to integrate the finite geometric identity above from 00 to 11. The integral of the remainder has absolute value at most 1/(N+1)1/(N+1), so it tends to zero. The integral of 1/(1+t)1/(1+t) from 00 to 11 is ln⁡2\ln 2.

At u=−1u=-1, the terms are all negative and the series becomes

−1−12−13−14−⋯ ,-1-\frac12-\frac13-\frac14-\cdots,

which diverges. Also, ln⁡(1+u)=ln⁡0\ln(1+u)=\ln 0 is undefined there. For ∣u∣>1|u|>1, the terms un/nu^n/n do not tend to zero, so the series diverges.

Value of uuSeries behaviorLogarithm
−1<u<1-1<u<1Converges to ln⁡(1+u)\ln(1+u)Defined
u=1u=1Converges to ln⁡2\ln 2Defined
u=−1u=-1DivergesUndefined
$u>1$

Worked example: approximating ln⁡(1.5)\ln(1.5)

Here, 1.5=1+0.51.5=1+0.5, so substitute u=0.5u=0.5:

ln⁡(1.5)=0.5−0.522+0.533−0.544+⋯ .\ln(1.5)=0.5-\frac{0.5^2}{2}+\frac{0.5^3}{3}-\frac{0.5^4}{4}+\cdots.

Using the first five terms gives

0.5−0.125+0.0416667−0.015625+0.00625≈0.4072917.0.5-0.125+0.0416667-0.015625+0.00625\approx 0.4072917.

The actual value is approximately 0.40546510.4054651. Since the terms decrease in magnitude and alternate in sign, the alternating-series error bound says the error after five terms is no larger than the next term:

0.566≈0.0026042.\frac{0.5^6}{6}\approx 0.0026042.

The actual difference is about 0.00182660.0018266, within that bound. More terms improve the approximation.

Common mistakes

  • Treating the radius as the whole interval. A radius of 11 guarantees convergence inside ∣u∣<1|u|<1 and divergence outside ∣u∣>1|u|>1. It does not settle either endpoint; test them separately.
  • Confusing the variable with the input to the logarithm. The series u−u2/2+u3/3−⋯u-u^2/2+u^3/3-\cdots represents ln⁡(1+u)\ln(1+u), not ln⁡u\ln u. For ln⁡x\ln x centered at 11, use u=x−1u=x-1.
  • Assuming every Taylor series represents its function everywhere the function exists. Taylor polynomials give local approximations, but the series must converge and its sum must be shown to equal the function. The geometric-series argument establishes that equality here.
  • Calling endpoint convergence absolute. At u=1u=1, the alternating harmonic series converges conditionally: its absolute-value series is the divergent harmonic series.

References

  1. Tom M. Apostol, Calculus, Vol. 1

Spotted an error? Send a correction. Read how we write and check articles in our editorial policy.