Taylor Series and Approximation
A power series is like a polynomial that never ends. Taylor series rewrite functions such as e^x and sin x as power series, which is how computers and calculators evaluate them.
Power series
A power series centred at \(a\) has the form \(\sum c_n (x - a)^n = c_0 + c_1(x - a) + c_2(x - a)^2 + \cdots\). For each value of \(x\), it is an ordinary series that may converge or diverge. So a power series defines a function wherever it converges. The geometric series \(\sum x^n = \frac{1}{1 - x}\) for \(|x| < 1\) is the simplest example.
\sum_{n=0}^{\infty} c_n (x-a)^nRadius and interval of convergence
Every power series converges in a symmetric interval around its centre: there is a radius \(R\) such that it converges for \(|x - a| < R\) and diverges for \(|x - a| > R\). \(R\) can be \(0\), a positive number, or infinite. Find it with the ratio test, then check the two endpoints separately, because the test is silent there.
Worked example: finding the radius
For \(\sum \frac{x^n}{n\,2^n}\), the ratio test gives \(\left|\frac{x}{2}\right|\frac{n}{n+1} \to \frac{|x|}{2}\), so it converges when \(|x| < 2\): \(R = 2\). At \(x = 2\) the series becomes \(\sum \frac{1}{n}\), which diverges; at \(x = -2\) it becomes the alternating harmonic series, which converges. The interval is \([-2, 2)\).
Calculus with power series
Inside its radius, a power series can be differentiated and integrated term by term, and the radius stays the same. Starting from \(\frac{1}{1 - x} = \sum x^n\): differentiating gives \(\frac{1}{(1-x)^2} = \sum n x^{n-1}\). Replacing \(x\) with \(-x\) and integrating gives \(\ln(1 + x) = x - \frac{x^2}{2} + \frac{x^3}{3} - \cdots\) for \(|x| < 1\).
Taylor series
If a function has a power series at \(a\), its coefficients must be \(c_n = \frac{f^{(n)}(a)}{n!}\), where \(f^{(n)}\) is the \(n\)th derivative. This series is the Taylor series of \(f\) at \(a\). Centred at \(a = 0\) it is called a Maclaurin series. The formula matches the function's value, slope, concavity, and every higher derivative at the centre.
f(x) = \sum_{n=0}^{\infty} \frac{f^{(n)}(a)}{n!}(x-a)^nThe essential Maclaurin series
\(e^x = 1 + x + \frac{x^2}{2!} + \frac{x^3}{3!} + \cdots\) for all \(x\). \(\sin x = x - \frac{x^3}{3!} + \frac{x^5}{5!} - \cdots\) and \(\cos x = 1 - \frac{x^2}{2!} + \frac{x^4}{4!} - \cdots\) for all \(x\). \(\frac{1}{1 - x} = 1 + x + x^2 + \cdots\) for \(|x| < 1\). Memorise these; most others can be built from them.
e^x = \sum_{n=0}^{\infty}\frac{x^n}{n!}Building new series from old ones
Substitute, multiply, differentiate, or integrate known series instead of computing derivatives. For \(e^{-x^2}\), replace \(x\) by \(-x^2\) in the series for \(e^x\): \(1 - x^2 + \frac{x^4}{2!} - \frac{x^6}{3!} + \cdots\). Integrating that term by term gives a series for \(\int_0^1 e^{-x^2}\,dx\), an integral with no elementary antiderivative that is central to statistics.
Taylor polynomials and approximation
Stopping a Taylor series after the \(x^n\) term gives the Taylor polynomial \(T_n(x)\), the best degree-\(n\) polynomial approximation near the centre. \(T_1\) is the tangent line. For small \(x\), \(\sin x \approx x\) and \(\cos x \approx 1 - \frac{x^2}{2}\), approximations physicists use constantly, for instance for a pendulum's small swings.
How good is the approximation?
Taylor's inequality bounds the error: if \(|f^{(n+1)}(x)| \le M\) near the centre, then the error \(|f(x) - T_n(x)| \le \frac{M}{(n+1)!}|x - a|^{n+1}\). For alternating series, the error is at most the first omitted term. To estimate \(e^{0.1}\) with \(T_2 = 1 + 0.1 + 0.005 = 1.105\), the error is under about \(0.0002\); the true value is \(1.10517\ldots\)
|R_n(x)| \le \frac{M}{(n+1)!}|x-a|^{n+1}Worked example: a Taylor series at another centre
Find the Taylor series of \(\ln x\) at \(a = 1\). The derivatives at 1 are \(f(1) = 0\), \(f'(1) = 1\), \(f''(1) = -1\), \(f'''(1) = 2\), and in general \(f^{(n)}(1) = (-1)^{n-1}(n - 1)!\). So \(\ln x = (x - 1) - \frac{(x-1)^2}{2} + \frac{(x-1)^3}{3} - \cdots\), valid for \(0 < x \le 2\).
Common mistakes
Forgetting the factorial in the Taylor coefficient. Forgetting to check endpoints of the interval of convergence. Assuming a Taylor series converges everywhere. Mixing up the series for \(\sin\) and \(\cos\): sine has only odd powers, cosine only even. Losing the alternating signs.
Try it yourself
Find the first three nonzero terms of the Maclaurin series for \(x\cos x\). Answer: multiply the cosine series by \(x\): \(x - \frac{x^3}{2} + \frac{x^5}{24}\). Then find the radius of convergence of \(\sum \frac{x^n}{n!}\). Answer: infinite; the ratio is \(\frac{|x|}{n + 1} \to 0\) for every \(x\).
Key takeaways
Power series converge within a radius \(R\), with endpoints checked separately. They can be differentiated and integrated term by term. Taylor coefficients are \(\frac{f^{(n)}(a)}{n!}\). Learn the series for \(e^x\), \(\sin x\), \(\cos x\), and \(\frac{1}{1-x}\), and build others from them. Taylor polynomials approximate functions, with error bounds from Taylor's inequality.
Flashcard
Power series centred at a
Check your understanding
What is the Maclaurin series for e^x, up to the x^3 term?
