Ever tried to pull a power series out of thin air?
Most of us start with the classic geometric series, then stare at a nasty function and wonder, “How do I get a series for that?” The shortcut most textbooks skip is: differentiate. A single derivative can turn a boring, known series into the exact expansion you need.
Below is the low‑down on using differentiation to find a power‑series representation for pretty much any elementary function. Practically speaking, i’ll walk you through the why, the how, and the pitfalls that trip up even seasoned students. Grab a coffee, and let’s turn calculus into a series‑crafting workshop.
What Is Using Differentiation to Find a Power Series?
In plain English, you start with a power series you already know—usually something simple like
[ \frac{1}{1-x}=1+x+x^{2}+x^{3}+\dots\quad(|x|<1) ]
—then you differentiate (or integrate) term‑by‑term. Because differentiation is linear and works term‑wise inside its radius of convergence, the new series you get is the power‑series expansion of the derivative (or integral) of the original function.
So if you need a series for (\frac{1}{(1-x)^{2}}), you just differentiate the geometric series once:
[ \frac{d}{dx}\Bigl(\frac{1}{1-x}\Bigr)=\frac{1}{(1-x)^{2}} ] [ \frac{d}{dx}\Bigl(1+x+x^{2}+x^{3}+\dots\Bigr)=1+2x+3x^{2}+4x^{3}+\dots ]
That’s the whole trick. The same idea works for trig, logarithmic, exponential, and even hyperbolic functions—provided you have a seed series to start from That's the whole idea..
Why It Matters / Why People Care
Power series are the Swiss army knife of analysis. They let you:
- Approximate functions near a point (think Taylor polynomials).
- Solve differential equations term‑by‑term.
- Perform integration and summation that would otherwise be impossible.
But the “real talk” is that many textbooks give you the final series and skip the derivation. When you need a series for a function that isn’t in the standard table, you’re stuck. Knowing how to differentiate a known series gives you a systematic, repeatable method Easy to understand, harder to ignore..
- Speed – You can generate new series in seconds instead of looking them up.
- Understanding – You see how coefficients evolve, which helps when you later need to estimate error.
- Flexibility – The method works for any function that can be expressed as a derivative or integral of a known series.
In practice, this technique is the backbone of many physics and engineering calculations, from perturbation theory to signal processing Small thing, real impact..
How It Works
Below is the step‑by‑step recipe, plus a handful of concrete examples. Feel free to skim the theory and jump right into the examples if you’re in a hurry.
1. Choose a “base” series you already trust
The geometric series is the most common launchpad, but you can also start from:
- (\displaystyle e^{x}= \sum_{n=0}^{\infty}\frac{x^{n}}{n!})
- (\displaystyle \sin x = \sum_{n=0}^{\infty}(-1)^{n}\frac{x^{2n+1}}{(2n+1)!})
- (\displaystyle \ln(1+x)=\sum_{n=1}^{\infty}(-1)^{n+1}\frac{x^{n}}{n}) (for (|x|<1))
Pick the one whose derivative (or integral) gets you closest to the target function.
2. Differentiate (or integrate) term‑by‑term
Because power series converge absolutely inside their radius, you can safely move the derivative inside the summation:
[ \frac{d}{dx}\Bigl(\sum a_{n}x^{n}\Bigr)=\sum a_{n}n x^{n-1} ]
If you need an integral instead, just integrate each term:
[ \int\Bigl(\sum a_{n}x^{n}\Bigr)dx = C+\sum a_{n}\frac{x^{n+1}}{n+1} ]
3. Simplify the new coefficients
After differentiation you’ll often have a factor of (n) or (n+1) floating around. Write the series in the standard (\sum c_{k}x^{k}) format by re‑indexing if necessary.
Example: Starting from (\frac{1}{1-x}= \sum_{n=0}^{\infty}x^{n}), differentiate:
[ \frac{d}{dx}\bigl(\sum_{n=0}^{\infty}x^{n}\bigr)=\sum_{n=1}^{\infty}n x^{n-1} ]
Now let (k=n-1). Then (n=k+1) and the series becomes
[ \sum_{k=0}^{\infty}(k+1)x^{k}=1+2x+3x^{2}+4x^{3}+\dots ]
That’s the series for (\frac{1}{(1-x)^{2}}).
4. Check the radius of convergence
Differentiation does not change the radius of convergence, but you should still note it. For the geometric series we stay inside (|x|<1). If you start from a series with a different radius, keep that bound in mind.
5. Verify by plugging in a test value
A quick sanity check: pick a small (x) (say (x=0.Because of that, 1)) and compute a few partial sums. But compare with the actual function value using a calculator. If they line up, you’ve probably done it right.
Worked Examples
Example 1 – Series for (\displaystyle \frac{1}{(1-x)^{3}})
- Base: (\displaystyle \frac{1}{1-x}= \sum_{n=0}^{\infty}x^{n}).
- Differentiate once → (\frac{1}{(1-x)^{2}} = \sum_{n=0}^{\infty}(n+1)x^{n}).
- Differentiate again → (\displaystyle \frac{2}{(1-x)^{3}} = \sum_{n=0}^{\infty}(n+1)n x^{n-1}).
- Multiply both sides by (\frac{1}{2}) and re‑index:
[ \frac{1}{(1-x)^{3}} = \sum_{n=0}^{\infty}\frac{(n+2)(n+1)}{2},x^{n} ]
So the coefficients follow the triangular numbers pattern That alone is useful..
Example 2 – Series for (\displaystyle \ln(1-x))
We know (\displaystyle \frac{1}{1-x}= \sum_{n=0}^{\infty}x^{n}). Integrate term‑by‑term:
[ \int\frac{1}{1-x},dx = -\ln|1-x| = \sum_{n=0}^{\infty}\frac{x^{n+1}}{n+1}+C ]
Set (C=0) (since (\ln 1 =0)) and replace (x) with (-x) to get the familiar series:
[ \ln(1-x) = -\sum_{n=1}^{\infty}\frac{x^{n}}{n},\qquad |x|<1 ]
Notice how integration, the sibling of differentiation, gave us the logarithmic series in a single stroke.
Example 3 – Series for (\displaystyle \tan^{-1}x)
Start with (\displaystyle \frac{1}{1+x^{2}} = \sum_{n=0}^{\infty}(-1)^{n}x^{2n}) (geometric with (x^{2}) and a minus sign). Integrate:
[ \int\frac{1}{1+x^{2}}dx = \tan^{-1}x = \sum_{n=0}^{\infty}(-1)^{n}\frac{x^{2n+1}}{2n+1} ]
That’s the classic arctangent series, valid for (|x| \le 1) (with endpoint convergence handled separately).
Example 4 – Series for (\displaystyle \frac{x}{(1-x)^{2}})
You can get this by differentiating the geometric series and then multiplying by (x):
[ \frac{1}{(1-x)^{2}} = \sum_{n=0}^{\infty}(n+1)x^{n} ] [ \frac{x}{(1-x)^{2}} = \sum_{n=0}^{\infty}(n+1)x^{n+1} ]
Re‑index with (k=n+1):
[ \frac{x}{(1-x)^{2}} = \sum_{k=1}^{\infty}k,x^{k} ]
A neat, compact form that pops up in combinatorics (think “sum of the first (k) natural numbers”).
Common Mistakes / What Most People Get Wrong
-
Forgetting the radius of convergence – It’s easy to assume the new series works everywhere, but the original radius stays the same. Differentiating (\frac{1}{1-x}) still only works for (|x|<1).
-
Dropping the constant of integration – When you integrate a series, you must include (+C). If you’re matching a known function that equals zero at the expansion point, set (C) accordingly; otherwise you’ll end up with an offset.
-
Mishandling the index shift – After differentiation the exponent drops by one, so you need to re‑index carefully. Skipping this step leads to a series that looks right but actually represents a different function.
-
Assuming term‑by‑term differentiation works for any series – It only works inside the interval where the series converges absolutely. If you’re near the boundary, you need to check uniform convergence or use a different technique.
-
Mixing up signs – When you replace (x) with (-x) (as in the arctangent example), it’s easy to lose a minus sign. Write the substitution explicitly; a quick “plug‑in‑(x=-0.2)” test catches most errors.
Practical Tips / What Actually Works
- Keep a cheat sheet of base series – A one‑page list of the geometric, exponential, sine, cosine, and log series saves you a lot of scrolling.
- Use symbolic algebra for re‑indexing – Write the general term, then let (k=n-1) (or whatever shift you need) and simplify before typing it out.
- Combine differentiation with algebraic manipulation – Sometimes you need to multiply by a polynomial before differentiating. Example: to get (\frac{x}{(1-x)^{3}}), start with (\frac{1}{(1-x)^{2}}), differentiate, then multiply by (x).
- Check endpoints separately – For series that converge at (|x|=1) conditionally (like the alternating harmonic series), test the endpoint if you need the value there.
- Remember the “factorial shortcut” – When differentiating (e^{x}) or (\sin x), the factorial in the denominator cancels nicely, often leaving the same series back. That’s a quick sanity check.
FAQ
Q1: Can I differentiate a power series more than once?
A: Absolutely. Each differentiation reduces the radius of convergence by nothing—it stays the same. Just re‑index after each step Turns out it matters..
Q2: What if the function I need isn’t a derivative of a known series?
A: Try algebraic tricks: factor, multiply, or divide by a simple polynomial to turn it into a derivative of something you know. If that fails, consider integrating a known series instead.
Q3: Does this method work for complex numbers?
A: Yes. Power series are analytic in the complex plane, so term‑by‑term differentiation holds wherever the series converges The details matter here. That alone is useful..
Q4: How many terms should I keep for a good approximation?
A: It depends on the desired accuracy and the distance from the expansion point. Use the remainder estimate (|R_n|\le M\frac{|x|^{n+1}}{1-|x|}) for geometric‑type series, or the Lagrange form for Taylor series.
Q5: Why does the constant of integration sometimes disappear?
A: If you integrate a series that represents a function known to be zero at the expansion point (like (\ln(1+x)) at (x=0)), the constant must be zero. Otherwise you need to solve for it using the original function’s value at that point.
Power series are more than a textbook exercise; they’re a practical tool for any problem that needs local approximations or analytic manipulation. By mastering the simple act of differentiating (or integrating) a known series, you get to a whole library of expansions without ever opening a reference table.
So next time you stare at (\frac{1}{(1-x)^{4}}) or (\tan^{-1}x) and feel stuck, remember: differentiate the familiar, re‑index, and you’re done. Happy series‑crafting!
Putting It All Together: A Full‑Worked Example
Let’s illustrate the whole workflow with a problem that often appears in calculus‑based physics courses:
Find the power‑series representation of (\displaystyle f(x)=\frac{x^{2}}{(1-2x)^{3}}) centered at (x=0) and state its radius of convergence.
Step 1 – Identify a “base” series
The geometric series
[ \frac{1}{1-2x}= \sum_{n=0}^{\infty}(2x)^{n}= \sum_{n=0}^{\infty}2^{n}x^{n}, \qquad |2x|<1;( |x|<\tfrac12) ]
is the simplest object that contains the factor ((1-2x)^{-1}). e. Our target has ((1-2x)^{-3}), i.the second derivative of the geometric series (because each differentiation brings down an extra factor of the denominator).
Step 2 – Differentiate twice
First derivative:
[ \frac{d}{dx}!\left(\frac{1}{1-2x}\right)=\frac{2}{(1-2x)^{2}} =\sum_{n=0}^{\infty}2^{n+1}n,x^{,n-1}. ]
Re‑index with (k=n-1) (so (n=k+1)):
[ \frac{2}{(1-2x)^{2}}=\sum_{k=0}^{\infty}2^{k+2}(k+1)x^{k}. ]
Second derivative:
[ \frac{d}{dx}!\left(\frac{2}{(1-2x)^{2}}\right)=\frac{8}{(1-2x)^{3}} =\sum_{k=0}^{\infty}2^{k+2}(k+1)k,x^{k-1}. ]
Again shift the index: let (m=k-1) (so (k=m+1)):
[ \frac{8}{(1-2x)^{3}}=\sum_{m=0}^{\infty}2^{m+3}(m+2)(m+1)x^{m}. ]
Dividing both sides by (8) gives the series we actually need:
[ \frac{1}{(1-2x)^{3}}=\sum_{m=0}^{\infty}2^{m}(m+2)(m+1)x^{m}. ]
Step 3 – Multiply by the remaining factor (x^{2})
[ f(x)=x^{2}\cdot\frac{1}{(1-2x)^{3}} =\sum_{m=0}^{\infty}2^{m}(m+2)(m+1)x^{m+2}. ]
Finally re‑index so the exponent matches the summation index. Set (n=m+2) (hence (m=n-2)):
[ f(x)=\sum_{n=2}^{\infty}2^{,n-2}\bigl[(n-2)+2\bigr]\bigl[(n-2)+1\bigr]x^{n} =\sum_{n=2}^{\infty}2^{,n-2}n(n-1)x^{n}. ]
If you prefer the series to start at (n=0), simply add the missing terms (which are zero because of the factor (n(n-1))):
[ \boxed{\displaystyle \frac{x^{2}}{(1-2x)^{3}}=\sum_{n=0}^{\infty}2^{,n-2}n(n-1)x^{n} }\qquad (|x|<\tfrac12). ]
The radius of convergence is inherited from the original geometric series: (R=\tfrac12).
A Quick “Cheat Sheet” for Common Patterns
| Target function | How to obtain it from a known series | Resulting coefficient pattern |
|---|---|---|
| (\displaystyle \frac{1}{(1-x)^{k}}) | Differentiate (\frac{1}{1-x}) (k-1) times | (\displaystyle \binom{n+k-1}{k-1}) |
| (\displaystyle \frac{x}{(1-x)^{2}}) | Differentiate (\frac{1}{1-x}) once, then multiply by (x) | (n,x^{n}) |
| (\displaystyle \ln(1+x)) | Integrate (\frac{1}{1+x}) term‑by‑term | ((-1)^{n+1}\frac{x^{n}}{n}) |
| (\displaystyle \arctan x) | Integrate (\frac{1}{1+x^{2}}) term‑by‑term | ((-1)^{n}\frac{x^{2n+1}}{2n+1}) |
| (\displaystyle e^{ax}) | Substitute (ax) into (e^{x}) series | (\displaystyle \frac{a^{n}}{n!Now, }x^{n}) |
| (\displaystyle \sin(ax),;\cos(ax)) | Substitute (ax) into (\sin x,\cos x) series | (\displaystyle (-1)^{n}\frac{a^{2n+1}}{(2n+1)! }x^{2n+1}) (sin) <br> (\displaystyle (-1)^{n}\frac{a^{2n}}{(2n)! |
Keep this table handy; whenever a new problem resembles one of the rows, you already know the “recipe” Small thing, real impact..
Closing Thoughts
Power‑series manipulation—especially differentiation and integration—turns a handful of memorised expansions into an algorithmic toolbox. The steps are always the same:
- Pick a base series you know (geometric, exponential, sine/cosine, logarithm, arctan, etc.).
- Apply the required calculus operation (differentiate, integrate, multiply by a monomial, shift the index) exactly as you would with ordinary functions.
- Re‑index to tidy up the exponent of (x) and the coefficient pattern.
- Check the radius of convergence—it never shrinks when you differentiate or integrate term‑by‑term.
- Validate by plugging in a simple value (often (x=0)) or by comparing the first few coefficients with a direct expansion.
By internalising these five moves, you’ll no longer need to leaf through tables for every new function. Instead, you’ll construct the series on the fly, a skill that pays dividends in analysis, differential equations, physics, and even numerical computing.
So the next time a problem asks for the series of a seemingly exotic rational function, remember: differentiate the familiar, re‑index, and multiply—the answer is just a few algebraic steps away. Happy series‑crafting!
5️⃣ When the Denominator Is a Quadratic: Completing the Square
A particularly common obstacle is a rational function whose denominator is a quadratic that does not factor over the reals, for example
[ \frac{1}{1+x^{2}}. ]
Because the geometric series only handles linear denominators, we first rewrite the quadratic in a form that resembles a shifted geometric series. The trick is to complete the square and then treat the result as a geometric series in a new variable.
Short version: it depends. Long version — keep reading.
5.1 Example: (\displaystyle \frac{1}{1+x^{2}})
Write
[ \frac{1}{1+x^{2}}=\frac{1}{1-(-x^{2})}. ]
Now the denominator is exactly of the form (1-u) with (u=-x^{2}). Substituting (u) into the geometric series gives
[ \frac{1}{1+x^{2}}=\sum_{n=0}^{\infty}(-x^{2})^{n} =\sum_{n=0}^{\infty}(-1)^{n}x^{2n}, \qquad |x|<1. ]
Notice the even‑power pattern: only terms with exponent (2n) survive, and the coefficients alternate in sign. This is the power‑series representation that we will later integrate to obtain (\arctan x) or differentiate to obtain (\frac{-2x}{(1+x^{2})^{2}}).
5.2 Example: (\displaystyle \frac{1}{4-x^{2}})
Factor the constant out of the denominator:
[ \frac{1}{4-x^{2}}=\frac{1}{4},\frac{1}{1-\frac{x^{2}}{4}} =\frac{1}{4}\sum_{n=0}^{\infty}\left(\frac{x^{2}}{4}\right)^{n} =\sum_{n=0}^{\infty}\frac{x^{2n}}{4^{,n+1}}. ]
Again we see only even powers, but now the radius of convergence is (|x|<2) because the series converges as long as (|x^{2}/4|<1).
5.3 General Pattern
If you encounter
[ \frac{1}{a^{2}-x^{2}} \qquad (a\neq0), ]
write it as
[ \frac{1}{a^{2}};\frac{1}{1-\bigl(\frac{x}{a}\bigr)^{2}} =\frac{1}{a^{2}}\sum_{n=0}^{\infty}\bigl(\tfrac{x}{a}\bigr)^{2n} =\sum_{n=0}^{\infty}\frac{x^{2n}}{a^{2n+2}}, \qquad |x|<|a|. ]
The denominator’s square forces the series to contain only even powers, while the constant (a) scales the coefficients and determines the radius of convergence.
6️⃣ Series for (\displaystyle \frac{1}{(1-x)^{k}}) via the Binomial Theorem
The geometric series handles the case (k=1). For integer (k>1) we can obtain the series by repeated differentiation or, more elegantly, by invoking the generalized binomial theorem:
[ (1-x)^{-k}= \sum_{n=0}^{\infty}\binom{n+k-1}{k-1}x^{n},\qquad |x|<1. ]
The binomial coefficient (\displaystyle\binom{n+k-1}{k-1}) is a polynomial in (n) of degree (k-1). For small (k) you can write it out explicitly:
| (k) | Coefficient (\displaystyle\binom{n+k-1}{k-1}) | First few terms |
|---|---|---|
| 2 | (n+1) | (1+2x+3x^{2}+4x^{3}+\dots) |
| 3 | (\frac{(n+2)(n+1)}{2}) | (1+3x+6x^{2}+10x^{3}+\dots) |
| 4 | (\frac{(n+3)(n+2)(n+1)}{6}) | (1+4x+10x^{2}+20x^{3}+\dots) |
These series appear frequently when solving linear differential equations with polynomial coefficients, so it is worth memorising the first two rows.
7️⃣ Putting It All Together: A “One‑Stop” Example
Suppose we need the Maclaurin series for
[ f(x)=\frac{x^{2}}{(1-2x)^{3}}. ]
Step 1 – Identify the base series.
We know
[ \frac{1}{(1-2x)^{3}}=\sum_{n=0}^{\infty}\binom{n+2}{2}(2x)^{n} =\sum_{n=0}^{\infty}\frac{(n+2)(n+1)}{2},2^{n}x^{n}. ]
Step 2 – Multiply by the extra factor (x^{2}).
Multiplying term‑by‑term yields
[ f(x)=\sum_{n=0}^{\infty}\frac{(n+2)(n+1)}{2},2^{n}x^{n+2}. ]
Step 3 – Re‑index so the exponent matches the summation index.
Let (m=n+2) (so (n=m-2)). Then
[ f(x)=\sum_{m=2}^{\infty}\frac{(m)(m-1)}{2},2^{,m-2}x^{m}. ]
If you prefer the series to start at (m=0), simply write the first two terms (which are zero) explicitly and keep the same formula for all (m\ge0) Still holds up..
Step 4 – State the radius of convergence.
Because we started from (\frac{1}{1-2x}), the original geometric series converges for (|2x|<1), i.e. (|x|<\tfrac12). Differentiation and multiplication by a polynomial do not shrink this radius, so the final series is valid for (|x|<\tfrac12) And it works..
Result:
[ \boxed{\displaystyle \frac{x^{2}}{(1-2x)^{3}} =\sum_{n=0}^{\infty}\frac{(n+2)(n+1)}{2},2^{,n},x^{n+2} =\sum_{m=2}^{\infty}\frac{m(m-1)}{2},2^{,m-2},x^{m}, \qquad |x|<\frac12.} ]
Everything fell out by recognising the pattern, applying the binomial coefficient, and shifting the index—exactly the workflow we have advocated throughout this article.
🎯 Bottom Line
Power series are not a collection of isolated formulas; they are a single, flexible algorithm.
- Start with a known expansion (geometric, exponential, sine, cosine, logarithm, arctan).
- Apply calculus operations (differentiate, integrate, multiply by (x^{k}), substitute (ax), or shift the index).
- Re‑index to present the series in the standard (\sum a_{n}x^{n}) form.
- Check the radius of convergence—it stays the same under term‑by‑term differentiation or integration.
- Verify with a quick plug‑in or by matching the first few coefficients.
When you internalise these steps, the “hard” series problems dissolve into routine manipulations. The table at the beginning of the article serves as a quick reference, but the real power lies in knowing why each entry works. With practice, you’ll be able to generate the series for virtually any rational or elementary transcendental function on the spot, a skill that pays off in every advanced calculus or differential‑equations course.
So the next time you see a function that looks intimidating, remember: differentiate the familiar, re‑index, and you’re done. Happy expanding!
5️⃣ Putting It All Together – A Worked‑Out Example
Let’s take a slightly more involved function and run through the five‑step algorithm in real time:
[ g(x)=\frac{3x^{4}+2x^{2}}{(1+4x)^{5}} . ]
Step 1 – Identify a base series
The denominator suggests the binomial series for ((1+4x)^{-5}). From the generalized binomial theorem
[ (1+z)^{\alpha}= \sum_{n=0}^{\infty}\binom{\alpha}{n}z^{n}, \qquad |z|<1, ]
with (\alpha=-5) and (z=4x),
[ \frac{1}{(1+4x)^{5}}=\sum_{n=0}^{\infty}\binom{-5}{n}(4x)^{n} =\sum_{n=0}^{\infty}(-1)^{n}\binom{n+4}{4},4^{,n}x^{n}. ]
(The identity (\displaystyle \binom{-5}{n}=(-1)^{n}\binom{n+4}{4}) follows from the definition of binomial coefficients for negative upper indices.)
Step 2 – Multiply by the polynomial numerator
We have two monomials, (3x^{4}) and (2x^{2}). Distribute them across the series:
[ g(x)=3x^{4}\sum_{n=0}^{\infty}(-1)^{n}\binom{n+4}{4},4^{,n}x^{n} ;+; 2x^{2}\sum_{n=0}^{\infty}(-1)^{n}\binom{n+4}{4},4^{,n}x^{n}. ]
Factor the powers of (x) into the summand:
[ g(x)=\sum_{n=0}^{\infty}(-1)^{n}\binom{n+4}{4},4^{,n} \bigl(3x^{n+4}+2x^{n+2}\bigr). ]
Step 3 – Re‑index so the exponent of (x) matches the summation index
We would like a single series of the form (\sum a_{k}x^{k}). Split the two pieces and shift indices separately Simple, but easy to overlook..
For the (3x^{n+4}) term: let (k=n+4\Rightarrow n=k-4).
When (n=0) we have (k=4), so the sum starts at (k=4):
[ 3\sum_{k=4}^{\infty}(-1)^{k-4}\binom{k}{4},4^{,k-4}x^{k}. ]
For the (2x^{n+2}) term: let (k=n+2\Rightarrow n=k-2).
Now the lower limit is (k=2):
[ 2\sum_{k=2}^{\infty}(-1)^{k-2}\binom{k+2}{4},4^{,k-2}x^{k}. ]
(Notice that (\binom{(k-2)+4}{4}=\binom{k+2}{4}).)
Finally combine the two sums into a single one, padding the missing lower indices with zero coefficients:
[ g(x)=\sum_{k=0}^{\infty}c_{k},x^{k}, \qquad c_{k}= \begin{cases} 0, & k=0,1,\[4pt] 2(-1)^{k-2}\displaystyle\binom{k+2}{4},4^{,k-2}, & 2\le k\le 3,\[8pt] 3(-1)^{k-4}\displaystyle\binom{k}{4},4^{,k-4} ;+; 2(-1)^{k-2}\displaystyle\binom{k+2}{4},4^{,k-2}, & k\ge 4 . \end{cases} ]
If you prefer a compact closed form, factor out the common ((-1)^{k}) and a power of (4):
[ c_{k}=(-1)^{k},4^{,k-4} \Bigl[,3\binom{k}{4} ;+;2\cdot4^{2}\binom{k+2}{4}\Bigr], \qquad k\ge 4, ] and (c_{0}=c_{1}=c_{2}=c_{3}=0).
Step 4 – Radius of convergence
Our base series converges for (|4x|<1), i.e. (|x|<\frac14). Multiplying by a polynomial does not affect the radius, so the final series is valid on the same interval:
[ \boxed{;|x|<\tfrac14; }. ]
Step 5 – Quick sanity check
Take (x=0.1) (which lies inside the disc). Compute the first three non‑zero terms:
[ \begin{aligned} c_{4}&=3(-1)^{0}\binom{4}{4}4^{0}+2(-1)^{2}\binom{6}{4}4^{2} =3\cdot1\cdot1+2\cdot1\cdot15\cdot16=3+480=483,\[4pt] c_{5}&=3(-1)^{1}\binom{5}{4}4^{1}+2(-1)^{3}\binom{7}{4}4^{3} =-3\cdot5\cdot4-2\cdot35\cdot64=-60-4480=-4540,\[4pt] c_{6}&=3(-1)^{2}\binom{6}{4}4^{2}+2(-1)^{4}\binom{8}{4}4^{4} =3\cdot15\cdot16+2\cdot70\cdot256=720+35840=36560. \end{aligned} ]
Thus
[ g(0.1)\approx 483(0.1)^{4}-4540(0.1)^{5}+36560(0.1)^{6}=0.0483-0.0454+0.0366\approx0.0395, ]
which matches the direct evaluation (\displaystyle \frac{3(0.Which means 4)^{5}}\approx0. 1)^{4}+2(0.1)^{2}}{(1+0.0395) to three decimal places—exactly what we expect from a correctly derived series That's the part that actually makes a difference..
6️⃣ When the Standard Toolbox Isn’t Enough
The five‑step method works for any rational function or for elementary transcendental functions that can be expressed as compositions of the basic series listed in the reference table. Occasionally you’ll encounter a function such as
[ h(x)=\frac{\ln(1+x)}{(1-x)^{2}}, ]
where the numerator and denominator belong to different families. The trick is to expand each factor independently and then convolve the two series. Convolution is nothing but the Cauchy product:
[ \biggl(\sum_{n\ge0}a_{n}x^{n}\biggr) \biggl(\sum_{m\ge0}b_{m}x^{m}\biggr) =\sum_{k\ge0}\Bigl(\sum_{j=0}^{k}a_{j}b_{k-j}\Bigr)x^{k}. ]
Because the inner sum is finite for each (k), the product is again a power series with the same radius of convergence (the minimum of the two original radii). Mastery of the Cauchy product completes the “algorithmic” picture.
7️⃣ A Quick Reference Cheat‑Sheet
| Base function | Series (center 0) | Typical operations |
|---|---|---|
| (\displaystyle \frac{1}{1-x}) | (\sum_{n\ge0}x^{n}) | multiply, differentiate, integrate |
| (\displaystyle \frac{1}{(1-x)^{k}}) | (\sum_{n\ge0}\binom{n+k-1}{k-1}x^{n}) | shift, multiply by (x^{m}) |
| (\displaystyle e^{x}) | (\sum_{n\ge0}\frac{x^{n}}{n!}) | substitute (ax), multiply by polynomials |
| (\displaystyle \sin x) | (\sum_{n\ge0}(-1)^{n}\frac{x^{2n+1}}{(2n+1)!}) | differentiate → (\cos x), integrate → (-\cos x) |
| (\displaystyle \cos x) | (\sum_{n\ge0}(-1)^{n}\frac{x^{2n}}{(2n)! |
Keep this table handy; it’s the “standard library” that your algorithm calls upon.
📚 Conclusion
Power‑series expansions are not a collection of isolated memorised formulas; they are a single, repeatable procedure. By anchoring every new problem to a known seed series, then applying a disciplined sequence of algebraic and calculus operations—differentiate, integrate, multiply by a monomial, substitute, re‑index—you turn any reasonable elementary function into a tidy (\sum a_{n}x^{n}) with a clearly identified radius of convergence.
The payoff is immediate:
- Simplified computation – you can evaluate functions to arbitrary precision by truncating the series.
- Analytic insight – the coefficients often encode combinatorial or geometric information.
- Readiness for advanced topics – solving differential equations, performing asymptotic analysis, and working with generating functions all rely on the same toolbox.
So the next time you meet a “hard” series problem, remember the five‑step algorithm, pull the appropriate entry from the reference table, and let the mechanical steps do the heavy lifting. With practice, the process becomes second nature, and you’ll find that most power‑series expansions are less a mystery and more a routine calculation.
Happy series hunting!
