Intuition The one core idea
A power series is a "polynomial that never stops" — where a polynomial just means a sum of whole-number powers of x each carrying a fixed number weight, like 3 x 2 + 5 x + 7 (defined properly in §2). This whole topic says: inside its safe zone, this never-ending polynomial can be differentiated and integrated one term at a time, and the size of the safe zone never changes.
This page assumes nothing . Every letter, symbol and picture the parent note parent topic leans on is built here from the ground up, in an order where each block rests on the one before it.
Before any formula, here is the plain-words meaning of each symbol. We earn each one properly in its own section below.
Symbol
Plain words
Section
∑
"add up a whole list"
§1
n
a counter: 0 , 1 , 2 , 3 , …
§1
t n
a generic term — the n -th item on the list
§1
∞
"keeps going forever" — the list never stops
§1
a n
the number weight on the n -th piece
§2
x
the input we plug in
§2
f , f ( x )
a machine turning input x into an output number
§2
( x − c ) n
the n -th building block, centred at c
§3
c
the centre point of the series
§3
∣ ⋅ ∣
absolute value: size of a number, sign dropped
§4
R
radius: half-width of the safe zone
§4
lim sup
"the biggest value a wobbly list keeps returning to"
§5
d x d , f ′
slope machine (derivative)
§6
∫
area machine (integral)
§7
C
unknown constant of integration
§7
Definition Sigma notation
The symbol ∑ (capital Greek "S", for Sum ) means add these up . The counter n underneath tells you where to start; the number on top tells you where to stop. Each item on the list is written t n — a generic term , meaning "whatever the n -th item happens to be."
∑ n = 0 3 t n = t 0 + t 1 + t 2 + t 3 .
Read it out loud as: "for n from 0 to 3 , add up each t n ." Each t n is one term — one item on the shopping list.
Definition The infinity symbol
∞
∞ is not a number you can reach; it is shorthand for "keeps going with no last step." Writing it as the top of a sum means the list of terms never ends :
∑ n = 0 ∞ t n = t 0 + t 1 + t 2 + t 3 + ⋯
Definition Convergence and divergence
Add the terms one at a time and watch the running total (the partial sum) s N = t 0 + t 1 + ⋯ + t N . If, as N grows without bound, these running totals get and stay arbitrarily close to one fixed number S — closer than any tiny gap you name, for all large enough N — we say the series converges to S and write ∑ n = 0 ∞ t n = S . If the running totals do not settle on any single number (they run off to infinity, or keep jumping around), the series diverges .
Intuition Does an endless sum even mean anything?
By that definition, yes — sometimes! If each term is much smaller than the last, the running totals home in on a single number. Think of walking half the remaining distance to a wall forever: 2 1 + 4 1 + 8 1 + ⋯ never overshoots 1 , and the running totals close in on exactly 1 . That "homing in" is precisely convergence, and it is the whole reason infinite sums are useful.
Look at Figure s01. The cyan dots are the running total after 1 , 2 , 3 , … terms of 2 1 + 4 1 + ⋯ . The white arrows show each dot creeping toward the amber dashed line at height 1 — a picture of "homing in" (convergence).
Common mistake "Infinite sum must be infinite."
Why it feels right: you're adding infinitely many things. The fix: if the things shrink fast enough, the running totals settle on a finite number — like the half-steps above adding to exactly 1 . That is convergence; only when the running totals fail to settle do we call it divergence.
Now we make each term depend on a number x we get to choose.
A polynomial is a finite sum of whole-number powers of x (x 0 = 1 , x 1 , x 2 , … ), each multiplied by a fixed number. Example: 3 x 2 + 5 x + 7 . The fixed numbers (3 , 5 , 7 ) are its weights; there is always a last term.
a n , variable x , function f
x is the input — the number we plug in. Picture it as a position on a horizontal number line.
a n is the coefficient : a fixed number weight glued to the n -th term. The list a 0 , a 1 , a 2 , … is decided in advance and does not depend on x . (So the generic term t n from §1 is now specifically a n ( x − c ) n .)
f names a function — a machine: you feed it an input x and it returns one output number, written f ( x ) ("f of x ").
A simple example: if every a n = 1 , the terms are 1 , x , x 2 , x 3 , … and
f ( x ) = ∑ n = 0 ∞ x n = 1 + x + x 2 + x 3 + ⋯
This is the Geometric series — the friendliest power series there is, and the parent note builds everything from it.
Intuition Why split into "weight
× power"?
Separating the fixed weight a n from the moving power of x is exactly how a polynomial works (3 x 2 + 5 x + 7 ). A power series is that idea with no last term — so all the polynomial tricks have a chance of carrying over. That "chance" is what this topic proves.
c
Instead of measuring from zero, we can measure from a chosen home base c . The block ( x − c ) is "how far x is from the centre c ", and ( x − c ) n raises that distance to the n -th power.
So the general power series in the parent note is
f ( x ) = ∑ n = 0 ∞ a n ( x − c ) n = a 0 + a 1 ( x − c ) + a 2 ( x − c ) 2 + ⋯
where f ( x ) (from §2) is the number the whole sum lands on when you feed in x .
Look at Figure s02. The amber curve is the true total 1 − x 1 ; the cyan/white curves are what you get by stopping after a few terms of 1 + x + x 2 + ⋯ . Each extra term hugs the amber curve over a wider stretch — an infinite polynomial is the smooth function, in the limit.
c instead of always 0 ?
If you're studying behaviour near a special point (say near x = 5 ), you want your building blocks to be small exactly there. Choosing c = 5 makes ( x − c ) tiny near 5 , so early terms dominate and the series is most accurate right where you care. When c = 0 this reduces to plain ∑ a n x n .
Plug in a big x and the powers ( x − c ) n explode; the running totals run off to infinity (diverge ). Plug in a small x and the terms shrink; the running totals settle (converge ). There is a cutoff distance — and distance ignores direction, which is what absolute value is for.
Definition Absolute value
∣ ⋅ ∣
∣ q ∣ ("the absolute value of q ") is the size of a number with its sign thrown away : ∣ 3 ∣ = 3 and ∣ − 3 ∣ = 3 . So ∣ x − c ∣ is the plain distance from x to c on the number line, whichever side x is on. Likewise ∣ a n ∣ is the size of the weight a n , ignoring whether it's positive or negative.
Definition Radius of convergence
R
There is a number R ≥ 0 (possibly ∞ ) such that:
if ∣ x − c ∣ < R (inside), the series converges ;
if ∣ x − c ∣ > R (outside), the series diverges .
The safe zone (interval of convergence) is the open interval ( c − R , c + R ) .
Intuition The degenerate case
R = 0 — and the centre is always safe
Notice that plugging x = c gives ∣ x − c ∣ = 0 , so every term after the first is a n ⋅ 0 n = 0 . The sum is just a 0 — always a finite number. So a power series always converges at its own centre , even when R = 0 (safe zone shrinks to the single point x = c ) — the strict "< R " only governs points other than the centre. At the opposite extreme R = ∞ the safe zone is the whole line.
Look at Figure s03. The cyan band is the safe zone of half-width R around the amber centre dot c ; the white squares mark the two edges c ± R , flagged "test by hand."
radius ?
On the real line the safe zone is a segment centred at c , and R is its half-width — literally the radius of that segment. The parent's headline result "R doesn't change under differentiation/integration" is a statement about this half-width . See Power series and radius of convergence .
Common mistake "Inside the safe zone means the two endpoints are fine too."
Why it feels right: the boundary looks like part of the zone. The fix: the two edges x = c ± R are a coin-toss — each must be tested by hand. This is exactly the wrinkle the parent's "DIRT — Test endpoints again" mnemonic warns about; see Abel's theorem (endpoint behaviour) .
How do we find R ? The Cauchy–Hadamard theorem does it from the weights alone:
R 1 = lim sup n → ∞ ∣ a n ∣ 1/ n .
Two new pieces here.
n -th root ∣ a n ∣ 1/ n
∣ a n ∣ 1/ n is "the number which, raised to the power n , gives ∣ a n ∣ ." It measures the geometric size of the weight per step — how fast the weights grow or shrink, on a "times-per-term" scale rather than "plus-per-term."
lim sup (limit superior)
Some lists never settle to one value — they wobble. lim sup is the highest value the wobble keeps coming back to in the long run (its "ceiling of persistent peaks"). If the list does settle, lim sup is just that ordinary limit.
Look at Figure s04. The cyan dots are a wobbly list of ∣ a n ∣ 1/ n values; the amber dashed line is the lim sup — the ceiling the peaks keep returning to even though individual dots dip below it.
lim sup and not a plain limit?
Because ∣ a n ∣ 1/ n can bounce forever (e.g. weights that alternate big/small). A plain limit might not exist, but the worst-case ceiling always does — and it's the ceiling that decides where the series first breaks. That's why Cauchy–Hadamard is bulletproof: it works even for wobbly coefficient lists.
The parent's whole radius proof rests on one fact: multiplying weights by the counter n barely nudges this n -th-root scale, because n 1/ n → 1 . We can sanity-check that number: even at n = 100 , 10 0 1/100 ≈ 1.047 — already almost 1 .
d x d f ( x ) , also written f ′ ( x ) , is the slope of the graph of f at each point — its instantaneous steepness. On a single power block the rule is the "bring-the-power-down" rule:
d x d x n = n x n − 1 .
Look at what that rule does to a term: it multiplies the weight by n and knocks the power down by one. That is precisely why the parent's differentiated series reads ∑ n a n ( x − c ) n − 1 — every term got hit by the same rule. The n = 0 term (a constant) has slope zero and vanishes, which is why the sum restarts at n = 1 .
d x d the tool for "term by term"?
The topic's promise is "treat an infinite polynomial like a normal one." The derivative is the operation whose rule on each block is dead simple. The deep question the topic answers is: are we allowed to apply that simple rule to infinitely many blocks and add up the results? (Answer: yes, inside the safe zone — thanks to Uniform convergence and the Weierstrass M-test .)
∫ f ( x ) d x is the area-accumulator : it undoes differentiation. Its rule on a single block bumps the power up and divides by the new power:
∫ x n d x = n + 1 x n + 1 + C .
The + C (the constant of integration ) is an unknown constant : many functions share the same slope-pattern, differing only by a fixed vertical shift, so integration leaves that shift undetermined until you pin it with one known value (like f ( 0 ) ). This is exactly why the parent fixes C using arctan 0 = 0 and ln ( 1 + 0 ) = 0 .
Applied to every term, this gives the parent's integrated series C + ∑ n + 1 a n ( x − c ) n + 1 : each weight divided by n + 1 , each power bumped up one.
Intuition Why does dividing by
n + 1 not change R ?
On the n -th-root scale of §5, dividing by n + 1 contributes a factor ( n + 1 ) − 1/ n → 1 — the same vanishing nudge as n 1/ n → 1 . The half-width R can't feel it. Meanwhile shrinking terms can rescue a broken endpoint — which is how integrating 1 + x 2 1 gains convergence at x = 1 to give π /4 .
Sigma sum, counter n, infinity
Convergence and divergence of running totals
Weights an times powers of x
Building block x minus c to the n
Absolute value distance and radius R
limsup persistent ceiling
Derivative the slope machine
Integral the area machine
Uniform convergence and Weierstrass M test
Derive arctan and ln series, solve ODEs
Test yourself — say the answer before revealing.
What does n = 0 ∑ ∞ t n mean in plain words? Add up the never-ending list t 0 + t 1 + t 2 + ⋯ ; it may still land on a finite number if terms shrink fast enough.
What precisely does it mean for a series to converge? Its running totals s N get and stay arbitrarily close to one fixed number S as N grows; if they never settle on a single number, it diverges.
What is a polynomial, and how does a power series differ? A polynomial is a finite sum of whole-number powers of x with fixed weights (e.g. 3 x 2 + 5 x + 7 ); a power series is the same idea with no last term.
In a n ( x − c ) n , which part is fixed in advance and which do you choose? a n (the weight) and c (the centre) are fixed; x is the input you plug in.
What does ∣ x − c ∣ measure, and why absolute value? The distance from x to c ; absolute value strips the sign so only size (distance), not direction, matters.
What is the radius of convergence R geometrically, and what happens at x = c ? The half-width of the safe interval ( c − R , c + R ) ; the series always converges at the centre x = c (it equals a 0 there), even if R = 0 .
Why use lim sup ∣ a n ∣ 1/ n rather than a plain limit? Because ∣ a n ∣ 1/ n may wobble forever and a plain limit need not exist; the lim sup ceiling always exists and pins down where the series first breaks.
State the one-block rules for d x d and ∫ . d x d x n = n x n − 1 (power down, times n ); ∫ x n d x = n + 1 x n + 1 + C (power up, divide by n + 1 ).
Why does the constant + C appear after integrating? Many functions share the same slope-pattern, differing by a vertical shift; C stays unknown until fixed by a known value like f ( 0 ) .
What single limit makes both radius proofs work? n 1/ n → 1 (and likewise ( n + 1 ) ± 1/ n → 1 ), so multiplying/dividing weights by n leaves the n -th-root scale unchanged.