Telescoping & Recursion

Study Sheet

Telescoping & Recursion

Sums that collapse and sequences that build

Making Long Computations Short

Tip
Telescoping sums

Partial fractions turn 1k(k+1)\dfrac{1}{k(k+1)} into 1k−1k+1\dfrac1k - \dfrac{1}{k+1}, so the sum telescopes: ∑k=1n1k(k+1)=1−1n+1\sum_{k=1}^{n} \dfrac{1}{k(k+1)} = 1 - \dfrac{1}{n+1}.

In plain terms. Write each term as a difference; everything in the middle cancels.

Example. ∑k=191k(k+1)=1−110=910\sum_{k=1}^{9} \dfrac{1}{k(k+1)} = 1 - \dfrac{1}{10} = \dfrac{9}{10}.

Why it works. If every term has the form uk−uk+1u_k - u_{k+1}, then in the sum each intermediate uku_k appears once with a plus sign (from term kk) and once with a minus sign (from term k−1k - 1), so only the first and last survive. Partial fractions is the tool that FINDS the form uk−uk+1u_k - u_{k+1} when a term is a rational function.

Concept
Linear recursions

When each state comes from the previous ones — like an=an−1+an−2a_n = a_{n-1} + a_{n-2} — compute forward from the base cases rather than hunting for a formula.

In plain terms. Build the answer step by step; ten steps of arithmetic beat an hour of algebra.

Example. Climbing 5 stairs taking 1 or 2 at a time: 1,2,3,5,81, 2, 3, 5, 8 — there are 88 ways (a Fibonacci-type recursion).

Why it works. A recursion says the count for size nn is built from counts for smaller sizes, so computing upward from the base cases is an exact method, not an approximation. On the AIME, where nn is usually below 2020, a table of values is often faster and safer than a closed form.

Example
Worked: steps of one, two or three

In how many ways can you climb a staircase of 1010 steps if each stride covers 11, 22 or 33 steps?

Spotting it. Condition on the last stride: it covered 11, 22 or 33 steps, leaving a staircase of n−1n - 1, n−2n - 2 or n−3n - 3.

Step 1. cn=cn−1+cn−2+cn−3c_n = c_{n-1} + c_{n-2} + c_{n-3} with c0=1c_0 = 1, c1=1c_1 = 1, c2=2c_2 = 2.

Step 2. c3,…,c10=4,7,13,24,44,81,149,274c_3, \dots, c_{10} = 4, 7, 13, 24, 44, 81, 149, 274. The answer is 274274.

Tip
Circle power tools together
ABCD7242015
The example to scale: AC=25AC = 25 is a diameter, and Ptolemy gives 25⋅BD=7⋅20+15⋅2425 \cdot BD = 7 \cdot 20 + 15 \cdot 24.

Ptolemy's theorem (AC⋅BD=AB⋅CD+AD⋅BCAC \cdot BD = AB \cdot CD + AD \cdot BC in a cyclic quadrilateral) and the Power of a Point (AP⋅PB=CP⋅PDAP \cdot PB = CP \cdot PD for chords through PP) often chain together in one problem.

In plain terms. In circle problems, name the theorem each length fact comes from and the algebra falls out.

Example. Cyclic ABCDABCD with diameter AC=25AC = 25, AB=7AB = 7, BC=24BC = 24, CD=20CD = 20, DA=15DA = 15: Ptolemy gives BD=7⋅20+15⋅2425=20BD = \dfrac{7 \cdot 20 + 15 \cdot 24}{25} = 20.

Side note
Where it goes wrong

Pairing the wrong sides in Ptolemy. The product of the DIAGONALS equals the sum of the products of OPPOSITE sides. In the example, using adjacent sides instead, AB⋅BC+CD⋅DA=168+300AB \cdot BC + CD \cdot DA = 168 + 300, gives BD=46825=18.72BD = \tfrac{468}{25} = 18.72 rather than 2020. And the theorem needs a cyclic quadrilateral; for any other quadrilateral the equality becomes a strict inequality.

Try it
Try it: seven stairs

In how many ways can you climb 77 stairs with strides of 11, 22 or 33 steps?

Show the answer

With cn=cn−1+cn−2+cn−3c_n = c_{n-1} + c_{n-2} + c_{n-3} and c0=c1=1c_0 = c_1 = 1, c2=2c_2 = 2: c3,…,c7=4,7,13,24,44c_3, \dots, c_7 = 4, 7, 13, 24, 44. The answer is 4444.

Finding the Difference That Telescopes

Tip
Partial fractions first

A rational term with factored denominator splits into pieces whose difference telescopes: 1k(k+2)=12(1k−1k+2)\dfrac{1}{k(k+2)} = \tfrac12\left(\dfrac1k - \dfrac1{k+2}\right).

With a gap of 22, two terms survive at each end instead of one: ∑k=1n1k(k+2)=12(1+12−1n+1−1n+2)\sum_{k=1}^{n}\dfrac{1}{k(k+2)} = \tfrac12\left(1 + \tfrac12 - \tfrac1{n+1} - \tfrac1{n+2}\right). Counting the survivors correctly is where the points are; writing out the first three and last three terms explicitly makes it visible.

Tip
Telescoping products

A product telescopes when each factor is a ratio of consecutive values: ∏k=2nk−1k=1n\prod_{k=2}^{n}\dfrac{k-1}{k} = \dfrac1n.

Factorial-heavy products usually hide such a ratio: ∏(1−1k2)=∏(k−1)(k+1)k2\prod\left(1 - \tfrac1{k^2}\right) = \prod\dfrac{(k-1)(k+1)}{k^2} splits into two telescoping products and leaves n+12n\dfrac{n+1}{2n}. Writing each factor as (something at k+1k+1)/(something at kk) is the deciding step.

Example
Worked: a gapped telescope

Compute ∑k=1981k(k+2)\displaystyle\sum_{k=1}^{98}\frac{1}{k(k+2)}.

Step 1 — 1k(k+2)=12(1k−1k+2)\dfrac{1}{k(k+2)} = \tfrac12\left(\dfrac1k - \dfrac1{k+2}\right).

Step 2 — the surviving terms are 12(1+12−199−1100)\tfrac12\left(1 + \tfrac12 - \tfrac1{99} - \tfrac1{100}\right).

Step 3 — =12(32−1999900)=1465119800= \tfrac12\left(\tfrac32 - \tfrac{199}{9900}\right) = \tfrac{14651}{19800}.

Example
Worked: a product of cubic ratios

∏k=210k3−1k3+1=mn\displaystyle\prod_{k=2}^{10}\frac{k^3 - 1}{k^3 + 1} = \frac mn in lowest terms. Find m+nm + n.

Spotting it. Factor: k3−1=(k−1)(k2+k+1)k^3 - 1 = (k - 1)(k^2 + k + 1) and k3+1=(k+1)(k2−k+1)k^3 + 1 = (k + 1)(k^2 - k + 1). The key observation is k2+k+1=(k+1)2−(k+1)+1k^2 + k + 1 = (k + 1)^2 - (k + 1) + 1: the quadratic factor on top at kk is the one on the bottom at k+1k + 1.

Step 1. The product splits as ∏k=210k−1k+1⋅∏k=210k2+k+1k2−k+1\displaystyle\prod_{k=2}^{10}\frac{k - 1}{k + 1} \cdot \prod_{k=2}^{10}\frac{k^2 + k + 1}{k^2 - k + 1}.

Step 2. The first is 1⋅210⋅11=155\dfrac{1 \cdot 2}{10 \cdot 11} = \dfrac{1}{55} (a gap-two telescope: 1,21, 2 survive on top, 10,1110, 11 below). The second telescopes to 102+10+122−2+1=1113=37\dfrac{10^2 + 10 + 1}{2^2 - 2 + 1} = \dfrac{111}{3} = 37.

Step 3. The product is 3755\dfrac{37}{55}, so m+n=92m + n = 92.

Side note
Where it goes wrong

Counting survivors wrong. With a gap of 22, TWO terms survive at each end: ∑k=198(1k−1k+2)\sum_{k=1}^{98}\left(\tfrac1k - \tfrac1{k+2}\right) keeps 1+121 + \tfrac12 at the front and −199−1100-\tfrac1{99} - \tfrac1{100} at the back. Keeping only 1−11001 - \tfrac1{100} (as if the gap were 11) is the standard error. Write out the first three and last three terms and cross off what cancels.

Try it
Try it: a telescoping product

Compute ∏k=210(1−1k)\displaystyle\prod_{k=2}^{10}\left(1 - \frac1k\right).

Show the answer

Each factor is k−1k\dfrac{k-1}{k}, so the product is 110\dfrac{1}{10}.

Telescoping Beyond Fractions

Tip
Find FF with tk=F(k+1)−F(k)t_k = F(k + 1) - F(k)

Useful differences: k⋅k!=(k+1)!−k!k \cdot k! = (k + 1)! - k!; 1k+k+1=k+1−k\dfrac{1}{\sqrt k + \sqrt{k + 1}} = \sqrt{k + 1} - \sqrt k; 1k(k+1)(k+2)=12[1k(k+1)−1(k+1)(k+2)]\dfrac{1}{k(k + 1)(k + 2)} = \dfrac12\left[\dfrac{1}{k(k + 1)} - \dfrac{1}{(k + 1)(k + 2)}\right]; k(k+1)!=1k!−1(k+1)!\dfrac{k}{(k + 1)!} = \dfrac1{k!} - \dfrac1{(k + 1)!}.

Why it works. If every term is F(k+1)−F(k)F(k + 1) - F(k), the sum from k=1k = 1 to nn is F(n+1)−F(1)F(n + 1) - F(1), because each intermediate FF appears once with each sign. The skill is guessing FF from the shape of the term: factorials suggest factorials, square roots suggest conjugates, a product of consecutive factors suggests a product with one factor fewer. Always test the guess at k=1k = 1.

Example
Worked: a factorial sum

Find the remainder when 1⋅1!+2⋅2!+3⋅3!+⋯+20⋅20!1 \cdot 1! + 2 \cdot 2! + 3 \cdot 3! + \cdots + 20 \cdot 20! is divided by 10001000.

Spotting it. k⋅k!=(k+1−1)k!=(k+1)!−k!k \cdot k! = (k + 1 - 1)k! = (k + 1)! - k!.

Step 1. The sum telescopes to 21!−1!=21!−121! - 1! = 21! - 1.

Step 2. 21!21! is divisible by 10001000 (it contains 232^3 and 535^3), so 21!−1≡−1≡999(mod1000)21! - 1 \equiv -1 \equiv 999 \pmod{1000}.

Example
Worked: three consecutive factors

∑k=181k(k+1)(k+2)=mn\displaystyle\sum_{k=1}^{8}\frac{1}{k(k + 1)(k + 2)} = \frac mn in lowest terms. Find m+nm + n.

Spotting it. A product of three consecutive factors: compare it with products of two. 1k(k+1)−1(k+1)(k+2)=(k+2)−kk(k+1)(k+2)=2k(k+1)(k+2)\tfrac{1}{k(k + 1)} - \tfrac{1}{(k + 1)(k + 2)} = \tfrac{(k + 2) - k}{k(k + 1)(k + 2)} = \tfrac{2}{k(k + 1)(k + 2)}.

Step 1. So the sum is 12[11⋅2−19⋅10]=12⋅45−190=2290=1145\tfrac12\left[\tfrac{1}{1 \cdot 2} - \tfrac{1}{9 \cdot 10}\right] = \tfrac12 \cdot \tfrac{45 - 1}{90} = \tfrac{22}{90} = \tfrac{11}{45}.

Step 2. m+n=56m + n = 56.

Example
Worked: roots in the denominator

∑k=11431(k+1)k+kk+1=mn\displaystyle\sum_{k=1}^{143}\frac{1}{(k + 1)\sqrt k + k\sqrt{k + 1}} = \frac mn in lowest terms. Find m+nm + n.

Spotting it. Factor kk+1\sqrt k\sqrt{k + 1} out of the denominator: (k+1)k+kk+1=kk+1(k+1+k)(k + 1)\sqrt k + k\sqrt{k + 1} = \sqrt k\sqrt{k + 1}\left(\sqrt{k + 1} + \sqrt k\right). Then rationalise.

Step 1. 1kk+1(k+1+k)=k+1−kkk+1=1k−1k+1\dfrac{1}{\sqrt k\sqrt{k + 1}(\sqrt{k + 1} + \sqrt k)} = \dfrac{\sqrt{k + 1} - \sqrt k}{\sqrt k\sqrt{k + 1}} = \dfrac{1}{\sqrt k} - \dfrac{1}{\sqrt{k + 1}}.

Step 2. The sum telescopes to 1−1144=11121 - \dfrac{1}{\sqrt{144}} = \dfrac{11}{12}, so m+n=23m + n = 23.

Side note
Where it goes wrong

Guessing a difference without testing it. It is tempting to write 1k(k+1)(k+2)=1k(k+1)−1(k+1)(k+2)\tfrac{1}{k(k + 1)(k + 2)} = \tfrac{1}{k(k + 1)} - \tfrac{1}{(k + 1)(k + 2)}, but at k=1k = 1 the left side is 16\tfrac16 and the right side is 12−16=13\tfrac12 - \tfrac16 = \tfrac13: off by a factor of 22. One substitution catches the missing 12\tfrac12.

Try it
Try it: conjugates

Compute ∑k=1801k+k+1\displaystyle\sum_{k=1}^{80}\frac{1}{\sqrt k + \sqrt{k + 1}}.

Show the answer

Each term is k+1−k\sqrt{k + 1} - \sqrt k, so the sum is 81−1=8\sqrt{81} - \sqrt1 = 8.