This deck covers weak, strong, structural, and well-founded induction, and shows that they are equivalent to the well-ordering principle. It targets the misconceptions that induction is circular, that the base case is optional, and that weak induction always suffices, along with the all-horses-are-the-same-color fallacy.
Subject: Foundations of Higher Mathematics · 122 slides · symbolic lesson
Open the interactive version of this deck · Homework for this lesson
Objectives
By the end of this deck you can:
1. State and apply the principle of induction: a base case plus an inductive step.
2. Use the induction hypothesis correctly and say exactly what it lets you assume.
3. Choose between weak induction, strong induction, and the well-ordering principle.
4. Run structural induction over recursively defined objects like formulas and trees.
5. Diagnose broken induction proofs, including the famous all-horses fallacy.
Warm-up
Discussion prompt
Before we open Proof Techniques II: Induction, Strong Induction & Well-Ordering: without looking back, what was the main idea of Proof Techniques I: Direct, Contrapositive, Contradiction, and what could you do by the end of it that you could not do before?
Hint: One sentence for the idea, one for the skill. If the second one is blank, that is the part to revisit.
Answer:
That deck shows how to prove a theorem three ways - directly, by contrapositive, and by contradiction - and then covers vacuous and trivial proofs, biconditionals, arguing without loss of generality, and disproof by counterexample. It targets the misconceptions that examples prove universal statements, that the converse is the same as the contrapositive, and that assuming the conclusion is a valid argument.
Concept
An induction proof establishes that one statement holds for every natural number at once.
Package the claim as a predicate: for each n there is a proposition, and we want all of them true.
\( P(n): \text{ the statement asserted about the natural number } n \)
predicate over the naturals — A family of propositions indexed by n, one for each natural number. Induction proves the whole infinite family from two finite obligations.
Counterexample
Discussion prompt
An induction proof establishes that one statement holds for every natural number at once.
That is stated as though it always holds. Do one of two things: produce a case where it fails, or say precisely what rules such a case out. "It just does" is not on the menu.
Hint: Hunt at the extremes first — zero, one, negative, empty, equal. If every extreme survives, the reason they survive is the proof.
Answer:
Package the claim as a predicate: for each n there is a proposition, and we want all of them true.
Concept
Verifying the claim for n equal to 1, 2, 3, and a million is not a proof. Infinitely many cases remain.
This is the for all trap from predicate logic: finitely many witnesses never settle a universal statement over an infinite domain.
Induction is the escape hatch: two finite checks that together cover the whole infinite family.
Analogy
Discussion prompt
Explain Why checking cases is not enough by analogy to something with no Foundations of Higher Mathematics in it at all — a queue, a recipe, a map, a bank balance, whatever fits. Then say where your analogy breaks.
Hint: An analogy that never breaks is not an analogy, it is the same idea wearing a hat. Find the seam — that is the part that is actually new.
Answer:
Verifying the claim for n equal to 1, 2, 3, and a million is not a proof. Infinitely many cases remain.
Picture it
Figure (svg): A row of upright dominoes with the leftmost one tipping into the next, on a baseline.
Discussion prompt
Read the picture before the words. What is this showing, and what is the one thing it is built to make obvious? Commit to an answer, then read on.
Hint: Name the parts, then say what changes between them — and if nothing changes, say what is being held still.
Answer:
Picture one domino for each natural number, standing in a long row.
Intuition
Picture one domino for each natural number, standing in a long row.
Figure (svg): A row of upright dominoes with the leftmost one tipping into the next, on a baseline.
Two facts guarantee they all fall: the first is pushed over, and each standing domino is close enough to topple its neighbor.
Push the first (the base case). Guarantee each knocks the next (the inductive step). Then every domino falls.
Explain it
Discussion prompt
Explain The mental model: a chain of dominoes to a student a year behind you. No notation, no jargon they have not met — and it still has to be true.
Hint: If your explanation needs a symbol they have never seen, you are describing the notation rather than the idea.
Answer:
Picture one domino for each natural number, standing in a long row.
Intuition
Suppose the base case holds and the step holds. Pick any n you like, say 47.
The base gives the first case. The step carries case 0 to 1, 1 to 2, and onward, one hop at a time.
Every particular n sits only finitely far from the base, so the two checks really do reach all of them.
Concept
Here is the rule written as a single logical statement.
\[ \big(P(0) \;\wedge\; \forall k\,[\,P(k) \rightarrow P(k+1)\,]\big) \;\rightarrow\; \forall n\, P(n) \]
Read it as: if the base holds, and every case implies the next, then all cases hold.
The base need not start at zero. If the claim only begins at some starting index, start the induction there.
Concept
The base case is where the chain starts. You prove the claim outright for the smallest index.
\( P(b) \text{ holds, where } b \text{ is the starting value, often } 0 \text{ or } 1 \)
It is usually a quick, concrete computation. Never skip it: a chain with no first domino never starts.
Concept
The step proves one implication, uniformly in k.
\( \forall k \;\; \big(P(k) \rightarrow P(k+1)\big) \)
You assume the claim at k and must derive the claim at k plus one. You do this once, for a generic k.
Proving an implication does not assert its hypothesis. You promise only: whenever a domino stands, it topples the next.
Concept
induction hypothesis — The assumption that the claim already holds at k, granted to you for free while you prove the claim at k plus one.
Beginners feel they are cheating by assuming what they want to prove. They are not.
You are not assuming the final goal, the universal statement. You assume one earlier instance to reach the next. That is the whole engine.
Intuition
The common worry: you assumed what you are trying to prove. You did not.
The goal is the universal statement, true for every n. The hypothesis is a single earlier instance.
Trading one earlier case for the next is a fair deal, not circularity. You never assume the whole family; you assume one member to earn its successor.
Concept
Every clean induction proof has the same visible moves.
Base case. State and check the smallest index.
Inductive step. Fix a generic k and state the induction hypothesis explicitly.
Derivation. Work from the hypothesis to the next case, then conclude for all n.
Ranking
Put in order
Put the moves of Worked example: the sum one through n into the order they have to happen.
Why: These are the moves of the worked example in the order it makes them, and each one is set up by the one before it. The left side is 1; the right side is 1 times 2 over 2, which is 1.
Worked example
Prove the closed form for the running total of the first n positive integers.
\[ P(n): \; \sum_{i=1}^{n} i = \frac{n(n+1)}{2} \]
Base case: check n equal to 1
Why: The left side is 1; the right side is 1 times 2 over 2, which is 1. They match.
\( \sum_{i=1}^{1} i = 1 = \frac{1 \cdot 2}{2} \)
Assume the induction hypothesis at k
Why: We grant ourselves the formula at k; this is the gift we build on.
\( \sum_{i=1}^{k} i = \frac{k(k+1)}{2} \)
Add the next term, k plus 1, to both sides
Why: The sum to k plus one is the sum to k plus the new term; substitute the hypothesis for the first piece.
\( \sum_{i=1}^{k+1} i = \frac{k(k+1)}{2} + (k+1) \)
Factor out k plus 1 and simplify
Why: Pull out the common factor and combine over 2 to reach the target shape.
\( = (k+1)\cdot\frac{k+2}{2} = \frac{(k+1)(k+2)}{2} \)
Verify this is the claim at k plus 1
Why: The result equals the formula with n replaced by k plus one, so the step closes. Base plus step gives the claim for all n at least 1.
\( \frac{(k+1)\big((k+1)+1\big)}{2} = \frac{(k+1)(k+2)}{2} \; \checkmark \)
Picture it
Animation
Shows: Each line of the worked example "the sum one through n", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: The result equals the formula with n replaced by k plus one, so the step closes. Base plus step gives the claim for all n at least 1.
Step zero
Discussion prompt
Worked example: the sum of the first n odd numbers — before any calculation: what is the plan? Name the moves in order, in plain English, without doing the arithmetic.
Hint: It starts with: Base case: n equal to 1
Answer:
Worked example
Add up the first n odd numbers and watch a perfect square appear.
\[ P(n): \; \sum_{i=1}^{n} (2i-1) = n^{2} \]
Base case: n equal to 1
Why: The first odd number is 1, and 1 squared is 1.
\( \sum_{i=1}^{1}(2i-1) = 1 = 1^{2} \)
Assume the hypothesis at k
Why: Grant the formula for the first k odd numbers.
\( \sum_{i=1}^{k}(2i-1) = k^{2} \)
Add the next odd number, 2k plus 1
Why: The next odd number is 2 times k plus one, minus one, which equals 2k plus 1; add it to both sides.
\( \sum_{i=1}^{k+1}(2i-1) = k^{2} + (2k+1) \)
Recognize the perfect-square trinomial
Why: The expression k squared plus 2k plus 1 factors as k plus one, squared.
\( k^{2} + 2k + 1 = (k+1)^{2} \)
Verify this is the claim at k plus 1
Why: We obtained exactly k plus one, squared, the formula at k plus one. The induction is complete for all n at least 1.
\( \sum_{i=1}^{k+1}(2i-1) = (k+1)^{2} \; \checkmark \)
Picture it
Animation
Shows: Each line of the worked example "the sum of the first n odd numbers", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: We obtained exactly k plus one, squared, the formula at k plus one. The induction is complete for all n at least 1.
Estimation
Predict first
A geometric sum of powers of two always lands one below the next power.
Commit before you compute: what does Worked example: powers of two, one short of the next come out to? A rough magnitude and the right form is enough — the point is to have something concrete to be wrong about.
Correct: Verify this is the claim at k plus 1
Why: A prediction you can defend turns the computation into a check rather than a leap of faith — and an answer that contradicts it is caught on the spot. The result is 2 to the quantity k plus one, plus one, minus one, matching the formula at k plus one.
Worked example
A geometric sum of powers of two always lands one below the next power.
\[ P(n): \; \sum_{i=0}^{n} 2^{i} = 2^{\,n+1} - 1 \]
Base case: n equal to 0
Why: The single term is 2 to the zero, which is 1; the right side is 2 to the one, minus one, also 1.
\( \sum_{i=0}^{0} 2^{i} = 1 = 2^{1} - 1 \)
Assume the hypothesis at k
Why: Grant the closed form up to the term 2 to the k.
\( \sum_{i=0}^{k} 2^{i} = 2^{\,k+1} - 1 \)
Add the next power of two
Why: Append the term 2 to the k plus one to both sides.
\( \sum_{i=0}^{k+1} 2^{i} = \big(2^{\,k+1} - 1\big) + 2^{\,k+1} \)
Combine the two equal powers
Why: Two copies of 2 to the k plus one make 2 to the k plus two.
\( 2\cdot 2^{\,k+1} - 1 = 2^{\,k+2} - 1 \)
Verify this is the claim at k plus 1
Why: The result is 2 to the quantity k plus one, plus one, minus one, matching the formula at k plus one. Done for all n at least 0.
\( 2^{\,(k+1)+1} - 1 = 2^{\,k+2} - 1 \; \checkmark \)
Picture it
Animation
Shows: Each line of the worked example "powers of two, one short of the next", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: The result is 2 to the quantity k plus one, plus one, minus one, matching the formula at k plus one. Done for all n at least 0.
Missing information
Discussion prompt
The sum of the first n cubes equals the square of the n-th triangular number.
What do you need to know — or decide — before the first line can be written? List everything the problem has to hand you.
Hint: Anything you would have to invent to get started is a thing the problem must supply.
Answer:
The left side is 1 cubed, which is 1; the right side is the square of 1 times 2 over 2, also 1.
Worked example
The sum of the first n cubes equals the square of the n-th triangular number.
\[ P(n): \; \sum_{i=1}^{n} i^{3} = \left(\frac{n(n+1)}{2}\right)^{2} \]
Base case: n equal to 1
Why: The left side is 1 cubed, which is 1; the right side is the square of 1 times 2 over 2, also 1.
\( 1^{3} = 1 = \left(\tfrac{1\cdot 2}{2}\right)^{2} \)
Assume the hypothesis at k
Why: Grant the closed form for the first k cubes.
\( \sum_{i=1}^{k} i^{3} = \left(\frac{k(k+1)}{2}\right)^{2} \)
Add the next cube
Why: Append the term k plus one, cubed, to both sides.
\( \sum_{i=1}^{k+1} i^{3} = \left(\frac{k(k+1)}{2}\right)^{2} + (k+1)^{3} \)
Factor out the common square
Why: Pull out one quarter of k plus one, squared; the bracket becomes k squared plus 4k plus 4, a perfect square.
\( = \frac{(k+1)^{2}}{4}\big(k^{2}+4k+4\big) = \left(\frac{(k+1)(k+2)}{2}\right)^{2} \)
Verify this is the claim at k plus 1
Why: The result is the triangular-number-squared formula with n equal to k plus one. The identity holds for all n at least 1.
\( \left(\frac{(k+1)\big((k+1)+1\big)}{2}\right)^{2} \; \checkmark \)
Picture it
Animation
Shows: Each line of the worked example "the sum of the first n cubes", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: The result is the triangular-number-squared formula with n equal to k plus one. The identity holds for all n at least 1.
Fill the middle
Fill in the blanks
From Worked example: the Tower of Hanoi move count — finish the line. Write what belongs on the right of the equals sign before you look.
P(n): \; T(n) = 2^{n} - 1
Why: Producing the right-hand side unprompted is the difference between recognising this line and being able to use it. No disks means no moves, and 2 to the zero, minus one, is 0.
Worked example
The standard recursive strategy moves n disks with a count that satisfies a simple recurrence. Solve it in closed form.
\( T(0) = 0, \qquad T(n) = 2\,T(n-1) + 1 \)
Claim: the closed form is one less than a power of two.
\( P(n): \; T(n) = 2^{n} - 1 \)
Base case: n equal to 0
Why: No disks means no moves, and 2 to the zero, minus one, is 0.
\( T(0) = 0 = 2^{0} - 1 \)
Assume the hypothesis at k
Why: Grant the closed form for k disks.
\( T(k) = 2^{k} - 1 \)
Apply the recurrence at k plus 1
Why: Substitute the hypothesis into the recurrence for T at k plus one.
\( T(k+1) = 2\big(2^{k} - 1\big) + 1 = 2^{\,k+1} - 1 \)
Verify this is the claim at k plus 1
Why: The result is 2 to the k plus one, minus one, matching the formula at k plus one. So n disks require that many moves, for all n at least 0.
\( T(k+1) = 2^{\,k+1} - 1 \; \checkmark \)
Picture it
Animation
Shows: Each line of the worked example "the Tower of Hanoi move count", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: The result is 2 to the k plus one, minus one, matching the formula at k plus one. So n disks require that many moves, for all n at least 0.
Concept
Nothing about induction requires an equation. The claim at each n can be an inequality, a divisibility, or any proposition.
For inequalities the step usually chains: bound the next quantity by something involving the k case, then apply the hypothesis.
The discipline is identical; the only change is that a middle line reads at most instead of equals.
Step zero
Discussion prompt
Worked example: Bernoulli's inequality — before any calculation: what is the plan? Name the moves in order, in plain English, without doing the arithmetic.
Hint: It starts with: Base case: n equal to 0
Answer:
Worked example
A workhorse inequality: a power of one plus x is at least its linear approximation.
\[ P(n): \; (1+x)^{n} \geq 1 + nx \qquad (x \geq -1) \]
Base case: n equal to 0
Why: Both sides equal 1, so the inequality holds as equality.
\( (1+x)^{0} = 1 \geq 1 + 0\cdot x \)
Assume the hypothesis at k
Why: Grant the inequality at exponent k.
\( (1+x)^{k} \geq 1 + kx \)
Multiply both sides by one plus x
Why: Because x is at least negative one, the factor one plus x is nonnegative, so the inequality direction is preserved.
\( (1+x)^{k+1} \geq (1 + kx)(1 + x) \)
Expand and drop the nonnegative square term
Why: Expanding gives 1 plus k plus one times x, plus k times x squared; that last term is nonnegative, so dropping it only weakens the bound.
\( (1+kx)(1+x) = 1 + (k+1)x + kx^{2} \geq 1 + (k+1)x \)
Verify this is the claim at k plus 1
Why: Chaining the last two lines gives the inequality at exponent k plus one, exactly the claim.
\( (1+x)^{k+1} \geq 1 + (k+1)x \; \checkmark \)
Picture it
Animation
Shows: Each line of the worked example "Bernoulli's inequality", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: Chaining the last two lines gives the inequality at exponent k plus one, exactly the claim.
Worked example
A divisibility claim, proved by the same base-and-step ritual.
\[ P(n): \; 3 \mid (n^{3} - n) \]
Base case: n equal to 0
Why: Zero cubed minus zero is 0, and 3 divides 0.
\( n^{3} - n = 0, \qquad 3 \mid 0 \)
Assume the hypothesis at k
Why: Grant that 3 divides k cubed minus k; write that quantity as 3 times an integer m.
\( k^{3} - k = 3m \text{ for some integer } m \)
Expand the k-plus-one expression
Why: Compute the cube of k plus one, minus k plus one, and group the known multiple.
\( (k+1)^{3} - (k+1) = (k^{3} - k) + 3k^{2} + 3k \)
Substitute the hypothesis and factor out 3
Why: Replace k cubed minus k by 3m; every remaining term already carries a factor of 3.
\( = 3m + 3k^{2} + 3k = 3\big(m + k^{2} + k\big) \)
Verify this is the claim at k plus 1
Why: The expression is 3 times an integer, so 3 divides it. The divisibility holds for all n at least 0.
\( 3 \mid \big((k+1)^{3} - (k+1)\big) \; \checkmark \)
Picture it
Animation
Shows: Each line of the worked example "three divides n cubed minus n", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: The expression is 3 times an integer, so 3 divides it. The divisibility holds for all n at least 0.
Anomaly
Predict first
A student writes this, and it looks reasonable:
Bogus claim: every natural number equals its successor. Watch a valid-looking step.
It is wrong. Say what breaks — and say it before you turn the page.
Correct: If k equals k plus one, then adding one gives k plus one equals k plus two, which is the case at k plus one.
The step only promises that one case implies the next. It never asserts that any case is true.
Why: If k equals k plus one, then adding one gives k plus one equals k plus two, which is the case at k plus one. The implication is logically valid.
Trap
Bogus claim: every natural number equals its successor. Watch a valid-looking step.
\( P(n): \; n = n + 1 \)
Do the step: assume the case at k, add 1 to both sides
Why: If k equals k plus one, then adding one gives k plus one equals k plus two, which is the case at k plus one. The implication is logically valid.
\( k = k+1 \;\Rightarrow\; k+1 = k+2 \)
The implication really is true, so the step checks out, yet the conclusion is absurd.
The step only promises that one case implies the next. It never asserts that any case is true.
Demand the base case
Why: The case at 0 says 0 equals 1, which is false. With no first domino the chain never starts and nothing is proved.
\( P(0): \; 0 = 1 \quad \text{is false} \)
Lesson: a valid step with a false base proves nothing. Always establish the base case.
Notation
Annotate
From Trap: the inductive step with no base case — read this one piece at a time. What is each part doing?
On: \( P(n): \; n = n + 1 \)
Anomaly
Predict first
A student writes this, and it looks reasonable:
A tempting way to prove an identity: write the equation you want and simplify both sides until they agree.
It is wrong. Say what breaks — and say it before you turn the page.
Correct: Writing the goal and reducing both sides to zero equals zero feels convincing.
A proof flows from what you already know to what you want. Never open by asserting the target equality.
Why: Writing the goal and reducing both sides to zero equals zero feels convincing.
Trap
A tempting way to prove an identity: write the equation you want and simplify both sides until they agree.
\( \text{Target } P(k+1): \; \sum_{i=1}^{k+1} i = \frac{(k+1)(k+2)}{2} \)
Manipulate the target equation itself
Why: Writing the goal and reducing both sides to zero equals zero feels convincing.
\( \sum_{i=1}^{k+1} i - \frac{(k+1)(k+2)}{2} \stackrel{?}{=} 0 \)
But you began by writing down the very thing to be proved as if it were already true. That is assuming the conclusion.
Derive forward from the hypothesis instead
Why: Start from the known case at k, add the new term, and transform until the k-plus-one statement emerges as a consequence.
\( \sum_{i=1}^{k+1} i = \underbrace{\sum_{i=1}^{k} i}_{\text{hypothesis}} + (k+1) \)
A proof flows from what you already know to what you want. Never open by asserting the target equality.
Concept
The base is the smallest index for which you claim the statement. Everything below it is simply not asserted.
Some claims are false for small inputs and only start holding later; then your base is that first true index.
\( \text{e.g. } 2^{n} > n^{2} \text{ holds for } n \geq 5, \text{ not before} \)
Match the base to the claim. Starting too low can force you to prove something false; starting too high leaves a gap.
Fill the middle
Fill in the blanks
From Trap: skipping the base check — finish the line. Write what belongs on the right of the equals sign before you look.
\frac\frac{(k+1)(k+2)}{2} + 1___ + 1 + (k+1) = ___
Why: Producing the right-hand side unprompted is the difference between recognising this line and being able to use it. Assuming the case at k, adding k plus one to both sides reproduces the same spurious plus-one at k plus one.
Trap
Alleged claim: for every n at least 1, the sum one through n equals the usual formula plus one.
\( Q(n): \; \sum_{i=1}^{n} i = \frac{n(n+1)}{2} + 1 \)
The step goes through untouched
Why: Assuming the case at k, adding k plus one to both sides reproduces the same spurious plus-one at k plus one. The step is self-consistent.
\( \frac{k(k+1)}{2} + 1 + (k+1) = \frac{(k+1)(k+2)}{2} + 1 \)
A self-consistent step can propagate a falsehood forever if you never check the base.
Check the base at n equal to 1
Why: The left side is 1; the right side is 1 plus 1, which is 2. They disagree, so the case at 1 is false.
\( \sum_{i=1}^{1} i = 1 \neq 2 = \frac{1\cdot 2}{2} + 1 \)
The base fails, so the claim is false. The base case is not a formality; it anchors the whole chain to reality.
Translation
\( \sum_{i=1}^{1} i = 1 \neq 2 = \frac{1\cdot 2}{2} + 1 \)
Draw it
Translate both ways. First write the expression above as a sentence with no symbols in it at all. Then cover it, and write your sentence back as notation. If the two versions disagree, the disagreement is the thing to fix.
Concept
Sometimes proving the next case needs facts about cases far earlier than k, not just the immediately preceding one.
Example: to factor n into primes you split it as a times b, where a and b are both smaller than n but usually far below k.
Ordinary induction only hands you the single case k. When you need the whole history below n, reach for strong induction.
Concept
Strong induction hands you every earlier case at once, not just the last one.
\[ \Big(\forall n\,\big[\,(\forall k < n\; P(k)) \rightarrow P(n)\,\big]\Big) \;\rightarrow\; \forall n\, P(n) \]
The hypothesis is now: the claim holds for all indices below n. From that richer assumption you prove the claim at n.
There is no separate base line written here: the smallest n has no cases below it, so its argument must stand alone, which is the base in disguise.
Concept
A surprising tactic: sometimes a claim will not induct, but a stronger claim will.
A stronger statement means a stronger induction hypothesis, and that extra assumption is exactly what the step needs to push through.
loaded induction — Proving a sharper statement than you were asked for, because the stronger hypothesis carries the inductive step where the weaker one stalls.
Definition probe
Sort into buckets
Every line below is part of the definition of predicate over the naturals or of loaded induction — one or the other, never both. Put each where it belongs.
Intuition
Weak induction lets each domino feel only the one right behind it.
Strong induction lets each domino feel the combined push of every domino that has already fallen.
Same guarantee, more leverage: use it whenever the next case decomposes into arbitrary smaller cases.
Estimation
Predict first
Show every integer from 2 upward is a product of primes.
Commit before you compute: what does Worked example: every integer at least 2 has a prime… come out to? A rough magnitude and the right form is enough — the point is to have something concrete to be wrong about.
Correct: Verify this is the claim at n
Why: A prediction you can defend turns the computation into a check rather than a leap of faith — and an answer that contradicts it is caught on the spot. In both cases n is written as a product of primes, using only cases strictly below n, exactly what strong induction supplies.
Worked example
Show every integer from 2 upward is a product of primes.
\[ P(n): \; n \text{ is a product of one or more primes} \]
Base case: n equal to 2
Why: Two is prime, so it is a product of a single prime, itself.
\( 2 = 2 \quad (\text{prime}) \)
Strong hypothesis: assume the claim for all j with 2 at most j, and j below n
Why: Every integer below n and at least 2 already factors into primes.
Case 1: n is prime
Why: Then n is a one-factor product of primes and we are done immediately.
Case 2: n is composite, so it splits below n
Why: Write n as a times b with both factors at least 2 and strictly below n; the strong hypothesis factors each, and we concatenate.
\( n = a\cdot b, \quad a = \textstyle\prod_i p_i, \; b = \textstyle\prod_j q_j \)
Verify this is the claim at n
Why: In both cases n is written as a product of primes, using only cases strictly below n, exactly what strong induction supplies. Holds for all n at least 2.
\( n = \Big(\textstyle\prod_i p_i\Big)\Big(\textstyle\prod_j q_j\Big) \; \checkmark \)
Picture it
Animation
Shows: Each line of the worked example "every integer at least 2 has a prime factorization", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: In both cases n is written as a product of primes, using only cases strictly below n, exactly what strong induction supplies. Holds for all n at least 2.
Ranking
Put in order
Put the moves of Worked example: every amount of 12 or more, using 4s and 5s into the order they have to happen.
Why: These are the moves of the worked example in the order it makes them, and each one is set up by the one before it. Twelve is three fours; thirteen is two fours and a five; fourteen is a four and two fives; fifteen is three fives.
Worked example
Show every integer at least 12 is a nonnegative combination of 4 and 5.
\[ P(n): \; n = 4a + 5b \text{ for some integers } a,b \geq 0 \]
Establish four base cases: 12, 13, 14, 15
Why: Twelve is three fours; thirteen is two fours and a five; fourteen is a four and two fives; fifteen is three fives. We need four bases because the step reaches back by 4.
\( 12 = 3(4),\; 13 = 2(4){+}5,\; 14 = 4{+}2(5),\; 15 = 3(5) \)
Strong hypothesis: assume the claim for all reachable amounts below n, with n at least 16
Why: Every expressible amount below n is already a combination of fours and fives.
Reduce n by 4
Why: Since n is at least 16, the amount n minus 4 is at least 12 and below n, so it is covered by the hypothesis.
\( n - 4 = 4a + 5b \)
Add one more four
Why: Adding a four to that combination expresses n itself.
\( n = 4(a+1) + 5b \)
Verify this is the claim at n
Why: n is a nonnegative combination of 4 and 5, using the case n minus 4 that the four base cases and the hypothesis guarantee. Holds for all n at least 12.
\( n = 4(a+1) + 5b \; \checkmark \)
Picture it
Animation
Shows: Each line of the worked example "every amount of 12 or more, using 4s and 5s", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: n is a nonnegative combination of 4 and 5, using the case n minus 4 that the four base cases and the hypothesis guarantee. Holds for all n at least 12.
Fill the middle
Fill in the blanks
From Worked example: bounding the sum of reciprocal squares — finish the line. Write what belongs on the right of the equals sign before you look.
\sum_1}^{n}\frac{1}{i^{2}} \leq 2 - \frac{1}{n} < 2 \; \checkmark}
Why: Producing the right-hand side unprompted is the difference between recognising this line and being able to use it. A sharper bound gives the step something to push against; this is loaded induction.
Worked example
Goal: bound the partial sums of reciprocal squares below 2. The bare bound will not induct, so we prove a stronger statement instead.
\( \text{Want: } \sum_{i=1}^{n} \frac{1}{i^{2}} < 2 \)
Strengthen the claim with a slack term
Why: A sharper bound gives the step something to push against; this is loaded induction.
\( P(n): \; \sum_{i=1}^{n} \frac{1}{i^{2}} \leq 2 - \frac{1}{n} \)
Base case: n equal to 1
Why: The left side is 1; the right side is 2 minus 1, which is 1. Equality holds.
\( \frac{1}{1^{2}} = 1 = 2 - \frac{1}{1} \)
Assume the strengthened hypothesis at k
Why: Grant the sharper bound at k.
\( \sum_{i=1}^{k} \frac{1}{i^{2}} \leq 2 - \frac{1}{k} \)
Add the next term and estimate it
Why: Add the reciprocal square at k plus one, then use the telescoping bound that it is at most one over k minus one over k plus one.
\( \frac{1}{(k+1)^{2}} \leq \frac{1}{k(k+1)} = \frac{1}{k} - \frac{1}{k+1} \)
Combine to reach the k-plus-one bound
Why: The minus one over k from the hypothesis cancels the plus one over k from the estimate, leaving the sharper bound at k plus one.
\( \sum_{i=1}^{k+1}\frac{1}{i^{2}} \leq 2 - \frac{1}{k} + \frac{1}{k} - \frac{1}{k+1} = 2 - \frac{1}{k+1} \)
Verify and recover the goal
Why: That is exactly the strengthened claim at k plus one; and since 2 minus one over n is below 2, the original bound follows for every n at least 1.
\( \sum_{i=1}^{n}\frac{1}{i^{2}} \leq 2 - \frac{1}{n} < 2 \; \checkmark \)
Picture it
Animation
Shows: Each line of the worked example "bounding the sum of reciprocal squares", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: That is exactly the strengthened claim at k plus one; and since 2 minus one over n is below 2, the original bound follows for every n at least 1.
Anomaly
Predict first
A student writes this, and it looks reasonable:
Same postage claim, but a hasty proof verifies only n equal to 12 as the base.
It is wrong. Say what breaks — and say it before you turn the page.
Correct: The step sends n back to n minus 4 and adds a four.
Then every n at least 16 lands on one of the four seeds after repeated subtraction. Verify the smallest few by hand to be safe.
Why: The step sends n back to n minus 4 and adds a four.
Trap
Same postage claim, but a hasty proof verifies only n equal to 12 as the base.
\( P(n): \; n = 4a + 5b, \quad a,b \geq 0 \)
Reduce n by 4 as before
Why: The step sends n back to n minus 4 and adds a four.
\( n \;\to\; n - 4 \)
Trace what a single base actually reaches
Why: Starting from 12 and jumping by 4 only produces 12, 16, 20, and so on. The amounts 13, 14, and 15 are never proved.
\( 12,\,16,\,20,\dots \text{ only} \)
Match the number of base cases to the step's reach
Why: The step reduces by 4, so you need four consecutive base cases to seed every residue: 12, 13, 14, 15.
\( n-4 \geq 12 \iff n \geq 16 \)
Then every n at least 16 lands on one of the four seeds after repeated subtraction. Verify the smallest few by hand to be safe.
Break the constraint
Discussion prompt
The rule this trap just fixed:
The step reduces by 4, so you need four consecutive base cases to seed every residue: 12, 13, 14, 15.
Now break it on purpose. Build a case that violates it and follow the consequences until something visibly fails. Where does the failure first show up — and would you have noticed it if you had not been looking?
Hint: The dangerous rules are the ones whose violation still produces an answer. If yours fails loudly, try to find one that fails quietly.
Answer:
The step sends n back to n minus 4 and adds a four.
Concept
well-ordering principle — Every nonempty set of natural numbers has a least element.
It sounds obvious, yet it is logically as strong as induction, and it is often the cleanest tool for existence and minimality arguments.
\[ \forall S \subseteq \mathbb{N},\; S \neq \varnothing \;\rightarrow\; \exists m \in S\; \forall s \in S\; (m \leq s) \]
Picture it
Figure (svg): Four dots stepping downward left to right, connected by arrows, ending above a baseline labeled hits bottom.
Discussion prompt
Read the picture before the words. What is this showing, and what is the one thing it is built to make obvious? Commit to an answer, then read on.
Hint: Name the parts, then say what changes between them — and if nothing changes, say what is being held still.
Answer:
You cannot strictly decrease through the naturals forever; every descending path hits bottom.
Intuition
You cannot strictly decrease through the naturals forever; every descending path hits bottom.
Figure (svg): Four dots stepping downward left to right, connected by arrows, ending above a baseline labeled hits bottom.
Equivalently: there is no infinite strictly decreasing sequence of natural numbers.
This is the engine behind minimal counterexample proofs: assume a counterexample exists, take the smallest one, and derive a smaller one, which is impossible.
Worked example
Use well-ordering directly, through a smallest-witness argument.
\[ \text{Claim: every } n \geq 2 \text{ has a prime divisor} \]
Collect the divisors of n that are at least 2
Why: n divides itself, so n belongs to this set; thus it is a nonempty set of naturals.
\( D = \{ d \geq 2 : d \mid n \}, \quad n \in D \)
By well-ordering, D has a least element p
Why: A nonempty subset of the naturals has a smallest member, so the smallest divisor of n that is at least 2 exists.
\( p = \min D \)
Argue that p is prime
Why: If p had a divisor e strictly between 1 and p, then e would also divide n and sit in D below p, contradicting minimality.
\( e \mid p,\; 1 < e < p \;\Rightarrow\; e \in D, \; e < p \)
Verify the claim
Why: So p has no divisor strictly between 1 and itself; p is prime and divides n. Every n at least 2 has a prime divisor.
\( p \text{ prime}, \; p \mid n \; \checkmark \)
Picture it
Animation
Shows: Each line of the worked example "every integer at least 2 has a prime divisor", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: So p has no divisor strictly between 1 and itself; p is prime and divides n. Every n at least 2 has a prime divisor.
Step zero
Discussion prompt
Worked example: existence of quotient and remainder — before any calculation: what is the plan? Name the moves in order, in plain English, without doing the arithmetic.
Hint: It starts with: Form the set of nonnegative remainders
Answer:
Worked example
Well-ordering delivers the division algorithm cleanly.
\[ \text{Given } a \geq 0,\; d > 0, \text{ find } q,r: \; a = dq + r,\; 0 \leq r < d \]
Form the set of nonnegative remainders
Why: Consider every value a minus d times q that is at least 0, as q ranges over the integers.
\( R = \{ a - dq \geq 0 : q \in \mathbb{Z} \} \)
R is nonempty
Why: Taking q equal to 0 gives a, which is at least 0, so a belongs to R.
\( a = a - d\cdot 0 \in R \)
By well-ordering, R has a least element r
Why: Let r be the smallest nonnegative remainder, with its witness q.
\( r = \min R = a - dq \)
Show that r is below d
Why: If r were at least d, then r minus d equals a minus d times q plus one, still nonnegative and smaller, contradicting minimality of r.
\( r \geq d \Rightarrow r - d = a - d(q+1) \in R, \; r-d < r \)
Verify the result
Why: So the remainder is nonnegative and below d, and a equals d times q plus r, the required quotient and remainder. Existence is proved.
\( a = dq + r,\; 0 \leq r < d \; \checkmark \)
Picture it
Animation
Shows: Each line of the worked example "existence of quotient and remainder", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: So the remainder is nonnegative and below d, and a equals d times q plus r, the required quotient and remainder. Existence is proved.
Anomaly
Predict first
A student writes this, and it looks reasonable:
Tempting overreach: every nonempty set has a least element, so just pick the smallest one.
It is wrong. Say what breaks — and say it before you turn the page.
Correct: The set of all integers has no least element; you can always go more negative.
Well-ordering is a special property of the natural numbers, not a universal law about ordered sets.
Why: The set of all integers has no least element; you can always go more negative.
Trap
Tempting overreach: every nonempty set has a least element, so just pick the smallest one.
Try it on the integers
Why: The set of all integers has no least element; you can always go more negative.
\( \mathbb{Z}: \; \dots, -3, -2, -1 \text{ descends forever} \)
Try it on the positive rationals
Why: The positive rationals have no least element either; between 0 and any candidate lies a smaller positive rational.
\( 0 < \tfrac{q}{2} < q \)
Well-ordering is a special property of the natural numbers, not a universal law about ordered sets.
Repair the argument
Why: Apply well-ordering only to a set of naturals: numerators, exponents, or counts extracted from the problem.
\( S \subseteq \mathbb{N},\; S \neq \varnothing \;\Rightarrow\; \min S \text{ exists} \)
Before you say take the smallest, confirm your set genuinely lives inside the naturals.
Notation
Annotate
From Trap: well-ordering only holds for the naturals — read this one piece at a time. What is each part doing?
On: \( S \subseteq \mathbb{N},\; S \neq \varnothing \;\Rightarrow\; \min S \text{ exists} \)
Concept
Weak induction, strong induction, and well-ordering are logically equivalent over the naturals: each one implies the others.
\[ \text{weak induction} \;\Longleftrightarrow\; \text{strong induction} \;\Longleftrightarrow\; \text{well-ordering} \]
So prove by strong induction and take a minimal counterexample are interchangeable styles. Pick whichever reads more cleanly.
Intuition
A minimal-counterexample proof secretly runs the well-ordering principle.
Assume the claim fails somewhere. Collect all failing indices; that is a nonempty set of naturals.
Well-ordering hands you the smallest failure. You then manufacture a still-smaller failure and reach a contradiction, so no failure can exist.
Hypothesis
Predict first
Worked example: well-ordering implies induction is about to be worked. State your hypothesis first: which rule or definition decides this one, and what is the first move it forces? Then watch whether the example agrees with you.
Correct: Suppose, for contradiction, the claim fails somewhere
Why: Let F be the set of naturals where the claim is false, and assume F is nonempty.
A hypothesis you wrote down is falsifiable; a vague sense of how it will go is not. If the example opens somewhere else, that gap is the thing worth chasing.
Worked example
Show that if the base and step both hold, well-ordering forces the claim for all n.
\[ \text{Assume } P(0) \text{ and } \forall k\,\big(P(k) \rightarrow P(k+1)\big) \]
Suppose, for contradiction, the claim fails somewhere
Why: Let F be the set of naturals where the claim is false, and assume F is nonempty.
\( F = \{ n \in \mathbb{N} : \neg P(n) \}, \quad F \neq \varnothing \)
Take the least failure m by well-ordering
Why: F is a nonempty set of naturals, so it has a smallest element m.
\( m = \min F \)
Rule out m equal to 0
Why: The base case makes the claim true at 0, so 0 is not in F; therefore m is at least 1 and m minus 1 is a natural number.
\( m \geq 1 \)
Use minimality, then the step
Why: Since m is least in F, the case m minus 1 is not in F, so it holds; the inductive step turns it into the case m.
\( P(m-1) \text{ true} \;\xrightarrow{\text{step}}\; P(m) \text{ true} \)
Verify the contradiction
Why: The case m is true, yet m was chosen as a point of failure, a contradiction. So F is empty and the claim holds for all n.
\( m \in F \text{ and } P(m) \text{ true} \;\Rightarrow\; \bot \; \checkmark \)
Picture it
Animation
Shows: Each line of the worked example "well-ordering implies induction", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: The case m is true, yet m was chosen as a point of failure, a contradiction. So F is empty and the claim holds for all n.
Concept
Many objects in computer science are built by a finite rulebook: start from atoms, then apply constructors.
inductively defined set — A set specified by base elements plus construction rules, containing exactly the objects reachable by finitely many rule applications.
Examples: the naturals (zero, then successor), lists (empty, then prepend), and well-formed formulas (variables, then connectives).
Matching
Match the pairs
Match each term to the definition this lesson gave it — not the one you would guess from the word.
Why: These are the working definitions of predicate over the naturals, induction hypothesis, loaded induction, well-ordering principle, inductively defined set as Proof Techniques II: Induction, Strong Induction & Well-Ordering uses them. Pairing them correctly is the test of whether you could state each one with the slide switched off.
Intuition
You define a function on such a set by recursion: give its value on the atoms, then on each constructor in terms of the parts.
You prove a property on the same set by structural induction: check the atoms, then check each constructor assuming the property on the parts.
Definition and proof march in lockstep over the very same rulebook. This is the Curry-Howard flavor: programs and proofs share one structure.
Concept
To prove a property holds for every element of an inductively defined set:
Base cases
Why: Prove the property for each atomic constructor.
Constructor steps
Why: For each building rule, assume the property on the immediate sub-parts and prove it for the assembled object.
The assumption on the sub-parts is the structural induction hypothesis, the same gift, indexed by structure instead of by a number.
Missing information
Discussion prompt
Let the left-count and right-count tally the parentheses in a formula. Claim: they are always equal.
What do you need to know — or decide — before the first line can be written? List everything the problem has to hand you.
Hint: Anything you would have to invent to get started is a thing the problem must supply.
Answer:
A bare variable contains no parentheses, so both counts are 0.
Worked example
Let the left-count and right-count tally the parentheses in a formula. Claim: they are always equal.
\[ P(\varphi): \; L(\varphi) = R(\varphi) \]
Base case: the formula is a propositional variable
Why: A bare variable contains no parentheses, so both counts are 0.
\( L(p) = 0 = R(p) \)
Constructor: negation wraps a formula
Why: The negation form adds one left and one right parenthesis to the sub-formula; the hypothesis on it keeps the counts equal.
\( L = L(\varphi){+}1,\quad R = R(\varphi){+}1 \)
Constructor: a binary connective joins two formulas
Why: This form adds one left and one right parenthesis around the two sub-formulas; assume the property on both.
\( L = L(\varphi){+}L(\psi){+}1, \quad R = R(\varphi){+}R(\psi){+}1 \)
Verify equality is preserved
Why: Using the equalities on the sub-formulas, both new totals agree. Every well-formed formula satisfies the property by structural induction.
\( L(\varphi){+}L(\psi){+}1 = R(\varphi){+}R(\psi){+}1 \; \checkmark \)
Picture it
Animation
Shows: Each line of the worked example "formulas have balanced parenthesis counts", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: Using the equalities on the sub-formulas, both new totals agree. Every well-formed formula satisfies the property by structural induction.
Estimation
Predict first
A full binary tree is either a single leaf, or an internal node joining two full binary subtrees. Count leaves and internal nodes.
Commit before you compute: what does Worked example: full binary trees, leaves versus internal… come out to? A rough magnitude and the right form is enough — the point is to have something concrete to be wrong about.
Correct: Verify equality
Why: A prediction you can defend turns the computation into a check rather than a leap of faith — and an answer that contradicts it is caught on the spot. That equals the tree's internal count plus one, the claim for the whole tree.
Worked example
A full binary tree is either a single leaf, or an internal node joining two full binary subtrees. Count leaves and internal nodes.
\[ \text{Claim: } \ell(T) = i(T) + 1 \]
Base case: the tree is a single leaf
Why: One leaf, no internal nodes: 1 equals 0 plus 1.
\( \ell = 1,\; i = 0,\; 1 = 0 + 1 \)
Constructor: join two subtrees under a new internal node
Why: Assume the property on the left and right subtrees; the whole tree's leaves are their leaf counts summed, and its internal nodes are their internal counts summed plus one.
\( \ell(T) = \ell(L){+}\ell(R), \quad i(T) = i(L){+}i(R){+}1 \)
Substitute both hypotheses
Why: Replace each subtree's leaf count by its internal count plus one.
\( \ell(T) = (i(L){+}1) + (i(R){+}1) = i(L){+}i(R){+}2 \)
Verify equality
Why: That equals the tree's internal count plus one, the claim for the whole tree. Structural induction completes the proof.
\( \ell(T) = i(T) + 1 \; \checkmark \)
Picture it
Animation
Shows: Each line of the worked example "full binary trees, leaves versus internal nodes", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: That equals the tree's internal count plus one, the claim for the whole tree. Structural induction completes the proof.
Anomaly
Predict first
A student writes this, and it looks reasonable:
Bogus claim: in every group of n horses, all have the same color. Proof by induction on n.
It is wrong. Say what breaks — and say it before you turn the page.
Correct: A single horse trivially matches itself.
The step silently assumes n is at least 2, so the overlap is nonempty, but the base only supplies n equal to 1. The missing 1-to-2 link breaks the chain.
Why: A single horse trivially matches itself.
Trap
Bogus claim: in every group of n horses, all have the same color. Proof by induction on n.
\( P(n): \; \text{any } n \text{ horses share one color} \)
The base at n equal to 1 is fine
Why: A single horse trivially matches itself.
The step overlaps two groups of size n
Why: Given n plus one horses, drop the last to get n same-colored horses, drop the first to get another n, and claim the shared middle forces all equal.
\( \{1,\dots,n\} \text{ and } \{2,\dots,n+1\} \text{ overlap} \)
Test the step at its first use, going from 1 to 2
Why: For two horses the two groups are the singletons at position 1 and position 2. They do not overlap, so nothing links the two colors.
\( \{1\} \cap \{2\} = \varnothing \)
The step silently assumes n is at least 2, so the overlap is nonempty, but the base only supplies n equal to 1. The missing 1-to-2 link breaks the chain.
Lesson: a step must be valid at the very first index where it is used. Check the base-to-next transition explicitly.
Two truths and a lie
Sort into buckets
Some of these hold up and some are the exact mistakes this lesson is built to prevent. Sort them.
prove an identity: write the equation you want and simplify both sides until they agree.Concept
loop invariant — A statement true just before the loop and preserved by every iteration; by induction on the iteration count it therefore holds at loop exit.
Establishing the invariant before the first pass is the base case. Showing that one pass preserves it is the inductive step.
At exit, induction guarantees the invariant together with the exit condition, which give correctness. This is how we reason about every loop and recursion.
Step zero
Discussion prompt
Worked example: correctness of square-and-multiply — before any calculation: what is the plan? Name the moves in order, in plain English, without doing the arithmetic.
Hint: It starts with: Base case: n equal to 0
Answer:
Worked example
The routine squares the result of the halved exponent, multiplying by the base once more when the exponent is odd. Prove it correct by strong induction on the exponent.
power(x, n):
if n == 0: return 1
h = power(x, n // 2)
if n is even: return h * h
else: return x * h * h\( \text{Claim: } \mathrm{power}(x,n) = x^{n} \)
Base case: n equal to 0
Why: The routine returns 1, and x to the zero is 1.
\( \mathrm{power}(x,0) = 1 = x^{0} \)
Strong hypothesis: assume correctness for all exponents below n
Why: In particular the recursive call on the floor of n over 2, which is strictly below n for n at least 1, returns the right value.
\( h = \mathrm{power}(x, \lfloor n/2 \rfloor) = x^{\lfloor n/2 \rfloor} \)
Even case: n equals 2m
Why: Then the floor of n over 2 is m and the routine returns h times h.
\( h\cdot h = x^{m}\cdot x^{m} = x^{2m} = x^{n} \)
Odd case: n equals 2m plus 1
Why: Then the floor of n over 2 is m and the routine returns x times h times h.
\( x\cdot h\cdot h = x\cdot x^{2m} = x^{2m+1} = x^{n} \)
Verify correctness
Why: Both parities yield x to the n, using only exponents below n, exactly what strong induction supplies. The algorithm is correct for all n at least 0.
\( \mathrm{power}(x,n) = x^{n} \; \checkmark \)
Picture it
Animation
Shows: Each line of the worked example "correctness of square-and-multiply", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: Both parities yield x to the n, using only exponents below n, exactly what strong induction supplies. The algorithm is correct for all n at least 0.
Concept
well-founded relation — A relation with no infinite strictly descending chain. Induction is valid over any set carrying such a relation.
Ordinary induction uses the naturals with less than; structural induction uses is a sub-part of; both are well-founded.
To prove a property everywhere, prove it at each element assuming it at all strictly smaller elements. Termination arguments are exactly this.
Explain it
Discussion prompt
Explain Well-founded induction generalizes them all to a student a year behind you. No notation, no jargon they have not met — and it still has to be true.
Hint: If your explanation needs a symbol they have never seen, you are describing the notation rather than the idea.
Answer:
Ordinary induction uses the naturals with less than; structural induction uses is a sub-part of; both are well-founded.
Ranking
Put in order
Put the moves of Worked example: Euclid's algorithm terminates into the order they have to happen.
Why: These are the moves of the worked example in the order it makes them, and each one is set up by the one before it. Take the second argument, a nonnegative integer, as the measure of progress.
Worked example
Euclid replaces the pair of arguments with a new pair. Prove the recursion cannot run forever.
gcd(a, b):
if b == 0: return a
return gcd(b, a mod b)Identify a decreasing measure
Why: Take the second argument, a nonnegative integer, as the measure of progress.
\( \mu(a,b) = b \in \mathbb{N} \)
Show the measure strictly decreases
Why: The recursive call uses second argument a mod b, and the remainder is strictly below b whenever b is positive.
\( 0 \leq (a \bmod b) < b \)
Invoke well-foundedness of the naturals
Why: A strictly decreasing sequence of naturals cannot be infinite; well-ordering forbids infinite descent.
\( b_0 > b_1 > b_2 > \cdots \text{ is impossible in } \mathbb{N} \)
Verify termination
Why: The measure must reach 0, at which point the second argument is 0 and the algorithm returns. So Euclid's algorithm always halts.
\( b = 0 \Rightarrow \text{return } a \; \checkmark \)
Picture it
Animation
Shows: Each line of the worked example "Euclid's algorithm terminates", appearing one at a time.
The same working the example does, in the order a tutor would write it.
Takeaway: The measure must reach 0, at which point the second argument is 0 and the algorithm returns. So Euclid's algorithm always halts.
Concept
When a problem has two parameters, order the pairs so decreasing the first, or the second with the first fixed, counts as smaller.
\[ (a,b) \prec (a',b') \iff a < a' \;\text{ or }\; (a = a' \text{ and } b < b') \]
This lexicographic order on pairs of naturals is well-founded, so induction over it is valid.
The Ackermann function and many nested-loop terminations are proved exactly this way, as a descent in a lexicographic measure.
Analogy
Discussion prompt
Explain Multi-variable and lexicographic induction by analogy to something with no Foundations of Higher Mathematics in it at all — a queue, a recipe, a map, a bank balance, whatever fits. Then say where your analogy breaks.
Hint: An analogy that never breaks is not an analogy, it is the same idea wearing a hat. Find the seam — that is the part that is actually new.
Answer:
When a problem has two parameters, order the pairs so decreasing the first, or the second with the first fixed, counts as smaller.
Concept
Induction extends past the finite numbers to the ordinals, a well-ordered scaffold that continues after all the naturals.
It splits into three kinds of case: the bottom, successor stages one after another, and limit stages that gather everything below them.
\( 0, \quad \alpha + 1, \quad \text{and limits } \lambda = \sup_{\beta < \lambda} \beta \)
The engine is unchanged: ordinals admit no infinite descent, so proving each case from all smaller cases proves everything. A full treatment waits for set theory.
Counterexample
Discussion prompt
Induction extends past the finite numbers to the ordinals, a well-ordered scaffold that continues after all the naturals.
That is stated as though it always holds. Do one of two things: produce a case where it fails, or say precisely what rules such a case out. "It just does" is not on the menu.
Hint: Hunt at the extremes first — zero, one, negative, empty, equal. If every extreme survives, the reason they survive is the proof.
Answer:
The engine is unchanged: ordinals admit no infinite descent, so proving each case from all smaller cases proves everything. A full treatment waits for set theory.
Pattern
1. Name the predicate and the starting index
Why: Write the exact statement you claim for each n, and where the claim begins.
2. Prove the base case or cases
Why: Check the smallest index outright, and use one base case per unit the step reaches back.
3. State the induction hypothesis precisely
Why: Weak: assume the single previous case. Strong: assume all cases below n. Say which one you are using.
4. Derive the next case from the hypothesis
Why: Work forward until the next statement appears; never open by assuming the goal itself.
5. Verify and conclude for all n
Why: Sanity-check the algebra and one concrete instance, then declare the universal statement.
Pattern
Diagnose from how the next case depends on earlier ones.
| If the case at n needs... | Use |
|---|---|
| only the single previous case | weak induction |
| arbitrary smaller cases, from splits or factors | strong induction |
| the sub-parts of a built object | structural induction |
| a smallest witness or counterexample | well-ordering |
| a decreasing measure to halt | well-founded induction |
Comparison
Comparison matrix
From Pattern: which flavor to reach for: refill the Use column from what you know. The rest of the table is as it appeared.
| If the case at n needs... | Use |
|---|---|
| only the single previous case | weak induction |
| arbitrary smaller cases, from splits or factors | strong induction |
| the sub-parts of a built object | structural induction |
| a smallest witness or counterexample | well-ordering |
| a decreasing measure to halt | well-founded induction |
Elimination
Eliminate the wrong options
In a strong-induction proof of a claim for n at least 1, while proving the case n, what exactly may you assume?
3 of these 4 are wrong. Strike them one at a time, and say what rules each one out before you strike the next. The survivor is the answer.
Survives elimination: A
Why: Strong induction grants the property at all indices strictly below the current n, from the base up to n minus 1; that is precisely its extra power over weak induction.
Check
Read the setup, then choose.
Check your understanding
In a strong-induction proof of a claim for n at least 1, while proving the case n, what exactly may you assume?
Answer: A
Why: Strong induction grants the property at all indices strictly below the current n, from the base up to n minus 1; that is precisely its extra power over weak induction.
Prediction
Predict first
A proof shows: if the case at n minus 3 holds then the case at n holds, and it wants the claim for all n at least 8. How many base cases are needed, and which?
Answer it in your own words, now, with nothing to choose from. The options are on the next slide — and picking the right one off a list is an easier skill than producing it.
Correct: Three base cases: n equal to 8, 9, and 10.
Why: The step reaches back by 3, so repeated subtraction sorts every n at least 8 into three chains starting at 8, 9, and 10; each chain needs its own verified seed.
Check
The step's reach determines how many seeds you need.
Check your understanding
A proof shows: if the case at n minus 3 holds then the case at n holds, and it wants the claim for all n at least 8. How many base cases are needed, and which?
Answer: B
Why: The step reaches back by 3, so repeated subtraction sorts every n at least 8 into three chains starting at 8, 9, and 10; each chain needs its own verified seed.
Check
Decide which principle the argument actually requires.
Check your understanding
You prove that every integer n at least 2 is a product of primes by splitting a composite n as a times b with both factors between 2 and n minus 1. Which induction is required?
Answer: B
Why: The factors a and b are generally far below n minus 1, so you must assume the claim for all smaller values, not just the predecessor; that is exactly strong induction.
Commit first
Predict first
In the all-horses-same-color argument, where does the proof actually break?
Commit to an answer, then rate it — certain, fairly sure, or guessing — and write the rating down before you turn the page.
Correct: The inductive step fails going from 1 horse to 2, where the two subgroups do not overlap.
Why: The step secretly needs a horse shared between the two subgroups, which requires n at least 2; at the first transition, 1 to 2, the overlap is empty, so the chain is severed there.
The rating matters as much as the answer: confident-and-wrong is the combination that survives revision, because nothing about it feels like it needs revisiting.
Check
The all-horses proof has a true base and clean algebra, yet a false conclusion.
Check your understanding
In the all-horses-same-color argument, where does the proof actually break?
Answer: A
Why: The step secretly needs a horse shared between the two subgroups, which requires n at least 2; at the first transition, 1 to 2, the overlap is empty, so the chain is severed there.
Check
Justify the key move of a smallest-counterexample proof.
Check your understanding
To prove a property of all naturals by minimal counterexample, you assume the set F of counterexamples is nonempty and take its least element. Which fact legitimizes this?
Answer: A
Why: F is a nonempty subset of the naturals, and the well-ordering principle guarantees exactly that such a set has a smallest element, which the argument then seizes.
Prediction
Predict first
To prove a property holds for every well-formed formula by structural induction, what must you establish?
Answer it in your own words, now, with nothing to choose from. The options are on the next slide — and picking the right one off a list is an easier skill than producing it.
Correct: The property on variables, and that each connective preserves it given the sub-formulas.
Why: Structural induction checks every base constructor, the variables, and every building rule, each connective, assuming the property on the immediate sub-formulas, covering exactly how a formula can be built.
Check
Recall what the structural principle actually obliges you to prove.
Check your understanding
To prove a property holds for every well-formed formula by structural induction, what must you establish?
Answer: A
Why: Structural induction checks every base constructor, the variables, and every building rule, each connective, assuming the property on the immediate sub-formulas, covering exactly how a formula can be built.
Elimination
Eliminate the wrong options
Trying to prove the partial sums of reciprocal squares stay below 2, the direct claim sum is below 2 fails to induct. What repairs the proof?
3 of these 4 are wrong. Strike them one at a time, and say what rules each one out before you strike the next. The survivor is the answer.
the sum is below 3.Survives elimination: A
Why: A stronger claim carries a stronger hypothesis, and the extra slack term is exactly what the inductive step consumes to close; this is loaded induction.
Check
Recall why bounding the reciprocal-square sum needed a stronger claim.
Check your understanding
Trying to prove the partial sums of reciprocal squares stay below 2, the direct claim sum is below 2 fails to induct. What repairs the proof?
the sum is below 3.Answer: A
Why: A stronger claim carries a stronger hypothesis, and the extra slack term is exactly what the inductive step consumes to close; this is loaded induction.
Connect it up
Draw it
One page, no notation unless you need it: draw how these connect — Pattern: the induction recipe · Pattern: which flavor to reach for · What induction actually proves · Why checking cases is not enough · The mental model: a chain of dominoes. Put an arrow wherever one of them is what makes another possible, and label the arrow with why.
Recap
You can prove a statement for every natural number from two finite obligations: a base case and an inductive step.
You know the induction hypothesis is a single earlier instance you get to assume, not the goal, and not circular.
You can pick the right tool: weak induction for the predecessor, strong induction for arbitrary smaller cases, structural induction for built objects, and well-ordering for smallest-witness arguments.
| Situation | Tool |
|---|---|
| needs the previous case | weak induction |
| needs all smaller cases | strong induction |
| needs sub-parts of a structure | structural induction |
| needs a least counterexample | well-ordering |
The deep point: weak induction, strong induction, and well-ordering are one principle wearing three faces. The naturals admit no infinite descent, and that single fact powers every proof in this deck.
Want this taught 1-on-1? Alexander tutors Foundations of Higher Mathematics — $55/session, free consultation.