Contrapositive and Contradiction

Study Sheet

Contrapositive and Contradiction

Proving by assuming the opposite

Contrapositive

Concept
"If PP then QQ" equals "if not QQ then not PP"

The contrapositive of an implication is logically equivalent to it, and often easier to prove directly — especially when the conclusion QQ is a "not" statement or when PP gives you nothing to compute with.

The contrapositive is not the converse. "If it rains, the ground is wet" has contrapositive "if the ground is dry, it is not raining" (true) and converse "if the ground is wet, it is raining" (false — sprinklers). Confusing the two is the single most common logical error, in and out of mathematics.

Example
When the hypothesis is useless

Prove: if n2n^2 is odd then nn is odd. Directly, "n2n^2 odd" gives n2=2k+1n^2 = 2k + 1 — and then what? Contrapositive: if nn is even, n=2mn = 2m, then n2=4m2=2(2m2)n^2 = 4m^2 = 2(2m^2) is even. Done. ■\blacksquare

Example
Negating an "or"

Prove: if xyxy is irrational, then xx is irrational or yy is irrational. The contrapositive negates both sides, and the negation of "xx irrational or yy irrational" is "xx rational AND yy rational". So we prove: if x=p/qx = p/q and y=r/sy = r/s are rational, then xy=prqsxy = \dfrac{pr}{qs} is rational (with qs≠0qs \ne 0). ■\blacksquare The contrapositive turned a statement about an unknown irrational into a computation with fractions.

Example
A quadratic expression

Prove: if n2−6n+5n^2 - 6n + 5 is even, then nn is odd. Contrapositive: if nn is even, then n2−6n+5n^2 - 6n + 5 is odd. With n=2kn = 2k: n2−6n+5=4k2−12k+5=2(2k2−6k+2)+1n^2 - 6n + 5 = 4k^2 - 12k + 5 = 2(2k^2 - 6k + 2) + 1, odd. ■\blacksquare

Side note
Where it goes wrong

Proving the converse by mistake. In the last example, "if nn is odd then n2−6n+5n^2 - 6n + 5 is even" is also true — but it is the converse, a different theorem. Writing it up proves nothing about the original. Before starting, write the contrapositive out in full and check that both halves are negated and swapped.

Try it
Try it: contrapositive

Prove: if 3n+23n + 2 is even, then nn is even.

Show the answer

Contrapositive: if nn is odd, n=2k+1n = 2k + 1, then 3n+2=6k+5=2(3k+2)+13n + 2 = 6k + 5 = 2(3k + 2) + 1 is odd. ■\blacksquare

Contradiction

Concept
Assume the claim is false, derive nonsense

To prove PP by contradiction: assume ¬P\neg P and deduce something impossible (1=01 = 0, a number that is both even and odd, a smaller element of a set that had a smallest element). Then ¬P\neg P cannot hold, so PP does. It is the right tool for "there is no…", "is irrational", "infinitely many".

Contradiction is more powerful than the contrapositive because you get to assume the WHOLE negation, including whatever structure "PP is false" carries — a smallest counterexample, an explicit fraction p/qp/q, a largest prime. Its cost is that a proof by contradiction, once the contradiction appears, tells you nothing about why PP is true. Prefer a direct proof when one exists; reach for contradiction when the negation gives you something concrete to work with.

Example
2\sqrt2 is irrational

Suppose 2=p/q\sqrt2 = p/q in lowest terms. Then p2=2q2p^2 = 2q^2, so p2p^2 is even, so pp is even (proved in What a Proof Is), p=2rp = 2r. Then 4r2=2q24r^2 = 2q^2, q2=2r2q^2 = 2r^2, so qq is even too. Both even contradicts lowest terms. ■\blacksquare

Example
Infinitely many primes

Suppose there are finitely many: p1,…,pkp_1, \ldots, p_k. Let N=p1⋯pk+1N = p_1\cdots p_k + 1. N>1N > 1 has a prime factor pp; pp is some pip_i; but NN leaves remainder 11 on division by pip_i, so pi∤Np_i \nmid N. Contradiction. ■\blacksquare

Tip
Minimal counterexample

To prove every nn has property PP: suppose not, and let n0n_0 be the SMALLEST failure (well-ordering). Use the fact that everything below n0n_0 works to show n0n_0 works too — contradiction.

Side note
Where it goes wrong

Assuming the negation and then proving PP directly without ever using the assumption. That is a direct proof wearing a costume; the "contradiction" is fake. If you never used ¬P\neg P, delete it.

Try it
Try it: contradiction

Prove that there is no largest even integer.

Show the answer

Suppose NN is the largest even integer. Then N+2N + 2 is even (as N=2kN = 2k gives N+2=2(k+1)N + 2 = 2(k + 1)) and larger than NN. Contradiction. ■\blacksquare

Contradiction With Numbers

Concept
Irrational means "not a ratio"

To prove a number is irrational, assume it equals p/qp/q with integers p,qp, q, clear the fraction, and derive something impossible about integers — usually a parity clash, a divisibility that cannot hold, or two different prime factorizations of one number.

Irrationality is a negative property: it says what a number is not. There is nothing to compute with until you assume the opposite, and then you have two integers and an equation between them. That is why almost every irrationality proof is a proof by contradiction, and why the contradiction almost always lives in the integers rather than in the original number.

Example
log⁡23\log_2 3 is irrational

Suppose log⁡23=p/q\log_2 3 = p/q with positive integers p,qp, q (it is positive since 3>13 > 1). Then 2p/q=32^{p/q} = 3, and raising to the power qq gives 2p=3q2^p = 3^q. The left side is even (as p≥1p \ge 1) and the right side is odd. Contradiction. ■\blacksquare

Example
A difference of squares

Prove there are no integers x,yx, y with x2−y2=10x^2 - y^2 = 10. Suppose there were. Then (x−y)(x+y)=10(x - y)(x + y) = 10. The two factors differ by 2y2y, an even number, so they have the same parity. They cannot both be odd, since their product is even; so both are even, and then their product is divisible by 44. But 4∤104 \nmid 10. Contradiction. ■\blacksquare

Example
2+3\sqrt2 + \sqrt3 is irrational

Suppose 2+3=t\sqrt2 + \sqrt3 = t with tt rational. Squaring, 5+26=t25 + 2\sqrt6 = t^2, so 6=t2−52\sqrt6 = \dfrac{t^2 - 5}{2} is rational. Now show 6\sqrt6 is not: if 6=p/q\sqrt6 = p/q in lowest terms, then p2=6q2p^2 = 6q^2 is even, so p=2ap = 2a; then 4a2=6q24a^2 = 6q^2, so 2a2=3q22a^2 = 3q^2, which makes 3q23q^2 even and hence qq even. Both pp and qq even contradicts lowest terms. ■\blacksquare

The proof has two layers, and each is a contradiction. The first layer reduces the claim to a simpler one (squaring removes one root); the second is the familiar lowest-terms argument. Finding that reduction is the real idea — squaring once turns a sum of two roots into a single root, which the standard argument can handle.

Side note
Where it goes wrong

"2\sqrt2 and 3\sqrt3 are irrational, so their sum is irrational." Sums of irrationals can be rational: 2+(2−2)=2\sqrt2 + (2 - \sqrt2) = 2. The conclusion above is true, but only the argument in the example proves it.

Try it
Try it: rational plus irrational

Let rr be rational and xx irrational. Prove that r+xr + x is irrational.

Show the answer

Suppose r+x=sr + x = s with ss rational. Then x=s−rx = s - r, a difference of two rationals, which is rational (ab−cd=ad−bcbd\tfrac ab - \tfrac cd = \tfrac{ad - bc}{bd}). That contradicts xx irrational. ■\blacksquare

Choosing a Method

Concept
Three routes to one theorem

For "if PP then QQ": a direct proof assumes PP and reaches QQ; a proof by contrapositive assumes not-QQ and reaches not-PP; a proof by contradiction assumes PP AND not-QQ and reaches any impossibility. Contradiction assumes the most, so it is the most permissive — and the least informative.

A practical rule: look at the conclusion. If QQ is a positive, computable property ("is even", "divides", "equals"), try direct first. If QQ is a negative ("is not", "is irrational", "no solution") or an "or", its negation is positive and concrete — try the contrapositive. If you need both PP and not-QQ at once, or the negation hands you an extreme object (a largest prime, a smallest counterexample), use contradiction.

Example
The same claim three ways

Claim: if 5n+45n + 4 is odd, then nn is odd.

Direct. n=(5n+4)−4n−4=(5n+4)−2(2n+2)n = (5n + 4) - 4n - 4 = (5n + 4) - 2(2n + 2), an odd number minus an even one, so odd.

Contrapositive. If n=2kn = 2k then 5n+4=10k+4=2(5k+2)5n + 4 = 10k + 4 = 2(5k + 2) is even.

Contradiction. Suppose 5n+45n + 4 is odd and n=2kn = 2k is even. Then 5n+4=2(5k+2)5n + 4 = 2(5k + 2) is even — impossible.

The third is the second with extra words; when a contradiction proof never uses PP until the last line, rewrite it as a contrapositive.

Example
When contradiction is the natural choice

Claim: there is no smallest positive rational number. Suppose rr were one. Then r/2r/2 is rational, positive, and smaller than rr. Contradiction. ■\blacksquare The negation handed us a specific object, the supposed smallest one, and we broke it by halving; neither a direct proof nor a contrapositive has anything to start from.

Side note
Where it goes wrong

Negating incorrectly. The negation of "nn is odd and prime" is "nn is even OR nn is not prime", not "nn is even and not prime". A contradiction proof built on the wrong negation assumes too much, and its "contradiction" proves nothing.

Try it
Try it: pick a method

Prove: if a+b≥20a + b \ge 20 for real numbers a,ba, b, then a≥10a \ge 10 or b≥10b \ge 10. Which method did you choose, and why?

Show the answer

Contrapositive: if a<10a < 10 and b<10b < 10, then a+b<20a + b < 20 — add the two inequalities. ■\blacksquare The conclusion was an "or", so its negation is an "and" of two usable inequalities, which is exactly what the contrapositive needs.

Why It Works

An implication P⇒QP \Rightarrow Q is false in exactly one situation: PP true, QQ false. The contrapositive not Q⇒not P\text{not } Q \Rightarrow \text{not } P is false in exactly one situation: "not QQ" true and "not PP" false — that is, QQ false and PP true. It is the same situation. Two statements that fail in exactly the same cases are true in exactly the same cases, so they are equivalent, and proving either one proves the other.

Every statement is true or false, and none is both. If assuming "not PP" leads, by valid steps, to a statement that is both true and false — "p/qp/q is in lowest terms, and p,qp, q are both even" — then "not PP" cannot be true, because valid steps from a true statement never reach a false one. So PP is true. The whole method rests on that one fact about valid reasoning, which is why every step after the assumption must be justified as carefully as in a direct proof.

Side note
The converse is different

The same row-by-row check shows why the converse is NOT equivalent: Q⇒PQ \Rightarrow P fails when QQ is true and PP false, a different row from the one where P⇒QP \Rightarrow Q fails. So one can be true while the other is false, as with "if 6∣n6 \mid n then 3∣n3 \mid n".