Home page for accesible maths 3 Mathematical proofs

Style control - access keys in brackets

Font (2 3) - + Letter spacing (4 5) - + Word spacing (6 7) - + Line spacing (8 9) - +

3.4 Proof by contradiction

The final proof technique which we shall cover — proof by contradiction — is the least intuitive. It is based on the conclusion of Example 2.3.8: for any statement variables pp and qq, the statements ‘‘¬(pq)\neg(p\Rightarrow q)’’ and ‘‘p&(¬q)p\ \&\ (\neg q)’’ are logically equivalent. In other words, negating both statements, we see that ‘‘pqp\Rightarrow q’’ is logically equivalent to ‘‘¬(p&(¬q))\neg\bigl(p\ \&\ (\neg q)\bigr)’’. Thus, to prove ‘‘pqp\Rightarrow q’’, we may instead show that the statement ‘‘p&(¬q)p\ \&\ (\neg q)’’ is false. We achieve this by showing that the assumption that pp and ¬q\neg q are both true leads to a contradiction (that is, a statement which is clearly false). Some examples will (hopefully) clarify this; further examples will be given later in the course.

Example 3.4.1

We saw in Example 2.4.4(iv) that the statement ‘‘there is a rational number which is greater than every natural number’’ can be written (r)(n)(r>n),(\exists\,r\in\mathbb{Q})(\forall n\in\mathbb{N})(r>n), and we claimed that this statement is false. To verify our claim, we must show that its negation is true. Now ¬((r)(n)(r>n))\neg\bigl((\exists\,r\in\mathbb{Q})(\forall n\in\mathbb{N})(r>n)\bigr) is logically equivalent to (r)(¬((n)(r>n)))(\forall r\in\mathbb{Q})\bigl(\neg\bigl((\forall n\in\mathbb{N})(r>n)\bigr)\bigr) which can be written ‘‘pqp\Rightarrow q’’, where pp represents the statement rr\in\mathbb{Q}, while qq stands for ¬((n)(r>n)).\neg\bigl((\forall n\in\mathbb{N})(r>n)\bigr).

We prove this by contradiction; that is, we assume that pp and ¬q\neg q are both true and seek a contradiction. Our assumptions state that rr\in\mathbb{Q} and (n)(r>n)(\forall n\in\mathbb{N})(r>n) after cancellation of the double negation. We see that r>1r>1, so that r=k/mr=k/m for some integers k>m1k>m\geqslant 1. Let n=k+1n=k+1\in\mathbb{N}. Then

n>kk/m=rn>k\geqslant k/m=r

which contradicts our assumption that r>n.r>n. Thus pp and ¬q\neg q cannot both be true, so if pp is true, then ¬q\neg q must be false, that is, qq is true. Hence the implication ‘‘pqp\Rightarrow q’’ is always true, as required. \Box

Example 3.4.2

The number 2\sqrt{2} is irrational.

Proof. Recall that, by definition, x=2x=\sqrt{2} is the unique positive real number such that x2=2x^{2}=2. Therefore, what we seek to prove can be rephrased as

((x(0,))&(x2=2))(x).\bigl(\bigl(x\in(0,\infty)\bigr)\ \ \&\ \ (x^{2}=2)\bigr)\ \ \Rightarrow\ \ (x% \notin\mathbb{Q}).

To give a proof by contradiction of this statement, we assume that

(x(0,))&(x2=2)&¬(x),\bigl(x\in(0,\infty)\bigr)\ \ \&\ \ (x^{2}=2)\ \ \&\ \ \neg(x\notin\mathbb{Q}),

that is, (x(0,))(x\in(0,\infty)) & (x2=2)(x^{2}=2) & (x)(x\in\mathbb{Q}). We can then write x=a/bx=a/b, where a, ba,\,b\in\mathbb{N}. After cancellation of any factors that aa and bb have in common, we may suppose that aa and bb have no common factors (except ±1\pm 1). Squaring both sides of the equation x=a/bx=a/b, we obtain 2=x2=a2/b22=x^{2}=a^{2}/b^{2}, so that a2=2b2a^{2}=2b^{2}. This shows that a2a^{2} is even (since 2b22b^{2} is clearly even), but then aa must be even (since odd times odd is odd), say a=2ca=2c for some cc\in\mathbb{N}. Thus, we have

2b2=a2=(2c)2=4c2.2b^{2}=a^{2}=(2c)^{2}=4c^{2}.

Dividing both sides of this equation by 22, we obtain b2=2c2b^{2}=2c^{2}, so that b2b^{2} is even (since 2c22c^{2} is clearly even), and thus bb is even as before. In conclusion, we see that aa and bb are both even, so 22 is a common factor of aa and bb, contradicting the assumption that they have no common factors. This contradiction proves that x.x\notin\mathbb{Q}. \Box