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) - +

3.2 Direct proof

Many results can be established directly simply by starting from the given assumptions and proceeding through a succession of inferences to deduce the desired conclusion. Workshop exercises and exam questions which invite such a proof will often begin with words like ‘‘verify’’ or ‘‘check’’.

To establish an implication of the form ‘‘pqp\Rightarrow q’’ (which may alternatively be phrased ‘‘pp implies qq’’, or ‘‘if pp, then qq’’) by direct proof, one should assume pp to be true and then deduce that qq is true.

Example 3.2.1

Verify that x=2+2x=\sqrt{2+\sqrt{2}} satisfies x4-4x2+2=0x^{4}-4x^{2}+2=0.

Proof. Let x=2+2x=\sqrt{2+\sqrt{2}}. Then we have x2=2+2x^{2}=2+\sqrt{2}, and so

x4=(x2)2=(2+2)2=22+(2)2+222=6+42,x^{4}=(x^{2})^{2}=\bigl(2+\sqrt{2}\bigr)^{2}=2^{2}+\bigl(\sqrt{2}\bigr)^{2}+2% \cdot 2\sqrt{2}=6+4\sqrt{2},

from which we conclude that

x4-4x2+2=(6+42)-4(2+2)+2=6+42-8-42+2=0,x^{4}-4x^{2}+2=\bigl(6+4\sqrt{2}\bigr)-4\bigl(2+\sqrt{2}\bigr)+2=6+4\sqrt{2}-8% -4\sqrt{2}+2=0,

as required. \Box

Example 3.2.2

Check that y=3e2x-2e3xy=3\,e^{2x}-2\,e^{3x} is a solution to the differential equation

d2ydx2-5dydx+6y=0.\frac{\mathrm{d}^{2}y}{\mathrm{d}x^{2}}-5\,\frac{\mathrm{d}y}{\mathrm{d}x}+6y=0. (3.2.1)

Proof. Let y=3e2x-2e3xy=3\,e^{2x}-2\,e^{3x}. Then we have

dydx=6e2x-6e3x  and  d2ydx2=12e2x-18e3x,\frac{\mathrm{d}y}{\mathrm{d}x}=6\,e^{2x}-6\,e^{3x}\qquad\text{and}\qquad\frac% {\mathrm{d}^{2}y}{\mathrm{d}x^{2}}=12\,e^{2x}-18\,e^{3x},

so that

d2ydx2-5dydx+6y\displaystyle\frac{\mathrm{d}^{2}y}{\mathrm{d}x^{2}}-5\,\frac{\mathrm{d}y}{% \mathrm{d}x}+6y =(12e2x-18e3x)-5(6e2x-6e3x)+6(3e2x-2e3x)\displaystyle=(12\,e^{2x}-18\,e^{3x})-5(6\,e^{2x}-6\,e^{3x})+6(3\,e^{2x}-2\,e^% {3x})
=12e2x-18e3x-30e2x+30e3x+18e2x-12e3x=0,\displaystyle=12\,e^{2x}-18\,e^{3x}-30\,e^{2x}+30\,e^{3x}+18\,e^{2x}-12\,e^{3x% }=0,

as required. \Box

Remark 3.2.3

In Example 3.2.2 we were able to check that y=3e2x-2e3xy=3\,e^{2x}-2\,e^{3x} is a solution to the differential equation (3.2.1) without knowing why one would ever consider such an expression for yy. To understand this choice of yy, one needs to know about the theory of differential equations which will be taught in MATH101–2.

Example 3.2.4

Let mm and nn be integers. If mm is even, then so is the product mnmn.

Proof. Every even number is a multiple of 22, and so we can write m=2km=2k for some kk\in\mathbb{Z}. Consequently we have mn=(2k)n=2(kn)mn=(2k)n=2(kn), which is even because knkn is an integer. \Box

Remark 3.2.5

Denote by EE the set of even integers. We can then rewrite the statement of Example 3.2.4 as

(mE)(n)(mnE),or as  ((mE)&(n))(mnE);(\forall m\in E)(\forall n\in\mathbb{Z})(mn\in E),\qquad\text{or as}\qquad% \bigl((m\in E)\ \&\ (n\in\mathbb{Z})\bigr)\Rightarrow(mn\in E);

all three statements mean the same.

In general, statements which begin with ‘‘for all’’ can be rephrased in the form ‘‘pqp\Rightarrow q’’ for suitable statements pp and qq, because if AA is a set and P(x)P(x) is any statement involving a variable xx, then the statements ‘‘(xA)(P(x))(\forall x\in A)\bigl(P(x)\bigr)’’ and ‘‘(xA)P(x)(x\in A)\Rightarrow P(x)’’ are logically equivalent.

Example 3.2.6

For every odd integer nn, the integer n2-2nn^{2}-2n is odd.

Proof. Denote by DD the set of odd integers. We then observe that this statement can be rewritten as

(nD)(n2-2nD).(n\in D)\Rightarrow(n^{2}-2n\in D).

To prove this, let nDn\in D. Then we can write n=2m+1n=2m+1 for some mm\in\mathbb{Z}, and consequently we have

n2-2n\displaystyle n^{2}-2n =(2m+1)2-2(2m+1)\displaystyle=(2m+1)^{2}-2(2m+1)
=4m2+4m+1-4m-2=4m2-1=2(2m2)-1,\displaystyle=4m^{2}+4m+1-4m-2=4m^{2}-1=2(2m^{2})-1,

so n2-2nn^{2}-2n is odd because 2(2m2)2(2m^{2}) is even. \Box

Sometimes a proof may involve separate consideration of different cases, which between them cover all possibilities; this allows us to make extra assumptions within each of the cases. When there are only a few possibilities, such as in the context of truth tables or parity models, this is a reasonable way to proceed. When there are many cases to consider, proof by consideration of all cases can be tedious and unilluminating.

Example 3.2.7

For every integer nn, the integer n2+nn^{2}+n is even.

Proof. Let nn\in\mathbb{Z}, and observe that n2+n=n(n+1)n^{2}+n=n(n+1). If nn is even, then so is the product n(n+1)n(n+1) by Example 3.2.4. Otherwise nn is odd; then n+1n+1 is even, and once again the product n(n+1)n(n+1) is even by Example 3.2.4. Thus for every nn\in\mathbb{Z}, n2+nn^{2}+n is even. \Box

Remark 3.2.8

In the proof of Example 3.2.7, the two ‘‘strands’’ only came back together at the very end. Consequently the whole statement ‘‘pqp\Rightarrow q’’ (where pp stands for the statement ‘‘nn\in\mathbb{Z}’’, while qq represents ‘‘n2+nn^{2}+n is even’’) was stated there to clarify matters. When carrying out proof by consideration of cases, it is helpful to the reader to know when all cases have been established and the proof is complete.

The statement qq in an implication of the form ‘‘pqp\Rightarrow q’’ may involve ‘‘\exists’’. In such cases the proof usually involves constructing the required element somehow.

Example 3.2.9

For every integer nn, there exists an integer mm such that n+m=0n+m=0.

Proof. We observe that this statement can be rewritten as

(n)((m)(n+m=0)).(n\in\mathbb{Z})\Rightarrow\bigl((\exists\,m\in\mathbb{Z})(n+m=0)\bigr).

To prove this, let nn\in\mathbb{Z}, and define m=-nm=-n; then mm\in\mathbb{Z}, and n+m=n+(-n)=0n+m=n+(-n)=0 as required. \Box

So far we have considered proofs of statements of the form ‘‘pqp\Rightarrow q’’. If we wish to prove a bi-implication, that is, a statement of the form ‘‘pqp\Leftrightarrow q’’, then we must prove both the forward implication ‘‘pqp\Rightarrow q’’ and its converse ‘‘qpq\Rightarrow p’’. These two parts of the proof may involve quite different arguments.

Example 3.2.10

Let aa, bb, ss and tt be real numbers, and suppose that aba\neq b. Then aa and bb are solutions to the quadratic equation x2+sx+t=0x^{2}+sx+t=0 if and only if a+b=-sa+b=-s and ab=tab=t.

Proof. ‘‘\Rightarrow’’. Suppose that aa and bb are solutions to the equation x2+sx+t=0x^{2}+sx+t=0. Since they are distinct, the standard formula for solving a quadratic equation implies that

a=-s+s2-4t2  and  b=-s-s2-4t2a=\frac{-s+\sqrt{s^{2}-4t}}{2}\qquad\text{and}\qquad b=\frac{-s-\sqrt{s^{2}-4t% }}{2}

(or the other way round, which would not affect our proof), and therefore we have

a+b\displaystyle a+b =-s+s2-4t2+-s-s2-4t2\displaystyle=\frac{-s+\sqrt{s^{2}-4t}}{2}+\frac{-s-\sqrt{s^{2}-4t}}{2}
=-s+s2-4t-s-s2-4t2=-s\displaystyle=\frac{-s+\sqrt{s^{2}-4t}-s-\sqrt{s^{2}-4t}}{2}=-s

and

ab\displaystyle ab =(-s+s2-4t2)(-s-s2-4t2)\displaystyle=\biggl(\frac{-s+\sqrt{s^{2}-4t}}{2}\biggr)\biggl(\frac{-s-\sqrt{% s^{2}-4t}}{2}\biggr)
=(-s)2-(s2-4t)24=s2-(s2-4t)4=4t4=t,\displaystyle=\frac{(-s)^{2}-\bigl(\sqrt{s^{2}-4t}\bigr)^{2}}{4}=\frac{s^{2}-(% s^{2}-4t)}{4}=\frac{4t}{4}=t,

as required.

‘‘\Leftarrow’’. Conversely, suppose that a+b=-sa+b=-s and ab=tab=t. Then b=-s-ab=-s-a, and hence

t=ab=a(-s-a)=-as-a2.t=ab=a(-s-a)=-as-a^{2}.

Rearranging this equation, we see that a2+sa+t=0a^{2}+sa+t=0, so that aa is a solution to the equation x2+sx+t=0x^{2}+sx+t=0, as desired. In a similar way we can verify that bb is a solution to this equation. \Box