Theorem – Lower sum <= upper sum

Contents
1. Theorem
2. Proof
3. Consequences of the theorem.

Theorem

Let P and Q be any two partitions of an interval [a,b] \subset \mathbb{R}, and let f be bounded on [a,b]. Then, we have that

L(f, Q) \leq U(f,P)

Ie, the lower sum is always less than or equal to the upper sum.

Proof

Recall from this theorem that, for two partitions S and Q of [a,b], where S is a refinement of Q, we have,

L(f, S) \leq L(f, Q) \leq U(f, Q) \leq U(f, S) \quad (1)

If we define S = A \cup B, then S is a refinement of both P and Q, and so with S = A \cup B, we can re-write (1) as two different sets of inequalities,

\begin{array}{cc} L(f, S) \leq L(f, P) \leq U(f, P) \leq U(f, S) \\\\  L(f, S) \leq L(f, Q) \leq U(f, Q) \leq U(f, S)  \end{array}

Hence, we have three cases. Firstly, if U(f, Q) < U(f, P), then L(f,Q)\leq U(f, Q) < U(f, P), so obviously L(f, Q) \leq U(f,P).

Secondly, if U(f, Q) = U(f, P), then it is obvious that L(f, Q) \leq U(f, P) (this just follows from (1)).

Finally, if U(f, Q) > U(f, P), or U(f, P) < U(f, Q), then we have three inequalities

\begin{array}{c} U(f, P) < U(f, Q) \\ L(f, P) \leq U(f, Q) \\ L(f, Q) + L(f, P) \leq U(f, Q) + U(f, P) \end{array}.

Adding the first two inequalities together gives,

U(f, P) \geq 2\times U(f, Q) - L(f, P). Secondly, re-writing the third inequality gives

U(f, P) \geq L(f, P) + L(f, Q) - U(f, Q)

Now, adding these two inequalities together, we obtain,

\begin{array}{cc} 2\times U(f, P) &\geq 2\times U(f, Q) - L(f, P) + L(f, P) + L(f, Q) - U(f, Q) \\ &= U(f, Q) +L(f, Q) \end{array}.

But, as we know that L(f, Q) \leq U(f, Q), we have

2\times U(f, P) \geq 2\times L(f, Q), which of course gives L(f, Q) \leq U(f, P)\text{.} ~\square

Consequences of the theorem

As a consequence of this theorem, let us assume that for a function f bounded on [a,b], that the upper integral of f, U(f), is lower than the lower integral, L(f). Ie,

U(f) < L(f)

But then, this would imply that there exist partitions P and Q such that

U(f, P) \leq L(f, Q) \quad(2).

However, we have just proven the negation of (2), and so it follows that no such partitions exist.