Question 8.T.4: (Darboux) Suppose the function f : [a, b] → R is bounded. Gi...
(Darboux)
Suppose the function f : [a, b] → \mathbb{R} is bounded. Given ε > 0, there is a δ > 0 such that, if P is any partition of [a, b] satisfying \left\|P\right\| < δ, then
U(f, P) − U(f) < ε, L(f) − L(f, P) < ε.
In the language of limits, we may (loosely) express this result as
\underset{\left\|P\right\|→0 }{\lim} U (f, P) = U (f ) , \underset{\left\|P\right\|→0 }{\lim} L (f, P) = L (f ) .
Learn more on how we answer questions.
Let ε > 0 be given. From the definition of U (f ) , there is a partition P_{0} = \left\{u_{0}, u_{1}, . . . u_{r}\right\} such that
U (f, P_{0}) < U (f) + ε/2. (8.9)
Since f is bounded, there is a number K such that |f(x)| ≤ K for all x ∈ [a, b] . Take δ_{1} = ε/8rK, and suppose P = \left\{x_{0}, x_{1}, x_{2}, …, x_{n}\right\} is any partition satisfying \left\|P\right\| < δ_{1}. Set P_{1} = P_{0} ∪ P and let us calculate U (f, P ) − U (f, P_{1}) by calculating the contribution of each subinterval \left[x_{i}, x_{i+1}\right] to this difference.
If the subinterval does not contain any member of P_{0}, then its contribution to U (f, P ) equals its contribution to U (f, P_{1}) and thereby contributes nothing to the difference. Suppose the subinterval contains an element of P_{0}. Since its contribution to U (f, P ) is at most K (x_{i+1} − x_{i}) , while its contribution to U (f, P_{1}) is at least −K(x_{i+1} − x_{i}), we see that its contribution to the difference does not exceed 2K (x_{i+1} − x_{i}). Hence
U (f, P ) − U (f, P_{1}) ≤ \sum\limits_{i∈J}{2K \left(x_{i+1} − x_{i}\right) ,}
where J = \left\{i : [x_{i}, x_{i+1}] ∩ P_{0} ≠ ∅\right\} .
Since each member of P_{0} can belong to at most two subintervals of P, J cannot have more than 2r members. Consequently
U (f, P ) − U (f, P_{1}) ≤ \sum\limits_{i∈J}{2K \left\|P\right\| ≤ 4Kr \left\|P\right\|< ε/2.}
With U (f, P_{1}) ≤ U (f, P_{0}) , we obtain
U (f, P ) − U (f, P_{0}) < ε/2
and deduce from (8.9) that
U (f, P ) − U (f) = [U (f, P ) − U (f, P_{0})] + [U (f, P_{0}) − U (f )] < ε.
In a similar fashion we can find a δ_{2} > 0 such that
\left\|P\right\| < δ_{2} ⇒ L (f ) − L (f, P ) < ε.
Now the number δ = \min \left\{δ_{1}, δ_{2}\right\} meets the requirements.