Skip to content
Topos Uranos

Summary

This lesson turns the idea of the preceding lesson into a definition: the area under a curve is the only number that lies between the sums from below and the sums from above. For a bounded function on a closed interval it defines partitions, their refinement and their norm, and the lower and upper sums, built with the infimum and the supremum of the function on each subinterval. It proves that adding points does not make the lower sum go down or the upper sum go up, and that every lower sum is less than or equal to every upper sum; by the completeness axiom, the supremum of the lower sums and the infimum of the upper ones exist, and they are the lower integral and the upper integral. If they coincide, the function is integrable and that number is its integral. It proves Riemann’s integrability criterion, also with sequences of partitions, and computes from the definition the integrals of a constant, of xx and of x2x^2. It shows that every area meeting the requirements of the preceding lesson is the integral, and with it defines the area under a graph. Finally, it defines Riemann sums, with an intermediate point in each subinterval, and their limit as the norm tends to zero, and proves Darboux’s theorem: a function is integrable if and only if its Riemann sums have a limit, and that limit is the integral; in particular, the sums with equal subintervals tend to the integral.

Learning objectives

  1. Define partitions, lower and upper sums, and the lower and upper integrals of a bounded function, and prove the refinement lemma and its consequences.
  2. Define the integrable function and its integral, and prove Riemann’s integrability criterion, also in its form with sequences of partitions.
  3. Define Riemann sums and their limit as the norm tends to zero, and prove Darboux’s theorem, which makes the two definitions of the integral equivalent.
  4. Compute integrals from the definition, choosing convenient partitions and intermediate points, and read the area under a graph as an integral.

Partitions and lower and upper sums

The lesson on the area problem enclosed several areas between sums of rectangles and ended with three pending tasks: to admit arbitrary divisions of the interval, to admit functions that are not monotonic, with the infimum and the supremum of the function on each subinterval as heights, and to know when the sums from below and from above come as close as desired. This lesson accomplishes the first two and gives the criterion for the third; the lesson on integrable functions will decide which functions meet it.

DefinitionPartition of an interval

Let a<ba < b. A partition of [a,b]\icc{a}{b} is a finite set P={x0,x1,…,xn}P = \{x_0, x_1, \ldots, x_n\} with a=x0<x1<⋯<xn=ba = x_0 < x_1 < \cdots < x_n = b. The intervals [xk−1,xk]\icc{x_{k - 1}}{x_k}, with kk from 11 to nn, are its subintervals; Δxk=xk−xk−1\Delta x_k = x_k - x_{k - 1} is the length of subinterval kk, and the norm of PP is the largest of those lengths:

∥P∥=max⁡(Δx1,…,Δxn)\norma{P} = \max(\Delta x_1, \ldots, \Delta x_n)

If PP and QQ are partitions of [a,b]\icc{a}{b} and P⊆QP \subseteq Q, we say that QQ is a refinement of PP. The partition with xk=a+kb−anx_k = a + k \frac{b - a}{n} is called uniform, with nn subintervals; its norm is b−an\frac{b - a}{n}.

Two partitions PP and QQ of [a,b]\icc{a}{b} always have a common refinement, their union P∪QP \cup Q, which is a finite set containing aa and bb whose points, in increasing order, form a partition. Each subinterval of a refinement of PP is contained in a subinterval of PP, and therefore refining does not increase the norm. The lengths of the subintervals always add up to b−ab - a: by the telescoping sum theorem, from the lesson on summations, progressions and induction, ∑k=1nΔxk=xn−x0=b−a\sum_{k=1}^{n} \Delta x_k = x_n - x_0 = b - a.

DefinitionLower and upper sums

Let ff be a function bounded on [a,b]\icc{a}{b}, that is, such that there exist numbers mm and MM with m≤f(x)≤Mm \leq f(x) \leq M for every xx in [a,b]\icc{a}{b}, and let P={x0,…,xn}P = \{x_0, \ldots, x_n\} be a partition of [a,b]\icc{a}{b}. For each kk, let mkm_k and MkM_k be the infimum and the supremum of the values f(x)f(x) with xk−1≤x≤xkx_{k - 1} \leq x \leq x_k. The lower sum and the upper sum of ff associated with PP are:

L(f,P)=∑k=1nmk Δxk,U(f,P)=∑k=1nMk ΔxkL(f, P) = \sum_{k=1}^{n} m_k \, \Delta x_k, \qquad U(f, P) = \sum_{k=1}^{n} M_k \, \Delta x_k

The numbers mkm_k and MkM_k exist by the completeness axiom, from the lesson on number systems: the set of values of ff on subinterval kk is nonempty and bounded above by MM, and it has a supremum; the set of the opposites of those values is bounded above by −m-m, and the opposite of its supremum is the greatest lower bound of the values, their infimum. Since mm is a lower bound and mkm_k the greatest one, m≤mkm \leq m_k; likewise, Mk≤MM_k \leq M; and mk≤f(xk)≤Mkm_k \leq f(x_k) \leq M_k. Multiplying by Δxk>0\Delta x_k > 0 and adding, with the theorem on the properties of summation, from the lesson on summations:

m(b−a)≤L(f,P)≤U(f,P)≤M (b−a)m(b - a) \leq L(f, P) \leq U(f, P) \leq M \, (b - a)

If ff takes no negative values, L(f,P)L(f, P) is the area of rectangles contained in the region under the graph, and U(f,P)U(f, P) that of rectangles covering it. If ff is increasing on [a,b]\icc{a}{b}, the value f(xk−1)f(x_{k - 1}) is a lower bound of the values on subinterval kk and is one of them, so it is their infimum; likewise, Mk=f(xk)M_k = f(x_k), and with the uniform partition the sums are the LnL_n and UnU_n of the preceding lesson.

LemmaRefinement lemma

Let ff be a function bounded on [a,b]\icc{a}{b}, and let PP and QQ be partitions of [a,b]\icc{a}{b} with P⊆QP \subseteq Q. Then:

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

When a subinterval is divided, the infimum on each piece is not less than the infimum on the whole subinterval, and the supremum is not greater. The figure shows, for a function that is not monotonic, what the lower sum gains and what the upper sum loses when a point is added, and a refinement with more points.

Proof

  1. mk Δxk=mk(y−xk−1)+mk(xk−y)≤m′(y−xk−1)+m′′(xk−y)→L(f,P)≤L(f,Q)m_k \, \Delta x_k = m_k (y - x_{k - 1}) + m_k (x_k - y) \leq m' (y - x_{k - 1}) + m'' (x_k - y) \rightarrow \resaltar{L(f, P) \leq L(f, Q)}

    Suppose first that QQ has only one point more than PP, a point yy with xk−1<y<xkx_{k - 1} < y < x_k. The other subintervals are common. Let m′m' and m′′m'' be the infima of ff on [xk−1,y]\icc{x_{k - 1}}{y} and on [y,xk]\icc{y}{x_k}. The number mkm_k is a lower bound of the values of ff on subinterval kk, and therefore of the values on each piece; since the infimum is the greatest lower bound, mk≤m′m_k \leq m' and mk≤m′′m_k \leq m''.

  2. M′ (y−xk−1)+M′′ (xk−y)≤Mk Δxk→U(f,Q)≤U(f,P)M' \, (y - x_{k - 1}) + M'' \, (x_k - y) \leq M_k \, \Delta x_k \rightarrow \resaltar{U(f, Q) \leq U(f, P)}

    Likewise, if M′M' and M′′M'' are the suprema on the two pieces, MkM_k is an upper bound of the values on each of them, and M′≤MkM' \leq M_k, M′′≤MkM'' \leq M_k. The middle inequality is that of the sums of one and the same partition.

  3. L(f,Q0)≤L(f,Q1)≤⋯≤L(f,Qj)≤U(f,Qj)≤⋯≤U(f,Q0)L(f, Q_0) \leq L(f, Q_1) \leq \cdots \leq L(f, Q_j) \leq U(f, Q_j) \leq \cdots \leq \resaltar{U(f, Q_0)}

    If QQ has jj points more than PP, they are added one at a time: P=Q0⊆Q1⊆⋯⊆Qj=QP = Q_0 \subseteq Q_1 \subseteq \cdots \subseteq Q_j = Q, each partition with one point more than the one before it. By induction on jj, with the first two moves at each addition, the inequalities combine.

CorollaryEvery lower sum is less than or equal to every upper sum

Let ff be a function bounded on [a,b]\icc{a}{b}. For every pair of partitions PP and QQ of [a,b]\icc{a}{b}, L(f,P)≤U(f,Q)L(f, P) \leq U(f, Q). The two partitions may be unrelated to each other, but their union refines both.

Proof

  1. L(f,P)≤L(f,P∪Q)≤U(f,P∪Q)≤U(f,Q)L(f, P) \leq L(f, P \cup Q) \leq U(f, P \cup Q) \leq \resaltar{U(f, Q)}

    The union P∪QP \cup Q is a refinement of PP and of QQ; the refinement lemma, applied twice, gives the inequalities.