Skip to content
Topos Uranos

Techniques for definite integrals

Summary

Many definite integrals are computed without an antiderivative: it is enough to compare them with another, equal integral that, added to or subtracted from the first, gives something simple. This lesson proves, from the lower and upper sums, that translating or reflecting an integrable function does not change its integral, and deduces from this three rules that hold for every integrable function: the parity rule on a symmetric interval, the periodicity rule (the integral over any period is the same) and the reflection x↦a+b−xx \mapsto a + b - x of an interval onto itself. For continuous functions, the change of variable of the preceding lesson adds the inversion x↦1xx \mapsto \frac{1}{x} on intervals with reciprocal endpoints. With parity and the antiderivatives of the unit on antiderivatives, the orthogonality relations of sines and cosines are proved; with the reduction formula for the powers of sine, the values of the Wallis integrals, the Wallis product and the order of magnitude of those integrals, which the unit on improper integrals will use.

Learning objectives

  1. Prove that the integral does not change when the function is translated or reflected, and deduce the rules of parity, of periodicity and of reflection about the midpoint for integrable functions.
  2. Use symmetries and the inversion x↦1xx \mapsto \frac{1}{x} to compute integrals without knowing an antiderivative of the integrand.
  3. Prove the orthogonality relations of sines and cosines on [−π,π]\icc{-\pi}{\pi}.
  4. Compute the Wallis integrals, prove the Wallis product and deduce the order of magnitude of WnW_n.

Translations and reflections

The theorem on change of variable in the definite integral, from the lesson on the fundamental theorem of calculus, requires the integrand to be continuous, because it is proved with an antiderivative. The rules that follow hold for every integrable function, and for that reason they are proved directly with the sums: when a partition is translated or reflected, its subintervals keep their lengths, and on each subinterval the translated or reflected function takes the same values as the original one.

LemmaTranslation and reflection of the integral

Let ff be a function integrable on [a,b]\icc{a}{b}, with a<ba < b, and let cc be a real number. Let g(x)=f(x+c)g(x) = f(x + c) for xx in [a−c,b−c]\icc{a - c}{b - c}, and h(x)=f(c−x)h(x) = f(c - x) for xx in [c−b,c−a]\icc{c - b}{c - a}. Then gg and hh are integrable on their intervals, and:

∫a−cb−cf(x+c) dx=∫abf(x) dx=∫c−bc−af(c−x) dx\int_{a - c}^{b - c} f(x + c) \, \mathrm{d}x = \int_a^b f(x) \, \mathrm{d}x = \int_{c - b}^{c - a} f(c - x) \, \mathrm{d}x

The lower and upper sums of gg and of hh are exactly those of ff. The figure shows a positive function on [a,b]\icc{a}{b} with the rectangles of a lower sum and its translate, with the same rectangles; a finer partition of the translated interval, which comes from another partition of [a,b]\icc{a}{b}; and a reflected one, with the same rectangles in reverse order.

Proof

  1. L(g,P′)=∑k=1nmk Δxk=L(f,P),U(g,P′)=∑k=1nMk Δxk=U(f,P)\dato{tr}{L(g, P') = \sum_{k=1}^{n} m_k \, \Delta x_k = L(f, P), \qquad U(g, P') = \sum_{k=1}^{n} M_k \, \Delta x_k = U(f, P)}

    The function ff is bounded, because the definition of integrable function and its integral, from the lesson on the Riemann integral, applies only to bounded functions, and gg and hh take the same values as ff, so they are bounded too. Let P={x0,x1,…,xn}P = \{x_0, x_1, \ldots, x_n\} be a partition of [a,b]\icc{a}{b} and P′P' the partition of [a−c,b−c]\icc{a - c}{b - c} formed by the points xk−cx_k - c, whose subintervals have the same lengths Δxk\Delta x_k. When xx runs through [xk−1−c,xk−c]\icc{x_{k - 1} - c}{x_k - c}, the point x+cx + c runs through [xk−1,xk]\icc{x_{k - 1}}{x_k}: the set of values of gg on subinterval kk of P′P' is the set of values of ff on subinterval kk of PP, and it has the same infimum mkm_k and the same supremum MkM_k, which exist by the completeness axiom, from the lesson on number systems.

  2. ∫a−cb−c‾g=∫ab‾f=∫abf=∫ab‾f=∫a−cb−c‾g→∫a−cb−cg=∫abf\underline{\int_{a - c}^{b - c}} g = \underline{\int_a^b} f = \int_a^b f = \overline{\int_a^b} f = \overline{\int_{a - c}^{b - c}} g \rightarrow \resaltar{\int_{a - c}^{b - c} g = \int_a^b f}

    Every partition QQ of [a−c,b−c]\icc{a - c}{b - c} is of that form: adding cc to its points gives a partition PP of [a,b]\icc{a}{b} with P′=QP' = Q. Therefore the set of lower sums of gg is the set of lower sums of ff, and the same holds for the upper ones; their suprema and infima, given by the completeness axiom, coincide. Since ff is integrable, its lower and upper integrals equal ∫abf\int_a^b f, and by the definition of integrable function, gg is integrable, with that integral.

  3. L(h,P′′)=∑k=1nmn+1−k Δxn+1−k=∑j=1nmj Δxj=L(f,P)→∫c−bc−ah=∫abfL(h, P'') = \sum_{k=1}^{n} m_{n + 1 - k} \, \Delta x_{n + 1 - k} = \sum_{j=1}^{n} m_j \, \Delta x_j = L(f, P) \rightarrow \resaltar{\int_{c - b}^{c - a} h = \int_a^b f}

    For the reflection, let yk=c−xn−ky_k = c - x_{n - k}, with kk from 00 to nn. Since the xkx_k increase, so do the yky_k, and they form a partition P′′P'' of [c−b,c−a]\icc{c - b}{c - a}. When xx runs through the subinterval [yk−1,yk]\icc{y_{k - 1}}{y_k}, the point c−xc - x runs through [xn−k,xn−k+1]\icc{x_{n - k}}{x_{n - k + 1}}, subinterval n−k+1n - k + 1 of PP, of the same length. Reversing the order, from the theorem on properties of summation, from the lesson on summations, progressions and induction, gives the same sums, and likewise for the upper ones; every partition of [c−b,c−a]\icc{c - b}{c - a} is of that form, because reflecting it again gives one of [a,b]\icc{a}{b}. The reasoning of the preceding move completes the argument.