A set of points \(\set{t_0, t_1, \ldots, t_n}\) such that \(a = t_0, b = t_n\text{,}\) and \(t_{i-1} \lt t_i\text{,}\) for all \(i\) is called a partition of the closed interval \([a,b]\text{.}\)
Give some partitions of \([2,4]\) and \([-1,0]\text{.}\) What would the points of a partition of \([0,1]\) be if \(t_i - t_{i-1}\) is the same for all \(i\text{?}\)
Given \([a,b]\text{,}\)\(n\in\N\text{,}\) and calling \(\frac{b-a}{n} = \Delta t\text{,}\) we may for convenience give the partition \(\set{a, a+\Delta t, a + 2\Delta t, \cdots, b - \Delta t, b}\) the name regular partition \(P_n\) (where \([a,b]\) is implicit).
Given a partition \(P = \set{t_0, t_1, \ldots, t_n}\) of \([a,b]\text{,}\) let \(y_i\) be the supremum of \(\setof{f(x)}{x\in [t_{i-1}, t_i]}\text{.}\) We say that the sum
Given a partition \(P = \set{t_0, t_1, \ldots, t_n}\) of \([a,b]\text{,}\) let \(z_i\) be the infimum of \(\setof{f(x)}{x\in [t_{i-1}, t_i]}\text{.}\) We say that the sum
Let \(f(x) = 0\) for each number \(x\in [0,1]\) except \(x=0\text{,}\) and let \(f(0) = 1\text{.}\) Show that \(U_p(f) \gt 0\) for all partitions of \([0,1]\text{,}\) but that for any positive number \(\varepsilon \gt 0\)
Aside
there is a partition \(P_{\varepsilon}\) such that \(U_{P_\varepsilon}(f) \lt \varepsilon\text{.}\) In addition, fully describe all lower sums of \(f\) on \([0,1]\text{.}\)
Let \(f(x) = x\text{,}\) and let \(P_n\) be the (regular) partition of \([0,1]\) given by \(\set{0, \frac{1}{n}, \frac{2}{n}, \ldots, \frac{n-1}{n}, 1}\text{.}\) Compute \(U_{P_5}(f)\) and \(L_{P_5}(f)\text{,}\) and then give a formula for \(U_{P_n}(f)\) and \(L_{P_n}(f)\text{.}\)
Suppose that \(f\) is bounded, with \(m\) a lower bound and \(M\) an upper bound of \(f([a,b])\text{.}\) Show that, for any partition \(P\) of \([a,b]\text{,}\)\(U_P(f) \le M\cdot (b-a)\) and \(L_P(f) \ge m\cdot (b-a)\text{.}\)
We say that a partition \(Q\) refines (or is a refinement of) a partition \(P\) if \(P\) and \(Q\) are both partitions of \([a,b]\) and \(P\subseteq Q\text{.}\)
If \(f\) is a bounded function on \([a,b]\) and \(P\) and \(Q\) are partitions of \([a,b]\) such that \(Q\) refines \(P\text{,}\) show that \(L_P(f) \le L_Q(f)\) and \(U_P(f) \ge U_Q(f)\text{.}\)
With \(f\) a bounded function on \([a,b]\text{,}\) we define the upper integral from \(a\) to \(b\) of \(f\) to be the infimum of the set of all upper sums for \(f\) on \([a,b]\text{.}\) We denote the upper integral by \(\overline{\int_a^b f}\text{.}\)
With \(f\) a bounded function on \([a,b]\text{,}\) we define the lower integral from \(a\) to \(b\) of \(f\) to be the supremum of the set of all lower sums for \(f\) on \([a,b]\text{.}\) We denote the lower integral by \(\underline{\int_a^b f}\text{.}\)
Suppose that \(f\) is not just bounded, but continuous. Suppose further that \(f(x) \ge 0\) for all \(x\in [a,b]\)and that for some \(z\in [a,b]\text{,}\)\(f(z) \gt 0\text{.}\) Then show that \(\overline{\int_a^b f} \gt 0\text{.}\)
For \(f\) as above, if \(\overline{\int_a^b f} = \underline{\int_a^b f}\text{,}\) then we say that \(f\) is integrable on \([a,b]\) and call this common value the integral of \(f\) from \(a\) to \(b\text{.}\) We denote the integral, if it exists, by
Traditionally when this value exists we say that \(f\) is Riemann integrable and the number \(\int_a^b f\) is the Riemann integral. Technically, however, it is the Darboux integral, with Riemann integrals coming from so-called Riemann sums. The two notions can be proved equivalent.
Since we saw above that the lower and upper integrals can be distinct, there exist nonintegrable bounded functions (in this sense). Luckily, there also exist integrable functions.
It is worth pondering why the information we have from previous examples is not quite enough to show that we have more integrable functions. After all, readers may be familiar with computing integrals. Historically, though, finding the values of many integrals (originally called quadrature since one wished to construct an actual square with the same area) was quite challenging. Since many applications in a variety of fields come directly from approximations involving the sums we have been computing, it would be best to have exact values for these applications, not just approximations from the sums. Keep this in mind as we progress.
Let \(f\) be a bounded function on \([a,b]\text{.}\) Show that if \(f\) is integrable then for all \(\varepsilon \gt 0\) there is a partition \(P\) of \([a,b]\) such that \(U_P(f) - L_P(f) \lt \varepsilon\text{,}\)or show the converse (both are true and useful).
This is not a lot of material to integrate. The next sets of theorems will vastly expand our repertoire of functions known to be integrable, though not yet of those with precise values at hand.
A function is nondecreasing if for each \(x \lt y\) in the domain of \(f\) we have \(f(x) \le f(y)\text{.}\) We say that it is (strictly) increasing if \(f(x) \lt f(y)\) always. The concept of nonincreasing is defined with \(\ge\) in place of \(\le\text{.}\)
You may want to write down a number of upper and lower sums to prepare for the previous problem. What previous problem will likely be easiest to use to prove integration theorems?
This example is clever, but still holds a bit of magic, since even a small change from \(\sqrt{1-x^2}\) might make it very hard. Could we integrate in general?
The preceding results mean that power functions, exponentials, and many other standard functions are integrable on certain intervals, even if we canβt finish computing them yet. The next theorems verify that most related functions are integrable on most typical intervals, and will help evaluate them too.
Let \(f\) and \(g\) be integrable on \([a,b]\text{.}\) Show that \(f+g\) is integrable on \([a,b]\text{.}\) Can you show the formula \(\int_a^b (f+g) = \int_a^b f + \int_a^b g\text{?}\)
Give two functions \(f\) and \(g\) which are integrable on \([0,1]\) such that \(fg\) is integrable on \([0,1]\) but \(\left(\int_0^1 f\right) \left(\int_0^1 g\right) \ne \int_0^1 fg\text{.}\)
We will now head straight for the punchline to the story of calculus, the βfundamental theoremsβ. Just as these were not recognized as such until quite some time after Newton and Leibniz, you may want to reevaluate what you think their importance is here.
Assume that \([a,b]\) is a closed interval and \(c\in (a,b)\text{.}\) Show that \(\overline{\int_a^c f} + \overline{\int_c^b f} = \overline{\int_a^b f}\text{.}\)
Let \(U\) be defined such that for each \(x\in [a,b]\text{,}\)\(U(x) = \overline{\int_a^x f(t) \, dt}\text{,}\) and let \(L(x) = \underline{\int_a^x f(t) \, dt}\text{.}\) Show that if \(f\) is continuous on \([a,b]\text{,}\) then \(U\) and \(L\) are differentiable at every \(c\in [a,b]\text{,}\) with \(U'(c) = f(c) = L'(c)\text{.}\)
In ProblemΒ 172, show that in fact \(U(x) = L(x)\text{,}\) and conclude that if \(f\) is continuous on \([a,b]\) then \(f\) is integrable on \([a,b]\text{.}\)
And finally we get the most beloved theorems of all. Their true importance lies in the inverse relationship between derivatives and integrals, and that one can evaluate integrals (which a priori come from the sums we did earlier) exactly using information about derivatives. You should be able to prove these without lots of subtle work, using the facts about lower and upper integrals.
Let \(f\) be a continuous function on \([a,b]\text{,}\) and let \(F\) be defined such that for each \(x\in [a,b]\text{,}\)\(F(x) = \int_a^x f\text{.}\) Then for each point \(x\in [a,b]\text{,}\)\(F\) has a derivative at \(x\) and we can compute \(F'(x) = f(x)\text{.}\)
Suppose that \(g\) is a function on \([a,b]\text{,}\) that \(g\) has a derivative at each point of \([a,b]\text{,}\) and that the function \(g'\) is itself continuous at each point in \([a,b]\text{.}\) Then show that \(\int_a^b g' = g(b) - g(a)\text{.}\)
These are the keys which made the efforts of Newton and Leibniz into the calculusβsomething that one could calculate with. It is worth noting that ProblemΒ 175 is true even if \(f'\) is just integrableβtruly, a great theorem.