0

I have been taking a university course in which they define the Riemann-Stieltjes Integral using upper sums and lower sums of a function and then we take their respective infimum and supremum.

In the course they have also mentioned Reimann sums and they have proved that if the limit of reimann sums, as the mesh of the tagged partitions go to zero, exists then it is Riemann-Stieltjes Integrable. But they said that the converse holds true when the function if continuous or the function you integrate if with respect to is continuous.

I also found this pdf https://www.math.mcgill.ca/labute/courses/255w03/L1.pdf Riemann-Stieltjes using the limit of reimann sums.

But according to our professor the two definitions are not strictly equivalent. So which definition is correct?

  • 1
    The sentence starting with "But they said.." seems to have got away from you. If you wrote out in terms of the function, it might help. – Thomas Andrews Aug 02 '24 at 19:16
  • I'd say this definition is the correct one: https://en.wikipedia.org/wiki/Lebesgue%E2%80%93Stieltjes_integration. – Mason Aug 03 '24 at 02:11

1 Answers1

1

You can define the Riemann integral in different ways. The one you are referring to, with the infimum and the supremum, uses the so-called Darboux sums. I will write two equivalent definitions of the Riemann integral and the usual definition of the Riemann-Stieltjes integral. Hopefully, with the second definition of the Riemann integral in mind, it will become clear that Riemann integrable implies Riemann-Stieltjes integrable.

Definition 1 of Riemann integral: Let $f:[a,b] \to \mathbb{R}$ be a bounded function. Consider a partition $P = \{a = x_0 < x_1 < \dots < x_n = b\}$ and define the following sums: $$L(f,P) := \sum_{i=1}^n m_i (x_i - x_{i-1}) \quad \text{and} \quad U(f,P) := \sum_{i=1}^n M_i (x_i - x_{i-1}),$$ where $m_i = \inf_{t \in [x_{i-1},x_i]} f(t)$ and $M_i = \sup_{t \in [x_{i-1},x_i]} f(t)$. Then, let $\mathcal{P}$ be the set of partitions of $[a,b]$ and set $$\mathcal{L} := \sup_{P \in \mathcal{P}} L(f,P) \quad \text{and} \quad \mathcal{U} := \inf_{P \in \mathcal{P}} U(f,P).$$ If $\mathcal{L} = \mathcal{U}$, then we say that $f$ is Riemann-integrable over $[a,b]$ and denote its integral by $\int_a^b f dx$.

Definition 2 of Riemann integral: Let $(P_n)$ be a sequence of partitions $\{a=x^n_0 < x^n_1 < \dots < x^n_n = b\}$ such that $|P_n| := \max_{i} |x_i - x_{i-1}|$ goes to zero as $n \to \infty$. The number $|P_n|$ is typically referred to as the mesh of $P_n$. Now, consider the following sums: $$S(P_n) := \sum_{i=1}^n f(\xi_i) (x_i - x_{i-1}),$$ where $\xi_i \in [x_{i-1},x_i]$ is an arbitrary test point over which $f$ is defined. If the limit $\lim_{n\to \infty} S(P_n)$ exists, then we say that $f$ is Riemann-integrable over $[a,b]$ and denote its integral by $\int_a^b fdx$.

Now, one can show that these two definitions are equivalent. There are two key insights to see this. First, note that we can "sandwich" the sums of Definition 2 with the lower and upper sums of Definition 1. Precisely, $$L(f,P) := \sum_{i=1}^n m_i (x_i - x_{i-1}) \leq \sum_{i=1}^n f(\xi_i) (x_i - x_{i-1}) \leq \sum_{i=1}^n M_i (x_i - x_{i-1}) =: U(f,P).$$ Second, given a sequence of partitions with vanishing mesh size $(P_n)$, we may construct a new sequence of partitions $(Q_n)$ with vanishing mesh size such that $Q_{n+1} \subset Q_n$, i.e. $Q_{n+1}$ is a refinement of $Q_n$. You can see this post for details or Apostol's Mathematical Analysis.

Regarding the Riemann-Stieltjes integral,...

Definition of Riemann-Stieltjes integral: Consider two real-valued functions $f:[a,b] \to \mathbb{R}$ and $g:[a,b] \to \mathbb{R}$. Additionally, let $(P_n)$ be a sequence of partitions such that $|P_n|\to 0$ as $n \to \infty$. Define the sums $$S(f,g,P_n) := \sum_{i=1}^n f(\xi_i)\big(g(x_{i}) - g(x_{i-1})\big).$$ If the limit $\lim_{n\to \infty} S(f,g,P_n)$ exists, then we say that the Riemann-Stieltjes integral of $f$ against $g$ over $[a,b]$ exists and we denote it by $\int_a^b f dg$.

Looking at Definition 2 of Riemann integral it should be clear now that the Riemann integral is a particular case of the Riemann-Stieltjes integral. Specifically, the Riemann-Stieltjes integral is a Riemann integral only when $g(x) = x$, i.e. the integrator is the identity function. If a function is Riemann-integrable, then it is immediately Riemann-Stieltjes integrable. The converse doesn't even make sense because with Riemann integrals you do not consider different integrators.

Oscar
  • 1,465