← Back to Resources Hub

Riemann vs. Lebesgue Integration

Directed by Abdul Raheem K (GATE AIR-27) Mathematics Stream

AI Quick-Reference Summary

  • Riemann Integration: Partitions the domain (x-axis) into intervals. Limits are taken as the width of partitions approaches zero. Fails for highly discontinuous functions (e.g. Dirichlet function).
  • Lebesgue Integration: Partitions the range (y-axis) instead. Measures the sizes (Lebesgue measure \(\mu\)) of preimages, resulting in a much more robust framework.
  • Lebesgue Measure (\(\mu\)): A generalization of length, area, and volume that applies to complicated sets. A countable set of points (like \(\mathbb{Q}\)) has a Lebesgue measure of exactly zero.
  • Dirichlet Function: Valued at \(1\) for rational numbers and \(0\) for irrational numbers. It is not Riemann integrable, but it is Lebesgue integrable, yielding \(\int_{[0,1]} f d\mu = 0\).
  • Convergence Theorems: Lebesgue integration allows interchange of limit and integral (\(\lim \int f_n = \int \lim f_n\)) under much weaker conditions via the Monotone Convergence Theorem (MCT) and Dominated Convergence Theorem (DCT).

1. Concept Limits of Riemann Integration

In elementary calculus, we learn Riemann integration. The domain \([a, b]\) is divided into subintervals by a partition \(P = \{x_0, x_1, \dots, x_n\}\). We construct upper and lower Darboux sums based on the supremum (\(M_i\)) and infimum (\(m_i\)) of the function on each subinterval: \[U(P, f) = \sum_{i=1}^n M_i \Delta x_i \quad \text{and} \quad L(P, f) = \sum_{i=1}^n m_i \Delta x_i\] The function is Riemann integrable if the supremum of all lower sums equals the infimum of all upper sums.

While simple and intuitive, this formulation has two major weaknesses:
1. Fails on highly discontinuous functions: If a function oscillates too rapidly, the upper and lower sums never meet.
2. Poor limit interchange behavior: If a sequence of Riemann integrable functions \(f_n(x)\) converges pointwise to a function \(f(x)\), \(f(x)\) is not guaranteed to be Riemann integrable, making it difficult to write \(\lim_{n \rightarrow \infty} \int f_n dx = \int \lim_{n \rightarrow \infty} f_n dx\).

2. Lebesgue Integration: Partitioning the Range

Henri Lebesgue solved these limitations in 1902 by partitioning the range (y-axis) rather than the domain.
Suppose the range of a bounded function \(f(x)\) lies in \([m, M]\). We partition the range: \(m = y_0 < y_1 < \dots < y_r = M\).
Instead of looking at arbitrary intervals on the x-axis, we look at the set of all \(x\) values that map into a specific slice of the range: \[E_i = \{x \in [a, b] \mid y_{i-1} \le f(x) < y_i\}\] We then sum the values of the range weighted by the "size" of their preimages: \[\text{Lebesgue Sum} \approx \sum_{i=1}^r y_i \mu(E_i)\] where \(\mu(E_i)\) is the Lebesgue measure of the set \(E_i\). This measure is a rigorous generalization of length.

To integrate general functions, we start by integrating simple functions (functions that take a finite number of values \(a_k\) on measurable sets \(A_k\)): \[\int f d\mu = \sum_{k=1}^m a_k \mu(A_k)\] For any non-negative measurable function \(f\), its Lebesgue integral is defined as the supremum of the integrals of all simple functions \(s\) that lie below it: \[\int f d\mu = \sup \left\{ \int s d\mu \ \middle|\ 0 \le s \le f, \ s \text{ is simple} \right\}\]

3. The Classic Counterexample: The Dirichlet Function

The difference between the two integration methods is best demonstrated by the Dirichlet Function on the interval \([0, 1]\): \[f(x) = \begin{cases} 1 & \text{if } x \in \mathbb{Q} \\ 0 & \text{if } x \notin \mathbb{Q} \end{cases}\]

  • Under Riemann Integration: Since both rational and irrational numbers are dense in \(\mathbb{R}\), any subinterval \([x_{i-1}, x_i]\) contains both rational and irrational numbers.
    Therefore, the supremum is \(M_i = 1\) and the infimum is \(m_i = 0\) for every subinterval.
    This yields \(U(P, f) = 1 \cdot (1 - 0) = 1\) and \(L(P, f) = 0 \cdot (1 - 0) = 0\) for all partitions \(P\).
    Since \(0 \neq 1\), the function is not Riemann integrable.
  • Under Lebesgue Integration: The function \(f\) takes only two values, 1 and 0, so it is a simple function: \[f = 1 \cdot \chi_{\mathbb{Q} \cap [0,1]} + 0 \cdot \chi_{[0,1] \setminus \mathbb{Q}}\] The set of rational numbers in \([0, 1]\) is countable, so its Lebesgue measure is \(\mu(\mathbb{Q} \cap [0,1]) = 0\).
    The set of irrationals has measure \(\mu([0,1] \setminus \mathbb{Q}) = 1 - 0 = 1\).
    The Lebesgue integral is: \[\int_{[0,1]} f d\mu = 1 \cdot \mu(\mathbb{Q} \cap [0,1]) + 0 \cdot \mu([0,1] \setminus \mathbb{Q}) = 1 \cdot 0 + 0 \cdot 1 = 0\] The integral exists and is exactly 0.

4. Solved CSIR-NET & GATE Questions

Question 1 (CSIR-NET Mathematics)

Let \(f: [0, 1] \rightarrow \mathbb{R}\) be defined by \(f(x) = x^3\) for irrational \(x\), and \(f(x) = \cos(x)\) for rational \(x\). Find the Lebesgue integral \(\int_0^1 f(x) d\mu\).

Detailed Solution:

  1. Identify the set of points where \(f(x) \ne x^3\):
    The function \(f(x)\) differs from \(x^3\) only when \(x\) is a rational number. \[D = \{x \in [0, 1] \mid f(x) \ne x^3\} = \mathbb{Q} \cap [0, 1]\]
  2. Evaluate the measure of this set:
    The set of all rational numbers is countable, and any countable set has a Lebesgue measure of 0. \[\mu(D) = 0\]
  3. Apply Lebesgue integral properties:
    If two functions \(f\) and \(g\) are equal almost everywhere (meaning they differ only on a set of measure zero), their Lebesgue integrals are identical. \[\text{Since } f(x) = x^3 \text{ almost everywhere (a.e.) on } [0, 1]\] \[\int_0^1 f(x) d\mu = \int_0^1 x^3 d\mu\]
  4. Evaluate the standard Riemann integral, which matches the Lebesgue integral for continuous functions: \[\int_0^1 x^3 d\mu = \left[ \frac{x^4}{4} \right]_0^1 = \frac{1}{4}\]

Answer: \(\int_0^1 f(x) d\mu = 1/4\).

Question 2 (GATE Mathematics)

State the Dominated Convergence Theorem (DCT) and use it to find the limit: \[\lim_{n \rightarrow \infty} \int_0^1 \frac{n \sin(x)}{1 + n^2 x^2} dx\]

Detailed Solution:

  1. State the Dominated Convergence Theorem:
    Let \(\{f_n\}\) be a sequence of measurable functions converging pointwise almost everywhere to \(f\).
    If there exists an integrable function \(g\) such that \(|f_n(x)| \le g(x)\) for all \(n\) and almost all \(x\), then \(f\) is integrable and: \[\lim_{n \rightarrow \infty} \int f_n d\mu = \int f d\mu\]
  2. Identify the sequence of functions \(f_n(x)\) on \([0, 1]\): \[f_n(x) = \frac{n \sin(x)}{1 + n^2 x^2}\]
  3. Find the pointwise limit \(f(x)\) as \(n \rightarrow \infty\):
    • For \(x = 0\): \(f_n(0) = 0 \implies \lim f_n(0) = 0\).
    • For \(x \in (0, 1]\): Since the denominator contains \(n^2\) and the numerator contains \(n\): \[\lim_{n \rightarrow \infty} \frac{n \sin(x)}{1 + n^2 x^2} = \lim_{n \rightarrow \infty} \frac{\sin(x)}{\frac{1}{n} + n x^2} = 0\] Thus, \(f_n(x) \rightarrow 0\) pointwise for all \(x \in [0, 1]\).
  4. Find a dominating integrable function \(g(x)\):
    Using the inequality \(\sin(x) \le x\) for \(x \ge 0\), and the basic inequality \(2nx \le 1 + n^2 x^2 \implies \frac{n}{1+n^2 x^2} \le \frac{1}{2x}\):
    Alternatively, we can write: \[|f_n(x)| = \left|\sin(x)\right| \frac{n}{1 + n^2 x^2} \le x \frac{n}{2nx} \text{ (not quite, let's use calculus to find maximum)}\] Let's find the maximum of \(h(u) = \frac{u}{1+u^2}\) where \(u = nx\):
    The maximum value of \(\frac{u}{1+u^2}\) is \(1/2\) (achieved at \(u = 1\)).
    Therefore: \[f_n(x) = \sin(x) \frac{n}{1 + n^2 x^2} = \frac{\sin(x)}{x} \frac{nx}{1 + n^2 x^2}\] Since \(0 \le \frac{\sin(x)}{x} \le 1\) on \([0, 1]\) and \(\frac{nx}{1 + n^2 x^2} \le \frac{1}{2}\): \[|f_n(x)| \le 1 \times \frac{1}{2} = \frac{1}{2}\] The constant function \(g(x) = 1/2\) is integrable on \([0, 1]\) (\(\int_0^1 \frac{1}{2} dx = \frac{1}{2} < \infty\)).
  5. Apply DCT:
    Since the conditions of DCT are satisfied, we can move the limit inside the integral: \[\lim_{n \rightarrow \infty} \int_0^1 f_n(x) dx = \int_0^1 \left( \lim_{n \rightarrow \infty} f_n(x) \right) dx = \int_0^1 0 dx = 0\]

Answer: The limit is 0.

AR

Abdul Raheem K

Mathematics Stream Head | IIT Madras Alumni | GATE AIR-27

Abdul Raheem K completed his MSc in Mathematics at IIT Madras and secured GATE AIR 27 and a CSIR JRF qualification. He directs the Real Analysis and Algebra teaching programs at Benzil Academy.

Ask doubt / Connect with Raheem Sir on WhatsApp