Dirichlet’s Theorem and Liouville’s Extension of Dirichlet’s Theorem

Dirichlet’s Theorem on multiple integrals and Liouville’s extension are two of the most computationally useful results in classical analysis — and two of the most under-taught. Most undergraduate analysis courses spend weeks on convergence and continuity but cover Dirichlet’s theorem in a single lecture, if at all. That’s a missed opportunity. The theorem turns triple integrals over awkward regions (simplexes, ellipsoids, cones) into clean products of Gamma functions, which means a student who masters it can solve problems in 30 seconds that take their classmates 30 minutes via direct integration. This guide is the long version I wish I’d had when I first learned this material — full statement, full proof sketch, multiple worked examples, and Liouville’s beautiful generalization.

The Gamma function is one of the most important special functions in mathematics. It extends the factorial to non-integer values. The Beta function is its close relative, and together they show up everywhere: probability theory, complex analysis, physics, and engineering.

This article covers both functions, their definitions, key properties, and the famous Dirichlet theorem that connects them to multidimensional integrals.

Why Dirichlet’s Theorem Is Worth Mastering

The single sentence that captures the theorem: integrals of the form \( \iiint x^{l-1} y^{m-1} z^{n-1} \, dV \) over the simplex \( x + y + z \le 1, \, x, y, z \ge 0 \) equal \( \dfrac{\Gamma(l)\Gamma(m)\Gamma(n)}{\Gamma(l+m+n+1)} \). That formula has direct applications in probability (Dirichlet distribution, used in Bayesian inference and topic modeling), in statistical mechanics (counting microstates of constrained energy distributions), in computer graphics (sampling points uniformly on a simplex), and in numerical analysis (constructing quadrature rules on tetrahedra). If you’re an applied mathematics, statistics, or machine-learning student in 2026, you will encounter the Dirichlet distribution in Latent Dirichlet Allocation (LDA topic models) and in Bayesian conjugate priors. The theorem on this page is the integration identity that makes all of that work.

The Gamma Function

You know that \( 5! = 5 \times 4 \times 3 \times 2 \times 1 = 120 \). But what’s \( 3.5! \)? Or \( (-0.5)! \)? The factorial function only works for non-negative integers.

The Gamma function solves this problem. It’s a smooth curve that passes through all the factorial values and fills in everything in between.

Euler’s Integral Definition

The most common definition uses what’s called Euler’s integral:

$$\Gamma(a) = \int_0^{\infty} e^{-t} t^{a-1} \, dt$$

This integral converges when \( a > 0 \). It’s both an improper integral (upper limit is infinity) and potentially singular at \( t = 0 \) when \( a < 1 \).

The connection to factorials is simple: for positive integers \( n \),

$$\Gamma(n) = (n-1)!$$

Notice the shift. \( \Gamma(5) = 4! = 24 \), not \( 5! \). This off-by-one thing trips people up constantly. Euler defined it this way, and we’re stuck with it.

Euler’s Infinite Product Definition

There’s another way to define the Gamma function that works for all complex numbers (except non-positive integers):

$$\Gamma(a) = \lim_{m \to \infty} \frac{1 \cdot 2 \cdot 3 \cdots m}{a(a+1)(a+2) \cdots (a+m)} m^a$$

This definition extends the Gamma function to negative non-integer values. It shows the function has simple poles at \( 0, -1, -2, -3, \ldots \)

Weierstrass’s Infinite Product Definition

The third definition comes from Weierstrass:

$$\frac{1}{\Gamma(a)} = a e^{a\gamma} \prod_{m=1}^{\infty} \left(1 + \frac{a}{m}\right) e^{-a/m}$$

Here, \( \gamma \) is the Euler-Mascheroni constant:

$$\gamma = \lim_{n \to \infty} \left(1 + \frac{1}{2} + \frac{1}{3} + \cdots + \frac{1}{n} – \ln n\right) \approx 0.5772157$$

This constant shows up everywhere in analysis. Nobody knows if it’s rational or irrational.

All three definitions are equivalent. They give the same values wherever they’re all defined. Use whichever one is most convenient for your problem.

Key Properties of the Gamma Function

The recursion relation is the most useful property:

$$\Gamma(a+1) = a \cdot \Gamma(a)$$

This is what makes it extend the factorial. Just like \( 5! = 5 \times 4! \), we have \( \Gamma(6) = 5 \cdot \Gamma(5) \).

Some specific values worth memorizing:

\( \Gamma(1) = 1 \)

\( \Gamma(1/2) = \sqrt{\pi} \) (this one is famous and useful)

\( \Gamma(n) = (n-1)! \) for positive integers

\( \Gamma(n) = \pm\infty \) for non-positive integers (simple poles)

The reflection formula connects values at \( a \) and \( 1-a \):

$$\Gamma(a)\Gamma(1-a) = \frac{\pi}{\sin(\pi a)}$$

And the duplication formula (Legendre’s formula):

$$\Gamma(a)\Gamma\left(a + \frac{1}{2}\right) = \frac{\sqrt{\pi}}{2^{2a-1}} \Gamma(2a)$$

The Beta Function

The Beta function takes two parameters and is defined as:

$$B(x, y) = \int_0^1 t^{x-1}(1-t)^{y-1} \, dt$$

This integral converges when both \( x > 0 \) and \( y > 0 \).

The Beta function is symmetric: \( B(x, y) = B(y, x) \).

The most important formula connects Beta to Gamma:

$$B(x, y) = \frac{\Gamma(x)\Gamma(y)}{\Gamma(x+y)}$$

This relationship is incredibly useful. It lets you convert complicated Beta integrals into Gamma function expressions, which are often easier to evaluate.

Alternative Forms of the Beta Function

The substitution \( t = \sin^2\theta \) gives the trigonometric form:

$$B(x, y) = 2\int_0^{\pi/2} \sin^{2x-1}\theta \cos^{2y-1}\theta \, d\theta$$

The substitution \( t = u/(1+u) \) gives the infinite integral form:

$$B(x, y) = \int_0^{\infty} \frac{u^{x-1}}{(1+u)^{x+y}} \, du$$

Both forms appear regularly in applications.

Dirichlet’s Theorem

Dirichlet’s theorem extends the Beta-Gamma relationship to triple integrals. It’s essential for evaluating certain volume integrals.

Statement

If \( V \) is the region where \( x \geq 0 \), \( y \geq 0 \), \( z \geq 0 \), and \( x + y + z \leq 1 \), then:

$$\iiint_V x^{l-1} y^{m-1} z^{n-1} \, dx \, dy \, dz = \frac{\Gamma(l)\Gamma(m)\Gamma(n)}{\Gamma(l+m+n+1)}$$

The region \( V \) is a tetrahedron with vertices at the origin and the three points \( (1,0,0) \), \( (0,1,0) \), and \( (0,0,1) \).

Proof

We evaluate the integral by setting up the limits carefully. The region is bounded by the coordinate planes and the plane \( x + y + z = 1 \).

$$\int_0^1 \int_0^{1-x} \int_0^{1-x-y} x^{l-1} y^{m-1} z^{n-1} \, dz \, dy \, dx$$

Integrating with respect to \( z \) first:

$$= \int_0^1 \int_0^{1-x} x^{l-1} y^{m-1} \left[\frac{z^n}{n}\right]_0^{1-x-y} dy \, dx$$

$$= \frac{1}{n} \int_0^1 \int_0^{1-x} x^{l-1} y^{m-1} (1-x-y)^n \, dy \, dx$$

Now substitute \( y = (1-x)t \), so \( dy = (1-x) \, dt \). When \( y = 0 \), \( t = 0 \). When \( y = 1-x \), \( t = 1 \).

$$= \frac{1}{n} \int_0^1 \int_0^1 x^{l-1} (1-x)^{m-1} t^{m-1} (1-x)^n (1-t)^n (1-x) \, dt \, dx$$

$$= \frac{1}{n} \int_0^1 \int_0^1 x^{l-1} (1-x)^{m+n} t^{m-1} (1-t)^n \, dt \, dx$$

The integrals separate:

$$= \frac{1}{n} \int_0^1 x^{l-1} (1-x)^{m+n} \, dx \times \int_0^1 t^{m-1} (1-t)^n \, dt$$

These are Beta functions:

$$= \frac{1}{n} \cdot B(l, m+n+1) \cdot B(m, n+1)$$

Converting to Gamma functions using \( B(x,y) = \frac{\Gamma(x)\Gamma(y)}{\Gamma(x+y)} \):

$$= \frac{1}{n} \cdot \frac{\Gamma(l)\Gamma(m+n+1)}{\Gamma(l+m+n+1)} \cdot \frac{\Gamma(m)\Gamma(n+1)}{\Gamma(m+n+1)}$$

Since \( \Gamma(n+1) = n \cdot \Gamma(n) \), we have:

$$= \frac{1}{n} \cdot \frac{\Gamma(l)\Gamma(m+n+1)}{\Gamma(l+m+n+1)} \cdot \frac{\Gamma(m) \cdot n \cdot \Gamma(n)}{\Gamma(m+n+1)}$$

The \( n \) cancels, and \( \Gamma(m+n+1) \) cancels:

$$= \frac{\Gamma(l)\Gamma(m)\Gamma(n)}{\Gamma(l+m+n+1)}$$

Which completes the proof. ∎

Liouville’s Extension of Dirichlet’s Theorem

Liouville generalized Dirichlet’s theorem to handle functions of \( x + y + z \) and different boundary conditions.

If \( x, y, z \) are all positive and \( h_1 < x + y + z < h_2 \), then:

$$\iiint_V x^{l-1} y^{m-1} z^{n-1} F(x+y+z) \, dx \, dy \, dz = \frac{\Gamma(l)\Gamma(m)\Gamma(n)}{\Gamma(l+m+n)} \int_{h_1}^{h_2} F(h) h^{l+m+n-1} \, dh$$

This is powerful. It reduces a triple integral involving \( F(x+y+z) \) to a single integral in \( h \). The geometric complexity gets absorbed into the Gamma function ratio.

Special Case

When \( F = 1 \), \( h_1 = 0 \), and \( h_2 = 1 \), Liouville’s extension reduces to Dirichlet’s original theorem (with a slight adjustment in the Gamma function argument).

Applications

These functions aren’t just mathematical curiosities. They appear constantly in real problems.

The Beta distribution in statistics uses the Beta function. It models probabilities of probabilities (like conversion rates, click-through rates, or any proportion with uncertainty).

The Gamma distribution models waiting times. It’s everywhere in queuing theory, reliability engineering, and Bayesian statistics.

In physics, the Gamma function appears in the volume of n-dimensional spheres, partition functions in statistical mechanics, and Feynman integrals in quantum field theory.

String theory uses the Beta function extensively. The Veneziano amplitude (one of the first string theory results) is literally a Beta function.

Suggested Reading

Calculus Made Easy

Silvanus P. Thompson (Author), Martin Gardner (Author)

Hardcover ‏ : ‎ 336 pages

View on Amazon.com

View on Amazon India

Frequently Asked Questions