Fact-checked by Grok 2 weeks ago

Beta function

The Beta function, also known as Euler's integral of the first kind, is a two-parameter special function in mathematics, defined for complex numbers x and y with positive real parts by the improper integral

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

^[1]^[2] This representation converges absolutely when \operatorname{Re}(x) > 0 and \operatorname{Re}(y) > 0, providing an analytic continuation of the function to the complex plane except for poles at non-positive integers.^[1] The Beta function is intimately connected to the Gamma function through the identity

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

which holds for \operatorname{Re}(x) > 0 and \operatorname{Re}(y) > 0, enabling evaluation via Gamma function properties and facilitating its use in broader analytic contexts.^[3]^[4] First systematically studied by Leonhard Euler and Adrien-Marie Legendre in the 18th century and later named by Jacques Binet in 1839, the function generalizes binomial coefficients for non-integer arguments, as B(m+1, n+1) = \frac{m! n!}{(m+n+1)!} for positive integers m and n.^[2]^[5] Beyond its foundational role in the theory of special functions and integral calculus—where it arises in evaluations of definite integrals and series expansions—the Beta function serves as the normalizing constant for the Beta probability distribution in statistics, with probability density function f(\theta; \alpha, \beta) = \frac{\theta^{\alpha-1} (1 - \theta)^{\beta-1}}{B(\alpha, \beta)} for \theta \in (0, 1).^[6]^[2] This connection underscores its applications in Bayesian inference, where Beta priors are conjugate to binomial likelihoods, and in modeling proportions or probabilities across fields like physics, engineering, and machine learning.^[7]^[8] Extensions and generalizations, such as the incomplete Beta function, further extend its utility in asymptotic analysis and numerical computations.^[1]

Definition and Integral Representations

Standard Integral Form

The beta function, denoted B(x, y), is defined for complex numbers x and y by the integral

B(x, y) = \int_{0}^{1} t^{x-1} (1-t)^{y-1} \, dt,

provided that \operatorname{Re}(x) > 0 and \operatorname{Re}(y) > 0. This representation is known as Euler's integral of the first kind.^[9] The integral converges under these conditions because the potential singularities at the endpoints t = 0 and t = 1 are integrable. Near t = 0, the integrand behaves like t^{\operatorname{Re}(x)-1}, and the integral \int_{0}^{\epsilon} t^{\operatorname{Re}(x)-1} \, dt remains finite for small \epsilon > 0 when \operatorname{Re}(x) > 0. Similarly, near t = 1, substituting u = 1 - t yields an integrand behaving like u^{\operatorname{Re}(y)-1}, which is integrable if \operatorname{Re}(y) > 0. The integrand is continuous and bounded on the compact interval (0, 1), so no issues arise in the interior.^[6] Leonhard Euler first introduced this integral form in 1729 in correspondence with Christian Goldbach, as part of his early work on generalizing the factorial to non-integer values.^[10] For positive integers m and n, the beta function simplifies to B(m, n) = \frac{(m-1)! (n-1)!}{(m+n-1)!}, which follows from repeated integration by parts or recognizing the integral as a ratio of factorials.^[9] For example, B(2, 3) = \frac{1! \cdot 2!}{4!} = \frac{1 \cdot 2}{24} = \frac{1}{12}.^[9]

Alternative Integral Forms

One alternative representation of the beta function, valid for \operatorname{Re}(x) > 0 and \operatorname{Re}(y) > 0, extends the integration limits to the positive real axis:

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

^[11] This form arises from the standard integral representation over [0, 1] through the substitution u = t / (1 - t), which transforms the finite interval into [0, \infty) while preserving the integrand structure after simplification.^[11] Another equivalent form employs trigonometric functions and is given by

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

again for \operatorname{Re}(x) > 0 and \operatorname{Re}(y) > 0.^[6] It is derived from the standard form via the change of variables t = \sin^2 \theta, which converts the powers of t and $1 - t into powers of sine and cosine, with the differential adjusting by a factor of 2 to yield the full beta function.^[6] These representations enhance computational versatility and connections to other integrals; the infinite form proves advantageous for evaluating improper integrals of rational functions, such as those appearing in the moments of the F-distribution, while the trigonometric form simplifies assessments of definite integrals involving trigonometric powers, including extensions of Wallis' formula for \pi.^[11]^[6]

Relation to Special Functions

Connection to Gamma Function

The beta function B(x, y) is fundamentally related to the gamma function \Gamma(z) by the identity

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

for complex numbers x and y with \Re(x) > 0 and \Re(y) > 0. This relation provides an alternative expression for the beta function in terms of the gamma function, which is defined as \Gamma(z) = \int_0^\infty t^{z-1} e^{-t} \, dt for \Re(z) > 0.^[1] To derive this identity, consider the product of two gamma functions:

\Gamma(x) \Gamma(y) = \left( \int_0^\infty t^{x-1} e^{-t} \, dt \right) \left( \int_0^\infty s^{y-1} e^{-s} \, ds \right) = \int_0^\infty \int_0^\infty t^{x-1} s^{y-1} e^{-(t+s)} \, ds \, dt.

Introduce the change of variables t = u v, s = u (1 - v), with Jacobian |J| = u, and limits u from 0 to \infty, v from 0 to 1. This yields

\Gamma(x) \Gamma(y) = \int_0^1 \int_0^\infty (u v)^{x-1} [u (1 - v)]^{y-1} e^{-u} u \, du \, dv = \int_0^1 v^{x-1} (1 - v)^{y-1} \, dv \int_0^\infty u^{x+y-1} e^{-u} \, du = B(x, y) \Gamma(x + y).

Thus,

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

^[3] This relation extends to an analytic continuation of the beta function to the entire complex plane except at points where x or y is a non-positive integer, where poles arise due to those of the gamma function. The gamma function itself is meromorphic with simple poles at non-positive integers, allowing the ratio to define a meromorphic function for B(x, y) via this expression beyond the original domain of convergence of the integral representation. The connection enables evaluation of the beta function using known properties of the gamma function, such as its values at half-integers. For instance, since \Gamma\left(\frac{1}{2}\right) = \sqrt{\pi} and \Gamma(1) = 1, it follows that B\left(\frac{1}{2}, \frac{1}{2}\right) = \frac{\left[ \Gamma\left(\frac{1}{2}\right) \right]^2}{\Gamma(1)} = \pi. This specific case arises in applications like the arc length of certain curves and probability distributions. More generally, properties like the reflection formula or duplication formula for the gamma function can be applied to compute or approximate beta values efficiently.

Links to Other Functions

The beta function admits a representation in terms of the Gauss hypergeometric function {_2F_1}, specifically

B(x,y) = x^{-1} \ {}_2F_1(x, 1-y; x+1; 1),

valid for \Re(x) > 0 and \Re(y) > 0. This expression arises from the hypergeometric series expansion of the incomplete beta function evaluated at the upper limit of 1, providing a pathway to compute the beta function using hypergeometric algorithms when the gamma function representation is less convenient.^[12] For positive integers m and n, the beta function generalizes relations involving binomial coefficients through its integral form. In particular,

B(m+1, n+1) = \int_0^1 t^m (1-t)^n \, dt = \frac{m! \, n!}{(m+n+1)!} = \frac{1}{(m+n+1) \binom{m+n}{m}}.

This identity highlights how the beta function extends combinatorial quantities like binomial coefficients to continuous parameters, enabling integral representations for their reciprocals scaled by the total degree.^[9] The beta function connects to the digamma function \psi(z), the logarithmic derivative of the gamma function, via differentiation of its logarithm:

\frac{\partial}{\partial x} \ln B(x,y) = \psi(x) - \psi(x+y).

This relation facilitates the study of the beta function's behavior near its poles and zeros without explicit computation of derivatives. For specific parameter values, the beta function evaluates to quantities that appear in elliptic integrals. Notably, the lemniscate constant, defined as the elliptic-type integral \int_0^1 (1 - t^4)^{-1/2} \, dt, equals \frac{1}{4} B\left( \frac{1}{4}, \frac{1}{2} \right).^[13] This connection demonstrates how the beta function's integral form captures arc lengths and periods in lemniscate curves, a special case of elliptic geometry.

Analytic Continuation and Properties

Domain of Convergence

The standard integral representation of the beta function converges absolutely for complex parameters x and y satisfying \operatorname{Re}(x) > 0 and \operatorname{Re}(y) > 0, as the integrand t^{x-1}(1-t)^{y-1} remains integrable near the endpoints t=0 and t=1 under these conditions.^[1] Outside this half-plane, the integral diverges, but the function admits analytic continuation to the entire complex plane via its expression in terms of the gamma function, yielding a meromorphic extension.^[1] This meromorphic continuation reveals that the beta function possesses simple poles precisely when x or y is a non-positive integer, arising from the corresponding simple poles of the gamma functions in the numerator of the relation B(x,y) = \Gamma(x)\Gamma(y)/\Gamma(x+y).^[14] At these points, the denominator \Gamma(x+y) may also contribute a pole, but the overall residue remains finite and nonzero unless cancellation occurs exactly, preserving the simple pole character.^[15] For non-integer parameters, the integrand of the defining integral is multi-valued due to the fractional powers, introducing branch points at t=0 and t=1; to ensure convergence and a well-defined value within the domain \operatorname{Re}(x) > 0, \operatorname{Re}(y) > 0, principal branches are employed, with branch cuts typically placed along the negative real axis for each power.^[1] The analytic continuation circumvents this multi-valuedness, resulting in a single-valued meromorphic function across \mathbb{C}^2 excluding the polar loci.^[16]

Differentiation Under the Integral

The beta function admits differentiation under the integral sign due to its integral representation B(x,y) = \int_0^1 t^{x-1} (1-t)^{y-1} \, dt for \operatorname{Re}(x) > 0 and \operatorname{Re}(y) > 0. Differentiating with respect to the parameter x yields the partial derivative in integral form:

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

where the interchange of derivative and integral is justified by dominated convergence or uniform integrability in the domain of convergence.^[1] Similarly, the partial derivative with respect to y is

\frac{\partial}{\partial y} B(x,y) = \int_0^1 t^{x-1} (1-t)^{y-1} \log(1-t) \, dt.

Using the relation to the gamma function, B(x,y) = \frac{\Gamma(x) \Gamma(y)}{\Gamma(x+y)}, the partial derivative simplifies to a closed form involving the digamma function \psi(z) = \frac{\Gamma'(z)}{\Gamma(z)}:

\frac{\partial}{\partial x} B(x,y) = B(x,y) \left[ \psi(x) - \psi(x+y) \right],

with an analogous expression for the derivative with respect to y: B(x,y) \left[ \psi(y) - \psi(x+y) \right]. This expression holds for \operatorname{Re}(x) > 0 and \operatorname{Re}(y) > 0. Higher-order partial derivatives can be obtained by repeated application of the Leibniz rule under the integral sign. For the kth partial derivative with respect to x,

\frac{\partial^k}{\partial x^k} B(x,y) = \int_0^1 t^{x-1} (1-t)^{y-1} (\log t)^k \, dt.

Equivalently, using the gamma representation and Faà di Bruno's formula for higher logarithmic derivatives, these involve polygamma functions \psi^{(m)}(z), the mth derivatives of the digamma function. For instance, the second partial derivative is

\frac{\partial^2}{\partial x^2} B(x,y) = B(x,y) \left[ \psi'(x) - \psi'(x+y) + \left( \psi(x) - \psi(x+y) \right)^2 \right],

where \psi'(z) is the trigamma function. These derivatives find application in the analysis of the beta distribution, where the partial derivatives of B(x,y) appear in expressions for the score function and Fisher information matrix during maximum likelihood estimation of the shape parameters x and y. For example, the expected value of the score involves terms like \psi(x) - \psi(x+y), linking to the distribution's mean x/(x+y).

Series, Product, and Asymptotic Expansions

Infinite Product Representation

The infinite product representation of the Beta function is derived from the Weierstrass canonical product form of the reciprocal Gamma function. Specifically, since B(x, y) = \frac{\Gamma(x) \Gamma(y)}{\Gamma(x + y)}, substituting the Weierstrass products yields an explicit multiplicative form for B(x, y).^[17] The Weierstrass representation for the reciprocal Gamma function is given by

\frac{1}{\Gamma(z)} = z \, e^{\gamma z} \prod_{n=1}^{\infty} \left(1 + \frac{z}{n}\right) e^{-z/n},

where \gamma is the Euler-Mascheroni constant and the product converges for all complex z except non-positive integers. Combining these for \Gamma(x), \Gamma(y), and \Gamma(x + y), the exponential and Euler constant terms cancel, resulting in the infinite product

B(x, y) = \frac{x + y}{x y} \prod_{n=1}^{\infty} \frac{1 + \frac{x + y}{n}}{\left(1 + \frac{x}{n}\right) \left(1 + \frac{y}{n}\right)}.

This representation holds for \operatorname{Re}(x) > 0 and \operatorname{Re}(y) > 0, where the product converges absolutely to the Beta function value. This product form is particularly useful in analytic studies, such as deriving the reflection formula for the Gamma function, \Gamma(z) \Gamma(1 - z) = \frac{\pi}{\sin(\pi z)}, which specializes to B(z, 1 - z) = \frac{\pi}{\sin(\pi z)} for $0 < \operatorname{Re}(z) < 1. By analyzing the behavior of the product terms near poles or using limits, the sine product representation can be connected, providing insight into the meromorphic structure of the Beta function.

Series Expansions

The beta function admits a binomial series expansion obtained from its integral representation B(x,y) = \int_0^1 t^{x-1} (1-t)^{y-1} \, dt. For \operatorname{[Re](/page/Re)}(x) > 0 and \operatorname{[Re](/page/Re)}(y) > 0, expand (1-t)^{y-1} using the binomial theorem:

(1-t)^{y-1} = \sum_{k=0}^\infty \binom{y-1}{k} (-t)^k,

where \binom{y-1}{k} = \frac{(y-1)(y-2) \cdots (y-k)}{k!}. Substituting and interchanging the sum and integral (justified by absolute convergence for \operatorname{[Re](/page/Re)}(y) > 0) yields

B(x,y) = \sum_{k=0}^\infty (-1)^k \binom{y-1}{k} \int_0^1 t^{x+k-1} \, dt = \sum_{k=0}^\infty (-1)^k \binom{y-1}{k} \frac{1}{x+k}.

This series converges in the specified domain.^[18] An equivalent power series representation arises from the connection to the Gaussian hypergeometric function, as detailed in prior sections on relations to special functions. Specifically,

B(x,y) = \frac{1}{x} \, {}_2F_1(x, 1-y; x+1; 1) = \frac{1}{x} \sum_{k=0}^\infty \frac{(x)_k (1-y)_k}{(x+1)_k} \frac{1}{k!},

where ( \cdot )_k denotes the Pochhammer symbol (rising factorial). The series at argument 1 converges for \operatorname{Re}(y) > 0. This form follows from the integral representation of the hypergeometric function applied to the beta integral.^[12] For expansions around specific points, consider small perturbations from a base point (x_0, y_0) with \operatorname{Re}(x_0) > 0, \operatorname{Re}(y_0) > 0. Using the relation B(x,y) = \Gamma(x) \Gamma(y) / \Gamma(x+y), the logarithmic derivative with respect to x gives

\frac{\partial}{\partial x} \ln B(x,y) = \psi(x) - \psi(x+y),

where \psi(z) is the digamma function. Thus, the first-order Taylor expansion around x = x_0 is

B(x_0 + h, y_0) = B(x_0, y_0) \exp\left( h [\psi(x_0) - \psi(x_0 + y_0)] + O(h^2) \right) \approx B(x_0, y_0) \left[ 1 + h (\psi(x_0) - \psi(x_0 + y_0)) \right],

with higher-order terms involving polygamma functions \psi^{(n)}(z). A similar expansion holds for perturbations in y. These approximations are useful for numerical evaluation near known points. Expansions around points with large x or y can leverage the above series, though their practical use often requires truncation or acceleration techniques for convergence.^[1]

Asymptotic Approximations

The beta function B(x, y) = \frac{\Gamma(x) \Gamma(y)}{\Gamma(x+y)} admits asymptotic approximations in various limiting regimes, particularly when the parameters x and y are large or when one is small while the other is fixed. These approximations are essential for numerical evaluation and theoretical analysis in regimes where exact computation is challenging.^[19] For large positive real values of x and y, Stirling's approximation provides a leading-order asymptotic expansion for the logarithms of the gamma functions involved. Specifically,

\log B(x, y) \approx (x + y - \tfrac{1}{2}) \log(x + y) - x \log x - y \log y + \tfrac{1}{2} \log(2\pi),

with higher-order terms available from the full Stirling series. This follows directly from the asymptotic form \log \Gamma(z) \approx (z - \tfrac{1}{2}) \log z - z + \tfrac{1}{2} \log(2\pi) for large |z| in |\arg z| < \pi - \delta, applied to each gamma factor and combined via the logarithmic relation for B(x, y). The approximation is uniform for x, y \to \infty with x/y bounded away from 0 and \infty.^[19] An equivalent leading approximation can be derived from the integral representation B(x, y) = \int_0^1 t^{x-1} (1-t)^{y-1} \, dt using Laplace's method, which concentrates the integral around the interior maximum of the exponent at t = x/(x+y). Substituting t = x/(x+y) + u / \sqrt{x+y} yields a Gaussian integral, resulting in

B(x, y) \sim \sqrt{\frac{2\pi x y}{ (x+y)^3 }} \left( \frac{x}{x+y} \right)^x \left( \frac{y}{x+y} \right)^y

as x, y \to \infty, matching the exponential form from Stirling up to subleading corrections. This method highlights the saddle-point contribution and extends to higher orders via further expansion of the phase function.^[20] When one parameter is small, say x \to 0^+ with y > 0 fixed, the gamma function representation gives B(x, y) \sim 1/x, since \Gamma(x) \sim 1/x and \Gamma(x+y) \sim \Gamma(y). More refined uniform approximations near this boundary regime, incorporating the next-order terms from the Laurent expansion of \Gamma(x), ensure accuracy across transitional values; for instance, a uniform expansion valid for small x and moderate y can be obtained by rescaling the integral near t = 0. These boundary-adjusted approximations, often derived via modified Laplace techniques or Watson's lemma, are particularly useful when the integrand's peak shifts toward the endpoint.

Transformations and Identities

Functional Equations

The beta function satisfies a reflection-like identity derived from the corresponding formula for the gamma function. Specifically, for $0 < \Re x < 1,

B(x, 1 - x) = \frac{\pi}{\sin (\pi x)}.

This follows directly from the relation B(x, y) = \Gamma(x) \Gamma(y) / \Gamma(x + y) and Euler's reflection formula \Gamma(x) \Gamma(1 - x) = \pi / \sin (\pi x). A duplication-type relation for the beta function arises as a special case of the gamma function's duplication formula, which is itself a particular instance of Gauss's multiplication theorem for m = 2. For \Re x > 0,

B\left(x, x + \frac{1}{2}\right) = 2^{1 - 2x} \sqrt{\pi}.

This identity connects the beta function evaluated at shifted arguments to a simple explicit form involving the square root of \pi. More generally, the beta function obeys a multiplication theorem, expressing B(mx, my) in terms of a product of beta functions with arguments shifted by multiples of $1/m. For positive integer m and \Re x > 0, \Re y > 0,

B(mx, my) = (2\pi)^{(m-1)/2} m^{-1/2} \prod_{k=0}^{m-1} B\left(x + \frac{k}{m}, y + \frac{k}{m}\right).

This generalization stems from applying Gauss's multiplication formula to each gamma function in the defining relation for the beta function. The reciprocal of the beta function is defined as

\frac{1}{B(x,y)} = \frac{\Gamma(x+y)}{\Gamma(x)\Gamma(y)},

where B(x,y) = \int_0^1 t^{x-1} (1-t)^{y-1} \, dt for \Re(x) > 0, \Re(y) > 0. This form arises directly from the relation between the beta and gamma functions and serves as the normalizing constant in various contexts, including probability theory. A key property involves its connection to the Gauss hypergeometric function via the summation theorem. Specifically,

\frac{1}{B(x,y)} = (x + y - 1) \ {}_2F_1(1 - x, 1 - y; 1; 1),

valid for \Re(x + y - 1) > 0, where the hypergeometric function at argument 1 is evaluated using the analytic continuation provided by the theorem. This representation highlights the reciprocal's ties to confluent hypergeometric series and facilitates derivations in special function theory. For positive integers m, n \geq 1, the reciprocal takes a combinatorial form:

\frac{1}{B(m,n)} = \binom{m+n-1}{m-1} = \frac{(m+n-1)!}{(m-1)!(n-1)!}.

This binomial coefficient interpretation underscores applications in combinatorics, such as counting lattice paths or polynomial coefficients. In inequalities, the reciprocal beta function appears in bounds derived from convexity properties of the gamma function. For instance, applying the AM-GM inequality to the parameters yields x + y \geq 2\sqrt{xy}, which, combined with log-convexity of \Gamma, implies lower bounds on $1/B(x,y). Such bounds are useful in optimization and approximation theory.

Incomplete Beta Function

Definition and Basic Properties

The incomplete beta function, often denoted as B(z; x, y), is defined for real parameters x > 0, y > 0, and $0 \leq z \leq 1 by the integral representation

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

This function represents the cumulative integral up to z of the kernel associated with the complete beta function B(x, y), which is recovered as the special case B(1; x, y) = B(x, y). A closely related quantity is the regularized incomplete beta function, defined as

I(z; x, y) = \frac{B(z; x, y)}{B(x, y)},

which normalizes the incomplete form by the complete beta function and satisfies $0 \leq I(z; x, y) \leq 1 for $0 \leq z \leq 1.^[21] Key properties include the boundary evaluation B(0; x, y) = 0 and the aforementioned B(1; x, y) = B(x, y), ensuring the function bridges the incomplete and complete cases seamlessly. Additionally, it exhibits symmetry via the substitution u = 1 - t, yielding B(z; x, y) = B(1 - z; y, x). For $0 < z < 1 and x, y > 0, the incomplete beta function is strictly increasing in z, with $0 < B(z; x, y) < B(x, y), reflecting its monotonic accumulation of the integrand over the interval. Basic bounds can be derived from the integrand's positivity, though tighter estimates depend on specific parameter regimes.

Integral Representations and Expansions

The regularized incomplete beta function I(z; a, b) admits a representation in terms of the Gauss hypergeometric function, which facilitates its series expansion and analytic continuation. Specifically,

I(z; a, b) = \frac{z^a}{a \, B(a, b)} \ {}_2F_1(a, 1 - b; a + 1; z),

valid for \operatorname{Re}(a) > 0 and |z| < 1. This form arises from integrating the binomial expansion of (1 - t)^{b-1} term by term in the defining integral and recognizing the resulting power series as the hypergeometric function. The series expansion of {}_2F_1 is then

{}_2F_1(a, 1 - b; a + 1; z) = \sum_{k=0}^\infty \frac{(a)_k (1 - b)_k}{(a + 1)_k} \frac{z^k}{k!},

providing a computational tool for small |z| or when convergence is favorable.^[12] Another useful representation is the continued fraction expansion, which converges rapidly for numerical evaluation, particularly when the hypergeometric series is inefficient. One explicit form expresses I(z; a, b) as

I(z; a, b) = \frac{z^a (1 - z)^b}{a B(a, b)} \cfrac{1}{1 + \cfrac{d_1}{1 + \cfrac{d_2}{1 + \cfrac{d_3}{1 + \ddots}}}},

where the partial numerators d_k are given by d_{2m} = \frac{m(b - m) z}{(a + 2m - 1)(a + 2m)} and d_{2m+1} = -\frac{(a + m)(a + b + m) z}{(a + 2m)(a + 2m + 1)} for m = 0, 1, 2, \dots. This expansion, derived from integral representations and Euler's continued fraction transformation applied to hypergeometric functions, alternates in accuracy: the convergents of order $4m and $4m+1 underestimate I(z; a, b), while those of order $4m+2 and $4m+3 overestimate it. It is particularly effective for z near 0 or 1 and moderate parameter values.^[12] For large parameters, asymptotic expansions provide approximations that capture the behavior of I(z; a, b) when a or b (or both) tend to infinity. When both a and b are large with ratio \lambda = b/a fixed, a uniform asymptotic expansion uses the error function centered at the transition point z \approx \lambda / (1 + \lambda), yielding

I(z; a, b) \sim \frac{1}{2} \operatorname{erfc}\left( \sqrt{a} \, \zeta \right) + O\left( a^{-1/2} \right),

with \zeta determined by the scaled variable near the maximum of the integrand. These expansions are derived from saddle-point analysis of the integral and are essential for applications in statistics where parameters grow with sample size. For fixed b and large a with z < 1 fixed, I(z; a, b) \to 0 exponentially.^[22]

Multivariate Generalizations

Dirichlet Beta Function

The Dirichlet beta function, denoted β(s), is a special function defined for complex numbers s with real part greater than 0 by the Dirichlet series

\beta(s) = \sum_{n=0}^{\infty} \frac{(-1)^n}{(2n+1)^s}.

This series converges absolutely for Re(s) > 0 and provides the principal definition of the function. The Dirichlet beta function admits analytic continuation to the entire complex plane.^[23] At positive integer values, the Dirichlet beta function takes on notable closed forms. For example, β(2) equals Catalan's constant G, defined as G = \sum_{n=0}^\infty (-1)^n / (2n+1)^2 \approx 0.915966, which arises in various combinatorial contexts. Similarly, β(3) = \pi^3 / 32 \approx 0.968946. These evaluations highlight the function's ties to transcendental constants and are obtained through Fourier series expansions or residue calculus.^[24]^[25]

Properties and Relations

The multivariate beta function, denoted B(x_1, \dots, x_k) for positive real parameters x_1, \dots, x_k > 0, is defined as the integral

B(x_1, \dots, x_k) = \int_{\Delta_{k-1}} \prod_{i=1}^k t_i^{x_i - 1} \, d\mathbf{t},

where \Delta_{k-1} = \{ (t_1, \dots, t_k) \in (0,1)^k \mid \sum_{i=1}^k t_i = 1 \} is the standard (k-1)-simplex and d\mathbf{t} denotes the Lebesgue measure on this domain.^[26] This integral representation generalizes the univariate beta function to higher dimensions and arises naturally in multivariate analysis.^[26] An explicit closed-form expression relates the multivariate beta function to the gamma function:

B(x_1, \dots, x_k) = \frac{\prod_{i=1}^k \Gamma(x_i)}{\Gamma\left( \sum_{i=1}^k x_i \right)}.

This formula serves as the normalizing constant for the Dirichlet distribution, a multivariate probability distribution on the simplex with density f(\mathbf{x}) = \frac{1}{B(x_1, \dots, x_k)} \prod_{i=1}^k x_i^{x_i - 1} for \mathbf{x} \in \Delta_{k-1}.^[26] The relation to the gamma function follows from the properties of independent gamma-distributed random variables whose normalized ratios yield the Dirichlet variables.^[26] The multivariate beta function exhibits symmetry in its parameters, satisfying B(x_1, \dots, x_k) = B(x_{\sigma(1)}, \dots, x_{\sigma(k)}) for any permutation \sigma of the indices, which is evident from the product form involving the gamma functions.^[26] Regarding marginal integrals, integrating the kernel \prod t_i^{x_i - 1} over a sub-simplex corresponding to a subset of variables yields a lower-dimensional multivariate beta function as the result; specifically, the marginal for a single variable reduces to the univariate beta function B(x_j, \sum_{i \neq j} x_i).^[26] The univariate beta function corresponds to the case k=2.^[26]

Applications

Probability Distributions

The beta function plays a central role in the definition of the beta distribution, a continuous probability distribution supported on the interval [0, 1] with shape parameters \alpha > 0 and \beta > 0. This distribution is widely used to model random proportions or probabilities, such as the success probability in repeated trials or proportions in Bayesian inference. The normalizing constant in its probability density function (PDF) is precisely the beta function B(\alpha, \beta), ensuring the total probability integrates to 1 over the support. The PDF of a beta-distributed random variable T \sim \operatorname{Beta}(\alpha, \beta) is given by

f(t; \alpha, \beta) = \frac{t^{\alpha-1} (1-t)^{\beta-1}}{B(\alpha, \beta)}, \quad 0 < t < 1.

^[27] The shape of this density depends on \alpha and \beta: for \alpha = \beta = 1, it reduces to the uniform distribution on [0, 1]; when \alpha > 1 and \beta > 1, it is unimodal and bell-shaped; and for \alpha < 1 or \beta < 1, it exhibits U-shaped or J-shaped forms near the boundaries.^[28] The raw moments of the beta distribution are directly expressed in terms of ratios of beta functions. Specifically, the kth raw moment is

\mathbb{E}[T^k] = \frac{B(\alpha + k, \beta)}{B(\alpha, \beta)},

for positive integer k, which simplifies to the product \prod_{i=0}^{k-1} \frac{\alpha + i}{\alpha + \beta + i}.^[29] This formula arises from integrating t^k against the PDF, leveraging the integral representation of the beta function. For k=1, it yields the mean \mathbb{E}[T] = \alpha / (\alpha + \beta); the variance follows from the second moment as \operatorname{Var}(T) = \alpha \beta / [(\alpha + \beta)^2 (\alpha + \beta + 1)].^[27] The cumulative distribution function (CDF) of the beta distribution is the regularized incomplete beta function:

F(z; \alpha, \beta) = I(z; \alpha, \beta) = \frac{B(z; \alpha, \beta)}{B(\alpha, \beta)},

where B(z; \alpha, \beta) = \int_0^z t^{\alpha-1} (1-t)^{\beta-1} \, dt is the incomplete beta function, for $0 \leq z \leq 1.^[30] This connection allows efficient numerical computation of probabilities using properties of the incomplete beta function. The beta distribution is intimately related to other common distributions through probabilistic transformations and Bayesian conjugacy. In Bayesian statistics, the beta distribution serves as the conjugate prior for the success probability p in a binomial likelihood: if prior beliefs about p follow \operatorname{Beta}(\alpha, \beta) and data consist of s successes in n trials, the posterior is \operatorname{Beta}(\alpha + s, \beta + n - s).^[31] Additionally, the beta distribution links to the F-distribution via the transformation Y = \frac{\beta}{\alpha} \cdot \frac{T}{1 - T}, where Y \sim F(2\alpha, 2\beta), providing a bridge between proportion modeling and variance ratio tests.^[32]

Combinatorial and Number Theory Uses

The beta function provides an integral representation that connects directly to binomial coefficients in combinatorics. For positive integers m and n, the beta function is given by

B(m, n) = \int_0^1 t^{m-1} (1-t)^{n-1} \, dt = \frac{(m-1)! (n-1)!}{(m+n-1)!}.

This expression relates to the binomial coefficient via

B(m, n) = \frac{1}{(m+n-1) \binom{m+n-2}{m-1}},

allowing the reciprocal of binomial coefficients to be expressed as a beta integral scaled by a linear factor. This representation is fundamental for deriving combinatorial identities, such as those arising from the binomial theorem and its generalizations, and facilitates probabilistic interpretations in discrete structures like random walks and lattice paths.^[33] In the context of hypergeometric sums, the beta function serves as a key component in the integral representation of the Gauss hypergeometric function, which encapsulates many combinatorial series. Specifically,

{}_2F_1(a, b; c; z) = \frac{\Gamma(c)}{\Gamma(b) \Gamma(c-b)} \int_0^1 t^{b-1} (1-t)^{c-b-1} (1 - z t)^{-a} \, dt, \quad \Re(c) > \Re(b) > 0,

where the normalizing factor \frac{\Gamma(c)}{\Gamma(b) \Gamma(c-b)} = \frac{1}{B(b, c-b)} is the reciprocal of the beta function. This form is used to evaluate sums involving binomial coefficients, such as those in the binomial series expansion or Saalschützian identities, providing closed forms for hypergeometric evaluations that appear in counting problems like committee formations or polytope volumes. The beta integral thus bridges continuous analysis to discrete combinatorial enumerations.^[34]

References

[1]
DLMF: §5.12 Beta Function ‣ Properties ‣ Chapter 5 Gamma ...
In this section all fractional powers have their principal values, except where noted otherwise. In (5.12.1)–(5.12.4) it is assumed R a > 0 and R b > 0.Missing: mathematics | Show results with:mathematics
[2]
[PDF] Beta Function and its Applications - University of Tennessee, Knoxville
Just as the gamma function for integers describes fac- torials, the beta function can define a binomial coeffi - cient after adjusting indices.
[3]
[PDF] Relationship Between the Gamma and Beta Functions
Claim: The gamma and beta functions are related as b(a, b) = Γ(a)Γ(b). Γ(a + b) . Proof of Claim: Γ(a)Γ(b) = (/ ∞. 0 xa− ...
[4]
[PDF] The Gamma function 1. Basics involving Γ(s), B(α, β), etc.
Sep 15, 2019 · Basics involving Γ(s), B(α, β), etc. We take the Gamma function Γ(s) to be defined by Euler's integral. Γ(s) = ∫ ∞.<|control11|><|separator|>
[5]
[PDF] Euler's beta integral in Pietro Mengoli's works - CORE
Mar 27, 2009 · The name Beta function was assigned later in 1839 by Jacques. Binet, see Cajori (1928–1929). For biographical information about Euler one may ...
[6]
[PDF] Gamma and Beta Integrals 1 Gamma Function - MIT
May 30, 2020 · This article presents an overview of the gamma and beta functions and their relation to a variety of integrals.
[7]
[PDF] The Beta and Dirichlet Distributions 7.4.1 The Beta Random Variable
is just a normalizing constant (ensures the pdf to integrates to. 1). It is called the Beta function, and so our random variable is called a Beta random ...
[8]
[PDF] Inference Probability - CMU School of Computer Science
Here the normalizing constant B(α, β) is the beta function: B(α, β) = Z 1. 0 θα−1(1 − θ)β−1 dθ. The support of the beta distribution is θ ∈ (0, 1), and ...<|control11|><|separator|>
[9]
Beta Function -- from Wolfram MathWorld
The beta function B(p,q) is the name used by Legendre and Whittaker and Watson (1990) for the beta integral (also called the Eulerian integral of the first ...Missing: standard 1729
[10]
Leonhard Euler (1707 - 1783) - Biography - MacTutor
He also studied beta and gamma functions, which he had introduced first in 1729. Legendre called these 'Eulerian integrals of the first and second kind' ...
[11]
Beta function - StatLect
The Beta function is a function of two variables that is often found in probability theory and mathematical statistics.
[12]
[PDF] Math 213a (Fall 2024) Yum-Tong Siu 1 GAMMA FUNCTION Gamma ...
B (x, y) = Γ(x) Γ(y). Γ(x + y) . We are going to derive the formula for the Beta function as a definite integral whose integrand depends on the variables x and ...
[13]
DLMF: §8.17 Incomplete Beta Functions ‣ Related Functions ...
In the case of §8.17 it is straightforward to continue most results analytically to other real values of a, b, and x, and also to complex values.
[14]
Lemniscate Constant -- from Wolfram MathWorld
The lemniscate constant is the quantity (OEIS A062539; Abramowitz and Stegun 1972; Finch 2003, p. 420), where G is Gauss's constant.
[15]
Euler beta function—Wolfram Documentation
Its singularities and discontinuities are restricted to the non-positive integers: ... The beta function series expansion at poles: Copy to clipboard. In[1]:=1.
[16]
[PDF] Higher order monotonicity in the context of beta and gamma functions
May 18, 2018 · simple poles on the non-positive integers. The function 1/Γ is ... which gives a meromorphic continuation of the beta function in C2.
[17]
[PDF] The beta function B(x, y) - Rutgers Physics
The Beta function is defined, for Re α > 0, Re β > 0, as. B(α, β) = ∫ 1. 0 dx ... There is lots more about Gamma functions in Whittaker and Watson,.<|control11|><|separator|>
[18]
5.8 Infinite Products ‣ Properties ‣ Chapter 5 Gamma Function
x ≠ 0 , − 1 , … . ⓘ. Symbols: Γ ⁡ ( z ) : gamma function, i : imaginary unit, k : nonnegative integer ...
[19]
[PDF] arXiv:1702.01129v1 [math.NA] 5 Feb 2017
Feb 5, 2017 · This allows to derive accurate polynomial expansions of ln(1+x) and of the incomplete beta function whit a broader range of applicability than ...
[20]
5.11 Asymptotic Expansions ‣ Properties ‣ Chapter 5 Gamma ...
3), or sometimes just its leading term, is known as Stirling's formula ... 5.10 Continued Fractions5.12 Beta Function · © 2010–2025 NIST / Disclaimer ...
[21]
[PDF] Asymptotic Methods for Integrals - CWI
Jul 17, 2014 · Γ(z) by Laplace's method . . . . . . . . . . . . . . . . . . . . . . . 66 ... 42.3 Asymptotic inversion of the incomplete beta function .
[22]
[PDF] Some inequalities for the beta function and its ratios - arXiv
Aug 4, 2025 · Gamma function, Beta function, Digamma function. 1. arXiv:2508.02327 ... 1)B(x, y). (16). Page 4. 4. J.M. TANOH DJE, E. A. K. SCHWINGER, AND ...
[23]
Incomplete Beta Function -- from Wolfram MathWorld
The incomplete beta function is implemented in the Wolfram Language as Beta[z, a, b]. It is given in terms of hypergeometric functions.
[24]
[PDF] The incomplete beta function
Integrals, Incomplete beta function, digamma function. ... The last integral was evaluated using (3.2). Differentiate this identity with respect to the parameter ...
[25]
8.18 Asymptotic Expansions of I x ⁡ ( a , b )
If b and x are fixed, with b > 0 and 0 < x < 1 , then as a → ∞ for each n = 0 , 1 , 2 , ... . If b = 1 , 2 , 3
[26]
[1804.09922] Arithmetic of Catalan's constant and its relatives - arXiv
Apr 26, 2018 · Here \beta(s)=\sum_{k=0}^\infty(-1)^k(2k+1)^{-s} denotes Dirichlet's beta function, so that \beta(2) is Catalan's constant. Comments: 9 ...
[27]
[2309.13134] An Alternative Approach to Computing $β(2k+1)$ - arXiv
Sep 22, 2023 · Abstract:This paper presents a new approach to evaluating the special values of the Dirichlet beta ... integral representation of \beta(2k).
[28]
sequences and series - Proving that $\frac{\pi^{3}}{32}=1-\sum_{k=1 ...
Dec 19, 2013 · 4 Answers. and is known as the Dirichlet beta function. As shown below, β(3)=π332.How to prove this identify for Catalan zeta function?Analytic continuation of Dirichlet beta funcion - Math Stack ExchangeMore results from math.stackexchange.com
[29]
Dirichlet distribution | Mean, covariance, proofs, derivations - StatLect
The Dirichlet distribution is a multivariate continuous probability distribution often used to model the uncertainty about a vector of unknown probabilities.Definition · Caveat · Relation to the Beta distribution · Marginal distributions<|separator|>
[30]
1.3.6.6.17. Beta Distribution - Information Technology Laboratory
The beta distribution's probability density function is defined by shape parameters p and q, and lower/upper bounds a and b. The standard form has a=0 and b=1.
[31]
23.2 - Beta Distribution | STAT 414 - STAT ONLINE
Let X 1 and X 2 have independent gamma distributions with parameters α , θ and β respectively. Therefore, the joint pdf of X 1 and X 2 is given by.
[32]
Probability Playground: The Beta Distribution
The function B(a, b) in the pdf and cdf denotes the beta function, while the function B(x; a, b) in the cdf denotes the incomplete beta function. Parameter ...
[33]
[PDF] STAT:5100 (22S:193) Statistical Inference I - University of Iowa
Oct 12, 2015 · Beta Distribution. • The normalizing constant of the Beta distribution is the Beta function. B(α, β) = Z 1. 0 tα−1(1 − t)β−1dt = Γ(α)Γ(β). Γ ...
[34]
[PDF] Conjugate priors: Beta and normal Class 15, 18.05
With a conjugate prior the posterior is of the same type, e.g. for binomial likelihood the beta prior becomes a beta posterior. Conjugate priors are useful ...
[35]
[PDF] Maximum Likelihood Estimation of the Parameters of the Beta ...
The family of beta distributions is related to many of the common statistical distributions, including the t, F, binomial and negative binomial distributions.
[36]
[PDF] arXiv:1808.04011v1 [math-ph] 12 Aug 2018
Aug 12, 2018 · Properties of the beta function. We will describe several properties ... Recall the binomial expansion. (52). (x + y)p = X k≥0 pk xp ...
[37]
[PDF] Accelerating the Hypergeometric Function with the Beta Integral to ...
Feb 12, 2024 · The beta integral is used to accelerate the hypergeometric function to derive new infinite series for π and gamma values, including Ramanujan- ...
[38]
[PDF] Representations of the Riemann zeta function - arXiv
Jan 31, 2019 · Integral representations of the Riemann zeta func- tion for odd-integer arguments. Journal of Computational and Applied Mathematics. 142, 435 ...
[39]
[PDF] Integral representations of the Legendre chi function - arXiv
and the Dirichlet beta function (2 )n β. ,. , are direct ... thus giving the integral representation of the Riemann zeta function for odd-integers arguments.
[40]
[PDF] Notes On An Approach To Apery's Constant - arXiv
Feb 26, 2024 · 3 Zeta(3). 8. 4 Zeta At Odd Integer Values. 16. 5 ... To summarize, here are a few of the representations of ζ(3) as trigonometric integral.
[41]
[PDF] arXiv:2403.00635v1 [math.CO] 1 Mar 2024
Mar 1, 2024 · So partitions separated by parity provide a new avenue for the exploration of combinatorial objects within the context of q-hypergeometric ...