Carlson's symmetric elliptic integral of the first kind #
This file defines Carlson's symmetric elliptic integral of the first kind (DLMF 19.16.1)
R_F(x, y, z) = ½ ∫₀^∞ (t + x)^(-1/2) (t + y)^(-1/2) (t + z)^(-1/2) dt
for complex x, y, z, with the principal branch of each factor (t + x) ^ (-1/2 : ℂ), together
with its degenerate case R_C(x, y) = R_F(x, y, y) (DLMF 19.2.17). Carlson's integrals are
symmetric replacements for Legendre's incomplete elliptic integrals: every Legendre integral is a
combination of them, and they are the input to the duplication algorithm for computing elliptic
integrals.
The integral defining R_F converges absolutely when x, y, z lie in Complex.slitPlane ∪ {0}
and at most one of them is 0. Since R_F is symmetric, the convergence theorem
Complex.integrableOn_carlsonRF_integrand is stated with the possibly vanishing argument
in the first position. The behavior of carlsonRF outside this domain is not characterized here;
where the integrand is not integrable, carlsonRF is the totalized value 0 of the Bochner
integral and carries no meaning.
Main results #
Complex.carlsonRF,Complex.carlsonRC— the integralsR_FandR_C.Complex.integrableOn_carlsonRF_integrand,Complex.integrableOn_carlsonRC_integrand— absolute convergence of the defining integrals ofR_FandR_C.Complex.carlsonRF_comm_left,Complex.carlsonRF_comm_right—R_Fis symmetric in its three arguments.Complex.carlsonRF_ofReal_mul— homogeneity of degree-1/2under positive real scaling.Complex.carlsonRF_self,Complex.carlsonRC_self— the diagonal valuesR_F(x, x, x) = R_C(x, x) = x^(-1/2).Complex.carlsonRF_pos,Complex.carlsonRC_pos—R_FandR_Care positive real numbers at nonnegative real arguments, at most one of the symmetric ones zero.Complex.carlsonRC_eq_integral— the defining integral ofR_C(DLMF 19.2.17).Complex.carlsonRC_ofReal_eq_arccos,Complex.carlsonRC_ofReal_eq_arcosh— the elementary evaluation ofR_Cat real arguments (DLMF 19.2.19, 19.2.20):R_C(x, y) = arccos √(x/y) / √(y - x)for0 ≤ x < y, andR_C(x, y) = arcosh √(x/y) / √(x - y)for0 < y < x.Complex.carlsonRF_ofReal_eq_integral— at nonnegative real arguments,R_Fis the real integral½ ∫₀^∞ (√(t + x) √(t + y) √(t + z))⁻¹ dt.
References #
- NIST Digital Library of Mathematical Functions, §19.16 and §19.2(iv), https://dlmf.nist.gov/19.16.
- B. C. Carlson, Special Functions of Applied Mathematics, Academic Press, 1977, Chapter 9.
- B. C. Carlson, Numerical computation of real or complex elliptic integrals, Numer. Algorithms 10 (1995), 13–26.
B. Braams' lean-codes (https://github.com/bjbraams/lean-codes) formalizes Carlson's book,
including R_F, on top of Carlson's general R-function, a Dirichlet average over a simplex. This
file instead works directly from the integral definition above and is not ported from it.
Carlson's symmetric elliptic integral of the first kind (DLMF 19.16.1):
R_F(x, y, z) = ½ ∫₀^∞ (t + x)^(-1/2) (t + y)^(-1/2) (t + z)^(-1/2) dt, with principal complex
powers of each factor.
The integral converges when x, y, z ∈ slitPlane ∪ {0} with at most one of them 0
(integrableOn_carlsonRF_integrand); its behavior outside this domain is not characterized here.
Where the integrand is not integrable, the value is the totalized junk value 0.
Instances For
Carlson's degenerate integral R_C(x, y) = R_F(x, y, y) (DLMF 19.2.17). Its integral form
½ ∫₀^∞ (t + x)^(-1/2) (t + y)⁻¹ dt is carlsonRC_eq_integral.
Instances For
Convergence #
Convergence of R_F. The integrand of carlsonRF x y z is integrable on (0, ∞) when
x ∈ slitPlane ∪ {0} and y, z ∈ slitPlane. By symmetry of the integrand this covers every
argument triple in slitPlane ∪ {0} with at most one zero entry.
Convergence of R_C. The integrand t ↦ (t + x)^(-1/2) (t + y)⁻¹ of
carlsonRC_eq_integral is integrable on (0, ∞) when x ∈ slitPlane ∪ {0} and
y ∈ slitPlane.
Symmetry and homogeneity #
Homogeneity of R_F. For real l > 0,
R_F(l x, l y, l z) = l^(-1/2) R_F(x, y, z). No hypothesis on x, y, z is needed: the substitution
t = l s identifies the two integrals whether or not they converge.
Diagonal values #
Real arguments #
R_F at nonnegative real arguments is a real integral:
R_F(x, y, z) = ½ ∫₀^∞ (√(t + x) √(t + y) √(t + z))⁻¹ dt for real x, y, z ≥ 0. The two integrands
agree on (0, ∞), so no convergence hypothesis is needed.
R_F is positive at real arguments: for real x ≥ 0 and y, z > 0, R_F(x, y, z) is a
positive real number (in the order ComplexOrder, so its imaginary part is 0). By symmetry, any
one of the three arguments may be the one that vanishes.