import numberdb.sage as numberdb
from sage.rings.rational_field import QQ
from sage.rings.complex_arb import ComplexBallField
def period_map(t, digits=100):
t = QQ(t)
if t < 0 or t == 1:
raise ValueError("t must be nonnegative and different from 1")
C = ComplexBallField(numberdb.bits(digits, losing=180))
i = C(0, 1)
if t == 0:
return C(QQ(1)/2, QQ(1)/2)
z = C(t**(-4))
s = (1-z).sqrt() if t > 1 else -i*(z-1).sqrt()
u = ((1-s)/2).sqrt()
m = 2*u/(1+u)
v = i*(1-m).elliptic_k()/m.elliptic_k()
return C(QQ(1)/4, v.imag()) if t == QQ(3)/4 else v
period_map(QQ(5)/6)
# Stored selection: t=0 and every reduced a/b != 1 with 1<=a,b<=6,
# ordered by height max(a,b), then by value within each height.
Arb ball arithmetic encloses exact rational coefficients, the algebraic pullback and the ratio of elliptic integrals at 513 working bits, with the denominator separated from zero. The 100-significant-digit decimals enclose the resulting component balls. The exact origin, exactly zero real parts for $t>1$, and the real component $1/4$ at $t=3/4$ are preserved; the component $1/4$ has an exact lambda-transformation certificate in mathematical-checks.md.
Every inexact fibre was also checked by separately implemented Picard–Fuchs Taylor continuation in mpmath at 145 and 180 decimal digits, using the Frobenius germ rather than the elliptic pullback. Finite eta products with analytic tail bounds check the modular inverse at every nonzero fibre, including propagated stored-interval uncertainty and inverse conditioning. Known-answer and deliberately wrong branch, normalization, parameter and enclosure controls are retained in the attached executable checks and results. The Picard–Fuchs convergence check is numerical, not a separate rigorous enclosure proof; production bounds come from Arb.