Claims · 11 dated claims on The design horocycle, each with its tag and its witness; newest 2026-10-05.
Proved No semigroup of SL(2, Z) containing T^m for some m != 0, T = [[1, 1], [0, 1]], and one element [[a, b], [c, d]] with c != 0 has every lower-left entry in S_F, -S_F or 0 at a digit set with fill < base: gamma T^(mk) gamma has lower-left entry c(a + d) + c^2 m k, whose absolute values meet [1, x] in at least x/(c^2 abs(m)) - abs(a + d)/abs(c) - 1 points, while A_F(x) <= 2 fill x^alpha with alpha < 1; so the design cusps contain the orbit of infinity under no semigroup that contains a translation and moves infinity. Witness: stack.md The design horocycle, No orbit
Proved The series psi^F f = sum f(Im gamma z) over the cosets with c in S_F is a function on the cylinder Gamma_infinity \ H; below the support of f its closed horocycle integral is h sum_(c in S_F) phi(c) g(c h^(1/2)), the smoothed m_F, and its first Fourier coefficient against e(x) is h sum_(c in S_F) mu(c) g(c h^(1/2)) + O(h^(3/2)), the Ramanujan sum at 1 being mu(c), a smoothed M_F whose leading term at fill < base is Z_F(1) F(-1/2) h^(1/2). Witness: stack.md The design horocycle, No orbit
Proved With a_F = (mu 1_(S_F)) * 1 and nu_F = (mu 1_(S_F)) * id, the design Eisenstein series Theta_F f(z) = sum a_F(gcd(c, d)) f(Im z/abs(cz + d)^2) over the lattice pairs equals sum_g a_F(g) psi_(f_g), f_g(u) = f(u/g^2), is invariant under SL(2, Z) with integral zeta(2) Z_F(2) F(-1) over the modular surface, its mean (3/pi) zeta(2) Z_F(2) F(-1) = (pi/2) Z_F(2) F(-1), and its closed horocycle integral at a height h below the support of f is h sum_c nu_F(c) g(c h^(1/2)); at the full digit set it is the incomplete Eisenstein series. Witness: stack.md The design horocycle, The design Eisenstein series
Verified The closed horocycle integral of Theta_F f, summed over every lattice pair weighted by a_F of its gcd, equals h sum_c nu_F(c) g(c h^(1/2)) at h = 4^(-k), k = 2..6, for the C^infinity bump, worst gap 7.5e-16 at base 3 {0,1} and 4.4e-16 at base 10 missing 7, over 1808 and 1298 weighted pairs, 3544 and 2564 arcs, at k = 6, the two routes sharing one quadrature rule; Theta_F f summed directly at 5 seeded points and 4 images of each under SL(2, Z) moves by at most 6.1e-13. Witness: lab/py/design-horocycle, verb bridge
Proved For a C^infinity window f on [1, 2] and E_F(y) = y^2 sum_c nu_F(c) f(cy) - F(2) Z_F(2): E_F(y) = y^2 sum_(d in S_F) mu(d) R_f(dy) with R_f(u) = sum_e e f(eu) - F(2) u^(-2); M_F(x) = O(x^b), 0 <= b < 2, gives E_F(y) = O(y^(2 - b)), so theta(F) = 1/2 gives E_F(y)/y^2 = O(A_F(1/y)^(1/2 + eps)); the rate y^(2 - b - eps) for every such window gives Z_F holomorphic on Re s > b, b >= 0; and holomorphy there with polynomial growth uniform on each strip b + eps <= Re s <= 3 gives the rate back. Witness: stack.md The design horocycle, The design meter
Proved The horocycle integral C_F(h) of Theta_F f has int_0^1 C_F(h) h^(s-2) dh = phi(s) zeta(2s) Z_F(2s) F(-s) on Re s > 1, phi(s) = xi(2s - 1)/xi(2s), so the design's scattering coefficient is phi(s) sum_g a_F(g) g^(-2s); on Re s > alpha/2 it is meromorphic with poles at most at s = 1 and at s = 1/2, the latter with coefficient zeta(0) Z_F(1), and any continuation further left is a continuation of Z_F(2s), which is not known; no Laplace eigenvalue and no transfer operator of the design enters. Witness: stack.md The design horocycle, The scattering coefficient
Proved At every digit set with fill < base, C_F(h) = (pi/2) Z_F(2) F(-1) - (1/2) Z_F(1) F(-1/2) h^(1/2) + O(h^(1 - b/2)) whenever M_F(x) = O(x^b), b < 1, so unconditionally with remainder O(h^(1 - alpha/2)) and under theta(F) = 1/2 with O(h^(1 - alpha/4 - eps)); the middle term is the cusp growth Theta_F f(x + iy) = (1/2) Z_F(1) F(-1/2) y^(1/2) + O(y^(alpha/2)) above height 1; at the full digit set, under M(x) = O(x^b) with b < 1, the expansion holds with the middle term absent because 1/zeta vanishes at s = 1. Witness: stack.md The design horocycle, The scattering coefficient
Verified At base 3 {0,1}, Z_F(1) lies in [0.662730, 0.680999], the partial sum to 2^24 plus a tail at most 0.005375, the exact reciprocal sum 0.000807 of the 16384 sixteen-digit elements above 2^24 and sum_(L > 16) (2/3)^(L-1) = 0.004567; and (C_F(h) - (pi/2) Z_F(2) F(-1))/h^(1/2) reads -0.093894, -0.066758, -0.071184, -0.077455, -0.072029 at h = 4^(-k), k = 4, 6, 8, 10, 12, against -(1/2) Z_F(1) F(-1/2) in [-0.072304, -0.070364], -0.071334 at the partial sum, while the full set reads 0.067150, -0.018951, -0.015743, -0.000692, -0.003755. Witness: lab/py/design-horocycle, verb bridge
Proved Sampling the horocycle along the design, int psi_f(x + ih) d mu_F(x) = sum_(c >= 1) sum_m c_c(m) hat G_c(m) conj(hat mu_F(m)) below the support of f, with c_c(m) the Ramanujan sum, the frequency-m term having, at m != 0, c-series sigma_(1-s)(m)/zeta(s), and hat mu_F(base m) = hat mu_F(m) at every integer m; no term of the expansion carries Z_F. Witness: stack.md The design horocycle, The sampled horocycle
Verified On y = 2^(-j), j in [8, 23], the least-squares slope of log of the per-octave rms of abs(E_F)/y^2 against log A_F(1/y), 15 octaves, windows balanced to F(2) = 0, reads 0.4918, 0.4986, 0.4371 at base 3 {0,1}, base 10 missing 7, base 5 {0,1,2} for C^infinity and 0.5509, 0.5189, 0.4426 for C^2, residuals 0.413, 0.300, 0.279 and lower and upper halves 0.1778 and 0.8896, 0.3324 and 0.4708, 0.5470 and 0.2417 for C^infinity; random signs on the support read 0.4685, 0.5470, 0.5489, abs(mu)0.9918, 0.9710, 0.9728, the full set 0.4970, 0.4956; a fit, and no exponent is read from it. Witness: lab/py/design-horocycle, verb meter
Conjecture For every digit set with 2 <= fill <= base - 1, E_F(y) = O(y^(2 - alpha/2 - eps)) for every C^infinity window and every eps > 0; read on three designs to 2^24, where the per-octave root mean square of abs(E_F)/y^2 stays between 0.0347 and 0.1562 times A_F(1/y)^(1/2) while that grows by factors 20.9 to 102.6, and at base 3 {0,1} climbs from 0.035 to 0.120 times it over the last seven octaves. Witness: stack.md The design horocycle, The meter read; lab/py/design-horocycle, verb meter