Mahler measures of $x^4+y^4+z^4+1+k xyz$

edit · history · discussion · files · short url · Mahler measure K3 surfaces hypergeometric functions integral special functions special values number theory
Numbers
$k$ 
$B(k)$
0:
0.4262783988175057909235214265961668730580067696296351075416064580265294512291165814891241883322429436
4:
1.353184954231269106150060654595395883532411255274121503198088564217949044461427480092749977792957654
5:
1.598871112971091495500568733187683332693518206073195607513819747579122677837231939366959414692791780
6:
1.786926566323026951751639615515459545917796131250457156841253290274980789092340338461140023943204898
7:
1.943354196611725959324960787691076915546887867377162299540808611644191119440279403582812075635695587
8:
2.077957459734832726297385754273187068151715983785367370007499626653728582604152140998534557574413257
9:
2.196302653733753938053349873223813085596405374142133081677713847553305373183465117064183308336748581
10:
2.301981911793937644574162893326220245753272892785748326016386558007567059299123017577739855158180410
11:
2.397483985326415607886393215236961530436283928188301593955104590461372317072620751721710081389093386
12:
2.484616561870926858537797015709213829830409518045092381016050101206351627415797321731988046438153441
13:
2.564738893298754853694166562045599750938822118435094566726501666772477942908173673669618319570690246
14:
2.638900930702086029472877434715083812837533152008093178734117281641659352505703480790525615559034630
15:
2.707931559437610163158316707868249548262081183509999225812040479354242968649832582318443765647864859
16:
2.772497096054115753165524075000651310836243136806612743567824933184042656997481247067361914279862816
$\sqrt[4]{648}$:
1.608320766458486457492054998981346769170932694651567962765031693616650070366027230122865925146095398
$\sqrt[4]{2304}$:
1.932934331174248133725586369252675719501825669089391778947403997732620175998157648509878909946153017
28:
3.332194747783616890562222793405014452794824894725492559099709174899965450126611940433875326619500730
Definition
For real $k\geq0$, let $P_k(x,y,z)=x^4+y^4+z^4+1+kxyz$. Its ordinary logarithmic Mahler measure is $B(k)=\int_{[0,1]^3}\log|P_k(e^{2\pi i t_1},e^{2\pi i t_2},e^{2\pi i t_3})|\,dt_1dt_2dt_3$, with natural logarithms and Haar probability measure [1], equation (2.6). Here $i^2=-1$.
Parameters
$k$
—   coefficient of xyz ($k\in\mathbb R$, $k\geq0$)
Formulas
(1)
For $k\geq4$, $B(k)=\log k-\sum_{n\geq1}(4n)!/[4n(n!)^4 k^{4n}]$, whose first correction is $6/k^4$ [1], Proposition 2.2; [2], Proposition 1.3(iii). The coefficient $(4n)!/(n!)^4$ is the constant term of $[(x^4+y^4+z^4+1)/(xyz)]^{4n}$.
(2)
Put $F(t)={}_3F_2(1/4,1/2,3/4;1,1;t)$, where ${}_pF_q$ denotes the generalized hypergeometric function [4]. Then $B(k)=\log k-\tfrac14\int_0^{256/k^4}(F(t)-1)\,dt/t$ for $k\geq4$, and $kB'(k)=F(256/k^4)$ for $k>4$. The origin is removable, with $(F(t)-1)/t\to3/32$.
(3)
$F(t)={}_2F_1(1/8,3/8;1;t)^2$ by Clausen's identity [5]. For $0\leq t<1$ the Gauss function is its real analytic branch from zero; at $t=1$ take the finite left limit.
(4)
The degree-four torus maps preserve Haar measure, so $B(0)=m(1+x+y+z)=7\zeta(3)/(2\pi^2)$, where $m$ uses the normalization in the definition and $\zeta$ is the Riemann zeta function. See the linear-polynomial measure. The linear-polynomial evaluation is recorded in [3], page 50.
(5)
Let $q=e^{2\pi i\tau}$, $\operatorname{Im}\tau>0$, and $\eta(\tau)=q^{1/24}\prod_{n\geq1}(1-q^n)$. For $f(\tau)=\eta(\tau)^2\eta(2\tau)\eta(4\tau)\eta(8\tau)^2=\sum_{n\geq1}a_nq^n$, put $L(f,3)=\sum_{n\geq1}a_n/n^3$, the uncompleted modular $L$-value. Then $B(4)=40\sqrt2\,L(f,3)/\pi^3$ [1], Corollary 2.6. The four further CM identities in [2], equations (1.9)–(1.12), use the same factor-four conversion.
Comments
(6)
Zeros on the torus use the ordinary integrable logarithm, with no regularization. The torus bound after division by $kxyz$ is $4/k$. Thus $k=4$ is the boundary of the exterior expansion; its ordinary finite value is included [2], Proposition 1.3(iii).
(7)
Rogers' auxiliary function is $f_4(u)=4B(u^{1/4})$, so the actual coefficient $k$ in this table corresponds to $u=k^4$, and $B(k)=f_4(k^4)/4$ [1]. Fourth-root-of-unity changes of the coefficient preserve the measure. The nonnegative representative is used here.
(8)
The identities $648^{1/4}$ and $2304^{1/4}$ mean the unique positive roots of $X^4-648$ in $(5,6)$ and $X^4-2304$ in $(6,7)$, respectively. The other source-selected fourth roots are $20736^{1/4}=12$ and $614656^{1/4}=28$ [2], Theorem 1.4.
(9)
The homogenization $x^4+y^4+z^4+w^4+kxyzw$ is a Dwork quartic pencil, giving the K3 geometric context. The measure, its normalized derivative period and a marked period ratio are different quantities.
(10)
The integer band $4\leq k\leq16$ is a bounded exploratory reference slice of primitive integer-coefficient quartics. The upper cutoff, four times the torus-contact coefficient, is noncanonical. The Fermat degeneration at zero and all four arithmetic comparisons in [2], Theorem 1.4, supplement it. The displayed exterior formulas make no prescription for the uncomputed ordinary values at $0<k<4$.
Programs
(P1)
Sage
import numberdb.sage
from sage.rings.rational_field import QQ
from sage.rings.integer_ring import ZZ
from sage.rings.complex_arb import ComplexBallField
# An exterior value at the actual polynomial coefficient k.
k = ZZ(5)
C = ComplexBallField(450)
u = k**4
B = (C(u).log()/4 - C(QQ(6)/u) * C(QQ(256)/u).hypergeometric(
    [QQ(5)/4, QQ(3)/2, QQ(7)/4, 1, 1], [2, 2, 2, 2])).real()
print(B)
# Attached generate.py handles the endpoint and Fermat case, and enumerates
# 0, all integers 4..16, positive roots 648^(1/4), 2304^(1/4), and 28.
References
[1]
M. D. Rogers, New ${}_5F_4$ hypergeometric transformations, three-variable Mahler measures, and formulas for $1/\pi$, Ramanujan Journal 18 (2009), 327–340; equations (2.6), (2.9), (2.21), Corollary 2.6. (arXiv)
[2]
D. Samart, Three-variable Mahler measures and special values of modular and Dirichlet $L$-series, arXiv version 2 (2013); Proposition 1.3(iii), Theorem 1.4. (arXiv)
[3]
C. J. Smyth, On measures of polynomials in several variables, Bulletin of the Australian Mathematical Society 23 (1981), 49–63; page 50. (doi)
Links
Similar tables
Holomorphic periods of hypergeometric motives of K3 type —   Template $2.4$ is the normalized derivative period $F(256/k^4)=kB'(k)$ on $k>4$.
Period map of the Dwork quartic K3 family —   That marked ratio of periods concerns the same quartic geometric context; it is distinct from this logarithmic measure.
Mahler measures of the reciprocal-sum family —   Rogers' equations (2.19) and (2.21) express the Bertin reciprocal-sum measure as a linear combination of quartic measures at transformed arguments [1].
Mahler measures of the squared reciprocal-product pencil —   That ordinary measure is Rogers' distinct function $f_3$ in the same three-function transformation theory [1].
Mahler measures of the body-centred cubic lattice family —   That measure $M$ gives Rogers' other auxiliary function $f_2(u)=2M(\sqrt u)$ for $u\geq64$ [1].
Mahler measures of short sums of monomials —   Its $n=4$ row equals the Fermat degeneration $B(0)$.
Values of the Riemann zeta function at rational numbers —   The value $\zeta(3)$ gives the Fermat degeneration in (4).
Values of Dirichlet L-functions at positive integers —   Primitive quadratic character values $L(\chi_{-4},2)$, $L(\chi_{-3},2)$ and $L(\chi_{-8},2)$ enter Samart's CM comparisons; $\chi_D(n)=(D/n)$ is the Kronecker character [2].
Data properties
Entries are of type: real number
Table is complete: no (it contains $k=0$, all integers $4\leq k\leq16$, $648^{1/4}$, $2304^{1/4}$ and $28$, giving 17 distinct coefficients. The integer band is a noncanonical exploratory slice; zero and four are intrinsic, and the four CM cases in Samart's Theorem 1.4 include the already listed integer $12$)
How they were obtained:

Exact rational constant-term series and Arb logarithms enclose every k>4 value at 433 working bits. In z=256/k^4 the positive hypergeometric coefficient ratios are less than one, giving an explicit geometric tail. Algebraic coefficients are evaluated using their exact fourth powers, so rounded roots are never substituted. At k=0 Arb zeta and pi enclose the Fermat evaluation.

more

At k=4 the integrated Clausen square is split at t=1/2. The Gauss connection formula at t=1 gives analytic and square-root series in s=1-t. Their positive coefficients are bounded by one; exact rational polynomial integration and a bound 400*2^(-N-1)/(N+1), with N=426, enclose the omitted high-half integral. Arb encloses gamma factors and algebraic prefactors. All resulting radii are below 1e-120 before 100 significant digits are stored. The attached mathematical-checks.md derives the bounds.

Independent mpmath integration of the Gauss square at 135 and 175 working decimal digits agrees for all 15 exterior coefficients. Independent PARI modular and primitive character L-functions at 140 and 180 digits verify the endpoint and all four CM comparisons, matching the newforms by Fourier coefficients. Independent mpmath zeta checks the Fermat value. Nonzero known-answer controls and deliberately wrong normalization, value and modular-form controls pass. These checks test the implementation and normalization; the rigorous uncertainty comes from the explicit bounds and Arb computation. Executable checks and numerical evidence are attached.