Part 1: The 163 Dimensions of the Moonshine Functions
By Tito Piezas III
Keywords: monstrous moonshine, moonshine functions, Monster group, primegenerating polynomials, McKayThompson series, pi formulas, 163.
Abstract: An expository work on 1) Conway, Norton, and Atkin’s intriguing discovery that the moonshine functions span a linear space of dimension 163; and 2) the functions’ connection to Ramanujantype pi formulas.
I. Introduction II. The 172 McKayThompson Series and Linear Dependencies III. More Pi Formulas? IV. A Continued Fraction Using e^{π√163}
I. Introduction
Consider the following primegenerating polynomials of form P(n) = an^{2}an+c,
P(n) = n^{2}n+41 P(n) = 2n^{2}2n+19 P(n) = 3n^{2}3n+23 P(n) = 4n^{2}4n+3
distinctly prime for n = {1 to c1}, with the first (studied by Euler in 1772 in the form n^{2}+n+41) for a remarkable 40 consecutive n. And the following functions and their series,
1) j(τ) = 1/q + 744 + 196884q + 21493760q^{2} + 864299970q^{3} + … (A007240)
2) r_{2A}(τ) = 1/q + 104 + 4372q + 96256q^{2} + 1240002q^{3} + … (A101558)
3) r_{3A}(τ) = 1/q + 42 + 783q + 8672q^{2} + 65367q^{3} + … (A030197)
4) r_{4A}(τ) = 1/q + 24 + 276q + 2048q^{2} + 11202q^{3} + … (A097340)
where (throughout this article),
q = e^{2πiτ} = exp(2πiτ)
The first is the jfunction j(τ), a modular form of weight zero, but without the constant term, all these are simply the McKayThompson series T_{1A} , T_{2A}, T_{3A}, and T_{4A} of the Monster group. Solving the polynomials for n at the point P(n) = 0, we get the complex roots,
n = (1+√163)/2 n = (2+√148)/4 = (1+√37)/2 n = (3+√267)/6 n = (4+√32)/8 = (1+√2)/2
where, consistent with the requirements for modular forms, we have chosen the root that is on the upper half of the complex plane. The first has fundamental discriminant d = 163, the largest (in absolute value) with class number h(d) = 1, while the next two d = {148, 267} = {4∙37, 3∙89} have class number h(d) = 2. Let n = τ, and plugging τ into q = exp(2πiτ), we get the very small real numbers,
q = exp(π √163) q = exp(π/2 √148) = exp(π√37) q = exp(π/3 √267) q = exp(π/4 √32) = exp(π√2)
which, when substituted into the appropriate series above, yields the integers,
1) j((1+√163)/2) = 640320^{3}
2) r_{2A}((1+√37)/2) = 14112^{2}
3) r_{3A}((3+√267)/6) = 300^{3}
4) r_{4A}((1+√2)/2) = 2^{7}
The first integer appears in the Chudnovsky brothers’ pi formula,
where one can see the 163, as well as in the famous,
e^{π√163} ≈ 640320^{3} + 743.9999999999992…
The second integer appears in a pi formula found by Ramanujan in 1912 (before the Monster group was even discovered),
where 4(37) also appears, as well as in the almostinteger,
e^{π/2√148} = e^{π√37} ≈ 14112^{2} + 103.99997…
This (and similar formulas) inspired the one found by Chudnovsky brothers. The third appears in another kind of pi formula,
where, one can see d = 267, and in the almostinteger,
e^{π/3√267} = e^{π√(89/3)} ≈ 300^{3} + 41.99997…
And finally, 2^{7} appears in a fourth kind,
Using his own approach, Ramanujan found examples for all four types. (See Pi Formulas and the Monster Group for the complete list of 36 known formulas where the denominator C is an integer.) We instead used four functions based on T_{1A}, T_{2A}, T_{3A}, T_{4A}, but there are in fact many more of them. And here’s a the astonishing fact first noticed by Conway and Norton: the functions [1], [2], [3], [4], and others, call these the moonshine functions, span a linear space of dimension 163! Interesting “coincidence” that it had to be that number, isn’t it?
II. The 172 McKayThompson Series and Linear Dependencies
To recall, the Monster group has 194 conjugacy classes so its character table has 194 rows and 194 columns. Each column represents one conjugacy class and linear combinations of the entries gives its McKayThompson series which yields, to quote John McKay or Mark Ronan, a “moonshine function” (hence the title of this article). Some are the same so is reduced to 172 series, with the two distinct conjugacy classes A & B of order 27 still included as an exceptional case. (Author note: My thanks to Michael Somos for clarifying certain points.)
Table 1. The 172 McKayThompson series of Monstrous Moonshine
The first dozen coefficients of each of the 172 series are given in Table 4, Values of Head Characters, (p. 334) of John Conway and Simon Norton’s 1979 paper, Monstrous Moonshine. Conveniently, an index of the McKayThompson series is also in the Online Encyclopedia of Integer Sequences. (The moonshine functions, in accordance with Atlas notation, have the letter corresponding to the conjugacy class in uppercase, while the nonmonstrous ones have it in lowercase. By accident, T_{24B} is missing in the index, but is still in the OEIS.)
Also, David Madore has calculated the first 3500 coefficients of all the 172 moonshine McKayThompson series, while Michael Somos has an extensive database about the relationships between the coefficients and more.
But there are linear dependencies between the 172 series,
Table 2. The 9 Linear Dependencies
Taking into account these nine linear dependencies, then the dimension can be reduced to 172 – 9 = 163. To bring this insight to life, in his blog, Lieven le Bruyn quotes p. 227 of Mark Ronan’s book Symmetry and the Monster,
“Conway recalls that, ‘As we went down into the 160s, I said ‘...let’s guess what number we will reach.’ They guessed it would be 163 – which has a very special property in number theory – and it was! There is no explanation for this. We don’t know whether it is merely a coincidence, or something more....”
The first six were found by Conway and Norton, while the last three were discovered by Oliver Atkin who also showed that there are no more. These nine are found in Conway and Norton’s paper, but in rather technical form. For example, dependency # 7 is given as,
T_{4} + T_{42} = 2T_{8}
so I had to translate it into English, so to speak, i.e. convert it into the notation used in the OEIS (as well as in Madore’s and Somos’ databases). We then find from the OEIS,
T_{4C}(q) = 1/q + 20q  62q^{3} + 216q^{5}  641q^{7} + … (A007248)
T_{4D}(q) = 1/q  12q + 66q^{3}  232q^{5} + 639q^{7} + … (A007249)
T_{8E}(q) = 1/q + 4q + 2q^{3}  8q^{5}  q^{7} + … (A029841)
and we can easily see from the first few coefficients that indeed,
T_{4C} + T_{4D} = 2T_{8E }
though it is proven by Atkin that the equality holds for all terms. Thus, just as one can consider the series T_{27B} as redundant being equal to T_{27A}, and since there are nine linear dependencies, then only 163 series as “moonshine functions” are really necessary to linearly express all possible values of the 172 series. Or, as was mentioned, these functions span a linear space of dimension 163.
In an email, McKay calls this as a “delicious coincidence”, while Norton opines this may be merely a coincidence. However, in Conway and Norton’s paper, in the last paragraph of Section 9 which deals with moonshine for other groups, it was asked if there is “…a period three automorphism for the case Γ_{0}(67)+”. (The discriminant d = 67 is, of course, the younger brother of d = 163 in the family of Heegner numbers.)
III. More Pi Formulas?
In addition to the very interesting fact that the moonshine functions span a linear space of dimension 163, one question we can ask is, since there are so many of them, it is possible to use other McKayThompson series T_{pA} in Ramanujantype pi formulas? Ramanujan’s four kinds of pi formulas are of the form,
where h_{p} is one of various products of three Pochhammer symbols (their factorial equivalents are given in this article) and C is either j(τ), r_{2A}(τ), r_{3A}(τ), r_{4A}(τ) or equivalently, the McKayThompson series T_{1A}, T_{2A}, T_{3A}, T_{4A} with an appropriate constant term. It is tempting to speculate that, perhaps using different Pochhammer symbols and an appropriate constant term, one can use, say, T_{7A} as well.
In fact, J. Guillera has extended Ramanujan’s results and found pi formulas of form,
where g_{p} is one of various products of five Pochhammer symbols, but apparently the denominators are not values of any McKayThompson series. See “A Compilation of RamanujanType Formulas for 1/π^{m}” for m > 1 for a list.
IV. A Continued Fraction Using e^{π√163}
First, given the curious primegenerating polynomial,
P(n) = 4n^{2}+163
which is prime for n = {0 to 19}. Solve for P(n) = 0, hence n = (1/2)√163, plug this into r_{4A}(n) given in the Introduction and we find that it is an algebraic number,
r_{4A}((1/2)√163) = x^{24}
where x ≈ 5.31863… is the real root of the simple cubic equation x^{3}6x^{2}+4x2 = 0. Second, given the Ramanujan constant R,
R = e^{π√163} ≈ 640320^{3} + 743.9999999999992…
and we see that R and x have the beautiful relationships,
and,
where q = 1/R = 1/e^{π√163}. The root x has the exact value, let τ = (1+√163)/2,
where η(τ) is the Dedekind eta function which partly explains the two relationships between R and x. While this continued fraction is not unique to d = 163 (the general form is by E. Heine, 1846), I thought of including it in this article since R is a wellknown constant. The significance of the integer 24 will be further elaborated in other articles. Furthermore, it is also the case that,
e^{π√43} ≈ 12^{3}(9^{2}1)^{3} + 743.999… e^{π√67} ≈ 12^{3}(21^{2}1)^{3} + 743.99999… e^{π√163} ≈ 12^{3}(231^{2}1)^{3} + 743.999999999999…
The reason for the squares within the cubes is a certain Eisenstein series. But that’s another story ...
 End  © July 2011, Tito Piezas III You can email author at tpiezas@gmail.com.
