New $_5F_4$ hypergeometric transformations, three-variable Mahler measures, and formulas for $1/\pi$
Mathew D. Rogers

TL;DR
This paper develops new hypergeometric transformations and identities linking three-variable Mahler measures, leading to explicit evaluations of $_5F_4$ functions and novel formulas for calculating 1/π.
Contribution
It introduces new $_5F_4$ hypergeometric transformations and identities connecting Mahler measures, enabling explicit evaluations and new formulas for 1/π.
Findings
Derived new $_5F_4$ hypergeometric identities.
Obtained explicit $_5F_4$ evaluations.
Produced new formulas for 1/π.
Abstract
New relations are established between families of three-variable Mahler measures. Those identities are then expressed as transformations for the hypergeometric function. We use these results to obtain two explicit evaluations, and several new formulas for .
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsMathematical functions and polynomials · Advanced Mathematical Identities · Iterative Methods for Nonlinear Equations
New hypergeometric transformations, three-variable Mahler
measures, and formulas for
Mathew D. Rogers
*Department of Mathematics, University of British Columbia
Vancouver, BC, V6T-1Z2, Canada
email: [email protected]*
Abstract
New relations are established between families of three-variable Mahler measures. Those identities are then expressed as transformations for the hypergeometric function. We use these results to obtain two explicit evaluations, and several new formulas for .
MSC: 33C20, 33C05, 11F66
1 Introduction
In this paper we will study the consequences of some recent results of Bertin. Recall that Bertin proved -series expansions for a pair of three-variable Mahler measures in [8]. As usual the Mahler measure of an -variable polynomial, , is defined by
[TABLE]
We will define and in terms of the following three-variable Mahler measures
[TABLE]
We can recover Bertin’s original notation by observing that , and after substituting in Eq. (1.2) we see that [8].
In Section 2 we will show how to establish a large number of interesting relations between , , and three more three-variable Mahler measures. For example, for sufficiently large Eq. (2.21) is equivalent to
[TABLE]
Rodriguez-Villegas briefly mentioned the Mahler measure on the last page of [17].
We will also show that identities like Eq. (1.3) are equivalent to transformations for the hypergeometric function. Recall that the generalized hypergeometric function is defined by
[TABLE]
where . We have restated Eq. (1.3) as a hypergeometric transformation in Eq. (2.24). As a special case of Eq. (1.3) we can also deduce that
[TABLE]
where , and is the usual -series of . We will conclude Section 2 with a brief discussion of some related, but still unproven, evaluations of the and hypergeometric functions.
It turns out that and are also closely related to Watson’s triple integrals. For appropriate values of , Watson showed that and reduce to products of elliptical integrals (for relevant results see [18], [13], [14], and [12]). In Section 3 we will use some related transformations to prove new formulas for . For example, we will show that
[TABLE]
Notice that this formula for involves the Domb numbers. Chan, Chan and Liu obtained a similar formula for involving Domb numbers in [10], we have recovered their result in (3.10). Zudilin and Yang also discovered some related formulas for in [20]. All of the transformations that we will utilize in Section 3 follow from differentiating the identities established in Section 2.
2 Identities between Mahler measures and transformations for the function
Bertin proved that both and have convenient -series expansions when is parameterized correctly. Before stating her theorem, we will define some notation. As usual let
[TABLE]
and define by
[TABLE]
Notice that if then , where is the Eisenstein series of weight on the full modular group [3].
Theorem 2.1**.**
(Bertin) For sufficiently small
[TABLE]
where
[TABLE]
In this section we will show that both and reduce to linear combinations of hypergeometric functions. We will accomplish this goal by first expressing each of the functions
[TABLE]
in terms of . We will then exploit those identities to establish linear relations between functions in the set . This is significant since , , and all reduce to hypergeometric functions. In particular this implies the non-trivial fact that both and also reduce to linear combinations of functions.
Proposition 2.2**.**
The following identities hold for sufficiently large:
[TABLE]
For
[TABLE]
and if
[TABLE]
*Proof. * We can prove each of these identities using a method due to Rodriguez-Villegas [17]. We will illustrate the proof of Eq. (2.7) explicitly. Rearranging the Mahler measure shows that
[TABLE]
If , then , hence by the Taylor series for the logarithm
[TABLE]
Notice that Eq. (2.7) holds whenever , since is harmonic in .
While Proposition 2.2 shows that the results in this paper easily translate into the language of hypergeometric functions, the relationship to Mahler measure is more important than simple pedagogy. Bertin proved that for certain values of the zero varieties of the (projectivized) polynomials from equations (1.1) and (1.2) define hypersurfaces. She also proved formulas relating the -functions of these surfaces at to rational multiples of the Mahler measures. Proposition 2.2 shows that these results imply explicit evaluations (see Corollary 2.6 for explicit examples). While it might also be interesting to interpret the polynomials from equations (2.4) through (2.6) in terms of hypersurfaces, we will not pursue that direction here.
Theorem 2.3**.**
For sufficiently small
[TABLE]
where
[TABLE]
The following inverse relations hold for sufficiently small:
[TABLE]
*Proof. *We can use Ramanujan’s theory of elliptic functions to verify the first half of this theorem. Recall that the elliptic nome is defined by
[TABLE]
It is a well established fact that is a rational function of whenever . For example if , then . Therefore we can verify Eq. (2.12) by differentiating with respect to , and by showing that the identity holds when .
Observe that when both sides of Eq. (2.12) approach . Differentiating with respect to yields
[TABLE]
This final identity follows from applying three well known formulas:
[TABLE]
We can verify equations (2.13) and (2.14) in a similar manner by using the fact that , and .
The crucial observation for proving equations (2.15) through (2.17) is the fact that satisfies the following functional equation for any prime :
[TABLE]
We will only need the case to prove Eq. (2.15):
[TABLE]
Notice that this last formula always allows us to eliminate from an equation. Applying the substitutions , , and to Eq. (2.12) yields
[TABLE]
Since this system of equations is non-singular, we can invert the matrix to recover Eq. (2.15). We can prove equations (2.16) and (2.17) in a similar fashion.
If we compare Theorem 2.3 with Bertin’s results we can deduce some obvious relationships between the Mahler measures. For example, combining Eq. (2.14) with Eq. (2.3), and combining Eq. (2.13) with Eq. (2.2), we find that
[TABLE]
Notice that many more identities follow from substituting equations (2.15) through (2.17) into formulas (2.2), (2.3), (2.12), (2.13) and (2.14). However, for the remainder of this section we will restrict our attention to equations (2.19) and (2.20). In particular, we will appeal to the theory of elliptic functions to transform those results into identities which depend on rational arguments.
If we let , then it is well known that is an algebraic function of for (for example see [4] or [5]). It follows immediately that , , , , and are also algebraic functions of . The following lemma lists several instances where those functions have rational parameterizations.
Lemma 2.4**.**
Suppose that , where . The following identities hold for sufficiently small:
[TABLE]
[TABLE]
Rational formulas also exist for certain values of and :
[TABLE]
The main difficulty with Lemma 2.4 is the fact that very few values of reduce to rational functions of . Consider the set as an example. While Lemma 2.4 shows that , , and are all rational with respect to , the formula for involves radicals. Recall that if , then
[TABLE]
where Since the curve is elliptic with conductor , it follows immediately that rational substitutions for will never reduce to a rational function. For the sake of legibility, we will therefore avoid all identities which involve those four functions simultaneously. By avoiding pitfalls of this nature, we can derive several interesting results from Lemma 2.4.
Theorem 2.5**.**
For sufficiently large
[TABLE]
*Proof. * These identities follow from applying Lemma 2.4 to equations (2.19) and (2.20). If we consider Eq. (2.19), then Lemma 2.4 shows that , , and are all rational functions of . Forming a resultant with respect to , we obtain
[TABLE]
Simplifying with the aid of a computer, this becomes
[TABLE]
If we choose so that , then , and a formula for follows in a similar fashion.
If we let with and sufficiently large, then Eq. (2.22) reduces to the following infinite series identity:
[TABLE]
Similarly, if we let then Eq. (2.21) is equivalent to:
[TABLE]
In Section 3 we will differentiate equations (2.23) and (2.24) to obtain several new formulas for . But first we will conclude this section by deducing some explicit evaluations.
Recall that for certain values of , Bertin evaluated and in terms of the -series of surfaces. She also proved equivalent formulas involving twisted cusp forms. Amazingly, her formulas correspond to cases where the right-hand sides of equations (2.21) and (2.22) collapse to one hypergeometric term. We can combine her results with equations (2.23) and (2.24) to deduce several new evaluations.
Corollary 2.6**.**
If , then
[TABLE]
If , then
[TABLE]
While many famous identities, such as Dougall’s formula [3], reduce special values of the function to gamma functions, equations (2.25) and (2.26) do not fit into this category. Rather these new formulas are higher dimensional analogues of Boyd’s conjectures. In particular, Boyd has conjectured large numbers of identities relating two-variable Mahler measures (that mostly reduce to functions) to the -series of elliptic curves [9]. The most famous outstanding conjecture of this type asserts that
[TABLE]
where
[TABLE]
and “” indicates numerical equality to at least decimal places. Recently, Kurokawa and Ochiai proved a formula [15] which simplifies this last conjecture to
[TABLE]
Of course it would be highly desirable to rigorously prove Boyd’s conjectures. Failing that, it might be interesting to search for more hypergeometric identities like equations (2.25) and (2.26). This line of thought suggests the following fundamental problem with which we shall conclude this section:
Open Problem: Determine every -series that can be expressed in terms of generalized hypergeometric functions with algebraic parameters.
3 New formulas for
In the previous section we produced several new transformations for the hypergeometric function. Now we will differentiate those formulas to obtain some new transformations, and several accompanying formulas for . The following formula is a typical example of the identities in this section:
[TABLE]
Ramanujan first proved identities like Eq. (3.1) in his famous paper “Modular equations and approximations to ” [16]. He showed that the following infinite series holds for certain constants , , and :
[TABLE]
Ramanujan determined many sets of algebraic values for , , and by expressing them in terms of the classical singular moduli and . He also stated (but did not prove) several formulas for where is replaced by for (for more details see [6] or [11]).
Ramanujan’s formulas for have attracted a great deal of attention because of their intrinsic beauty, and because they converge extremely quickly. For example, Mathematica calculates using a variant of a Ramanujan-type formula due to the Chudnovsky brothers [19]:
[TABLE]
More recent mathematicians including Yang and Zudilin have derived formulas for which are not hypergeometric, but still similar to Eq. (3.3). For example, Yang showed that
[TABLE]
and Zudilin gave many infinite series for containing nested sums of binomial coefficients [20]. All of the formulas that we will prove, including Eq. (3.1), are essentially of this type. Before proving the next theorem, we will point out that equation (3.4) appears implicitly in the work of Chan, Chan and Liu, and can be obtained by combining equations (4.5) and (4.6) in their paper [10].
Theorem 3.1**.**
For sufficiently small
[TABLE]
If is sufficiently small
[TABLE]
*Proof. * Applying the operator to Eq. (2.23), and then simplifying yields
[TABLE]
Eq. (3.4) then follows from applying a transformation:
[TABLE]
We can prove Eq. (3.5) in a similar manner by differentiating Eq. (2.24) and then using
[TABLE]
Equation (3.6) can each be derived in three steps. First square both sides of the following identity (see Corollary 6.2 in [5]):
[TABLE]
When is sufficiently small, we can apply the following transformation
[TABLE]
and then conclude by setting . Similarly, equation (3.7) follows from combining another identity (which can be shown to follow from Theorem 9.15 in [5]):
[TABLE]
with a similar transformation
[TABLE]
and then setting .
While the infinite series in Theorem 3.1 are not hypergeometric since they involve nested binomial sums, they are still interesting. In particular, those formulas easily translate into unexpected integrals involving powers of modified Bessel functions. For sufficiently small, Eq. (3.4) is equivalent to
[TABLE]
where is the modified Bessel function of the first kind. Recall the series expansions for and :
[TABLE]
Eq. (3.8) is surprising because there is no known hypergeometric expression [2]. It is therefore not obvious that the Laplace transform of should equal a hypergeometric function. M. Lawrence Glasser has kindly pointed out that equation (3.8) is essentially a well known result, and that a variety of similar integrals have also been studied by Joyce [14], Glasser and Montaldi [12], and others.
Finally, we will list a few formulas for . Notice that equation (3.10) first appeared in the work of Chan, Chan and Liu [10].
Corollary 3.2**.**
Let , then the following formulas are true:
[TABLE]
Let , then the following identity holds:
[TABLE]
*Proof. * We can use Eq. (3.4) to easily deduce that if
[TABLE]
then , and . Since the left-hand side of this last formula equals when , it is easy to verify equations (3.9) through (3.11) [11].
We can verify Eq. (3.12) in a similar manner by combining Eq. (3.5) with Ramanujan’s formula
[TABLE]
4 Conclusion
We will conclude the paper by suggesting two future projects. Firstly, it would be desirable to determine whether or not a rational series involving exists for . Secondly, it might be interesting to consider the Mahler measure
[TABLE]
since arises from Ramanujan’s theory of signature .
Acknowledgements
The author would like to thank David Boyd for the many helpful discussions and encouragement. The author also thanks Fernando Rodriquez-Villegas and Marie José Bertin for the useful discussions. The author thanks Wadim Zudilin for the useful communications and for the reference [20]. The author also thanks Heng Huat Chan and Zhiguo Liu for the reference and their useful remarks [10]. The author extends his gratitude to M. Lawrence Glasser for the references [12] and [14]. Finally, thanks to Jianyun Shan for pointing out [13].
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[1] G. E. Andrews and B.C. Berndt, Ramanujan’s Lost Notebook, Part I, Springer-Verlag, New York, 2005.
- 2[2] C. M. Bender, D. C. Brody and B. K. Meister, On powers of Bessel functions, J. Math. Phys. 44(2003), no. 1, 309–314.
- 3[3] B.C. Berndt, Ramanujan’s Notebooks, Part II, Springer-Verlag, New York, 1989.
- 4[4] B.C. Berndt, Ramanujan’s Notebooks, Part III, Springer-Verlag, New York, 1991.
- 5[5] B.C. Berndt, Ramanujan’s Notebooks, Part V, Springer-Verlag, New York, 1998.
- 6[6] B.C. Berndt, S. Bhargava and F.G. Garvan, Ramanujan’s theories of elliptic functions to alternative bases, Trans. Amer. Math. Soc. 347 (1995), 4163-4244.
- 7[7] M. J. Bertin, Mahler’s measure and L-series of K 3 𝐾 3 K 3 hypersurfaces, (Preprint).
- 8[8] M. J. Bertin, Mesure de Mahler D’Hypersurfaces K 3 𝐾 3 K 3 , (Preprint 2005). See http://arxiv.org/abs/math/0501153
