Abstract
Several interesting entries from page 196 in the lost notebook are examined. These relate to two of Ramanujan’s papers on integrals.
Access provided by Autonomous University of Puebla. Download chapter PDF
Keywords
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.
10.1 Introduction
On page 196 in his lost notebook, Ramanujan lists several identities that are related to the Riemann zeta function, Dirichlet L-series, and periodic zeta functions. Some of the identities are connected to previous results of Ramanujan in [256] and [258], but none of the identities on page 196 can be found in these papers. Furthermore, all of the identities on page 196 are new. The purpose of this chapter is to examine all of these interesting identities. Two of the identities were examined and generalized in a paper that the second author wrote with H.H. Chan and Y. Tanigawa [47].
10.2 Identities for Series Related to ζ(2) and L(1, χ)
At the top of page 196 in [269], Ramanujan records three identities related to ζ(2), and at the bottom of the page, he states a similar result related to L(1, χ), where χ is the nonprincipal primitive character modulo 4. In each of the first three identities, the coefficient 4 of the series on the right-hand side must be replaced by 2. We record the results in corrected form.
Entry 10.2.1 (p. 196)
Let Re x ≥ 0. Then
where the principal value of the square root is taken.
Proof. We assume throughout the proof that x ≥ 0. The more general result for Re x ≥ 0 will then hold by analytic continuation. We begin with the familiar theta transformation formula, which is found in Ramanujan’s notebooks [268], [39, p. 43, Entry 27(i)]. It will be convenient, however, to use the formulation, for Ret > 0,
which is found in Titchmarsh’s treatise [306, p. 22, Eq. (2.6.3)], for example. Integrate both sides of (10.2.2) over [0, x], invert the order of integration and summation by absolute convergence, and multiply both sides by − π to reach the identity
In comparing (10.2.3) with (10.2.1), we see that we must address the integrals on the right side of (10.2.3). First, set t = x ∕ u and then set n 2 u = (n + tx)2. Hence,
When examining (10.2.1) in relation to (10.2.3) and (10.2.4), we see that it remains to show that
Integrating the latter integral by parts, in particular integrating 1 ∕ (n + tx)2 and differentiating the exponential, we readily find that for Rex > 0,
On the other hand, after a little trickery and then a direct integration, we find that
From these two calculations, we see that (10.2.5) has been demonstrated, and so the proof is complete.
In Chap. 15 of his second notebook [268], [38, p. 306, Theorem 3.1], Ramanujan stated a general asymptotic formula for
as x → 0 + . If we set p = 2 and m = − 1, and replace x by πx in this asymptotic formula, we find that
as x → 0 + , which should be compared with (10.2.1). In [38, pp. 306–308], a proof of Ramanujan’s general asymptotic formula was obtained by contour integration. In the course of this proof, the error term, i.e., o(1) in (10.2.6), is represented by a certain contour integral. It seems to be very difficult, however, to transform this contour integral into the expression involving the infinite series on the right-hand side of (10.2.1).
Entry 10.2.2 (p. 196)
Let Re x ≥ 0. Then
and
where the principal value of the square root is taken.
Proof. As in the previous proof, we shall assume that x ≥ 0; an appeal to analytic continuation then establishes Entry 10.2.2 for Rex ≥ 0. We shall prove (10.2.7) and (10.2.8) with x replaced by y. In (10.2.1), replace x by z = x + iy, with y ≥ 0. Let θ = argz. Then (10.2.1) takes the form
Now,
from which we see that
Setting x = 0 and θ = \( \frac{1}{2}\pi \), we find that
If we now use (10.2.10) in (10.2.9), we deduce (10.2.7) with x replaced by y.
A similar calculation of Im E(x, y) followed by setting x = 0 and θ = \( \frac{1}{2}\pi \) yields (10.2.8) with x replaced by y.
Entry 10.2.3 (p. 196)
For x ≥ 0,
Proof. We begin by specializing the well-known theta relation for an odd primitive character [101, p. 70, Eq. (9)]. In our case, this odd primitive character is the real nonprincipal character modulo 4. Accordingly, for t > 0,
Integrate both sides of (10.2.12) over [0, x], and then multiply both sides by − π ∕ 4 to deduce that
where we utilized Leibniz’s series for π ∕ 4 and made the substitution t = x ∕ u in the integrals on the right side. Next, set (2n + 1)2 u = (2n + 1 + 2tx)2. Then
If we substitute (10.2.14) into (10.2.13), we obtain (10.2.11) to complete the proof.
10.3 Analogues of Gauss Sums
We now offer three claims from the middle of page 196 of [269]. These were first proved in a more general setting by Berndt, Chan, and Tanigawa [47]. More precisely, they evaluate the sum
where m and k are positive integers, in several ways, obtaining evaluations in terms of trigonometric functions, Stirling numbers of the second kind, and ballot numbers. On page 196, Ramanujan considers only the case m = 1.
Entry 10.3.1 (p. 196)
Let a be an even positive integer. Then
We first prove (10.3.3) assuming the truth of (10.3.1) and (10.3.2).
Proof of (10.3.3) of Entry 10.3.1 . Using the addition formulas for sin and cos, we easily find that
We evaluate the more general series
where r is an even positive integer. In order to effect these evaluations, we need to introduce periodic Bernoulli numbers.
Let A = { a n }, − ∞ < n < ∞, denote a sequence of numbers with period k. Then the periodic Bernoulli numbers B n (A), n ≥ 0, can be defined [66, p. 55, Proposition 9.1], for | z | < 2π ∕ k, by
Furthermore [66, p. 56, Eq. (9.5)], for each positive integer n,
where B n (x), n ≥ 0, denotes the nth Bernoulli polynomial. We say that A is even if a n = a − n for every integer n.
The complementary sequence B = { b n }, − ∞ < n < ∞, is defined by [66, p. 32]
It is easily checked that if A is even, then B is even, and that (10.3.6) holds if and only if
Now set
If A and r are even and if r ≥ 2, then [66, p. 49, Eq. (6.25)]
From (10.3.6) and (10.3.7), we see that the sequences A and B are not symmetric. Thus, we note from above that since A is even,
We are now ready to state general evaluations in closed form for S a (r) and T a (r).
Theorem 10.3.1.
If S a (r) and T a (r) are defined by (10.3.4) and if r and a are even positive integers, then
and
In our work below, we need the value of the Gauss sum [54, p. 43, Exercise 5]
where c is an even positive integer.
Before proceeding further, we show that (10.3.1) and (10.3.2) are special cases of (10.3.9) and (10.3.10), respectively. Let r = 2 in Theorem 10.3.1. Recall that B 2(x) = x 2 − x + \( \frac{1}{6}\pi\). Then
upon the use of (10.3.11) twice.
The proof of (10.3.2) follows along the same lines, but note that in this case, by (10.3.11),
Proof of Theorem 10.3.1 . Let
which is an even periodic sequence with period a, since a is even. Then, from (10.3.6) and (10.3.11),
Therefore, by (10.3.5), with B in place of A,
If we substitute (10.3.12) into (10.3.8) and simplify, we deduce (10.3.9).
The proof of (10.3.10) is analogous to that for (10.3.9). Now we set
which of course is even, and repeat the same kind of argument that we gave above.
We now provide another evaluation of the series on the left-hand sides of (10.3.1) and (10.3.2) in closed form. However, we obtain evaluations in entirely different forms from those claimed by Ramanujan in Entry 10.3.1.
Theorem 10.3.2.
Let a be an even positive integer, a ≥ 2. Then
and
Proof. Setting n = ka + j, 0 ≤ k < ∞, 1 ≤ j ≤ a, we find that
Singling out the term for j = a ∕ 2 and noting that the terms in the outer sum with indices j and a − j are identical, we find from (10.3.15) that
say. There remains the evaluation of U(j, a).
First observe that if for − ∞ < k ≤ − 1, we set k = − r − 1, then
It therefore suffices to evaluate the bilateral sum in (10.3.17).
To evaluate U(j, a), recall the partial fraction decomposition
Differentiating once above, we find that
Putting z = r ∕ k in (10.3.18), we deduce that
Putting (10.3.19) in (10.3.16), we complete the proof of (10.3.13).
The proof of (10.3.14) follows along exactly the same lines. In analogy with (10.3.17), we now easily deduce that
By the same identical argument that we used above, we conclude that
We record a few examples to illustrate Theorem 10.3.2, namely,
Equating the evaluations of S a (2) and T a (2) in (10.3.13) and (10.3.14), respectively, with those in (10.3.1) and (10.3.2), we obtain identities that would be surprising if we had not known of their origins, namely,
and
Note that on the left-hand sides above, the sums contain only trigonometric functions, while on the right-hand sides the sums contain both polynomials and trigonometric functions. Trigonometric identities involving polynomials in the summands appear to be rare. The sums on both sides of the identities may be regarded as new analogues of Gauss sums.
In fact, Ramanujan states a second equality for the sum on the left side of (10.3.3). We slightly reformulate this result in the next entry.
Entry 10.3.4 (p. 196)
If a is an even positive integer, then
Proof. Our proof depends on two results from Ramanujan’s papers [256, 262]. First, if a is an even positive integer [262, Eq. (17)], [267, p. 132], then
Now, from [256, Eq. (50)], [267, p. 67],
If we substitute (10.3.22) in (10.3.21) and then multiply both sides of the resulting equality by 4π 2 ∕ a 3 ∕ 2, we deduce that
which is easily seen to be equivalent to (10.3.20).
References
B.C. Berndt, Ramanujan’s Notebooks, Part II, Springer-Verlag, New York, 1989.
B.C. Berndt, Ramanujan’s Notebooks, Part III, Springer-Verlag, New York, 1991.
B.C. Berndt, H.H. Chan, and Y. Tanigawa, Two Dirichlet series evaluations found on page 196 of Ramanujan’s lost notebook, Math. Proc. Cambridge Philos. Soc. 153 (2012), 341–360.
B.C. Berndt, R.J. Evans, and K.S. Williams, Gauss and Jacobi Sums, John Wiley, New York, 1998.
B.C. Berndt and L. Schoenfeld, Periodic analogues of the Euler-Maclaurin and Poisson summation formulas with applications to number theory, Acta Arith. 28 (1975), 25–68.
H. Davenport, Multiplicative Number Theory, 3rd ed., Springer-Verlag, New York, 2000.
S. Ramanujan, Some definite integrals connected with Gauss’s sums, Mess. Math. 44 (1915), 75–85.
S. Ramanujan, Some definite integrals, J. Indian Math. Soc. 11 (1915), 81–87.
S. Ramanujan, On certain infinite series, Mess. Math. 45 (1916), 11–15.
S. Ramanujan, Collected Papers, Cambridge University Press, Cambridge, 1927; reprinted by Chelsea, New York, 1962; reprinted by the American Mathematical Society, Providence, RI, 2000.
S. Ramanujan, Notebooks (2 volumes), Tata Institute of Fundamental Research, Bombay, 1957; second ed, 2012.
S. Ramanujan, The Lost Notebook and Other Unpublished Papers, Narosa, New Delhi, 1988.
E.C. Titchmarsh, The Theory of the Riemann Zeta-Function, Clarendon Press, Oxford, 1951.
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 2013 Springer Science+Business Media New York
About this chapter
Cite this chapter
Andrews, G.E., Berndt, B.C. (2013). Identities Related to the Riemann Zeta Function and Periodic Zeta Functions. In: Ramanujan's Lost Notebook. Springer, New York, NY. https://doi.org/10.1007/978-1-4614-4081-9_10
Download citation
DOI: https://doi.org/10.1007/978-1-4614-4081-9_10
Published:
Publisher Name: Springer, New York, NY
Print ISBN: 978-1-4614-4080-2
Online ISBN: 978-1-4614-4081-9
eBook Packages: Mathematics and StatisticsMathematics and Statistics (R0)