Abstract
On page 206 in his lost notebook, Ramanujan recorded the following enigmatic identity for his theta function \(\varphi (q)\):
We give the three missing terms. In addition, we calculate the class invariant \(G_{343}\) and further special values of \(\varphi (e^{-n\pi }),\) for \(n = 7, 21, 35,\) and 49.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Ramanujan’s general theta function f(a, b) is defined by [24, p. 197], [4, p. 34]
which provides an alternative formulation [5, p. 3] for the classical theta functions \((\theta _i(z,q))_{i=1}^4\) in [34, pp. 462–465]. The symmetry reflected in the definition of f(a, b) is inherited by its representation by the Jacobi triple product identity [17, pp. 176–183], [24, p. 197], [4, p. 35], which states that
where
In Ramanujan’s notation, the theta function \(\varphi (q)\) is defined by [24, p. 197], [4, p. 36]
where the series and product representations are straightforward from (1) and (2), respectively. Furthermore, we set [24, p. 197], [4, p. 37]
If \(q = \exp (-\pi \sqrt{n})\), for a positive rational number n, the class invariant \(G_n\) of Ramanujan [21, 22, pp. 23–39], [7, p. 183] and Weber [33, p. 114] is defined by
For odd n, Ramanujan’s values for \(G_n\) are listed in [7, pp. 189–199], Weber’s list is in [33, pp. 721–726], and motivation is in [12]. The class invariants are algebraic [15, pp. 214, 257].
A fundamental result in the theory of elliptic functions is that for a positive rational n,
Here, \({_2}F{_1}\) is the ordinary or Gaussian hypergeometric function [34, pp. 24, 281], K is the complete elliptic integral of the first kind [34, pp. 499–500], [4, p. 102], [9], and \(k_n\) is a singular value or singular modulus [34, pp. 525–527], [12, 14, 19, 7, p. 183] of the elliptic integral K. The singular values are algebraic [1]. Ramanujan used the notation \(\alpha _n := k_n^2\) [12, 7, p. 183]. This statement is given more generally in [24, p. 207], [4, p. 101, Entry 6]. An overview of the theory of elliptic functions can be found in [4, p. 102], [9, 12, 7, pp. 323–324].
For a positive rational n, a positive integer d, and \(q = \exp (-\pi \sqrt{n})\), in the theory of modular equations, the multiplier m of degree d is defined by
The multiplier m can be defined more generally, as given in [4, p. 230], [9, 12, 7, p. 324]. In our case, m defined in (7) is algebraic [37]. An overview of the theory of modular equations can be found in [4, pp. 213–214], [9, 12, 7, p. 185].
It is classical [34, pp. 524–525], and it was also discovered by Ramanujan [24, p. 207], [4, p. 103], [23, p. 248], [9, 7, p. 325], that
For a positive rational n, Ramanujan recorded his values for \(\varphi (e^{-n\pi })\) in terms of \(\varphi (e^{-\pi })\), but in view of (8), \(\varphi (e^{-n\pi })\) is therefore determined explicitly. At scattered places in his notebooks, Ramanujan recorded some values of \(\varphi (e^{-n\pi })\) when n is a power of two, namely, for \(n = 1, 2, 4, 1/2, 1/4\) [23, p. 248], [7, p. 325]; and when \(n \ge 3\) is an odd integer, namely, for \(n = 3, 5, 7, 9,\) and 45 [23, pp. 284, 285, 297, 287, 312], [9, 7, pp. 327–337]. The values at powers of two are parts of more general results from Ramanujan’s second notebook [24, p. 210], [4, pp. 122–123]. The evaluations for the odd values were established by Berndt and Chan [9, 7, pp. 327–337]. They also determined the values for \(n = 13, 27,\) and 63.
Selberg and Chowla [26] showed that for any singular value \(k_n\), the elliptic integral \(K(k_n\)) is expressible in terms of gamma functions. J. M. Borwein and Zucker [14, 37, 13, p. 298] evaluated \(K(k_n)\), for \(n = 1, \dots , 16\). Thus, by (6), we have the value of \(\varphi (e^{-\pi \sqrt{n}})\) in these cases. We give two theta function values, corresponding to \(k_3\) and \(k_7\) [19, 38], respectively:
and
where \(B(x,y) \mathop {:=} \Gamma (x)\Gamma (y)/\Gamma (x + y)\), for \({\text {Re}}(x),{\text {Re}}(y) \mathop {>} 0\), is the beta function [34, pp. 253–256].
If we would like to calculate \(\varphi (e^{-\pi \sqrt{r}})\) explicitly, for a positive integer r, then if r is square-free and the corresponding values \(k_r\) and \(K(k_r)\) are known, we can use (6). If r is not square-free and the value \(\varphi (e^{-\pi \sqrt{n}})\) is known, where n is the square-free part of r and \(\sqrt{r} = d\sqrt{n}\), for a positive integer d, then we can calculate \(\varphi (e^{-d\pi \sqrt{n}})\), with appropriate modular equations of degree d, which contains the class invariants \(G_{n}\) and \(G_{d^2n}\), with known explicit values, and the multiplier m. There are other particular methods, as we see next, in Entry 1.1.
On page 206 in his lost notebook [25], Ramanujan recorded the following identities.
Entry 1.1
(p. 206) Let
-
(i)
\(\frac{\varphi (q^{1/7})}{\varphi (q^7)} = 1 + u + v + w\). Then
-
(ii)
\(p := uvw = \frac{8q^2 (-q; q^2)_\infty }{(-q^7; q^{14})^7_\infty }\) and
-
(iii)
\(\frac{\varphi ^8(q)}{\varphi ^8(q^7)} - (2 + 5p)\frac{\varphi ^4(q)}{\varphi ^4(q^7)} + (1 - p)^3 = 0.\) Furthermore,
-
(iv)
\(u = \bigg (\frac{\alpha ^2 p}{\beta }\bigg )^{1/7}, \quad v = \bigg (\frac{\beta ^2 p}{\gamma }\bigg )^{1/7}, \text {\quad and \quad } w = \bigg (\frac{\gamma ^2 p}{\alpha }\bigg )^{1/7},\) where \(\alpha , \beta ,\) and \(\gamma \) are the roots of the equation
-
(v)
\(r(\xi ) := \xi ^3 + 2\xi ^2\bigg (1 + 3p - \frac{\varphi ^4(q)}{\varphi ^4(q^7)}\bigg ) + \xi p^2(p+4) - p^4 = 0.\) For example,
-
(vi)
\(\varphi (e^{-7\pi \sqrt{7}}) = 7^{-3/4}\varphi (e^{-\pi \sqrt{7}})\Big \{1 + (\quad )^{2/7} + (\quad )^{2/7} + (\quad )^{2/7}\Big \}.\)
We remark that (i)–(v) hold for \(|q| < 1\), with \(q \ne 0\) in (iv). If \(q = 0\), then \(u = v = w = 0\). Parts (i)–(v) were proved by Son [27, 2, pp. 180–194], [28, pp. 198–200],
Part (i) is recorded in Ramanujan’s second notebook [24, p. 239], [4, p. 303] as well, in the form of
from where the values of u, v, and w can be determined [27, 2, p. 181], [28, p. 198] as
since they are not clearly defined in (iv). For (i) to hold, we could give u, v, and w in any arbitrary order, but throughout the paper, we use the definitions in (11).
In (vi), Ramanujan gave an enigmatic identity, as a fragmentary example, where on the right-hand side there are three missing terms. Note that Ramanujan used the exponent 2/7, instead of 1/7, as he should have according to (iv). It turns out that this is correct, so it is likely that Ramanujan knew something about the structure of the terms. Ramanujan wrote \(7^{3/4}\) instead of \(7^{-3/4}\) on the right-hand side. We have corrected this.
Berndt [8], Son [28], and Andrews and Berndt [2, p. 181] leave the problem of the three missing terms open. They wonder why Ramanujan did not record the terms in (vi). We cannot answer this question, but Ramanujan gave us the procedure in (i)–(v) that helps us to find them. The equations in Entry 1.1(i)–(v) can be interpreted as the following:
-
(i)
Our aim is to find the values of u, v, and w for a given \(|q| < 1\).
-
(ii)
Calculate p.
-
(iii)
Solve the quadratic equation for \(\varphi ^4(q)/\varphi ^4(q^7)\) and choose the correct root.
-
(v)
By solving the cubic equation \(r(\xi ) = 0\), find \(\alpha , \beta ,\) and \(\gamma \).
-
(iv)
By using \(\alpha , \beta ,\) and \(\gamma \) in the correct order, construct u, v, and w.
Before we take these steps, we give some preliminaries in Sects. 2 and 3. Ramanujan gave us no hints on which is the correct choice for \(\varphi ^4(q)/\varphi ^4(q^7)\) in (iii), and how to find the correct order of the roots of r in (v). Possibly, he stopped after solving (v), since the exponents 2/7 on the right-hand side of (vi) become apparent after one finds a proper representation for the roots of r, but before their correct order is determined. This part needs most of our preparation; thus we give lemmas on the correct order of \(\alpha , \beta ,\) and \(\gamma \) in Sect. 3. Our main result is in Sect. 4, where we give the three missing terms of (vi). In Sect. 5, we give a closed-form expression for the class invariant \(G_{343}\). In Sect. 6, we calculate the special values of \(\varphi (e^{-n\pi }),\) for \(n = 7, 21,\) and 35, and the value of \(\varphi (e^{-7\pi \sqrt{3}})\). In Sect. 7, we conclude our article with the value of \(\varphi (e^{-49\pi })\), given as a second example of Ramanujan’s type for (i).
2 Preliminaries
We recall the transformation formula for \(\varphi (e^{-\pi \sqrt{n}})\).
Lemma 2.1
If n is a positive rational number, then
Proof
The transformation formula for \(\varphi (q)\) states that [24, p. 199], [4, p. 43] if \(a, b > 0\) with \(ab = \pi \), then
The lemma is the special case for \(a^2 = \pi /\sqrt{n}\). \(\square \)
Ramanujan gave some properties of \(G_n\). We need the following.
Lemma 2.2
If n is a positive rational number, then \(G_n = G_{1/n}\).
Proof
See Ramanujan’s paper [21, 22, pp. 23–39] or Yi’s thesis [35, pp. 18–19]. \(\square \)
Our next lemma helps us to find p in Entry 1.1(ii). It gives a connection between p and \(G_n\).
Lemma 2.3
If \(q = \exp (-\pi \sqrt{n})\), for a positive rational number n, then
Proof
From Entry 1.1(ii), (4), and (5), we have
\(\square \)
The Chebyshev polynomial \(U_n\) of the second kind [20, pp. 3–4] is defined for \(\left| \cos \theta \right| \le 1\) by
The polynomial \(U_n\) satisfies the recurrence relation \(U_0(x) = 1, U_1(x) = 2x\),
which extends the definition to all complex values x. Thus, \(U_n\) is a polynomial of degree n, with real, distinct roots, which are symmetric about zero. The first few cases are listed in [16, p. 994].
Lemma 2.4
The roots of \(U_n\) are
Proof
This follows directly from the definition (12). \(\square \)
3 The order of the roots
In Entry 31 of Chapter 16 of Ramanujan’s second notebook [24, p. 200], [4, pp. 48–49], the following general theorem is stated. Let \({\mathcal {U}}_k = a^{k(k+1)/2}b^{k(k-1)/2}\) and \({\mathcal {V}}_k = a^{k(k-1)/2}b^{k(k+1)/2}\) for each nonnegative integer k. Then, for a positive integer n,
For a positive integer n, and \(|q| < 1\), \(q \ne 0\), let
where the series representation is obtained by (1). Note by (3) that \(u_0 = \varphi (q^n)\). By setting \((a, b) = (q^{1/n}, q^{1/n})\) into (13), we obtain
We consider \(f(a,b) = af(a^{-1}, a^2b)\), for \(|ab| < 1\) and \(a \ne 0\), which is followed by Entry 18(i),(iv) in [24, p. 197], [4, pp. 34–35]. Because of this, we have \(u_k = u_{n-k},\) for \(k = 1,\dots ,n-1.\) Since \(u_0\) is nonzero, for each odd integer n,
For \(n = 7\), we arrive at Entry 1.1(i), and for u, v, and w, defined in (11), we have
For finding the three missing terms, it is enough to handle u, v, and w for \(0< q < 1\), in which case these values are real. Thus in this section, we state our lemmas under this condition.
First, we would like to show that for \(0< q < 1\), the values \(u_k\) defined in (14) are in descending order, for \(k = 0,\dots ,\lfloor n/2\rfloor \), where \(\lfloor n/2 \rfloor \) is the largest integer r, such that \(r \le n/2\). For this purpose, we now overview some classical results.
The third Jacobi theta function \(\theta _3(z, q)\) is defined by [34, pp. 463–464]
or equivalently, \(\theta _3(z \mid \tau )\) is defined by
With \(q = e^{\pi i \tau }\), (16) and (17) are equal, and \(|q| < 1\) if and only if \({\text {Im}} \tau > 0\). We use both notations, depending on whether we would like to indicate the dependence on q or \(\tau \). Using the identity \(e^{2niz} + e^{-2niz} = 2\cos (2nz)\), we can rewrite (16) as [34, pp. 463–464]
A straightforward calculation shows the connection between the third Jacobi theta function and Ramanujan’s general theta function. From (16) and (1), we find that [4, p. 3]
and from (17) and (1), we have
In the following lemma, we show a monotonicity property of the third Jacobi theta function.
Lemma 3.1
If m is an integer and \(0< q < 1\), then
-
(i)
if \(z \in [m\pi , m\pi + (\pi /2)]\), then \(\theta _3(z,q)\) is strictly monotonically decreasing in z, and
-
(ii)
if \(z \in [m\pi - (\pi /2), m\pi ]\), then \(\theta _3(z,q)\) is strictly monotonically increasing in z.
Proof
Since \(\theta _3(z, q)\) has period \(\pi \) in z [34, p. 463], it is enough to show that the statement is true for \(m = 0\). Because \(\theta _3(z, q)\) is an even function of z [34, p. 464], it is enough to show one of the two cases.
To prove (i), we consider \(\theta _3(z,q)\) for \(z \in [0, \pi /2]\) and \(0< q < 1\). The zeros of \(\theta _3(z,q)\) are of the form of \(z = (k + (1/2))\pi + (\ell + (1/2))\pi \tau \), where \({\text {Im}} \tau > 0\), for all integer values of k and \(\ell \) [34, pp. 465–466], therefore it has no zeros for \(z \in [0, \pi /2]\). Since the series for \(\theta _3(z,q)\) in (16) is a series of analytic functions, uniformly convergent in any bounded domain of values of z [34, p. 463], \(\theta _3(z, q)\) is therefore a continuous function. Furthermore, from (16) or (18), \(\theta _3(z,q)\) is a real-valued function, for that \(\theta _3(0, q) = \sum _{n = -\infty }^{\infty } q^{n^2} > 0\), thus by the contrapositive of the intermediate value theorem, we find that \(\theta _3(z, q)\) is positive for \(z \in [0, \pi /2]\).
From (19) and from the Jacobi triple product identity (2), we find that [34, p. 469]
Since the resulting series converges uniformly [34, pp. 471, 479], we may differentiate the logarithm of (21) with respect to z. Denoting the first partial derivative of \(\theta _3(z, q)\) with respect to z by \(\theta '_3(z, q)\), and taking the logarithmic derivative of (21), we find that [34, p. 489]
Now, note that the denominator of the summand on the right-hand side of (22) is positive, since
Thus, the sign of the sum is depending only on the sign of \(\sin 2z\). Since \(\theta _3(z,q) > 0\), and since \(\sin 2z = 0\) for \(z \in \{0, \pi /2\}\) and \(\sin 2z > 0\) for \(z \in (0, \pi /2)\), we find that \(\theta '_3(z,q) = 0\) for \(z \in \{0, \pi /2\}\) and \(\theta '_3(z,q) < 0\) for \(z \in (0,\pi /2)\). Since \(\theta _3(z, q)\) is continuous for \(z \in [0, \pi /2]\), we conclude that \(\theta _3(z,q)\) is strictly monotonically decreasing for \(z \in [0, \pi /2]\). \(\square \)
Now, we prove the needed monotonicity property of the values \(u_k\) defined in (14).
Lemma 3.2
If n is a nonnegative integer and \(0< q < 1\), then \(u_k\) is positive and strictly monotonically decreasing for \(k = 0, \dots , n/2\), when n is even, and for \(k = 0, \dots , (n-1)/2\), when n is odd.
Proof
Let n and q be fixed with the given conditions. From (14), we deduce that \(u_k\) is positive. To prove the monotonicity, we rewrite \(u_k\) in terms of the third Jacobi theta function. From (14), (1), (17), and (20), we find that
where here and in the rest of the proof \(k = 0, \dots , \lfloor n/2 \rfloor \),
Since \(\log q < 0\), we have \({\text {Im}} \tau > 0\). Note that \(\tau \) is independent of k, and that \(e^{\pi i \tau } = e^{-\pi / C} = q^n\).
Now, we apply Jacobi’s imaginary transformation formula [34, pp. 474–476], [6, pp. 140–141]
where by (24),
and
where C is some positive value independent of k,
Since \(e^{-\pi i / \tau } = e^{-\pi C} \in (0,1)\) and \((z_k/\tau ) = (k\pi /n)\) is a strictly monotonically increasing sequence in \([0,\pi /2]\), applying Lemma 3.1(i) with \(m = 0\), we complete the proof. \(\square \)
The following lemma is a corollary of Lemma 3.2 for u, v, and w defined in (11).
Lemma 3.3
For \(0< q < 1\),
-
(i)
\(2> u> v> w > 0\),
-
(ii)
\(0< p < 8\).
Proof
To prove (i), we represent u, v, and w as in (15), and we use Lemma 3.2 with \(n = 7\). Part (ii) follows from (i) with \(p = uvw\), as it is defined in Entry 1.1(ii). \(\square \)
We need the following two statements on the values of u, v, and w defined in (11).
Lemma 3.4
For \(|q| < 1\),
-
(i)
\(u^3v + v^3w + w^3u = 2\bigg (\frac{\varphi ^4(q)}{\varphi ^4(q^7)} - 3p - 1\bigg )\) and
-
(ii)
\(u^7 + v^7 + w^7 = \frac{\varphi ^8(q)}{\varphi ^8(q^7)} - 7(p - 2)\frac{\varphi ^4(q)}{\varphi ^4(q^7)} + 7p^2 -49p - 15.\)
Proof
See Son’s article [27] or the Andrews–Berndt book [2, pp. 185–186, 194].
Next, we give the correct root of Entry 1.1(iii) for \(\varphi ^4(q)/\varphi ^4(q^7)\), when \(0< q < 1\).
Lemma 3.5
For \(0< q < 1\),
Proof
From Lemma 3.3(i), we know that \(u, v, w > 0\), thus by Lemma 3.4(i) we have
By solving Entry 1.1(iii) for \(\varphi ^4(q)/\varphi ^4(q^7)\), we find that only the given solution satisfies this. \(\square \)
The cubic polynomial r defined in Entry 1.1(v) has the following property.
Lemma 3.6
For \(0< q < 1\), r has three distinct positive roots.
Proof
We recall that if a cubic polynomial with real coefficients has a positive discriminant, then it has three distinct real roots. For \(|q| < 1\), depending on the value of \(\varphi ^4(q)/\varphi ^4(q^7)\) from Entry 1.1(iii), r has one of the following two possible discriminants:
From Lemma 3.5, we know that for \(0< q < 1\), the discriminant of r is \(\Delta _{-}\). It follows by elementary algebra that
It is clear that for \(p > 0\), we have \(\Delta _{+} > 0\), and for \(p> 0\) and \(p \ne 8\), we have \(p^{10}(p - 8)^6 > 0\). Since by Lemma 3.3(ii) if \(0< q < 1\), then \(0< p < 8\); thus we find that \(\Delta _{-} > 0\).
From Lemma 3.3 we know that if \(0< q < 1\), then \(p > 0\) and \(u, v, w > 0\). Thus, from the construction in Entry 1.1(iv) it follows that the roots of r are positive. \(\square \)
The next lemma helps to choose the correct order of the roots of r in the case of \(0< q < 1\).
Lemma 3.7
For \(0< q < 1\), suppose that the roots of r are given in order \((\alpha , \beta , \gamma )\) such that they satisfy the following two conditions:
-
(i)
\(\frac{\alpha ^2 p}{\beta } + \frac{\beta ^2 p}{\gamma } + \frac{\gamma ^2 p}{\alpha } = \frac{\varphi ^8(q)}{\varphi ^8(q^7)} - 7(p - 2)\frac{\varphi ^4(q)}{\varphi ^4(q^7)} + 7p^2 -49p - 15\) and
-
(ii)
\(\frac{\alpha ^2}{\beta }> \frac{\beta ^2}{\gamma } > \frac{\gamma ^2}{\alpha }\).
Then
Condition (i) guarantees the correct order of \(\alpha , \beta \), and \(\gamma \) in Entry 1.1(iv), so that Entry 1.1(i) holds. Condition (ii) provides the correct order of u, v, and w, so that (11) holds.
Proof
First, for each possible order of \(\alpha , \beta ,\) and \(\gamma \), consider the set of possible values of u, v, and w given in Entry 1.1(iv). For \((\alpha , \beta , \gamma ), (\beta , \gamma , \alpha ), (\gamma , \alpha , \beta )\) and for \((\gamma , \beta , \alpha ), (\beta , \alpha , \gamma ), (\alpha , \gamma , \beta )\), we have
respectively. Thus, it is enough to consider the order \((\alpha , \beta , \gamma )\) and its reverse \((\gamma , \beta , \alpha )\). By Lemma 3.4(ii), we know that for at least one of these two sets it is true that the sum of their seventh powers fulfills the condition in (i). We show that exactly one of the two sets fulfills it. Suppose that
After rearrangement, we find that
but this is contradiction, since p is positive by Lemma 3.3(ii) and \(\alpha , \beta ,\) and \(\gamma \) are distinct, positive numbers by Lemma 3.6.
Since p is positive by Lemma 3.3(ii), the condition (ii) guarantees that \(u> v > w\), which holds by Lemma 3.3(i). \(\square \)
Lastly, we need the following trigonometric identity.
Lemma 3.8
We have
Proof
Let \(\theta := \pi / 7\), and let
By using power-reduction [16, p. 32] and product-to-sum [16, p. 29] identities, we derive that
Since we know [3] that \(b - c - a = -1/2\) and \(abc = 1/8\), the proof is complete. \(\square \)
4 The three missing terms
We complete Ramanujan’s enigmatic septic theta function identity Entry 1.1(vi).
Theorem 4.1
We have
Proof
We use the results in Entry 1.1(i)–(v) with \(q = \exp (-\pi /\sqrt{7})\). First, by using Lemma 2.1, we rewrite Entry 1.1(i) as
With \(G_7\) given in [18, 30, 31, 7, p. 189], by Lemma 2.2, we find that \(G_{1/7} = G_7 = 2^{1/4}\). By Lemma 2.3 with \(n = 1/7\), for Entry 1.1(ii) we have
Next, we solve the equation in Entry 1.1(iii). By Lemma 3.5, or in this special case by Lemma 2.1, we have
Now, we have all the coefficients of r given in Entry 1.1(v). We have to determine the zeros of
With an appropriate polynomial transformation, we relate r to \(U_6\) defined in (12) or given in [16, p. 994]. Note that for \(\xi \ne 0\),
and \(r(0) = U_6(0) = -1\). From Lemma 2.4, we know that \(U_6\) has the roots \(\xi _k = \cos (k\pi /7),\) for \(k = 1,\dots , 6\), for which \(\xi _k \ne 0\) and \(|\xi _k| = |\xi _{7-k}|\). Thus, we find that
Lastly, we determine the appropriate order of the roots \(\alpha , \beta ,\) and \(\gamma \) in Entry 1.1(iv). The choice
is correct, since the condition Lemma 3.7(i) holds by Lemma 3.8, and Lemma 3.7(ii) is satisfied by the inequalities \(0< \cos (3\pi /7)< \cos (2\pi /7)< \cos (\pi /7) < 1\). We arrive at
which completes the proof. \(\square \)
By combining the value (10) and Theorem 4.1, we find the evaluation of \(\varphi (e^{-7\pi \sqrt{7}})\), i.e.,
5 The class invariant \(G_{343}\) in closed-form
Berndt [8], Son [27], and Andrews and Berndt [2, p. 181] proposed the explicit value of the class invariant \(G_{343}\) as an open problem. Actually, Watson showed in [30, 31] that \(G_{343} = 2^{1/4}x\), where \(x^7 - 7x^6 - 7x^5 - 7x^4 - 1 = 0\). This can be proved by using a modular equation of degree 7, given in Entry 19(ix) of Chapter 19 of Ramanujan’s second notebook [24, p. 240], [4, p. 315], [11, Lemma 3.5], [36, Theorem 2.4]. Watson [31] solved this septic polynomial as well. Thus, we have
where
and
In this section, we give a closed-form expression for \(G_{343}\), based on our previous results.
Theorem 5.1
We have
where
and
Proof
By taking \(q = \exp (-\pi \sqrt{7})\), the expression for m is obtained by Theorem 4.1 as
For Entry 1.1(ii), by Lemma 2.3 with \(n = 7\) and with \(G_7 = 2^{1/4}\) [18, 30, 31, 7, p. 189], we find that
from which we have \(G_{343} = 2^{1/4}p^{-1/7}\). On the other hand, for Entry 1.1(iii), we have
After rearrangement, the following cubic polynomial in p can be deduced:
We solve this equation, and choose the only real root, which is given by (28). \(\square \)
6 Examples for Entry 1.1(iii)
The Borwein brothers [13, p. 145] observed first [12] that class invariants can be used to calculate certain values of \(\varphi (e^{-n\pi })\). By using ideas from Berndt’s proof [4, p. 347] of Entry 1(iii) of Chapter 20 of Ramanujan’s second notebook [24, p. 241], [4, p. 345], one can deduce that [7, p. 330, (4.5)], [9, (3.10)]
Similarly, as in [7, p. 339, (8.11)] and [7, p. 334, (5.7)], we have
and
There are two groups of values for \(\varphi (q)\), which can by deduced from Entry 1.1. The first one is from Entry 1.1(iii), and the second is from Entry 1.1(i). Now, in the spirit of Entry 1.1(iii), we give a result for \(\varphi (e^{-7\pi \sqrt{n}})/\varphi (e^{-\pi \sqrt{n}})\), which is similar to those in (29)–(31), and then we calculate the values of \(\varphi (e^{-7\pi }), \varphi (e^{-7\pi \sqrt{3}}), \varphi (e^{-21\pi }),\) and \(\varphi (e^{-35\pi })\).
Lemma 6.1
If n is a positive rational number, then
For
we define
Using m(p), we can state Lemma 6.1 as
Note that m(p) is a multiplier of degree 7, defined in (7), with the substitution \(n \mapsto (49n)^{-1}\).
Proof
We apply Entry 1.1(ii),(iii). For \(q = \exp (-\pi /\sqrt{49n})\), by Lemmas 2.3 and 2.2, we find that \(p = 2\sqrt{2}G_{49n}/G_n^7\). By using Lemma 3.5 with Lemma 2.1, we complete the proof. \(\square \)
The next result is from Ramanujan’s first notebook [23, p. 297], [7, p. 328], and proved first by Berndt and Chan [9, 7, pp. 336–337]. Our proof uses Entry 1.1(iii), but all of these proofs depend on some of the modular equations given in Entry 19 of Chapter 19 in Ramanujan’s second notebook [24, p. 240], [4, pp. 314–324]. The value of \(\varphi (e^{-7\pi })\), in terms of \(\varphi (e^{-\pi })\) given in (8), is stated as follows.
Theorem 6.2
We have
Proof
We apply Lemma 6.1 with \(n = 1\). From [7, p. 189], \(G_1 = 1\), and from [21, 22, p. 26], [7, p. 191],
Using \(4 + \sqrt{7} = (\sqrt{7} + 1)^2/2\), from (32), we find that
We make two observations. Let a be a real number, and let
Then, straightforward algebra shows that
and for \(a \ne 0\),
Now, set \(a := (28)^{1/4} = \sqrt{2} \cdot 7^{1/4}\). Comparing (36) with (37), we see that \(p = p_a\). Furthermore, note in (38) and (39) that the terms with a factor of \(a^4 - 28\) vanish. Thus, from (33)–(39), we find that
\(\square \)
The next theorems appear to be new.
Theorem 6.3
We have
Proof
We apply Lemma 6.1 with \(n = 3\). From [31, 7, p. 189], \(G_3 = 2^{1/12}\), and from [21, 22, p. 28], [32, 7, p. 194],
Thus, from (32), we find that
The next observation follows by elementary algebra. Let a be a real number, and let
and
Then
where
Now, set \(a := (756)^{1/6} = 2^{1/3} \cdot \sqrt{3} \cdot 7^{1/6}\). A straightforward calculation shows that
Thus, by comparing (41) with (42), we see that \(p = p_a\). Furthermore, note in (44) that the term with a factor of \(a^6 -756\) vanishes. Thus, from (33), (34), and (41)–(44), with some simplification, we find that
\(\square \)
By combining the value (9) and Theorem 6.3, we find the evaluation of \(\varphi (e^{-7\pi \sqrt{3}})\), i.e.,
The next values are derived in terms of \(\varphi (e^{-\pi })\) given in (8).
Theorem 6.4
We have
where
and m(p) is given in (33).
Proof
We apply Lemma 6.1 with \(n = 9\). From [21, 22, p. 24], [7, p. 189],
and from [21, 22, p. 29], [10, 7, p. 197],
An elementary calculation shows that
From (32), \(p = 2\sqrt{2}G_{441}/G_9^7\). Using the values for \(G_9\) and \(G_{441}\) given above, we deduce (45).
From [23, p. 284], [9, 7, pp. 327, 329–331],
Combining Lemma 6.1 and (47) completes the proof. \(\square \)
Another expression for \(\varphi (e^{-21\pi })\) can be obtained by using (29) with \(n = 49\) and with the values \(G_{441}\) from (46), and \(G_{49}\) from (35). Combining the result with the value of \(\varphi (e^{-7\pi })\) from Theorem 6.2, after some simplification we find that
Theorem 6.5
We have
where
and m(p) is given in (33).
Proof
We apply Lemma 6.1 with \(n = 25\). From [21, 22, p. 26], [7, p. 190],
and from [21, 22, p. 30], [29, 7, p. 199],
A simple calculation shows that
We use
as it is given by Watson [29], which also can be verified directly. From (32), \(p = 2\sqrt{2}G_{1225}/G_{25}^7\). Using the values for \(G_{25}\) and \(G_{1225}\) given above, we deduce (48).
From [23, p. 285], [9, 7, pp. 327–329],
Combining Lemma 6.1 and (50) completes the proof. \(\square \)
Another expression for \(\varphi (e^{-35\pi })\) can be obtained by using (30) with \(n = 49\) and with the values \(G_{1225}\) from (49), and \(G_{49}\) from (35). Combining the result with the value of \(\varphi (e^{-7\pi })\) from Theorem 6.2, after some simplification we find that
7 Examples for Entry 1.1(i)
Now, we review our proof for Theorem 4.1, which is in the form of Entry 1.1(i). Then, we give the value of \(\varphi (e^{-49\pi })\), as a second illustration of Entry 1.1(i). We use the results from the proof of Theorem 6.2.
Let \(a \in \{0, (28)^{1/4}\}\). After (37), let
Since the second term on the right-hand side of (38) vanishes for \(a \in \{0, (28)^{1/4}\}\), after (40), let
Furthermore, let
Comments on the proof of Theorem 4.1 Set \(a := 0\). Using the notations of the proof of Theorem 4.1, from (51), (52), and (53), we find that \(p = p_a = 1\), \(\varphi ^4(q)/\varphi ^4(q^7) = m_a^2 = 7\), and
By Lemma 2.4, we know that \(\cos (\pi /7)\) is a root of \(U_6\). Thus, by using the power-reduction formula [16, p. 32]
we can factor r over \({\mathbb {Q}}(\cos (\pi /7))\) as
where
In the same manner as in the proof of Lemma 3.8, we deduce
where the order of the roots is determined by Lemmas 3.7 and 3.8. We construct u, v, and w, and the proof is complete. \(\square \)
After these preliminaries, we derive the value of \(\varphi (e^{-49\pi })\). The corresponding polynomial r, and its roots are more complicated than in Ramanujan’s example in Entry 1.1(vi). We used Mathematica for polynomial factorization and numerical evaluations.
Theorem 7.1
We have
where
and
Proof
We use the results in Entry 1.1(i)–(v) with \(q = \exp (-\pi /7)\). First, by using Lemma 2.1, we rewrite Entry 1.1(i) as in (55).
Then, set \(a := (28)^{1/4} = \sqrt{2} \cdot 7^{1/4}\), and consider \(p_a, m_a,\) and \(r_a\) from (51), (52), and (53). By Lemma 2.2, and by Lemma 2.3 with \(n = 1/49\), for Entry 1.1(ii), we have \(p = 2\sqrt{2}G_{49}/G_1^7\). Thus, comparing (36) with (51), we see that \(p = p_a = (a^2/2) + a + 1\). For Lemma 1.1(iii), by using Lemmas 3.5 and 2.1, we arrive at
where the last equation is obtained by the comparison of (40) and (52). Thus, by comparing Entry 1.1(v) with (53), we find that \(r(\xi ) = r_a(\xi )\), where we remind readers that
Now, because of Lemma 2.4, following (54), we factor \(r_a\) over \({\mathbb {Q}}(\cos (\pi /7), a)\) as
where
Numerical evaluations exclude all possible orders of \(\alpha ,\beta ,\) and \(\gamma \), which do not meet the conditions in Lemma 3.7(i),(ii). Thus, we find that \((\alpha , \beta , \gamma )\) is the correct order of the roots, and by Entry 1.1(iv), the proof is complete.
From Theorem 7.1, similarly as we have seen in Theorem 5.1, the value of \(G_{2401}\) can be determined. As from Theorem 6.2 in Theorem 7.1, by using Theorems 6.3, 6.4, and 6.5, analogous results can be obtained for \(\varphi (e^{-49\pi \sqrt{3}}), \varphi (e^{-147\pi }),\) and \(\varphi (e^{-245\pi })\), respectively. These values can be expressed by using the solutions of the corresponding cubic polynomials, but their structure seems much more complicated.
Based on a remark at the end of Section 12 of Chapter 20 of Ramanujan’s second notebook [24, p. 247], [4, p. 400], we believe that the septic identity in Entry 1.1 is a special case of a much more general result. We will continue our investigation in this direction with the description of the analogous cubic and quintic identities.
References
Abel, N.H.: Recherches sur les fonctions elliptiques. J. Reine Angew. Math. 3, 160–190 (1828)
Andrews, G.E., Berndt, B.C.: Ramanujan’s Lost Notebook. Part II. Springer, New York (2009)
Bankoff, L., Garfunkel, J.: The heptagonal triangle. Math. Magn. 46, 7–19 (1973)
Berndt, B.C.: Ramanujan’s Notebooks. Part III. Springer, New York (1991)
Berndt, B.C.: Ramanujan’s theory of theta-function. In: Ram Murty, M. (ed.) Theta Functions, From the Classical to the Modern, Centre de Recherches Mathématiques Proceedings and Lecture Notes, vol. 1, pp. 1–63. American Mathematical Society, Providence (1993)
Berndt, B.C.: Ramanujan’s Notebooks. Part IV. Springer-Verlag, New York (1994)
Berndt, B.C.: Ramanujan’s Notebooks. Part V. Springer, New York (1998)
Berndt, B.C.: The remaining 40% of Ramanujan’s Lost Notebook. In Number Theory and its Applications, Surikaisekikenkyuusho Kokyuuroku, No. 1060, pp. 111–118, RIMS Kyoto University, Kyoto, (1998)
Berndt, B.C., Chan, H.H.: Ramanujan’s explicit values for the classical theta function. Mathematika 42(2), 278–294 (1995)
Berndt, B.C., Chan, H.H., Zhang, L.-C.: Ramanujan’s class invariants and cubic continued fraction. Acta Arith. 73(1), 67–85 (1995)
Berndt, B.C., Chan, H.H., Zhang, L.-C.: Ramanujan’s class invariants, Kronecker’s limit formula, and modular equations. Trans. Am. Math. Soc. 349(6), 2125–2173 (1997)
Berndt, B.C., Chan, H.H., Zhang, L.C.: Ramanujan’s class invariants with applications to the values of \(q\)-continued fractions and theta functions. In: Ismail, M.E.H., Masson, D.R., Rahman, M. (eds.) Special Functions, \(q\)-Series and Related Topics, Fields Institute Communications Series, vol. 14, pp. 37–53. American Mathematical Society, Providence (1997)
Borwein, J.M., Borwein, P.B.: Pi and the AGM. Wiley, New York (1987)
Borwein, J.M., Zucker, I.J.: Fast evaluation of the gamma function for small rational fractions using complete elliptic integrals of the first kind. IMA J. Numer. Anal. 12(4), 519–526 (1992)
Cox, D.A.: Primes of the Form \(x^2 + ny^2\). Wiley, New York (1989)
Gradshteyn, I.S., Ryzhik, I.M.: Table of Integrals, Series, and Products, 7th edn. Academic Press, New York (2007)
Jacobi, C.G.J.: Fundamenta nova theoriae functionum ellipticarum. Sumptibus fratrum Borntræger, Regiomonti (1829)
Joubert, P.: Sur la théorie des fonctions elliptiques et son application à la théorie des nombres. Comptes rendus 50, 907–912 (1860)
Joyce, G.S., Zucker, I.J.: Special values of the hypergeometric series. Math. Proc. Camb. Philos. Soc. 109(2), 257–261 (1991)
Mason, J.C., Handscomb, D.C.: Chebyshev Polynomials. CRC Press, Boca Raton (2003)
Ramanujan, S.: Modular equations and approximations to \(\pi \). Quart. J. Math. 45, 350–372 (1914)
Ramanujan, S.: Collected Papers of Srinivasa Ramanujan. Cambridge Univ. Press, Cambridge (1927)
Ramanujan, S.: Notebooks of Srinivasa Ramanujan, vol. I. Tata Institute of Fundamental Research, Bombay (1957)
Ramanujan, S.: Notebooks of Srinivasa Ramanujan, vol. II. Tata Institute of Fundamental Research, Bombay (1957)
Ramanujan, S.: The Lost Notebook and Other Unpublished Papers. Narosa, New Delhi (1988)
Selberg, A., Chowla, S.: On Epstein’s zeta function. J. Reine Angew. Math. 227, 86–110 (1967)
Son, S.H.: Septic theta function identities in Ramanujan’s Lost Notebook. Acta Arith. 98(4), 361–374 (2001)
Son, S.H.: Ramanujan’s symmetric theta functions in his lost notebook. In: D. Dominici, D., Maier, R.S. (eds.) Special Functions and Orthogonal Polynomials, American Mathematical Society, Providence, RI. Contemp. Math. 471, 187–202. (2008)
Watson, G.N.: Some singular moduli (II). Quart. J. Math. 3(1), 189–212 (1932)
Watson, G.N.: Singular moduli (4). Acta Arith. 1(2), 284–323 (1935)
Watson, G.N.: Singular moduli (3). Proc. Lond. Math. Soc. 40(1), 83–142 (1936)
Watson, G.N.: Singular moduli (5). Proc. Lond. Math. Soc. 42(1), 377–397 (1937)
Weber, H.: Lehrbuch der Algebra, Dritter Band, 2nd edn. Druck und Verlag von Friedrich Vieweg und Sohn, Braunschweig (1908)
Whittaker, E.T., Watson, G.N.: A Course of Modern Analysis, 4th edn. Cambridge University Press, Cambridge (1950)
Yi., J.: The Construction and Applications of Modular Equations. PhD thesis, University of Illinois at Urbana–Champaign, Urbana, Illinois (2001)
Zhang, L.-C.: Ramanujan’s class invariants, Kronecker’s limit formula and modular equations (III). Acta Arith. 82(4), 379–392 (1997)
Zucker, I.J.: The evaluation in terms of \(\Gamma \)-functions of the periods of elliptic curves admitting complex multiplication. Math. Proc. Camb. Philos. Soc. 82(1), 111–118 (1977)
Zucker, I.J., Joyce, G.S.: Special values of the hypergeometric series II. Math. Proc. Camb. Philos. Soc. 131(2), 309–319 (2001)
Acknowledgements
I started to work on this paper as an independent scholar in Győr, Hungary, and finalized it at University of Tromsø – The Arctic University of Norway, in Tromsø, Norway. The comments of Professors Bruce C. Berndt and Trygve Johnsen are highly appreciated.
Funding
Open Access funding provided by UiT The Arctic University of Norway
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Rebák, Ö. The three missing terms in Ramanujan’s septic theta function identity. Ramanujan J 60, 885–911 (2023). https://doi.org/10.1007/s11139-022-00584-2
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11139-022-00584-2