Eisenstein series#Identities involving Eisenstein series

{{about|holomorphic Eisenstein series in dimension 3|the non-holomorphic case|Real analytic Eisenstein series|the higher dimensional case|Siegel Eisenstein series}}

Eisenstein series, named after German mathematician Gotthold Eisenstein,{{Cite web |title=Gotthold Eisenstein - Biography |url=https://mathshistory.st-andrews.ac.uk/Biographies/Eisenstein/ |access-date=2023-09-05 |website=Maths History |language=en}} are particular modular forms with infinite series expansions that may be written down directly. Originally defined for the modular group, Eisenstein series can be generalized in the theory of automorphic forms.

Eisenstein series for the modular group

Image:Gee_three_real.jpeg. Negative numbers are black.]]

Image:Gee_three_imag.jpeg

Let {{mvar|τ}} be a complex number with strictly positive imaginary part. Define the holomorphic Eisenstein series {{math|G_2k(τ)}} of weight {{math|2k}}, where {{math|k ≥ 2}} is an integer, by the following series:{{Cite journal |last=Gekeler |first=Ernst-Ulrich |date=2011 |title=PARA-EISENSTEIN SERIES FOR THE MODULAR GROUP GL(2, 𝔽q[T]) |url=https://www.jstor.org/stable/taiwjmath.15.4.1463 |journal=Taiwanese Journal of Mathematics |volume=15 |issue=4 |pages=1463–1475 |doi=10.11650/twjm/1500406358 |s2cid=119499748 |issn=1027-5487|doi-access=free }}

: $G_{2k}(\tau) = \sum_{ (m,n)\in\Z^2\setminus\{(0,0)\}} \frac{1}{(m+n\tau )^{2k}}.$

This series absolutely converges to a holomorphic function of {{mvar|τ}} in the upper half-plane and its Fourier expansion given below shows that it extends to a holomorphic function at {{math|τ {{=}} i∞}}. It is a remarkable fact that the Eisenstein series is a modular form. Indeed, the key property is its {{math|SL(2, $\mathbb{Z}$ )}}-covariance. Explicitly if {{math|a, b, c, d ∈ $\mathbb{Z}$ }} and {{math|ad − bc {{=}} 1}} then

: $G_{2k} \left( \frac{ a\tau +b}{ c\tau + d} \right) = (c\tau +d)^{2k} G_{2k}(\tau)$

{{Hidden

| style = border:1px solid black; text-align:left;width: 70%;

| multiline=

| headerstyle = text-align:left;

| header = (Proof)

| contentstyle = text-align:left;

| content =

: $\begin{align}
G_{2k}\left(\frac{a\tau+b}{c\tau+d}\right) &= \sum_{(m,n) \in \Z^2 \setminus \{(0,0)\}} \frac{1}{\left(m+n\frac{a\tau+b}{c\tau+d}\right)^{2k}} \\
&= \sum_{(m,n) \in \Z^2 \setminus \{(0,0)\}} \frac{(c\tau+d)^{2k}}{(md+nb+(mc+na)\tau)^{2k}} \\
&= \sum_{\left(m',n'\right) = (m,n)\begin{pmatrix}d \ \ c\\b \ \ a\end{pmatrix}\atop (m,n)\in \Z^2 \setminus \{(0,0)\}} \frac{(c\tau+d)^{2k}}{\left(m'+n'\tau\right)^{2k}}
\end{align}$

If {{math|ad − bc {{=}} 1}} then

: $\begin{pmatrix}d & c\\b & a\end{pmatrix}^{-1} = \begin{pmatrix}\ a & -c\\-b & \ d\end{pmatrix}$

so that

: $(m,n) \mapsto (m,n)\begin{pmatrix}d & c\\b & a\end{pmatrix}$

is a bijection {{math| $\mathbb{Z}$ ² → $\mathbb{Z}$ ²}}, i.e.:

: $\sum_{\left(m',n'\right) = (m,n)\begin{pmatrix}d \ \ c\\b \ \ a\end{pmatrix}\atop (m,n)\in \Z^2 \setminus \{(0,0)\}} \frac{1}{\left(m'+n'\tau\right)^{2k}} = \sum_{\left(m',n'\right)\in \mathbb{Z}^2 \setminus \{(0,0)\}} \frac{1}{(m'+n'\tau)^{2k}} = G_{2k}(\tau)$

Overall, if {{math|ad − bc {{=}} 1}} then

: $G_{2k}\left(\frac{a\tau+b}{c\tau+d}\right) = (c\tau+d)^{2k} G_{2k}(\tau)$

and {{math|G_2k}} is therefore a modular form of weight {{math|2k}}. Note that it is important to assume that {{math|k ≥ 2}}, otherwise it would be illegitimate to change the order of summation, and the {{math|SL(2, $\mathbb{Z}$ )}}-invariance would not hold. In fact, there are no nontrivial modular forms of weight 2. Nevertheless, an analogue of the holomorphic Eisenstein series can be defined even for {{math|k {{=}} 1}}, although it would only be a quasimodular form.

}}

Note that {{math|k ≥ 2}} is necessary such that the series converges absolutely, whereas {{math|k}} needs to be even otherwise the sum vanishes because the {{math|(-m, -n)}} and {{math|(m, n)}} terms cancel out. For {{math|k {{=}} 1}} the series converges but it is not a modular form.

Relation to modular invariants

The modular invariants {{math|g₂}} and {{math|g₃}} of an elliptic curve are given by the first two Eisenstein series:{{Cite journal |last1=Obers |first1=N. A. |last2=Pioline |first2=B. |date=2000-03-07 |title=Eisenstein Series in String Theory |journal=Classical and Quantum Gravity |volume=17 |issue=5 |pages=1215–1224 |doi=10.1088/0264-9381/17/5/330 |issn=0264-9381|arxiv=hep-th/9910115 |bibcode=2000CQGra..17.1215O |s2cid=250864942 }}

: $\begin{align} g_2 &= 60 G_4 \\ g_3 &= 140 G_6 .\end{align}$

The article on modular invariants provides expressions for these two functions in terms of theta functions.

Recurrence relation

Any holomorphic modular form for the modular group{{Cite journal |last1=Mertens |first1=Michael H. |last2=Rolen |first2=Larry |date= 2015|title=Lacunary recurrences for Eisenstein series |journal= Research in Number Theory|volume= 1|issue= |pages= |doi=10.1007/s40993-015-0010-x |arxiv=1504.00356 |issn=2363-9555|doi-access=free }} can be written as a polynomial in {{math|G₄}} and {{math|G₆}}. Specifically, the higher order {{math|G_2k}} can be written in terms of {{math|G₄}} and {{math|G₆}} through a recurrence relation. Let {{math|d_k {{=}} (2k + 3)k! G_{2k + 4}}}, so for example, {{math|d₀ {{=}} 3G₄}} and {{math|d₁ {{=}} 5G₆}}. Then the {{mvar|d_k}} satisfy the relation

: $\sum_{k=0}^n {n \choose k} d_k d_{n-k} = \frac{2n+9}{3n+6}d_{n+2}$

for all {{math|n ≥ 0}}. Here, $n \choose k$ is the binomial coefficient.

The {{math|d_k}} occur in the series expansion for the Weierstrass's elliptic functions:

: $\begin{align}
\wp(z) &=\frac{1}{z^2} + z^2 \sum_{k=0}^\infty \frac {d_k z^{2k}}{k!} \\
&=\frac{1}{z^2} + \sum_{k=1}^\infty (2k+1) G_{2k+2} z^{2k}.
\end{align}$

Fourier series

Image:Eisenstein_4.jpg

Image:Eisenstein_6.jpg

Image:Eisenstein_8.jpg

Image:Eisenstein_10.jpg

Image:Eisenstein_12.jpg

Image:Eisenstein_14.jpg

Define {{math|q {{=}} e^2πiτ}}. (Some older books define {{mvar|q}} to be the nome {{math|q {{=}} e^{{pi}}iτ}}, but {{math|q {{=}} e^2{{pi}}iτ}} is now standard in number theory.) Then the Fourier series of the Eisenstein{{Cite journal |last=Karel |first=Martin L. |date=1974 |title=Fourier Coefficients of Certain Eisenstein Series |url=https://www.jstor.org/stable/1971017 |journal=Annals of Mathematics |volume=99 |issue=1 |pages=176–202 |doi=10.2307/1971017 |jstor=1971017 |issn=0003-486X}} series is

: $G_{2k}(\tau) = 2\zeta(2k) \left(1+c_{2k}\sum_{n=1}^\infty \sigma_{2k-1}(n)q^n \right)$

where the coefficients {{math|c_2k}} are given by

: $\begin{align}
c_{2k} &= \frac{(2\pi i)^{2k}}{(2k-1)! \zeta(2k)} \\[4pt]
&= \frac {-4k}{B_{2k}} = \frac 2 {\zeta(1-2k)}.
\end{align}$

Here, {{math|B_n}} are the Bernoulli numbers, {{math|ζ(z)}} is Riemann's zeta function and {{math|σ_p(n)}} is the divisor sum function, the sum of the {{mvar|p}}th powers of the divisors of {{mvar|n}}. In particular, one has

: $\begin{align}
G_4(\tau)&=\frac{\pi^4}{45} \left( 1+ 240\sum_{n=1}^\infty \sigma_3(n) q^{n} \right) \\[4pt]
G_6(\tau)&=\frac{2\pi^6}{945} \left( 1- 504\sum_{n=1}^\infty \sigma_5(n) q^n \right).
\end{align}$

The summation over {{mvar|q}} can be resummed as a Lambert series; that is, one has

: $\sum_{n=1}^{\infty} q^n \sigma_a(n) = \sum_{n=1}^{\infty} \frac{n^a q^n}{1-q^n}$

for arbitrary complex {{math|{{abs|q}} < 1}} and {{mvar|a}}. When working with the q-expansion of the Eisenstein series, this alternate notation is frequently introduced:

: $\begin{align}
E_{2k}(\tau)&=\frac{G_{2k}(\tau)}{2\zeta (2k)}\\
&= 1+\frac {2}{\zeta(1-2k)}\sum_{n=1}^{\infty} \frac{n^{2k-1} q^n}{1-q^n} \\
&= 1- \frac{4k}{B_{2k}}\sum_{n=1}^{\infty} \sigma_{2k-1}(n)q^n \\
&= 1 - \frac{4k}{B_{2k}} \sum_{d,n \geq 1} n^{2k-1} q^{nd}. \end{align}$

Identities involving Eisenstein series

= As theta functions =

Source:{{Cite web |title=How to prove this series identity involving Eisenstein series? |url=https://math.stackexchange.com/questions/4284880/how-to-prove-this-series-identity-involving-eisenstein-series |access-date=2023-09-05 |website=Mathematics Stack Exchange |language=en}}

Given {{math|q {{=}} e^2{{pi}}iτ}}, let

: $\begin{align}
E_4(\tau)&=1+240\sum_{n=1}^\infty \frac {n^3q^n}{1-q^n} \\
E_6(\tau)&=1-504\sum_{n=1}^\infty \frac {n^5q^n}{1-q^n} \\
E_8(\tau)&=1+480\sum_{n=1}^\infty \frac {n^7q^n}{1-q^n}
\end{align}$

and define the Jacobi theta functions which normally uses the nome {{math|e^{{pi}}iτ}},

: $\begin{align}
a&=\theta_2\left(0; e^{\pi i\tau}\right)=\vartheta_{10}(0; \tau) \\
b&=\theta_3\left(0; e^{\pi i\tau}\right)=\vartheta_{00}(0; \tau) \\
c&=\theta_4\left(0; e^{\pi i\tau}\right)=\vartheta_{01}(0; \tau)
\end{align}$

where {{math|θ_m}} and {{math|ϑ_ij}} are alternative notations. Then we have the symmetric relations,

: $\begin{align}
E_4(\tau)&= \tfrac{1}{2}\left(a^8+b^8+c^8\right) \\[4pt]
E_6(\tau)&= \tfrac{1}{2}\sqrt{\frac{\left(a^8+b^8+c^8\right)^3-54(abc)^8}{2}} \\[4pt]
E_8(\tau)&= \tfrac{1}{2}\left(a^{16}+b^{16}+c^{16}\right) = a^8b^8 +a^8c^8 +b^8c^8
\end{align}$

Basic algebra immediately implies

: $E_4^3-E_6^2 = \tfrac{27}{4}(abc)^8$

an expression related to the modular discriminant,

: $\Delta = g_2^3-27g_3^2 = (2\pi)^{12} \left(\tfrac{1}{2}a b c\right)^8$

The third symmetric relation, on the other hand, is a consequence of {{math|E₈ {{=}} E{{su|b=4|p=2}}}} and {{math|a⁴ − b⁴ + c⁴ {{=}} 0}}.

= Products of Eisenstein series =

Eisenstein series form the most explicit examples of modular forms for the full modular group {{math|SL(2, $\mathbb{Z}$ )}}. Since the space of modular forms of weight {{math|2k}} has dimension 1 for {{math|2k {{=}} 4, 6, 8, 10, 14}}, different products of Eisenstein series having those weights have to be equal up to a scalar multiple. In fact, we obtain the identities:{{Cite journal |last1=Dickson |first1=Martin |last2=Neururer |first2=Michael |date= 2018|title=Products of Eisenstein series and Fourier expansions of modular forms at cusps |journal=Journal of Number Theory |volume=188 |pages=137–164 |doi=10.1016/j.jnt.2017.12.013|arxiv=1603.00774 |s2cid=119614418 }}

: $E_4^2 = E_8, \quad E_4 E_6 = E_{10}, \quad E_4 E_{10} = E_{14}, \quad E_6 E_8 = E_{14}.$

Using the {{mvar|q}}-expansions of the Eisenstein series given above, they may be restated as identities involving the sums of powers of divisors:

: $\left(1+240\sum_{n=1}^\infty \sigma_3(n) q^n\right)^2 = 1+480\sum_{n=1}^\infty \sigma_7(n) q^n,$

hence

: $\sigma_7(n)=\sigma_3(n)+120\sum_{m=1}^{n-1}\sigma_3(m)\sigma_3(n-m),$

and similarly for the others. The theta function of an eight-dimensional even unimodular lattice {{math|Γ}} is a modular form of weight 4 for the full modular group, which gives the following identities:

: $\theta_\Gamma (\tau)=1+\sum_{n=1}^\infty r_{\Gamma}(2n) q^{n} = E_4(\tau), \qquad r_{\Gamma}(n) = 240\sigma_3(n)$

for the number {{math|r_Γ(n)}} of vectors of the squared length {{math|2n}} in the E8 lattice.

Similar techniques involving holomorphic Eisenstein series twisted by a Dirichlet character produce formulas for the number of representations of a positive integer {{mvar|n}}' as a sum of two, four, or eight squares in terms of the divisors of {{mvar|n}}.

Using the above recurrence relation, all higher {{math|E_2k}} can be expressed as polynomials in {{math|E₄}} and {{math|E₆}}. For example:

: $\begin{align}
E_{8} &= E_4^2 \\
E_{10} &= E_4\cdot E_6 \\
691 \cdot E_{12} &= 441\cdot E_4^3+ 250\cdot E_6^2 \\
E_{14} &= E_4^2\cdot E_6 \\
3617\cdot E_{16} &= 1617\cdot E_4^4+ 2000\cdot E_4 \cdot E_6^2 \\
43867 \cdot E_{18} &= 38367\cdot E_4^3\cdot E_6+5500\cdot E_6^3 \\
174611 \cdot E_{20} &= 53361\cdot E_4^5+ 121250\cdot E_4^2\cdot E_6^2 \\
77683 \cdot E_{22} &= 57183\cdot E_4^4\cdot E_6+20500\cdot E_4\cdot E_6^3 \\
236364091 \cdot E_{24} &= 49679091\cdot E_4^6+ 176400000\cdot E_4^3\cdot E_6^2 + 10285000\cdot E_6^4
\end{align}$

Many relationships between products of Eisenstein series can be written in an elegant way using Hankel determinants, e.g. Garvan's identity

: $\left(\frac{\Delta}{(2\pi)^{12}}\right)^2=-\frac{691}{1728^2\cdot250}\det \begin{vmatrix}E_4&E_6&E_8\\ E_6&E_8&E_{10}\\ E_8&E_{10}&E_{12}\end{vmatrix}$

where

: $\Delta=(2\pi)^{12}\frac{E_4^3-E_6^2}{1728}$

is the modular discriminant.{{cite arXiv|last=Milne|first=Steven C.|year=2000|eprint=math/0009130v3|title=Hankel Determinants of Eisenstein Series}} The paper uses a non-equivalent definition of $\Delta$ , but this has been accounted for in this article.

= Ramanujan identities =

Srinivasa Ramanujan gave several interesting identities between the first few Eisenstein series involving differentiation.{{Cite journal |last1=Bhuvan |first1=E. N. |last2=Vasuki |first2=K. R. |date=2019-06-24 |title=On a Ramanujan's Eisenstein series identity of level fifteen |url=https://doi.org/10.1007/s12044-019-0498-4 |journal=Proceedings - Mathematical Sciences |language=en |volume=129 |issue=4 |pages=57 |doi=10.1007/s12044-019-0498-4 |s2cid=255485301 |issn=0973-7685}} Let

: $\begin{align}
L(q)&=1-24\sum_{n=1}^\infty \frac {nq^n}{1-q^n}&&=E_2(\tau) \\
M(q)&=1+240\sum_{n=1}^\infty \frac {n^3q^n}{1-q^n}&&=E_4(\tau) \\
N(q)&=1-504\sum_{n=1}^\infty \frac {n^5q^n}{1-q^n}&&=E_6(\tau),
\end{align}$

then

: $\begin{align}
q\frac{dL}{dq} &= \frac {L^2-M}{12} \\
q\frac{dM}{dq} &= \frac {LM-N}{3} \\
q\frac{dN}{dq} &= \frac {LN-M^2}{2}.
\end{align}$

These identities, like the identities between the series, yield arithmetical convolution identities involving the sum-of-divisor function. Following Ramanujan, to put these identities in the simplest form it is necessary to extend the domain of {{math|σ_p(n)}} to include zero, by setting

: $\begin{align}\sigma_p(0) = \tfrac12\zeta(-p) \quad\Longrightarrow\quad
\sigma(0) &= -\tfrac{1}{24}\\
\sigma_3(0) &= \tfrac{1}{240}\\
\sigma_5(0) &= -\tfrac{1}{504}.
\end{align}$

Then, for example

: $\sum_{k=0}^n\sigma(k)\sigma(n-k)=\tfrac5{12}\sigma_3(n)-\tfrac12n\sigma(n).$

Other identities of this type, but not directly related to the preceding relations between {{mvar|L}}, {{mvar|M}} and {{mvar|N}} functions, have been proved by Ramanujan and Giuseppe Melfi,{{cite book|authorlink=Srinivasa Ramanujan|last=Ramanujan|first=Srinivasa|chapter=On certain arithmetical functions|pages=136–162|title=Collected Papers|date=1962|publisher=Chelsea|location=New York, NY}}{{cite book|authorlink=Giuseppe Melfi|first=Giuseppe|last=Melfi|contribution=On some modular identities|title=Number Theory, Diophantine, Computational and Algebraic Aspects: Proceedings of the International Conference held in Eger, Hungary|publisher=Walter de Grutyer & Co.|date=1998|pages=371–382}} as for example

: $\begin{align}
\sum_{k=0}^n\sigma_3(k)\sigma_3(n-k)&=\tfrac1{120}\sigma_7(n) \\
\sum_{k=0}^n\sigma(2k+1)\sigma_3(n-k)&=\tfrac1{240}\sigma_5(2n+1) \\
\sum_{k=0}^n\sigma(3k+1)\sigma(3n-3k+1)&=\tfrac19\sigma_3(3n+2).
\end{align}$

Generalizations

Automorphic forms generalize the idea of modular forms for general Lie groups; and Eisenstein series generalize in a similar fashion.

Defining {{math|O_K}} to be the ring of integers of a totally real algebraic number field {{mvar|K}}, one then defines the Hilbert–Blumenthal modular group as {{math|PSL(2,O_K)}}. One can then associate an Eisenstein series to every cusp of the Hilbert–Blumenthal modular group.