5.5 Depth one
5.5.1 Explicit formulae
It is convenient to introduce some special notation: we will write ๐ด๐,๐(๐) โถ=๐ด(๐,0,โฆ,0
โโโ
๐โ1
;๐) =
โซ
1 0
(2๐๐๐ก)๐โ1
(๐โ 1)! ๐๐(๐ก, ๐)๐๐ก, and
๎ญ(๐, ๐;๐) โถ=โ
๐โฅ0 ๐โฅ1
๐ด๐,๐(๐)
(2๐๐)๐โ1๐๐โ1๐๐โ1 = โ
๐โฅ1โซ
1 0
(๐ก๐)๐โ1 (๐โ 1)!๐น
( ๐ก, ๐
2๐๐;๐ )
๐๐ก
= โซ
1 0
๐๐ก๐๐น (
๐ก, ๐ 2๐๐;๐
) ๐๐ก,
where we mention, once for all, that byโซ01we mean theregularized integraldescribed in the definition of elliptic MZVs.
First of all, note that one can reduce all depth one elliptic MZVs to the๐ด๐,๐โs:
๐ด(0,โฆ,0
โโโ
๐
, ๐,0,โฆ,0
โโโ
๐
;๐) =
= (2๐๐)๐+๐
โซ
1 0
๐ก๐(1 โ๐ก)๐
๐!๐ ! ๐๐(๐ก, ๐) =
โ๐ ๐=0
(2๐๐)๐ โ๐(โ1)๐ (๐ โ๐)!
(๐+๐ ๐
)
๐ด๐,๐+๐+1(๐). (5.87) Let us now see what is the analogue of Proposition5.3.1for the depth 1 case.
Lemma 5.5.1.
๐
๐๐๎ญ(๐, ๐;๐) = (1 โ๐๐)๎ณ(๐, ๐) โ๐ ๐
๐๐๎ญ(๐, ๐;๐) (5.88) Proof. Using the mixed heat equation, the substitution๐ฬ = ๐โ2๐๐and Lemma 5.3.1we get17
๐
๐๐ โซ
1 0
๐๐ก๐๐น (
๐ก, ๐ 2๐๐;๐
)
๐๐ก= 1 2๐๐โซ
1 0
๐๐ก๐ ๐
๐๐ก๐ ฬ๐
๐น(๐ก, ฬ๐;๐)๐๐ก
= 1 2๐๐
( ๐๐ ๐
๐ ฬ๐
๐น(1, ฬ๐;๐) โ ๐
๐ ฬ๐
๐น(0, ฬ๐;๐) )
โ ๐ 2๐๐โซ
1 0
๐๐ก๐ ๐
๐ ฬ๐
๐น(๐ก, ฬ๐;๐)๐๐ก
= (1 โ๐๐)๎ณ(๐, ๐) โ๐ ๐
๐๐๎ญ(๐, ๐;๐). (5.89)
โก
We want to make more precise in this setting the remark that elliptic MZVs can be written as iterated integrals of Eisenstein series. As always, we write๐ด๐,๐(๐) = ๐ดโ๐,๐+๐ด0๐,๐(๐), as well as๎ญ(๐, ๐;๐) =๎ญโ(๐, ๐) +๎ญ0(๐, ๐;๐)at the level of generating
17As in the proof of Enriquezโs Proposition5.3.1, we omit the proof that the basic properties of iterated integrals are still valid for the regularized ones. All details can be found in [44].
functions. Note that, by (5.37), we have that for all๐โ 1 ๐ดโ๐,๐= (2๐๐)๐B๐
๐!๐! , (5.90)
and that, by Corollary5.2.1and (5.22),
๐ดโ1,๐= (2๐๐)๐โ1( ๐๐ 2(๐โ 1)! โ
โ๐2โโ1
โ
๐=1
๐(2๐+ 1) (๐โ 2๐โ 1)!(2๐๐)2๐
)
. (5.91)
We will denote the generating series of the exponentially small part๐พ0๐ of the Eisen-stein series๐พ๐ by
๎ณ0(๐, ๐) โถ=โ
๐โฅ1
๐๐พ0๐+1๐๐โ1. (5.92)
Moreover, we denote18
ฮ๐ฟ๐,๐(๐) โถ= ฮ(0,โฆ,0
โโโ
๐โ1
, ๐;๐) =
โซ[๐,โ1โ]
๐พ๐๐พ0โฏ๐พ0
โโโโโ
๐โ1
and
ฮ๐ ๐,๐(๐) โถ= ฮ(๐,0,โฆ,0
โโโ
๐โ1
;๐) =
โซ[๐,โ1โ]
๐พ0โฏ๐พ0
โโโโโ
๐โ1
๐พ๐
Again, we write
ฮ๐ฟ๐,๐(๐) = ฮ๐ฟ,โ๐,๐ (๐) + ฮ๐ฟ,0๐,๐(๐), ฮ๐ ๐,๐(๐) = ฮ๐ ,โ๐,๐ (๐) + ฮ๐ ,0๐,๐(๐).
It is obvious that
ฮ๐ฟ,0๐,๐(๐) =
โซ[๐,๐โ]๐พ0๐๐พ0โฏ๐พ0
โโโโโ
๐โ1
(5.93)
and that
ฮ๐ฟ,โ๐,๐ (๐) = B๐(2๐๐๐)๐
๐!๐! . (5.94)
Moreover, one can easily check that ฮ๐ ,0๐,๐(๐) = (2๐๐)๐โ1
(๐โ 1)! โซ
๐โ
๐
๐ง๐โ1๐พ0๐(๐ง)๐๐ง (5.95) and that
ฮ๐ ,โ๐,๐ (๐) = ฮ๐ฟ,โ๐,๐ (๐) = B๐(2๐๐๐)๐
๐!๐! . (5.96)
It is important to keep in mind thatฮ๐ฟ๐,๐(๐)andฮ๐ ๐,๐(๐)vanish identically whenever๐ is odd. Moreover, we want to stress the fact that one can compute the๐-expansion of anyฮ0(๐1,โฆ, ๐๐;๐), and this is the simplest way to get the๐-expansion of A-elliptic MZVs, as already noticed in [16]. In depth one the๐-expansion (5.85) just reads
ฮ๐ฟ,0๐,๐(๐) = โ 2 (๐โ 1)!
โ
๐,๐โฅ1
๐๐โ๐โ1
๐๐ ๐๐๐. (5.97)
18As we have already done in the previous section, here byโซ ๐พ๐we meanโซ ๐พ๐(๐ง)๐๐ง.
Let us now introduce the generating functions
๎ฑ(๐, ๐;๐) โถ= โ
๐,๐โฅ1
(โ1)๐โ1 (2๐๐)๐โ1
(๐+๐โ 1)!
(๐โ 1)! ฮ๐ฟ๐+๐,๐(๐)๐๐โ1๐๐โ1, (5.98)
๎ฒ(๐, ๐;๐) โถ= โ
๐,๐โฅ1
1 (2๐๐)๐โ1
(๐+๐โ 1)!
(๐โ 1)! ฮ๐ ๐+๐,๐(๐)๐๐โ1๐๐โ1. (5.99) As before, we write
๎ฑ0(๐, ๐;๐) = โ
๐,๐โฅ1
(โ1)๐โ1 (2๐๐)๐โ1
(๐+๐โ 1)!
(๐โ 1)! ฮ๐ฟ,0๐+๐,๐(๐)๐๐โ1๐๐โ1 and
๎ฒ0(๐, ๐;๐) = โ
๐,๐โฅ1
1 (2๐๐)๐โ1
(๐+๐โ 1)!
(๐โ 1)! ฮ๐ ,0๐+๐,๐(๐)๐๐โ1๐๐โ1.
The main consequence of the lemma above (this consequence, stated in a different way, is already contained in [66]) is the following:
Proposition 5.5.1.
๎ญ0(๐, ๐;๐) = (๐๐ โ 1)๎ฑ0(๐, ๐;๐) (5.100) Proof.Iterating the statement of the proposition, and noting that, for every๐โฅ1,
โซ[๐,โ1โ]
๐ดโ๐,๐๐๐ง1๐๐ง2โฏ๐๐ง๐ =๐(๐), we deduce that
๎ญ0(๐, ๐;๐) = (๐๐ โ 1)โ
๐โฅ1
๐๐โ1 (2๐๐)๐โ1
๐๐โ1
๐๐๐โ1โซ[๐,๐โ]๎ณ0(๐, ๐ง1)๐๐ง1๐พ0๐๐ง2โฏ๐พ0๐๐ง๐. (5.101) Expanding the generating series๎ณ0, we get
๎ญ0(๐, ๐;๐) = (๐๐ โ 1) โ
๐,๐โฅ1
(โ1)๐โ1๐!
(๐โ๐)!(2๐๐)๐โ1ฮ๐ฟ,0๐+1,๐(๐)๐๐โ๐๐๐โ1
= (๐๐ โ 1) โ
๐,๐โฅ1
(โ1)๐โ1(๐+๐โ 1)!
(๐โ 1)!(2๐๐)๐โ1 ฮ๐ฟ,0๐+๐,๐(๐)๐๐โ1๐๐โ1,
where to get(โ1)๐โ1 in the second equality we have used that๐+๐must be even.
This concludes the proof.
โก
Corollary 5.5.1. For any๐โฅ1and any๐โฅ1we have ๐ด๐,๐(๐) =๐ดโ๐,๐+(โ1)๐โ1
(๐โ 1)!
โ๐โ1 ๐=1
(2๐๐)๐โ๐(๐+๐โ 1)!
(๐โ๐)! ฮ๐ฟ,0๐+๐,๐(๐). (5.102) Proof. It is a straightforward comparison term by term of the left hand side of the proposition with the right hand side.
โก
Remark 5.5.2. This corollary, together with (5.97), the formulae for the constant term ๐ดโ๐,๐ and formula (5.87), lead to completely explicit formulae for the ๐-expansion of
all A-elliptic MZVs of depth one, which allow to approximate them numerically to high precision19. We have used this to check numerically all results given in the rest of this section.
Remark 5.5.3. By formula (5.102) we deduce that all A-elliptic MZVs of the form ๐ด1,๐(๐)includeฮ(2, ๐) = โ2 log(๐(๐))as part of their Fourier expansion, and therefore all๐ต1,๐(๐) โถ= ๐ต(1,0,โฆ,0
โโโ
๐โ1
;๐) will include log(๐) in the asymptotic expansion. Note that for the modified
๐ดฬ1,๐(๐) โถ=๐ด1,๐(๐) โ(2๐๐)๐โ2
(๐โ 1)!๐ด1,2(๐) (5.103) the termฮ(2, ๐)disappears.
Another consequence of the proposition above is that we can easily invert the rรดle of A-elliptic MZVs and iterated integrals of Eisenstein series:
Corollary 5.5.2. For any๐โฅ1and any1โค๐โค๐โ 1we have ฮ๐ฟ,0๐,๐(๐) = (โ1)๐โ๐โ1(๐โ๐โ 1)!
(๐โ 1)!
๐+1โ
๐=2
B๐+1โ๐(2๐๐)๐โ๐โ๐
(๐+ 1 โ๐)! ๐ด0๐โ๐,๐(๐). (5.104) Proof. Note that๐๐ โ 1 = ๐ +๐2โ2 + โฆ, and this corresponds to the fact that length one elliptic MZVs are constant. This means that, if we want to invert equation (5.100), we need to rescale both sides, i.e. to multiply both sides by๐. This yields
๐
๐๐ โ 1๎ญ0(๐, ๐;๐) =๎ฑ0(๐, ๐;๐), (5.105) and comparing term by term this equation we are done.
โก
Now we want to deduce explicit formulae for B-elliptic MZVs of depth one. To do this, we need to know the behaviour ofฮ๐ฟ๐,๐(๐)under๐ โฆโ1โ๐. Let us introduce another piece of notation. For any operator ฮ acting on a space of formal power series, we setฮ(๐) โถ= ฮโฆ โฏ โฆฮ
โโโโโ
๐
, and exp(ฮ) โถ= โ
๐โฅ0ฮ(๐)โ๐!. Then we have the fol-lowing two lemmas.
Lemma 5.5.4.
๎ฒ0(๐, ๐;๐) =๎ฑ0(๐+๐๐ , ๐;๐) (5.106) Proof.By equation (5.95), we can write
๎ฒ0(๐, ๐;๐) =โ
๐โฅ1
๐๐โ1 (๐โ 1)!
๐๐โ1
๐๐๐โ1โซ
๐โ
๐
๐ง๐โ1๎ณ0(๐, ๐ง)๐๐ง (5.107)
19One can easily get500digits in less than a second with PARI GP.
Since๐โซ[๐,๐โ]๐พ0๐(๐ง)๐๐ง= โ๐พ0๐(๐), repeatedly integrating by parts gives
โ
๐โฅ1
๐๐โ1 (๐โ 1)!
๐๐โ1
๐๐๐โ1 โซ
๐โ
๐
๐ง๐โ1๎ณ0(๐, ๐ง)๐๐ง=
= exp( ๐๐ ๐
๐๐ ) โ
๐โฅ1
๐๐โ1 (2๐๐)๐โ1
๐๐โ1
๐๐๐โ1 โซ[๐,๐โ]๎ณ0(๐, ๐ง1)๐๐ง1๐พ0๐๐ง2โฏ๐พ0๐๐ง๐, (5.108) and this, as we have already seen in the proof of Proposition5.5.1, leads to the iden-tity
๎ฒ0(๐, ๐;๐) = exp( ๐๐ ๐
๐๐ )
๎ฑ0(๐, ๐;๐).
By Taylorโs theoremexp( ๐d
dx
)๐(๐ฅ) =๐(๐ฅ+๐). This concludes the proof.
โก
By an abuse of notation, we denote by ๎ฒ0(๐, ๐; โ1โ๐) the exponentially small (with respect to๐ โ ๐โ) part of the function๎ฒ0(๐, ๐; โ1โ๐).20 This will not create any confusion in what follows.
Lemma 5.5.5.
๎ฒ0(๐, ๐; โ1โ๐) =๎ฒ0(๐ ,โ๐;๐) (5.109) Proof.We have already seen that we can write
๎ฒ0(๐, ๐;๐) =โ
๐โฅ1
๐๐โ1 (๐โ 1)!
๐๐โ1
๐๐๐โ1โซ
๐โ
๐
๐ง๐โ1๎ณ0(๐, ๐ง)๐๐ง.
Therefore, using the transformation properties of๎ณ, we find that the exponentially small part of๎ฒ0(๐, ๐; โ1โ๐)is equal to
โ
๐โฅ1
(โ1)๐โ1๐๐โ1 (๐โ 1)!
๐๐โ1
๐๐๐โ1โซ
๐โ
๐
๐ง1โ๐๎ณ0(๐ง๐, ๐ง)๐๐ง.
Expanding this expression with respect to๐, and using that๐+๐must be even, we
find โ
๐,๐โฅ1
(โ1)๐โ1(๐+๐โ 1)!
(๐โ 1)!(๐โ 1)! โซ
๐โ
๐
๐ง๐โ1๐พ0๐+๐(๐ง)๐๐ง ๐๐โ1๐๐โ1. Comparing this expression with the definition of๎ฒconcludes the proof.
โก
Comparing the coefficients of the identity in Lemma5.5.4gives Corollary 5.5.3. For each๐, ๐ โฅ1we have
ฮ๐ ,0๐+๐,๐(๐) =
โ๐โ1 ๐=0
(โ1)๐+๐โ1(2๐๐๐)๐
๐! ฮ๐ฟ,0๐+๐,๐โ๐(๐), (5.110)
ฮ๐ฟ,0๐+๐,๐(๐) =
โ๐โ1 ๐=0
(โ1)๐+๐โ1(2๐๐๐)๐
๐! ฮ๐ ,0๐+๐,๐โ๐(๐). (5.111)
Moreover, Lemma5.5.5implies
20In general the latter yields some non-exponentially small term, as we will see later.
Corollary 5.5.4. For๐, ๐ โฅ1,
ฮ๐ ,0๐+๐,๐(โ1โ๐) = (โ1)๐โ1(2๐๐)๐โ๐(๐โ 1)!
(๐โ 1)! ฮ๐ ,0๐+๐,๐(๐) (5.112) These two facts together lead to
Corollary 5.5.5. Let๐โฅ1and1โค๐ โค๐โ 1. Then ฮ๐ ,0๐,๐(โ1โ๐) = (2๐๐)๐โ1(๐โ 1 โ๐)!
(๐โ 1)!
โ๐โ๐ ๐=1
(โ1)๐+๐๐๐โ๐โ๐
(๐โ๐โ๐)!(2๐๐)๐โ1ฮ๐ฟ,0๐,๐ (๐) (5.113) and
ฮ๐ฟ,0๐,๐(โ1โ๐) = (2๐๐)๐โ1
โ๐โ1 ๐=0
(๐โ๐+๐โ 1)!
๐!(๐โ๐โ 1)!
๐โ๐+๐โ
๐=1
(โ1)๐โ1๐๐โ๐โ๐
(๐โ๐+๐โ๐)!(2๐๐)๐โ1ฮ๐ฟ,0๐,๐(๐) (5.114) We are now ready to put all the pieces together. Let us consider
๐ต๐,๐(๐) =๐ต(๐,0,โฆ,0
โโโ
๐โ1
;๐).
We write
๐ต๐,๐(๐) =๐ต๐,๐โ(๐) +๐ต0๐,๐(๐),
meaning as always that๐ต๐,๐โ(๐)denotes the first Laurent polynomials in the asymp-totic expansion of๐ต๐,๐(๐)(plus a logarithmic term, in case๐= 1, cf. Remark5.5.3) and ๐ต๐,๐0 (๐)denotes the exponentially small part. Let us consider the generating function
๎ฎ0(๐, ๐;๐) โถ= โ
๐,๐โฅ1
๐ต0๐,๐(๐)
(2๐๐)๐โ1๐๐โ1๐๐โ1. Then we have
Theorem 5.5.6.
๎ฎ0(๐, ๐;๐) = (๐๐ โ 1)๎ฑ0(
โ๐๐,โ๐โ ๐ ๐;๐
)
. (5.115)
Proof.By Proposition5.5.1and Lemma5.5.4we have
๎ญ0(๐, ๐; โ1โ๐) = (๐๐ โ 1)๎ฑ0(๐, ๐; โ1โ๐) = (๐๐ โ 1)๎ฒ0( ๐+ ๐
๐, ๐; โ1โ๐ )
, where in this case๎ฒ0(๐, ๐; โ1โ๐)really means that we consider๎ฒ0(๐, ๐;๐)and then we invert๐. Using Lemma5.5.5and again Lemma5.5.4we conclude that
๎ฎ0(๐, ๐;๐) = (๐๐ โ 1)๎ฒ0(
๐ ,โ๐โ๐ ๐;๐
)
= (๐๐ โ 1)๎ฑ0(
โ๐๐,โ๐โ๐ ๐;๐
) , where in this case๎ฒ0(๐, ๐; โ1โ๐) was meant to be the exponentially small part of
๎ฒ0(๐, ๐; โ1โ๐).
โก
The consequence of this theorem is that we are now able to express B-elliptic MZVs of depth one in terms of iterated integrals of Eisenstein series, and therefore in terms of A-elliptic MZVs:
Corollary 5.5.6. For๐, ๐โฅ1we have ๐ต๐,๐(๐) =๐ต๐,๐โ(๐)+
(2๐๐)๐โ1 (๐โ 1)!
โ๐โ1 ๐=1
(๐+๐โ 1)!
(๐โ๐)!
โ๐โ1 ๐=0
(โ1)๐โ๐โ1(๐+๐โ 1)!
๐!(๐โ๐โ 1)!
โ๐+๐ ๐=1
(โ1)๐โ1๐๐โ๐
(2๐๐)๐โ1(๐+๐โ๐)!ฮ๐ฟ,0๐+๐,๐(๐), (5.116) and thus
๐ต๐,๐(๐) =๐ต๐,๐โ(๐) + (โ1)๐โ1(2๐๐)๐โ1 (๐โ 1)!
โ๐โ1 ๐=1
1 (๐โ๐)!
โ๐โ1 ๐=0
(โ1)๐(๐+๐โ 1)!
๐!(๐โ๐โ 1)! ร
ร
โ๐+๐ ๐=1
(๐+๐โ๐โ 1)!๐๐โ๐ (๐+๐โ๐)!
๐+1โ
๐=2
B๐+1โ๐
(๐+ 1 โ๐)!(2๐๐)๐โ1๐ด0๐+๐โ๐,๐(๐). (5.117) Proof. One can prove the first statement either by comparing term by term the equation of Theorem5.5.6, or by using the corollaries of Proposition5.5.1, Lemma5.5.4 and Lemma5.5.5. The second statement follows from the first plus Corollary5.5.2.
โก
In particular, one can see that the range of powers of ๐ agrees with the range predicted by Theorem5.3.3. It is useful to write down the formulae given in the corollary, together with the explicit formula for the constant term, for the simplest case of length 2:
๐ต๐,2(๐) = ๐ต๐,2โ(๐) +
โ๐ ๐=1
(โ1)๐โ1๐!๐๐โ๐
(๐โ๐)!(2๐๐)๐โ2ฮ๐ฟ,0๐+1,๐(๐)
= ๐ต๐,2โ(๐) + (โ1)๐โ1๐!
โ๐ ๐=1
๐๐โ๐
โ๐+1 ๐=2
B๐+1โ๐
(๐+ 1 โ๐)!(2๐๐)๐โ2๐ด0๐+1โ๐,๐(๐) The first non-trivial examples of B-elliptic MZVs (recall that in order to have interesting examples we need๐+๐odd, and we prefer to avoid B-elliptic MZVs with log(๐)in the expansion) given by our construction are
๐ต3,2(๐) = โ(2๐๐)2
720 ๐3โ ๐(3)
2๐๐ โ 6๐(4) (2๐๐)2
1
๐ + 6๐๐๐2ฮ๐ฟ,04,1(๐) โ 6๐ฮ๐ฟ,04,2(๐) + 3
๐๐ฮ๐ฟ,04,3(๐), ๐ต2,3(๐) = (2๐๐)3
720 ๐2โ๐๐
2 ๐(2) โ 2๐(3)
๐ โ3๐(4) ๐๐
1
๐2 โ 48๐2ฮ๐ฟ,04,1(๐) โ 36๐๐ฮ๐ฟ,04,2(๐) +12
๐ ฮ๐ฟ,04,3(๐),
๐ต5,2(๐) = (2๐๐)2
30240๐5โ ๐(5)
(2๐๐)3 โ 10๐(6)
(2๐๐)4๐ + 10๐๐๐4ฮ๐ฟ,06,1(๐)
โ 20๐3ฮ๐ฟ,06,2(๐) + 30
๐๐๐2ฮ๐ฟ,06,3(๐) โ 120๐
(2๐๐)2ฮ๐ฟ,06,4(๐) + 120
(2๐๐)3ฮ๐ฟ,06,5(๐),
๐ต4,3(๐) = โ(2๐๐)3 30240๐4โ1
2 ๐(4)
2๐๐ โ 4๐(5)
(2๐๐)2๐ โ 20๐(6) (2๐๐)3๐2
โ 40๐๐๐2ฮ๐ฟ,06,2(๐) + 120๐ฮ๐ฟ,06,3(๐) โ 180
๐๐ ฮ๐ฟ,06,4(๐) + 480
(2๐๐)2๐ฮ๐ฟ,06,5(๐),
๐ต3,4(๐) = โ(2๐๐)4
5040 ๐3โ๐๐ ๐(3)
6 โ 3๐(4)
2๐ โ 3๐(5)
๐๐๐2 โ 20๐(6)
(2๐๐)2๐3 +(2๐๐)3๐2 2 ฮ๐ฟ,04,1(๐)
โ (2๐๐)2๐ฮ๐ฟ,04,2(๐) + 2๐๐ฮ๐ฟ,04,3(๐) + 120๐๐ฮ๐ฟ,06,3(๐) โ 360
๐ ฮ๐ฟ,06,4(๐) + 360
๐๐๐2ฮ๐ฟ,06,5(๐),
๐ต2,5(๐) = (2๐๐)5
5040 ๐2โ (2๐๐)3๐(2)
24 โ (2๐๐)2๐(3)
3๐ โ3๐๐ ๐(4)
๐2 โ4๐(5)
๐3 โ 5๐(6) ๐๐๐4
โ (2๐๐)3ฮ๐ฟ,04,2(๐) + 2(2๐๐)2
๐ ฮ๐ฟ,04,3(๐) โ 240๐๐
๐2 ฮ๐ฟ,06,4(๐) + 480
๐3 ฮ๐ฟ,06,5(๐).