HAL Id: hal-00288656
https://hal.archives-ouvertes.fr/hal-00288656
Submitted on 18 Jun 2008
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of
sci-entific research documents, whether they are
pub-lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
symmetric power L-functions in the level aspect
Guillaume Ricotta, Emmanuel Royer
To cite this version:
Guillaume Ricotta, Emmanuel Royer. Lower order terms for the one-level densities of symmetric
power L-functions in the level aspect. Acta Arithmetica, Instytut Matematyczny PAN, 2010, 141 (2),
pp.153-170. �hal-00288656�
GUILLAUMERICOTTAANDEMMANUELROYER
Abstra t. In[12 ℄,theauthorsdetermined,amongotherthings,themaintermsfor
theone-leveldensities forlow-lyingzerosofsymmetri power
L
-fun tionsinthelevel aspe t. In this paper, the lower orderterms of these one-level densities are found. The ombinatorial di ulties,whi hshouldariseinsu h ontext,are drasti allyre-du edthanks to Chebyshevpolynomials,whi hare the hara tersof theirredu ible representations of
SU(2)
.Contents
1. Introdu tion and statementof theresult 1
1.1. Des ription of the familiesof
L
-fun tionsstudied 11.2. One-level densitiesof thesefamilies 3
2. Chebyshev polynomials and He ke eigenvalues 4
3. Riemann's expli itformulafor symmetri power
L
-fun tions 64. Proofof Theorem A 8
Appendix A. Some ommentson an aestheti identity 10
Appendix B. S.J. Miller'sidentityand Cheby hev polynomials 11
Referen es 11
A knowledgements The rst author is nan ed by the ANR proje t Aspe ts
Arithmé-tiquesdes Matri es Aléatoires et duChaos Quantique. He would like tothank the
Uni-versity of Nottingham,where this work has been nished, for its hospitality. The se ond
one is supported by the ANR proje t Modunombres. Both authors would like to thank
the anonymousreferee of [12℄ for suggesting them this problem.
1. Introdu tion and statement of the result
1.1. Des ription ofthefamiliesof
L
-fun tionsstudied. Thepurposeofthispaper isto ompute the lowerorder termsofsome parti ular statisti sasso iated to low-lyingzerosofseveralfamiliesofsymmetri power
L
-fun tionsinthelevelaspe t: theone-level densities. First of all, we give a short des ription of these families. To any primitiveholomorphi uspform
f
of primelevelq
and even weight 1κ > 2
(see [12 , 2.1℄ forthe automorphi ba kground) sayf ∈ H
∗
κ
(q)
, one an asso iate itsr
-th symmetri powerL
-fun tion denoted byL(Sym
r
f, s)
for any integer
r > 1
. It is given by the following absolutely onvergent andnon-vanishingEuler produ t ofdegreer + 1
onℜe s > 1
L(Sym
r
f, s) =
Y
p∈P
L
p
(Sym
r
f, s)
Date:VersionofJune18,2008. 1
Inthispaper,theweight
κ
isaxed evenintegerand thelevelq
goestoinnityamongtheprime numbers.where
L
p
(Sym
r
f, s) =
r
Y
i=0
1 −
α
f
(p)
i
β
f
(p)
r−i
p
s
−1
for any prime number
p
. From now on,α
f
(p)
,β
f
(p)
are the Satake parameters off
at the prime numberp
and(λ
f
(n))
n>1
is its sequen e of He ke eigenvalues, whi h is arithmeti ally normalised:λ
f
(1) = 1
and|λ
f
(p)| 6 2
for any primep
. We also dene [1, (3.16) and (3.17)℄ a lo alfa tor at∞
whi h is given by a produ t ofr + 1
Gamma fa torsnamelyL
∞
(Sym
r
f, s) =
Y
06a6(r−1)/2
Γ
R
(s + (2a + 1)(κ − 1)/2) Γ
R
(s + 1 + (2a + 1)(κ − 1)/2)
(1.1) ifr
isodd andL
∞
(Sym
r
f, s) = Γ
R
(s + µ
κ,r
)
Y
16a6r/2
Γ
R
(s + a(κ − 1)) Γ
R
(s + 1 + a(κ − 1))
(1.2) ifr
iseven whereµ
κ,r
=
(
1
ifr(κ − 1)/2
isodd,0
otherwise. The ompletedL
-fun tionis dened byΛ(Sym
r
f, s) = (q
r
)
s/2
L
∞
(Sym
r
f, s)L(Sym
r
f, s)
and
q
r
isthearithmeti ondu tor. Wewillneedsome ontrolontheanalyti behaviour
ofthisfun tion. Unfortunately,su hinformationisnot urrentlyknowninallgenerality.
We sumup our main assumption inthefollowing statement.
Hypothesis
Nice(r, f )
The fun tionΛ (Sym
r
f, s)
is a ompleted
L
-fun tion in the sense that it satises thefollowing ni e analyti properties:•
it an be extended to an holomorphi fun tion of order1
onC
,•
it satises a fun tionalequation of the shapeΛ(Sym
r
f, s) = ε (Sym
r
f ) Λ(Sym
r
f, 1 − s)
where the sign
ε (Sym
r
f ) = ±1
of the fun tionalequation isgiven by
ε (Sym
r
f ) =
(
+1
ifr
iseven,
ε
f
(q) × ε(κ, r)
otherwise (1.3) withε(κ, r) = i(
r+1
2
)
2
(κ−1)+
r+1
2
=
i
κ
ifr ≡ 1 (mod 8)
,−1
ifr ≡ 3 (mod 8),
−i
κ
ifr ≡ 5 (mod 8),
+1
ifr ≡ 7 (mod 8)
andε
f
(q) = −
√
qλ
f
(q) = ±1
.Remark 1Hypothesis
Nice(r, f )
is known forr = 1
(E. He ke [3, 4 , 5℄),r = 2
thanks to theworkofS.Gelbartand H.Ja quet[2 ℄andr = 3, 4
fromtheworksofH.Kimand F. Shahidi[9 , 8,7 ℄.We aim at studying the lower order terms of the one-level density for the family of
L
-fun tions given by[
q
prime{L(Sym
r
f, s), f ∈ H
κ
∗
(q)}
1.2. One-leveldensities of these families. Thepurposeofthisworkistodetermine
thelowerordertermsoftheone-leveldensitiesasso iatedtothesefamiliesof
L
-fun tions. Let us give the statement of our result, in whi hν
is a positive real number,Φ
is an even S hwartz fun tion,whose Fourier transformΦ
b
is ompa tly supported in[−ν, +ν]
(denoted byΦ ∈ S
ν
(R)
) andf
is a primitive holomorphi usp form of prime levelq
and even weightκ > 2
for whi h hypothesisNice(r, f )
holds2
. We refer to [12 , 2.2℄
fortheprobabilisti ba kground. Notethat, thankstoFourier inversionformula,su h a
fun tion
Φ
an be extendedto an entire even fun tion whi h satises∀s ∈ C,
Φ(s) ≪
n
exp (ν|ℑm s|)
(1 + |s|)
n
(1.4)for any integer
n > 0
. Theone-level density (relativelytoΦ
) ofSym
r
f
isdened byD
1,q
[Φ; r](f ) =
X
ρ, Λ(Sym
r
f,ρ)=0
Φ
log (q
r
)
2iπ
ℜe ρ −
1
2
+ i ℑm ρ
wherethesumisoverthenon-trivialzeros
ρ
ofL(Sym
r
f, s)
repeatedwithmultipli ities.
Theasymptoti expe tation oftheone-leveldensityis bydenition
lim
q
primeq→+∞
X
f ∈H
∗
κ
(q)
ω
q
(f )D
1,q
[Φ; r](f )
whereω
q
(f ) =
Γ(κ−1)
(4π)
κ−1
hf,f i
q
isthe harmoni weight off
. Beforestatingour result,let us dene thefollowing onstants:C
PNT=
1 +
Z
+∞
1
θ(t) − t
t
2
dt
,
(1.5)C =
X
p∈P
log p
p
3/2
− p
,
(1.6)C
∞
= −(r + 1) log π + C
Γ
(1.7)where
θ
is the rstChebyshevfun tion:θ(t) =
X
p
primep6t
log p,
C
Γ
=
X
06a6(r−1)/2
Γ
′
Γ
1
4
+
(2a + 1)(κ − 1)
4
+
Γ
′
Γ
1
4
+
1
2
+
(2a + 1)(κ − 1)
4
(1.8) ifr
isodd andC
Γ
=
Γ
′
Γ
1
4
+
µ
κ,r
2
+
X
16a6r/2
Γ
′
Γ
1
4
+
a(κ − 1)
2
+
Γ
′
Γ
1
4
+
1
2
+
a(κ − 1)
2
(1.9) ifr
iseven.TheoremA Let
r > 1
be anyintegerandε = ±1
. We assumethathypothesisNice(r, f )
holds for any prime numberq
and any primitive holomorphi usp form of levelq
and even weightκ > 2
. Letν
1,max
(r, κ, θ
0
) =
1 −
1
2(κ − 2θ
0
)
2
r
2
2-with
θ
0
= 7/64
. Ifν < ν
1,max
(r, κ, θ
0
)
then the asymptoti expe tation of the one-level density isb
Φ(0) +
(−1)
r+1
2
Φ(0)
+
C
∞
− 2(−1)
r
C
PNT− 2δ
2|r
C
b
Φ(0)
log q
r
+ O
1
log
3
(q
r
)
.
Remark 2The main terms of the asymptoti expe tation of these one-level densities
have already been found in [12 ℄ (see Theorem B). The new information is the lower
orderterms namelyterms ofsize
1/ log (q
r
)
.
Remark 3Note that
θ
0
= 7/64
is the best known approximation towards Ramanujan-Peterson-Selberg's onje ture (see [12 , hypothesisH
2
(θ)
page 16℄) thanks to theworks of H.Kim,F. Shahidi andP.Sarnak ([8 ,7℄). The valueθ = 0
is expe ted.Remark 4 It is lear from the proof of Theorem A that the same result holds for the
signed familieswiththesame restri tion onthe support asin[12℄.
Remark 5 Theparti ular ase
r = 1
hasalreadybeen investigated byS.J. Miller[11 ℄. NotationWe writeP
fortheset ofprimenumbersandthe mainparameterinthispaper isa primenumberq
, whose name is the level, whi h goes to innityamongP
. Thus, iff
andg
aresomeC
-valuedfun tionsofthereal variablethenthenotationsf (q) ≪
A
g(q)
orf (q) = O
A
(g(q))
mean that|f(q)|
issmallerthan a onstant whi honlydepends onA
timesg(q)
at least forq
a large enough prime number.2. Chebyshev polynomials and He ke eigenvalues
Re allthatthegeneral fa tsaboutholomorphi uspforms anbefoundin[12 ,2.1℄.
Let
p 6= q
aprime number andf ∈ H
∗
κ
(q)
. Denote byχ
St
the hara ter ofthe standard representationSt
ofSU(2)
. Bythework of Deligne, thereexistsθ
f,p
∈ [0, π]
su h thatλ
f
(p) = χ
St
e
iθ
f,p
0
0
e
−iθ
f,p
.
Moreoverthe multipli ativityrelation reads
λ
f
(p
ν
) = χ
Sym
ν
e
iθ
f,p
0
0
e
−iθ
f,p
= X
ν
χ
St
e
iθ
f,p
0
0
e
−iθ
f,p
= X
ν
(λ
f
(p))
(2.1)where
χ
Sym
ν
isthe hara terofthe irredu iblerepresentationSym
ν
St
of
SU(2)
andthe polynomialsX
ν
aredened bytheir generatingseriesX
ν>0
X
ν
(x)t
ν
=
1
1 − xt + t
2
.
(2.2)Theyareequivalentely dened by
X
ν
(2 cos θ) =
sin ((ν + 1)θ)
sin (θ)
.
(2.3)These polynomials are known as Chebyshev polynomials of se ond kind. Ea h
X
ν
has degreeν
, is even ifν
is even and odd otherwise. The family(X
ν
)
ν>0
is a basis for the polynomialve tor spa eQ[T ]
,orthonormal withrespe tto theinnerprodu thP, Qi
ST
=
1
π
Z
2
−2
P (x)Q(x)
r
1 −
x
2
4
dx.
Thefollowing propositionlists Chebyshevpolynomials'neededproperties forthis work.
•
If̟ > 0
is any integerthenX
r
̟
=
r̟
X
j=0
x(̟, r, j)X
j
(2.4) withx(̟, r, j) = hX
r
̟
, X
j
i
ST
=
2
π
Z
π
0
sin
̟
((r + 1)θ) sin ((j + 1)θ)
sin
̟−1
(θ)
dθ.
(2.5) In parti ular,x(̟, r, j) =
0
ifj ≡ r̟ + 1 (mod 2),
(
̟
̟/2
)
1+̟/2
if̟
is even,r = 1
andj = 0
. (2.6)•
Ifα
isa omplex number of norm1
andn > 0
isan integer thenα
n
+ α
−n
=
2X
0
(α + α
−1
)
ifn = 0
,X
1
(α + α
−1
)
ifn = 1
,X
n
(α + α
−1
) − X
n−2
(α + α
−1
)
otherwise. (2.7)•
Ifα
isa omplex number of norm1
andr, n > 1
are someintegers thenS(α; n, r) =
r
X
j=0
α
n(2j−r)
= δ
2|r
+
X
16j6r
j≡r
(mod 2)
α
jn
+ α
−jn
(2.8)=
X
06j6r
j≡r
(mod 2)
X
jn
(α + α
−1
) − X
jn−2
(α + α
−1
)
(2.9)= X
r
(α
n
+ α
−n
)
(2.10) whereX
−1
= X
−2
= 0
by onvention.•
Ifr > 1
andn > 1
are someintegers thenX
06j6r
j≡r
(mod 2)
[X
jn
− X
jn−2
] =
r
X
j=0
(−1)
j
X
n−2
j
X
n(r−j)
(2.11) whereX
−1
= X
−2
= 0
by onvention.•
Ifℓ > 0
isan integerthenX
ℓ
=
X
06u6ℓ
u≡ℓ
(mod 2)
(−1)
(ℓ−u)/2
(ℓ + u)/2
u
T
u
.
(2.12)Proof of proposition2.1. Therst point follows fromthefa tthat
X
̟
r
isanpolynomial ofdegreer̟
,whi hisevenifr̟
isevenandoddotherwise. Thus,(2.4)istheexpansion ofthis polynomialinthe orthonormal basis(X
j
)
06j6r̟
. The se ondpoint follows from theequality2 cos (nθ) sin (θ) = sin ((n + 1)θ) − sin ((n − 1)θ).
If
α = exp (iθ)
thenthis equality ombined with(2.3) leadto2 cos (nθ) = X
n
(2 cos θ) − X
n−2
(2 cos θ),
whi histhe desiredresultsin e
2 cos θ = α + α
−1
and
2 cos (nθ) = α
n
+ α
−n
. Thethird
pointis adire t onsequen e of these ondone, ofthedire t omputation
S(α; n, r) =
α
n(r+1)
− α
−n(r+1)
and of
X
r
(α
n
+ α
−n
) = X
r
(2 cos (nθ)) =
α
n(r+1)
− α
−n(r+1)
α
n
− α
−n
if
α = exp (iθ)
. The fourthpoint is easily dedu edfromthefa tthatS(α; n, r) =
r
X
j=0
(−1)
j
X
n−2
j
(α + α
−1
)X
n(r−j)
(α + α
−1
)
for any omplex number
α
of norm1
. Letus prove theprevious equality. A ording to [13, Page 727, rstand se ondequations℄,X
r>0
X
nr
(α + α
−1
)t
r
=
1 + X
n−2
(α + α
−1
)t
X
r>0
X
r
(α
n
+ α
−n
)t
r
.
Asa onsequen e,X
nr
(α + α
−1
) = X
r
(α
n
+ α
−n
) + X
n−2
(α + α
−1
)X
r−1
(α
n
+ α
−n
),
whi himpliesX
r
(α
n
+ α
−n
) =
r
X
j=0
(−1)
j
X
n−2
j
(α + α
−1
)X
n(r−j)
(α + α
−1
).
Thelast point isobtained by developping (2.2) asan entire series in
x
. 3. Riemann'sexpli it formula for symmetri powerL
-fun tions TostudyD
1,q
[Φ; r](f )
foranyΦ ∈ S
ν
(R)
,wetransformthissumoverzerosintoasum overprimes inthenextproposition. Inotherwords,weestablishanexpli itformulaforsymmetri power
L
-fun tions.Proposition 3.1 Let
r > 1
andf ∈ H
∗
κ
(q)
for whi hhypothesisNice(r, f )
holds andletΦ ∈ S
ν
(R)
. We haveD
1,q
[Φ; r](f ) =
b
Φ(0) +
(−1)
r+1
2
Φ(0)
+
Φ(0)
b
log (q
r
)
C
∞
+ 2(−1)
r+1
C
PNT− 2δ
2|r
C
+ P
q
1
[Φ; r](f ) +
r−1
X
m=0
(−1)
m
P
q
2
[Φ; r, m](f ) + P
q
3
[Φ; r](f ) + O
1
log
3
(q
r
)
whereC
PNT is dened in (1.5) ,C
in (1.6) ,C
∞
in (1.7) whereasP
q
1
[Φ; r](f ) = −
2
log (q
r
)
X
p∈P
p∤q
λ
f
(p
r
)
log p
√
p
Φ
b
log p
log (q
r
)
,
P
q
2
[Φ; r, m](f ) = −
2
log (q
r
)
X
p∈P
p∤q
λ
f
p
2(r−m)
log p
p
Φ
b
2 log p
log (q
r
)
P
q
3
[Φ; r](f ) = −
2
log (q
r
)
X
p∈P
p∤q
X
n>3
X
16j6r
j≡r
(mod 2)
λ
f
(p
jn
) − λ
f
(p
jn−2
)
log p
p
n/2
Φ
b
n log p
log (q
r
)
for any integer
m ∈ {0, . . . , r − 1}
. Proof of proposition3.1. LetG(s) = Φ
log (q
r
)
2iπ
s −
1
2
.
From [6,eq. (4.11) and (4.14)℄we get
D
1,q
[Φ; r](f ) = b
Φ(0) − (r + 1)
log π
log q
r
Φ(0)
b
−
log q
2
r
X
p∈P
+∞
X
m=1
r
X
j=0
α
f
(p)
jm
β
f
(p)
(r−j)m
b
Φ
m log p
log q
r
log p
p
m/2
+
Φ(0)
b
log q
r
r
X
j=0
Γ
′
Γ
1
4
+
µ
j
2
+ O
1
log
3
q
.
(3.1)Letusfo usonthethirdterm in(3.1) . Notthatthe ontributionoftheprime
q
isgiven by−
2
r
+∞
X
m=1
λ
f
(q)
r
√
q
m
b
Φ
m
r
≪
1
q
(r+1)/2
and for
p 6= q
we user
X
j=0
α
f
(p)
jm
β
f
(p)
(r−j)m
= S (α
f
(p); m, r)
withthenotation of (2.8) . We obtain
S (α
f
(p); 1, r) = X
r
α
f
(p) + α
f
(p)
−1
= λ
f
(p
r
)
a ordingto (2.1) andS (α
f
(p); 2, r) =
X
06j6r
j≡r
(mod 2)
X
2j
α
f
(p) + α
f
(p)
−1
− X
2j−2
α
f
(p) + α
f
(p)
−1
=
r
X
j=0
(−1)
j
X
2(r−j)
α
f
(p) + α
f
(p)
−1
(
f.(2.11))
=
r−1
X
m=0
(−1)
m
λ
f
p
2(r−m)
+ (−1)
r
.
Asa onsequen e,X
p∈P
p6=q
+∞
X
m=1
r
X
j=0
α
f
(p)
jm
β
f
(p)
(r−j)m
=
X
p∈P
p∤q
λ
f
(p
r
) log p
p
1/2
Φ
b
log p
log (q
r
)
+
X
p∈P
p∤q
r−1
X
m=0
(−1)
m
λ
f
p
2(r−m)
!
log p
p
Φ
b
log (p
2
)
log (q
r
)
+ (−1)
r
X
p∈P
p∤q
log p
p
Φ
b
log (p
2
)
log (q
r
)
+
X
p∈P
p∤q
X
n>3
S (α
f
(p); n, r)
log p
p
n/2
Φ
b
log (p
n
)
log (q
r
)
.
(3.2)summation, thisterm equals, up to
O(q
−0.9
)
,(−1)
r
Z
+∞
1
θ(t)
t
2
b
Φ
2 log t
log (q
r
)
−
log (q
2
r
)
Φ
b
′
2 log t
log (q
r
)
dt := S
3
.
Then,S
3
= (−1)
r
Z
+∞
1
b
Φ
2 log t
log (q
r
)
−
log (q
2
r
)
Φ
b
′
2 log t
log (q
r
)
dt
t
+ (−1)
r
Z
+∞
1
θ(t) − t
t
b
Φ
2 log t
log (q
r
)
−
log (q
2
r
)
Φ
b
′
2 log t
log (q
r
)
dt
t
.
Sin e
Φ(u) = b
b
Φ(0) + O(u
2
)
andΦ
b
′
(u) ≪ |u|
,we getS
3
= (−1)
r
log (q
r
)
2
Z
+∞
0
b
Φ(u) du − (−1)
r
Z
+∞
0
b
Φ
′
(u) du + (−1)
r
Φ(0)
b
Z
+∞
1
θ(t) − t
t
2
dt
+ O
1
log
2
(q
r
)
and nallyS
3
= (−1)
r
log (q
r
)
4
Φ(0) + (−1)
r
Φ(0)
b
1 +
Z
+∞
1
θ(t) − t
t
2
dt
+ O
1
log
2
(q
r
)
.
We nallytake areofthe fourth term of (3.2). A ording to (2.1) and (2.8) ,we have
S (α
f
(p); n, r) = δ
2|r
+
X
16j6r
j≡r
(mod 2)
λ
f
(p
jn
) − λ
f
(p
jn−2
)
.
Onemayremark that
X
p∈P
p∤q
X
n>3
log p
p
n/2
Φ
b
n log p
log (q
r
)
=
X
p∈P
X
n>3
log p
p
n/2
Φ (0) + O
b
1
log
3
(q
r
)
sin e
Φ(u) = b
b
Φ(0) + O(u
2
)
. Then, we easily get
X
p∈P
X
n>3
log p
p
n/2
=
X
p∈P
log p
p
3/2
− p
.
4. Proof of Theorem ATheaim of thispartis to determine anasymptoti expansion of
X
f ∈H
∗
κ
(q)
ω
q
(f )D
1,q
[Φ; r](f ) = E
h
q
(D
1,q
[Φ; r]) .
A ording to proposition3.1and theproofof [12 , eq. (4.6) and (4.7)℄, if
ν <
1 −
1
2(κ − 2θ)
2
r
2
(4.1) thenE
h
q
(D
1,q
[Φ; r]) =
b
Φ(0) +
(−1)
r+1
2
Φ(0)
+
Φ(0)
b
log (q
r
)
C
∞
+ 2(−1)
r+1
C
PNT− 2δ
2|r
C
+ E
h
q
P
q
3
[Φ; r](f )
+ O
1
log
3
(q
r
)
.
(4.2)The rst term in (4.2) is the main term given in Theorem A. We now estimate the
penultemate term of (4.2)via thetra e formulagiven in[12 ,Proposition 2.2℄:
E
h
q
P
q
3
[Φ; r]
= P
3
q,new
[Φ; r] + P
3
q,old
[Φ; r]
(4.3) whereP
3
q,new
[Φ; r] = −
2
log (q
r
)
X
p∈P
p∤q
X
n>3
X
16j6r
j≡r
(mod 2)
∆
q
(p
jn
, 1) − ∆
q
(p
jn−2
, 1)
log p
p
n/2
Φ
b
n log p
log (q
r
)
andP
3
q,old
[Φ; r] =
2
q log (q
r
)
X
ℓ|q
∞
1
ℓ
X
p∈P
p∤q
X
n>3
X
16j6r
j≡r
(mod 2)
∆
1
(p
jn
ℓ
2
, 1) − ∆
1
(p
jn−2
ℓ
2
, 1)
log p
p
n/2
Φ
b
n log p
log (q
r
)
.
Form 6= 1
we have∆
k
(m, 1) = 2πi
κ
X
c>1
k|c
S(m, 1; c)
c
J
κ−1
4π
√
m
c
where
S(m, 1; c)
is a Kloosterman sum. Let us estimate the new part whi h an be written asP
3
q,new
[Φ; r] = −
2(2πi
κ
)
log (q
r
)
X
16j6r
j≡r
(mod 2)
X
n>3
P
3
q,new
[Φ; r, jn] − P
3
q,new
[Φ; r, jn − 2]
whereP
3
q,new
[Φ; r, k] =
X
p∈P
p6=q
log p
p
n/2
Φ
b
log p
log q
r/n
!
X
c>1
q|c
S(p
k
, 1; c)
c
J
κ−1
4π
p
p
k
c
!
.
(4.4)By [12,lemma 3.10℄,the
c
-sum in(4.4) isbounded byτ (q)
√
q
√
p
k
q
1/2
ifp > q
2/k
,√
p
k
q
κ−1
otherwise. We dedu eX
n>3
P
3
q,new
[Φ; r, jn] ≪
τ (q)
q
κ−1/2
X
n>3
X
p6q
rν/n
1
p
n/2
p
rn(κ−1)/2
log p
≪
τ (q)
q
κ−1/2
X
36n6νr log q/log 2
1
n
q
νr[((κ−1)r−1)n/2+1]/n
≪
τ (q)
q
κ−1/2
q
νr[(κ−1)r−1]/2
q
νr/3
log log(3q)
≪
1
q
1/2
assoon as
ν < 2/r
2
(and in parti ular if (4.1) is satised). We make thesame
ompu-tationsfor
jn − 2
andnd thenthatP
3
q,new
[Φ; r, k]
is anadmissible error term. Theold partisP
3
q,old
[Φ; r] =
2(2πi
κ
)
q log (q
r
)
X
16j6r
j≡r
(mod 2)
X
n>3
P
3
q,old
[Φ; r, jn] − P
3
q,old
[Φ; r, jn − 2]
whereP
3
q,old
[Φ; r, k] =
X
p∈P
p6=q
log p
p
n/2
Φ
b
log p
log q
r/n
X
ℓ|q
∞
1
ℓ
∆
1
(p
k
ℓ
2
, 1).
From [12,eq (3.2) and(3.3)℄ we have
X
ℓ|q
∞
1
ℓ
∆
1
(p
k
ℓ
2
, 1) 6 2(k + 1)
sothatX
n>3
P
3
q,old
[Φ; r, jn] ≪ 1
and similaryfor
P
3
q,old
[Φ; r, jn − 2]
. FinallyE
h
q
P
q
3
[Φ; r]
enters the
O(1/ log
3
q
r
)
term.
Appendix A. Some omments onan aestheti identity
Itis possible to prove onindu tion on
k
0
>
1
thefollowing equalityinQ
[T ]
:X
2k
0
− X
2k
0
−2
=
k
X
0
−1
j=0
X
16k
j
<k
j−1
<···<k
1
<k
0
(−1)
j
"
j−1
Y
i=0
2k
i
k
i
− k
i+1
#
T
2k
j
−
2k
j
k
j
.
(A.1) Asa onsequen e,ifK > 1
thenX
2K+1
− X
2K−1
= (−1)
K
T
1 +
X
16k
0
6K
(−1)
k
0
X
2k
0
− X
2k
0
−2
.
(A.2)Now, use(2.4) with
r = 1
(sothatX
1
= T
)to getfrom (A.1) theequalityX
2k
0
− X
2k
0
−2
=
k
X
0
−1
j=0
X
16k
j
<k
j−1
<···<k
1
<k
0
(−1)
j
"
j−1
Y
i=0
2k
i
k
i
− k
i+1
#
2k
j
X
ℓ=0
x(2k
j
, 1, ℓ)X
ℓ
−
2k
j
k
j
X
0
and ompare the oe ientsof
X
0
to obtain, thanksto (2.6)the equalityk
0
−1
X
j=0
X
16k
j
<k
j−1
<···<k
1
<k
0
(−1)
j
"
j−1
Y
i=0
2k
i
k
i
− k
i+1
#
2k
j
k
j
k
j
1 + k
j
= 0.
We ould have expressed formulas (A.1) and (A.2) in terms of Fourier oe ients of
primitive forms to determine the lower order terms. However, this is denitely not the
bestway topro eedsin e it onsistsinde omposingthepolynomial
X
K
− X
K−2
inthe anoni al basis ofQ[T ]
and de omposing again ea h element of this anoni al basisin theChebyshev basis(X
ℓ
)
ℓ∈N
.Appendix B. S.J.Miller'sidentity and Cheby hev polynomials
S.J. Miller([10 ,Equation(3.12) Page 6℄) re ently proved that
α
f
(p)
K
+ β
f
(p)
K
=
X
06k6K
k≡K
(mod 2)
c
K,k
λ
f
(p)
k
(B.1)where
c
K,k
= 0
ifk ≡ K + 1 (mod 2)
andc
0,0
= 0,
c
2K,0
= 2(−1)
K
(K > 1),
c
2K,2L
=
2(−1)
K+L
K(K + L − 1)!
(2L)!(K − L)!
(1 6 L 6 K),
c
2K+1,2L+1
=
(−1)
K+L
(2K + 1)(K + L)!
(2L + 1)!(K − L)!
(0 6 L 6 K).
We would like to give a qui k proof of this identity, the ru ial tool being Cheby hev
polynomials.
Proof of equation (B.1) . We knowthat
α
f
(p)
K
+ β
f
(p)
K
= X
K
(λ
f
(p)) − X
K−2
(λ
f
(p))
for
K > 2
a ordingto (2.7) . Thus, theproof onsists in de omposing thepolynomialX
K
− X
K−2
inthe anoni al basisofQ[T ]
. This an be done via(2.12) . Itentailsthatα
f
(p)
K
+β
f
(p)
K
=
X
06u6K−2
u≡K
(mod 2)
(−1)
(K−u)/2
(K + u)/2
u
+
(K + u)/2 − 1
u
λ
f
(p)
u
+
X
K−16u6K
u≡K
(mod 2)
(−1)
(K−u)/2
(K + u)/2
u
λ
f
(p)
u
,
whi his anequivalent formulationof (B.1) .
Remark B.1 Equation(B.1) ouldbeusedtore overthelower orderterms omingfrom
P
q
3
[Φ; r]
but,on e again, it is not the most lever way to pro eed sin e it would imply de omposingthepolynomialsX
K
−X
K−2
inthe anoni albasisofQ[T ]
atthebeginning ofthepro ess andde omposingthe polynomialsT
j
inthebasis
(X
r
)
r>0
justbeforethe endoftheproofinordertobeabletoapplysometra eformulafortheFourier oe ientsof uspforms.
Referen es
[1℄ J.CogdellandP.Mi hel, Onthe omplexmomentsof symmetri power
L
-fun tionsats
= 1
,Int. Math.Res.Not.(2004),no. 31,15611617. MRMR2035301(2005f:11094)[2℄ StephenGelbartandHervéJa quet,Arelationbetweenautomorphi representationsof
GL(2)
andGL(3)
,Ann.S i.É oleNorm.Sup.(4)11(1978),no.4,471542. MRMR533066(81e:10025) [3℄ E.He ke,ÜberdieBestimmungDiri hlets herReihendur hihreFunktionalglei hung,Math.Ann.112(1936),no.1,664699. MRMR1513069
[4℄ ,ÜberModulfunktionen unddieDiri hlets hen Reihenmit Eulers her Produktentwi klung.
I,Math.Ann.114(1937),no.1,128. MRMR1513122
[5℄ ,ÜberModulfunktionen unddieDiri hlets hen Reihenmit Eulers her Produktentwi klung. II,Math.Ann.114(1937),no.1,316351. MRMR1513142
[6℄ Henryk Iwanie , WenzhiLuo,and PeterSarnak,Low lyingzeros of familiesof
L
-fun tions, Inst. HautesÉtudesS i.Publ.Math.(2000),no.91, 55131(2001). MRMR1828743 (2002h:11081) [7℄ Henry H.Kim, Fun toriality forthe exterior square ofGL
4
and thesymmetri fourth ofGL
2
,J.[8℄ Henry H. Kim and Freydoon Shahidi, Cuspidality of symmetri powers with appli ations, Duke Math.J.112(2002),no.1,177197. MRMR1890650 (2003a:11057)
[9℄ , Fun torialprodu tsfor
GL
2
× GL
3
andthe symmetri ube forGL
2
, Ann.ofMath. (2) 155 (2002), no. 3, 837893, With an appendix by Colin J. Bushnell and Guy Henniart. MRMR1923967 (2003m:11075)
[10℄ StevenJ.Miller, Anidentity forsumsofpolylogarithm fun tions.
[11℄ , One- and two-level densities for rational families of ellipti urves: eviden e for the
underlying group symmetries, Compos. Math. 140 (2004), no. 4, 952992. MR MR2059225 (2005 :11085)
[12℄ Guillaume Ri otta and Emmanuel Royer, Statisti s for low-lying zeros of symmetri power
l
-fun tionsinthelevelaspe t,(2007),submitted,availableathttp://arxiv.org/abs/math/07 03760 .[13℄ EmmanuelRoyerandJieWu,Spe ialvaluesofsymmetri power