Auxiliary Fns Hal 072009

Auxiliary functions in transcendence proofs
Michel Waldschmidt
To cite this version:

Michel Waldschmidt. Auxiliary functions in transcendence proofs. 2009. <hal-00405120>
HAL Id: hal-00405120

https://hal.archives-ouvertes.fr/hal-00405120
Submitted on 24 Jul 2009
HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
SASTRA Ramanujan Lectures The Ramanujan Journal
Updated: March 13, 2009
Auxiliary functions in transcendental number theory.

Michel Waldschmidt
Contents
1 Explicit functions 3
1.1 Liouville . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Hermite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Padé approximation . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 Hypergeometric methods . . . . . . . . . . . . . . . . . . . . . . . 7
2 Interpolation methods 7
2.1 Weierstraß question . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Integer–valued entire functions . . . . . . . . . . . . . . . . . . . 9
2.3 Transcendence of eπ . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.4 Lagrange interpolation . . . . . . . . . . . . . . . . . . . . . . . . 12
3 Auxiliary functions arising from the Dirichlet’s box principle 13

3.1 Thue–Siegel lemma . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1.1 The origin of the Thue–Siegel Lemma . . . . . . . . . . . 14
3.1.2 Siegel, Gel’fond, Schneider . . . . . . . . . . . . . . . . . . 15
3.1.3 Schneider–Lang Criterion . . . . . . . . . . . . . . . . . . 18
3.1.4 Higher dimension: several variables . . . . . . . . . . . . . 20
3.1.5 The six exponentials Theorem . . . . . . . . . . . . . . . 22
3.1.6 Modular functions . . . . . . . . . . . . . . . . . . . . . . 23
3.2 Universal auxiliary functions . . . . . . . . . . . . . . . . . . . . 24
3.2.1 A general existence theorem . . . . . . . . . . . . . . . . . 24
3.2.2 Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.3 Mahler’s Method . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4 Interpolation determinants 28
4.1 Lehmer’s Problem . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.2 Laurent’s interpolation determinants . . . . . . . . . . . . . . . . 29
5 Bost slope inequalities, Arakelov’s Theory 30
1
Acknowledgement
The author is grateful to Paul Voutier for a number of clever remarks on a
preliminary version of this survey.
Abstract
We discuss the role of auxiliary functions in the development of transcendental
number theory.
Earlier auxiliary functions were completely explicit (§ 1). The earliest tran-
scendence proof is due to Liouville (§ 1.1), who produced the first explicit ex-
amples of transcendental numbers at a time where their existence was not yet
known; in his proof, the auxiliary function is just a polynomial in one variable.
Hermite’s proof of the transcendence of e (1873) is much more involved, the
auxiliary function he builds (§ 1.2) is the first example of the Padé approxi-
mants (§ 1.3), which can be viewed as a far reaching generalization of continued
fraction expansion [13, 14]. Hypergeometric functions (§ 1.4) are among the
best candidates for using Padé approximations techniques.
Another tool which plays the role of auxiliary functions is produced by in-
terpolation formulae (§ 2). They occurred in the theory after a question by
Weierstraß (§ 2.1) on the so-called exceptional set Sf of a transcendental func-
tion f , which is the set of algebraic numbers α such that f (α) is algebraic. The
answer to his question is that any set of algebraic numbers is the exceptional
set of some transcendental function f ; this shows that one should add further
conditions in order to get transcendence criteria. One way is to replace algebraic
number by rational integer: this gives rise to the study of integer–valued entire
functions (§ 2.2) with the works of G. Pólya (1915), A.O. Gel’fond (1929) and
many others. The connexion with transcendental number theory may not have
been clear until the solution by A.O. Gel’fond in 1929 of the question of the
transcendence of eπ , a special case of Hilbert’s seventh problem (§ 2.3). Along
these lines, recent developments are due to T. Rivoal, who renewed forgotten
rational interpolation formulae (1935) of R. Lagrange (§ 2.4).
The simple (but powerful) construction by Liouville was extended to several
variables by A. Thue (§ 3.1.1), who introduced the Dirichlet’s box principle (pi-
geonhole principle) (§ 3) into the topic of Diophantine approximation in the early
1900’s. In the 1920’s, Siegel (§ 3.1) developed this idea and applied it in 1932 to
transcendental number theory. This gave rise to the Gel’fond–Schneider method
(§ 3.1.2) which produces the Schneider–Lang Criterion in one (§ 3.1.3) or sev-
eral (§ 3.1.4) variables. Among many developments of this method are results
on modular functions (§ 3.1.6). Variants of the auxiliary functions produced
by Dirichlet’s Box Principle are universal auxiliary functions, which have small
Taylor coefficients at the origin (§ 3.2). Another approach, due to K. Mahler
(§ 3.3), involves auxiliary functions whose existence is deduced from linear al-
gebra instead of Thue–Siegel Lemma.
In 1991 M. Laurent introduced interpolation determinants (§ 4). Two years
2
later J.B. Bost used Arakhelov theory (§ 5) to prove slope inequalities, which
dispenses of the choice of bases.
1 Explicit functions
1.1 Liouville
The first examples of transcendental numbers were produced by Liouville [48]
in 1844. At that time, it was not yet known that transcendental numbers exist.
The idea of Liouville is to show that all algebraic real numbers α are badly
approximated by rational numbers. The simplest example is a rational number
α = a/b: for any rational number p/q 6= a/b, the inequality

a p
− ≥ 1
b q bq
holds. For an irrational real number x, on the contrary, for any ǫ > 0 there
exists a rational number p/q such that

p ǫ
0 < x − ≤ ·

q q
This yields an irrationality criterion, which is the basic tool for proving the
irrationality of specific numbers: a real number x is irrational if and only if
there exists a sequence (pn /qn )n≥0 of distinct rational numbers with

pn
lim qn x −
= 0.
n→∞ qn
This criterion is not too demanding: the quality of the approximation is not very
strong. Indeed for an given an irrational number x, it is known that there exist
much better rational approximations, since there exist infinitely many rational
numbers p/q for which
x − p ≤ 1 ·

q 2q 2
It is a remarkable fact that on the one hand there exist such sharp approxima-
tions, and on the other hand we are usually not able to produce (or at least to
show the existence of) much weaker rational approximations. Also in examples
like ζ(3), there is a single known explicit sequence of rational approximations
which arises from the irrationality proofs – of course, once the irrationality is
established, the existence of other much better approximations follows, but so
far no one is able to produce an explicit sequence of such approximations.
Liouville extended the irrationality criterion into a transcendence criterion.
The proof by Liouville involves the irreducible polynomial f ∈ Z[X] of the given
irrational algebraic number α. Since α is algebraic, there exists an irreducible
3
polynomial f ∈ Z[X] such that f (α) = 0. Let d be the degree of f . For p/q ∈ Q
the number q d f (p/q) is a non–zero rational integer, hence
1
|f (p/q)| ≥ ·
qd
On the other hand, it is easily seen that there exists a constant c(α) > 0,
depending only on α (and its irreducible polynomial f ), such that

p
|f (p/q)| ≤ c(α) α − .

q
An explicit value for a suitable c(α) is given for instance in Exercice 3.6 of [104].
Therefore
′

α − p ≥ c (α)

q qd
with c′ (α) = 1/c(α).
Let ξ be a real number such that, for any κ > 0, there exists a rational
number p/q with q ≥ 2 satisfying

p 1
0 < ξ − < ·
q qκ
It follows from Liouville’s inequality that ξ is transcendental. Real numbers

satisfying this assumption are called Liouville’s numbers. The first examples
[48] produced by Liouville in 1844 involved properties of continued fractions,
but already in the first part of his 1844 note he considered series
X 1
ℓm!
m≥1
for ℓ ∈ Z≥2 . Seven years later in [49] he refers to a letter from Goldbach to
Euler for numbers
X km
10m!
m≥1
where (km )m≥1 is a sequence of integers in the range {0, . . . , 9}. Next (p. 140
of [49]) he uses the same argument to prove the irrationality of
X 1
m≥1
ℓm2
whose transcendence was proved only in 1996 by Yu. V. Nesterenko [61, 63].
We consider below (§ 3.1.1) extensions of Liouville’s result by Thue, Siegel,
Roth and Schmidt.
4
1.2 Hermite
During his course at the École Polytechnique in 1815 (see [84]), Fourier gave a
simple proof for the irrationality of e, which can be found in many textbooks (for
instance Th. 47 Chap. 4 § 7 of Hardy and Wright [36]). The idea is to truncate
the Taylor expansion at the origin of the exponential function. In this proof, the
auxiliary function is the tail of the Taylor expansion of the exponential function
N
X zn
ez −
n=0
n!
which one specializes at z = 1. As noticed by F. Beukers, the proof becomes even

shorter if one specializes at z = −1. This proof has been revisited by Liouville
in 1840 [47] who succeeded to extend the argument and to prove that e2 is
not a quadratic number. This result is quoted by Hermite in his memoir [37].
Fourier’s argument produces rational approximations to the number e, which
are sharp enough to prove the irrationality of e but not the transcendence.
The denominators of these approximations are N !. One idea of Hermite is
to look for other rational approximations. Instead of the auxiliary functions
ez − A(z) for A ∈ Q[z], Hermite introduces more general auxiliary functions
R(z) = B(z)ez − A(z). He finds a polynomial B such that the Taylor expansion
at the origin of B(z)ez has a large gap: he calls A(z) the polynomial part of
the expansion before the gap, so that the auxiliary function R(z) has a zero of
high multiplicity at the origin. Hermite gives explicit formulae for A, B and R,
in particular the polynomials A and B have rational coefficients – the question
is homogeneous, one may multiply by a denominator to get integer coefficients.
Also he obtains upper bounds for these integer coefficients (they are not too
large) and for the modulus of the remainder (which is small on a given disc).
As an example, given r ∈ Q \ {0} and ǫ > 0, one can use this construction to
show the existence of A, B and R with 0 < |R(r)| < ǫ. Hence er 6∈ Q. This
gives another proof of Lambert’s result on the irrationality of er for r ∈ Q \ {0},
and this proof extends to the irrationality of π as well [12, 62].
Hermite [37] goes much further, since he obtains the transcendence of e.
To achieve this goal, he considers simultaneous rational approximations to the
exponential function, in analogy with Diophantine approximation. The idea is
as follows. Let B0 , B1 , . . . , Bm be polynomials in Z[x]. For 1 ≤ k ≤ m, define
Rk (x) = B0 (x)ekx − Bk (x).
Set bj = Bj (1), 0 ≤ j ≤ m and
R = a1 R1 (1) + · · · + am Rm (1).
The numbers aj and bj are rational integers, hence
a0 b0 + a1 b1 + · · · + am bm = b0 (a0 + a1 e + a2 e2 + · · · + am em ) − R
also. Therefore, if one can prove 0 < |R| < 1, then one deduces
a0 + a1 e + · · · + am em 6= 0.
5
Hermite’s construction is more general: he produces rational approximations to
the functions 1, eα1 x , . . . , eαm x , when α1 , . . . , αm are pairwise distinct complex
numbers. Let n0 , . . . , nm be rational integers, all ≥ 0. Set N = n0 + · · · + nm .
Hermite constructs explicitly polynomials B0 , B1 , . . . , Bm , with Bj of degree
N − nj , such that each of the functions
B0 (z)eαk z − Bk (z), (1 ≤ k ≤ m)
has a zero at the origin of multiplicity at least N .

Such functions are now known as Padé approximations of the second kind
(or of type II).
1.3 Padé approximation

In his thesis in 1892, H.E. Padé studied systematically the approximation of
complex analytic functions by rational functions. See Brezinski’s papers [13, 14],
where further references to previous works by Jacobi (1845), Faá de Bruno
(1859), Sturm, Brioschi, Sylvester, Frobenius (1870), Darboux (1876), Kro-
necker (1881) are given.
There are two dual points of view, giving rise to the two types of Padé
Approximants [24].
Let f0 , . . . , fm be complex functions which are analytic near the origin and
n0 , . . . , nm be non–negative rational integers. Set N = n0 + · · · + nm .
Padé approximants of type II are polynomials B0 , . . . , Bm with Bj having
degree ≤ N − nj , such that each of the functions
Bi (z)fj (z) − Bj (z)fi (z) (0 ≤ i < j ≤ m)
has a zero of multiplicity > N .

Padé approximants of type I are polynomials P1 , . . . , Pm with Pj of degree
≤ nj such that the function
P1 (z)f1 (z) + · · · + Pm (z)fm (z)
has a zero at the origin of multiplicity at least N + m − 1.

For type I as well as type II, existence of Padé approximants follow from
linear algebra: one compares the number of equations which are produced by
the vanishing conditions on the one hand, with the number of coefficients of
P , considered as unknowns, on the other. Unicity of the solution, up to a
multiplicative constant (the linear system of equations is homogeneous) is true
only in specific cases (perfect systems): it amounts to proving that the matrix
of the system of equations is regular.
These approximants were also studied by Ch. Hermite for the exponentials
functions in 1873 and 1893; later, in 1917, he gave further integral formulae for
the remainder. For transcendence purposes, Padé approximants of type I have
been used for the first time in 1932 by K. Mahler [51], who gave effective version
of the transcendence theorems by Hermite, Lindemann and Weierstraß.
6
In the theory of Diophantine approximation, there are transference theo-
rems, initially due to Khintchine (see for instance [16, 44]). Similar transference
properties for Padé approximation have been considered by H. Jager [39] and
J. Coates [18, 19].
1.4 Hypergeometric methods

Explicit Padé approximations are known only for restricted classes of functions;
however, when they are available, they often produce very sharp Diophantine
estimates. Among the best candidates for having explicit Padé Approxima-
tions are the hypergeometric functions. A. Thue [88] developed this idea in the
early 20th Century and was able to solve explicitly several classes of Diophantine
equations. There is a contrast between the measures of irrationality for instance
which can be obtained by hypergeometric methods and those produced by other
methods, like Baker’s method (§ 3.1.2): typically, hypergeometric methods pro-
duce numerical constants with one or two digits (when the expected value is
something like 2), where Baker’s method produces constants with several hun-
dreds digits. On the other hand, Baker’s method works in much more general
situations. Compared with the Thue–Siegel–Roth–Schmidt’s method (§ 3.1.1),
it has the great advantage of being explicit.
Among many contributors to this topic, we quote A. Thue, C.L. Siegel,
A. Baker, G.V. Chudnovskii, M. Bennett, P. Voutier, G. Rhin, C. Viola, T. Rivoal. . .
These works also involve sorts of auxiliary functions (integrals) depending on
parameters which needs to be suitably selected in order to produce sharp esti-
mates.
Chapter 2 of [24] deals with effective constructions in transcendental number
theory and includes two sections (§ 6 and § 7) on generalized hypergeometric
functions and series.
2 Interpolation methods
We discuss here another type of auxiliary function which occurred in works
related with a question of Weierstraß on the exceptional set of a transcendental
entire function. Recall that an entire function is a complex valued function
which is analytic in C. A function f is algebraic (over C(z)) if f is a solution of
a functional equation P (z, f (z)) = 0 for some non–zero polynomial P ∈ C[X, Y ].
An entire function is algebraic if and only if it is a polynomial. A function which
is not algebraic is called transcendental1 .
1 A polynomial whose coefficients are not all algebraic numbers is an algebraic function,
namely is algebraic over C(z), but is a transcendental element over Q(z). However, as soon
as a polynomial assumes algebraic values at infinitely many points (including derivatives), its
coefficients are algebraic. Therefore for the questions we consider it makes no difference to
consider algebraicity of functions over C or over Q.
7
2.1 Weierstraß question
Weierstraß (see [53]) initiated the question of investigating the set of algebraic
numbers where a given transcendental entire function f takes algebraic values.
Denote by Q the field of algebraic numbers (algebraic closure of Q in C).
For an entire function f , we define the exceptional set Sf of f as the set of
algebraic numbers α such that f (α) is algebraic:

Sf := α ∈ Q ; f (α) ∈ Q .
For instance Hermite–Lindemann’s Theorem on the transcendence of log α and

eβ for α and β algebraic numbers is the fact that the exceptional set of the
function ez is {0}. Also the exceptional set of ez + e1+z is empty, by the
Theorem of Lindemann–Weierstrass. The exceptional set of functions like 2z or
eiπz is Q, as shown by the Theorem of Gel’fond and Schneider.
The exceptional set of a polynomial is Q if the polynomial has algebraic
coefficients, otherwise it is finite. Also any finite set of algebraic numbers is
the exceptional set of some entire function: for s ≥ 1 the set {α1 , . . . , αs } is
the exceptional set of the polynomial π(z − α1 ) · · · (z − αs ) ∈ C[z] and also of
the transcendental entire function (z − α2 ) · · · (z − αs )ez−α1 . It does not seem
easy (without Schanuel’s conjecture) to produce explicit examples of functions
having either Z≥0 or Z as exceptional set.
The study of exceptional sets started in 1886 with a letter of Weierstrass to
Strauss. It was later developed by Strauss, Stäckel, Faber – see [53]. Further
results are due to van der Poorten, Gramain, Surroca and others (see [35, 86]).
Among the results which were obtained, a typical one is the following: if A is
a countable subset of C and if E is a dense subset of C, there exist transcendental
entire functions f mapping A into E.
Also van der Poorten noticed in [89] that there are transcendental entire
functions f such that Dk f (α) ∈ Q(α) for all k ≥ 0 and all algebraic α.
The question of possible sets Sf has been solved in [38]: any set of algebraic
numbers is the exceptional set of some transcendental entire function. Also mul-
tiplicities can be included, as follows: define the exceptional set with multiplicity
of a transcendental entire function f as the subset of (α, t) ∈ Q × Z≥0 such that
f (t) (α) ∈ Q. Here f (t) stands for the t-th derivative of f .
Then any subset of Q × Z≥0 is the exceptional set with multiplicities of some
transcendental entire function f . More generally, the main result of [38] is the
following:
Let A be a countable subset of C. For each pair (α, s) with α ∈ A,

and s ∈ Z≥0 , let Eα,s be a dense subset of C. Then there exists a
transcendental entire function f such that
s
d
f (α) ∈ Eα,s
dz
for all (α, s) ∈ A × Z≥0 .
8
One may replace C by R: it means that one may take for the sets Eα,s dense
subsets of R, provided that one requires A to be a countable subset of R.
The proof is a construction of an interpolation series (see § 2.2) on a sequence
where each w occurs infinitely often. The coefficients of the interpolation series
are selected recursively to be sufficiently small (and nonzero), so that the sum
f of the series is a transcendental entire function.
This process yields uncountably many such functions. One may also require
that they are algebraically independent over C(z) together with their derivatives.
Further, at the same time, one may request each of these functions f to have a
small growth, meaning that f satisfies |f |R ≤ |g|R for all R ≥ 0, where g is a
given transcendental function with g(0) 6= 0.
A simple measure for the growth of an entire function f is the real valued
function R 7→ |f |R , where
|f |R = sup |f (z)|.
|z|=R
An entire function f has an order of growth ≤ ̺ if for all ǫ > 0 the inequality
|f |R ≤ exp R̺+ǫ

holds for sufficiently large R.

As a very special case (selecting A to be the set Q of algebraic numbers and
each Eα,s to be either Q or its complement in C), one deduces the existence of
uncountably many algebraic independent transcendental entire functions f such
that any Taylor coefficient at any algebraic point α takes a prescribed value,
either algebraic or transcendental.
2.2 Integer–valued entire functions

In 1915, G. Pólya [68] initiated the study of integer–valued entire functions; he
proved that if f is a transcendental entire function such that f (n) ∈ Z for all
n ∈ Z≥0 , then
1
lim sup log |f |R > 0. (2.1)
R→∞ R
An example is the function 2z for which the lower bound is log 2. A stronger ver-
sion of the fact that 2z is the “smallest” entire transcendental function mapping
the positive integers to rational integers is the estimate
lim sup 2−R |f |R ≥ 1,

R→∞
which is valid under the same assumptions as for (2.1).

The method involves interpolation series: given an entire function f and a
sequence of complex numbers (αn )n≥1 , define inductively a sequence (fn )n≥0 of
entire functions by f0 = f and, for n ≥ 0,
fn (z) = fn (αn+1 ) + (z − αn+1 )fn+1 (z).
9
Define, for j ≥ 0,
Pj (z) = (z − α1 )(z − α2 ) · · · (z − αj ).
One gets an expansion
f (z) = A(z) + Pn (z)fn (z),
where
A = a0 + a1 P1 + · · · + an−1 Pn−1 ∈ C[z] and an = fn (αn+1 ) (n ≥ 0).
Conditions for such an expansion to be convergent as n → ∞ are known – see
for instance [32].
Such interpolation series give formulae for functions with given values at
the sequence of points αn ; when some of the αn ’s are repeated, these formulae
involve the successive derivatives of the function at the given point. For instance,
for a constant sequence αn = z0 for all n ≥ 1, one obtains the Taylor series
expansion of f at z0 . These formulae have been studied by I. Newton and J-L.
Lagrange.
Analytic formulae for the coefficients an and for the remainder fn follow from
Cauchy’s residue Theorem. Indeed, let x, z, α1 , . . . , αn be complex numbers
with x 6∈ {z, α1 , . . . , αn }. Starting from the easy relation
1 1 z − α1 1 ,
= + · (2.2)
x−z x − α1 x − α1 x − z
one deduces by induction the next formula due to Hermite:
n−1
1 X (z − α1 )(z − α2 ) · · · (z − αj )
=
x−z j=0
(x − α1 )(x − α2 ) · · · (x − αj+1 )
(z − α1 )(z − α2 ) · · · (z − αn ) 1
+ · ·
(x − α1 )(x − α2 ) · · · (x − αn ) x − z
Let D be an open disc containing α1 , . . . , αn , let C denote the circumference of
D, let D′ be an open disc containing the closure of D. Assume f is analytic in
D′ . Then
1 f (x)dx
Z
aj = (0 ≤ j ≤ n − 1)
2iπ C (x − α1 )(x − α2 ) · · · (x − αj+1 )
and
1 f (x)dx
Z
fn (z) = ·
2iπ C (x − α1 )(x − α2 ) · · · (x − αn )(x − z)
Pólya applies these formulae to prove that if f is an entire function which does
not grow too fast and satisfies f (n) ∈ Z for n ∈ Z≥0 , then the coefficients an in
the expansion of f at the sequence (αn )n≥1 = {0, 1, 2, . . .} vanish for sufficiently
large n, hence f is a polynomial.
Further works on this topic, using a variety of methods, are due to G.H. Hardy,
G. Pólya, D. Sato, E.G. Straus, A. Selberg, Ch. Pisot, F. Carlson, F. Gross,. . .
– and A.O. Gel’fond (see § 2.3).
10
2.3 Transcendence of eπ
Pólya’s study of the growth of transcendental entire functions taking integral
values at the positive rational integers was extended to the Gaussian integers
by A.O. Gel’fond in 1929 [30]. By means of interpolation series at the points
in Z[i], he proved that if f is a transcendental entire function which satisfies
f (α) ∈ Z[i] for all α ∈ Z[i], then
1
lim sup log |f |R ≥ γ. (2.3)
R→∞ R2
The example of the Weierstraß sigma function attached to the lattice Z[i]
Y z ωz + z22
σ(z) = z 1− e 2ω
ω
ω∈Z[i]\{0}
(which is the Hadamard canonical product with a simple zero at any point of
Z[i]), shows that the constant γ cannot be larger than π/2. Often for such
problems, dealing with a discrete subset of C, replacing integer values by zero
values gives some hint of what should be expected, at least for the order of
growth (the exponent 2 of R2 in the left hand side of formula (2.3)), if not for
the value of the constant (the number γ in the right hand side of formula (2.3)).
Other examples of Hadamard canonical products are
Y z z/n
z 1− e = −eγz Γ(−z)−1
n
n≥1
for the set Z>0 = {1, 2, . . . } of positive integers and

Y z z/n
z 1− e = π −1 sin(πz)
n
n∈Z\{0}
for the set Z of rational integers.

The initial admissible value computed by A.O. Gel’fond in 1929 for γ in (2.3)
was pretty small, namely γ = 10−45 . It was improved by several mathematicians
including Fukasawa, Gruman, Masser, until 1981 when F. Gramain [34] reached
γ = π/(2e), which is best possible, as shown by D.W. Masser [54] one year
earlier. See [1] for recent results and extensions to number fields.
This work of Gel’fond’s [30] turns out to have fundamental consequences on
the development of transcendental number theory, due to its connexion with
the number eπ . Indeed, the assertion that the number
eπ = 23, 140 692 632 779 269 005 729 086 367 . . .
is irrational is equivalent to saying that the function eπz cannot take all its values
in Q(i) when the argument z ranges over Z[i]. By expanding the function eπz
into an interpolation series at the Gaussian integers, Gel’fond was able to prove
the transcendence of eπ . More generally Gel’fond proved the transcendence of
11
αβ for α and β algebraic, α 6= 0, α 6= 1 and β imaginary quadratic. In 1930,
Kuzmin extended the proof
√
to the case where β is real quadratic, thus proving
the transcendence of 2 2 . The same year, Boehle proved that if β is algebraic
2 d−1
of degree d ≥ 2, then one at least of the d − 1 numbers αβ , αβ , . . . , αβ is
transcendental (see for instance [25]).
The next important step came from Siegel’s introduction of further ideas in
the theory (see § 3.1.2).
We conclude this subsection by noting that our knowledge of the Diophantine
properties of the number eπ is far from being complete. It was proved recently by
Yu.V. Nesterenko (see § 3.1.6) that the two numbers π and eπ are algebraically
independent, but there is no proof so far that eπ is not a Liouville number.
2.4 Lagrange interpolation

Newton-Lagrange interpolation (§ 2.2) of a function yields a series of polynomi-
als, namely linear combinations of products (z − α1 ) · · · (z − αn ). Another type
of interpolation has been devised in [40] by another Lagrange (René and not
Joseph–Louis), in 1935, who introduced instead a series of rational fractions.
Starting from the formula
1 α−β x−β z−α 1
= + · ·
x−z (x − α)(x − β) x − α z − β x − z
in place of (2.2), iterating and integrating as in § 2.2, one deduces an expansion

n−1
X (z − α1 ) · · · (z − αj )
f (z) = bj + Rn (z).
j=0
(z − β1 ) · · · (z − βj )
This approach has been developed in 2006 [69] by T. Rivoal, who applies it to
the Hurwitz zeta function
∞
X 1
ζ(s, z) = (s ∈ C, ℜe(s) > 1, z ∈ C).
(k + z)s
k=1
He expands ζ(2, z) as a Lagrange series in
z 2 (z − 1)2 · · · (z − n + 1)2
·
(z + 1)2 · · · (z + n)2
He shows that the coefficients of the expansion belong to Q + Qζ(3). This

enables him to produce a new proof of Apéry’s Theorem on the irrationality of
ζ(3).
Further, he gives a new proof of the irrationality of log 2 by expanding
∞
X (−1)k
k+z
k=1
12
into a Lagrange interpolation series. Furthermore, he gives a new proof of the
irrationality of ζ(2) by expanding the function
∞
X 1 1
−
k k+z
k=1
as a Hermite–Lagrange series in
2
z(z − 1) · · · (z − n + 1)
·
(z + 1) · · · (z + n)
It is striking that these constructions yield exactly the same sequences of rational
approximations as the one produced by methods which look very much different
[26].
Further developments of the interpolation methods should be possible. For
instance Taylor series are the special case of Hermite’s formula with a single
point and multiplicities — they give rise to Padé approximants. Multiplicities
could also be introduced in Lagrange–Rivoal interpolation.
3 Auxiliary functions arising from the Dirich-

let’s box principle
3.1 Thue–Siegel lemma
Here is a translation of a statement p. 213 of Siegel’s paper [81]:
Lemma 3.1 (Thue–Siegel). Let
y1 = a11 x1 + · · · + a1n xn
..
.
ym = am1 x1 + · · · + amn xn
be m linear forms in n variables with rational integer coefficients. Assume n >
m. Let A ∈ Z>0 be an upper bound for the absolute values of the mn coefficients
akl . Then the system of homogeneous linear equations y1 = 0, . . . , ym = 0 has
a solution in rational integers x1 , . . . , xn , not all of which are 0, with absolute
values less than 1 + (nA)m/(n−m) .
The fact that there is a non-trivial solution is a consequence of linear alge-
bra, thanks to the assumption n > m. The point here is that Dirichlet’s box
principe shows the existence of a non-trivial solution satisfying an explicit upper
bound. The estimate for that solution is essential in the proofs due to A. Thue,
C.L. Siegel, and later A.O. Gel’fond, Th. Schneider, A. Baker, W.M. Schmidt
and others (however, K. Mahler devised another method where such estimate
is not required, linear algebra suffices - see § 3.3).
The initial proof by Thue and Siegel relied on the box principle. More
sophisticated arguments have been introduced by K. Mahler, using geometry of
numbers, and they yield to a number of developments which we do not survey
here (see for instance [75]).
13
3.1.1 The origin of the Thue–Siegel Lemma
The first improvement of Liouville’s inequality was reached by A. Thue in 1909
[87, 88]. Instead of evaluating the values at p/q of a polynomial in a single
variable (viz. the irreducible polynomial of the given algebraic number α), he
considers two approximations p1 /q1 and p2 /q2 of α and evaluates at the point
(p1 /q1 , p2 /q2 ) a polynomial P in two variables. This polynomial P ∈ Z[X, Y ] is
constructed (or rather is shown to exist) by means of Dirichlet’s box principle
(Lemma 3.1). The required conditions are that P has zeroes of sufficiently large
multiplicity at (0, 0) and at (p1 /q1 , p2 /q2 ). The multiplicity is weighted: this is
what Thue called index of P at a point. The estimate for the coefficients of the
solution of the system of linear equations in Lemma 3.1 plays an important role
in the proof.
One of the main difficulties that Thue had to overcome was to produce a
zero estimate, in order to find a non–zero value of some derivative of P .
A crucial feature of Thue’s argument is that he needs to select a second
approximation p2 /q2 depending on a first one p1 /q1 . Hence a first very good
approximation p1 /q1 is required to produce an effective result from this method.
In general, such arguments lead to sharp estimates for all p/q with at most one
exception. This approach has been worked out by J.W.S. Cassels, H. Davenport
and others to deduce upper bounds for the number of solutions of certain Dio-
phantine equations. However, these results are not effective, meaning that they
do not yield complete solutions of these equations. More recently, E. Bombieri
has produced examples where a sufficiently good approximation exists for the
method to work in an effective way. Later, he produced effective refinements
to Liouville’s inequality by extending the argument (see [24] Chap. 1 § 5.4).
Further improvement of Thue’s method were obtained by C.L. Siegel in the
1920’s: he developed Thue’s method and succeeded in refining his irrationality
measure for algebraic real numbers. In 1929, Siegel [81], thanks to a further
sharpening of his previous estimate, derived his well known theorem on integer
points on curves: the set of integral points on a curve of genus ≥ 1 is finite.
The introduction of the fundamental memoir [81] of C.L. Siegel in 1929
stresses the importance of Thue’s idea involving the pigeonhole principle. In the
second part of this fundamental paper he extends the Lindemann–Weierstraß
Theorem (on the algebraic independence of eβ1 , . . . , eβn when β1 , . . . , βn are Q-
linearly independent algebraic numbers) from the usual exponential function to
a wide class of entire functions which he calls E-functions. He also introduces
the class of G-functions which has been extensively studied since 1929. See
also his monograph in 1949 [83], Shidlovskii’s book [80] and the Encyclopaedia
volume by Feldman and Nesterenko [24] for E–functions, André’s book [3] and
[24] Chap. 5 § 7 for G–functions. Among many developments related to G
functions are the works of Th. Schneider, V.G. Sprindzuck and P. Dèbes related
to algebraic functions (see [24] Chap. 5 § 7).
The work of Thue and Siegel on the rational approximation to algebraic
numbers was extended by many a mathematician, including Th. Schneider,
A.O. Gel’fond, F. Dyson, until K. F. Roth obtained in 1955 a result which is
14
essentially optimal. For his proof he introduces polynomials in many variables.
A powerful higher dimensional generalization of Thue–Siegel–Roth’s Theo-
rem, the Subspace Theorem, was obtained in 1970 by W.M. Schmidt [75]; see
also [7, 9]. Again, the proof involves a construction of an auxiliary polynomial in
several variables, and one of the most difficult auxiliary results is a zero estimate
(index theorem).
Schmidt’s Subspace Theorem together with its variants (including effective
estimates for the exceptional subspaces as well as results involving several val-
uations) have a large number of applications: to Diophantine approximation
and Diophantine equations, to transcendence and algebraic independence, to
the complexity of algebraic numbers – see [7].
We give here only a simplified statement of this fundamental result, which
is already quite deep and very powerful.
Theorem 3.2 (Schmidt’s Subspace Theorem - simplified form). For m ≥ 2

let L1 , . . . , Lm be independent linear forms in m variables with algebraic coeffi-
cients. Let ǫ > 0. Then the set
{x = (x1 , . . . , xm ) ∈ Zm ; |L1 (x) · · · Lm (x)| ≤ |x|−ǫ }
is contained in the union of finitely many proper subspaces of Qm .
3.1.2 Siegel, Gel’fond, Schneider

In 1932, C.L. Siegel [82] obtained the first results on the transcendence of elliptic
integrals of the first kind (a Weierstrass elliptic function cannot have simulta-
neously algebraic periods and algebraic invariants g2 , g3 ), by means of a very
ingenious argument which involved an auxiliary function whose existence fol-
lows from the Dirichlet’s box principle. This idea turned out to be crucial in
the development of transcendental number theory.
The seventh of the 23 problems raised by D. Hilbert in 1900 is to prove
the transcendence of the numbers αβ for α and β algebraic (α 6= 0, α 6= 1,
β 6∈ Q). In this statement, αβ stands for exp(β log α), where log α is any2
logarithm of α. The solution was achieved independently by A.O. Gel’fond [31]
and Th. Schneider√
[76] in 1934. Consequences, already quoted by Hilbert, are
2
the facts that
√
2 and eπ are transcendental. The question of the arithmetic
nature of 2 2 was considered by L. Euler in 1748 [22].
The proofs by Gel’fond and Schneider are different, but both of them rest
on some auxiliary function which arises from Dirichlet’s box principle, following
Siegel’s contribution to the theory.
Let us argue by contradiction and assume that α, β and αβ are all algebraic,
with α 6= 0, α 6= 1, β 6∈ Q. Define K = Q(α, β, αβ ). By assumption, K is a
number field.
2 The assumption α 6= 1 can be replaced by the weaker assumption log α 6= 0. That means
that one can take α = 1, provided that we select for log α a non–zero multiple of 2iπ The
result allowing α = 1 is not more general: it amounts to the same to take α = −1, provided
that one replaces β by 2β.
15
A.O. Gel’fond’s proof [31] rests on the fact that the two entire functions ez
and eβz are algebraically independent, they satisfy differential equations with
algebraic coefficients and they take simultaneously values in K for infinitely
many z, viz. z ∈ Z log α.
Th. Schneider’s proof [76] is different: he notices that the two entire functions
z and αz = ez log α are algebraically independent, they take simultaneously
values in K for infinitely many z, viz. z ∈ Z + Zβ. He makes no use of
differential equations, since the coefficient log α which occurs by derivating the
function αz is not algebraic.
Schneider introduces a polynomial A(X, Y ) ∈ Z[X, Y ] in two variables and
considers the auxiliary function
F (z) = A(z, αz )
at the points m + nβ: these values γmn are in the number field K = Q(α, β, αβ ).
Gel’fond also introduces also a polynomial A(X, Y ) ∈ Z[X, Y ] in two vari-
ables and considers the auxiliary function
F (z) = A(ez , eβz );
the values γmn at the points m log α of the derivatives F (n) (z) are again in the
number field K.
With these notations, the proofs are similar: the first step is the existence
of a non–zero polynomial A, of partial degrees bounded by L1 and L2 , say,
such that the associated numbers γmn vanish for certain values of m and n, say
0 ≤ m < M , 0 ≤ n < N . This amounts to showing that a system of linear
homogeneous equations has a non–trivial solution; linear algebra suffices for the
existence. In this system of equations, the coefficients are algebraic numbers
in the number field K, the unknowns are the coefficients of the polynomial A,
and we are looking for a solution in rational integers. There are several options
at this stage: one may either require only coefficients in the ring of integers
of K, in which case the assumption L1 L2 > M N suffices. An alternative way
is to require the coefficients to be in Z, in which case one needs to assume
L1 L2 > M N [K : Q].
This approach is not quite sufficient for the next steps: one will need es-
timates for the coefficients of this auxiliary polynomial A. This is where the
Thue–Siegel Lemma occurs into the picture: by assuming that the number of
unknowns, namely L1 L2 , is slightly larger than the number of equations, say
twice as large, this lemma produces a bound, for a non–trivial solution of the
homogeneous linear system, which is sharp enough for the rest of the proof.
The second step is an induction: one proves that γmn vanishes for further
values of (m, n). Since there are two parameters (m, n), there are several options
for this extrapolation (increasing m, or n, or m+n for instance), but anyway the
idea is that if F has sufficiently many zeroes, then F takes rather small values on
some disc (Schwarz Lemma), and so do its derivatives (Cauchy’s inequalities).
Further, an element of K which is sufficiently small should vanish (by a Liouville
16
type inequality, or a so-called size inequality, or else the product formula – see
for instance [41, 9, 24, 104] among many references on this topic).
For the last step, there are also several options: one may perform the induc-
tion with infinitely many steps and use an asymptotic zero estimate, or else stop
after a finite number of steps and prove that some determinant does not vanish.
The second method is more difficult and this is the one Schneider succeeded to
complete, but his proof can be simplified by pursuing the induction forever.
There is a duality between the two methods. In Gel’fond ’s proof, replace L1
and L2 by S1 and S2 , and replace M and N by T0 and T1 ; hence the numbers
γmn which arise are
t0
d
e(s1 +s2 β)z z=t1 log α

dz
while in Schneider’s proof replacing L1 and L2 by T1 and T2 , and replacing M
and N by S0 and S1 , then the numbers γmn which arise are
z t0 αt1 z z=s1 +s2 β

It is easily seen that they are the same, namely
(s1 + s2 β)t0 αt1 s1 (αβ )t1 s2 .
See [99] and § 13.7 of [104].

Gel’fond–Schneider Theorem was extended in 1966 by A. Baker [4], who
proved the more general result that if log α1 , . . . , log αn are Q–linearly inde-
pendent logarithms of algebraic numbers, then the numbers 1, log α1 , . . . , log αn
are linearly independent over Q. The auxiliary function used by Baker may be
considered as a function of several variables or as a function of a single com-
plex variable, depending on the point of view (cf. [95]). The analytic estimate
(Schwarz lemma) involves merely a single variable. The differential equations
can be written with a single variable with transcendental coefficients. By intro-
ducing several variables, only algebraic coefficients occur. See also § 3.1.4.
Indeed, assume that α1 , . . . , αn , αn+1 , β0 , . . . , βn are algebraic numbers which
satisfy
β0 + β1 log α1 + βn log αn = log αn+1
for some specified values of the logarithms of the αj . Then the n + 2 functions
of n variables
z0 , ez1 , . . . , ezn , eβ0 z0 +β1 z1 +···+βn zn
satisfy differential equations with algebraic coefficients and take algebraic values
at the integral multiples of the point
(1, log α1 , . . . , log αn ) ∈ Cn .
This situation is therefore an extension of the setup in Gel’fond’s solution of

Hilbert’s seventh problem, and Baker’s method can be viewed as an exten-
sion of Gel’fond’s method. The fact that all points are on a complex line
17
C(1, log α1 , . . . , log αn ) ⊂ Cn means that Baker’s method requires only tools
from the theory of one complex variable.
On the other hand, the corresponding extension of Schneider’s method re-
quires several variables: under the same assumptions, consider the functions
z0 , z1 , . . . , zn , ez0 α1z1 · · · αnzn
and the points in the subgroup of Cn+1 generated by
{0} × Zn + Z(β0 , β1 , . . . , βn ).

Since Baker’s Theorem includes the transcendence of e, there is no hope to prove

it without introducing the differential equation of the exponential function - in
order to use the fact that the last function involves e and not another number
in the factor ez0 , we also take derivatives with respect to z0 . For this method,
we refer to [104]. The duality between Baker’s method and Schneider’s method
in several variables is explained below in § 3.2.
3.1.3 Schneider–Lang Criterion

In 1949, [78] Th. Schneider produced a very general statement on algebraic
values of analytic functions, which can be used as a principle for proofs of
transcendence. This statement includes a large number of previously known
results like the Hermite–Lindemann and Gel’fond–Schneider Theorems. It also
contains the so–called Six exponentials Theorem 3.8 (which was not explicitly in
the literature then). To a certain extent, such statements provide partial answers
to Weierstraß question (see § 2.1) that exceptional sets of a transcendental
function are not too large; here one puts restrictions on the functions, while in
Pólya’s work concerning integer–valued entire functions, the assumptions were
mainly on the points and the values (the mere condition on the functions were
that they have a finite order of growth).
A few years later, in his book [79] on transcendental numbers, Schneider gave
variants of this statement, which lose some generality but gained in simplicity.
Further simplifications were introduced by S. Lang in 1964 and the state-
ment which is reproduced in his book on transcendental numbers [41] is the
so-called Criterion of Schneider–Lang (see also the appendix of [42] as well as
[95] Th. 3.3.1).
Theorem 3.3. Let K be a number field and f1 , . . . , fd be entire functions in C.

Assume that f1 and f2 are algebraically independent over K and have finite order
of growth. Assume also that they satisfy differential equations: for 1 ≤ i ≤ d,
assume fi′ is a polynomial in f1 , . . . , fd with coefficients in K. Then the set S
of w ∈ C such that all fi (w) are in K is finite3 .
3 For simplicity, we consider only entire functions; the results extends to meromorphic
functions, and this is important for applications, for instance to elliptic functions. To deal
with functions which are analytic in a disc only is also an interesting issue [90, 91, 92, 93, 94].
18
This statement includes the Hermite–Lindemann Theorem on the transcen-
dence of eα : take
K = Q(α, eα ), f1 (z) = z, f2 (z) = ez , S = {mα ; m ∈ Z},
as well as the Gel’fond –Schneider Theorem on the transcendence of αβ following

Gel’fond’s method: take
K = Q(α, β, αβ ), f1 (z) = ez , f2 (z) = eβz , S = {m log α ; m ∈ Z}.
This criterion 3.3 does not include some of the results which are proved by
means of Schneider’s method (for instance it does not contain the Six Expo-
nentials Theorem 3.8), but there are different criteria (not involving differential
equations) for that purpose (see for instance [41, 95, 104]).
Here is the idea of the proof of the Schneider–Lang Criterion 3.3. We argue
by contradiction: assume f1 and f2 take simultaneously their values in the
number field K for different values w1 , . . . , wS ∈ C. We want to show that there
exists a non–zero polynomial P ∈ Z[X1 , X2 ] such that the function P (f1 , f2 )
is the zero function: this will contradict the assumption that f1 and f2 are
algebraically independent.
The first step is to show that there exists a non–zero polynomial P ∈
Z[X1 , X2 ] such that F = P (f1 , f2 ) has a zero of high multiplicity, say ≥ T ,
at each ws , (1 ≤ s ≤ S): we consider the system of ST homogeneous linear
equations
t
d
F (ws ) = 0 for 1 ≤ s ≤ S and 0 ≤ t < T (3.4)
dz
where the unknowns are the coefficients of P . If we require that the partial
degrees of P are strictly less than L1 and L2 respectively, then the number
of unknowns is L1 L2 . Since we are looking for a polynomial P with rational
integer coefficients, we need to introduce the degree [K : Q] of the number field
K. As soon L1 L2 > T S[K : Q], there is a non–trivial solution. Further, the
Thue–Siegel Lemma produces an upper bound for the coefficients of P .
The next step is an induction: the goal is to prove that F = 0. By induction
on T ′ ≥ T , one proves
t
d
F (ws ) = 0 for 1 ≤ s ≤ S and 0 ≤ t < T ′ .
dz
One already knows that it holds for T ′ = T by (3.4).
The proof of this induction is the same as the ones by Gel’fond and Schneider
of the transcendence of αβ , (see § 3.1.2), combining an analytic upper bound
(Schwarz Lemma) and an arithmetic lower bound (Liouville’s inequality). At
the end of the induction, one deduces F = 0, which is the desired contradiction
with the algebraic independence of f1 and f2 .
As we have seen in § 3.1.2, such a scheme of proof is characteristic of the
Gel’fond–Schneider’s method.
19
The main analytic argument is Schwarz Lemma for functions of one variable,
which produces an upper bound for the modulus of an analytic function having
many zeroes. One also requires Cauchy’s inequalities in order to bound the
moduli of the derivatives of the auxiliary function.
In this context, a well known open problem raised by Th. Schneider (this is
the second in the list of his 8 problems from his book [79]) is related with his
proof of the transcendence of j(τ ), where j is the modular function defined in the
upper half plane ℑm(z) > 0 and τ is an algebraic point in this upper half plane
which is not imaginary quadratic. Schneider himself proved the transcendence
of j(τ ), but his proof is not direct, it rests on the use of elliptic functions (one
may apply the Schneider–Lang Criterion for meromorphic functions). There-
fore, the question is to prove the same result by using modular functions. In
spite of recent progress on transcendence and modular functions (see § 3.1.6),
this problem is still open. The difficulty lies in the analytic estimate and the
absence of a suitable Schwarz Lemma - the best results on this topic are due to
I.Wakabayashi [90, 91, 92, 93, 94].
3.1.4 Higher dimension: several variables

In 1941, Th. Schneider [77] obtained an outstanding result on the values of
Euler’s Gamma and Beta functions: for any rational numbers a and b such that
none of a, b and a + b is an integer, the number
Γ(a)Γ(b)
B(a, b) =
Γ(a + b)
is transcendental.
His proof involves a generalization of Gel’fond’s method to several variables
and yields a general transcendence criterion for functions satisfying differential
equations with algebraic coefficients and taking algebraic values at the points of
a large Cartesian product. He applies this criterion to the theta functions asso-
ciated with the Jacobian of the Fermat curves. His transcendence results apply
more generally to yield transcendence results on periods of abelian varieties.
After a suggestion of P. Cartier, S. Lang extended the classical results on
the transcendence of the values of the classical exponential function to the ex-
ponential function of commutative algebraic groups. During this process, he
generalized the one dimensional Schneider–Lang criterion 3.3 to several vari-
ables [41] Chap. IV. In this higher dimensional criterion, the conclusion is that
the set of exceptional values in Cn cannot contain a large Cartesian product.
This was the generalization to several variables of the fact that in the one di-
mensional case the exceptional set is finite (with a bound for the number of
elements).
A simplified version of the Schneider–Lang Criterion in several variables for
Cartesian products is the following ([104] Theorem 4.1).
Theorem 3.5 (Schneider–Lang Criterion in several variables). Let d and n be
two integers with d > n ≥ 1, K be a number field, and f1 , . . . , fd be algebraically
20
independent entire functions of finite order of growth. Assume, for 1 ≤ ν ≤
n and 1 ≤ i ≤ d, that the partial derivative (∂/∂zν )fi of fi belongs to the
ring K[f1 , . . . , fd ]. Further, let (y 1 , . . . , y n ) be a basis of Cn over C. Then the
numbers
1 ≤ i ≤ d, (s1 , . . . , sn ) ∈ Zn

fi (s1 y 1 + · · · + sn y n ),
do not all belong to K.

Besides the consequences already derived by Schneider in 1941, Lang gave
further consequences of this result to commutative algebraic groups, especially
abelian varieties [41] Chap. IV and [98] Chap. 5.
It is interesting from an historical point of view to notice that Bertrand and
Masser [6] succeeded in 1980 to deduce Baker’s Theorem form the Schneider–
Lang Criterion 3.5 (see [104] § 4.2). They could also prove the elliptic analog
of Baker’s result and obtain the linear independence, over the field of algebraic
numbers, of elliptic logarithms of algebraic points - at that time such a result
was available only in the case of complex multiplication (by previous work of
D.W. Masser).
According to [41], Historical Note of Chap. IV, M. Nagata suggested that in
the higher dimensional version of the Schneider–Lang Criterion the conclusion
could be that the exceptional set of points where all the functions take simul-
taneously values in a number field K is contained in an algebraic hypersurface,
the bound for the number of points being replaced by a bound for the degree of
the hypersurface. This program was fulfilled by E. Bombieri in 1970 [8].
Theorem 3.6 (Bombieri). Let d and n be two integers with d > n ≥ 1, K be
a number field, and f1 , . . . , fd be algebraically independent entire functions of
finite order of growth. Assume, for 1 ≤ ν ≤ n and 1 ≤ i ≤ d, that the partial
derivative (∂/∂zν )fi of fi belongs to the ring K[f1 , . . . , fd ]. Then the set of
points w ∈ Cn where the d functions f1 , . . . , fd all take values in K is contained
in an algebraic hypersurface.
Bombieri produces an upper bound for the degree of such an hypersurface
(see also [98] Th. 5.1.1). His proof [8] involves different tools, including L2 –
estimates by L. Hörmander for functions of several variables. One main difficulty
that Bombieri had to overcome was to generalize Schwarz Lemma to several
variables, and his solution involves an earlier work by E. Bombieri and S. Lang
[10], where they use Lelong’s theory of the mass of zeroes of analytic functions
in a ball. Chapter 7 of [98] is devoted to this question. The next statement
(Proposition 3.7) follows from the results in Chapter 7 of [98]. Given a finite
subset S of Cn and an integer t ≥ 1, one defines ωt (S) as the smallest degree of
a polynomial having a zero at each point of S of multiplicity at least t. Then
the sequence ω(t)/t has a limit Ω(S) which satisfies
1 1
ωt (S) ≤ Ω(S) ≤ ωt (S) ≤ ω1 (S)
t+n−1 t
for all t ≥ 1.
21
Proposition 3.7 (Schwarz Lemma in several variables). Let S be a finite subset
of Cn and ǫ be a positive real number. There exists a positive real number
r0 = r0 (S, ǫ) such that, for any positive integer t, any real numbers R and r
with R > r ≥ r0 and any entire function having a zero of multiplicity ≥ t at
each point of S,
t(Ω(S)−ǫ)
4nr
|f |r ≤ |f |R .
R
More recent results on Schwarz Lemma in several variables are due to D. Roy
[70, 71, 72, 73].
3.1.5 The six exponentials Theorem

Here is the six exponentials Theorem, due to Siegel, Lang and Ramachandra
[41, 95, 104, 98].
Theorem 3.8 (Six exponentials Theorem). Let x1 , . . . , xd be complex numbers
which are linearly independent over Q and let y1 , . . . , yℓ be also complex numbers
which are linearly independent over Q. Assume dℓ > d + ℓ. Then one at least
of the numbers
exi yj , (1 ≤ i ≤ d, 1 ≤ j ≤ ℓ) (3.9)
is transcendental.
The condition dℓ > d + ℓ in integers d and ℓ means that the relevant cases
are d = 2 and ℓ = 3 or d = 3 and ℓ = 2 (there is a symmetry), hence the name
(since ℓd = 6 in these cases). We refer to [41, 95] for more information on this
topic.
The classical proof by Schneider’s method involves an auxiliary function of
the form
F (z) = P (ex1 z , . . . , exd z ),
where the existence of the polynomial P follows from Dirichlet’s box principle
and the Thue–Siegel Lemma. The conditions which are required are that F
vanishes at many points of the form s1 y1 + . . . + sℓ yℓ with varying integers
s1 , . . . , sℓ , the induction shows that F vanishes at more points of this form,
until one deduces that F vanishes at all such points, and the conclusion then
easily follows. One could avoid an infinite induction by using a zero estimate.
A variant is to use an interpolation determinant (see § 4 and [104] Chap. 2).
As an example of upper bound for an interpolation determinant, here is the
statement of Lemma 2.8 in [104].
Lemma 3.10. Let ϕ1 , . . . , ϕL be entire functions in C, ζ1 , . . . , ζL be elements

of C, σ1 , . . . , σL nonnegative integers, and 0 < r ≤ R be real numbers, with
|ζµ | ≤ r (1 ≤ µ ≤ L). Then the absolute value of the determinant

d σµ
∆ = det ϕλ (ζµ )
dz 1≤λ,µ≤L
22
is bounded from above by
−L(L−1)/2+σ1 +···+σL L σµ
R Y d
|∆| ≤ L! max sup
ϕλ (z) .
r 1≤µ≤L |z|=R dz
λ=1
Another method of proof of the six exponentials Theorem 3.8 is proposed

in [101, 103]. The starting idea is to consider the function of two complex
variables ezw . If all numbers in (3.9) are algebraic, then this function takes
algebraic values (in fact in a number field) at all points (z, w) where z is of
the form t1 x1 + · · · + td xd and w of the form s1 y1 + . . . + sℓ yℓ with integers
t1 , . . . , td , s1 , . . . , sℓ . The method does not work like this: a single function ezw
in two variables does not suffice, one needs several functions. For this reason
one introduces redundant variables. Letting zh and wk be new variables, one
investigates the values of the functions are ezh wk at the points of Cartesian
products.
Such a device has been introduced in transcendental number theory in 1981
by P. Philippon [65] for giving a proof of an algebraic independence result an-
nounced by G.V. Chudnovskii - at that time the original proof was very com-
plicated and the approach by Philippon, using a transcendence criterion from
É. Reyssat, was a dramatic simplification. Now much sharper results are known.
In 1981, introducing several variables was called Landau’s trick, which is a ho-
mogeneity argument: letting the number of variables tend to infinity enables
one to kill error terms.
The generalisation of the six exponentials Theorem to higher dimension, also
with multiplicities, is one of the main topics of [104].
Further auxiliary functions occur in the works on algebraic independence by
A.O. Gel’fond [33], G.V. Chudnovskii and others. A reference is [63].
3.1.6 Modular functions

The solution by the team at St Etienne [5] of the problems raised by Mahler
and Manin on the transcendence of the values of the modular function J(q) for
algebraic values of q in the unit disc4 involves an interesting auxiliary function:
the general scheme of proof is the one of Gel’fond and Schneider. They construct
their auxiliary function by means of the Thue–Siegel Lemma as a polynomial
in z and J(z) having a high multiplicity zero at the origin, say ≥ L. They
consider the exact order of multiplicity M , which by construction is ≥ L, and
they bound |z −M F (z)| in terms of |z| (this amounts to using the easiest case of
Schwarz Lemma in one variable with a single point). They apply this estimate
to z = q S where S is the smallest integer for which F (q S ) 6= 0. Liouville’s
estimate gives the conclusion. See also [63], Chap. 2.
A similar construction was performed by Yu. V. Nesterenko in 1996 (see
[61] and [63] Chap. 3), when he proved the algebraic independence of π, eπ
and Γ(1/4): his main result is that for any q in the open set 0 < |q| < 1, the
4 The connexion between J(q) and j(τ ) is j(τ ) = J(e2iπτ ).
23
transcendence degree of the field

Q q, P (q), Q(q), R(q)
is at least 3. Here, P , Q, R are the classical Ramanujan functions, which are

sometimes denoted as E2 , E4 and E6 (Eisenstein series). The auxiliary function
F is a polynomial in the four functions z, E2 (z), E4 (z), E6 (z) ; like for the
théorème stéphanois on the transcendence of J(q), it is constructed by means
of the Thue–Siegel Lemma so that it has a zero of large multiplicity, say M .
In order to apply a criterion for algebraic independence, Nesterenko needs to
establish an upper bound for M , and this is not an easy result. An alternative
argument due to Philippon (see [63] Chap. 4) is to apply a measure of algebraic
independence of Faisant and Philibert [23] on numbers of the form ω/π and
η/π.
3.2 Universal auxiliary functions

3.2.1 A general existence theorem
In the Gel’fond–Schneider method, the auxiliary function is constructed by
means of the Thue–Siegel Lemma, and the requirement is that it has many
zeroes (multiplicity are there in Gel’fond’s method, not in Schneider’s method).
There is an alternative construction which was initiated in a joint work with
M. Mignotte in 1974 [60], in connexion with quantitative statements related with
transcendence criteria like the Schneider–Lang criterion. This approach turned
out to be specially efficient in another context, namely in extending Schneider’s
method to several variables [96]. The idea is to require that the auxiliary func-
tion F has small Taylor coefficients at the origin; it follows that its modulus on
some discs will be small, hence its values (including derivatives, if one wishes)
at points in such a disc will also be small. Combining Liouville’s estimate with
Cauchy’s inequality for estimating the derivatives, one deduces that F has a lot
of zeroes, more than would be reached by the Dirichlet’s box principle. At this
stage there are several options; the easy case is when a sharp zero estimate is
known: we immediately reach the conclusion without any further extrapolation:
in particular there is no need of Schwarz Lemmas in several variables. This is
what happend in [96] for exponential functions in several variables, the zero
estimate being due to D.W. Masser [55]. This result, dealing with products of
multiplicative groups (tori), can be extended to commutative algebraic groups
[97], thanks to the zero estimate of Masser and Wüstholz [57, 58].
A zero estimate is a statement which gives a lower bound for the degree of a
polynomial which vanishes at given points (when multiplicities are introduced
this is sometimes calles a multiplicity estimate). D.W. Masser developed the
study of zero estimates, and P. Philippon, G. Wüstholz, Yu.V. Nesterenko were
among those who contributed to the theory. Introductions to this subject have
been written by D. Roy (Chapters 5 and 8 of [104] and Chapter 11 of [63]). Also
Chapter 5 of [4] is devoted to multiplicity estimates.
24
A Schwarz Lemma is a statement which gives an upper bound for the max-
imum modulus of an analytic function which vanishes at given points - multi-
plicities may be there. We gave an example in Proposition 3.7.
When the function has only small values at these points, instead of zeoes,
one speaks either of a small value Lemma or of an approximate Schwarz Lemma
[66].
A Schwarz Lemma implies a zero estimate, and a small value Lemma implies
a Schwarz Lemma. However, the assumptions for obtaining a small value Lemma
for instance are usually stronger than for only a Schwarz Lemma: as an example,
in one variable there is no need to introduce an assumption on the distance of
the given points for a Schwarz Lemma, while for a small value Lemma it is
requested.
This construction of universal auxiliary functions is developed in [99] (see
Lemme 2.1) and [100].
Here is Proposition 4.10 of [104].
Proposition 3.11. Let L and n be positive integers, N , U , V , R, r positive
real numbers and ϕ1 , . . . , ϕL entire functions in Cn . Define W = N + U + V
and assume
L
R X
W ≥ 12n2 , e ≤ ≤ eW/6 , |ϕλ |R ≤ eU
r
λ=1
and n
(2W )n+1 ≤ LN log(R/r) .
Then there exist rational integers p1 , . . . , pL , with
0 < max |pλ | ≤ eN ,
1≤λ≤L
such that the function F = p1 ϕ1 + · · · + pL ϕL satisfies

|F |r ≤ e−V .
For an application to algebraic independence, see [63] Prop. 3.3 Ch. 14.
3.2.2 Duality
In the papers [99] (see Lemme 2.1) and [100], a dual construction is performed,
where auxiliary analytic functionals are constructed, and the duality between the
methods of Schneider and Gel’fond can be explained by means of the Fourier–
Borel transform (see [99], especially Lemmas 3.1 and 7.6, and [104], § 13.7). In
the special case of exponential polynomials, it reduces to the relation
Dσ z τ etz (s) = Dτ (z σ esz ) (t),

(3.12)
where σ, τ , s, t, z stand for tuples
σ = (σ1 , . . . , σn ) ∈ Zn≥0 , τ = (τ1 , . . . , τn ) ∈ Zn≥0 ,
s = (s1 , . . . , sn ) ∈ Cn , t = (t1 , . . . , tn ) ∈ Cn ,
z = (z1 , . . . , zn ) ∈ Cn
25
and σ1 σn
∂ ∂
Dσ = ···
∂z1 ∂zn
(see [99] Lemma 3.1 and [104] Corollary 13.21), which is a generalization in
several variables of the formula
σ τ
d τ tz d
(z σ esz ) (t)

z e (s) =
dz dz
for t, s in C and σ, τ non–negative integers, both sides being

min{τ,σ}
X σ!τ !
tσ−k sτ −k est .
k!(τ − k)!(σ − k)!
k=0
Another (less special) case of (3.12) is the duality between Schneider’s method
in several variables and Baker’s method. In Baker’s method one considers the
values at (t, t log α1 , . . . , t log αn ) of
τ0 τn
∂ ∂
z0σ es1 z1 · · · esn zn es0 (β0 z0 +···+βn zn ) ,

···
∂z0 ∂zn
while Schneider’s method in several variables deals with the values of

σ
∂
z0τ0 z1τ1 · · · znτn (ez0 α1z1 · · · αnzn )t

∂z0
at the points
(s0 β0 , s1 + s0 β1 , . . . , sn + s0 βn ),
and these values are just the same, namely
min{τ0 ,σ}
X σ!τ0 !
tσ−k (s0 β0 )τ0 −k (s1 +s0 β1 )τ1 · · · (sn +s0 βn )τn (α0s0 · · · αnsn )t
k!(τ0 − k)!(σ − k)!
k=0
with
α0 = eβ0 α1β1 · · · αnβn .
This is a special case of (3.12) with
τ = (τ0 , . . . , τn ) ∈ Zn+1
≥0 , σ = (σ0 , 0, . . . , 0) ∈ Zn+1
≥0 ,
t = (t, t log α1 , . . . , t log αn ) ∈ Cn+1 , s = (s0 β0 , s1 +s0 β1 , . . . , sn +s0 βn ) ∈ Cn+1 ,

z = (z0 , . . . , zn ) ∈ Cn+1 .
The construction of universal auxiliary functions is one of the tools in D. Roy’s
approach to Schanuel’s Conjecture in [70, 71].
26
3.3 Mahler’s Method
In 1929, K. Mahler [50] developed an original method to prove the transcendence
of values of functions satisfying certain types of functional equations. This
method was somehow forgotten, for instance it is not quoted among the 440
references of the survey paper [25] by N.I. Fel’dman, and A.B. Šidlovskiı̆. After
the publication of the paper [52] by Mahler, several mathematicians (including
K. Kubota, J.H. Loxton and A.J. van der Poorten) extended the method (see the
Lecture Notes [64] by K. Nishioka for further references). The construction of
the auxiliary function is similar to what is done in Gel’fond–Schneider’s method,
with a main difference: in place of the Thue–Siegel Lemma, only linear algebra
is required. No estimate for the coefficients of the auxiliary polynomial is needed
in Mahler’s method.
The following example is taken from § 1.1 of [64]. Let d ≥ 2 be a rational
integer and α an algebraic number with 0 < |α| < 1. Let us prove that the
number
∞
X k
αd
k=0
is transcendental, a result due to K. Mahler [50]. The basic remark is that the
function
∞
X k
f (z) = zd
k=0
d
satisfies the functional equation f (z ) = f (z) − z. It is not difficult to check
also that f is a transcendental function, which means that if P is a non–zero
polynomial in C[X, Y ], then the function F (z) = P z, f (z) is not the zero
function. From linear algebra, it follows that if L is a sufficiently large integer,
there exist a non–zero polynomial P in Z[X,
Y ] of partial degrees ≤ L such that
the associated function F (z) = P z, f (z) has a zero at the origin of multiplicity
> L2 . Indeed, the existence of P amounts to showing the existence of a non–
trivial solution to a system of L2 +1 homogeneous linear equations with rational
coefficients in (L + 1)2 unknowns.
If T denotes the multiplicity of F at the origin, then the limit
k k
lim F (αd )α−d T
k→∞
k
is a non–zero constant. This produces an upper bound for |F (αd )| when k is a
sufficiently large positive integer, and this upper bound is not compatible with
Liouville’s inequality. Hence f (α) is transcendental.
Another instructive example of application of Mahler’s method is given by
D.W. Masser in the first of his Cetraro’s lectures [56].
The numbers whose transcendence is proved by this method are not Liou-
ville numbers, but they are quite well approximated by algebraic numbers. In
the example we discussed, the number f (α) is very well approximated by the
27
algebraic numbers
K
X k
αd , (K > 0).
k=0
4 Interpolation determinants
An interesting development of the saga of auxiliary functions took place in
1991, thanks to the introduction of interpolation determinants by M. Laurent.
Its origin goes back to earlier works on a question raised by Lehmer which we
first introduce.
4.1 Lehmer’s Problem

Let θ be a non–zero algebraic integer of degree d. Mahler’s measure of θ is
d
Y Z 1
2iπt
M (θ) = max(1, |θi |) = exp log |f (e |dt ,
i=1 0
where θ = θ1 and θ2 , · · · , θd are the conjugates of θ and f the monic irreducible

polynomial of θ in Z[X].
From the definition one deduces M (θ) ≥ 1. According to a well-known and
easy result of Kronecker, M (θ) = 1 if and only if θ is a root of unity.
D.H. Lehmer [46] asked whether there is a constant c > 1 such that M (θ) < c
implies that θ is a root of unity.
Among many tools which have been introduced to answer this question,
we only quote some of them which are relevant for our concern. In 1977, M.
Mignotte [59] used ordinary Vandermonde determinants to study algebraic num-
bers whose conjugates are close to the unit circle. In 1978, C.L. Stewart [85]
sharpend earlier results by Schinzel and Zassenhaus (1965) and Blanksby and
Montgomery (1971) by introducing an auxiliary function (whose existence fol-
lows from the Thue–Siegel Lemma) and using an extrapolation similar to what
is done in the Gel’fond–Schneider method.
Refined estimates were obtained by E. Dobrowolski in 1979 [21] using Stew-
art’s approach together with congruences modulo p. He achieved the best un-
conditional result known so far in this direction (apart from some marginal
improvements on the numerical value for c: There is a constant c such that, for
θ a non–zero algebraic integer of degree d,
M (θ) < 1 + c(log log d/ log d)3
implies that θ is a root of unity.

In 1982, D. Cantor and E.G. Straus [15] revisited this method of Stewart and
Dobrowolski by replacing the auxiliary function by a generalised Vandermonde
determinant. The sketch of proof is the following: in Dobrowolski’s proof, there
is a zero lemma which can be translated into a statement that some matrix
28
has a maximal rank; therefore some determinant is not zero. This determinant
is bounded from above by means of Hadamard’s inequality; the upper bound
depends on M (θ). On the other hand, the absolute value ofQthis determinant
is shown to be big, because it has many factors of the form i,j |θip − θj |k , for
many primes p. The lower bound makes use of a Lemma due to Dobrowolski:
For θ not a root of unity, Y p
|θi − θj | ≥ pd
i,j
for any prime p.

One may also prove the lower bound by means of a p–adic Schwarz Lemma:
a function (here merely a polynomial) with many zeroes has a small (p–adic)
absolute value. In this case the method is similar to the earlier one, with analytic
estimates on one side and arithmetic ones (Liouville type, or product formula)
on the other. See [104] § 3.6.5 and § 3.6.6.
Dobrowolski’s result has been extended to several variables by F. Amoroso
and S. David in [2] – the higher dimensional version is much more involved.
We refer to S. David’s survey [20] for further references on this topic. We only
notice the role of the generalized Dirichlet exponents ωt and Ω (see [98] § 1.3
and Chap. 7), which occurred in transcendence statements in connexion with
multidimensional Schwarz Lemmas (cf. Proposition 3.7).
4.2 Laurent’s interpolation determinants

In 1991, M. Laurent [43] discovered that one may get rid of the Dirichlet’s box
principle in Gel’fond–Schneider’s method by means of his interpolation determi-
nants. In the classical approach, there is a zero estimate (or vanishing estimate,
also called multiplicity estimate when derivatives are there) which shows that
some auxiliary function cannot have too many zeroes. This statement can be
converted into the non–vanishing of some determinant. Laurent works directly
with this determinant: a Liouville-type estimate produces a lower bound, the
remarkable fact is that analytic estimates like Schwarz Lemma produce sharp
upper bounds. Again, the analytic estimates depend only in one variable (even
if the determinant is a value of a function in many variables, it suffices to restrict
this function to a complex line – see [104] § 6.2). Therefore this approach is spe-
cially efficient when dealing with functions of several variables, where Schwarz
Lemmas are lacking.
Baker’s method is introduced in [104] § 10.2 with interpolation determinants
and in § 10.3 with auxiliary functions.
Interpolation determinants are easy to use when a sharp zero estimate is
available. If not, it is more tricky to prove the analytic estimate. However,
it is possible to perform extrapolation in the transcendence methods involving
interpolation determinants: an example is a proof of Pólya’s theorem (2.1) by
means of interpolation determinants [102].
Proving algebraic independence results by means of interpolation determi-
nants was quoted as an open problem in [99] p. 257: at that time (1991) aux-
29
iliary functions (or auxiliary functionals) were required, together with a zero
estimate (or an interpolation estimate). The first solution to this problem in
1997 [74] (see also [104] Corollary 15.10) makes a detour via measures of simulta-
neous approximations: such measures can be proved by means of interpolation
determinants, and they suffice to produce algebraic independence statements
(small transcendence degree: at least two numbers in certain given sets are al-
gebraically independent). Conjecturally, this method should also produce large
transcendence degree results - see [104] Conjecture 15.31. See also the work by
P. Philippon [67].
Another approach, due to M. Laurent and D. Roy [45] (see also [63] Chap. 13)
is based on the observation that in algebraic independence proofs, the determi-
nants which occur produce sequences of polynomials having small values to-
gether with their derivatives at a given point. By means of a generalization
of Gel’fond’s transcendence criterion involving multiplicities, M. Laurent and
D. Roy succeeded to get algebraic independence results. Further generalisa-
tions of Gel’fond’s criterion involving not only multiplicities, but also several
points (having some structure, either additive or multiplicative) are being in-
vestigated by D. Roy in connection with his original strategy towards a proof
of Schanuel’s Conjecture [70, 71].
5 Bost slope inequalities, Arakelov’s Theory

Interpolation determinants require choices of bases. A further tool has been
introduced by J-B. Bost in 1994 [11], where bases are no more required: the
method is more intrinsic. His argument rests on Arakelov’s Theory, which is
used to produce slope inequalities. This new approach is specially interesting for
results on abelian varieties obtained by transcendence methods, the examples
developed by Bost being related with the work of D. Masser and G. Wüstholz on
periods and isogenies of abelian varieties over number fields. Further estimates
related to Baker’s method and measures of linear independence of logarithms
of algebraic points on abelian varieties have been achieved by E. Gaudron [27,
28, 29] using Bost’s approach. An introduction to Bost method can be found in
the Bourbaki lecture [17] by A. Chambert-Loir in 2002.
References
[1] M. Ably and M. M’zari, Interpolation polynomiale sur un ordre d’un
corps de nombres, The Ramanujan J., 17 (2008), pp. 281–304.
[2] F. Amoroso and S. David, Le problème de Lehmer en dimension

supérieure, J. reine angew. Math., 513 (1999), pp. 145–179.
[3] Y. André, G-functions and geometry, Aspects of Mathematics, E13,

Friedr. Vieweg & Sohn, Braunschweig, 1989.
30
[4] A. Baker and G. Wüstholz, Logarithmic forms and Diophantine ge-
ometry, New Mathematical Monographs 9. Cambridge University Press,
2007.
[5] K. Barré-Sirieix, G. Diaz, F. Gramain, and G. Philibert, Une
preuve de la conjecture de Mahler-Manin, Invent. Math., 124 (1996),
pp. 1–9.
[6] D. Bertrand and D. Masser, Linear forms in elliptic integrals, Invent.
Math., 58 (1980), pp. 283–288.
[7] Y. Bilu, The many faces of the subspace theorem [after Adamczewski,
Bugeaud, Corvaja, Zannier. . . ]. Séminaire Bourbaki, 59e année (2006–
2007), No 967; Astérisque 317 (2008) 1–38, 2006.
[8] E. Bombieri, Algebraic values of meromorphic maps, Invent. Math., 10
(1970), pp. 267–287.
[9] E. Bombieri and W. Gubler, Heights in Diophantine geometry, New
Mathematical Monographs 4. Cambridge: Cambridge University Press,
2006.
[10] E. Bombieri and S. Lang, Analytic subgroups of group varieties, Invent.
Math., 11 (1970), pp. 1–14.
[11] J.-B. Bost, Périodes et isogénies des variétés abéliennes sur les corps
de nombres (d’après D. Masser et G. Wüstholz), séminaire Bourbaki, vol.
1994/95, Astérisque, (1996), pp. 115–161, Exp. No. 795, 4.
[12] R. Breusch, A proof of the irrationality of π, The American Mathemat-
ical Monthly, 61 (1954), pp. 631–632.
[13] C. Brezinski, The long history of continued fractions and Padé approx-
imants, in Padé approximation and its applications (Amsterdam, 1980),
vol. 888 of Lecture Notes in Math., Springer, Berlin, 1981, pp. 1–27.
[14] , History of continued fractions and Padé approximants, vol. 12 of

Springer Series in Computational Mathematics, Springer-Verlag, Berlin,
1991.
http://www.emis.de/cgi-bin/MATH-item?0714.01001.
[15] D. C. Cantor and E. G. Straus, On a conjecture of D. H. Lehmer,
Acta Arith., 42 (1982/83), pp. 97–100. Correction: ibid., 42 3 (1983),
327.
[16] J. W. S. Cassels, An introduction to Diophantine approximation, Hafner

Publishing Co., New York, 1972. Facsimile reprint of the 1957 edition,
Cambridge Tracts in Mathematics and Mathematical Physics, No. 45.
31
[17] A. Chambert-Loir, Théorèmes d’algébricité en géométrie diophantienne
(d’après J.-B. Bost, Y. André, D. & G. Chudnovsky), séminaire Bourbaki,
vol. 2000/2001, exp. no. 886, Astérisque, 282 (2002), pp. 175–209.
[18] J. Coates, On the algebraic approximation of functions. I, II, III,
IV, Nederl. Akad. Wetensch. Proc. Ser. A 69=Indag. Math., 28 (1966),
pp. 421–434, 435–448, 449–461 and 29 (1967), pp. 205–212.
[19] , Approximation in algebraic function fields of one variable, J. Aus-
tral. Math. Soc., 7 (1967), pp. 341–355.
[20] S. David, On the height of subvarieties of group varieties. To appear,
2009.
[21] E. Dobrowolski, On a question of Lehmer and the number of irreducible
factors of a polynomial, Acta Arith., 34 (1979), pp. 391–401.
[22] L. Euler, Introductio in Analysin Infinitorum. (Opera Omnia. Series
Prima: Opera Mathematica, Volumen Novum.), Societas Scientiarum
Naturalium Helveticae, Geneva, 1945.
www.math.dartmouth.edu/∼euler/pages/E101.html .
[23] A. Faisant and G. Philibert, Quelques résultats de transcendance liés
à l’invariant modulaire j, J. Number Theory, 25 (1987), pp. 184–200.
[24] N. I. Fel′ dman and Y. V. Nesterenko, Transcendental numbers, in
Number Theory, IV, vol. 44 of Encyclopaedia Math. Sci., Springer, Berlin,
1998, pp. 1–345.
[25] N. I. Fel′ dman and A. B. Šidlovskiı̆, The development and present
state of the theory of transcendental numbers, Uspehi Mat. Nauk, 22
(1967), pp. 3–81.
[26] S. Fischler, Irrationalité de valeurs de zêta (d’après Apéry, Rivoal,. . . ),
Astérisque, (2004), pp. 27–62.
[27] É. Gaudron, Mesure d’indépendance linéaire de logarithmes dans un
groupe algébrique commutatif, C. R. Acad. Sci. Paris Sér. I Math., 333
(2001), pp. 1059–1064.
[28] , Formes linéaires de logarithmes effectives sur les variétés
abéliennes, Ann. Sci. École Norm. Sup. (4), 39 (2006), pp. 699–773.
[29] , Étude du cas rationnel de la théorie des formes linéaires de loga-
rithmes, J. Number Theory, 127 (2007), pp. 220–261.
[30] A. O. Gel′ fond, Sur les propriétés arithmétiques des fonctions entières,
Tôhoku Math. Journ., 30 (1929), pp. 280–285.
[31] , Sur le septième problème de D. Hilbert, Izv. Akad. Nauk SSSR, 7
(1934), pp. 623–630.
32
[32] , Isčislenie konečnyh raznosteı̆, Gosudarstv. Izdat. Tehn.-Teor. Lit.,
Moscow-Leningrad, 1952. French transl., Calcul des différences finies,
Dunod (1963); engl. transl., Calculus of finite differences, Hindustan Pub-
lishing Corp. (1971).
[33] , Transcendental and algebraic numbers, Translated from the first
Russian edition by Leo F. Boron, Dover Publications Inc., New York,
1960.
[34] F. Gramain, Sur le théorème de Fukasawa-Gel′ fond, Invent. Math., 63

(1981), pp. 495–506.
[35] , Fonctions entières arithmétiques, in Séminaire d’analyse 1985–1986
(Clermont-Ferrand, 1985–1986), Univ. Clermont-Ferrand II, Clermont,
1986, pp. 9, Exp. No. 9.
[36] G. H. Hardy and E. M. Wright, An introduction to the theory of
numbers, Oxford University Press, Oxford, sixth ed., 2008. Revised by D.
R. Heath-Brown and J. H. Silverman.
[37] C. Hermite, Sur la fonction exponentielle, C. R. Acad. Sci. Paris, 77

(1873), pp. 18–24, 74–79, 226–233, 285–293. Œuvres de Charles Her-
mite, Paris: Gauthier-Villars, 1905-1917. University of Michigan Histori-
cal Math Collection
http://name.umdl.umich.edu/AAS7821.0001.001.
[38] J. Huang, D. Marques, and M. Mereb, Algebraic values of transcen-

dental functions at algebraic points.
http://arxiv.org/abs/0808.2766, 2008.
[39] H. Jager, A multidimensional generalization of the Padé table. I, II, III,
IV, V, VI, Nederl. Akad. Wetensch. Proc. Ser. A 67 = Indag. Math., 26
(1964), pp. 193–198, 199–211, 212–225, 227–239, 240–244, 245–249.
[40] R. Lagrange, Mémoire sur les séries d’interpolation, Acta Math., 64
(1935), pp. 1–80.
[41] S. Lang, Introduction to transcendental numbers, Addison-Wesley Pub-

lishing Co., Reading, Mass.-London-Don Mills, Ont., 1966.
[42] , Algebra, vol. 211 of Graduate Texts in Mathematics, Springer-

Verlag, New York, third ed., 2002.
[43] M. Laurent, Sur quelques résultats récents de transcendance, Astérisque,
(1991), pp. 209–230 (1992). Journées Arithmétiques, 1989 (Luminy, 1989).
[44] , On transfer inequalities in Diophantine approximation, in “Analytic

Number Theory - Essays in Honour of Klaus Roth”, Cambridge University
Press, 2009, pp. 306–314.
33
[45] M. Laurent and D. Roy, Criteria of algebraic independence with mul-
tiplicities and approximation by hypersurfaces, J. reine Angew. Math., 536
(2001), pp. 65–114.
[46] D. H. Lehmer, Factorization of certain cyclotomic functions, Annals of
Math., 34 (1933), pp. 461–479.
[47] J. Liouville, Sur l’irrationalité du nombre e = 2, 718 . . . , J. Math. Pures

Appl., 1 (1840), p. 192. Addition à la note sur l’irrationalité du nombre
e, id., 193–194;
http://portail.mathdoc.fr/JMPA/.
[48] , Mémoires et communications des Membres et des correspondants

de l’Académie, C.R. Acad. Sci. Paris, 18 (1844), pp. 883–885; Nouvelle
démonstration d’un théorème sur les irrationelles algébriques, inséré dans
le Compte rendu de la dernière séance, id., pp. 910–911.
[49] , Sur des classes très étendues de quantités dont la valeur n’est ni
algébrique, ni même réductible à des irrationnelles algébriques, J. Math.
Pures et Appl., 16 (1851), pp. 133–142.
http://portail.mathdoc.fr/JMPA/.
[50] K. Mahler, Arithmetische Eigenschaften der Lösungen einer Klasse von
Funktionalgleichungen, Math. Ann., 101 (1929), pp. 342–366.
[51] , Zur Approximation der Exponentialfunktion und des Logarithmus.

I,II, J. für Math., 166 (1931), pp. 118–136, 137–150.
[52] , Remarks on a paper by W. Schwarz, J. Number Theory, 1 (1969),

pp. 512–521.
[53] , Lectures on transcendental numbers, Springer-Verlag, Berlin, 1976.

Lecture Notes in Mathematics, Vol. 546.
[54] D. W. Masser, Sur les fonctions entières à valeurs entières, C. R. Acad.
Sci. Paris Sér. A-B, 291 (1980), pp. A1–A4.
[55] , On polynomials and exponential polynomials in several complex
variables, Invent. Math., 63 (1981), pp. 81–95.
[56] , Heights, transcendence, and linear independence on commutative

group varieties, in Diophantine approximation (Cetraro, 2000), vol. 1819
of Lecture Notes in Math., Springer, Berlin, 2003, pp. 1–51.
[57] D. W. Masser and G. Wüstholz, Zero estimates on group varieties.
I, Invent. Math., 64 (1981), pp. 489–516.
[58] , Zero estimates on group varieties. II, Invent. Math., 80 (1985),
pp. 233–267.
34
[59] M. Mignotte, Entiers algébriques dont les conjugués sont proches du cer-
cle unité, in Séminaire Delange-Pisot-Poitou, 19e année: 1977/78, Théorie
des nombres, Fasc. 2, Secrétariat Math., Paris, 1978, pp. 6, Exp. No. 39.
http://www.numdam.org/.
[60] M. Mignotte and M. Waldschmidt, Approximation des valeurs
de fonctions transcendantes, Nederl. Akad. Wetensch. Proc. Ser. A
78=Indag. Math., 37 (1975), pp. 213–223.
[61] Y. V. Nesterenko, Modular functions and transcendence questions,
Mat. Sb., 187 (1996), pp. 65–96.
[62] , A simple proof of the irrationality of π, Russ. J. Math. Phys., 13

(2006), p. 473.
[63] Y. V. Nesterenko and P. Philippon, Introduction to algebraic in-
dependence theory, vol. 1752 of Lecture Notes in Mathematics, Springer-
Verlag, Berlin, 2001.
[64] K. Nishioka, Mahler functions and transcendence, vol. 1631 of Lecture

Notes in Mathematics, Springer-Verlag, Berlin, 1996.
[65] P. Philippon, Indépendance algébrique de valeurs de fonctions exponen-

tielles p-adiques, J. reine Angew. Math., 329 (1981), pp. 42–51.
[66] P. Philippon, Une approche méthodique pour la transcendance et

l’indépendance algébrique de valeurs de fonctions analytiques, J. Number
Theory, 64 (1997), pp. 291–338.
[67] , Approximations algébriques des points dans les espaces projectifs.
I, J. Number Theory, 81 (2000), pp. 234–253.
[68] G. Pólya, Über ganzwertige ganze Funktionen, Rend. Circ. Mat. Palermo,
40 (1915), pp. 1–16.
[69] T. Rivoal, Applications arithmétiques de l’interpolation Lagrangienne,

Intern. J. Number Th., 5 (2009), pp. 185–208.
[70] D. Roy, An arithmetic criterion for the values of the exponential function,
Acta Arith., 97 (2001), pp. 183–194.
[71] , Une formule d’interpolation en deux variables, J. Théor. Nombres
Bordeaux, 13 (2001), pp. 315–323. 21st Journées Arithmétiques (Rome,
1999).
[72] , Interpolation formulas and auxiliary functions, J. Number Theory,

94 (2002), pp. 248–285.
[73] , Interpolation sur des perturbations d’ensembles produits, Bull. Soc.
Math. France, 130 (2002), pp. 387–408.
35
[74] D. Roy and M. Waldschmidt, Approximation diophantienne et
indépendance algébrique de logarithmes, Ann. Sci. École Norm. Sup. (4),
30 (1997), pp. 753–796.
[75] W. M. Schmidt, Diophantine approximation, vol. 785, Lecture Notes in
Mathematics. Berlin-Heidelberg-New York: Springer-Verlag, 1980.
[76] T. Schneider, Transzendenzuntersuchungen periodischer Funktionen. I.

Transzendenz von Potenzen, J. reine angew. Math., 172 (1934), pp. 65–69.
[77] , Zur Theorie der Abelschen Funktionen und Integrale, J. reine

angew. Math., 183 (1941), pp. 110–128.
[78] , Ein Satz über ganzwertige Funktionen als Prinzip für Transzen-
denzbeweise, Math. Ann., 121 (1949), pp. 131–140.
[79] , Einführung in die transzendenten Zahlen, Springer-Verlag, Berlin,
1957. French Transl., Introduction aux nombres transcendants, Gauthier-
Villars, Paris (1959).
[80] A. B. Šidlovskiı̆, Transcendental numbers, vol. 12 of de Gruyter Studies

in Mathematics, Walter de Gruyter & Co., Berlin, 1989. Translated from
the Russian by Neal Koblitz, With a foreword by W. Dale Brownawell.
[81] C. L. Siegel, Über einige Anwendungen diophantischer Approximatio-

nen, Abhandlungen Akad. Berlin, Nr. 1 (1929), p. 70 S.
[82] , Über die Perioden elliptischer Funktionen, J. f. M., 167 (1932),

pp. 62–69.
[83] , Transcendental numbers, Princeton, N. J.: Princeton University
Press, 1949.
[84] J. Stainville, Mélanges d’analyse algébrique et de géométrie.

http://www.bibnum.education.fr/mathematiques/
melange-danalyse-algebrique-et-de-geometrie, 1815.
[85] C. L. Stewart, Algebraic integers whose conjugates lie near the unit
circle, Bull. Soc. Math. France, 106 (1978), pp. 169–176.
[86] A. Surroca, Valeurs algébriques de fonctions transcendantes, Int. Math.

Res. Not., Art. ID 16834 (2006), p. 31.
[87] A. Thue, Über Annäherungswerte algebraischer Zahlen., J. für Math.,

135 (1909), pp. 284–305.
[88] A. Thue, Selected mathematical papers, Universitetsforlaget, Oslo, 1977.

[89] A. J. Van der Poorten, Transcendental entire functions mapping ev-
ery algebraic number field into itself, J. Austral. Math. Soc., 8 (1968),
pp. 192—193.
36
[90] I. Wakabayashi, Meilleures estimations possibles pour
l’ensemble des points algébriques de fonctions analytiques, in
Seminar on number theory, 1983–1984 (Talence, 1983/1984),
Univ. Bordeaux I, Talence, 1984, pp. 8, Exp. No. 19.
http://www.digizeitschriften.de/resolveppn/GDZPPN002545349&L=2.
[91] , On the optimality of certain estimates for algebraic values of ana-

lytic functions, J. Austral. Math. Soc. Ser. A, 39 (1985), pp. 400–414.
[92] , Algebraic values of meromorphic functions on Riemann surfaces,
J. Number Theory, 25 (1987), pp. 220–229.
[93] , Algebraic values of functions on the unit disk, in Prospects of math-
ematical science (Tokyo, 1986), World Sci. Publishing, Singapore, 1988,
pp. 235–266.
[94] , An extension of the Schneider-Lang theorem, in Seminar on Dio-
phantine Approximation (Japanese) (Yokohama, 1987), vol. 12 of Sem.
Math. Sci., Keio Univ., Yokohama, 1988, pp. 79–83.
[95] M. Waldschmidt, Nombres transcendants, Springer-Verlag, Berlin,

1974. Lecture Notes in Mathematics, Vol. 402.
[96] , Transcendance et exponentielles en plusieurs variables, Invent.

Math., 63 (1981), pp. 97–127.
[97] , Sous-groupes analytiques de groupes algébriques, Ann. of Math. (2),

117 (1983), pp. 627–657.
[98] , Nombres transcendants et groupes algébriques, Astérisque, (1987),
p. 218. With appendices by Daniel Bertrand and Jean-Pierre Serre.
[99] , Fonctions auxiliaires et fonctionnelles analytiques. I, II, J. Analyse

Math., 56 (1991), pp. 231–254, 255–279.
[100] , Constructions de fonctions auxiliaires, in Approximations dio-

phantiennes et nombres transcendants (Luminy, 1990), de Gruyter, Berlin,
1992, pp. 285–307.
[101] , Integer valued functions on products, J. Ramanujan Math. Soc., 12

(1997), pp. 1–24.
[102] , Extrapolation with interpolation determinants, in Special functions
and differential equations (Madras, 1997), Allied Publ., New Delhi, 1998,
pp. 356–366.
[103] , Integer valued entire functions on Cartesian products, in Num-

ber theory in progress, Vol. 1 (Zakopane-Kościelisko, 1997), de Gruyter,
Berlin, 1999, pp. 553–576.
37
[104] , Diophantine approximation on linear algebraic groups, vol. 326
of Grundlehren der Mathematischen Wissenschaften, Springer-Verlag,
Berlin, 2000.
Michel WALDSCHMIDT
Université P. et M. Curie (Paris VI)
Institut de Mathématiques CNRS UMR 7586
Théorie des Nombres Case 247
175, rue du Chevaleret
F–75013 PARIS
e-mail: miw@math.jussieu.fr
URL: http://www.math.jussieu.fr/∼miw/
38

Auxiliary Fns Hal 072009

Uploaded by

Copyright:

Available Formats

Auxiliary Fns Hal 072009

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Auxiliary Fns Hal 072009

Uploaded by

Copyright:

Available Formats

Auxiliary functions in transcendence proofs

To cite this version:

HAL Id: hal-00405120

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

Updated: March 13, 2009

Auxiliary functions in transcendental number theory.

3 Auxiliary functions arising from the Dirichlet’s box principle 13

5 Bost slope inequalities, Arakelov’s Theory 30

It follows from Liouville’s inequality that ξ is transcendental. Real numbers

which one specializes at z = 1. As noticed by F. Beukers, the proof becomes even

has a zero at the origin of multiplicity at least N .

1.3 Padé approximation

Bi (z)fj (z) − Bj (z)fi (z) (0 ≤ i < j ≤ m)

has a zero of multiplicity > N .

P1 (z)f1 (z) + · · · + Pm (z)fm (z)

has a zero at the origin of multiplicity at least N + m − 1.

1.4 Hypergeometric methods

For instance Hermite–Lindemann’s Theorem on the transcendence of log α and

Let A be a countable subset of C. For each pair (α, s) with α ∈ A,

for all (α, s) ∈ A × Z≥0 .

holds for sufficiently large R.

2.2 Integer–valued entire functions

lim sup 2−R |f |R ≥ 1,

which is valid under the same assumptions as for (2.1).

fn (z) = fn (αn+1 ) + (z − αn+1 )fn+1 (z).

for the set Z>0 = {1, 2, . . . } of positive integers and

for the set Z of rational integers.

2.4 Lagrange interpolation

in place of (2.2), iterating and integrating as in § 2.2, one deduces an expansion

He expands ζ(2, z) as a Lagrange series in

He shows that the coefficients of the expansion belong to Q + Qζ(3). This

3 Auxiliary functions arising from the Dirich-

Theorem 3.2 (Schmidt’s Subspace Theorem - simplified form). For m ≥ 2

{x = (x1 , . . . , xm ) ∈ Zm ; |L1 (x) · · · Lm (x)| ≤ |x|−ǫ }

is contained in the union of finitely many proper subspaces of Qm .

3.1.2 Siegel, Gel’fond, Schneider

F (z) = A(ez , eβz );

z t0 αt1 z z=s1 +s2 β

It is easily seen that they are the same, namely

(s1 + s2 β)t0 αt1 s1 (αβ )t1 s2 .

See [99] and § 13.7 of [104].

(1, log α1 , . . . , log αn ) ∈ Cn .

This situation is therefore an extension of the setup in Gel’fond’s solution of

z0 , z1 , . . . , zn , ez0 α1z1 · · · αnzn

and the points in the subgroup of Cn+1 generated by

Since Baker’s Theorem includes the transcendence of e, there is no hope to prove

3.1.3 Schneider–Lang Criterion

Theorem 3.3. Let K be a number field and f1 , . . . , fd be entire functions in C.

K = Q(α, eα ), f1 (z) = z, f2 (z) = ez , S = {mα ; m ∈ Z},

as well as the Gel’fond –Schneider Theorem on the transcendence of αβ following

K = Q(α, β, αβ ), f1 (z) = ez , f2 (z) = eβz , S = {m log α ; m ∈ Z}.

3.1.4 Higher dimension: several variables

do not all belong to K.

3.1.5 The six exponentials Theorem

Lemma 3.10. Let ϕ1 , . . . , ϕL be entire functions in C, ζ1 , . . . , ζL be elements

Another method of proof of the six exponentials Theorem 3.8 is proposed

3.1.6 Modular functions

is at least 3. Here, P , Q, R are the classical Ramanujan functions, which are

3.2 Universal auxiliary functions