1 Introduction
A real number is called transcendental if it is not a root of any nonzero polynomial with integer coefficients. Since the pioneering work of Liouville in 1844, transcendental numbers have occupied a central position in number theory. Liouville not only constructed the first explicit transcendental numbers, but also introduced a remarkable class of numbers that admit exceptionally accurate rational approximations. These numbers, now known as Liouville numbers, provided the first examples of transcendental real numbers and remain among the most important objects in transcendence theory.
The subject developed dramatically during the twentieth century. Among the most celebrated achievements are the theorems of Gel’fond [Reference Gel’fond3] and Schneider [Reference Schneider6], which independently solved Hilbert’s seventh problem. They proved that if
$\alpha $
and
$\beta $
are algebraic numbers with
$\alpha \neq 0,1$
and
$\beta $
irrational, then
$\alpha ^\beta $
is transcendental. These results established deep connections between algebraic and transcendental numbers and demonstrated the power of analytic methods in number theory. Since then, transcendence theory has expanded into a rich discipline concerned with algebraic independence, Diophantine approximation, special values of analytic functions and the arithmetic properties of exponential expressions.
A real irrational number x is called a Liouville number if, for every
$N\in \mathbb N$
, there exist integers p and q with
$q\ge 2$
such that
We denote by
$\mathcal L$
the set of all Liouville numbers.
It is classical that every Liouville number is transcendental. From the viewpoint of measure theory, however, the set
$\mathcal L$
is extremely small: it has Lebesgue measure zero. From the viewpoint of topology, the situation is very different. A standard application of Baire’s theorem shows that
$\mathcal L$
is a dense
$G_\delta $
subset of
$\mathbb R$
. Consequently,
$\mathcal L$
is topologically large despite being measure-theoretically negligible.
This contrast between measure and category has become a recurring theme in modern analysis and number theory. A classical account of such phenomena can be found in Oxtoby’s monograph Measure and Category [Reference Oxtoby5], which systematically develops the striking differences between measure-theoretic and category-theoretic notions of largeness. In transcendence theory, Baire category methods have repeatedly proved useful for establishing the abundance of exceptional classes of transcendental numbers. A particularly attractive example is provided by Steuding [Reference Steuding, Sander, Steuding and Steuding7], who demonstrated how functional-analytic and Baire category techniques can be used to obtain results concerning Liouville numbers and related transcendence phenomena. Such methods complement more classical arithmetic approaches and reveal that many transcendence questions possess a natural topological dimension.
The arithmetic structure of
$\mathcal L$
is surprisingly rich. In a celebrated theorem, Erdős [Reference Erdős2] proved that every real number can be expressed as a sum and, if nonzero, as a product of two Liouville numbers. Later, Burger [Reference Burger1] extended these ideas to a broad class of two-variable functions, showing in particular that every positive real number
$r\neq 1$
can be written in the form
$r=\sigma ^\tau $
for suitable Liouville numbers
$\sigma $
and
$\tau .$
These results illustrate the remarkable flexibility of Liouville numbers and suggest that transcendental values frequently arise in exponential constructions.
Despite this abundance of representations, the behaviour of the self-power map
$f(x)=x^x$
remains poorly understood. In particular, it is not known whether
$x^x$
must be transcendental whenever x is a Liouville number. A partial answer was obtained by Marques and Ribeiro [Reference Marques and Ribeiro4], who proved transcendence for a dense
$G_\delta $
subclass of Liouville numbers, namely the class of
$\varepsilon $
-strong Liouville numbers.
The purpose of the present paper is to investigate the self-power map from a topological perspective. Rather than studying individual Liouville numbers, we examine the size and structure of the sets
Our main results show that
$\mathcal T$
is comeagre and has full Lebesgue measure in
$(0,\infty )$
, while
$\mathcal G$
is a dense
$G_\delta $
subset of
$(0,\infty )$
and has Lebesgue measure zero. Thus, the self-power map exhibits a striking category–measure dichotomy: transcendental values occur on a set that is both topologically and measure-theoretically large, whereas Liouville values occur on a set that is topologically large but measure-theoretically negligible.
Throughout the paper, we write
$ \overline {\mathbb {Q}}\cap \mathbb {R}$
for the set of real algebraic numbers. This avoids possible confusion with the standard notation
$\overline {\mathbb {Q}}$
, which usually denotes the full algebraic closure of
$\mathbb {Q}$
in
$\mathbb C$
.
2 Results and proofs
The following result, although classical, is included for completeness.
Lemma 2.1. The set
$\mathcal L$
of Liouville numbers is a dense
$G_\delta $
subset of
$\mathbb R$
. Moreover,
$\mathcal L$
has Lebesgue measure zero.
Proof. For
$N,Q\in \mathbb N$
, with
$Q\ge 2$
, define
Each
$U_{N,Q}$
is open. It is also dense in
$\mathbb R$
, since it contains all rational numbers
$p/q$
with
$q\ge Q$
, and these are dense in
$\mathbb R$
.
We claim that
Indeed, the inclusion from right to left is immediate. Conversely, if
$x\in \mathcal L$
, then x is irrational. Given N and Q, choose
$M>N$
. By the Liouville property, there exists
$p/q$
such that
For fixed Q, the denominators q occurring in such approximations cannot remain bounded as
$M\to \infty $
, since x is irrational. Hence, for some such approximation,
$q\ge Q$
, and then
Thus,
$x\in U_{N,Q}$
. The first factor is a countable intersection of open dense sets. Also,
$\mathbb R\setminus \mathbb Q$
is a dense
$G_\delta $
, since
Therefore,
$\mathcal L$
is a dense
$G_\delta $
subset of
$\mathbb R$
.
Finally, the classical Liouville theorem implies that every Liouville number is transcendental. Since the set of algebraic real numbers is countable, this does not by itself give measure zero. The standard Borel–Cantelli argument gives measure zero: for fixed
$N>2$
, the total length of the intervals with
$q\ge Q$
over any bounded interval is bounded by a constant multiple of
which tends to
$0$
as
$Q\to \infty $
. Hence, the set of real numbers satisfying the Liouville inequalities for this fixed
$N>2$
has measure zero on every bounded interval. Since
$\mathcal L$
is contained in that set,
$\mathcal L$
has Lebesgue measure zero.
Theorem 2.2. The set
is comeagre in
$(0,\infty )$
(in particular, it is a dense
$G_\delta $
) and it has full Lebesgue measure in
$(0,\infty )$
.
Proof. Let
$f\colon (0,\infty )\to (0,\infty )$
be the function
$f(x)=x^x=\exp (x\log x)$
. The derivative of f is
$f'(x)=x^x(\log x+1)$
, and thus f is strictly decreasing on
$(0,e^{-1})$
and strictly increasing on
$(e^{-1},\infty )$
, with a unique global minimum
$f(e^{-1})=e^{-1/e}$
. Consequently, for each
$\alpha \in \mathbb {R},$
the level set
$F_\alpha = f^{-1}(\{\alpha \})\subset (0,\infty )$
has cardinality
$|F_\alpha |\leq 2$
.
Consider the set
Since each
$F_\alpha $
is finite and
$ \overline {\mathbb {Q}}\cap \mathbb {R}$
is countable, A is countable. Hence, A is meagre and has Lebesgue measure
$0$
. Therefore, its complement
$\mathcal {T}=(0,\infty )\setminus A$
is comeagre (equivalently, a dense
$G_\delta $
) and has full Lebesgue measure.
Corollary 2.3. The set
is a dense
$G_\delta $
subset of
$(0,\infty )$
.
Proof. By Theorem 2.2,
$\mathcal T$
is a dense
$G_\delta $
subset of
$(0,\infty )$
. By Lemma 2.1,
$\mathcal L\cap (0,\infty )$
is also a dense
$G_\delta $
. Since the intersection of two dense
$G_\delta $
subsets of a Baire space is again a dense
$G_\delta $
, the result follows.
Theorem 2.4. The set
is a dense
$G_\delta $
subset of
$(0,\infty )$
and has Lebesgue measure zero.
Proof. Let
$f:(0,\infty )\to (0,\infty )$
be given by
$f(x)=x^x$
. The function f is continuous, and
$f'(x)=x^x(\log x+1)$
. Hence,
$f'(x)=0$
only at
$x=e^{-1}$
. Since
$\mathcal L$
is a
$G_\delta $
subset of
$\mathbb R$
and f is continuous, the preimage
$ \mathcal G=f^{-1}(\mathcal L)$
is a
$G_\delta $
subset of
$(0,\infty )$
.
We next prove density. Let
$I\subset (0,\infty )$
be a nonempty open interval. Since f is continuous and nonconstant on I, the image
$f(I)$
contains a nonempty open interval J. Since
$\mathcal L$
is dense in
$\mathbb R$
, we have
$J\cap \mathcal L\neq \varnothing $
. Choose
$y\in J\cap \mathcal L$
. Since
$J\subset f(I)$
, there exists
$x\in I$
such that
$f(x)=y$
. Therefore
$x\in I\cap \mathcal G$
. Hence,
$I\cap \mathcal G\neq \varnothing $
, so
$\mathcal G$
is dense in
$(0,\infty )$
.
It remains to prove that
$\mathcal G$
has Lebesgue measure zero. Let
$c=e^{-1}$
. Since
${f'(x)\neq 0}$
for every
$x\neq c$
, every compact interval
$K\subset (0,\infty )\setminus \{c\}$
has the property that
${|f'(x)|\ge m_K>0}$
for all
$x\in K$
. Since K is an interval not containing c, the derivative
$f'$
has constant sign on K. Moreover, by the mean value theorem,
Therefore,
$f|_K$
is injective and its inverse branch is Lipschitz. In particular, the inverse branch
$(f|_K)^{-1}$
is Lipschitz on
$f(K)$
.
Since
$\mathcal L$
has Lebesgue measure zero, the set
$\mathcal L\cap f(K)$
has measure zero. Since
$f|_K$
is injective and its inverse branch is Lipschitz, the image of every null set under
$(f|_K)^{-1}$
is a null set. Since
$\mathcal L\cap f(K)$
has measure zero,
also has measure zero.
Now,
$(0,\infty )\setminus \{c\}$
is a countable union of compact intervals of the above type. For example,
where empty intervals are ignored. Therefore,
$f^{-1}(\mathcal L)\setminus \{c\}$
has measure zero. Adding the single point c does not change the measure, and hence
Thus,
$\mathcal G$
is a dense
$G_\delta $
subset of
$(0,\infty )$
and has Lebesgue measure zero.
Corollary 2.5. The set
is a dense
$G_\delta $
subset of
$(0,\infty )$
.
Proof. By Theorem 2.4,
$\mathcal G$
is a dense
$G_\delta $
subset of
$(0,\infty )$
. By Lemma 2.1,
$\mathcal L\cap (0,\infty )$
is also a dense
$G_\delta $
subset of
$(0,\infty )$
. Hence,
$\mathcal G\cap \mathcal L$
is the intersection of two dense
$G_\delta $
subsets of the Baire space
$(0,\infty )$
and is therefore again a dense
$G_\delta $
.
Acknowledgement
The authors thank the referee for valuable comments that improved the presentation of the paper.



