Galois Rational Maps Overview

• Galois rational maps are rational functions whose field extension is Galois, exhibiting deck groups that act transitively on fibers.
• They are studied through mechanisms such as pullbacks, arboreal representations, and field intersections, linking dynamics with function-field theory.
• Research connects one-dimensional phenomena with higher-dimensional birational geometry and cohomological invariants, highlighting exceptional symmetry constraints.

In current research, Galois rational maps are studied through several closely related mechanisms: Galois extensions of rational function fields, deck transformation groups of rational coverings, monodromy of iterated preimage towers, and rational pullbacks of branched covers. The central one-dimensional case concerns rational functions $X\in \mathbb C(z)$ that are Galois coverings of $\mathbb{CP}^1$, equivalently those for which $\mathbb C(z)/\mathbb C(X)$ is Galois and the deck group acts transitively on fibers; adjacent literatures analyze arboreal and dynatomic Galois groups of iterates, pullback generation of $G$-covers of $\mathbb P^1$, and generically finite rational covers between higher-dimensional $K$-trivial varieties (Pakovich, 31 Mar 2026, Leung et al., 2024, Dèbes et al., 2018, Verni, 18 Jul 2025).

1. Galois coverings of the projective line

For a holomorphic map $X:E\to R$, the deck transformation group is

$$G_X=\operatorname{Aut}(E,X)=\{\mu\in \operatorname{Aut}(E)\mid X\circ \mu=X\}.$$

The map $X$ is a Galois covering if $G_X$ acts transitively on fibers of $\mathbb{CP}^1$0. Equivalently, the extension $\mathbb{CP}^1$1 is Galois, and then

$\mathbb{CP}^1$2

In the compact-surface setting, one also has the degree formula

$\mathbb{CP}^1$3

and conversely this equality characterizes Galois coverings (Pakovich, 31 Mar 2026).

For rational maps on $\mathbb{CP}^1$4, Galois coverings are exactly quotient maps by finite subgroups of $\mathbb{CP}^1$5. The finite subgroups are cyclic, dihedral, tetrahedral, octahedral, and icosahedral. This classification has two immediate consequences emphasized in the literature. First, if a rational Galois covering is indecomposable, then it must be Möbius-conjugate to

$\mathbb{CP}^1$6

Second, every degree-two rational map is Galois, with deck group of order $\mathbb{CP}^1$7, so low degree by itself does not force rigid intersection or monodromy behavior (Pakovich, 31 Mar 2026).

The subgroup/intermediate-cover correspondence is fundamental. Every subgroup $\mathbb{CP}^1$8 corresponds to a factorization

$\mathbb{CP}^1$9

where $\mathbb C(z)/\mathbb C(X)$0 is Galois with $\mathbb C(z)/\mathbb C(X)$1. This correspondence underlies decomposition arguments for rational maps and is one reason Galois coverings serve as the organizing objects in the function-field approach.

2. Field intersections, functional equations, and equivariance

A basic problem asks, for rational functions $\mathbb C(z)/\mathbb C(X)$2 with $\mathbb C(z)/\mathbb C(X)$3 and

$\mathbb C(z)/\mathbb C(X)$4

when the intersection field $\mathbb C(z)/\mathbb C(X)$5 is nontrivial, and when the extension degree

$\mathbb C(z)/\mathbb C(X)$6

attains its minimal possible value. By Lüroth’s theorem every intermediate subfield of $\mathbb C(z)/\mathbb C(X)$7 is of the form $\mathbb C(z)/\mathbb C(X)$8, so

$\mathbb C(z)/\mathbb C(X)$9

is equivalent to the existence of rational functions $G$0 such that

$G$1

Under the standing assumption $G$2, one always has

$G$3

and the equality case is governed by fiber products and “good solutions” of $G$4 (Pakovich, 31 Mar 2026).

Pakovich gives a complete characterization when one map is Galois. If $G$5 is a Galois covering, then

$G$6

if and only if $G$7 factors as

$G$8

where $G$9 is $\mathbb P^1$0-equivariant in the sense that there exists an automorphism

$\mathbb P^1$1

such that

$\mathbb P^1$2

$\mathbb P^1$3 is a rational Galois covering, and the group generated by $\mathbb P^1$4 and $\mathbb P^1$5 is finite of order

$\mathbb P^1$6

When both maps are Galois coverings, nontrivial intersection occurs if and only if $\mathbb P^1$7 is finite, and the minimal value $\mathbb P^1$8 occurs if and only if

$\mathbb P^1$9

(Pakovich, 31 Mar 2026).

A common misconception is that nontrivial intersection already forces this equivariant/Galois structure. It does not. The degree-two phenomenon already produces counterexamples: since any rational map of degree $K$0 is Galois and two involutions can generate a dihedral group of arbitrarily large order, one can obtain

$K$1

arbitrarily large with $K$2. Pakovich also gives an explicit example with

$K$3

for which $K$4 but $K$5 is neither Galois nor $K$6-equivariant, so the minimal-degree criterion fails (Pakovich, 31 Mar 2026).

A complementary reduction in low transcendence degree is available for arbitrary fields. If $K$7 satisfies

$K$8

then

$K$9

for some $X:E\to R$0, some $X:E\to R$1 that is either homogeneous and primitive or zero, and some primitive nonconstant pair $X:E\to R$2. In the case $X:E\to R$3, one has

$X:E\to R$4

and if $X:E\to R$5 is minimal, then $X:E\to R$6 is algebraically closed in $X:E\to R$7 (Bondt, 2015). This does not by itself produce a Galois classification, but it isolates a one-variable intermediate field that is often the natural starting point for one.

3. Arboreal Galois groups of iterated rational maps

In arithmetic dynamics, the Galois theory of a rational map $X:E\to R$8 of degree $X:E\to R$9 is often encoded by the tower of preimage fields of a base point $$G_X=\operatorname{Aut}(E,X)=\{\mu\in \operatorname{Aut}(E)\mid X\circ \mu=X\}.$$0: $$G_X=\operatorname{Aut}(E,X)=\{\mu\in \operatorname{Aut}(E)\mid X\circ \mu=X\}.$$1 The corresponding arboreal representation is the injective homomorphism

$$G_X=\operatorname{Aut}(E,X)=\{\mu\in \operatorname{Aut}(E)\mid X\circ \mu=X\}.$$2

where $$G_X=\operatorname{Aut}(E,X)=\{\mu\in \operatorname{Aut}(E)\mid X\circ \mu=X\}.$$3 is the rooted tree of iterated preimages (Leung et al., 2024).

For post-critically finite rational maps, non-abelianity is the dominant phenomenon. If $$G_X=\operatorname{Aut}(E,X)=\{\mu\in \operatorname{Aut}(E)\mid X\circ \mu=X\}.$$4 is a number field, $$G_X=\operatorname{Aut}(E,X)=\{\mu\in \operatorname{Aut}(E)\mid X\circ \mu=X\}.$$5 is a PCF rational map of degree $$G_X=\operatorname{Aut}(E,X)=\{\mu\in \operatorname{Aut}(E)\mid X\circ \mu=X\}.$$6, and $$G_X=\operatorname{Aut}(E,X)=\{\mu\in \operatorname{Aut}(E)\mid X\circ \mu=X\}.$$7 is non-preperiodic, then

$$G_X=\operatorname{Aut}(E,X)=\{\mu\in \operatorname{Aut}(E)\mid X\circ \mu=X\}.$$8

is not abelian. The proof uses infinitely many non-Archimedean places of periodic reduction, a local splitting principle for irreducible polynomials with abelian Galois group, and equidistribution of small points on Berkovich projective lines. Combined with the result of Ferraguti–Ostafe–Zannier cited there, this shows that potentially abelian polynomial arboreal pairs are confined to the PCF and preperiodic regime (Leung et al., 2024).

A different non-abelianity criterion is archimedean. If $$G_X=\operatorname{Aut}(E,X)=\{\mu\in \operatorname{Aut}(E)\mid X\circ \mu=X\}.$$9 has a real archimedean place corresponding to an embedding $X$0, if $X$1 is surjective, if $X$2 is nonperiodic, and if the Julia set of $X$3 is not contained in $X$4, then

$X$5

is not abelian. The mechanism is that abelianity would force “no partial splitting” at the real place, hence total reality of an infinite backward orbit, and equidistribution would then force the canonical measure, and therefore the Julia set, to be supported on the real locus, contradicting the hypothesis (Leung, 2024).

This criterion becomes concrete for both polynomials and Lattès maps. For $X$6 of degree $X$7, the conditions

$X$8

and “$X$9 is nonempty and is contained in the critical interval $G_X$0” are equivalent. For certain Lattès maps arising from duplication on elliptic curves

$G_X$1

the associated rational map

$G_X$2

is surjective on $G_X$3, while $G_X$4, so the associated arboreal Galois groups are non-abelian for nonperiodic real base points (Leung, 2024).

4. Specialization and dynatomic Galois groups

For PCF rational functions, one can compare the generic iterated Galois group over $G_X$5 with specialized groups over $G_X$6. If

$G_X$7

and

$G_X$8

then under a $G_X$9-group hypothesis on the first-level generic Galois group, there exists an integer $\mathbb{CP}^1$00, depending on $\mathbb{CP}^1$01 and $\mathbb{CP}^1$02, such that

$\mathbb{CP}^1$03

In particular, for a PCF quadratic rational function over a number field, one has

$\mathbb{CP}^1$04

for all $\mathbb{CP}^1$05 outside a thin set (Benedetto et al., 2023).

For the family

$\mathbb{CP}^1$06

with $\mathbb{CP}^1$07 PCF, the criterion is fully finite-level: if $\mathbb{CP}^1$08 is the size of the forward orbit of the critical point $\mathbb{CP}^1$09, then

$\mathbb{CP}^1$10

if and only if

$\mathbb{CP}^1$11

Here $\mathbb{CP}^1$12 is the compositum of the degree-$\mathbb{CP}^1$13 extensions of $\mathbb{CP}^1$14 contained in $\mathbb{CP}^1$15 (Benedetto et al., 2023).

Periodic-point Galois theory is encoded instead by dynatomic polynomials. For a rational map $\mathbb{CP}^1$16,

$\mathbb{CP}^1$17

cuts out points of exact period $\mathbb{CP}^1$18, and when $\mathbb{CP}^1$19 is separable its Galois group is the $\mathbb{CP}^1$20th dynatomic group. For quadratic rational maps with a critical point of exact period $\mathbb{CP}^1$21, there are two normal forms over $\mathbb{CP}^1$22: $\mathbb{CP}^1$23 For both families, the generic third dynatomic Galois group is

$\mathbb{CP}^1$24

For period $\mathbb{CP}^1$25, the generic group for $\mathbb{CP}^1$26 is

$\mathbb{CP}^1$27

whereas for $\mathbb{CP}^1$28 it is the full

$\mathbb{CP}^1$29

Exceptional specializations are controlled by explicit rational parametrizations of subgroup fixed fields, so the variation of the dynatomic Galois group becomes a Diophantine problem on auxiliary curves (Krumm et al., 2023).

5. Rational pullbacks of Galois covers

Another major use of rational maps is as pullback operators on covers of $\mathbb{CP}^1$30. If

$\mathbb{CP}^1$31

is a $\mathbb{CP}^1$32-regular cover and

$\mathbb{CP}^1$33

is a nonconstant rational map, the pullback cover $\mathbb{CP}^1$34 is defined by the normalization of the fiber product. If $\mathbb{CP}^1$35 is given by $\mathbb{CP}^1$36, then $\mathbb{CP}^1$37 is given by

$\mathbb{CP}^1$38

(Dèbes et al., 2018).

The pullback operation is highly constrained. If $\mathbb{CP}^1$39 has branch point number $\mathbb{CP}^1$40 and $\mathbb{CP}^1$41 has branch point number $\mathbb{CP}^1$42, then

$\mathbb{CP}^1$43

and similarly the genus does not decrease under pullback. This monotonicity is one ingredient in the classification of groups whose $\mathbb{CP}^1$44-covers are generated by pullback from a bounded family (Dèbes et al., 2018).

The sharp theorem is that the finite subgroups of $\mathbb{CP}^1$45 are exactly the finite groups $\mathbb{CP}^1$46 for which there exists an integer $\mathbb{CP}^1$47 such that every $\mathbb{CP}^1$48-Galois cover of $\mathbb{CP}^1$49 can be obtained as a rational pullback of a cover with at most $\mathbb{CP}^1$50 branch points. For $\mathbb{CP}^1$51, one well-chosen cover with at most $\mathbb{CP}^1$52 branch points already suffices. For

$\mathbb{CP}^1$53

no bounded branch-point family is regularly parametric, and allowing the branch point number to grow produces genuinely new Galois realizations over $\mathbb{CP}^1$54 (Dèbes et al., 2018).

This also resolves the geometric Beckmann–Black lifting property: the statement that any two $\mathbb{CP}^1$55-Galois covers of $\mathbb{CP}^1$56 are pullbacks of another $\mathbb{CP}^1$57-cover holds only for

$\mathbb{CP}^1$58

The spherical groups

$\mathbb{CP}^1$59

are therefore exceptional not only as finite Möbius groups but also as the only groups for which rational pullback is universally generative (Dèbes et al., 2018).

6. Arithmetic rigidity and higher-dimensional analogues

Arithmetic constraints on rational maps often isolate the same symmetry-rich families that appear in Galois constructions. If $\mathbb{CP}^1$60 is a number field and $\mathbb{CP}^1$61 is a rational map of degree $\mathbb{CP}^1$62 whose multipliers all lie in $\mathbb{CP}^1$63, then $\mathbb{CP}^1$64 is a power map, a Chebyshev map, or a Lattès map. Power maps and Chebyshev maps have only integer multipliers, and for a Lattès map there exists an imaginary quadratic field $\mathbb{CP}^1$65 such that all multipliers lie in $\mathbb{CP}^1$66; they are all integers if and only if the map is flexible. These families are described there as finite quotients of affine maps on cylinders or tori and are sometimes called exceptional (Huguin, 2022).

In the real rational Jacobian setting, the Galois case is equally rigid. If $\mathbb{CP}^1$67 is an everywhere-defined rational nonsingular map and the extension

$\mathbb{CP}^1$68

is Galois, then $\mathbb{CP}^1$69 is invertible if and only if $\mathbb{CP}^1$70 is birational. More generally, invertibility forces the extension degree to be odd and forces the automorphism group

$\mathbb{CP}^1$71

to be trivial, so a nontrivial Galois extension is incompatible with invertibility (Campbell, 2012).

A higher-dimensional birational theory studies generically finite dominant rational maps

$\mathbb{CP}^1$72

between smooth projective varieties with trivial canonical bundle. Such a map is Galois when the induced field extension

$\mathbb{CP}^1$73

is Galois, and one introduces the birational deck group

$\mathbb{CP}^1$74

together with the monodromy group defined from the Galois closure. In this setting there are strong hyper-Kähler restrictions: if $\mathbb{CP}^1$75 is a hyper-K manifold with

$\mathbb{CP}^1$76

and $\mathbb{CP}^1$77 is a Calabi–Yau rational cover, then

$\mathbb{CP}^1$78

so any Galois cover is an isomorphism; moreover, if

$\mathbb{CP}^1$79

is a Calabi–Yau Galois cover of a hyper-K manifold and

$\mathbb{CP}^1$80

then the branch divisor $\mathbb{CP}^1$81 is $\mathbb{CP}^1$82-exceptional (Verni, 18 Jul 2025).

A cohomological analogue appears for varieties associated to central simple algebras. If $\mathbb{CP}^1$83 and $\mathbb{CP}^1$84 generate the same cyclic subgroup of $\mathbb{CP}^1$85, then there are rational embeddings between the associated Brauer–Severi varieties and between their norm hypersurfaces; in fact this condition is equivalent to dominant rational maps both ways between the Brauer–Severi varieties, to stable birationality, and to birationality of the norm hypersurfaces $\mathbb{CP}^1$86 and $\mathbb{CP}^1$87 (Novaković, 2016). This suggests a broader use of “Galois rational maps” in which rational maps are controlled by descent data and cohomological invariants rather than only by deck groups of maps to $\mathbb{CP}^1$88.

Taken together, these strands show that Galois rational maps form a convergent theme rather than a single definition. In one dimension they are governed by deck groups, function-field intersections, and pullback rigidity; in arithmetic dynamics they appear as arboreal and dynatomic Galois groups of iterates; and in higher-dimensional birational geometry they are constrained by monodromy, Kodaira dimension, and hyper-Kähler or Brauer-theoretic structure. Across these settings, the recurring exceptional objects are those with quotient or algebraic-group origin, and the recurring obstruction is that apparently mild symmetry or arithmetic hypotheses force a drastic reduction in possible rational maps.