Cremona Group¶

The group of birational self-transformations of projective n-space over a fixed field, equivalently the field automorphisms of a purely transcendental function field.

Version: v2 · 2026-09-19 · History
Domain-specific #: 1588
Domain group: Formal Sciences
Origin domain: Mathematics
Subdomains: Birational Geometry, Groups of Birational Transformations → Mathematics
Aliases: Cremona transformation group

Core Idea¶

Fix a field \(k\) and a positive integer \(n\). The Cremona group of rank \(n\) over \(k\), customarily written \(\operatorname{Cr}_n(k)\), is the group of birational self-transformations of projective \(n\)-space:

\[ \operatorname{Cr}_n(k)=\operatorname{Bir}_k(\mathbb P^n_k). \]

Equivalently, after choosing affine coordinates, it is the group of \(k\)-automorphisms of the purely transcendental function field:

\[ \operatorname{Cr}_n(k)\cong \operatorname{Aut}_k k(x_1,\ldots,x_n). \]

These are two presentations of the same object. A birational map is defined on a dense open part of projective space and has a rational inverse on another dense open part; it need not be defined, regular, or one-to-one at every point. In homogeneous coordinates an element can be represented by \(n+1\) homogeneous polynomials of the same degree with no common nonconstant factor, subject to the requirement that the induced rational map have a rational inverse. Two representatives that agree on a dense open set determine the same birational transformation.^[1]

Composition supplies the group law. Rational formulas are substituted and then common factors are cancelled; the identity rational map is neutral; a birational inverse supplies the group inverse. The group therefore gathers all rational changes of coordinates of one projective space into a single algebraic object. The fixed field and dimension are part of the identity: changing either can change elements, subgroups, conjugacy classes, generation results, and topology.

This abstraction is operationally autonomous, not just the observation that one named set happens to be a group. Researchers use \(\operatorname{Cr}_n(k)\) as the ambient object for factorizing birational maps, classifying finite and algebraic subgroups, comparing conjugacy classes, measuring degree growth, regularizing group actions on birational models, and testing algebraic or topological simplicity. Those practices recur across many transformations and varieties. The construct remains domain-specific because rational varieties, function fields, indeterminacy, blow-ups, and birational equivalence are constitutive rather than interchangeable cargo.

How would you explain it like I'm…

Move-and-Undo Club

Imagine rules made from math formulas that slide and stretch all the dots in a space to new spots. Each rule has an undo rule that puts them back, though at a few special spots the rule gets stuck and can't say where a dot goes. The Cremona group is the big collection of all these move-and-undo rules.

Reversible Formula Moves

Mathematicians like to study ways of moving the points of a space around using formulas made of fractions, where another such formula can undo the move. A formula like that might break at a few special spots, a bit like dividing by zero, and that's allowed, as long as it works almost everywhere. The Cremona group is the collection of all such undoable formula-moves on one kind of space called projective space. Doing one move and then another gives another move in the collection, and every move has an undo move, which is what makes it a "group."

Birational Maps of Projective Space

In algebra, a group is a collection of transformations you can combine and undo. The Cremona group collects all "birational" transformations of projective space: maps given by fractions of polynomials that have an inverse of the same kind. Unlike ordinary rearrangements, these maps don't have to be defined at every point or be one-to-one everywhere; they only need to work and be reversible on a large open part of the space. Two formulas that agree on such a large part count as the same map. Combining maps means substituting one formula into another and cancelling common factors. Which number system you use (the field) and how many dimensions you have are part of the definition; changing either one changes the group.

Fix a field k and a positive integer n. The Cremona group Cr_n(k) is the group of birational self-maps of projective n-space over k, equivalently the group of k-automorphisms of the rational function field k(x_1, ..., x_n); these are two presentations of the same object. A birational map is defined on a dense open subset and has a rational inverse defined on another dense open subset; it need not be regular or injective everywhere. In homogeneous coordinates, an element is given by n+1 homogeneous polynomials of equal degree with no common nonconstant factor, such that the induced rational map has a rational inverse, and representatives agreeing on a dense open set define the same element. The group law is composition: substitute formulas and cancel common factors, with the identity map as neutral element and the birational inverse as group inverse. Thus Cr_n(k) gathers all rational coordinate changes of one projective space into a single algebraic object. Researchers use it to factor birational maps, classify finite and algebraic subgroups, compare conjugacy classes, study degree growth, regularize group actions on birational models, and test algebraic or topological simplicity. Changing k or n can change the elements, subgroups, conjugacy classes, generation results, and topology.

Structural Signature¶

the base field \(k\) — the coefficients, rational points, field automorphisms, and admissible birational maps are all relative to it;
the dimension \(n\) — fixes the transcendence degree and the projective space \(\mathbb P^n_k\);
the rational model — projective \(n\)-space, or equivalently its function field \(k(x_1,\ldots,x_n)\);
the dense-open equality rule — rational maps agreeing wherever both are defined on some dense open subset represent the same map;
the birational self-map — a dominant rational map from the model to itself having a rational inverse;
the homogeneous presentation — same-degree coordinate polynomials without a common factor, modulo common nonzero scalar and equality as rational maps;
the indeterminacy locus — the proper subset where a chosen rational presentation does not define a point;
composition with reduction — substitution of rational formulas followed by removal of common factors;
the identity and birational inverse — the neutral map and two-sided inverse in the rational-map category;
the field-automorphism presentation — pullback on rational functions identifies the geometric group with \(\operatorname{Aut}_k k(x_1,\ldots,x_n)\), subject to the conventional covariance choice;
the regular subgroup — \(\operatorname{PGL}_{n+1}(k)=\operatorname{Aut}_k(\mathbb P^n_k)\) sits inside \(\operatorname{Cr}_n(k)\);
birational models and resolution — blow-ups or other birational modifications expose base points, exceptional loci, and induced actions;
group-level diagnostics — generators, relations, subgroups, normal subgroups, conjugacy, degree growth, and actions on geometric or combinatorial spaces.

Recognition test. A case instantiates this abstraction only if one field and dimension are fixed, every element is a birational self-map of the same projective space (or the corresponding function-field automorphism), and composition and rational inverses are the group operations. A lone rational map, maps between unrelated varieties, or only the regular projective automorphisms does not suffice.

What It Is Not¶

Not one Cremona transformation. An individual birational self-map is an element; the Cremona group is the full collection with composition.
Not the group of regular automorphisms alone. Degree-one projective transformations form \(\operatorname{PGL}_{n+1}(k)\). For \(n\ge 2\), nonlinear birational transformations make the Cremona group much larger.
Not every rational self-map. A dominant rational map such as a generic degree-two map may lack a rational inverse and therefore is not a group element.
Not all birational maps between varieties. The source and target must be the fixed \(\mathbb P^n_k\), or must be transported to it through a declared birational model.
Not an everywhere-defined action on points. An element can have indeterminacy points and contract subvarieties. The group law belongs to rational maps, not to total bijections of the underlying point set.
Not the general automorphism-group construction. It is the particular automorphism group of \(k(x_1,\ldots,x_n)\), equivalently the birational automorphism group of projective space, together with a specialized geometric practice.
Not the de Jonquières group. That is a subgroup preserving a rational fibration or pencil under specified conventions.
Not invariant under silently changing \(k\) or \(n\). Results over an algebraically closed field need not extend to the real numbers, finite fields, or nonclosed fields; the plane group is not a stand-in for higher rank.
Not a linear algebraic group in the ordinary finite-type sense. The Cremona group contains finite-dimensional algebraic subgroups, but the entire group is an infinite-dimensional transformation object with special topologies and parameter spaces.^[1]

Scope of Application¶

The primary scope is birational geometry. A rational variety \(X\) of dimension \(n\) is birational to \(\mathbb P^n_k\); a chosen birational map transports \(\operatorname{Bir}_k(X)\) into a conjugate presentation of \(\operatorname{Cr}_n(k)\). This lets geometers replace coordinate-heavy rational transformations by one ambient group while keeping track of how different models resolve indeterminacy.

In plane geometry, the group organizes classical factorization. Over an algebraically closed field, the Noether–Castelnuovo theorem generates \(\operatorname{Cr}_2(k)\) from \(\operatorname{PGL}_3(k)\) and the standard quadratic involution. The group is also the setting for homaloidal nets, base-point multiplicities, de Jonquières subgroups, finite subgroup classification, and presentations by generators and relations.^[1]

In algebraic dynamics, iterating one element produces a sequence of degrees. On surfaces, the asymptotic rate

\[ \lambda(f)=\lim_{m\to\infty}\deg(f^m)^{1/m} \]

is the first dynamical degree in the projective-plane setting. It is invariant under birational conjugacy and distinguishes bounded, polynomial, and exponential growth regimes under appropriate hypotheses.^[1] The group view turns iteration, conjugation, and invariant geometric structures into connected questions.

The group also supports geometric group theory and group-action questions: which abstract groups embed, which elements generate normal subgroups, how subgroups act on Picard–Manin hyperbolic space, and whether the ambient group is simple in an algebraic or topological sense. Higher dimensions bring qualitatively new exceptional loci and factorization behavior; they must not be inferred mechanically from the plane case.^[2]^[3]

Clarity¶

Use five checks. First, state \(k\) and \(n\). Second, identify either \(\mathbb P^n_k\) or \(k(x_1,\ldots,x_n)\). Third, verify rational invertibility, not merely dominance. Fourth, distinguish equality on dense opens from equality of one chosen coordinate tuple. Fifth, name whether a claim concerns abstract group structure, a topology on the group, degree-filtered parameter spaces, or an action on a birational model.

For example, the formula

\[ \sigma[x:y:z]=[yz:xz:xy] \]

defines a quadratic birational involution of \(\mathbb P^2\). It is undefined at the three coordinate points, yet \(\sigma^2\) equals the identity as a rational map after common factors are cancelled. Calling it a pointwise bijection of the whole projective plane is false; calling it an element of \(\operatorname{Cr}_2(k)\) is correct.

Dimension one is a diagnostic boundary. Every birational self-map of \(\mathbb P^1\) has degree one, hence \(\operatorname{Cr}_1(k)=\operatorname{PGL}_2(k)\). The nonlinear phenomenon begins in dimension at least two. A proposed example that uses a nonlinear invertible rational function in one variable has failed the birational-inverse check.

Manages Complexity¶

The construct compresses innumerable rational coordinate substitutions into one compositional system. Instead of treating each formula independently, one can ask whether it lies in a standard subgroup, how it factors into generators, which conjugacy class contains it, what its iterates do, and which invariants survive a change of birational coordinates.

Resolution of indeterminacy supplies a second compression. A plane birational map that is opaque as a rational formula can be lifted through blow-ups so that its effect on divisor classes becomes linear. The resulting action on Picard or Picard–Manin spaces records degrees and base-point multiplicities. This does not make the original map regular on \(\mathbb P^2\); it moves the analysis to a model whose geometry exposes the hidden operations.

The field-automorphism presentation supplies a third bridge. A geometric birational self-map and an algebraic automorphism of the rational-function field are not two unrelated topics. Switching presentations lets one use field theory to prove geometric facts or use exceptional divisors and linear systems to understand field automorphisms.

Abstract Reasoning¶

If a homogeneous presentation has degree one, its element lies in \(\operatorname{PGL}_{n+1}(k)\); the converse holds for regular automorphisms of projective space.
If a rational self-map lacks a rational inverse, closure in a transformation semigroup may hold, but membership in the Cremona group fails.
If \(f\) and \(g\) are represented by coordinate tuples, the raw degree of \(f\circ g\) can be below \(\deg(f)\deg(g)\) because substitution may create a common factor that must be cancelled.
If a property is preserved under conjugation \(f\mapsto hfh^{-1}\), it describes birational dynamics independently of the chosen rational coordinate frame.
If an element has indeterminacy, that does not defeat group membership; it instead requires interpreting inverse and composition on dense open sets.
If a subgroup preserves a pencil or rational fibration, it may lie in a conjugate of a de Jonquières-type subgroup; preservation is extra structure, not automatic for all Cremona elements.
If a claim proved for \(\operatorname{Cr}_2(\mathbb C)\) uses algebraic closure or surface resolution, it cannot be exported to \(\operatorname{Cr}_n(k)\) without checking field and dimension hypotheses.
If \(\deg(f^m)\) grows exponentially, its dynamical degree exceeds one; bounded or polynomial degree growth belongs to different dynamical regimes.
If \(\operatorname{Cr}_n(k)\) is called “simple,” the relevant category must be named: absence of abstract normal subgroups and absence of closed normal subgroups in a chosen topology are different assertions.

Knowledge Transfer¶

Exact transfer occurs among rational models of the same function field. If \(X\) is \(k\)-rational and \(\phi:X\dashrightarrow\mathbb P^n_k\) is birational, then conjugation by \(\phi\) identifies \(\operatorname{Bir}_k(X)\) with a presentation of \(\operatorname{Cr}_n(k)\). The isomorphism depends on the chosen birational identification, while conjugacy-invariant conclusions do not.

Many tools transfer to other birational automorphism groups: resolve indeterminacy, study induced actions on divisor groups, classify elements by growth, and analyze subgroups through preserved fibrations or geometric models. The exact label Cremona group should normally remain reserved for the rational function field or projective-space case.

Cross-domain metaphor is weaker. A collection of reversible transformations in physics, computation, or design may instantiate the prime Group or Symmetry, but it does not instantiate a Cremona group without a projective rational variety, function field, birational maps, and rational inverses. The portable residue is composable reversible transformation; the Cremona identity is mathematical cargo.

Examples¶

Dimension one. For \(\mathbb P^1_k\), every element is a linear fractional transformation

\[ x\longmapsto \frac{ax+b}{cx+d},\qquad ad-bc\ne0, \]

with scalar-equivalent matrices identified. Thus \(\operatorname{Cr}_1(k)=\operatorname{PGL}_2(k)\). This boundary case contains no genuinely nonlinear birational self-map.^[1]

The standard quadratic involution. In \(\operatorname{Cr}_2(k)\), \([x:y:z]\mapsto[yz:xz:xy]\) has degree two and three base points. On the dense torus where \(xyz\ne0\), it is equivalent to coordinate inversion, and applying it twice returns the original point as a rational map. Together with projective linear transformations it generates the plane Cremona group over an algebraically closed field.^[1]

A monomial element. An integer matrix \(A=(a_{ij})\in\operatorname{GL}_n(\mathbb Z)\) defines a birational transformation of the algebraic torus by monomials. Extending it rationally to projective space embeds \(\operatorname{GL}_n(\mathbb Z)\) into \(\operatorname{Cr}_n(k)\). Invertibility of the exponent matrix is the rational-inverse certificate.

A non-example. The map \(\mathbb P^1\to\mathbb P^1\), \(x\mapsto x^2\), is dominant but generically two-to-one when characteristic does not collapse the example. It has no rational inverse and is not in \(\operatorname{Cr}_1(k)\).

Degree growth. Iterating a plane Cremona transformation can keep degree bounded, make it grow linearly or quadratically, or make it grow exponentially. The limit defining \(\lambda(f)\) converts the sequence into a conjugacy invariant and connects the rational formulas to an isometric action on Picard–Manin space.^[1]

Structural Tensions¶

Formula versus rational-map class. Coordinate tuples are concrete, but common factors and dense-open equality make them nonunique. Diagnose at the function-field or rational-map level.
Group inverse versus pointwise bijection. Every element has a birational inverse, while individual points may be indeterminate or lie on contracted loci. Diagnose on compatible dense opens.
Regular versus birational symmetry. \(\operatorname{PGL}_{n+1}\) is computationally familiar but omits nonlinear transformations. Diagnose by degree and base loci.
Fixed model versus resolved model. Blow-ups clarify a transformation but change the variety on which it becomes regular. Track the conjugating birational maps explicitly.
Dimension two versus higher dimension. Surface methods and generators are unusually strong. Diagnose every theorem for its dimension assumptions.
Algebraic versus topological simplicity. Nontrivial normal subgroups can coexist with topological simplicity when those subgroups are not closed. State the topology before using “simple.”^[4]^[3]
Instance versus autonomous abstraction. The group is an automorphism-group instance, yet its base-point calculus, factorization, degree-growth, subgroup, and birational-model practices form a stable specialist apparatus. Retain the node only with those roles, not as a bare proper name.

Structural–Framed Character¶

Structural; aggregate 0.18. Field, dimension, rational inverse, dense-open equality, and composition determine membership without institutional or evaluative judgment. The same definition is independently recoverable from projective geometry and function-field algebra.

It nevertheless remains domain-specific. “Projective space,” “rational function field,” “birational,” “indeterminacy,” and “blow-up” are literal mathematical commitments. Removing them leaves the already-cataloged Group, Automorphism Group, Transformation, and Symmetry structures rather than a new prime.

Structural Core vs. Domain Accent¶

The structural core is all admissible reversible self-transformations of one object, closed under composition. This core is supplied by Automorphism Group and ultimately Group. Conjugacy, subgroup, orbit, generator, and invariant reasoning then follows from those general structures.

The domain accent is what selects the transformations and makes the node useful: the object is \(k(x_1,\ldots,x_n)\) or \(\mathbb P^n_k\); equality is rational equality on dense opens; admissibility means birationality; coordinates may have base loci; blow-ups expose the action; degree and exceptional divisors supply diagnostics. Those conditions are not decoration. They distinguish the Cremona group from regular projective automorphisms, generic automorphism groups, and arbitrary rational-map semigroups.

This entry is a kind of Group.

Group is the broader abstraction: composition is associative, the identity is present, and every element has a rational inverse. Algebraically, \(\operatorname{Cr}_n(k)\) sits still more tightly under Automorphism Group, because it equals \(\operatorname{Aut}_k k(x_1,\ldots,x_n)\).

Transformation describes the element-level mappings, while Symmetry explains invariants and conjugacy. Resolution, Iteration, and Growth illuminate common lines of analysis.

Relationships to Other Abstractions¶

Current abstraction Cremona Group Domain-specific

Parents (1) — outgoing typed relations

Cremona Group is a kind of Group Prime

The minimal currently valid the broader abstraction is the live prime Group.

Neighborhood in Abstraction Space¶

Cremona Group sits in a sparse region of the domain-specific corpus (85^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (2769 abstractions)

Nearest neighbors

Algebraic Variety — 0.84
Grothendieck–Riemann–Roch theorem — 0.81
N-Square Identity — 0.81
Norm Form — 0.80
Algebraic Cycle — 0.80

Computed from structural-signature embeddings · 2026-10-25

Not to Be Confused With¶

Automorphism Group is the closest accepted workspace abstraction but is not yet present in the validator's frozen catalog. It gives the general construction \(X\mapsto\operatorname{Aut}(X)\); it does not specify the rational function field, projective geometry, base loci, degree growth, or dimension-sensitive theorems. Eventual implementation should re-evaluate it as a tighter parent.

Group covers reversible composability but not which transformations qualify. Symmetry covers invariance under transformations but not the full birational self-map group. Transformation covers individual mappings rather than their complete group.

\(\operatorname{PGL}_{n+1}(k)\) is the regular degree-one subgroup. A de Jonquières group preserves a fibration. The birational automorphism group \(\operatorname{Bir}(X)\) of a nonrational variety is not normally called a Cremona group. A semigroup of dominant rational self-maps may contain noninvertible elements. Each neighbor omits a load-bearing role.

References¶

[1] Serge Cantat. “The Cremona Group.” In Algebraic Geometry: Salt Lake City 2015, Proceedings of Symposia in Pure Mathematics 97.1, American Mathematical Society, 2018, pp. 101–142. Defines the equivalent function-field, affine, and projective incarnations; treats indeterminacy, subgroups, plane generators, Picard–Manin actions, and degree growth. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g

[2] Jérémy Blanc, Stéphane Lamy, and Susanna Zimmermann. “Quotients of Higher-Dimensional Cremona Groups.” Acta Mathematica 226.2 (2021): 211–318. Gives higher-dimensional quotient and nonsimplicity results under stated field hypotheses. registry ↩ Show verification details

[3] Serge Cantat and Stéphane Lamy. “Normal Subgroups in the Cremona Group.” Acta Mathematica 210 (2013): 31–94. Establishes normal-subgroup phenomena for the plane Cremona group. registry ↩a ↩b

[4] Jérémy Blanc and Susanna Zimmermann. “Topological Simplicity of the Cremona Groups.” American Journal of Mathematics 140.5 (2018): 1297–1309. Distinguishes topological simplicity from abstract simplicity. registry ↩ Show verification details

[5] Maria Alberich-Carramiñana. Geometry of the Plane Cremona Maps. Lecture Notes in Mathematics 1769, Springer, 2002. Systematic treatment of plane Cremona maps, base points, and linear systems. registry