How statement and proof provenance work
The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.
- Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
- AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
- AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.
These labels describe origin, not correctness: citations and verification chips remain separate evidence.
Cayley Graphs, Word Metrics and Quasi-Isometry
1 · Prerequisites
- Binary Operations, Monoids, Groups and Subgroups
- Completeness, Completion, and Uniform Continuity
- Construction of the Natural Numbers
- Construction of the Real Numbers via Cauchy Sequences
- Construction of the Real Numbers via Dedekind Cuts
- Cosets, Index and Lagrange's Theorem
- Countability and Uncountability
- Finite Counting, Factorials and Binomial Coefficients
- Foundations of the Real Numbers for Analysis
- Free Groups and Presentations
- Graphs, Walks and Connectivity
- Group Actions, Orbits, Stabilisers and Cayley's Theorem
- Group Homomorphisms and the Isomorphism Theorems
- Limits of Real Functions
- Metric Spaces
- Monotone Sequences, Bolzano-Weierstrass, and Cauchy Completeness
- Normal Subgroups and Quotient Groups
- Order, Zorn's Lemma, and the Axiom of Choice
- Relations, Functions, and Quotients
- Roots, Rational Powers, and Classical Inequalities
- Sequences and Limits
- Suprema and Infima
- The ZFC Axioms and the Basic Set Constructions
2 · Summary
Published finite simple graph conventions, the walk and path definitions, and the tree characterisation supply the combinatorial background, while the group-action page supplies Cayley's theorem: every group is isomorphic to a subgroup of and the algebra pages supply Free group on a set of generators, A free basis of a group, and semidirect products. The metric background comes from Metric space: iff , symmetry, and the triangle inequality; pseudometric and ultrametric, Isometry, isometric embedding, and the subspace metric on a subset, Lipschitz map, -Hölder map for rational , and contraction, and the basic boundedness and ball conventions that the word metric and quasi-isometry language reuse.
The page first removes the finiteness restriction from graphs, defines the path metric, and compares it with the published finite-graph distance. It then defines directed and undirected Cayley graphs, proves their connectedness, regularity, and left-translation symmetry, and identifies the word metric with the path metric of the Cayley graph. The free-group section characterises tree Cayley graphs through reduced words. The final section defines coarse Lipschitz maps, quasi-isometries, separated nets, quasi-geodesics, and quasi-isometry type, proves the equivalence and invariance statements among those notions, and records finiteness as a geometric property of finitely generated groups.
3 · Logical flowchart
4 · Definitions, theorems and proofs
Simple graphs on an arbitrary vertex set
Definition
A simple graph is an ordered pair in which is a set and
The elements of are the vertices and the elements of are the edges. Thus a simple graph has no loops and no parallel edges.
This is the same incidence convention as A finite simple graph is a finite vertex set together with a set of two-element vertex subsets, with the word "finite" removed from the hypothesis on . Whenever a statement on this page is made for a simple graph, it allows an arbitrary vertex set unless finiteness is stated separately.
Walks, paths, connectedness and components in a simple graph on an arbitrary vertex set
Definition
Let be a simple graph in the sense of Simple graphs on an arbitrary vertex set.
A walk, closed walk, trail, path, and cycle in are defined exactly as in Walks, closed walks, trails, paths and cycles, with length equal to the number of traversed edges, with no finiteness hypothesis on . In particular, a walk is a finite vertex list with for each , and a path is a walk with distinct vertices.
Vertices are connected when some path in has endpoints and . The graph is connected when and every two vertices of are connected.
For , the connected component of is the induced subgraph on
These are the same notions as in Connected graphs and connected components defined by the existence of vertex paths, again with the finiteness hypothesis removed.
Every walk contains a path between the same endpoints, of no greater length
Statement
Every walk contains a path between the same endpoints, of no greater length.
Facts & Assumptions
Given: The hypotheses of the Statement.
A simple graph is a pair with any set and a set of two-element subsets of (Simple graphs on an arbitrary vertex set).
A walk of length in a simple graph is a finite vertex list with consecutive vertices adjacent; a path is a walk with distinct vertices; the graph is connected when it is nonempty and every two vertices are joined by a path (Walks, paths, connectedness and components in a simple graph on an arbitrary vertex set).
Every nonempty subset has a least element: there is with for all . (The well-ordering principle).
Proof
Among the walks from to of length at most that of the given one, the well-ordering principle selects one of least length.
If a vertex occurred twice in the selected walk, deleting the segment between the two occurrences would give a shorter such walk, contradicting minimality.
So the selected walk repeats no vertex and is a path with the same endpoints and no greater length; finiteness of the vertex set is never used.
The path metric of a connected simple graph
Definition
Let be a connected simple graph. For vertices , connectedness supplies at least one path from to , and every such path has a length in . By The well-ordering principle, the set of these lengths has a least element. The path metric of is therefore the function
Here is the number of edges traversed by the path , as in Walks, closed walks, trails, paths and cycles, with length equal to the number of traversed edges. The proof that is a metric in the sense of Metric space: iff , symmetry, and the triangle inequality; pseudometric and ultrametric is The path metric of a connected simple graph is a metric on its vertex set.
The path metric of a connected simple graph is a metric on its vertex set
Statement
The path metric of a connected simple graph is a metric on its vertex set.
Facts & Assumptions
Given: The hypotheses of the Statement.
The path metric of a connected simple graph assigns to two vertices the least length of a path joining them (The path metric of a connected simple graph).
A walk of length in a simple graph is a finite vertex list with consecutive vertices adjacent; a path is a walk with distinct vertices; the graph is connected when it is nonempty and every two vertices are joined by a path (Walks, paths, connectedness and components in a simple graph on an arbitrary vertex set).
Every walk in a simple graph contains a path with the same endpoints and of no greater length (Every walk contains a path between the same endpoints, of no greater length).
A metric on a set satisfies separation, symmetry and the triangle inequality (Metric space: iff , symmetry, and the triangle inequality; pseudometric and ultrametric).
A simple graph is a pair with any set and a set of two-element subsets of (Simple graphs on an arbitrary vertex set).
Proof
The only path of length zero joins a vertex to itself, so the distance vanishes exactly on the diagonal.
Reversing a path preserves its length, so the distance is symmetric.
Concatenating two shortest paths gives a walk of the summed length, which the previous lemma replaces by a path no longer, giving the triangle inequality.
Steps 1.1, 1.2 and 1.3 establish separation, symmetry and the triangle inequality, so the path metric is a metric.
A bijection of vertex sets is an isometry for the path metrics if and only if it is a graph isomorphism
Statement
Let and be connected simple graphs and let be a bijection. Then is an isometry for the path metrics if and only if it is a graph isomorphism.
Facts & Assumptions
Given: Connected simple graphs and a bijection .
A simple graph is a pair with any set and a set of two-element subsets of (Simple graphs on an arbitrary vertex set).
A walk of length in a simple graph is a finite vertex list with consecutive vertices adjacent; a path is a walk with distinct vertices; the graph is connected when it is nonempty and every two vertices are joined by a path (Walks, paths, connectedness and components in a simple graph on an arbitrary vertex set).
The path metric of a connected simple graph assigns to two vertices the least length of a path joining them (The path metric of a connected simple graph).
The path metric of a connected simple graph is a metric on its vertex set (The path metric of a connected simple graph is a metric on its vertex set).
A function is an isometric embedding when it preserves every distance, and it is an isometry when it is also bijective (Isometry, isometric embedding, and the subspace metric on a subset).
A graph isomorphism is a vertex bijection such that is an edge exactly when is an edge, for all distinct (Graph isomorphisms, automorphisms and graph complements).
Proof
A graph isomorphism carries paths to paths of the same length in both directions, so it preserves the minima defining the metric.
Conversely an isometry preserves distance one, and distance one is exactly adjacency, so it preserves the edge relation in both directions.
On a finite vertex set the graph notions agree, and on connected graphs the two path distances agree
Statement
On a finite vertex set the simple-graph, walk, path, cycle, connectedness and component notions on this page agree with the published finite-graph notions. If the graph is connected, its path metric agrees with the published graph distance; more generally, the same equality holds after restricting to any connected component.
Facts & Assumptions
Given: The hypotheses of the Statement.
A simple graph is a pair with any set and a set of two-element subsets of (Simple graphs on an arbitrary vertex set).
A walk of length in a simple graph is a finite vertex list with consecutive vertices adjacent; a path is a walk with distinct vertices; the graph is connected when it is nonempty and every two vertices are joined by a path (Walks, paths, connectedness and components in a simple graph on an arbitrary vertex set).
A walk of length is a finite vertex list (Walks, closed walks, trails, paths and cycles, with length equal to the number of traversed edges).
Vertices and of a graph are connected, or reachable from one another, when a path in has endpoints and . Equivalently, a walk joins them, because turns such a walk into a path. (Connected graphs and connected components defined by the existence of vertex paths).
For vertices in one connected component, the published graph distance is the least length of a path joining them (Graph distance within a component, eccentricity, diameter and girth, including the acyclic convention).
These are the metric axioms, so shortest-path distance is a metric on . (Shortest-path distance is a metric on every connected component).
By deleting zero or more closed segments from its vertex list, one obtains a path from to of length at most . If repeats a vertex, the resulting path can be chosen to have length strictly less than . (Every walk between two vertices contains a path between the same endpoints).
The path metric of a connected simple graph assigns to two vertices the least length of a path joining them (The path metric of a connected simple graph).
The path metric of a connected simple graph is a metric on its vertex set (The path metric of a connected simple graph is a metric on its vertex set).
Every walk in a simple graph contains a path with the same endpoints and of no greater length (Every walk contains a path between the same endpoints, of no greater length).
Proof
A simple graph whose vertex set is finite is a finite simple graph in the published sense, clause by clause.
The walk, path, cycle, connectedness and component notions defined here read verbatim as the published ones on a finite vertex set, and the walk-to-path lemma specialises to the published one.
The published graph distance is the least length of a path in the same sense. It therefore agrees with the path metric when the graph is connected and, for a disconnected graph, with the path metric of each connected component. The metric statement specialises to the published component-wise one.
Locally finite graphs and vertex degree without a finiteness hypothesis
Definition
Let be a simple graph and let . Its neighbourhood is
If is finite, its cardinality is the degree of and is written
The graph is locally finite if is finite for every vertex . Equivalently, every vertex has finite degree.
In a connected locally finite graph every ball of the path metric is finite
Statement
In a connected locally finite graph every ball of the path metric is finite.
Facts & Assumptions
Given: The hypotheses of the Statement.
A graph is locally finite when every vertex has finitely many neighbours (Locally finite graphs and vertex degree without a finiteness hypothesis).
The path metric of a connected simple graph assigns to two vertices the least length of a path joining them (The path metric of a connected simple graph).
is the open ball, the closed ball and the sphere of centre and radius . The radius is always a strictly positive real; a ball of radius or of negative radius is never written in this library. (Open ball, closed ball and sphere in a metric space).
A walk of length in a simple graph is a finite vertex list with consecutive vertices adjacent; a path is a walk with distinct vertices; the graph is connected when it is nonempty and every two vertices are joined by a path (Walks, paths, connectedness and components in a simple graph on an arbitrary vertex set).
A set is finite when for some . (The cardinality of a finite set).
Every complete ordered field is Archimedean: for every real there is a natural number with (Every complete ordered field is Archimedean).
A subset of a finite set is finite (A subset of a finite set is finite, with , and equality holds if and only if ).
Proof
Fix a vertex . For each natural number let . Then , so is finite.
Assume is finite. If , choose a path from to of minimal length . If then . If , then and lies in . Hence . Each set is finite by local finiteness, so the right-hand side is a finite union of finite sets and is therefore finite; thus is finite.
By step 1.2, if is finite then so is .
Therefore every is finite. Now let . By the Archimedean property choose a natural number with . Because is the length of a path, it is a natural number; so implies . Hence , and [L6] makes finite.
Cycles, trees and forests in a simple graph on an arbitrary vertex set
Definition
Let be a simple graph.
A cycle in is a cycle in the sense of Walks, closed walks, trails, paths and cycles, with length equal to the number of traversed edges, that is, a closed walk of length at least whose vertices are distinct except for the repeated endpoint.
The graph is acyclic if it contains no cycle.
A tree is a connected acyclic simple graph.
A forest is an acyclic simple graph.
A nonempty simple graph is a tree if and only if each pair of vertices is joined by exactly one path
Statement
A nonempty simple graph is a tree if and only if each pair of vertices is joined by exactly one path.
Facts & Assumptions
Given: A nonempty simple graph.
A cycle is a closed walk of length at least three with distinct vertices apart from its endpoints; a forest is a simple graph with no cycle and a tree is a connected forest (Cycles, trees and forests in a simple graph on an arbitrary vertex set).
A walk of length in a simple graph is a finite vertex list with consecutive vertices adjacent; a path is a walk with distinct vertices; the graph is connected when it is nonempty and every two vertices are joined by a path (Walks, paths, connectedness and components in a simple graph on an arbitrary vertex set).
A simple graph is a pair with any set and a set of two-element subsets of (Simple graphs on an arbitrary vertex set).
Every walk in a simple graph contains a path with the same endpoints and of no greater length (Every walk contains a path between the same endpoints, of no greater length).
Proof
If paths are unique, nonemptiness together with the existence of a path between each pair makes the graph connected; a cycle of length at least three would give two distinct paths between two of its vertices.
Conversely, in a tree two distinct paths with the same endpoints first diverge at some index and first meet again later, and the two segments between those points form a cycle.
Labelled directed graphs, their underlying simple graphs, and label-preserving isomorphisms
Definition
Let be a set of labels. A labelled directed graph with label set is a pair consisting of a vertex set and a set of labelled arcs
An element is an arc from to labelled by . Distinct labels may connect the same ordered pair of vertices; this is why a labelled directed graph is treated here as a digraph variant rather than as a simple graph.
The underlying simple graph of is the simple graph on whose edge set consists of the unordered pairs for which some labelled arc or exists, with loops discarded.
If and are labelled directed graphs with the same label set , a label-preserving directed graph isomorphism is a bijection such that
for all and .
The directed labelled Cayley graph of a group with respect to a subset
Definition
Let be a group and let . The directed labelled Cayley graph of with respect to is the labelled directed graph
with vertex set and labelled arc set
Thus each generator supplies from every vertex a directed edge labelled to the vertex . If , the graph has a labelled loop at every vertex.
The Cayley graph of a group with respect to a subset
Definition
Let be a group with identity and let . The Cayley graph of with respect to is the simple graph
with vertex set and edge set
Equivalently, it is the underlying simple graph of The directed labelled Cayley graph of a group with respect to a subset after forgetting directions and labels and then deleting the loops coming from the identity. The exclusion of is load bearing: it is what keeps the Cayley graph a simple graph in the sense of Simple graphs on an arbitrary vertex set.
Finitely generated groups
Definition
Let be a group. A subset generates if in the sense of The subgroup generated by a subset, the cyclic subgroup , and cyclic groups.
The group is finitely generated if some finite subset generates . Such an is a finite generating set of .
A Cayley graph is connected if and only if the subset generates the group
Statement
A Cayley graph is connected if and only if the subset generates the group.
Facts & Assumptions
Given: The hypotheses of the Statement.
The Cayley graph of a group with respect to a subset has vertex set and edge set (The Cayley graph of a group with respect to a subset).
A walk of length in a simple graph is a finite vertex list with consecutive vertices adjacent; a path is a walk with distinct vertices; the graph is connected when it is nonempty and every two vertices are joined by a path (Walks, paths, connectedness and components in a simple graph on an arbitrary vertex set).
An elementary cancellation deletes two adjacent letters or . A word is reduced if no elementary cancellation applies. (Words in an alphabet with formal inverses, elementary cancellation, and reduced words).
Proof
A walk from the identity to spells an expression for as a product of elements of and their inverses, and each such expression is a walk.
So the component of the identity is exactly the generated subgroup, and left translation carries it onto the component of any vertex; connectedness therefore means the subgroup is everything.
Cayley-graph neighbourhoods are equipotent, and local finiteness is equivalent to finiteness of the symmetrised subset
Statement
Let be a group, , and . Left translation gives a bijection between the neighbourhoods of any two vertices of . The graph is locally finite exactly when is finite; in that case it is regular of finite degree .
Facts & Assumptions
Given: A group , a subset , and .
The Cayley graph of a group with respect to a subset has vertex set and edge set (The Cayley graph of a group with respect to a subset).
A graph is locally finite when every vertex has finitely many neighbours (Locally finite graphs and vertex degree without a finiteness hypothesis).
The degree of is , equivalently the number of edges incident with . A graph is -regular when every vertex has degree ; it is cubic when it is -regular. (Adjacency, incidence, open and closed neighbourhoods, vertex degree, minimum degree and maximum degree).
A set is finite when for some . (The cardinality of a finite set).
Proof
The neighbours of are the elements with in the symmetrised set minus the identity, and left multiplication by is a bijection from the neighbours of to those of .
Thus one neighbourhood is finite exactly when all are, which occurs exactly when is finite. In that case the degree is defined at every vertex and equals , so the graph is regular of that finite degree.
Left translation acts on a Cayley graph by label-preserving automorphisms, freely on vertices
Statement
Left translation acts on a Cayley graph by label-preserving automorphisms, freely on vertices.
Facts & Assumptions
Given: The hypotheses of the Statement.
The directed labelled Cayley graph of a group and a subset has vertex set and an arc from to labelled for each and (The directed labelled Cayley graph of a group with respect to a subset).
The Cayley graph of a group with respect to a subset has vertex set and edge set (The Cayley graph of a group with respect to a subset).
A labelled directed graph is a vertex set with a set of arcs each carrying a label, and a labelled isomorphism is a vertex bijection preserving arcs and labels (Labelled directed graphs, their underlying simple graphs, and label-preserving isomorphisms).
A left action of on is a function , written , such that (Left group actions, transitive actions, and faithful actions).
A left action of a group on a set is free when (A free group action has no nonidentity element fixing a point).
Proof
Left multiplication by sends the arc from to to the arc from to , which carries the same label.
It is bijective on vertices with inverse left multiplication by , and the assignment is a homomorphism into the label-preserving automorphism group.
It fixes a vertex only when , that is only when is the identity, so the action is free.
The label-preserving automorphism action on a Cayley graph is the left regular representation of Cayley's theorem
Statement
The label-preserving automorphism action on a Cayley graph is the left regular representation of Cayley's theorem.
Facts & Assumptions
Given: The hypotheses of the Statement.
Left translation acts on a Cayley graph by label-preserving automorphisms, freely on vertices (Left translation acts on a Cayley graph by label-preserving automorphisms, freely on vertices).
The Cayley graph of a group with respect to a subset has vertex set and edge set (The Cayley graph of a group with respect to a subset).
Every group is isomorphic to the subgroup of formed by its left translations . (Cayley's theorem: every group is isomorphic to a subgroup of ).
A left action of on is a function , written , such that (Left group actions, transitive actions, and faithful actions).
Group isomorphisms, automorphisms and the set . (Group isomorphisms, automorphisms and the set ).
Proof
Cayley’s theorem embeds a group in the symmetric group on its underlying set by left multiplication, and the action just constructed has exactly those permutations.
So the label-preserving automorphism action is that embedding followed by the inclusion of the automorphism group in the symmetric group; the Cayley graph refines the regular action rather than supplying a second embedding theorem.
For a finite group the Cayley graph is a finite simple graph in the published sense and the two distances agree
Statement
Let be a finite group and let be a finite generating set. Then is a connected finite simple graph in the published sense, and its path metric agrees with the published graph distance.
Facts & Assumptions
Given: A finite group and a finite generating set .
The Cayley graph of a group with respect to a subset has vertex set and edge set (The Cayley graph of a group with respect to a subset).
On a finite vertex set these graph notions are the published ones and the path metric is the published graph distance (On a finite vertex set the graph notions agree, and on connected graphs the two path distances agree).
The path metric of a connected simple graph assigns to two vertices the least length of a path joining them (The path metric of a connected simple graph).
Let and lie in the same connected component of a graph . Their distance is (Graph distance within a component, eccentricity, diameter and girth, including the acyclic convention).
A set is finite when for some . (The cardinality of a finite set).
Proof
The vertex set is the group, which is finite, and the edges are two-element subsets, so the published definition applies verbatim; because generates , the Cayley graph is connected.
Both distances are the least length of a path in the same sense, so they agree.
Word length of a group element with respect to a generating set
Definition
Let be a group and let be a generating set. For , consider the set
This set is nonempty because generates . Indeed, the set of all finite products of elements of contains the identity, is closed under products and inverses, and contains , so it is a subgroup containing ; conversely every subgroup containing contains all such products. It is therefore exactly by The subgroup generated by a subset, the cyclic subgroup , and cyclic groups. By The well-ordering principle, it has a least element. The word length of with respect to is that least element and is written
Thus exactly when is the empty product, that is, the identity of . When is finite, such expressions are obtained by evaluating words in the formal alphabet of Words in an alphabet with formal inverses, elementary cancellation, and reduced words; distinct formal words can evaluate to the same product when two letters represent the same group element of .
Word length is defined on every element and satisfies the subadditivity, inversion and vanishing laws
Statement
Word length is defined on every element and satisfies the subadditivity, inversion and vanishing laws.
Facts & Assumptions
Given: The hypotheses of the Statement.
The word length is the least such that is a product of elements of (Word length of a group element with respect to a generating set).
Every nonempty subset has a least element: there is with for all . (The well-ordering principle).
Proof
If generates, the set of lengths of expressions of is a nonempty subset of the natural numbers, so it has a least element.
Concatenating expressions gives subadditivity, and reversing an expression while inverting each letter gives equality of the lengths of and .
The empty expression has length zero and represents only the identity, so word length vanishes exactly there.
The word metric of a group with respect to a generating set
Definition
Let be a group and let be a generating set. The word metric of with respect to is the function
By Word length is defined on every element and satisfies the subadditivity, inversion and vanishing laws, this function satisfies the metric axioms of Metric space: iff , symmetry, and the triangle inequality; pseudometric and ultrametric. Thus is a metric space, called the word metric space of with respect to .
The word metric is a left-invariant metric and coincides with the path metric of the Cayley graph
Statement
The word metric is a left-invariant metric and coincides with the path metric of the Cayley graph.
Facts & Assumptions
Given: The hypotheses of the Statement.
The word metric of with respect to is (The word metric of a group with respect to a generating set).
Word length is defined on every element and satisfies , , and exactly when is the identity (Word length is defined on every element and satisfies the subadditivity, inversion and vanishing laws).
The word length is the least such that is a product of elements of (Word length of a group element with respect to a generating set).
A metric on a set satisfies separation, symmetry and the triangle inequality (Metric space: iff , symmetry, and the triangle inequality; pseudometric and ultrametric).
The Cayley graph of a group with respect to a subset has vertex set and edge set (The Cayley graph of a group with respect to a subset).
A Cayley graph is connected if and only if its defining subset generates the group (A Cayley graph is connected if and only if the subset generates the group).
The path metric of a connected simple graph assigns to two vertices the least length of a path joining them (The path metric of a connected simple graph).
The path metric of a connected simple graph is a metric on its vertex set (The path metric of a connected simple graph is a metric on its vertex set).
Proof
The three length laws transported by are exactly the three metric axioms.
Left invariance is immediate, since .
A walk of length from to in the Cayley graph is the same datum as an expression for of length once identity generators are discarded, so the two minima agree.
Balls of a word metric are finite if and only if the generating set is finite
Statement
Balls of a word metric are finite if and only if the generating set is finite.
Facts & Assumptions
Given: The hypotheses of the Statement.
The word metric of with respect to is (The word metric of a group with respect to a generating set).
The word length is the least such that is a product of elements of (Word length of a group element with respect to a generating set).
The word metric is a left-invariant metric and coincides with the path metric of the Cayley graph (The word metric is a left-invariant metric and coincides with the path metric of the Cayley graph).
In a locally finite connected graph every ball of the path metric is finite (In a connected locally finite graph every ball of the path metric is finite).
Every vertex of a Cayley graph has the same degree, and the graph is locally finite exactly when the symmetrised generating set is finite (Cayley-graph neighbourhoods are equipotent, and local finiteness is equivalent to finiteness of the symmetrised subset).
is the open ball, the closed ball and the sphere of centre and radius . The radius is always a strictly positive real; a ball of radius or of negative radius is never written in this library. (Open ball, closed ball and sphere in a metric space).
A set is finite when for some . (The cardinality of a finite set).
A group is finitely generated when some finite subset generates it (Finitely generated groups).
Write Then is finite and , the right-hand product being the -valued one of. (The product rule: , and ).
Proof
If is finite the Cayley graph is locally finite, so balls of its path metric are finite; left invariance moves this to every centre.
If is infinite then the open ball of radius about the identity contains every element of , because each such element has word length ; so that ball is infinite.
The word metric is the largest left-invariant metric in which each generator and its inverse lie within distance one of the identity
Statement
The word metric is the largest left-invariant metric in which each generator and its inverse lie within distance one of the identity.
Facts & Assumptions
Given: The hypotheses of the Statement.
The word metric of with respect to is (The word metric of a group with respect to a generating set).
The word length is the least such that is a product of elements of (Word length of a group element with respect to a generating set).
Word length is defined on every element and satisfies , , and exactly when is the identity (Word length is defined on every element and satisfies the subadditivity, inversion and vanishing laws).
The word metric is a left-invariant metric and coincides with the path metric of the Cayley graph (The word metric is a left-invariant metric and coincides with the path metric of the Cayley graph).
Every metric satisfies the triangle inequality (Metric space: iff , symmetry, and the triangle inequality; pseudometric and ultrametric).
Proof
For a competing left-invariant metric with the stated property, the triangle inequality along a shortest expression bounds the competing distance by the word length.
The word metric itself gives every symmetrised generator distance one from the identity, so it satisfies the constraint and dominates every competitor.
Right translation by a fixed element displaces every point of a word metric space by exactly the word length of that element
Statement
Right translation by a fixed element displaces every point of a word metric space by exactly the word length of that element.
Facts & Assumptions
Given: The hypotheses of the Statement.
The word metric of with respect to is (The word metric of a group with respect to a generating set).
The word length is the least such that is a product of elements of (Word length of a group element with respect to a generating set).
The word metric is a left-invariant metric and coincides with the path metric of the Cayley graph (The word metric is a left-invariant metric and coincides with the path metric of the Cayley graph).
Proof
Left invariance gives for every , so right translation by displaces every point by the same amount.
Hence right translation by is at bounded distance from the identity map, and it is the identity exactly when that displacement is zero.
The Cayley graph of a free group with respect to a free basis is a tree
Statement
The Cayley graph of a free group with respect to a free basis is a tree.
Facts & Assumptions
Given: The hypotheses of the Statement.
The Cayley graph of a group with respect to a subset has vertex set and edge set (The Cayley graph of a group with respect to a subset).
A Cayley graph is connected if and only if its defining subset generates the group (A Cayley graph is connected if and only if the subset generates the group).
A cycle is a closed walk of length at least three with distinct vertices apart from its endpoints; a forest is a simple graph with no cycle and a tree is a connected forest (Cycles, trees and forests in a simple graph on an arbitrary vertex set).
A simple graph is a tree if and only if every two of its vertices are joined by exactly one path (A nonempty simple graph is a tree if and only if each pair of vertices is joined by exactly one path).
A free group on a set is a group together with a map such that, for every group and every function , there is a unique group homomorphism satisfying (Free group on a set of generators).
The subset is a free basis of if is a free group on the set in the sense of. (A free basis of a group).
An elementary cancellation deletes two adjacent letters or . A word is reduced if no elementary cancellation applies. (Words in an alphabet with formal inverses, elementary cancellation, and reduced words).
The reduced words on form a group when the product of reduced words is their concatenation followed by free reduction. (Reduced words form the free group on an alphabet).
Every class in contains exactly one reduced word. (Every class in contains exactly one reduced word).
Proof
A free basis generates, so the Cayley graph is connected.
Suppose it contains a cycle of length . The successive quotients are basis letters or their inverses, and distinctness of the vertices makes the corresponding word reduced.
That reduced word is nonempty and evaluates to the identity, contradicting uniqueness of normal form; so the graph is a tree.
With respect to a free basis, the word length of an element is the length of its reduced word
Statement
With respect to a free basis, the word length of an element is the length of its reduced word.
Facts & Assumptions
Given: The hypotheses of the Statement.
The word length is the least such that is a product of elements of (Word length of a group element with respect to a generating set).
Word length is defined on every element and satisfies , , and exactly when is the identity (Word length is defined on every element and satisfies the subadditivity, inversion and vanishing laws).
The Cayley graph of a free group with respect to a free basis is a tree (The Cayley graph of a free group with respect to a free basis is a tree).
An elementary cancellation deletes two adjacent letters or . A word is reduced if no elementary cancellation applies. (Words in an alphabet with formal inverses, elementary cancellation, and reduced words).
The reduced words on form a group when the product of reduced words is their concatenation followed by free reduction. (Reduced words form the free group on an alphabet).
Every class in contains exactly one reduced word. (Every class in contains exactly one reduced word).
The subset is a free basis of if is a free group on the set in the sense of. (A free basis of a group).
Proof
A reduced word of length is an expression of length , so the word length is at most the reduced length.
An expression shorter than the reduced word would free-reduce to a second reduced word for the same element, contradicting uniqueness of normal form.
If no product of two members of a generating set is the identity and the Cayley graph is a tree, the set is a free basis
Statement
If no product of two members of a generating set is the identity and the Cayley graph is a tree, the set is a free basis.
Facts & Assumptions
Given: The hypotheses of the Statement.
The Cayley graph of a group with respect to a subset has vertex set and edge set (The Cayley graph of a group with respect to a subset).
A cycle is a closed walk of length at least three with distinct vertices apart from its endpoints; a forest is a simple graph with no cycle and a tree is a connected forest (Cycles, trees and forests in a simple graph on an arbitrary vertex set).
A simple graph is a tree if and only if every two of its vertices are joined by exactly one path (A nonempty simple graph is a tree if and only if each pair of vertices is joined by exactly one path).
The Cayley graph of a free group with respect to a free basis is a tree (The Cayley graph of a free group with respect to a free basis is a tree).
A free group on a set is a group together with a map such that, for every group and every function , there is a unique group homomorphism satisfying (Free group on a set of generators).
The subset is a free basis of if is a free group on the set in the sense of. (A free basis of a group).
The reduced words on form a group when the product of reduced words is their concatenation followed by free reduction. (Reduced words form the free group on an alphabet).
Every class in contains exactly one reduced word. (Every class in contains exactly one reduced word).
An elementary cancellation deletes two adjacent letters or . A word is reduced if no elementary cancellation applies. (Words in an alphabet with formal inverses, elementary cancellation, and reduced words).
Proof
The universal property gives a surjection from the free group on onto restricting to the identity on .
Assume, for contradiction, that the kernel is nontrivial, and take a shortest nonempty reduced word in it; its length is at least two, since the map is injective on .
Length exactly two is excluded by the hypothesis that no product of two members of is the identity.
Length at least three gives distinct partial products, by minimality, and these form a cycle in the Cayley graph, contradicting that it is a tree; so the kernel is trivial.
Geodesics and geodesic metric spaces
Definition
Let be a metric space and let .
A geodesic segment from to is a map for some real such that
Necessarily , by substituting and .
The metric space is geodesic if every two points of are joined by a geodesic segment.
Coarse Lipschitz maps and quasi-isometric embeddings
Definition
Let and be metric spaces and let .
The map is coarse Lipschitz if there are reals and such that
It is a quasi-isometric embedding if there are reals and such that
Thus a quasi-isometric embedding is a coarse Lipschitz map whose distances are also controlled from below, up to the same kind of additive error.
Bounded distance between two maps into a metric space
Definition
Let be a set, let be a metric space, and let .
The maps and are at bounded distance from one another if there is a real such that
In that case one also says that and are at finite distance.
Bounded distance is an equivalence relation and is preserved by pre-composition and by post-composition with a coarse Lipschitz map
Statement
Bounded distance is an equivalence relation and is preserved by pre-composition and by post-composition with a coarse Lipschitz map.
Facts & Assumptions
Given: The hypotheses of the Statement.
Two maps into a metric space are at bounded distance when the distance between their values is bounded uniformly (Bounded distance between two maps into a metric space).
A map is -coarse Lipschitz when , and an -quasi-isometric embedding when in addition (Coarse Lipschitz maps and quasi-isometric embeddings).
Proof
Reflexivity, symmetry and transitivity follow from the metric axioms with the bounds added.
Pre-composition changes no value, so it preserves the bound exactly.
Post-composition with an -coarse Lipschitz map multiplies the bound by and adds ; without the coarse Lipschitz hypothesis the bound need not survive.
Coarsely dense subsets, quasi-inverses and quasi-isometries
Definition
Let and be metric spaces.
A subset is coarsely dense in if there is a real such that for every there is an with . Thus every point of lies within one uniform bound of an actual point of . When is nonempty this is equivalent, after enlarging the bound if necessary, to boundedness of the point-to-set distance from Bounded subset, diameter, distance from a point to a set, and distance between two sets in a metric space. The quantified form also covers the empty space without invoking the undefined expression .
Let and be coarse Lipschitz maps. Then is a quasi-inverse of if both composites and , and and , are at bounded distance in the sense of Bounded distance between two maps into a metric space.
A quasi-isometry is a coarse Lipschitz map that admits a coarse Lipschitz quasi-inverse. Two metric spaces are quasi-isometric if some quasi-isometry between them exists.
Composites of coarse Lipschitz maps and of quasi-isometric embeddings are again such, with explicit constants
Statement
Composites of coarse Lipschitz maps and of quasi-isometric embeddings are again such, with explicit constants.
Facts & Assumptions
Given: The hypotheses of the Statement.
A map is -coarse Lipschitz when , and an -quasi-isometric embedding when in addition (Coarse Lipschitz maps and quasi-isometric embeddings).
Proof
Substituting the inequality for the inner map into the one for the outer map bounds the composite above with constants and .
The same substitution on the lower bounds gives the other inequality, with the constants recorded rather than absorbed.
A map at bounded distance from a quasi-isometric embedding is one, with the additive constant enlarged
Statement
A map at bounded distance from a quasi-isometric embedding is one, with the additive constant enlarged.
Facts & Assumptions
Given: The hypotheses of the Statement.
A map is -coarse Lipschitz when , and an -quasi-isometric embedding when in addition (Coarse Lipschitz maps and quasi-isometric embeddings).
Two maps into a metric space are at bounded distance when the distance between their values is bounded uniformly (Bounded distance between two maps into a metric space).
Bounded distance is an equivalence relation, is preserved by pre-composition, and is preserved by post-composition with a coarse Lipschitz map (Bounded distance is an equivalence relation and is preserved by pre-composition and by post-composition with a coarse Lipschitz map).
Proof
Two applications of the triangle inequality, one at each argument, relate the two maps’ distances up to twice the bound.
Enlarging the additive constant by twice the bound gives both inequalities, with the multiplicative constant unchanged.
A quasi-isometric embedding with coarsely dense image has a quasi-inverse quasi-isometric embedding
Statement
Assume the Axiom of Choice (The Axiom of Choice).
A quasi-isometric embedding with coarsely dense image has a quasi-inverse quasi-isometric embedding.
Facts & Assumptions
Given: The hypotheses of the Statement, including the Axiom of Choice.
A subset is coarsely dense when every point of the space is within a fixed distance of it, and a quasi-isometry is a coarse Lipschitz map admitting a coarse Lipschitz quasi-inverse (Coarsely dense subsets, quasi-inverses and quasi-isometries).
A map is -coarse Lipschitz when , and an -quasi-isometric embedding when in addition (Coarse Lipschitz maps and quasi-isometric embeddings).
Two maps into a metric space are at bounded distance when the distance between their values is bounded uniformly (Bounded distance between two maps into a metric space).
Every family of nonempty sets has a choice function >. (The Axiom of Choice).
Proof
Let be an -quasi-isometric embedding whose image is -coarsely dense. By the definition of coarse density, for every the set is nonempty, so the Axiom of Choice gives a map with for every .
For , the upper inequality for gives so . Likewise is at most , so is a quasi-isometric embedding.
By step 1.1 the composite is at bounded distance at most from . Also so for every ; hence is at bounded distance from . Therefore is a coarse Lipschitz quasi-inverse of .
A map is a quasi-isometry exactly when it is a quasi-isometric embedding with coarsely dense image
Statement
Assume the Axiom of Choice (The Axiom of Choice).
A map is a quasi-isometry exactly when it is a quasi-isometric embedding with coarsely dense image.
Facts & Assumptions
Given: The hypotheses of the Statement, including the Axiom of Choice.
A subset is coarsely dense when every point of the space is within a fixed distance of it, and a quasi-isometry is a coarse Lipschitz map admitting a coarse Lipschitz quasi-inverse (Coarsely dense subsets, quasi-inverses and quasi-isometries).
Under the Axiom of Choice, a quasi-isometric embedding with coarsely dense image admits a quasi-inverse quasi-isometric embedding (A quasi-isometric embedding with coarsely dense image has a quasi-inverse quasi-isometric embedding).
Proof
If a map is a quasi-isometric embedding with coarsely dense image, the previous theorem supplies a quasi-inverse quasi-isometric embedding, so the map is a quasi-isometry.
Conversely, if is a quasi-inverse of and for every , then every target point lies within distance of , so the image of is coarsely dense.
Being quasi-isometric is reflexive, symmetric and transitive
Statement
Being quasi-isometric is reflexive, symmetric and transitive.
Facts & Assumptions
Given: The hypotheses of the Statement.
A subset is coarsely dense when every point of the space is within a fixed distance of it, and a quasi-isometry is a coarse Lipschitz map admitting a coarse Lipschitz quasi-inverse (Coarsely dense subsets, quasi-inverses and quasi-isometries).
Bounded distance is an equivalence relation, is preserved by pre-composition, and is preserved by post-composition with a coarse Lipschitz map (Bounded distance is an equivalence relation and is preserved by pre-composition and by post-composition with a coarse Lipschitz map).
Composites of coarse Lipschitz maps and of quasi-isometric embeddings are again such, with explicit constants (Composites of coarse Lipschitz maps and of quasi-isometric embeddings are again such, with explicit constants).
Proof
For every metric space , the identity map is coarse Lipschitz and is its own quasi-inverse, so is quasi-isometric to itself.
If is a quasi-inverse of a quasi-isometry , then is a quasi-inverse of , so is quasi-isometric to .
If has quasi-inverse and has quasi-inverse , then is coarse Lipschitz and is a coarse Lipschitz quasi-inverse of it: the composites are at bounded distance from the relevant identities by compatibility of bounded distance with composition. Thus quasi-isometry is transitive.
The quasi-isometry group of a metric space
Definition
Let be a metric space. Consider the set of all quasi-isometries . Two such maps are identified when they are at bounded distance, an equivalence relation by Bounded distance is an equivalence relation and is preserved by pre-composition and by post-composition with a coarse Lipschitz map.
The set of these equivalence classes is written
It is called the quasi-isometry group of . The fact that composition of maps descends to these equivalence classes and makes into a group is proved in Quasi-isometries modulo bounded distance form a group, and a quasi-isometry induces an isomorphism of these groups.
Quasi-isometries modulo bounded distance form a group, and a quasi-isometry induces an isomorphism of these groups
Statement
Quasi-isometries modulo bounded distance form a group, and a quasi-isometry induces an isomorphism of these groups.
Facts & Assumptions
Given: The hypotheses of the Statement.
The quasi-isometry group of a metric space is the set of quasi-isometries of it modulo bounded distance (The quasi-isometry group of a metric space).
Bounded distance is an equivalence relation, is preserved by pre-composition, and is preserved by post-composition with a coarse Lipschitz map (Bounded distance is an equivalence relation and is preserved by pre-composition and by post-composition with a coarse Lipschitz map).
A subset is coarsely dense when every point of the space is within a fixed distance of it, and a quasi-isometry is a coarse Lipschitz map admitting a coarse Lipschitz quasi-inverse (Coarsely dense subsets, quasi-inverses and quasi-isometries).
A group is a monoid in which every element is invertible. (Group and abelian group).
Group isomorphisms, automorphisms and the set . (Group isomorphisms, automorphisms and the set ).
A binary relation on is an equivalence relation when it is reflexive on , symmetric and transitive, that is, when it is (Equivalence relation, equivalence class, and the quotient set ).
Proof
Composition is well defined on bounded-distance classes, by the compatibility lemma.
Associativity is inherited from composition of maps, the class of the identity is neutral, and the class of any chosen quasi-inverse is a two-sided inverse, because the composites are at bounded distance from the identities by definition of quasi-isometry.
If has quasi-inverse , then is a homomorphism whose inverse is ; the compatibility lemma shows both are well defined on bounded-distance classes.
Bilipschitz embeddings and bilipschitz equivalences of metric spaces
Definition
Let and be metric spaces and let .
The map is a bilipschitz embedding if there is a real such that
It is a bilipschitz equivalence if it is in addition bijective (Injection, surjection, bijection).
Two metric spaces are bilipschitz equivalent if some bilipschitz equivalence between them exists.
Every isometry is a bilipschitz equivalence and every bilipschitz equivalence is a quasi-isometry, and two metrics on one set are Lipschitz equivalent exactly when the identity is a bilipschitz equivalence between them
Statement
Every isometry is a bilipschitz equivalence and every bilipschitz equivalence is a quasi-isometry, and two metrics on one set are Lipschitz equivalent exactly when the identity is a bilipschitz equivalence between them.
Facts & Assumptions
Given: The hypotheses of the Statement.
A map is a bilipschitz embedding when for some , and a bilipschitz equivalence when it is a bijective such map with bilipschitz inverse (Bilipschitz embeddings and bilipschitz equivalences of metric spaces).
A map is -coarse Lipschitz when , and an -quasi-isometric embedding when in addition (Coarse Lipschitz maps and quasi-isometric embeddings).
A subset is coarsely dense when every point of the space is within a fixed distance of it, and a quasi-isometry is a coarse Lipschitz map admitting a coarse Lipschitz quasi-inverse (Coarsely dense subsets, quasi-inverses and quasi-isometries).
Isometric embedding and isometry. A function is an isometric embedding if (Isometry, isometric embedding, and the subspace metric on a subset).
- and are topologically equivalent if they have the same metric topology: - and are uniformly equivalent if for every real there are reals and such that, for all , - and are Lipschitz equivalent if there are reals with (Topologically, uniformly and Lipschitz equivalent metrics on a set).
If and are Lipschitz equivalent, they are uniformly equivalent. (Lipschitz equivalence implies uniform equivalence implies topological equivalence).
Proof
An isometry is a bilipschitz equivalence with multiplicative constant one.
If is a bilipschitz equivalence, then and are both coarse Lipschitz with additive constant zero, and the composites are the identities; hence is a quasi-isometry.
Two metrics on one set are Lipschitz equivalent exactly when the identity between them is a bilipschitz equivalence; this is a comparison of two definitions written on the same data.
Separated nets in a metric space
Definition
Let be a metric space and let .
For a real , the subset is -separated if
For a real , the subset is an -net in if for every there is an with . For nonempty this is equivalent, after enlarging if needed, to a uniform bound on the point-to-set distance of Bounded subset, diameter, distance from a point to a set, and distance between two sets in a metric space. The empty subset is therefore an -net exactly when is empty.
A separated net in is a subset that is -separated for some and is an -net for some .
Two metric spaces are quasi-isometric if and only if each contains a separated net and the two nets are bilipschitz equivalent
Statement
Assume the Axiom of Choice (The Axiom of Choice).
Two metric spaces are quasi-isometric if and only if each contains a separated net and the two nets are bilipschitz equivalent.
Facts & Assumptions
Given: The hypotheses of the Statement, including the Axiom of Choice.
A subset of a metric space is a separated net when its points are uniformly separated and it is coarsely dense (Separated nets in a metric space).
A map is -coarse Lipschitz when , and an -quasi-isometric embedding when in addition (Coarse Lipschitz maps and quasi-isometric embeddings).
Under the Axiom of Choice, a map is a quasi-isometry if and only if it is a quasi-isometric embedding with coarsely dense image (A map is a quasi-isometry exactly when it is a quasi-isometric embedding with coarsely dense image).
A map is a bilipschitz embedding when for some , and a bilipschitz equivalence when it is a bijective such map with bilipschitz inverse (Bilipschitz embeddings and bilipschitz equivalences of metric spaces).
Every family of nonempty sets has a choice function >. (The Axiom of Choice).
Under the Axiom of Choice, every nonempty poset in which every chain has an upper bound has a maximal element (Zorn's lemma).
Being quasi-isometric is a reflexive, symmetric and transitive relation on metric spaces (Being quasi-isometric is reflexive, symmetric and transitive).
A subset inherits the ambient metric, and its inclusion is an isometric embedding (Isometry, isometric embedding, and the subspace metric on a subset).
Proof
Let and be quasi-isometric, and by [L2] choose a quasi-isometric embedding with coarsely dense image. Fix constants , and such that is -quasi-isometric and is -coarsely dense. Put . Consider the poset of -separated subsets of , ordered by inclusion. It is nonempty because is -separated, and the union of a chain of -separated subsets is again -separated; so Zorn's lemma gives a maximal -separated subset . If some satisfied , then would still be -separated, contradicting maximality. Hence every point of lies within distance of , so is a separated net in .
For distinct one has and so the restriction is bilipschitz. Also, if , choose with and then choose with ; then So is a separated net in , bilipschitz equivalent to .
Conversely, let and be separated nets, and let be a bilipschitz equivalence. By [L7] the inclusions and are isometric embeddings, and because and are nets those inclusions have coarsely dense image; hence [L2] makes them quasi-isometries. By [L5] the map is a quasi-isometry. Transitivity of quasi-isometry now gives , so and are quasi-isometric.
The nonempty metric spaces quasi-isometric to a one-point space are exactly those of finite diameter
Statement
The nonempty metric spaces quasi-isometric to a one-point space are exactly those of finite diameter.
Facts & Assumptions
Given: The hypotheses of the Statement.
A subset is coarsely dense when every point of the space is within a fixed distance of it, and a quasi-isometry is a coarse Lipschitz map admitting a coarse Lipschitz quasi-inverse (Coarsely dense subsets, quasi-inverses and quasi-isometries).
A map is -coarse Lipschitz when , and an -quasi-isometric embedding when in addition (Coarse Lipschitz maps and quasi-isometric embeddings).
Bounded subset. is bounded if or there are and a real with . (Bounded subset, diameter, distance from a point to a set, and distance between two sets in a metric space).
Proof
Let be nonempty and bounded, so for some and . The constant map and the map with are coarse Lipschitz; one has , and for every the distance between and is less than . Hence is a quasi-isometry.
Conversely, if is a quasi-isometry and is a quasi-inverse, then for some every satisfies . Thus and is bounded, hence has finite diameter.
The identity map between the word metrics of two finite generating sets is a bilipschitz equivalence
Statement
The identity map between the word metrics of two finite generating sets is a bilipschitz equivalence.
Facts & Assumptions
Given: The hypotheses of the Statement.
The word metric of with respect to is (The word metric of a group with respect to a generating set).
The word length is the least such that is a product of elements of (Word length of a group element with respect to a generating set).
Word length is defined on every element and satisfies , , and exactly when is the identity (Word length is defined on every element and satisfies the subadditivity, inversion and vanishing laws).
The word metric is a left-invariant metric and coincides with the path metric of the Cayley graph (The word metric is a left-invariant metric and coincides with the path metric of the Cayley graph).
A map is a bilipschitz embedding when for some , and a bilipschitz equivalence when it is a bijective such map with bilipschitz inverse (Bilipschitz embeddings and bilipschitz equivalences of metric spaces).
- and are topologically equivalent if they have the same metric topology: - and are uniformly equivalent if for every real there are reals and such that, for all , - and are Lipschitz equivalent if there are reals with (Topologically, uniformly and Lipschitz equivalent metrics on a set).
A group is finitely generated when some finite subset generates it (Finitely generated groups).
A set is finite when for some . (The cardinality of a finite set).
Proof
Let be the largest word length in the second metric of a member of the first symmetrised set; finiteness of that set is exactly what makes the maximum exist.
Expanding an element of length in the first metric and applying the triangle inequality along the expression bounds its second length by .
Exchanging the roles of the two sets gives the reverse inequality, so the identity is a bilipschitz equivalence and the two metrics are Lipschitz equivalent.
The quasi-isometry type of a finitely generated group
Definition
Let be a finitely generated group. Choose a finite generating set and form the word metric space of The word metric of a group with respect to a generating set.
By The identity map between the word metrics of two finite generating sets is a bilipschitz equivalence, the word metrics are bilipschitz equivalent; by Every isometry is a bilipschitz equivalence and every bilipschitz equivalence is a quasi-isometry, and two metrics on one set are Lipschitz equivalent exactly when the identity is a bilipschitz equivalence between them they are therefore quasi-isometric. Hence the quasi-isometry class of is independent of the chosen finite generating set .
This well-defined quasi-isometry class is the quasi-isometry type of .
A finitely generated group is finite if and only if it is quasi-isometric to a point
Statement
A finitely generated group is finite if and only if it is quasi-isometric to a point.
Facts & Assumptions
Given: The hypotheses of the Statement.
A finitely generated group is quasi-isometric to a metric space when its word metric for some, equivalently every, finite generating set is (The quasi-isometry type of a finitely generated group).
Balls of a word metric are finite if and only if the generating set is finite (Balls of a word metric are finite if and only if the generating set is finite).
The nonempty metric spaces quasi-isometric to a one-point space are exactly those of finite diameter (The nonempty metric spaces quasi-isometric to a one-point space are exactly those of finite diameter).
Bounded subset. is bounded if or there are and a real with . (Bounded subset, diameter, distance from a point to a set, and distance between two sets in a metric space).
A set is finite when for some . (The cardinality of a finite set).
Proof
A finite group has finite diameter in any word metric, hence is quasi-isometric to a point.
Conversely finite diameter with a finite generating set makes the whole group a ball of finite radius, and such balls are finite.
A subgroup of finite index in a finitely generated group is finitely generated, and its inclusion is a quasi-isometry
Statement
A subgroup of finite index in a finitely generated group is finitely generated, and its inclusion is a quasi-isometry.
Facts & Assumptions
Given: The hypotheses of the Statement.
A group is finitely generated when some finite subset generates it (Finitely generated groups).
The word length is the least such that is a product of elements of (Word length of a group element with respect to a generating set).
Word length is defined on every element and satisfies , , and exactly when is the identity (Word length is defined on every element and satisfies the subadditivity, inversion and vanishing laws).
The word metric of with respect to is (The word metric of a group with respect to a generating set).
A map is -coarse Lipschitz when , and an -quasi-isometric embedding when in addition (Coarse Lipschitz maps and quasi-isometric embeddings).
A subset is coarsely dense when every point of the space is within a fixed distance of it, and a quasi-isometry is a coarse Lipschitz map admitting a coarse Lipschitz quasi-inverse (Coarsely dense subsets, quasi-inverses and quasi-isometries).
For a subgroup , the set of distinct left cosets is a partition of : every element belongs to a left coset, every coset is nonempty, and two left cosets are either equal or disjoint. (The left cosets of a subgroup partition the group).
A set is finite when for some . (The cardinality of a finite set).
Proof
Fix a finite left transversal for in that contains the identity; the cosets partition the group and there are finitely many of them.
Rewriting a product of ambient generators by pushing transversal representatives to the right expresses each subgroup element in the finite set of products that fall back into the subgroup, so that set generates it.
Write each uniquely as with and . If with and , , then is one of the finitely many subgroup elements used in step 2.1; so in the word metric of for the generating set built there. Chaining along a shortest -expression gives , so is Lipschitz. Since is the identity on and every lies within of the inclusion of , the composites and are the identity on and at bounded distance from the identity on respectively. Therefore the inclusion is a quasi-isometry with quasi-inverse .
The quotient map by a finite normal subgroup is a quasi-isometry of word metric spaces
Statement
The quotient map by a finite normal subgroup is a quasi-isometry of word metric spaces.
Facts & Assumptions
Given: The hypotheses of the Statement.
A group is finitely generated when some finite subset generates it (Finitely generated groups).
The word length is the least such that is a product of elements of (Word length of a group element with respect to a generating set).
Word length is defined on every element and satisfies , , and exactly when is the identity (Word length is defined on every element and satisfies the subadditivity, inversion and vanishing laws).
The word metric of with respect to is (The word metric of a group with respect to a generating set).
A map is -coarse Lipschitz when , and an -quasi-isometric embedding when in addition (Coarse Lipschitz maps and quasi-isometric embeddings).
A subset is coarsely dense when every point of the space is within a fixed distance of it, and a quasi-isometry is a coarse Lipschitz map admitting a coarse Lipschitz quasi-inverse (Coarsely dense subsets, quasi-inverses and quasi-isometries).
The quotient group, or factor group, has the left cosets (The quotient group and coset product ).
The subgroup is normal in when (Normal subgroup: invariance under conjugation).
A set is finite when for some . (The cardinality of a finite set).
Proof
The image of a finite generating set generates the quotient, so the quotient map does not increase word length and is one-Lipschitz.
Let , which exists because is finite. For each coset , choose a representative of minimal word length in that coset, breaking ties lexicographically in the fixed finite alphabet ; then the quotient map satisfies . If with , then , so Chaining along a shortest quotient expression gives , so is Lipschitz.
For every , the elements and lie in the same coset, so and therefore . Thus and is at bounded distance from . Therefore is a quasi-isometry with quasi-inverse .
Quasi-geodesics and quasi-geodesic metric spaces
Definition
Let be a metric space.
A map from an interval is a -quasi-geodesic if , , and
A -quasi-geodesic segment from to is such a map with domain a closed bounded interval and with , .
The metric space is quasi-geodesic if there are constants and such that every two points of are the endpoints of some -quasi-geodesic segment.
The vertex set of a connected simple graph with its path metric is a -quasi-geodesic space
Statement
The vertex set of a connected simple graph with its path metric is a -quasi-geodesic space.
Facts & Assumptions
Given: The hypotheses of the Statement.
A -quasi-geodesic is a -quasi-isometric embedding of a closed real interval, and a space is -quasi-geodesic when every two of its points are joined by one (Quasi-geodesics and quasi-geodesic metric spaces).
The path metric of a connected simple graph assigns to two vertices the least length of a path joining them (The path metric of a connected simple graph).
The path metric of a connected simple graph is a metric on its vertex set (The path metric of a connected simple graph is a metric on its vertex set).
A geodesic of length in a metric space is an isometric embedding of the interval , and the space is geodesic when every two points are the endpoints of one (Geodesics and geodesic metric spaces).
A map is -coarse Lipschitz when , and an -quasi-isometric embedding when in addition (Coarse Lipschitz maps and quasi-isometric embeddings).
It is written and called the integer part, or floor, of . (Integer part: for every real there is exactly one integer with ).
Proof
Given two vertices, parametrise a path realising their distance by sending each integer point of to the corresponding vertex and each intermediate real to the nearer endpoint of its unit subinterval.
The distances so obtained differ from those of the interval by at most one, so the parametrisation is a -quasi-isometric embedding.
Hence every two vertices are joined by a -quasi-geodesic, so the space is -quasi-geodesic and not geodesic unless it is a single point.
A group with the word metric of any generating set is a -quasi-geodesic space
Statement
A group with the word metric of any generating set is a -quasi-geodesic space.
Facts & Assumptions
Given: The hypotheses of the Statement.
The word metric is a left-invariant metric and coincides with the path metric of the Cayley graph (The word metric is a left-invariant metric and coincides with the path metric of the Cayley graph).
The vertex set of a connected simple graph with its path metric is a -quasi-geodesic space (The vertex set of a connected simple graph with its path metric is a -quasi-geodesic space).
The word metric of with respect to is (The word metric of a group with respect to a generating set).
A group is finitely generated when some finite subset generates it (Finitely generated groups).
Proof
The word metric is the path metric of the Cayley graph, which is connected because the set generates.
The previous proposition then applies verbatim.
The composite of a quasi-geodesic with a quasi-isometric embedding is a quasi-geodesic, with computed constants
Statement
The composite of a quasi-geodesic with a quasi-isometric embedding is a quasi-geodesic, with computed constants.
Facts & Assumptions
Given: The hypotheses of the Statement.
A -quasi-geodesic is a -quasi-isometric embedding of a closed real interval, and a space is -quasi-geodesic when every two of its points are joined by one (Quasi-geodesics and quasi-geodesic metric spaces).
A map is -coarse Lipschitz when , and an -quasi-isometric embedding when in addition (Coarse Lipschitz maps and quasi-isometric embeddings).
Composites of coarse Lipschitz maps and of quasi-isometric embeddings are again such, with explicit constants (Composites of coarse Lipschitz maps and of quasi-isometric embeddings are again such, with explicit constants).
Proof
A quasi-geodesic is by definition a quasi-isometric embedding of a closed interval.
The composition lemma applied to two quasi-isometric embeddings gives the conclusion, with the constants that lemma records.
A coarse Lipschitz map between word metric spaces of finitely generated groups is Lipschitz
Statement
A coarse Lipschitz map between word metric spaces of finitely generated groups is Lipschitz.
Facts & Assumptions
Given: The hypotheses of the Statement.
A map is -coarse Lipschitz when , and an -quasi-isometric embedding when in addition (Coarse Lipschitz maps and quasi-isometric embeddings).
The word length is the least such that is a product of elements of (Word length of a group element with respect to a generating set).
The word metric of with respect to is (The word metric of a group with respect to a generating set).
The word metric is a left-invariant metric and coincides with the path metric of the Cayley graph (The word metric is a left-invariant metric and coincides with the path metric of the Cayley graph).
A group is finitely generated when some finite subset generates it (Finitely generated groups).
- is Lipschitz with constant , where and , if is Lipschitz if it is Lipschitz with some such constant. (Lipschitz map, -Hölder map for rational , and contraction).
Proof
Two points at distance one differ by a single generator, so their images are at distance at most .
Chaining along a shortest expression bounds the image distance by times the source distance.
So the map is Lipschitz with constant ; the argument uses that the word metric takes integer values and that adjacent points are at distance one, so it does not extend to an arbitrary metric source.
A bijective quasi-isometry between word metric spaces of finitely generated groups is a bilipschitz equivalence
Statement
A bijective quasi-isometry between word metric spaces of finitely generated groups is a bilipschitz equivalence.
Facts & Assumptions
Given: The hypotheses of the Statement.
A subset is coarsely dense when every point of the space is within a fixed distance of it, and a quasi-isometry is a coarse Lipschitz map admitting a coarse Lipschitz quasi-inverse (Coarsely dense subsets, quasi-inverses and quasi-isometries).
A coarse Lipschitz map between word metric spaces of finitely generated groups is Lipschitz (A coarse Lipschitz map between word metric spaces of finitely generated groups is Lipschitz).
A map is -coarse Lipschitz when , and an -quasi-isometric embedding when in addition (Coarse Lipschitz maps and quasi-isometric embeddings).
A map is a bilipschitz embedding when for some , and a bilipschitz equivalence when it is a bijective such map with bilipschitz inverse (Bilipschitz embeddings and bilipschitz equivalences of metric spaces).
- is Lipschitz with constant , where and , if is Lipschitz if it is Lipschitz with some such constant. (Lipschitz map, -Hölder map for rational , and contraction).
Proof
Let be a bijective quasi-isometry, and let be a coarse Lipschitz quasi-inverse. If for every and is -coarse Lipschitz, then for and one has So the set-theoretic inverse is coarse Lipschitz.
Both and are therefore coarse Lipschitz, hence Lipschitz by the previous proposition.
If has Lipschitz constant and has Lipschitz constant , then for all , so is a bilipschitz equivalence.
Quasi-isometry invariants and geometric properties of finitely generated groups
Definition
Let be a property of finitely generated groups.
The property is a quasi-isometry invariant if, whenever and are finitely generated and have the same quasi-isometry type, the truth of for implies its truth for .
Equivalently, depends only on the quasi-isometry type of the group. In geometric group theory one also says that such a property is geometric.
Finiteness is a geometric property of finitely generated groups
Statement
Finiteness is a geometric property of finitely generated groups.
Facts & Assumptions
Given: The hypotheses of the Statement.
A quasi-isometry invariant is a map on finitely generated groups constant on quasi-isometry classes, and a property is geometric when its indicator is such an invariant (Quasi-isometry invariants and geometric properties of finitely generated groups).
A finitely generated group is finite if and only if it is quasi-isometric to a one-point space (A finitely generated group is finite if and only if it is quasi-isometric to a point).
A finitely generated group is quasi-isometric to a metric space when its word metric for some, equivalently every, finite generating set is (The quasi-isometry type of a finitely generated group).
Being quasi-isometric is a reflexive, symmetric and transitive relation on metric spaces (Being quasi-isometric is reflexive, symmetric and transitive).
Proof
Two quasi-isometric finitely generated groups are simultaneously quasi-isometric to a point, by transitivity of the relation.
Being quasi-isometric to a point characterises finiteness, so the indicator of finiteness is a quasi-isometry invariant and finiteness is a geometric property.
5 · Examples, counterexamples and false statements
None yet.
Sources
- C. Loh, Geometric Group Theory: An Introduction (2015 course version), Section 3.1
- C. Loh, Geometric Group Theory: An Introduction (2015 course version), 264 pp.
- C. Drutu and M. Kapovich, Geometric Group Theory (with an appendix by B. Nica), 837 pp.
- C. Loh, Geometric Group Theory: An Introduction (2015 course version), Section 5.2
- C. Loh, Geometric Group Theory: An Introduction (2015 course version), Sections 3.1-3.2
- C. Drutu and M. Kapovich, Geometric Group Theory, Section 7.9
- C. Loh, Geometric Group Theory: An Introduction (2015 course version), Section 3.2
- C. Loh, Geometric Group Theory: An Introduction (2015 course version), Section 5.3
- C. Drutu and M. Kapovich, Geometric Group Theory, Section 8.1
- C. Loh, Geometric Group Theory: An Introduction (2015 course version), Section 5.1
- C. Loh, Geometric Group Theory: An Introduction (2015 course version), Section 5.5