Support-Preserving Endomorphisms
- Support-preserving endomorphisms are *-endomorphisms of graph C*-algebras that preserve the diagonal MASA via permutation unitaries and combinatorial structures.
- They implement the Cuntz-Krieger relations by leveraging localized unitaries and finite-dimensional subalgebras, with invertibility characterized by nilpotency in associated finite rings.
- These endomorphisms connect the combinatorial aspects of graph theory with Weyl group structures, offering insights into inner versus outer automorphisms and shift-commutation dynamics.
A support-preserving endomorphism of a graph -algebra is a -endomorphism that globally preserves the diagonal maximal abelian subalgebra (MASA) , often implemented by unitaries subject to compatibility conditions with the graph structure. These endomorphisms, also called "permutative" when realized by permutation unitaries, have deep connections to the structure theory of -algebras, Weyl groups, and the interplay of combinatorial and operator-algebraic methods (Conti et al., 2011).
1. The Structure of Graph -Algebras and the Diagonal MASA
Given a countable directed graph , the graph -algebra is generated by
- Mutually orthogonal projections .
- Partial isometries 0 subject to the Cuntz-Krieger relations:
- (GA1) 1, 2 for 3.
- (GA2) 4.
- (GA3) 5 when 6.
The diagonal MASA 7 is the abelian 8-subalgebra generated by projections 9 for finite paths 0, 1. Under standard hypotheses (no sinks, every loop has an exit), 2 is maximal abelian. The algebra admits a canonical gauge action with fixed-point subalgebra 3 (the core AF-algebra).
2. Endomorphisms from Unitaries and the Support-Preserving Condition
Let
4
Every 5 induces a 6-endomorphism 7 defined by
8
Since 9 commutes with 0 for all 1, these 2 satisfy the Cuntz-Krieger relations. If 3 lies in the minimal unitization of the algebraic part of the core 4, 5 is injective. The semigroup law is 6. Invertibility of 7 is equivalent to 8 belonging to the image of the strict extension of 9.
Localized endomorphism: If 0 and belongs to the finite-dimensional span of 1, then 2 is localized at level 3.
Support-preserving or permutative endomorphisms are those 4 such that 5; typically these arise when 6 is a permutation unitary acting combinatorially at some level.
3. Weyl Groups and Their Combinatorial Structure
The Weyl group 7 of a graph 8-algebra 9 is defined as
0
where 1 are automorphisms preserving 2 globally, and 3 those acting as the identity on 4. 5 is countable and discrete for graphs with no sinks and all loops having exits.
Permutative automorphisms arising from permutation unitaries 6 and automorphisms of the underlying graph 7 generate a significant combinatorial subgroup of 8.
The restricted Weyl group 9 further requires automorphisms to preserve the core 0: 1 Structural results include:
- Every automorphism in 2 arises as 3 for a unique 4 lying in 5.
- 6 as topological groups.
- 7 is generated (semidirectly) by permutation unitaries in 8 and graph-automorphisms.
- If the relative commutant of 9 in 0 is trivial, the only inner automorphisms in the restricted group come from the finite subgroup 1.
4. Invertibility Criteria and Combinatorial Analysis
For 2 lying in the finite-dimensional algebra 3, the following are equivalent ((Conti et al., 2011), Thm 5.1):
- 4 is an automorphism of 5 and 6 is localized.
- The sequence 7 stabilizes as 8.
- The finite ring 9 generated by 0 for 1 is nilpotent.
- The limit space 2 obtained by inductive intersections lies in 3.
For the restriction 4, a parallel criterion applies: 5 is an automorphism iff the finite-dimensional subring 6 generated by compressions 7 to 8 is nilpotent—corresponding to a descending sequence of subspaces in 9 falling eventually into 0.
5. Diagrams and the Combinatorial Approach to Permutative Endomorphisms
Permutation unitaries at exactly level 1 correspond combinatorially to disjoint families of permutations 2 indexed by vertices 3. The action of 4 on 5 and the finite rings 6 can be encoded in finite labeled graphs.
Two combinatorial conditions arise:
- Condition (b): 7 iff the associated directed graph on pairs 8 for paths 9 has the property that no two distinct fixed points exist under sufficiently long colored paths. Equivalently, a partial order forces upward flow of each "off-diagonal" vertex.
- Condition (d): 00 iff in a digraph on pairs 01 with edges labeled by 02 when 03 and 04 agree on their terminal letter, all cycles except for trivial loops are excluded.
The composite theorem ((Conti et al., 2011), Thm 6.4): Let 05 be a permutation unitary at level 06. Then 07 is an automorphism of 08 if and only if both (b) and (d) hold. Notably, (b) 09 invertible; combining (d) yields invertibility of 10.
6. Outer Automorphism Criteria and Shift-Commutation
An automorphism in 11 is inner if and only if the implementing unitary lies in the finite permutation-unitary subgroup 12. Two mechanisms are used to distinguish outerness:
- Gauge-action rigidity: If the only unitaries normalizing 13 are those in 14, then every support-preserving automorphism 15 preserving 16 must come from 17.
- Shift-commutation: Given 18, the induced homeomorphism 19 of the spectrum 20 satisfies an eventual commutation with the one-sided shift 21; specifically, for some 22,
23
where 24.
Corollary: If 25 in the combinatorial subgroup of 26 has infinite order, it represents an infinite-order element in 27.
7. Illustrative Examples
Two canonical examples illustrate the criteria for support-preserving endomorphisms:
- Fibonacci-graph: For a graph with two vertices and three edges in a two-cycle plus tail configuration, at level 28 there exists a permutation unitary 29 for which Condition (b) holds but (d) fails. Thus, 30 is an automorphism, but 31 is a proper (non-surjective) support-preserving endomorphism.
- Simple Kirchberg-algebra: For a strongly connected 4-vertex graph with 32, a non-graph-automorphism permutation unitary 33 at level 34 gives 35 of order 36 in 37, with both (b) and (d) satisfied, hence 38.
These cases explicitly employ the combinatorial digraph constructions to verify invertibility and restriction to the diagonal.
For full proofs, detailed constructions, and comprehensive combinatorial models, see Sections 3–6 of (Conti et al., 2011).