reconstructing étale groupoids from their algebras · • groupoid reconstruction asks the...

165
History and the case of groups Previous results The methods Reconstructing ´ etale groupoids from their algebras Benjamin Steinberg (City College of New York) December 5, 2017 Facets of Irreversibility

Upload: others

Post on 22-Oct-2020

5 views

Category:

Documents


0 download

TRANSCRIPT

  • History and the case of groups Previous results The methods

    Reconstructing étale groupoids from their

    algebras

    Benjamin Steinberg (City College of New York)

    December 5, 2017Facets of Irreversibility

  • History and the case of groups Previous results The methods

    Outline

    History and the case of groups

    Previous results

    The methods

  • History and the case of groups Previous results The methods

    Overview

    • Groupoid reconstruction asks the question: when is anétale groupoid determined by the pair consisting of itsalgebra and its diagonal subalgebra.

  • History and the case of groups Previous results The methods

    Overview

    • Groupoid reconstruction asks the question: when is anétale groupoid determined by the pair consisting of itsalgebra and its diagonal subalgebra.

    • Renault showed that a diagonal-preserving isomorphism ofreduced C∗-algebras induces an isomorphism of groupoidsfor topologically principal étale groupoids.

  • History and the case of groups Previous results The methods

    Overview

    • Groupoid reconstruction asks the question: when is anétale groupoid determined by the pair consisting of itsalgebra and its diagonal subalgebra.

    • Renault showed that a diagonal-preserving isomorphism ofreduced C∗-algebras induces an isomorphism of groupoidsfor topologically principal étale groupoids.

    • A number of authors have studied the analogous questionfor ample groupoid algebras over rings.

  • History and the case of groups Previous results The methods

    Overview

    • Groupoid reconstruction asks the question: when is anétale groupoid determined by the pair consisting of itsalgebra and its diagonal subalgebra.

    • Renault showed that a diagonal-preserving isomorphism ofreduced C∗-algebras induces an isomorphism of groupoidsfor topologically principal étale groupoids.

    • A number of authors have studied the analogous questionfor ample groupoid algebras over rings.

    • We present here what we believe is the best result onecan get from the present methodology.

  • History and the case of groups Previous results The methods

    Setup

    • R is a unital commutative ring.

  • History and the case of groups Previous results The methods

    Setup

    • R is a unital commutative ring.

    • G is a Hausdorff ample groupoid.

  • History and the case of groups Previous results The methods

    Setup

    • R is a unital commutative ring.

    • G is a Hausdorff ample groupoid.

    • As an R-module RG = Cc(G(1), R).

  • History and the case of groups Previous results The methods

    Setup

    • R is a unital commutative ring.

    • G is a Hausdorff ample groupoid.

    • As an R-module RG = Cc(G(1), R).

    • The product is convolution:

    fg(γ) =∑

    d(γ)=d(α)

    f(γα−1)g(α).

  • History and the case of groups Previous results The methods

    Setup

    • R is a unital commutative ring.

    • G is a Hausdorff ample groupoid.

    • As an R-module RG = Cc(G(1), R).

    • The product is convolution:

    fg(γ) =∑

    d(γ)=d(α)

    f(γα−1)g(α).

    • Cc(G(0), R) sits inside of RG with convolution restricting

    to pointwise multiplication.

  • History and the case of groups Previous results The methods

    Setup

    • R is a unital commutative ring.

    • G is a Hausdorff ample groupoid.

    • As an R-module RG = Cc(G(1), R).

    • The product is convolution:

    fg(γ) =∑

    d(γ)=d(α)

    f(γα−1)g(α).

    • Cc(G(0), R) sits inside of RG with convolution restricting

    to pointwise multiplication.

    • We call it the diagonal subalgebra DR(G ).

  • History and the case of groups Previous results The methods

    Setup

    • R is a unital commutative ring.

    • G is a Hausdorff ample groupoid.

    • As an R-module RG = Cc(G(1), R).

    • The product is convolution:

    fg(γ) =∑

    d(γ)=d(α)

    f(γα−1)g(α).

    • Cc(G(0), R) sits inside of RG with convolution restricting

    to pointwise multiplication.

    • We call it the diagonal subalgebra DR(G ).

    • DR(G ) is commutative.

  • History and the case of groups Previous results The methods

    Diagonal-preserving isomorphisms

    • An isomorphism Φ: RG → RG ′ is diagonal-preserving ifΦ(DR(G )) = DR(G

    ′).

  • History and the case of groups Previous results The methods

    Diagonal-preserving isomorphisms

    • An isomorphism Φ: RG → RG ′ is diagonal-preserving ifΦ(DR(G )) = DR(G

    ′).

    • An isomorphism φ : G → G ′ induces a diagonal-preservingisomorphism.

  • History and the case of groups Previous results The methods

    Diagonal-preserving isomorphisms

    • An isomorphism Φ: RG → RG ′ is diagonal-preserving ifΦ(DR(G )) = DR(G

    ′).

    • An isomorphism φ : G → G ′ induces a diagonal-preservingisomorphism.

    • We are interested in when the converse holds.

  • History and the case of groups Previous results The methods

    Diagonal-preserving isomorphisms

    • An isomorphism Φ: RG → RG ′ is diagonal-preserving ifΦ(DR(G )) = DR(G

    ′).

    • An isomorphism φ : G → G ′ induces a diagonal-preservingisomorphism.

    • We are interested in when the converse holds.

    • So we want to know: when does the pair (RG , DR(G ))determine G ?

  • History and the case of groups Previous results The methods

    Diagonal-preserving isomorphisms

    • An isomorphism Φ: RG → RG ′ is diagonal-preserving ifΦ(DR(G )) = DR(G

    ′).

    • An isomorphism φ : G → G ′ induces a diagonal-preservingisomorphism.

    • We are interested in when the converse holds.

    • So we want to know: when does the pair (RG , DR(G ))determine G ?

    • We work in the category of rings: we just ask Φ to be aring isomorphism.

  • History and the case of groups Previous results The methods

    Diagonal-preserving isomorphisms

    • An isomorphism Φ: RG → RG ′ is diagonal-preserving ifΦ(DR(G )) = DR(G

    ′).

    • An isomorphism φ : G → G ′ induces a diagonal-preservingisomorphism.

    • We are interested in when the converse holds.

    • So we want to know: when does the pair (RG , DR(G ))determine G ?

    • We work in the category of rings: we just ask Φ to be aring isomorphism.

    • This is not much of a loss of generality because DR(G )knows a lot about R.

  • History and the case of groups Previous results The methods

    Diagonal-preserving isomorphisms

    • An isomorphism Φ: RG → RG ′ is diagonal-preserving ifΦ(DR(G )) = DR(G

    ′).

    • An isomorphism φ : G → G ′ induces a diagonal-preservingisomorphism.

    • We are interested in when the converse holds.

    • So we want to know: when does the pair (RG , DR(G ))determine G ?

    • We work in the category of rings: we just ask Φ to be aring isomorphism.

    • This is not much of a loss of generality because DR(G )knows a lot about R.

    • We do not work with the ∗-ring structure.

  • History and the case of groups Previous results The methods

    How about groups?

    • Groups are just one-object ample groupoids.

  • History and the case of groups Previous results The methods

    How about groups?

    • Groups are just one-object ample groupoids.

    • RG, for G a group, is the usual group algebra.

  • History and the case of groups Previous results The methods

    How about groups?

    • Groups are just one-object ample groupoids.

    • RG, for G a group, is the usual group algebra.

    • R is the diagonal subalgebra.

  • History and the case of groups Previous results The methods

    How about groups?

    • Groups are just one-object ample groupoids.

    • RG, for G a group, is the usual group algebra.

    • R is the diagonal subalgebra.

    • So we are asking: when does RG ∼= RH as R-algebrasimply G ∼= H?

  • History and the case of groups Previous results The methods

    How about groups?

    • Groups are just one-object ample groupoids.

    • RG, for G a group, is the usual group algebra.

    • R is the diagonal subalgebra.

    • So we are asking: when does RG ∼= RH as R-algebrasimply G ∼= H?

    • This is the classical isomorphism problem for group rings(goes back to 1940s).

  • History and the case of groups Previous results The methods

    Negative results

    • Any two finite abelian groups of the same order haveisomorphic complex algebras by Fourier Analysis.

  • History and the case of groups Previous results The methods

    Negative results

    • Any two finite abelian groups of the same order haveisomorphic complex algebras by Fourier Analysis.

    • So a group cannot be recovered from its group algebra ingeneral.

  • History and the case of groups Previous results The methods

    Negative results

    • Any two finite abelian groups of the same order haveisomorphic complex algebras by Fourier Analysis.

    • So a group cannot be recovered from its group algebra ingeneral.

    • Clearly C is too big.

  • History and the case of groups Previous results The methods

    Negative results

    • Any two finite abelian groups of the same order haveisomorphic complex algebras by Fourier Analysis.

    • So a group cannot be recovered from its group algebra ingeneral.

    • Clearly C is too big.

    • It is more hopeful ZG1 ∼= ZG2 =⇒ G1 ∼= G2.

  • History and the case of groups Previous results The methods

    Negative results

    • Any two finite abelian groups of the same order haveisomorphic complex algebras by Fourier Analysis.

    • So a group cannot be recovered from its group algebra ingeneral.

    • Clearly C is too big.

    • It is more hopeful ZG1 ∼= ZG2 =⇒ G1 ∼= G2.

    • Hertweck found two non-isomorphic groups of order221 · 9728 with isomorphic integral group rings (Annals2001).

  • History and the case of groups Previous results The methods

    Negative results

    • Any two finite abelian groups of the same order haveisomorphic complex algebras by Fourier Analysis.

    • So a group cannot be recovered from its group algebra ingeneral.

    • Clearly C is too big.

    • It is more hopeful ZG1 ∼= ZG2 =⇒ G1 ∼= G2.

    • Hertweck found two non-isomorphic groups of order221 · 9728 with isomorphic integral group rings (Annals2001).

    • These groups cannot be recovered from their group ringsover any base ring.

  • History and the case of groups Previous results The methods

    Positive results

    • There seems to be only one general method to show thata group algebra determines the group.

  • History and the case of groups Previous results The methods

    Positive results

    • There seems to be only one general method to show thata group algebra determines the group.

    • If r ∈ R× and g ∈ G, then rg ∈ (RG)×.

  • History and the case of groups Previous results The methods

    Positive results

    • There seems to be only one general method to show thata group algebra determines the group.

    • If r ∈ R× and g ∈ G, then rg ∈ (RG)×.

    • These are call trivial units.

  • History and the case of groups Previous results The methods

    Positive results

    • There seems to be only one general method to show thata group algebra determines the group.

    • If r ∈ R× and g ∈ G, then rg ∈ (RG)×.

    • These are call trivial units.

    • RG has no non-trivial units if every unit is trivial.

  • History and the case of groups Previous results The methods

    Positive results

    • There seems to be only one general method to show thata group algebra determines the group.

    • If r ∈ R× and g ∈ G, then rg ∈ (RG)×.

    • These are call trivial units.

    • RG has no non-trivial units if every unit is trivial.

    • That is ψ in the diagram

    G

    ψ

    (RG)×

    (RG)×/R×

    is an isomorphism.

  • History and the case of groups Previous results The methods

    No non-trivial units

    • If RG and RH have no non-trivial units and RG ∼= RHas R-algebras, then G ∼= (RG)×/R× ∼= (RH)×/R× ∼= H .

  • History and the case of groups Previous results The methods

    No non-trivial units

    • If RG and RH have no non-trivial units and RG ∼= RHas R-algebras, then G ∼= (RG)×/R× ∼= (RH)×/R× ∼= H .

    • In fact, if RG has no non-trivial units and RG ∼= RH asR-algebras, then RH has no non-trivial units.

  • History and the case of groups Previous results The methods

    No non-trivial units

    • If RG and RH have no non-trivial units and RG ∼= RHas R-algebras, then G ∼= (RG)×/R× ∼= (RH)×/R× ∼= H .

    • In fact, if RG has no non-trivial units and RG ∼= RH asR-algebras, then RH has no non-trivial units.

    • The proof uses that a group is a basis for its group ring.

  • History and the case of groups Previous results The methods

    No non-trivial units

    • If RG and RH have no non-trivial units and RG ∼= RHas R-algebras, then G ∼= (RG)×/R× ∼= (RH)×/R× ∼= H .

    • In fact, if RG has no non-trivial units and RG ∼= RH asR-algebras, then RH has no non-trivial units.

    • The proof uses that a group is a basis for its group ring.

    • So if RG has no non-trivial units, then G is determined byits group ring up to diagonal-preserving isomorphism.

  • History and the case of groups Previous results The methods

    No non-trivial units

    • If RG and RH have no non-trivial units and RG ∼= RHas R-algebras, then G ∼= (RG)×/R× ∼= (RH)×/R× ∼= H .

    • In fact, if RG has no non-trivial units and RG ∼= RH asR-algebras, then RH has no non-trivial units.

    • The proof uses that a group is a basis for its group ring.

    • So if RG has no non-trivial units, then G is determined byits group ring up to diagonal-preserving isomorphism.

    • But which group rings have no non-trivial units?

  • History and the case of groups Previous results The methods

    Group rings with no non-trivial units

    • If G has the unique product property (upp), RG has nonon-trivial units for any integral domain R.

  • History and the case of groups Previous results The methods

    Group rings with no non-trivial units

    • If G has the unique product property (upp), RG has nonon-trivial units for any integral domain R.

    • Left orderable groups have upp.

  • History and the case of groups Previous results The methods

    Group rings with no non-trivial units

    • If G has the unique product property (upp), RG has nonon-trivial units for any integral domain R.

    • Left orderable groups have upp.

    • This includes torsion-free abelian groups, free groups andbraid groups.

  • History and the case of groups Previous results The methods

    Group rings with no non-trivial units

    • If G has the unique product property (upp), RG has nonon-trivial units for any integral domain R.

    • Left orderable groups have upp.

    • This includes torsion-free abelian groups, free groups andbraid groups.

    • Upp groups are torsion-free.

  • History and the case of groups Previous results The methods

    Group rings with no non-trivial units

    • If G has the unique product property (upp), RG has nonon-trivial units for any integral domain R.

    • Left orderable groups have upp.

    • This includes torsion-free abelian groups, free groups andbraid groups.

    • Upp groups are torsion-free.

    • Higman proved if ZG has no non-trivial units, Z[G× C2]has no non-trivial units.

  • History and the case of groups Previous results The methods

    Group rings with no non-trivial units

    • If G has the unique product property (upp), RG has nonon-trivial units for any integral domain R.

    • Left orderable groups have upp.

    • This includes torsion-free abelian groups, free groups andbraid groups.

    • Upp groups are torsion-free.

    • Higman proved if ZG has no non-trivial units, Z[G× C2]has no non-trivial units.

    • He also proved that ZG has no non-trivial units if G isfinite abelian of exponent dividing 4 or 6 or if G is aquaternion group.

  • History and the case of groups Previous results The methods

    Zero divisors

    • Let R be an integral domain.

  • History and the case of groups Previous results The methods

    Zero divisors

    • Let R be an integral domain.

    • If G has torsion, RG has zero divisors.

  • History and the case of groups Previous results The methods

    Zero divisors

    • Let R be an integral domain.

    • If G has torsion, RG has zero divisors.

    • If G is torsion-free, then RG having no non-trivial unitsimplies RG has no zero divisors.

  • History and the case of groups Previous results The methods

    Zero divisors

    • Let R be an integral domain.

    • If G has torsion, RG has zero divisors.

    • If G is torsion-free, then RG having no non-trivial unitsimplies RG has no zero divisors.

    • So no non-trivial units + no zero divisors = no non-trivialunits + torsion-free.

  • History and the case of groups Previous results The methods

    Zero divisors

    • Let R be an integral domain.

    • If G has torsion, RG has zero divisors.

    • If G is torsion-free, then RG having no non-trivial unitsimplies RG has no zero divisors.

    • So no non-trivial units + no zero divisors = no non-trivialunits + torsion-free.

    • Kaplansky’s unit conjecture says if G is torsion-free, thenRG has no non-trivial units.

  • History and the case of groups Previous results The methods

    Zero divisors

    • Let R be an integral domain.

    • If G has torsion, RG has zero divisors.

    • If G is torsion-free, then RG having no non-trivial unitsimplies RG has no zero divisors.

    • So no non-trivial units + no zero divisors = no non-trivialunits + torsion-free.

    • Kaplansky’s unit conjecture says if G is torsion-free, thenRG has no non-trivial units.

    • Kaplansky’s unit conjecture implies Kaplansky’s zerodivisor conjecture.

  • History and the case of groups Previous results The methods

    Zero divisors

    • Let R be an integral domain.

    • If G has torsion, RG has zero divisors.

    • If G is torsion-free, then RG having no non-trivial unitsimplies RG has no zero divisors.

    • So no non-trivial units + no zero divisors = no non-trivialunits + torsion-free.

    • Kaplansky’s unit conjecture says if G is torsion-free, thenRG has no non-trivial units.

    • Kaplansky’s unit conjecture implies Kaplansky’s zerodivisor conjecture.

    • Note that zero divisors are irrelevant to the group ringisomorphism problem.

  • History and the case of groups Previous results The methods

    Beyond integral domains

    • If RG has no non-trivial units, it determines G for anyring R.

  • History and the case of groups Previous results The methods

    Beyond integral domains

    • If RG has no non-trivial units, it determines G for anyring R.

    • However, if G 6= 1 and RG has no non-trivial units Rmust be indecomposable and reduced.

  • History and the case of groups Previous results The methods

    Beyond integral domains

    • If RG has no non-trivial units, it determines G for anyring R.

    • However, if G 6= 1 and RG has no non-trivial units Rmust be indecomposable and reduced.

    • R is indecomposable if 0, 1 are its only idempotents (R isnot a direct product).

  • History and the case of groups Previous results The methods

    Beyond integral domains

    • If RG has no non-trivial units, it determines G for anyring R.

    • However, if G 6= 1 and RG has no non-trivial units Rmust be indecomposable and reduced.

    • R is indecomposable if 0, 1 are its only idempotents (R isnot a direct product).

    • R is reduced if 0 is its only nilpotent.

  • History and the case of groups Previous results The methods

    Beyond integral domains

    • If RG has no non-trivial units, it determines G for anyring R.

    • However, if G 6= 1 and RG has no non-trivial units Rmust be indecomposable and reduced.

    • R is indecomposable if 0, 1 are its only idempotents (R isnot a direct product).

    • R is reduced if 0 is its only nilpotent.

    • Integral domains are indecomposable and reduced.

  • History and the case of groups Previous results The methods

    Beyond integral domains

    • If RG has no non-trivial units, it determines G for anyring R.

    • However, if G 6= 1 and RG has no non-trivial units Rmust be indecomposable and reduced.

    • R is indecomposable if 0, 1 are its only idempotents (R isnot a direct product).

    • R is reduced if 0 is its only nilpotent.

    • Integral domains are indecomposable and reduced.

    • So are coordinate rings of connected affine varieties, e.g.,C[x, y]/(xy).

  • History and the case of groups Previous results The methods

    Beyond integral domains

    • If RG has no non-trivial units, it determines G for anyring R.

    • However, if G 6= 1 and RG has no non-trivial units Rmust be indecomposable and reduced.

    • R is indecomposable if 0, 1 are its only idempotents (R isnot a direct product).

    • R is reduced if 0 is its only nilpotent.

    • Integral domains are indecomposable and reduced.

    • So are coordinate rings of connected affine varieties, e.g.,C[x, y]/(xy).

    • It follows from a result of Neher if kG has no non-trivialunits for every field k, then RG has no non-trivial unitsfor every indecomposable reduced ring R.

  • History and the case of groups Previous results The methods

    Beyond integral domains

    • If RG has no non-trivial units, it determines G for anyring R.

    • However, if G 6= 1 and RG has no non-trivial units Rmust be indecomposable and reduced.

    • R is indecomposable if 0, 1 are its only idempotents (R isnot a direct product).

    • R is reduced if 0 is its only nilpotent.

    • Integral domains are indecomposable and reduced.

    • So are coordinate rings of connected affine varieties, e.g.,C[x, y]/(xy).

    • It follows from a result of Neher if kG has no non-trivialunits for every field k, then RG has no non-trivial unitsfor every indecomposable reduced ring R.

    • This applies to upp groups and left orderable groups.

  • History and the case of groups Previous results The methods

    Groupoids: previous results

    • In the following results R is an integral domain.

  • History and the case of groups Previous results The methods

    Groupoids: previous results

    • In the following results R is an integral domain.

    • Brown, Clark, an Huef proved that a diagonal-preserving∗-ring isomorphism of Leavitt path algebras implies anisomorphism of their path groupoids.

  • History and the case of groups Previous results The methods

    Groupoids: previous results

    • In the following results R is an integral domain.

    • Brown, Clark, an Huef proved that a diagonal-preserving∗-ring isomorphism of Leavitt path algebras implies anisomorphism of their path groupoids.

    • Leavitt path algebras are the ring theoretic versions ofgraph C∗-algebras.

  • History and the case of groups Previous results The methods

    Groupoids: previous results

    • In the following results R is an integral domain.

    • Brown, Clark, an Huef proved that a diagonal-preserving∗-ring isomorphism of Leavitt path algebras implies anisomorphism of their path groupoids.

    • Leavitt path algebras are the ring theoretic versions ofgraph C∗-algebras.

    • They are defined by the same groupoids as in the C∗-case.

  • History and the case of groups Previous results The methods

    Groupoids: previous results

    • In the following results R is an integral domain.

    • Brown, Clark, an Huef proved that a diagonal-preserving∗-ring isomorphism of Leavitt path algebras implies anisomorphism of their path groupoids.

    • Leavitt path algebras are the ring theoretic versions ofgraph C∗-algebras.

    • They are defined by the same groupoids as in the C∗-case.

    • Ara, Bosa, Hazrat, and Sims proved a diagonal-preservingring isomorphism between algebras of topologicallyprincipal groupoids implies an isomorphism of groupoids.

  • History and the case of groups Previous results The methods

    Groupoids: previous results

    • In the following results R is an integral domain.

    • Brown, Clark, an Huef proved that a diagonal-preserving∗-ring isomorphism of Leavitt path algebras implies anisomorphism of their path groupoids.

    • Leavitt path algebras are the ring theoretic versions ofgraph C∗-algebras.

    • They are defined by the same groupoids as in the C∗-case.

    • Ara, Bosa, Hazrat, and Sims proved a diagonal-preservingring isomorphism between algebras of topologicallyprincipal groupoids implies an isomorphism of groupoids.

    • This is the ring theoretic analogue of Renault’s result.

  • History and the case of groups Previous results The methods

    Topologically principal versus effective

    • G is topologically principal if there is a dense set ofobjects with trivial isotropy.

  • History and the case of groups Previous results The methods

    Topologically principal versus effective

    • G is topologically principal if there is a dense set ofobjects with trivial isotropy.

    • G is effective if Int(Iso(G )) = G (0).

  • History and the case of groups Previous results The methods

    Topologically principal versus effective

    • G is topologically principal if there is a dense set ofobjects with trivial isotropy.

    • G is effective if Int(Iso(G )) = G (0).

    • Topologically principal implies effective.

  • History and the case of groups Previous results The methods

    Topologically principal versus effective

    • G is topologically principal if there is a dense set ofobjects with trivial isotropy.

    • G is effective if Int(Iso(G )) = G (0).

    • Topologically principal implies effective.

    • The converse is true for second countable groupoids.

  • History and the case of groups Previous results The methods

    Topologically principal versus effective

    • G is topologically principal if there is a dense set ofobjects with trivial isotropy.

    • G is effective if Int(Iso(G )) = G (0).

    • Topologically principal implies effective.

    • The converse is true for second countable groupoids.

    • Brown, Clark, Farthing and Sims produced an effectivegroupoid with every isotropy group isomorphic to Z.

  • History and the case of groups Previous results The methods

    Topologically principal versus effective

    • G is topologically principal if there is a dense set ofobjects with trivial isotropy.

    • G is effective if Int(Iso(G )) = G (0).

    • Topologically principal implies effective.

    • The converse is true for second countable groupoids.

    • Brown, Clark, Farthing and Sims produced an effectivegroupoid with every isotropy group isomorphic to Z.

    • Guess: for any group G, there is an effective groupoidwith every isotropy group isomorphic to G.

  • History and the case of groups Previous results The methods

    Topologically principal versus effective

    • G is topologically principal if there is a dense set ofobjects with trivial isotropy.

    • G is effective if Int(Iso(G )) = G (0).

    • Topologically principal implies effective.

    • The converse is true for second countable groupoids.

    • Brown, Clark, Farthing and Sims produced an effectivegroupoid with every isotropy group isomorphic to Z.

    • Guess: for any group G, there is an effective groupoidwith every isotropy group isomorphic to G.

    • For results on simplicity, Cuntz-Krieger uniquenesstheorems, etc., effective is the right notion for groupoidalgebras.

  • History and the case of groups Previous results The methods

    Cartan subalgebras

    • G is effective iff DR(G ) is a maximal commutativesubring of RG .

  • History and the case of groups Previous results The methods

    Cartan subalgebras

    • G is effective iff DR(G ) is a maximal commutativesubring of RG .

    • If G is effective and there is a diagonal-preservingisomorphism RG → RG ′, then G ′ is effective.

  • History and the case of groups Previous results The methods

    Cartan subalgebras

    • G is effective iff DR(G ) is a maximal commutativesubring of RG .

    • If G is effective and there is a diagonal-preservingisomorphism RG → RG ′, then G ′ is effective.

    • Thus if there was a version of Ara et al. for effectivegroupoids, then effective groupoids would be determinedby their algebra and diagonal subalgebra.

  • History and the case of groups Previous results The methods

    Cartan subalgebras

    • G is effective iff DR(G ) is a maximal commutativesubring of RG .

    • If G is effective and there is a diagonal-preservingisomorphism RG → RG ′, then G ′ is effective.

    • Thus if there was a version of Ara et al. for effectivegroupoids, then effective groupoids would be determinedby their algebra and diagonal subalgebra.

    • It is not immediately obvious that being topologicallyprincipal is invariant under diagonal-preservingisomorphism.

  • History and the case of groups Previous results The methods

    Carlsen and Rout

    • The best prior result is due to Carlsen and Rout.

  • History and the case of groups Previous results The methods

    Carlsen and Rout

    • The best prior result is due to Carlsen and Rout.

    • They consider groupoids having a dense set of objectswhose isotropy group rings over R have no zero divisorsand no non-trivial units.

  • History and the case of groups Previous results The methods

    Carlsen and Rout

    • The best prior result is due to Carlsen and Rout.

    • They consider groupoids having a dense set of objectswhose isotropy group rings over R have no zero divisorsand no non-trivial units.

    • They prove that a diagonal-preserving isomorphismbetween algebras of groupoids in this class implies agroupoid isomorphism.

  • History and the case of groups Previous results The methods

    Carlsen and Rout

    • The best prior result is due to Carlsen and Rout.

    • They consider groupoids having a dense set of objectswhose isotropy group rings over R have no zero divisorsand no non-trivial units.

    • They prove that a diagonal-preserving isomorphismbetween algebras of groupoids in this class implies agroupoid isomorphism.

    • This recovers Ara et al. since the group ring of the trivialgroup has no zero divisors and no non-trivial units.

  • History and the case of groups Previous results The methods

    Carlsen and Rout

    • The best prior result is due to Carlsen and Rout.

    • They consider groupoids having a dense set of objectswhose isotropy group rings over R have no zero divisorsand no non-trivial units.

    • They prove that a diagonal-preserving isomorphismbetween algebras of groupoids in this class implies agroupoid isomorphism.

    • This recovers Ara et al. since the group ring of the trivialgroup has no zero divisors and no non-trivial units.

    • It also recovers the Leavitt path algebra result of Brownet al. without the ∗-condition.

  • History and the case of groups Previous results The methods

    Carlsen and Rout

    • The best prior result is due to Carlsen and Rout.

    • They consider groupoids having a dense set of objectswhose isotropy group rings over R have no zero divisorsand no non-trivial units.

    • They prove that a diagonal-preserving isomorphismbetween algebras of groupoids in this class implies agroupoid isomorphism.

    • This recovers Ara et al. since the group ring of the trivialgroup has no zero divisors and no non-trivial units.

    • It also recovers the Leavitt path algebra result of Brownet al. without the ∗-condition.

    • Isotropy groups in path groupoids are either trivial or Z(hence orderable).

  • History and the case of groups Previous results The methods

    Carlsen and Rout

    • The best prior result is due to Carlsen and Rout.

    • They consider groupoids having a dense set of objectswhose isotropy group rings over R have no zero divisorsand no non-trivial units.

    • They prove that a diagonal-preserving isomorphismbetween algebras of groupoids in this class implies agroupoid isomorphism.

    • This recovers Ara et al. since the group ring of the trivialgroup has no zero divisors and no non-trivial units.

    • It also recovers the Leavitt path algebra result of Brownet al. without the ∗-condition.

    • Isotropy groups in path groupoids are either trivial or Z(hence orderable).

    • This result does not cover effective groupoids and groupswith no non-trivial units but with torsion.

  • History and the case of groups Previous results The methods

    Our goals

    • We aim to improve on Carlsen and Rout in several ways.

  • History and the case of groups Previous results The methods

    Our goals

    • We aim to improve on Carlsen and Rout in several ways.

    • We want our groupoids to be determined up toisomorphism by their algebras and diagonal subalgebrasamongst all groupoids, not just a special subclass.

  • History and the case of groups Previous results The methods

    Our goals

    • We aim to improve on Carlsen and Rout in several ways.

    • We want our groupoids to be determined up toisomorphism by their algebras and diagonal subalgebrasamongst all groupoids, not just a special subclass.

    • We want to drop the no zero divisor condition since it isnot used in the group ring case.

  • History and the case of groups Previous results The methods

    Our goals

    • We aim to improve on Carlsen and Rout in several ways.

    • We want our groupoids to be determined up toisomorphism by their algebras and diagonal subalgebrasamongst all groupoids, not just a special subclass.

    • We want to drop the no zero divisor condition since it isnot used in the group ring case.

    • We want to work with the interior isotropy groups (theisotropy groups of the interior of the isotropy bundle) inorder to handle the effective case.

  • History and the case of groups Previous results The methods

    Our goals

    • We aim to improve on Carlsen and Rout in several ways.

    • We want our groupoids to be determined up toisomorphism by their algebras and diagonal subalgebrasamongst all groupoids, not just a special subclass.

    • We want to drop the no zero divisor condition since it isnot used in the group ring case.

    • We want to work with the interior isotropy groups (theisotropy groups of the interior of the isotropy bundle) inorder to handle the effective case.

    • We want to allow more general coefficient rings.

  • History and the case of groups Previous results The methods

    Our goals

    • We aim to improve on Carlsen and Rout in several ways.

    • We want our groupoids to be determined up toisomorphism by their algebras and diagonal subalgebrasamongst all groupoids, not just a special subclass.

    • We want to drop the no zero divisor condition since it isnot used in the group ring case.

    • We want to work with the interior isotropy groups (theisotropy groups of the interior of the isotropy bundle) inorder to handle the effective case.

    • We want to allow more general coefficient rings.

    • We accomplish this by generalizing the property of agroup ring having no non-trivial units to groupoid rings.

  • History and the case of groups Previous results The methods

    Our goals

    • We aim to improve on Carlsen and Rout in several ways.

    • We want our groupoids to be determined up toisomorphism by their algebras and diagonal subalgebrasamongst all groupoids, not just a special subclass.

    • We want to drop the no zero divisor condition since it isnot used in the group ring case.

    • We want to work with the interior isotropy groups (theisotropy groups of the interior of the isotropy bundle) inorder to handle the effective case.

    • We want to allow more general coefficient rings.

    • We accomplish this by generalizing the property of agroup ring having no non-trivial units to groupoid rings.

    • We then give concrete conditions to have this abstractproperty.

  • History and the case of groups Previous results The methods

    A concrete theorem

    Theorem (BS)

    Let R be indecomposable, G have a dense set of objectswhose interior isotropy group rings have no non-trivial units

    and G ′ be any ample groupoid. The following are equivalent.

    1. G ∼= G ′.

    2. There is a diagonal-preserving isomorphismΦ: RG → RG ′.

  • History and the case of groups Previous results The methods

    A concrete theorem

    Theorem (BS)

    Let R be indecomposable, G have a dense set of objectswhose interior isotropy group rings have no non-trivial units

    and G ′ be any ample groupoid. The following are equivalent.

    1. G ∼= G ′.

    2. There is a diagonal-preserving isomorphismΦ: RG → RG ′.

    • This implies Carlsen-Rout because having no non-trivialunits passes to subgroups.

  • History and the case of groups Previous results The methods

    A concrete theorem

    Theorem (BS)

    Let R be indecomposable, G have a dense set of objectswhose interior isotropy group rings have no non-trivial units

    and G ′ be any ample groupoid. The following are equivalent.

    1. G ∼= G ′.

    2. There is a diagonal-preserving isomorphismΦ: RG → RG ′.

    • This implies Carlsen-Rout because having no non-trivialunits passes to subgroups.

    • It covers effective groupoids because the interior isotropygroups are all trivial.

  • History and the case of groups Previous results The methods

    A concrete theorem

    Theorem (BS)

    Let R be indecomposable, G have a dense set of objectswhose interior isotropy group rings have no non-trivial units

    and G ′ be any ample groupoid. The following are equivalent.

    1. G ∼= G ′.

    2. There is a diagonal-preserving isomorphismΦ: RG → RG ′.

    • This implies Carlsen-Rout because having no non-trivialunits passes to subgroups.

    • It covers effective groupoids because the interior isotropygroups are all trivial.

    • It covers Leavitt path algebras over indecomposablereduced rings. Just indecomposable is needed undercondition (L).

  • History and the case of groups Previous results The methods

    Recovering groupoids

    • If RG has no non-trivial units, we recover G as(RG)×/R×.

  • History and the case of groups Previous results The methods

    Recovering groupoids

    • If RG has no non-trivial units, we recover G as(RG)×/R×.

    • This is the unit group mod the diagonal units.

  • History and the case of groups Previous results The methods

    Recovering groupoids

    • If RG has no non-trivial units, we recover G as(RG)×/R×.

    • This is the unit group mod the diagonal units.

    • We have to replace this with something else for groupoids.

  • History and the case of groups Previous results The methods

    Recovering groupoids

    • If RG has no non-trivial units, we recover G as(RG)×/R×.

    • This is the unit group mod the diagonal units.

    • We have to replace this with something else for groupoids.

    • A groupoid doesn’t “live” inside its algebra like a groupdoes.

  • History and the case of groups Previous results The methods

    Recovering groupoids

    • If RG has no non-trivial units, we recover G as(RG)×/R×.

    • This is the unit group mod the diagonal units.

    • We have to replace this with something else for groupoids.

    • A groupoid doesn’t “live” inside its algebra like a groupdoes.

    • We work with inverse semigroups instead.

  • History and the case of groups Previous results The methods

    Local bisections

    • A local bisection U of G is an open subset of G (1) withd |U , r |U injective.

  • History and the case of groups Previous results The methods

    Local bisections

    • A local bisection U of G is an open subset of G (1) withd |U , r |U injective.

    • The compact local bisections form an inverse semigroupΓc(G ) under setwise multiplication.

  • History and the case of groups Previous results The methods

    Local bisections

    • A local bisection U of G is an open subset of G (1) withd |U , r |U injective.

    • The compact local bisections form an inverse semigroupΓc(G ) under setwise multiplication.

    • It has been observed by Exel, Lawson and Lenz andprobably others that G ∼= G ′ iff Γc(G ) ∼= Γc(G

    ′) forample groupoids.

  • History and the case of groups Previous results The methods

    Local bisections

    • A local bisection U of G is an open subset of G (1) withd |U , r |U injective.

    • The compact local bisections form an inverse semigroupΓc(G ) under setwise multiplication.

    • It has been observed by Exel, Lawson and Lenz andprobably others that G ∼= G ′ iff Γc(G ) ∼= Γc(G

    ′) forample groupoids.

    • G is the tight (or ultrafilter) groupoid of Γc(G ).

  • History and the case of groups Previous results The methods

    Local bisections

    • A local bisection U of G is an open subset of G (1) withd |U , r |U injective.

    • The compact local bisections form an inverse semigroupΓc(G ) under setwise multiplication.

    • It has been observed by Exel, Lawson and Lenz andprobably others that G ∼= G ′ iff Γc(G ) ∼= Γc(G

    ′) forample groupoids.

    • G is the tight (or ultrafilter) groupoid of Γc(G ).

    • Γc(G ) embeds in RG via U 7→ χU .

  • History and the case of groups Previous results The methods

    Local bisections

    • A local bisection U of G is an open subset of G (1) withd |U , r |U injective.

    • The compact local bisections form an inverse semigroupΓc(G ) under setwise multiplication.

    • It has been observed by Exel, Lawson and Lenz andprobably others that G ∼= G ′ iff Γc(G ) ∼= Γc(G

    ′) forample groupoids.

    • G is the tight (or ultrafilter) groupoid of Γc(G ).

    • Γc(G ) embeds in RG via U 7→ χU .

    • Γc(G ) spans RG but is not a basis.

  • History and the case of groups Previous results The methods

    Local bisections

    • A local bisection U of G is an open subset of G (1) withd |U , r |U injective.

    • The compact local bisections form an inverse semigroupΓc(G ) under setwise multiplication.

    • It has been observed by Exel, Lawson and Lenz andprobably others that G ∼= G ′ iff Γc(G ) ∼= Γc(G

    ′) forample groupoids.

    • G is the tight (or ultrafilter) groupoid of Γc(G ).

    • Γc(G ) embeds in RG via U 7→ χU .

    • Γc(G ) spans RG but is not a basis.

    • We aim to recover Γc(G ).

  • History and the case of groups Previous results The methods

    Local bisections

    • A local bisection U of G is an open subset of G (1) withd |U , r |U injective.

    • The compact local bisections form an inverse semigroupΓc(G ) under setwise multiplication.

    • It has been observed by Exel, Lawson and Lenz andprobably others that G ∼= G ′ iff Γc(G ) ∼= Γc(G

    ′) forample groupoids.

    • G is the tight (or ultrafilter) groupoid of Γc(G ).

    • Γc(G ) embeds in RG via U 7→ χU .

    • Γc(G ) spans RG but is not a basis.

    • We aim to recover Γc(G ).

    • If G is a group, Γc(G) = G ∪ {0}.

  • History and the case of groups Previous results The methods

    Local bisections

    • A local bisection U of G is an open subset of G (1) withd |U , r |U injective.

    • The compact local bisections form an inverse semigroupΓc(G ) under setwise multiplication.

    • It has been observed by Exel, Lawson and Lenz andprobably others that G ∼= G ′ iff Γc(G ) ∼= Γc(G

    ′) forample groupoids.

    • G is the tight (or ultrafilter) groupoid of Γc(G ).

    • Γc(G ) embeds in RG via U 7→ χU .

    • Γc(G ) spans RG but is not a basis.

    • We aim to recover Γc(G ).

    • If G is a group, Γc(G) = G ∪ {0}.

    • What inverse semigroup replaces (RG)×?

  • History and the case of groups Previous results The methods

    The normalizer of the diagonal

    • Let N = {f ∈ RG | ∃f ′ ∈ RG , ff ′f =f, fDR(G )f

    ′ ∪ f ′DR(G )f ⊆ DR(G )}.

  • History and the case of groups Previous results The methods

    The normalizer of the diagonal

    • Let N = {f ∈ RG | ∃f ′ ∈ RG , ff ′f =f, fDR(G )f

    ′ ∪ f ′DR(G )f ⊆ DR(G )}.

    • N is the normalizer of the diagonal subalgebra.

  • History and the case of groups Previous results The methods

    The normalizer of the diagonal

    • Let N = {f ∈ RG | ∃f ′ ∈ RG , ff ′f =f, fDR(G )f

    ′ ∪ f ′DR(G )f ⊆ DR(G )}.

    • N is the normalizer of the diagonal subalgebra.

    • If G is a group, N = (RG)× ∪ {0}.

  • History and the case of groups Previous results The methods

    The normalizer of the diagonal

    • Let N = {f ∈ RG | ∃f ′ ∈ RG , ff ′f =f, fDR(G )f

    ′ ∪ f ′DR(G )f ⊆ DR(G )}.

    • N is the normalizer of the diagonal subalgebra.

    • If G is a group, N = (RG)× ∪ {0}.

    • N is an inverse semigroup containing Γc(G ).

  • History and the case of groups Previous results The methods

    The normalizer of the diagonal

    • Let N = {f ∈ RG | ∃f ′ ∈ RG , ff ′f =f, fDR(G )f

    ′ ∪ f ′DR(G )f ⊆ DR(G )}.

    • N is the normalizer of the diagonal subalgebra.

    • If G is a group, N = (RG)× ∪ {0}.

    • N is an inverse semigroup containing Γc(G ).

    • All the idempotents of N are in DR(G ) and hencecommute.

  • History and the case of groups Previous results The methods

    The normalizer of the diagonal

    • Let N = {f ∈ RG | ∃f ′ ∈ RG , ff ′f =f, fDR(G )f

    ′ ∪ f ′DR(G )f ⊆ DR(G )}.

    • N is the normalizer of the diagonal subalgebra.

    • If G is a group, N = (RG)× ∪ {0}.

    • N is an inverse semigroup containing Γc(G ).

    • All the idempotents of N are in DR(G ) and hencecommute.

    • If R is indecomposable, E(DR(G )) = E(N) = E(Γc(G )).

  • History and the case of groups Previous results The methods

    The normalizer of the diagonal

    • Let N = {f ∈ RG | ∃f ′ ∈ RG , ff ′f =f, fDR(G )f

    ′ ∪ f ′DR(G )f ⊆ DR(G )}.

    • N is the normalizer of the diagonal subalgebra.

    • If G is a group, N = (RG)× ∪ {0}.

    • N is an inverse semigroup containing Γc(G ).

    • All the idempotents of N are in DR(G ) and hencecommute.

    • If R is indecomposable, E(DR(G )) = E(N) = E(Γc(G )).

    • Previous papers prove these facts by explicitly describingthe elements of N using their full hypotheses.

  • History and the case of groups Previous results The methods

    The normalizer of the diagonal

    • Let N = {f ∈ RG | ∃f ′ ∈ RG , ff ′f =f, fDR(G )f

    ′ ∪ f ′DR(G )f ⊆ DR(G )}.

    • N is the normalizer of the diagonal subalgebra.

    • If G is a group, N = (RG)× ∪ {0}.

    • N is an inverse semigroup containing Γc(G ).

    • All the idempotents of N are in DR(G ) and hencecommute.

    • If R is indecomposable, E(DR(G )) = E(N) = E(Γc(G )).

    • Previous papers prove these facts by explicitly describingthe elements of N using their full hypotheses.

    • Our proof is direct and elementary.

  • History and the case of groups Previous results The methods

    Normal subsemigroups

    • RG has no non-trivial units when G is isomorphic to unitsmod diagonal units.

  • History and the case of groups Previous results The methods

    Normal subsemigroups

    • RG has no non-trivial units when G is isomorphic to unitsmod diagonal units.

    • The analogue for us should be normalizers mod diagonalnormalizers.

  • History and the case of groups Previous results The methods

    Normal subsemigroups

    • RG has no non-trivial units when G is isomorphic to unitsmod diagonal units.

    • The analogue for us should be normalizers mod diagonalnormalizers.

    • If S is an inverse semigroup, K ≤ S is a normal inversesubsemigroup if:

  • History and the case of groups Previous results The methods

    Normal subsemigroups

    • RG has no non-trivial units when G is isomorphic to unitsmod diagonal units.

    • The analogue for us should be normalizers mod diagonalnormalizers.

    • If S is an inverse semigroup, K ≤ S is a normal inversesubsemigroup if:1. E(K) = E(S) (K is full);

  • History and the case of groups Previous results The methods

    Normal subsemigroups

    • RG has no non-trivial units when G is isomorphic to unitsmod diagonal units.

    • The analogue for us should be normalizers mod diagonalnormalizers.

    • If S is an inverse semigroup, K ≤ S is a normal inversesubsemigroup if:1. E(K) = E(S) (K is full);2. sKs−1 ⊆ K for all s ∈ S.

  • History and the case of groups Previous results The methods

    Normal subsemigroups

    • RG has no non-trivial units when G is isomorphic to unitsmod diagonal units.

    • The analogue for us should be normalizers mod diagonalnormalizers.

    • If S is an inverse semigroup, K ≤ S is a normal inversesubsemigroup if:1. E(K) = E(S) (K is full);2. sKs−1 ⊆ K for all s ∈ S.

    • If φ : S → T is a homomorphism, ker φ = φ−1(E(T )) is anormal subsemigroup.

  • History and the case of groups Previous results The methods

    Normal subsemigroups

    • RG has no non-trivial units when G is isomorphic to unitsmod diagonal units.

    • The analogue for us should be normalizers mod diagonalnormalizers.

    • If S is an inverse semigroup, K ≤ S is a normal inversesubsemigroup if:1. E(K) = E(S) (K is full);2. sKs−1 ⊆ K for all s ∈ S.

    • If φ : S → T is a homomorphism, ker φ = φ−1(E(T )) is anormal subsemigroup.

    • Normal subsemigroups generalize normal subgroups.

  • History and the case of groups Previous results The methods

    Normal subsemigroups

    • RG has no non-trivial units when G is isomorphic to unitsmod diagonal units.

    • The analogue for us should be normalizers mod diagonalnormalizers.

    • If S is an inverse semigroup, K ≤ S is a normal inversesubsemigroup if:1. E(K) = E(S) (K is full);2. sKs−1 ⊆ K for all s ∈ S.

    • If φ : S → T is a homomorphism, ker φ = φ−1(E(T )) is anormal subsemigroup.

    • Normal subsemigroups generalize normal subgroups.

    • Idempotent-separating congruences are determined byappropriate normal subsemigroups.

  • History and the case of groups Previous results The methods

    Idempotent-separating congruences

    • A congruence or homomorphism on S isidempotent-separating if it is injective on idempotents.

  • History and the case of groups Previous results The methods

    Idempotent-separating congruences

    • A congruence or homomorphism on S isidempotent-separating if it is injective on idempotents.

    • An idempotent-separating congruence is uniquelydetermined by its kernel.

  • History and the case of groups Previous results The methods

    Idempotent-separating congruences

    • A congruence or homomorphism on S isidempotent-separating if it is injective on idempotents.

    • An idempotent-separating congruence is uniquelydetermined by its kernel.

    • In particular, it is injective iff its kernel is E(S)(idempotent-pure).

  • History and the case of groups Previous results The methods

    Idempotent-separating congruences

    • A congruence or homomorphism on S isidempotent-separating if it is injective on idempotents.

    • An idempotent-separating congruence is uniquelydetermined by its kernel.

    • In particular, it is injective iff its kernel is E(S)(idempotent-pure).

    • K ⊳ S is the kernel of an idempotent-separatingcongruence iff aa−1 = a−1a for all a ∈ K.

  • History and the case of groups Previous results The methods

    Idempotent-separating congruences

    • A congruence or homomorphism on S isidempotent-separating if it is injective on idempotents.

    • An idempotent-separating congruence is uniquelydetermined by its kernel.

    • In particular, it is injective iff its kernel is E(S)(idempotent-pure).

    • K ⊳ S is the kernel of an idempotent-separatingcongruence iff aa−1 = a−1a for all a ∈ K.

    • D(N) = N ∩DR(G ) is a normal subsemigroup of N .

  • History and the case of groups Previous results The methods

    Idempotent-separating congruences

    • A congruence or homomorphism on S isidempotent-separating if it is injective on idempotents.

    • An idempotent-separating congruence is uniquelydetermined by its kernel.

    • In particular, it is injective iff its kernel is E(S)(idempotent-pure).

    • K ⊳ S is the kernel of an idempotent-separatingcongruence iff aa−1 = a−1a for all a ∈ K.

    • D(N) = N ∩DR(G ) is a normal subsemigroup of N .

    • Since D(N) is commutative it satisfies the abovecondition.

  • History and the case of groups Previous results The methods

    Idempotent-separating congruences

    • A congruence or homomorphism on S isidempotent-separating if it is injective on idempotents.

    • An idempotent-separating congruence is uniquelydetermined by its kernel.

    • In particular, it is injective iff its kernel is E(S)(idempotent-pure).

    • K ⊳ S is the kernel of an idempotent-separatingcongruence iff aa−1 = a−1a for all a ∈ K.

    • D(N) = N ∩DR(G ) is a normal subsemigroup of N .

    • Since D(N) is commutative it satisfies the abovecondition.

    • So there is an idempotent-separating quotientπ : N → N/D(N).

  • History and the case of groups Previous results The methods

    The local bisection hypothesis

    • The restriction of π to Γc(G ) is injective since thediagonal elements of Γc(G ) are exactly the idempotents.

  • History and the case of groups Previous results The methods

    The local bisection hypothesis

    • The restriction of π to Γc(G ) is injective since thediagonal elements of Γc(G ) are exactly the idempotents.

    • So we have a commutative diagram

    Γc(G )

    ψ

    N

    π

    N/D(N)

    with ψ injective.

  • History and the case of groups Previous results The methods

    The local bisection hypothesis

    • The restriction of π to Γc(G ) is injective since thediagonal elements of Γc(G ) are exactly the idempotents.

    • So we have a commutative diagram

    Γc(G )

    ψ

    N

    π

    N/D(N)

    with ψ injective.• For groups, ψ is an isomorphism iff RG has no non-trivialunits.

  • History and the case of groups Previous results The methods

    The local bisection hypothesis

    • The restriction of π to Γc(G ) is injective since thediagonal elements of Γc(G ) are exactly the idempotents.

    • So we have a commutative diagram

    Γc(G )

    ψ

    N

    π

    N/D(N)

    with ψ injective.• For groups, ψ is an isomorphism iff RG has no non-trivialunits.

    • This is equivalent to saying each unit has singletonsupport, i.e., is a local bisection.

  • History and the case of groups Previous results The methods

    The local bisection hypothesis

    • The restriction of π to Γc(G ) is injective since thediagonal elements of Γc(G ) are exactly the idempotents.

    • So we have a commutative diagram

    Γc(G )

    ψ

    N

    π

    N/D(N)

    with ψ injective.• For groups, ψ is an isomorphism iff RG has no non-trivialunits.

    • This is equivalent to saying each unit has singletonsupport, i.e., is a local bisection.

    • We say G satisfies the local bisection hypothesis if thesupport of each element of N is a local bisection.

  • History and the case of groups Previous results The methods

    The local bisection hypothesis

    • The restriction of π to Γc(G ) is injective since thediagonal elements of Γc(G ) are exactly the idempotents.

    • So we have a commutative diagram

    Γc(G )

    ψ

    N

    π

    N/D(N)

    with ψ injective.• For groups, ψ is an isomorphism iff RG has no non-trivialunits.

    • This is equivalent to saying each unit has singletonsupport, i.e., is a local bisection.

    • We say G satisfies the local bisection hypothesis if thesupport of each element of N is a local bisection.

    • This occurs iff ψ is an isomorphism.

  • History and the case of groups Previous results The methods

    The local bisection hypothesis II

    • The construction G 7→ N/D(N) is clearly functorial withrespect to diagonal-preserving isomorphisms.

  • History and the case of groups Previous results The methods

    The local bisection hypothesis II

    • The construction G 7→ N/D(N) is clearly functorial withrespect to diagonal-preserving isomorphisms.

    • So if G and G ′ satisfy the local bisection hypothesis andtheir algebras are isomorphic via a diagonal-preservingisomorphism, then Γc(G ) ∼= Γc(G

    ′) and hence G ∼= G ′.

  • History and the case of groups Previous results The methods

    The local bisection hypothesis II

    • The construction G 7→ N/D(N) is clearly functorial withrespect to diagonal-preserving isomorphisms.

    • So if G and G ′ satisfy the local bisection hypothesis andtheir algebras are isomorphic via a diagonal-preservingisomorphism, then Γc(G ) ∼= Γc(G

    ′) and hence G ∼= G ′.

    • In fact, diagonal-preserving isomorphisms preserve thelocal bisection hypothesis and so the assumption is onlyneeded for one of the groupoids.

  • History and the case of groups Previous results The methods

    The local bisection hypothesis II

    • The construction G 7→ N/D(N) is clearly functorial withrespect to diagonal-preserving isomorphisms.

    • So if G and G ′ satisfy the local bisection hypothesis andtheir algebras are isomorphic via a diagonal-preservingisomorphism, then Γc(G ) ∼= Γc(G

    ′) and hence G ∼= G ′.

    • In fact, diagonal-preserving isomorphisms preserve thelocal bisection hypothesis and so the assumption is onlyneeded for one of the groupoids.

    • The proof is a bit like the group ring case but is trickierbecause Γc(G ) is not linearly independent in RG .

  • History and the case of groups Previous results The methods

    The local bisection hypothesis II

    • The construction G 7→ N/D(N) is clearly functorial withrespect to diagonal-preserving isomorphisms.

    • So if G and G ′ satisfy the local bisection hypothesis andtheir algebras are isomorphic via a diagonal-preservingisomorphism, then Γc(G ) ∼= Γc(G

    ′) and hence G ∼= G ′.

    • In fact, diagonal-preserving isomorphisms preserve thelocal bisection hypothesis and so the assumption is onlyneeded for one of the groupoids.

    • The proof is a bit like the group ring case but is trickierbecause Γc(G ) is not linearly independent in RG .

    • We use instead the order structure of Γc(G ).

  • History and the case of groups Previous results The methods

    The interior of the isotropy bundle

    • Let H = Int(Iso(G )).

  • History and the case of groups Previous results The methods

    The interior of the isotropy bundle

    • Let H = Int(Iso(G )).

    • Then G (0) ⊆ H ⊆ G and these are open subgroupoids.

  • History and the case of groups Previous results The methods

    The interior of the isotropy bundle

    • Let H = Int(Iso(G )).

    • Then G (0) ⊆ H ⊆ G and these are open subgroupoids.

    • Thus DR(G ) ≤ RH ≤ RG .

  • History and the case of groups Previous results The methods

    The interior of the isotropy bundle

    • Let H = Int(Iso(G )).

    • Then G (0) ⊆ H ⊆ G and these are open subgroupoids.

    • Thus DR(G ) ≤ RH ≤ RG .

    • Moreover, DR(G ) = DR(H ).

  • History and the case of groups Previous results The methods

    The interior of the isotropy bundle

    • Let H = Int(Iso(G )).

    • Then G (0) ⊆ H ⊆ G and these are open subgroupoids.

    • Thus DR(G ) ≤ RH ≤ RG .

    • Moreover, DR(G ) = DR(H ).

    Theorem (BS)

    G satisfies the local bisection hypothesis iff H does.

  • History and the case of groups Previous results The methods

    The interior of the isotropy bundle

    • Let H = Int(Iso(G )).

    • Then G (0) ⊆ H ⊆ G and these are open subgroupoids.

    • Thus DR(G ) ≤ RH ≤ RG .

    • Moreover, DR(G ) = DR(H ).

    Theorem (BS)

    G satisfies the local bisection hypothesis iff H does.

    • So a groupoid is determined by its algebra and itsdiagonal subalgebra if the interior of its isotropy bundlesatisfies the local bisection hypothesis.

  • History and the case of groups Previous results The methods

    The interior of the isotropy bundle

    • Let H = Int(Iso(G )).

    • Then G (0) ⊆ H ⊆ G and these are open subgroupoids.

    • Thus DR(G ) ≤ RH ≤ RG .

    • Moreover, DR(G ) = DR(H ).

    Theorem (BS)

    G satisfies the local bisection hypothesis iff H does.

    • So a groupoid is determined by its algebra and itsdiagonal subalgebra if the interior of its isotropy bundlesatisfies the local bisection hypothesis.

    • I’m not convinced that this condition has a simplerreformulation.

  • History and the case of groups Previous results The methods

    The interior of the isotropy bundle

    • Let H = Int(Iso(G )).

    • Then G (0) ⊆ H ⊆ G and these are open subgroupoids.

    • Thus DR(G ) ≤ RH ≤ RG .

    • Moreover, DR(G ) = DR(H ).

    Theorem (BS)

    G satisfies the local bisection hypothesis iff H does.

    • So a groupoid is determined by its algebra and itsdiagonal subalgebra if the interior of its isotropy bundlesatisfies the local bisection hypothesis.

    • I’m not convinced that this condition has a simplerreformulation.

    • The key point is f ∈ N and α, β ∈ supp(f) impliesd(α) = d(β) ⇐⇒ r(α) = r(β).

  • History and the case of groups Previous results The methods

    Some sufficient conditions

    • If there is a dense set of objects such that the interiorisotropy group rings have no non-trivial units, it is easy toshow that H satisfies the local bisection hypothesis.

  • History and the case of groups Previous results The methods

    Some sufficient conditions

    • If there is a dense set of objects such that the interiorisotropy group rings have no non-trivial units, it is easy toshow that H satisfies the local bisection hypothesis.

    • If G (0) has a dense set X of isolated points, then Gsatisfies the local bisection hypothesis iff the isotropygroup ring at each x ∈ X has no non-trivial units.

  • History and the case of groups Previous results The methods

    Some sufficient conditions

    • If there is a dense set of objects such that the interiorisotropy group rings have no non-trivial units, it is easy toshow that H satisfies the local bisection hypothesis.

    • If G (0) has a dense set X of isolated points, then Gsatisfies the local bisection hypothesis iff the isotropygroup ring at each x ∈ X has no non-trivial units.

    • A number of inverse semigroup universal groupoids havethis property.

  • History and the case of groups Previous results The methods

    Some sufficient conditions

    • If there is a dense set of objects such that the interiorisotropy group rings have no non-trivial units, it is easy toshow that H satisfies the local bisection hypothesis.

    • If G (0) has a dense set X of isolated points, then Gsatisfies the local bisection hypothesis iff the isotropygroup ring at each x ∈ X has no non-trivial units.

    • A number of inverse semigroup universal groupoids havethis property.

    • An element of RH belongs to N iff its restriction toeach isotropy group is either 0 or a unit and its support isa local bisection iff each restriction is either 0 or a trivialunit.

  • History and the case of groups Previous results The methods

    Some sufficient conditions

    • If there is a dense set of objects such that the interiorisotropy group rings have no non-trivial units, it is easy toshow that H satisfies the local bisection hypothesis.

    • If G (0) has a dense set X of isolated points, then Gsatisfies the local bisection hypothesis iff the isotropygroup ring at each x ∈ X has no non-trivial units.

    • A number of inverse semigroup universal groupoids havethis property.

    • An element of RH belongs to N iff its restriction toeach isotropy group is either 0 or a unit and its support isa local bisection iff each restriction is either 0 or a trivialunit.

    • So for group bundles, the local bisection hypothesis isquite similar to the no non-trivial unit hypothesis forgroup rings.

  • History and the case of groups Previous results The methods

    Graded groupoids

    • One can consider more generally an ample groupoid Gwith a cocycle c : G → G.

  • History and the case of groups Previous results The methods

    Graded groupoids

    • One can consider more generally an ample groupoid Gwith a cocycle c : G → G.

    • Then RG is G-graded.

  • History and the case of groups Previous results The methods

    Graded groupoids

    • One can consider more generally an ample groupoid Gwith a cocycle c : G → G.

    • Then RG is G-graded.

    • The question is whether the pair (RG , DR(G )), taken upto graded isomorphism determines G and the cocycle.

  • History and the case of groups Previous results The methods

    Graded groupoids

    • One can consider more generally an ample groupoid Gwith a cocycle c : G → G.

    • Then RG is G-graded.

    • The question is whether the pair (RG , DR(G )), taken upto graded isomorphism determines G and the cocycle.

    • We show this is the case if the homogeneous componentof 1 satisfies the local bisection condition.

  • History and the case of groups Previous results The methods

    Graded groupoids

    • One can consider more generally an ample groupoid Gwith a cocycle c : G → G.

    • Then RG is G-graded.

    • The question is whether the pair (RG , DR(G )), taken upto graded isomorphism determines G and the cocycle.

    • We show this is the case if the homogeneous componentof 1 satisfies the local bisection condition.

    • The proof is mostly the same but we work with gradedinverse semigroups.

  • History and the case of groups Previous results The methods

    Graded groupoids

    • One can consider more generally an ample groupoid Gwith a cocycle c : G → G.

    • Then RG is G-graded.

    • The question is whether the pair (RG , DR(G )), taken upto graded isomorphism determines G and the cocycle.

    • We show this is the case if the homogeneous componentof 1 satisfies the local bisection condition.

    • The proof is mostly the same but we work with gradedinverse semigroups.

    • Carlsen and Rout and Ara et al. also worked in the gradedsetting.

  • History and the case of groups Previous results The methods

    The end

    Thank you for your attention!

    History and the case of groupsPrevious resultsThe methods