Cloak and dagger
Chris Heunen
1 / 34 Algebra and coalgebra
Increasing generality:
I Vector space with bilinear (co)multiplication
I (Co)monoid in monoidal category
I (Co)monad: (co)monoid in functor category
I (Co)algebras for a (co)monad
Interaction between algebra and coalgebra?
2 / 34 I Purpose of cloak is to obscure presence or movement of dagger
I Dagger, a concealable and silent weapon: dagger categories
I Cloak, worn to hide identity: Frobenius law
Cloak and dagger
I Situation involving secrecy or mystery
3 / 34 I Dagger, a concealable and silent weapon: dagger categories
I Cloak, worn to hide identity: Frobenius law
Cloak and dagger
I Situation involving secrecy or mystery
I Purpose of cloak is to obscure presence or movement of dagger
3 / 34 Cloak and dagger
I Situation involving secrecy or mystery
I Purpose of cloak is to obscure presence or movement of dagger
I Dagger, a concealable and silent weapon: dagger categories
I Cloak, worn to hide identity: Frobenius law
3 / 34 Dagger
4 / 34 † Dagger: functor Cop → C with A† = A on objects, f †† = f on maps
f f † A −→ BB −→ A
Dagger category: category equipped with dagger
† −1 I Invertible computing: groupoid, f = f † op I Possibilistic computing: sets and relations, R = R
I Partial invertible computing: sets and partial injections † T I Probabilistic computing: doubly stochastic maps, f = f † T I Quantum computing: Hilbert spaces, f = f † † I Second order: dagger functors F (f) = F (f )
I Unitary representations:[G, Hilb]†
Dagger Method to turn algebra into coalgebra: self-duality Cop ' C
5 / 34 † −1 I Invertible computing: groupoid, f = f † op I Possibilistic computing: sets and relations, R = R
I Partial invertible computing: sets and partial injections † T I Probabilistic computing: doubly stochastic maps, f = f † T I Quantum computing: Hilbert spaces, f = f † † I Second order: dagger functors F (f) = F (f )
I Unitary representations:[G, Hilb]†
Dagger Method to turn algebra into coalgebra: self-duality Cop ' C
† Dagger: functor Cop → C with A† = A on objects, f †† = f on maps
f f † A −→ BB −→ A
Dagger category: category equipped with dagger
5 / 34 † op I Possibilistic computing: sets and relations, R = R
I Partial invertible computing: sets and partial injections † T I Probabilistic computing: doubly stochastic maps, f = f † T I Quantum computing: Hilbert spaces, f = f † † I Second order: dagger functors F (f) = F (f )
I Unitary representations:[G, Hilb]†
Dagger Method to turn algebra into coalgebra: self-duality Cop ' C
† Dagger: functor Cop → C with A† = A on objects, f †† = f on maps
f f † A −→ BB −→ A
Dagger category: category equipped with dagger
† −1 I Invertible computing: groupoid, f = f
5 / 34 I Partial invertible computing: sets and partial injections † T I Probabilistic computing: doubly stochastic maps, f = f † T I Quantum computing: Hilbert spaces, f = f † † I Second order: dagger functors F (f) = F (f )
I Unitary representations:[G, Hilb]†
Dagger Method to turn algebra into coalgebra: self-duality Cop ' C
† Dagger: functor Cop → C with A† = A on objects, f †† = f on maps
f f † A −→ BB −→ A
Dagger category: category equipped with dagger
† −1 I Invertible computing: groupoid, f = f † op I Possibilistic computing: sets and relations, R = R
5 / 34 † T I Probabilistic computing: doubly stochastic maps, f = f † T I Quantum computing: Hilbert spaces, f = f † † I Second order: dagger functors F (f) = F (f )
I Unitary representations:[G, Hilb]†
Dagger Method to turn algebra into coalgebra: self-duality Cop ' C
† Dagger: functor Cop → C with A† = A on objects, f †† = f on maps
f f † A −→ BB −→ A
Dagger category: category equipped with dagger
† −1 I Invertible computing: groupoid, f = f † op I Possibilistic computing: sets and relations, R = R
I Partial invertible computing: sets and partial injections
5 / 34 † T I Quantum computing: Hilbert spaces, f = f † † I Second order: dagger functors F (f) = F (f )
I Unitary representations:[G, Hilb]†
Dagger Method to turn algebra into coalgebra: self-duality Cop ' C
† Dagger: functor Cop → C with A† = A on objects, f †† = f on maps
f f † A −→ BB −→ A
Dagger category: category equipped with dagger
† −1 I Invertible computing: groupoid, f = f † op I Possibilistic computing: sets and relations, R = R
I Partial invertible computing: sets and partial injections † T I Probabilistic computing: doubly stochastic maps, f = f
5 / 34 † † I Second order: dagger functors F (f) = F (f )
I Unitary representations:[G, Hilb]†
Dagger Method to turn algebra into coalgebra: self-duality Cop ' C
† Dagger: functor Cop → C with A† = A on objects, f †† = f on maps
f f † A −→ BB −→ A
Dagger category: category equipped with dagger
† −1 I Invertible computing: groupoid, f = f † op I Possibilistic computing: sets and relations, R = R
I Partial invertible computing: sets and partial injections † T I Probabilistic computing: doubly stochastic maps, f = f † T I Quantum computing: Hilbert spaces, f = f
5 / 34 I Unitary representations:[G, Hilb]†
Dagger Method to turn algebra into coalgebra: self-duality Cop ' C
† Dagger: functor Cop → C with A† = A on objects, f †† = f on maps
f f † A −→ BB −→ A
Dagger category: category equipped with dagger
† −1 I Invertible computing: groupoid, f = f † op I Possibilistic computing: sets and relations, R = R
I Partial invertible computing: sets and partial injections † T I Probabilistic computing: doubly stochastic maps, f = f † T I Quantum computing: Hilbert spaces, f = f † † I Second order: dagger functors F (f) = F (f )
5 / 34 Dagger Method to turn algebra into coalgebra: self-duality Cop ' C
† Dagger: functor Cop → C with A† = A on objects, f †† = f on maps
f f † A −→ BB −→ A
Dagger category: category equipped with dagger
† −1 I Invertible computing: groupoid, f = f † op I Possibilistic computing: sets and relations, R = R
I Partial invertible computing: sets and partial injections † T I Probabilistic computing: doubly stochastic maps, f = f † T I Quantum computing: Hilbert spaces, f = f † † I Second order: dagger functors F (f) = F (f )
I Unitary representations:[G, Hilb]†
5 / 34 † I Evil: demand equality A = A of objects
I Dagger category theory different beast: isomorphism is not the correct notion of ‘sameness’
“Homotopy type theory” Univalent foundations program, 2013
Never bring a knife to a gun fight
I Terminology after (physics) notation (but beats identity-on-objects-involutive-contravariant-functor)
6 / 34 I Dagger category theory different beast: isomorphism is not the correct notion of ‘sameness’
Never bring a knife to a gun fight
I Terminology after (physics) notation (but beats identity-on-objects-involutive-contravariant-functor)
† I Evil: demand equality A = A of objects
“Homotopy type theory” Univalent foundations program, 2013
6 / 34 Never bring a knife to a gun fight
I Terminology after (physics) notation (but beats identity-on-objects-involutive-contravariant-functor)
† I Evil: demand equality A = A of objects
I Dagger category theory different beast: isomorphism is not the correct notion of ‘sameness’
“Homotopy type theory” Univalent foundations program, 2013
6 / 34 −1 † I dagger isomorphism (unitary): f = f
† I dagger monic (isometry): f ◦ f = id
I dagger equalizer / kernel / biproduct
† † † −1 † I monoidal dagger category: (f ⊗ g) = f ⊗ g , α = α
What about monoids??
Way of the dagger
Motto: “everything in sight ought to cooperate with the dagger”
7 / 34 I dagger equalizer / kernel / biproduct
† † † −1 † I monoidal dagger category: (f ⊗ g) = f ⊗ g , α = α
What about monoids??
Way of the dagger
Motto: “everything in sight ought to cooperate with the dagger”
−1 † I dagger isomorphism (unitary): f = f
† I dagger monic (isometry): f ◦ f = id
7 / 34 † † † −1 † I monoidal dagger category: (f ⊗ g) = f ⊗ g , α = α
What about monoids??
Way of the dagger
Motto: “everything in sight ought to cooperate with the dagger”
−1 † I dagger isomorphism (unitary): f = f
† I dagger monic (isometry): f ◦ f = id
I dagger equalizer / kernel / biproduct
7 / 34 What about monoids??
Way of the dagger
Motto: “everything in sight ought to cooperate with the dagger”
−1 † I dagger isomorphism (unitary): f = f
† I dagger monic (isometry): f ◦ f = id
I dagger equalizer / kernel / biproduct
† † † −1 † I monoidal dagger category: (f ⊗ g) = f ⊗ g , α = α
7 / 34 Way of the dagger
Motto: “everything in sight ought to cooperate with the dagger”
−1 † I dagger isomorphism (unitary): f = f
† I dagger monic (isometry): f ◦ f = id
I dagger equalizer / kernel / biproduct
† † † −1 † I monoidal dagger category: (f ⊗ g) = f ⊗ g , α = α
What about monoids??
7 / 34 Cloaks are worn
8 / 34 I algebra A with finitely many minimal right ideals
I algebra A with nondegenerate A ⊗ A → k with [ab, c] = [a, bc]
I algebra A with comultiplication δ : A → A ⊗ A satisfying
(id ⊗ µ) ◦ (δ ⊗ id) = (µ ⊗ id) ◦ (id ⊗ δ)
I algebra A with equivalent left and right regular representations
“Theorie der hyperkomplexen Gr¨oßen I” Sitzungsberichte der Preussischen Akademie der Wissenschaften 504–537, 1903
“On Frobeniusean algebras II” Annals of Mathematics 42(1):1–21, 1941
Frobenius algebra Many definitions over a field k:
I algebra A with functional A → k, kernel without left ideals
9 / 34 I algebra A with nondegenerate A ⊗ A → k with [ab, c] = [a, bc]
I algebra A with comultiplication δ : A → A ⊗ A satisfying
(id ⊗ µ) ◦ (δ ⊗ id) = (µ ⊗ id) ◦ (id ⊗ δ)
I algebra A with equivalent left and right regular representations
“Theorie der hyperkomplexen Gr¨oßen I” Sitzungsberichte der Preussischen Akademie der Wissenschaften 504–537, 1903
“On Frobeniusean algebras II” Annals of Mathematics 42(1):1–21, 1941
Frobenius algebra Many definitions over a field k:
I algebra A with functional A → k, kernel without left ideals
I algebra A with finitely many minimal right ideals
9 / 34 I algebra A with comultiplication δ : A → A ⊗ A satisfying
(id ⊗ µ) ◦ (δ ⊗ id) = (µ ⊗ id) ◦ (id ⊗ δ)
I algebra A with equivalent left and right regular representations
“Theorie der hyperkomplexen Gr¨oßen I” Sitzungsberichte der Preussischen Akademie der Wissenschaften 504–537, 1903
“On Frobeniusean algebras II” Annals of Mathematics 42(1):1–21, 1941
Frobenius algebra Many definitions over a field k:
I algebra A with functional A → k, kernel without left ideals
I algebra A with finitely many minimal right ideals
I algebra A with nondegenerate A ⊗ A → k with [ab, c] = [a, bc]
9 / 34 I algebra A with equivalent left and right regular representations
“Theorie der hyperkomplexen Gr¨oßen I” Sitzungsberichte der Preussischen Akademie der Wissenschaften 504–537, 1903
“On Frobeniusean algebras II” Annals of Mathematics 42(1):1–21, 1941
Frobenius algebra Many definitions over a field k:
I algebra A with functional A → k, kernel without left ideals
I algebra A with finitely many minimal right ideals
I algebra A with nondegenerate A ⊗ A → k with [ab, c] = [a, bc]
I algebra A with comultiplication δ : A → A ⊗ A satisfying
(id ⊗ µ) ◦ (δ ⊗ id) = (µ ⊗ id) ◦ (id ⊗ δ)
9 / 34 Frobenius algebra Many definitions over a field k:
I algebra A with functional A → k, kernel without left ideals
I algebra A with finitely many minimal right ideals
I algebra A with nondegenerate A ⊗ A → k with [ab, c] = [a, bc]
I algebra A with comultiplication δ : A → A ⊗ A satisfying
(id ⊗ µ) ◦ (δ ⊗ id) = (µ ⊗ id) ◦ (id ⊗ δ)
I algebra A with equivalent left and right regular representations
“Theorie der hyperkomplexen Gr¨oßen I” Sitzungsberichte der Preussischen Akademie der Wissenschaften 504–537, 1903
“On Frobeniusean algebras II” Annals of Mathematics 42(1):1–21, 1941
9 / 34 P −1 I comultiplication g 7→ h gh ⊗ h P −1 I both sides of Frobenius law evaluate to k gk ⊗ kh on g ⊗ h
So Frobenius algebra incorporates finite group representation theory
Frobenius law in algebra
Any finite group G induces Frobenius group algebra A:
I A has orthonormal basis {g ∈ G}
I multiplication g ⊗ h 7→ gh
I unit e
10 / 34 Frobenius law in algebra
Any finite group G induces Frobenius group algebra A:
I A has orthonormal basis {g ∈ G}
I multiplication g ⊗ h 7→ gh
I unit e
P −1 I comultiplication g 7→ h gh ⊗ h P −1 I both sides of Frobenius law evaluate to k gk ⊗ kh on g ⊗ h
So Frobenius algebra incorporates finite group representation theory
10 / 34 I Frobenius property is independent of base field k!
I Extension of scalars: if l extends k, then A Frobenius over k iff l ⊗k A Frobenius over l
I Restriction of scalars: if l extends k, then A Frobenius over l iff A Frobenius over k
Frobenius law in algebra
Frobenius algebras are wonderful:
I left and right Artinian
I left and right self-injective
11 / 34 Frobenius law in algebra
Frobenius algebras are wonderful:
I left and right Artinian
I left and right self-injective
I Frobenius property is independent of base field k!
I Extension of scalars: if l extends k, then A Frobenius over k iff l ⊗k A Frobenius over l
I Restriction of scalars: if l extends k, then A Frobenius over l iff A Frobenius over k
11 / 34 I Coding theory:
I Hamming weight of linear code and dual code related I code isomorphism that preserves Hamming weight is monomial
I Geometry: cohomology rings of compact oriented manifolds are Frobenius
“Combinatorial properties of elementary abelian groups” Radcliffe College, Cambridge, 1962
Frobenius law in mathematics
I Number theory: commutative Frobenius algebras are Gorenstein
“Modular elliptic curves and Fermat’s last theorem” Annals of Mathematics 142(3):443–551, 1995
12 / 34 I Geometry: cohomology rings of compact oriented manifolds are Frobenius
Frobenius law in mathematics
I Number theory: commutative Frobenius algebras are Gorenstein
I Coding theory:
I Hamming weight of linear code and dual code related I code isomorphism that preserves Hamming weight is monomial
“Modular elliptic curves and Fermat’s last theorem” Annals of Mathematics 142(3):443–551, 1995
“Combinatorial properties of elementary abelian groups” Radcliffe College, Cambridge, 1962
12 / 34 Frobenius law in mathematics
I Number theory: commutative Frobenius algebras are Gorenstein
I Coding theory:
I Hamming weight of linear code and dual code related I code isomorphism that preserves Hamming weight is monomial
I Geometry: cohomology rings of compact oriented manifolds are Frobenius
“Modular elliptic curves and Fermat’s last theorem” Annals of Mathematics 142(3):443–551, 1995
“Combinatorial properties of elementary abelian groups” Radcliffe College, Cambridge, 1962
12 / 34 Computes manifold invariants via Pachner moves:
⇐⇒
I Topological quantum field theory depends only on topology 2d TQFTs ' commutative Frobenius algebras
I RT construction turns modular tensor category into 3d TQFT
“A geometric approach to two-dimensional conformal field theory” University of Utrecht, 1989
“Invariants of 3-manifolds via link polynomials and quantum groups” Inventiones Mathematicae 103(3):547–597, 1991
“P. L. homeomorphic manifolds are equivalent by elementary shellings” European Journal of Combinatorics 12(2):129–145, 1991
Frobenius law in physics Quantum field theory: replace particles by fields; state space varies over space-time
13 / 34 Computes manifold invariants via Pachner moves:
⇐⇒
I RT construction turns modular tensor category into 3d TQFT
“Invariants of 3-manifolds via link polynomials and quantum groups” Inventiones Mathematicae 103(3):547–597, 1991
“P. L. homeomorphic manifolds are equivalent by elementary shellings” European Journal of Combinatorics 12(2):129–145, 1991
Frobenius law in physics Quantum field theory: replace particles by fields; state space varies over space-time
I Topological quantum field theory depends only on topology 2d TQFTs ' commutative Frobenius algebras
“A geometric approach to two-dimensional conformal field theory” University of Utrecht, 1989
13 / 34 Computes manifold invariants via Pachner moves:
⇐⇒
“P. L. homeomorphic manifolds are equivalent by elementary shellings” European Journal of Combinatorics 12(2):129–145, 1991
Frobenius law in physics Quantum field theory: replace particles by fields; state space varies over space-time
I Topological quantum field theory depends only on topology 2d TQFTs ' commutative Frobenius algebras
I RT construction turns modular tensor category into 3d TQFT
“A geometric approach to two-dimensional conformal field theory” University of Utrecht, 1989
“Invariants of 3-manifolds via link polynomials and quantum groups” Inventiones Mathematicae 103(3):547–597, 1991
13 / 34 Frobenius law in physics Quantum field theory: replace particles by fields; state space varies over space-time
I Topological quantum field theory depends only on topology 2d TQFTs ' commutative Frobenius algebras
I RT construction turns modular tensor category into 3d TQFT Computes manifold invariants via Pachner moves:
⇐⇒
“A geometric approach to two-dimensional conformal field theory” University of Utrecht, 1989
“Invariants of 3-manifolds via link polynomials and quantum groups” Inventiones Mathematicae 103(3):547–597, 1991
“P. L. homeomorphic manifolds are equivalent by elementary shellings” European Journal of Combinatorics 12(2):129–145, 1991 13 / 34 Cloak
14 / 34 Equivalently: a monoid (A, µ, η) such that µ† is a homomorphism of (A, µ, η)-modules (in braided monoidal category)
It can be:
I commutative: µ ◦ β = µ (in braided monoidal category) † † I symmetric: η ◦ µ ◦ β = η ◦ µ (in braided monoidal category) † I special / strongly separable: µ ◦ µ = id † I normalizable: µ ◦ µ invertible, positive, and central
Dagger Frobenius structures: definition In a dagger monoidal category: a dagger Frobenius structure consists of an object A and maps µ: A ⊗ A → A and η : I → A satisfying
(µ ⊗ id) ◦ µ = (id ⊗ µ) ◦ µ µ ⊗ (id ⊗ η) = id = µ ⊗ (η ⊗ id) (µ ⊗ id) ◦ (id ⊗ µ†) = (id ⊗ µ) ◦ (µ† ⊗ id)
15 / 34 It can be:
I commutative: µ ◦ β = µ (in braided monoidal category) † † I symmetric: η ◦ µ ◦ β = η ◦ µ (in braided monoidal category) † I special / strongly separable: µ ◦ µ = id † I normalizable: µ ◦ µ invertible, positive, and central
Dagger Frobenius structures: definition In a dagger monoidal category: a dagger Frobenius structure consists of an object A and maps µ: A ⊗ A → A and η : I → A satisfying
(µ ⊗ id) ◦ µ = (id ⊗ µ) ◦ µ µ ⊗ (id ⊗ η) = id = µ ⊗ (η ⊗ id) (µ ⊗ id) ◦ (id ⊗ µ†) = (id ⊗ µ) ◦ (µ† ⊗ id)
Equivalently: a monoid (A, µ, η) such that µ† is a homomorphism of (A, µ, η)-modules (in braided monoidal category)
15 / 34 Dagger Frobenius structures: definition In a dagger monoidal category: a dagger Frobenius structure consists of an object A and maps µ: A ⊗ A → A and η : I → A satisfying
(µ ⊗ id) ◦ µ = (id ⊗ µ) ◦ µ µ ⊗ (id ⊗ η) = id = µ ⊗ (η ⊗ id) (µ ⊗ id) ◦ (id ⊗ µ†) = (id ⊗ µ) ◦ (µ† ⊗ id)
Equivalently: a monoid (A, µ, η) such that µ† is a homomorphism of (A, µ, η)-modules (in braided monoidal category)
It can be:
I commutative: µ ◦ β = µ (in braided monoidal category) † † I symmetric: η ◦ µ ◦ β = η ◦ µ (in braided monoidal category) † I special / strongly separable: µ ◦ µ = id † I normalizable: µ ◦ µ invertible, positive, and central
15 / 34 “Frobenius algebras and 2D topological quantum field theories” Cambridge University Press, 2003
is free symmetric monoidal category on a Frobenius algebra
(2d TQFT is just a monoidal functor (Cob, +) → (FHilb, ⊗))
Frobenius algebra example: cobordisms Category of cobordisms:
I objects are 1-dimensional compact manifolds
I arrows are 2-dimensional compact manifolds with boundary
=
16 / 34 (2d TQFT is just a monoidal functor (Cob, +) → (FHilb, ⊗))
“Frobenius algebras and 2D topological quantum field theories” Cambridge University Press, 2003
Frobenius algebra example: cobordisms Category of cobordisms:
I objects are 1-dimensional compact manifolds
I arrows are 2-dimensional compact manifolds with boundary
=
is free symmetric monoidal category on a Frobenius algebra
16 / 34 Frobenius algebra example: cobordisms Category of cobordisms:
I objects are 1-dimensional compact manifolds
I arrows are 2-dimensional compact manifolds with boundary
=
is free symmetric monoidal category on a Frobenius algebra
(2d TQFT is just a monoidal functor (Cob, +) → (FHilb, ⊗))
“Frobenius algebras and 2D topological quantum field theories” Cambridge University Press, 2003
16 / 34 I direct sums of Frobenius structures are Frobenius structures, so L all finite-dimensional C*-algebras i Mni are Frobenius
I conversely: all normalizable Frobenius structures are C* L I in particular: commutative Frobenius structures are i M1 that is, choice of orthonormal basis
“Categorical formulation of finite-dimensional quantum algebras” Communications in Mathematical Physics 304(3):765–796, 2011
“A new description of orthogonal bases” Mathematical Structures in Computer Science 23(3):555–567, 2013
Frobenius algebra example: C*-algebras In the category of finite-dimensional Hilbert spaces:
I Mn is a monoid under µ: eij ⊗ ekl 7→ δjkeil † P I µ : eij 7→ k eik ⊗ ekj satisfies Frobenius law: X eij ⊗ ekl 7→ δjk eim ⊗ eml m
17 / 34 I conversely: all normalizable Frobenius structures are C* L I in particular: commutative Frobenius structures are i M1 that is, choice of orthonormal basis
“Categorical formulation of finite-dimensional quantum algebras” Communications in Mathematical Physics 304(3):765–796, 2011
“A new description of orthogonal bases” Mathematical Structures in Computer Science 23(3):555–567, 2013
Frobenius algebra example: C*-algebras In the category of finite-dimensional Hilbert spaces:
I Mn is a monoid under µ: eij ⊗ ekl 7→ δjkeil † P I µ : eij 7→ k eik ⊗ ekj satisfies Frobenius law: X eij ⊗ ekl 7→ δjk eim ⊗ eml m
I direct sums of Frobenius structures are Frobenius structures, so L all finite-dimensional C*-algebras i Mni are Frobenius
17 / 34 L I in particular: commutative Frobenius structures are i M1 that is, choice of orthonormal basis
“A new description of orthogonal bases” Mathematical Structures in Computer Science 23(3):555–567, 2013
Frobenius algebra example: C*-algebras In the category of finite-dimensional Hilbert spaces:
I Mn is a monoid under µ: eij ⊗ ekl 7→ δjkeil † P I µ : eij 7→ k eik ⊗ ekj satisfies Frobenius law: X eij ⊗ ekl 7→ δjk eim ⊗ eml m
I direct sums of Frobenius structures are Frobenius structures, so L all finite-dimensional C*-algebras i Mni are Frobenius
I conversely: all normalizable Frobenius structures are C*
“Categorical formulation of finite-dimensional quantum algebras” Communications in Mathematical Physics 304(3):765–796, 2011
17 / 34 Frobenius algebra example: C*-algebras In the category of finite-dimensional Hilbert spaces:
I Mn is a monoid under µ: eij ⊗ ekl 7→ δjkeil † P I µ : eij 7→ k eik ⊗ ekj satisfies Frobenius law: X eij ⊗ ekl 7→ δjk eim ⊗ eml m
I direct sums of Frobenius structures are Frobenius structures, so L all finite-dimensional C*-algebras i Mni are Frobenius
I conversely: all normalizable Frobenius structures are C* L I in particular: commutative Frobenius structures are i M1 that is, choice of orthonormal basis
“Categorical formulation of finite-dimensional quantum algebras” Communications in Mathematical Physics 304(3):765–796, 2011
“A new description of orthogonal bases” Mathematical Structures in Computer Science 23(3):555–567, 2013
17 / 34 † −1 I µ = {(h, (h ◦ f, f)) | cod(h) = cod(f)} satisfies Frobenius law
I conversely: all dagger Frobenius structures are groupoids
Frobenius algebra example: groupoids
In the category of sets and relations:
I Morphism set of groupoid G is monoid under µ = {((g, f), g ◦ f) | dom(g) = cod(f)}
η = {idx | x object of G}
“Relative Frobenius algebras are groupoids” Journal of Pure and Applied Algebra 217:114–124, 2013
“Quantum and classical structures in nondeterministic computation” Quantum Interaction, LNAI 5494:143–157, 2009
18 / 34 I conversely: all dagger Frobenius structures are groupoids
Frobenius algebra example: groupoids
In the category of sets and relations:
I Morphism set of groupoid G is monoid under µ = {((g, f), g ◦ f) | dom(g) = cod(f)}
η = {idx | x object of G}
† −1 I µ = {(h, (h ◦ f, f)) | cod(h) = cod(f)} satisfies Frobenius law
“Relative Frobenius algebras are groupoids” Journal of Pure and Applied Algebra 217:114–124, 2013
“Quantum and classical structures in nondeterministic computation” Quantum Interaction, LNAI 5494:143–157, 2009
18 / 34 Frobenius algebra example: groupoids
In the category of sets and relations:
I Morphism set of groupoid G is monoid under µ = {((g, f), g ◦ f) | dom(g) = cod(f)}
η = {idx | x object of G}
† −1 I µ = {(h, (h ◦ f, f)) | cod(h) = cod(f)} satisfies Frobenius law
I conversely: all dagger Frobenius structures are groupoids
“Relative Frobenius algebras are groupoids” Journal of Pure and Applied Algebra 217:114–124, 2013
“Quantum and classical structures in nondeterministic computation” Quantum Interaction, LNAI 5494:143–157, 2009
18 / 34 Cloak hides dagger
19 / 34 B
I Morphisms f : A → B depicted as boxes f A I Composition: stack boxes vertically
I Tensor product: stack boxes horizontally
I Dagger: turn box upside-down
Coherence isomorphisms melt away
Graphical calculus “Notation which is useful in private must be given a public value and that it should be provided with a firm theoretical foundation”
“The geometry of tensor calculus I” Advances in Mathematics 88(1):55–112, 1991
20 / 34 Coherence isomorphisms melt away
Graphical calculus “Notation which is useful in private must be given a public value and that it should be provided with a firm theoretical foundation”
B
I Morphisms f : A → B depicted as boxes f A I Composition: stack boxes vertically
I Tensor product: stack boxes horizontally
I Dagger: turn box upside-down
“The geometry of tensor calculus I” Advances in Mathematics 88(1):55–112, 1991
20 / 34 Graphical calculus “Notation which is useful in private must be given a public value and that it should be provided with a firm theoretical foundation”
B
I Morphisms f : A → B depicted as boxes f A I Composition: stack boxes vertically
I Tensor product: stack boxes horizontally
I Dagger: turn box upside-down
Coherence isomorphisms melt away
“The geometry of tensor calculus I” Advances in Mathematics 88(1):55–112, 1991
20 / 34 Complete: diagrams isotopic iff equal in category of Hilbert spaces
“Finite-dimensional Hilbert spaces are complete for dagger compact categories” Logical Methods in Computer Science 8(3:6):1–12, 2012
Graphical calculus
Sound: isotopic diagrams represent equal morphisms
k k g h h = (k ⊗ id) ◦ (g ⊗ h†) ◦ f = g f f
“A survey of graphical languages for monoidal categories” New Structures for Physics, LNP 813:289–355, 2011
21 / 34 Graphical calculus
Sound: isotopic diagrams represent equal morphisms
k k g h h = (k ⊗ id) ◦ (g ⊗ h†) ◦ f = g f f
Complete: diagrams isotopic iff equal in category of Hilbert spaces
“A survey of graphical languages for monoidal categories” New Structures for Physics, LNP 813:289–355, 2011
“Finite-dimensional Hilbert spaces are complete for dagger compact categories” Logical Methods in Computer Science 8(3:6):1–12, 2012 21 / 34 Frobenius law becomes:
=
“Ordinal sums and equational doctrines” Seminar on triples and categorical homology theory, LNCS 80:141–155, 1966
“Two-dimensional topological quantum field theories and Frobenius algebras” Journal of Knot Theory and its Ramifications 5:569–587, 1996
Frobenius law graphically Instead of box, will draw for multiplication A ⊗ A → A of monoid.
= = =
22 / 34 Frobenius law graphically Instead of box, will draw for multiplication A ⊗ A → A of monoid.
= = =
Frobenius law becomes:
=
“Ordinal sums and equational doctrines” Seminar on triples and categorical homology theory, LNCS 80:141–155, 1966
“Two-dimensional topological quantum field theories and Frobenius algebras” Journal of Knot Theory and its Ramifications 5:569–587, 1996
22 / 34 In particular:
= =
Spider theorem Any connected diagram built from the components of a special ( = ) Frobenius algebra equals the following normal form:
23 / 34 Spider theorem Any connected diagram built from the components of a special ( = ) Frobenius algebra equals the following normal form:
In particular:
= =
23 / 34 Hence any Frobenius structure is self-dual
A
A∗ =
A
Dual objects
Note:
= =
24 / 34 Dual objects
Note:
= =
Hence any Frobenius structure is self-dual
A
A∗ =
A
24 / 34 † I Involution on Mn is precisely a 7→ a In category of sets and relations:
I Canonical duals: : 1 → X × X given by {(∗, (x, x)) | x ∈ X} −1 I Involution on groupoid is precisely {(g, g ) | g ∈ G}
I Decorated graphical calculus:
h ◦ g f h ◦ g f f idx f idx
−1 (h ◦ g) ◦ f f −1 g = = h ◦ (g ◦ f) = f
h g ◦ f h g ◦ f idy f idy f
Dual objects: examples In category of finite-dimensional Hilbert spaces: ∗ P ∗ I Canonical duals: : C → H ⊗ H given by 1 7→ i ei ⊗ ei
25 / 34 In category of sets and relations:
I Canonical duals: : 1 → X × X given by {(∗, (x, x)) | x ∈ X} −1 I Involution on groupoid is precisely {(g, g ) | g ∈ G}
I Decorated graphical calculus:
h ◦ g f h ◦ g f f idx f idx
−1 (h ◦ g) ◦ f f −1 g = = h ◦ (g ◦ f) = f
h g ◦ f h g ◦ f idy f idy f
Dual objects: examples In category of finite-dimensional Hilbert spaces: ∗ P ∗ I Canonical duals: : C → H ⊗ H given by 1 7→ i ei ⊗ ei
† I Involution on Mn is precisely a 7→ a
25 / 34 −1 I Involution on groupoid is precisely {(g, g ) | g ∈ G}
I Decorated graphical calculus:
h ◦ g f h ◦ g f f idx f idx
−1 (h ◦ g) ◦ f f −1 g = = h ◦ (g ◦ f) = f
h g ◦ f h g ◦ f idy f idy f
Dual objects: examples In category of finite-dimensional Hilbert spaces: ∗ P ∗ I Canonical duals: : C → H ⊗ H given by 1 7→ i ei ⊗ ei
† I Involution on Mn is precisely a 7→ a In category of sets and relations:
I Canonical duals: : 1 → X × X given by {(∗, (x, x)) | x ∈ X}
25 / 34 I Decorated graphical calculus:
h ◦ g f h ◦ g f f idx f idx
−1 (h ◦ g) ◦ f f −1 g = = h ◦ (g ◦ f) = f
h g ◦ f h g ◦ f idy f idy f
Dual objects: examples In category of finite-dimensional Hilbert spaces: ∗ P ∗ I Canonical duals: : C → H ⊗ H given by 1 7→ i ei ⊗ ei
† I Involution on Mn is precisely a 7→ a In category of sets and relations:
I Canonical duals: : 1 → X × X given by {(∗, (x, x)) | x ∈ X} −1 I Involution on groupoid is precisely {(g, g ) | g ∈ G}
25 / 34 Dual objects: examples In category of finite-dimensional Hilbert spaces: ∗ P ∗ I Canonical duals: : C → H ⊗ H given by 1 7→ i ei ⊗ ei
† I Involution on Mn is precisely a 7→ a In category of sets and relations:
I Canonical duals: : 1 → X × X given by {(∗, (x, x)) | x ∈ X} −1 I Involution on groupoid is precisely {(g, g ) | g ∈ G}
I Decorated graphical calculus:
h ◦ g f h ◦ g f f idx f idx
−1 (h ◦ g) ◦ f f −1 g = = h ◦ (g ◦ f) = f
h g ◦ f h g ◦ f idy f idy f
25 / 34 I is embedding: = =
I preserves multiplication: = =
I preserves unit: = =
“On the theory of groups as depending on the equation θn = 1” Philosophical Magazine 7(42):40–47, 1854
: one size fits all
Any monoid (A, ) embeds into A∗ ⊗ A by e :=
Pairs of pants If an object A has a dual, then A∗ ⊗ A is a monoid A∗ ⊗ A
A∗ ⊗ A := :=
A∗ ⊗ A A∗ ⊗ A
26 / 34 I is embedding: = =
I preserves multiplication: = =
I preserves unit: = =
Pairs of pants: one size fits all If an object A has a dual, then A∗ ⊗ A is a monoid A∗ ⊗ A
A∗ ⊗ A := :=
A∗ ⊗ A A∗ ⊗ A
Any monoid (A, ) embeds into A∗ ⊗ A by e :=
“On the theory of groups as depending on the equation θn = 1” Philosophical Magazine 7(42):40–47, 1854 26 / 34 I preserves multiplication: = =
I preserves unit: = =
Pairs of pants: one size fits all If an object A has a dual, then A∗ ⊗ A is a monoid A∗ ⊗ A
A∗ ⊗ A := :=
A∗ ⊗ A A∗ ⊗ A
Any monoid (A, ) embeds into A∗ ⊗ A by e :=
I is embedding: = =
“On the theory of groups as depending on the equation θn = 1” Philosophical Magazine 7(42):40–47, 1854 26 / 34 I preserves unit: = =
Pairs of pants: one size fits all If an object A has a dual, then A∗ ⊗ A is a monoid A∗ ⊗ A
A∗ ⊗ A := :=
A∗ ⊗ A A∗ ⊗ A
Any monoid (A, ) embeds into A∗ ⊗ A by e :=
I is embedding: = =
I preserves multiplication: = =
“On the theory of groups as depending on the equation θn = 1” Philosophical Magazine 7(42):40–47, 1854 26 / 34 Pairs of pants: one size fits all If an object A has a dual, then A∗ ⊗ A is a monoid A∗ ⊗ A
A∗ ⊗ A := :=
A∗ ⊗ A A∗ ⊗ A
Any monoid (A, ) embeds into A∗ ⊗ A by e :=
I is embedding: = =
I preserves multiplication: = =
I preserves unit: = =
“On the theory of groups as depending on the equation θn = 1” Philosophical Magazine 7(42):40–47, 1854 26 / 34 Theorem in a monoidal dagger category: a monoid is Frobenius ⇔ its Cayley embedding is involutive
e e = = = i
“Reversible monadic computing” MFPS 2015
“Geometry of abstraction in quantum computation” Dagstuhl Seminar Proceedings 09311, 2010
Why the Frobenius law? Maps I → A∗ ⊗ A correspond to maps A → A. So in presence of dagger, A∗ ⊗ A is involutive monoid!
∗ A A∗
i :=
A A
27 / 34 e e = = = i
Why the Frobenius law? Maps I → A∗ ⊗ A correspond to maps A → A. So in presence of dagger, A∗ ⊗ A is involutive monoid!
∗ A A∗
i :=
A A
Theorem in a monoidal dagger category: a monoid is Frobenius ⇔ its Cayley embedding is involutive
“Reversible monadic computing” MFPS 2015
“Geometry of abstraction in quantum computation” Dagstuhl Seminar Proceedings 09311, 2010 27 / 34 Why the Frobenius law? Maps I → A∗ ⊗ A correspond to maps A → A. So in presence of dagger, A∗ ⊗ A is involutive monoid!
∗ A A∗
i :=
A A
Theorem in a monoidal dagger category: a monoid is Frobenius ⇔ its Cayley embedding is involutive
e e = = = i
“Reversible monadic computing” MFPS 2015
“Geometry of abstraction in quantum computation” Dagstuhl Seminar Proceedings 09311, 2010 27 / 34 Dagger likes cloak
28 / 34 dagger
Frobenius Frobenius † Frobenius unitary
Frobenius Frobenius
Frobenius monads
I Let C be a monoidal category
I A monad is a monoid in [C, C]
I A monad T on a C is strong when equipped with a natural transformation A ⊗ T (B) → T (A ⊗ B)
I Theorem: There is an adjunction between monoids in C and strong monads on C.
A 7→ − ⊗ A T (I) ← T [
29 / 34 Frobenius monads
I Let C be a monoidal dagger category
I A Frobenius monad is a Frobenius monoid in [C, C]†
I A Frobenius monad T on a C is strong when equipped with a unitary natural transformation A ⊗ T (B) → T (A ⊗ B)
I Theorem: There is an equivalence between Frobenius monoids in C and strong Frobenius monads on C.
A 7→ − ⊗ A T (I) ← T [
29 / 34 dagger dagger Frobenius
F
Conversely, if a monad on C is Frobenius then it is of this form.
Algebras
Let C and D be categories, F : C → D and G: D → C be functors with F a G. Then GF is monad with:
Kl(GF ) D EM(GF ) a a a C
“Frobenius monads and pseudomonoids” Journal of Mathematical Physics 45:3930-3940, 2004
“Frobenius algebras and ambidextrous adjunctions” Theory and Applications of Categories 16:84–122, 2006
30 / 34 Conversely, if a monad on C is Frobenius then it is of this form.
Algebras
Let C and D be dagger categories, F : C → D and G: D → C be dagger functors with F a G. Then GF is Frobenius monad with:
Kl(GF ) D F EM(GF ) a a a C
“Frobenius monads and pseudomonoids” Journal of Mathematical Physics 45:3930-3940, 2004
“Frobenius algebras and ambidextrous adjunctions” Theory and Applications of Categories 16:84–122, 2006
30 / 34 Algebras
Let C and D be dagger categories, F : C → D and G: D → C be dagger functors with F a G. Then GF is Frobenius monad with:
Kl(GF ) D F EM(GF ) a a a C
Conversely, if a monad on C is Frobenius then it is of this form.
“Frobenius monads and pseudomonoids” Journal of Mathematical Physics 45:3930-3940, 2004
“Frobenius algebras and ambidextrous adjunctions” Theory and Applications of Categories 16:84–122, 2006
30 / 34 They form the largest subcategory of EM(T ) that inherits dagger.
Frobenius-Eilenberg-Moore algebras A Frobenius-Eilenberg-Moore algebra for a Frobenius monad T is an Eilenberg-Moore algebra (A, a) with
T (a)† T (A) T 2(A) µ† µ T 2(A) T (A) T (a)
For T = − ⊗ A, looks like:
=
31 / 34 Frobenius-Eilenberg-Moore algebras A Frobenius-Eilenberg-Moore algebra for a Frobenius monad T is an Eilenberg-Moore algebra (A, a) with
T (a)† T (A) T 2(A) µ† µ T 2(A) T (A) T (a)
For T = − ⊗ A, looks like:
=
They form the largest subcategory of EM(T ) that inherits dagger. 31 / 34 I “Cloak is everywhere”: Frobenius is omnipotent sweet spot between interesting and not general enough
I “Cloak hides dagger”: Frobenius law is coherence condition between dagger and closure
I “Dagger likes cloak”: Frobenius monads are dagger adjunctions (free) algebra categories again have dagger
Summary
What have we learnt?
I “Cloak and dagger”: Frobenius law and dagger categories both relate algebra and coalgebra
32 / 34 I “Cloak hides dagger”: Frobenius law is coherence condition between dagger and closure
I “Dagger likes cloak”: Frobenius monads are dagger adjunctions (free) algebra categories again have dagger
Summary
What have we learnt?
I “Cloak and dagger”: Frobenius law and dagger categories both relate algebra and coalgebra
I “Cloak is everywhere”: Frobenius is omnipotent sweet spot between interesting and not general enough
32 / 34 I “Dagger likes cloak”: Frobenius monads are dagger adjunctions (free) algebra categories again have dagger
Summary
What have we learnt?
I “Cloak and dagger”: Frobenius law and dagger categories both relate algebra and coalgebra
I “Cloak is everywhere”: Frobenius is omnipotent sweet spot between interesting and not general enough
I “Cloak hides dagger”: Frobenius law is coherence condition between dagger and closure
32 / 34 Summary
What have we learnt?
I “Cloak and dagger”: Frobenius law and dagger categories both relate algebra and coalgebra
I “Cloak is everywhere”: Frobenius is omnipotent sweet spot between interesting and not general enough
I “Cloak hides dagger”: Frobenius law is coherence condition between dagger and closure
I “Dagger likes cloak”: Frobenius monads are dagger adjunctions (free) algebra categories again have dagger
32 / 34 Any Kleisli morphism? No, only FEM-coalgebras A → T (A)! Consider exception monad T = − + E
“Classical and quantum structuralism” Semantic Techniques in Quantum Computation 29–69, 2009
“Handling algebraic effects” Logical Methods in Computer Science 9(4):23, 2013
Quantum measurement
n n Fix orthonormal basis on C so T = − ⊗ C is Frobenius monad on category of Hilbert spaces. Measurement is map A → T (A).
33 / 34 Consider exception monad T = − + E
“Handling algebraic effects” Logical Methods in Computer Science 9(4):23, 2013
Quantum measurement
n n Fix orthonormal basis on C so T = − ⊗ C is Frobenius monad on category of Hilbert spaces. Measurement is map A → T (A). Any Kleisli morphism? No, only FEM-coalgebras A → T (A)!
“Classical and quantum structuralism” Semantic Techniques in Quantum Computation 29–69, 2009
33 / 34 Quantum measurement
n n Fix orthonormal basis on C so T = − ⊗ C is Frobenius monad on category of Hilbert spaces. Measurement is map A → T (A). Any Kleisli morphism? No, only FEM-coalgebras A → T (A)! Consider exception monad T = − + E A η f
A + E (A, a)
I intercept exception e: execute fe, or f if no exception I handler for T specifies EM-algebra (A, a) and f : A → A I vertical arrows are Kleisli maps, dashed one EM-map
“Classical and quantum structuralism” Semantic Techniques in Quantum Computation 29–69, 2009
“Handling algebraic effects” Logical Methods in Computer Science 9(4):23, 2013 33 / 34 Consider exception monad T = − + E
Quantum measurement
n n Fix orthonormal basis on C so T = − ⊗ C is Frobenius monad on category of Hilbert spaces. Measurement is map A → T (A). Any Kleisli morphism? No, only FEM-coalgebras A → T (A)!
A η f
n A ⊗ C (A, a)
I “handle outcome x: execute fx, or f if no measurement” I handler for T specifies EM-algebra (A, a) and f : A → A I vertical arrows are Kleisli maps, dashed one EM-map
“Classical and quantum structuralism” Semantic Techniques in Quantum Computation 29–69, 2009
“Handling algebraic effects” Logical Methods in Computer Science 9(4):23, 2013 33 / 34 Consider exception monad T = − + E
Quantum measurement
n n Fix orthonormal basis on C so T = − ⊗ C is Frobenius monad on category of Hilbert spaces. Measurement is map A → T (A). Any Kleisli morphism? No, only FEM-coalgebras A → T (A)!
A η f
n A ⊗ C (A, a)
I Kleisli maps A → T (B) ‘build’ effectful computation I FEM-algebras T (B) → B are destructors ‘handling’ the effects I Effectful computation for Frobenius monad happens in FEM(T )
“Classical and quantum structuralism” Semantic Techniques in Quantum Computation 29–69, 2009
“Handling algebraic effects” Logical Methods in Computer Science 9(4):23, 2013 33 / 34 34 / 34