Actions of Hochschild cohomology

in representation theory of finite groups

Fajar Yuliawan

Fakultas Matematika dan Ilmu Pengetahuan Alam

Institut Teknologi Bandung

8 November 2018
 Outline

• Group algebras and modules

• Categories in representation of finite groups

• Actions of the Hochschild cohomology

 Group algebras

• k field, e.g Q, R, Fp = Z/pZ, Fpn , etc

• G finite group, e.g. {1}, Cp ∼
= Z/pZ, Cp × Cp , Sn , etc
 Group algebras

• k field, e.g Q, R, Fp = Z/pZ, Fpn , etc

• G finite group, e.g. {1}, Cp ∼
= Z/pZ, Cp × Cp , Sn , etc

• Group algebra kG : vector space with basis G , and

multiplication ‘induced’ by multiplication in G , i.e.
 
X 
kG = λg g | λg ∈ k
 
g ∈G
 Group algebras

• k field, e.g Q, R, Fp = Z/pZ, Fpn , etc

• G finite group, e.g. {1}, Cp ∼
= Z/pZ, Cp × Cp , Sn , etc

• Group algebra kG : vector space with basis G , and

multiplication ‘induced’ by multiplication in G , i.e.
 
X 
kG = λg g | λg ∈ k
 
g ∈G

and   !
X X XX
 λg g  µh h = λg µh gh
g ∈G h∈G g ∈G h∈G
 Group algebras
Examples
(1) G = {1}, kG = {λ · 1 | λ ∈ k} and (λ · 1)(µ · 1) = (λµ) · 1.
 Group algebras
Examples
(1) G = {1}, kG = {λ · 1 | λ ∈ k} and (λ · 1)(µ · 1) = (λµ) · 1.
Thus
kG ∼ = k, λ · 1 →
7 λ
 Group algebras
Examples
(1) G = {1}, kG = {λ · 1 | λ ∈ k} and (λ · 1)(µ · 1) = (λµ) · 1.
Thus
kG ∼ = k, λ · 1 →
7 λ
(2) G = C2 = {1, a} with a2 = 1, kG = {λ · 1 + µ · a | λ, µ ∈ k},
and

(λ1 · 1 + µ1 · a)(λ2 · 1 + µ2 · a)
=λ1 λ2 · 1 + λ1 µ2 · a + µ1 λ2 · a + µ1 µ2 · 1
=(λ1 λ2 + µ1 µ2 ) · 1 + (λ1 µ2 + λ2 µ1 ) · a.
 Group algebras
Examples
(1) G = {1}, kG = {λ · 1 | λ ∈ k} and (λ · 1)(µ · 1) = (λµ) · 1.
Thus
kG ∼ = k, λ · 1 →
7 λ
(2) G = C2 = {1, a} with a2 = 1, kG = {λ · 1 + µ · a | λ, µ ∈ k},
and

(λ1 · 1 + µ1 · a)(λ2 · 1 + µ2 · a)
=λ1 λ2 · 1 + λ1 µ2 · a + µ1 λ2 · a + µ1 µ2 · 1
=(λ1 λ2 + µ1 µ2 ) · 1 + (λ1 µ2 + λ2 µ1 ) · a.

Thus

kG ∼
= k[X ]/hX 2 − 1i, λ · 1 + µ · a 7→ λ + µX
 Group algebras

(3) G = C2 and char(k) = 2 then kG ∼

= k[X ]/hX 2 i.
 Group algebras

(3) G = C2 and char(k) = 2 then kG ∼

= k[X ]/hX 2 i.

In general, G = (Cp )n = Cp × Cp × · · · Cp and char(k) = p,

then

kG ∼ p p
= k[X1 , X2 , . . . , Xn ]/hX1 , X2 , . . . , Xnp i
 Group algebras

(3) G = C2 and char(k) = 2 then kG ∼

= k[X ]/hX 2 i.

In general, G = (Cp )n = Cp × Cp × · · · Cp and char(k) = p,

then

kG ∼ p p
= k[X1 , X2 , . . . , Xn ]/hX1 , X2 , . . . , Xnp i
Observation: kG is commutative if and only if G is Abelian, e.g.
kS3 is not commutative
 Modules over group algebras
k field, G finite group, kG group algebra
 Modules over group algebras
k field, G finite group, kG group algebra
• (left) kG -module: vector space M together with
‘multiplication’

kG × M → M, (a, m) 7→ a · m,

satisfying the ‘usual’ modules axioms, e.g. 1 · m = m,

(a + b)m = am + bm, (ab)m = a(bm),
a(λm + µn) = λam + µan, etc
 Modules over group algebras
k field, G finite group, kG group algebra
• (left) kG -module: vector space M together with
‘multiplication’

kG × M → M, (a, m) 7→ a · m,

satisfying the ‘usual’ modules axioms, e.g. 1 · m = m,

(a + b)m = am + bm, (ab)m = a(bm),
a(λm + µn) = λam + µan, etc

Example
1. The zero vector space (of dimension 0)
P  P
2. k is a kG -module with g ∈G λg · g · µ = g ∈G λg µ
3. kG is a kG -module with the usual multiplication
 Homomorphism of kG -modules

k field, G finite group, kG group algebra

M, N be kG -modules
 Homomorphism of kG -modules

k field, G finite group, kG group algebra

M, N be kG -modules
• kG -homomorphism: k-linear map f : M → N s.t.
f (a · m) = a · f (m) for all a ∈ kG , m ∈ M,
 Homomorphism of kG -modules

k field, G finite group, kG group algebra

M, N be kG -modules
• kG -homomorphism: k-linear map f : M → N s.t.
f (a · m) = a · f (m) for all a ∈ kG , m ∈ M,

or equivalently f (g · m) = g · f (m) for all g ∈ G and m ∈ M.

 Homomorphism of kG -modules

k field, G finite group, kG group algebra

M, N be kG -modules
• kG -homomorphism: k-linear map f : M → N s.t.
f (a · m) = a · f (m) for all a ∈ kG , m ∈ M,

or equivalently f (g · m) = g · f (m) for all g ∈ G and m ∈ M.

• Composition of two kG -homomorphisms is again a

kG -homomorphism, and
 Homomorphism of kG -modules

k field, G finite group, kG group algebra

M, N be kG -modules
• kG -homomorphism: k-linear map f : M → N s.t.
f (a · m) = a · f (m) for all a ∈ kG , m ∈ M,

or equivalently f (g · m) = g · f (m) for all g ∈ G and m ∈ M.

• Composition of two kG -homomorphisms is again a

kG -homomorphism, and

the class of all kG -modules together with kG -homomorphism

between them form a category, denoted by Mod(kG ).
 Definition of categories

A category C consists of
• a ‘class’ of objects Ob(C)
• a set of morphism HomC (X , Y ) for each pair X , Y ∈ Ob(C)
• a function, called composition

◦ : HomC (X , Y ) × HomC (Y , Z ) → HomC (X , Z )

(f , g ) 7→ g ◦ f

for each triple X , Y , Z ∈ Ob(C)

satisfying certain axioms.
 Definition of categories

A category C consists of
• a ‘class’ of objects Ob(C)
• a set of morphism HomC (X , Y ) for each pair X , Y ∈ Ob(C)
• a function, called composition

◦ : HomC (X , Y ) × HomC (Y , Z ) → HomC (X , Z )

(f , g ) 7→ g ◦ f

for each triple X , Y , Z ∈ Ob(C)

satisfying certain axioms.

A ‘morphism’ between categories is called a functor

Category Set
Category Mod(k)
Functor
 Category of kG -modules

• Usual objective: Classify the indecomposable in mod(kG ).

 Category of kG -modules

• Usual objective: Classify the indecomposable in mod(kG ).

Theorem [Krull-Schmidt]. Every f.d. module M in mod(kG )
can be decomposed uniquely into direct sum of
indecomposable modules,

M = M1 ⊕ M2 ⊕ · · · ⊕ Mn .
 Category of kG -modules

• Usual objective: Classify the indecomposable in mod(kG ).

Theorem [Krull-Schmidt]. Every f.d. module M in mod(kG )
can be decomposed uniquely into direct sum of
indecomposable modules,

M = M1 ⊕ M2 ⊕ · · · ⊕ Mn .

• In ‘ordinary representation theory’:

Theorem [Maschke]. If char(k) - |G |, every indecomposable
kG -module is irreducible/simple, i.e. kG is semisimple
 Category of kG -modules

• Usual objective: Classify the indecomposable in mod(kG ).

Theorem [Krull-Schmidt]. Every f.d. module M in mod(kG )
can be decomposed uniquely into direct sum of
indecomposable modules,

M = M1 ⊕ M2 ⊕ · · · ⊕ Mn .

• In ‘ordinary representation theory’:

Theorem [Maschke]. If char(k) - |G |, every indecomposable
kG -module is irreducible/simple, i.e. kG is semisimple
• In ’modular representation theory’:
If char(k) | |G |, the group algebra kG is usually of ‘wild
representation type’, hence it is impossible to classify the
indecomposables modules.
 Categories in modular representation theory

k field, G finite group and char(k) | |G |

Fact: kG is f.d. self-injective algebra, hence projective = injective
 Categories in modular representation theory

k field, G finite group and char(k) | |G |

Fact: kG is f.d. self-injective algebra, hence projective = injective
1. Stable module category StMod(kG ) or Mod(kG )
• objects: kG -modules
• morphisms: HomkG (M, N) = Hom(M, N)/ PHom(M, N),

PHom(M, N) = {f | f factors through some proj. ob.}

 Categories in modular representation theory

k field, G finite group and char(k) | |G |

Fact: kG is f.d. self-injective algebra, hence projective = injective
1. Stable module category StMod(kG ) or Mod(kG )
• objects: kG -modules
• morphisms: HomkG (M, N) = Hom(M, N)/ PHom(M, N),

PHom(M, N) = {f | f factors through some proj. ob.}

2. Derived category D(Mod kG ) of the module category

• objects: ‘complexes’ of kG -modules
• morphisms: equivalence classes of ‘roof’, i.e. M → X ← N
with N → X a ‘quasi-isomorphism’
 Categories in modular representation theory

k field, G finite group and char(k) | |G |

Fact: kG is f.d. self-injective algebra, hence projective = injective
1. Stable module category StMod(kG ) or Mod(kG )
• objects: kG -modules
• morphisms: HomkG (M, N) = Hom(M, N)/ PHom(M, N),

PHom(M, N) = {f | f factors through some proj. ob.}

2. Derived category D(Mod kG ) of the module category

• objects: ‘complexes’ of kG -modules
• morphisms: equivalence classes of ‘roof’, i.e. M → X ← N
with N → X a ‘quasi-isomorphism’
3. Homotopy category of injective kG -modules K(Inj kG )
 Homotopy category of injectives

• Objects: complexes of injective kG -modules

where I n are injective kG -modules and d n ◦ d n−1 = 0.

 Homotopy category of injectives

• Objects: complexes of injective kG -modules

where I n are injective kG -modules and d n ◦ d n−1 = 0.

• Morphism: chain maps (of degree 0) modulo null-homotopy
maps

{chain maps I → J}
HomK(Inj kG ) (I , J) =
{null-homotopic maps I → J}
Morphisms in homotopy category of injectives
Morphisms in homotopy category of injectives
Morphisms in homotopy category of injectives
Morphisms in homotopy category of injectives
 Homotopy category of injectives

Theorem (Krause 2005)

There exists a diagram of six functors

satisfying some ‘nice’ properties.

 Homotopy category of injectives

Theorem (Krause 2005)

There exists a diagram of six functors

satisfying some ‘nice’ properties.

Remark
The category K (Inj kG ) contains a copy of Mod(kG ) and two
copies of D(Mod(kG ))
 Homotopy category of injectives

• Aim: Study K(Inj kG )

• Plan: Give some algebraic structure on the set of morphism
between two objects
 Homotopy category of injectives

• Aim: Study K(Inj kG )

• Plan: Give some algebraic structure on the set of morphism
between two objects

Benson-Iyengar-Krause theory (2008):

If R is a graded-commutative notherian ring ‘acting’ on
K(Inj kG ) then there is a ‘local cohomology’ functor

Γp : K(Inj kG ) → K(Inj kG )

for each homogeneous prime ideals p ⊆ R.

 Homotopy category of injectives

• Aim: Study K(Inj kG )

• Plan: Give some algebraic structure on the set of morphism
between two objects

Benson-Iyengar-Krause theory (2008):

If R is a graded-commutative notherian ring ‘acting’ on
K(Inj kG ) then there is a ‘local cohomology’ functor

Γp : K(Inj kG ) → K(Inj kG )

for each homogeneous prime ideals p ⊆ R.

Define the support of an object X in K(Inj kG ) to be

suppR (X ) = {p ∈ Spech (R) | Γp (X ) = 0}.

 Some homological algebras

A an f.d. algebra over a field k and X , Y A-modules

• an injective resolution of X is a complex of injective modules

d0 d1
iX = · · · → 0 → I 0 −→ I 1 −→ I 2 → · · ·
together with a map η : X → I 0 such that the sequence
η d0 d1
→ I 0 −→ I 1 −→ I 2 → · · ·
0→X −
is exact.
• In particular, iX is an object in K(Inj A)
 Some homological algebras

• The n-th extension group Extn (X , Y ) is

ExtnA (X , Y ) = HomK(Inj A) (iX , iY [n]),

where iX , iY are injective resolution of X , Y , resp.

 Some homological algebras

• The n-th extension group Extn (X , Y ) is

ExtnA (X , Y ) = HomK(Inj A) (iX , iY [n]),

where iX , iY are injective resolution of X , Y , resp.

Thus, an element of ExtnA (X , Y ) looks like:
 Hochschild cohomology

A an f.d. algebra over a field k, Ae the enveloping algebra of A

(i.e. (A, A)-bimodules = Ae -modules)
 Hochschild cohomology

A an f.d. algebra over a field k, Ae the enveloping algebra of A

(i.e. (A, A)-bimodules = Ae -modules)
• M an (A, A)-bimodule. The n-th Hochschild cohomology of A
with coefficient in M is

HH(A, M) = ExtnAe (A, M) = HomK(Inj Ae ) (iA, iM[n])

 Hochschild cohomology

A an f.d. algebra over a field k, Ae the enveloping algebra of A

(i.e. (A, A)-bimodules = Ae -modules)
• M an (A, A)-bimodule. The n-th Hochschild cohomology of A
with coefficient in M is

HH(A, M) = ExtnAe (A, M) = HomK(Inj Ae ) (iA, iM[n])

• The Hochschild cohomology ring of A is the graded ring

M
HH∗ (A, A) = HHn (A, A)
n∈Z

with multiplication defined using composition and shift.

 Hochschild cohomology

• If f ∈ HHm (A, A), i.e. f : iA → iA[m]

and g ∈ HHn (A, A), i.e. g : iA → iA[n],
 Hochschild cohomology

• If f ∈ HHm (A, A), i.e. f : iA → iA[m]

and g ∈ HHn (A, A), i.e. g : iA → iA[n],

then f · g is defined as the composition

f g [m]
iA −
→ iA[m] −−−→ iA[m + n]
in HHm+n (A, A)
 Hochschild cohomology

• If f ∈ HHm (A, A), i.e. f : iA → iA[m]

and g ∈ HHn (A, A), i.e. g : iA → iA[n],

then f · g is defined as the composition

f g [m]
iA −
→ iA[m] −−−→ iA[m + n]
in HHm+n (A, A)

Theorem [Gerstenhaber 1963]. The Hochschild cohomology ring of

A is a graded-commutative ring.

Theorem [Evens 1961, Venkov 1959, Ginzburg-Kumar 1993,

Friedlander-Suslin 1997]. The Hochschild cohomology ring of A is
noetherian.
 The Action

k field and G finite group, A = kG

• Idea: Use tensor product over A
If X is an (A, A)-bimodule and M an A-module, then X ⊗A M
is an A-module.
 The Action

k field and G finite group, A = kG

• Idea: Use tensor product over A
If X is an (A, A)-bimodule and M an A-module, then X ⊗A M
is an A-module.
• Extend to tensor product on the homotopy category:

− ⊗A − : K(Mod Ae ) × K(Mod A) → K(Mod A)

 The Action

k field and G finite group, A = kG

• Idea: Use tensor product over A
If X is an (A, A)-bimodule and M an A-module, then X ⊗A M
is an A-module.
• Extend to tensor product on the homotopy category:

− ⊗A − : K(Mod Ae ) × K(Mod A) → K(Mod A)

• Restrict to the homotopy category of injectives

− ⊗A − : K(Inj Ae ) × K(Inj A) → K(Inj A)

 The Action

k field and G finite group, A = kG

• Idea: Use tensor product over A
If X is an (A, A)-bimodule and M an A-module, then X ⊗A M
is an A-module.
• Extend to tensor product on the homotopy category:

− ⊗A − : K(Mod Ae ) × K(Mod A) → K(Mod A)

• Restrict to the homotopy category of injectives

− ⊗A − : K(Inj Ae ) × K(Inj A) → K(Inj A)

 The Action
A = kG , K = K(Inj A),
M
Hom∗K (X , Y ) = Hom∗K (X , Y [n])
n∈Z
 The Action
A = kG , K = K(Inj A),
M
Hom∗K (X , Y ) = Hom∗K (X , Y [n])
n∈Z

• Apply the bifunctor to the morphisms, get:

HH∗ (A, A) × Hom∗K (X , Y ) → Hom∗K (X , Y ),

where a ‘homogeneous’ pair f : iA → iA[m] and g : X → Y [n]
is mapped to
f ⊗g
X ∼
= iA ⊗A X −−→ iA[m] ⊗A Y [n] ∼
= Y [m + n]
 The Action
A = kG , K = K(Inj A),
M
Hom∗K (X , Y ) = Hom∗K (X , Y [n])
n∈Z

• Apply the bifunctor to the morphisms, get:

HH∗ (A, A) × Hom∗K (X , Y ) → Hom∗K (X , Y ),

where a ‘homogeneous’ pair f : iA → iA[m] and g : X → Y [n]
is mapped to
f ⊗g
X ∼
= iA ⊗A X −−→ iA[m] ⊗A Y [n] ∼
= Y [m + n]

• Get (graded) HH∗ (A, A)-module structure on Hom∗K (X , Y ) for

each pair X , Y ∈ K = K(Inj A).
Main Theorem
 Conclusion and Outlook

• The Hochschild cohomology ring acts on the homotopy

category of injective kG -modules
• Since the stable category and the derived category is
‘contained’ in the homotopy category of injectives, we get also
actions on both categories (for free)
• The above action can be generalized to arbitrary f.d.
self-injective algebra over a field k
• Using B-I-K’s machinery, get local cohomology functor and
support theory for all categories above
 Conclusion and Outlook

• The Hochschild cohomology ring acts on the homotopy

category of injective kG -modules
• Since the stable category and the derived category is
‘contained’ in the homotopy category of injectives, we get also
actions on both categories (for free)
• The above action can be generalized to arbitrary f.d.
self-injective algebra over a field k
• Using B-I-K’s machinery, get local cohomology functor and
support theory for all categories above
An open problem:
• Does the above action generalize to arbitrary f.d. algebras?
Thank You!