Deformations of Algebras

Part III Essay

Abstract
The Deformation Theory of Associative Algebras was initiated by Murray Gerstenhaber in
the 1960s, to parallel Analytical Deformation Theory. In his papers, Gerstenhaber described
the rich connection of deformations of associative algebras and their Hochschild cohomology.
After a brief look at Hochschild cohomology we will define deformations of algebras as done
by Gerstenhaber and review a series of his original results. After introducing the Gertenhaber
bracket, we will look at the modern description of deformations via the Maurer-Cartan equation
and DGLAs.

Contents
1 Hochschild Cohomology 1

2 Deformations of Associative Algebras 4

3 Equivalence of Deformations 7

4 Circle Product and Gerstenhaber Bracket 10

5 DGLA Philosophy 17

6 Prelude to Deformation Quantization 23

Notation: Unless otherwise indicated, the following notations hold throughout this work:

- K will denote a field, ⊗ will denote the tensor product over field K, ⊗K .

- A will denote an associative K-Algebra, as defined in (1.1). Furthermore, a, b, c and a1 , a2 , . . . , an

will always denote elements of the algebra A in question. We will refer to the multiplication map
of algebra A by π : A ⊗ A → A and for ease of notation, we will denote π(a1 ⊗ a2 ) by a1 a2 .

- A⊗n denotes the tensor product of n copies of A: A ⊗ A ⊗ · · · ⊗ A, and if n = 0, A⊗0 = K.

- We will refer to A[[t]] as a K-vectorspace by notation A[[t]], and refer to it as an algebra by Aπ .

This notation is justified in section (2).

- From section (3) onwards, by a deformation, we mean an formal deformation.

- By a graded vector space structure on V , we refer to a decomposition V = ⊕n≥0 Vn where Vn are

vectorspaces, elements vn ∈ Vn are referred to as homogeneous elements with |vn | := n.

I
1 HOCHSCHILD COHOMOLOGY 1

1 Hochschild Cohomology
Since the deformation theory of associative algebras is tied in with Hochschild cohomology, we will
first give a brief description of this cohomology theory. Our introduction to the topic is largely based
on [22] and further detail can be found there.
Definition 1.1. For a field K, a K-vectorspace A is called an associative K-algebra or simply a
K-algebra, if we have a K-bilinear map π : A ⊗ A → A, such that

π(π(a ⊗ b) ⊗ c) = π(a ⊗ π(b ⊗ c))

for all a, b, c ∈ A.
A simple example of associative algebras is any field K regarded as K-vectorspace and its multi-
plication. Other important examples include the group ring KG for a group G, and the universal en-
veloping algebra U(g) of a Lie algebra g. Furthermore, basic notions such as algebra homomorphisms
and modules are defined as usual to respect both the vectorspace structure and the multiplication. For
a detailed account about associative algebras, one can refer to [17]. Additionally, recall that an A-
bimodule M , has both left and right A-actions which respect each other, i.e. (a1 m)a2 = a1 (ma2 )
for a1 , a2 ∈ A and m ∈ M .
To define the Hochschild cohomology of an algebra A, we must first restrict our attention to the
sequence of A-bimodules

d1
0 / HomK (K, A) / HomK (A, A) d2 / HomK (A⊗2 , A) d3 / · · · (1)

with maps dn : HomK (A⊗n−1 , A) → HomK (A⊗n , A) defined by

n−1
X
dn (f )(a1 ⊗ a2 ⊗ · · · ⊗ an ) = a1 f (a2 ⊗ · · · ⊗ an ) + (−1)i f (a1 ⊗ · · · ⊗ ai ai+1 ⊗ · · · ⊗ an )
i=1
+ (−1)n f (a1 ⊗ · · · ⊗ an−1 )an
(2)

Notice that every map in HomK (K, A) is uniquely determined by where in A the unit of K is taken
to. Hence, HomK (K, A) ∼ = A as an A-bimodule and d1 : A → HomK (A, A) is described by

(d1 (a0 )) (a1 ) = a1 a0 − a0 a1 (3)

Furthermore, observe that dn+1 dn = 0 for any n ≥ 1:

dn+1 dn (f )(a1 ⊗ a2 ⊗ · · · ⊗ an+1 ) = a1 dn (f )(a2 ⊗ · · · ⊗ an+1 )

Xn
+ (−1)i dn (f )(a1 ⊗ · · · ⊗ ai ai+1 ⊗ · · · ⊗ an+1 ) + (−1)n+1 dn (f )(a1 ⊗ · · · ⊗ an )an+1
i=1
= (a1 a2 f (a3 ⊗ · · · ⊗ an+1 )) (1 − 1) + (a1 f (a2 ⊗ · · · ⊗ an )an+1 ) (−1)n + (−1)n+1

n
!
X
i−1
+ (−1) a1 f (a2 ⊗ · · · ⊗ ai ai+1 ⊗ · · · ⊗ an+1 ) (1 − 1)
i=2
n−1
!
X
(−1)i f (a1 ⊗ · · · ⊗ ai ai+1 ⊗ · · · ⊗ an )an+1 (−1)n + (−1)n+1


+
i=1
X
+ (−1)i+j f (a2 ⊗ · · · ⊗ aj aj+1 ⊗ · · · ⊗ ai ai+1 ⊗ · · · ⊗ an+1 )
i>j
1 HOCHSCHILD COHOMOLOGY 2

X
+ (−1)i+j−1 f (a2 ⊗ · · · ⊗ ai ai+1 ⊗ · · · ⊗ aj aj+1 ⊗ · · · ⊗ an+1 )
j>i

+ (f (a1 ⊗ · · · ⊗ an−1 )an an+1 ) (−1)2n + (−1)2n+1 = 0

Hence, Im(dn ) ⊆ Ker(dn+1 ) and for those familiar with Homological language, sequence (1) is a
complex.

Definition 1.2. For K-algebra A, we define its n th Hochschild Cohomology group, HH n (A), as the
homology of complex (1) by
.
HH n (A) = H n HomK (A⊗· , A) = Ker(dn+1 ) Im(dn )

for all n ≥ 0 and d0 is taken to be zero map. Moreover, elements in Ker(dn+1 ) are called Hochschild
n -Cocycle and elements in Im(dn ) are called Hochschild n -Coboundaries. Furthermore, elements
in HomK (A⊗n , A) are referred to as Hochschild n -Cochains.

Remark 1.3. Although we have presented the complex (1) and the maps di out of thin air, they arise
rather naturally from the bar complex of A-bimodule A, and the correspondence of A-bimodules and
left Ae -modules, where Ae ∼ = A ⊗ Aop is the enveloping algebra of A. A detailed account of the
definition can be found in Chapter 1 of [22].

Even in low degrees Hochschild cohomology carries rich detail of the algebra:
At degree 0: By identity (3)

HH 0 (A) = Ker(d1 )/Im(d0 ) ∼

= Ker(d1 ) = {a0 ∈ A | a1 a0 − a0 a1 = 0}

Hence, HH 0 (A) = Z(A), the center of A.

At degree 1: Hochschild 1-cocycles correspond to K-derivations of A. If f ∈ Ker(d2 ) then

0 = d2 (f )(a1 ⊗ a2 ) = a1 f (a2 ) + (−1)f (a1 a2 ) + f (a1 )a2

⇔ f (a1 a2 ) = a1 f (a2 ) + f (a1 )a2 (4)

for any a1 , a2 ∈ A. Notice, (4) is called the Leibniz rule and any K-linear map in HomK (A, A)
which satisfies it is said to be a K-derivation of A.
At degree 2: A map f ∈ HomK (A⊗2 , A) is a Hochschild 2-cocycle if d3 (f ) = 0:

⇒ 0 = d3 (f )(a1 ⊗ a2 ⊗ a3 ) = a1 f (a2 ⊗ a3 ) − f (a1 a2 ⊗ a3 ) + f (a1 ⊗ a2 a3 ) − f (a1 ⊗ a2 )a3

which is equivalent to f satisfying

a1 f (a2 ⊗ a3 ) + f (a1 ⊗ a2 a3 ) = f (a1 a2 ⊗ a3 ) + f (a1 ⊗ a2 )a3 (5)

for any a1 , a2 , a3 ∈ A. Identity (5) will be essential to deformation theory as we will see in the next
section.
On the calculation of Hochschild cohomology: Because K is a field, and any K-vectorspace is pro-
jective as a K-module, Theorem 9.1.5 from [21] implies that the Hochschild cohomology of algebra
A can be computed via the Ext functor. In particular,

HH n (A) ∼
= ExtnAe (A, A)

where Ae = A ⊗ Aop and there exists a correspondence between A-bimodule and left Ae -modules.
Thereby, the Hochschild cohomology as defined in (1.2) is equal to the cohomology of any projective
resolution, of A-bimodules, of A and can be calculated accordingly.
1 HOCHSCHILD COHOMOLOGY 3

Example 1.4. Let A = K[]/(2 ), where Char(K) 6= 2. We can calculate its Hochschild cohomology
via the following sequence of A-bimodules:

··· / A ⊗ A τ− / A ⊗ A τ+
/ A ⊗ A τ− / A ⊗ A π /A /0 (6)

Where if l : A → A is left multiplication by  and r : A → A, right multiplication by , then

τ+ = l ⊗ IdA + IdA ⊗ r and τ− = l ⊗ IdA − IdA ⊗ r

Thereby, τ− τ+ = τ+ τ− = 0 since 2 = 0. Observe that A ⊗ A is a free A-bimodules and projective:

This is clear from the A-bimodule and Ae -module correspondence mentioned earlier and the fact that
Ae = A ⊗ A since A is commutative. Moreover, since Char(K) 6= 2:

Ker(π) = Im(τ− ) = Ker(τ+ ) = SpanK {1 ⊗  −  ⊗ 1,  ⊗ }

Im(τ+ ) = Ker(τ− ) = SpanK {1 ⊗  +  ⊗ 1,  ⊗ }

Hence, sequence (6) is exact and thereby is a projective resolution. We calculate its cohomology by
applying HomA-bi (−, A) to the truncated complex

··· / A ⊗ A τ− / A ⊗ A τ+
/ A ⊗ A τ− / A ⊗ A /0

⇓
∗
τ− ∗
τ+
0 / HomA-bi (A ⊗ A, A) / HomA-bi (A ⊗ A, A) / HomA-bi (A ⊗ A, A) / ···
(7)
Further, observe that

HomA-bi (A ⊗ A, A) = {ga |a ∈ A, where ga (1 ⊗ 1) = a} ∼

=A

since an A-bimodule map in HomA-bi (A ⊗ A, A) is determined by where 1 ⊗ 1 is taken to. By an

easy computation

τ−∗ (ga ) = ga (τ− ) = 0 τ+∗ (ga ) = ga (τ+ ) = 2ga

so that τ−∗ = 0 and we can re-write (7) as

0 /A 0 /A 2 /A 0 /A 2 /A / ··· (8)

Hence, HH0 (A) ∼ = A, HH2m−1 (A) = Ker(2) ∼ = A and HH2m (A) ∼ = A/A for m ≥ 1. A
similar result can be derived for A = K[]/(n ), for any n, and can be found in Example 1.1.16 of
[22].

To calculate the Hochschild cohomology of more complicated examples, we need a more exten-
sive Homological toolbox and notation, hence the above example will suffice and most importantly,
will tie in nicely with the deformation Theory as introduced in the next chapters. One can refer to
[22] and [18] for the calculation of the Hochschild cohomology of a large range of algebras.
2 DEFORMATIONS OF ASSOCIATIVE ALGEBRAS 4

2 Deformations of Associative Algebras

Given an associative algebra A over a field K, we can define the vectorspace of formal power series
A[[t]] = A ⊗ K[[t]]. The aim of deformation theory is to define a multiplication on A[[t]], in order to
enrich it with an associative algebra structure.
Observe that since elements α ∈ A[[t]] look like power series (i.e. α = i≥0 ai ti for some ai ∈ A),
P
a K[[t]]-bilinear map ν : A[[t]] ⊗ A[[t]] → A[[t]] is uniquely determined by how it acts on elements
of form a ⊗ b with a, b ∈ A. Moreover, for a ⊗ b with a, b ∈ A, νi (a ⊗ b) are uniquely determined in
the expansion
ν(a ⊗ b) = ν0 (a ⊗ b) + ν1 (a ⊗ b)t + ν2 (a ⊗ b)t2 + · · ·
Further, since ν is K[[t]]-bilinear, νi : A ⊗ A → A as determined by ν are also K-bilinear.
In order for the new multiplication to be a deformation of the old, we require ν0 to be the multipli-
cation of A. In other words, when t = 0, A[[t]] with the multiplication ν is just the original algebra
A.

Definition 2.1. (I) For K-algebra A, we say µ = i≥0 µi ti with µi ∈ HomK (A, A) is a one-
P
parameter deformation of A if µ0 is the multiplication is the multiplication of A. We refer to
A[[t]] with multiplication µ by Aµ .

(II) We say Aµ is an formal deformation if

µ(µ(a ⊗ b) ⊗ c) = µ(a ⊗ µ(b ⊗ c)) (9)

for a, b, c ∈ A.

If µ = π, we get the ring of formal power series A[[t]] with the multiplication of A extended to
A[[t]], t-linearly. We will refer to this algebra as Aπ , to avoid confusion between using A[[t]] as an
algebra and as a vectorspace.

Example 2.2. Let A = K[]/(2 ). For any α ∈ K, we define ρα ∈ HomK (A⊗2 , A) on the basis of
A, {1, }, by
ρα (1 ⊗ 1) = ρα (1 ⊗ ) = ρα ( ⊗ 1) = 0, ρα ( ⊗ ) = α (10)
Thus for any sequence ∆ = {αi }∞ i
P
i=1 of elements αi ∈ K, let µ∆ := π + i≥1 αi t . For any sequence
∆, µ∆ is a one-parameter deformation; however, its associativity is less clear. Take the sequence to
be non-zero only on the first term i.e. for α ∈ K consider µα := π + ρα t, then it is easy to check that
Aµα is a formal deformation: Firstly, observe that we have relations

µα (1 ⊗ 1) = 1, µα ( ⊗ 1) = µα (1 ⊗ ) = , µα ( ⊗ ) = αt (11)

We only need to check associativity on the basis {1, }, and the relations above imply that µα = π
unless both entries are . Therefore, we only need to check the following cases

• µα (µα (1 ⊗ ) ⊗ ) =µα (µα ( ⊗ 1) ⊗ ) = µα (µα ( ⊗ ) ⊗ 1) = αt

=µα (1 ⊗ µα ( ⊗ )) = µα ( ⊗ µα (1 ⊗ )) = µα ( ⊗ µα ( ⊗ 1))
• µα (µα ( ⊗ ) ⊗ ) =αt = µα ( ⊗ µα ( ⊗ ))

Hence Aµα are formal deformations and from the relations in (11) it should be clear that we have an
isomorphism of K-algebras Aµα ∼ = K[][[t]]/(2 − αt). Observe that “at t = 0”, the deformation
gives our original algebra: (Aµα )t=0 ∼
= K[]/(2 ) = A.
2 DEFORMATIONS OF ASSOCIATIVE ALGEBRAS 5

By looking at the coefficient of tn in the associativity condition (9) of µ, we see that Aµ is

associative if and only if
n
X n
X
µi (µn−i (a ⊗ b) ⊗ c) = µi (a ⊗ µn−i (b ⊗ c)) (12)
i=0 i=0

holds for all n ≥ 0. Since µ0 is the multiplication of associative algebra A, (12) clearly holds for
n = 0. Further, we can rewrite (12) as
n−1
X n−1
X
µi (µn−i (a ⊗ b) ⊗ c) − µi (a ⊗ µn−i (b ⊗ c))
i=1 i=1
(13)
=aµn (b ⊗ c) + µn (a ⊗ bc) − µn (a ⊗ b)c − µn (ab ⊗ c)

Also, the RHS of (13) is equal to the Hochschild 3-coboundary of µn , d3 µn (a, b, c), where the map
d3 : HomK (A⊗2 , A) → HomK (A⊗3 , A) is as defined in (2) in the last section.
n−1
X n−1
X
µi (µn−i (a ⊗ b) ⊗ c) − µi (a ⊗ µn−i (b ⊗ c)) = d3 µn (a, b, c) (?n )
i=1 i=1

Definition 2.3. In the expansion µ = π + i≥1 µi ti of a one-parameter deformation µ, the first

P

non-zero coefficient µn ∈ HomK (A⊗2 , A) is called the infinitesimal of µ.

Observe that if µn is the infinitesimal of an associative deformation µ, all the terms on the LHS
of (?n ) will be zero, and thereby d3 µn = 0. Hence:
Proposition 2.4. If µ is a formal deformation of algebra A, then the infinitesimal of µ is a Hochschild
2-cocycle.
Definition 2.5. For algebra A, we say Hochschild 2-cocycle f ∈ Ker(d3 ) is integrable if there exists
an formal deformation of A, µ, such that f is the infinitesimal of µ.
Example 2.6. As we saw in Example (2.2), Aµα with µα = π + ρα t is a formal deformation for any
α ∈ K. We can confirm Proposition (2.4): The expansion

d3 ρα (a ⊗ b ⊗ c) = aρα (b ⊗ c) − ρα (ab ⊗ c) + ρα (a ⊗ bc) − ρα (a ⊗ b)c

is zero for any combination of a, b, c ∈ {1, }: since if at least one of a, b, c is 1, then two of the terms
are zero and the other two cancel out. In the case that a = b = c = , then the two middle terms of
the expansion are zero and the other two terms cancel out. Therefore, ρα ∈ Ker(d3 ) and since Aµα
are formal deformations, then ρα are integrable.
Remark 2.7.P If Hochschild 2-cocycle µn is integrable, and occurs as the infinitesimal of deformation
µ = π + i≥1 µi ti , then in the can appear as the coefficient of tm for any m ≥ 1 and form a
deformation µ0 with µ0i = 0 for 1 ≤ i ≤ m − 1 and µ0m+i = µn+i for i ≥ 0. Therefore, in the
literature when dealing with integrability problems, the infinitesimal is often assumed to occur at
n = 1.
The natural question to ask is; which Hochschild 2-cocycles are integrable? Gerstenhaber at-
tacked this problem inductively: we assume that for µ1 , µ2 , . . . , µn−1 ∈ HomK (A⊗2 , A), equations
(?i : 1 ≤ i ≤ n − 1) hold. Then, we would like to find a µn ∈ HomK (A⊗2 , A) such that (?n )
holds. Of course this is only possible if and only if the LHS of (?n ) is a Hochschild 3-coboundary
and lies in Im(d3 ). Hence, from this point of view, the LHS of equation (?n ) can be thought of as
the ‘obstruction’ to our deformation problem.
2 DEFORMATIONS OF ASSOCIATIVE ALGEBRAS 6

Pn−1 i
Definition 2.8. If µ : A[[t]] → A[[t]] is defined by decomposition µ = π + i≥0 µi t , where
⊗2
µi ∈ HomK (A , A) and equations (?i : 1 ≤ i ≤ n − 1) hold, we call (A, µ) a truncated
deformation and the LHS of (?n ), the (n − 1)
1)-th obstruction.

In section (4), we will show that the (n − 1)-th obstruction of a truncated deformation is a
Hochschild 3-cocycle, but we will first need to introduce the Gerstenhaber Bracket.

Example 2.9. For ρα as defined in Example (2.2), since Im(ρα ) lies in K for any α ∈ K, and
ρα (−, −) is zero when one of the entries is from K, then for any α, β ∈ K

ρα (ρβ (a ⊗ b) ⊗ c) = 0 = ρα (a ⊗ ρβ (b ⊗ c))

holds for a, b, c ∈ {1, } and thereby for any a, b,Pc ∈ A = K[]/(2 ). Hence for any finite sequence
n−1
∆ = {αi }n−1
i=1 , the (n − 1)th obstruction of π +
i
i≥1 ραi t vanishes:

n−1
X n−1
X
ραi (ραn−i (a ⊗ b) ⊗ c) − ραi (a ⊗ ραn−i (b ⊗ c)) = 0 − 0 = 0
i=1 i=1

Moreover, by Example (2.6), we know d3 ρα = 0 for any α ∈ K. Hence, we can extend the truncated
deformation via any ρα i.e. for any α ∈ K, equation
n−1
X n−1
X
ραi (ραn−i (a ⊗ b) ⊗ c) − ραi (a ⊗ ραn−i (b ⊗ c)) = d3 ρα
i=1 i=1

holds. So for any sequence ∆ = {αi }∞ i

P
i=1 of elements αi ∈ K, µ∆ = π + i≥1 αi t defines a formal
2
deformation of A = K[]/( ).

Remark 2.10. Much of the vocabulary of the theory such as integrability, obstructions and infinites-
imals were chosen by Gerstenhaber to parallel the Analytical Theory of Deformations. Notice that if
a one-parameter deformation µ is formal, it defines an associative algebra structure on vectorspace
A[[t]], where t is an indeterminate variable or ‘parameter’. Given this structure, we could then let
t be any α element of A and define a new associative multiplication on A; however, the issue which
i
P
arises is the convergence of an expression i≥0 ai α in A, which is the difficulty of the analytical
theory. This is why Gerstenhaber chooses to work over the ring of formal power series. In fact, in [6]
Gerstenhaber works over K((t)), while we have P chosen to work over K[[t]]. Additionally, it should
be clear that given a truncated deformation (A, n−1 i
i≥0 µi t ), we can define an associative algebra
n
structure on the vectorspace A ⊗ K[t]/(t ). Moreover, in this deformation, parameter t can take the
value of any element of A with order n, since we are taking finite sums and do not need to worry about
convergence. We discuss the generalisation of deformations to A ⊗ R, where K-algebra R satisfies
certain conditions in section (5).

Example 2.11. The deformations Aµα from Example (2.2) can also be considered as deformations
over K[t]/(t2 ) and over K[t]/(t2 ), Aµα ∼
= K[, t]/(2 − αt, t2 ). As mentioned in the Remark above,
we can replace parameter t by any element of order 2 in K. If K = R, α = 1 and t = −1, then

Aµ1 ∼
= K[, t]/(2 − t, t2 ) ⇒ (Aµ1 )t=−1 ∼
= R[]/(2 + 1) ∼
=C

Notation: Since we are only concerned with formal deformations from this point on, we will
simply write ‘a deformation Aµ ’, instead of a formal deformation.
3 EQUIVALENCE OF DEFORMATIONS 7

3 Equivalence of Deformations
Since a deformation of K-algebra A induces an associative algebra structure on vectorspace A[[t]],
the next natural question to ask is: When do two formal deformation induce isomorphic algebras?
This question is resolved by defining the notion of equivalence between deformations.
Definition 3.1. We say one-parameter deformations Aµ and Aλ are equivalent if there exists a K[[t]]-
linear map φ : A[[t]] → A[[t]] such that
X
φ = φ0 + φ1 t + φ2 t2 + · · · = φi t i
i≥0

where φi ∈ HomK (A, A) and φ0 = IdA , and

φ(µ(a ⊗ b)) = λ(φ(a) ⊗ φ(b)) (14)

or equivalently, diagram
µ
A[[t]] ⊗ A[[t]] / A[[t]] (15)
φ⊗φ φ
 
A[[t]] ⊗ A[[t]]
λ / A[[t]]

commutes.
Remark 3.2. It is known that an element α = i≥0 ri xi is invertible in the ring of formal power
P
series R[[x]] if and only if r0 is invertible. The inverseof α can be constructed inductively; β =
P i −1 −1 Pn−1
i≥0 bi x , where b0 = r0 and bn = −r0 i=0 ri bi , so that αβ = 1.

P Therefore, by the above Remark, any linear map φ ∈ HomK (A, A)[[t]] of form φ = IdA +
i
i≥1 φi t is invertible and a K-vectorspace isomorphism of A[[t]]. Moreover, its inverse has the form
φ0 = IdA + i≥1 φ0i ti and equivalence, as defined above, is symmetric and indeed an equivalence
P
relation, since reflexiveness and transitivity clearly hold. Furthermore, if Aµ and Aλ are formal
deformations and equivalent via a map φ, then by (14), φ is also K-algebra isomorphism.
Definition 3.3. We say deformation Aµ is trivial, if it is equivalent to Aπ , and we call algebra A
rigid if all its formal deformations are trivial.
Given an equivalence of deformations φ : Aµ → Aλ , consider the coefficient of tn in (14):
X X
φi (µj (a ⊗ b)) = λi (φj (a) ⊗ φk (b)) (16)
i+j=n i+j+k=n

where φ0 is the identity map on A. Since µ0 and λ0 are the multiplication of A, then we can rewrite
the above equation as
X X
φi (µj (a ⊗ b)) − λi (φj (a) ⊗ φk (b)) = aφn (b) + φn (a)b − φn (ab) (17)
i+j=n i+j+k=n
i6=0 j,k6=n

Notice that the RHS of (17) is the Hochschild coboundary of φn ∈ HomK (A, A):
X X
φi (µj (a ⊗ b)) − λi (φj (a) ⊗ φk (b)) = d2 φn (a ⊗ b) (??n )
i+j=n i+j+k=n
i6=0 j,k6=n

Hence, (??1 ) holding implies the following result:

3 EQUIVALENCE OF DEFORMATIONS 8

Proposition 3.4. If deformations Aµ and Aλ are equivalent via a map φ as above, then µ1 − λ1 =
d2 φ1 .
Corollary 3.5. If deformation Aµ is trivial, then µ1 is a Hochschild 2-coboundary.
Proof. Let φ : Aµ → Aπ be an equivalence. By looking at the last Proposition, λ1 = 0 and
µ1 = d2 φ1 ∈ Im(d2 ).

Remark 3.6. Recall from Remark (2.10) that a deformation over K[t]/(t2 ) has form µ = π + µ1 t.
Further by Proposition (2.4) and (?1 ) holding for µ, we know that µ1 ∈ Ker(d3 ). On the other
hand, since t2 = 0, then Proposition (3.4) implies that two deformations µ and λ over K[t]/(t2 ) are
equivalent if and only if µ1 − λ1 ∈ Im(d2 ). Hence, we have the bijection
Formal deformations ofA over K[t]/(t2 ) upto equivalence ←→ HH 2 (A)

(18)
Example 3.7. Let A = K[]/(2 ) and Char(K) 6= 2. If ϕ ∈ HomK (A, A), then
d2 ϕ(1 ⊗ 1) =1ϕ(1) − ϕ(1) + ϕ(1)1 = ϕ(1)
d2 ϕ(1 ⊗ ) =1ϕ() − ϕ() + ϕ(1) = ϕ(1) = d2 ϕ( ⊗ 1)
d2 ϕ( ⊗ ) =ϕ() − ϕ(0) + ϕ() = 2ϕ()
Recall that deformations µα = π + ρα t for α ∈ K are formal. Moreover, we know from Example
(2.6) that ρα are 2-cocycles for any α ∈ K and by the definition of ρα , we can deduce that if α 6= 0,
then
ρα ( ⊗ ) = α ∈
/ A
However, by the above calculation for any ϕ ∈ HomK (A, A), d2 ϕ( ⊗ ) ∈ A. Therefore, if α 6= 0,
then ρα is not a coboundary. By Corollary (3.5), this implies that deformations Aµα are nontrivial
if α 6= 0. Furthermore, since ρα − ρβ = ρα−β is not a coboundary if α 6= β, then {ρα | α ∈ K}
represent distinct elements in HH 2 (A). Additionally, by Proposition (3.4), {Aµα | α ∈ K} must be a
set of two by two non-equivalent deformations of A. Also, recall from Example (1.4) that HH 2 (A) ∼
=
K. Hence,
HH 2 (A) = {ρα + Im(d2 ) | α ∈ K}
Also as mentioned in Example (2.11), Aµα can be considered as deformations of A over K[t]/(t2 ).
In this case, Remark (3.6) implies that {Aµα | α ∈ K} are all the deformations of A over K[t]/(t2 ),
upto equivalence.
Corollary (3.5) does not imply that the infinitesimal of a trivial deformation is a coboundary,
since the infinitesimal might appear as the coefficient of tn for some n bigger than 1. However, we
will show in the next theorem that a non-trivial deformation, must be equivalent to a deformation
whose infinitesimal is not a coboundary. But firstPwe must observe that if Aµ is a deformation and
φt ∈ HomK (A, A)[[t]] has the form φt = IdA + i≥1 φi ti , then φt µ(φ−1 −1
t ⊗ φt ) is a deformation:

• Clearly since µ0 is the multiplication of A and φ−1 0 i

P
t has form IdA + i≥1 φi t by Remark (3.2),
then φt µ(φ−1 −1 0 i
P
t ⊗ φt ) has form π + i≥1 µi t and thereby is a one-parameter deformation.

• Secondly, µ0 = φt µ(φ−1 −1
t ⊗ φt ) is associative:

µ0 (µ0 (a ⊗ b) ⊗ c) =µ0 φt µ φ−1 −1

t (a) ⊗ φt (b) ⊗ c
=φt µ φ−1 −1 −1 −1



t φt µ φt (a) ⊗ φt (b) ⊗ φt (c)
=φt µ µ φ−1 −1 −1



t (a) ⊗ φt (b) ⊗ φt (c) (19)
=φt µ φ−1 −1 −1



t (a) ⊗ µ φt (b) ⊗ φt (c)
=φt µ φ−1 −1 −1 −1
= µ0 (a ⊗ µ0 (b ⊗ c))



t (a) ⊗ φt φt µ φt (b) ⊗ φt (c)
3 EQUIVALENCE OF DEFORMATIONS 9

Theorem 3.8. If deformation Aµ is a non-trivial deformation, then it is equivalent to a deformation

Aλ such that the infinitesimal of λ, λn ∈ Ker(d3 ), is non-vanishing in HH 2 (A, A) i.e. λn is not a
coboundary.

Proof. By Proposition (2.4), the infinitesimal of any deformation is a 2-cocycle. Let Aµ be a non-
trivial deformation. If its infinitesimal µn is a 2-coboundary, then µn ∈ Im(d2 ) and µn = d2 f0 for
some f0 ∈ HomK (A, A).
Let φ0 : A[[t]] → A[[t]] be defined by φ0 = IdA + f0 tn . By Remark (3.2), φ0 is invertible and by
(19), µ0 = φ0 µ(φ−1 −1
0 ⊗ φ0 ) is an equivalent deformation.
Claim. If µ0m is the infinitesimal of µ0 , then m > n.
The claim, if proved, directly implies the theorem: since either the infinitesimal of µ0 is not a cobound-
ary and the theorem holds, or we can repeat the process, to get an equivalent deformation. Assume that
by repeating the process, we always get a coboundary as the infinitesimal. Then we get a sequence of
(i)
equivalent deformations {µ(i) }i≥0 where the infinitesimals of µ(i) , µmi , are all 2-coboundaries with
(i)
µmi = d2 fi , where µ(i+1) = φi µ(i) (φ−1 −1
i ⊗ φi ) for φi = IdA + fi t
mi and µ = µ. Since the claim
0
implies that {mi }i≥0 is a strictly increasing sequence, then
X
Φ := IdA + f0 tn + f1 tm1 + · · · = IdA + fi tmi
i≥0

is well-defined in HomK (A, A)[[t]]. Since Φ can be seen as the composition of all φi , the claim also
implies that the infinitesimal of λ := Φµ(Φ−1 ⊗ Φ−1 ) must be the coefficient of a tM , where M is
larger than all mi . Because {mi }i≥0 is strictly increasing, then the coefficient of all ti must be zero.
So λ = π and Aµ is a trivial deformation, which is a contradiction.
Proof of claim: Observe that by Remark (3.2) φ0 = IdA + f0 tn is invertible and

φ−1 n −1 n+1
0 = IdA − f0 t + φn+1 t + ···

for some φ−1

n+i ∈ HomK (A, A), where i ≥ 1. So

µ0 (a ⊗ b) =φ0 µ(φ−1 −1 n n
0 (a) ⊗ φ0 (b)) = φ0 µ ((a − f0 (a)t + · · · ) ⊗ (b − f0 (b)t + · · · ))
=φ0 µ (a ⊗ b − (f0 (a) ⊗ b)tn − (a ⊗ f0 (b))tn + · · · )
=φ0 (ab − f0 (a)btn − af0 (b)tn + µn (a ⊗ b)tn + · · · )
=ab − f0 (a)btn − af0 (b)tn + µn (a ⊗ b)tn + f0 (ab)tn + · · ·
=ab + (f0 (ab) − f0 (a)b − af0 (b) + d2 f0 (a ⊗ b)) tn + · · ·

where we’ve omitted the powers tr of t with r > n. By the definition of d2 , the coefficient of tn in
the above expression is zero. Therefore, the infinitesimal of µ0 occurs at µ0m with m > n.

Corollary 3.9. If HH 2 (A) = 0, then A is rigid.

Let g be a finite dimensional semisimple Lie algebra, then HH 2 (U(g)) = 0, where U(g) is the
universal enveloping algebra of g and by Corollary (3.9), U(g) is rigid. For a detailed proof of why
HH 2 (U(g)) = 0 refer to Exercise 2.8.1(c) in [20]. Another example of rigid algebras are separable
algebras: when K is a field, algebra A is said to be separable if it is a finite product of simple algebras
whose centres are separable field extensions of K. In Section 14 of [4], six equivalent definitions of
separable algebras are presented, one being the absolute projectivity of A as an A-bimodule, which
implies HH n (A) = 0 for n > 0. Another equivalent definition requires the existence of an idem-
potent in Ae satisfying certain properties. In Example 2.6 of [18], the matrix ring Mn (K) and group
ring KG, where the order of G is invertible in K, are shown to be separable by finding this idempo-
tent. The importance of separable algebras in deformation theory is due to the main result of section
4 CIRCLE PRODUCT AND GERSTENHABER BRACKET 10

14 of [4], which states that if S is a separable subalgebra of A, then for any deformation of A, we
can find an equivalent deformation µ such that S is fixed, i.e for s ∈ S and a ∈ A, µ(s, a) = sa
and µ(a, s) = as. As a consequence of this Theorem, if algebra A is unital and 1 is its unit, then
K = K1 is a separable subalgebra of A, and for any deformation, we can find an equivalent defor-
mation such that element 1 acts as a unit with respect to the new multiplication. The preservation of
units by deformation was actually proved first in Theorem 17 of [7] and again implied by the relative
preservation of general idempotents described in section 20 of [9]. Other results not covered here
are Gerstenhaber and Schack’s work in [9] and [4] on diagrams of algebras and their deformations,
where the mentioned result on separable subalgebras translates to the deformation of monomorphism
S ,→ A. Furthermore, another result described in Theorem 17 of [7] is the preservation of invertabil-
ity of elements by deformations over K((t)) and that deformations of division rings are themselves
division rings. Further works of Gerstenhaber on division rings can be found in [7] and [8].

Remark 3.10. Although U(g) is rigid as an algebra, it is also enriched with a coalgebra structure
which can have non-trivial deformations1 . Descriptions of the coalgebra structure of U(g) can be
found in [12]. Specifically, section XVI.5 describes Quantum Enveloping Algebras which are defor-
mations of U(g) as a bialgebra. By the above, these deformations will be trivial with respect to its
algebra structure.

4 Circle Product and Gerstenhaber Bracket

In [5], Gerstenhaber defines a Graded Lie structure on C∗ (A) := ⊕n≥0 HomK (A⊗n , A), by what
is now commonly called the Gerstenhaber Bracket. First we need to introduce the circle product as
defined in the same paper2 .

Definition 4.1. For f ∈ HomK (A⊗m , A) and g ∈ HomK (A⊗n , A), we define

(I) the circle product f ◦ g ∈ HomK (A⊗m+n−1 , A) by

(f ◦ g)(a1 ⊗ · · · ⊗ am+n−1 ) = (20)

X m
(−1)(n−1)(i−1) f (a1 ⊗ · · · ⊗ g(ai ⊗ · · · ⊗ ai+n−1 ) ⊗ · · · ⊗ am+n−1 )
i=1

(II) and the cup product f ^ g ∈ HomK (A⊗m+n , A) by

(f ^ g)(a1 ⊗ · · · ⊗ am+n ) = f (a1 ⊗ · · · ⊗ am )g(am+1 ⊗ · · · ⊗ am+n ) (21)

Recall that in section (1), we defined maps dn : HomK (A⊗(n−1) , A) → HomK (A⊗n , A) such
that dn+1 dn = 0. This construction corresponds to a map d∗ : C∗ (A) → C∗ (A), where for a
homogeneous element a ∈ C∗ (A) = ⊕n≥0 HomK (A⊗n , A) with |a| = n, d∗ (a) = dn (a) holds and
d∗ is extended linearly to all of C∗ (A).

Lemma 4.2. For f ∈ HomK (A⊗m , A) and g ∈ HomK (A⊗n , A) ,

d∗ (f ◦ g) = (−1)(n−1) d∗ f ◦ g + f ◦ d∗ g + (−1)mn+n−1 f ^ g + (−1)n g ^ f (22)

holds.
1
Refer to Remark (4.5), where we briefly describe how analogous deformation theories can be defined for coalgebras
and bialgebras.
2
As Gerstenhaber himself remarks in [5], the paper was originally intended to be a part of his paper on deformation
theory.
4 CIRCLE PRODUCT AND GERSTENHABER BRACKET 11

Proof. For a1 ⊗ a2 ⊗ · · · ⊗ am+n ∈ A⊗(m+n) , the LHS of (22) expands as bellow:

[d∗ (f ◦ g)](a1 ⊗ · · · ⊗ am+n ) =a1 (f ◦ g)(a2 ⊗ · · · ⊗ am+n )

m+n−1
X
+ (−1)i (f ◦ g)(a1 ⊗ · · · ⊗ ai ai+1 ⊗ · · · ⊗ am+n )
i=1
+ (−1)m+n (f ◦ g)(a1 ⊗ · · · ⊗ am+n−1 )am+n

Since (n − 1)i = (n − 1)(i − 1) + (n − 1), the RHS of (22) expands as bellow:

(−1)(n−1) [d∗ f ◦ g](a1 ⊗ · · · ⊗ am+n ) =

m+1
X
(−1)(n−1)i d∗ f (a1 ⊗ · · · ⊗ g(ai ⊗ . . . ai+n−1 ) ⊗ · · · ⊗ am+n )
i=1
m+1
X
= (−1)(n−1)i a1 f (a2 ⊗ · · · ⊗ g(ai ⊗ . . . ai+n−1 ) ⊗ · · · ⊗ am+n )
i=2
+ (−1)n−1 g(a1 ⊗ · · · ⊗ an )f (an+1 ⊗ · · · ⊗ am+n ) [= −(−1)n g ^ f ]
Xm
+ (−1)(n−1)i+m+1 f (a1 ⊗ · · · ⊗ g(ai ⊗ . . . ai+n−1 ) ⊗ · · · ⊗ am+n−1 ) am+n
i=1
+ (−1)(n−1)(m+1)+m+1 f (a1 ⊗ · · · ⊗ am )g(am+1 ⊗ · · · ⊗ am+n ) = −(−1)mn+n−1 f ^ g
 

m+1
X
+ (−1)(n−1)i+i−1 f (a1 ⊗ · · · ⊗ ai−1 g(ai ⊗ . . . ai+n−1 ) ⊗ · · · ⊗ am+n )
i=2
m
X
+ (−1)(n−1)i+i f (a1 ⊗ · · · ⊗ g(ai ⊗ . . . ai+n−1 )ai+n ⊗ · · · ⊗ am+n )
i=1
m+1
X X
+ (−1)(n−1)i+j f (a1 ⊗ · · · ⊗ aj aj+1 ⊗ · · · ⊗ g(ai ⊗ . . . ai+n−1 ) ⊗ · · · ⊗ am+n )
i=1 j<i−1
or j>i+n−1
m
X
[f ◦d∗ g](a1 ⊗ · · · ⊗ am+n ) = (−1)n(i−1) f (a1 ⊗ · · · ⊗ d∗ g(ai ⊗ . . . ai+n ) ⊗ · · · ⊗ am+n )
i=1
m
X
= (−1)n(i−1) f (a1 ⊗ · · · ⊗ ai g(ai+1 ⊗ . . . ai+n ) ⊗ · · · ⊗ am+n )
i=1
m
X
+ (−1)n(i−1)+(n+1) f (a1 ⊗ · · · ⊗ g(ai+1 ⊗ . . . ai+n−1 )ai+n ⊗ · · · ⊗ am+n )
i=1
Xm Xn
+ (−1)n(i−1)+j f (a1 ⊗ · · · ⊗ g(ai ⊗ · · · ⊗ ai+j−1 ai+j ⊗ · · · ⊗ ai+n ) ⊗ · · · ⊗ am+n )
i=1 j=1

Hence, we see that terms of form f (· · · ai g() · · · ) and f (· · · g()ai+n · · · ) in the RHS cancel out and
by the definition of the circle product, after the addition of (−1)mn+n−1 f ^ g + (−1)n g ^ f , the
remaining terms on the RHS are equal to the expansion of the LHS.

Now we touch on the important relation between associativity and the circle product. Let f, g ∈
HomK (A⊗2 , A), then

f ◦ g(a ⊗ b ⊗ c) = f (g(a ⊗ b) ⊗ c) − f (a ⊗ g(b ⊗ c)) (23)

4 CIRCLE PRODUCT AND GERSTENHABER BRACKET 12

Hence, for µi ∈ HomK (A⊗2 , A), where 0 ≤ i ≤ n − 1,

n−1
X n−1
X n−1
X
µi (µn−i (a ⊗ b) ⊗ c) − µi (a ⊗ µn−i (b ⊗ c)) = µi ◦ µn−i (a ⊗ b ⊗ c) (∗)
i=1 i=1 i=1

Recall that the LHS of the equation above is the (n-1)-th obstruction for a truncated obstruction. This
identity will help us show these obstructions are 3-cocycles:
Theorem 4.3. If (A, µ) with µ = n−1 i
P
i≥0 µi t is a truncated deformation, as defined in (2.8), then the
(n-1)th obstruction
n−1
X n−1
X
On−1 (a ⊗ b ⊗ c) := µi (µn−i (a ⊗ b) ⊗ c) − µi (a ⊗ µn−i (b ⊗ c))
i=1 i=1

is a Hochschild 3-cocycle, i.e. d4 On−1 = 0.

Proof. Since (A, µ) is a truncated deformation, then (?k : 1 ≤ k ≤ n − 1) hold and
k−1
X k−1
X
µi (µk−i (a ⊗ b) ⊗ c) − µi (a ⊗ µk−i (b ⊗ c)) = d3 µk (a ⊗ b ⊗ c)
i=1 i=1

for 1 ≤ k ≤ n − 1. Then by Lemma (22)

n−1
! n−1
X X
(−1)2−1 d∗ µi ◦ µn−i + µi ◦ d∗ µn−i


d∗ µi ◦ µn−i =
i=1 i=1
n−1
X
(−1)4+2−1 µi ^ µn−i + (−1)2 µn−i ^ µi


+
i=1
n−1
XX i−1 n−1
X n−i−1
X
=− µj ◦ µi−j ◦ µn−i + µi ◦ µj ◦ µn−i−j
i=1 j=1 i=1 j=1
n−1
X n−1
X
− µi ^ µn−i + µn−i ^ µi
i=1 i=1
X X
=− µ i ◦ µj ◦ µk + µi ◦ µj ◦ µk + 0 = 0
i+j+k=n i+j+k=n
i,j,k≥1 i,j,k≥1
Pk−1
By equation (∗), On−1 = i=1 µi ◦ µk−i . Thereby, d∗ On−1 = 0 and the (n-1)-th obstruction is a
Hochschild 3-cocycle.

Corollary 4.4. For algebra A, if HH 3 (A) = 0 then any Hochschild 2-cocycle is integrable.
Proof. Since by the above Theorem, any obstruction to extending a truncated deformation with the 2-
cocycle as its infinitesimal, is a 3-cocycle and if HH 3 (A) = 0, the obstruction must be 3-coboundary,
which is what equation (?n ) required.

Remark 4.5. In Gerstenhaber’s papers on deformations of associative algebras, he comments that a

similair theory can be applied to Lie algebras. In section 1 of [6] he demonstrates that the infinitesi-
mal of a Lie algebra deformation must be a 2-cocycle in the respective cohomology theory. Indeed, in
[10] and [11], T. Fox describes a deformation theory on T-algebras for any Monad/Triple (T, η, ν)
on the category of K-vectorspaces, using Triple cohomology. Furthermore, in section 8 of [11], even
the circle product is described for a general triple. Hence, any result up to now holds for general
triples on K-vectorspaces with their respective cohomology theories.
4 CIRCLE PRODUCT AND GERSTENHABER BRACKET 13

Example 4.6. We now have the sufficient tools to look at a more sophisticated example; A = K[x, y].
Over field K, A has basis B = {xα y β |α, β ≥ 0}. We can define the bilinear map ψ : A ⊗ A → A on
the basis via   
∂a ∂b
ψ(a ⊗ b) = (24)
∂x ∂y
∂ ∂
where a, b ∈ B and ∂x , ∂y are partial derivations in the usual sense. We can easily check that
d3 ψ = 0 on the basis and thereby on all of A: If a, b, c ∈ B

d3 ψ(a ⊗ b ⊗ c) =aψ(b ⊗ c) − ψ(ab ⊗ c) + ψ(a ⊗ bc) − ψ(a ⊗ b)c

∂b ∂c ∂ab ∂c ∂a ∂bc ∂a ∂b
=a − + − c
∂x ∂y ∂x ∂y ∂x ∂y ∂x ∂y
∂b ∂c ∂b ∂c ∂a ∂c ∂a ∂c ∂a ∂b ∂a ∂b
=a −a −b + b+ c− c=0
∂x ∂y ∂x ∂y ∂x ∂y ∂x ∂y ∂x ∂y ∂x ∂y

Hence, ψ ∈ Ker(d3 ). Moreover, ψ is integrable and can occur as an infinitesimal. This is because
HH 3 (A) = 0: In fact for a polynomial algebra K[x1 , x2 , . . . , xm ],

HH n (K[x1 , x2 , . . . , xm ]) = K[x1 , x2 , . . . , xm ] ⊗ n (SpanK {y1 , y2 , . . . , ym })

V
(25)

where n V denotes the nth exterior power of V V= SpanK {y1 , y2 , . . . , ym }. Therefore, when m = 2
V
variables are at work, the 3rd exterior power 3 (SpanK {y1 , y2 }) has a basis of elements yi1 ∧
yVi2 ∧ yi3 , where ij ∈ {1, 2} and at least two coincide, giving yi1 ∧ yi2 ∧ yi3 = 0. Consequently,
3 (Span {y , y }) = 0 and HH 3 (A) = 0. Result (25) is a specific case of the celebrated
K 1 2
Hochschild-Kostant-Rosenberg Theorem [Theorem 9.4.7 [21]] and a detailed proof of the presented
version can be found in Example 2.1.3 of [22].
Since HH 3 (A) = 0, by Corollary (4.4), any 2-cocycle is integrable and thereby the 1st obstruction to
extending µ = π + ψt vanishes i.e. there exists a bilinear map µ2 ∈ HomK (A⊗2 , A) which satisfies

d3 µ2 (a ⊗ b ⊗ c) =O1 (a ⊗ b ⊗ c) = ψ ◦ ψ(a ⊗ b ⊗ c)
  
∂a ∂b ∂b ∂c
=ψ(ψ(a ⊗ b) ⊗ c) − ψ(a ⊗ ψ(b ⊗ c)) = ψ ⊗c −ψ a⊗
∂x ∂y ∂x ∂y
2
∂ a ∂b ∂c 2
∂a ∂ b ∂c 2
∂a ∂ b ∂c 2
∂a ∂b ∂ c
= 2 + − −
∂x ∂y ∂y ∂x ∂x∂y ∂y ∂x ∂y∂x ∂y ∂x ∂x ∂y 2
∂ 2 a ∂b ∂c ∂a ∂b ∂ 2 c
= 2 −
∂x ∂y ∂y ∂x ∂x ∂y 2

Observe that unlike Example (2.9) where A was K[]/(2 ) and the obstruction to extending π + ρα
was zero and trivially solvable, here the existence of µ2 satisfying the above equation is non-trivial.
On the other hand, similarly to the case of ρα , all non-vanishing 2-cocycles are of form ψκ where
κ ∈ A = K[x, y] and   
∂a ∂b
ψκ (a ⊗ b) = κ (26)
∂x ∂y
One can show that for polynomial algebras, all 2-coboundaries are those 2-cocycles which are
symmetric. Starting with any symmetric bilinear map in Ker(d3 ), one can construct a map in
HomK (A, A) so that its image under d2 is the bilinear map. For the details of this method refer
to Theorem 3.1 in [16]. Furthermore, observe that when κ 6= 0, ψκ is not symmetric:

ψκ (x ⊗ y) = κ 6= 0 = ψκ (y ⊗ x)
4 CIRCLE PRODUCT AND GERSTENHABER BRACKET 14

So ψκ ∈
/ Im(d2 ) and additionally since
Vψ κ1 − ψκ2 = ψκ1 −κ2 , then ψκ + Im(d2 ) represent distinct
2
elements of HH (A). Further, since 2 (SpanK {y1 , y2 }) = SpanK {y1 ∧ y2 } ∼
= K, then by (25),
HH (A) ∼
2
= K[x, y]. Hence,

HH 2 (K[x, y]) = {ψκ + Im(d2 )|κ ∈ K[x, y]} ∼

= K[x, y]

Therefore, for each κ ∈ K[x, y], we have a non-trivial formal deformation µκ = π + i≥1 µ(κ,i) ti
P
such that µκ,1 = ψκ . Furthermore, by Proposition (3.4) these deformations are pairwise non-
equivalent since ψκ1 − ψκ2 = ψκ1 −κ2 ∈ / Im(d2 ) if κ1 6= κ2 . It is essential to notice that since
ψκ is not symmetric when κ 6= 0, then algebra Aµκ is a non-commutative. Deformations of this type
feed into the subject of Non-commutative Geometry: it is well known that K[x, y] can be viewed as
the algebra of polynomial functions on the affine 2-space K2 . With this idea in mind, one can assume
that the deformed algebra must be the algebra of functions on a non-commutative geometry. More
detail on this philosophy can be found in [20]. In order for the geometric aspects of the algebra to
be preserved after deformation, more constraints are needed. In fact, the example of ψ as presented
here arises naturally when looking at the Poisson structure on K[x, y]. We will discuss this side of the
theory briefly in the last section.

In [5], Gerstenhaber shows that a pre-Lie algebra, a vectorspace with a circle product satisfying
certain properties, give rise to a Lie bracket and Lie structure on the vectorspace as follows:

Definition 4.7. For f ∈ HomK (A⊗m , A) and g ∈ HomK (A⊗n , A), we define the Gerstenhaber
bracket [f, g] ∈ HomK (A⊗m+n−1 , A) as

[f, g] = f ◦ g − (−1)(m−1)(n−1) g ◦ f (27)

Lemma 4.8. For f ∈ HomK (A⊗m , A), g ∈ HomK (A⊗n , A) and h ∈ HomK (A⊗p , A), the follow-
ing hold:

(I) [f, g] = −(−1)(m−1)(n−1) [g, f ]

(II) (−1)(m−1)(p−1) [f, [g, h]] + (−1)(n−1)(m−1) [g, [h, f ]] + (−1)(p−1)(n−1) [h, [f, g]] = 0

(III) d∗ [f, g] = (−1)(n−1) [d∗ f, g] + [f, d∗ g]

Proof. (I) clearly follows from the definition of the bracket. (II) follows from

[f, [g, h]] =f ◦ [g, h] − (−1)(m−1)(n+p−2) [g, h] ◦ f

=f ◦ g ◦ h − (−1)(n−1)(p−1) f ◦ h ◦ g − (−1)(m−1)(n+p−2) g ◦ h ◦ f
+ (−1)(n−1)(p−1)+(m−1)(n+p−2) h ◦ g ◦ f

So when multiplied by its coefficient in (II):

(−1)(m−1)(p−1) [f, [g, h]] =(−1)(m−1)(p−1) f ◦ g ◦ h − (−1)(p−1)(m−1+n−1) f ◦ h ◦ g

− (−1)(m−1)(n+p−2+p−1) g ◦ h ◦ f
+ (−1)(n−1)(p−1)+(m−1)(n+p−2+p−1) h ◦ g ◦ f
⇒ (−1)(m−1)(p−1) [f, [g, h]] =(−1)(m−1)(p−1) f ◦ g ◦ h − (−1)(p−1)(m+n−2) f ◦ h ◦ g (F)
− (−1)(m−1)(n−1) g ◦ h ◦ f + (−1)(n−1)(p+m−2) h ◦ g ◦ f

Observe that the first and third terms in (F) are cyclic permutations of f, g and h and the second and
fourth terms in (F) are the non-cyclic permutations. Hence, since the powers of −1 in the coefficients
4 CIRCLE PRODUCT AND GERSTENHABER BRACKET 15

respect the permutations, the terms on the LHS of (II) will all cancel out.
(III) follows from identity (22) in Lemma (4.2):
 
d∗ [f, g] = d∗ f ◦ g − (−1)(m−1)(n−1) g ◦ f
=(−1)(n−1) d∗ f ◦ g + f ◦ d∗ g + (−1)mn+n−1 f ^ g + (−1)n g ^ f
 
− (−1)(m−1)(n−1) (−1)(m−1) d∗ g ◦ f + g ◦ d∗ f + (−1)mn+m−1 g ^ f + (−1)m f ^ g
   
=(−1)(n−1) d∗ f ◦ g − (−1)(n−1)m g ◦ d∗ f + f ◦ d∗ g − (−1)(m−1)n d∗ g ◦ f
 
+ (−1)mn+n−1 − (−1)(mn−n−m+1)+m f ^ g
 
+ (−1)n − (−1)(mn−n−m+1)+mn+m−1 g ^ f
=(−1)(n−1) [d∗ f, g] + [f, d∗ g] + 0 + 0 = (−1)(n−1) [d∗ f, g] + [f, d∗ g]

Similar to d∗ , since the Gerstenhaber bracket is defined on the homogeneous elements of C∗ (A) =
⊕n≥0 HomK (A⊗n , A), it can be linearly extended to act on all of C∗ (A). Furthermore, the Gersten-
haber bracket is well-defined on Hochschild cohomology. First, observe that by Lemma (4.8) (III),
for f and g as in the Lemma, we have

d∗ [d∗ f, g] = (−1)n−1 [0, g] + [d∗ f, d∗ g] = [d∗ f, d∗ g] (Θ)

n n n
= (−1) ((−1) [d∗ f, d∗ g] + 0) = (−1) d∗ [f, d∗ g]

Theorem 4.9. The Gerstenhaber Bracket is well-defined on HH ∗ (A) = ⊕n≥0 HH n (A).

Proof. (I) Let f ∈ Ker(dm+1 ) and g ∈ Ker(dn+1 ), then

d∗ [f, g] = [d∗ f, g] + [f, d∗ g] = 0 + 0 = 0

So [f, g] = 0 ∈ Ker(dn+m+1 ). Furthermore, if d∗ α ∈ Im(dm ) and d∗ β ∈ Im(dn ), then by Lemma

(4.8) (III) and (Θ),

[f + d∗ α, g + d∗ β] =[f, g] + [d∗ α, g] + [f, d∗ β] + [d∗ α, d∗ β]

=[f, g] + (−1)n−1 (d∗ [α, g] − [α, d∗ g]) + d∗ [f, β] − (−1)n−2 [d∗ f, β]

+ d∗ [d∗ α, β]

and since f, g ∈ Ker(d∗ ), then

[f + d∗ α, g + d∗ β] =[f, g] + (−1)n−1 d∗ [α, g] + d∗ [f, β] + d∗ [α, d∗ β]

=[f, g] + d∗ (−1)n−1 [α, g] + [f, β] + [d∗ α, β]

So [−, −] is well-defined on HH ∗ (A).

Remark 4.10. We have only used the cup product here to ease the notation in our proofs; however,
the Cup product ^ has several important properties on its own. In fact, HH ∗ (A) has a graded
commutative ring structure with respect to the cup product3 , which was one of the major results in
Gerstenhaber’s paper [5]. Due to this rich structure and the connection of the cup product and the
bracket, which were explored by Gerstenhaber in [5], an algebra with a cup product and a bracket,
satisfying similair conditions, is often called a Gerstenhaber algebra or G-algebra. For more detail,
refer to Definition 1.4.6 of [22].
3
With a shift of 1 compared to the grading of HH ∗ (A) as a DGLA as introduced in Theorems (4.12) and (4.14)
4 CIRCLE PRODUCT AND GERSTENHABER BRACKET 16

In fact both C∗ (A) and HH ∗ (A) are enriched with the structure of graded Lie algebras via the
Gerstenhaber bracket:

Definition 4.11. We say a graded K-vectorspace g = ⊕n∈Z gn is a graded Lie algebra if there exists
a bilinear [−, −] : g ⊗ g → g which satisfies [gm , gn ] ⊂ gm+n and

(I) [a, b] = −(−1)|a||b| [b, a] (Anti-commutativity)

(II) (−1)|a||c| [a, [b, c]] + (−1)|b||a| [b, [c, a]] + (−1)|c||b| [c, [a, b]] = 0 (Graded Jacobi identity)

for homogeneous elements a, b, c ∈ g. We say a K-vectorspace g is differential graded Lie algebra

or DGLA, if in addition to having a graded Lie algebra structure, we have a map d : g → g such that
d is a differential i.e. d2 = 0 and for homogeneous element a, da is homogeneous with |da| = |a| + 1.
Further the differential and the bracket interact by the graded Leibniz rule:

d[a, b] = [da, b] + (−1)|a| [a, db] (28)

Notation. For a graded vectorspace V = ⊕n∈Z Vn , we denote the shift of V by i, by V +i =

⊕n∈Z Vn+i , where Vn+i = Vn+i . Also observe any Z≥0 -graded vectorspace ⊕n≥0 Vn is Z-graded with
Vn = 0 for n < 0.

Theorem 4.12. Both C∗ (A)+1 and HH ∗ (A)+1 form graded Lie algebras with respect to the Ger-
stenhaber bracket.

Notice that although identity (III) from Lemma (4.8) resembles the graded Leibniz rule, it does
not agree with it completely. This can be fixed by taking the differential map δ∗ : C∗ (A) → C∗ (A)
defined on homogeneous elements f ∈ HomK (A⊗m , A) by δ∗ (f ) = (−1)m−1 d∗ (f ), and extended
to C∗ (A). Clearly, since d2∗ = 0 and d∗ is well-defined on the Hochschild complex HH ∗ (A), then δ∗
is a differential and is well-defined on HH ∗ (A) as well.

Lemma 4.13. For f ∈ HomK (A⊗m , A), g ∈ HomK (A⊗n , A),

δ∗ [f, g] = [δ∗ f, g] + (−1)m−1 [f, δ∗ g] (29)

holds.

Proof. This follows from (III) of Lemma (4.8):

 
δ∗ [f, g] =(−1)(m+n−1)−1 d∗ [f, g] = (−1)m+n (−1)(n−1) [d∗ f, g] + [f, d∗ g]
=(−1)(m−1) [d∗ f, g] + (−1)m−1+n−1 [f, d∗ g]
=[(−1)(m−1) d∗ f, g] + (−1)m−1 [f, (−1)n−1 d∗ g] = [δ∗ f, g] + (−1)m−1 [f, δ∗ g]

Theorem 4.14. Both C∗ (A)+1 and HH ∗ (A)+1 have DGLA structures with respect to the Gersten-
haber bracket and the differential δ∗ .

It is worth mentioning that maps δ∗ and d∗ , when restricted to HH ∗ (A)+1 , are just the zero map.
The zero map when looked at as a differential is often referred to as the zero differential.

Remark 4.15. Many modern texts on deformation theory focused on DGLAs are often unclear about
the choice of δ∗ . Since in the literature, it is a known fact that complexes C∗ (A)+1 and HH ∗ (A)+1
form DGLAs with the Gerstenhaber bracket, not much attention is given to this choice of the differ-
ential. In [20] and [14], when mentioning the ”Hochschild DGLA”, the differential is taken falsly
5 DGLA PHILOSOPHY 17

as d∗ . However, [3] chooses to introduce the Hochschild differential map as δ∗ instead of d∗ . The
text [15] was the only text the author found, that this change was explicitly stated. The importance of
this choice, in addition to not agreeing with the common form of the Leibniz rule, is its effect on the
Maurer-Cartan equation which we will introduce in the next section. If we take d∗ as the differential,
then the format of the Maurer-Cartan changes by a negative sign as indicated in equation (4.3.1) in
[22]. We have chosen to introduce δ∗ at this point to both respect the older theory and notation set by
Gerstenhaber and Hochschild, and connect it with the modern theory.
A statement that is sometimes mentioned, is that ”the Gerstenhaber bracket enriches C∗ (A)+1
and HH ∗ (A)+1 with DGLA structures”. This is because the differentials d∗ and δ∗ can be defined in
terms of the bracket:
Lemma 4.16. For f ∈ HomK (A⊗m , A), [f, π] = −d∗ f .
Proof. By the Definition of the bracket [f, π] ∈ HomK (A⊗m+2−1 , A) and
 
[f, π](a1 ⊗ · · · ⊗am+1 ) = f ◦ π − (−1)(m−1)(2−1) π ◦ f (a1 ⊗ · · · ⊗ am+1 )
m
X
= (−1)(i−1)(2−1) f (a1 ⊗ · · · ⊗ ai ai+1 ⊗ · · · ⊗ am+1 )
i=1
 
− (−1)m−1 f (a1 ⊗ · · · ⊗ am )am+1 + (−1)(m−1) a1 f (a2 ⊗ · · · ⊗ am+1 )
= − d∗ (f )

Hence, we can write operation d∗ as (−1)[−, π]. On the other hand, since δ∗ (f ) = (−1)m−1 d∗ (f )
for f ∈ HomK (A⊗m , A), then by Lemma (4.8) (I),
δ∗ (f ) = (−1)(m−1) (−[f, π]) = −(−1)(m−1)(2−1) [f, π] = [π, f ] (30)
Which gives us a re-interpretation of δ∗ as [π, −]. This identity is the last ingredient we need to
describe deformations in terms of DGLAs.

5 DGLA Philosophy
By Theorem (4.14), C∗ (A)+1 = ⊕n≥−1 HomK (A⊗n+1 , A) has a DGLA structure with the Gersten-
haber bracket and differential map δ∗ . Operations [−, −] and d∗ can both be t-linearly extended to
C∗ (A)+1 [[t]] = ⊕n≥−1 HomK (A⊗n+1 , A)[[t]] and induce a DGLA structure on C∗ (A)+1 [[t]]. Let µ
be a one-parameter deformation of algebra A, thereby
X
µ = µ0 + µ1 t + µ2 t2 + · · · = µi t i
i≥0

where µi ∈ HomK (A⊗2 , A). Hence µ ∈ HomK (A⊗2 , A)[[t]].

Recall the relation between the circle product and associativity from (23). Since we have linearly
extended the Gerstenhaber bracket (and the circle product), then
µ(µ(a ⊗ b) ⊗ c) − µ(a ⊗ µ(b ⊗ c)) = µ ◦ µ

Suppose that Char(K) 6= 2. Then

[µ, µ] = µ ◦ µ − (−1)2−1 µ ◦ µ = 2µ ◦ µ
and thereby:
5 DGLA PHILOSOPHY 18

Proposition 5.1. If Char(K) 6= 2. Then µ ∈ HomK (A⊗n , A)[[t]] defines a formal deformation if
and only if [µ, µ] = 0.
Recall that if µ is a deformation, then µ0 = π where π ∈ HomK (A⊗2 , A) is the multiplication
of algebra A. So µ = π + µ0 where µ0 ∈ tHomK (A⊗2 , A)[[t]].

[µ, µ] = [π + µ0 , π + µ0 ] = [π, π] + [π, µ0 ] + [µ0 , π] + [µ0 , µ0 ]

Since π is associative, then [π, π] = 0. Further by Lemma (4.8) (I), [µ0 , π] = −(−1)(2−1)(2−1) [π, µ0 ].
Hence
[µ, µ] = 2[π, µ0 ] + [µ0 , µ0 ]
By Lemma (4.16) and the resulting identity (30), [π, f ] = δ∗ (f ). Hence,

[µ, µ] = 2δ∗ (f ) + [µ0 , µ0 ]

Corollary 5.2. If Char(K) 6= 2. Then µ = π + µ0 with µ0 ∈ tHomK (A⊗n , A)[[t]] defines a formal
deformation if and only if
2δ∗ (f ) + [µ0 , µ0 ] = 0 (MC)
holds.
Equation (MC) is called the Maurer-Cartan equation.
Since every deformation µ = π + µ0 is determined uniquely by µ0 (not upto isomorphism), we have
the bijection

{Formal deformations of A, µ} ←→ {µ0 ∈ tHomK (A⊗n , A)[[t]] satisfying (MC)} (31)

By what we mentioned at the start of the section and Theorem (4.14), C∗ (A)+1 [[t]] has a DGLA
structure and so does tC∗ (A)+1 [[t]]
Definition 5.3. For a DGLA g = ⊕n∈Z gn , we say α ∈ g1 is a Maurer-Cartan element if it satisfies
the Maurer-Cartan Equation with the respective bracket and differential of the DGLA. The set of
Maurer-Cartan elements of g is denoted by MCE(g)
In our case, g = tC∗ (A)+1 [[t]], and MCE(g) ⊂ tHomK (A⊗2 , A)[[t]]. So we can rewrite the
bijection (31) as

{Formal deformations of A, µ} ←→ MCE tC∗ (A)+1 [[t]]

(32)

Notation. From now on we will denote DGLA tC∗ (A)+1 [[t]] by C.

Remark 5.4. If a homogeneous element α of a DGLA g = ⊕n∈Z gn were to satisfy the Maurer-Cartan
equation, then δα and [α, α] must have the same degree:

|δα| = |α| + 1 = 2|α| = |[α, α]| ⇒ |α| = 1

Since we are interested in deformations up to equivalence, we look to describe the equivalence of

deformations in terms of DGLAs. Recall from (3.1) that an equivalence of deformations, φt , is of the
form
φt = IdA + φ1 t + φ2 t2 + · · ·
where φi ∈ HomK (A, A).
Assume Char(K) = 0.
Since over characteristic 0, log and exp are well defined. Then let ϕt = log(φt ) ∈ HomK (A, A)[[t]],
ϕ2t
IdA + φ1 t + φ2 t2 + · · · = φt = exp(ϕt ) = IdA + ϕt + + ···
2
5 DGLA PHILOSOPHY 19

Also recall the expansion of the logarithm function

a2 a3
log(1 + a) = a − + + ···
2 3
Which implies ϕt belongs to tHomK (A, A)[[t]], since a = φt − IdA ∈ tHomK (A, A)[[t]]. So any
equivalence of deformations φt can be written as exp(ϕt ) where ϕt ∈ C0 = tHomK (A, A)[[t]]. Now,
let ϕ ∈ C0 = tHomK (A, A)[[t]], and α0 ∈ MCE(C). Then by bijection (32), π + α0 is a deformation
and by (19)
β = exp(ϕ)(π + α0 )(exp(−ϕ) ⊗ exp(−ϕ))
is also a deformation, since exp(−ϕ) = exp(ϕ)−1 . By bijection (32), β = π + β 0 where β 0 is a
Maurer-Cartan element.
π + β 0 =exp(ϕ)(π + α0 )(exp(−ϕ) ⊗ exp(−ϕ))
⇒ exp(ϕ)(π + α0 )(exp(−ϕ) ⊗ exp(−ϕ)) − π = β 0 ∈ MCE(C)
Hence, elements of exp(C0 ) = {exp(ϕ)|ϕ ∈ C0 } act on MCE(C) via
exp(ϕ) y α0 = exp(ϕ)(π + α0 )(exp(−ϕ) ⊗ exp(−ϕ)) − π ∈ MCE(C) (33)
Moreover, exp(C0 ) forms a group and this action respects that group structure: Observe that in any
DGLA g = ⊕n∈Z gn , g0 and [−, −] form a Lie algebra, since the bracket takes elements of degree 0
to an element of degree 0, [−, −] : g0 ⊗ g0 → g0 .
Theorem 5.5. (Baker-Campbell-Hausdorff formula)[Theorem 2.1 [23]] If h is a Lie algebra over
a field K with characteristic 0, and α, β ∈ h, then exp(α)exp(β) = exp(γ) for γ, a formal infinite
sum of elements in h; in particular
∞
X (−1)n+1 X [αr1 , β s1 , . . . , αrn , β sn ]
γ = log (exp(α)exp(β)) = (34)
( i=1 ri + si ) ( ni=1 ri si )
Pn Q
n
n=1 r1 +s. 1 >0
..
rn +sn >0

where
[αr1 , β s1 , . . . , αrn , β sn ] = [α, [α, [· · · , [α, [β, [β, [· · · , [β,[a, [· · · ]]]]
| {z }| {z }
r1 s1

In the above Theorem, if γ is well-defined in h i.e. the infinite sum converges to an element of
h, then exp(g0 ) forms a group with multiplication. In our case, it is important to notice that since
α, β ∈ tHomK (A, A)[[t]], the power of t in [αr1 , β s1 , . . . β sn ] is at least ni=1 ri +si ≥ n. Hence, for
P
any n, the coefficient of tn is a finite sum, and log(exp(α)exp(β)) is well-defined and is an element
of exp(C0 ) = tHomK (A, A)[[t]].
Now we show that the action of exp(C0 ) on MCE(C), as described in (33), respects the group structure
of exp(C0 ):
exp(β) y exp(α) y µ0 = exp(β) y exp(α)(π + µ0 )(exp(−α) ⊗ exp(−α)) − π

=exp(β)(π + exp(α)(π + µ0 )(exp(−α) ⊗ exp(−α)) − π)(exp(−β) ⊗ exp(−β)) − π

= (exp(β)exp(α)) (π + µ0 )(exp(−α)exp(−β) ⊗ exp(−α)exp(−β)) − π
= (exp(β)exp(α)) y µ0
Hence, deformations are equivalent if and only if their corresponding Maurer-Cartan elements are in
the same orbit of exp(C0 ). So we have the bijection
( ) ( )
Formal deformations of A, µ, over K[[t]] Orbits of MCE (C0 ) , under
←→ (35)
upto equivalence the group action of exp(C0 )
5 DGLA PHILOSOPHY 20

The equivalence defined by the action of exp(g0 ) on MCE(C0 ) as described in (33), is called Gauge
equivalence/action and exp(g0 ) is referred to as the Gauge group. In fact, by taking advantage of
group-theoretic notation, bijection (35) is often presented as
( )
Formal deformations ofA, µ, over K[[t]] MCE(C0 )
←→ (36)
upto equivalence exp(C0 )

The above bijection is the cornerstone of the so called “DGLA Philosophy” in deformation theory.
Notice that the problem of classifying deformations of A over K[[t]] completely translates to a prob-
lem on the DGLA, C0 = tC∗ (A)+1 [[t]]. The DGLA philosophy as described by Mannetti in the first
page of [14] states that “over characteristic 0, every deformation problem is governed by a DGLA via
solutions of the Maurer-Cartan equation modulo gauge action”.
Generalising our work:
For algebra A, let L := C∗ (A)+1 and observe that the DGLA we have been working with is

C = L ⊗ tK[[t]] = ⊕ HomK (A⊗n+1 , A) ⊗ tK[[t]]

n≥−1

• Firstly, notice that K-algebra K[[t]] has a decomposition K[[t]] = K⊕(t) as a K-vectorspace, where
(t) = tK[[t]] is an ideal of K[[t]].

Definition 5.6. For a field K, we say K-algebra R is an augmented algebra4 if R ∼

= K ⊕ R+ , as a
K-vectorspace, for an ideal R+ .

We used the fact that K[[t]] = K ⊕ (t) is augmented so that the definition of a one-parameter
deformation over K[[t]] would make sense: We are using the existence of a decomposition

A ⊗ K[[t]] = (A ⊗ K) ⊕ (A ⊗ (t))

for the new multiplication on A ⊗ K[[t]] to take form µ = i≥0 µi ti with µ0 = π. For an augmented
P
K-algebra R = K⊕R+ , multiplication µ on A⊗R will decompose into µ = µ0 +µ0 with Im(µ0 ) ⊆ A
and Im(µ0 ) ⊆ (A ⊗ R+ ). With these conditions, a deformation makes sense by requiring µ0 to be
the original multiplication of A.
• Secondly, observe that at the start of this section where we extend the Gerstenhaber bracket and
circle product t-linearly to tC∗ (A)+1 [[t]], we are using the commutativity of tK[[t]]. In fact, if we
look at tC∗ (A)+1 [[t]] as C∗ (A)+1 ⊗ tK[[t]] = ⊕n≥−1 HomK (A⊗n+1 , A) ⊗ tK[[t]], we are extending
these operations via
(f ⊗ a) ◦ (g ⊗ b) = (f ◦ g ⊗ ab)
where we would previously write

(f tm ) ◦ (gtn ) = (f ◦ g)tm+n

We require the commutativity of tK[[t]] in order for the anti-commutativity condition from Definition
(4.11) to hold in C∗ (A)+1 ⊗ tK[[t]]:

(f ◦ g ⊗ ab) = (f ⊗ a) ◦ (g ⊗ b) = − (−1)|f ||g| (g ⊗ b) ◦ (f ⊗ a)

= − (−1)|f ||g| (g ◦ f ⊗ ba) = (f ◦ g ⊗ ba)

i
P
• Throughout all our calculations, we have used the fact that any sum of the form i≥0 αi t with
4
An equivalent definition can be found in [20], Definition 1.10.10.
5 DGLA PHILOSOPHY 21

αi ∈ K, represents an element in K[[t]]. Another way to put this phenomena is to say for any
sequence (α0 , α1 , α2 , . . . ) with αi in K, there exists an element α ∈ K[[t]] such that
n−1
X
αi ti mod R+
n


α≡
i≥0

where R+ = (t) = tK[[t]].

Definition 5.7. For ideal I / R consider the inverse sequence
R/I  R/I 2  R/I 3  · · ·
The inverse limit of the sequence, R̂ = lim n
←−R/I is called the completion of R with respect to the
I -adic topology.
By the definition of the inverse limit:
R̂ = {(r0 , r1 , r2 , . . . , ) : rn ∈ R/I n and rn+1 ≡ rn (mod I n )} (37)
It is easy to see that there exists a natural map ι : R → R̂. Hence, the assumption we have used
for K[[t]] is in fact equivalent to ι being surjective, where R̂ is the completion of augmented ring
R = K ⊕ R+ with respect to ideal R+ . In fact for R = K[[t]], ι is an isomorphism and in order to
fully generalise our work, we also require ι to be injective:
• An important fact that we used in this section was that the Baker-Campbell-Hausdorff formula is
well-defined on L0 ⊗ tK[[t]]. The property of K[[t]] that we used, is that ideal tK[[t]] is pronilpotent:
n = 0.
Definition 5.8. We say ideal R+ of ring R is pronilpotent5 if ∩n≥0 R+
In fact, the ideal R+ being pronilpotent is equivalent to the map ι : R → R̂ being injective.
Definition 5.9. We say ring R with ideal I / R is complete with respect to the I -adic topology, if the
natural map ι : R → R̂ is an isomorphism. For an augmented ring R = K ⊕ R+ , we simply say R
is a complete augmented ring if it is complete with respect to the R+ -adic topology.
To elaborate further, after Theorem (5.5) we argued that the number of elements in the expansion
of log(exp(α)exp(β)) with the nth power of t is finite. For a pronilpotent ideal R
+ , since ∩n≥0 R+n =

0, the number of elements in the expansion intersecting with R+ r \ ∩r n

n≥0 R+ are finite. And by
(37), elements in R ∼ = R̂ can be written as sequences (r0 , r1 , r2P, . . . , ), where rn ∈ R/R+ n and
n
rn+1 ≡ rn (mod R+ ), which via ι correspond to formal sums r0 + n≥0 (rn+1 − rn ) in R, where
r
\ ∩rn≥0 R+ n


rn+1 − rn ∈ R+
and r0 ∈ K. So the infinite sum in the BCH formula is well-defined if R+ is pronilpotent and R is
complete with respect to R+ .
Furthermore, we have continuously utilized K[[t]] being complete with respect to the (t)-adic topol-
ogy when extending addition to A ⊗ K[[t]], t-linearly. In fact, this t-linear extension is often written
ˆ sometimes called the topological tensor product, and constructed as follows
using the notation ⊗,
n
ˆ = lim A ⊗R R/R+
A⊗R ←−
n→∞

However, since t-linear extension over K[[t]] is commonly used and well understood by Mathemati-
cians, the author has chosen to avoid this bit of detail and notation until now.
By the above explanations, we can generalise our deformation theory to any commutative complete
augmented K-algebra R = K ⊕ R+ :
5
The definition of pronilpotent ideals as presented here is due to [19].
5 DGLA PHILOSOPHY 22

Definition 5.10. For algebra A with L = C∗ (A)+1 and commutative complete augmented K-algebra
ˆ ⊗(A
R = K ⊕ R+ , a formal deformation of A over R, is a map µ : (A⊗R) ˆ ⊗R)ˆ ˆ such that
→ A⊗R
µ is associative and
µ ≡ π (mod R+ )

Furthermore, since our theory on DGLAs can be generalised, we have a bijection

( )
Formal deformations of A,µ, over R, MCE(L⊗R ˆ +)
←→ (38)
upto equivalence ˆ +)
exp(L0 ⊗R

Observe that examples of algebras satisfying the required conditions include R = K[t]/(tn ) for
n ≥ 2, with R+ = tK[t](tn ), where their deformation theory sometimes referred to as (n − 1)-th
order deformations, as well as R = K[[t1 , t2 , . . . , tn ]] with R+ = (t1 , t2 , . . . , tn ) giving us several-
parameter deformations.

Example 5.11. Recall from Remark (2.10), that when R = K[t]/(t2 ), there exists a bijection between
deformations of an algebra A over K[t]/(t2 ) and HH 2 (A). We can confirm this via the DGLA method
as well: Let L = C∗ (A)+1 and µ0 ∈ L1 ⊗ tK[t]/(t2 ). Then µ0 = αt for α ∈ HomK (A⊗2 , A) and

[µ0 , µ0 ] = 2µ0 ◦ µ0 = 2(α ◦ α)t2 = 0

So µ0 ∈ L1 ⊗ (t) is a Maurer-Cartan element if and only if δ∗ (µ0 ) = 0 = −d3 µ0 and

MCE(L ⊗ tK[t]/(t2 )) = Ker(d3 )

Further the gauge action is induced by elements exp(γ) where γ ∈ L0 ⊗ (t) = tHomK (A, A).
Hence, since t2 = 0, such a γ has form γ = ϕt where ϕ ∈ HomK (A, A) and exp(ϕt) = IdA + ϕt.
Thereby, the gauge action described in (33) becomes

exp(ϕt) y αt =exp(ϕt)(π + αt)(exp(−ϕt) ⊗ exp(−ϕt)) − π

=(IdA + ϕt)(π + αt)((IdA − ϕt) ⊗ (IdA − ϕt)) − π
=π((IdA − ϕt) ⊗ (IdA − ϕt)) + α(IdA ⊗ IdA )t + ϕπ(IdA ⊗ IdA )t − π
=π − π − π(ϕ ⊗ IdA )t − π(IdA ⊗ ϕ)t + αt + ϕ(π)t
=(α − d2 ϕ)t

Thereby,
MCE L ⊗ tK[t]/(t2 ) ∼

.
= Ker(d3 ) Im(d ) = HH 2 (A)
exp (L0 ⊗ tK[t]/(t2 )) 2

Which confirms bijection (18):

( )
Formal deformations ofA, over K[t]/(t2 ) MCE L ⊗ tK[t]/(t2 )


←→ ←→ HH 2 (A)
upto equivalence exp (L0 ⊗ tK[t]/(t2 ))

Remark 5.12. In many papers on deformation theory and DGLAs, including [14], Def L (R) is first
defined in the case where R = K⊕R+ and R+ is a nilpotent ideal, or equivalently R is local Artinian
with residue field K. Observe that for a commutative complete augmented ring R = K ⊕ R+ , the
rings R/R+ n all have a nilpotent maximal ideal R and R is equal to the inverse limit of these rings.
+
Hence, once we define the theory for nilpotent R+ , we can extend the theory to our case where R+ is
pronilpotent and R a commutative complete augmented ring.
6 PRELUDE TO DEFORMATION QUANTIZATION 23

We denote the set on the RHS of (38) by

MCE(L ⊗ R+ )
Def L (R) :=
exp(L0 ⊗ R+ )
For each DGLA L, Def L : ArtK → Set, where ArtK is the category of local Artinian K-algebras
and Set the category of sets, is functorial and for each DGLA we obtain a functor Def L called the
deformation functor. More detail on this construction can be found in [14] and [3].
A natural question to ask is when two algebras have the same deformation theory, which translates to
their cochain DGLAs admitting the same Deformation functor.
Definition 5.13. We say a morphism of DGLAs Φ : L → N is a quasi-isomorphism is the induced
morphism on their cohomology DGLAs, H(Φ) : H(L) → H(N ) is an isomorphism.
Theorem 5.14. [Corollary 3.2 [14]] Let Φ : L → N be a quasi-isomorphism of DGLAs. Then the
induced morphism Def L → Def N is an isomorphism.
Observe that in our case H C∗ (A)+1 = HH ∗ (A)+1 and thereby, over characteristic 0, the

deformation theory of an algebra is completely unique to its Hochschild cohomology.

6 Prelude to Deformation Quantization

Much of the interest in Algebraic Deformation Theory is due to its connections with deformations
in Geometry and its connections with Theoretical Physics. In fact parameter t is often denoted by ~,
since the case of interest in Physics is when the parameter equals Planck’s constant. The deforma-
tion philosophy in Mathematical Physics as described in the survey [2] says: ”Intuitively, classical
mechanics is the limit of quantum mechanics when ~ = h/2π goes to zero”. As described in section
III of the same paper, Classical mechanics is concerned with the theory of Poisson manifolds and
the deformation of interest are deformations of the algebra of functions on the manifold. Hence, we
devote the last section of this essay to Poisson algebras and Deformation Quantization:
Definition 6.1. A commutative associative algebra A is said to be a Poisson algebra if A is equipped
with a Lie bracket {−, −} such that

{a, bc} = {a, b}c + b{a, c} (39)

holds. The bracket {−, −} is then called the Poisson bracket on A.

Any commutative algebra can in fact be endowed with a Poisson structure via its Hochschild
cohomology:
Theorem 6.2. Let A be a commutative algebra and ρ a Hochschild 2-cocycle such that [ρ, ρ] = 0,
then the K-bilinear map {−, −} defined by

{a, b} = ρ(a ⊗ b) − ρ(b ⊗ a) (40)

defines a Poisson bracket on A.

Proof. First, we show {−, −} is a Lie bracket. Observe that since [ρ, ρ] = 0, then
1 1
ρ(ρ(a ⊗ b) ⊗ c) − ρ(a ⊗ ρ(b ⊗ c)) = ρ ◦ ρ(a, b, c) = [ρ, ρ](a, b, c) = 0
2 2
and

{a, {b, c}} ={a, ρ(b ⊗ c) − ρ(c ⊗ b)}

6 PRELUDE TO DEFORMATION QUANTIZATION 24

=ρ(a ⊗ ρ(b ⊗ c)) − ρ(a ⊗ ρ(c ⊗ b)) − ρ(ρ(b ⊗ c) ⊗ a) + ρ(ρ(c ⊗ b) ⊗ a)

then, the Jacobi identity follows directly:

{a, {b, c}}+{b, {c, a}} + {c, {a, b}} =

ρ(a ⊗ ρ(b ⊗ c)) − ρ(a ⊗ ρ(c ⊗ b)) − ρ(ρ(b ⊗ c) ⊗ a) + ρ(ρ(c ⊗ b) ⊗ a)
+ρ(b ⊗ ρ(c ⊗ a)) − ρ(b ⊗ ρ(a ⊗ c)) − ρ(ρ(c ⊗ a) ⊗ b) + ρ(ρ(a ⊗ c) ⊗ b)
+ρ(c ⊗ ρ(a ⊗ b)) − ρ(c ⊗ ρ(b ⊗ a)) − ρ(ρ(a ⊗ b) ⊗ c) + ρ(ρ(b ⊗ a) ⊗ c) = 0

Furthermore, since d3 ρ = 0 and A is commutative, then {a, bc} = {a, b}c + b{a, c} follows from

0 = d3 ρ(a ⊗ b ⊗ c) = aρ(b ⊗ c) − ρ(ab ⊗ c) + ρ(a ⊗ bc) − ρ(a ⊗ b)c

⇒ {a, bc} =ρ(a ⊗ bc) − ρ(bc ⊗ a) = ρ(a ⊗ cb) − ρ(cb ⊗ a)
=d3 ρ(a ⊗ c ⊗ b) + d3 ρ(c ⊗ b ⊗ a) + ρ(ac ⊗ b) + ρ(a ⊗ c)b − cρ(b ⊗ a) − ρ(c ⊗ ba)
=d3 ρ(a ⊗ c ⊗ b) + d3 ρ(c ⊗ b ⊗ a) − d3 ρ(c ⊗ a ⊗ b) + cρ(a ⊗ b) − ρ(c ⊗ a)b
+ ρ(a ⊗ c)b − cρ(b ⊗ a) = 0 + 0 − 0 + {a, b}c + {a, c}b = {a, b}c + b{a, c}

Observe that for a commutative algebra A, and ϕ ∈ HomK (A, A)

d2 ϕ(a ⊗ b) = aϕ(b) − ϕ(ab) + ϕ(a)b = ϕ(b)a − ϕ(ba) + bϕ(a) = d2 ϕ(b ⊗ a)

Hence, for Hochschild 2-cocycle ρ ∈ Ker(d3 ), ρ and ρ + d2 ϕ induce the same Poisson bracket via
(40).

P of ai Poisson algebra A with

Definition 6.3. A deformation quantization Poisson bracket {−, −} is
⊗n
a formal deformation Aµ with µ = π + i≥1 µi t and µi ∈ HomK (A , A) such that6

µ(a ⊗ b) − µ(b ⊗ a) ≡ {a, b}t (mod t2 ) (41)

Example 6.4. For A = K[x, y], recall the definition of map ψ from Example (4.6). As we demon-
strated previously, ψ is a Hochschild 2-cocycle. In fact it is well-known that
∂a ∂b ∂b ∂a
{a, b} = − = ψ(a ⊗ b) − ψ(b ⊗ a)
∂x ∂y ∂x ∂y
defines a Poisson bracket on A, as confirmed by Theorem (6.2). In fact in Example (4.6), we showed
that the truncated deformation π + ψt can extend to a formal deformation Aµ1 , giving a deformation
quantization of K[x, y] with respect to this Poisson structure. A similar deformation quantization
exists for any polynomial algebra and the formal deformation is referred to as the Moyal product
[Example 1.4.1 [13]]. Although, in this case, the Moyal product gives an example of deformation
quantization if its definition required the Poisson bracket to occur as the coefficient of t. This is
slightly different from our example, but both can be recovered from each other.

A Poisson manifold is a smooth manifold M with a Poisson bracket on its algebra of smooth
functions C ∞ (M). The main question of Quantization theory is “Does an arbitrary Poisson manifold
have a deformation quantization?”, where a deformation quantization of a Poisson manifold M is
∞
a deformation quantization of C (M), µ = π + i≥1 µi ti where µi are bidifferential operators.
P
This question was first worked on for the case of Symplectic manifolds in [1], where the idea of
6 1
In many texts, there is an additional coefficient of 2
on the LHS of (41), to indicate it being the anti-symmetric part of
bilinear map µ.
REFERENCES 25

deformation quantization was first pushed forward. Formal deformations of the mentioned form
with the additional property of having differential operators as coefficients are often referred to as
?-products. The solution in the case of Poisson manifolds is due to M. Kontsevich and Theorem
(5.14) is key to his solution. In [13], Kontsevich describes a quasi-isomorphism of two DGLAs, the
Hochschild complex restricted to differential operators, and the complex of poly-vector fields on M.
Maurer-Cartan elements of the first DGLA, as we’ve seen in section (5), are formal ?-products on
C ∞ (M) and the MC elements of the second DGLA are formal Poisson deformations of C ∞ (M). A
quasi-isomorphism of graded differential algebras was already known from the Hochschild-Konstant-
Rosenberg Theorem, but it did not respect the brackets of both sides. This issue was solved by
Kontsevich in [13], where he extends this map to a quasi-isomorphism of L∞ -algebras. DGLAs are
a specific case of L∞ -algebras. Hence by Theorem (5.14), this results to a bijection between formal
?-products and formal Poisson deformations on C ∞ (M), which by Kontsevich’s map then implies
the existence of a deformation quantization. The account of the proof described here is influenced by
the sketches given in [19] and section 8 of [3] and the proof in full detail can be found in [13].

References
[1] F. Bayen, M. Flato, C. Fronsdal, A.Lichnerowicz, D. Sternheimer, Deformation theory and quan-
tization I, Annals of Physics, Vol. 111, Issue 1, 1978, 61-110

[2] P. Bonneau, M. Gerstenhaber, A. Giaquinto and D. Sternheimer, Quantum groups and deforma-
tion quantization: Explicit approaches and implicit aspects, Journal of Mathematical Physics, Vol.
45, 2004, 3703-3741

[3] M. Doubek, M. Markl, P. Zima, Deformation Theory (Lecture Notes), Archivum Mathematicum,
Vol 43, 2007, issue 5, 333-371

[4] M. Gerstenhaber, S. D. Schack, Algebraic cohomology and deformation theory In: M.

Hazewinkel, M. Gerstenhaber Deformation Theory of Algebras and Structures and Applications,
NATO ASI series, Vol. 247, Springer, Dordrecht

[5] M. Gerstenhaber, The cohomology structure of an associative ring, Annals of Mathematics, Vol
78, No. 2, Sep 1963, 267-288

[6] M. Gerstenhaber, Deformations of Rings and Algebras, Ann. of Mathematics, 2nd Series, Vol.79,
No. 1, Jan 1964, 59-103

[7] M. Gerstenhaber, Deformations of Rings and Algebras III, Annals of Mathematics, 2nd Series,
Vol. 88, No. 1, Jul. 1968, 1-34

[8] M. Gerstenhaber, On the construction of division rings by the deformation of fields, Proceedings
of the National Academy of Sciences of the United States of America, Vol. 55, No. 4, Apr. 1966,
690-692

[9] M. Gerstenhaber, S. D. Schack, On the deformation of algebra morhpisms and diagrams, Trans-
actions of the American Mathematical Society, Vol. 279, N. 1, Sep 1983, 1-50

[10] T. F. Fox, Algebraic deformations and triple cohomology, Proceedings of American Mathemat-
ical Society., Vol 78, No. 4, Apr. 1980, 467-472

[11] T. F. Fox, An Introduction to algebraic deformation theory, Journal of Pure and Applied Algebra,
Vol. 84, 1993, 17-41
REFERENCES 26

[12] C. Kassel Quantum Groups, Vol. 155 of Graduate Texts in Mathematics, Springer-Verlag, 1995

[13] M. Kontsevich, Deformation quantization of Poisson manifolds, I, 1997,arxiv: q-alg/9709040

[14] M. Mannetti, Deformation theory via differential graded Lie algebras, 2005,
arxiv:math/0507284v1

[15] M. Mannetti, Differential graded Lie algebras, www1.mat.uniroma1.it/people/manetti/DT2011/

DGLA.pdf

[16] M. Penkava, P. Vanhaecke, Deformation Quantization of Polynomial Poisson Algebras, 1998,

arxiv: 9804022v1

[17] R. Pierce, Associative algebras, Vol. 88 of Graduate Texts in Mathematics, Springer-Verlag,

1982

[18] M. J. Redondo, Hochschild cohomology: some methods for computations, Resenhas IME-USP,
Vol. 5, No. 2, 2001, 113-137

[19] T. Schedler, Deformations of algebras in noncommutative algebraic geometry, 2012,

arxiv:1212.0914v1

[20] T. Schedler, Deformations of algebras in noncommutative geometry, 2015, arxiv:1212.0914v2

[21] C. A. Weibel, An introduction to homological algebra (Cambridge Studies in Advanced Mathe-

matics 38), Cambridge University Press. Reprint 1997

[22] S. Witherspoon, An Introduction to Hochschild Cohomology, www.math.tamu.edu/

sarah.witherspoon/pub/HH-18March2018.pdf

[23] R. Zhang, The Baker-Campbell-Hausdorff Formula, www.math.columbia.edu/ rzhang/files/BC

HFormula.pdf