a

r
X
i
v
:
1
0
0
8
.
4
0
7
6
v
2


[
m
a
t
h
.
R
A
]


1
0

S
e
p

2
0
1
0
NON COMMUTATIVE TRUNCATED POLYNOMIAL
EXTENSIONS
JORGE A. GUCCIONE, JUAN J. GUCCIONE, AND CHRISTIAN VALQUI
Abstract. We introduce the notion of non commutative truncated polyno-
mial extension of an algebra A. We study two families of these extensions.
For the rst one we obtain a complete classication and for the second one,
which we call upper triangular, we nd that the obstructions to inductively
construct them, lie in the Hochschild homology of A, with coecients in a
suitable A-bimodule.
Introduction
Let k be a commutative ring and let A, C be unitary k-algebras. By denition,
a twisted tensor product of A with C over k, is an algebra structure dened on
A
k
C, with unit 1 1, such that the canonical maps i
A
: A A
k
C and
i
C
: C A
k
C are algebra maps satisfying ac = i
A
(a)i
C
(c). This structure was
introduced independently in [Ma1] and [Tam], and it has been formerly studied by
many people with dierent motivations (In addition to the previous references see
also [B-M1], [B-M2], [Ca], [C-S-V], [C-I-M-Z], [G-G], [Ma2], [J-L-P-V], [VD-VK]).
A number of examples of classical and recently dened constructions in ring theory
ts into this construction. For instance, Ore extensions, skew group algebras, smash
products, etcetera (for the denitions and properties of these structures we refer to
[Mo] and [Ka]). On the other hand, it has been applied to braided geometry and
it arises as a natural representative for the product of noncommutative spaces, this
being based on the existing duality between the categories of algebraic ane spaces
and commutative algebras, under which the cartesian product of spaces corresponds
to the tensor product of algebras. And last, but not least, twisted tensor products
arise as a tool for building algebras starting with simpler ones.
Given algebras A and C, a basic problem is to determine all the twisted tensor
products of A with C. To our knowledge, the rst paper in which this problem was
attacked in a systematic way was [C], in which C. Cibils studied and completely
solved the case C = kk. Subsequently, in [J-L-N-S], the methods developed in [C]
were extended to cover the case C = k k (n-times). Meanwhile, in [G-G-V],
some partial results were obtained in the cases C = k[x] and C = k[[x]].
In this paper we consider this problem when C is a truncated polynomial alge-
bra k[y]/y
n
. We call these twisted tensor products non commutative truncated
polynomial extensions of A, because they have underlying module A[y]/y
n
and
include A and k[y]/y
n
as subalgebras.
2000 Mathematics Subject Classication. Primary 16S10; Secondary 16S80.
Key words and phrases. Twisting maps, Polynomial rings.
Jorge A. Guccione and Juan J. Guccione research were supported by UBACYT X095 and PIP
112-200801-00900 (CONICET).
Christian Valqui research was supported by PUCP-DGI-2010-0025, Lucet 90-DAI-L005, SFB
478 U. Munster, Konrad Adenauer Stiftung.
The second author thanks the appointment as a visiting professor C atedra Jose Tola Pasquel
and the hospitality during his stay at the PUCP.
1
2 JORGE A. GUCCIONE, JUAN J. GUCCIONE, AND CHRISTIAN VALQUI
It is well known that there is a canonical bijection between the twisted tensor
products of A with C and the so called twisting maps s: C
k
A A
k
C. So
each twisting map s is associated with a twisted tensor product of A with C over
k, which will be denoted by A
s
C.
It is evident that each k-linear map s: k[y]/y
n

k
A A
k
k[y]/y
n
de-
termines and it is determined by k-linear maps
r
j
: A A (0 j, r < n) such
that
(0.1) s(y
r
a) =
n1

j=0

r
j
(a) y
j
.
The map s so dened is a twisting map if the maps
r
j
satisfy suitable conditions.
In particular, we will see that B := ker
1
0
should be a subalgebra of A, and
1
0
a
nilpotent right B-linear map.
The main results of this paper are the following: Theorem 3.2, which determines
all the twisting maps such that
- B is a subalgebra of the center of A,
- s(B
k
A) A
k
B,
- there exist h 2 and x A such that
h
0
= 0 and
h1
0
(x) is invertible,
and Theorem 4.1, which establish that the obstruction to extend a twisting map
s
n
:
k[y]
y
n

k
A A
k
k[y]
y
n

with
1
0
= 0 to one
s
n+1
:
k[y]
y
n+1

k
A A
k
k[y]
y
n+1

,
lies in the Hochschild cohomology of A with coecients in a suitable bimodule.
We will call these non commutative polynomial extensions upper triangular. An
intersting fact of these extensions is that the evaluation in y = 0 is an algebra
homomorphism from A
s
k[y]/y
n
to A. As we point out in Remark 4.9, Theo-
rem 4.1 also can be used to construct a type of non commutative extensions of an
algebra A by power series.
From now on we assume implicitly that all the maps are k-linear maps, all the
algebras are over k, and the tensor product over k is denoted by , without any
subscript.
The paper is organized as follows: in Section 1) we make a quick review of the
basic general properties of twisted tensor products and twisting maps, we determine
necessary and sucient conditions for a family of maps
r
j
: A A (0 j, r < n),
in order that the map
s:
k[y]
y
n

A A
k[y]
y
n

,
dened by the formula (0.1), is a twisting map, and we introduce a canonical
representation of an arbitrary non commutative truncated polynomial extension
A
s
k[y]/y
n
, of an algebra A, in the matrix algebra M
n
(A). In Section 2), we
study a broad family of non commutative truncated polynomial extensions, which
includes those with
1
0
= 0. In Section 3) we classify the non commutative truncated
polynomial extensions with
1
0
,= 0 that satises a few natural conditions. Finally,
in Section 4), we consider the non commutative truncated polynomial extensions
with
1
0
= 0. These can be constructed inductively. For this, the main tool is
Theorem 4.1. Using it, we obtain several families of these sort of extensions. In
NON COMMUTATIVE TRUNCATED POLYNOMIAL EXTENSIONS 3
particular, all extensions of a truncated polynomial algebra k[x]/x
m
satisfying
s(y x) xk[x]/x
m
yk[y]/y
n
.
1. Some basic facts
This section is divided in two parts. In the rst one, we review the denitions
of twisted tensor products and twisting maps, and we establish some of the basic
results about these structures. For the proofs we refer to [C-S-V], [VD-VK] and
[C-I-M-Z]. Recall from the introduction that a non commutative truncated poly-
nomial extension of an algebra A is a twisted tensor product A
s
k[y]/y
n
. In the
second one, we start the study of these extensions, by determining the conditions
that a family of maps
r
j
: A A (0 j, r < n) must fulll in order that the map
s:
k[y]
y
n

A A
k[y]
y
n

,
given by
s(y
r
a) =
n1

j=0

r
j
(a) y
j
,
is a twisting map.
1.1. General remarks. Let A and C be algebras. Let
A
,
A
,
C
and
C
be the
multiplication and unit maps of A and C, respectively. A twisted tensor product of
A with C is an algebra structure on the k-module A C, such that the canonical
maps
i
A
: A A C and i
C
: C AC
are algebra homomorphisms and

(i
A
i
C
) = id
AC
, where denotes the multi-
plication map of the twisted tensor product.
Assume we have a tensor product of A with C. Then, the map
s: C A A C,
dened by s :=

(i
C
i
A
), satises:
(1) s

(
C
A) = A
C
and s

(C
A
) =
A
C,
(2) s

(
C
A) = (A
C
)

(s C)

(C s),
(3) s

(C
A
) = (
A
C)

(A s)

(s A).
A map satisfying these conditions is called a twisting map. Conversely, if
s: C A A C
is a twisting map, then AC becomes a twisted tensor product via

s
:= (
A

C
)

(A s C).
This algebra will be denoted A
s
C. Furthermore, these constructions are inverse
to each other.
The following result is useful in order to check that a map s: C A AC is
a twisting map, and will be used implicitly in this paper.
Proposition 1.1. Let s: CA AC be a map satisfying conditions (1) and (2).
If (c
i
)
iI
generates C as an algebra and
s(c
i
aa

) = (
A
C)

(A s)

(s A)(c
i
a a

)
for all a, a

A and each index i, then s is a twisting map.

4 JORGE A. GUCCIONE, JUAN J. GUCCIONE, AND CHRISTIAN VALQUI
1.2. Non commutative truncated polynomial extensions. In the sequel we
x C = k[y]/y
n
. Let A be a k-algebra and s: C A A C a k-linear map.
The equations
s(y
r
a) =
n1

j=0

r
j
(a) y
j
dene k-linear maps
r
j
: A A for 0 j, r < n. Moreover, we put
r
j
= 0 if r n
and 0 j < n. Note that the
r
j
s are dened for r 0 and 0 j < n.
Proposition 1.2. The following assertions are equivalent:
(1) The map s is a twisting map.
(2) (a)
0
j
=
j0
id.
(b)
r
j
(1) =
jr
.
(c) For j < n and 0 < r < n,

r
j
(ab) =
n1

i=0

r
i
(a)
i
j
(b). (Product law)
(d) For j < n, r > 1 and 0 < i < r,

r
j
=
j

l=0

i
l

ri
jl
. (Composition law)
(3) (a)
0
j
=
j0
id.
(b)
1
j
(1) =
j1
.
(c) For j < n,

1
j
(ab) =
n1

i=0

1
i
(a)
i
j
(b).
(d) For j < n and r > 1,

r
j
=
j

l=0

1
l

r1
jl
.
Proof. (1) (2) We know that s is a twisting map if and only if
(a) s(1 a) = a 1,
(b) s(y
r
1) = 1 y
r
,
(c) s(y
r
ab) = (
A
C)

(A s)

(s A)(y
r
a b),
(d) s(y
r
y
t
a) = (A
C
)

(s C)

(C s)(y
r
y
t
a),
for 0 < r, t < n and a, b A. But a direct computation shows that (a) (2)(a),
(b) (2)(b), (c) (2)(c) and (d) (2)(d).
(2) (3) This is trivial.
(3) (2) First note that (2)(b) follows immediately from (3)(a), (3)(b) and (3)(d).
We now prove that condition (2)(d) holds. For i = 1 and r > 1 this is the same
as (3)(d). We suppose that (2)(d) is true for a xed i > 0 and all r > i, and we
prove it for i + 1 and all r > i + 1. Fix r > i + 1. Then

r
j
=
j

h=0

1
h

r1
jh
by (3)(d)
=
j

h=0
jh

u=0

1
h

i
u

ri1
jhu
by inductive hypothesis
NON COMMUTATIVE TRUNCATED POLYNOMIAL EXTENSIONS 5
=
j

h=0
j

l=h

1
h

i
lh

ri1
jl
setting l := u + h
=
j

l=0
l

h=0

1
h

i
lh

ri1
jl
=
j

l=0

i+1
l

ri1
jl
. by (3)(d)
So (2)(d) is true. It remains to check that (2)(c) is also true. For r = 1 it is the
same as (3)(c). Suppose (2)(c) holds for a xed r with 1 r < n 1. Then

r+1
j
(ab) =
j

l=0

1
l
_

r
jl
(ab)
_
by (3)(d)
=
j

l=0

1
l
_
n1

i=0

r
i
(a)
i
jl
(b)
_
by inductive hypothesis
=
j

l=0
n1

i=0
n1

m=0

1
m
(
r
i
(a))
m
l
_

i
jl
(b)
_
by (3)(c)
=
n1

i=0
n1

m=0

1
m
(
r
i
(a))
j

l=0

m
l
_

i
jl
(b)
_
=
n1

m=0
n1

i=0

1
m
(
r
i
(a))
m+i
j
(b) by (2)(a) and (2)(d)
=
n1

m=0
n1

u=m

1
m
_

r
um
(a)
_

u
j
(b)
setting u := m + i, since

u
j
= 0 for u n
=
n1

u=0
u

m=0

1
m
_

r
um
(a)
_

u
j
(b)
=
n1

u=0

r+1
u
(a)
u
j
(b). by (3)(d)
This nishes the proof.
Remark 1.3. Let B = ker
1
0
. By items (3)(a), (3)(b), the Product law and the
Composition law, B is a subalgebra of A and
1
0
is a right B-linear map.
Remark 1.4. The Composition Law is valid for r 0 and 0 i r. This follows
immediately from item (2)(a) and will be used freely throughout the paper
Remark 1.5. From item (3)(d) of the above proposition it follows easily by induction
on r that
(1.2)
r
j
=

u
1
,...,ur0
u
1
++ur=j

1
u1

1
ur
for all r 1. In particular
r
0
=
1
0

1
0
(r times).
Corollary 1.6. For each 0 j < n, let
1
j
: A A be a k-linear map satisfying

1
j
(1)=
1j
. Set
0
j
:=
0j
id and

r
j
:=

u
1
,...,ur0
u
1
++ur=j

1
u1

1
ur
for r > 1 and j < n.
6 JORGE A. GUCCIONE, JUAN J. GUCCIONE, AND CHRISTIAN VALQUI
If
n
j
= 0 for all j < n and

1
j
(ab) =
n1

i=0

1
i
(a)
i
j
(b) for a, b A and j < n,
then the maps
r
j
satisfy the equivalent conditions of Proposition 1.2.
Proof. By hypothesis we know that (3)(b) and (3)(c) of Proposition 1.2 hold. More-
over, by the denition of the
r
j
s, it is clear that the maps
r
j
satisfy items (3)(a)
and (3)(d) of the same proposition, and that
r
j
= 0 for r n.
Remark 1.7. Notice that when
1
0
= 0, then the condition
n
j
= 0 for j < n is
automatically satised.
Associated with a twisting map
s: C A A C
we have the matrix M M
n
(End
k
(A)) given by
M =
_
_
_
_
_
id 0 . . . 0

1
0

1
1
. . .
1
n1
.
.
.
.
.
.
.
.
.

n1
0

n1
1
. . .
n1
n1
_
_
_
_
_
.
Moreover, for a A we dene the matrix M(a) M
n
(A) as the evaluation of M
in a. That is
M(a)
ij
:=
i
j
(a) (0 i, j < n)
Corollary 1.8. The matrices M(a) fulll:
(1) M(1) = Id.
(2) M(ab) = M(a)M(b).
Proof. This follows from the Product law and the fact that
r
j
(1) =
j1
.
Theorem 1.9. The formulas (a) := M(a) for a A, and
(y) :=
_
_
_
_
_
0 1 0
.
.
.
.
.
.
.
.
.
.
.
.
0 0 1
0 0 0
_
_
_
_
_
(the nilpotent Jordan matrix J
0
),
dene a faithful representation : A
s
C M
n
(A).
Proof. Since (y)
n
= 0, in order to check that denes an algebra map, we only
need to verify that
(y)(a) = (
1
0
(a)) + (
1
1
(a))(y) + + (
1
n1
(a))(y)
n1
.
But note that
_
J
0
M(b)
_
ij
=
_
M(b)
i+1,j
for i < n 1,
0 otherwise,
and
_
M(b)J
u
0
_
ij
=
_
M(b)
i,ju
for j u,
0 otherwise,
and so
_
(y)(a)
_
ij
=
_
J
0
M(a)
_
ij
=
i+1
j
(a)
NON COMMUTATIVE TRUNCATED POLYNOMIAL EXTENSIONS 7
=
j

u=0

i
ju
_

1
u
(a)
_
=
j

u=0
M
_

1
u
(a)
_
i,ju
=
n1

u=0
_
M
_

1
u
(a)
_
J
u
0
_
ij
=
n1

u=0
_
(
1
u
(a))(y)
u
_
ij
,
where the second equality is valid also in the case i = n 1, since
_
J
0
M(a)
_
n1,j
= 0 =
n
j
(a).
The injectivity follows from the fact that the composition of with the surjection
onto the rst row gives the canonical linear isomorphism A
s
C A
n
.
2. A family of twisting maps
Recall that C = k[y]/y
n
. Let A be a k-algebra. The aim of this section is to
study the broad family of twisting maps s: C A AC satisfying the following
conditions:
A1) There exist 1 h n and x A such that
h
0
= 0 and q :=
h1
0
(x) is
right cancelable,
A2)
1
0
is an endomorphism of B-bimodules, where B := ker
1
0
.
Actually Condition A1) is used throughout all the section without explicit mention,
but Condition A2) is not used until Lemma 2.11.
Remark 2.1. By Remark 1.3, we know that
1
0
is a right B-linear map. So, Con-
dition A2) is automatically fullled if B is included in the center of A. Moreover

h1
0
(A) is a right ideal of B. So if q is right invertible we can assume that q = 1.
Remark 2.2. For some results we will need to ask that q is invertible.
Remark 2.3. The family that we are going to consider includes all the twisting maps
with
1
0
= 0. However the results we establish in this section only are relevant when

1
0
,= 0.
In the sequel, for every a A we let M
(h)
(a) denote the hh-submatrix of M(a)
formed by the rst h rows and columns of M(a), and we x both, x and q. Note
that, by the Composition Law, q B.
Lemma 2.4. For each j = 0, . . . , h 1 and i h, the map
i
j
= 0. Consequently,
M
(h)
(ab) = M
(h)
(a)M
(h)
(b) for all a, b A.
Proof. By the Composition law it suces to check this for i = h. For 0 j < h,
let b
j
:=
h1j
0
(x). By the Composition law,
(2.3)
r
0
(b
j
) =
_
q if r = j,
0 if r > j.
Let a be an arbitrary element of A. Since, by the Product law,
0 =
h
0
(ab
1
) =
n1

i=0

h
i
(a)
i
0
(b
1
) =
h
1
(a)q,
8 JORGE A. GUCCIONE, JUAN J. GUCCIONE, AND CHRISTIAN VALQUI
we have
h
1
(a) = 0. Then, again by the Product law,
0 =
h
0
(ab
2
) =
n1

i=0

h
i
(a)
i
0
(b
2
) =
h
2
(a)q,
and so
h
2
(a) = 0, etcetera.
Proposition 2.5. Let 0 < l n/h. For each i lh and j < lh, the map
i
j
vanishes.
Proof. We proceed by induction on l. For l = 1 the result is the previous lemma.
Assuming it is true for l 1,

i
j
=
j

u=0

h
u

ih
ju
= 0,
for each i (l +1)h and j < (l +1)h, as we want, since in each summand
h
u

ih
ju
one of the factors vanishes.
The previous result can be rephrased by saying that the matrix M has the
following shape:
M =
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
id 0 . . . 0

1
0

1
1
. . .
1
h1
.
.
.
.
.
.
.
.
.
.
.
.

h1
0

h1
1
. . .
h1
h1
. . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . .
0

1
n1
.
.
.

h1
n1
0 0 0

h
h

h
2h1
.
.
.
.
.
.
.
.
.

2h1
h

2h1
2h1


.
.
.
.
.
.
0 0
.
.
.
0

n1
n1
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
Corollary 2.6. The elements b
j
=
h1j
0
(x) (j = 0, . . . , h 1) introduced in the
proof of Lemma 2.4, satisfy

lh+r
lh
(b
j
) =
_

lh
lh
(b
jr
) if r j,
0 otherwise,
for 0 l
n1
h
.
Proof. By the Composition law and Proposition 2.5,

lh+r
lh
(b
j
) =
lh

u=0

lh
u
_

r
lhu
(b
j
)
_
=
lh
lh
_

r
0
(b
j
)
_
.
The assertion follows now from the denition of the b
i
s and equality (2.3).
Corollary 2.7. If q is invertible, then h divides n.
Proof. By Remark 2.1 we can assume q = 1. Now, by Corollary 2.6 and Proposi-
tion 1.2, we know that
lh+j
lh
(b
j
) =
lh
lh
(1) = 1. But, if h n, then the case j = nlh
with l =
n1
h
leads to
n
lh
(b
j
) = 1, which is impossible, since
n
lh
= 0.
NON COMMUTATIVE TRUNCATED POLYNOMIAL EXTENSIONS 9
Lemma 2.8. Let D be a k-algebra, and let g : D D be a k-linear map. Assume
that g
h
= 0 and that there exists x D such that q := g
h1
(x) is invertible. Suppose
also that E := ker g is a k-subalgebra of D and g is a right E-linear map. Then D
is a free right E-module of rank h. Moreover B= x, g(x), . . . , g
h1
(x) is a basis.
Proof. Consider a null combination
h1

i=0
g
i
(x)
i
= 0,
with coecients in E. Applying g
h1
to both sizes of this equality, we see that
g
h1
(x)
0
= q
0
= 0. Hence,
0
= 0. Now, applying successively g
h2
, . . . , g
1
, we
get
1
= 0, . . . ,
h1
= 0. So Bis linearly independent. It remains to check that B
generates D as a right E-module. Note that q
1
E, because g(q
1
)q = g(1) = 0.
We will prove by induction on i that there exist
0
, . . . ,
i1
E, such that
(2.4) g
hi
(a) = g
hi
(x)
0
+ g
hi+1
(x)
1
+ + g
h1
(x)
i1
for i = 0, . . . , h.
The case i = 0 is trivial, since g
h
(a) = 0. Assume that (2.4) holds for a xed i < h
and set
(2.5) a
i
:= a x
0
g(x)
1
g
i1
(x)
i1
.
From (2.4) it follows immediately that g
hi
(a
i
) = 0. Hence g
hi1
(a
i
) E, which
implies that
i
:= q
1
g
hi1
(a
i
) E. Consequently, by (2.5),
g
hi1
(a) = g
hi1
(x)
0
+ g
hi
(x)
1
+ + g
h2
(x)
i1
+ g
h1
(x)
i
,
since g
hi1
(a
i
) = q
i
= g
h1
(x)
i
.
Theorem 2.9. If q is invertible, then A is a right free B-module. Furthermore
B= x,
1
0
(x), . . . ,
h1
0
(x) is a basis.
Proof. Apply the lemma with D = A, E = B and g =
1
0
.
Corollary 2.10. Let k be a eld and let A be a nite dimensional k-algebra. If
there exists a twisting map s: C A A C satisfying Condition A1), then
dim
k
(A) = h dim
k
(B).
Proof. This follows from Theorem 2.9, since in a nite dimensional k-algebra each
right cancelable element q is invertible.
Lemma 2.11. If b B, then M
(h)
(b) = bI
h
.
Proof. When h = 1, then M
(h)
=
0
0
= id
A
, and the result is trivial. Assume that
h > 1. Note that
1
0
(b) = 0 implies
i
0
(b) = 0 for all i > 0. Let b
j
(j = 1, . . . , h 1)
be as in Lemma 2.4. Consider the matrix
M
(h)
(b
1
) =
_
_
_
_
_
b
1
0 . . . 0
q
1
1
(b
1
) . . .
1
h1
(b
1
)
0
.
.
.
.
.
.
.
.
.
0
h1
1
(b
1
) . . .
h1
h1
(b
1
)
_
_
_
_
_
.
By Condition A2) and Lemma 2.4,
bq = b
1
0
(b
1
) =
1
0
(bb
1
) =
_
M
(h)
(b)M
(h)
(b
1
)
_
10
=
1
1
(b)q,
and so
1
1
(b) = b, by Condition A1). The same matrix product at the entries (j, 0)
for j = 2, . . . , h1, combined with the facts that
j
0
(b
1
) = 0 and
j
0
is left B-linear,
yields
j
1
(b)q = 0, and so
j
1
(b) = 0. Now, since
0
0
(b) =
1
1
(b) = b and
j
i
(b) = 0 for
i = 0, 1 and j ,= i, the equalities
b
j
0
(b
2
) = M
(h)
(bb
2
)
j0
=
_
M
(h)
(b)M
(h)
(b
2
)
_
j0
j = 1, . . . , h 1,
10 JORGE A. GUCCIONE, JUAN J. GUCCIONE, AND CHRISTIAN VALQUI
give
2
2
(b) = b and
j
2
(b) = 0 for j ,= 2. Proceeding in the same way successively
with M
(h)
(b
3
), . . . , M
(h)
(b
n1
), we obtain the desired result.
Proposition 2.12. For i, j = 0, . . . , h 1, the map
i
j
: A A is left and right
B-linear.
Proof. We only check the left linearity, since the right one is similar. Let b B
and a A. By Lemmas 2.4 and 2.11,

i
j
(ba) =
_
M
(h)
(b)M
(h)
(a)
_
ij
=
_
bI
h
M
(h)
(a)
_
ij
= bM
(h)
(a)
ij
= b
i
j
(a),
as we want.
Proposition 2.13. For each b B the matrix M(b) is upper triangular. Moreover,

lh
lh+u
(b) =
lh+1
lh+u+1
(b) = =
lh+hu1
lh+h1
(b) for l <
_
n
h
_
and u < h,
and

n/hh
n/hh+u
(b) = =
nu1
n1
(b) for u < n
_
n
h
_
h if h does not divide n.
Proof. In order to check that M(b) is upper triangular it suces to verify that

lh+i
lh+j
(b) = 0 for l 0, 0 j < h and i > j. If i h, this follows from Pro-
position 2.5. So, we can assume that j < i < h. Let i

= lh+i and j

= lh+j. We
have

j
(b) =
i1

v=0

lh
j

v
_

i
v
(b)
_
+
u

v=i

lh
j

v
_

i
v
(b)
_
= 0,
because
i
v
(b) = 0 for b B and v < i, by Lemma 2.11, and
lh
j

v
= 0 for v > j,
by Proposition 2.5. Now we are going to prove the equalities. Assume rst that
l < n/h. Let
0 v < h u, v

= lh + v and u

= lh + u + v.
Then

u
(b) =
h1

j=0
j=v

lh
u

j
_

v
j
(b)
_
+
lh
lh+u
_

v
v
(b)
_
+
u

j=h

lh
u

j
_

v
j
(b)
_
=
lh
lh+u
(b),
since
v
j
(b) = 0 for j < h, j ,= v,
lh
u

j
= 0 for j h, and
v
v
(b) = b. The case
l = n/h is similar, but we must take 0 v < n n/hh u.
Lemma 2.14. Let s: C E E C be a twisting map and let 2 j
0
< n. If

1
j
=
1j
id for j < j
0
, then for all i < n the following facts hold:
(1)
i
l
= 0 for l < i and
i
i
= id.
(2)
i
i+l
= 0 for 0 < l < min(j
0
1, n i).
(3)
i
i+j01
= i
1
j0
for i n j
0
.
Proof. Note rst that item (1) follows from formula (1.2) and the fact that
1
0
= 0
and
1
1
= id. We now prove item (2). Again by formula (1.2)

i
i+l
=

u
1
,...,u
i
0
u
1
++u
i
=i+l

1
u1

1
ui
Suppose
i
i+l
,= 0. Then some
1
u1

1
ui
,= 0. Since
1
j
=
1j
id for j < j
0
, each
u
v
is 1 or greater or equal than j
0
. But since l 1 and u
1
+ +u
i
= i +l, there is
at least one u
v
greater or equal than j
0
. But then u
1
+ +u
i
i +j
0
1 > i +l,
which is a contradiction. We nally prove item (3). We proceed by induction on i.
The case i = 1 is trivial. Assume that i > 1 and that the result is valid for i 1. By
NON COMMUTATIVE TRUNCATED POLYNOMIAL EXTENSIONS 11
item (1), we know that
i1
i+j01v
= 0 for v > j
0
. Moreover,
1
v
= 0 for 1 < v < j
0
.
Hence,

i
i+j01
=
i+j01

v=0

i1
i+j01v

1
v
=
i1
i+j02

1
1
+
i1
i1

1
j0
= (i 1)
1
j0
+
1
j0
= i
1
j0
,
where the third equality follows from item (1) and the inductive hypothesis.
Proposition 2.15. Let s: C A A C be a twisting map. Assume that B is
included in the center of A and that h is greater than 1 and cancelable in B. Then,
s(C B) B C if and only if s(c b) = b c for all c C and b B.
Proof. Suppose that s(C B) B C, which implies that
i
j
(B) B for all i, j.
By Lemma 2.11 we know that
1
1
(b) = b for all b B. Hence, by items (1) and (2)
of Lemma 2.14, in order to nish the proof its suces to check that
1
j
= 0 on B,
for all j 2. By Lemma 2.11 this is clear for 1 < j < h, and we are going to prove
it for j h by induction on j. So, we assume that
1
j
(b) = 0, for all b B and
h 1 j < j
0
, and we consider two cases:
a) If j
0
= lh for some l 1, then

h
lh
(bx) =
lh+h1

j=0

h
j
(b)
j
lh
(x) by Prop. 2.5
= b
h
lh
(x) +
h
lh+h1
(b)
lh+h1
lh
(x) by Lemma 2.14
= b
h
lh
(x) +
h
lh+h1
(b)
lh
lh
(1). by Corollary 2.6
= b
h
lh
(x) +
h
lh+h1
(b). by Prop. 1.2
On the other hand,

h
lh
(xb) =
lh

j=1

h
j
(x)
j
lh
(b) by Lemma 2.14 and Prop. 1.2
=
h
1
(x)
1
lh
(b) +
h
lh
(x)b by Lemma 2.14
=
h
lh
(x)b. by Lemma 2.4
So,
h
lh+h1
(b) = 0, since b is central, and then h
1
lh
(b) = 0, by item (3) of
Lemma 2.14. But this implies that
1
lh
(b) = 0, since h is cancelable in B.
b) If j
0
= lh + j for some l 1 and 1 j < h, then on one hand

1
lh
(bb
j
) =
lh+h1

i=0

1
i
(b)
i
lh
(b
j
) by Prop. 2.5
=
1
1
(b)
1
lh
(b
j
) +
lh+h1

i=j0

1
i
(b)
i
lh
(b
j
) by hypothesis
=
1
1
(b)
1
lh
(b
j
) +
1
lh+j
(b)
lh
lh
(1) by Corollary 2.6
= b
1
lh
(b
j
) +
1
lh+j
(b), by Prop. 1.2 and Lemma 2.11
and, on the other hand,

1
lh
(b
j
b) =
n1

i=1

1
i
(b
j
)
i
lh
(b) by Prop. 1.2
=
1
lh
(b
j
)b. by Lemma 2.14
12 JORGE A. GUCCIONE, JUAN J. GUCCIONE, AND CHRISTIAN VALQUI
Hence
1
lh+j
(b) = 0, which concludes the inductive step and nishes the proof of
the proposition.
Remark 2.16. Not all the non commutative truncated polynomial extensions satisfy
the hypothesis of Corollary 2.7 or Proposition 2.15. For instance, take A = k[x]
and n = 3. The following twisting map is not of this type

1
0
(b
0
+ xb
1
) = b
1
x
2
for b
0
, b
1
k[x
2
],

1
1
(x
r
) = (1)
r
x
r
,

1
2
(x
r
) = (1)
r1
x
r1
,
Note that B = k[x
2
], h = 2 does not divides n and s(C B) B C.
Proposition 2.17. Let D be a k-algebra, h > 1 a divisor of n and g
l
: D D
(0 l < n/h) k-linear maps satisfying g
l
(1) = 0. For each j 0, let g
j
0
denote the
j-fold composition of g
0
. Assume that g
h
0
= 0 and that there exists x D such that
q := g
h1
0
(x) is invertible. Suppose furthermore that E := ker g
0
is a k-subalgebra
of D and that g
0
is a right E-linear map. Then, there exists at most one twisting
map s: C D D C, such that
1
lh
= g
l
.
Proof. Since q
1
E, because
g
0
(q
1
)q = g
0
(q
1
q) = 0,
replacing x by xq
1
, we can assume that g
h1
0
(x) = 1, and we do it. For 0 j < h,
let b
j
:= g
hj1
0
(x). Note that b
0
= 1. By formula (1.2), the maps
i
j
, with
i 2, are determined by the
1
u
s, with u j. Moreover, by Corollary 2.6 and
Proposition 1.2,

lh+r
lh
(b
j
) =
_
1 if r = j
0 if r > j.
Hence, by the Product law,

1
lh
(ab
j
) =
n1

i=0

1
i
(a)
i
lh
(b
j
) =
lh+j1

i=0

1
i
(a)
i
lh
(b
j
) +
1
lh+j
(a),
and so, for each j 1, the map
1
lh+j
is determined by the
1
i
s with i < lh+j.
Remark 2.18. Proposition 2.17 is valid under the weaker hypothesis that q is right
cancelable if we demand that s(C E) E C. Note that, by Proposition 2.15,
then, necessarily,
i
j
(b) =
ij
b for all b E and i, j (but we do not assume that h
divides n).
Under suitable hypothesis it is possible to say more about the maps
i
j
. For
instance we have the following result, which will not be used in the sequel.
Proposition 2.19. Let l
n1
h
. Assume that
(2.6)
lh
lh
(ba) =
lh
lh
(b)
lh
lh
(a) for all a A and b B,
and that there exist q
l

h1
0
(A) for 0 l < n/h, such that
lh
lh
(q
l
) is right cance-
lable in A. Then

lh+i
lh+j
(b) = 0 for b B and 0 i < j < h.
Proof. Take x
l
A such that
h1
0
(x
l
) = q
l
and set b
(l)
j
:=
h1j
0
(x
l
) for 0 j < h.
By the Composition law and Proposition 2.5,

lh+k
lh
=
lh

u=0

lh
lhu

k
u
=
lh
lh

k
0
for all k 0.
NON COMMUTATIVE TRUNCATED POLYNOMIAL EXTENSIONS 13
So,
(2.7)
lh+k
lh
(b
(l)
j
) =
lh
lh
_

k+h1j
0
(x
l
)
_
=
lh
lh
_

kj
0
(q
l
)
_
=
_

lh
lh
(q
l
) if k = j,
0 if k > j,
where the last equality for k > j, follows from the fact that q
l
B ker
kj
0
.
Moreover, by Proposition 2.13 we know that for b B,
(2.8)
u
v
(b) = 0 for u > v.
Hence,

lh
lh
(b)
lh
lh
(b
(l)
j
) =
lh
lh
_
bb
(l)
j
_
by (2.6)
=
_
M(b)M(b
(l)
j
)
_
lh,lh
by Corollary 1.8
=
j1

k=0

lh
lh+k
(b)
lh+k
lh
(b
(l)
j
) +
lh
lh+j
(b)
lh
lh
(q
l
), by (2.7) and (2.8)
for all b B. Thus
j1

k=1

lh
lh+k
(b)
lh+k
lh
(b
(l)
j
) +
lh
lh+j
(b)
lh
lh
(q
l
) = 0.
Now, an easy induction on j yields
lh
lh+j
(b) = 0 for j = 1, . . . , h 1. By Proposi-
tion 2.13 this nishes the proof.
Remark 2.20. By the Product law and Propositions 2.5 and 2.13,

lh
lh
(ab) =
n1

j=0

lh
j
(a)
j
lh
(b) =
lh
lh
(a)
lh
lh
(b) for all a A, b B and l <
_
n
h
_
.
Hence, the rst hypothesis of the previous proposition is automatically fullled
when B is included in the center of A. In fact
lh
lh
(b) B, since

1
0
_

lh
lh
(b)
_
=
lh+1
lh
(b) = 0
by the Composition law and Proposition 2.13.
If (2.6) is valid for all l
n1
h
, then from the previous result, Proposition 2.13
and the fact that
0
0
= id, it follows that the matrix M(b) has the following shape
M(b) =
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
b 0 0
0 b 0
.
.
.
.
.
.
.
.
.
.
.
.
0 0 b
0

1
h
(b)
.
.
.
0

1
n1
(b)
.
.
.
0 . . . . . . 0

h
h
(b) 0 0
0
h
h
(b) 0
.
.
.
.
.
.
.
.
.
.
.
.
0 0
h
h
(b)
0 0

2h
2h
(b)
.
.
.

.
.
.
.
.
.
.
.
.
0
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
for each b B.
14 JORGE A. GUCCIONE, JUAN J. GUCCIONE, AND CHRISTIAN VALQUI
3. Twisting maps with
1
0
,= 0
Recall that C = k[y]/y
n
and A is an arbitrary k-algebra. The aim of this
section is to determine all the twisting maps s: C A A C with
1
0
,= 0
satisfying:
- There exist 1 < h n and x A such that
h
0
= 0 and
h1
0
(x) is invertible,
- B := ker
1
0
is included in the center of A,
- s(C B) B C.
Remark 3.1. By the results established in the previous section, we know that if h
is cancelable in B, then h[n, s induces the ip on C B (hence, by the Product
law, the
i
j
are B-linear maps), s is determined by the maps
1
lh
(0 l < n/h) and
A is free over B with basis 1 = b
0
, . . . , b
n1
, where b
j
=
hj1
0
(x).
The next theorem says that the
1
lh
s can be chosen arbitrarily.
Theorem 3.2. Let h > 1 be a divisor of n and let g
l
: A A (0 l < n/h)
be k-linear maps satisfying g
l
(1) = 0. Assume that g
h
0
= 0 and that there exists
x such that q := g
h1
0
(x) is invertible. Suppose furthermore that B := ker g
0
is a
k-subalgebra of the center of A and that the g
l
s are B-linear maps. Then there
exists a unique twisting map s: C A A C such that
1
lh
= g
l
.
Proof. Since 0 = g
0
(q
1
q) = g
0
(q
1
)q = 0, we have q
1
B. Replacing x by
q
1
x, we can assume g
h1
0
(x) = 1. We set
0
j
:=
0j
id and, based on the proof of
Proposition 2.17, for increasing l we dene
r
j
for r 1 and lh j < lh + h, as
follows:
- First
1
lh
:= g
l
,
- Then
r
lh
for r 2, using formula (1.2),
- Then,
1
lh+j
by

1
lh+j
(a) :=
1
lh
(ab
j
)
lh+j1

k=0

1
k
(a)
k
lh
(b
j
) for 1 j < h,
where b
j
=
h1j
0
(x).
- Finally,
r
j
for r 2 and lh + 1 j < lh + h, using formula (1.2).
By construction the maps
r
j
are B-linear and
1
j
(1) =
1j
. Hence, by Corollary 1.6,
in order to prove the theorem it suces to show that
n
j
= 0 for j < n and that
the maps
1
j
s satisfy the Product law. To carry out this task, we will need to use
the Composition law (which follows immediately from the denition of the
r
j
s)
and that
r
j
(1) =
rj
for all r (which follows easily from the case r = 1, using
formula (1.2). Next we will check the Product law for every block of
1
j
s with
lh j < lh + h, and that
n
j
= 0 for j < n, in ve steps.
First step: Check that
lh+h1
lh
(x) = 1.
Second step: Verify the Product law for
1
lh
.
Third step: Verify the Product law for
r
j
with r > 1 and j lh.
Fourth step: Verify the Product law for
1
lh+1
, . . . ,
1
lh+h1
.
Fifth step: Check that
i
j
= 0 for j < lh + h and i lh + h.
For l = 0, we have:
First step. This is true by assumption.
NON COMMUTATIVE TRUNCATED POLYNOMIAL EXTENSIONS 15
Second step. Since the maps
r
j
are B-linear and, by Lemma 2.8, b
j
: 0 j < h
is a B-basis of A, it is sucient to show that

1
0
_
ab
j
_
=
n1

k=0

1
k
(a)
k
0
(b
j
) for 0 j < h.
For j = 0 this follows from the fact that
k
0
(1) =
k0
, while, for j > 0, this follows
from the denition of the
1
j
s and the facts that
j
0
(b
j
) = 1 and
k
0
(b
j
) = 0 for
k > j.
Third step. Assuming that the result is valid for r and proceeding as when we
checked item (2)(c) in the part (3) (2) of the proof of Proposition 1.2, we obtain

r+1
0
(ab) =
1
0
_

r
0
(ab)
_
=
n1

u=0

r+1
u
(a)
u
0
(b).
Fourth step. Let 0 < j < h. Assume that the Product law holds for
1
i
with i < j.
Then,

1
j
(ab) =
1
0
(abb
j
)
j1

i=0

1
i
(ab)
i
0
(b
j
)
=
n1

u=0

1
u
(a)
u
0
(bb
j
)
j1

i=0
n1

u=0

1
u
(a)
u
i
(b)
i
0
(b
j
) by inductive hypothesis
=
n1

u=0

1
u
(a)
_

u
0
(bb
j
)
j1

i=0

u
i
(b)
i
0
(b
j
)
_
=
n1

u=0

1
u
(a)
_
n1

i=0

u
i
(b)
i
0
(b
j
)
j1

i=0

u
i
(b)
i
0
(b
j
)
_
by Second and Third
steps
=
n1

u=0

1
u
(a)
u
j
(b),
where for the last equality we use that
j
0
(b
j
) = 1 and
i
0
(b
j
) = 0 for i > j.
Fifth step. This follows from the proof of Lemma 2.4, since it requires only maps
r
j
,
with r 0 and 0 j < h, satisfying
h
0
= 0, the Product law and the Composition
law (and not the existence of a twisting map).
Next, assuming we have carried out the ve steps until l 1, we execute the ve
steps for l.
First step. By the Composition law,

lh+h1
lh
(x) =
lh

u=0

h
u
_

lh1
lhu
(x)
_
=
h
h
(1) = 1,
since
h
u
= 0 for u < h,
lh1
lhu
= 0 for u > h and
lh1
lhh
(x) = 1.
Second step. Since the maps
r
j
are B-linear and, by Lemma 2.8, b
j
: 0 r < h
is a B-basis of A, it is sucient to show that

1
lh
_
ab
j
_
=
n1

k=0

1
k
(a)
k
lh
(b
j
) for 0 j < h.
16 JORGE A. GUCCIONE, JUAN J. GUCCIONE, AND CHRISTIAN VALQUI
For j = 0 this follows from the fact
k
lh
(1) =
k,lh
. Let now j > 0. By the Fifth
step for l 1,
(3.9)
lh+j
lh
(b
j
) =
lh

i=0

lh
i
_

j
lhi
(b
j
)
_
=
lh
lh
_

j
0
(b
j
)
_
= 1
and
(3.10)
lh+r
lh
(b
j
) =
lh

i=0

rj
i
_

lh+j
lhi
(b
j
)
_
=
rj
0
_

lh+j
lh
(b
j
)
_
= 0 for r > j.
The denition of the
1
lh+j
s for 1 j h 1 yields the desired result.
Third step. Assuming that the result is valid for r and arguing as in the case l = 0
we get

r+1
j
(ab) =
j

k=0

1
k
_

r
jk
(ab)
_
=
n1

u=0

r+1
u
(a)
u
j
(b).
Fourth step. Assume that the Product law holds for
1
i
with i < lh + j. Then, by
the denition of
1
lh+j
, the inductive hypothesis and the Third step,

1
lh+j
(ab) =
1
lh
(abb
j
)
lh+j1

i=0

1
i
(ab)
i
lh
(b
j
)
=
n1

u=0

1
u
(a)
u
lh
(bb
j
)
lh+j1

i=0
n1

u=0

1
u
(a)
u
i
(b)
i
lh
(b
j
)
=
n1

u=0

1
u
(a)
_

u
lh
(bb
j
)
lh+j1

i=0

u
i
(b)
i
lh
(b
j
)
_
=
n1

u=0

1
u
(a)
_
n1

i=0

u
i
(b)
i
lh
(b
j
)
lh+j1

i=0

u
i
(b)
i
lh
(b
j
)
_
=
n1

u=0

1
u
(a)
n1

i=lh+j

u
i
(b)
i
lh
(b
j
)
=
n1

u=0

1
u
(a)
u
lh+j
(b),
where the last equality follows from (3.9) and (3.10).
Fifth step. By the Composition law, for each i lh + h and j < lh + h,

i
j
=
j

u=0

h
u

ih
ju
= 0,
since
h
u
= 0 for u < h and
ih
ju
= 0 for u h.
Now we give an algorithm to construct non commutative truncated polynomial
extensions of a k-algebra A:
(1) Take a subalgebra B of the center of A such that A is a free B-module with
basis b
0
= 1, b
1
, . . . , b
h1
,
(2) Take C = k[y]/y
n
, where n is a multiple of h,
(3) Finally, choose a family g
l
: A A (1 l < n/h) of B-linear maps satisfy-
ing g
l
(1) = 0.
NON COMMUTATIVE TRUNCATED POLYNOMIAL EXTENSIONS 17
Then, there is a unique twisting map s: C A AC such that
s(y a) =
n1

j=0

1
j
(a) y
j
,
where
-
1
0
: A A is the B-linear map dened by g
0
(b
i
) = b
i1
for i 1 and
g
0
(1) = 0,
-
1
lh
= g
l
for 1 l < n/h,
-
1
lh+j
: A A is the B-linear map dened by

1
lh+j
(a) :=
1
lh
(ab
j
)
lh+j1

k=0

1
k
(a)
k
lh
(b
j
) for 1 j < h.
As was said before, all the twisting maps s: C A AC such that
- B := ker
1
0
is a subalgebra of the center of A,
- s(B A) AB,
- there exist h 2 and x A such that
h
0
= 0 and
h1
0
(x) is invertible,
are of this type. In particular, for all such algebras, h[n and A is free over B.
4. Upper triangular twisting maps
The aim of this section is to study twisting maps s: C A A C with

1
0
= 0. Under this assumption the low dimensional Hochschild cohomology plays
a prominent role. The obstructions to inductively construct twisting maps are
cohomology classes. For the sake of simplicity, given a twisting map with
1
0
= 0,
we set :=
1
1
. Moreover, we let
m
denote the m-fold composition of with itself.
Note that formula (1.2) implies
i
j
= 0 for j < i and
i
i
=
i
. In particular M is
upper triangular. Therefore, as in the introduction, we call these twisting map and
the corresponding twisted product, upper triangular. Note moreover that, by the
Product law, is an algebra endomorphism. Throughout this section Z(A) denotes
the center of A and we set
j
:=
j
.
From now on we set C
n
:= k[y]/y
n
, and we let

A

n denote the k-module A

endowed with the A-bimodule structure given by a b c := (a)b
n
(c).
Theorem 4.1. Let s
n
: C
n
A A C
n
be an upper triangular twisting map
and let
i
j
(i 0 and 0 j < n) be the family of k-linear endomorphisms of A
associated with s
n
. Consider the map F : A A

A

n, dened by
F(a b) =
n1

i=2

1
i
(a)
i
n
(b),
where
(4.11)
i
n
:=

u
1
,...,u
i
1
u
1
++u
i
=n

1
u1

1
ui
for i 2.
Then, F is a normalized cocycle in the canonical Hochschild cochain complex of A
with coecients in

A

n. Moreover, there exists an upper triangular twisting map

s
n+1
: C
n+1
A A C
n+1
,
with the same
1
j
s as s
n
for j = 0, . . . , n1, if and only if [F] = 0 in H
2
(A,

n).
In this case F = b
2
(
1
n
), where b
2
is the Hochschild coboundary.
18 JORGE A. GUCCIONE, JUAN J. GUCCIONE, AND CHRISTIAN VALQUI
Proof. It is easy to check that
i
n
(1) = 0 for 1 < i < n. Hence F is normal. We
next prove that it is a cocycle. In fact
b
3
(F)(a b c) =(a)F(b c) F(ab c) + F(a bc) + F(a b)
n
(c)
=
n1

i=2

1
1
(a)
1
i
(b)
i
n
(c)
n1

i=2

1
i
(ab)
i
n
(c)
+
n1

i=2

1
i
(a)
i
n
(bc)
n1

i=2

1
i
(a)
i
n
(b)
n
n
(c)
=
n1

i=2

1
1
(a)
1
i
(a)
i
n
(b)
n1

i=2
i

l=1

1
l
(a)
l
i
(b)
i
n
(c)
+
n1

i=2
n

l=i

1
i
(a)
i
l
(b)
l
n
(c)
n1

i=2

1
i
(a)
i
n
(b)
n
n
(c)
=
n1

i=2
i

l=2

1
l
(a)
l
i
(b)
i
n
(c) +
n1

i=2
n1

l=i

1
i
(a)
i
l
(b)
l
n
(c)
=0.
Now, note that, since
1
0
= 0,

u
1
,...,u
i
0
u
1
++u
i
=n

1
u1

1
ui
,
is well dened for each i 2 (independently of the value assigned to
1
n
), and gives

i
n
. Hence, we can use Corollary 1.6 to conclude that there exists s
n+1
satisfying
the required conditions, if and only if there is a k-linear map
1
n
: A A fullling
(4.12)
1
n
(ab) =
1
1
(a)
1
n
(b) +
1
n
(a)
n
n
(b) +
n1

i=2

1
i
(a)
i
n
(b),
or, equivalently, b
2
(
1
n
)(a b) = F(a b). In fact, the maps
i
j
(i 2 and j < n)
are the same as for s
n
, and so, if a, b A and j < n, then

1
j
(ab) =
n1

i=0

1
i
(a)
i
j
(b) since s
n
is a twisting map
=
n

i=0

1
i
(a)
i
j
(b) because
n
j
= 0.
Finally, the equality

1
n
(ab) =
n

i=0

1
i
(a)
i
n
(b) for a, b A
is the same as (4.12).
Next, we are going to describe the rst serious obstruction to construct an upper
triangular twisting map, when is the identity map. In the following result
j
1
means the j-fold composition of
1
.
Corollary 4.2. Under the assumptions of Theorem 4.1, if = id, then the follow-
ing facts hold:
NON COMMUTATIVE TRUNCATED POLYNOMIAL EXTENSIONS 19
(1) Any derivation
1
of A denes a twisting map s
3
: C
3
A A C
3
, via

1
0
= 0,
1
1
= id and
1
2
=
1
. Moreover, all upper triangular twisting maps
from C
3
A to AC
3
, with
1
1
= id, are of this type.
(2) Any pair of derivations
1
,
2
of A gives a twisting map s
4
: C
4
A AC
4
,
via
1
0
= 0,
1
1
= id,
1
2
=
1
and
1
3
=
2
1
+
2
. Moreover, all upper
triangular twisting maps from C
4
A to A C
4
, with
1
1
= id, are of this
type.
(3) Let
1
and
2
be derivatons of A. Consider the upper triangular twisting
map s
4
: C
4
A A C
4
, dened by
1
and
2
. Then, there exists an
upper triangular twisting map s
5
: C
5
A A C
5
, with the same
1
1
,
1
2
and
1
3
as s
4
, if and only if [
1
] [
2
] = 0 in H
2
(A, A).
Proof. Let F : A A A be as in Theorem 4.1. When n = 2, we have F = 0.
When n = 3, a direct computation shows that F = 2
1

1
= b
2
(
2
1
). Finally,
when n = 4, we have:
F(a b) =
1
2
(a)
2
4
(b) +
1
3
(a)
3
4
(b)
=
1
(a)(3
2
1
(b) + 2
2
(b)) + 3(
2
1
(a) +
2
(a))
1
(b)
= b
2
(
3
1
+ 2
1

2
)(a b) + (
2

1
)(a b).
Items (1), (2) and (3) follow now immediately from Theorem 4.1.
Proposition 4.3. Let be an endomorphism of A. Assume that there exist
b
2
, . . . , b
n1
A such that
2
(b
2
),
3
(b
3
), . . . ,
n1
(b
n1
) are not zero divisors
and b
j
, (b
j
) : j = 2, . . . , n 1 Z(A). Then, given elements a
2
, . . . , a
n1
A,
there is at most one upper triangular twisting map s: C
n
A AC
n
with
1
1
=
and
1
j
(b
j
) = a
j
for j = 2, . . . , n 1.
Proof. Since
1
0
= 0, it follows easily from Corollary 1.6 that, for each j n, the
maps
1
0
, . . . ,
1
j1
dene a twisting map s
j
: C
j
A A C
j
. It is clear that
in order to complete the proof we only need to check that the uniqueness of s
j
implies the one of s
j+1
. By Theorem 4.1, to carry out this task it suces to show
that if : A

A

j is a derivation (that is, a 1-cocycle in the cochain Hochschild

complex) which vanish in b
j
, then = 0. But, from the equalities
(a)
j
(b
j
) = (ab
j
) = (b
j
a) = (b
j
)(a) = (a)(b
j
) for all a A,
it follows that = 0, since
j
(b
j
) is not a zero divisor of A.
Lemma 4.4. Let s
l
: C
l
A AC
l
and F be as in Theorem 4.1, and let b Z(A).
Dene G
b
: AA A by G
b
(a) := F(a b) F(b a). Then,
b
2
(G
b
)(a a

) = F(a a

)
l
(b) (b)F(a a

).
Proof. Since
b
2
(G
b
)(a a

) =(a)F(a

b) F(aa

b) + F(a b)
l
(a

)
(a)F(b a

) + F(b aa

) F(b a)
l
(a

)
and F is a cocycle,
0 =b
3
(F)(a a

b) + b
3
(F)(b a a

) b
3
(F)(a b a

)
=(a)F(a

b) F(aa

b) + F(a a

b) F(a a

)
l
(b)
+ (b)F(a a

) F(ba a

) + F(b aa

) F(b a)
l
(a

)
(a)F(b a

) + F(ab a

) F(a ba

) + F(a b)
l
(a

)
=b
2
(G
b
)(a a

) + (b)F(a a

) F(a a

)
l
(b),
as desired.
20 JORGE A. GUCCIONE, JUAN J. GUCCIONE, AND CHRISTIAN VALQUI
Proposition 4.5. Let s
l
: C
l
A AC
l
be an upper triangular twisting map and
let b Z(A). If
l
(b) is invertible and (b),
l
(b) Z(A), then, for any a Z(A),
there is a unique twisting map
s
l+1
: C
l+1
A A C
l+1
with
1
l
(b) = a and the same
1
j
s as s
l
for j = 0, . . . , l 1.
Proof. We set
1
l
:= a
l
(b)
1

l
+
l
(b)
1
G
b
. Using Lemma 4.4, it is easy to check
that this map fullls b
2
(
1
l
) = F, where b
2
is the coboundary of the Hochschild
complex of A with coecients in

A

l . Then Theorem 4.1 and Proposition 4.3

guarantee the existence and uniqueness of such s
l+1
.
Theorem 4.6. Let : A A be an endomorphism. Assume that there exist
b
2
, . . . , b
n1
Z(A) such that (b
j
),
j
(b
j
) Z(A) and
j
(b
j
) is invertible for
all j. Then, given elements a
2
, . . . , a
n1
Z(A), there is a unique upper triangular
twisting map
s: C
n
A A C
n
,
with
1
1
= and
1
j
(b
j
) = a
j
, for j = 2, . . . , n 1.
Proof. Apply the previous proposition successively to l = 2, 3, . . . , n 1.
Example 4.7. It is not dicult to nd examples in which the hypothesis of the
previous theorem are fullled, for instance, we have the following cases in which
b
j
= b for all j:
(1) Let k be a characteristic zero eld, A = k[x
1
, . . . , x
r
] and take b = x
1
for
all j in the theorem. Let be an algebra morphism with (x
1
) = x
1
+ ,
for some k 0. Then
j
(b) = (1 j) is invertible.
(2) Let K/k be a eld extension and assume there exists an Gal(K/k) and
b K such that b ,=
i
(b) for 1 i n 2. Then
j
(b) is invertible for
1 < j < n.
Remark 4.8. For upper triangular twisting maps, the map =
1
1
seems more
important than in the case
1
0
,= 0. For instance, it is an endomorphism of algebras,
moreover it is easy to check that if is an injective map, then s is also injective;
and that if s is surjective, then so is . Hence, when A is nite dimensional, s is
bijective if and only if is.
Remark 4.9. Let A be a k-algebra. Consider the k-module A[[y]] consisting of the
power series with coecients in A. It is easy to check that having an associative
and unitary algebra structure on A[[y]] such that:
- A and k[[y]] are unitary subalgebras of A[[y]],
-
_

i=0
a
i
y
i
_
y =

i=0
a
i
y
i+1
,
- The canonical surjection A[[y]] A is a morphism of algebras,
is the same that having a associative and unitary algebra structure on each A[y]/y
n

such that
- A and k[y]/y
n
are unitary subalgebras of A[y]/y
n
,
- The multiplication map takes a y
i
to ay
i
,
- The canonical surjection A[y]/y
n
A is a morphism of algebras,
in such a way that the canonical maps

n
:
A[y]
y
n+1

A[y]
y
n

are k-algebra morphisms. Hence, in order to construct such an algebra structure

on A[[y]], Theorem 4.1 can be applied.
NON COMMUTATIVE TRUNCATED POLYNOMIAL EXTENSIONS 21
4.1. Non commutative truncated polynomial extensions of k[x]/x
m
.
Theorem 4.10. Let P
1
, , P
n1
be polynomials in A = k[x]/x
m
such that x
divides P
j
for all j. Then there exists a unique upper triangular twisting map
s
n
: C
n
A A C
n
with
1
j
(x) = P
j
for each j.
Proof. For n = 1 the result is trivial (the unique twisting map is the ip). Suppose
that it is true for n = l, and that x divides
i
j
(x) for all i, j < l. Arguing as in the
proof of Theorem 4.1, we see that in order to construct s
l+1
, it suces to dene

1
l
satisfying the Product law. Using the same construction as in that proof, we
obtain maps
2
l
, . . . ,
l
l
, satisfying the Product law. Consider the matrix
M
(l)
(x) :=
_
_
_
_
_
_
_
x 0 0 0
0
1
1
(x)
1
l1
(x) P
l
0 0
2
l1
(x)
2
l
(x)
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
0 0 0
l
l
(x)
_
_
_
_
_
_
_
,
and take
1
l
(x
h
) :=
_
M
(l)
(x)
h
_
1l
. By the formula of the matrix product
(4.13)
1
l
(x
u+v
) =
l

j=1

1
j
(x
u
)
j
l
(x
v
),
provided that u + v < m. To conclude that (4.13) also holds when u + v m it
suces to verify that x divides M
(l)
(x). We leave this task to the reader.
Corollary 4.11. If P
1
, P
2
, is a sequence of polynomials in A = k[x]/x
m
such
that x divides P
j
for all j, then there exists an algebra structure on A[[y]] such that
- A and k[[y]] are unitary subalgebras of A[[y]],
-
_

i=0
a
i
y
i
_
y =

i=0
a
i
y
i+1
,
- The canonical surjection A[[y]] A is a morphism of algebras,
- yx =

i=1
P
i
y
i
.
Proof. Apply Remark 4.9.
References
[B-M1] T. Brezi nski and S. Majid, Coalgebra bundles, Communications in Mathematical
Physics, Vol 191 (1998) 467492.
[B-M2] T. Brezi nski and S. Majid, Quantum geometry of algebra factorizations and coalgebra
bundles, Communications in Mathematical Physics, Vol 213 (2000) 491521.
[C-S-V] A. Cap, H. Schichl and J. Vanzura, On twisted tensor products of algebras, Communi-
cations in Algebra, Vol 23 (1995) 47014735.
[Ca] P. Cartier, Produits tensoriels tordus, Expose au Seminaire des groupes quantiques de
l

Ecole Normale Superieure, Paris (1991-1992).
[C-I-M-Z] S. Cnepel, B. Ion, G. Militar u, S. Zhu, The factorisation problem and smash biprod-
ucts of algebras and coalgebras, Algebr. Represent. Theory, Vol 3 (2000) 1442.
[C] C. Cibils, Non-commutive duplicates of nite sets, Journal of Algebra Appl., Vol 5(3)
(2006) 361377.
[G-G] J. A. Guccione and J. J. Guccione, Hochschild Homology of Twisted Tensor Products,
K-Theory, vol 18 (4) (1999) 363-400.
[G-G-V] J. A. Guccione, J. J. Guccione and C. Valqui, Twisted planes, Commun. Algebra, Vol
38 (5) (2010) 1930-1956.
[J-L-N-S] P. Jara, J. Lopez Pe na, G. Navarro and D. Stefan, On the classication of twisting
maps between k
n
and k
m
, arXiv:0805.2874v3 [math.RA] 24 Sep 2009.
[J-L-P-V] P. Jara Martinez, J. Lopez Pe na, F. Panaite and F. Van Oystaeyen, On iterated twisted
tensor products of algebras, Internat. J. Mathematics, vol 19(9) (2008) 10531101.
[Ka] C. Kassel, Quantum groups, Graduate Texts in Mathematics, Vol 155 Springer-Verlag,
New Xork (1995).
22 JORGE A. GUCCIONE, JUAN J. GUCCIONE, AND CHRISTIAN VALQUI
[Ma1] S. Majid, The coendomorphism bialgebra of an algebra, J. Fac Sci. Univ. Tokio Sect.
IA Math, vol 34 (1990) 425456.
[Ma2] S. Majid, Algebras and Hopf algebras in braided categories, Advances in Hopf algebras,
Marcel Dekker (1994) 55105.
[Mo] S. Montgomery, Hopf algebras and their actions on rings, CBMS Regional Conference
Series in Mathematics, vol 82 AMS Providence Rhode Island (1993).
[Tam] D. Tambara, Physics for algebraists: Non-commutative and non-cocommutative Hopf
algebras by a bicrossproduct construction, J. Algebra, vol 130 (1990) 1764.
[VD-VK] A. Van Daele and S. Van Keer, The Yang Baxter and the pentagon equation, Compositio
Math., Vol 91 (1994) 201221.
Departamento de Matem atica, Facultad de Ciencias Exactas y Naturales, Pabell on 1,
Ciudad Universitaria, (1428) Buenos Aires, Argentina.
E-mail address: [email protected]
Departamento de Matem atica, Facultad de Ciencias Exactas y Naturales, Pabell on 1,
Ciudad Universitaria, (1428) Buenos Aires, Argentina.
E-mail address: [email protected]
Pontificia Universidad Cat olica del Per u - Instituto de Matem atica y Ciencias Afi-
nes, Secci on Matem aticas, PUCP, Av. Universitaria 1801, San Miguel, Lima 32, Per u.
E-mail address: [email protected]