This article was downloaded by: [University of Bath]

On: 03 October 2014, At: 03:32

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Complex Variables and Elliptic

Equations: An International Journal
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/gcov20

Multicomplex hyperfunctions
a a
A. Vajiac & M.B. Vajiac
a
Faculty of Mathematics, Schmid College of Science , Chapman
University , One University Drive, Orange , CA 92866 , USA
Published online: 19 Aug 2011.

To cite this article: A. Vajiac & M.B. Vajiac (2012) Multicomplex hyperfunctions,
Complex Variables and Elliptic Equations: An International Journal, 57:7-8, 751-762, DOI:
10.1080/17476933.2011.603419

To link to this article: http://dx.doi.org/10.1080/17476933.2011.603419

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
 Complex Variables and Elliptic Equations
Vol. 57, Nos. 7–8, July–August 2012, 751–762

Multicomplex hyperfunctionsy
A. Vajiac* and M.B. Vajiac

Faculty of Mathematics, Schmid College of Science, Chapman University,

One University Drive, Orange, CA 92866, USA
Communicated by M. Luna-Elizarrarás
(Received 26 May 2011; final version received 26 June 2011)
Downloaded by [University of Bath] at 03:32 03 October 2014

In this article, we extend our previous work on hyperfunction theory built

on the space of analytic functions of one or several bicomplex variables to
the multicomplex setting BCn.
Keywords: hypercomplex analysis; hyperfunctions
AMS Subject Classifications: 16E05, 35C15, 13P10, 35N05

1. Introduction
There has been a recent surge of interest in the study of bicomplex and multicomplex
numbers in general in the past decade. This article continues the program introduced
by the authors and their collaborators that endeavours to develop many facets of
the theory of functions of bicomplex and multicomplex variables. We focus on
constructing a hyperfunction theory on the space of multicomplex variables.
During the past decades, a few isolated works analysed either the properties of
bicomplex numbers, or the properties of holomorphic functions of bicomplex
variables, and, without pretense of completeness, we direct the attention of the
reader first to the to book of Price [1], where a full foundation of the theory of
bicomplex numbers was given, then to the work of Shapiro, Rochon and Ryan [2–7].
In our recent paper [8] we studied the properties of holomorphic functions of a
multicomplex variable and in [9] we developed a hyperfunction theory on the space
of holomorphic functions of a bicomplex variable. We studied some of the algebraic
properties of the system of differential equations satisfied by such functions and, in
the spirit of the methodology introduced and developed in [10], we were able to
deduce some new duality theorems which parallel the well-known result of Koethe–
Martineau–Grothendieck for one and several complex variables.
As it is well-known (see, e.g. [11]) the Koethe theorem (and his many
generalizations) are the first step in the development of the theory of hyperfunctions
in R and in Rn. Similarly, a good portion of [10] is devoted to show how the algebraic

*Corresponding author. Email: avajiac@chapman.edu

yDedicated to Professor Michael Shapiro in honour of his 65th birthday.

ISSN 1747–6933 print/ISSN 1747–6941 online

ß 2012 Taylor & Francis
http://dx.doi.org/10.1080/17476933.2011.603419
http://www.tandfonline.com
 752 A. Vajiac and M.B. Vajiac

study of the Cauchy–Fueter system can be used to obtain an interesting theory of

hyperfunctions as boundary values of functions defined on the space of quaternions,
and satisfy the Cauchy–Fueter system (such functions are known in the literature as
regular functions). Thus, it is natural to imagine that it should be possible to build a
suitable theory of hyperfunctions as a byproduct of the theory of holomorphic
functions of one, several bicomplex variables, and, in this article, of one
multicomplex variable.
In order to set the stage, we introduce in the same fashion as [9] the key
definitions and results for the case of holomorphic functions of complex variables
and for the case of regular functions of quaternionic variables (see, e.g. [10,11] for all
the necessary details). To begin with, classical hyperfunctions are defined on Rn as a
purely n-codimensional subvariety of Cn. With this terminology we indicate the fact
that Rn is an initial surface for the Cauchy–Kovalevsky problem for the Cauchy–
Riemann system with real analytic data. In other words, if f is a real analytic function
Downloaded by [University of Bath] at 03:32 03 October 2014

defined on Rn, then there exists only one holomorphic extension of f to a complex
neighbourhood of Rn in Cn.
The definition of hyperfunctions, in particular, is strictly related to the Koethe
duality theorem (strictly speaking, the Koethe only proved the theorem in the case of
one variable) which states that if K is a compact subset of Cn, and if O denotes the
sheaf of holomorphic functions, then

HnK ðCn ; OÞ ffi ðOðK ÞÞ0 ,

where HnK ðCn ; OÞ indicates the cohomology of Cn supported by K, or equivalently the

relative cohomology Hn(Cn, Cn n K;O). Hyperfunctions on open sets U  Rn are then
defined as

BðU Þ ¼ HnU ðCn ; OÞ

and it can be shown that the collection {U, B(U )} is a flabby sheaf on Rn. In the case
of n ¼ 1 variable, the situation is much more explicit, and hyperfunctions are actually
pairs of holomorphic functions, as one can see through the equivalent definition

OðC n U Þ
BðU Þ ¼ :
OðCÞ
The quaternionic case turns out to be more complex. To begin with, we do not have
an easily identified initial surface for the Cauchy–Kovalevsky problem for the
Cauchy–Fueter system, unless we restrict our attention to the case of one
quaternionic variable, in which case it is easy to show that the space H ~ of purely
imaginary quaternions (which is isomorphic to R ) is such a hypersurface in H ¼ R4.
3

More in general, we know from [10] that any initial surface must have dimension
2n þ 1, and we study in detail such initial surfaces in our [12]. Nevertheless, if R
denotes the sheaf of regular functions in n variables, and if S is a suitable dual sheaf,
one can prove an analogue of the Koethe’s duality theorem, namely, for every
compact set K  Hn,

H2n1
K ðHn ; RÞ ffi ðSðK ÞÞ0 ,
 Complex Variables and Elliptic Equations 753

which in the case of one and two quaternionic variables becomes significantly
simpler, as R and S are now isomorphic (this is proved in [13], but it is not true
for n  3).
In [9,14] we were able to prove that, for any compact K in the space BCn, the
following duality theorem can be proved for the sheaf H of holomorphic functions of
several bicomplex variables:
H3n ðBCn , HÞ ffi ðHðK ÞÞ0 :
It should be pointed out that the various exponents that appear in these duality
results (n in the complex case, 2n  1 in the quaternionic case, and 3n in the
bicomplex case) are in fact the codimensions of the characteristic varieties associated
to the defining systems of differential equations, as summarized in the following
table.
Downloaded by [University of Bath] at 03:32 03 October 2014

Cn Hn BCn BCn
Codimension of the characteristic variety
n 2n  1 3n 2n  1
Duality
n n
HnK ðCn ; OÞ 2n1
HK ðHn ; RÞ H3n
K ðBC ; HÞ
2 1
HK ðBCn ; Hn Þ
ffi ðOðK ÞÞ0 ffi ðSðK ÞÞ0 ffi ðHðK ÞÞ0 ffi ðHn ðK ÞÞ0

This table shows the well-known fact that hyperfunctions are defined on varieties
whose dimensions equal the dimensions of the corresponding characteristic varieties
and the nature of the duality theorems that are available in the complex, quaternionic,
bicomplex and multicomplex case. The main purpose of this article is to develop the
results leading to the last column of the above table. In Section 2, we briefly summarize
the main results from the theory of a multicomplex variable, properties of the space of
multicomplex variables, as well as a brief theory of holomorphic functions and their
properties. These results are all carefully studied in [15]; the authors would like to point
the interested reader to [9,14] for a detailed explanations of the algebraic results on the
system that defines holomorphic functions in one and several bicomplex variables, as
well as one multicomplex variable.
In Section 3 we define hyperfunctions on R as a codimension 1 variety in BCn,
and we prove the fundamental properties of the sheaf they form. Finally, Section 4 is
devoted to the development of an abstract Dolbeault complex, and to the special
form that such complex takes when particularized to the case of holomorphic
functions of multicomplex variable. We will show that this complex allows an explicit
reformulation of bicomplex hyperfunctions in terms of suitable differential forms, in
a way that is as explicit as the one available for classical hyperfunctions (as well as
quaternionic hyperfunctions in dimension one and two).

2. The space BCn and multicomplex holomorphy

This section follows closely our previous work [8], in collaboration with
D.C. Struppa, on the multicomplex spaces BCn, for n  2. For the purpose of
 754 A. Vajiac and M.B. Vajiac

self-containment and consistency in this article, we recall the notions and properties
of holomorphic functions of a multicomplex variable.
The first case is when we only have one imaginary unit, say i, in which case the
space BC1 is the usual complex plane C. Since, in what follows, we work with
different complex planes, generated by different imaginary units, we denote such a
space also by Ci, in order to clarify which is the imaginary unit used in the space
itself.
We now turn to the definition of the multicomplex spaces, BCn, for values of
n  2. These spaces are defined by taking n commuting imaginary units i1, i2, . . . , in,
i.e. i2a ¼ 1, and iaib ¼ ibia for all a, b. Since the product of two commuting imaginary
units is a hyperbolic unit, and since the product of an imaginary unit and a
hyperbolic unit is an imaginary unit, we see that these units will generate a group An
of 2n elements: one is the identity 1, 2n1 elements are imaginary units and 2n1  1
are hyperbolic units. Then the algebra generated over the real numbers by An is the
Downloaded by [University of Bath] at 03:32 03 October 2014

multicomplex space BCn which forms a ring under the usual addition and
multiplication operations. The ring P BCn generates a real algebra, and each of its
elements can be written as Z ¼ I 2 An ZI I, where ZI are real numbers. This last
representation of Z 2 BCn is complicated, therefore we make use of the nesting
properties of BCn as follows.
Following [1], it is natural to define the n-dimensional multicomplex space as
  
BCn :¼ Zn ¼ Zn1,1 þ in Zn1,2  Zn1,1 , Zn1,2 2 BCn1 ,
with the natural operations of addition and multiplication. Since BCn1 can be
defined in a similar way, we recursively obtain, at the k-th level,
X Y
n
Zn ¼ ðit Þt 1 Zk,I
jIj¼nk t¼kþ1

where Zk,I 2 BCk, I ¼ (kþ1, . . . , n), and j 2 {1, 2}.

Because of the existence of n imaginary units, we define multiple types of
conjugations as follows:
8 X Y n
>
> ðit Þt 1 ðil Þl 1 Zk,I if l 4 k
>
>
>
< jIj¼nk t¼kþ1
i t6¼l
Zn l ¼
>
> X Y n
i
>
> ðit Þt 1 Zk,I l if l 5 k
>
: t¼kþ1
jIj¼nk
t6¼l

and
il
i2 ...
i ...il
Zn 1 ¼ Zin1 :

It turns out that

i
X X
n X X
k
i
Zn l ¼ l,i ð1Þl 1 ck Zk,I þ l,i ck Zk,I l :
jIj¼nk i¼kþ1 jIj¼nk i¼1
 Complex Variables and Elliptic Equations 755

These conjugations have also been defined in [16,17], where the authors prove that
the composition of these conjugates forms a commutative group of cardinality 2n
isomorphic with Zn2 . Moreover, they show that BCn is a subalgebra of the Clifford
algebra ClR(0, 2n).
In the multicomplex space BCn there are several subsets of idempotent elements
that are organized at each ‘nested’ level BCk inside BCn as follows. Denoting by
1 þ ik il 1  ik il
ekl :¼ , ekl :¼ ,
2 2
it is easy to check that these elements are idempotent. Each of the following sets:
S1 :¼ fen1,n , en1,n g,
S2 :¼ fen2,n1  S1 , en2,n1  S1 g,
..
Downloaded by [University of Bath] at 03:32 03 October 2014

.
Sn1 :¼ fe12  Sn2 , e12  Sn2 g ,
has 2k idempotents, for all k. It is possible to immediately verify that in each set Sk,
the product of any two idempotents is zero. Therefore, we have several idempotent
representations of Zn 2 BCn, as follows.
THEOREM 2.1 Any Zn 2 BCn can be written as
2k
X
Zn ¼ Znk,j ej ,
j¼1

where Znk, j 2 BCnk and ej 2 Sk, for any k ¼ 1,. . . , n  1.

In particular, for k ¼ n  1, any Zn 2 BCn is written as
Zn ¼ 1 e1 þ 2 e2 þ    þ 2n1 e2n1 , ð1Þ
where j 2 Ci1.
Due to the fact that the product of two idempotents is 0 at each level Sk, we will
have many zero divisors in BCn organized in ‘singular cones’. The topology of the
space is difficult, but just like in the case of BC2 we can circumvent this by avoiding
the zero divisors to define the derivative of a multicomplex function as follows.
Definition 2.2 Let  be an open set of BCn and let Zn,0 2 . A function F :  ! BCn
has a multicomplex derivative at Zn,0 if
 
lim ðZn  Zn,0 Þ1 FðZn Þ  FðZn,0 Þ ¼: F 0 ðZn,0 Þ
Zn !Zn,0

exists whenever Zn  Zn,0 is invertible in BCn.

Functions which admit a multicomplex derivative at each point in their domain
are called multicomplex holomorphic, and it can be shown that this is equivalent to
require that they admit a power series expansion in Zn [1, Section 47].
We denote by O(BCn) the space of multicomplex holomorphic functions.
A multicomplex holomorphic function F 2 O(BCn) can be split into F ¼ U þ inV,
where U, V are holomorphic functions of two BCn1 variables, i.e. they admit BCn1
 756 A. Vajiac and M.B. Vajiac

(partial) derivatives in both variables. There is an equivalent notion of multicomplex

holomorphy, which is more suitable to our computational algebraic purposes. The
following theorem follows immediately (the differential operators that appear in the
statement are defined in detail in the next few sections, but it is not difficult to
imagine their actual definition):
THEOREM 2.3 Let  be an open set in BCn and F :  ! BCn such that
F ¼ U þ inV 2 C1(). Then F is multicomplex holomorphic if and only if:
(1) U and V are multicomplex holomorphic in both multicomplex BCn1 variables
Zn1,1 and Zn1,2.
(2) @Zn1,1 U ¼ @Zn1,2 V and @Zn1,1 V ¼ @Zn1,2 U on ; these equations are called the
multicomplex Cauchy–Riemann conditions.
In the idempotent representation, F is BCn holomorphic if and only if the
function U (respectively V ) is a BCn1 holomorphic function in  :¼ Zn1,1  in1
Downloaded by [University of Bath] at 03:32 03 October 2014

Zn1,2 (respectively in  :¼ Zn1,1 þ in1Zn1,2). We can apply recursively this

reasoning for any idempotent set Sk. For example, in the last idempotent
representation set Sn1, using Equation (1) for Z, we get

F ¼ U1 e1 þ    þ U2n1 e2n1 :
We obtain that F is BCn holomorphic if and only if the functions U‘ are complex
holomorphic functions depending only on the Ci1 variable ‘, for all ‘. More
generally, we obtain the following corollary.
COROLLARY 2.4 A multicomplex function F is holomorphic if and only if in any
idempotent representation along the sets S‘ the corresponding multicomplex functions
‘ ‘
U‘ : C2 ° C are multicomplex holomorphic in C2 and, moreover, the multicomplex
Cauchy–Riemann equations are also valid. This last condition equivalent to U‘ are
functions depending only on their corresponding multicomplex variables ‘ 2 BC‘.
The following theorem is an immediate consequence of the nesting behaviour of
BCn and it is easily proven by induction.
THEOREM 2.5 A C1-function F :   BCn ! BCn is holomorphic on  if and only if
@F
I
¼0
@Z
for all sets I of imaginary units such that card(I )  n and I is in increasing order of
indices of complex units.
This theorem can be given an interesting interpretation, similarly to what is
proved by Rochon [5], as follows. We obtain the fact that the multicomplex function
F can be written as
X
FðZn Þ ¼ fI ðz1 , . . . , z2n1 ÞI, ð2Þ
jIjn1

for all Zn 2 , where fI are complex valued functions defined on an open subset
n1
of C2 . With this convention and notation we obtain the following theorem,
in which the C1-condition for F can be eliminated.
 Complex Variables and Elliptic Equations 757

THEOREM 2.6 A function F :   BCn ! BCn (written as in (2)) is holomorphic on 

if and only if fI are complex holomorphic functions and they respect the corresponding
Cauchy–Riemann-type conditions.
Theorem 2.6 indicates that multicomplex holomorphic functions are solutions of
certain overdetermined systems of linear constant coefficients differential equations.
Let us denote, as customary in this field, by P(D) the matrix of differential operators
that act on functions, and whose kernels consist of holomorphic functions of
multicomplex variables.
Since holomorphic functions of multicomplex variables are nothing but vectors
of differentiable functions satisfying a homogeneous system of constant coefficients
differential equations, we immediately obtain the following result (which was stated
for BC2 in [14]).
PROPOSITION 2.7 Holomorphic functions on BCn form a sheaf of rings Hn.
Downloaded by [University of Bath] at 03:32 03 October 2014

The cohomological properties of the sheaf of holomorphic functions on BC2

were derived in [14], by using some standard tools described in detail in [10]. We can
easily show that those same ideas apply to holomorphic functions on BCn, for any
value of n.
In BCn we have 2n  1 differential operators (associated to the various
conjugates), which we relabel as

@Z1 , . . . , @ 2n 1 :
Z

It is not difficult to show that the resolution one obtains in this general case is
Koszul-like, and more precisely one has
2n 1 n
0 ! R2 ! R2ð2 1Þ ! R2ð
n
2 Þ ! R2ð2 31Þ !   
2n 1
   ! R2ð 2 Þ ! R2ð2n 1Þ ! R2 ! M ! 0 ,

where R is a polynomial ring in 2n complex variables, and M ¼ R2/hPti. Moreover, a

n
direct computation yields Extj(M, R) ¼ 0 for 0  j  2n  2, and Ext2 1 ðM, RÞ 6¼ 0.
COROLLARY 2.8 The characteristic variety of M has dimension 1.
In addition, one obtains the following precise result.
THEOREM 2.9 The first Betti number of the resolution is
 
2ð2n1  1Þ  
b1 ¼ 2 þ 4 2n1  1 ¼ ð2n  2Þð2n  1Þ:
2

In an analogous way as in [9], one can prove the following duality theorem.
THEOREM 2.10 If K is a bounded convex set in BCn, then Hj(BCn, Hn) ¼ 0, for
1  j  2n  2. If K is compact, then
n
H2K 1 ðBCn , Hn Þ ’ ðHn ðK ÞÞ0 : ð3Þ

This space allows us to define the space of multicomplex hyperfunctions, as

follows in the next section.
 758 A. Vajiac and M.B. Vajiac

3. Multicomplex hyperfunctions
The standard duality theorem given by isomorphism (3) suggests that the one-
variable multicomplex BCn hyperfunction theory should be defined in codimension
2n  1, i.e. on R.
As mentioned in Section 1, a key argument in developing a hyperfunction theory
is the Cauchy–Kovalevsky extension. Specifically, every real analytic function
f : R ! C extends uniquely to a holomorphic function F :   C ! C, in a
neighbourhood  of the real axis. Similarly, in several complex variables, every
real analytic function f : Rn ! C extends uniquely to a holomorphic function
e ! H, where H
F :   Cn ! C. It is easy to see that every real analytic function f : H e
is the algebra of purely imaginary quaternions, extends uniquely to a regular
function F : H ! H. The case of several quaternionic variables is much more
complicated, and we refer the reader to [12] for its discussion.
We prove below a similar extension theorem in the case of functions of one
Downloaded by [University of Bath] at 03:32 03 October 2014

multicomplex variable.
THEOREM 3.1 Every real analytic function f : R ! BCn extends uniquely to a
bicomplex holomorphic function F :   BCn ! BCn, where  is a neighbourhood
of R in BCn.
Proof Let f : R ! BCn be a real analytic function. In the idempotent representation
in Sn1,
f ðxÞ ¼ u1 ðxÞe1 þ    þ u2n1 ðxÞe2n1 :
The functions u‘ extend uniquely to complex Ci1 holomorphic functions u~‘ ðzÞ,
where z ¼ x þ i1 y 2 Ci1 , such that

u~‘ R ¼ u‘ , ‘ ¼ 1, . . . , 2n1 ,

via the inclusion of R  Ci1 , x ° z ¼ x þ i10. Now consider the extension of f to

f~ : 1 ! BC2 in a neighbourhood of the real axis R in Ci1 :
~ ¼ u~ 1 ðzÞe1 þ    þ u~ 2n1 ðzÞe2n1 :
fðzÞ
Let Zn ¼ 1 e1 þ    þ 2n1 e2n1 2 BCn . Define the extension F of f~ to BCn to be
~ 1 Þe1 þ    þ fð
FðZn Þ :¼ fð ~ 2n1 Þe2n1
! !
2n1
X 2n1
X
¼ u~‘ ð‘ Þe‘ e1 þ    þ u~‘ ð‘ Þe‘ e2n1
‘¼1 ‘¼1
2n1
X
¼ u~‘ ð‘ Þe‘ : ð4Þ
‘¼1

Because ‘ are independent variables, it follows that the holomorphic functions

u~‘ ð‘ Þ depend only on ‘, which is equivalent to the requirement that F is BCn
holomorphic. 
By construction, it is immediate that FR ¼ f. Uniqueness of the bicomplex
extension follows immediately from the uniqueness of the restrictions to the complex
holomorphic functions. g
 Complex Variables and Elliptic Equations 759

COROLLARY 3.2 For n4m, every BCm holomorphic function F :   BCm ! BCn
extends uniquely to a BCn holomorphic function in a neighbourhood 0 of  in BCn.
Proof The proof is similar to the one in the above theorem. A BCm holomorphic
function F with values in BCn has an idempotent decomposition with respect to the
set Snm, and all of its components in this setup extend uniquely to a function F~
defined on a subset of BCn following the method described in (4). g
Example 3.3 Consider, on the real axis R, the function f ðxÞ ¼ x1 which is real
analytic outside the origin. The process described in the previous theorem shows that
its bicomplex BC2 holomorphic extension (4) is given by
1 1
FðZÞ ¼ e12 þ e12 ¼ Z1 ,
 
where Z ¼  e12 þ  e12 . The bicomplex function F is singular at the origin, but not
Downloaded by [University of Bath] at 03:32 03 October 2014

only there, i.e. along the bicomplex divisors of zero.

Continuing the procedure (4), we can extend f to a function G :  ! BC3,
1 1 1 1
GðZ3 Þ ¼ e12 e23 þ e12 e23 þ e12 e23 þ e12 e23 ¼ Z1
3 ,
1 1 2 2
a tricomplex holomorphic in a neighbourhood of the real axis in BC3. We can
continue this extension to BCn, for any n  2.
The following results are almost immediate consequences of our earlier
work [9,14].
THEOREM 3.4 The flabby dimension of Hn is 2n  1.
Proof The flabby dimension is immediately 2n  1, and the vanishing of the
Extj(M, R) guarantees the conclusion. g
The next result is a standard consequence of the flabby dimensionality of Hn
(see, e.g. [10]).
COROLLARY 3.5 For any open set   BCn we have

Hj ð; Hn Þ ¼ 0, j  2n  1:

One can strengthen this result if a convexity hypothesis is added to the

open set .
THEOREM 3.6 For a open convex set   BCn we have

Hj ð; Hn Þ ¼ 0, j  1:

Proof The case of j  2n  1 follows from the previous corollary, and does not
require any condition on the open set . For j ¼ 1, . . . , 2n  1 the vanishing of
Extj(M, R) and the fact that  is convex guarantees the conclusion, as follows.
We have

Ker Pj ðDÞ
Hj ð; Hn Þ ¼ ,
Im Pj1 ðDÞ
 760 A. Vajiac and M.B. Vajiac

where the kernel Ker Pj(D) is the space of BCn functions satisfying the compatibility
conditions for the system Pj1(D). For convex  it follows that these functions
are images of functions by Pj1(D). This proof is analogous to the proof of
Theorem 4.5 in [12]. g
As in the case of quaternionic hyperfunctions, we cannot define multicomplex
hyperfunctions as quotients of appropriate cohomology spaces, but we can define the
multicomplex hyperfunctions in R as follows.
Definition 3.7 Let A  R be an open set and consider  a BCn neighbourhood of A
in BCn, i.e. A ¼  \ R. Then we define
n
BBCn ðAÞ :¼ H2A 1 ð; Hn Þ:
THEOREM 3.8 The above definition does not depend on .
Downloaded by [University of Bath] at 03:32 03 October 2014

Proof The restriction map U U

V :  ðU; Hn Þ !  ðV; Hn Þ defined as V ðÞ ¼ jV is
an isomorphism at the cohomology level since  is supported in A. g
As shown in [11], the cohomological properties of the sheaf Hn imply the
following theorem.
THEOREM 3.9 The collection f, BBCn ðÞg defines a flabby sheaf on R.

4. Dolbeault complexes
Let E be the space of infinitely differentiable functions on Euclidean X of large
enough dimension. Let Pj and #j, for j ¼ 1, . . . , m, be differential operators with
constant coefficients and differential symbols, respectively.
For any function f 2 E we define Tj( f ) :¼ Pj( f )#j. We assume that the symbols #j
satisfy #j#i ¼ #i#j. The operators Pj commute, i.e. Pj(Pi( f )) ¼ Pi(Pj( f )) for every
differentiable function f.
@f
A simple example is X ¼ C, m ¼ 1 and Tð f Þ ¼ @z dz, obvious extension to X ¼ Cm,
@f
for m41. Another example is X ¼ H, m ¼ 1 and Tð f Þ ¼ @q dq. For m41: much more
delicate.
We define a complex of operators and differential forms as follows. At the first
stage we define E 1 to be simply the space {E#1 þ    þ E#m}, and the operator
D0 : E ! E 1 is defined by
D0 ð f Þ ¼ P1 ð f Þ#1 þ    þ Pm ð f Þ#m ¼ T1 ð f Þ þ    þ Tm ð f Þ:
At the second stage, E 2 is defined as the space of differential forms of degree two, i.e.

E 2 ¼ E#1 #2 þ E#1 #3 þ    þ E#m1 #m ,

m
constituted of 2 summands. The operator D1 : E 1 ! E 2 is defined by
X
m
D1 ð f1 #1 þ    þ fm #m Þ ¼ Pi ð fj Þ#i #j :
i, j¼1

It is immediate that D1D0 ¼ 0, i.e. we have the first step of a complex.

 Complex Variables and Elliptic Equations 761

We continue by simply applying the standard rules of differential forms. We

build the spaces E s of differential forms for s ¼ 0, . . . , m. The operators Ds : E s ! E sþ1
are defined by
X
m
Ds ð!Þ ¼ Ti ð!Þ,
i¼1
s
 , Ds} consisting of
for any s-differential form !. In this way we get a complex: E :¼ {E
s m
m þ 1 non-trivial spaces, each space E being a direct sum of s summands, and m
non-trivial operators. This complex is non-exact, and a cohomology is defined as
usual by setting, for s ¼ 1, . . . , m
Ker Ds : E s ! E sþ1
Hs ðEÞ ¼ ,
Im Ds1 : E s1 ! E s
Downloaded by [University of Bath] at 03:32 03 October 2014

and
H0 ðEÞ ¼ Ker D0 : E ! E 1 :

In order to apply this abstract construction for our multicomplex spaces, we

choose the space X and the operators Tj so that the abstract construction is
applicable to the study of holomorphic functions of one multicomplex variable. It is
quite natural to take m ¼ 2n  1 and choose as the corresponding operators those
generated by all the different conjugations. For example, in the case m ¼ 3, we have
@f i @f i @f i i
T1 ð f Þ ¼ i1
dZ 1 , T2 ð f Þ ¼ i2
dZ 2 , T3 ð f Þ ¼ i1 i2
dZ 1 2 :
@Z @Z @Z
In the general case we get
@f I
TI ð f Þ ¼ I
dZ ,
@Z
where I is a set of imaginary units.
It is immediate to see that the operators commute on differentiable functions on
the space of multicomplex numbers, and that the complex we obtain is composed of
2m spaces of differential forms. To exemplify, in the case m ¼ 2 we have the
following:
. the space E of infinitely differentiable functions
. the space E 1 of one form (which is the sum of three isomorphic spaces)
. the space E 2 of two form (also the sum of three isomorphic spaces)
. the space E 3 of three forms, which is isomorphic to the space E itself.
The study of the cohomology groups is concerned with the conditions for
cocycles to be coboundaries, which are exactly the compatibility conditions which we
obtained above. We note here that the complex E offers a fine resolution of the
sheaf Hn:
n
0 ! Hn ! E ! E 1 !    ! E 2 1
! 0:

This fact allows us to calculate the cohomology of the sheaf Hn by simply using the
resolution.
 762 A. Vajiac and M.B. Vajiac

Acknowledgements
The authors are grateful to Professors D.C. Struppa and M. Shapiro for the collaborative
work and many stimulating conversations on the topics of bicomplex numbers and of
hyperholomorphic functions.

References

[1] G.B. Price, An Introduction to Multicomplex Spaces and Functions, Monographs and
Textbooks in Pure and Applied Mathematics, Vol. 140, Marcel Dekker, Inc., New York,
1991.
[2] M.E. Luna-Elizarraras and M. Shapiro, On modules over bicomplex and hyperbolic
numbers, to appear in Applied Complex and Quaternionic Approximation, World Sci.,
Singapore.
Downloaded by [University of Bath] at 03:32 03 October 2014

[3] D. Rochon and M. Shapiro, On algebraic properties of bicomplex and hyperbolic numbers,
An. Univ. Oradea Fasc. Mat. 11 (2004), pp. 71–110.
[4] K.S. Charak, D. Rochon, and N. Sharma, Normal families of bicomplex holomorphic
functions, Fractals 17(3) (2009), pp. 257–268.
[5] D. Rochon, On a relation of bicomplex pseudoanalytic function theory to the complexified
stationary Schrodinger equation, Complex Variables Elliptic Eqns. 53 (2008), pp. 501–521.
[6] J. Ryan, Complexified Clifford analysis, Complex Variables Elliptic Eqns. 1 (1982),
pp. 119–149.
[7] J. Ryan, C2 extensions of analytic functions defined in the complex plane, Adv. Appl.
Clifford Algebras 11 (2001), pp. 137–145.
[8] D.C. Struppa, A. Vajiac, and M.B. Vajiac, Holomorphy in multicomplex spaces, submitted
to the Proceedings of the IWOTA 2010, Birkhauser Berlin.
[9] F. Colombo, I. Sabadini, D.C. Struppa, A. Vajiac, and M.B. Vajiac, Bicomplex
hyperfunctions, Ann. Mat. Pura Appl. 190 Number 2 (2011), pp. 247–261.
[10] F. Colombo, I. Sabadini, F. Sommen, and D.C. Struppa, Analysis of Dirac Systems and
Computational Algebra, Birkhäuser, Boston, 2004.
[11] G. Kato and D.C. Struppa, Fundamentals of Algebraic Microlocal Analysis,
Marcel Dekker, New York, 1999.
[12] F. Colombo, A. Damiano, I. Sabadini, and D.C. Struppa, Quaternionic hyperfunctions on
five-dimensional varieties in H2, J. Geom. Anal. 17(3) (2007), pp. 435–454.
[13] W.W. Adams, C.A. Berenstein, P. Loustaunau, I. Sabadini, and D.C. Struppa, Regular
functions of several quaternionic variables and the Cauchy–Fueter complex, J. Geom. Anal.
9(1) (1999), pp. 1–15.
[14] F. Colombo, I. Sabadini, D.C. Struppa, A. Vajiac, and M.B. Vajiac, Singularities of
functions of one and several bicomplex variables, Ark. Mat. (Institut Mittag-Leffler),
Online First, September 2010.
[15] A. Sudbery, Quaternionic Analysis, Math. Proc. Camb. Phil. Soc. 85 (1979), pp. 199–225.
[16] V. Garant-Pelletier and D. Rochon, On a generalized Fatou–Julia theorem in multicomplex
spaces, Fractals 17(2) (2009), pp. 241–255.
[17] V. Garant-Pelletier, Ensembles de Mandelbrot et de Julia remplis classiques, generalises aux
espaces mulricomplexes et theoreme de Fatou–Julia generalise, Master Thesis, Universite
de Montreal, 2011.