Multicomplex Ideals, Modules and Hilbert Spaces
Multicomplex Ideals, Modules and Hilbert Spaces
Multicomplex Ideals, Modules and Hilbert Spaces
Hilbert Spaces
Derek Courchesne1 and Sébastien Tremblay2
April 2024
Abstract
In this article we study some algebraic aspects of multicomplex num-
bers Mn . A canonical idempotent representation defined in terms of n
multicomplex conjugates is introduced. This representation facilitates
computations in this algebra and makes it possible to introduce a gen-
eralized conjugacy, i.e. a composition of the n multicomplex conjugates,
as well as a multicomplex norm. The ideals of the ring of multicomplex
numbers are then studied. Multicomplex free modules and their linear op-
erators are introduced. Finally, we develop multicomplex Hilbert spaces.
Keywords. Multicomplex numbers, Commutative ring, C∗ -algebra, Canonical
multicomplex representation, Modules on multicomplex, Hilbert spaces,
1 Introduction
For over two decades, bicomplex numbers BC ≃ M2 and its subring of hyper-
bolic numbers (also known as perplex, duplex or pseudo-complex numbers) have
garnered significant attention in mathematical research [7, 11, 16, 20, 22, 23, 29]
and found applications in various domains of physics, including quantum me-
chanics [19, 24, 30, 31, 37, 38] and general relativity [2, 14, 34] as well as in
engineering [28, 40]. The pioneering work of Corrado Segre in 1892 introduced
an infinite set of algebras, termed bicomplex M2 , tricomplex M3 , and so forth,
extending the realm of complex numbers C ≃ M1 . However, despite the sub-
stantial interest and utilization of bicomplex algebra, particularly underscored
by Price’s comprehensive text [27], a notable gap in literature persists concern-
ing the general case of muticomplex algebra Mn .
While Price’s seminal work extensively covers bicomplex numbers mainly
from the point of view of mathematical analysis, it merely touches upon the
broader concept of multicomplex numbers in its final chapter. Multicomplex
algebra, characterized by an arbitrary and finite number of commutative imagi-
nary units i1 , i2 , . . . , in beyond the traditional two of bicomplex numbers (or one
of complex numbers), remained largely unexplored. This can be attributed pri-
marily to the daunting computational challenges associated with multicomplex
calculations, as exemplified by the approach adopted by Price.
1
In light of these considerations, this article endeavors to address this lacuna
by embarking on a study of multicomplex algebra. Our objectives are threefold:
firstly, to meticulously investigate the algebraic structure of multicomplex num-
bers – unlike the mathematical analysis approach of Price (see last chapter in
[27]) – thereby extending the existing discourse beyond bicomplex numbers. Sec-
ondly, to introduce a canonical idempotent representation in terms of the multi-
complex conjugaisons facilitating computationally operations on multicomplex
numbers. And finally, to furnish readers with the requisite mathematical foun-
dations essential for the advancement and application of multicomplex algebra
in diverse realms encompassing fundamental mathematics, physics or applied
sciences.
η ∈ M0 ≃ R : η = x0 ,
η ∈ M1 ≃ C : η = x0 + x1 i1 ,
η ∈ M2 : η = x0 + x1 i1 + x2 i2 + x12 i1 i2 ,
(2)
η ∈ M3 : η = x0 + x1 i1 + x2 i2 + x3 i3 + x12 i1 i2 + x13 i1 i3
+ x23 i2 i3 + x123 i1 i2 i3 ,
..
.
2
where the coefficients x are real and M2 , M3 are the bicomplex and the tricom-
plex algebras, respectively [27]. One can write the general expression of η ∈ Mn
in terms of the power set P {1, . . . , n} :
X
η= xA iA , Pn := P {1, . . . , n} , xA ∈ R,
A∈Pn
where the empty set in Pn is associated with the index zero and i0 := 1, the sin-
gleton {k} ∈ Pn is associated with the index k, the set {k, l} ∈ Pn is associated
with the indices kl such that ikl := ik il , etc.
where †0 is the identity map. We combine these conjugations with the compo-
sition operation denoted ◦ and defined as follow :
3
Moreover, we also have the following relations for the multiplicative unit 1 and
the unit in expressed in terms of γn and γn′ :
1 = γn + γn′ ,
(7)
in = −in−1 (γn − γn′ ).
Hence {γn , γn′ } forms a basis of Mn over Mn−1 . Let η, ζ ∈ Mn with respective
components η1 , η2 and ζ1 , ζ2 in Mn−1 relative to that new basis. Then from (6)
addition and multiplication in Mn are done componentwise:
Γn := γ2 γ3 · · · γn , n ≥ 2. (9)
Proposition 1. For n ≥ 2 the set Γ‡n has 2n−1 distinct elements. If these
elements are represented by εk for k = 1, . . . , 2n−1 we have
4
such that Γ‡2 contains two distinct elements {ε1 = γ2 , ε2 = γ2′ }. Moreover,
for 1 ≤ k, l ≤ 2n−1 , where the symbol δkl is the usual Kronecker delta and
Λn := †1 †2 · · · †n is the composition of all single conjugates in Mn .
Proof. We proceed by induction over n. For n = 2 we have from the proof
of Proposition 1 that ε1 = γ2 and ε2 = γ2′ , hence εk εl = δkl εl for k, l = 1, 2.
†1 †2
Moreover, εΛ1 = γ2
2
= γ2 = ε 1 , ε Λ ′
2 = ε2 and ε1 + ε2 = γ2 + γ2 = 1.
2
Suppose now that all three properties of the proposition are satisfied for
n
n ≥ 2. Then from Proposition 1 and Remark 2 we have Γ‡n+1 = {ε̃k }2k=1 such
that for 1 ≤ k, l ≤ 2n−1 we obtain
ε̃k · ε̃l = εk γn+1 · εl γn+1 = εk εl γn+1 = δkl εl γn+1 = δkl ε̃l ,
′ ′ ′ ′
ε̃2n−1 +k · ε̃2n−1 +l = εk γn+1 · εl γn+1 = εk εl γn+1 = δkl εl γn+1 = δkl ε̃2n−1 +l ,
′
ε̃k · ε̃2n−1 +l = εk γn+1 · εl γn+1 = 0,
5
which proves (i). Moreover, we have
Λ Λ Λ
ε̃k n+1 = (εk γn+1 )Λn+1 = εΛn n+1 n+1
k γn+1 = εk γn+1 = εk (γn+1 )
†n †n+1
= εk γn+1 = ε̃k
Λ n+1
and ε̃2n−1 +k = ε̃2
n−1 +k by a similar calculation, which demonstrate (ii). Finally,
we have
n n−1 n−1 n−1 n−1
2
X 2X 2X 2X 2X
′
ε̃j = ε̃k + ε̃2n−1 +k = εk γn+1 + εk γn+1
j=1 k=1 k=1 k=1 k=1
n−1
2X
′
= (γn+1 + γn+1 ) εk = 1
k=1
where the change of variables of Remark 2 was use. Hence, the set Γ‡n+1 =
n
{ε̃k }2k=1 is a basis of Mn+1 over C.
Theorem 5 (Canonical idempotent representation). For n ≥ 2, any multicom-
plex element η ∈ Mn can be represented by
n−1
2X
η= zk εk , zk ∈ C(ii )
k=1
6
Remark 6. To simplify the notation for the rest of the paper we will consider
the element i1 as the usual imaginary complex number i. Therefore, we consider
P2n−1
C(ii ) ≃ C and η = k=1 zk εk with zk ∈ C in the last theorem. Moreover for
any z ∈ C(ii ) ≃ C we have z †1 = z, where the bar represents the usual complex
conjugaison.
2.5 Projections
P2n−1
For any multicomplex number η = k=1 zk εk ∈ Mn written in the canonical
idempotent representation, we introduce the jth multicomplex projection as the
function Pj : Mn → C such that Pj (η) = zj . In what follows, the caret notation
̂ for indices will be used in relation with the jth projection in the following
manner :
n−1
2X
η̂ := Pj (η) and η = ηk̂ εk . (11)
k=1
i.e. the jth projection is the identity map when applied on complex elements.
The projection operator is a linear operator for addition and multiplication in
Mn , i.e. for all η, ζ ∈ Mn
Conversely, we have
η̂ = 0 ⇔ εj η = εj η̂ = 0 (12)
holds even if η 6= 0. The set of zero divisors, denoted by M−1
n ,
is then represented
by
n−1
2X
−1
Mn = η = ηk̂ εk | η̂ = 0 for at least one j .
k=1
P2n−1
For any η = k=1 ηk̂ εk ∈ Mn \M−1
n it’s now easy to obtain its inverse: η
−1
=
P2n−1 −1
k=1 ηk̂ εk .
7
2.6 Multiperplex subalgebra
An important subalgebra called the multiperplex numbers and denoted Dn is
the one made up by multicomplex numbers invariant under the Λn composition
of conjugates :
Dn = {ζ ∈ Mn | ζ Λ = ζ}.
Remark 7. For the rest of this article we will omit index n in Λ except where
necessary for clarity.
Since all εk from the canonical representation are invariant under Λ (see
Proposition 3), Γ‡n is a basis of Dn . Expanding the equality ζ Λ = ζ we can
see that a multicomplex number is in the subalgebra Dn if and only if all its
components in the canonical representation are real :
n−1 n−1
2X 2X
Λ
ζ =ζ ⇔ ζk̂ εk = ζk̂ εk ⇔ ζk̂ = ζk̂ , (13)
k=1 k=1
This is also known as the product order (or componentwise order) from [10].
We denote D+ n the set of all positive multiperplex numbers i.e.
D+
n = {η ∈ Dn | η ≥ 0}.
As the multiplication acts componentwise in the canonical basis, the square root
is distributed over the components such that
n−1 n−1
p 2X q 2X
Λ
kηk := η η = 2
|ηk̂ | εk = |ηk̂ |εk ∈ D+
n
k=1 k=1
8
and we know that the result is a positive multiperplex number since |ηĵ | ≥ 0
for all j = 1, . . . , 2n−1 . The value obtained is independent of any basis as it is
always possible to write it back in the canonical basis, and the definition of the
norm itself is independent of any particular representation. For all η, ζ ∈ Mn
we have
n−1 n−1 n−1
2X 2X 2X
kη + ζk = |ηk̂ + ζk̂ |εk ≤ |ηk̂ |εk + |ζk̂ |εk = kηk + kζk
k=1 k=1 k=1
since |ηĵ + ζĵ | ≤ |ηĵ | + |ζĵ | for all j. The others norm properties (absolute
homogeneity and positiveness) can be verified in the same way directly from
the complex case.
The multicomplex algebra Mn equipped with this norm and the conjugate
Λ as the involution is a C∗ -algebra. This fact comes once again from the prop-
erties of complex numbers combined to the properties of Λ as a composition of
conjugates.
3 Multicomplex ideals
3.1 Multiperplex ideals
A multicomplex ring ideal is a subring I of Mn such that the set ηI := {η ζ |
ζ ∈ I} is contained in I for all η ∈ Mn (the multiperplex ring ideal is defined in
the same way on Dn ). Starting with the description of multiperplex ideals, we
will see later that we can get the multicomplex ideals from the complexification
operation on vector spaces, thus covering the study of both at the same time.
From properties (i) and (iii) of Proposition 3, the basis elements Γ‡n =
n−1
{εk }2k=1 of the canonical idempotent representation are an orthogonal decompo-
sition of the identity [13]. This decomposition allow us to write the multiperplex
ring as the following finite direct sum:
n−1
2M
Dn = Dn ε k . (15)
k=1
Each term Dn εj of this sum is the principal ideal generated by the corresponding
single element εj . Since this decomposition is also a basis of the space, for all
P2n−1
η ∈ Dn we have η εj = k=1 ηk̂ εk εj = η̂ εj and Dn εj = Rεj , meaning that
all principal ideals generated by an element of the canonical idempotent basis
are of the form Rεj and the space Dn is a direct sum of these ideals. We will
show that not only Dn , but all multiperplex ideals can be characterized in the
same way.
Proposition 8. Let I 6= {0} be a proper multiperplex ideal of Dn . Then all
non-zero elements of I are zero divisors.
9
Proof. Let η ∈ I a non-zero element. Suppose that η is not a zero divisor,
then η −1 η = 1 ∈ I. Hence, the presence of identity in I implies I = Dn , a
contradiction.
Proposition 9. The principal ideal Rεj is minimal for j = 1, . . . , 2n−1 .
Proof. Let I ⊂ Rεj be a subideal. Either I = {0} or there exists a real number
x 6= 0 such that xεj ∈ I. Then for all y ∈ R, (y/x) · xεj = yεj ∈ I and
I = Rεj .
Lemma 10. A nontrivial multiperplex ideal contains at least one minimal ideal
Rεj .
Proof. Let I 6= {0} a multiperplex ideal. Then there exists a non-zero element
η ∈ I with at least one non-vanishing projection η̂ . Then for all x ∈ R
xεj xη̂ εj
η = = xεj ∈ I ⇒ Rεj ⊆ I.
η̂ η̂
M
Lemma 11. For any subset of indices J ⊆ {1, . . . , 2n−1 }, the direct sum Rεj
j∈J
X
is a multiperplex principal ideal generated by the element εj .
j∈J
Theorem
M 12. A multiperplex ideal I 6= {0} is a principal ideal of the form
Rεj , where J ⊆ {1, . . . , 2n−1 }.
j∈J
L
Proof. Let I be a nontrivial multiperplex ideal and j∈J Rεj ⊆ I the largest
principal
L ideal contained inL
I (from Lemma 10, I contains at least oneLRεk ). If
I\ j∈J Rεj = ∅ then L I = j∈J Rεj which is the desired result. If I\ L j∈J Rεj 6=
∅, then (since 0 ∈ j∈J Rεj ) there exists a non-zero element η ∈ I\ j∈J Rεj
with a non vanishing projection ηk̂ . The existence of this element in the ideal
implies that Rεk ⊆ I and
M
Rεj ⊕ Rεk ⊆ I.
j∈J
L
L the fact that j∈J Rεj is the largest direct sum contained in I.
Contradicting
Thus I = j∈J Rεj and is a principal ideal from Lemma 11.
10
Having characterized the multiperplex ideals in terms of the minimal ideals,
we turn our attention to the orthogonal hyperplanes Hj defined as
Hj := {η ∈ Dn | η εj = 0}. (16)
This set corresponds to all multiperplex numbers with a vanishing jth projec-
tion. We show easily that Hj is an ideal of the multiperplex since ζ Hj ⊆ Hj
for all ζ ∈ Dn .
Proposition 13. For j = 1, . . . , 2n−1 the orthogonal hyperplane Hj is a maxi-
mal multiperplex ideal.
Proof. Suppose Hj is not maximal, then there is a proper ideal I such that
Hj ⊂ I. Let η ∈ I\Hj , then η̂ 6= 0 (otherwise η ∈ Hj ). This implies that
Rεj ⊂ I and
Hj ⊕ Rεj = Dn ⊆ I.
Thus we conclude that all orthogonal hyperplanes are maximal ideals.
Proposition 14. For any proper subset of indices ∅ 6= J ⊂ {1, . . . , 2n−1 }, we
have M \
Rεj = Hk .
j∈J k∈J ∁
From these two last propositions and Theorem 12, we see that all multiper-
plex ideals are written in terms of a direct sum of minimal ideals Rεj as well as
intersections of maximal ideals Hj . Moreover, there are no other ideals distinct
from the Rεj and Hj that are also minimal or maximal and the characterization
of multiperplex is thus completed.
11
Proposition 15. The complexification of a multiperplex ideal is a multicomplex
ideal.
Proof. Let ID ∈ I(Dn ). Considering the complexification ID ⊕ iID of ID , we
suppose that d1 + id2 ∈ ID ⊕ iID . Then, from (14) any η ∈ Mn can be written
as η = d′1 + id′2 where d′1 , d′2 ∈ Dn . We find
η (d1 + id2 ) = (d′1 + id′2 ) (d1 + id2 ) = (d′1 d1 − d′2 d2 ) + i(d′1 d2 + d′2 d1 ).
Since ID is an ideal, all products of the right-hand side of the equation are in
ID and η (d1 + id2 ) ∈ ID ⊕ iID .
From this proposition, the complexification can be seen as a well-defined
function from I(Dn ) to I(Mn ).
Let us now define the set operator R by
We note that for any multicomplex ideal IM (or multicomplex subring) then
R(IM ) is nonempy since 0 ∈ IM and 0Λ = 0.
Proposition 16. Let IM be a multicomplex ideal. Then R(IM ) is a multiperplex
ideal such that R is a mapping from the set of multicomplex ideals to the set of
multiperplex ideals, i.e. R : I(Mn ) → I(Dn ).
Proof. Let IM ∈ I(Mn ). Then for all ζ ∈ Dn and η ∈ R(IM ), we have
(ζ η)Λ = ζ Λ η Λ = ζ η
12
Lemma 18. The mapping R : I(Mn ) → I(Dn ) is the inverse of the complexi-
fication from I(Dn ) to I(Mn ), i.e. for all ID ∈ I(Dn ) and IM ∈ I(Mn )
⇔ d2 = 0 and η ∈ ID .
Now, let ζ = d′1 + id′2 ∈ R(IM )⊕ iR(IM ). Since R(IM ) ⊆ IM from the definition
of R, ζ is in IM . Conversely, if ζ = d1 + id2 ∈ IM then from Proposition 17,
ζ Λ = d1 − id2 ∈ IM and
ζ + ζΛ ζ − ζΛ
= d1 ∈ IM , = d2 ∈ IM .
2 2i
Since d1 , d2 are multiperplex elements in IM , they are invariant under Λ, thus
d1 , d2 ∈ R(IM ) and ζ ∈ R(IM ) ⊕ iR(IM ).
From ring theory, the intersection of two ideals is an ideal and this oper-
ation defines an algebraic structure on I(Dn ) and I(Mn ). More specifically,
(I(Dn ), ∩) and (I(Mn ), ∩) are both monoı̈ds with respective identity elements
Dn and Mn .
Theorem 19. The complexification is a one-to-one correspondence from I(Dn )
to I(Mn ) preserving the intersection, i.e. for any I1 , I2 ∈ I(Dn ),
Proof. The existence of the inverse from Lemma 18 is sufficient to conclude that
the complexification is bijective between I(Dn ) and I(Mn ). If I1 and I2 are two
multiperplex ideals then η = d1 + id2 ∈ (I1 ∩ I2 ) ⊕ i(I1 ∩ I2 ) if and only if
d1 , d2 ∈ I1 ∩ I2 ⇔ d1 , d2 ∈ I1 and d1 , d2 ∈ I2
13
3.3 Translation to the multicomplex ring
From Theorem 19, we get the multicomplex minimal and maximal ideals from
the complexification of Rεj and Hj .
Mn /IJ = {η + IJ | η ∈ Mn } = {ζ + IJ | ζ ∈ IJ ∁ } ≃ IJ ∁ .
The last part comes from the canonical representation which let us separate η
into two distinct sums : one contained in IJ and the other one in IJ ∁ .
4 Free Mn -module
4.1 Bases and subspaces
From [33], any free module over a commutative ring has a well-defined rank. As
Mn is a commutative algebra, any two basis sets of a free Mn -module have the
same (fixed) cardinality m.
Let W be a free Mn -module and {|wl i}m l=1 a basis made up of m < ∞
elements. This describes the finite-dimensional case, and an element |ψi of W
is written as a linear combination of the basis elements :
m
X
|ψi = ηl |wl i, ηl ∈ Mn . (22)
l=1
14
An important subset V ⊂ W is the set of all elements with coefficients restricted
to the field of complex numbers
nX
m o
V := zl |wl i | zl ∈ C . (23)
l=1
This space V is a m-dimensional vector space over C, see [30]. We can use
the canonical idempotent representation of multicomplex numbers to further
develop the expression of any |ψi in W . For each ηl ∈ Mn we have :
n−1
2X
ηl = zl,k̂ εk , zl,k̂ ∈ C. (24)
k=1
and we set
m
X
|ψik̂ := zl,k̂ |wl i. (26)
l=1
Hence, for a given basis of W any element |ψi ∈ W can be written uniquely as
n−1
2X
|ψi = εk |ψik̂ , |ψik̂ ∈ V. (27)
k=1
From that representation of any element |ψi ∈ W , we can define the ket pro-
jector Pj : W → V as
Pj |ψi := |ψî , j = 1, . . . , 2n−1 . (28)
Remark 21. Without ambiguity, we use the same notation Pj both for the
multicomplex projector Pj : Mn → C defined in (11) and the module projector
Pj : W → V defined in (28).
One can show that Pj : W → V is quasilinear, i.e. for any |ψi, |φi ∈ W and
α ∈ Mn :
Pj (|ψi + |φi) = Pj |ψi + Pj |φi = |ψî + |φî
(29)
Pj (α|ψi) = Pj (α)Pj |ψi = α̂ |ψî ,
where Pj (α) is the multicomplex projector.
Remark 22. The definitions of V in (23) and the module projector Pj in (28)
depend on the choice of the basis {|wl i}m
k=1 since each |wl i could be expanded in
a new basis with multicomplex coefficients. The elements of V or any projected
ket |ψî would not necessarily have complex coefficients when written in this new
basis.
15
We say that a ket |ψi is in the null cone of W if for at least one 1 ≤ j ≤ 2n−1 ,
|ψî = 0. Using that framework, we now turn our attention to the properties of
an arbitrary basis of W .
Theorem 23. No basis elements of a free Mn -module can belong to the null
cone.
Proof. Let |wp i be an element of a basis of W . We can write
n−1
2X
|wp i = εk |wp ik̂ .
k=1
If we suppose that |wp i is in the null cone, then |wp il̂ = 0 for at least one l̂.
This implies
εl |wp i = εl |wp il̂ = 0,
but this last equation contradicts linear independence of the basis.
For a given basis {wl }m
l=1 of W , let us now define
m
X
εk V := εk zl |wl i | zl ∈ C , k = 1, . . . , 2n−1 .
l=1
W = ε1 V ⊕ ε2 V ⊕ · · · ⊕ ε2n−1 V.
16
Pm
Since this last equation is valid for every k = 1, . . . , 2n−1 , we have l=1 βl |sl i =
0. The set {|sl i}m
l=1 being a basis of W , for l = 1, . . . , m we have βl = 0 such
that αl = 0 and {Pj |sl i}m
l=1 is a linearly independent set.
We now show that this set span V . Let |φi ∈ V and consider the ket
|ψi = εj |φi ∈ W.
m
Since
Pm the multicomplex span of {|sl i}l=1 is W , there exists βl ∈ Mn such that
l=1 βl |sl i = |ψi. Therefore,
m
X m
X
|φi = Pj |ψi = Pj βl |sl i = Pj (βl )Pj |sl i.
l=1 l=1
The two bases are related by a multicomplex nonsingular matrix (γlp ) [20]
such that m
X
|wl i = γlp |w
ep i, γlp ∈ Mn .
p=1
17
The matrix itself is then
n−1 n−1
2X 2X
A = (Aij )m×m = aij k̂ εk = εk (aij k̂ )m×m . (32)
m×m
k=1 k=1
n−1
2X
det A = εk det Ak̂ .
k=1
(k)
where the columns Ci are complex. Since the determinant is a multilinear
function, for C1 we have
n−1 n−1
2X 2X
(k) (k)
det A = det εk C1 , C2 , . . . , Cm = εk det C1 , C2 , . . . , Cm .
k=1 k=1
From the previous theorem we can see that det A = 0 if and only if det Al̂ = 0
for l = 1, . . . , m. Moreover, det A is in the null cone if det Al̂ = 0 for at least
one l.
Definition 28. A multicomplex square matrix is singular if its determinant is
in the null cone.
18
Theorem 29. The inverse A−1 of a multicomplex square matrix A exists if and
n−1
2X
−1 −1
only if A is nonsingular. Then A is given by A = εk (Ak̂ )−1 .
k=1
Remark 31. Note that for an arbitrary ket |ψi ∈ W , we have APj |ψi ∈ W
and Pj A|ψi ∈ V , hence a multicomplex linear operator A do not commute with
Pj in general.
Theorem 32. A multicomplex linear operator A : W → W commutes with the
ket projector Pj : W → V for 1 ≤ j ≤ 2n−1 if and only if AW ⊆ V .
Proof. Suppose A and Pj commute, then for all |ψi ∈ W we have
and
APj |ψi = A|ψî
such that [A, Pj ] = 0.
19
Definition 33. A multicomplex linear operator A : W → W belongs to the null
cone if for at least one 1 ≤ j ≤ 2n−1 we have  = 0.
Definition 34. Let A : W → W be a multicomplex linear operator and let
A|ψi = λ|ψi, λ ∈ Mn
where |ψi is not in the null cone. Then λ is called an eigenvalue of A and |ψi
is the corresponding eigenket.
Expanding the expression A|ψi = λ|ψi we get
n−1 n−1 n−1 n−1
2X 2X 2X 2X
A εk |ψik̂ = λ εk |ψik̂ ⇒ εk A|ψik̂ = εk λ|ψik̂ . (36)
k=1 k=1 k=1 k=1
This last equation means that the eigenvalue of a ket projection of A is the
corresponding multicomplex projection of λ.
Theorem 35. Let A and B be two multicomplex linear operators. Then for all
1 ≤ j ≤ 2n−1 we have the following properties:
Pj (A + B)|ψi = Pj (A|ψi + B|ψi) = Â |ψî + B̂ |ψî = Â |ψi + B̂ |ψi
Pj AB|ψi = Pj A|φi = Â |φî = Â Pj |φi = Â Pj B|ψi = Â B̂ |ψi.
Once again, an equivalent proof of the next theorem is found in [20], since
the case with bicomplex numbers is easily generalized to that of multicomplex
numbers from the previous results.
Theorem 36. The action of a linear multicomplex operator on W can be rep-
resented by a multicomplex matrix.
20
5 Multicomplex Hilbert Spaces
5.1 Scalar Product
Definition 37. The multicomplex scalar product is a function associating a
multicomplex number to each pair of elements |ψi, |φi ∈ W which for all |χi ∈ W
and α ∈ Mn satisfies:
1. (|ψi, |φi + |χi) = (|ψi, |φi) + (|ψi, |χi);
2. (|ψi, α|φi) = α(|ψi, |φi);
3. (|ψi, |φi) = (|φi, |ψi)Λ ;
4. (|ψi, |ψi) ∈ D+
n and (|ψi, |ψi) = 0 if and only if |ψi = 0.
Here D+n is the set of multicomplex numbers with real and positive com-
ponents in the idempotent canonical representation. The jth projection of the
multicomplex scalar product is denoted
(· , ·)̂ := Pj (· , ·) (38)
and from Definition 37 it follows that any projection is itself a well defined
standard scalar product on the associated vector space εj V (as well as on V ),
i.e.
(|ψi, |φi + |χi)̂ = (|ψi, |φi)̂ + (|ψi, |χi)̂ , (|ψi, α|φi)̂ = α̂ (|ψi, |φi)̂ ,
(|ψi, |φi)̂ = (|φi, |ψi)̂ and (|ψi, |ψi)̂ ∈ R+ , (|ψi, |ψi)̂ = 0 ⇔ |ψî = 0.
(39)
Proposition 38. A ket |ψi ∈ W is in the null cone if and only if its multicom-
plex scalar product (|ψi, |ψi) is in the null cone.
Proof. It’s a direct consequence of the last equivalence in (39).
Theorem 39. Let |ψi, |φi ∈ W , then
n−1
2X
(|ψi, |φi) = (|ψik̂ , |φik̂ )k̂ εk .
k=1
Proof.
n−1 n−1 n−1 n−1
2X 2X 2X 2X
(|ψi, |φi) = |ψik̂ εk , |φil̂ εl = (|ψik̂ , |φil̂ ) εΛ
k εl
k=1 l=1 k=1 l=1
n−1 n−1 n−1
2X 2X 2X
= (|ψik̂ , |φik̂ )εk = (|ψik̂ , |φik̂ )̂ εj εk
k=1 k=1 j=1
n−1
2X
= (|ψik̂ , |φik̂ )k̂ εk .
k=1
21
We know that V and εj V for all j = 1, . . . , 2n−1 are finite-dimensional
vector spaces on the complex C. Furthermore, for each j, (· , ·)̂ is a standard
complex scalar product on both V and εj V , which means that each of these
spaces equipped with the projected scalar product is a finite-dimensional Hilbert
space. From Theorem 24, W is a direct sum of the spaces εj V and is thus itself
a standard Hilbert space equipped with the following scalar product :
n−1 n−1 n−1
2X 2X 2X
|ψk i, |φl i := (|ψk i, |φk i)k̂ , |ψk i, |φk i ∈ εk V. (40)
C
k=1 l=1 k=1
This function induce a norm and a metric from which W is a complete metric
space. There is an important distinction to make between the scalar products
defined on W in Definition 37 and in equation (40): one is multicomplex-valued
and the other is complex-valued respectively. Moreover, from expression (40)
the complex-valued scalar product is clearly induced from the multicomplex
one, but the converse is also possible. Indeed, take (· , ·)C as an independently
defined complex-valued scalar product on W (considered as a vector space over
C), then
n−1
2X
(|ψi, |φi) := (|ψik̂ , |φik̂ )C εk (41)
k=1
22
Theorem 42. Any ket |ψi ∈ W not in the null cone can be normalized.
Proof. For any |ψi ∈ W not in the null cone, we have (|ψi, |ψi) ∈ D+
n and this
scalar product has strictly real positive components:
n−1
2X
(|ψi, |ψi) = ak ε k , ak > 0.
k=1
Thus the multicomplex operator satisfies the same basic properties of the adjoint
on an usual complex vector space. Moreover,
Now let |ψi, |φi ∈ W . We define the operator |φihψ| so that its action on an
arbitrary ket |χi ∈ W is given by
From the generalized Riesz theorem, the action of hψ| on a ket is a linear func-
tional and always gives a scalar, which means that the operator |φihψ| itself is
linear.
As in a standard Hilbert space, the identity operator can be written in terms
of any orthonormal basis {|ul i}m
l=1 of W :
m
X
|ul ihul | = id. (44)
l=1
23
Indeed, the actions of the left-hand side on any basis element |up i is given by
m
X m
X m
X
|ul ihul | |up i = |ul ihul |up i = δlp |ul i = |up i. (45)
l=1 l=1 l=1
and for any projection [Pj (A)]∗ = Pj (A∗ ) = Pj (A). It implies that the projec-
tion of a self-adjoint operator on W is itself a self-adjoint operator on V .
Theorem 45. Let A be a self-adjoint operator on W . Then the eigenvalues
of A associated to an eigenket not in the null cone are all in the set Dn of
multicomplex numbers with real components.
Proof. Let |ψi be an eigenket of A not in the null cone. From equation (37),
If A is self-adjoint then its projections are also self-adjoint and this implies that
λ̂ is a real number for j = 1, . . . , 2n−1 .
Theorem 46. Two eigenkets of a multicomplex self-adjoint operator are or-
thogonal if the difference of the two associated eigenvalues is not in the null
cone.
Proof. Let |ψi and |φi be two eigenkets of a self-adjoint operator A on W with
associated eigenvalues λ and λ′ respectively. Then
24
Proof. Any projection Pj (A) = Â is a self-adjoint linear operator on V . Apply-
ing the standard spectral decomposition theorem, for all j = 1, . . . , 2n−1 there
exists an orthonormal set {|ψl î }m
l=1 of eigenkets of  which is also a basis of
V with respect to the scalar product (· , ·)ĵ . For
n−1
2X
|ψl i := |ψl ik̂ εk
k=1
where λl,k̂ represents the complex eigenvalue associated with the eigenket |ψl ik̂ .
6 Conclusion
The multicomplex canonical basis seems to be the first step to a full understand-
ing of the multicomplex number space. Not only is it the most natural basis
to represent the principal ideals of this structure, but also a way to greatly
simplify the algebra as both addition and multiplication become component-
wise. The multicomplex version of modules and Hilbert spaces share the same
general properties of their complex counterparts, with some differences or more
specific cases due to the presence of zero divisors. We expect new possible in-
terpretations and a broadening of the theory in any further development where
multicomplex numbers are applied.
References
[1] Bryan Berman, Shannon L. Capps, Isaiah Sauvageau, Eric Gao, Sebas-
tian D. Eastham, and Ryan P. Russell. ISORROPIA-MCX: Enabling sen-
sitivity analysis with multicomplex variables in the aerosol thermodynamic
model, ISORROPIA. Earth and Space Science, 10(6), 2023.
[2] Gunther Caspar, Thomas Schönebach, Peter Otto Hess, Mirko Schäfer,
and Walter Greiner. Pseudo-complex general relativity: Schwarzschild,
reissner–nordstöm and kerr solutions. International Journal of Modern
Physics E, 21(02):1250015, 2012.
[3] Johan Ceballos, Nicolás Coloma, Antonio Di Teodoro, Diego Ochoa-
Tocachi, and Francisco Ponce. Fractional multicomplex polynomials. Com-
plex Analysis and Operator Theory, 16(4), 2022.
25
[4] Julia Cen and Andreas Fring. Multicomplex solitons. Journal of Nonlinear
Mathematical Physics, 27(1):17–35, 2019.
[5] J. B. Conway. A Course in Functional Analysis. Graduate Texts in Math-
ematics. Springer, Berlin, 2nd edition, 1990.
[6] Lokenath Debnath and Piotr Mikusinski. Hilbert Spaces with Applications.
Elsevier Academic Press, Amsterdam, 2005.
[7] Aiad Elgourari, Allal Ghanmi, and Mohammed Souid El Ainin. Bicomplex
Frames. Sahand Communications in Mathematical Analysis, 18(3):69–89,
2021.
[8] Vincent Garant-Pelletier. Ensembles de mandelbrot et de julia remplis
classiques, généralisés aux espaces multicomplexes et théorème de fatou-
julia généralisé, 2011. Mémoire de maı̂trise.
[9] J. Garza and H. Millwater. Higher-order probabilistic sensitivity calcula-
tions using the multicomplex score function method. Probabilistic Engi-
neering Mechanics, 45:1–12, 2016.
[10] Sudhir R. Ghorpade and Balmohan V. Limaye. A Course in Multivariable
Calculus and Analysis. Springer, New York, 2010.
[11] Adam Hammam. The bicomplex dual fractional hankel transform. Complex
Analysis and Operator Theory, 18(2), 2024.
[12] B. Hartley and T. O. Hawkes. Rings, Modules and Linear Algebra. Chap-
man and Hall, 1970.
[13] Michiel Hazewinkel, Nadiya Gubareni, and Vladimir V. Kirichenko. Alge-
bras, Rings and Modules, volume 1 of Mathematics and Its Applications.
Kluwer Academic, 2004.
[14] Peter O. Hess and Walter Greiner. Pseudo-complex general relativity. In-
ternational Journal of Modern Physics E, 18(01):51–77, 2009.
[15] I. L. Kantor and A. S. Solodovnikov. Hypercomplex Numbers : an Elemen-
tary Introduction to Algebras. Springer-Verlag, New York, 1989.
[16] Romesh Kumar and Kulbir Singh. Maximal ideals in a bicomplex algebra
and bicomplex Gelfand-Mazur theorem. Complex Variables and Elliptic
Equations, 67(10):2545—-2552, 2022.
[17] Serge Lang. Complex Analysis. Graduate Texts in Mathematics. Springer-
Verlag, London, 4th edition, 1999.
[18] Gregory Lantoine, Ryan P Russell, and Thierry Dargent. Using multicom-
plex variables for automatic computation of high-order derivatives. ACM
Transactions on Mathematical Software (TOMS), 38(3):1–21, 2012.
26
[19] Raphaël Gervais Lavoie, Louis Marchildon, and Dominic Rochon. The bi-
complex quantum harmonic oscillator. Nuovo Cimento Della Societa Ital-
iana Di Fisica B, 125(10):1173–1192, 2010.
[20] Raphaël Gervais Lavoie, Louis Marchildon, and Dominic Rochon. Finite-
dimensional bicomplex hilbert spaces. Adv. Appl. Clifford Algebras,
21(3):561–581, 2011.
[21] S. Lipschutz and M. Lipson. Schaum’s Outline of Linear Algebra. McGraw-
Hill, New York, 4th edition, 2008.
[22] M. Elena Luna-Elizarrarás, Michael Shapiro, Daniele C. Struppa, and
Adrian Vajiac. Bicomplex Holomorphic Functions. Frontiers in Mathe-
matics. Birkhauser Cham, Switzerland, first edition, 2015.
[23] M.E. Luna-Elizarrarás, M. Shapiro, D.C. Struppa, and A. Vajiac. Complex
laplacian and derivatives of bicomplex functions. Complex Anal. Oper.
Theory, 7:1675–1711, 2013.
[24] J. Mathieu, L. Marchildon, and D. Rochon. The bicomplex quantum
coulomb potential problem. Canadian Journal of Physics, 91(12):1093–
1100, 2013.
[25] Harry Millwater and David Wagner. Multicomplex wave functions for lin-
ear and nonlinear schrödinger equations. Adv. Appl. Clifford Algebras,
27(2):1857–1879, 2017.
[26] Juan David Navarro, Harry R. Millwater, Arturo Montoya, and David Re-
strepo. Arbitrary-order sensitivity analysis in phononic metamaterials us-
ing the multicomplex taylor series expansion method coupled with bloch’s
theorem. J. Appl. Mech, 89(2), 2022.
[27] G. Baley Price. An introduction to Multicomplex Spaces and Functions.
Marcel Dekker Inc., New York, 1991.
[28] Thomas Reum and Hannes Toepfer. Introduction of the bicomplex analysis
in the finite element method applied to electromagnetic far-field calcula-
tions. In 2019 19th International Symposium on Electromagnetic Fields
in Mechatronics, Electrical and Electronic Engineering (ISEF), pages 1–2,
2019.
[29] D. Rochon and M. Shapiro. On algebraic properties of bicomplex and hy-
perbolic numbers. Analele Universitatii Oradea, Fasc. Matematica, 11:71–
110, 2004.
[30] D. Rochon and S. Tremblay. Bicomplex quantum mechanics : I. The Gener-
alized Schrödinger Equation. Adv. Appl. Clifford Algebras, 14(2):231–248,
2004.
27
[31] D. Rochon and S. Tremblay. Bicomplex quantum mechanics : II. The
Hilbert Space. Adv. Appl. Clifford Algebras, 16(2):135–157, 2006.
[32] Steven Roman. Advanced Linear Algebra. Graduate Texts in Mathematics.
Springer, New York, 2nd edition, 2005.
[33] Joseph J. Rotman. Advanced Modern Algebra. Prentice Hall, 2002.
[34] M. Schäfer, P.O. Hess, and W. Greiner. Geometry of pseudo-complex gen-
eral relativity. Astronomische Nachrichten, 335(6-7):751—-756, 2014.
[35] Daniele C. Struppa, Adrian Vajiac, and Mihaela B. Vajiac. Differential
equations in multicomplex spaces. In Swanhild Bernstein, Uwe Kähler,
Irene Sabadini, and Frank Sommen, editors, Hypercomplex Analysis: New
Perspectives and Applications, pages 213–227, Cham, 2014. Springer Inter-
national Publishing.
[36] Kyle A. Theaker and Robert A. Van Gorder. Multicomplex wave functions
for linear and nonlinear schrödinger equations. Adv. Appl. Clifford Algebras,
27(2):1857–1879, 2017.
[37] S. Ulrych. Relativistic quantum physics with hyperbolic numbers. Physics
Letters B, 625(3):313–323, 2005.
[38] S. Ulrych. Considerations on the hyperbolic complex Klein–Gordon equa-
tion. Journal of Mathematical Physics, 51(6):063510, 06 2010.
[39] A. Vajiac and M.B. Vajiac. Multicomplex hyperfunctions. Complex Vari-
ables and Elliptic Equations, 57(7-8):751–762, 2012.
[40] Valkova-Jarvis Zlatka, Mihaylova Dimitriya, and Stoynov Viktor. Efficient
orthogonal bicomplex bilinear DSP algorithm design. Electrical, Control
and Communication Engineering, 16(1):30—-36, 2020.
1
DÉPARTEMENT DE MATHÉMATIQUES ET D’INFORMATIQUE, UNIVERSITÉ
DU QUÉBEC, TROIS-RIVIÈRES, QC, CANADA
Email address: derek.courchesne@uqtr.ca
2
DÉPARTEMENT DE MATHÉMATIQUES ET D’INFORMATIQUE, UNIVERSITÉ
DU QUÉBEC, TROIS-RIVIÈRES, QC, CANADA
Email address: sebastien.tremblay@uqtr.ca
28